Book - Directed Enzyme Evolution Advancees and Applications PDF

Miguel Alcalde Editor
Directed
Enzyme
Evolution:
Advances and
Applications
Directed Enzyme Evolution: Advances
and Applications
Miguel Alcalde
Editor
Directed Enzyme
Evolution: Advances
and Applications
Editor
Miguel Alcalde
Department of Biocatalysis
Institute of Catalysis and Petrochemistry, CSIC
Madrid
Spain
ISBN 978-3-319-50411-7 ISBN 978-3-319-50413-1 (eBook)

DOI 10.1007/978-3-319-50413-1
Library of Congress Control Number: 2017932555
© Springer International Publishing AG 2017

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology
now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the
editors give a warranty, express or implied, with respect to the material contained herein or for any errors
or omissions that may have been made.
Printed on acid-free paper
This Springer imprint is published by Springer Nature

The registered company is Springer International Publishing AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
The last couple of decades have witnessed a true revolution in protein engineering
that very few could have foreseen, a revolution delivered by the hand of directed
evolution. This reliable, robust and effective methodology has enabled us to design
enzymes with improved or even novel properties, a mere pipedream only a few
years ago. Through consecutive rounds of random mutation, recombination and
screening/selection, enzymes tailored a la carte are meeting the needs of different
industrial processes, addressing the challenges of working at high temperature or
extreme pHs, in non-natural environments or in the presence of strong inhibitors.
More significantly, through laboratory evolution we can now harness the catalytic
promiscuity of many enzymes to undertake non-natural chemistry for a range of
applications that lie within the biotechnology rainbow. Conversely, the rise of
directed evolution has demonstrated that the tailoring of enzymes, metabolic path-
ways or even whole microorganisms is now no longer beyond our reach.
Although mimicking the natural process of evolution on a laboratory scale might
make us feel ashamed of our immense ignorance regarding protein function (for
example through the identification of substitutions of interest that we could have
never predicted by rational analysis), dozens of enzymes are being rapidly engi-
neered in a manner that is establishing solid links between the structure–function
relationship of proteins and their biotechnological applications. The advent of com-
putational protein engineering, the constant expansion of protein and gene data-
bases, together with the birth of more sophisticated (ultra)high-throughput screening
protocols and new library creation methods, is paving the way to expand directed
enzyme evolution beyond the limits of nature. As such, the number of enzymes
undergoing laboratory evolution is becoming overwhelming, covering many aspects
of biotechnology, and contributing to the rapid development of systems and syn-
thetic biology.
As we reach the 25th anniversary of the discovery of directed enzyme evolution,
this book presents some case studies of evolved enzymes, while updating concepts
and methods within this vigorous research ground. The outstanding panel of con-
tributors in Directed Enzyme Evolution: Advances and Applications has ensured the
compilation of some remarkable examples of enzymes improved by evolution in
just a single volume, as well as providing an interesting collection of the cutting-
edge approaches and methods currently available or soon to be added to the directed
evolution toolbox. The first part of the book (Chaps. 1, 2, 3, 4, 5, 6 and 7) focuses
v
vi Preface
on several evolutionary success stories, involving tryptophan synthases, therapeutic

and stereoselective enzymes, CO2-fixing enzymes, unspecific peroxygenases, phy-
tases as well as the directed evolution of whole cells. From Chaps. 8 to 10, library
creation methods, in silico/computational enzyme design and ancestral enzyme res-
urrection are described in depth, while offering clues to their applications and
prospects.
I truly hope that Directed Enzyme Evolution: Advances and Applications will
become a valuable benchside book for professors, researchers and students working
and/or lecturing in the field of protein engineering and biotechnology, complement-
ing other practical texts and volumes on this fascinating field of research.
Madrid, Spain Miguel Alcalde

Contents
1 Directed Evolution of an Allosteric Tryptophan Synthase to Create a

Platform for Synthesis of Noncanonical Amino Acids�� 1
Javier Murciano-Calles, Andrew R. Buller,
and Frances H. Arnold
2 Engineering Therapeutic Enzymes�� 17
Stefan Lutz, Elsie Williams, and Pravin Muthu
3 Recent Advances in Directed Evolution of Stereoselective
Enzymes�� 69
Manfred T. Reetz
4 Improving CO2 Fixation by Enhancing
Rubisco Performance�� 101
Robert H. Wilson and Spencer M. Whitney
5 Directed Evolution of Unspecific Peroxygenase�� 127
Patricia Molina-Espeja, Patricia Gómez de Santos,
and Miguel Alcalde
6 Recent Advances in Directed Phytase Evolution and Rational
Phytase Engineering�� 145
Amol V. Shivange and Ulrich Schwaneberg
7 Strain Development by Whole-Cell Directed Evolution�� 173
Tong Si, Jiazhang Lian, and Huimin Zhao
8 Back to Basics: Creating Genetic Diversity�� 201
Kang Lan Tee and Tuck Seng Wong
9 Resurrected Ancestral Proteins as Scaffolds
for Protein Engineering �� 229
Valeria A. Risso and Jose M. Sanchez-Ruiz
10 Molecular Modeling in Enzyme Design, Toward In Silico Guided
Directed Evolution �� 257
Emanuele Monza, Sandra Acebes, M. Fátima Lucas,
and Victor Guallar
vii
Directed Evolution of an Allosteric
Tryptophan Synthase to Create 1
a Platform for Synthesis of Noncanonical
Amino Acids
Javier Murciano-Calles, Andrew R. Buller,
and Frances H. Arnold
Abstract
Tryptophan and its derivatives are important natural products and have many
biochemical and synthetic applications. However, the more elaborate these
derivatives are, the more complex the synthesis becomes. In this chapter, we
summarize the development of an engineered enzymatic platform for synthesis
of diverse tryptophan analogs. This endeavor utilizes the tryptophan synthase
(TrpS) enzyme, an α2β2 heterodimeric protein complex that catalyzes the last
two steps in the biosynthetic pathway of tryptophan. Although the synthetically
useful reaction (indole + Ser = Trp) takes place in the β-subunit (TrpB), the
exquisite allosteric regulation of this enzyme impedes the use of isolated TrpB
due to its dramatically decreased activity in the absence of the α-subunit (TrpA).
This chapter discusses our efforts to engineer TrpB to serve as a general plat-
form for the synthesis of noncanonical amino acids. We used directed evolution
to enhance the activity of TrpB from Pyrococcus furiosus (PfTrpB), so that it
can act as a stand-alone biocatalyst. Remarkably, we found that mutational acti-
vation mimics the allosteric activation induced by binding of TrpA. Toward our
goal of expanding the substrate scope of this reaction, we activated other homo-
logs with the same mutations discovered for PfTrpB. We found improved cata-
lysts for the synthesis of 5-substituted tryptophans, an important biological
motif. Finally, we performed directed evolution of TrpB for synthesis of
β-branched amino acids, a group of products whose chemical syntheses are par-
ticularly challenging.
J. Murciano-Calles • A.R. Buller • F.H. Arnold (*)

Division of Chemistry and Chemical Engineering, California Institute of Technology,
Pasadena, CA, USA
e-mail: [email protected]
© Springer International Publishing AG 2017 1

M. Alcalde (ed.), Directed Enzyme Evolution: Advances and Applications,
DOI 10.1007/978-3-319-50413-1_1
2 J. Murciano-Calles et al.
1.1 Introduction
In addition to being one of the standard 20 proteinogenic α-amino acids, tryptophan

(Trp) is the immediate precursor of important biomolecules such as the neurotrans-
mitter serotonin [1], vitamin B3 [2], and auxin phytohormones [3, 4]. In biosyn-
thetic pathways of more complex natural products, modified tryptophan is frequently
the core of the final biomolecule [5–9]. Tryptophan derivatives have also been used
in chemical biology for a variety of applications [10]. These noncanonical amino
acids (ncAAs) also serve as intermediates in the production of pharmaceuticals by
chemical synthesis [11]. It is therefore important to develop efficient routes to pre-
paring these compounds.
Tryptophan synthase (TrpS) has been used to make a wide variety of Trp ana-
logs. TrpS catalyzes the last steps in the de novo pathway for Trp in all three domains
of life. This enzyme synthesizes Trp from 3-indole-d-glycerol phosphate (IGP) and
l-serine (Ser) in two steps that take place in two separate subunits of the heterodi-
meric enzyme [12]. TrpS has been used for the synthesis of many modified trypto-
phans by reaction of Ser and an appropriate nucleophile, typically a substituted
indole [13–20]. However, these reactions frequently proceed with low yields (below
50%). Researchers have tried to expand the substrate scope to nitrogen nucleophiles
through protein engineering, but those efforts concomitantly boosted an abortive
side deamination reaction that prevented efficient use of the catalyst [21].
We believed that TrpS would be a good starting point to develop a platform for
synthesizing a wider variety of Trp analogs than had been reported. In particular, we
sought to generate catalysts for C-C, C-N, and C-S bond-forming reactions to make
ncAAs from inexpensive starting materials using low catalysts loadings. Our
approach was to first simplify the heterodimeric enzyme complex and engineer
TrpB to function as a single, stand-alone enzyme. To help the reader understand this
strategy, we describe the sophisticated mechanism of allosteric regulation that gov-
erns native TrpS activity.
1.1.1 TrpS
In the first steps of the native catalytic cycle, IGP binds in the α-subunit (TrpA), and
Ser binds in the β-subunit (TrpB), where it forms a covalent Schiff base linkage to
the cofactor, pyridoxal 5’-phosphate (PLP). Once bound, Ser undergoes dehydra-
tion to form an electrophilic amino acrylate intermediate. TrpA then catalyzes the
retro-aldol cleavage of IGP, releasing indole, which diffuses through a tunnel con-
necting the two subunits and enters the TrpB active site. There it reacts with the
amino acrylate to yield l-tryptophan (Fig. 1.1).
The efficient functioning of TrpS requires that all of the mechanistic steps be
carefully synchronized [22]. In essence, both subunits have an open conformation
that permits the entry of substrates, but exhibits a slow rate of catalysis, and a closed
conformation that accelerates intermediate chemical steps, but cannot bind sub-
strate or release product. Each subunit’s equilibrium between open and closed states
1 Directed Evolution of the Tryptophan Synthase 3
Fig. 1.1 Overall reaction catalyzed by TrpS. The α-subunit (green) cleaves IGP in a retro-aldol
reaction that releases indole, which diffuses through a molecular tunnel into the β-subunit (orange).
There, indole reacts with Ser in a PLP-dependent β-substitution reaction to yield Trp
is allosterically modulated by the other subunit, thus ensuring that intermediates are
not released prematurely. In particular, the indole must be available to TrpB imme-
diately upon formation of the amino acrylate intermediate, which could otherwise
decompose through hydrolysis. However, if indole were released before the sub-
units are in a fully closed state, it would leak into the cellular medium, whereupon
it would diffuse through the membrane and be lost. To accomplish the necessary
synchronicity, each subunit acts as an allosteric effector to the other [12, 22].
The molecular basis for this synchronization comprises structural transitions in
both subunits [23–25]. In TrpA, the majority of residues in the αL6 loop switch
from disordered to well ordered, forming a closed state (Fig. 1.2). In TrpB, the
structural change is more significant; around 20% of the residues change conforma-
tion. The so-called COMM domain, which refers to the region of TrpB that medi-
ates the communication between subunits, undergoes a rigid-body motion and
rotates between open, partially closed, and fully closed conformations (Fig. 1.2).
Closure in the COMM domain impedes access to the active site from solution and
stabilizes the closed conformation in TrpA. Concurrently with the COMM domain
motion, the conformation of TrpB in the interface between the two subunits is also
altered. These conformational changes combine to form a 25-Å tunnel between both
subunits, allowing indole to diffuse from TrpA to the TrpB active site.
TrpS thus exhibits a sophisticated allosteric control mechanism in which both
subunits play a fundamental role. However, only the β-subunit performs catalysis in
the β-substitution reaction that is useful to make ncAAs. Unfortunately, the
Fig. 1.2 Structural transitions in the conformational switch from open to closed states in
Salmonella typhimurium TrpS. The open conformation is in orange (PDB ID: 1KFK), and the
closed conformation is in blue (PDB ID: 2J9X). In TrpA (light gray), many residues in loop αL6
are disordered in the open conformation, but most become well ordered upon the switch to the
closed conformation. In TrpB (dark gray), the arrows indicate the direction of the motion of the
COMM domain from the open to the closed state. The PLP is represented in sticks
allosteric regulation has hindered the use of isolated TrpB, whose activity in isola-
tion is seriously diminished compared to the full TrpS complex. Consequently, we
sought to engineer a stand-alone TrpB for efficient catalysis in the absence of
TrpA. We hypothesized that such a simplified biocatalyst could serve as an effica-
cious starting point for further engineering to expand reactivity.
1.2 ctivation of TrpB from Pyrococcus furiosus by Directed

A
Evolution
The first task in engineering a stand-alone TrpB catalyst was to identify a suitable
parent for directed evolution. Most studies of TrpS have been done with the homo-
log from Salmonella typhimurium (StTrpS), a mesophilic organism. We decided to
use TrpS from Pyrococcus furiosus, a thermophilic archaeon that survives at
100 °C. The ability of this enzyme to function at high temperature (75 °C) enables
solubilization of highly hydrophobic substrates such as indole without addition of
cosolvent. Another significant advantage of engineering a protein from a thermo-
philic organism is the possibility of accumulating more mutations that boost activity
but may be destabilizing [26].
We wanted to evolve TrpB to catalyze its native reaction, the condensation of
indole and Ser to give Trp, more efficiently as a stand-alone enzyme. For this we
developed a high-throughput assay that measures the change in absorbance at
290 nm as indole is converted to Trp [27]. The use of a thermostable protein is
highly advantageous for screening at this wavelength, because the background
absorption from E. coli proteins can be reduced through heat treatment at 75 °C,
which yields moderately pure PfTrpB enzymes.
We constructed a random mutagenesis library of PfTrpB by error-prone PCR and

measured the activity of 528 clones. From this library, we identified 20 clones (3.8%
of all variants) with at least a 35% increase in Vmax relative to the wild-type enzyme
[28]. The most active of these contained a single Thr→Ser mutation that increased
the catalytic efficiency of PfTrpB on indole by 20-fold, which is even greater than
the change induced by TrpA binding. Twelve activating mutations were recom-
bined, and screening yielded a new variant, PfTrpB4D11, that retained the T292S
mutation and incorporated four more (E17G, I68V, F274S, T321A). This enzyme,
which has a further 2.6-fold increase in catalytic efficiency, was used as the parent
for a final round of random mutagenesis, from which we identified PfTrpB0B2 har-
boring the single additional P12L mutation that increased the catalytic efficiency to
3.3 × 105 M−1 s−1 with indole, 83-fold higher than PfTrpB and threefold higher than
the allosterically activated PfTrpS complex [28].
We next investigated whether the increased activity of this stand-alone TrpB was
achieved through the same mechanism as the binding of TrpA. Several lines of
evidence suggested this was the case. We performed our screening under saturating
conditions of each substrate and nevertheless observed a coupled decrease in the KM
for each substrate. PfTrpA binding not only causes a ~threefold increase in kcat but
also a decrease in KM values for Ser and indole, by twofold and fourfold, respec-
tively, suggesting a similar mechanism of activation at work during both effector
binding and mutational activation.
To further understand how PfTrpA regulates PfTrpB, we used X-ray crystallog-
raphy and UV-vis spectroscopy to probe conformational changes and the steady-
state distribution of intermediates in the active site. Importantly, our data showed
that the structural and spectroscopic properties of PfTrpB and PfTrpS are very simi-
lar to those reported in the distantly-related but well-studied enzyme from
Salmonella typhimurium [25]. The UV-vis spectrum of the internal aldimine
(i.e., when the PLP is bound to the catalytic lysine) has a λmax of ~412 nm
(Fig. 1.3a). Ser binding to PfTrpB is associated with the large-scale conformational
rearrangement of the COMM domain into a partially closed state and accumulation
of an external aldimine intermediate, which shifts λmax to 428 nm (Fig. 1.3a). When
PfTrpA is bound, the Ser-bound spectrum shifts to a characteristic λmax at 350 nm,
consistent with stabilization of the amino acrylate intermediate. Structural studies
with StTrpS show that this species is stabilized by a fully closed state of the COMM
domain (Fig. 1.2), which also has an increased affinity for indole [12]. As can be
seen in Fig. 1.3b, the spectrum after addition of Ser to PfTrpB0B2 corresponds to
amino acrylate intermediate, further supporting the hypothesis that mutations
increase activity through the same mechanism as allosteric effector binding, i.e.,
stabilization of the closed conformational state.
Each of these experiments probed changes in PfTrpB during its native catalytic
cycle with Ser and indole. We found that PfTrpA binding also increases the relative
rate of product formation upward of 200-fold with different indole analogs.
Therefore, it was of interest to know whether mutations were also activating for
ncAA synthesis. We found that our stand-alone PfTrpB0B2 catalyst was also broadly
activated for eight different indole analogs in C-C and C-N bond-forming reactions
O
E(Aex1)
a E(Ain)
λmax = 412 nm
K82 K82
HO O
– λmax = 428 nm
+ NH2 H +
N N
H H
– –
O O
P P
N CH3 N CH3
O
E104
O–
NH H
N O
O
–
HO O
– +
O N
+ O H
N –
H O
– –
E(A-A) O P
O
P λmax = 350 nm N
N CH3
N CH3 OH H
P
b N CH3
0.1
0.09 PfTrpB PfTrpS PfTrpB0B2
0.08
0.07
Absorbance
0.06
0.05
0.04
0.03
0.02
0.01
0
300 350 400 450 500 550 300 350 400 450 500 550 300 350 400 450 500 550
Wavelength (nm)
Fig. 1.3 Spectroscopic signatures of the TrpB catalytic cycle. (a) Different intermediates in the
catalytic cycle of TrpB are shown with their corresponding λmax. Protonation states are assigned
according to Caulkins et al. [29] from study of StTrpS. Spectroscopic data are broadly similar for
the two enzymes, supporting this assignment. (b) UV-vis spectra of PfTrpB in isolation, in the
native complex, and in the stand-alone engineered protein PfTrpB0B2. Spectra recorded with 20 μM
of protein (black lines) and after addition of 20 mM Ser (gray dashes). The shifts in λmax indicate
that PfTrpB accumulates E(Aex1) at steady state, whereas PfTrpS and PfTrpB0B2 accumulate the
E(A-A) intermediate
[28]. This trend was similar to the rate enhancement induced by PfTrpA binding,
with some differences emerging between the two enzyme systems. PfTrpB0B2 was
moderately faster at C-N bond-forming reactions with indole and indazole, whereas
PfTrpS was moderately faster with 5-bromoindole. Hence, mutations that increased
activity with indole were broadly activating with other substrates. Again, PfTrpB0B2
resembles the catalytic features described for PfTrpS.
Given the considerable interest in understanding allosteric phenomena [30–
34], a compelling question arose: are activating mutations confined to a distinct
allosteric pathway or interface? The first round of evolution identified 27
Fig. 1.4 Sites where

mutation of PfTrpB is
associated with increased
activity. In the structure of
PfTrpS (PDB ID: 5E0K),
TrpA is shown in pink, and
TrpB is shown in green and
blue. PLP is depicted in
sticks. Residues where
mutations were found that
gave validated increases in
Vmax during high-throughput
screening are shown as
spheres. There is an
abundance of sites in the
COMM domain (colored in
blue) and at the subunit
interface that give rise to
increases in activity
mutations distributed across 20 improved clones. Although the effect of each

mutation has not been measured individually, the data provide a broad picture of
PfTrpB activation (Fig. 1.4). It was described that some residues undergo switch-
like behavior upon effector binding in StTrpS, but just one of the 27 possible
activating mutations (N166D) was found along the homologous route for PfTrpB
[12, 35, 36]. Taking a more generous view of what an allosteric “pathway” might
look like, 17 of the 27 activating mutations (63%) were found at either the pro-
tein-protein interface or within the COMM domain. These regions comprise 31%
of the total PfTrpB sequence, indicating modest enrichment within the areas pre-
viously thought to control allosteric signaling in TrpB. While the mutations have
a positive effect on PfTrpB catalysis, it is not clear how or why they influence the
rate of the reaction. Indeed, almost 40% of activating mutations are outside of
any region that one might a priori think has an impact on allosteric signaling or
catalysis.
From the study of PfTrpB activation, it is clear that mutations can reproduce
complex conformational changes induced by effector binding. Similar effects were
previously shown in a handful of other examples from the literature. Shi and Kay
identified mutations for the activation of the bacterial HslV protease [37]. The pro-
teolytic activity of HslV is increased ~200-fold upon binding of its partner protein
HslU, which is essential for its role in cellular protein degradation [38]. A sensitive
NMR analysis was used to monitor chemical shift perturbations in Ile, Leu, Met,
Val, and Thr residues in HslV upon HslU binding. From these data, a pair of helices
that undergo conformational changes at the HslU binding site was identified. A
small panel of conservative mutations at positions within these helices was con-
structed, and, strikingly, six of the mutations increased catalytic efficiency, the larg-
est by ~20-fold. NMR analysis showed that increases in activity were correlated
with substantial chemical shift perturbations similar to the effects of native effector
binding.
Another example is the activation of LovD, an acyl transferase that transfers an
α-methylbutyrate group that is covalently tethered to an acyl carrier protein, LovF
[39, 40]. With substrate surrogates that are not bound to LovF, the acyltransferase
activity of LovD is greatly reduced, indicating that the carrier protein also serves as
an allosteric activator. Tang and collaborators employed nine rounds of directed
evolution to increase activity on a nonnatural substrate in the absence of the protein
effector as well as increase the thermal stability and tolerance to organic solvent
[40]. They assessed the aggregate effect of these mutations on LovD conforma-
tional dynamics with molecular dynamics (MD) simulations. Their results sug-
gested that the activity of LovD is altered upon LovF binding through the
stabilization of a closed and catalytically active conformation. Interestingly, simula-
tions suggested that engineered LovD enzymes experience comparable conforma-
tion changes in the absence of their effector. Testing this hypothesis experimentally
would have been exceptionally difficult without further modification of LovD, as
there is no chromophore (like PLP for TrpB) that enables one to probe the steady
state of the catalytic cycle directly.
1.3 ctivation of TrpB from Other Species by Transfer

A
of Mutations
With our stand-alone PfTrpB in hand, we wished to test whether the mutational
activation of TrpB could be generalized to TrpBs from other organisms. All known
TrpBs are subject to allosteric regulation by TrpA [12, 41, 42], and it is plausible
that the allosteric mechanisms are conserved across TrpS homologs. Furthermore,
our interest in expanding the substrate scope of TrpB might be helped by assessing
other TrpB homologs, since enzyme homologs often exhibit different substrate
scopes [43, 44]. For example, native StTrpS is a poor catalyst for N-alkylation,
whereas PfTrpS is moderately proficient. However, we did not wish to repeat the
effort required to activate PfTrpB (screening ~3100 clones) for the other homologs.
Instead we tested whether activating mutations in PfTrpB0B2 have the same effects
when transferred to TrpBs from other species.
Successful transfer of beneficial mutations among homologous proteins has
been reported numerous times, although not for allosteric properties, as far as
we know. For instance, multiple sequence alignments of homologs allow the
identification of consensus residues that tend to be thermostabilizing within the
protein family [45]. Also, mutations that alter nicotinamide cofactor specificity
can be transferred to homologous enzymes [46]; in this case, structural and
sequence analyses of an entire protein family provided specialized “recipes” to
change specificity from NADPH to NADH. However, engineering allostery

may be significantly more complex than transferring mutations that enhance
thermostability, where the mutational effects are largely additive, or specificity,
where the effects are usually more localized. Allosteric regulation occurs
through a dynamic mechanism that transfers information between the allosteric
binding site and the enzyme active site and involves many, if not all, residues in
the protein. Often, experimental evidence establishes that a residue that partici-
pates in transmitting this information is not conserved across different homo-
logs. This holds even when the allosteric mechanisms are superficially similar,
and evolution frequently causes homologous proteins to develop different allo-
steric mechanisms [47].
To test the transferability of the allostery-mimicking mutations, we selected
diverse TrpB homologs with differing sequence identities and well separated in the
phylogenetic tree [48]. The closest homolog to PfTrpB tested was from
Archaeoglobus fulgidus, AfTrpB (72% sequence identity), which is a thermophilic
archaeon. We also selected the TrpB from Thermotoga maritima (TmTrpB, 64%
sequence identity), a thermophile that belongs to the bacterial domain of life. Lastly,
we chose the TrpB from Escherichia coli (EcTrpB, 57% sequence identity), a meso-
phile. The sequence alignment of the homologs showed some differences at the
positions of the activating mutations in PfTrpB0B2. Importantly, two mutations in
PfTrpB0B2 were already present as the wild-type residues in two of the homologs,
A321 in TmTrpB (mutation T321A in PfTrpB0B2) and S297 in EcTrpB (mutation
T292S in PfTrpB0B2).
Making the 0B2 mutations in the homologs led to varied levels of catalytic effi-
ciency. AfTrpB0B2 was indeed activated, with a ~20-fold increase in catalytic effi-
ciency with respect to wild type. Similar to P. furiosus enzymes, the UV-vis spectrum
after addition of Ser to AfTrpB showed a λmax at 428 nm, reflecting accumulation of
the external aldimine intermediate. After addition of Ser to the 0B2 mutant, the
spectrum was similar to AfTrpS, with a λmax at 350 nm [49] corresponding to the
amino acrylate intermediate. Hence, these mutations, which mimic the binding of
the allosteric protein partner in PfTrpB, appear to have the same effect in
AfTrpB. However, this was not the case for the other two mutant homologs, where
the catalytic efficiencies of TmTrpB0B2 and EcTrpB0B2 decreased by more than 40%
relative to the wild-type enzymes.
To determine whether a subset of the mutations could activate the other two
homologs, we made and screened recombination libraries of the 0B2 mutations for
TmTrpB and EcTrpB. With this strategy, we found three activating mutations
(P19G, I69V, and T292S) for TmTrpB. T292S was the most activating single muta-
tion, conferring a sixfold increase in catalytic efficiency. All the variants containing
this mutation showed a UV-vis spectrum after addition of Ser similar to the native
TmTrpS complex [49]. The most activated TmTrpB variant harbored the three muta-
tions and exhibited a tenfold increase in catalytic efficiency.
Activation of the last, E. coli homolog was more challenging: the recombination
library of the 0B2 mutations in EcTrpB produced no increased activity. The T292S
mutation had a prominent effect in the other TrpB homologs, but Ser is the native
residue in the equivalent position of EcTrpB. We made a saturation mutagenesis

library at this site but again did not find any activated variants. In a final attempt to
activate this homolog, we returned to the initial random mutagenesis performed on
PfTrpB, where 27 activating mutations were found [28]. We chose to test the muta-
tions of the double mutant PfTrpBM144T N166T because these residues are highly con-
served across all known TrpBs and are located in the COMM domain. These
mutations activated EcTrpB, giving more than a twofold increase in catalytic effi-
ciency. We tested the effects of these two mutations in the other homologs and
found that all were activated, with a two- to fivefold increase in catalytic efficiency.
The UV-vis spectra after addition of Ser to PfTrpBM144T N166D or EcTrpBM149T N171D did
not have a λmax at 350 nm and instead showed an accumulation of the external aldi-
mine species. However, after addition of Ser to the corresponding AfTrpB and
TmTrpB mutants, the spectra showed a λmax at 350 nm, characteristic of the amino
acrylate species. These results suggest that this set of mutations also mimics the
allosteric activation exerted by TrpA binding, although in some of the homologs,
this activation does not completely reach TrpS-like behavior [49]. Similarly, in the
initial evolution of PfTrpB, the activating T292S mutation alone was not sufficient
to shift the spectrum to the amino acrylate species, which required four more
mutations.
Once we accumulated this panel of stand-alone TrpB enzymes, we sought to
compare their substrate profiles with substituted indoles. We were particularly inter-
ested in accessing Trp derivatives with substituents in the 5 position because this
site is often substituted in biologically relevant Trp-based compounds. For instance,
the halogenase PyrH chlorinates tryptophan in the 5 position during the biosynthe-
sis of the antifungal antibiotic pyrroindomycin B [8]. In the biosynthesis of the
neurotransmitter serotonin and the hormone melatonin, a tryptophan hydroxylase
mono-oxygenates tryptophan in the 5 position [1]. Halogenation or borylation in
this position would generate analogs that could serve as intermediates for other
biologically active compounds through cross-coupling reactions [50–52]. Different
TrpS homologs have shown a substantial decrease in activity with 5-substituted
indoles bearing a substituent bulkier than fluorine [17, 18].
We tested our panel of activated stand-alone biocatalysts with 5-bromoindole.
Remarkably, one of the new activated TrpB, TmTrpBM145T N167D, was ~sixfold faster
than PfTrpB0B2, the first enzyme we engineered (Table 1.1). We tested eight more
5-substituted indoles with the two enzymes and saw that TmTrpBM145T N167D was
always faster than PfTrpB0B2, from 1.4- to 7.5-fold (Table 1.1). As shown in Table
1.1, the yields for 5-chlorotryptophan and 5-bromotryptophan are 93% and 88%,
respectively. Previous use of StTrpS for these two reactions reported yields of 61%
for 5-chlorotryptophan and only 33% for 5-bromotryptophan [17]. Moreover, the
chemical syntheses of these compounds described to date involve multiple steps,
generate racemic products, and have final yields below 50% [6, 53–55]. Hence, our
panel of stand-alone TrpBs is a useful set of biocatalysts for the synthesis of ncAAs
and should be a fertile starting point for the further development of biocatalysts to
make more synthetically challenging ncAAs.
Table 1.1 Yields and total turnovers of the reactions catalyzed by TmTrpBM145T N167D to obtain
5-substituted tryptophan derivatives. The rates relative to PfTrpB0B2 are also included
TmTrpBM145T N167D
X OH or
X NH2
PfTrpB0B2
+ O
N H2N OH
H OH PLP, KPi buffer O
5% DMSO/H2O N
75 °C H
Reaction catalyzed by TmTrpBM145T N167D
X Rate relative to PfTrpB 0B2
Yield (%) Total turnovers
Cl 3.0 93 9300
Br 5.6 88 4400
NO2 7.5 25 1250
B(OH)2 1.8 38 1900
CHO 1.9 32 1600
CN 4.5 49 2450
OH 1.4 93 9300
CH3 1.4 91 9100
OCH3 1.5 76 7600
1.4 irected Evolution of PfTrpB for the Synthesis

D
of β-Branched Amino Acids
β-Branched amino acids are particularly desirable because the additional substitu-
ent at Cβ alters the conformational properties of peptides and small molecules that
bear it. For example, the clinically important antibiotic daptomycin features a
β-methylglutamate residue, and the activity of the drug is greatly diminished with-
out this modification [56, 57]. Unfortunately, chemical synthesis of β-branched
amino acids is particularly challenging, owing to the need for both enantio- and
diastereoselective transformations.
One strategy to access β-branched ncAAs would be to substitute Ser in the reac-
tion with Thr. While the nucleophile scope of TrpB has been well explored and our
panel of stand-alone variants can catalyze an array of substitutions, synthesis of Trp
analogs by replacing Ser had not been reported. We screened our stand-alone
PfTrpB catalysts for activity with a variety of Ser analogs and found that the natural
amino acid l-threonine (Thr) can replace Ser, yielding (2S,3S)-β-methyltryptophan
(β-MeTrp) in a single step (Fig. 1.5). Previous attempts to perform reactions with
Thr required the use of the strong nucleophile benzyl mercaptan, which proceeded
with greatly diminished activity compared to Ser, and the stereochemistry was
unknown [20].
β-MeTrp is an intermediate in the natural biosynthetic pathways to maremycin
and streptonigrin [7, 9]. Studies have shown that this intermediate is produced in
L-Threonine Indole (2S,3S)-β-MeTrp

O
O 4 3 OH
OH 5 TrpB
+ 2
6 NH2
NH2 N PLP
HO 7 H1
N
H
Fig. 1.5 β-Substitution of Thr with indole yields β-MeTrp. This reaction is catalyzed at trace
levels by PfTrpB, but the level increases with directed evolution. Activity was also observed with
a variety of indole analogs
four steps from Trp, using three different enzymes with S-adenosylmethionine as
the methyl source. Chemical synthetic routes to this ncAA have relatively good
selectivity but require multiple steps, making it challenging to apply for the produc-
tion of diverse β-MeTrp analogs. Therefore, we sought to evolve our platform for
the synthesis of these challenging β-branched amino acids using Thr as a nonnative
substrate.
We first explored the PfTrpB lineage of stand-alone enzymes for activity with
Thr. Wild-type PfTrpB yielded just 66 turnovers in 24 h. The single mutant
PfTrpBT292S showed ~sixfold enhanced activity, and reactions with variant PfTrpB4D11
(containing T292S, E17G, I68V, F274S, and T321A mutations) yielded 660 turn-
overs. Interestingly, PfTrpB0B2 gave slightly slower rates than PfTrpB4D11, despite
having higher activity with Ser [28]. Therefore, we selected PfTrpB4D11 for directed
evolution to increase activity with Thr.
We screened 352 clones from a random mutagenesis library of PfTrpB4D11 and
identified six missense mutations in five clones with increased Vmax. The most
active variant, PfTrpB4G1, has a single additional mutation, F95L. Recombination
of all of the mutations and screening resulted in variant PfTrpB2B9, which has muta-
tions I16V, F95L, and V384A and significantly enhanced activity with Thr.
Interestingly, the I16V mutation is adjacent to E17G present in the parent, and the
F95L mutation is adjacent to the COMM domain of TrpB. Combined, these muta-
tions provide at least a 6000-fold boost in productivity compared to the wild-type
PfTrpB enzyme.
This engineered enzyme has several positive features as a biocatalyst. As we
screened for Vmax, some of the increases in activity in cell lysates were due to boosts
in the expression of soluble enzyme. Hence, the protein can now be prepared at
~350 mg/L of culture, facilitating larger-scale reactions. The enzyme also retains
high activity at elevated temperatures, allowing for high substrate loading. However,
reactions with a single equivalent of each substrate give only 44% conversion to
products, which corresponds to 2220 total turnovers (TTN). UV-vis spectroscopy
showed that an abortive deamination reaction is occurring when Thr is incubated
with PfTrpB2B9. With the addition of more equivalents of Thr (up to ten), we
observed >99% conversion based on indole and up to 8200 TTN. Additionally, this
reaction proceeds with >99% enantiomeric and diastereomeric excess, highlighting

the exquisite selectivity of the enzyme.
Using these reaction conditions, we explored the nucleophile scope of this new
β-substitution reaction. Indoles with methylation at the 2 and 6 positions (number-
ing of indole is shown in Fig. 1.5) were well tolerated by the enzyme, yielding 6400
and 1100 turnovers, respectively. We also observed product formation with 4- and
5-fluoroindole, with ~3.4-fold lower TTN for the 4-fluoroindole, which is more
electron deficient at C-3 than the 5-fluoroindole analog. However, we did not
observe product formation using 5-chloro-, 5-bromo-, or 6-hydroxyindoles, demon-
strating a reduced substrate scope in the new reaction.
We observed that 7-azaindole reacted with 220 TTN and also found a secondary
product corresponding to N-alkylation. This regioselectivity is identical to that of
indazole, which we found reacts exclusively in an N-alkylation reaction that pro-
ceeds with 500 TTN. Lastly, we observed a productive reaction with thiophenol in
1300 TTN, demonstrating that the enzyme can catalyze stereoselective C-C, C-N,
and C-S bond-forming reactions.
The development of this enzymatic platform for the synthesis of β-branched ncAAs
was greatly facilitated by previous efforts to engineer a stand-alone enzyme. At all
stages, only a single enzyme was necessary to catalyze the reaction, and this simplified
system supported expression of the catalyst at high levels, ~350 mg of PfTrpB2B9 per L
culture, which will facilitate production of these important synthons [58].
1.5 Summary and Outlook
We have engineered the heterodimeric enzyme TrpS into a single enzyme platform
for the synthesis of ncAAs from simple starting materials. This was accomplished
by identifying mutations in TrpB that recapitulate the effects that are induced by its
allosteric binding partner, TrpA. We identified dozens of mutations that are activat-
ing in PfTrpB and showed that some of them could be transferred to homologs to
produce an array of catalysts with varied properties. These new stand-alone TrpB
enzymes can be evolved readily for altered function, as we demonstrated for the
β-substitution of Thr. All of this was accomplished with mutations that were identi-
fied by screening random mutant libraries, and the mutations are distributed
throughout the protein. Hence, each of the catalysts has a unique active site that can
be engineered to increase activity with a particular nucleophile or electrophile. We
believe this approach will continue to yield superior catalysts for the biocatalytic
production of desirable ncAAs.
Acknowledgments We gratefully thank Sabine Brinkmann-Chen for a critical reading of this

chapter. We also thank David K. Romney for his helpful discussions during the elaboration of the
chapter. Javier Murciano-Calles acknowledges financial support from the Alfonso Martín Escudero
Foundation. This work was funded through the Jacobs Institute for Molecular Engineering for
Medicine and Ruth Kirschstein NIH Postdoctoral Fellowship F32GM110851 (to Andrew
R. Buller).
References
1. Zhang J, Wu C, Sheng J, Feng X (2016) Molecular basis of 5-hydroxytryptophan synthesis in
Saccharomyces cerevisiae. Mol Biosyst 12(5):1432–1435
2. Ikeda M et al (1965) Studies on the biosynthesis of nicotinamide adenine dinucleotide: II. A
role of picolinic carboxylase in the biosynthesis of nicotinamide adenine dinucleotide from
tryptophan in mammals. J Biol Chem 240(3):1395–1401
3. Stepanova AN et al (2008) TAA1-Mediated auxin biosynthesis is essential for hormone cross-
talk and plant development. Cell 133(1):177–191
4. Tao Y et al (2008) Rapid synthesis of auxin via a new tryptophan-dependent pathway is
required for shade avoidance in plants. Cell 133(1):164–176
5. Barry SM et al (2012) Cytochrome P450-catalyzed l-tryptophan nitration in thaxtomin phyto-
toxin biosynthesis. Nat Chem Biol 8(10):814–816
6. Kieffer ME, Repka LM, Reisman SE (2012) Enantioselective synthesis of tryptophan deriva-
tives by a tandem Friedel-Crafts conjugate addition/asymmetric protonation reaction. J Am
Chem Soc 134(11):5131–5137
7. Kong D et al (2016) Identification of (2S,3S)-β-methyltryptophan as the real biosynthetic
intermediate of antitumor agent streptonigrin. Sci Rep 6:20273
8. Zehner S et al (2005) A regioselective tryptophan 5-halogenase is involved in pyrroindomycin
biosynthesis in Streptomyces rugosporus LL-42D005. Chem Biol 12(4):445–452
9. Zou Y et al (2013) Stereospecific biosynthesis of β-methyltryptophan from l-tryptophan fea-
tures a stereochemical switch. Angew Chem Int Ed 52(49):12951–12955
10. Lang K, Chin JW (2014) Cellular incorporation of unnatural amino acids and bioorthogonal
labeling of proteins. Chem Rev 114(9):4764–4806
11. Patel R (2013) Biocatalytic synthesis of chiral alcohols and amino acids for development of
pharmaceuticals. Biomolecules 3(4):741
12. Dunn MF (2012) Allosteric regulation of substrate channeling and catalysis in the tryptophan
synthase bienzyme complex. Arch Biochem Biophys 519(2):154–166
13. Corr MJ, Smith DRM, Goss RJM (2016) One-pot access to l-5,6-dihalotryptophans and l-
alknyltryptophans using tryptophan synthase. Tetrahedron 72(45):7306–7310
14. Ferrari D, Niks D, Yang L-H, Miles EW, Dunn MF (2003) Allosteric communication in the
tryptophan synthase bienzyme complex: roles of the β-subunit aspartate 305−arginine 141 salt
bridge. Biochemistry 42(25):7807–7818
15. Goss RJM, Newill PLA (2006) A convenient enzymatic synthesis of l-halotryptophans. Chem
Commun 47:4924–4925
16. Perni S, Hackett L, Goss RJ, Simmons MJ, Overton TW (2013) Optimisation of engineered
Escherichia coli biofilms for enzymatic biosynthesis of l-halotryptophans. AMB Express
3(1):1–10
17. Smith DRM et al (2014) The first one-pot synthesis of l-7-iodotryptophan from 7-iodoindole and
serine, and an improved synthesis of other l-7-halotryptophans. Org Lett 16(10):2622–2625
18.
Tsoligkas AN et al (2011) Engineering biofilms for biocatalysis. Chembiochem
12(9):1391–1395
19. Winn M, Roy AD, Grüschow S, Parameswaran RS, Goss RJM (2008) A convenient one-step
synthesis of l-aminotryptophans and improved synthesis of 5-fluorotryptophan. Bioorg Med
Chem Lett 18(16):4508–4510
20. Esaki N, Tanaka H, Miles EW, Soda K (1983) Enzymatic synthesis of S-substituted l-cysteines
with tryptophan synthase of Escherichia coli. Agric Biol Chem 47(12):2861–2864
21. Ferrari D, Yang LH, Miles EW, Dunn MF (2001) β-D305A Mutant of tryptophan syn-
thase shows strongly perturbed allosteric regulation and substrate specificity. Biochemistry
40(25):7421–7432
22. Niks D et al (2013) Allostery and substrate channeling in the tryptophan synthase bienzyme
complex: evidence for two subunit conformations and four quaternary states. Biochemistry
52(37):6396–6411
23. Barends TRM et al (2008) Structure and mechanistic implications of a tryptophan synthase
quinonoid intermediate. Chembiochem 9(7):1024–1028
24. Lai J et al (2011) X-ray and NMR Crystallography in an enzyme active site: the indoline qui-
nonoid intermediate in tryptophan synthase. J Am Chem Soc 133(1):4–7
25. Ngo H et al (2007) Allosteric regulation of substrate channeling in tryptophan synthase: modu-
lation of the l-serine reaction in stage I of the β-reaction by α-site ligands. Biochemistry
46(26):7740–7753
26. Bloom JD, Labthavikul ST, Otey CR, Arnold FH (2006) Protein stability promotes evolvabil-
ity. Proc Natl Acad Sci U S A 103(15):5869–5874
27. Lane AN, Kirschner K (1983) The catalytic mechanism of tryptophan synthase from

Escherichia coli. Eur J Biochem 129(3):571–582
28. Buller AR et al (2015) Directed evolution of the tryptophan synthase β-subunit for stand-alone
function recapitulates allosteric activation. Proc Natl Acad Sci 112(47):14599–14604
29. Caulkins BG et al (2015) Catalytic roles of beta Lys87 in tryptophan synthase: N-15 solid state
NMR studies. Biochim Biophys Acta Protein Proteomics 1854(9):1194–1199
30. Sol A, Tsai C-J, Ma B, Nussinov R (2009) The origin of allosteric functional modulation:
multiple pre-existing pathways. Structure 17(8):1042–1050
31. Gerek ZN, Ozkan SB (2011) Change in allosteric network affects binding affinities of PDZ
domains: analysis through perturbation response scanning. PLoS Comput Biol 7(10):e1002154
32. McLeish Tom CB, Cann Martin J, Rodgers Thomas L (2015) Dynamic transmission of
protein allostery without structural change: spatial pathways or global modes? Biophys
J 109(6):1240–1250
33. Suel GM, Lockless SW, Wall MA, Ranganathan R (2003) Evolutionarily conserved networks
of residues mediate allosteric communication in proteins. Nat Struct Mol Biol 10(1):59–69
34. Woods KN, Pfeffer J (2016) Using THz spectroscopy, evolutionary network analysis methods,
and MD simulation to map the evolution of allosteric communication pathways in c-type lyso-
zymes. Mol Biol Evol 33(1):40–61
35. Raboni S, Bettati S, Mozzarelli A (2005) Identification of the geometric requirements for allo-
steric communication between the α- and β-Subunits of tryptophan synthase. J Biol Chem
280(14):13450–13456
36. Weyand M, Schlichting I, Herde P, Marabotti A, Mozzarelli A (2002) Crystal structure of the
βSer178 → Pro mutant of tryptophan synthase: a “knock-out” allosteric enzyme. J Biol Chem
277(12):10653–10660
37. Shi L, Kay LE (2014) Tracing an allosteric pathway regulating the activity of the HslV prote-
ase. Proc Natl Acad Sci 111(6):2140–2145
38. Yoo SJ et al (1996) Purification and characterization of the heat shock proteins HslV and HslU that
form a new ATP-dependent protease in Escherichia coli. J Biol Chem 271(24):14035–14040
39. Gao X et al (2009) Directed evolution and structural characterization of a simvastatin synthase.
Chem Biol 16(10):1064–1074
40. Jiménez-Osés G et al (2014) The role of distant mutations and allosteric regulation on LovD
active site dynamics. Nat Chem Biol 10(6):431–436
41. Hettwer S, Sterner R (2002) A novel tryptophan synthase β-subunit from the hyperthermophile
Thermotoga maritima: quaternary structure, steady-state kinetics, and putative physiological
role. J Biol Chem 277(10):8194–8201
42. Hiyama T, Sato T, Imanaka T, Atomi H (2014) The tryptophan synthase β-subunit paralogs
TrpB1 and TrpB2 in Thermococcus kodakarensis are both involved in tryptophan biosynthesis
and indole salvage. FEBS J 281(14):3113–3125
43. Dunn MR, Otto C, Fenton KE, Chaput JC (2016) Improving polymerase activity with unnatu-
ral substrates by sampling mutations in homologous protein architectures. ACS Chem Biol
11(5):1210–1219
44. Khanal A, Yu McLoughlin S, Kershner JP, Copley SD (2015) Differential effects of a mutation
on the normal and promiscuous activities of orthologs: implications for natural and directed
evolution. Mol Biol Evol 32(1):100–108
45. Lehmann M, Wyss M (2001) Engineering proteins for thermostability: the use of sequence
alignments versus rational design and directed evolution. Curr Opin Biotechnol 12(4):371–375
46. Brinkmann-Chen S et al (2013) General approach to reversing ketol-acid reductoisomerase
cofactor dependence from NADPH to NADH. Proc Natl Acad Sci U S A 110(27):10946–10951
47. Kuriyan J, Eisenberg D (2007) The origin of protein interactions and allostery in colocaliza-
tion. Nature 450(7172):983–990
48. Merkl R (2007) Modelling the evolution of the archaeal tryptophan synthase. BMC Evol Biol
7(1):1–20
49. Murciano-Calles J, Romney DK, Brinkmann-Chen S, Buller AR, Arnold FH (2016) A panel
of TrpB biocatalysts derived from tryptophan synthase through the transfer of mutations that
mimic allosteric activation. Angew Chem Int Ed 55(38):11577–11581
50. Durak LJ, Payne JT, Lewis JC (2016) Late-stage diversification of biologically active mol-
ecules via chemoenzymatic C–H functionalization. ACS Catal 6(3):1451–1454
51. Pathak TP, Miller SJ (2013) Chemical tailoring of teicoplanin with site-selective reactions.
J Am Chem Soc 135(22):8415–8422
52. Roy AD, Grüschow S, Cairns N, Goss RJM (2010) Gene expression enabling synthetic diver-
sification of natural products: chemogenetic generation of pacidamycin analogs. J Am Chem
Soc 132(35):12243–12245
53. Blaser G, Sanderson JM, Batsanov AS, Howard JAK (2008) The facile synthesis of a series of
tryptophan derivatives. Tetrahedron Lett 49(17):2795–2798
54. Konda-Yamada Y et al (2002) Convenient synthesis of 7′ and 6′-bromo-d-tryptophan and
their derivatives by enzymatic optical resolution using d-aminoacylase. Tetrahedron 58(39):
7851–7861
55. Ma C, Liu X, Yu S, Zhao S, Cook JM (1999) Concise synthesis of optically active ring-A
substituted tryptophans. Tetrahedron Lett 40(4):657–660
56. Fowler VG et al (2006) Daptomycin versus standard therapy for bacteremia and endocarditis
caused by Staphylococcus aureus. N Engl J Med 355(7):653–665
57. Nguyen KT et al (2006) Combinatorial biosynthesis of novel antibiotics related to daptomycin.
Proc Natl Acad Sci U S A 103(46):17462–17467
58. Herger M, van Roye P, Romney DK, Brinkmann-Chen S, Buller AR, Arnold FH (2016)
Synthesis of β-branched tryptophan analogs with an engineered variant of tryptophan syn-
thase. J Am Chem Soc 138(27):8388–8391
Engineering Therapeutic Enzymes
2
Stefan Lutz, Elsie Williams, and Pravin Muthu
Abstract
Biologics constitute a rapidly growing category of pharmaceutical drug prod-
ucts. With over 100 clinically approved therapeutics and many more in develop-
ment, protein-based compounds represent an important subset among these
biomacromolecular drug candidates, ranging from antibodies, anticoagulants,
growth factors, hormones, interferons, and interleukins to enzymes. While
recombinant gene technology has traditionally played a key role in development
and production of these therapeutics, protein engineering offers an additional
dimension for tailoring the biochemical and biophysical properties of proteins to
the specific needs of clinical applications. Given the tremendous potential of
protein engineering methods to alter and improve the function of biocatalysts,
our review focuses on recent examples highlighting the advances and challenges
in applying these techniques toward the engineering of therapeutic enzymes.
More specifically, our review will focus on three categories of therapeutic
enzymes: pharmaceutical enzymes where the protein itself constitutes the thera-
peutic agent, prodrug-activating enzymes where the protein indirectly triggers a
clinical effect, and diagnostic enzymes where a protein’s superior selectivity and
specificity offer advantages over traditional analytical methods.
2.1 Introduction
Proteins are involved in the vast majority of biological processes inside and outside
the cellular environment. They have continuously been selected and optimized by
Darwinian evolution to facilitate specific chemical reactions over millions of years,
S. Lutz (*) • E. Williams • P. Muthu

Department of Chemistry, Emory University, 1515 Dickey Drive, Atlanta, GA 30084, USA

DOI 10.1007/978-3-319-50413-1_2
18 S. Lutz et al.
enabling them to perform structural, catalytic, and regulatory functions with impres-
sive efficiency, specificity, and selectivity. Recognizing the tremendous potential
and value of these biomacromolecules, efforts by scientists and engineers to harness
their exquisite performance to manufacture medicinal, industrial, and consumer
products date back over 100 years. While the scope of early studies was often lim-
ited by low availability of particular proteins, the development of recombinant DNA
technology in the early 1970s represented a paradigm shift, allowing not only for
the deliberate manipulation of genes that encode for particular proteins but also for
the reproducible and scalable heterologous expression of targeted proteins. These
capabilities marked a new era for fundamental and applied studies of native pro-
teins, and they also set the stage for a more creative, synthetic approach toward
studying and tailoring protein structure and function a decade later. As discussed
here and in other parts of this book, the generation of protein variants via site-
directed and random mutagenesis, as well as in vitro recombination, again revolu-
tionized the field, providing an effective strategy toward probing the contributions
of individual amino acids or protein domains to overall function. At the same time,
these methods could be applied for altering the functional properties of proteins, in
principle enabling the tailoring of proteins toward almost any desirable target
function.
Given their key role in the chemistry of life, proteins are also often intimately
involved in cellular malfunction that can lead to disease, as well as processes tar-
geted by drugs and toxins. In the spirit of “fighting fire with fire,” these same pro-
teins offer opportunities for the development of new, potentially highly effective
treatments and diagnostics to overcome, compensate for, or simply monitor pertur-
bations in a biological system. In fact, the exploitation of native and engineered
proteins for such therapeutic applications has been highly successful [1, 2]. In 2011,
almost 100 proteins were approved for clinical use in the United States and European
Union with sales exceeding US$100bln. When sorted by molecular mechanisms, a
clear majority of these protein-based therapeutics fall into the category of non-
covalent binders, consisting of genuine unmodified proteins and monoclonal anti-
bodies. Meanwhile, the category of proteins affecting covalent bonds listed 21
enzymes [2]. This latter category represents a particularly exciting group of thera-
peutics as enzymes, not limited to stoichiometric interactions, can due to their cata-
lytic activity dramatically amplify their impact on host cells. Consequently,
biological efficacy is enhanced even at low doses of the therapeutics, while costs
and undesirable side effects are minimized. However, therapeutic enzymes also
present unusual challenges as they are exogenous proteins, introduced into a host
cell to catalyze a reaction of clinical benefit. Such function comes with unique
design criteria distinct from general drug and protein design. On one hand, these
enzymes can be tuned to levels of specificity and selectivity typically not achievable
with traditional small-molecule drugs. On the other hand, they must minimize inter-
ference with native cellular functions, remain functional under physiological condi-
tions (e.g., blood plasma), exhibit suitable pharmacokinetic profiles, and elicit no or
2 Engineering Therapeutic Enzymes 19
minimal immunogenic response from patient. These additional constraints signifi-

cantly increase the complexity and challenge for engineering therapeutic enzymes.
It is for these reasons that the last two decades have seen serious efforts to identify
novel biocatalysts for therapeutic applications that far exceed the abovementioned
list of enzymes with clinical approval.
In this article, we are reviewing developments of therapeutic enzymes with a
particular focus on studies involving engineering of the actual protein. Additional
opportunities for functional improvements of proteins including posttranslational
modifications such as PEGylation or fusion to other proteins are not discussed.
While some of the enzymes discussed have been approved or are currently in clini-
cal testing, others remain under development in the laboratory or have been discon-
tinued. We believe that they all represent interesting examples of the challenges and
opportunities for tailoring enzymes for therapeutic applications. In addition, many
studies provide valuable insights into the protein’s structure-function relationships.
For our discussion, we have organized these enzymes into three subcategories:
pharmaceutical enzymes, prodrug-activating enzymes, and diagnostic enzymes.
2.1.1 Pharmaceutical Enzymes
The subcategory of pharmaceutical enzymes represents proteins that themselves

constitute the therapeutic agent. Representatives include proteases, esterases, and
nucleases, as well as amino acid-processing enzymes. The clinical efficacy of these
enzymes typically benefits from enhanced catalytic activity and pharmacokinetics,
as well as reduced immunogenicity. While improvements of activity and stability
are relatively straightforward and are routinely addressed by directed evolution
methods [3, 4], immunogenicity remains the most common reason for drug failure.
Historically, the majority of preclinical research on therapeutic enzymes were done
with nonhuman proteins and often resulted in termination of clinical trials due to an
observed immune response [5]. In some cases, simply altering the protein origin to
the human homolog was sufficient. For example, initial studies on DNaseI as a
therapeutic agent to treat cystic fibrosis were done with the bovine enzyme [6].
Pulmozyme® (Genentech) is its human homolog and displayed the intended thera-
peutic effect without observed immunogenicity [7]. For many other enzymes, the
reduction of immune response is not straightforward and remains a significant chal-
lenge. While various approaches to identify potential epitopes have been explored,
no general solution to protein immunogenicity exists to date.
2.1.2 Enzymes for Suicide Gene Therapy
A more recent application of therapeutic enzymes is toward the development of

improved chemotherapies through the introduction of foreign genes. For
20 S. Lutz et al.
decades, chemotherapy has been a component of standard of treatment for can-

cer patients [8]. However, the high systemic toxicity of the therapeutic agents
remains a dose- limiting constraint. Gene-directed enzyme-prodrug therapy
(GDEPT) involves the intratumoral delivery and expression of so-called suicide
genes, encoding for enzymes which, in combination with a prodrug, elicit a
cytotoxic effect [9]. First introduced by Mootlen et al. [10], GDEPT places
strict constraints on the therapeutic enzyme, requiring high activity and speci-
ficity for the unnatural prodrug but no significant interference with native cel-
lular functions. For these reasons, wild-type enzymes were generally found to
be poor GDEPT candidates [11]. Protein engineering of native enzymes has
been proposed for GDEPT and has been advanced in four specific cases:
2′-deoxyribonucleoside kinases and purine nucleoside phosphorylase that oper-
ate in conjunction with nucleoside analog prodrugs, cytosine deaminase that
utilizes prodrugs originally developed as fungicides, and nitroreductases that
activate custom prodrugs. Beyond these four extensively studied enzyme-
prodrug systems, more than 20 other GDEPT systems deploying natural enzymes
are described in the literature [12, 13].
2.1.3 Diagnostic Enzymes
The inherent selectivity and specificity of enzymes can offer significant advantages
over traditional analytical methods for the detection of analytes and biomarkers.
The functional properties of such enzyme-based biosensors not only allow for direct
identification of targeted molecules in complex biological samples; the sensor-
target interactions also impacts the catalytic performance of the enzyme, amplifying
the effect of a single, bound analyte by manyfold and dramatically boosting sensor
sensitivity. Besides the need for increased stability which is typically addressed by
enzyme immobilization, engineering of diagnostic enzymes has largely focused on
changing and optimizing the specificity for particular analytes, as exemplified by
variants of cholinesterases for the detection of organophosphate and carbamates
found in pesticides and nerve agents.
2.2 Proteases
Proteases play key roles in regulating fundamental biological processes, approxi-

mately 2% of the human genome encodes for protease enzymes, and they are the
molecular targets for many established drug classes [14]. Perhaps not surprising
then is the variety of medical conditions for which researchers are pursuing engi-
neered proteases as therapeutic enzymes. These include leukemia [15], vitreomacu-
lar adhesion [16], and, as will be expanded upon below, cardiovascular disorders,
multiple neurological disorders, and celiac disease.
2.2.1 P
roteases as Anticoagulant Enzymes for Treatment
of Cardiovascular Disorders
2.2.1.1 Urokinase Plasminogen Activator and Tissue Plasminogen

Activator
The use of proteases to treat cardiovascular disorders is the most established thera-
peutic use of this enzyme class. Urokinase plasminogen activator (u-PA) was the
first enzyme to be approved for therapeutic use by the FDA in 1978 [17] not only
ushering in proteases as a class of drugs for thrombolytics but demonstrating the
commercial viability of enzymes as drugs. u-PA is able to break down blood clots
as its proteolytic activity cleaves the zymogen plasminogen into the active serine
protease plasmin. Active plasmin then cleaves fibrin, a fibrous protein that is a cru-
cial competent of blood clots, eventually dissolving the clot. While u-PA was a
significant advance for proteolytic-based therapy, it is a tissue plasminogen activa-
tor (t-PA), first trailed in humans in 1984 [18] and approved for myocardial infarc-
tion and eventually for stroke, that is now more widely used. t-PA also converts
plasminogen to active plasmin, but it preferentially cleaves plasminogen that is
bound to fibrin, allowing t-PA to be systemically delivered but still effect local fibri-
nolysis at indicated areas [19]. However, t-PA is rapidly inactivated by endogenous
inhibitors, most importantly plasminogen activator inhibitor-1 (PAI-1), as part of a
natural feedback regulation to prevent against rampant fibrinolysis [20], and recom-
binant infused t-PA is reported to have a biological half-life of only 6 min [21].
Improving both the biological half-life and specificity toward fibrin bound plas-
minogen were clear aims for which to engineer second-generation t-PA proteases.
In 2000, the FDA approved tenecteplase (TNKase®) developed by Genentech. This
improved variant contains three modifications from the wild-type t-PA sequence
[22]. The first modification was a replacement of the residues at positions 296–299
with Ala, identified via a strategy in which charged residues close together in the
primary sequence of t-PA were systematically substituted for Ala [23]. This variant
shows sharply decreased PAI-1 inhibition and increased specificity and activity for
plasmin bound to fibrin. Additional mutations T103N [24] and N117Q modified
N-linked glycosylation sites and reduced plasma clearance [22]. These sequence
modifications resulted in a modest improvement of half-life, from 6 to 18 min in
clinical trials [21]. Unfortunately, not all attempts to improve t-PA bioavailability
have been successful. The variant lanoteplase showed both an increased half-life
and a decreased PA1-I binding [25]; however, when tested in large-scale clinical
trial, this appears to have triggered an increase in intracranial bleeding that halted
development of the protease [25], highlighting the unfortunate reality that even
when an engineering campaign successfully meets an identified target, the resulting
therapeutic enzyme is not assured of in vivo clinical success.
2.2.1.2 Engineered Thrombin

Thrombin is a protease that can trigger both procoagulation and anticoagulation
processes. In an unbound state, thrombin proteolytically converts fibrinogen into
22 S. Lutz et al.
fibrin allowing the formation of blood clots. Fibrin glue, a combination of fibrino-
gen and native thrombin enzyme, has a long-standing use as a surgical aid to wound
healing [26]. However, upon binding to the endothelial cell receptor thrombomodu-
lin, thrombin “switches” to an anticoagulant activity as it can now activate the
zymogen protein C, converting it into activated protein C (APC) that continues the
protease cascade and triggers downstream anticoagulation pathways. Multiple engi-
neering efforts have focused on separating out these anti- and procoagulant activi-
ties to develop thrombin as an anticoagulant protease for potential therapeutic use
[27–29]. One variant of interest is a double mutant W215A/E217A that has demon-
strated efficacy as an anticoagulant in both preclinical primate [30] and rodent mod-
els [31] of ischemic stroke and thrombosis. This rationally designed variant
combined two single mutations (W215A [32] and E217A [27]) that had previously
been characterized as interfering with proteolysis of fibrinogen but not protein
C. The double mutant shows a 20,000-fold lower activity for fibrin yet still activates
protein C comparably to wild-type enzyme.
2.2.2 Procoagulant Protease Engineering
Factor VII (FVII) is a protease in the procoagulant protease cascade required to

form blood clots. Recombinant FVIIa, marketed as NovoSeven® by Novo Nordisk,
was approved by the FDA for the treatment of hemophilia in 1999. While the
wild-type formulation has been demonstrated as a safe and effective treatment for
procoagulation, there is a potential market for variants with increased and sus-
tained procoagulant activity. Persson et al. consolidated rationally identified sub-
stitutions from previous works, resulting in a triple mutant (V158D/E296V/
M298Q) of FVIIa [33]. The activity of the wild-type FVIIa is stimulated by asso-
ciation with tissue factor (TF), which is locally present at sites of vascular injury.
The triple mutant has amino acid substitutions that force the protease to adopt a
conformation similar to the TF-induced structure, increasing the TF-independent
activity of the enzyme. Unfortunately, in phase III clinical trials, several patients
developed anti-drug antibodies eliciting a neutralizing response, and development
of this variant has been halted [34]. Harvey et al. also used protein engineering to
improve FVIIa for a separate property [35]. The wild-type FVIIa has a relatively
low affinity for the membranes of platelets, reducing its in vivo efficacy. The
group was guided by previous literature, suggesting that membrane affinity of
FVIIa was mediated through the the γ-carboxyglutamic acid (Gla) domain [36].
Site-saturation mutagenesis of this 40 amino acid residue domain was carried out
and a light-scattering assay used to evaluate proteins for increase binding to phos-
pholipid vesicles. The resulting lead variant contained four substitutions (P10Q/
K32E/D33F/A34E) and progressed to phase III clinical trials for the treatment of
hemophilia. This study was also halted, as patients again developed neutralizing
antibodies toward the engineered protease [37].
The clinical outcome of the modified FVIIa variants highlights a difficulty of

engineering enzymes as therapeutic agents. Both studies were halted as a precau-
tionary measure due to an unforeseen immune response, resulting from genetic
modification of a human wild-type sequence. The development of non-immunogenic
enzyme variants with alternate properties remains a consistent challenge for the
engineering of therapeutic enzymes.
2.2.3 Alzheimer’s Disease
A role for therapeutic proteases in the treatment of multiple neurological disorders

has been investigated. One such disorder is Alzheimer’s disease. While the molecu-
lar basis for Alzheimer’s is not fully understood, the disease is characterized by
buildup of β-amyloid (Aβ) peptide-based plaques in the brain. Many therapeutic
strategies in advanced clinical trials seek to target these Aβ plaques through immu-
notherapy or inhibition of the secretases that produce them from amyloid precursor
protein [38]. An alternative strategy is the use of proteases to break down Aβ
plaques – either through pharmacological upregulation of endogenous protease
activity [39] or gene therapy-based approaches designed to increase expression of
Aβ plaque degrading proteases [40, 41]. A large concern, and a general hurdle for
many protease-based therapies, is off-target toxicity arising from the often broad
substrate specificity of proteases. As such, increasing specificity is a principle aim
of many protease engineering campaigns, such as those focused on increasing the
specificity of neprilysin [42, 43], a human zinc metalloprotease able to degrade Aβ
plaques. Site-directed mutagenesis of active site and solvent accessible residues was
carried out, and beneficial single mutations were identified by screening for
increased activity with Aβ peptide sequences and decreased activity on eight known
native peptide substrates. These were then screened combinatorially resulting in a
lead variant with two mutations, G399V/G174K, that showed a 20-fold increase in
kcat/Km relative to wild-type neprilysin and a 2.6- to 3200-fold decrease in kcat/Km on
a panel of native peptide substrates [44]. In separate work, a member of the trypsin-
like serine protease superfamily, human kallikrein 7 (hK7), has also demonstrated
an in vitro ability to cleave peptides at a sequence in the core of the Aβ peptide, and
Guerrero et al. reported on efforts to increase this catalytic activity and specificity
toward Aβ by engineering the protease to prefer phenylalanine upstream of the pep-
tide cleavage site instead of the native tyrosine preference [45]. In a yeast-based
expression system, a randomized 107-sized mutant library was screened using a
protease-activated fluorescent reporter. The resulting lead hK7 variant showed a
10–30-fold increase in selectivity toward the Aβ peptide sequence versus native
peptide substrates. However, this was due primarily to the loss of activity toward
tyrosine-containing native peptide sequences rather than increased Aβ peptide
activity. The variant demonstrated reduced toxicity toward both the yeast expression
system and neuronal cells co-cultured with purified protein.
24 S. Lutz et al.
2.2.4 B
otulinum Neurotoxins for Neurological
and Non-neurological Indications
In 1989, botulinum neurotoxin serotype A (BoNT/A) from Clostridium botulinum

was approved for use by the FDA for the treatment of strabismus (cross eyes) and
blepharospams (eyelid spasms), although it is now perhaps most commonly known
for its cosmetic use in attenuating frown lines. BoNT/A is currently also used in a
wide number of approved and off-label indications including dystonias, migraines,
overactive bladder, excessive sweating or drooling, and muscle conditions associ-
ated with diseases such as cerebral palsy, Parkinson’s disease, and multiple sclerosis
[46, 47]. The molecular basis of the neuro-inhibitory effect of BoNTs is due to a
targeted protease catalytic activity. The mature form of BoNTs contains three
domains – a light chain zinc endopeptidase domain attached via a disulfide bond to
a heavy chain-binding domain and translocation domain. These allow the BoNTs to
bind to peripheral cholinergic nerve terminals and eventually enter the nerve cell
cytosol where the zinc endopeptidase then cleaves specific N-ethylmaleimide-
sensitive fusion protein attachment protein receptor (SNARE) proteins. As these
SNARE proteins are essential for docking and fusion of synaptic vesicles, this
impairment of SNARE function inhibits acetylcholine release and consequently
neurotransmitter signal transduction. While no recombinant BoNTs are currently
approved as therapeutics, there are ongoing efforts to engineer BoNTs. In-depth
coverage of this body of work is beyond the scope of this review but has recently
been comprehensively reviewed by Masuyer et al. [48]. BoNT modifications have
been made with a wide range of aims including to improve safety of current treat-
ments [49], to increase efficacy [50, 51], or to modify cell targeting so as to expand
the therapeutic use of BoNTs to other neurons or to entirely non-neurological con-
ditions. This engineering has involved both amino acid substitutions in the catalytic
protease domain and exploitation of the modular nature of the multi-domain BoNTs
to “switch out” different translocations and/or binding domains both from other
BoNT serotypes [51, 52] or to replace them with completely unrelated cell-binding
proteins [53, 54], including antibodies [55] (Table 2.1).
Examples of BoNT domain engineering include the work of Wang et al. in which
a BoNT/A variant was generated with an inactive protease domain and an artifi-
cially attached active endopeptidase domain from a serotype E BoNT [52]. This
hybrid variant was designed to combine the long-acting delivery of BoNT/A pro-
teins with the protease action of BoNT/E, which by cleaving at a different site than
BoNT/A in its SNARE target (SNAP25) gives greater potential for pain alleviation
than the activity of native BoNT/A. In this example the hybrid BoNT was still tar-
geted toward a neuronal cell; however, it has been demonstrated that BoNTs can be
retargeted toward non-neuronal cell types when attached to an appropriate protein
partner. These potential therapeutic proteins are designed on the basis that SNARE
proteins are essential not just for release of neurotransmitter vesicles but also in
vesicle-based secretion pathways of non-neuronal cell types. That BoNTs protease
activity could be retargeted was first demonstrated by an in vitro chemical attach-
ment of the endopeptidase domain of BoNT/A to a wheat germ agglutinin protein
Table 2.1 Summary of engineered proteases
Therapeutic use Enzyme Modification Properties and examples of clinical use Therapeutic product Reference
Anticoagulant Urokinase (None) Fibrinolysis for treatment of pulmonary Abbokinase (approved
embolism 1983)
t-PA (None) Fibrinolysis for treatment of pulmonary Alteplase/Activase®
embolism, acute ischemic stroke and (approved 1987)
myocardial infarction
t-PA T103N/N117Q/K296A/ Reduced in vivo inhibition and Tenecteplase/TNKase Keyt (1994) [22]
H297A/ R298A/R299A increased half-life (approved 2000)
t-PA N117G, deletion of Increased half-life Lanoteplase Larsen (1991) [56]
fibronectin and epidermal (discontinued)
growth factor domain
2 Engineering Therapeutic Enzymes
Thrombin W215A/E217A Decreased procoagulant fibrinogen Cantwell (2000)

activation but maintains protein C [30]
activation, triggering anticoagulant
protease cascade
Procoagulant Thrombin (None) Activates fibrinogen, promoting Recothrom® (approved
coagulation at damaged tissues for 2008)
wound treatment
FVIIa (None) Induced by tissue factor, a membrane NovoSeven® (approved
signal protein, to promote coagulation 1999)
for hemorrhage treatment
FVIIa V158D/E296V/M298Q Mutations force FVIIa into tissue factor NN1731 Persson (2001)
activated conformation (discontinued) [33]
FVIIa P10Q/K32E/D33F/A34E Improved affinity for membranes and BAY 86–6150 Mahlangu (2016)
tissue factor activation (discontinued) [37]
Alzheimer’s Neprilysin G399V/G174K Increased specificity for proteolysis of Webster (2014)
disease β-amyloid peptide sequences [43]
Human Increased specificity for proteolysis of Guerrero (2016)
Kallikrein 7 β-amyloid peptide sequences [45]
25
(continued)
26
Table 2.1 (continued)
Therapeutic use Enzyme Modification Properties and examples of clinical use Therapeutic product Reference
®
Neuromuscular BoNT/A (None) Cleaves SNAP-25 complex, inhibiting Botox (approved
disorders neurotransmission for treatment of 1989)
muscle dystonias and multiple other
medical conditions
BoNT/A L256E/V258P and A308L Altered specificity, instead cleaving Sikorra (2016)
SNAP-23 complex to potentially inhibit [57]
pathogenic secretion in non-neuronal
cells
BoNT/E K224D Altered specificity, instead cleaving Chen (2009) [58]
SNAP-23 complex to potentially inhibit
pathogenic secretion in non-neuronal
cells
BoNT/A and E Fusion of BoNT/E protease Longer lasting pain attenuation Wang (2011) [52]
hybrid domain and BoNT/A
translocation and binding
domains
BoNT/D- Fusion of BoNT/D protease Inhibition of pituitary growth hormone Somm (2012) [59]
GRRH hybrid and translocation domain to hypersecretion in animal model of
GRRH binding peptide acromegaly
Celiac disease Endoproteases (none) Combination of two unmodified ALV003 (phase II) Tye-Din (2010)
EP-B2 and glutamine and proline-specific [60]
SC-PEP proteases to break down dietary gliadin
SC-PEP I581V/F459Y/M511L/I406V Improved activity at low pH and Ehren (2008) [61]
resistance to pepsin degradation
Kumamolisin V119D/S262K/N291D/ Improved specificity for (PQ) dipeptide Gordon (2012)
D293T/G319S/D358G/ motif [62]
D368H
S. Lutz et al.
that has a generic ability to be internalized in vitro by multiple neuronal cell types
and by pancreatic B cells. Once internalized, the BoNT/A-containing protein atten-
uated insulin secretion [53]. Following this, more specific cell targeting for thera-
peutic benefit has been investigated. For example, in pursuit of a therapy to decrease
the pituitary growth hormone hypersecretion seen in acromegaly, an adult-onset
endocrine disease, the endopeptidase and translocation domain of BoNT serotype D
was attached to a peptide that binds to growth hormone-releasing hormone recep-
tors (GRRH). This led to in vitro internalization of the hybrid protein into GRRH-
containing cells and cleavage of the target intracellular SNARE. Rats injected with
the hybrid protein showed markedly attenuated growth hormone secretion and syn-
thesis [59]. The potential expansion of BoNT-based therapies has also been pursued
by amino acid substitution of residues within the endopeptidase domain to change
the specificity toward the targeted SNARE from the neuronal SNAP25 cleaved by
BoNT/A to that of the non-neuronal SNAP23 [57, 58]. In the first engineering effort
of this type, Chen and Barberi used rational engineering to identify one amino acid
substitution, K224D, that could be made to the endopeptidase domain of BoNT/E to
enable it to cleave not just SNAP25 but also SNAP23.
2.2.5 Digestive Proteases
In a separate context, proteases have also been investigated for the treatment of
celiac disease. Celiac disease is a chronic illness in which dietary gluten triggers an
autoimmune response and subsequent inflammation and damage of the intestine in
genetically susceptible individuals. The current treatment is elimination of gluten
from the patient’s diet entirely, as no curative therapy exists [63]. More specifically
this inflammatory immune response is triggered by peptides that result from the
incomplete digestion of gliadin – gliadin being a significant protein component of
gluten. Due to its high proline and glutamine content, gliadin is unusually resistant
to breakdown by human digestive proteases [64]. A potential therapeutic treatment
is the digestion of these immunogenic peptides by exogenously delivered proteases.
Currently the enzyme mix ALV003, a combination of EP-B2 (a protease expressed
in germinating barley seeds) and SC-PEP (a proline-specific protease from the bac-
teria Sphingomonas capsulata), is in phase II clinical trials as an oral enzyme ther-
apy [60]. A therapeutically useful protease would need to be delivered orally, be
stable and active in the low pH stomach environment, be resistant to degradation
from endogenous proteases, and show sufficient activity and specificity toward the
disease-triggering proline- and glutamine-rich oligopeptides, such that these oligo-
peptides are degraded even when present with other dietary proteins that could com-
pete as substrates. In addition to examination of native proteases – such as with
ALV003 – there have been efforts to use enzyme engineering to improve upon some
of these features.
Ehren et al. focused on SC-PEP, the proline-specific protease being trialed as part
of ALV003 [61]. Their main aims were to improve SC-PEP activity in the
28 S. Lutz et al.
physiologically relevant, acidic environment (pH 4.5) and to increase resistance to

pepsin degradation. They used structural data, published literature, and sequence
analysis of 100 PEP homologs to identify 30 target amino acid substitutions that
were then combined to generate a small enzyme library of 47 SC-PEP variants –
each carrying three to five individual amino acid substitutions. These variants were
tested for protease activity at pH 4.5 and in the presence of pepsin. The contribution
of each individual amino acid substitution to these factors was scored and this infor-
mation used to select the mutations that were then recombined to build a second
library of 48 variants. Of these, 48% were more resistant to pepsin degradation than
wild-type SC-PEP, and 54% showed a greater than 10% increase in degradation of
the model peptide at pH 4.5. This demonstrates the increase in overall library qual-
ity that can be gained from empirical testing of even a small number of mutations:
in the first round of library testing, only 13% of library members demonstrated
increased pepsin resistance and 14.8% showed improved activity.
In a separate approach, Gorden et al. engineered the protease kumamolisin from
the acidophilic bacterium Alicyclobacillus sendaiensis [62]. Database searches
identified this enzyme as a promising initial candidate for engineering as it is
highly active at pH 2–4 and a temperature of 37 °C, cleaves peptides after the
dipeptide recognition sequences Pro/Arg or Pro/Lys, and exhibits slight activity for
the motif common in gliadin, Pro/Gln. The RosettaDesign software was used to
model the tetrapeptide Pro/Gln/Leu/Pro into the binding pocket of the crystal
structure of kumamolisin. This leads to the design of a library of 261 variants car-
rying various substitutions selected from examining 75 residues lining the enzyme
active site. While 20% of the library showed a decreased ability to cleave the
model substrate, 30% retained wild-type activity and 50% showed an activity
increase. The most active variant (V119D/S262K/N291D/D293T/G319S/D358G/
D368H) showed 100-fold greater proteolytic activity for a model Pro-Gln-rich
peptide as compared to the native enzyme and roughly 800-fold increased specific-
ity as it no longer cleaved a model peptide substrate containing the native Pro/Arg
motif. In an extension of this work, the same group again used Rosetta software to
further enhance the activity of this improved kumamolisin variant against two glia-
din peptides known to be highly immunogenic – one rich in Pro/Gln/Gln, the other
in Pro/Gln/Leu [65].
2.2.6 Protease Engineering Strategies
In regard to protease engineering in general, a recent review highlights a range of

powerful high-throughput strategies that have been used to engineer protease activ-
ity [66]. These include phage-based systems [67] and FACS with protease-activated
fluorescent reporters [44, 45, 68] or antibiotic resistance proteins [69]. While not
directly focused on engineering of proteases for therapeutic applications, it should
be noted that these strategies developed from tuning substrate specificities or
increasing catalytic efficiency of model proteases such as the E. coli OmpT [70] or
the tobacco protease TEV [69, 71, 72] may in the future be applied to the challenges
of engineering proteases as therapeutics. Equally interesting is the work of research-

ers using enzyme engineering tools to elucidate the mutational paths through which
proteases can evolve resistance to protease inhibitor drugs [73].
2.3 Ribonucleases
Ribonucleases (RNases) have been investigated as anticancer therapeutics for over

60 years [74]. Differences in surface structures between normal and malignant cells
are thought to be responsible for the selective intake of exogenous RNases and con-
sequent interference with RNA metabolism. Specifically, the enzymes’ cytotoxic
properties are conferred by rapid RNA degradation in targeted cells, stalling cell
cycle progression and leading to apoptosis. To date, the RNase therapeutic to
advance furthest in clinical trials is ranpirnase (Onconase), an unmodified enzyme
from the RNase A superfamily isolated from the Rana pipiens frog. Onconase
showed promising results when used to treat malignant mesothelioma in phase II
trials but subsequently failed in advanced tests [75, 76]. Nevertheless, it is currently
still under clinical investigation as a possible antiviral for the treatment of human
papillomavirus.
To increase the therapeutic effects with respect to current and future therapeutic
applications of RNases, protein engineering has been used to generate variants with
decreased binding affinities toward cytosolic ribonuclease inhibitor protein (RI)
[77]. RI is a 50-kDa horseshoe-shaped protein that binds extremely tightly to many
RNase A family members and greatly diminishes RNase cytotoxicity [78].
Engineered RNases that maintain catalytic activity but exhibit decreased binding to
human RI include mutants of the human enzymes RNase 1 and RNase 5 (angio-
genin), as well as bovine pancreatic RNase A [79–84]. Employing rational design
strategies, the examination of crystal structures of RNase-RI complexes allowed
researchers to identify key residues at the protein-protein interface. Amino acid
substitutions at these positions reduce the binding affinity of RI and consequently
minimize RNase inhibition. More recently, these engineered human RNases have
also been investigated in combination with cancer-targeting antibodies as non-
immunogenic, cytotoxic warheads [84, 85].
2.4 Amino Acid-Degrading Enzymes
A number of enzymes have been investigated as potential anticancer agents due to

their ability to metabolize amino acids. The mechanistic basis underlying the thera-
peutic effect is a depletion of the endogenous amino acid pool. The resulting auxot-
rophy in one or more specific amino acids forces the tumor to rely on exogenous
sources to supplement the limiting amino acids or else halt protein synthesis, which
triggers apoptosis in the malignant cells (for recent reviews see [86, 87]). Of par-
ticular interest for this resource-depletion strategy are enzymes for the degradation
of L-arginine (Arg) and L-asparagine (Asn).
30 S. Lutz et al.
2.4.1 Asparagine-Degrading Enzymes
Asparaginase II from E. coli (EcAII) is currently a key component in the treatment

of childhood acute lymphoblastic leukemia (ALL). Marked under the trade name
Elspar [88], EcAII hydrolyzes Asn to Asp and ammonia, which results in depletion
of serum levels of Asn and triggers lymphoblast apoptosis. Consequently, overall
survival rates of childhood acute lymphoblastic leukemia are relatively high at close
to 90% [89], yet immunogenicity of the bacterial enzyme can cause undesirable side
effects [90] and renders some patients ineligible or unable to continue treatment. In
attempts to alleviate these effects, Cantor et al. pursued an engineering strategy
aimed at generating an EcAII mutant with decreased immune response while retain-
ing catalytic activity [91]. Starting with in silico analysis of EcAII to predict three
putative T-cell epitopes, four-residue fragments within each epitope were then sub-
jected to saturation mutagenesis to disrupt interactions with major histocompatibil-
ity complex (MHC-II). The resulting mutant libraries were overexpressed in E. coli
and enriched for enzymes retaining Asn hydrolase activity using a neutral drift strat-
egy. More specifically, the selection of E. coli expressing active EcAII variants was
based on the host cell’s ability to hydrolyze Asn to Asp, which was required for
concomitant expression of GFP. After targeting each predicted epitope, the lead
EcAII mutant showed eight amino acid changes. Although catalytic performance
for the mutant was slightly impaired as reflected by the threefold drop in specific
activity due to an increase in the apparent Michaelis-Menten constant, this variant
did display decreased immunogenicity. Tests in a transgenic mouse model that
expressed human leukocyte antigen-II molecules indicated a tenfold reduction in
anti-EcAII IgG levels as compared to wild-type enzyme. Separately, increased ther-
apeutic benefit has also been pursued via mutagenesis studies on asparaginase
homologs from other prokaryotes [92]. These studies included investigating the
contribution of secondary glutaminase activity in Helicobacter pylori asparaginase
mutants to cytotoxicity, as well as on increasing the catalytic activity and cytotoxic-
ity of the thermostable Pyrococcus furiosus asparaginase [93]. Finally, efforts have
been reported to generate a thermostable variant of the Erwinia chrysanthemi aspar-
aginase [94]. The wild-type enzyme is currently used as a second-line therapeutic
treatment for ALL [95].
2.4.2 Arginine-Degrading Enzymes
Arginine depletion has also been investigated, in particular for the treatment of
hepatocellular carcinomas and melanomas lacking the enzyme argininosuccinate
synthetase 1 [87]. Of interest are a number of prokaryotic arginine deiminases
(ADIs) that hydrolyze Arg to citrulline and ammonium. Most clinically advanced is
a mycobacterium ADI in PEGylated form [96–99]. To date however the majority of
engineering efforts have focused on the more recently characterized highly cyto-
toxic PpADI from Pseudomonas plecoglossicida, thoroughly reviewed by Han
et al. [100, 101]. Targets for protein engineering have included improving catalytic
performance and thermostability, as well as shifting the enzyme’s pH optimum from

its native pH 6 closer to physiological conditions to prevent dramatic activity losses
observed for wild-type PpADI [102]. These engineering efforts have been aided by
a newly developed screening assay that can assess PpADI activities at physiologi-
cally relevant low levels of Arg (~100 μM) [103]. Briefly, PpADI mutants were
expressed in an E. coli strain which controls GFP expression via a promoter that can
be repressed by Arg. Host cells carrying improved ADI variants hence could easily
be identified as they showed enhanced GFP expression and could be enriched by
FACS. This high-throughput assay allowed for the screening of about eight million
variants over three iterative rounds of directed evolution to arrive at a mutant with
an almost three orders of magnitude higher kcat/S0.5 compared to wild-type enzyme.
Furthermore, these engineered ADIs possessed significant catalytic activity at phys-
iological arginine levels (100 μM), whereas the native enzyme showed no detect-
able activity at this substrate concentration.
Addressing concerns over immunogenicity of prokaryotic ADIs in therapeutic
applications, human enzymes that can metabolize Arg have also been investigated.
While mammals do not have direct ADI homologs, the human enzyme arginase I
(hArgI) that hydrolyzes Arg to urea and ornithine was identified as a possible sub-
stitute [104]. However, preliminary studies of the native enzyme revealed Km values
for arginine in the millimolar range. Furthermore, the pH optimum of hArgI near
pH 9.5 is outside the physiological range [105]. A less conventional but effective
strategy to overcome these limitations has focused on the enzyme’s cofactor rather
than amino acid mutagenesis [106]. Native hArgI requires a Mn2+ cofactor to gener-
ate a metal-activated water for attack on the guanidinium carbon of Arg. Cofactor
replacement with Co2+ resulted in a pKa shift by one pH unit and generated an
enzyme variant with a tenfold increase in kcat/Km at pH 7.4. Separately, serum half-
life was markedly improved by PEGylation of the enzymes [107].
2.5 Deoxyribonucleoside Kinases
Beyond applications of engineered enzymes as therapeutic agents themselves, their

potential role as prodrug-activating catalysts has flourished over the last decade.
One of the examples are 2′-deoxyribonucleoside kinases (dNKs), enzymes that are
part of the nucleoside salvage pathway and play a critical role in therapeutic appli-
cations of nucleoside analog (NA) prodrugs [108]. The salvage pathway enables
mammalian cells to recycle scavenged 2′-deoxynucleosides from the environment
via transmembrane uptake and three consecutive, dNK-catalyzed phosphoryl-
transfer steps [109]. The process complements de novo nucleotide biosynthesis to
supply triphosphate anabolites for DNA replication and other cellular functions. In
addition, the salvage pathway is responsible for the intracellular activation of NAs,
synthetic mimics (Fig. 2.1) of the natural DNA building blocks that upon phos-
phorylation to their corresponding triphosphates turn into competitive substrates for
low-fidelity DNA polymerases and reverse transcriptase found in cancer cells and
viruses, respectively [110].
32 S. Lutz et al.
Fig. 2.1 Nucleoside analogs as prodrugs for antiviral and cancer therapy: aciclovir (1), ganciclo-
vir (2), AZT (3), d4T (4), ddT (5), and L-FMAU (6)
Problems with NA activation arise due to the high substrate specificity of the
host’s endogenous kinases in general and the initial phosphorylation by dNKs in
particular. Inefficient phosphorylation of NAs not only reduces the potency of exist-
ing prodrugs but can result in the accumulation of cytotoxic reaction intermediates.
It is also responsible for the failure of a large number of potential NA prodrug can-
didates in vivo [111]. One solution to overcome the shortcomings of the phosphory-
lation cascade has been the coadministration of prodrugs with exogenous,
broad-specificity kinases via suicide gene therapy [10, 112]. While biochemical and
preclinical experiments have demonstrated the effectiveness of the strategy in prin-
ciple, the studies also uncovered limitations arising from the dNKs’ significantly
lower activity for NAs compared to their performance with the native substrates. In
addition, the broad substrate specificity of exogenous kinases raised concerns as it
interferes with the tightly regulated 2′-deoxyribonucleoside metabolism [113].
Studies have hence focused on identifying engineered NA kinases for the selective
and efficient activation of these prodrugs.
2.5.1 Thymidine Kinase from Herpes Simplex Virus
Herpes simplex virus type-1 thymidine kinase (HSVtk) was the first candidate for
NA prodrug activation and remains one of the most commonly used enzymes in
advanced (preclinical) studies [114–116]. Besides its native role in viral replication,
HSVtk is a natural candidate for application in conjunction with NA prodrugs due
to its thousandfold higher activity and elevated substrate promiscuity compared to
mammalian dNKs [117]. In fact, the phosphorylative activation of aciclovir (ACV,

1) and ganciclovir (GCV, 2) (Fig. 2.1) by wild-type HSVtk made these NA prodrugs
early therapeutics for herpes infections [118]. Nevertheless, the enzyme’s effective-
ness in antiviral treatment is limited by its two to three orders of magnitude prefer-
ence for native metabolites over NAs. Engineering HSVtk for greater prodrug
affinity could significantly enhance the clinical value of this enzyme.
In early studies, Loeb and coworkers conducted a series of directed evolution
experiments to increase the NA specificity of HSVtk. Targeting the regions flanking
two putative nucleoside binding sites (positions 162–164 and 171–173) by random
mutagenesis, HSVtk variants with increased sensitivity for 1 and 2 were identified
via genetic selection of million-member libraries using the thymidine kinase-
deficient E. coli strain KY895 [119–121]. These initial mutation studies improved
dNK activity and shifted substrate specificities upon amino acid substitution at
Phe161 alone, as well as in combination with changes in neighboring positions
(Table 2.2) [121]. In vitro assays with two selected candidates, variants 30 and 75,
show a dramatic rise in the activity ratio for NA over thymidine, and these func-
tional gains could be confirmed in mammalian cell culture experiments [139]. A
follow-up study by Black et al. to reevaluate the five amino acid replacements in
variants 30 and 75 identified a further optimized variant termed SR39 [124]. In vitro
and in vivo, variant SR39 exhibited the greatest sensitivity to 2 among engineered
HSVtks, increasing cytotoxicity by ~300-fold over wild-type HSVtk. Variant SR39
by itself, as well as in combination with other enzymes such as guanylate kinase
[140], has since demonstrated its effectiveness in various cancer models, making it
a benchmark for suicide gene therapy [141].
Guided by structural information and emerging NA drug resistance, several other
groups have explored amino acid replacements in other positions of HSVtk to find
variants with improved activity and specificity for NAs. Drake and coworkers
explored the impact of amino acid replacements at Gln125. Conservative amino
acid substitutions (Q125N) at that position showed moderate specificity changes in
favor of 2 over thymidine [123]. Meanwhile, Balzarini and coworkers revisited the
highly conserved positions 167 and 168 [125, 142]. The introduction of bulky amino
acid side chains including A167Y/F and A168H not only created an effective steric
block for native thymidine but resulted in unexpectedly preservation and even
enhancement of activity for 2. Beyond these mutagenesis studies focused on the
active site of HSVtk, Christians et al. performed a DNA shuffling experiment to
enhance the activity of the viral kinase for 3′-azidothymidine (AZT, 3) [122]. The
study used homologous thymidine kinases from HSV type-1 and type-2, as well as
their chimeras as parents for in vitro recombination and screening for increase 3
sensitivity in E. coli KY895. Four iterative rounds of laboratory evolution identified
lead kinase variant (Cycle4 TK), a patchwork of multiple sequence fragments from
the two parents and a few additional amino acid changes originating from random
mutations. Functionally, Cycle4 TK was found to be a generalist with almost identi-
cal kinetic parameters for 3 and thymidine, representing a roughly tenfold improve-
ment and a 25-fold decline in catalytic performance for 3 and thymidine,
respectively.
Table 2.2 Overview of engineered 2′-deoxynucleoside kinases as prodrug-activating catalysts in antiviral and cancer therapy and reporter for noninvasive PET
34
imaging techniques
Enzyme Product/reference Modification Observed property
Thymidine kinase Wild type (None) Highly active and promiscuous
(herpes simplex virus 2′-deoxyribonucleoside kinase
type 1) Black (1993) [120] P155A/F161V; F161I/C Improved activity for natural pyrimidine
nucleosides, ACV, GCV, and AZT
Black (1996) [121] L159I/I160L/F161A/A168Y/L169F and I160L/ Variants 30 and 75: lower thymidine activity but
F161L/A168V/L169M enhanced activity for ACV and GCV
Christians (1999) [122] Chimeric enzyme Cycle4 TK: improved activity for AZT
Hinds (2000) [123] Q125N Lower thymidine activity but enhanced activity
for GCV
Black (2001) [124] L159I/I160F/F161L/A168F/L169M Variants SR39: superior activity for GCV
Balzarini (2006) [125] A167F and A168H Eliminates thymidine activity but enhanced
activity for GCV
2′-Deoxyribonucleoside Munch-Petersen (1998) (None) Highly active, broad-specificity kinase
kinase (Drosophila [126]
melanogaster) Knecht (2000) [127] N45D/N64D Variant MuD: increased specificity for AZT and
ddC
Knecht (2002) [128] V84A/M88R/A110D Increased specificity for purine nucleosides
Knecht (2007) [129] N45D/N64D/N210D/ L239P Variants B5 and B10: increased specificity for
purine analogs
Gerth (2007) [130] Chimeric enzyme Improved activity for d4T
Solaroli (2007) [131] M88R Increased cytotoxicity of purine NAs
Liu (2009) [132] T85M/E172V/Y179F/H193Y Variant R4.V3: increased catalytic performance
for ddT
Liu (2010) [133] L66F/Y70M/E172I/Y175W Variant RosD7: increased catalytic performance
for ddT
(continued)
S. Lutz et al.
Thymidine kinase type-2 Wild type (None) High specificity for thymine and uracil
(Homo sapiens) 2′-deoxynucleosides
Campbell (2012) [134] N93D/L109F Improved activity for L-FMAU
2′-Deoxycytidine kinase Wild type (None) Moderate activity for 2′-deoxycytidine and
(Homo sapiens) purine nucleosides
Sabini (2007) [135] A100V/R104M/D133A Broader substrate specificity including thymidine
Iyidogan (2008) [136] D47E/R104Q/D133G/N163I/F242L Variant epTK6: highly promiscuous
2′-deoxyribonucleoside kinase
Hazra (2008) [137] R104M/D133A Improved activity for L-nucleosides
Muthu (2014) [138] R104M/V130T/D133A/L191A Variant B6-II: improved activity for L-FMAU
2 Engineering Therapeutic Enzymes
35
36 S. Lutz et al.
2.5.2 2′-Deoxyribonucleoside Kinase from Drosophila

melanogaster
The 2′-deoxyribonucleoside kinase from Drosophila melanogaster (DmdNK) is

expressed in fruit fly embryos and possesses the highest catalytic activity and broad-
est substrate specificity among known mammalian dNKs [126]. Although primarily
a pyrimidine kinase, DmdNK also shows significant activity for purine nucleosides.
The wild-type enzyme tolerates a variety of 2′-deoxyribose derivatives, making it
an appealing candidate for therapeutic applications in suicide gene therapy [143]. A
number of protein engineering studies have further tweaked its functional perfor-
mance toward efficient and specific activation of NAs.
In searching for DmdNK variants with increased sensitivity toward multiple
NAs, Knecht et al. followed earlier HSVtk engineering approaches by deploying
random mutagenesis and in vivo selection in E. coli KY895 [127]. The lead candi-
date emerging from these experiments was DmdNK variant MuD with two amino
acid changes (N45D/N64D). Expression of MuD in the bacterial host raised its
sensitivity to 3 and 2′,3′-dideoxycytidine (ddC) by >300-fold and >tenfold, respec-
tively. Kinetic analysis suggested that these functional gains could largely be
attributed to the enzyme’s loss of function for native 2′-deoxynucleosides while
leaving kinetic parameters for NAs mostly unchanged. A rationale for the observed
changes was not immediately obvious; both Asn are surface residues and distal to
the active site. A subsequent crystallographic study suggested that N64D destabi-
lizes the neighboring LID region, a conformationally flexible segment that contrib-
utes to substrate binding via hydrogen-bonding interactions with the substrates
3′-OH group [144]. In separate experiments, the same group focused on raising
enzyme activity for purine substrates by employed rational design based on struc-
tural comparisons of HSVtk and DmdNK [128]. The work identified a fruit fly
enzyme variant with changes in three active site residues (V84A/M88R/A110D)
that showed slightly improved kinetic performance for purine nucleosides. This
enzyme and related DmdNK variants were later tested for activation of purine
NAs, demonstrating >100-fold enhanced cytotoxicity of purine arabinofuranosides
in vitro and in cell cultures [131]. In follow-up work, DNA shuffling experiments
led to several new variants with improved activity for purine NAs cladribine and
fludarabine 9 (Fig.2.2) [129]. Although causing declines in specific activity, the
amino acid changes resulted in variants with favorable catalytic performance with
NAs relative to the natural substrates. The benefit of such functional changes was
confirmed in cancer models for two lead variants: B5 and B10 (Table 2.2). These
two variants exhibited up to thousandfold greater sensitivity for the two purine
NAs. However, the studies also revealed significant variability in kinase perfor-
mance in different cell lines, highlighting the limitations of laboratory evolution
experiments that rely on an E. coli-based surrogate system over screening kinase
variants in a specific cancer cell line. Beyond the potential functional role of the
previously reported amino acid replacements at positions 64 and 84, the basis for
the observed contributions of positions 210 and 239 is not obvious as they are
located in the conformationally poorly defined C-terminal region. Nevertheless,
Fig. 2.2 Purine nucleoside analogs as prodrugs for PNP-based anticancer therapy: MePR (7),
FdA (8), fludarabine (9), and Me(talo)PR (10)
the region’s importance to enzyme function was independently confirmed by a

study using nonhomologous recombination to generate chimera of DmdNK and
human thymidine kinase 2 [130]. Multiple hybrid enzymes with C-terminal
sequences from the fruit fly enzyme were found that exhibited significantly
improved activity for the anti-HIV prodrug 2′,3′-didehydro-3′-deoxythymidine
(d4T, 4).
More recently, two novel methods were applied to dNK engineering by our
group. To overcome the functional bias and limitations related to the use of the
auxotrophic E. coli KY895, we initially developed a FACS-based approach to
screen directed evolution libraries of dNKs for 2′,3′-dideoxythymidine (ddT, 5)
activity [132]. DmdNK libraries were created by random mutagenesis and DNA
shuffling, followed by expression of the corresponding kinase variants in E. coli.
Upon incubation with a fluorescent analog of ddT, host cells expressing a func-
tional kinase accumulated the fluorophor and could be isolated by FACS. Four
iterative rounds of directed evolution yielded top-performing variant R4.V3
(Table 2.2) that showed a 20-fold preference for 5 over thymidine. Secondly, we
experimented with in silico redesign of DmdNK using RosettaDesign software to
accelerate the discovery process and eliminate the need for high-throughput
screening [133]. Active site modeling with ddT suggested several alternative solu-
tions for tuning the enzyme’s substrate specificity. Experimental evaluation of
these computational designs led to variant RosD7 which showed roughly eight-
fold preference for ddT over native substrate and required testing of only about
two dozen DmdNK variants.
While DmdNK, HSVtk, and their variants are strong candidates for potential
applications in suicide gene therapy, their nonhuman origin remains a significant
limitation due to concerns over immunogenic response. Engineering efforts have
therefore increasingly focused on tailoring the properties of human dNKs.
2.5.3 2′-Deoxycytidine Kinase from Homo sapiens
Among the four native dNKs in human, 2′-deoxycytidine kinase (dCK) is the most
promiscuous enzyme and catalyzes the phosphorylation of 2′-deoxycytidine and
both purine homologs, as well as several NAs [110]. Even though dCK has rela-
tively poor catalytic activity compared to DmdNK and HSVtk, its broad substrate
38 S. Lutz et al.
specificity and inherently reduced immunogenicity make it an attractive target for

protein engineering.
Initial efforts to improve the catalytic efficiency of dCK were based on crystal-
lographic data and previous results from DmdNK engineering studies. Sabini et al.
focused on three residues, Ala100, Arg104, and Asp133, that correspond to posi-
tions 84, 88, and 110 in DmdNK [135]. These residues form an elaborate hydrogen-
bonding network responsible for the effective discrimination of thymidine as a
substrate. A simple amino acid swap in these positions (A100V/R104M/D133A)
broadened the substrate specificity of dCK. Besides improved activity for 2′-deoxy-
cytidine, the dCK variant showed unprecedented activity for thymidine and several
additional NAs. In a separate study, our group and others implemented a combina-
tion of rational design and random mutagenesis to more systematically explore the
impact of amino acid replacements in these three positions on the substrate specific-
ity of dCK [136, 137]. Besides confirming the key role of positions 104 and 133 in
respect to substrate specificity, these experiments identified a number of alternate,
functionally more advantageous amino acid substitutions. Of particular interest was
variant epTK6 (Table 2.2) which showed superior catalytic performance with a
diverse collection of NAs.
2.5.4 From Suicide Gene Therapy to PET Reporter Systems
The development of tailored dNKs for prodrug activation is not the only strategy to
overcome the often rate-limiting initial phosphorylation step for NAs. Significant
efforts have also been devoted to the development of synthetic alternatives. While
direct cellular uptake of phosphorylated nucleosides is not feasible, esterification of
NA monophosphates can temporarily mask the electrostatic charges of the phos-
phate group and facilitate diffusion across the cell membrane [145]. Upon cellular
entry, ester hydrolysis removes the masking groups and liberates the corresponding
monophosphate anabolite. As this phosphate ester strategy has proven increasingly
effective for the delivery of various NA prodrugs [146], efforts to engineer dNKs
have gradually shifted to other applications, including the chemo-enzymatic synthe-
sis of nucleoside triphosphate and the development of effective reporter systems for
positron emission tomography (PET) imaging [147].
PET is a versatile method for noninvasive visualization of biological processes in
living subjects and is based on two-component systems; an isotopically labeled
small-molecule reporter (typically 18F) and a reporter gene expressed within target
cells that will interact and trap the radionuclide reporter whose accumulation can be
detected via PET. Engineered dNKs, applied in combination with 18F-labeled NAs,
offer a powerful strategy to monitor, for example, whole-body tissue distribution of
viral vectors in gene and oncolytic therapies, as well as adoptive cell-based thera-
pies [148, 149]. The concept was first implemented with wild-type HSVtk or variant
SR39 as reporter genes in combination with [18F]-labeled probes such as
2′-fluoro-2′-deoxy-1-β-L-arabinofuranosyl-5-methyluracil (L-FMAU, 6) and found

widespread application in cell culture studies, as well as in translational work with
animals and humans [150–152]. While phosphorylation of 6 by highly specific,
endogenous dNKs occurs only at very low levels, the herpes enzymes activate it
albeit with moderate efficiency. Over the last decade, the development of novel PET
reporter kinases has hence focused on enzymes with high activity and specificity for
such PET probes. In addition, engineering efforts have shifted to human dNKs to
minimize risks associated with immune responses in clinical applications.
Initial experiments with human PET kinase/reporter systems largely relied on
previously reported dCK variants [153]. As discussed above, these variants with
amino acid changes in positions 104 and 133 not only exhibit activity for thymidine
and 2′-deoxyuridine but were found to also effectively phosphorylate NAs with
L-ribose sugars, a key feature of many potential PET probes [136, 137]. Side-by-
side comparison of these dCK variants and HSVtk SR39 in cell culture and animal
model studies suggested an on par performance of some human kinase variants
[154]. While these findings were encouraging, kinetic data also indicated that in
spite of functional improvements for 6, these enzymes still favored native metabo-
lites. This limits their clinical value as substrate competition for the active site
reduces imaging sensitivity and requires higher doses of radionuclide probes. The
next generation of high-contrast PET reporter systems demands bioorthogonal
dNKs, enzymes that effectively activate L-isomeric NA probes while discriminating
against native 2′-deoxynucleosides.
Attempting to meet those demands, we employed a semi-rational engineering
strategy to create an L-selective dCK variant [138]. Leveraging preexisting kinetic
data, we used computational design to create a small, focused library of 16 dCK
variants and evaluated them with D- and L-nucleosides. Lead variant B6-II showed
a two- and tenfold preference for 6 over D-2′-deoxycytidine and D-thymidine,
respectively. These early successes have led to a second round of dCK engineering,
using design-of-experiment methodology to systematically evaluate roughly 200
dCK sequences with up to 11 substitutions. In this ongoing study, kinase variants
with up to tenfold improved activity for 6, and almost 100-fold preference for
L-nucleosides has been identified (Muthu and Lutz unpublished). In separate exper-
iments, Lavie and coworkers have explored engineered variants of human mito-
chondrial thymidine kinase type 2 (TK2) as an alternative to human dCK in PET
imaging. TK2 is a native pyrimidine nucleoside kinase with moderate catalytic
activity for NAs including L-nucleosides [155], yet technical challenges with heter-
ologous protein expression have so far prohibited the enzyme’s structural analysis
and greatly limited protein engineering efforts. Building on results from PET stud-
ies with wild-type TK2 [156], Campbell et al. used a TK2 homology model to
identify two key functional residues in the enzyme activity site [134]. Amino acid
substitutions in positions Asn93 and Leu109 (N93D/L109F) not only lowered the
TK2 variant’s turnover of native D-nucleosides but also raised its activity for 6
in vitro, as well as in cell culture and mice studies.
40 S. Lutz et al.
2.6 Cytosine Deaminases
Cytosine deaminases (CDAs) are another family of nucleoside-metabolizing

enzymes that have undergone tailoring by protein engineering for potential ther-
apeutic applications. CDAs are part of the pyrimidine salvage pathway, catalyz-
ing the hydrolytic deamination of cytosine to uracil and ammonia. More
important from a medicinal standpoint, they are only found in prokaryotes and
fungi. The native enzymes accept 5-fluorocytosine (5FC) as a substrate, con-
verting the nucleobase analog into 5-fluorouracil, which is a very potent inhibi-
tor of thymidine synthase [157]. Inhibition of thymidine synthase in turn disrupts
de novo production of thymidine monophosphate and consequently stalls
nucleic acid metabolism. The small size and high cell permeability of 5FC
makes this prodrug and its CDA-based activation a highly promising anticancer
treatment.
Protein engineering of CDAs has mostly focused on increased catalytic activ-
ity and enzyme stability. In early work, Mahan et al. used structural information
to identify a lid region in E. coli codA (residues 310–320) which undergoes a
conformational rearrangement as part of the enzyme’s catalytic cycle [158].
Targeting the region by alanine-scanning mutagenesis, variant D314A showed a
20-fold decrease for native cytosine activity and a simultaneous twofold increase
for 5-fluorocytosine activity. In a follow-up study, the same authors used the
increased 5FC sensitivity of E. coli host cells expressing improved enzyme vari-
ants as a negative selection assay to evaluate a randomized whole-gene library
[159]. The study found one variant with greatly enhanced activity for 5FC, car-
rying two amino acid replacements (Q102R/D314G). Further analysis revealed
that only the substitution at position 314 contributed to the observed functional
changes. This result prompted the authors to prepare a site-saturation mutagen-
esis library for a comprehensive evaluation of that position, yielding D314S as
an additional option for CDA variants with enhanced 5FC activation. These latter
engineering studies were complemented with crystallographic analysis of
selected enzyme variants to better understand the functional improvements. The
observed structural changes suggest small but important alterations in the active
site organization, as well as possible changes to the dynamics of the lid region.
More recently, CDA variants with improved selectivity for 5FC were reported by
Fuchita et al. [160]. The authors employed targeted mutagenesis in two regions
of codA (residues 149–159 and 310–320) and evaluated the million-member
library through negative selection. In the top-performing CDA variant, three
amino acid substitutions (V152A/F316C/D317G) caused an approximately
20-fold shift in substrate preference from cytosine to 5FC. Kinetic studies of the
variant revealed a combination of changes to KM and kcat values. Meanwhile,
crystallographic analysis confirmed that none of the three residues was in direct
contact with the substrate but seemed to affect enzyme activity through subtle
conformational changes of neighboring residues.
In addition to the studies of bacterial deaminases, yeast CDA makes an attractive
prodrug-activating enzyme due to its overall superior catalytic performance.
However, its potential in therapeutic applications is compromised by the native

enzymes’ limited stability. To this end, Korkegian et al. took a computational route
to improve the stability of CDA from Saccharomyces cerevisiae [161]. Using
RosettaDesign to predict amino acid substitutions for stabilizing the protein struc-
ture, their experiments identified two neighboring substitutions (A23L and I140L)
and a third, distant change (V108I) which together improved the packing of the pro-
tein’s hydrophobic core. Improved core packing in the redesigned enzyme raised its
temperature of unfolding (Tm) by 10 °C and translated into a 30-fold increased half-
life at 50 °C without compromising catalytic performance. In a separate study by the
same team, CDA variants with improved activity for 5FC were sought via random-
ized mutagenesis of 11 amino acid residues [162]. These positions were selected
based on their location in highly conserved regions of the protein. After diversifica-
tion by primer-based codon scrambling, the resulting protein library was tested for
improved variants in a two-stage genetic complementation assay, using a codA-defi-
cient E. coli strain. Positive selection of library members via growth on cytosine
minimal media was followed by negative selection for 5FC sensitive variants. The
experiments identified three distinct variants with single amino acid substitutions at
either position 92, position 93, or position 98, yet only D92E conferred increased
5FC sensitivity in subsequent mammalian cell culture. The kinetics of purified yeast
CDA (D92E) showed little change in the Michaelis-Menten parameters, yet bio-
physical studies indicated an increased Tm by 4 °C which could be rationalized by the
residues’ position at the homodimer interface. Interestingly, the detected gains in
activity in vivo are likely due to elevated protein stability. However, the limits of
such a strategy for improving enzyme activity could be seen when the D92E substi-
tution was combined with the computationally designed CDA variant. While the
quadruple mutant showed additive behavior in regard to protein stability, the activity
of the variant declined.
In summary, recent protein engineering studies have demonstrated the potential
for functional improvements in CDAs. Concerns over the immunogenicity of ectop-
ically expressed CDAs in patients remain to be addressed and have hampered clini-
cal trials. Additionally, 5FC by itself is a potent bactericide, and oral administration
of the prodrug could affect a patient’s intestinal flora, potentially resulting in unin-
tended side effects. Nevertheless, clinical studies of dNK-CDA conjugates have
been reported [163, 164] (Table 2.3).
2.7 Purine Nucleoside Phosphorylases
In purine metabolism, purine nucleoside phosphorylase (PNP) catalyzes the revers-

ible phosphorolysis of inosine and guanosine into the corresponding nucleobase and
ribose-1-phosphate [172]. PNPs can be divided into two categories based on their
structures: trimeric and hexameric. Trimeric PNPs are found in both prokaryotes
and eukaryotes and exhibit high specificity for 6-oxopurine nucleobases. Hexameric
PNPs are exclusive to prokaryotes and accept a broader range of nucleobase sub-
strates including 6-amino and 6-oxopurines. Given these differences in substrate
42 S. Lutz et al.
Table 2.3 Non kinase-based GDEPT systems modified by protein engineering

Product/
Enzyme reference Modification Observed property
Cytosine Wild type (None) Robust enzyme with
deaminase catalytic activity for
(E. coli) cytosine >> 5-fluorocytosine
(5FC)
Mahan (2004) D314A/G/S Shift in catalytic activity
[158] favoring 5FC over cytosine
due to altered lid dynamics
Fuchita (2009) V152A/F316C/D317G Shift in catalytic activity
[160] favoring 5FC over cytosine
via distal amino acid
changes
Cytosine Wild type (None) Fragile enzyme with high
deaminase catalytic activity for
(S. cerevisiae) 5-fluorocytosine (5FC)
Korkegian A23L/V108L/I140L Increased Tm by 10 °C due
(2005) [161] to improved hydrophobic
core packing; unchanged
catalytic performance.
Increased Tm by 10 °C due
to improved hydrophobic
core packing; unchanged
catalytic performance
Stolworthy D92E Increased Tm by 4 °C due to
(2008) [162] stabilization of homodimer
interface; unchanged
Purine Wild type (None) Moderate catalytic
nucleoside performance on a broad
phosphorylase range of substrates
(E. coli) Bennett (2003) M64V >100-fold improved
[165] catalytic activity for
nucleoside analog due to
reshaping of active site
binding pocket
Purine Wild type (None) Activity for 6-oxopurine
nucleoside substrates
phosphorylase Stockler (1997) N243D/E201Q Altered substrate specificity
(human) [166] for 6-aminopurine substrates
Purine Wild type (None) Six- and 12-fold improved
nucleoside catalytic activity for
phosphorylases fludarabine compared to
(S. sofataricus) E. coli PNP. Six- and
12-fold improved catalytic
activity for fludarabine
compared to E. coli PNP

Product/
Enzyme reference Modification Observed property
Nitroreductases Wild type (None) Molecular target for several
(E. coli) (NfsB) bactericides
NfsB: Grove F125K Improved activity for
(2003) [167] CB1954
NfsB: Guise T41Q/N71S/F124T Improved activity for
(2007) [168] CB1954
YieF: Barack T160N/Q175L/Y230A/ Improved activity for
(2006) [169] Y239N CB1954
NfsB: T41L/F70A Improved activity for
Jaberipour CB1954
(2010) [170]
FRaseI: Swe A120V/F124G Improved activity for
(2012) [171] CB1954
specificity, initial efforts to utilize PNP in therapeutic applications focused on the

hexameric class. Conceptionally, PNPs are delivered into a tumor via suicide gene
therapy where their broader substrate specificity is exploited to catalyze the cleav-
age of nontoxic nucleoside prodrugs into cytotoxic purine analogs [173]. The hexa-
meric PNP from E. coli (DeoD) was found to be sufficiently promiscuous to activate
a number of nucleoside prodrugs including 9-β-D-[2′-deoxyribofuranosyl]-6-
methylpurine (MePR, 7) and 2-fluoro-2′-deoxy-adenosine (FdA, 8), as well as the
ribose analog 9-β-D-arabino-furanosyl-2-fluoroadenine (FaraA, 9). No phospho-
rolysis of these prodrugs was detected with trimeric human PNP.
Despite these initial successes, PNP-based prodrug activation faces a major chal-
lenge due to the prodrug toxicity to a patient’s intestinal microbiome. In early stud-
ies, these undesirable side effects on the gut flora were targeted by protein
engineering of DeoD. Two strategies to enhance the enzyme’s therapeutic efficacy
were pursued: improving catalytic activity to allow for lower prodrug doses and
altering substrate specificity to activate prodrugs not converted by any wild-type
PNP. Specifically, molecular modeling of 7 and derivatives (such as Me (talo)PR;
10) in DeoD revealed the importance of sugar puckering in productive substrate
binding and identified a steric clash between Met64 and the C6′-methyl group of 10.
While this structural incompatibility makes 10 a poor substrate for native PNPs, the
substitution of Met64 with several smaller hydrophobic amino acids led to variant
Met64Val possessing an enlarged active site binding pocket [165]. Detailed kinetic
analysis of the Met64Val variant showed that catalytic performance with 10 had
improved by over two orders of magnitude, while activity for 7 was largely
unchanged. These in vitro gains also translated into greater potency in vivo.
Treatment of mice bearing D54 tumors with 10 in the presence of this PNP variant
raised the cytotoxicity of the prodrug by at least tenfold without exhibiting any
significant toxicity in non-PNP-transduced cells.
44 S. Lutz et al.
Separately, fludarabine (9) has demonstrated potential as a prodrug in PNP-based

suicide gene therapy. To date, studies have largely focused on nucleoside analog
activation by native DeoD, demonstrating in vitro efficacy in cell cultures and
mouse models [174, 175]. However, the modest activity of E. coli PNP clearly lim-
its the prodrug potency, requiring administration in high doses that often cause seri-
ous side effects. A recent report of two PNPs from the hyperthermophilic archaean
Sulfolobus solfataricus offers an interesting new starting point for future engineer-
ing efforts to improve phosphorolysis of 9 [176]. Compared to DeoD, the two
enzymes possess six- and 12-fold higher catalytic efficiency, respectively. Sequence-
structure comparison may allow for the identification of critical functional residues
and interactions that can guide the redesign of the E. coli enzyme by rational meth-
ods and directed evolution.
Finally, the high substrate specificity of trimeric enzymes in general, and the
human PNP in particular, can be altered by protein engineering. Relying on crystal-
lographic data, Glu201 and Asn243 were identified as the two key residues respon-
sible for the strict preference for 6-oxopurine nucleoside substrates [166]. A change
in the hydrogen bond donor/acceptor pattern through site-directed mutagenesis at
these two positions (E201Q and N243D) resulted in the predicted reversal of sub-
strate preference. The high catalytic activity of the unnatural human 6-aminopurine
nucleoside phosphorylase and its reduced immunogenicity compared to the bacterial
PNPs offer an attractive starting point for further engineering and optimization.
2.8 Nitroreductases
A prominent non-nucleoside-based GDEPT system is the use of bacterial nitrore-

ductase enzymes to activate DNA-damaging nitroaromatic prodrugs [177]. Bacterial
nitroreductases are a class of flavin-dependent oxidoreductases with broad substrate
specificities that have made them of interest for a number of biotechnological appli-
cations including bioremediation [178] and transgenic cell labeling and ablation
[179]. Most GDEPT work has focused on the nitroreductase NfsB from E. coli and
its activation of the prodrug 5-(aziridin-1-yl)-2,4-dinitrobenzamide (CB1954),
which when reduced is converted to a highly cytotoxic DNA cross-linker [180].
Phase I and II trials of CB1954 in conjunction with a replication-defective adenovi-
rus expressing E. coli NfsB showed that treatment was safe, and a decline in levels
of prostate-specific antigen in some patients suggested some effect; however, over-
all it was deemed that greater therapeutic efficacy would be desirable [181].
In an initial engineering effort, Grove et al. examined a previously solved crystal
structure of NfsB [182] and used this to target six residues around the active site for
saturation mutagenesis, with the resulting single mutants screened for improved
CB1954 activation via monitoring CB1954-dependent E. coli growth inhibition
[167]. Fourteen of these beneficial single mutants were then combined to generate a
small 53-member library of double mutants [170]. The most effective mutants, T41L/
N71S and T41L/F70A, were reported to be 14–17-fold more potent than wild-type
NfsB_Ec at sensitizing SKOV3 human cancer cells to CB1954. While these efforts
focused on fairly low-throughput E. coli growth inhibition-based screens, a substan-
tially larger library (~1,000,000 variants) made from simultaneous randomization of
multiple active site residues was screened using a system in which nitroreductase
variants were expressed in E. coli from chromosomally inserted bacteriophages.
When active nitroreductase variants reduced CB1954, the subsequent DNA damage
triggered entry into the phage lytic cycle, bursting the E. coli and allowing collection
of the bacteriophage carrying improved nitroreductases [168]. Other groups have
focused on engineering nitroreductases other than NfsB, such as the NfsB homolog
FRaseI from Vibrio fischeri [171] or an unrelated E. coli nitroreductase YieF. While
YieF was not improved via a CB1954-focused campaign, a variant Y6 (T160N/
Q175L/Y230A/Y239N) that had been evolved by successive rounds of random
mutagenesis for improved chromate reduction also serendipitously showed an
increased ability to sensitize HeLa cells to CB1954 [169].
2.9 holinesterases: Bridging Protein Therapeutics

C
and Biosensors
Cholinesterases (ChE) are a broad family of enzymes that selectively hydrolyze

cholinesters into choline and the corresponding carboxylic acid. In humans, there
are two types of ChEs: acetylcholinesterase (AChE) and butyrylcholinesterase
(BuChE) [183]. AChE is the primary ChE in the body and catalyzes the breakdown
of the neurotransmitter acetylcholine. Found at neuromuscular junctions and synap-
tic clefts, the membrane-associated enzyme is responsible for hydrolysis of the sig-
naling molecule which terminates synaptic transmission. In contrast, BuChE is
more ubiquitously found throughout the body and possesses broad specificity for
choline-based esters. Its natural biological function is largely unknown. Because of
their essential functions, inhibition of ChE by drugs or poisons including organo-
phosphates and carbamates (Fig. 2.3) has been the subject of many studies including
protein engineering. These engineering studies initially focused on probing the
human cholinesterases to explain and predict the consequences of exposure to
inhibitors [184]. More recently, efforts have shifted toward alternate applications of
ChEs as therapeutic agents, both as potential drug candidates and as next-generation
biosensors.
2.9.1 Butyrylcholinesterase
The absence of an identifiable critical physiological function largely resulted in

scientific neglect of BuChE for over 50 years [185]. However, the enzyme’s fate
changed with the discovery of its effectiveness as a stoichiometric scavenger for
nerve agents. In addition, animals injected with exogenous BuChE at concentra-
tions exceeding endogenous levels by a thousandfold showed no adverse effects.
46 S. Lutz et al.
These findings have driven the development of prophylactic treatments for indi-
viduals at risk of exposure to toxins including organophosphates and carbamates
[186–189].
Beyond these therapeutic applications of native enzyme, BuChE was observed to
hydrolyze cocaine, albeit with low efficiency [190]. A number of protein engineer-
ing studies have subsequently focused on generating BuChE variants with enhanced
substrate specificity for possible treatment of general cocaine addiction and acute
toxicity. These efforts have generally been limited to rational design studies rather
than directed evolution as protein production of human BuChE requires mammalian
expression systems for proper glycosylation and tetramer assembly, significantly
complicating the experimental evaluation of large libraries of enzyme variants by
high-throughput methods.
Focusing on the natural amino acid variability in the active site of BuChEs
from different mammals, Xie et al. investigated the functional impact of specific
amino acid substitutions in about a dozen positions of the human enzyme [191].
All but one variant showed lower hydrolytic activity for cocaine. The substitu-
tion of Ala328 for Tyr raised the variant’s activity by fourfold over wild-type
enzyme. The moderate functional gain could be rationalized by computational
modeling, showing conformational changes that eliminate steric constraints in
the active site, permitting for more effective cocaine binding. Separately, MD
simulations and in silico design studies focused on amino acid positions Ala328
and Tyr332 for enhanced cocaine hydrolysis by BuChE [192–194]. These works
were motivated by kinetic data for native BuChE showing a thousandfold faster
conversion of unnatural (+)-cocaine over its natural enantiomer [195]. Molecular
modeling implicated the two amino acids in reducing steric hindrance and form-
ing critical cation-π interactions for productive binding of the unnatural enan-
tiomer but not for natural (−)-cocaine. Simple inversion of the amino acid side
chains (A328Y/Y332A or A328Y/Y332G) resulted in 10–20-fold higher cata-
lytic performance of the BuChE variant with (−)-cocaine. Independently, amino
acid variations at positions 328 and 332 in combination with two additional
unspecified substitutions were reported as AME359 [196]. This BuChE variant
showed fourfold higher activity than previous double mutants, and while details
of the work were not disclosed at the time, Pan et al. later listed its amino acid
replacements as F227A/S287G/A328W/Y332M [197]. That latest study
deployed molecular dynamics simulations to capture the effects of amino acid
changes on the enzyme’s ability to stabilize the transition state of BuChE-
catalyzed cocaine hydrolysis. These modeling efforts led to a BuChE variant
with four amino acid changes, A199S, S287G, A328W, and Y332G. Experimental
evaluation of this variant yielded a cocaine hydrolase with almost 500-fold
improved catalytic performance over wild-type enzyme. Given the functional
improvements summarized above, administration of BuChE variants in patients
effectively reduces the half-life of cocaine from 45 to 90 min to just a few sec-
onds [198]. Tests with these engineered enzymes for acute cocaine toxicity have
shown promise for clinical efficacy in animal models [199, 200] and phase I
clinical trials in humans [198].
2.9.2 Acetylcholinesterase
AChE serves as a key component in neurotransmission. As such, it is a primary

molecular target of organophosphates and carbamates and represents the basis for
the biological activity of many pesticides and nerve agents. Similarly to BuChE, the
administration of exogenous AChE has been demonstrated to be an effective stoi-
chiometric scavenger to counter acute toxicity yet is not practical as a potential
therapeutic due to the enzyme’s critical biological function and potential serious
side effects associated with high doses of AChE. Instead, AChE has found utility as
a biosensor to detect pesticides, nerve agents, and related compounds [201]. These
AChE-based biosensors offer a highly tunable, selective, and robust solution for the
identification of single or multiple chemical agents directly from field samples; an
unmet need in environmental safety as the detection of organophosphates and car-
bamates is traditionally based on GC-MS methods [202]. While delivering accurate
results, these analytical techniques require time-consuming sample preparation,
specialized and expensive equipment, and highly trained personnel. An enzyme-
based sensory system, capitalizing on the biocatalyst’s high substrate specificity,
selectivity, and potential for signal amplification, offers an attractive alternative for
detection of low levels of analytes. This concept was first demonstrated by two
groups in the early 1980s, using immobilized AChE from the electrical eel
(Electrophorus electricus) to monitor the presence of organophosphate and carba-
mate pesticides in water samples [203, 204]. Eel AChE was chosen for its high cata-
lytic activity combined with commercial availability of the enzyme. The presence of
pesticides at picomolar concentration (parts per billion) could be detected via the
reduction in ChE activity.
Building on the idea of an AChE-based biosensor, Villatte et al. screened a small
collection of AChE homologs from different species for their sensitivity to organo-
phosphate and carbamate insecticides (Fig. 2.3) [205]. Among these wild-type
enzymes, the homolog from fruit fly (Drosophila melanogaster) showed the highest
sensitivity to inhibitors, exceeding the responsiveness of eel AChE by up to eight-
fold for 14 out of the 19 tested insecticides. Further active site analysis of the fruit
Fig. 2.3 Chemical structures of organophosphates and carbamates used as insecticides and nerve
agents: paraoxon (11), dichlorvos (12), pirimiphos-methyl (13), malathion (14), carbofuran (15),
sarin (16), soman (17), cyclohexylsarin (18), and VX (19)
48 S. Lutz et al.
fly enzyme by computational modeling indicated a critical π-stacking interaction

between the enzyme and inhibitors. The model suggested tighter inhibitor binding
upon substitution of Phe for Tyr408, one of the hydrophobic residues in the sub-
strate binding pocket. Experimental evaluation of this amino acid replacement
showed moderate improvements in inhibitor sensitivity, raising the AChE variant’s
responsiveness to 12-fold over eel AChE. These functional gains led to more com-
prehensive engineering efforts of the fruit fly enzyme via a series of site-directed
mutagenesis experiments [206]. Seeking to identify variants with enhanced inhibi-
tor sensitivity, the study targeted sites known to control insecticide toxicity, key
positions in the active site identified by inhibitor docking studies, and functionally
important residues as determined by alanine scanning. Each of these strategies
yielded multiple variants with increased sensitivity over fruit fly AChE, yet gains
were typically <tenfold. Possible synergistic effects from combinations of individ-
ual beneficial changes were also explored to further improve biosensor perfor-
mance. Bundling of favorable amino acid substitutions in two positions (E69Y/
Y71D) led to a 300-fold increased sensitivity for dichlorvos (12) compared to the
parental enzyme. The two residues are located in a loop region at the active site
entrance thought to control access to the catalytic center of the enzyme. Amino acid
changes are likely to cause changes to the loop’s conformational flexibility and
consequently impact its biological function. Besides increased inhibitor binding,
improvements of overall AChE stability were investigated in a separate study [207].
The fruit fly enzyme contains eight Cys residues; seven of them form intra- or inter-
molecular disulfide bonds. Replacement of the eighth, unpaired Cys, which is
located at position 290 in the core structure of the enzyme, with Val resulted in
small to moderate (<threefold) improved lifetime at elevated temperatures and in
the presence of chemical denaturants. The functional gains could be rationalized
through reduced disulfide exchange and increased hydrophobic core packing.
More recently, a separate, monomeric AChE from parasitic nematode
Nippostrongylus brasiliensis (ACheE B) has emerged as a promising protein engi-
neering template for the generation of biosensors with high sensitivity and broad
specificity [208]. ACheE B can be heterologously expressed in Pichia pastoris, a
significant advantage for engineering efforts compared to the previously discussed
AChEs which depend on insect or mammalian expression systems for suitable pro-
tein yields [209]. Guided by previous engineering studies, amino acid changes at ten
positions in or near the active site were explored by single- and double-site varia-
tions. The study targeted primarily residues with large hydrophobic side chains,
replacing them with smaller counterparts to widen the active site cleft and enlarge
the substrate binding pocket. The resulting AChE B variants were then screened
against a collection of 14 potential inhibitors and identified several candidates with
enhanced sensitivity for organophosphonates. Functional gains typically ranged in
the three- to tenfold range but showed up to 100-fold improvements for pirimiphos-
methyl (13), raising the sensitivity of the nematode enzyme to levels comparable
with the fruit fly and eel AChE standards in the field. Interestingly, the impact of
individual amino acid substitutions on the inhibition patterns of these three enzymes
showed some notable differences in spite of their high sequence homology. These
differences highlight the challenge of accurately predicting the impact of amino

acid substitutions. Overall, ACheE B is an attractive alternative platform for biosen-
sor applications, given its ability to be expressed in yeast, as well as its on par func-
tional performance with and superior stability over current AChE standards.
While none of the engineering efforts have generated a true AChE generalist
with broad and uniform inhibitor sensitivity, the assembly of sensor arrays with
multiple enzymes was reported to cover a broader spectrum and complex mixtures
of analytes. To accurately detect and quantify paraoxon (11) and carbofuran (15) in
binary mixtures, Bachmann and Schmid initially built multielectrode thick-film
sensors that carried four different wild-type AChEs (eel, bovine, rat, and fruit fly)
[210]. Analyzing the responses from individual enzymes with the help of machine
learning algorithms, they were able to reliably measure nanomolar concentrations
of each compound. The same team subsequently replaced the native enzymes with
four engineered variants of fruit fly AChE [211]. These variants were carefully
selected based on their divergent sensitivity for a variety of organophosphate and
carbamate pesticides. The resulting biosensor showed roughly fourfold lower detec-
tion limits than its counterpart based on wild-type enzymes and underscores the
potential benefits of lab-customized AChEs with tailored inhibitor sensitivity.
2.9.3 Other Esterases
Cholinesterase represents an important target for organophosphate-based pesticide

and nerve agents (Fig. 2.3) and can be deployed as stoichiometric scavengers to
counter the toxic efforts of these inhibitors. However, they fail to be effective (cata-
lytic) tools for destroying these compounds. Nature offers an alternative for cata-
lytic detoxification of organophosphates though. Environmental exposure of
microbes by decades of pesticide use has resulted in the emergence of enzymes that
break down these molecules. Among the most extensively studied enzymes in this
category are phosphotriesterases (PTEs), β/α-barrel proteins isolated from
Flavobacterium sp. and Pseudomonas diminuta [213–215]. PTEs have no known
natural substrates but very effectively hydrolyze a broad range of pesticides includ-
ing paraoxon (11). They also show low to moderate activity for neurotoxic warfare
agents such as soman (17) [216].
Over the last 20 years, a number of PTE engineering studies have attempted to
improve the catalytic degradation of organophosphates (Table 2.4). By and large,
these studies have focused on the small and large substrate binding pockets in the
active site, consisting of residues Ile106, Trp131, Phe132, Phe306, Ser308, and
Phe309, as well as His254, His257, Leu303, and Met317, respectively (Fig. 2.4). In
one of the first papers, Raushel and coworkers employed site-directed mutagenesis
to prepare 11 PTE variants with polar residues in the small binding pocket [217].
The modification accelerated cleavage of phosphorus-fluorine bonds by tenfold.
The gain was rationalized by favorable interactions of the fluoride with the proton
donors and more productive substrate orientation in the active site. Continuing their
exploration of the active site by rational design, they performed a systematic
50 S. Lutz et al.
Table 2.4 Protein engineering of organophosphate hydrolases

Phosphotriesterase Wild type (None) Effective hydrolase
(P. diminuta and for paraoxon and
Flavobac. sp) related insecticides
Watkins (1997) [217] F132Y/H and Improved activity for
F132H/F306H diisopropyl
fluorophosphate
Chen-Goodspeed I106G/A; F132G/A; Multiple PTE variants
(2001) [218] H254F/Y; H257Y/W; with enhanced activity
L271F/Y; S308G/A; and relaxed or
M317F/Y/W reversed
stereoselectivity
Cho (2002) [219] A14T/A80V/K185R/ Variant 22A11:
H257Y/I274N improved activity for
methyl parathion
Griffiths (2003) [220] I106T/F132L Variant h5: twofold
improvement for
paraoxon hydrolysis
Hill (2003) [221] H254G/H257W/ ~1000-fold increased
L303T activity for soman
hydrolysis
Cho (2004) [222] A14T/L17P/A80V/ Variant B3561:
V116I/ K185R/ improved activity for
A203T/I274N/P342S broad range of
substrates including
chlorpyrifos
Tsai (2010) [223] H254Q/H257F; ~10–500-fold
H257Y/L303T; improved activity for
H254G/H257W/ various G and V-type
L303T nerve agents
Bigley (2013) [224] I106C/F132V/ Variant L7ep-3a:
H254Q/H257Y/ ~100-fold improved
A270V/L272M/ activity for VX nerve
I274N/S308L agent
Cherny (2013) [225] K77A/A80V/F132E/ Variant G5-C23: 100
T173N/G208D/ to 1000-fold improved
H254G/I274N catalytic performance
for V-type nerve
agents
Phosphotriesterase Wild type (None) Effective hydrolase
(Agrobacterium for dimethyl
radiobacter) organophosphates,
phosmet, and fenthion
Jackson (2009) [226] W131H/F132A Increased activity for
chlorfenvinphos
hydrolysis
Naqvi (2014) [227] S308L/Y309A ~5000-fold increased
activity for
degradation of
malathion

Paraoxonase 1 (New Aharoni (2004) [228] I126Y/M130L/ Variant G3C9: highly
Zealand rabbit liver) K137S/L142V/ expressed in E. coli
A301G/A320V/ with wild-type-like
M341L/V343I catalytic performance.
Variant G3C9: highly
expressed in E. coli
with wild-type-like
G3C9 + L69V/ Variant G3C9.10 and
E218D or G19R/ G3C9.49: 40–50-fold
S193P/N287D/ higher catalytic
V346A efficiency for
paraoxon hydrolysis
Amitai (2006) [229] G3C9 + L69V, Increased activity for
H115W or V346A various toxic
organophosphates by
~100–300-fold
Gupta (2011) [230] G3C9 + L69G/ Variant 4E9: highly
S111T/H115W/ efficient hydrolase for
H134R/F222S/T332S degradation of
G-agents
Fig. 2.4 Key amino acid residues in substrate binding pockets of PTE with bound substrate analog
(PDB access#: 1dpm [212]). The diethoxy-p-nitrophenyl-phosphonate substrate analog is colored
in blue, while the position of the two zinc atoms is shown as gray spheres
52 S. Lutz et al.
evaluation of the functional impact of altering amino acids in both pockets [218].
Their study demonstrated that significant control over reactivity and stereoselectiv-
ity of PTE variants could be gained by shrinking or enlarging these two pockets.
Rather than being limited to a few specific amino acid substitutions, Hill et al. tar-
geted His254, His257, and Leu303 by multisite-saturation mutagenesis to explore
the enzyme’s hydrolytic activity with a chromogenic analog of 17 [221]. Substitutions
in these positions led to a triple variant with an almost thousandfold enhanced cata-
lytic efficiency. While a rationale for the functional gains was not immediately
obvious from x-ray crystallography, MD simulations suggested an overall optimiza-
tion of substrate orientation in the active site [223]. Merging the list of target sites
and multisite-saturation mutagenesis, the same group re-evaluated their PTE librar-
ies against a larger collection of chromogenic nerve agent analogs [231]. As in the
previous study by Hill et al., candidates with substitutions in the large binding
pocket emerged as the best-performing variants, showing one or two orders of mag-
nitude improved hydrolytic activity for analogs of 16, 17, and 18.
To expand the search for favorable amino acid changes beyond a few selected
sites chosen by rational design, the development of surface display, in vitro com-
partmentalization, and genetic complementation methods for high-throughput
screening of PTEs set the stage for large-scale, in vitro directed evolution experi-
ments. Chen and coworkers generated large combinatorial libraries of PTE variants
by DNA shuffling and deployed an E. coli cell surface display system to find an
improved hydrolase for the paraoxon analog methyl parathion [219]. Two rounds of
in vitro evolution yielded 22A11, a variant with five amino acid replacements that
showed a 25-fold higher specific activity for the test substrate than its parental
PTE. The location of the substitutions suggests that their impact on enzyme func-
tion results from a combination of direct and indirect effects; H257Y alters the size
and shape of the substrate binding pocket, while A80V, L185R, and I274N are
surface-exposed residues that might affect protein folding and stability. Using wild-
type PTE and 22A11 as templates, the authors subsequently used the same labora-
tory evolution scheme to search for a hydrolase of chlorpyrifos, another major
pesticide [222]. The study identified variant B3561, a descendent of 22A11 with a
total of eight amino acid changes including four substitutions originally found in the
parental sequence. The catalytic efficiency of B3561 for chlorpyrifos was increased
by >700-fold over wild-type PTE (twofold over 22A11) and was on par with the
hydrolysis rate of paraoxon by wild PTE. Separately, Griffiths and Tawfik adopted
their in vitro compartmentalization method for PTE library screening, using
microbead-based surface immobilization in combination with FACS [220, 232].
The method was deployed to analyze four multisite-saturation mutagenesis librar-
ies, targeting key positions in the small binding pocket of PTE for variants with
improved paraoxon hydrolysis. The kinetic analysis of lead variant h5 with substitu-
tions in positions 106 and 132 resulted in a 70-fold increase in catalytic rate but
compromised substrate binding, enhancing the overall catalytic performance of the
variant by a moderate twofold over wild-type PTE. Finally, Raushel and coworkers
exploited an in vivo selection scheme for library analysis, using an E. coli strain
whose growth was controlled by an artificial phosphonate assimilation pathway
[233]. The selection of large PTE libraries with several phosphonate substrates
modeled after G-type nerve agents (16–18) yielded numerous enzyme variants with
improved catalytic performance, yet none of these hits surpassed the previously
reported variant with H257Y/L303T [231].
The moderate successes of large-scale directed evolution experiments led to a
trend reversal in PTE engineering. More recent studies have instead been relying on
smaller, more focused libraries that can be screened by medium-throughput,
microtiterplate-based colorimetric assays. Besides changes in engineering strategy,
target substrates have also shifted to include the degradation of V-type nerve agents
including VX (19). In contrast to 11 and G-type nerve toxins, organophosphates
classified as V-agents possess a thiolo leaving group, which makes them poor sub-
strates for existing hydrolases. Starting with PTE variant (H254Q/H257F), Raushel
and coworkers employed iterative rounds of multisite-saturation and random muta-
genesis to identify variant L7ep-3a with six additional amino acid changes and a
100-fold enhanced catalytic efficiency for 19 [224]. The x-ray structure analysis
revealed substantial remodeling of the active site for more effective substrate inter-
actions [234]. Meanwhile, Cherny et al. merged traditional laboratory-based
directed evolution with in silico remodeling of PTE to search for a novel hydrolase
with broad specificity for V-type nerve agents [225]. Key positions in the enzyme
active site were identified based on the results from previous engineering studies
and supplemented with modeling data obtained from RosettaDesign software.
These amino acid variations were sampled iteratively over five rounds and led to the
identification of several lead candidates including PET variant G5-C23 whose cata-
lytic performance and stereoselectivity for 19 matched the previously reported vari-
ant L7ep-3a. In addition, G5-C23 also shows broad substrate specificity for two
other V-type nerve agents (RVX and CVX) and G-type neurotoxins (16–18).
Besides extensive engineering studies on PTE from Pseudomonas diminuta,
similar efforts have been undertaken to tailor functional homologs from
Agrobacterium radiobacter and mammalian sources. The native hydrolase from
Agrobacterium (OpdA) showed superior catalytic activity over PTE for several
organophosphates, making it a potential candidate for medical treatment of pesti-
cide poisoning [235]. To further expand the enzyme’s substrate specificity, Jackson
et al. employed computational docking to locate steric constraints that interfered
with productive substrate binding in the active site [226, 227]. Initially, modifica-
tions in the small binding pocket (W131H/F132A) all but eliminated the enzyme’s
stereoselectivity for the E- and Z-isomers of chlorfenvinphos and raised the cata-
lytic efficiency by roughly ten- and 500-fold, respectively. Subsequently, unfavor-
able sterics deemed responsible for the low-level hydrolysis of 14 by OpdA was
eliminated by multisite-saturation mutagenesis (CASTing). The leading OpdA vari-
ant S308L/Y309A possessed an expanded binding pocket which allowed for a
favorable substrate binding geometry and resulted in ~5000-fold rate enhancement.
In mammals, members of the serum paraoxonase (PON) family primarily catalyze
the hydrolysis of esters and lactones yet were found to exhibit residual activity for
organophosphates. To improve the latter activity, Aharoni et al. first used DNA shuf-
fling of four native PON1s from human, rat, mouse, and rabbit to select for highly
54 S. Lutz et al.
expressed recombinants [228]. The emerging lead candidate G3C9 was largely
derived from rabbit but carried eight amino acid substitutions that appear to facili-
tate more effective protein folding and increase stability due to reorganization of
hydrophobic regions of the enzyme. The hydrolytic activity of G3C9 for a variety of
organophosphates was then further increased by random mutagenesis [228, 229]. To
further advance the functional performance of these PON1 against selected G-type
nerve agents, the authors used six rounds of targeted and random mutagenesis in
combination with plate-based and FACS-based screening to analyze their library
with a fluorogenic coumarin analog of 18 [230]. Variant 4E9 emerged as one of the
top-performing hydrolases, containing six amino acid changes located primarily in
the enzyme active site. Beyond its high efficiency in cyclohexylsarin degradation,
4E9 showed similarly high catalytic activity for related G-type nerve agents.
Extending their functional studies to animal models, the authors also demonstrated
significantly increased survival rates for mice prophylactically treated with this
enzyme.
Conclusions
The utilization of protein-based biologics in general and engineered enzymes in
particular offers tremendous opportunities for the development of highly effec-
tive and specific therapies. Whether the enzyme itself is the drug, is responsible
for the activation of prodrugs, or serves as detection device for disease markers,
the functional versatility of proteins and their evolvability makes these biologi-
cal macromolecules a powerful and almost limitless resource for clinical appli-
cations. Over the last two decades, strategies to engineer therapeutic enzymes
have largely mirrored the advances in the field of protein engineering. While
early efforts focused on few, site-specific changes, the introduction of high-
throughput screening techniques led to the expansion of library size and more
comprehensive sampling of proteins by directed evolution. However, the first
rule for directed evolution that “you get what you select for” has proven particu-
larly true for therapeutic enzymes. Screening large libraries of enzyme variants
in bacterial surrogate systems rather than their eventual (mammalian cell) host
is technically much simpler and affordable but has, in more than once instance,
resulted in underperforming lead candidates when subsequently tested in clini-
cal settings. It is with that background that newly emerging protein engineering
strategies based on computational design, machine learning algorithms, and
greater mechanistic and structural insights offer exciting future opportunities
for the development of the next-generation protein-based biologics. These
methods provide critical guidance for protein redesign, enabling the generation
of small, focused libraries whose functional diversity can be assessed in com-
plex assays that more accurately approximate the cellular environment and
maximize the functional gains. Building on the past successes and failures of
engineered enzymes in therapeutic applications reviewed here, the future holds
much promise for tailored enzymes to detect and tackle disease and environ-
mental threats to human health.
References
1. Leader B, Baca QJ, Golan DE (2008) Protein therapeutics: a summary and pharmacological
classification. Nat Rev Drug Discov 7(1):21–39
2. Dimitrov DS (2012) Therapeutic proteins. Methods Mol Biol 899:1–26
3. Kurtzman AL, Govindarajan S, Vahle K, Jones JT, Heinrichs V, Patten PA (2001) Advances in
directed protein evolution by recursive genetic recombination: applications to therapeutic pro-
teins. Curr Opin Biotechnol 12(4):361–370
4. Vasserot AP, Dickinson CD, Tang Y, Huse WD, Manchester KS, Watkins JD (2003) Optimization
of protein therapeutics by directed evolution. Drug Discov Today 8(3):118–126
5. McCafferty J, Glover DR (2000) Engineering therapeutic proteins. Curr Opin Struct Biol
10(4):417–420
6. Shak S, Capon DJ, Hellmiss R, Marsters SA, Baker CL (1990) Recombinant human DNase I
reduces the viscosity of cystic fibrosis sputum. Proc Natl Acad Sci U S A 87(23):9188–9192
7. Quan JM, Tiddens HA, Sy JP, McKenzie SG, Montgomery MD, Robinson PJ, Wohl ME,
Konstan MW, Pulmozyme Early Intervention Trial Study G (2001) A two-year randomized,
placebo-controlled trial of dornase alfa in young patients with cystic fibrosis with mild lung
function abnormalities. J Pediatr 139(6):813–820
8. Corrie PG (2008) Cytotoxic chemotherapy: clinical aspects. Medicine 36(1):24–28
9. Greco O, Dachs GU (2001) Gene directed enzyme/prodrug therapy of cancer: historical
appraisal and future prospectives. J Cell Physiol 187(1):22–36
10. Moolten FL (1986) Tumor chemosensitivity conferred by inserted herpes thymidine kinase
genes: paradigm for a prospective cancer control strategy. Cancer Res 46(10):5276–5281
11. Encell LP, Landis DM, Loeb LA (1999) Improving enzymes for cancer gene therapy. Nat
Biotechnol 17(2):143–147
12. Duarte S, Carle G, Faneca H, De Lima MCP, Pierrefite-Carle V (2012) Suicide gene therapy
in cancer: where do we stand now? Cancer Lett 324(2):160–170
13. Zarogoulidis, P, Darwiche, K, Sakkas, A, Yarmus, L, Huang, H, Li, Q, Freitag, L, Zarogoulidis,
K, Malecki, M (2013) Suicide gene therapy for cancer – current strategies. J Genet Syndr Gene
Ther 4 pii:16849
14. Hamada Y, Kiso Y (2016) New directions for protease inhibitors directed drug discovery.
Biopolymers 106(4):563–579
15. Kurokawa M, Ito T, Yang CS, Zhao C, Macintyre AN, Rizzieri DA, Rathmell JC, Deininger
MW, Reya T, Kornbluth S (2013) Engineering a BCR-ABL-activated caspase for the selective
elimination of leukemic cells. Proc Natl Acad Sci U S A 110(6):2300–2305
16. Noppen B, Fonteyn L, Aerts F, De Vriese A, De Maeyer M, Le Floch F, Barbeaux P, Zwaal R,
Vanhove M (2014) Autolytic degradation of ocriplasmin: a complex mechanism unraveled by
mutational analysis. Protein Eng Des Sel 27(7):215–223
17. Craik CS, Page MJ, Madison EL (2011) Proteases as therapeutics. Biochem J 435(1):1–16
18. Collen D, Lijnen HR (2009) The tissue-type plasminogen activator story. Arterioscler Thromb
Vasc Biol 29(8):1151–1155
19. Hoylaerts M, Rijken DC, Lijnen HR, Collen D (1982) Kinetics of the activation of plasmino-
gen by human tissue plasminogen activator. Role of fibrin. J Biol Chem 257(6):2912–2919
20. Andreasen PA, Egelund R, Petersen HH (2000) The plasminogen activation system in tumor
growth, invasion, and metastasis. Cell Mol Life Sci 57(1):25–40
21. Semba CP, Sugimoto K, Razavi MK, Society of C, Interventional R (2001) Alteplase and
tenecteplase: applications in the peripheral circulation. Tech Vasc Interv Radiol 4(2):99–106
22. Keyt BA, Paoni NF, Refino CJ, Berleau L, Nguyen H, Chow A, Lai J, Pena L, Pater C, Ogez
J et al (1994) A faster-acting and more potent form of tissue plasminogen activator. Proc Natl
Acad Sci U S A 91(9):3670–3674
23. Bennett WF, Paoni NF, Keyt BA, Botstein D, Jones AJ, Presta L, Wurm FM, Zoller MJ (1991)
High resolution analysis of functional determinants on human tissue-type plasminogen activa-
tor. J Biol Chem 266(8):5191–5201
56 S. Lutz et al.
24. Refino CJ, Paoni NF, Keyt BA, Pater CS, Badillo JM, Wurm FM, Ogez J, Bennett WF (1993)
A variant of t-PA (T103N, KHRR 296-299 AAAA) that, by bolus, has increased potency and
decreased systemic activation of plasminogen. Thromb Haemost 70(2):313–319
25. Llevadot J, Giugliano RP (2000) Pharmacology and clinical trial results of lanoteplase in acute
myocardial infarction. Exp Opin Inv Drugs 9(11):2689–2694
26. Lundblad RL, Bradshaw RA, Gabriel D, Ortel TL, Lawson J, Mann KG (2004) A review of
the therapeutic uses of thrombin. Thromb Haemost 91(5):851–860
27. Gibbs CS, Coutre SE, Tsiang M, Li WX, Jain AK, Dunn KE, Law VS, Mao CT, Matsumura
SY, Mejza SJ et al (1995) Conversion of thrombin into an anticoagulant by protein engineer-
ing. Nature 378(6555):413–416
28. Tsiang M, Paborsky LR, Li WX, Jain AK, Mao CT, Dunn KE, Lee DW, Matsumura SY,
Matteucci MD, Coutre SE, Leung LL, Gibbs CS (1996) Protein engineering thrombin for
optimal specificity and potency of anticoagulant activity in vivo. Biochemistry
35(51):16449–16457
29. Marino F, Pelc LA, Vogt A, Gandhi PS, Di Cera E (2010) Engineering thrombin for selective
specificity toward protein C and PAR1. J Biol Chem 285(25):19145–19152
30. Gruber A, Cantwell AM, Di Cera E, Hanson SR (2002) The thrombin mutant W215A/E217A
shows safe and potent anticoagulant and antithrombotic effects in vivo. J Biol Chem
277(31):27581–27584
31. Berny-Lang MA, Hurst S, Tucker EI, Pelc LA, Wang RK, Hurn PD, Di Cera E, McCarty OJ,
Gruber A (2011) Thrombin mutant W215A/E217A treatment improves neurological outcome
and reduces cerebral infarct size in a mouse model of ischemic stroke. Stroke
42(6):1736–1741
32. Arosio D, Ayala YM, Di Cera E (2000) Mutation of W215 compromises thrombin cleavage of
fibrinogen, but not of PAR-1 or protein C. Biochemistry 39(27):8095–8101
33. Persson E, Kjalke M, Olsen OH (2001) Rational design of coagulation factor VIIa variants with
substantially increased intrinsic activity. Proc Natl Acad Sci U S A 98(24):13583–13588
34. Lentz SR, Ehrenforth S, Karim FA, Matsushita T, Weldingh KN, Windyga J, Mahlangu JN,
adept™2 investigators (2014) Recombinant factor VIIa analog in the management of hemo-
philia with inhibitors: results from a multicenter, randomized, controlled trial of vatreptacog
alfa. J Thromb Haemost 12 (8):1244–1253
35. Harvey SB, Stone MD, Martinez MB, Nelsestuen GL (2003) Mutagenesis of the gamma-
carboxyglutamic acid domain of human factor VII to generate maximum enhancement of the
membrane contact site. J Biol Chem 278(10):8363–8369
36. Neuenschwander PF, Morrissey JH (1994) Roles of the membrane-interactive regions of factor
VIIa and tissue factor. The factor VIIa Gla domain is dispensable for binding to tissue factor
but important for activation of factor X. J Biol Chem 269(11):8007–8013
37. Mahlangu J, Paz P, Hardtke M, Aswad F, Schroeder J (2016) TRUST trial: BAY 86-6150 use
in haemophilia with inhibitors and assessment for immunogenicity. Haemophilia. doi:10.1111/
hae.12994
38. Qian X, Hamad B, Dias-Lalcaca G (2015) The Alzheimer disease market. Nat Rev Drug
Discov 14(10):675–676
39. Jacobsen JS, Comery TA, Martone RL, Elokdah H, Crandall DL, Oganesian A, Aschmies S,
Kirksey Y, Gonzales C, Xu J, Zhou H, Atchison K, Wagner E, Zaleska MM, Das I, Arias RL,
Bard J, Riddell D, Gardell SJ, Abou-Gharbia M, Robichaud A, Magolda R, Vlasuk GP,
Bjornsson T, Reinhart PH, Pangalos MN (2008) Enhanced clearance of Abeta in brain by
sustaining the plasmin proteolysis cascade. Proc Natl Acad Sci U S A 105(25):8754–8759
40. Leissring MA, Farris W, Chang AY, Walsh DM, Wu X, Sun X, Frosch MP, Selkoe DJ (2003)
Enhanced proteolysis of beta-amyloid in APP transgenic mice prevents plaque formation, sec-
ondary pathology, and premature death. Neuron 40(6):1087–1093
41. Spencer B, Marr RA, Rockenstein E, Crews L, Adame A, Potkar R, Patrick C, Gage FH,
Verma IM, Masliah E (2008) Long-term neprilysin gene transfer is associated with reduced
levels of intracellular Abeta and behavioral improvement in APP transgenic mice. BMC
Neurosci 9:109
42. Sexton T, Hitchcook LJ, Rodgers DW, Bradley LH, Hersh LB (2012) Active site mutations
change the cleavage specificity of neprilysin. PLoS One 7(2):e32343
43. Webster CI, Burrell M, Olsson LL, Fowler SB, Digby S, Sandercock A, Snijder A, Tebbe J,
Haupts U, Grudzinska J, Jermutus L, Andersson C (2014) Engineering neprilysin activity and
specificity to create a novel therapeutic for Alzheimer’s disease. PLoS One 9(8):
e104001
44. Eckman EA, Adams SK, Troendle FJ, Stodola BA, Kahn MA, Fauq AH, Xiao HD, Bernstein
KE, Eckman CB (2006) Regulation of steady-state beta-amyloid levels in the brain by nepri-
lysin and endothelin-converting enzyme but not angiotensin-converting enzyme. J Biol Chem
281(41):30471–30478
45. Guerrero JL, O’Malley MA, Daugherty PS (2016) Intracellular FRET-based screen for rede-
signing the specificity of secreted proteases. ACS Chem Biol 11(4):961–970
46. Dressler D (2012) Clinical applications of botulinum toxin. Curr Opin Microbiol

15(3):325–336
47. Orsini M, Leite MA, Chung TM, Bocca W, de Souza JA, de Souza OG, Moreira RP, Bastos
VH, Teixeira S, Oliveira AB, Moraes Bda S, Matta AP, Jacinto LJ (2015) Botulinum neuro-
toxin type A in neurology: update. Neurol Int 7(2):5886
48. Masuyer G, Chaddock JA, Foster KA, Acharya KR (2014) Engineered botulinum neurotoxins
as new therapeutics. Annu Rev Pharmacol Toxicol 54:27–51
49. Vazquez-Cintron E, Tenezaca L, Angeles C, Syngkon A, Liublinska V, Ichtchenko K, Band P
(2016) Pre-clinical study of a novel recombinant botulinum neurotoxin derivative engineered
for improved safety. Sci Rep 6:30429
50. Rummel A, Mahrhold S, Bigalke H, Binz T (2011) Exchange of the H(CC) domain mediating
double receptor recognition improves the pharmacodynamic properties of botulinum neuro-
toxin. FEBS J 278(23):4506–4515
51. Wang J, Zurawski TH, Bodeker MO, Meng J, Boddul S, Aoki KR, Dolly JO (2012) Longer-
acting and highly potent chimaeric inhibitors of excessive exocytosis created with domains
from botulinum neurotoxin A and B. Biochem J 444(1):59–67
52. Wang J, Zurawski TH, Meng J, Lawrence G, Olango WM, Finn DP, Wheeler L, Dolly JO
(2011) A dileucine in the protease of botulinum toxin A underlies its long-lived neuroparaly-
sis: transfer of longevity to a novel potential therapeutic. J Biol Chem 286(8):6375–6385
53. Chaddock JA, Purkiss JR, Friis LM, Broadbridge JD, Duggan MJ, Fooks SJ, Shone CC, Quinn
CP, Foster KA (2000) Inhibition of vesicular secretion in both neuronal and nonneuronal cells
by a retargeted endopeptidase derivative of Clostridium botulinum neurotoxin type A. Infect
Immun 68(5):2587–2593
54. Fonfria E, Donald S, Cadd VA (2016) Botulinum neurotoxin A and an engineered derivate
targeted secretion inhibitor (TSI) A enter cells via different vesicular compartments. J Recept
Signal Transduct Res 36(1):79–88
55. Ma H, Meng J, Wang J, Hearty S, Dolly JO, O’Kennedy R (2014) Targeted delivery of
a SNARE protease to sensory neurons using a single chain antibody (scFv) against the
extracellular domain of P2X(3) inhibits the release of a pain mediator. Biochem J 462(2):
247–256
56. Larsen GR, Timony GA, Horgan PG, Barone KM, Hensen KS, Augus LB, Stoudemire JB
(1991) Protein engineering of novel plasminogen activators with increased thrombolytic
potency in rabbits relative to activase. J Biol Chem 266(1):8156–8161
57. Sikorra S, Litschko C, Muller C, Thiel N, Galli T, Eichner T, Binz T (2016) Identification and
characterization of botulinum neurotoxin a substrate binding pockets and their re-engineering
for human SNAP-23. J Mol Biol 428(2 Pt A):372–384
58. Chen S, Barbieri JT (2009) Engineering botulinum neurotoxin to extend therapeutic interven-
tion. Proc Natl Acad Sci U S A 106(23):9180–9184
59. Somm E, Bonnet N, Martinez A, Marks PM, Cadd VA, Elliott M, Toulotte A, Ferrari SL,
Rizzoli R, Huppi PS, Harper E, Melmed S, Jones R, Aubert ML (2012) A botulinum toxin-
derived targeted secretion inhibitor downregulates the GH/IGF1 axis. J Clin Invest
122(9):3295–3306
58 S. Lutz et al.
60. Tye-Din JA, Anderson RP, Ffrench RA, Brown GJ, Hodsman P, Siegel M, Botwick W,
Shreeniwas R (2010) The effects of ALV003 pre-digestion of gluten on immune response and
symptoms in celiac disease in vivo. Clin Immunol 134(3):289–295
61. Ehren J, Govindarajan S, Moron B, Minshull J, Khosla C (2008) Protein engineering of improved
prolyl endopeptidases for celiac sprue therapy. Protein Eng Des Sel 21(12):699–707
62. Gordon SR, Stanley EJ, Wolf S, Toland A, Wu SJ, Hadidi D, Mills JH, Baker D, Pultz IS,
Siegel JB (2012) Computational design of an alpha-gliadin peptidase. J Am Chem Soc
134(50):20513–20520
63. Tack GJ, Verbeek WH, Schreurs MW, Mulder CJ (2010) The spectrum of celiac disease:
epidemiology, clinical aspects and treatment. Nat Rev Gastroenterol Hepatol 7(4):204–213
64. Shan L, Molberg O, Parrot I, Hausch F, Filiz F, Gray GM, Sollid LM, Khosla C (2002)
Structural basis for gluten intolerance in celiac sprue. Science 297(5590):2275–2279
65. Wolf C, Siegel JB, Tinberg C, Camarca A, Gianfrani C, Paski S, Guan R, Montelione G, Baker
D, Pultz IS (2015) Engineering of Kuma030: a gliadin peptidase that rapidly degrades immu-
nogenic gliadin peptides in gastric conditions. J Am Chem Soc 137(40):13106–13113
66. Guerrero JL, Daugherty PS, O’Malley MA (2016) Emerging technologies for protease engi-
neering: new tools to clear out disease. Biotechnol Bioeng. doi:10.1002/bit.26066 [Epub ahead
of print]
67. Esvelt KM, Carlson JC, Liu DR (2011) A system for the continuous directed evolution of
biomolecules. Nature 472(7344):499–503
68. Hill ME, MacPherson DJ, Wu P, Julien O, Wells JA, Hardy JA (2016) Reprogramming cas-
pase-7 specificity by regio-specific mutations and selection provides alternate solutions for
substrate recognition. ACS Chem Biol 11(6):1603–1612
69. Sandersjoo L, Kostallas G, Lofblom J, Samuelson P (2014) A protease substrate profiling
method that links site-specific proteolysis with antibiotic resistance. Biotechnol
J 9(1):155–162
70. Varadarajan N, Rodriguez S, Hwang BY, Georgiou G, Iverson BL (2008) Highly active and
selective endopeptidases with programmed substrate specificities. Nat Chem Biol
4(5):290–294
71. Kostallas G, Samuelson P (2010) Novel fluorescence-assisted whole-cell assay for engineering
and characterization of proteases and their substrates. Appl Environ Microbiol
76(22):7500–7508
72. Carrico ZM, Strobel KL, Atreya ME, Clark DS, Francis MB (2016) Simultaneous selection
and counter-selection for the directed evolution of proteases in E. coli using a cytoplasmic
anchoring strategy. Biotechnol Bioeng 113(6):1187–1193
73. Dickinson BC, Packer MS, Badran AH, Liu DR (2014) A system for the continuous directed
evolution of proteases rapidly reveals drug-resistance mutations. Nat Commun 5:5352
74. Ledoux L (1955) Action of ribonuclease on two solid tumours in vivo. Nature

176(4470):36–37
75. Vogelzang NJ, Aklilu M, Stadler WM, Dumas MC, Mikulski SM (2001) A phase II trial of
weekly intravenous ranpirnase (Onconase), a novel ribonuclease in patients with metastatic
kidney cancer. Investig New Drugs 19(3):255–260
76. Mikulski SM, Costanzi JJ, Vogelzang NJ, McCachren S, Taub RN, Chun H, Mittelman A,
Panella T, Puccio C, Fine R, Shogen K (2002) Phase II trial of a single weekly intravenous
dose of ranpirnase in patients with unresectable malignant mesothelioma. J Clin Oncol
20(1):274–281
77. Rutkoski TJ, Raines RT (2008) Evasion of ribonuclease inhibitor as a determinant of ribonu-
clease cytotoxicity. Curr Pharm Biotechnol 9(3):185–189
78. Kobe B, Deisenhofer J (1996) Mechanism of ribonuclease inhibition by ribonuclease inhibitor
protein based on the crystal structure of its complex with ribonuclease A. J Mol Biol
264(5):1028–1043
79. Leland PA, Schultz LW, Kim BM, Raines RT (1998) Ribonuclease A variants with potent
cytotoxic activity. Proc Natl Acad Sci U S A 95(18):10407–10412
80. Rutkoski TJ, Kurten EL, Mitchell JC, Raines RT (2005) Disruption of shape-complementarity
markers to create cytotoxic variants of ribonuclease A. J Mol Biol 354(1):41–54
81. Johnson RJ, Chao TY, Lavis LD, Raines RT (2007) Cytotoxic ribonucleases: the dichotomy of
Coulombic forces. Biochemistry 46(36):10308–10316
82. Johnson RJ, McCoy JG, Bingman CA, Phillips GN Jr, Raines RT (2007) Inhibition of human
pancreatic ribonuclease by the human ribonuclease inhibitor protein. J Mol Biol
368(2):434–449
83. Cremer C, Braun H, Mladenov R, Schenke L, Cong X, Jost E, Brummendorf TH, Fischer R,
Carloni P, Barth S, Nachreiner T (2015) Novel angiogenin mutants with increased cytotoxicity
enhance the depletion of pro-inflammatory macrophages and leukemia cells ex vivo. Cancer
Immunol Immunother 64(12):1575–1586
84. Cong X, Cremer C, Nachreiner T, Barth S, Carloni P (2016) Engineered human angiogenin
mutations in the placental ribonuclease inhibitor complex for anticancer therapy: insights from
enhanced sampling simulations. Protein Sci. doi:10.1002/pro.2941
85. Riccio G, D’Avino C, Raines RT, De Lorenzo C (2013) A novel fully human antitumor immu-
noRNase resistant to the RNase inhibitor. Protein Eng Des Sel 26(3):243–248
86. Phillips MM, Sheaff MT, Szlosarek PW (2013) Targeting arginine-dependent cancers with
arginine-degrading enzymes: opportunities and challenges. Cancer Res Treat 45(4):251–262
87. Patil MD, Bhaumik J, Babykutty S, Banerjee UC, Fukumura D (2016) Arginine dependence
of tumor cells: targeting a chink in cancer’s armor. Oncogene. doi:10.1038/onc.2016.37
88. Boissel N, Sender LS (2015) Best practices in adolescent and young adult patients with acute
lymphoblastic leukemia: a focus on asparaginase. Young Adult Oncol 4(3):118–128
89. Pulte D, Gondos A, Brenner H (2009) Improvement in survival in younger patients with acute
lymphoblastic leukemia from the 1980s to the early 21st century. Blood 113(7):1408–1411
90. Liu C, Kawedia JD, Cheng C, Pei D, Fernandez CA, Cai X, Crews KR, Kaste SC, Panetta JC,
Bowman WP, Jeha S, Sandlund JT, Evans WE, Pui CH, Relling MV (2012) Clinical utility and
implications of asparaginase antibodies in acute lymphoblastic leukemia. Leukemia
26(11):2303–2309
91. Cantor JR, Yoo TH, Dixit A, Iverson BL, Forsthuber TG, Georgiou G (2011) Therapeutic
enzyme deimmunization by combinatorial T-cell epitope removal using neutral drift. Proc Natl
Acad Sci U S A 108(4):1272–1277
92. Maggi M, Chiarelli LR, Valentini G, Scotti C (2015) Engineering of Helicobacter pylori
L-asparaginase: characterization of two functionally distinct groups of mutants. PLoS One
10(2):e0117025
93. Bansal S, Srivastava A, Mukherjee G, Pandey R, Verma AK, Mishra P, Kundu B (2012)
Hyperthermophilic asparaginase mutants with enhanced substrate affinity and antineoplastic
activity: structural insights on their mechanism of action. FASEB J 26(3):1161–1171
94. Kotzia GA, Labrou NE (2009) Engineering thermal stability of L-asparaginase by in vitro
directed evolution. FEBS J 276(6):1750–1761
95. Figueiredo L, Cole PD, Drachtman RA (2016) Asparaginase Erwinia chrysanthemi as a com-
ponent of a multi-agent chemotherapeutic regimen for the treatment of patients with acute
lymphoblastic leukemia who have developed hypersensitivity to E. coli-derived asparaginase.
Expert Rev Hematol 9(3):227–234
96. Feun L, Savaraj N (2006) Pegylated arginine deiminase: a novel anticancer enzyme agent.
Expert Opin Investig Drugs 15(7):815–822
97. Glazer ES, Piccirillo M, Albino V, Di Giacomo R, Palaia R, Mastro AA, Beneduce G, Castello
G, De Rosa V, Petrillo A, Ascierto PA, Curley SA, Izzo F (2010) Phase II study of pegylated
arginine deiminase for nonresectable and metastatic hepatocellular carcinoma. J Clin Oncol
28(13):2220–2226
98. Ott PA, Carvajal RD, Pandit-Taskar N, Jungbluth AA, Hoffman EW, Wu BW, Bomalaski JS,
Venhaus R, Pan L, Old LJ, Pavlick AC, Wolchok JD (2013) Phase I/II study of pegylated argi-
nine deiminase (ADI-PEG 20) in patients with advanced melanoma. Investig New Drugs
31(2):425–434
99. Tomlinson BK, Thomson JA, Bomalaski JS, Diaz M, Akande T, Mahaffey N, Li T, Dutia MP,
Kelly K, Gong IY, Semrad T, Gandara DR, Pan CX, Lara PN Jr (2015) Phase I trial of arginine
deprivation therapy with ADI-PEG 20 plus docetaxel in patients with advanced malignant
solid tumors. Clin Cancer Res 21(11):2480–2486
60 S. Lutz et al.
100. Liu Y-M, Sun Z-H, Ni Y, Zheng P, Liu Y-P, Meng F-J (2008) Isolation and identification of
an arginine deiminase producing strain Pseudomonas plecoglossicida CGMCC2039. World
J Microbiol Biotechnol 24(10):2213–2219
101. Han RZ, Xu GC, Dong JJ, Ni Y (2016) Arginine deiminase: recent advances in discovery,
crystal structure, and protein engineering for improved properties as an anti-tumor drug. Appl
Microbiol Biotechnol 100(11):4747–4760
102. Ni Y, Li Z, Sun Z, Zheng P, Liu Y, Zhu L, Schwaneberg U (2009) Expression of arginine
deiminase from Pseudomonas plecoglossicida CGMCC2039 in E. coli and its anti-tumor
activity. Curr Microbiol 58(6):593–598
103. Cheng F, Kardashliev T, Pitzler C, Shehzad A, Lue H, Bernhagen J, Zhu L, Schwaneberg U
(2015) A competitive flow cytometry screening system for directed evolution of therapeutic
enzyme. ACS Synth Biol 4(7):768–775
104. Ni Y, Schwaneberg U, Sun ZH (2008) Arginine deiminase, a potential anti-tumor drug.
Cancer Lett 261(1):1–11
105. Kuhn NJ, Talbot J, Ward S (1991) pH-sensitive control of arginase by Mn(II) ions at submi-
cromolar concentrations. Arch Biochem Biophys 286(1):217–221
106. Stone EM, Glazer ES, Chantranupong L, Cherukuri P, Breece RM, Tierney DL, Curley SA,
Iverson BL, Georgiou G (2010) Replacing Mn(2+) with Co(2+) in human arginase i enhances
cytotoxicity toward l-arginine auxotrophic cancer cell lines. ACS Chem Biol 5(3):333–342
107. Stone E, Chantranupong L, Gonzalez C, O’Neal J, Rani M, VanDenBerg C, Georgiou G
(2012) Strategies for optimizing the serum persistence of engineered human arginase I for
cancer therapy. J Control Release 158(1):171–179
108. Jordheim LP, Durantel D, Zoulim F, Dumontet C (2013) Advances in the development of
nucleoside and nucleotide analogues for cancer and viral diseases. Nat Rev Drug Discov
12(6):447–464
109. Baldwin SA, Beal PR, Yao SY, King AE, Cass CE, Young JD (2004) The equilibrative nucle-
oside transporter family, SLC29. Pflugers Arch 447(5):735–743
110. Arner ES, Eriksson S (1995) Mammalian deoxyribonucleoside kinases. Pharmacol Ther
67(2):155–186
111. Shi J, McAtee JJ, Schlueter Wirtz S, Tharnish P, Juodawlkis A, Liotta DC, Schinazi RF
(1999) Synthesis and biological evaluation of 2′,3′-didehydro-2′,3′- dideoxy-5-fluorocytidine
(D4FC) analogues: discovery of carbocyclic nucleoside triphosphates with potent inhibitory
activity against HIV-1 reverse transcriptase. J Med Chem 42(5):859–867
112. Culver KW, Ram Z, Wallbridge S, Ishii H, Oldfield EH, Blaese RM (1992) In vivo gene
transfer with retroviral vector-producer cells for treatment of experimental brain tumors.
Science 256(5063):1550–1552
113. Song S, Pursell ZF, Copeland WC, Longley MJ, Kunkel TA, Mathews CK (2005) DNA
precursor asymmetries in mammalian tissue mitochondria and possible contribution to
mutagenesis through reduced replication fidelity. Proc Natl Acad Sci U S A
102(14):4990–4995
114. Trask TW, Trask RP, Aguilar-Cordova E, Shine HD, Wyde PR, Goodman JC, Hamilton WJ,
Rojas-Martinez A, Chen SH, Woo SL, Grossman RG (2000) Phase I study of adenoviral
delivery of the HSV-tk gene and ganciclovir administration in patients with current malignant
brain tumors. Mol Ther 1(2):195–203
115. Ji N, Weng D, Liu C, Gu Z, Chen S, Guo Y, Fan Z, Wang X, Chen J, Zhao Y, Zhou J, Wang
J, Ma D, Li N (2016) Adenovirus-mediated delivery of herpes simplex virus thymidine kinase
administration improves outcome of recurrent high-grade glioma. Oncotarget
7(4):4369–4378
116. Wheeler LA, Manzanera AG, Bell SD, Cavaliere R, McGregor JM, Grecula JC, Newton HB, Lo
SS, Badie B, Portnow J, Teh BS, Trask TW, Baskin DS, New PZ, Aguilar LK, Aguilar-Cordova
E, Chiocca EA (2016) Phase II multicenter study of gene-mediated cytotoxic immunotherapy
as adjuvant to surgical resection for newly diagnosed malignant glioma. Neuro-Oncology
18(8):1137–1145
117. Wigler M, Silverstein S, Lee LS, Pellicer A, Cheng Y, Axel R (1977) Transfer of purified
herpes virus thymidine kinase gene to cultured mouse cells. Cell 11(1):223–232
118. Skorenski M, Sienczyk M (2014) Anti-herpesvirus agents: a patent and literature review
(2003 to present). Expert Opin Ther Patents 24(8):925–941
119. Munir KM, French DC, Dube DK, Loeb LA (1992) Permissible amino acid substitutions
within the putative nucleoside binding site of Herpes Simplex virus type 1 encoded thymidine
kinase established by random sequence mutagenesis [corrected]. J Biol Chem
267(10):6584–6589
120. Black ME, Loeb LA (1993) Identification of important residues within the putative nucleo-
side binding site of HSV-1 thymidine kinase by random sequence selection: analysis of
selected mutants in vitro. Biochemistry 32(43):11618–11626
121. Black ME, Newcomb TG, Wilson HM, Loeb LA (1996) Creation of drug-specific Herpes
Simplex virus type 1 thymidine kinase mutants for gene therapy. Proc Natl Acad Sci U S A
93(8):3525–3529
122. Christians FC, Scapozza L, Crameri A, Folkers G, Stemmer WP (1999) Directed evolution of
thymidine kinase for AZT phosphorylation using DNA family shuffling. Nat Biotechnol
17(3):259–264
123. Hinds TA, Compadre C, Hurlburt BK, Drake RR (2000) Conservative mutations of gluta-
mine-125 in Herpes Simplex virus type 1 thymidine kinase result in a ganciclovir kinase with
minimal deoxypyrimidine kinase activities. Biochemistry 39(14):4105–4111
124. Black ME, Kokoris MS, Sabo P (2001) Herpes Simplex virus-1 thymidine kinase mutants
created by semi-random sequence mutagenesis improve prodrug-mediated tumor cell killing.
Cancer Res 61(7):3022–3026
125. Balzarini J, Liekens S, Solaroli N, El Omari K, Stammers DK, Karlsson A (2006) Engineering
of a single conserved amino acid residue of Herpes Simplex virus type 1 thymidine kinase
allows a predominant shift from pyrimidine to purine nucleoside phosphorylation. J Biol
Chem 281(28):19273–19279
126. Munch-Petersen B, Piskur J, Sondergaard L (1998) Four deoxynucleoside kinase activities
from Drosophila melanogaster are contained within a single monomeric enzyme, a new mul-
tifunctional deoxynucleoside kinase. J Biol Chem 273(7):3926–3931
127. Knecht W, Munch-Petersen B, Piskur J (2000) Identification of residues involved in the spec-
ificity and regulation of the highly efficient multisubstrate deoxyribonucleoside kinase from
Drosophila melanogaster. J Mol Biol 301(4):827–837
128. Knecht W, Sandrini MP, Johansson K, Eklund H, Munch-Petersen B, Piskur J (2002) A few
amino acid substitutions can convert deoxyribonucleoside kinase specificity from pyrimi-
dines to purines. EMBO J 21(7):1873–1880
129. Knecht W, Rozpedowska E, Le Breton C, Willer M, Gojkovic Z, Sandrini MP, Joergensen T,
Hasholt L, Munch-Petersen B, Piskur J (2007) Drosophila deoxyribonucleoside kinase
mutants with enhanced ability to phosphorylate purine analogs. Gene Ther
14(17):1278–1286
130. Gerth ML, Lutz S (2007) Non-homologous recombination of deoxyribonucleoside kinases
from human and Drosophila melanogaster yields human-like enzymes with novel activities.
J Mol Biol 370(4):742–751
131. Solaroli N, Johansson M, Balzarini J, Karlsson A (2007) Enhanced toxicity of purine nucleo-
side analogs in cells expressing Drosophila melanogaster nucleoside kinase mutants. Gene
Ther 14(1):86–92
132. Liu L, Li Y, Liotta D, Lutz S (2009) Directed evolution of an orthogonal nucleoside analog
kinase via fluorescence-activated cell sorting. Nucleic Acids Res 37(13):4472–4481
133. Liu L, Murphy P, Baker D, Lutz S (2010) Computational design of orthogonal nucleoside
kinases. Chem Commun (Camb) 46(46):8803–8805
134. Campbell DO, Yaghoubi SS, Su Y, Lee JT, Auerbach MS, Herschman H, Satyamurthy N,
Czernin J, Lavie A, Radu CG (2012) Structure-guided engineering of human thymidine
kinase 2 as a positron emission tomography reporter gene for enhanced phosphorylation of
non-natural thymidine analog reporter probe. J Biol Chem 287(1):446–454
135. Sabini E, Hazra S, Konrad M, Burley SK, Lavie A (2007) Structural basis for activation of
the therapeutic L-nucleoside analogs 3TC and troxacitabine by human deoxycytidine kinase.
Nucleic Acids Res 35(1):186–192
62 S. Lutz et al.
136. Iyidogan P, Lutz S (2008) Systematic exploration of active site mutations on human deoxy-
cytidine kinase substrate specificity. Biochemistry 47(16):4711–4720
137. Hazra S, Sabini E, Ort S, Konrad M, Lavie A (2009) Extending thymidine kinase activity to
the catalytic repertoire of human deoxycytidine kinase. Biochemistry 48(6):1256–1263
138. Muthu P, Chen HX, Lutz S (2014) Redesigning human 2′-deoxycytidine kinase enantioselec-
tivity for L-nucleoside analogues as reporters in positron emission tomography. ACS Chem
Biol 9(10):2326–2333
139. Kokoris MS, Sabo P, Adman ET, Black ME (1999) Enhancement of tumor ablation by a
selected HSV-1 thymidine kinase mutant. Gene Ther 6(8):1415–1426
140. Ardiani A, Sanchez-Bonilla M, Black ME (2010) Fusion enzymes containing HSV-1 thymi-
dine kinase mutants and guanylate kinase enhance prodrug sensitivity in vitro and in vivo.
Cancer Gene Ther 17(2):86–96
141. Karjoo Z, Chen X, Hatefi A (2016) Progress and problems with the use of suicide genes for
targeted cancer therapy. Adv Drug Deliv Rev 99(Pt A):113–128
142. Balzarini J, Liekens S, Esnouf R, De Clercq E (2002) The A167Y mutation converts the
Herpes Simplex virus type 1 thymidine kinase into a guanosine analogue kinase. Biochemistry
41(20):6517–6524
143. Slot Christiansen L, Munch-Petersen B, Knecht W (2015) Non-viral deoxyribonucleoside
kinases – diversity and practical use. J Genet Genomics 42(5):235–248
144. Welin M, Skovgaard T, Knecht W, Zhu C, Berenstein D, Munch-Petersen B, Piskur J, Eklund
H (2005) Structural basis for the changed substrate specificity of Drosophila melanogaster
deoxyribonucleoside kinase mutant N64D. FEBS J 272(14):3733–3742
145. Hecker SJ, Erion MD (2008) Prodrugs of phosphates and phosphonates. J Med Chem
51(8):2328–2345
146. Thornton P, Kadri H, Micolli A, Mehellou Y (2016) Nucleoside phosphate and phosphonate
prodrug clinical candidates. J Med Chem. doi:10.1021/acs.jmedchem.6b00523
147. Eriksson S, Kierdaszuk B, Munch-Petersen B, Oberg B, Johansson NG (1991) Comparison
of the substrate specificities of human thymidine kinase 1 and 2 and deoxycytidine kinase
toward antiviral and cytostatic nucleoside analogs. Biochem Biophys Res Commun 176(2):
586–592
148. Serganova I, Ponomarev V, Blasberg R (2007) Human reporter genes: potential use in clinical
studies. Nucl Med Biol 34(7):791–807
149. Yaghoubi SS, Campbell DO, Radu CG, Czernin J (2012) Positron emission tomography reporter
genes and reporter probes: gene and cell therapy applications. Theranostics 2(4):374–391
150. Gambhir SS, Bauer E, Black ME, Liang Q, Kokoris MS, Barrio JR, Iyer M, Namavari M,
Phelps ME, Herschman HR (2000) A mutant Herpes Simplex virus type 1 thymidine kinase
reporter gene shows improved sensitivity for imaging reporter gene expression with positron
emission tomography. Proc Natl Acad Sci U S A 97(6):2785–2790
151. Jacobs A, Tjuvajev JG, Dubrovin M, Akhurst T, Balatoni J, Beattie B, Joshi R, Finn R,
Larson SM, Herrlinger U, Pechan PA, Chiocca EA, Breakefield XO, Blasberg RG (2001)
Positron emission tomography-based imaging of transgene expression mediated by
replication-conditional, oncolytic Herpes Simplex virus type 1 mutant vectors in vivo. Cancer
Res 61(7):2983–2995
152. Dempsey MF, Wyper D, Owens J, Pimlott S, Papanastassiou V, Patterson J, Hadley DM,
Nicol A, Rampling R, Brown SM (2006) Assessment of 123I-FIAU imaging of herpes sim-
plex viral gene expression in the treatment of glioma. Nucl Med Commun 27(8):611–617
153. Likar Y, Zurita J, Dobrenkov K, Shenker L, Cai S, Neschadim A, Medin JA, Sadelain M,
Hricak H, Ponomarev V (2010) A new pyrimidine-specific reporter gene: a mutated human
deoxycytidine kinase suitable for PET during treatment with acycloguanosine-based cyto-
toxic drugs. J Nucl Med 51(9):1395–1403
154. Gil JS, Machado HB, Campbell DO, McCracken M, Radu C, Witte ON, Herschman HR
(2013) Application of a rapid, simple, and accurate adenovirus-based method to compare
PET reporter gene/PET reporter probe systems. Mol Imaging Biol 15(3):273–281
155. Wang L, Munch-Petersen B, Herrstrom Sjoberg A, Hellman U, Bergman T, Jornvall H,

Eriksson S (1999) Human thymidine kinase 2: molecular cloning and characterisation of the
enzyme activity with antiviral and cytostatic nucleoside substrates. FEBS Lett
443(2):170–174
156. Ponomarev V, Doubrovin M, Shavrin A, Serganova I, Beresten T, Ageyeva L, Cai C, Balatoni
J, Alauddin M, Gelovani J (2007) A human-derived reporter gene for noninvasive imaging in
humans: mitochondrial thymidine kinase type 2. J Nucl Med 48(5):819–826
157. Mullen CA, Kilstrup M, Blaese RM (1992) Transfer of the bacterial gene for cytosine deami-
nase to mammalian cells confers lethal sensitivity to 5-fluorocytosine: a negative selection
system. Proc Natl Acad Sci U S A 89(1):33–37
158. Mahan SD, Ireton GC, Stoddard BL, Black ME (2004) Alanine-scanning mutagenesis reveals
a cytosine deaminase mutant with altered substrate preference. Biochemistry
43(28):8957–8964
159. Mahan SD, Ireton GC, Knoeber C, Stoddard BL, Black ME (2004) Random mutagenesis and
selection of Escherichia coli cytosine deaminase for cancer gene therapy. Protein Eng Des
Sel 17(8):625–633
160. Fuchita M, Ardiani A, Zhao L, Serve K, Stoddard BL, Black ME (2009) Bacterial cytosine
deaminase mutants created by molecular engineering show improved 5-fluorocytosine-
mediated cell killing in vitro and in vivo. Cancer Res 69(11):4791–4799
161. Korkegian A, Black ME, Baker D, Stoddard BL (2005) Computational thermostabilization of
an enzyme. Science 308(5723):857–860
162. Stolworthy TS, Korkegian AM, Willmon CL, Ardiani A, Cundiff J, Stoddard BL, Black ME
(2008) Yeast cytosine deaminase mutants with increased thermostability impart sensitivity to
5-fluorocytosine. J Mol Biol 377(3):854–869
163. Rogulski KR, Kim JH, Kim SH, Freytag SO (1997) Glioma cells transduced with an
Escherichia coli CD/HSV-1 TK fusion gene exhibit enhanced metabolic suicide and radio-
sensitivity. Hum Gene Ther 8(1):73–85
164. Chang JW, Lee H, Kim E, Lee Y, Chung SS, Kim JH (2000) Combined antitumor effects of
an adenoviral cytosine deaminase/thymidine kinase fusion gene in rat C6 glioma.
Neurosurgery 47(4):931–938
165. Bennett EM, Anand R, Allan PW, Hassan AE, Hong JS, Levasseur DN, McPherson DT,
Parker WB, Secrist JA 3rd, Sorscher EJ, Townes TM, Waud WR, Ealick SE (2003) Designer
gene therapy using an Escherichia coli purine nucleoside phosphorylase/prodrug system.
Chem Biol 10(12):1173–1181
166. Stoeckler JD, Poirot AF, Smith RM, Parks RE Jr, Ealick SE, Takabayashi K, Erion MD
(1997) Purine nucleoside phosphorylase. 3. Reversal of purine base specificity by site-
directed mutagenesis. Biochemistry 36(39):11749–11756
167. Grove JI, Lovering AL, Guise C, Race PR, Wrighton CJ, White SA, Hyde EI, Searle PF
(2003) Generation of Escherichia coli nitroreductase mutants conferring improved cell sen-
sitization to the prodrug CB1954. Cancer Res 63(17):5532–5537
168. Guise CP, Grove JI, Hyde EI, Searle PF (2007) Direct positive selection for improved nitro-
reductase variants using SOS triggering of bacteriophage lambda lytic cycle. Gene Ther
14(8):690–698
169. Barak Y, Thorne SH, Ackerley DF, Lynch SV, Contag CH, Matin A (2006) New enzyme for
reductive cancer chemotherapy, YieF, and its improvement by directed evolution. Mol Cancer
Ther 5(1):97–103
170. Jaberipour M, Vass SO, Guise CP, Grove JI, Knox RJ, Hu L, Hyde EI, Searle PF (2010)
Testing double mutants of the enzyme nitroreductase for enhanced cell sensitisation to
prodrugs: effects of combining beneficial single mutations. Biochem Pharmacol
79(2):102–111
171. Swe PM, Copp JN, Green LK, Guise CP, Mowday AM, Smaill JB, Patterson AV, Ackerley
DF (2012) Targeted mutagenesis of the Vibrio fischeri flavin reductase FRase I to improve
activation of the anticancer prodrug CB1954. Biochem Pharmacol 84(6):775–783
64 S. Lutz et al.
172. Bzowska A, Kulikowska E, Shugar D (2000) Purine nucleoside phosphorylases: properties,

functions, and clinical aspects. Pharmacol Ther 88(3):349–425
173. Sorscher EJ, Peng S, Bebok Z, Allan PW, Bennett LL Jr, Parker WB (1994) Tumor cell
bystander killing in colonic carcinoma utilizing the Escherichia coli DeoD gene to generate
toxic purines. Gene Ther 1(4):233–238
174. Martiniello-Wilks R, Wang XY, Voeks DJ, Dane A, Shaw JM, Mortensen E, Both GW,
Russell PJ (2004) Purine nucleoside phosphorylase and fludarabine phosphate gene-directed
enzyme prodrug therapy suppresses primary tumour growth and pseudo-metastases in a
mouse model of prostate cancer. J Gene Med 6(12):1343–1357
175. Lukenbill J, Kalaycio M (2013) Fludarabine: a review of the clear benefits and potential
harms. Leuk Res 37(9):986–994
176. Cacciapuoti G, Bagarolo ML, Martino E, Scafuri B, Marabotti A, Porcelli M (2016) Efficient
fludarabine-activating PNP from archaea as a guidance for redesign the active site of E. coli
PNP. J Cell Biochem 117(5):1126–1135
177. Williams EM, Little RF, Mowday AM, Rich MH, Chan-Hyams JV, Copp JN, Smaill JB,
Patterson AV, Ackerley DF (2015) Nitroreductase gene-directed enzyme prodrug therapy:
insights and advances toward clinical utility. Biochem J 471(2):131–153
178. Rylott EL, Budarina MV, Barker A, Lorenz A, Strand SE, Bruce NC (2011) Engineering
plants for the phytoremediation of RDX in the presence of the co-contaminating explosive
TNT. New Phytol 192(2):405–413
179. White DT, Mumm JS (2013) The nitroreductase system of inducible targeted ablation facili-
tates cell-specific regenerative studies in zebrafish. Methods 62(3):232–240
180. Bridgewater J, Springer C, Knox R, Minton N, Michael N, Collins M (1995) Expression of
the bacterial nitroreductase enzyme in mammalian cells renders them selectively sensitive to
killing by the prodrug CB1954. Eur J Cancer 31(13):2362–2370
181. Patel P, Young JG, Mautner V, Ashdown D, Bonney S, Pineda RG, Collins SI, Searle PF, Hull
D, Peers E, Chester J, Wallace DM, Doherty A, Leung H, Young LS, James ND (2009) A
phase I/II clinical trial in localized prostate cancer of an adenovirus expressing nitroreductase
with CB1954. Mol Ther 17(7):1292–1299
182. Lovering AL, Hyde EI, Searle PF, White SA (2001) The structure of Escherichia coli nitro-
reductase complexed with nicotinic acid: three crystal forms at 1.7 A, 1.8 A and 2.4 A resolu-
tion. J Mol Biol 309(1):203–213
183. Vellom DC, Radic Z, Li Y, Pickering NA, Camp S, Taylor P (1993) Amino acid residues
controlling acetylcholinesterase and butyrylcholinesterase specificity. Biochemistry
32(1):12–17
184. Schwarz M, Glick D, Loewenstein Y, Soreq H (1995) Engineering of human cholinesterases
explains and predicts diverse consequences of administration of various drugs and poisons.
Pharmacol Ther 67(2):283–322
185. Lockridge O (2015) Review of human butyrylcholinesterase structure, function, genetic vari-
ants, history of use in the clinic, and potential therapeutic uses. Pharmacol Ther 148:34–46
186. Lenz DE, Yeung D, Smith JR, Sweeney RE, Lumley LA, Cerasoli DM (2007) Stoichiometric
and catalytic scavengers as protection against nerve agent toxicity: a mini review. Toxicology
233(1–3):31–39
187. Parikh K, Duysen EG, Snow B, Jensen NS, Manne V, Lockridge O, Chilukuri N (2011)
Gene-delivered butyrylcholinesterase is prophylactic against the toxicity of chemical war-
fare nerve agents and organophosphorus compounds. J Pharmacol Exp Ther
337(1):92–101
188. Saxena A, Sun W, Fedorko JM, Koplovitz I, Doctor BP (2011) Prophylaxis with human
serum butyrylcholinesterase protects guinea pigs exposed to multiple lethal doses of soman
or VX. Biochem Pharmacol 81(1):164–169
189. Terekhov S, Smirnov I, Bobik T, Shamborant O, Zenkova M, Chernolovskaya E, Gladkikh D,
Murashev A, Dyachenko I, Palikov V, Palikova Y, Knorre V, Belogurov A Jr, Ponomarenko N,
Blackburn GM, Masson P, Gabibov A (2015) A novel expression cassette delivers efficient
production of exclusively tetrameric human butyrylcholinesterase with improved pharmaco-
kinetics for protection against organophosphate poisoning. Biochimie 118:51–59
190. Brimijoin S, Gao Y (2012) Cocaine hydrolase gene therapy for cocaine abuse. Future Med
Chem 4(2):151–162
191. Xie W, Altamirano CV, Bartels CF, Speirs RJ, Cashman JR, Lockridge O (1999) An improved
cocaine hydrolase: the A328Y mutant of human butyrylcholinesterase is 4-fold more effi-
cient. Mol Pharmacol 55(1):83–91
192. Sun H, El Yazal J, Lockridge O, Schopfer LM, Brimijoin S, Pang YP (2001) Predicted
Michaelis-Menten complexes of cocaine-butyrylcholinesterase. Engineering effec-
tive butyrylcholinesterase mutants for cocaine detoxification. J Biol Chem 276(12):
9330–9336
193. Sun H, Pang YP, Lockridge O, Brimijoin S (2002) Re-engineering butyrylcholinesterase as a
cocaine hydrolase. Mol Pharmacol 62(2):220–224
194. Hamza A, Cho H, Tai HH, Zhan CG (2005) Molecular dynamics simulation of cocaine
binding with human butyrylcholinesterase and its mutants. J Phys Chem B 109(10):
4776–4782
195. Gatley SJ (1991) Activities of the enantiomers of cocaine and some related compounds as
substrates and inhibitors of plasma butyrylcholinesterase. Biochem Pharmacol
41(8):1249–1254
196. Pancook JD, Pecht G, Ader M, Mosko M, Lockridge O, Watkins JD (2003) Application of
directed evolution technology to optimize the cocaine hydrolase activity of human butyryl-
cholinesterase. FASEB J 17(4):A565–A565
197. Pan Y, Gao D, Yang W, Cho H, Yang G, Tai HH, Zhan CG (2005) Computational redesign of
human butyrylcholinesterase for anticocaine medication. Proc Natl Acad Sci U S A
102(46):16656–16661
198. Cohen-Barak O, Wildeman J, van de Wetering J, Hettinga J, Schuilenga-Hut P, Gross A,
Clark S, Bassan M, Gilgun-Sherki Y, Mendzelevski B, Spiegelstein O (2015) Safety, pharma-
cokinetics, and pharmacodynamics of TV-1380, a novel mutated butyrylcholinesterase treat-
ment for cocaine addiction, after single and multiple intramuscular injections in healthy
subjects. J Clin Pharmacol 55(5):573–583
199. Brimijoin S, Gao Y, Anker JJ, Gliddon LA, Lafleur D, Shah R, Zhao Q, Singh M, Carroll ME
(2008) A cocaine hydrolase engineered from human butyrylcholinesterase selectively blocks
cocaine toxicity and reinstatement of drug seeking in rats. Neuropsychopharmacology
33(11):2715–2725
200. Chen X, Zheng X, Zhou Z, Zhan CG, Zheng F (2016) Effects of a cocaine hydrolase engi-
neered from human butyrylcholinesterase on metabolic profile of cocaine in rats. Chem Biol
Interact. doi:10.1016/j.cbi.2016.05.003
201. Kim K, Tsay OG, Atwood DA, Churchill DG (2011) Destruction and detection of chemical
warfare agents. Chem Rev 111(9):5345–5403
202. Alder L, Greulich K, Kempe G, Vieth B (2006) Residue analysis of 500 high priority pesti-
cides: better by GC-MS or LC-MS/MS? Mass Spectrom Rev 25(6):838–865
203. Bhattacharya S, Alsen C, Kruse H, Valentin P (1981) Detection of organo-phosphate insecti-
cide by an immobilized-enzyme system. Environ Sci Technol 15(11):1352–1355
204. Razumas VJ, Kulys JJ, Malinauskas AA (1981) High-sensitivity bioamperometric determi-
nation of organo-phosphate insecticides. Environ Sci Technol 15(3):360–361
205. Villatte F, Marcel V, Estrada-Mondaca S, Fournier D (1998) Engineering sensitive acetylcho-
linesterase for detection of organophosphate and carbamate insecticides. Biosens Bioelectron
13(2):157–164
206. Boublik Y, Saint-Aguet P, Lougarre A, Arnaud M, Villatte F, Estrada-Mondaca S, Fournier D
(2002) Acetylcholinesterase engineering for detection of insecticide residues. Protein Eng
15(1):43–50
207. Fremaux I, Mazeres S, Brisson-Lougarre A, Arnaud M, Ladurantie C, Fournier D (2002)
Improvement of Drosophila acetylcholinesterase stability by elimination of a free cysteine.
BMC Biochem 3:21
208. Schulze H, Muench SB, Villatte F, Schmid RD, Bachmann TT (2005) Insecticide detection
through protein engineering of Nippostrongylus brasiliensis acetylcholinesterase B. Anal
Chem 77(18):5823–5830
66 S. Lutz et al.
209. Hussein AS, Chacon MR, Smith AM, Tosado-Acevedo R, Selkirk ME (1999) Cloning,
expression, and properties of a nonneuronal secreted acetylcholinesterase from the parasitic
nematode Nippostrongylus brasiliensis. J Biol Chem 274(14):9312–9319
210. Bachmann TT, Schmid RD (1999) A disposable multielectrode biosensor for rapid simulta-
neous detection of the insecticides paraoxon and carbofuran at high resolution. Anal Chim
Acta 401(1–2):95–103
211. Bachmann TT, Leca B, Vilatte F, Marty JL, Fournier D, Schmid RD (2000) Improved multi-
analyte detection of organophosphates and carbamates with disposable multielectrode bio-
sensors using recombinant mutants of Drosophila acetylcholinesterase and artificial neural
networks. Biosens Bioelectron 15(3–4):193–201
212. Vanhooke JL, Benning MM, Raushel FM, Holden HM (1996) Three-dimensional structure of
the zinc-containing phosphotriesterase with the bound substrate analog diethyl
4-methylbenzylphosphonate. Biochemistry 35(19):6020–6025
213. Serdar CM, Gibson DT, Munnecke DM, Lancaster JH (1982) Plasmid involvement in para-
thion hydrolysis by Pseudomonas diminuta. Appl Environ Microbiol 44(1):246–249
214. Scanlan TS, Reid RC (1995) Evolution in action. Chem Biol 2(2):71–75
215. Iyer R, Iken B (2015) Protein engineering of representative hydrolytic enzymes for remedia-
tion of organophosphates. Biochem Eng J 94:134–144
216. Dumas DP, Durst HD, Landis WG, Raushel FM, Wild JR (1990) Inactivation of organophos-
phorus nerve agents by the phosphotriesterase from Pseudomonas diminuta. Arch Biochem
Biophys 277(1):155–159
217. Watkins LM, Mahoney HJ, McCulloch JK, Raushel FM (1997) Augmented hydrolysis of
diisopropyl fluorophosphate in engineered mutants of phosphotriesterase. J Biol Chem
272(41):25596–25601
218. Chen-Goodspeed M, Sogorb MA, Wu F, Raushel FM (2001) Enhancement, relaxation, and
reversal of the stereoselectivity for phosphotriesterase by rational evolution of active site resi-
dues. Biochemistry 40(5):1332–1339
219. Cho CM, Mulchandani A, Chen W (2002) Bacterial cell surface display of organophosphorus
hydrolase for selective screening of improved hydrolysis of organophosphate nerve agents.
Appl Environ Microbiol 68(4):2026–2030
220. Griffiths AD, Tawfik DS (2003) Directed evolution of an extremely fast phosphotriesterase
by in vitro compartmentalization. EMBO J 22(1):24–35
221. Hill CM, Li WS, Thoden JB, Holden HM, Raushel FM (2003) Enhanced degradation of
chemical warfare agents through molecular engineering of the phosphotriesterase active site.
J Am Chem Soc 125(30):8990–8991
222. Cho CM, Mulchandani A, Chen W (2004) Altering the substrate specificity of organophos-
phorus hydrolase for enhanced hydrolysis of chlorpyrifos. Appl Environ Microbiol
70(8):4681–4685
223. Tsai PC, Fan Y, Kim J, Yang L, Almo SC, Gao YQ, Raushel FM (2010) Structural determi-
nants for the stereoselective hydrolysis of chiral substrates by phosphotriesterase.
Biochemistry 49(37):7988–7997
224. Bigley AN, Xu C, Henderson TJ, Harvey SP, Raushel FM (2013) Enzymatic neutralization of
the chemical warfare agent VX: evolution of phosphotriesterase for phosphorothiolate hydro-
lysis. J Am Chem Soc 135(28):10426–10432
225. Cherny I, Greisen P Jr, Ashani Y, Khare SD, Oberdorfer G, Leader H, Baker D, Tawfik DS
(2013) Engineering V-type nerve agents detoxifying enzymes using computationally focused
libraries. ACS Chem Biol 8(11):2394–2403
226. Jackson CJ, Weir K, Herlt A, Khurana J, Sutherland TD, Horne I, Easton C, Russell RJ, Scott
C, Oakeshott JG (2009) Structure-based rational design of a phosphotriesterase. Appl Environ
Microbiol 75(15):5153–5156
227. Naqvi T, Warden AC, French N, Sugrue E, Carr PD, Jackson CJ, Scott C (2014) A 5000-fold
increase in the specificity of a bacterial phosphotriesterase for malathion through combinato-
rial active site mutagenesis. PLoS One 9(4):e94177
228. Aharoni A, Gaidukov L, Yagur S, Toker L, Silman I, Tawfik DS (2004) Directed evolution of
mammalian paraoxonases PON1 and PON3 for bacterial expression and catalytic specializa-
tion. Proc Natl Acad Sci U S A 101(2):482–487
229. Amitai G, Gaidukov L, Adani R, Yishay S, Yacov G, Kushnir M, Teitlboim S, Lindenbaum
M, Bel P, Khersonsky O, Tawfik DS, Meshulam H (2006) Enhanced stereoselective hydroly-
sis of toxic organophosphates by directly evolved variants of mammalian serum paraoxonase.
FEBS J 273(9):1906–1919
230. Gupta RD, Goldsmith M, Ashani Y, Simo Y, Mullokandov G, Bar H, Ben-David M, Leader
H, Margalit R, Silman I, Sussman JL, Tawfik DS (2011) Directed evolution of hydrolases for
prevention of G-type nerve agent intoxication. Nat Chem Biol 7(2):120–125
231. Tsai PC, Bigley A, Li Y, Ghanem E, Cadieux CL, Kasten SA, Reeves TE, Cerasoli DM,
Raushel FM (2010) Stereoselective hydrolysis of organophosphate nerve agents by the bacte-
rial phosphotriesterase. Biochemistry 49(37):7978–7987
232. Sepp A, Tawfik DS, Griffiths AD (2002) Microbead display by in vitro compartmentalisa-
tion: selection for binding using flow cytometry. FEBS Lett 532(3):455–458
233. Tsai PC, Fox N, Bigley AN, Harvey SP, Barondeau DP, Raushel FM (2012) Enzymes for the
homeland defense: optimizing phosphotriesterase for the hydrolysis of organophosphate
nerve agents. Biochemistry 51(32):6463–6475
234. Bigley AN, Mabanglo MF, Harvey SP, Raushel FM (2015) Variants of phosphotriesterase for
the enhanced detoxification of the chemical warfare sgent VR. Biochemistry 54(35):
5502–5512
235. Horne I, Sutherland TD, Harcourt RL, Russell RJ, Oakeshott JG (2002) Identification of an
opd (organophosphate degradation) gene in an Agrobacterium isolate. Appl Environ
Microbiol 68(7):3371–3376
Recent Advances in Directed Evolution
of Stereoselective Enzymes 3
Manfred T. Reetz
Abstract
Directed evolution of enzymes provides a prolific source of biocatalysts for
asymmetric reactions in organic chemistry and biotechnology. Nowadays, the
real challenge in this research area is the development of mutagenesis methods
and strategies which ensure the formation of small and highest-quality mutant
libraries requiring minimal screening effort. This chapter constitutes a critical
analysis of recent developments. Saturation mutagenesis at sites lining the bind-
ing pocket using highly reduced amino acid alphabets has emerged as the supe-
rior approach for evolving stereoselectivity.
3.1 Introduction
Asymmetric catalysis stands at the heart of modern synthetic organic chemistry.

The three options are chiral transition metal catalysts, organocatalysts, and enzymes.
The latter have been applied in synthetic organic chemistry for a century, but bioca-
talysis has not been accepted as a routine technique for several reasons. However,
during the last two decades, notable progress has been made in bioprocess develop-
ment, reactor design, downstream processing, immobilization, improved expression
systems, and genome mining for identifying new enzymes [1]. Nevertheless, seri-
ous problems persisted due to the following often observed limitations:
• Poor or wrong stereoselectivity

• Limited substrate scope
• Insufficient activity
M.T. Reetz
Department of Chemistry, Philipps-Universität, Marburg, Germany

DOI 10.1007/978-3-319-50413-1_3
70 M.T. Reetz
NO2 O NO2
O O NO2
R H2O R R
O OH
+ O + HO
CH3 lipase CH3 CH3
(S)-2 (R)-1 3
rac-1(R = n-C8H17)
Scheme 3.1 Model reaction used in the first study of directed evolution of a stereoselective
enzyme [6]
The so-called rational design based on appropriate site-specific mutagenesis has

been shown to be successful in some cases [2], but directed evolution has clearly
emerged as the more general and reliable approach [3]. As in genetic optimization
of thermostability [4], three basically different gene mutagenesis methods can be
applied in order to enhance or invert stereoselectivity: error-prone polymerase chain
reaction (epPCR), saturation mutagenesis, and/or DNA shuffling [3, 5]. The first
example of directed evolution of an enantioselective enzyme concerned the hydro-
lytic kinetic resolution of rac-1, catalyzed by the lipase from Pseudomonas aerugi-
nosa (PAL) (Scheme 3.1) [6]. WT PAL leads to low enantioselectivity slightly
favoring the formation of (S)-2, the selectivity factor amounting to only E = 1.4.
Four cycles of epPCR at low mutation rate with introduction of a single point
mutation in each round enhanced enantioselectivity to E = 11 (S). Since the fifth
cycle resulted in marginal improvement (E = 15), which is far from practical appli-
cation, different mutagenesis strategies were developed. The combination of epPCR,
saturation mutagenesis, and DNA shuffling afforded a variant characterized by six
point mutations, showing a selectivity factor of E = 51 and a 250% increase in activ-
ity [7]. Only one of the mutations occurred near the active site, five being remote. A
theoretical analysis based on QM/MM showed a relay mechanism to be operating.
More importantly, it was predicted that only two of the six mutations are necessary
for high enantioselectivity. Indeed, the respective double mutant proved to be even
more effective (E = 62) [8].
These observations demonstrated that the genetic approach utilizing epPCR,
saturation mutagenesis, and DNA shuffling is successful, but not efficient, a great
deal of time-consuming screening being necessary (50,000 transformants). It was
also possible to invert enantioselectivity in favor of (R)-2, but this also involved
excessive screening [9]. At the time, several other groups joined efforts in general-
izing directed evolution of stereoselectivity using the same strategies, as summa-
rized in a 2004 review [10]. However, efficacy was not a focal point of research.
Since the screening step is the bottleneck in the overall directed evolution process,
methods and strategies for generating smaller and smarter libraries had to be
developed.
After several years of research using PAL and other enzymes, saturation muta-
genesis at sites lining the binding pocket as part of the combinatorial active-site
saturation test (CAST) emerged as the optimal strategy (Scheme 3.2a) [11].
CAST is a convenient acronym to distinguish it from saturation mutagenesis at
other (remote) sites for different purposes. When the “hits” in initial CAST libraries
3 Recent Advances in Directed Evolution of Stereoselective Enzymes 71
a b
Scheme 3.2 (a) Systematization of CASTing; A, B, C, etc. denote potential randomization sites,

each comprising one, two, or more residues lining the binding pocket. (b) Two-, three-, and four-
site ISM schemes
Table 3.1 Difference in screening effort when applying NNK (encoding all 20 canonical amino
acids) versus NDT (encoding 12 amino acids: Phe, Leu, Ile, Val, Tyr, His, Asn, Asp, Cys, Arg, Ser,
Gly) for 95% library coverage [5]
NNK NDT
Nr. of amino acid Transformants Transformants
positions at one site Codons needed Codons needed
1 32 94 12 34
2 1028 3066 144 430
3 32,768 98,163 1728 5175
4 >1.0 × 106 >3.1 × 106 >2.0 × 104 >6.2 × 104
5 >3.3 × 107 >1.0 × 108 >2.5 × 105 >5.5 × 105
6 >1.0 × 109 >3.2 × 109 >2.9 × 106 >8.9 × 106
7 >3.4 × 1010 >1. × 1011 >3.5 × 107 >1.1 × 108
8 >1.0 × 1012 >3.3 × 1012 >4.2 × 108 >1.3 × 109
9 >3.5 × 1013 >1.0 × 1014 >5.1 × 109 >1.5 × 1010
10 >1.1 × 1015 >3.4 × 1015 >6.1 × 1010 >1.9 × 1011
still display insufficient enantioselectivity, a recursive process is recommended: iter-

ative saturation mutagenesis (ISM) [12]. Scheme 3.2b shows the case of two-, three-,
and four-site ISM systems involving two, six, and 24 pathways. It is not necessary to
explore all theoretically possible pathways, but some may be more productive than
others. Since saturation mutagenesis at large randomization sites requires excessive
screening for 95% library coverage, reduced amino acid alphabets were introduced
[13]. In order to remind the reader of the relationship between the size of a random-
ization site, the nature of the amino acid alphabet, and the screening effort for 95%
library coverage, Table 3.1 is included here which illustrates the difference between
NNK and, e.g., NDT codon degeneracy, which encode 20 and 12 canonical amino
acids, respectively [5]. Any other reduced amino acid alphabet that the researcher
may want to use can be analyzed statistically in the same manner using the CASTER
computer aid [5], which is based on the Patrick/Firth algorithm [14]. The Nov metric
72 M.T. Reetz
can also be used, in this case identifying the nth best mutant [15]. CAST/ISM should
be guided by X-ray structures (or homology models) and sequence data. Initial
NNK-based saturation mutagenesis at individual CAST positions, requiring the
screening of only one 96-format microtiter plate, also provides information for
choosing a reduced amino acid alphabet in subsequent mutagenesis experiments.
The lessons learned from these methodological developments, reported in a
series of studies using different enzyme types [3f, 5], were then used as a guide in
the final directed evolution study of PAL for comparison purposes [16]. It was dis-
covered that CAST/ISM provides a notably improved triple mutant showing E = 594
(S) in the model reaction involving the kinetic resolution of rac-1, while requiring
the screening of less than 10,000 transformants. This highlights the progress in
methodology development. In the ISM study, three CAST sites A, B, and C, were
designed, and NDT codon degeneracy was applied, leading to the highly improved
triple mutant [16]. This variant has no superfluous mutations. The origin of the
unusually high degree of enantioselectivity was traced on a molecular level to strong
cooperative mutational effects. Such synergistic effects (more than additivity) were
later found in other ISM-based studies as well [17]. In an independent study utiliz-
ing a galactosidase as the enzyme, saturation mutagenesis was likewise shown to be
more efficient than DNA shuffling [18].
Since the publication of the best results concerning PAL, further progress in
methodology development has been achieved [5, 19, 20]. The primary focus was
placed on utilizing the smallest possible reduced amino acid alphabets, again for the
purpose of minimizing screening while maximizing library quality. In doing so, two
different strategies can be applied (Scheme 3.3). According to strategy 1, one and
the same reduced amino acid alphabet is used for the whole CAST randomization
site in a single experiment. In contrast, strategy 2 calls for a different reduced amino
one reduced amino

acid alphabet for entire
randomization site
optionally
screen ISM improved
hit(s) hit(s)
Strategy 1
WT different reduced amino

acid alphabet at
each position
Strategy 2
optionally
screen ISM improved
hit(s) hit(s)
Scheme 3.3 Two strategies for applying saturation mutagenesis in order to manipulate

stereoselectivity
acid alphabet for each position of a multi-residue site, a single experiment also
being involved. In both cases, ISM can be applied for further optimization.
Nowadays, automated GC or HPLC can handle typically 2000–3000 transfor-
mants within a few days. An on-plate pretest for activity is nevertheless recom-
mended, the much smaller number of hits then being analyzed for stereoselectivity
by chiral GC or HPLC. The question whether to choose a reduced amino acid alpha-
bet such as NDT in combination with, e.g., two-residue randomization sites fol-
lowed by ISM, or to opt for a much smaller reduced amino acid alphabet encoding
only one, two, or three amino acids in combination with larger randomization sites,
e.g., four to ten residues, has been addressed [20–22]. Triple code saturation muta-
genesis (TCSM) using three- or four-residue CAST randomization sites appears to
be the strategy of choice as shown by several recent studies [22]. In these cases, the
initial CAST libraries often harbor stereoselective variants that fulfill all require-
ments for practical applications or require only one ISM step for final fine-tuning.
Nevertheless, more experience is needed for final assessments. When applying
CAST/ISM, several guidelines are recommended:
• Library design by the CASTER computer aid (http://www.kofo.mpg.de/en/

research/biocatalysis) or the GLUE-IT metric (http://guinevere.otago.ac.nz/cgi-
bin/aef/glue-IT.pl), both available free of charge
• Guidance by structural, mechanistic, and (consensus) sequence data [20–22]
• Use of an on-plate pretest for activity followed by chiral GC or HPLC analysis
for enantioselectivity [21]
• Application of the quick quality control [23a] or quantitative Q-values [23b] in
order to avoid screening something that does not exist
• Application of pooling techniques for reducing the screening effort [23a, 24]
• Techno-economical analysis which considers, inter alia, the number, quality, and
cost of primers used in designing and generating mutant libraries [25]
The CAST/ISM-based approach has proven to be particularly efficient, fast, and

reliable [5, 19, 20], but other gene mutagenesis methods continue to be used.
Unfortunately, comparative experiments are generally not made. In some ISM-
based studies, a final round of epPCR was added for further (small) improvements,
as in the case of directed evolution of a glycosidase [26]. In other investigations,
only epPCR and/or DNA shuffling were employed. In the sections that follow,
selected recent examples of directed evolution of stereoselectivity using different
gene mutagenesis techniques and strategies are critically analyzed.
3.2 Cytochrome P450 Monooxygenases
Cytochrome P450 monooxygenases (CYPs) catalyze several synthetically useful

reaction types, including asymmetric oxidative hydroxylation R-H → R-OH, olefin
epoxidation, and sulfoxidation. Several reviews of CYP protein engineering have
appeared [27]. The first study featuring the manipulation of enantioselectivity of a
74 M.T. Reetz
HO HO
N
P450pyr N N
+
4 (S)-5 (R)-5
Scheme 3.4 Model reaction used in the first case of directed evolution of a P450 monooxygenase
leading to high regio- and stereoselectivity [28]
cytochrome P450 monooxygenase while maintaining essentially complete regiose-

lectivity concerned the P450-pyr catalyzed oxidative hydroxylation of N-benzyl-
pyrrolidine (4) with formation of product 5 (Scheme 3.4) [28]. WT P450-pyr shows
complete regioselectivity in favor of the 3-hydroxy product, but with a mere 43% ee
(S). The substrate appears to be bound in two energetically similar poses within
reach of the catalytically active high spin heme-Fe=O (Cpd I) in the large binding
pocket. The mechanism is known to involve a radical abstraction of the H-atom with
formation of a carbon-centered radical followed by rapid C-O bond formation. The
ideal O-H-C angle has been calculated to be about 130ο [29].
Using a homology model, 17 residues were identified within 5 Å of the heme-
docked substrate [28a]. With the exception of residues C366 and G256, they were
subjected individually to saturation mutagenesis using NNK codon degeneracy.
This would require for 95% library coverage the screening of only 94 transformants
in each case. In fact, an excess of 180 transformants were screened to ensure even
higher coverage [28a]. Whereas variant F403L improves (S)-selectivity to 65% ee,
a single point mutation (N100S) induces the reversal of enantioselectivity. In order
to enhance (R)-selectivity, a simplified version of ISM was applied by employing
mutant N100S as the template in saturation mutagenesis at other CAST sites. The
best double mutant proved to be N100S/T186I with an enantioselectivity of 83% ee
(R) and no trade-off in regioselectivity.
In a subsequent study, (S)-selectivity was improved by exploring more of protein
sequence space [28b]. First, the crystal structure of WT P450-pyr was determined
and used as a guide in choosing CAST sites. Twenty residues, A77, I82, I83, L98,
P99, N100, I102, A103, S182, D183, T185, T186, L251, V254, G255, D258, T259,
L302, M305, and F403, were targeted using the same simplified ISM-based strat-
egy. Only nine initial libraries had to be generated, since the remaining 11 had
already been obtained in the previous study. The best variant I83H/M305Q/A77S
was obtained in three rounds of iterative saturation mutagenesis (ISM), showing an
ee-value of 98% (S) at fully maintained regioselectivity and little trade-off in activ-
ity [28b]. It would be interesting to test TCSM [22] in this reaction.
Particularly challenging goals arise when regio- and enantioselectivity need to be
evolved. In a study dedicated to solving this problem, P450-BM3 was employed [30]. It
was found that WT P450-BM3 catalyzes the hydroxylation of cyclohexene-1-carboxylic
OH OH
S-selective R-selective
P450-BM3 CO2Me P450-BM3
CO2Me 6 CO2Me
(S)-7 mutant mutant (R)-7
Scheme 3.5 Model reaction used in the first case of directed evolution of a P450 monooxygen-
ase leading to high regio- and stereoselectivity with optional formation of either enantiomeric
product [30]
Fig. 3.1 The 24 P450-BM3 residues considered for saturation mutagenesis. They were assigned
to three categories marked in green (residues closest to the heme), blue (residues further away but
still lining the binding pocket), and yellow (residues at entrance to the large binding pocket) [30]
acid methyl ester (6) with insufficient regioselectivity (84%) and poor enantioselectivity
(34%) ee in slight favor of (R)-7) (Scheme 3.5).
In order to achieve practical levels of regioselectivity as well as >95% (R)- and
(S)-selectivity on an optional basis, ISM was applied. On the basis of the crystal
structure of P450-BM3, a total of 24 CAST residues were identified (first- and
second-sphere residues) (Fig. 3.1) [30].
NNK-based saturation mutagenesis was initially applied at 23 of the 24 posi-
tions, leading to a limited set of improved mutants with enhanced and reversed
76 M.T. Reetz
OH OH OH
OH OH OH
P450 BM3
9 +
variants HO
10
11 12 OH 13
β-cembrenediol
Scheme 3.6 P450-BM3 catalyzed oxidative hydroxylation of the diterpenoid ß-cembrene-

diol [32]
enantioselectivity. This information formed the basis of ISM optimization. For

example, the gene of the best (R)-selective variant was chosen as the template for
saturation mutagenesis at other sites which had also favored (R)-selectivity, and the
analogous procedure was applied for reversing (S)-selectivity. This strategy pro-
vided selective variants (97–98% ee) favoring (R)- and (S)-7, respectively, and dis-
playing regioselectivities in the range of 93–98% [30].
ISM was also applied to P450-BM3 catalyzed oxidative hydroxylation of ste-
roids as part of “late-stage oxidation” [31]. For example, testosterone was used as
the substrate, for which 2ß- and 15ß-selective variants were evolved showing >95%
regio- and diastereoselectivity.
Yet another example of late-stage oxidative hydroxylation concerns P450-BM3
catalyzed reactions of the monocyclic diterpenoid ß-cembrenediol (11), which is
essentially not accepted by the WT (2% conversion) (Scheme 3.6) [32]. At the time
of this project, a number of other protein engineering studies utilizing various struc-
turally different substrates were available, the mutational data being of significant
help in the new endeavor [27]. This is an important point, because it is logical to
learn from past experience and not to start from “scratch” in each new project. For
example, it was known that residue F87 is a sensitive position, smaller amino acids
generally being necessary to enable substrate acceptance of sterically demanding
compounds because the phenyl group of F87 prevents complete access to the cata-
lytically active heme-Fe=O (Cpd I). Therefore, known variants F87A and F87G
[27] were tested first, which indeed led to acceptable levels of activity. These were
then used as templates for site-specific mutagenesis at selected first-sphere CAST
residues A74, L75, V78, I263, A264, and L437. Iterative site-specific mutagenesis
was performed in several rounds, in each case a new point mutation being intro-
duced [32]. A total of 29 variants were produced by this technique, which resembles
ISM without the need to screen libraries. Structure-guided successive site-specific
mutagenesis with the creation of minimally sized libraries constitutes a fusion of
rational design and directed evolution. Whereas the particular choice of the muta-
tions limits structural diversity drastically, the results proved to be of practical inter-
est in this project. Several variants were generated in this way, F87A/I263L
catalyzing complete regioselective hydroxylation at position C-9 with an 89:11 dia-
stereomeric ratio. The triple mutant L75A/V78A/F87G enables hydroxylation at
position C-10 (97% regioselectivity) with a 74:26 diastereomeric ratio. Assignment
of the absolute configuration at the new stereogenic centers was not reported [32].
H
OH
O
IV-H4
O
C7(S): 100% O
H
O
6a
H O H
OH
O
O II-H10
7 parent P450 O
O C7(R): 100% O
O C7(S): 83%
H
H C7(R): 10% O
O C6a : 7% OH
O
O H
X-E12 O
artemisinin
C6a(S): 94% O
O
H
O
Scheme 3.7 Oxidative hydroxylation of artemisinin by P450-BM3 mutants [33]
This study nicely shows that very small semi-rationally designed mutant libraries
may well suffice, provided sufficient previous knowledge of mutational effects is
available. Indeed, after years of protein engineering of P450-BM3, the huge muta-
tional data serves as a convenient guide when targeting new substrates using ratio-
nal design or directed evolution.
Another interesting example of directed evolution of a CYP concerns the late-
stage regio- and stereoselective hydroxylation of the antimalaria drug artemisinin
catalyzed by P450-BM3 (Scheme 3.7) [33]. Here again a semi-rational approach
was implemented using saturation mutagenesis at first-sphere CAST sites, this time
based on initial P450 fingerprinting followed by fingerprint-driven reactivity predic-
tions and final ISM experiments. WT P450-BM3 does not accept the substrate.
First, 125,000 transformants were screened for activity using not the actual sub-
strate artemisinin, but five semisynthetic chromogenic probes, which gave rise to
1950 active variants that accept such sterically demanding substrates (criterion:
“>10% of parent enzyme activity on at least one of the fingerprint probes”), and 522
functionally unique variants (criterion: “larger than 20% variation on at least one of
the five fingerprint components compared to the parent enzyme and any other mem-
ber of the library”) [33]. After correlating the P450-BM3 fingerprints with the actual
artemisinin reactivity, 75 variants remained and were tested as catalysts for oxida-
tion of the real substrate by HPLC analysis. The best hit was variant FL#62, which
was found to have 16 point mutations. This work was then followed by several ISM
experiments. The result of all of these efforts is summarized in Scheme 3.7 [33].
Whereas the “parent” enzyme showed 83% C7(S)-, 10% C7(R)-, and 7% C6a-
selectivities, notable improvements were achieved in the final saturation mutagen-
esis experiments. The tendency to hydroxylate at C6 and C7 was maximized by
directed evolution. If for some reason a different position were to be the target, then
the challenging question of how to achieve such regioselectivity would arise.
78 M.T. Reetz
In the directed evolution of CYPs, as in other enzyme types, a clear trend to uti-
lize saturation mutagenesis has emerged [27, 34]. However, some researchers in
recent CYP studies rely on epPCR and site-specific mutagenesis. An example is the
evolution of P450-BM3 mutants that hydroxylate the contraceptive drug norethis-
terone at the 15ß- or 16ß-position [35a]. The combination of epPCR and DNA shuf-
fling has also been used, as in activity optimization of the 13-hydroperoxide lyase
CYP74B; the products are used for the production of C6-aldehydes [35b].
To date, the concept of P450-catalyzed late-stage oxidative hydroxylation of natu-
ral products or synthetic compounds suffers from the lack of predictability as to where
hydroxylation will occur. When designing a synthetic pathway in natural product syn-
thesis, it is generally unclear whether P450-catalyzed hydroxylation will occur at the
desired position. In such a situation, the initial library must be diverse enough to har-
bor variants that provide many different regioisomers. If the desired product is formed
to a small extent, then the respective mutant can be chosen for further mutational
improvements with the aim of turning the minor into the major product.
3.3 Baeyer-Villiger Monooxygenases
The first directed evolution study of a Baeyer-Villiger monooxygenase (BVMO)

concerned the optimization of cyclohexanone monooxygenase (CHMO) as the cata-
lyst in the oxidative desymmetrization of 4-hydroxycyclohexanone [36]. One of the
best variants evolved by epPCR, F345S, was then used in the desymmetrization of
structurally different ketones (Table 3.2) [37]. It can be seen that in essentially all
Table 3.2 CHMO variant Phe432Ser as the catalyst in oxidative desymmetrization [37]
Substrate ee (%) Substrate ee (%)
O 94 O 96
O O
O O
O 99 O
O >99
Cl O Cl O
O
O 91 O O >99
O O
O
O 97 O
O >99
O O
O
OH HO
O 78 O
O
O 99
O
O
H3C OH HO CH3
Scheme 3.8 R R
Oxidative kinetic O
resolution catalyzed BVMO O
O
by mutants of the O2
Baeyer-Villiger
monooxygenase
PAMO [40]
rac-14a R = H (R)- or (S)-15a R = H
b R = Cl b R = Cl
Scheme 3.9 Sequence alignment of eight BVMOs (loop 441–444 in gray box) [40]
cases, high enantioselectivity as well as chemoselectivity was achieved without

having to perform additional mutagenesis experiments.
These results are impressive when utilizing whole cells, but CHMO suffers from
insufficient thermostability. The discovery of the unusually robust phenyl acetone
monooxygenase (PAMO) aroused a great deal of interest, although its substrate
scope was shown to be very narrow [38]. Such substrates as cyclohexanone or deriv-
atives thereof are not accepted by this BVMO, apparently because the binding
pocket is too small to accommodate such compounds. In order to solve this prob-
lem, a series of CAST-based directed evolution studies have appeared [39].
A bioinformatics-based study of PAMO as the catalyst in oxidative kinetic resolu-
tion of compounds 14a-b deserves special attention for three reasons (Scheme 3.8)
[40]: (1) It not only utilized structural data for choosing appropriate CAST sites, but
also (2) sequence alignment information in order to derive optimal reduced amino acid
alphabets, and (3) a different codon degeneracy according to strategy 2 as outlined in
Scheme 3.3 (Introduction). Ketone 14a was used in all screening assays, the best
evolved variant then being tested in the oxidative kinetic resolution of 14b as well [40].
Guided by the PAMO crystal structure [38b], four residues in loop 441–444 next
to the binding pocket were chosen as CAST sites. NNK-based randomization of a
four-residue CAST site would require the screening of 3.1 million transformants for
95% library coverage, while NDT codon degeneracy would still call for excessive
screening( ≈ 62,000 transformants) (Table 3.1, Introduction). Therefore, eight
Baeyer-Villiger monooxygenases were aligned, the loop region 441–444 being of
interest (Scheme 3.9) [40]. A limited number of amino acids are conserved at the
80 M.T. Reetz
Table 3.3 Codon degeneracies chosen at each position in the PAMO loop 441–444. Degenerate
codons: A (adenine), B (cytosine/guanine/thymine), C (cytosine), G (guanine), S (cytosine/gua-
nine), K (guanine/thymine), N (adenine/cytosine/guanine/thymine. WT amino acids are shown in
parentheses [40]
Amino acid Codon Encoded amino Oversampling for 95%
positions degeneracy acids Codons coverage
441 KCA A, (S) 864 2587
442 KBG S, (A), L, V, W, G
443 BGC F, H, (L), V, Y, G,
D, R, C
444 NSC (S), A, P, T, R, G, C
four positions: Ser and Ala (position 441); Ala, Val, Gly, and Leu (position 442);
Leu, Phe, Gly, and Tyr (position 443); and Ser, Ala, Cys, and Thr (position 444).
Consequently, these amino acids were used as building blocks at the respective
positions of the four-residue randomization site, these reduced amino acid alphabets
minimizing the screening effort drastically.
Codon degeneracies were then designed for matching the amino acids occurring at
these four positions while also introducing a limited number of additional amino acids
for randomization experiments in order to minimize primer costs and enhance diver-
sity (Table 3.3) [40]. WT amino acid is maintained throughout. Interestingly, at posi-
tion 441, KCA codon degeneracy correlates with the introduction of only one new
amino acid (Ala). At positions 442–444, structural diversity was designed to be higher.
After screening a mere 1700 transformants (95% library coverage would require
2587), several active variants were discovered, PAMO mutant Ser441Ala/
Ala442Trp/Leu443Tyr/Ser444Thr displaying the highest activity and enantioselec-
tivity (E = 70 in favor of (R)-15a) (Scheme 3.8) [40]. This variant proved to be an
even better catalyst for the oxidation of the p-chloro-derivative 14b (E >200). As in
the case of 14a, WT PAMO does not accept this substrate. It was concluded that
bioinformatics can be used to define a reduced amino acid alphabet and that a dif-
ferent designed reduced amino acid alphabet can well be effective at each position
within a multi-residue randomization site in a single saturation mutagenesis experi-
ment (strategy 2 in Scheme 3.3). This approach was later employed in the directed
evolution of other enzyme types [21b, 22].
It should be mentioned that following these and other CAST-based studies of
BVMOs [39], examples of epPCR as an alternative were reported. For example,
BmoFI from Pseudomonas fluorescens DSM 50106 was used as the catalyst in the
oxidative kinetic resolution of rac-16 with preferential formation of (S)-17 (Scheme
3.10) [41]. Enantioselectivity of the WT ranges between E = 55 (at small scale) and
E = 71 (growing E. coli cells in shake flasks). Application of epPCR and screening
3500 transformants provided several improved mutants displaying E = 77–92. Upon
combining the respective point mutations, an excellent variant was identified,
His51Leu/Ser136Leu displaying a selectivity factor of E = 86 [41].
In a very different approach not relying on CASTing, epPCR, or DNA shuffling,
a theoretically predicted remote two-residue site comprising positions 93 and 94 in
O
O OH O OH
+ O 5
5 5
OH
rac-16 (R)-16 (S)-17
Scheme 3.10 Oxidative kinetic resolution catalyzed by mutants of the Baeyer-Villiger monooxy-
genase BmoFI [41]
S S S S
PAMO + + O O
O O
Me Me Me Me
18 (R)-19 (S)-19 20
Scheme 3.11 Asymmetric sulfoxidation catalyzed by mutants of the Baeyer-Villiger monooxy-

genase PAMO [22a]
PAMO was subjected to saturation mutagenesis with the expectation of an allosteric

effect in the absence of an effector molecule [42]. Several four-substituted cyclo-
hexanone derivatives (methyl, ethyl, n-butyl, tert-butyl) were chosen as substrates
which are not accepted by WT PAMO. This procedure proved to be surprisingly
successful, the double mutant Gln93Asn/Pro94Asp showing 97–98% (R)-selectivity
for the methyl, ethyl, and n-butyl derivatives in desymmetrization reactions. The
tert-butyl derivative was not accepted due to steric factors. Instead of an effector
molecule inducing allostery, it is the mutational change that is causing a conforma-
tional reorganization at the binding site. Calculations of Karplus-type covariance
maps [43] of the double mutant versus WT PAMO showed that the expected confor-
mational motions were indeed occurring [42]. Deconvolution of Gln93Asn/
Pro94Asp showed that the single mutants Gln93Asn and Pro94Asp are inactive,
signaling pronounced cooperative mutational effects.
To date, the concept of focusing on mutationally induced remote allosteric effects
has not been applied to other substrates. In further optimization, the double mutant
Gln93Asn/Pro94Asp could be used as a template for CASTing. However, research-
ers have preferred to concentrate from the very beginning on CAST-based ISM
when new projects are initiated. For example, using this strategy, PAMO was
evolved as the catalyst in asymmetric sulfoxidation reactions using prochiral thio-
ether 18 as the substrate (Scheme 3.11) [22a]. WT PAMO leads to the preferential
formation of (S)-19 with notable selectivity (90% ee). The primary goal was to
evolve inverted enantioselectivity in favor of (R)-19.
Six potential randomization residues were first identified on the basis of the
PAMO crystal structure [38b] and past studies which identified “hot spots”: P440,
A442, and L443 (as CAST loop residues) and V54, I67, and Q152 (as traditional
CAST sites). One possible strategy would be to group them into three two-residue
randomization sites and to apply ISM. In this study, a different procedure was chosen
[22a]. In exploratory experiments, residues V54, I67, Q152, and A442 were chosen
for individual saturation mutagenesis. Instead of applying traditional NNK codon
degeneracy, the “smart-intelligent” library construction according to Tang [44] was
82 M.T. Reetz
ee %100 ZGZ-1 (92% ee) ZGZ-1 (95% ee) ZGZ-1 (94% ee)
B B A
75 A A
P440F/A442F/L443D P440F/A442F/L443I
50
R
25
25
50
S
B
I67A
75
100 WT (90% ee)
Scheme 3.12 Two ISM pathways in the directed evolution of PAMO as catalyst in the asymmet-
ric sulfoxidation of 18, WT → A → B providing two (R)-selective variants ZGZ-1 and ZGZ-2 and
WT → B → A leading to an equally (R)-selective variant ZGZ-3 [22a]
used. Four pairs of primers with degenerate codons of NDT (encoding 12 amino
acids N, S, I, H, R, L, Y, C, F, D, G, and V), codon degeneracy VMA (encoding six
amino acids E, A, Q, P, K, T), codon degeneracy ATG (encoding M), and codon
degeneracy TGG (encoding W) were considered at the target sites [22a]. They were
mixed in a ratio of 12:6:1:1, and for each library creation, only 60 transformants had
to be screened for 95% coverage. Whereas saturation mutagenesis at V54 failed to
provide any hits, the other libraries harbored improved hits.
Following these initial experiments, a two-site ISM scheme starting from WT
PAMO was designed involving site A (P440/A442/L443) with reduced amino acid
alphabets and site B (I67 as a hot spot identified earlier) with 20 amino acids. The
reduced amino acid alphabet applied at P440 involved three building blocks as the
components of a triple code in addition to the WT amino acid, Tyr, Leu, and Phe.
This is an example of triple code saturation mutagenesis (TCSM) [22a] as part of
strategy 2 (Scheme 3.3). The results of this minimal search in protein sequence
space are summarized in Scheme 3.12 [22a]. It can be seen that pathway WT → A
→ B provided two variants showing highly reversed enantioselectivity: ZGZ-1
(I67C/P440F/A442F/L443D (92% ee) and ZGZ-2 (I67Q/P440F/A442N/L443I
(95% ee) in favor of (R)-19. The opposite pathway WT → B → A was also success-
ful, leading to variant I67A/P440Y/A442V/L443I with 94% ee (R). Overoxidation
with formation of the sulfone 20 occurred to only a minimal degree.
Fig. 3.2 Fitness pathway landscape featuring 24 pathways which by necessity lead from the
(S)-selective WT PAMO to the best (R)-selective mutant ZGZ-2 allowing the formation of (R)-19
[22a]. The green line denotes a typical trajectory lacking any local minima, and the red one a trajec-
tory characterized by at least one local minimum. The upward climb is associated with 3.9 kcal/mol
Complete reversal of enantioselectivity is impressive, because the change in

energy ΔΔG# in going from WT PAMO (90% ee, S) to variant ZGZ-2 (95% ee, R)
amounts to 3.9 Kcal/mol. Deconvolution of this highly (R)-selective quadruple
mutant ZGZ-2 (I67Q/P440F/A442N/L443I) led to the surprising discovery that the
respective single mutants are all (S)-selective: I67Q (69% ee), P440F (97% ee),
A442N (69% ee), and L443I (98% ee) [22a]. If these four (S)-selective single
mutants had been generated separately by other means, few researchers would com-
bine them in order to generate the opposite (R)-selectivity! This kind of synergistic
nonadditive effects continues to be observed whenever time and effort are invested
in deconvolution. It is an indication of the efficacy of ISM [17].
Exhaustive deconvolution was performed, meaning the generation of variants
formed by combining the four point mutations in the form of all theoretically pos-
sible double and triple mutants. It allowed a fitness pathway landscape to be con-
structed, comprising 4! = 24 experimental pathways which lead from (S)-selective
WT PAMO to (R)-selective variant ZGZ-2 (Fig. 3.2) [22a]. Six of the 24 pathways
have no local minima along the respective trajectories, meaning the absence of
libraries which contain no variants with improved enantioselectivity. Eighteen path-
ways are characterized by at least one local minimum along the evolutionary trajec-
tory. Analysis of the intermediate stages of all 24 pathways revealed strong
cooperative mutational effects. Thus, much can be learned from deconvolution stud-
ies of this kind. In combination with MD/docking computations, they throw light on
the origin of stereoselectivity while also illuminating the efficacy of ISM. It should
be noted that this type of “constrained” fitness landscape [22a, 45] is different from
constructing all theoretically possible pathways of an ISM scheme, as was imple-
mented experimentally in the case of a four-site ISM system with 24 pathways [46].
The latter has been termed “unconstrained” fitness pathway landscape [17, 46], in
which every trajectory leads to a different result (see Sect. 3.5).
84 M.T. Reetz
3.4 Lipases and Esterases
Directed evolution has been extensively applied to lipases and esterases in order to
manipulate stereoselectivity, substrate scope, and activity [3]. As delineated in the
Introduction, the lipase from Pseudomonas aeruginosa (PAL) is the most systemati-
cally studied stereoselective enzyme in the field of directed evolution. In the final
PAL study [16], which included deconvolution experiments, ISM was applied to the
original model reaction involving the hydrolytic kinetic resolution of rac-1 with
preferential formation of (S)-2 (Scheme 3.1). Based on the crystal structure of PAL,
a three-site ISM scheme comprising two-residue CAST sites A, B, and C was con-
sidered. Following exploratory mutagenesis experiments, the best pathway proved
to be WT → B → A which provided the triple mutant 1B2 (Leu162Asn/Met16Ala/
Leu17Phe) showing a selectivity factor of E = 94 (S) and notably enhanced activity:
WT PAL (kcat = 37×10−3 s−1; kcat/Km = 43.5 s−1 M−1) versus variant 1B2
(kcat = 1374×10−3 s−1; kcat/Km = 4041 s−1 M−1) [16].
This result was achieved in the following way: In the first ISM step, site B
(Leu159/Leu162) was randomized using NNK codon degeneracy, leading to single
mutant Leu162Asn with E = 8 (S). It was employed as the template for DNT-based
saturation mutagenesis at site A (Met16/Leu17), which like NDT involves 12 amino
acids as building blocks. The ISM pathway is illustrated in Scheme 3.13 [16].
Scheme 3.13 seems to suggest that the second mutational introduction, Met16Ala/
Leu17Phe, contributes most to the overall result. However, this assumes classical
mutational additivity [17]. Therefore, deconvolution was performed by generating
600
Leu 162Asn/Met16Ala/Leu17phe
500
100
90
80
70
E-value
A(Met16/Leu17:DNT)
60
50
40
30
Scheme 3.13 ISM
pathway WT → B → A in 20
the directed evolution of
PAL as the catalyst in the 10
hydrolytic kinetic Leu162asn
B(Leu159/Leu162:NNK)
resolution of rac-1 [16] 0 WT
NO2 NO2
O O
O O
CH3 rac-21 CH3 rac-22
NO2 NO2
O O
O O
CH3 rac-23 CH3 rac-24
Scheme 3.14 Model compounds used in the directed evolution study of CALA-catalyzed hydro-
lytic kinetic resolution [21b]
the double mutant Met16Ala/Leu17Phe and testing it in the model reaction.

Surprisingly, it proved to be a poor catalyst with E = 2.6 (S). Thus, a strong coopera-
tive mutational effect is operating. An MD/docking analysis uncovered the molecu-
lar basis of this epistatic synergism. It involves the creation of an extended H-bond
network as a consequence of the interaction of Leu162Asn in concert with
Met16Ala/Leu17Phe [16].
This study required the screening of less than 10,000 transformants at a time
when subsequent optimization of saturation mutagenesis strategies and techniques
such as triple code saturation mutagenesis (TCSM) [22b, c] were not yet available.
It is likely that TCSM as applied to PAL would require even less screening.
PAL served as a useful model system to test various mutagenesis strategies in a
comparative manner. However, for several reasons, it is not likely to become a lipase
of practical utility in organic chemistry. The situation is very different in the case of
Candida antarctica lipase B (CALB), one of the most popular enzymes in bioca-
talysis [1]. It has been employed in ISM-based directed evolution in order to expand
substrate scope and to invert enantioselectivity [47].
The homolog CALA has also been subjected to ISM-based directed evolution in
order to accept chiral phenyl-substituted carboxylic acid esters [48], but the bulkier
analogs of the ibuprofen type were not accepted. Therefore, a different strategy was
tested following strategy 2 (Scheme 3.3). The plan was to use a single amino acid as
building block (in addition to WT) at most of the positions of a nine-residue site in
a single mutagenesis experiment [21b]. The hydrolytic kinetic resolution of rac-21
was chosen for screening, and racemic esters 22–24 were also tested following the
mutagenesis experiments (Scheme 3.14). The goal was enhancing activity and
enantioselectivity while minimizing screening.
All experiments began with the use of the triple mutant F149Y/I150N/F233G as
template, obtained in the earlier ISM-based study [48]. Substrate 21 was docked into
the CALA binding pocket in the oxyanion form (tetrahedral intermediate at Ser184),
leading to the identification of nine residues at the acyl binding region: positions 149,
150, 215, 221, 225, 233, 234, 237, and 431 (Fig. 3.3) [21b]. Other residues near this
large CAST site were not considered because they are highly conserved as shown by
86 M.T. Reetz
Fig. 3.3 View of CALA binding pocket harboring covalently bound substrate 21 as an oxyanion
[21b]. Nine CAST residues with the respective reduced amino acid alphabet(s) used in saturation
mutagenesis are shown, the original WT amino acids being underlined
an alignment analysis. In view of the PAMO study [40], this would not necessarily
be mandatory. Structure-based decisions were made regarding the different reduced
amino acid alphabets at the nine randomization positions.
Substrate 21 is sterically so demanding that it is not accepted by the triple mutant
with acceptable rate. Thus, small amino acids were mostly chosen for saturation
mutagenesis. In the earlier CALA study [48], mutations Phe149Tyr and Ile150Asn
had been shown to be important for high enantioselectivity toward similar sub-
strates. Therefore, at these positions, Tyr and Asn were chosen as the respective
building blocks (Table 3.4) [21b]. At position 233, three amino acids were employed
(in addition to WT). A certain degree of intuition was involved in some of the
decisions.
A single highly condensed library was created using appropriately designed
primers, followed by screening about 2400 transformants in the model reaction of
rac-21 (≈90% library coverage). Only a few variants proved to be active toward
substrate rac-21 [21b]. The best hit was a penta-substituted variant Thr221Ser/
Leu225Val/Phe233Cys/Gly237Ala/Phe431Val in which four different amino acids
were introduced at five different positions. It shows high (S)-stereoselectivity
(E = 100). This and other CALA variants were tested in the hydrolytic kinetic reso-
lution of the substrates 21–24, which likewise resulted in acceptable levels of enan-
tioselectivity [21b]. The alternative of applying conventional NNK codon degeneracy
Table 3.4 Reduced amino acid alphabets Position WT residue Alternative residue(s)
used in simultaneous saturation
149 Phe Tyr
mutagenesis of a nine-residue
randomization site of CALA [21b] 150 Ile Asn
215 Pro Ala
221 Thr Ser
225 Leu Val
233 Phe Cys/Gly/Val
234 Ala Gly
237 Gly Ala
431 Phe Val
O O O
HO HO HO
Scheme O OH O
3.15 Model
hydrolytic kinetic
+
resolution catalyzed
by mutants of the
esterase RspE [50a] rac-25 (R)-26 (S)-25
encoding all 20 canonical amino acids would have required for 95% library cover-
age the screening of 1014 potentially enantioselective transformants. A limited num-
ber of deconvolution experiments revealed cooperative mutational effects. This
suggested that the particular variant would not be accessible by ISM. The same
strategy was later successfully applied to CALA as the catalyst in acylating kinetic
resolution of chiral alcohols [49].
It can be concluded that both approaches to the use of reduced amino acid alpha-
bets are successful, strategy 1 as well as strategy 2 as delineated in Scheme 3.3. It
would be interesting to test strategy 1 using triple code saturation mutagenesis.
Whether such a procedure would also allow reversal of enantioselectivity is cur-
rently a matter of speculation.
Esterases have also been targeted by directed evolution for improving stereose-
lectivity [3], recent studies utilizing various approaches including epPCR alone
[50a], epPCR in combination with saturation mutagenesis [50b], saturation muta-
genesis alone [50c, d], or site-specific mutagenesis in combination with saturation
mutagenesis [50e]. One study is featured here which relied solely on epPCR. The
carboxyl esterase from Rhodobacter sphaeroides (RspE) was subjected to three
recursive rounds of epPCR at low mutation rate, the model reaction being the hydro-
lytic kinetic resolution of methyl mandelate (rac-25) with preferential formation of
the carboxylic acid (R)-26 (Scheme 3.15) [50a]. WT RspE shows a selectivity factor
of E = 3.1(R).
In each epPCR cycle, 4000–6000 transformants were screened for activity using
an on-plate pH-dependent color test followed by conventional ee-determination. In
this way, it was possible to boost (R)-selectivity to E = 30.3 (Scheme 3.16) [50a].
This study is a new example of iterative epPCR in the successful attempt to evolve
88 M.T. Reetz
35
E=30.8
30 CVH
Enantioselectivity (E)
(Asn62Cys/Met121val/Leu145His)
25
20
E=14
15 CH
E=8.7 (Asn62Cys/Leu145His)
10
YH
5 E=3.1 (Asn62Tyr/Leu145His)
WT
0
Evolution step
Scheme 3.16 Hydrolytic kinetic resolution of rac-25 with preferential formation of (R)-26
(Scheme 3.15), catalyzed by RspE mutants which were evolved by recursive epPCR [50a]
H2O O HO OH
O
+
PhO ANEH PhO PhO
rac-27 (R)-27 (S)-28
Scheme 3.17 ANEH-catalyzed hydrolytic kinetic resolution used in the construction of a com-
plete four-site ISM system featuring 24 evolutionary pathways [46]
enhanced stereoselectivity of an esterase. Reversal of enantioselectivity was not

reported, but the best mutant in the model reaction showing E = 30.3 was used in the
kinetic resolution of structurally related substrates including acetates of chiral alco-
hols with selectivity factors of up to E = 92. It would be interesting to see how well
CASTing/ISM would perform in this enzyme system.
3.5 Epoxide Hydrolases
Several recent protein engineering studies of epoxide hydrolases have contributed

to methodology development in laboratory evolution [3f, 5]. The goal of one of
them was the exploration of all 24 pathways of a four-site ISM scheme [46]. To
date, it is the only case of complete exploration of such an ISM system. The hydro-
lytic kinetic resolution of rac-27 was chosen as the model reaction, catalyzed by
mutants of the epoxide hydrolase from Aspergillus niger (ANEH) (Scheme 3.17).
WT is a poor catalyst in this reaction (E = 4.6 in slight favor of (S)-28).
Four CAST randomization sites were considered on the basis of the WT ANEH
crystal structure, each comprising two residues. Subsequently all saturation mutagen-
esis libraries were constructed using NDT codon degeneracy encoding 12 amino acids
(Phe, Leu, Ile, Val, Tyr, His, Asn, Asp, Cys, Arg, Ser, and Gly), which is a balanced
“cocktail” of polar/nonpolar, charged/non-charged, hydrophobic/non-hydrophobic,
and aromatic/nonaromatic amino acids as building blocks. All 24 pathways provided

in the final saturation mutagenesis rounds notably improved variants characterized by
different sequences [46]. Some pathways proved to be more productive than others.
The 12 best pathways resulting in selectivity factors in the range E = 78–159 are shown
in Scheme 3.18a; the 12 pathways leading to the least improved variants (E = 28–78)
are pictured in Scheme 3.18b. These results suggest that if the researcher is faced with
the question of choosing an appropriate ISM pathway, an arbitrary choice has a high
probability of providing improved variants. This explains why in essentially all ISM
studies reported thus far arbitrarily chosen pathways led to notably improved variants
[5]. Nevertheless, superior variants may have been missed. It can therefore be con-
cluded that ISM systems should be designed in a way that involves less pathways and
therefore less mutant libraries, which correlates with a lower number of decisions for
the researcher to make. Indeed, methodology development since the publication of
this study has focused, inter alia, on step-economy [21, 22].
Noteworthy is another feature of the experimental results collected in Scheme
3.18. In several ISM pathways, libraries occurred which failed to harbor any
improved variants. This phenomenon signals a local minimum in the fitness land-
scape. Such “dead ends” may occur in any directed evolution project irrespective of
the mutagenesis method [3]. In the present case, ISM was not abandoned, but an
inferior mutant was used as the template in the subsequent saturation mutagenesis
experiment at the next randomization site. This trick led to notably improved
mutants. This unique way of escaping from a local minimum is reminiscent of neu-
tral drift [4d, 51] or the Eigen concept of quasi-species [52a]. The latter has been
invoked occasionally in directed evolution studies [52b].
Maximal step-economy would involve the generation of a single saturation
mutagenesis library, which should be small and ideally harbor both (R)- and (S)-
selective mutants. If the degree of stereoselectivity should not be fully satisfactory,
one ISM round could then be undertaken for fine-tuning. Along these lines, the
directed evolution of a different epoxide hydrolase was reported recently, namely,
limonene epoxide hydrolase (LEH) as the catalyst in the model hydrolytic desym-
metrization of cyclohexene oxide (29) with formation of (R,R)- and (S,S)-30
(Scheme 3.19) [21a]. WT LEH shows poor enantioselectivity with minimal prefer-
ence for (S,S)-30, the enantiomeric ratio (er) amounting to a mere 48:52 (4% ee).
Based on the crystal structure of WT LEH, ten CAST residues were identified for
saturation mutagenesis (Leu74, Phe75, Met78, Ile80, Leu103, Leu114, Ile116,
Phe134, Phe139, and Leu147) [21a]. Tyr53 activates the substrate by forming an
H-bond to the epoxide O-atom, Asp101 being mainly responsible for positioning
water which initiates the rate-determining SN2 reaction with ring opening.
Saturation mutagenesis using NNK codon degeneracy (20-amino-acid alphabet)
or NDT codon degeneracy (12-amino-acid alphabet) would require the screening of
about 1015 or 1011 transformants for 95% coverage, respectively. The use the small-
est amino acid alphabet, a single amino acid, would call for only ≈3000 transfor-
mants. This would reduce structural diversity dramatically, suggesting that such a
strategy would fail. Nevertheless, it could be successful if the right decision is made
concerning the choice of the amino acid. Such an approach constitutes single codon
90 M.T. Reetz
a 160
GUY-228
150
140
130 GUY-259 D
GUY-216
GUY-197
120
110 GUY 168 GUY-230

GUY-224
B* GUY-212
100 F* GUY-200
GUY-194 GUY-215
GUY-199 GUY-204
90
E value
GUY 145 F*
80 GUY 150 D
F* GUY 157
D E E
70 E
E
F* E
60 GUY-121 GUY 144 GUY 146 GUY 148 GUY 152
F*
E E GUY 159
50 F*
F* B* GUY 176
40 F* GUY 128
GUY 123 B* GUY 156 B* GUY 167
GUY 130 GUY 132
30 GUY 107 GUY 124 D B*
B* D GUY 134
E GUY 113 D B*
20 B*
GUY 111 GUY 127 GUY 116
F*
B* E D
10 D WT
F*
0
b 80
GUY-186 GUY-191
GUY-237
F*
70 GUY-139
F* GUY-181
GUY-141
GUY-207
D
60 D GUY 192
GUY-193 GUY-240
GUY 137 F*
GUY-188 GUY-121 D
E
50 E D B*
GUY 120 GUY-143
GUY 138 B*
GUY-163
E value
B*
40 E F* B*
F* D GUY-238
E
GUY-219 B* GUY-223
30 GUY-220
GUY 119 GUY-130 B* GUY 184
B* F*
GUY 153 GUY-165
D D GUY-179
GUY-161 D F*
GUY 107 E GUY 134 GUY 135
20 GUY-131 D
E GUY 113
GUY 127 D
F* E
B*
10 E GUY 116
GUY 111
D WT F*
Scheme 3.18 Experimental exploration of a complete 24-pathway ISM system involving the
ANEH-catalyzed hydrolytic kinetic resolution of rac-27 (Scheme 3.17) [46]. (a) Portion of the
24-pathway ISM scheme featuring the 12 most productive pathways leading to ANEH variants
displaying E = 78–159 (S); (b) portion of the 24-pathway ISM scheme showing the 12 least pro-
ductive pathways providing ANEH variants with E = 28–78 (S)
LEH OH OH
O +
H2O
OH OH
29 (R,R-30) (S,S-30)
Scheme 3.19 Hydrolytic desymmetrization catalyzed by LEH mutants evolved by applying

single codon saturation mutagenesis (SCSM) [21a] and more recently using triple codon saturation
mutagenesis (TCSM) [22b]
99:1 SZ19 SZ31 SZ60 SZ80

er=92:8 er=93:7 er=92:8 er=91:9
90:10
F75V/L103V/L114
80:20 V/I116V/F139V
L74V/L103V/L114
L103V/L114V/I116V/
70:30 V/I116V/F139V L114V/I116V/
F139V/L147V
(S,S)
F139V
60:40
WT
50:50 er=52:48
40:60
(R,R)
M78V/I80V/ I80V/L114V/L147V
30:70 L147V I80V/L114V
20:80
10:90 SZ57 SZ91

SZ42
er=12:88 er=15:85
er=12:88
1:99
Scheme 3.20 Results of generating one single codon saturation mutagenesis (SCSM) library on
the basis of valine as the sole building block, hydrolytic desymmetrization of cyclohexene oxide
(29) serving as the model reaction [21a]
saturation mutagenesis (SCSM) as part of strategy 1 in Scheme 3.3 and was first
tested in the directed evolution of LEH [21a]. In doing so, the choice of the single
building block was crucial. The crystal structure of LEH reveals that most of the
amino acids surrounding the binding pocket are hydrophobic. Therefore, valine,
having a hydrophobic and sterically demanding side chain, was chosen as the small-
est reduced amino acid alphabet for SCSM in a single saturation mutagenesis exper-
iment at the ten-residue site. The results following the screening of about 3200
transformants are shown in Scheme 3.20.
It can be seen that both (R,R)- and (S,S)-selective occur in one and the same small
but high-quality library. The number of introduced valines ranges between two and
five, depending upon the particular variant. In a single ISM step, fine-tuning was
performed, boosting the enantiomeric ratio to 98:2 (96% ee). Crystal structures of
the variants with and without product in combination with MD/docking computa-
tions uncovered the origin of stereoselectivity [21a]. In a control experiment, the use
92 M.T. Reetz
SZ348 SZ347 SZ351 20

100
ee=99 ee=99 ee=97
80 WT
0
I80Y/I116F ee=4
(S,S)
60
I80Y/I116V
(S,S)
40 I80Y/L114V - 20
I116V
20
-40 I80V/L114F
(R,R)
WT
0
ee=4
-20 -60
(R,R)
-40 I80V/L114F
-60 -80
M78V M32V M78V/L147F
M32V
-80 SZ343 SZ343 SZ376
-100 SZ529 SZ367 SZ389
-100 ee=89 ee=89 ee=92
ee=97 ee=94 ee=97
Scheme 3.21 The results of triple code saturation mutagenesis (TCSM) when applied to LEH as
the catalyst in the hydrolytic desymmetrization of cyclohexene oxide (29) [22b]
of serine failed completely, which supports the original hypothesis regarding the
successful choice of valine.
Single code saturation mutagenesis cannot be expected to be general. Therefore,
triple code saturation mutagenesis (TCSM) was developed [22b, c]. A reduced
amino acid alphabet of three members at a ten-residue randomization site would
require for 95% library coverage excessive screening. Therefore, if so many CAST
residues are chosen, they need to be grouped into smaller randomization sites. As
part of strategy 1 (Scheme 3.3), this approach was first tested with LEH as the cata-
lyst in the same model reaction involving the desymmetrization of epoxide 29
(Scheme 3.19). The ten previously identified CAST residues were grouped into
three randomization sites: (A) (V83/L114/I116), (B) (L74/M78/L147), and (C)
(M32/L35/L103) [22b].
By considering the crystal structure of LEH, it was logical to test valine, phenyl-
alanine, and tyrosine as the reduced amino acid alphabet in TCSM, even if the previ-
ous saturation mutagenesis experiments using these amino acids as building blocks
had not been performed. The V-F-Y triple code was applied to the three randomiza-
tion sites A, B, and C in separate experiments, requiring for 95% library coverage
the screening of only 576, 192, and 192 transformants, respectively. The best results
were obtained in library A (Scheme 3.21). As before, one and the same library har-
bors both (R,R)- and (S,S)-selective variants, but this time much better results were
obtained. (S,S)-selectivity amounts to 96–99% ee in three different variants, while
the best (R,R)-mutant results in 89% ee which was boosted to 97% ee by ISM. X-ray
structures of selected mutants, MD/docking computations, and kinetic data were
reported, which shed light on the origin of mutational effects [22b].
3.6 Transaminase
Transaminases are enzymes that catalyze the reductive amination of prochiral

ketones with formation of the respective chiral primary amines [1]. An impressive
industrial example of directed evolution was reported by Codexis, specifically in the
F F
F i-PrNH2 Acetone F
O O O NH2
N N N N
N N (R)
N F N F
Transaminase/PLP
F3C 31 F 3C (R)-32 99.95% ee
Scheme 3.22 Asymmetric reductive amination with formation of the antidiabetic therapeutic
drug sitagliptin (R)-32, catalyzed by mutants of transaminase ATA-117 [53]
asymmetric reductive amination of ketone 31 with formation of the antidiabetic

drug sitagliptin (32) (Scheme 3.22) [53].
The transaminase ATA-117 was chosen as the enzyme, which is related to the
structurally well-characterized homolog from Arthrobacter sp. Both were known to
be (R)-selective in the reductive amination of methyl ketones and small cyclic
ketones. ATA-117 proved to be (R)-selective in the desired reaction as well, but
activity was extremely low [53]. Thus, the goal was to enhance activity while main-
taining stereoselectivity. At the beginning of the project, the industrial researchers
did not use the “real” substrate 31, but first resorted to in vitro coevolution which
means testing simpler but still structurally related compounds (substrate walking)
[54]. With the help of a homology model of ATA-117, docking computations were
carried out which allowed reasonable choices for randomizing sites lining the bind-
ing pocket (CAST sites). NNK-based saturation mutagenesis led to variant S223P
with an 11-fold increase in activity in the reaction of a simplified model ketone. The
mutant was then employed as a template for ISM experiments using the “real” sub-
strate 31 [53]. Docking computations indicated that the trifluoromethyl group inter-
acts with residues V69, F122, T283, and A284. Therefore, four NNK-based
saturation mutagenesis libraries were generated separately at these four positions.
Moreover, a combinatorial library using several residues simultaneously was cre-
ated. The combinatorial library harbored an active variant characterized by four
point mutations lining “small” and “large” parts of the binding pocket. Double
mutants F122I/V69G, F122I/A284G, F122V/V69G, F122V/A284G, F122L/V69G,
and F122L/A284G were the best hits. They all contain the parent mutation
S223P. Activity was still quite low, yet without point mutation S223P, no activity
whatsoever resulted, as shown by a deconvolution experiment.
The gene of the most active variant was then used as the parent for the next round
of ISM, followed by combining the beneficial mutations from the small-pocket and
large-pocket saturation mutagenesis libraries. This provided a variant with 12 point
mutations and a 75-fold increase in activity. Subsequently, 11 additional rounds of
mutagenesis/screening were performed using DNA shuffling, epPCR, rational
design, and saturation mutagenesis at second-sphere sites. Process development
was also performed in parallel. A total of 36,480 transformants were screened using
an LC/MS/MS screen. The best variant was shown to have 27 point mutations. In
50% DMSO, this catalyst converts 200 g/L of the prositagliptin ketone 31 to sita-
gliptin (32) with >99.95% ee (R) [53].
94 M.T. Reetz
The catalytic performance of the best ATA-117 variant under operating condi-
tions underscores the success of the project. However, it is difficult to assess the
efficacy of the applied mutagenesis approach. It is not clear whether the order of the
mutagenesis cycles in the overall multistep process was actually planned or whether
corrections in the strategy had to be undertaken. Why were the particular mutagen-
esis events chosen in the reported order?
3.7 Alternative Mutagenesis Approaches
As noted in the Introduction, such mutagenesis techniques such as epPCR or DNA

shuffling can also be applied in order to manipulate the stereoselectivity of enzymes [3,
5, 10], but in several comparative studies, these have been shown to be less efficient
[16, 18]. Nevertheless, following several cycles of saturation mutagenesis, it may be
useful to add one round of epPCR for activity enhancement [26]. Other approaches
such as neutral drift [4d, 51], domain swapping [55], and circular permutation [56]
likewise deserve mention, but these strategies have not been applied very often to the
evolution of stereoselectivity. Neutral drift can be used for identifying superior starting
points for protein engineering by exploring accessible sequence space on the basis of
recursive cycles of (random) mutagenesis and screening or selection. The technique
identifies accumulating mutations which are neutral for the native function but which
may prove to be useful for novel catalytic profiles. An example is the evolution of pro-
miscuity by turning a ß-glucuronidase into a ß-galactosidase [51c].
Domain swapping was originally used in the study of natural evolution and for
addressing mechanistic questions in protein science, but it has also been applied
occasionally in directed evolution [3, 55]. For example, a glycosyltransferase was
engineered for different substrate specificity [55b]. However, it is not well suited for
enhancing or inverting stereoselectivity. A special form of this technique is circular
permutation in which the N- and C-termini of an enzyme are relocated [56]. A semi-
nal example concerns the engineering of enhanced activity of the lipase from
Candida antarctica (CALB) [56a, b]. New locations of the N- and C-termini in WT
CALB were designed to occur at positions 282 and 283 in hope of influencing local
backbone flexibility and perhaps active site accessibility. Indeed, this led to higher
activity [56b], but enantioselectivity was not addressed at the time nor in a subse-
quent review [56a].
Conclusions and Perspectives
This chapter provides a summary of the most important recent methodological
developments in the directed evolution of stereoselective enzymes as catalysts in
synthetic organic chemistry and biotechnology. Rather than being comprehen-
sive, five different types of enzymes were chosen which illustrate important
recent advances in strategies and methods.
It can safely be concluded that structure-guided saturation mutagenesis at
sites in vicinity of the active site (CASTing) is a distinctly logical approach to
reshape the bonding pockets of enzymes in the quest to manipulate stereoselec-

tivity, substrate scope, and activity. The use of reduced amino acid alphabets at
relatively large CAST randomization sites has emerged as the preferred strategy,
allowing for the creation of small yet high-quality mutant libraries requiring less
screening than in the past. In this respect, triple code saturation mutagenesis
(TCSM) appears to be an optimal compromise between limited structural diver-
sity and screening effort (bottleneck of directed evolution). TCSM allows for
step-economy, since initial mutant libraries already contain notably improved
(R)- as well as (S)-variants. If needed, further tuning is possible by iterative satu-
ration mutagenesis (ISM). It has been demonstrated that it is better to strive for
higher library coverage at reduced structural diversity regarding the chosen
amino acid alphabet rather to maintain maximum structural diversity using NNK
codon degeneracy at the same screening effort correlating with considerably
lower library coverage [13a].
One of the remaining fundamental challenges in directed evolution is the
development of a general guide for optimizing more than one or two enzyme
parameters, e.g., stereo-/regioselectivity, substrate scope, rate, and thermostabil-
ity. Efforts are underway in several laboratories.
References
1. (a) Drauz K, Gröger H, May O (eds) (2012) Enzyme catalysis in organic synthesis, 3rd edn.
Wiley-VCH, Weinheim; (b) Faber K (2011) Biotransformations in organic chemistry, 6th edn.
Springer, Heidelberg; (c) Liese A, Seelbach K, Wandrey C (eds) (2006) Industrial biotransfor-
mations, Wiley-VCH, Weinheim.
2. (a) Pleiss J (2012) Rational design of enzymes. In: Drauz K, Gröger H, May O (eds) Enzyme
catalysis in organic synthesis, 3rd edn. Wiley-VCH, Weinheim, p 89–117; (b) Ema T, Nakano
Y, Yoshida D, Kamata S, Sakai T (2012) Redesign of enzyme for improving catalytic activity
and enantioselectivity toward poor substrates: manipulation of the transition state. Org Biomol
Chem 10:6299–6308; (c) Steiner K, Schwab H (2012) Recent advances in rational approaches
for enzyme engineering. Comput Struct Biotechnol J 2:e201209010.
3. Reviews of directed evolution: (a) Bommarius AS (2015) Biocatalysis, a status report. Annu
Rev Chem Biomol Eng 6:319–345; (b) Denard CA, Ren H, Zhao H (2015) Improving and
repurposing biocatalysts via directed evolution. Curr Opin Chem Biol 25:55–64; (c) Currin
A, Swainston N, Day PJ, Kell DB (2015) Synthetic biology for the directed evolution of
protein biocatalysts: navigating sequence space intelligently. Chem Soc Rev 44:1172–1239;
(d) Gillam EMJ, Copp JN, Ackerley DF (eds) (2014) Directed evolution library creation. In:
Methods in molecular biology, Humana Press, Totowa; (e) Widersten M (2014) Protein engi-
neering for development of new hydrolytic biocatalysts. Curr Opin Chem Biol 21:42–47; (f)
Reetz MT (2012) Directed evolution of enzymes. In Drauz K, Gröger H, May O (eds) Enzyme
catalysis in organic synthesis, 3rd edn. Wiley-VCH, Weinheim, p 119–190
4. Review of directed evolution of protein thermostability: (a) Arnold FH (1998) Design by
directed evolution. Acc Chem Res 31:125–131; (b) Petrounia IP, Arnold FH (2000) Designed
evolution of enzymatic properties. Curr Opin Biotechnol 11: 325–330; (c) Polizzi KM,
Bommarius AS, Broering JM, Chaparro-Riggers JF (2007) Stability of biocatalysts. Curr Opin
Chem Biol 11:220–225; (d) Tokuriki N, Tawfik DS (2009) Stability effects of mutations and
protein evolvability. Curr Opin Struct Biol 19:596–604
96 M.T. Reetz
5. Reviews of directed evolution of stereoselectivity[3f]: (a) Reetz MT (2011) Laboratory evolu-

tion of stereoselective enzymes: a prolific source of catalysts for asymmetric reactions. Angew
Chem Int Ed 50:138–174; (b) Reetz MT (2013) Biocatalysis in organic chemistry and biotech-
nology: past, present, and future. J Am Chem Soc 135:12480–12496
6. Reetz MT, Zonta A, Schimossek K, Liebeton K, Jaeger KE (1997) Creation of enantiose-
lective biocatalysts for organic chemistry by in vitro evolution. Angew Chem Int Ed Eng
36:2830–2832
7. Reetz MT, Wilensek S, Zha D, Jaeger KE (2001) Directed evolution of an enantioselec-
tive enzyme through combinatorial multiple-cassette mutagenesis. Angew Chem Int Ed
40:3589–3591
8. Reetz MT, Puls M, Carballeira JD, Vogel A, Jaeger KE, Eggert T, Thiel W, Bocola M, Otte N
(2007) Learning from directed evolution: further lessons from theoretical investigations into
cooperative mutations in lipase enantioselectivity. Chembiochem 8:106–112
9. Zha DX, Wilensek S, Hermes M, Jaeger KE, Reetz MT (2001) Complete reversal of enantiose-
lectivity of an enzyme-catalyzed reaction by directed evolution. Chem Commun 24:2664–2665
10. Reetz MT (2004) Controlling the enantioselectivity of enzymes by directed evolution: practi-
cal and theoretical ramifications. Proc Natl Acad Sci U S A 101:5716–5722
11. Reetz MT, Bocola M, Carballeira JD, Zha DX, Vogel A (2005) Expanding the range of sub-
strate acceptance of enzymes: combinatorial active-site saturation test. Angew Chem Int Ed
44:4192–4196
12. (a) Reetz MT, Wang LW, Bocola M (2006) Directed evolution of enantioselective enzymes:
iterative cycles of CASTing for probing protein-sequence space. Angew Chem Int Ed
45:1236–1241; (b) Reetz MT (2005) Evolution im Reagenzglas: Neue Perspektiven für die
Weiße Biotechnologie. Tätigkeitsberichte der Max-Planck-Gesellschaft 327–331
13. (a) Reetz MT, Kahakeaw D, Lohmer R (2008) Addressing the numbers problem in directed
evolution. Chembiochem 9:1797–1804; (b) Clouthier CM, Kayser MM, Reetz MT (2006)
Designing new Baeyer-Villiger monooxygenases using restricted CASTing. J Org Chem
71:8431–8437
14. (a) Firth AE, Patrick WM (2008) GLUE-IT and PEDEL-AA: new programmes for analyzing
protein diversity in randomized libraries. Nucleic Acids Res 36 (Web Server issue):W281–285;
(b) Denault M, Pelletier JN (2007) Protein library design and screening: working out the
probabilities. In: Arndt KM, Müller KM (eds) Protein engineering protocols, Humana Press,
Totowa, p 127–154
15. Nov Y (2012) When second best is good enough: another probabilistic look at saturation muta-
genesis. Appl Environ Microbiol 78:258–262
16. Reetz MT, Prasad S, Carballeira JD, Gumulya Y, Bocola M (2010) Iterative saturation muta-
genesis accelerates laboratory evolution of enzyme stereoselectivity: rigorous comparison
with traditional methods. J Am Chem Soc 132:9144–9152
17. Reetz MT (2013) The importance of additive and non-additive mutational effects in protein
engineering. Angew Chem Int Ed 52:2658–2666
18. Parikh MR, Matsumura I (2005) Site-saturation mutagenesis is more efficient than DNA
shuffling for the directed evolution of beta-fucosidase from beta-galactosidase. J Mol Biol
352:621–628
19. Acevedo-Rocha CG, Hoebenreich S, Reetz MT (2014) Iterative saturation mutagenesis: a
powerful approach to engineer proteins by systematically simulating Darwinian evolution.
Methods Mol Biol 1179:103–128
20. Sun Z, Wikmark Y, Bäckvall J-E, Reetz MT (2016) New concepts for increasing the efficiency
in directed evolution of stereoselective enzymes. Chem Eur J 22:5046–5054
21. (a) Sun Z, Lonsdale R, Kong XD, Xu JH, Zhou J, Reetz MT (2015) Reshaping an enzyme
binding pocket for enhanced and inverted stereoselectivity: use of smallest amino acid alpha-
bets in directed evolution. Angew Chem Int Ed 54:12410–12415; (b) Sandström AG, Wikmark
Y, Engström K, Nyhlen J, Bäckvall JE (2012) Combinatorial reshaping of the Candida antarc-
tica lipase A substrate pocket for enantioselectivity using an extremely condensed library. Proc
Natl Acad Sci U S A 109:78–83
22. (a) Zhang ZG, Lonsdale R, Sanchis J, Reetz MT (2014) Extreme synergistic mutational effects
in the directed evolution of a baeyer-villiger monooxygenase as catalyst for asymmetric sulf-
oxidation. J Am Chem Soc 136:17262–17272; (b) Sun Z, Lonsdale R, Wu L, Li G, Li A,
Wang J, Zhou J, Reetz MT (2016) Structure-guided triple code saturation mutagenesis: effi-
cient tuning of the stereoselectivity of an epoxide hydrolase. ACS Catal 6:1590–1597; (c)
Sun Z, Lonsdale R, Ilie A, Li G, Zhou J, Reetz MT (2016) Catalytic asymmetric reduction of
difficult-to-reduce ketones: triple code saturation mutagenesis of an alcohol dehydrogenase.
ACS Catal 6:1598–1605
23. (a) Bougioukou DJ, Kille S, Taglieber A, Reetz MT (2009) Directed Evolution of an enanti-
oselective enoate-reductase: testing the utility of iterative saturation mutagenesis. Adv Synth
Catal 351:3287–3305; (b) Sullivan B, Walton AZ, Stewart JD (2013) Library construction and
evaluation for site saturation mutagenesis. Enzym Microb Technol 53:70–77
24. Polizzi KM, Parikh M, Spencer CU, Matsumura I, Lee JH, Realff MJ, Bommarius AS (2006)
Pooling for improved screening of combinatorial libraries for directed evolution. Biotechnol
Prog 22:961–967
25. Acevedo-Rocha CG, Reetz MT, Nov Y (2015) Economical analysis of saturation mutagenesis
experiments. Sci Rep 5:10654
26. Kwan DH, Constantinescu I, Chapanian R, Higgins MA, Kotzler MP, Samain E, Boraston
AB, Kizhakkedathu JN, Withers SG (2015) Toward efficient enzymes for the generation of
universal blood through structure-guided directed evolution. J Am Chem Soc 137:5695–5705
27. (a) Whitehouse CJ, Bell SG, Wong LL (2012) P450(BM3) (CYP102A1): connecting the
dots. Chem Soc Rev 41:1218–1260; (b) Fasan R (2012) Tuning P450 enzymes as oxidation
catalysts. ACS Catal 2:647–666; (c) Bernhardt R, Urlacher VB (2014) Cytochromes P450 as
promising catalysts for biotechnological application: chances and limitations. Appl. Microbiol
Biotechnol 98:6185–6203; (d) Holtmann D, Fraaije MW, Arends IW, Opperman DJ,
Hollmann F (2014) The taming of oxygen: biocatalytic oxyfunctionalisations. Chem Commun
50:13180–13200; (e) Roiban GD, Reetz MT (2015) Expanding the toolbox of organic chem-
ists: directed evolution of P450 monooxygenases as catalysts in regio- and stereoselective
oxidative hydroxylation. Chem Commun 51:2208–2224; (f) Girvan HM, Munro AW (2016)
Applications of microbiol cytochrome P450 enzymes in biotechnology and synthetic biology.
Curr Opin Chem Biol 31:136–145
28. (a) Tang WL, Li Z, Zhao H (2010) Inverting the enantioselectivity of P450pyr monooxygenase
by directed evolution. Chem Commun 46:5461–5463; (b) Pham SQ, Pompidor G, Liu J, Li
XD, Li Z (2012) Evolving P450pyr hydroxylase for highly enantioselective hydroxylation at
non-activated carbon atom. Chem Commun 48:4618–4620
29. Lonsdale R, Harvey JN, Mulholland AJ (2010) Inclusion of dispersion effects significantly
improves accuracy of calculated reaction barriers for cytochrome P450 catalyzed reactions.
J Phys Chem Lett 1:3232–3237
30. Agudo R, Roiban GD, Reetz MT (2012) Achieving regio- and enantioselectivity of P450-
catalyzed oxidative CH activation of small functionalized molecules by structure-guided
directed evolution. Chembiochem 13:1465–1473
31. Kille S, Zilly FE, Acevedo JP, Reetz MT (2011) Regio- and stereoselectivity of P450-catalysed
hydroxylation of steroids controlled by laboratory evolution. Nat Chem 3:738–743
32. Le-Huu P, Heidt T, Claasen B, Laschat S, Urlacher VB (2015) Chemo-, regio-, and stereose-
lective oxidation of the monocyclic diterpenoid beta-cembrenediol by P450 BM3. ACS Catal
5:1772–1780
33. Zhang K, Shafer BM, Demars MD 2nd, Stern HA, Fasan R (2012) Controlled oxidation of
remote sp3 C-H bonds in artemisinin via P450 catalysts with fine-tuned regio- and stereoselec-
tivity. J Am Chem Soc 134:18695–18704
34. (a) Chen MM, Snow CD, Vizcarra CL, Mayo SL, Arnold FH (2012) Comparison of ran-
dom mutagenesis and semi-rational designed libraries for improved cytochrome P450 BM3-
catalyzed hydroxylation of small alkanes. Prot Eng Des Sel 25:171–178; (b) Ritter C, Nett
N, Acevedo-Rocha CG, Lonsdale R, Kraling K, Dempwolff F, Hoebenreich S, Graumann
PL, Reetz MT, Meggers E (2015) Bioorthogonal enzymatic activation of caged compounds.
98 M.T. Reetz
Angew Chem Int Ed 54:13440–13443; (c) Dennig A, Lulsdorf N, Liu H, Schwaneberg

U (2013) Regioselective o-hydroxylation of monosubstituted benzenes by P450 BM3.
Angew Chem Int Ed 52:8459–8462; (d) Agudo R, Roiban GD, Lonsdale R, Ilie A, Reetz
MT (2015) Biocatalytic route to chiral acyloins: P450-catalyzed regio- and enantioselec-
tive alpha-hydroxylation of ketones. J Org Chem 80:950–956; (e) Hoebenreich S, Zilly FE,
Acevedo-Rocha CG, Zilly M, Reetz MT (2015) Speeding up directed evolution: combining
the advantages of solid-phase combinatorial gene synthesis with statistically guided reduc-
tion of screening effort. ACS Synth Biol 4:317–331; (f) Roiban GD, Agudo R, Reetz MT
(2014) Cytochrome P450 catalyzed oxidative hydroxylation of achiral organic compounds
with simultaneous creation of two chirality centers in a single C-H activation step. Angew
Chem Int Ed 53:8659–8663; (g) Hu S, Huang J, Mei LH, Yu Q, Yao SJ, Jin ZH (2010) Altering
the regioselectivity of cytochrome P450 BM-3 by saturation mutagenesis for the biosynthe-
sis of indirubin. J Mol Catal B-Enzym 67:29–35; (h) Nguyen KT, Virus C, Gunnewich N,
Hannemann F, Bernhardt R (2012) Changing the regioselectivity of a P450 from C15 to C11
hydroxylation of progesterone. Chembiochem 13:1161–1166; (i) Yang Y, Liu J, Li Z (2014)
Engineering of P450pyr hydroxylase for the highly regio- and enantioselective subterminal
hydroxylation of alkanes. Angew Chem Int Ed 53:3120–3124
35. (a) Reinen J, Vredenburg G, Klaering K, Vermeulen NPE, Commandeur JNM, Honing M,
Vos JC (2015) Selective whole-cell biosynthesis of the designer drug metabolites 15-or
16-ß-hydroxynorethisterone by engineered Cytochrome P450 BM3 mutants. J Mol Catal
B-Enzym 121:64–74; (b) Bruhlmann F, Bosijokovic B, Ullmann C, Auffray P, Fourage L,
Wahler D (2013) Directed evolution of a 13-hydroperoxide lyase (CYP74B) for improved
process performance. J Biotechnol 163:339–345
36. Reetz MT, Brunner B, Schneider T, Schulz F, Clouthier CM, Kayser MM (2004) Directed
evolution as a method to create enantioselective cyclohexanone monooxygenases for catalysis
in Baeyer-Villiger reactions. Angew Chem Int Ed 43:4075–4078
37. Mihovilovic MD, Rudroff F, Winninger A, Schneider T, Schulz F, Reetz MT (2006) Microbial
Baeyer-Villiger oxidation: stereopreference and substrate acceptance of cyclohexanone mono-
oxygenase mutants prepared by directed evolution. Org Lett 8:1221–1224
38. (a) Fraaije MW, Wu J, Heuts DP, van Hellemond EW, Spelberg JH, Janssen DB (2005)
Discovery of a thermostable Baeyer-Villiger monooxygenase by genome mining. Appl
Microbiol Biotechnol 66:393–400; (b) Malito E, Alfieri A, Fraaije MW, Mattevi A (2004)
Crystal structure of a Baeyer-Villiger monooxygenase. Prod Natl Acad Sci 101:13157–13162
39. Reviews of directed evolution of Baeyer-Villiger monooxygenases: (a) Zhang ZG, Parra LP,
Reetz MT (2012) Protein engineering of stereoselective Baeyer-Villiger monooxygenases.
Chem Eur J 18:10160–10172; (b) Balke K, Kadow M, Mallin H, Sass S, Bornscheuer UT
(2012) Discovery, application and protein engineering of Baeyer-Villiger monooxygenases for
organic synthesis. Org Biomol Chem 10:6249–6265
40. Reetz MT, Wu S (2008) Greatly reduced amino acid alphabets in directed evolution: making
the right choice for saturation mutagenesis at homologous enzyme positions. Chem Commun
43:5499–5501
41. Kirschner A, Bornscheuer UT (2006) Kinetic resolution of 4-hydroxy-2-ketones catalyzed by
a Baeyer-Villiger monooxygenase. Angew Chem 45:7004–7006
42. Wu S, Acevedo JP, Reetz MT (2010) Induced allostery in the directed evolution of an enanti-
oselective Baeyer-Villiger monooxygenase. Prod Natl Acad Sci USA 107:2775–2780
43. Ichiye T, Karplus M (1991) Collective motions in proteins – a covariance analysis of atomic
fluctuations in molecular-dynamics and normal mode simulations. Proteins 11:205–217
44. Tang L, Gao H, Zhu X, Wang X, Zhou M, Jiang R (2012) Construction of “small-intelli-
gent” focused mutagenesis libraries using well-designed combinatorial degenerate primers.
BioTechniques 52:149–158
45. Reetz MT, Sanchis J (2008) Constructing and analyzing the fitness landscape of an experimen-
tal evolutionary process. Chembiochem 9:2260–2267
46. Gumulya Y, Sanchis J, Reetz MT (2012) Many pathways in laboratory evolution can lead to
improved enzymes: how to escape from local minima. Chembiochem 13:1060–1066
47. Wu Q, Soni P, Reetz MT (2013) Laboratory evolution of enantiocomplementary Candida ant-

arctica lipase B mutants with broad substrate scope. J Am Chem Soc 135:1872–1881
48. Engström K, Nyhlen J, Sandström AG, Bäckvall JE (2010) Directed evolution of an enantiose-
lective lipase with broad substrate scope for hydrolysis of alpha-substituted esters. J Am Chem
Soc 132:7038–7042
49. Wikmark Y, Svedendahl Humble M, Bäckvall JE (2015) Combinatorial library based engi-
neering of Candida antarctica lipase A for enantioselective transacylation of sec-alcohols in
organic solvent. Angew Chem Int Ed 54:4284–4288
50. (a) Ma J, Wu L, Guo F, Gu J, Tang X, Jiang L, Liu J, Zhou J, Yu H (2013) Enhanced enanti-
oselectivity of a carboxyl esterase from Rhodobacter sphaeroides by directed evolution. Appl
Microbiol Biotechnol 97:4897–4906; (b) Luan ZJ, Li FL, Dou S, Chen Q, Kong XD, Zhou
JH, Yu HL, Xu JH (2015) Substrate channel evolution of an esterase for the synthesis of cilas-
tatin. Catal Sci Technol 5:2622–2629; (c) Godinho LF, Reis CR, van Merkerk R, Poelarends
GJ, Quax WJ (2012) An esterase with superior activity and enantioselectivity towards
1,2-o-isopropylideneglycerol esters obtained by protein design. Adv Synth Catal 354:3009–
3015; (d) Nobili A, Tao Y, Pavlidis IV, van den Bergh T, Joosten HJ, Tan T, Bornscheuer UT
(2015) Simultaneous use of in silico design and a correlated mutation network as a tool to
efficiently guide enzyme engineering. Chembiochem 16:805–810
51. (a) Gupta RD, Tawfik DS (2008) Directed enzyme evolution via small and effective neutral
drift libraries. Nat Methods 5:939–942; (b) Kaltenbach M, Tokuriki, N (2014) Generation
of effective libraries by neutral drift. Methods Molec Biol 1179:69–81; (c) Smith WS,
Hale JR, Neylon C (2011) Applying neutral drift to the directed molecular evolution of a
ß-glucuronidase into a ß-galactosidase: two different evolutionary pathways lead to the same
variant. BMC Res Notes 4:138; (d) Bernath-Levin K, Shainsky J, Sigawi L, Fishman A (2014)
Directed evolution of nitrobenzene dioxygenase for the synthesis of the antioxidant hydroxy-
tyrosol. Appl Microbiol Biotechnol 98:4975–4985
52. (a) Eigen M, McCaskill J, Schuster P (1988) Molecular quasi-species. J Phys Chem 92:6881–
6891; (b) Kurtovic S, Mannervik B (2009) Identification of emerging quasi-species in directed
enzyme evolution. Biochemistry 48:9330–9339
53. Savile CK, Janey JM, Mundorff EC, Moore JC, Tam S, Jarvis WR, Colbeck JC, Krebber A,
Fleitz FJ, Brands J, Devine PN, Huisman GW, Hughes GJ (2010) Biocatalytic asymmetric syn-
thesis of chiral amines from ketones applied to sitagliptin manufacture. Science 329:305–309
54. Chen Z, Zhao H (2005) Rapid creation of a novel protein function by in vitro coevolution.
J Mol Biol 348:1273–1282
55. (a) Ostermeier M, Benkovic SJ (2000) Evolution of protein function by domain swapping.
Adv Protein Chem 55:29–77; (b) Park AH, Park HY, Sohng JK, Lee HC, Liou K, Yoon YJ,
Km BG (2009) Expanding substrate specificity of GT-B fold glycosyltransferase via domain
swapping and high-throughput screening. Biotechnol Bioeng 102:988–994
56. (a) Yu Y, Lutz S (2011) Circular permutation: a different way to engineer enzyme structure
and function. Trends Biotechnol 29:18–25; (b) Qian Z, Lutz S (2005) Improving the catalytic
activity of Candida antarctica lipase B by circular permutation. J Am Chem Soc 127:13466–
13467; (c) Yu Y, Lutz S (2010) Improved triglyceride transesterification by circular permu-
tated Candida antarctica lipase B. Biotechnol Bioeng 105:44–50.
Improving CO2 Fixation by Enhancing
Rubisco Performance 4
Robert H. Wilson and Spencer M. Whitney
Abstract
The photosynthetic enzyme linking the inorganic and organic phases of the
biosphere is ribulose-1,5-bisphosphate [RuBP] carboxylase/oxygenase

(Rubisco). The complicated catalytic chemistry of Rubisco slows its CO2 fixa-
tion rate, allows for competitive inhibition by oxygen and permits the production
of misfire products that can self-inhibit activity. Significant effort has been
invested into better understanding the structure-function details of Rubisco as
improving its performance is recognised as a viable means to enhance the photo-
synthetic efficiency and yield potential of crops. While rational design approaches
have still been unable to provide catalysis enhancing solutions, modern directed
evolution tools are posing a promising conduit to improving Rubisco. Advances
in the design of effective selection systems for mutagenic Rubisco library screen-
ing have strategically increased their focus on using Escherichia coli. The inher-
ent sensitivity of E. coli viability to the pentose sugar substrate of Rubisco,
RuBP, is being exploited in an increasingly effective manner to select for Rubisco
mutants with increased activity. Here we review the differing directed evolution
technologies used to evolve Rubisco, examine the merits of available high-
throughput Rubisco-dependent E. coli (RDE) selection systems and postulate
approaches for improving their functionality.
R.H. Wilson • S.M. Whitney (*)

Research School of Biology, The Australian National University,
Acton, Australian Capital Territory 2601, Australia

DOI 10.1007/978-3-319-50413-1_4
102 R.H. Wilson and S.M. Whitney
4.1 Rubisco: A Target for Improvement
4.1.1 A Rate-Limiting Enzyme in Photosynthesis
The CO2-fixing enzyme ribulose-1,5-bisphosphate carboxylase/oxygenase

(Rubisco) is the catalytic gatekeeper linking the inorganic and organic phases of the
biosphere’s carbon cycle. Rubisco initiates carbon assimilation in the Calvin cycle
by fixing CO2 to the pentose sugar substrate ribulose-1,5-bisphosphate (RuBP) and
cleaving it into two molecules of 3-phosphoglycerate (3-PGA, Fig. 4.1a). The
3-PGA is the precursor to carbohydrate synthesis that is required for energy and
growth. In plants and algae the carboxylation reaction of Rubisco occurs at a slow
pace (~1–5 cycles per second) resulting in its catalytic properties often limiting the
rate of photosynthesis and growth in these organisms [9, 45]. To compensate for this
shortcoming, many photosynthetic organisms require high amounts of Rubisco to
meet their metabolic needs. For example, Rubisco can comprise up to 50% of the
soluble protein in rice and wheat leaves [9, 68]. This high investment in Rubisco is
therefore critical to supporting primary productivity in the global food chain which
results in it being the most abundant enzyme on Earth [16].
Rubisco is a bifunctional enzyme as it also catalyses RuBP oxygenation to
produce 3-PGA and 2-phosphoglycolate (2-PG). As 2-PG is toxic to growth,

organisms have evolved metabolic mechanisms to recycle it [15]. In plants this
occurs via photorespiration, a multi-organelle pathway requiring energy and emitting
CO2 – resulting in the loss of fixed carbon (Fig. 4.1a). In C3 crops like rice and wheat,
this carbon loss typically amounts to 25% of the CO2 assimilated by photosynthesis
[45]. As a consequence of these catalytic inefficiencies, improving the performance
of Rubisco has been an objective spanning more than three decades [17, 53, 85].
Fig. 4.1 Photosynthesis, Rubisco and Rubisco activase. (a) Summary of oxygenic photosynthesis
where the light reactions produce O2 from H2O photolysis and the derived energy is used in ana-
bolic reactions such as the Calvin cycle in chloroplasts. The cycle begins with the carboxylation of
ribulose-1,5-bisphosphate (RuBP) to produce 3-phosphoglycerate (3-PGA) that is reduced to
glyceraldehyde-3-phosphate, the precursor for carbohydrate (CHO) synthesis or RuBP regenera-
tion. Oxygenation of RuBP by Rubisco produces 2-phosphoglycolate (2-PG) whose recycling to
3-PGA by photorespiration occurs at the cost of energy and release of fixed CO2 (shown in grey).
(b) Distribution of the differing Rubisco forms in nature, their subunit structure, gene composition
(rbcL ± rbcS), genome location and copy number. The multiplicity of ancillary proteins that influ-
ence Rubisco biogenesis varies throughout nature. Examples include the protein folding GroEL-
GroES or corresponding Cpn60/Cpn20-23/Cpn10 (CPN, chloroplasts) and thermosome (TSM)
chaperonin complexes, Rubisco accumulation factors 1 and 2 (Raf1, Raf2), RbcX and the stromal
protein bundle sheath defective2 protein (BSD2) (Reviewed in [30]). (c) Carbamylation of each
catalytic site via the binding of CO2 (C) and then Mg2+ (M) is required to form an active ternary
complex (ECM) that can bind RuBP and initiate its carboxylation or oxygenation. Inactive sites
form when RuBP binds before carbamylation (ER) or when carbamylated catalytic sites are occu-
pied by inhibitory sugar phosphate ligands (ECMI). Example inhibitors are xylulose-1,5-
bisphosphate (XuBP, a catalytic misfire product) and carboxyarabinitol 1-phosphate (CA1P,
regulatory “shade” inhibitor in plants) whose removal is facilitated by Rubisco activase via transi-
tory intermolecular interactions powered by ATP hydrolysis
4 Improving CO2 Fixation by Enhancing Rubisco Performance 103
4.1.2 Natural Diversity in Rubisco Form and Function
Over 50 years of research has provided valuable knowledge regarding Rubisco’s

genetic and structural diversity in nature. Confidence that improving Rubisco for
agricultural applications is a feasible objective stems from the discovery of more
a Photosynthesis Growth
Light capture &
conversion (CHO)n
RuBP CO2
O2
Light Carbon metabolism energy
Energy (e.g. Calvin Cycle) Rubisco
thylakoids (e.g. NADPH, ATP) 2-PG CO2
e- H+
3-PGA
H2O O2
b The genetics and structural diversity of Rubisco
Vascular plants Cyanobacteria (α-*/β-#) Proteobacteria Archaea

Green algae Proteobacteria, Dinoflagellates
Non-green algae rbcL
rbcL plastome (n=1) rbcL nucleus (n≥1)
rbcL rbcS nucleus (n≥1) TSM
rbcS nucleus (n>2)
nucleus/plastome (n=1) CPN /GroELS (?)
CPN (?)
RbcX (?) CPN /GroELS
Raf1 Form I RbcX# Form II Form III
Raf2 Raf1#
BSD2 Raf2* RuBP
Form 1B L 8 S8 Form 1A-D L2 L6 L8 L2 L10
c The activation, catalysis and metabolic regulation of Rubisco
CO2
O2
CO2 Mg2+ RuBP
M 3-PGA
201Lys slow fast C + 2PG
RuBP
Mg2+
3-PGA
E CO2 ECM ECMR (x2)
RuBP XuBP
(CA1P) XuBP
RuBP
(CA1P)
RuBP XuBP
M
ADP C ADP
ATP ATP
ER ECMI
Inactive Active Rubisco activase

catalytic site catalytic site (AAA+ regulatory protein)
efficient versions of Rubisco found naturally in some red algae [6, 83]. Of benefit is
the oligomeric structure of Rubisco in plants and algae, as well as many photosyn-
thetic prokaryotes (e.g. cyanobacteria and many photolithotrophic bacteria), which
share the same subunit complement – eight ~50 kDa large (L) subunits and eight
~14–17 kDa small (S) subunits [4]. The L8S8 hexameric holoenzyme is termed form
I Rubisco (Fig. 4.1b) and constitutes the most abundant Rubisco isoform in the bio-
sphere. In plants and green algae, the S-subunits are coded by multiple rbcS genes
and the L-subunit by a single rbcL gene in the plastid genome (plastome) [85]. Each
S-subunit is transported into the chloroplast for assembly with L8 cores in a highly
complex process that requires many known and suspected molecular partners [30,
88] (Fig. 4.1b). While some of these assembly requirements appear conserved for
some L8S8 prokaryotic Rubiscos, the requirements for Rubisco biogenesis in non-
green algae are unclear and are not met by leaf chloroplasts [30, 83]. Associated with
their different assembly requirements are likely the preserved coding of Rubisco as a
rbcL – rbcS operon in the nongreen algae plastome (Fig. 4.1b) [1].
While the S-subunit of L8S8 Rubisco is critical for catalysis and enzyme stability,
other forms of Rubisco found in nature comprise only L-subunits [4]. Form II
Rubisco exists as L2 dimers or oligomers of dimers (Fig. 4.1b). The best studied
form II Rubisco is the L2 Rubisco from the bacterium Rhodospirillum rubrum. Form
III Rubisco can be found in Archaea and have no role in CO2 assimilation but are
capable of sustaining photosynthetic growth when transformed into leaf chloro-
plasts [2, 89]. Form III Rubisco functions to metabolise the RuBP produced in some
archaea during the nucleotide salvaging pathways [63] and can exist as L2 or as
RuBP-induced tetrameric/pentameric assemblages of L2 (Fig. 4.1b) [2].
4.1.3 A Highly Conserved Catalytic Site and Mechanism
Crystal structure examples for all three Rubisco forms have been solved [4, 9].
Consistent with a conserved catalytic mechanism, all forms of Rubisco show a strik-
ing conservation in L-subunit tertiary structure and amino acid arrangement in the
catalytic site. This conservation occurs despite the L-subunit primary sequence dif-
fering up to 70%. The structural conservation of the catalytic site suggests that the
basic scaffold of Rubisco is essential for function, although attributing what specific
sequences elicit the vast catalytic variation found in the Rubisco superfamily con-
tinues to prove challenging (Table 4.1). The catalytic sites form at the interfaces
between adjoining pairs of L-subunits that orient head to tail to form a L2 – the mini-
mal functional unit for Rubisco. Conserved amino acid residues from the C-terminal
domain of one L-subunit and the N-terminal domain of the other L-subunit contrib-
ute to the two catalytic sites in each L2 unit [4].
Prior to catalysis Rubisco must first be activated (Fig. 4.1c). Activation requires
the binding of a CO2 molecule to the ε-amino group of a conserved L-subunit
201
lysine (K201) in the catalytic site [10, 28]. Carbamylation of K201 produces an
anionic carbamate that is stabilised by Mg2+ binding resulting in an activated (car-
bamylated) ternary “ECM” complex (E for enzyme catalytic site, C for
Table 4.1 Catalytic variation at 25 °C among the Rubisco superfamily
Organisms with a CCM are shaded in grey

n.d not determined
a
Young et al. [92]
b
Sharwood et al. [67]
c
Leggat et al. [42]
d
Whitney and Andrews [81]
e
Alonso et al. [2] and Wilson et al. [89]
f
Values measured at 10 °C
carbamylated K201, M for Mg2+, Fig. 4.1c). In this active state, the catalytic site can
bind RuBP and form an enediol intermediate that provides a nucleophilic site at the
C2 carbon of RuBP for binding substrate CO2 (carboxylation) or O2 (oxygenation).
A series of hydration, protonation and cleavage reactions follow that result in for-
mation of the respective 3-PGA or 2-PG products (Fig. 4.1a). The complexity of this
catalytic chemistry, that is compounded by the inability of Rubisco to bind either
gas substrate, has posed a significant challenge to both nature and scientific research
in identifying solutions for improving Rubisco performance [53]. Nevertheless, the
large natural diversity in the catalytic properties of Rubisco (Table 4.1) confirms
evolutionary adaptation of the enzyme. For example, Rubisco from organisms with
CO2 concentrating mechanisms (CCM) that elevate CO2 concentrations around the
enzyme typically have higher CO2 fixation rates (kcatC) but lower CO2 affinities
(i.e. a higher Km for CO2, KC) [67], although this does not appear to be the case for
diatom (nongreen microalgae) Rubisco (Table 4.1) [92].
4.1.4 What Constitutes a “Better” Rubisco?
Throughout the literature are misconceptions about what constitutes a better

Rubisco. Improving kcatC is a desired change in organisms containing a CCM able
to provide saturating levels of CO2 that can compensate for any accompanying
reductions in CO2 affinity (i.e. increasing KC) or decline in the enzyme specificity
for CO2 over O2 (Sc/o) [67]. In organisms lacking a CCM (e.g. C3 plants), improving
kcatC does not necessarily constitute a better Rubisco. For example, the photosyn-
thesis and growth of C3 plants producing the high kcatC Rubisco from a cyanobac-
terium (Table 4.1) will not exceed the endogenous plant Rubisco unless the multiple
components of a functional CCM are co-introduced [57]. Therefore, according to
the C3 photosynthesis models of Farquhar et al. (1980) [20], a better Rubisco within
the context of a C3 plant is one with an improvement in Sc/o that is accompanied by
an increase in carboxylation efficiency under ambient O2 (defined as kcatC/KC21%O2)
[6, 67]. Importantly, modest declines in kcatC can be tolerated if sufficiently offset
by substantial improvement in CO2 affinity (i.e. low KC21%O2).
For many Rubisco isoforms the values for SC/O tend to be inversely correlated
with kCcat suggesting a catalytic trade-off between these two parameters. This fea-
ture has questioned the feasibility of engineering a faster and more CO2-specific
enzyme [64, 76]. However, kinetic surveys indicate this relationship significantly
diverges when considering Rubisco isoforms other than those from vascular plants
and a small sampling of algae and proteobacteria. For example, the SC/O-kCcat
inverse correlation strongly diverges when Rubisco from cyanobacteria, Archaea
and nongreen algae are considered [89, 92]. This is best illustrated by the slightly
lower kCcat, but vastly higher SC/O of Rubisco from the red algae Griffithsia monilis
that has the potential to enhance the photosynthetic efficiency of plants with a C3
physiology like wheat and rice by up to 30% [45, 83]. The feasibility of uncoupling
this relationship by directed evolution has also been demonstrated by the identifica-
tion of point mutations in archaeal Rubisco from Methanococcoides burtonii that
improve both kcatC and Sc/o [89].
4.1.5 Factors Limiting Mutagenic Study of Rubisco
The discovery that vascular plant Rubisco is not the pinnacle of evolution has engen-
dered confidence that the challenge of improving the enzyme for agricultural appli-
cations is not insurmountable. Efforts to improve form I Rubisco function have been
hindered by the inherent complexity of eukaryote L8S8 Rubisco biogenesis [30]. In
particular the L-subunit folding and assembly needs are not met in E. coli [33], even
when co-expressed with cognate chaperonin [11] or the Rubisco-specific assembly
chaperones RbcX and Rubisco accumulation factor 1 (Raf1) [21]. Most Rubisco
mutagenic studies have therefore focused on Rubisco from cyanobacteria and pro-
teobacteria that can be functionally expressed in E. coli, albeit to varying degrees of
competency [51]. Direct mutagenic testing of Rubisco in photosynthetic hosts such
as Rhodobacter sphaeroides (proteobacterium, [19]), Synechococcus sp. PCC6803
(cyanobacteria, [3]) and Chlamydomonas reinhardtii (green algae, [61]) has been
facilitated by the availability of rbcL ± rbcS-deficient mutants (Fig. 4.2a).
Unfortunately the viability of these photosynthetic hosts for screening Rubisco
mutagenic libraries is primarily confined by their low transformation efficiency (Fig.
4.2a). Similarly, while leaf chloroplast transformation provides the only system for
study of recombinant plant Rubiscos, the high cost and slowness of this technology
limit its usefulness for random mutagenesis studies of Rubisco [50, 85, 86].
4.2 Directed Evolution of Rubisco
Exploration of Rubisco sequence space to identify solutions that improve

performance has been a long-standing challenge. Our extensive knowledge of
Rubisco structure and function still remains insufficient to allow improved
performance by rational design. The success of directed evolution in numerous
enzyme bioengineering endeavours has lured researchers in both academia and
industry into applying these tools to improve Rubisco performance [8, 23, 51,
89, 93]. Of particular appeal is the level of control available through both unbi-
ased mutagenesis and curated selection using directed evolution and the capac-
ity to successively sample sequence space to make incremental improvements
towards a desired function [12, 13]. The last two decades has seen increasing
success in these endeavours with examples of forms I, II and III Rubisco having
been subject to directed evolution. As summarised in Table 4.2, most studies
have focused on cyanobacteria L8S8 Rubisco with particular attention made on
mutating the rbcL gene and only in a few cases including rbcS [8, 50, 69]. The
general outcomes for some key mutagenic screening studies are described in
Table 4.2.
4.2.1 The Challenges of Identifying an “Improved Rubisco”
A key consideration in any directed evolution application is identifying a suitable

screening system. This requires consideration on whether the screening system
meets the desired throughput, has sufficient fidelity to avoid selection of “false posi-
tives” and is suitably sensitive to identify minor and incremental improvements in a
desired trait (i.e. successive increases in fitness). With regard to Rubisco, selection
systems incorporating an in vitro catalytic screening component are typically com-
promised by throughput and fidelity. The challenges are primarily associated with
being able to accurately measure Rubisco content and quantify the multiplicity of
kinetic parameters that equate to an improved enzyme. In organisms with a CCM,
this might equate to simply increasing kcatC; in non-CCM-containing hosts,
increases in Sc/o and kcatC/KC21%O2 without extensive undermining of kcatC are typi-
cally required (Sect. 4.1.4).
A lack of suitable infrastructure and experimental familiarity in how to correctly
measure Rubisco content and kinetics continues to confuse the kinetic literature. As
a consequence, large differences in the kinetics for a Rubisco are often reported
between studies. For example, large errors in the reported kinetic properties of a
hybrid plant Rubisco comprising the sunflower L-subunit and tobacco S-subunits
[36] and for Thermococcus kodakaraensis form III Rubisco [18] have been cor-
rected in subsequent studies [40, 66]. With regard to Rubisco directed evolution
studies, some mutants have been incorrectly interpreted as catalytically enhanced
[52] rather than “solubility” mutants with enhanced L-subunit folding and assembly
[26]. These mistakes highlight how accurately measuring changes in Rubisco
synthesis can be problematic, especially using polyacrylamide gel electrophoresis

methods that can be misleading without appropriate controls and experimental
rigour [87]. Potential problems also arise measuring Rubisco kinetics using com-
mercial sources of RuBP contaminated with inhibitory sugar phosphates that reduce
the reliability of the catalytic measurements [68].
a Photosynthetic host (<103 mutants per screen)

Rubisco null host Rubisco library
rbcL ± rbcS Low transformation

efficiency
Heterotrophic; growth (in dark) Screen for colonies with
on external carbon source improved autotrophic growth
b Escherichia coli (wt) (<104 cfu per plate, false positives prominent)
Glycolysis/gluconeogenesis
hexoses glucose gald-3-P gly-1,3-P2 pyruvate growth
3-PGA
PPP
CO2 (+) Rubisco
nucleic PRK
acids rib-5-P RuBP(-) no growth
c MM1E. coli (>3 x 105 cfu per plate, false positives rare) malate,
glycerol cas amino acids
-
pA
glucose gald-3-P gly-1,3-P2 pyruvate growth

ga
PPP PRK Rubisco

nucleic rib-5-P RuBP 3-PGA
acids CO2(O2) (2PG) GM
arabinose IPTG
No Rubisco Wildtype Rubisco Enhanced mutant Rubisco

(control)
d MM1E. coli (RbcX)

(+) S-subunit Rubisco
MM1E. coli (GroEL/S)
(+)
L L
(unfolded) (folded) L2 L8 L8 S8
Table 4.2 The biochemical properties of Rubisco mutants selected by directed evolution
Rubisco origin Main mutations identified and effect on Rubisco References
Selection host: Rhodobacter capsulatus
Synechococcus PCC6301 L-subunit – F342V: twofold increase in host [69]
(L8S8) doubling time, 50% reduction to RuBP affinity
(KmRuBP)
Synechococcus PCC6301 L-subunit – A375V: improved CO2 and reduced [60]
(L8S8) O2 affinities, carboxylation rate 12% of wild type
Selection host: Chlamydomonas reinhardtii
Chlamydomonas reinhardtii L-subunit – A99T, A281S, D352G: improved [93]
(L8S8) specificity, carboxylation rate and affinity for CO2
Nicotiana tabacum (L8S8) L-subunit – A99T, A281S, D352G: little influence [25]
on catalytic properties of tobacco Rubisco
Selection host: Escherichia coli (RDE)
Synechococcus PCC6301 L-subunit – M259T, F342S: four- to ninefold [26]
(L8S8) increase in Rubisco assembly
Rhodospirillum rubrum (L2) H44N, D117V: altered affinity for CO2 and O2 [47]
Synechococcus PCC6301 L-subunit – L161M, M169L: up to 11-fold [51]
(L8S8) increase in Rubisco assembly
Synechococcus PCC7002 S-subunit – E49V, D82G: increased [8]
(L8S8) carboxylation efficiency by 45%
Synechococcus PCC6301 L-subunit – F140I: 2.9-fold increase in [14]
(L8S8) carboxylation efficiency, 9% decrease in
specificity
Methanococcoides burtonii E138V, K332E: 2.8–3.6-fold improvement to [89]
(L2–L10) carboxylation efficiency in air
Fig. 4.2 Directed evolution selection screens for Rubisco, past and present. (a) Low transforma-
tion efficiencies have limited the versatility, and success, of using Rubisco-deficient photosyn-
thetic host mutants to screen Rubisco libraries. (b) By exploiting the sensitivity of E. coli to RuBP
toxicity (−), expressing phosphoribulokinase (PRK, converts ribulose 5-phosphate, rib-5-P, from
the pentose phosphate pathway, PPP, to RuBP) can be used to visually select for increased Rubisco
activity (+, faster RuBP turnover) via improved cell viability (faster colony growth). A high pro-
portion of false positives reduce the sensitivity and throughput of this selection approach.
Maximum practical colony-forming units (cfu) able to be screened per plate are indicated in brack-
ets. (c) Interruption of glycolysis/gluconeogenesis in E. coli via glyceraldehyde-3-phosphate dehy-
drogenase deletion (gapA−) produced a higher fidelity Rubisco-dependent E. coli (RDE) selection
system (MM1) that requires a functional PRK (arabinose inducible) and Rubisco (IPTG inducible)
shunt to bypass glycolysis and allow carbon metabolism to flow from supplied hexose substrate
through to the citric acid cycle for energy and growth [47]. Trace levels of malate, casamino acids
and glycerol are needed to support early cell division upon induction of the PRK-Rubisco shunt.
The 3-PGA or 2-PG produced by Rubisco catalysis is metabolised in E. coli by glycolysis or gly-
colate metabolism (GM; see biocyc.org/ECOLI), respectively [54]. Shown are the colony pheno-
types of RDE-MM1 cells producing no Rubisco (left), Synechococcus PCC6301 Rubisco (middle)
and the higher solubility M259T PCC6301 mutant (right) [26]. (d) Alternative selection screens
using MM1 co-expressing either additional GroEL-GroES chaperonin complexes (which stimu-
late (+) L-subunit folding) or the Rubisco chaperone RbcX (which “staple” post-chaperonin folded
L-subunits into stable L2 units to facilitate L8 formation) altered the selectivity for novel
Synechococcus PCC6803 Rubisco mutants with differing catalytic and biophysical properties [14]
4.2.2 R
ubisco Mutant Selection Using a Photosynthesis
Deficient Host
Directed evolution selection systems require the target molecule under selection to
be identifiable via a screen that can report a desired functional change. As Rubisco
is essential for photosynthesis, initial directed evolution selection systems utilised
an available Rubisco deletion mutant of the photoheterotrophic bacterium
Rhodobacter capsulatus (strain SBI-II) [19]. Although non-photosynthetic, the
Rubisco-deficient SBI-II R. capsulatus cells can grow heterotrophically on media
such as peptone yeast extract medium [69]. Upon transforming with mutagenic
cyanobacteria Rubisco libraries, photosynthetic potential can be screened through
growth on Ormerod’s minimal medium and by varying the growth CO2 levels
(Fig. 4.2a). While able to effectively identify L-subunit residues that influence the
catalytic properties of cyanobacteria (Table 4.2), the approach was limited in
throughput by the very low transformation competency of the SBI-II cells [59], an
impediment that also limits the versatility of a new Rubisco-deletion strain of the
bacterium Ralstonia eutropha for directed evolution studies [62] due to low (4 × 103
cfu/μg DNA) electroporation efficiency [70].
The availability of a non-photosynthetic rbcL deletion strain of Chlamydomonas
reinhardtii (strain MX3312) has also been exploited for use in a directed evolution
study of Rubisco from this green algae [93]. Like R. capsulatus, heterotrophic
growth of the Rubisco null MX3312 strain can be sustained in the dark on acetate-
containing medium. The practicality of the system to screen mutagenic Rubisco
libraries is again limited by the low chloroplast transformation efficiency of
C. reinhardtii (Fig. 4.2a). Nevertheless, the approach proved successful in identify-
ing Rubisco mutants with improvements in carboxylation efficiency (Table 4.2).
Unfortunately, independent testing showed one mutant comprising three amino acid
mutations was not translatable to influencing the kinetic properties of Rubisco in
tobacco, the model C3 plant primarily used for Rubisco mutagenic testing [25].
4.2.3 Selection Using Rubisco-Dependent E. coli
The high transformation efficiency and fast growth rate of E. coli make it a favoured
host for many directed evolution applications. The value of a high-throughput bac-
terial selection system for thoroughly exploring the adaptive potential of Rubisco
has been a long-standing objective [46]. A potential caveat is that the range of
Rubisco types for experimentation may be limited to the prokaryotic and archaeal
isoforms capable of functional expression in E. coli [50].
Current Rubisco selection systems in E. coli require phosphoribulokinase (PRK)
co-expression. The PRK catalyses the production of RuBP via ATP-dependent
phosphorylation of the ribulose-5-phosphate (R5P) produced in the pentose phos-
phate pathway (Fig. 4.2b). In E. coli RuBP cannot be metabolised and its accumula-
tion is toxic to growth [32]. The cause of this toxicity remains unclear, although it
does not appear due to reductions in R5P availability since deletion of

6-phosphogluconate dehydrogenase in the pentose phosphate pathway needed to
produce R5P has no discernible effect on E. coli growth [35]. Expression of Rubisco
in E.coli-PRK-expressing cells can alleviate RuBP toxicity by conversion into
3-PGA, a natural metabolite of the glycolysis/gluconeogenesis pathway, or 2-PG
that can be converted into glycolate, a usable carbon source for E. coli growth [54]
(Fig. 4.2c). The dependence of E.coli-PRK-expressing cells on Rubisco activity for
survival forms the basis of all Rubisco-dependent E. coli (RDE) selection systems.
The initial application of RDE utilised a cyanobacterial PRK under the control
of an arabinose-inducible BAD promoter in plasmid pACYC184 [52]. When
expressed in wild-type E. coli, this simple RDE screen isolated a small range of
Synechococcus PCC6301 (cyanobacteria) L-subunit mutants that all shared a Met-
259-Thr (M259 T) substitution [52]. The M259T mutation was shown to enhance
L8S8 Rubisco biogenesis ~fivefold in E. coli and modestly improve the overall
kinetic properties of Synechococcus PCC6301 Rubisco [26] (Table 4.2). More
recent success with this RDE system found E49V and D82G mutations in the
Synechococcus PCC6301 S-subunit improved carboxylation efficiency by 50%
[8]. A continuing limitation of these studies is the low fidelity of the RDE system
due to >100:1 ratio of false to bona fide Rubisco mutants produced. This necessi-
tates lower cell plating densities to avoid excess colony formation thereby signifi-
cantly increasing the processing time and cost of analysis (Fig. 4.2b). The foremost
cause of false positives appears to be silencing of PRK synthesis via mutations or
transposon insertion in the prkA gene – thus reverting the cells to a wild-type
growth rate [8, 50, 52].
A higher fidelity RDE with increased throughput was developed by Mueller-
Cajar et al. (2007) that incorporates a shunt comprising PRK and Rubisco where
both enzymes are necessary for carbon metabolism and cell growth (Fig. 4.2c). The
RR1 E. coli strain in this alternative RDE system (called RDE-MM1) contains a
deletion in glyceraldehyde-3-phosphate dehydrogenase (gapA−) that inhibits car-
bon flux through the glycolysis/gluconeogenesis cycle. The PRK-Rubisco shunt
bridges the gapA− metabolic gap to enable carbon flux from hexose carbon sources
in the growth media through to the citric acid cycle for energy and growth ([47],
Fig. 4.2c). An arabinose-inducible BAD promoter is used to provide stringent regu-
lation of PRK expression [37], while Rubisco production is regulated by an IPTG-
inducible lac promoter (Fig. 4.2c). As PRK expression is directly linked to MM1
survival in this RDE system, the production of false positives derived from “PRK
silencing” is largely avoided [50]. The improved fidelity of the MM1-RDE system
has successfully isolated form II R. rubrum L2 Rubisco mutants that unveiled con-
served Sc/o determining amino acids [47], identified novel L-subunit mutations that
improve cyanobacterial L8S8 assembly (“solubility”) and affinity for RuBP [51] and
improved carboxylation efficiencies [14] (Table 4.2). More recent use of the MM1-
RDE in directed evolution of the archaeal Methanococcoides burtonii L10 Rubisco
(MbR) successfully selected two MbR mutants with significant improvements in
kcatC, Sc/o and kcatC/KC21%O2 (Table 4.2) [89].
4.2.4 E
xpressing Molecular Partners of Rubisco Can Alter
the Kinetic Outcome
The majority of attempts to evolve Synechococcus PCC6301 Rubisco in E. coli have

identified L-subunit mutations that improve cellular Rubisco levels (Table 4.2).
These mutations included M259T [52], I174V, Q212L and F345I/L [51] and C172G,
V189I, S398C and I465V (and various substitutions at amino acid F345, [14]). The
mechanism behind the increased production of assembled L8S8 Rubisco is uncer-
tain. It is hypothesised they enhance the interaction of the nascent L-subunit chains
with the E. coli GroEL-GroES chaperonin complex [14, 26] or provide the mono-
mers more time in vivo to form the more stable L2 to L8 complexes to which the
soluble S-subunits can rapidly self-associate and form functional L8S8 complexes. A
recent study highlighted the differential influence E. coli chaperonin (GroEL-
GroES) and the Rubisco-specific assembly chaperone RbcX have on Synechococcus
PCC6301 L-subunit evolution in RDE-MM1 cells [14]. As summarised in Fig. 4.2d,
both ancillary proteins improve rates of cyanobacterial L8S8 biogenesis, the GroEL-
GroES facilitate L-subunit folding [24] and the RbcX stabilise post-chaperonin
folded L-subunits into L2-(RbcX2)2 units that form L8 cores for S-subunit binding
[30]. In RDE-MM1, over-expression of GroEL-GroES had little influence on the
Rubisco mutational range selected despite stimulating cell growth through increased
Rubisco biogenesis. In contrast, selection of some Rubisco mutants was prevented
in RDE-MM1 cells producing RbcX due to apparently constraining the selection of
mutations to solvent exposed surface amino acids. These findings support the notion
that the requirement of form I Rubisco for a range of assembly factors (Fig. 4.1b)
with suitable compatibility [84] has had a pervasive influence on Rubisco evolution,
possibly constraining its potential to enhance catalysis [14, 30, 50, 53].
4.2.5 Limitations with Using Fusion Marker Selection
Efforts to mutate eukaryotic form I Rubisco to facilitate its assembly in E. coli have
yet to yield success [11, 47]. This apparent impasse is most likely due to the com-
plementarity requirements of eukaryotic Rubisco with its multiple assembly factors
for subunit folding, stabilisation and assembly into L8S8 holoenzyme (Fig. 4.1b, [30,
88]). This potential hurdle has however not averted attempts to develop approaches
for identifying solubility enabling eukaryotic L-subunit mutants. In theory this
could be achieved using RDE-MM1 to select for mutations that generate Rubisco
activity. The feasibility of this approach depends on modulating the RDE-MM1
sensitivity to detect very low levels of L8S8 assembly – a possibility still untested.
Instead we have trialled the viability of using reporter protein fusions to screen for
scalable changes in Rubisco solubility. Past success has demonstrated the potential
to detect improved protein assembly (solubility) using reporter fusions that increase
antibiotic resistance [72], enhance fluorescence [39] or alter the amount of dye-
based reporter produced [71, 80]. In the case of Rubisco, our research has shown
L-subunit fusions comprising a truncated N- or C-terminal 85-amino-acid (9 kDa)
beta-galactosidase alpha fragment proved insensitive to reporting variations in
Evolving for solubility using a Rubisco L-subunit fusion
rbcL Marker rbcS Marker rbcL rbcS
NSLAV…
∆Zα folA Ab
(85 aa) (160 aa) (> 250 aa)
Desired
fidelity of
screen white blue TMPsen TMPres Absen Abres
↑Rubisco synthesis ↑Rubisco synthesis ↑Rubisco synthesis
Observed L2 and L8S8 Rubisco Untested
screen fidelity assembly inhibited (feasibility?)
Fig. 4.3 Limitations in using L-subunit fusions for selecting improved Rubisco solubility. Studies
in our laboratory have tested fusion constructs using various selective markers fused in frame to the
N- or C-termini of R. rubrum or Synechococcus PCC6301 L-subunits, the latter comprising a
rbcL-rbcS operon as shown. Fusions with the alpha peptide of E. coli β-galactosidase (Zα) per-
formed poorly as a scalable screen of Rubisco assembly as colonies producing little or no assem-
bled Rubisco were vivid blue. Similarly, fusions with E. coli dihydrofolate reductase (folA)
conferred resistance to trimethoprim (TMP) that did not correlate with the levels of L2 or L8S8
Rubisco produced. The viability of using L-subunit fusions comprising antibiotic (Ab) markers has
not been tested
solubility by blue-white screening (Fig. 4.3). Even Rubisco mutants lacking assem-
bly capabilities produced vivid blue colonies through X-gal conversion making the
fidelity of the screen untenable for visually detecting changes in Rubisco solubility.
The use of a 160-amino-acid (18 kDa) dihydrofolate reductase (DHFR) fusion in
selecting for trimethoprim resistance linked changes in enzyme solubility has been
successful in identifying acetyltransferase mutants with improved solubility [44].
Such an approach proved unsuccessful with R. rubrum L2 and cyanobacterial L8S8
Rubisco as the additional DHFR sequence prevented L-subunit folding and holoen-
zyme assembly (Fig. 4.3). In hindsight this strategy was flawed from inception due
to the GroEL-GroES folding requirements of both Rubisco L-subunits [43] and
DHFR [27] and that the 71 kDa size of the L-DHFR and DHFR-L fusions exceed
the ~60 kDa protein folding limit of the E. coli chaperonin cage [31, 58]. The pro-
hibitive influences fusion reporter proteins have on L-subunit folding and assembly
make this a challenging, if not untenable, approach for detecting changes in Rubisco
solubility in E. coli.
4.3 Evolutionary Outcomes, Applications and Limitations
A major caveat with Rubisco directed evolution studies to date is correlating changes
in catalysis with changes in the growth performance of the host. While the RDE
systems provide the potential for higher throughput of mutagenic library screening,
its reliance on RuBP detoxification to report Rubisco mutants with improved cata-
lytic potential remains inexact.
4.3.1 A Need to Improve RDE Selection
The low levels of false positives selected using the RDE-MM1 system appear linked
to a heightened sensitivity of MM1 to PRK expression [47]. This phenotype may
stem from the diminished viability of the gapA− MM1 strain that requires slow
growth (23 °C) on minimal media for prolonged periods (6–16 days) which dra-
matically impedes experimental throughput [50, 65]. The MM1 strain also suffers
from a low transformation efficiency as growth in LB, or media with a high sugar
content, causes cell lysis [34]. It is also unclear why RDE-MM1 requires CO2 con-
centrations 50–65 times atmospheric levels for Rubisco selection. Possibly the high
CO2 may help maintain the activation status of the Rubisco catalytic sites (Fig. 4.1c)
and/or facilitate permissible rates of RuBP carboxylation in the bacterial cytoplasm
[50]. Somewhat paradoxically even under such high CO2 conditions, there remains
the potential for the selecting Rubisco mutants with increased RuBP oxygenation,
not carboxylation, properties (Table 4.2). It is presumed this can arise as the 2PG
produced can be metabolised into glycolate and used by E. coli for growth [47, 54].
Although increasing oxygenation rates is undesired within the context of improving
photosynthesis, such mutagenic outcomes help to better understand Rubisco
structure-function.
4.3.2 S
election Using a Photosynthetic Host: Is It Really
Advantageous?
An advantage of using a photosynthetic organism in Rubisco directed evolution

applications is the selected enzymes are supported by cognate, or near-cognate,
chaperoning resources of the host. Theoretically this should allow, for example, the
evolution of eukaryote Rubisco in a eukaryote host. As indicated in Sect. 4.2.2,
however, the throughput of such selection systems is relatively slow and costly.
There also appear a number of unresolved ambiguities in some photosynthetic host
studies that likely arise from challenges described in Sect. 4.2.1 associated with
accurately measuring Rubisco content and catalysis. For example, the reductions in
CO2 fixation rate, CO2 affinity and poorer specificity of the F342 V Synechococcus
PCC6301 Rubisco mutant selected in the R. capsulatus SBI-II Rubisco null line are
inconsistent with it benefiting photosynthetic growth [69]. The other F342I and
M259 T mutants selected in the R. capsulatus screen correlate with those repeatedly
selected in RDE studies due to enhanced cyanobacteria L8S8 assembly (Table 4.2).
It therefore seems likely the R. capsulatus screen successfully identified
Synechococcus PCC6301 Rubisco mutants with improved L8S8 biogenesis not ben-
eficial changes in catalysis (Table 4.2). A subsequent directed evolution study using
Chlamydomonas as a selection host identified three L-subunit mutations that
improved Rubisco catalysis in the algae with the same substitutions also able to
enhance tobacco Rubisco catalysis [93]. This finding was not supported in an
independent study where the mutations had no appreciable influence on tobacco

Rubisco kinetics, leaf photosynthesis or plant growth [25]. This brings into question
the versatility of mutagenic studies of L8S8 Rubisco from Chlamydomonas, or pho-
tosynthetic prokaryotes, with regard to providing solutions of direct translatable
benefit to Rubisco in vascular plants.
4.3.3 Translational Success to Improving Photosynthesis
The last few years have seen the successful translational application of RDE-derived
Rubisco mutants to enhance the productivity of a photosynthetic host. Traditionally
the most direct route for improving RDE fitness using Synechococcus PCC6301
Rubisco has been the selection of solubility-enhancing mutants (Sect. 4.2.4). In
contrast, using RDE-MM1 Durão et al. [14] identified point mutations in the
Synechococcus PCC6301 L-subunit that increased carboxylation efficiency (kcatC/
KC21%O2) by either 70% (V189I) or nearly threefold (F140I), the latter mutation not
affecting L8S8 assembly. When transformed into Synechocystis PCC6803 cells, the
F140I mutation improved the rate of photosynthesis by 40% and supported wild-
type growth rates in cells comprising ~20% less Rubisco [14]. While a highly moti-
vating outcome, the underpinning question is why had the mutation not already
been selected during evolution? Possibly the photosynthesis improvements are not
sustainable under alternative, nonoptimal, growth conditions – a hypothesis for
future experimental scrutiny.
A recent directed evolution study has highlighted the potential of improving the
performance of archaeal Methanococcoides burtonii L10 Rubisco (MbR) in a photo-
synthetic role [89]. As indicated in Sect. 4.1.2, form III archaeal Rubisco serves a
non-photosynthetic function, comprises only L-subunits and is highly soluble when
expressed in E. coli [2, 38, 91]. The alternative biological function of archaeal
Rubisco to metabolise RuBP produced as a by-product during nucleotide metabo-
lism [22, 63] has seen it evolve distinctive, somewhat variable, catalytic properties
relative to contemporary Rubisco isoforms [74]. This includes uniquely high affini-
ties for RuBP and atypically low carboxylation properties [2]. These atypical prop-
erties relative to photosynthetic Rubisco suggest archaeal Rubisco has been subject
to alternative evolutionary trajectories, possibly improving its potential for carbox-
ylation enhancement [2, 50]. Consistent with this hypothesis, two MbR mutants
(E138V and K332E) with three- to fourfold improvements in carboxylation effi-
ciency in air (kcatC/KC21%O2) were isolated from a single round of selection in the
RDE-MM1 system ([89]; Table 4.2). When expressed in tobacco chloroplasts, both
mutant MbR variants supported higher rates of photosynthesis and much faster
plant growth compared to control plants producing wild-type MbR. The efficient
expression of archaeal Rubisco in leaves, and overall success with evolving MbR
catalysis, inspires further rounds of evolution to improve catalytic parameters along
evolutionary trajectories that further enhance its photosynthetic potential.
4.3.4 Rubisco Production Can Be Toxic to E. coli
An important consideration of the RDE systems is being able to carefully regulate

PRK expression to ensure the intercellular RuBP production meets the desired
level of Rubisco activity sought. While the BAD promoter has proven suitable to
fine-tuning PRK expression by varying arabinose induction ([47], Fig. 4.2c), little
attention has been made to modulating Rubisco expression. This may require
closer attention when one considers the expression of form I [52] or form II [82]
Rubisco can itself impede, sometimes prevent, E. coli growth. As shown in
Fig. 4.4, the growth of XL-1Blue E. coli cells expressing R. rubrum or Synechococcus
PCC6301 Rubisco is increasingly impeded under rising IPTG induction. In con-
trast, growth is prevented in cells producing the highly soluble M. burtonii L2 or the
F345I mutated Synechococcus PCC6301 L8S8 Rubiscos. Awareness on how
Rubisco toxicity impacts the fidelity of RDE selection remains unexplored. For
example, it may preclude selection of activity-enhancing mutants that additionally
escalate Rubisco biogenesis. Alternatively it may beneficially focus selection of
catalytic enhancements that have little or no influence on Rubisco biogenesis,
somewhat akin to the F140I Synechococcus PCC6301 mutant identified by Durão
et al. (2015) [14].
Form of Rubisco expressed

R. rubrum SePCC6301 SePCC6301F345I M. burtonii
pTrcHisB L2 L8S8 L8S8 L2
(vector
control)
0 nM
IPTG
21 nM
IPTG
2.1 µM
IPTG
Fig. 4.4 High levels of Rubisco expression can inhibit E. coli growth. The differential influence
of Rubisco expression on E. coli growth was determined by spot plating of XL1-blue cells trans-
formed with pTrcHisB plasmids coding the Rubisco forms shown. Shown is the extent of growth
on LB agar containing 200 μg/mL Amp and varying IPTG concentrations (shown) after 5 days in
air at room temperature. Synechococcus PCC6301 (SePCC6301), mutation Phe415Ile (F345I)
improves Rubisco assembly by eight- to 15-fold in E. coli [51]
4.4 ew In Vivo Screening Strategies for Rubisco Directed

N
Evolution
The above examples of possible limitations in the versatility and sensitivity of

RDE-MM1 incite further development of the selection system to increase its
efficiency and broaden the scope of suitable Rubisco clients amenable to directed
evolution study. In this section we explore possible ways to meet these objectives
both in directed evolution studies of Rubisco and its activity-regulating protein
Rubisco activase (Fig. 4.1c).
4.4.1 Avoiding PRK Escape Mutants by Dual Selection
Antibiotic resistance markers are common selective agents used in molecular biol-
ogy to detect and maintain genetically transformed cells. This selective trait can also
be employed to detect recombinant protein production by fusing it to proteins cod-
ing antibiotic resistance. While there appears little merit in using Rubisco-reporter
protein fusions (Sect. 4.2.5), linking an antibiotic resistance onto PRK may avoid
selection of false positives that currently plague RDE systems using wild-type E.
coli (Sect. 4.2.3). These false positives primarily arise from interruption of PRK
expression via integration of varying transposon elements into differing regions of
the prkA gene (Fig. 4.5a–c). By appending an in-frame resistance marker to the
C-terminus of PRK (possibly via a flexible peptide linker), transposon interruption
to PRK synthesis will relinquish antibiotic resistance, preventing cell growth on
selective media (Fig. 4.5d, option 1). Such a strategy requires the PRK, and resis-
tance marker remains active as a PRK-AbR fusion peptide (possibly necessitating a
flexible peptide linker peptide). Notably the Synechococcus PCC6301 PRK typi-
cally used in RDE studies is a homodimer [75]. Therefore, the antibiotic marker
should ideally function as a monomer to avoid steric hindrance and oligomerisation
assembly issues that impede PRK functionality. The feasibility of developing suit-
able PRK-AbR fusion peptides that can improve the fidelity and throughput poten-
tial of the RDE systems using wild-type E. coli remains to be explored.
4.4.2 Tailoring to the Assembly Needs of Eukaryotic Rubisco
Rubisco isoforms from crop plants – especially grain crops of significant nutritional
and agricultural value – pose idyllic targets for catalytic modification by directed
evolution tools. However, their extensive requirement for compatible ancillary pro-
teins during L8S8 biogenesis, incorporating chaperonin folding and downstream
assembling factors (Fig. 4.1b), precludes their functional production in E. coli [30, 47].
As efforts so far have been unable to mutate plant L-subunits to meet the folding and
assembly machinery of E. coli [11, 51], it is likely the inclusion of one or more of
Rubisco’s chloroplast assembly factors may be needed to facilitate Rubisco solubil-
ity in E. coli. Fortunately the last decade has seen dramatic advances in our
understanding of L8S8 Rubisco biogenesis [30, 89]. Unlike the E. coli chaperonin
GroEL cylindrical ring and heptameric GroES cap, the chaperonins of chloroplasts
are structurally more diverse [7, 78, 79]. They comprise heptameric Cpn60 cages
comprising α and β subunits capped by hetero-oligomeric complexes of Cpn10 and
Cpn20 subunits. While the feasibility of expressing chloroplast chaperonins in E.
coli has been shown [7, 78], their ability to fold Rubisco in this context remains
uncertain. Possibly the expression of chloroplast chaperonin subunits in an RDE
may allow directed evolution of eukaryotic Rubisco (Fig. 4.5d). Vascular plant
Rubisco may however require expression of additional assembly factors in the
a XL1Blue-pACPRK b pACPRK transposon mutants

3 days, 25°C
5 11 6 7 1,3,12,13
pACYC184
lac Spinach prk A (993 bp)
2 10 4,9 8
No IPTG 0.1mM IPTG
c Transposon type, insertion location in prkA and orientation

Colony # prkA sequence (5’) transposon
5 (45) AAAGG - TAGACTGGCC… IS2 ... CTACCGGGCT
2 (87) TACTA - GGGGTTTGAG…Tn1000 ... CTCAAACCCC
7 (524) TGATTG - GGGGTTTGAG…Tn1000 ... CTCAAACCCC
1,3,12,13 (568) GGCAAAGTGT
6 (505) AAGCAGCA CTGAGAGATC… IS10 ... GATTCATCAG
11 (434) ACAGTC
4,9 (568) GGCAAAGTGT
8 (736) AACGAGGT CTGATGAATC… IS10 ... GATCTCTCAG
10 (157) GCTTTAGA
d Enhancing the fidelity and screening diversity of the RDE
1 Transposon
PRK Ab inactivation no growth Selective marker
RuBP sensitive
2 3
CPN Rubisco RCA RuBP CO2
Raf1 Improved mutant
Raf2 3PGA growth
Rubisco
BSD2 ATP ADP (2PG)
RbcX ER CO2 ECM (O )
2
inactive Mg2+ active
PRK mutants
rib-5-P RuBP growth
Transposon (false positives)
PRK
PPP inactivation
Glycolysis/gluconeogenesis of PRK
1 Avoid PRK escape mutants

2 Enhance / enable L8S8 biogenesis
3 Improve metabolic maintenance of Rubisco
RDE. Examples include Rubisco accumulation factors 1 and 2 (Raf1, Raf2), RbcX

and/or the bundle sheath defective2 (BSD2) protein (Figs. 4.1c and 4.5d, option 2)
that serve differing, sometimes crucial, roles in L8S8 biogenesis [21, 30, 84].
Engineering an RDE with the required multiplicity of assembly factors to facilitate
higher plant Rubisco (or algae Rubisco) assembly in E. coli or within in vitro recon-
stitution systems is clearly a significant, and risky, challenge but one of extremely
high reward if able to provide a suitable platform for high-throughput mutagenic
study of Rubisco from diverse origins.
4.4.3 I s There a Need for Metabolic Regulation of Rubisco

in an RDE System?
An outcome yet to be considered for improving Rubisco fitness using an in vivo

screen is if the selected mutation(s) influence(s) the extent Rubisco activity is inhib-
ited by pentose phosphate sugars, in particular its own substrate RuBP. As shown in
Fig. 4.1c, formation of inactive “ER” Rubisco arises when RuBP binds to non-
carbamylated catalytic sites. Recent work has highlighted the pervasive influence
ER formation affects Rubisco activity in plant and algal chloroplasts, reducing the
proportion of catalytically active Rubisco sites by 10–70% [68, 92]. In the absence
of exogenous RuBP, the rate of substrate dissociation from ER occurs within a half-
time of 0.5–5 min, depending on Rubisco isoform (Pearce 2006). However, ER
formation is sustained under saturating RuBP (Sharwood 2016). Therefore, to coun-
ter the formation of ER, and other catalytically inactive sugar phosphate inhibited
Rubisco complexes (ECMI, Fig. 4.1c), photosynthetic organisms utilise isoforms of
the Rubisco activase (RCA) enzyme family to reverse this inhibition [5, 9, 48, 50].
While having differences in oligomeric structure and mode of action, all RCA
homologues are part of the AAA+ protein superfamily and use the power of ATP
Fig. 4.5 Improving the function of RDE selection. Expanding the potential of RDE selection that
suffers from (a–c) a high frequency of transposon induced silencing of PRK in wild-type E. coli
by (d) utilising a PRK-antibiotic fusion (option 1), improving Rubisco activity through co-
expression of complimentary biogenesis factors (option 2) and/or Rubisco activase (option 3). (a)
Comparing growth after 3 days at 23 °C of ~2 × 103 XL1-Blue-pACPRK cells on non-inducing (no
IPTG) and spinach PRK-inducing (0.1 mM IPTG) LB media. The large colonies arising under
inducing conditions were due to (b) alleviation of RuBP toxicity by inactivation of spinach PRK
expression via (c) transposon insertion at differing positions in the spinach prkA gene in
pACPRK. Regions of the prkA sequences were repeated upstream (5′, as shown) and downstream
(3′, not shown) at the transposon insertion sites with the position of the first nucleotide relative to
the prkA initiator codon shown in parenthesis. The partial flanking sequence (ten nucleotide) of the
IS2 (white triangle), Tn1000 (white arrow) and IS10 (black arrow) transposon elements are shown
in italics and their relative orientation (ori) indicated by the arrow direction. The selection of these
false positives might be avoided by (d) fusing an in-frame antibiotic selectable marker to PRK. The
fidelity, selection stringency and variety of Rubisco substrates that can be screened by a RDE
might be increased through co-expression of ancillary Rubisco folding/assembly factors and/or a
compatible Rubisco activase (RCA) (described in Sects. 4.4.1, 4.4.2, and 4.4.3)
hydrolysis to facilitate sugar phosphate release from Rubisco (Fig. 4.1c) via poorly
understood transitory interactions [29, 49, 73, 77].
The necessity for RCA to regulate ER (and ECMI) levels in chloroplasts of illu-
minated leaves arises from the saturating steady-state levels of RuBP (>2 mM in
tobacco chloroplasts, [56]) that resemble the concentration of Rubisco catalytic
sites (~2.5–3.5 mM). By analogy, RDE-MM1 cells producing abundant levels of
Rubisco (>5% of the cellular soluble protein) also retain saturating RuBP levels
(0.2–0.9 mM) [47]. Untested is what proportion of the Rubisco pool in the RDE
cells is represented by inactive ER complexes. In vascular plants the reduction or
elimination of RCA results in a high-CO2-requiring phenotype where growth is CO2
assimilation limited due to low levels of activated, catalytically competent, Rubisco
within the stroma [48, 90]. Somewhat akin to this is the extraordinarily high CO2
requirement by all RDE systems for Rubisco activity selection to be undertaken in
air supplemented with >1% (v/v) CO2. As summarised in Fig. 4.5d (option 3), these
observations suggest directed evolution studies using RDE may benefit, possibly
alter, the evolutionary outcomes, through co-expressing a RCA that is complimen-
tary to the mutated client Rubisco under examination.
Potential examples of how ER formation might influence RDE selection arise
from the study of Mueller-Cajar et al. [47]. Here the content and catalytic properties
of the D117V/H and H44Q/N R. rubrum L2 Rubisco mutants did not accord with
increased rates of RuBP fixation or improved substrate affinities. How the L2
mutants improved RDE-MM1 fitness has remained an enigma. Both D117 and H44
are conserved among form II Rubisco and form a hydrogen bond that influences the
positioning of a conserved E48 (E60 in plant Rubisco). During catalysis E48 forms
a salt bridge with the conserved K329 (K334 in plant Rubisco) residue located at the
apex of loop 6 that closes over the catalytic site to initiate catalysis. Mutation of
D117 and H44 is therefore thought to alter E48 positioning and perturb loop 6 clo-
sure [47]. Conceivably, the altered properties to loop 6 closure conferred by the
D117V/H and H44Q/N mutations may alter the capability for inactive ER complex
formation. This hypothesis, as well as testing how RCA co-expression influences
Rubisco mutant selection in an RDE system, poses considerations for future analy-
sis (Fig. 4.5d, option 3).
4.4.4 A Role for RDE-MM1 in Directed Evolution Studies of RCA
The last few years has seen rapid expansion in our appreciation into the molecular
diversity of RCA in nature, in regard to variations in quaternary structure, phylo-
genetic distribution, mechanistic machinery and recognition specificity among
both form I and II Rubisco [29, 48, 49, 73, 77]. Should RCA co-expression benefit
directed evolution studies of Rubisco using RDE (Sect. 4.4.3), it may be possible
to adapt RDE cells to screen mutated rca gene libraries for evolved RCA function.
Already a directed evolution study has successfully identified more thermostable
plant RCA mutants [41]. However, such in vitro activity assay approaches suffer
from low throughput (<7 × 103 mutants) and are highly labour intensive, limita-
tions that theoretically would be averted via an in vivo RDE Rubisco screen.
Potential enzymatic phenotypes to select for are RCA mutations that unveil infor-
mation about residues that enable (or stimulate) recognition and binding with
cognate or heterologous Rubisco and those that improve RCA thermostability.
Both objectives need to consider two key limitations using RDE systems for
Rubisco bio-selection. Firstly, while all forms of RCA can be produced in E. coli
[29, 48, 49, 73, 77], functional Rubisco expression is limited to those sourced
from Archaea and some bacteria but not plants and algae (Sect. 4.1.5). This
limitation might be avertible by using Rubisco chimers comprising regions of
eukaryotic Rubisco subunit sequence transposed into cyanobacterial Rubisco.
A comparable approach using Chlamydomonas Rubisco chimers (produced by
chloroplast transformation) has helped identify how compatibility between amino
acids located in the equatorial region of plant L8S8 (L-subunit residues 89–94) and
those in the short “specificity” helix (H9) of RCA confers the ability of a plant
RCA to distinguish Rubisco from Solanaceae and non-Solanaceae species [55].
A second limitation of an RDE Rubisco screen is whether the temperature can be
modulated to select for improvements in RCA thermostability. Currently RDE
selection requires low temperatures to slow growth and optimise the fidelity of
mutant selection [50], a limitation potentially overcome using a PRK-AbR fusion
(Sect. 4.4.1).
4.5 Summary
The last decade has witnessed a wealth of Rubisco engineering studies aimed at
improving CO2 fixation in crops to increase productivity in response to growing
concerns of global food security [17, 45, 53, 67, 85]. The projections are dire – an
estimated global population of nine billion by 2050 requires we produce more food
in the coming four decades than has been produced in mankind’s entire history [23].
While modern directed evolution tools are beginning to show increasing promise
towards improving Rubisco performance, they need to advance at a faster pace if we
are to reach the improvement level required. Here we propose strategies for improv-
ing the fidelity, throughput and expansion of client Rubisco diversity amenable to
RDE screening using PRK-antibiotic fusions, through co-expression of appropriate
factors to facilitate Rubisco assembly in E. coli and possibly through expression of
the regulatory protein Rubisco activase (Fig. 4.5d). If directed evolution is to play a
role in improving Rubisco to enhance crop photosynthesis and yield potential, then
there is a need to steer away from the continuing focus on high kcatC prokaryotic
Rubisco whose CCM requirements defy what is currently achievable in chloro-
plasts. There is a need to refocus efforts to developing capabilities to screen muta-
genic libraries of more efficient Rubiscos from plants and red algae and fully
evaluate the evolutionary potential for enhancing the carboxylation properties of
non-photosynthetic Rubisco from Archaea.
References
1. Allen JF, de Paula WBM, Puthiyaveetil S, Nield J (2011) A structural phylogenetic map for
chloroplast photosynthesis. Trends Plant Sci 16(12):645–655
2. Alonso H, Blayney MJ, Beck JL, Whitney SM (2009) Substrate-induced assembly of
Methanococcoides burtonii D-ribulose-1, 5-bisphosphate carboxylase/oxygenase dimers into
decamers. J Biol Chem 284(49):33876–33882
3. Amichay D, Levitz R, Gurevitz M (1993) Construction of a Synechocystis PCC6803 mutant
suitable for the study of variant hexadecameric ribulose bisphosphate carboxylase/oxygenase
enzymes. Plant Mol Biol 23:465–476
4. Andersson I, Backlund A (2008) Structure and function of Rubisco. Plant Physiol Biochem
46(3):275–291
5. Andralojc PJ, Madgwick PJ, Tao Y, Keys A, Ward JL, Beale MH, Loveland JE, Jackson PJ,
Willis AC, Gutteridge S, Parry MAJ (2012) 2-Carboxy-D-arabinitol 1-phosphate (CA1P)
phosphatase: evidence for a wider role in plant Rubisco regulation. Biochem J 442:733–742
6. Andrews TJ, Whitney SM (2003) Manipulating ribulose bisphosphate carboxylase/oxygenase
in the chloroplasts of higher plants. Arch Biochem Biophys 414(2):159–169
7. Bai C, Guo P, Zhao Q, Lv Z, Zhang S, Gao F, Gao L, Wang Y, Tian Z, Wang J, Yang F, Liu C
(2015) Protomer roles in chloroplast chaperonin assembly and function. Mol Plant 8(10):
1478–1492
8. Cai Z, Liu G, Zhang J, Li Y (2014) Development of an activity-directed selection system
enabled significant improvement of the carboxylation efficiency of Rubisco. Protein Cell
5(7):552–562
9. Carmo-Silva E, Scales JC, Madgwick PJ, Parry MAJ (2015) Optimizing Rubisco and its regu-
lation for greater resource use efficiency. Plant Cell Environ 38(9):1817–1832
10. Cleland WW, Andrews TJ, Gutteridge S, Hartman FC, Lorimer GH (1998) Mechanism of
Rubisco – the carbamate as general base. Chem Rev 98(2):549–561
11. Cloney LP, Bekkaoui DR, Hemmingsen SM (1993) Co-expression of plastid chaperonin genes
and a synthetic plant Rubisco operon in Escherichia coli. Plant Mol Biol 23(6):1285–1290
12. Cobb RE, Sun N, Zhao H (2013) Directed evolution as a powerful synthetic biology tool.
Methods 60(1):81–90
13. Currin A, Swainston N, Day PJ, Kell DB (2015) Synthetic biology for the directed evolution of
protein biocatalysts: navigating sequence space intelligently. Chem Soc Rev 44(5):1172–1239
14. Durão P, Aigner H, Nagy P, Mueller-Cajar O, Hartl FU, Hayer-Hartl M (2015) Opposing
effects of folding and assembly chaperones on evolvability of Rubisco. Nat Chem Biol
11(2):148–155
15. Eisenhut M, Ruth W, Haimovich M, Bauwe H, Kaplan A, Hagemann M (2008) The photores-
piratory glycolate metabolism is essential for cyanobacteria and might have been conveyed
endosymbiontically to plants. Proc Natl Acad Sci 105(44):17199–17204
16. Ellis RJ (1979) The most abundant protein in the world. Trends Biochem Sci 4(4):241–244
17. Evans JR (2013) Improving photosynthesis. Plant Physiol 162(4):1780–1793
18. Ezaki S, Maeda N, Kishimoto T, Atomi H, Imanaka T (1999) Presence of a structurally novel
type Ribulose-bisphosphate carboxylase/oxygenase in the hyperthermophilic
Archaeon,Pyrococcus kodakaraensis KOD1. J Biol Chem 274(8):5078–5082. doi:10.1074/
jbc.274.8.5078
19. Falcone DL, Tabita FR (1991) Expression of endogenous and foreign ribulose 1,5-bisphosphate
carboxylase-oxygenase (Rubisco) genes in a Rubisco deletion mutant of Rhodobacter sphaer-
oides. J Bacteriol 173(6):2099–2108
20. Farquhar G, von Caemmerer Sv, Berry J (1980) A biochemical model of photosynthetic CO2
assimilation in leaves of C3 species. Planta 149 (1):78–90
21. Feiz L, Williams-Carrier R, Wostrikoff K, Belcher S, Barkan A, Stern DB (2012) Ribulose-
1,5-bis-phosphate carboxylase/oxygenase accumulation factor1 is required for holoenzyme
assembly in maize. Plant Cell 24(8):3435–3446
22. Finn MW, Tabita FR (2004) Modified pathway to synthesize ribulose 1,5-bisphosphate in
methanogenic archaea. J Bacteriol 186(19):6360–6366
23. Furbank RT, Quick WP, Sirault XRR (2015) Improving photosynthesis and yield potential in
cereal crops by targeted genetic manipulation: prospects, progress and challenges. Field Crop
Res 182:19–29
24. Goloubinoff P, Gatenby AA, Lorimer GH (1989) GroE heat-shock proteins promote assembly
of foreign prokaryotic ribulose bisphosphate carboxylase oligomers in Escherichia coli.
Nature 337(6202):44–47
25. Gready J, Kannappan B, Agrawal A, Street K, Stalker DM, Whitney S (2013) Status of options
for improving photosynthetic capacity through promotion of Rubisco performance: Rubisco
natural diversity and re-engineering, and other parts of C3 pathways. Paper presented at the
Proceedings of a workshop held at the Australian National University, Canberra, Australian
Capital Territory, Australia, 2–4 Sept 2009
26. Greene DN, Whitney SM, Matsumura I (2007) Artificially evolved Synechococcus PCC6301
Rubisco variants exhibit improvements in folding and catalytic efficiency. Biochem J 404(3):
517–524
27. Groß M, Robinson CV, Mayhew M, Hartl FU, Radford SE (1996) Significant hydrogen
exchange protection in GroEL-bound DHFR is maintained during iterative rounds of substrate
cycling. Proc Sci 5(12):2506–2513
28. Hartman FC, Harpel MR (1994) Structure, function, regulation, and assembly of D-ribulose-
1,5-bisphosphate carboxylase/oxygenase. Annu Rev Biochem 63:197–234
29. Hasse D, Larsson AM, Andersson I (2015) Structure of Arabidopsis thaliana Rubisco acti-
vase. Acta Crystallogr Sect D: Biol Crystallogr 71(Pt 4):800–808
30. Hauser T, Popilka L, Hartl FU, Hayer-Hartl M (2015) Role of auxiliary proteins in Rubisco
biogenesis and function. Nat Plants 1
31. Hayer-Hartl M, Bracher A, Hartl FU (2016) The GroEL-GroES chaperonin machine: a
nano-cage for protein folding. Trends Biochem Sci 41(1):62–76. doi:10.1016/j.
tibs.2015.07.009
32. Hudson GS, Morell MK, Arvidsson YBC, Andrews TJ (1992) Synthesis of spinach phospho-
ribulokinase and ribulose 1, 5-bisphosphate in Escherichia coli. Aust J Plant Physiol
19:213–221
33. Hwang S-R, Tabita FR (1989) Cloning and expression of the chloroplast-encoded rbcL
and rbcS genes from the marine diatom Cylindrotheca sp. strain N1. Plant Mol Biol 13:
69–79
34. Irani M, Maitra P (1974) Isolation and characterization of Escherichia coli mutants defective
in enzymes of glycolysis. Biochem Biophys Res Commun 56(1):127–133
35. Jiao Z, Baba T, Mori H, Shimizu K (2003) Analysis of metabolic and physiological responses
to gnd knockout in Escherichia coli by using C-13 tracer experiment and enzyme activity
measurement. FEMS Microbiol Lett 220(2):295–301
36. Kanevski I, Maliga P, Rhoades DF, Gutteridge S (1999) Plastome engineering of ribulose-1,5-
bisphosphate carboxylase/oxygenase in tobacco to form a sunflower large subunit and tobacco
small subunit hybrid. Plant Physiol 119(1):133–141
37. Khlebnikov A, Datsenko KA, Skaug T, Wanner BL, Keasling JD (2001) Homogeneous expres-
sion of the PBAD promoter in Escherichia coli by constitutive expression of the low-affinity
high-capacity AraE transporter. Microbiologica 147(12):3241–3247
38. Kitano K, Maeda N, Fukui T, Atomi H, Imanaka T, Miki K (2001) Crystal structure of a novel-
type archaeal rubisco with pentagonal symmetry. Structure 9(6):473–481
39. Klenk C, Ehrenmann J, Schütz M, Plückthun A (2016) A generic selection system for improved
expression and thermostability of G protein-coupled receptors by directed evolution. Sci Rep
6:28133
40. Kreel NE, Tabita FR (2015) Serine 363 of a hydrophobic region of Archaeal Ribulose 1,5-bispho-
sphate carboxylase/oxygenase from Archaeoglobus fulgidus and Thermococcus kodakaraensis
affects CO2/O2 substrate specificity and oxygen sensitivity. PLoS ONE 10(9):e0138351
41. Kurek I, Chang TK, Bertain SM, Madrigal A, Liu L, Lassner MW, Zhu GH (2007) Enhanced
thermostability of Arabidopsis Rubisco activase improves photosynthesis and growth rates
under moderate heat stress. Plant Cell 19(10):3230–3241
42. Leggat W, Whitney S, Yellowlees D (2004) Is coral bleaching due to the instability of the
zooxanthellae dark reactions? Symbiosis 37(1–3):137–153
43. Liu C, Young AL, Starling-Windhof A, Bracher A, Saschenbrecker S, Rao BV, Rao KV,
Berninghausen O, Mielke T, Hartl FU (2010) Coupled chaperone action in folding and assem-
bly of hexadecameric Rubisco. Nature 463(7278):197–202
44. Liu J-W, Boucher Y, Stokes HW, Ollis DL (2006) Improving protein solubility: the use of the
Escherichia coli dihydrofolate reductase gene as a fusion reporter. Protein Expr Purif
47(1):258–263
45. Long Stephen P, Marshall-Colon A, Zhu X-G (2015) Meeting the global food demand of the
future by engineering crop photosynthesis and yield potential. Cell 161(1):56–66
46. Morell MK, Paul K, Kane HJ, Andrews TJ (1992) Rubisco: maladapted or misunderstood?
Aust J Bot 40:431–441
47. Mueller-Cajar O, Morell M, Whitney SM (2007) Directed evolution of Rubisco in Escherichia
coli reveals a specificity-determining hydrogen bond in the form II enzyme. Biochemist
46(49):14067–14074
48. Mueller-Cajar O, Stotz M, Bracher A (2014) Maintaining photosynthetic CO2 fixation via
protein remodelling: the Rubisco activases. Photosynth Res 119(1–2):191–201
49. Mueller-Cajar O, Stotz M, Wendler P, Hartl FU, Bracher A, Hayer-Hartl M (2011) Structure
and function of the AAA+ protein CbbX, a red-type Rubisco activase. Nature 479(7372):
194–199
50. Mueller-cajar O, Whitney SM (2008) Directing the evolution of Rubisco and Rubisco activase:
first impressions of a new tool for photosynthesis research. Photosynth Res 98(1–3):667–675
51. Mueller-Cajar O, Whitney SM (2008) Evolving improved Synechococcus Rubisco functional
expression in Escherichia coli. Biochem J 414(2):205–214
52. Parikh MR, Greene DN, Woods KK, Matsumura I (2006) Directed evolution of Rubisco
hypermorphs through genetic selection in engineered E. coli. Protein Eng Des Sel 19(3):
113–119
53. Parry MAJ, Andralojc PJ, Scales JC, Salvucci ME, Carmo-Silva AE, Alonso H, Whitney SM
(2013) Rubisco activity and regulation as targets for crop improvement. J Exp Bot
64(3):717–730
54. Pellicer MT, Nunez MF, Aguilar J, Badia J, Baldoma L (2003) Role of 2-phosphoglycolate
phosphatase of Escherichia coli in metabolism of the 2-phosphoglycolate formed in DNA
repair. J Bacteriol 185(19):5815–5821
55. Portis AR, Li CS, Wang DF, Salvucci ME (2008) Regulation of Rubisco activase and its inter-
action with Rubisco. J Exp Bot 59(7):1597–1604
56. Price GD, Evans JR, von Caemmerer S, Yu J-W, Badger MR (1995) Specific reduction of
chloroplast glyceraldehyde-3-phosphate dehydrogenase activity by antisense RNA reduces
CO2 assimilation via a reduction in ribulose bisphosphate regeneration in transgenic tobacco
plants. Planta 195:369–378
57. Price GD, Howitt SM (2014) Plant science: towards turbocharged photosynthesis. Nature
513(7519):497–498
58. Sakikawa C, Taguchi H, Makino Y, Yoshida M (1999) On the maximum size of proteins to stay
and fold in the cavity of GroEL underneath GroES. J Biol Chem 274(30):21251–21256
59. Satagopan S, Chan S, Perry LJ, Tabita FR (2014) Structure-function studies with the unique
hexameric Form II ribulose-1, 5-bisphosphate carboxylase/oxygenase (Rubisco) from
Rhodopseudomonas palustris. J Biol Chem 289(31):21433–21450
60. Satagopan S, Scott SS, Smith TG, Tabita FR (2009) A Rubisco mutant that confers growth
under a normally “inhibitory” oxygen concentration. Biochemist 48(38):9076–9083
61. Satagopan S, Spreitzer RJ (2004) Substitutions at the Asp-473 Latch Residue of

Chlamydomonas ribulosebisphosphate carboxylase/oxygenase cause decreases in carboxyl-
ation efficiency and CO2/O2 specificity. J Biol Chem 279(14):14240–14244
62. Satagopan S, Tabita FR (2016) RubisCO selection using the vigorously aerobic and metaboli-
cally versatile bacterium Ralstonia eutropha. FEBS J 283:2869–2880
63. Sato T, Atomi H, Imanaka T (2007) Archaeal type III RuBisCOs function in a pathway for
AMP metabolism. Science 315(5814):1003–1006
64. Savir Y, Noor E, Milo R, Tlusty T (2010) Cross-species analysis traces adaptation of Rubisco
toward optimality in a low-dimensional landscape. Proc Natl Acad Sci 107(8):3475–3480
65. Seta FD, Boschi-Muller S, Vignais M, Branlant G (1997) Characterization of Escherichia coli
strains with gapA and gapB genes deleted. J Bacteriol 179(16):5218–5221
66. Sharwood R, von Caemmerer S, Maliga P, Whitney S (2008) The catalytic properties of hybrid
Rubisco comprising tobacco small and sunflower large subunits mirror the kinetically equiva-
lent source Rubiscos and can support tobacco growth. Plant Physiol 146:83–96
67. Sharwood RE, Ghannoum O, Whitney SM (2016) Prospects for improving CO2 fixation in
C3-crops through understanding C4-Rubisco biogenesis and catalytic diversity. Curr Opin Plant
Biol 31:135–142
68. Sharwood RE, Sonawane BV, Ghannoum O, Whitney SM (2016) Improved analysis of C4 and
C3 photosynthesis via refined in vitro assays of their carbon fixation biochemistry. J Exp Bot
67(10):3137–3148
69. Smith SA, Tabita FR (2003) Positive and negative selection of mutant forms of prokaryotic (cya-
nobacterial) ribulose-1, 5-bisphosphate carboxylase/oxygenase. J Mol Biol 331(3):557–569
70. Solaiman DK, Swingle BM, Ashby RD (2010) A new shuttle vector for gene expression in
biopolymer-producing Ralstonia eutropha. J Microbiol Methods 82(2):120–123
71. Soo VW, Hanson-Manful P, Patrick WM (2011) Artificial gene amplification reveals an abun-
dance of promiscuous resistance determinants in Escherichia coli. Proc Natl Acad Sci
108(4):1484–1489
72. Stemmer WPC (1994) DNA shuffling by random fragmentation and reassembly: in vitro
recombination for molecular evolution. Proc Natl Acad Sci 91:10747–10751
73. Stotz M, Mueller-Cajar O, Ciniawsky S, Wendler P, Hartl FU, Bracher A, Hayer-Hartl M
(2011) Structure of green-type Rubisco activase from tobacco. Nat Struct Mol Biol
18(12):1366–1370
74. Tabita FR, Hanson TE, Li H, Satagopan S, Singh J, Chan S (2007) Function, structure, and
evolution of the Rubisco-like proteins and their Rubisco homologs. Microbiol Mol Biol Rev
71(4):576–599
75. Tamoi M, Miyazaki T, Fukamizo T, Shigeoka S (2005) The Calvin cycle in cyanobacteria is
regulated by CP12 via the NAD(H)/NADP(H) ratio under light/dark conditions. Plant J 42(4):
504–513
76. Tcherkez GGB, Farquhar GD, Andrews TJ (2006) Despite slow catalysis and confused sub-
strate specificity, all ribulose bisphosphate carboxylases may be nearly perfectly optimized.
Proc Natl Acad Sci 103:7246–7251
77. Tsai YC, Lapina MC, Bhushan S, Mueller-Cajar O (2015) Identification and characterization
of multiple rubisco activases in chemoautotrophic bacteria. Nat Commun 6:8883. doi:10.1038/
ncomms9883
78. Tsai YC, Mueller-Cajar O, Saschenbrecker S, Hartl FU, Hayer-Hartl M (2012) Chaperonin
cofactors, Cpn10 and Cpn20, of green algae and plants function as hetero-oligomeric ring
complexes. J Biol Chem 287(24):20471–20481
79. Vitlin Gruber A, Nisemblat S, Azem A, Weiss C (2013) The complexity of chloroplast chap-
eronins. Trends Plant Sci 18(12):688–694
80. Wang HH, Isaacs FJ, Carr PA, Sun ZZ, Xu G, Forest CR, Church GM (2009) Programming
cells by multiplex genome engineering and accelerated evolution. Nature 460(7257):894–898
81. Whitney SM, Andrews TJ (1998) The CO2/O2 specificity of single-subunit ribulose-

bisphosphate carboxylase from the dinoflagellate, Amphidinium carterae. Aust J Plant Physiol
25(2):131–138
82. Whitney SM, Andrews TJ (2001) Plastome-encoded bacterial ribulose-1, 5-bisphosphate car-
boxylase/oxygenase (RubisCO) supports photosynthesis and growth in tobacco. Proc Natl
Acad Sci 98(25):14738–14743
83. Whitney SM, Baldet P, Hudson GS, Andrews TJ (2001) Form I Rubiscos from non-green algae
are expressed abundantly but not assembled in tobacco chloroplasts. Plant J 26(5):535–547
84. Whitney SM, Birch R, Kelso C, Beck JL, Kapralov MV (2015) Improving recombinant
Rubisco biogenesis, plant photosynthesis and growth by coexpressing its ancillary RAF1
chaperone. Proc Natl Acad Sci 112(11):3564–3569
85. Whitney SM, Houtz RL, Alonso H (2011) Advancing our understanding and capacity to engi-
neer nature’s CO2-sequestering enzyme, Rubisco. Plant Physiol 155(1):27–35
86. Whitney SM, Sharwood RE (2008) Construction of a tobacco master line to improve Rubisco
engineering in chloroplasts. J Exp Bot 59(7):1909–1921
87. Whitney SM, Sharwood RE (2014) Plastid transformation for Rubisco engineering and proto-
cols for assessing expression. Methods Mol Biol 1132:245–262
88. Wilson R, Whitney S (2015) Photosynthesis: getting it together for CO2 fixation. Nat Plants
1:15130
89. Wilson RH, Alonso H, Whitney SM (2016) Evolving Methanococcoides burtonii archaeal
Rubisco for improved photosynthesis and plant growth. Sci Report 6:22284
90. Yamori W, von Caemmerer S (2009) Effect of Rubisco activase deficiency on the temperature
response of CO2 assimilation rate and rubisco activation state: insights from transgenic tobacco
with reduced amounts of Rubisco activase. Plant Physiol 151(4):2073–2082
91. Yoshida S, Atomi H, Imanaka T (2007) Engineering of a type III rubisco from a hyperthermo-
philic archaeon in order to enhance catalytic performance in mesophilic host cells. Appl
Environ Microbiol 73(19):6254–6261
92. Young JN, Heureux AMC, Sharwood RE, Rickaby REM, Morel FMM, Whitney SM (2016)
Large variation in the Rubisco kinetics of diatoms reveals diversity among their carbon-
concentrating mechanisms. J Exp Bot 67(11):3445–3456
93. Zhu X-G, Kurek I, Liu L (2010) Engineering photosynthetic enzymes involved in CO2-
assimilation by gene shuffling. In: Rebeiz C, Benning C, Bohnert H et al (eds) Advances in
photosynthesis and respiration, The chloroplast, vol 31. Springer, Dordrecht, pp 307–322
Directed Evolution of Unspecific
Peroxygenase 5
Patricia Molina-Espeja, Patricia Gómez de Santos,
and Miguel Alcalde
Abstract
Unspecific peroxygenase (UPO) is a heme-thiolate peroxidase with mono(per)
oxygenase activity for the selective oxyfunctionalization of C-H bonds. Fueled
by catalytic concentrations of H2O2, which acts as both oxygen donor and as final
electron acceptor, this stable, soluble, and extracellular enzyme is a potential
biocatalyst for dozens of transformations that are of considerable interest in
organic synthesis. In this chapter we describe the main attributes of this versatile
enzyme while reflecting on the directed evolution campaigns recently followed
in our laboratory that set out to enhance the functional expression of UPO in
yeast and improve the activity, as well as approximating its properties to the
required industrial standards.
5.1 Introduction
At the beginning of the twenty-first century, the first true natural peroxygenase was
isolated from the basidiomycete Agrocybe aegerita (AaeUPO1) [54], an edible
mushroom that produces white rot [24]. After its initial misclassification as an
unusual alkaline lignin peroxidase, later as an haloperoxidase and often as an aro-
matic peroxygenase (APO), it was finally recognized as unspecific peroxygenase
(UPO) and considered as the first member of a new sub-subclass of oxidoreductases
(EC 1.11.2.1) (Table 5.1). What distinguished UPO in this sense is its ability to
insert oxygen into unactivated carbon atoms (both in aliphatic and aromatic com-
pounds) with high regio- and enantioselectivity, as well as its broad substrate
P. Molina-Espeja • P.G. de Santos • M. Alcalde (*)

Department of Biocatalysis, Institute of Catalysis, Consejo Superior de Investigaciones
Científicas (CSIC), 28049 Madrid, Spain

DOI 10.1007/978-3-319-50413-1_5
128 P. Molina-Espeja et al.
Table 5.1 EC-IUBMB 1.11 classification
specificity and the sole catalytic requirement for hydrogen peroxide, acting as both
final electron acceptor and as oxygen donor (as demonstrated in studies using
H218O2) [3, 27, 29, 30, 33, 55]. Previously, the only enzymes that could efficiently
perform such oxyfunctionalizations were the cytochrome P450 monooxygenases
(EC 1.14), albeit with an inconvenience that they rely on redox cofactors (NAD(P)
H) and auxiliary flavoproteins (Fig. 5.1). Moreover, P450s are usually associated to
membranes, while UPOs are soluble and extracellular. As such, from a catalytic
point of view, UPO is considered as the “missing link” between P450s and the chlo-
roperoxidase of Caldariomyces fumago (CPO, EC 1.11.1.10), being all of them
heme-thiolate enzymes (i.e., with a cysteine residue as the axial ligand of the heme
group). In this regard, strong efforts have been made to exploit the peroxide shunt
pathway of P450s, through which, and like UPO, the enzyme works only supplied
by H2O2 [13, 26]. However, the poor efficiency of this route, even for ad hoc evolved
P450s, and the poor stability in the presence of H2O2 are two major hurdles that are
not easily overcome [17, 51, 53]. CPO does naturally use H2O2 as a co-oxidant and
oxygen donor, and hence, both CPO and UPO are placed within the group of heme-
thiolate peroxidases. Nevertheless, when the traits of UPO and CPO are inspected
5 Directed Evolution of Unspecific Peroxygenase 129
Fig. 5.1 Comparison between P450s and UPO. P450s (left, red) are intermembrane proteins with
poor stability, which depend on auxiliary proteins (flavoproteins, pink hexagon) and redox cofac-
tors as an electron source and that are partially deviated to the formation of unproductive oxygen
species that reduces the reaction yield. By contrast, UPO (right, blue) is soluble and stable and
only needs H2O2, overcoming the aforementioned disadvantages of P450s
in more detail, the substrate spectrum of the latter is much more restricted as it is
incapable of oxygenating carbon atoms of alicyclic/aromatic rings or n-alkanes,
which are very inert in chemical terms [20, 22].
AaeUPO1 has a relatively novel sequence, with little homology to P450s and
sharing only 27% identity with CPO up to 35% in its N-terminal domain [44, 54].
In terms of other UPOs, more than 2500 putative UPO sequences from different
fungi have been detected in genomic databases [21, 23]. Moreover, two additional
UPOs have hitherto been identified and characterized, produced by Coprinus
(Coprinellus) radians (CraUPO) [2] and Marasmius rotula (MroUPO) [18, 49].
Other fungi are also considered to be potential producers of UPOs, although they
are still to be fully characterized [24]. In addition, Novozymes recently expressed a
putative UPO gene from Coprinopsis cinerea (rCciUPO) in Aspergillus oryzae [4],
as well as an UPO from an undetermined mold (rNOVO) [46]. Given their wide-
spread distribution in fungi, UPOs have been sorted into two structural groups: short
and long UPOs. Short UPOs (like MroUPO) are found throughout the world of
fungi, with molecular weights of ~26 kDa and a histidine residue as a charge stabi-
lizer at the active site. By contrast, long UPOs (like AaeUPO1 or CraUPO) are
found in basidiomycetes and ascomycetes, with molecular weights of ~44 kDa, an
internal disulfide bridge and an arginine residue as a charge stabilizer [24, 44].
The role of UPO in its natural context remains uncertain, although several activi-
ties have been proposed. This enzyme might be involved in the synthesis of different
metabolites, like antibiotics, as well as in detoxification processes or in the later
stages of lignin and humus degradation [20, 24, 31]. Therefore, UPO can be included
within the ligninolytic enzyme consortium involved in natural wood decay, along
with high-redox potential peroxidases (versatile, lignin, and manganese peroxi-
dases), laccases, and H2O2-supplying enzymes like aryl-alcohol oxidase.
5.2 Biochemical and Structural Features
To date, 16 UPO sequences have been detected in the A. aegerita genome, although
as yet only AaeUPO1 has been characterized in depth [24, 44, 54]. With up to 22%
glycosylation, this 46 kDa extracellular protein has different isoelectric points (pI)
that range from 4.9 to 5.7. Its spectroscopic characteristics are determined by a peak
at 420 nm (a Soret band representative of heme-containing proteins) and two max-
ima at 572 and 540 nm (CT1 and CT2 charge transfer bands, respectively) [54]. The
crystallographic structure of AaeUPO1 was recently published at ~2 Å resolution
[47], comprising ten α-helices and five ß-sheets, the latter formed by very few resi-
dues (Fig. 5.2). UPO possesses a heme group at the active site (protoporphyrin IX)
whose iron is hexacoordinated showing the fifth position associated with a cysteine
sulfur (Cys36) that acts as a proximal (axial) ligand, while the sixth position is asso-
ciated to a water molecule (distal ligand) (Fig. 5.2a). The axial ligand in CPO and
P450s is also a cysteine, unlike other heme-containing proteins such as classic per-
oxidases (His, His+Asp, Tyr, Tyr+Arg are the other possibilities) [50], and it is
assumed that this axial Cys is responsible for the oxygen transfer reaction. The
C-terminal is stabilized by a disulfide bridge formed between cysteines 278 and
319. UPO possesses a region for halide binding, close to the entrance to the catalytic
pocket, and a structural Mg2+ ion (Fig. 5.2b). The entrance of the substrate to the
catalytic pocket proceeds along a highly hydrophobic funnel-shaped channel that is
made up of ten aromatic residues (nine Phe and one Tyr), these leading the substrate
to the active center. Of these aromatic residues, five are of particular interest: three
Phe residues that orient the substrate (Phe69, Phe121, and Phe199) and two more
that act by delimiting its entrance to the channel (Phe76 and Phe191; Fig. 5.2c, d)
[47]. Finally, the Glu196 and Arg189 of UPO act as an acid-base pair involved in
catalysis.
UPO can catalyze peroxygenation reactions (i.e., insertion of an oxygen atom
through two-electron oxidation, peroxygenase, or mono(per)oxygenase activity) or
peroxidation reactions (i.e., abstraction of an electron, peroxidase, or peroxidative
activity), and it also possesses certain catalase activity. As selective C-H oxyfunc-
tionalizations are among the most solicited reactions in organic synthesis, the per-
oxygenase activity of UPO has been studied in depth, with 300 positive substrates
already tested (a number that continues to grow). Accordingly, the peroxygenase
reactions catalyzed by UPO include aromatic, alkylic, and aliphatic hydroxylations;
aromatic and aliphatic olefin epoxidations; ether cleavage; N-dealkylations; sulfoxi-
dations; N-oxidations; and brominations (Fig. 5.3). Indeed, the exploitation of such
peroxygenase activity is generating great industrial interest. One of the areas in
which UPO can be extremely useful is in the pharmaceutical sector, and indeed, the
ability of this enzyme to produce drugs and human drug metabolites (HDMs) has
been confirmed [5, 28, 35, 48]. While human P450s are known to metabolize drugs
into complex HDMs in the liver, for which it is mandatory to perform pharmacoki-
netic and pharmacodynamic studies, its chemical synthesis is complicated. However,
engineered UPOs could be used to produce HDMs with high selectivity in a more
economic and efficient manner. Other interesting applications include the synthesis
b c
d e
Fig. 5.2 AaeUPO1 crystal structure (PDB: 2YOR). (a) General view. The disulfide bridge and the
axial Cys36 are marked in green, the catalytic Phe in pink (delimiting the entrance for substrates
Phe76 and Phe191 (dark pink); orienting the substrate to the heme Phe69, Phe121, and Phe199
(light pink)), the structural Mg2+ in light pink, the acid-base pair (Glu196-Arg189) is represented
in light blue, the FeIII of the heme is in red, and the protoporphyrin IX of the heme is in CPK color-
ing. (b) Catalytic pocket of AaeUPO1 with detail of the heme environment. (c) Coordination of the
Mg2+. (d, e) Frontal view of the entrance to the heme channel with and without the surface,
respectively
Fig. 5.3 Activities of UPO
of limonene, cyclohexanone, and hydroxylated fatty acids (fatty alcohols) to be

used in cosmetics, food, solvents, polymers, or other materials (e.g., nylon) [19, 45,
46]. In this sense, the synthesis of 2,5-furandicarboxylic acid (FDCA), a renewable
building block of particular interest for the production of biopolymers to replace
traditional polyesters obtained from crude oil, has recently been achieved from
5-hydroxymethylfurfural through a cascade reaction that combined aryl-alcohol
oxidase and UPO [12]. Moreover, the potential use of UPO for the functionalization
of aliphatic alkanes into alcohols should not be underestimated, since this remains
a clear challenge in contemporary chemistry. Besides, the functionalization of aro-
matic hydrocarbons is very relevant in obtaining added value products, for example,
that of benzene to phenol [27] or of naphthalene to 1-naphthol [33]. Another practi-
cal example is the transformation of aromatic and aliphatic olefins into their corre-
sponding epoxides, important intermediates in chemical synthesis due to their great
versatility (e.g., to produce cosmetics, surfactants, or agrochemicals).
Although native UPO has been studied in the synthesis of the aforementioned
compounds, engineering will be essential to improve its performance so that it may
fully compete with chemical catalysts. Thus, one of the main goals for the design of
this enzyme should include the modeling of its selectivity (e.g., for the terminal
hydroxylation of alkanes or fatty acids), increasing its total turnover numbers in any
given process, a reduction of its peroxidase activity as a side effect that can lead to
unwanted by-products, and enhancement of its oxidative and operational stability or
of its activity in the presence of organic cosolvents. To achieve such goals, it is first
necessary to heterologously and functionally express the enzyme in a host suited to
the tailoring of its properties by directed evolution.
5.3 Directed UPO Evolution
5.3.1 The Kickoff: Evolution Toward Functional Expression
One particular challenge in the directed evolution of ligninolytic enzymes (also

known as ligninases) is to achieve their functional expression in the standard heter-
ologous hosts used. All attempts to achieve functional expression of ligninases in
Escherichia coli have ended up in the formation of inclusion bodies, and although
they can be refolded in vitro for structure-function relationship studies, this solution
is not practical given the high-throughput context in which directed evolution exper-
iments take place. Differences in codon usage, missing chaperones, and the lack of
suitable machinery to induce complex posttranslational modifications (e.g., glyco-
sylation, processing of the N- and/or C-terminal ends) preclude the use of this bac-
terial host in the directed evolution of ligninases. Indeed, all attempts to express
UPO in bacteria carried out to date have failed, not even achieving in vitro refolding
from inclusion bodies (Prof. A.T. Martinez, CIB-CSIC, personal communication).
Conversely, the baker’s yeast Saccharomyces cerevisiae is a more appropriate
host to express eukaryotic ligninases as its physiology is much closer to that of its
native fungi. As such, S. cerevisiae has been used for the directed evolution of the
versatile peroxidase, medium- and high-redox potential laccases, aryl-alcohol oxi-
dase, and now UPO ([1] and references herein). Among the main advantages offered
by S. cerevisiae is its high transformation efficiency (yielding individual colonies
with 107–108 transformants/μg DNA), as well as a broad portfolio of episomal uni-
and bidirectional vectors, an efficient secretory system that directs the exocytosis of
proteins into the culture medium (bypassing the lysis steps commonly used in E.
coli), and a high frequency of homologous DNA recombination that incorporates
proofreading to avoid the introduction of unwanted mutations. This latter feature
allows us to employ a palette of library creation methods in directed evolution
experiments.
Even considering all these advantages, the basal functional expression of ligni-
nases in yeast is weak, and UPO is not an exception (in the mU/L range). This is
mostly due to differences in the proteases and/or the signal peptidases along the
secretory route, coupled to the longer residence time in the Golgi apparatus. This
Golgi retention is frequently associated to enhanced glycosylation that slows down

its trafficking, often making the foreign protein toxic to the host. Therefore, we must
first apply directed evolution as a “molecular purge” to foment the secretion of lig-
ninases in yeast and, only subsequently, to drive them toward more specific
challenges.
The main strategies and tools used for the directed evolution of UPO in S. cere-
visiae include (i) the design of dual screening assays for both peroxygenase and
peroxidase activities, (ii) the optimization of microculture conditions, (iii) signal
peptide switching, and (iv) combined mutagenesis and DNA recombination for
standard and focused evolution. Given the success achieved in this case study (see
below), it is likely that a similar approach could now be adopted for CPO, the other
heme-thiolate peroxidase discovered in the 1960s that has resisted the efforts of
protein engineers to enhance its functional expression for decades.
When one is trying to drive evolution for secretion, the main objective is to main-
tain or even improve the different activities of the targeted enzyme while accumulat-
ing mutations that enhance secretion, without jeopardizing stability. To avoid the
enzyme becoming too specific for a given substrate in the course of evolution for
secretion, it is advisable to simultaneously screen with different compounds. In our
efforts to direct the evolution of UPO, we used a dual screening assay for peroxy-
genase and peroxidase activities, while the kinetic thermostability was also mea-
sured during rescreening to rule out potential destabilizing mutations. Thanks to this
approach, the final secretion mutant of the evolutionary pathway turned out to have
equivalent kinetic parameters, stability, and spectroscopic features to the wild-type
UPO (see below).
Since high-throughput culture conditions (i.e., the mutant libraries are grown and
screened in 96 well plates) are far from optimal (in terms of stirring, oxygen avail-
ability, etc.), it is crucial to adjust some factors to ensure the best point of departure
for directed UPO evolution, such as strain selection, incubation times, medium (i.e.,
heme supply and exogenous source of magnesium), and temperature [39].
It is also worth testing different signal peptides that might favor the trafficking of
the foreign polypeptide in yeast. Indeed, in addition to the native signal peptide of
UPO (n), a native prepro-leader of the α-factor from S. cerevisiae (α) [7] and a
mutant of the latter that improves secretion (α*) [37] were attached to the native
mature UPO in this study. The α-factor prepro-leader is commonly used for heter-
ologous expression in yeast, and in particular, we have used it for the successful
engineering of different laccases, versatile peroxidases, and aryl-alcohol oxidases
[11, 14, 37, 56]. Surprisingly, the best result was obtained with the native UPO sig-
nal peptide (n-UPO), which produced 149 mU/L, followed by the α-UPO (74 mU/L)
and the α*-UPO (negligible). It is possible that the evolved α* was not so effective
when attached to mature UPO because the beneficial mutations were selected when
it was linked to a fungal laccase [37], adjusting both polypeptides ad hoc to the
requirements of the yeast’s secretory pathway (i.e., the evolved prepro-leader and
the laccase). Hence, this evolved α* leader cannot be considered as a universal pep-
tide for the expression of other ligninases in yeast. Nevertheless, the α* leader does
work when it is attached to laccases from several different sources and with distinct
redox potentials (unpublished material), which suggests a strong connection must
exist between the evolved signal peptide and the group of enzymes to which it is
attached when pursuing improved secretion.
We very recently constructed chimeric prepro-leaders that combine regions of
the α-factor and the K1 killer toxin prepro-leaders and that could be suitable starting
points to enhance the secretion of other UPOs [56]. Introducing variability into the
signal leader could be also an effective strategy to augment UPO expression, and we
have approached this by using a homemade library creation method called
MORPHING (mutagenic organized recombination process for homologous in vivo
grouping) [16]. This simple tool, supported by the high frequency of homologous
DNA recombination of S. cerevisiae, allows us to introduce random mutations and
recombination events in stretches as small as ~20 amino acids while keeping the rest
of the sequence intact. The mutational loads can be adjusted for each specific region
according to its length, such that in vivo splicing of the different segments to the
linearized plasmid in just a single transformation step occurs by using overhangs of
~40 bp that flank each PCR product. We have taken advantage of MORPHING to
engineer versatile peroxidases [16], aryl-alcohol oxidase [56], rubisco (unpublished
material), and UPO for different biotechnological needs, highlighting the usefulness
of this method for focused evolution.
MORPHING at the UPO signal peptide yielded three almost consecutive and
independent mutations (discovered in single mutants) that significantly enhanced
secretion [16]. Those changes (F12Y, A14V, and R15G) lie in the hydrophobic core
of this sequence, and they were combined with a previously identified A21D muta-
tion [39]. In conjunction, these changes diminish the hydrophobicity of the region,
and they may favor interplay between the signal peptide and the signal recognition
particle (SRP) in the endoplasmic reticulum pathway. This evolved signal peptide
was appropriate to obtain a reliable breakdown in the properties of the final variant.
Thus, it was attached to the native UPO in order to achieve sufficient expression in
yeast to compare its kinetics and secretion with the evolved UPO counterpart. In
fact, we cannot exclude the use of this evolved leader for future efforts where new
UPO genes from different sources must be functionally expressed.
Besides focusing evolution on the signal leader, in our experience it is important
to combine random mutagenesis and DNA recombination to the whole DNA
sequence so that beneficial mutations introduced into independent clones can be
easily incorporated into the same gene scaffold in a drive toward total activity
improvement (TAI, the product of specific activity and secretion). Accordingly, in
five rounds of evolution, we mixed in vivo shuffling and error-prone PCR (with
distinct mutational loads), as well as we used DNA polymerases with different
mutational bias and thereafter performing in vivo assembly of mutant libraries. As
a representative example of the significance of this approach, in the second round of
evolution, we took advantage of the distances between the beneficial mutations
found in the improved mutants from the first generation (1A11, L67F; 3C2, I248V-
F311L) to favor homologous recombination in yeast (Fig. 5.4). In this way, clones
Fig. 5.4 Evolutionary tree of AaeUPO1 toward functional expression and 1-naphthol production.
The new mutations that appeared in every generation (G) are underlined: TAI total activity
improvement in fold for ABTS, NAI naphthalene activity improvement, NDR peroxygenase activ-
ity/peroxidase activity ratio
could be selected from this new round that incorporated the mutational backbone
formed by L67F-I248V-F311, as well as new mutations introduced into either the
leader sequence or the mature protein.
The final UPO variant in this evolutionary route toward secretion, PaDa-I, accu-
mulated four mutations in the signal peptide and five more in the mature protein.
PaDa-I had similar biochemical properties to the wild-type UPO in terms of
N-terminal processing, isoelectric point, pH activity profiles, stability, and spectro-
scopic features. Its degree of glycosylation was slightly enhanced (from 22% in
wild-type UPO to 30% in PaDa-I), possibly due to the tendency of S. cerevisiae to
hyperglycosylate foreign proteins. In terms of the catalytic constants with the sub-
strates tested (ABTS, NBD, veratryl alcohol, benzyl alcohol, and H2O2), both
enzymes exhibited similar values, although the overall secretion of PaDa-I by
S. cerevisiae increased 1114-fold to ~8 mg/L. When we transferred this mutant to
the methylotrophic yeast Pichia pastoris for overproduction in a bioreactor, these
values were enhanced to 217 mg/L, with the variant conserving all the evolved
properties. It is worth noting that P. pastoris is not a suitable host for directed evolu-
tion, not least because of the lack of reliable episomal vectors and the poor transfor-
mation efficiencies. After some process engineering, we achieved ~1 g/L UPO from
P. pastoris in a bioreactor (unpublished material), which comes closer to the appli-
cation of this tandem-yeast expression system in practical industrial cases (S. cere-
visiae for directed evolution and P. pastoris for large-scale fermentation) [40].
The use of an enzymatic cascade for the atom-efficient in situ generation of H2O2
was recently described with PaDa-I. In this approach, methanol works as a sacrifi-
cial electron acceptor in a cascade reaction with alcohol oxidase and formaldehyde
dismutase to yield three equivalents of H2O2 for the complete oxidation of ethylben-
zene by the UPO mutant. The total turnover numbers achieved in this system were
as high as ~300,000 mol product mol biocatalyst−1, emphasizing the potential appli-
cation of this type of cascade to minimize UPO oxidative inactivation [42].
The PaDa-I crystal structure has just been resolved at a resolution of 1.5 Å (in
collaboration with Prof. Julia Sanz, IQFR-CSIC, Madrid: unpublished data) high-
lighting some changes, including (i) a complete N-terminal in which the first three
residues are lacking in the wild-type UPO crystal structure as a consequence of their
potential proteolytic cleavage in A. aegerita EPG/LPP [47] and (ii) the F311L muta-
tion that influences the position of Phe76, broadening the entrance to the catalytic
pocket. More crystal structures have been resolved recently in association with an
array of relevant compounds (naphthalene, veratryl alcohol, benzyl alcohol,
2,6-dimethoxyphenol, styrene, acetanilide, and diclofenac), with all the aromatic
rings located at a van der Waals distance to the heme group. All of these are in the
pipeline for future structure-guided evolution experiments.
Among the available directed evolution tools, the use of neutral genetic drift
can help reveal promiscuous activities along with improved stabilities through the
accumulation of a set of neutral mutations. However, unless ultrahigh-throughput
screening assays are at hand, this method is very time-consuming as the genera-
tion of a neutral network (i.e., a polymorphic population of neutral variants with
a substantial number of substitutions) requires considerable experimental effort.
We have coupled in vivo shuffling with genetic drift to harness the benefits of
neutral evolution when engineering UPO ([43] and unpublished material). Using
PaDa-I as departure point, we shuffled genetic drift, and after only eight rounds,
the UPO sequence was modified by up to 11% (with an average of ~2 neutral
mutations per clone analyzed). Among the most promising neutral variants, we
identified clones with improved peroxygenase activity for different substrates,
those with reduced peroxidase activity and those with enhanced stability (over
5 °C in the T50).
5.3.2 Synthesis of 1-Naphthol: A Practical Case Study
A recent example of UPO evolution toward an industrial application was that of the
efficient synthesis of 1-naphthol, a compound that is very relevant in the agrochemi-
cal, textile, and pharma sectors [6, 25]. The production of 1-naphthol traditionally
requires chemical catalysts which, apart from their low efficiency and selectivity,
consume considerable energy and generate toxic waste [8–10, 34, 52]. Wild-type
UPO can oxygenate naphthalene yielding an intermediate epoxide that spontane-
ously hydrolyses at acid pH to generate 1-naphthol with only minor traces of
2-naphthol (Fig. 5.5) [32, 33]. Unfortunately, like all phenolic compounds, naph-
thols are substrates for the peroxidase activity of UPO, such that they are one elec-
tron oxidized by the enzyme to produce phenoxyl radicals and quinones. These
products subsequently undergo nonenzymatic homopolymerization that affects the
reaction yield. Hence, the question that arises is whether the peroxygenase activity
of UPO can be conserved while quenching the unwanted peroxidase activity in this
Fig. 5.5 Naphthalene oxygenation by UPO. First, naphthalene is transformed to 1,2-naphthalene

oxide, an unstable epoxide that splits into 1-naphthol and minor traces of 2-naphthol. 1-naphthol
is again used as substrate by the peroxidase activity of UPO, giving rise to 1,4-naphthoquinone.
The resulting quinones undergo nonenzymatic polymerization reactions that produce visible pre-
cipitates. 2-naphthol can also follow a similar reaction mechanism (not shown)
process. To address this challenge, we prepared a dual screening assay based on the
detection of 1-naphthol produced by UPO (positive selection criteria) and on the
one-electron oxidation of phenolics (negative selection criteria). Accordingly, only
those variants with improved/persistent peroxygenase activity to transform naph-
thalene into 1-naphthol and with reduced peroxidase activity to oxidize phenolics
were selected as candidates for further engineering [41]. As in the previous evolu-
tionary campaign, we also paid special attention to maintaining the robustness of
the protein by measuring kinetic thermostability during rescreenings. Using this
approach, we performed directed evolution using mutagenic PCR, in vivo shuffling,
and StEP, identifying a new mutant named JaWa (G241D-R257K) with notably
weaker peroxidase activity (3- to 12-fold lower in function of the substrate), slightly
improved peroxygenase activity on naphthalene (~1.5-fold better catalytic effi-
ciency), and enhanced thermostability (by 2 °C). As a result, JaWa had total turn-
over numbers of 50,000 mol product mol biocatalyst−1 and a high regioselectivity
(97%) for 1-naphthol production. Computational analysis by QM/MM indicated
that the mutation at position 241 rotates the backbone of the α-helix in which the
acid-base pair (Glu196-Arg189) lies such that Arg189 is closer to the substrate in
the catalytic site. In addition, the use of PELE (protein energy landscape explora-
tion) simulations [36] indicated that the distance between the naphthalene and the
heme was shorter for JaWa than for the parental PaDa-I, improving substrate
affinity.
Due to its unique features, this JaWa variant has been benchmarked in our labo-
ratory, along with other UPOs, for the synthesis of several phenolic HDMs like
5′-hydroxypropranolol 5′-OHP-, a process that requires the use of expensive radical
scavengers like ascorbic acid to revert the peroxidase activity. Indeed, JaWa by far
surpassed the performance of wild-type UPO, even in the absence of ascorbic acid,
and it is currently being subjected to new rounds of structure-guided evolution for
this synthetic application where peroxidase activity must be suppressed (unpub-
lished material).
In this regard, it is noteworthy that the reduction in peroxidase activity is fre-
quently associated to a drastic drop in stability, and as such, we are careful to select
stable variants while decreasing undesired one-electron oxidation. For example, we
identified the 3F10 mutant (T120P) in the course of evolution for secretion, with a
~fourfold improved peroxygenase to peroxidase ratio albeit at the cost of its stabil-
ity (a decreased in T50 of ~7 °C). Similarly, we applied MORPHING to structural
surface motives of UPO chosen by computational analysis in order to identify resi-
dues possibly involved in a long-range electron transfer (LRET) pathway to the
heme for the peroxidase activity. Through this approach we again found changes at
the same position but this time mutated to distinct residues (T120I; T120A) and
with a drastic drop in stability. Further combinatorial saturation mutagenesis experi-
ments unveiled the co-existence of several oxidations sites for peroxidative and per-
oxygenative activities in UPO [38]. We also explored several potential oxidizable
residues involved in such LRET by QM/MM, including a mutated Trp24 (W24F).
However, this change reduced both peroxidase and peroxygenase activities, empha-
sizing the complexity of this issue.
Our future endeavors to fully suppress peroxidase activity will include the intro-
duction of several stabilizing mutations together with those that suppress peroxi-
dase activity in a drive to obtain a unique mono(per)oxygenase with little peroxidase
activity (ongoing work).
Conclusions
Since C-H oxyfunctionalization is one of the most desired reactions in organic
synthesis, the use of UPO can potentially supplant chemical procedures to pro-
duce fine chemicals, drugs, or building blocks. Whether or not UPO becomes the
next-generation replacement of P450s remains to be seen. Nevertheless, since its
discovery in 2004, it has become clear that UPO may help simplify some com-
plex reactions that are particularly significant in organic chemistry and that until
now have been the exclusive terrain of P450s. UPO is extracellular and stable,
and it harnesses the limited peroxide shunt pathway of P450s in a highly profi-
cient manner, achieving thousands of total turnover numbers within a range of
selective oxygen transfer reactions. Even though UPO reunites some very attrac-
tive features, its practical use in real biotechnological processes is still to be
established. However, it is easy to envisage how the directed evolution of UPO
could represent a vehicle to transfer this biocatalyst from the laboratory bench to
industry. Indeed, UPO has now been adapted to the secretory machinery of S.
cerevisiae and P. pastoris by means of directed evolution, and it has been further
engineered for the synthesis of agrochemicals and HDMs, which we are confi-
dent is only the beginning.
We foresee that the robust tandem-yeast expression system designed for UPO
evolution will allow us to sculpt new enzyme attributes, involving both strong
peroxygenase activity with minimal oxidative inactivation and the full suppres-
sion of peroxidase activity. We truly visualize a future in which the catalog of
UPO applications will be widened toward nonnatural chemistry or activities in
nonnatural environments and whereby it may also be integrated into a fully con-
solidated bioprocessing microbe (white-rot yeast) in a drive toward more eco-
friendly solutions for the production of chemicals and fuels [15].
Acknowledgments We acknowledge the funding and financial support obtained from the
European Commission projects FP7-KBBE-2013-7-613549-INDOX, H2020-BBI-PPP-2015-2-
720297-ENZOX2, the COST-Action CM1303, and the Spanish Government projects BIO2013-
43407-R-DEWRY and BIO2016-79106-R-LIGNOLUTION.
References
1. Alcalde M (2015) Engineering the ligninolytic enzyme consortium. Trends Biotechnol
33:155–162
2. Anh DH, Ullrich R, Benndorf D, Svatoś A, Muck A, Hofrichter M (2007) The coprophilous
mushroom Coprinus radians secretes a haloperoxidase that catalyzes aromatic peroxygen-
ation. Appl Environ Microbiol 73:5477–5485
3. Aranda E, Kinne M, Kluge M, Ullrich R, Hofrichter M (2009) Conversion of dibenzothio-
phene by the mushrooms Agrocybe aegerita and Coprinellus radians and their extracellular
peroxygenases. Appl Microbiol Biotechnol 82:1057–1066
4. Babot ED, del Río JC, Kalum L, Martínez AT, Gutiérrez A (2013) Oxyfunctionalization of
aliphatic compounds by a recombinant peroxygenase from Coprinopsis cinerea. Biotechnol
Bioeng 110:2323–2332
5. Barková K, Kinne M, Ullrich R, Hennig L, Fuchs A, Hofrichter M (2011) Regioselective
hydroxylation of diverse flavonoids by an aromatic peroxygenase. Tetrahedron
67:4874–4878
6. Booth G (2012) Naphthalene derivatives. In: Ullmann’s encyclopedia of industrial chemistry.
Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim.
7. Brake AJ, Merryweather JP, Coit DG, Heberlein UA, Masiarz FR, Mullenbach GT, Urdea MS,
Valenzuela P, Barr PJ (1984) α-factor-directed synthesis and secretion of mature foreign pro-
teins in Saccharomyces cerevisiae. Biochemistry 81:4642–4646
8. Calinescu I, Avram R (1994a) 1-Naphthol production from tetraline. 1. Tetraline catalytic-
oxidation to 1-tetralene. Rev Chem 45:97–103
9. Calinescu I, Avram R (1994b) 1-naphthol production from tetraline. 3. Tetralone dehydrogena-
tion to 1-naphthol. Rev Chem 45:865–867
10. Calinescu I, Avram R, Lovu H (1994) 1-naphthol production from tetraline. 2. Tetraline
catalytic-oxidation to 1-tetralone – kinetic constants determination. Rev Chem 45:299–305
11. Camarero S, Pardo I, Cañas AI, Molina P, Record E, Martínez AT, Martínez MJ, Alcalde M
(2012) Engineering platforms for directed evolution of laccase from Pycnoporus cinnabari-
nus. Appl Environ Microbiol 78:1370–1384
12. Carro J, Ferreira P, Rodríguez L, Prieto A, Serrano A, Balcells B, Ardá A, Jiménez-Barbero
J, Gutiérrez A, Ullrich R, Hofrichter M, Martínez AT (2015) 5-hydroxymethylfurfural
conversion by fungal aryl-alcohol oxidase and unspecific peroxygenase. FEBS J 282:

3218–3229
13. Cirino PC, Arnold FH (2002) Protein engineering of oxygenases for biocatalysis. Curr Opin
Chem Biol 6:130–135
14. Garcia-Ruiz E, Gonzalez-Perez D, Ruiz-Dueñas FJ, Martínez AT, Alcalde M (2012) Directed
evolution of a temperature-, peroxide- and alkaline pH-tolerant versatile peroxidase. Biochem
J 441:487–498
15. Gonzalez-Perez D, Alcalde M (2014) Assembly of evolved ligninolytic genes in Saccharomyces
cerevisiae. Bioengineered 5:254–263
16. Gonzalez-Perez D, Molina-Espeja P, Garcia-Ruiz E, Alcalde M (2014) Mutagenic organized
recombination process by homologous in vivo grouping (MORPHING) for directed enzyme
evolution. PLoS ONE 9:e90919
17. Grinkova YV, Denisov IG, McLean MA, Sligar SG (2013) Oxidase uncoupling in heme mono-
oxygenases: human cytochrome P450 CYP3A4 in nanodiscs. Biochem Biophys Res Commun
430:1223–1227
18. Gröbe G, Ullrich R, Pecyna MJ, Kapturska D, Friedrich S, Hofrichter M, Scheibner K (2011)
High-yield production of aromatic peroxygenase by the agaric fungus Marasmius rotula.
AMB Express 2011:1–31
19. Gutiérrez A, Babot ED, Ullrich R, Hofrichter M, Martínez AT, del Río JC (2011) Regioselective
oxygenation of fatty acids, fatty alcohols and other aliphatic compounds by a basidiomycete
heme-thiolate peroxidase. Arch Biochem Biophys 514:33–43
20. Hofrichter M, Ullrich R (2006) Heme-thiolate haloperoxidases: versatile biocatalysts with bio-
technological and environmental significance. Appl Microbiol Biotechnol 71:276–288
21. Hofrichter M, Ullrich R, Pecyna MJ, Liers C, Lundell T (2010) New and classic families of
secreted fungal heme peroxidases. Appl Microbiol Biotechnol 87:871–897
22. Hofrichter M, Ullrich R (2013) Oxidations catalyzed by fungal peroxygenases. Curr Opin
Chem 19:116–125
23. Hofrichter M, Ullrich R, Kellner H, Upadhyay RC and Scheibner K (2014) Fungal unspecific
peroxygenases: a new generation of oxygen-transferring biocatalysts. In: Proceedings of the
8th international conference on Mushroom Biology and Mushroom Products (ICMBMP8),
pp 172–181
24. Hofrichter M, Kellner H, Pecyna MJ and Ullrich R (2015) Fungal unspecific peroxygenases:
heme-thiolate proteins that combine peroxidase and cytochrome P450 properties. In: Hrycay
EG, Bandiera SM (eds) Monooxygenase, peroxidase and peroxygenase properties and mecha-
nisms of cytochrome P450. Adv Exp Med Biol 851:341–368.
25. Jegannathan KR, Nielsen PH (2013) Environmental assessment of enzyme use in industrial
production – a literature review. J Clean Prod 42:228–240
26. Joo H, Lin Z, Arnold FH (1999) Laboratory evolution of peroxide-mediated cytochrome P450
hydroxylation. Nature 399:670–673
27. Karich A, Kluge M, Ullrich R, Hofrichter M (2013) Benzene oxygenation and oxidation by the
peroxygenase of Agrocybe aegerita. AMB Express 3:5
28. Kinne M, Poraj-Kobielska M, Aranda E, Ullrich R, Hammel KE, Scheibner K, Hofrichter M
(2009) Regioselective preparation of 5-hydroxypropranolol and 4′-hydroxydiclofenac with a
fungal peroxygenase. Bioorg Med Chem Lett 19:3085–3087
29. Kinne M, Poraj-Kobielska M, Ralph SA, Ullrich R, Hofrichter M, Hammel KE (2009)
Oxidative cleavage of diverse ethers by an extracellular fungal peroxygenase. J Biol Chem
284:29343–29349
30. Kinne M, Zeisig C, Ullrich R, Kayser G, Hammel KE, Hofrichter M (2010) Stepwise oxygen-
ations of toluene and 4-nitrotoluene by a fungal peroxygenase. Biochem Biophys Res Commun
397:18–21
31. Kinne M, Poraj-Kobielska M, Ullrich R, Nousiainen P, Sipilä J, Scheibner K, Hammel KE,
Hofrichter M (2011) Oxidative cleavage of non-phenolic ß-0-4 lignin model dimers by an
extracellular aromatic peroxygenase. Holzforschung 65:673–679
32. Kluge M, Ullrich R, Scheibner K, Hofrichter M (2007) Spectrophotometric assay for detection
of aromatic hydroxylation catalyzed by fungal haloperoxidase-peroxygenase. Appl Microbiol
Biotechnol 75:1473–1478
33. Kluge M, Ullrich R, Dolge C, Scheibner K, Hofrichter M (2009) Hydroxylation of naphtha-
lene by aromatic peroxygenase from Agrocybe aegerita proceeds via oxygen transfer from
H2O2 and intermediary epoxidation. Appl Microbiol Biotechnol 81:1071–1076
34. Kudo K, Ohmae T, Uno A (1976) U S Patent 3, 935, 282
35. Lucas F, Babot ED, Cañellas M, del Río JC, Kalum L, Ullrich R, Hofrichter M, Guallar V,
Martínez AT, Gutiérrez A (2016) Molecular determinants for selective C25-hydroxylation of
vitamins D2 and D3 by fungal peroxygenases. Catal Sci Technol 6:288–295
36. Madadkar-Sobhani A, Guallar V (2013) PELE web server: atomistic study of biomolecular
systems at your fingertips. Nucleic Acids Res 41:322–328
37. Mate DM, García-Burgos C, García-Ruiz E, Ballesteros AO, Camarero S, Alcalde M (2010)
Laboratory evolution of high-redox potential laccases. Chem Biol 17:1030–1041
38. Mate DM, Palomino MA, Mollina-Espeja P, Martin-Diaz J, Alcalde M (2017) Modification of
the peroxygenative: peroxidative activity ratio in the unspecific peroxygenase from Agrocybe
aegerita by structure-guided evolution. Protein Eng Des Sel. In press. https://doi.org/10.1093/
protein/gzw073
39. Molina-Espeja P, Garcia-Ruiz E, Gonzalez-Perez D, Ullrich R, Hofrichter M, Alcalde M
(2014) Directed evolution of unspecific peroxygenase from Agrocybe aegerita. Appl Environ
Microbiol 80:3496–3507
40. Molina-Espeja P, Ma S, Mate DM, Ludwig R, Alcalde M (2015) Tandem-yeast expression
system for engineering and producing unspecific peroxygenase. Enzym Microb Technol
73-74:29–33
41. Molina-Espeja P, Cañellas M, Plou FJ, Hofrichter M, Lucas F, Guallar V, Alcalde M (2016)
Synthesis of 1-naphthol by a natural peroxygenase engineered by directed evolution.
ChemBioChem 17:341–349
42. Ni Y, Fernández-Fueyo E, Gomez Baraibar A, Ullrich R, Hofrichter M, Yanase H, Alcalde M,
van Berkel WJH, Hollmann F (2016) Peroxygenase-catalyzed oxyfunctionalization reactions
promoted by the complete oxidation of methanol. Angew Chem Int Ed Eng 54:1–5
43. Paret C, Martin-Diaz J, Alcalde M (2015) Shuffling the neutral drift of unspecific peroxygen-
ase expressed in yeast. Master degree dissertation in Molecular and Cellular Biology,
Universidad Autónoma de Madrid
44. Pecyna MJ, Ullrich R, Bittner B, Clemens A, Scheibner K, Schubert R, Hofrichter M (2009)
Molecular characterization of aromatic peroxygenase from Agrocybe aegerita. Appl Microbiol
Biotechnol 84:885–897
45. Peter S, Kinne M, Ullrich R, Kayser G, Hofrichter M (2013) Epoxidation of linear, branched
and cyclic alkenes catalyzed by unspecific peroxygenase. Enzym Microb Technol
52:370–376
46. Peter S, Karich A, Ullrich R, Gröbe G, Scheibner K, Hofrichter M (2014) Enzymatic one-pot
conversion of cyclohexane into cyclohexanone: comparison of four fungal peroxygenases.
J Mol Catal B Enzym 103:47–51
47. Piontek K, Strittmatter E, Ullrich R, Gröbe G, Pecyna MJ, Kluge M, Scheibner K, Hofrichter
M, Plattner DA (2013) Structural basis of substrate conversion in a new aromatic peroxygen-
ase cytochrome P450 functionality with benefits. J Biol Chem 288:34767–34776
48. Poraj-Kobielska M, Kinne M, Ullrich R, Scheibner K, Kayser G, Hammel KE, Hofrichter M
(2011) Preparation of human drug metabolites using fungal peroxygenases. Biochem
Pharmacol 82:789–796
49. Poraj-Kobielska M (2013) Conversion of pharmaceuticals and other drugs by fungal peroxy-
genases. PhD Dissertation. Technical University of Dresden, Germany
50. Rydberg P, Sigfridsson E, Ryde U (2004) On the role of the axial ligand in heme proteins: a
theoretical study. J Biol Inorg Chem 9:203–223
51. Sakaki T (2012) Practical application of cytochrome P450. Biol Pharm Bull 35:844–849
52. Schuster L, Seid B (1979) U S Patent 4, 171, 459
53. Shoji O, Watanabe Y (2014) Peroxygenase reactions catalyzed by cytochromes P450. J Biol
Inorg Chem 19:529–539
54. Ullrich R, Nüske J, Scheibner K, Spantzel J, Hofrichter M (2004) Novel haloperoxidase from
the agaric basidiomycete Agrocybe aegerita oxidizes aryl alcohols and aldehydes. Appl
Environ Microbiol 70:4575–4581
55. Ullrich R, Dolge C, Kluge M, Hofrichter M (2008) Pyridine as novel substrate for regioselec-
tive oxygenation with aromatic peroxygenase from Agrocybe aegerita. FEBS Lett 582:
4100–4106
56. Viña-Gonzalez J, Gonzalez-Perez D, Ferreira P, Martinez AT, Alcalde M (2015) Focused
directed evolution of aryl-alcohol oxidase in Saccharomyces cerevisiae by using chimeric sig-
nal peptides. Appl Environ Microbiol 81:6451–6462
Recent Advances in Directed Phytase
Evolution and Rational Phytase 6
Engineering
Amol V. Shivange and Ulrich Schwaneberg
Abstract
Phytases are hydrolytic enzymes that initiate stepwise removal of phosphate
from phytate. Phytate is the major phosphorous storage compound in cereal
gains, oilseeds, and legumes and is indigestible by monogastric animals such as
poultry and swine. Supplementation of phytase in animal feed proved to improve
animal nutrition and decrease phosphorous pollution. Several phytases were
discovered in the last century, and today a highly competitive market situation
emerged the demands for phytases that are redesigned to excellently match
industrial demands. Phytase engineering by directed evolution and rational
design has offered a robust approach to tailor-made phytases with high specific
activity, broad thermal and pH profile, and protease resistance. In this chapter,
we summarized challenges and successful approaches employed in phytase
engineering. Factors influencing phytase thermostability, pH stability, pH optima,
and protease resistance have been discussed with respect to structural perspective
and potential molecular mechanism for improvement. Importance of cooperative
substitutions and a way to identify these interactions are discussed. Recent
development in screening technology and molecular insights in combining key
beneficial substitutions are detailed. In addition, strategies and approaches for
rapid and efficient evolution of phytases and to understand structure function
relationships on a molecular level have been proposed.
A.V. Shivange, PhD (*)

Division of Biology and Biological Engineering, California Institute of Technology,
Mail Code 156-29, 1200 E. California Blvd, Pasadena, CA 91125, USA
U. Schwaneberg
Lehrstuhl für Biotechnologie, RWTH Aachen University,
Worringerweg 1, 2074 Aachen, Germany

DOI 10.1007/978-3-319-50413-1_6
146 A.V. Shivange and U. Schwaneberg
6.1 Introduction
Phytase catalyzes a partial or complete removal of inorganic phosphate from phytic

acid (myo-inositol hexakisphosphates) or its salt phytate by stepwise hydrolysis of
phosphomonoester bonds (Fig. 6.1). Phytate is a primary storage form of phosphate
in plants, stored in cereal gains, oilseeds, and legumes which accounts for 50–80%
of total phosphate in the plant [1]. Unfortunately, phytate is indigestible by mono-
gastric animals such as swine, poultry, fish, and human beings due to lack of ade-
quate phytase activity in the gastrointestinal tract. In addition, a strong chelating
property of phytate makes complexes with divalent cations, proteins, and amino
acids [2, 3]. Consequently, supplementation of digestible phosphate is required
resulting in a large amount of inorganic phosphate excretion (2% of the dry matter
content) causing environmental pollution (60 million pigs are produced in the
United States annually, and each pig excretes 1.23 kg of phosphorous in the full life
cycle) [4]. Supplementation of phytase as a feed additive has proven to improve
nutrition uptake and reduce phosphate excretion by 50% [5]. An estimated supple-
mentation with 300–600 phytase activity units/kg of the diet may replace phospho-
rous (1.0 g/kg of diet) supplied as monocalcium phosphate [4]. Additionally,
supplemental phytase improves calcium, zinc, and iron utilization by animals [6, 7].
The first phytase was reported in 1907 [4] and the phytase research persisted slowly
over a century because the use of phytase provided no cost benefits over the use of
inorganic phosphorous as a feed supplement. In the last decade, a potential sixfold
increase in inorganic phosphorous cost (phosphate rocks forecasted to be eventually
mined out [8]), awareness, and recent worldwide regulations on controlling agricul-
tural pollution has made phytase supplementation much more cost-effective and
attractive. Currently, phytases represent more than 60% of the total feed enzyme
market [9] with a market value of US ∼$350 million per year and an impressive
growth rate of 10% per annum [10]. In addition to the feed industry, phytases have
a great potential in food processing [11], biofuel production [12], alcohol produc-
tion [13], human nutrition [1], and synthesis of lower myo-inositol phosphate [14].
Fig. 6.1 Schematic illustration of the phytase hydrolytic reaction. A stepwise removal of phos-
phate is catalyzed by hydrolytic cleavage of a phosphomonoester bond (bold lines) releasing inor-
ganic phosphate, less phosphorylated myo-inositol, divalent metal ions, and bound proteins
6 Recent Advances in Directed Phytase Evolution and Rational Phytase Engineering 147
Phytases are widely distributed and found in animals, plants, and microorgan-
isms. Based on the crystal structure and catalytic mechanism, a commonly accepted
phytase nomenclature divides them into four classes: histidine acid phosphatase
(HAP), β-propeller phytase (BPP), purple acid phosphatase (PAP), and protein tyro-
sine phosphatase (PTP) also referred as cysteine phytase. Several extensively stud-
ied and commercialized fungal and bacterial phytases belong to HAP class that
shares conserved active site sequence motifs (RHGXRXP and HD) and two struc-
tural domains (α-domain and α/β-domain). Another extensively studied phytase
from BPP class, also known as alkaline phytases, has β-sheet propeller structure,
exhibits strict activity for calcium-phytate complex, and shows high thermostability
[15]. Industrial application of phytases requires high specific activity, higher ther-
mal tolerance for feed pelleting process (60–80 °C), and higher pH stability and
protease resistance to withstand gastric environments. Phytases from HAP class
possess high specific activity (100–3000 U/mg) but have moderate thermostability,
whereas phytases from BPP class are more thermostable but exhibit lower catalytic
efficiency (specific activity <50 U/mg). Therefore, an ideal phytase may rarely, if at
all, be found in nature, and there is a constant need to evolve phytases for industrial
demands. In this chapter, we summarize recent development in phytase engineering
by directed evolution and rational design strategies. Novel strategies and approaches
for rapid and efficient evolution of phytases are additionally discussed based on suc-
cess stories in phytase reengineering.
6.2 Challenges in Directed Phytase Evolution
6.2.1 L
imitations of Random Mutagenesis Methods Employed
for Directed Phytase Evolution
Despite the advances in “smart” mutant library generation and screening methods,
directed phytase evolution campaigns have been a challenging and laborious pro-
cess. Thermostability improvement for phytases has been a primary goal for indus-
trial demands of phytases. The main challenge in designing directed phytase
evolution experiment for improving thermal resistance is selection of a “right”
method for mutagenesis and screening. Over the past decade, several phytases have
been evolved for higher thermostability. A key observation in the success stories of
directed phytase evolution has been that the majority of amino acid substitutions are
“chemically” different than the wild-type amino acids, for example, glycine to
aspartic acid, aspartic acid to asparagine [16], lysine to glutamate/methionine [17],
threonine to lysine, glutamine to histidine, and lysine to glutamine [18]. This chemi-
cal diversity in the substitution pattern indicates a prerequisite to employ a muta-
genesis method that generates diverse mutational spectra enriched with functional
traits. Organization of the genetic code and the naturally occurring mutational bias
favor strongly chemically similar substitution patterns [19] agreeing with the ambi-
guity reduction theory [20], which proposes that the genetic code was organized to
reduce incorporation of chemically diverse substitutions. Widely used
polymerase-based mutagenesis methods have a bias toward transitions (mutation

from purine ➔ purine or pyrimidine ➔ pyrimidine), resulting in hampered chemical
diversity in the generated mutant library. For example, a transition at the third posi-
tion of a codon encodes in almost all the codons the wild-type amino acid. In con-
trast, transversion (A/G ➔ T/C or T/C ➔ A/G)-enriched mutagenesis methods have
been shown to generate chemical diversity enriched with diverse mutational spectra
[21, 22]. Screening several epPCR libraries (6800 clones) failed to yield a thermo-
stable variant from Yersinia mollaretii phytase (Ymphytase); however, generating a
chemically diverse library with high mutational load resulted in an identification of
a thermostable Ymphytase variant M1 with moderate screening efforts (2350
clones) [22]. Two (T77K and V298M) out of five (D52N, T77K, K139E, G187S,
V298M) substitutions found in M1 were unobtainable by the employed epPCR
method. These results provide an insight that employing a mutagenesis method
which incorporates chemically diverse substitutions will enrich the library popula-
tion with functional phytase variants and reduce screening efforts in the directed
phytase evolution.
6.2.2 P
hytase Fitness Landscape: Improving More than One
Property in Directed Phytase Evolution
Enzyme evolution for industrial demands frequently requires an optimization of

more than one property. An ideal phytase would be thermostable, highly active,
protease resistant and low pH resistant. Directed phytase evolution campaigns have
usually been aimed to improve one property at a time. An essential property for
industrial feed pelleting process is the high thermal tolerance. A known fact is that
the phytases from thermophilic organism either have undesirable high optimal tem-
perature (>50 °C) or lower specific activity [23], whereas mesophilic phytases have
higher specific activity but moderate thermostability [17]. A “fitness” for phytases
can be defined as a sequence that satisfies all the criteria for a phytase to function as
desired. Criteria in the screening process may include high activity at low pH and a
high thermal and protease resistance. Phytase evolution can then be envisioned as
an uphill walk from one functional phytase to the desired fitness (from Wt to M6 in
Fig. 6.2a). The fitness landscape for a phytase with more than one criterion in which
two properties are being improved is shown in Fig. 6.2. It is obvious that a fitness
landscape for one property (fitness versus combination of substitutions) cannot
overlap with other property of interest indicating a separate fitness landscape exists
for each property. Thermostability and activity improvement of a phytase are
inversely proportional to each other and can be correlated well with the flexibility of
the enzyme. Reducing the flexibility of a phytase by improving intramolecular non-
covalent interactions has been shown to enhance thermal resistance [18, 24], while
high flexibility in the active site has been proposed to be a prerequisite for higher
catalytic efficiency [25]. On a path of uphill walk of the phytase thermostability fit-
ness (Fig. 6.2a), several variants that have low or even undesirable fitness for activ-
ity exist (Fig. 6.2b). During iterative rounds of directed evolution, it is crucial to
a b
Fig. 6.2 A fitness landscape of combining key substitutions in directed Ymphytase evolution. The
mapping of genotype (combination of substitutions) to phenotype (experimentally determined fit-
ness for desired property) for thermostability (a) and phytase activity (b) revealed a nonoverlap-
ping fitness landscape, indicating a separate fitness landscape exists for each property. A walk from
blue, to cyan, to green, to yellow, and to red represents an increased phytase fitness. On the uphill
walk for thermostable variant (from Wt to M6), several variants with undesirable activity exist.
Epistatic effects play a major role in optimizing enzymes for more than one fitness; a combination
of substitution in M3 improved two properties as a result of epistatic interactions
select phytase variants with more than one selection pressure and combine the key
substitutions to identify variants with higher fitness for more than one property
(M3 in Fig. 6.2).
6.2.3 I mportance of Cooperative Substitutions (Epistatic

Effects) in Directed Phytase Evolution
Epistatic effects can be defined as nonadditive interactions between substitutions

that affect enzyme properties. These substitutions act cooperatively and can only be
discovered if more than one amino acid is being mutated simultaneously. Over the
past decades, it is becoming clear that there are two pathways in directed evolution,
(i) a widely known pathway of accumulating beneficial substitutions in each round
of mutagenesis and screening and (ii) a pathway in which cooperative/epistatic
effects between substitutions result in additive or synergistic effects. Epistatic
effects in directed evolution have been underestimated despite their high impor-
tance [18, 26–28]. The latter might be attributed to (i) a low mutational load of
mutant libraries (classical directed evolution experiment), (ii) a low robustness
toward mutations depending on the protein structure and catalyzed reaction, and
(iii) methodological limitations to incorporate more than one substitution per pro-
tein. Classical directed evolution achieves accumulation of a beneficial substitution
in iterative rounds of mutagenesis and screening, in which enzymes are often
improved by exchanging one amino acid per round of evolution [29]. Several
improved phytases possess more than one residue substitution per round of directed
evolution [30]. A study elucidating individual role of each substitution as well as the
role of each combination of substitutions revealed cooperative effects between sub-
stitutions. A detrimental substitution G187S (affecting activity and thermostability)
in Ymphytase improved thermostability in combination with distantly located T77K

substitution, supporting the hypothesis that non-beneficial substitutions may also be
vital in directed phytase evolution. Another substitution Q154H had no effect on
activity; however, the specific activity was increased by ~200 U/mg after being
combined with G187S-T77K variant [18]. A surface substitution K265E had a neg-
ligible effect on Bacillus phytase activity; however, the specific activity was
improved when K265E was combined with either D24G or S51G (active site substi-
tutions) [28]. Interestingly, a multisite saturation of five distantly located sites that
were identified in a directed phytase evolution improved thermostability as well as
pH stability of Ymphytase that might be due to possible epistatic effects [26]. The
sites for saturation mutagenesis were identified from a previous directed evolution
experiment employing random mutagenesis method. A threefold higher pH stability
(at pH 2.8/3 h) and a small improvement in thermal resistance were obtained when
compared to previously identified combination of substitutions (M1). Discovery of
a novel combination by screening only 1100 clones from an OmniChange library
proved to be a promising strategy to re-interrogate the substitutions for possible
epistatic effects. These results demonstrate an interesting and vital role of the sec-
ond pathway in directed evolution and encourage to incorporate strategies to study
and improve enzymes by cooperative substitutions.
6.3 xploring Phytase Engineering by Directed Evolution

E
and Rational Design Strategies
Directed evolution and rational design strategies have emerged concomitantly as an

ideal method to tailor-make a phytase for industrial demands. Over the past years,
several microbial phytases with high catalytic efficiency and moderate thermosta-
bility have been discovered [31]. Especially, the phytases (HAP) from the
Enterobacteriaceae family have attracted industrial interest due to their high spe-
cific activity. The major goal in phytase design has been engineering phytases with
higher activity and thermostability. Phytases engineered by directed evolution and
rational design strategies have been summarized in Tables 6.1 and 6.2, respectively.
Structural flexibility plays an important role in enzyme activity and thermostability
and usually depends on the nature of intra-protein interactions. Improving thermo-
stability typically affects activity of phytases and vice versa; therefore, implement-
ing a “smart” strategy is a prerequisite in engineering phytases for industrial
applications. A flowchart on the design and discovery of novel phytases is shown in
Fig. 6.3 illustrating several successful approaches used to engineer phytases.
6.3.1 Engineering Thermal Resistance in Phytases
6.3.1.1 Hydrogen Bonding Network

Hydrogen bonds in a protein structure are considered a dominant factor for thermal
stabilization [47, 48]. Several directed phytase evolution experiments yielded
Table 6.1 Directed phytase evolution campaigns employed in improving phytases for industrial applications
Phytase Specific Mutagenesis
name activity method (clones
Organism name (family) (U/mg) screened) Tm (°C) Substitutions Improvements Potential mechanism Reference
Aspergillus phyA Parent: epPCR (5500) NA E156G, 61% higher E156G, T236A: [32]
niger N25 (HAP) 204,514 T236A, specific activity decreased number of
Q396R hydrogen bonds ➔
improved phytase
flexibility
Variant: 81% higher Q396R: improved
330,064 catalytic affinity to negatively
efficiencya charged phytate
Escherichia AppA2 wt: 1003 epPCR (5000) wt: 62 Variant-1: 20% increased K46E: increased [17]
coli (HAP) K46E residual activity hydrogen bonds
at 80 °C/10 min
Variant-1: Variant-1: Variant-2: 6–7 °C increased K65E: increased
742 68.5 K65E, K97M, Tm hydrogen bonds ➔
S209G stabilizing interaction
between α- and
α/β-domains
Variant-2: Variant-2: 56% and 152% K97M/S209G: reduced
905 69.8 improved structural hindrance to
catalytic neighboring residues
efficiency
(continued)
6 Recent Advances in Directed Phytase Evolution and Rational Phytase Engineering
151
152

Escherichia AppA wt and Gene site wt: 63.7 Q84W, 12 °C increased Reduced flexibility in [33, 34]
coli (HAP) Phy9X saturation Y277D, Tm loops and random coiled
1700 mutagenesis W68E, K97C, structures
(GSSM) Phy9X: R181Y, 3.5-fold higher Increased surface polarity
(158,608) 75.7 N226C, A95P, gastric stability Elimination of
S168E N-glycosylation site
close to the active site ➔
reduced steric hindrance
for substrate entry
Penicillium phyA wt: 32 epPCR with NA 2–28: T11A, ~Fourfold Increased number of [35]
sp.Q7 (HAP) MnCl2 and then G56E, L65F, increased intra-protein hydrogen
epPCR with Q144H, specific activity bonds
dITP (5300) L151S
Variant 2–249: T11A, 55% (2–28) and L151S: improved
2–28: 133.3 H37Y, G56E, 74% (2–249) catalytic residue (H352)
L65F, Q144H, increased exposure to the substrate
L151S, residual activity
N354D at 80 °C/5 min
Variant Retained pepsin N354D: decreased
2–249: resistance hydrogen bonds to
136.6 (96.7% activity catalytic residue D353 ➔
after 2 h at 0.01 improved activity
pepsin/phytase Reduced pKa of the side
(w/w) ratio) chain (N354D) affected
pKa of the catalytic
center ➔ shift on the
optimal temperature
A.V. Shivange and U. Schwaneberg
Escherichia AppA wt: 1610 epPCR (NA) NA I408L 23.3% increased Stabilizing terminal part [36]
coli (HAP) I408L: residual activity of the phytase
1653 at 80 °C/5 min
Yersinia Ymphytase wt: 1073 SeSaM (2350) wt: 63 D52N, T77K, ~20% increased T77K: incorporation of a [18, 22]
mollaretii (HAP) K139E, residual activity salt bridge
G187S, at 58 °C/20 min
M1: 993 M1: 64.5 V298M 1.5 °C increased G187S: improved
Tm hydrogen bonding
network to an adjust loop
Yersinia Ymphytase wt: 1043 OmniChange wt: 59 D52E, 32% increased D52E, G187S, K139E: [26]
mollaretii (HAP) (1100) K139T, residual activity increased number of
G187S, at 58 °C/20 min hydrogen bonds
Variant Omni1: V298F 2 °C increased V298F: incorporation of
Omni1: 61 Tm aromatic-aromatic
1034 Twofold higher interactions
pH stability (pH Possible epistatic
2.8; 3 h) interactions
(continued)
153
154

Yersinia Ymphytase wt: 1073 SeSaM (1600) wt: 58.5 M3: T77K, M6: 54% Increased intra-protein [18]
mollaretii (HAP) G187S, increased hydrogen bonding
Q154H residual activity network
at 58 °C/20 min
M3: 1274 SSM (200) M3: 61 M6: T77K, M6: 3 °C T77K: incorporated salt
Q154H, increased Tm bridge ➔ stabilizing two
G187S, adjacent helices
M6: 1017 SDM M6: 61.5 K289Q Specific activity T77K and K289Q:
of M3 increased increased interactions
by ~200 U/mg between adjacent helices
(holding tightly) ➔
reduced flexibility in the
loop situated next to the
helix ➔ improved
thermostability
KeySIDE Q154H: improved
solvent exposure ➔
improved thermostability
Increased flexibility of
the active site loop
epistatic effects between
T77K and G187S
(thermostability) and
T77K, G187S, Q154H
(activity)
Bacillus phy168 wt: 13.8 epPCR (NA) NA D24G, K70R, 42.8% D24G: widening of the [28]
subtilis 168 (BPP) K111E, improvement in binding pocket and
N121S specific activity elimination of steric
hindrance for phytate
binding
Variant: 2.3-fold Epistatic effects between
19.7 improved D24G and K265E/
catalytic K265N and S51G and
efficiency K265E
Retention time to Increased number of
reach 40% hydrogen bonds ➔
residual activity improved thermostability
at 80 °C was
improved by
17 min
Synthetic Ymphytase wt: 1043 ProCoS (1050) NA Synthetic: 34 Broadened pH 20% increased negative [52]
substitutions optima polar surface
ProCoS-2: 3.8-fold 8% increased positive
401 improved pH polar surface
stability at pH pH stability increased by
2.8 (37 °C/3 h) increased fraction of
polar surface and
decreasing neutral
surface (20%)
a
kcat/Km, Tm melting temperature, HAP histidine acid phosphatase, BPP β-propeller phytase, NA not available
155
Table 6.2 Approaches employed to improve phytases by rational and semi-rational design
156
Phytase
name Specific activity
Organism name (family) (U/mg) Tm (°C) Substitutions Rationale Strategy Results Reference
Synthetic Consensus Parent: NA Parent: Synthetic Consensus Design of a Optimum [24]
phytase 56–63 sequence consensus temperature
(HAP) sequence from increased from
13 fungal 45–55 to 71 °C
Consensus: ~30 Consensus phytase 15–22 °C
78 sequences increased Tm
Stabilized a
surface loop and
N-terminus by
increased number
of hydrogen
bonds
Escherichia coli EcAppA wt: 2163 NA N204A Sequence Altering binding N204A or S206A: [34]
(HAP) N204A or S206A: S206A alignment with pocket and elimination of
~2370 the consensus glycosylation N-glycosylation
KKG ➔
of highly pattern based on site close to the
NQT
active phytases Citrobacter active site ➔
HPP ➔ NGT phytases reduced steric
P173S hindrance for
substrate entry
G311S 33% increased
residual activity
at 66 °C/5 min for
the triple mutant
Aspergillus niger PhyA wt: 80 wt: 63.3 A58E, P65S, Structure Adopting 20% increased [37]
(HAP) Q191R, alignment with hydrogen residual activity
T271R a thermostable bonding network at 100 °C/10 min
Variant: 83 Variant: phytase and ionic 7 °C increased
~73 interactions form Tm
a thermostable Improved affinity
homolog (decreased Km)
Improved
inorganic
phosphate release
from phytate in
soybean meal
Synthetic Beta- FTE: ~28 NA Synthetic Structure- A homology Optimum [38]
propeller guided model of the temperature
phytase consensus consensus between 45 and
(BPP) sequence sequence from 70 °C
FTEII: ~28 15 bacillus Variants
phytases was possessed higher
used to design thermostability at
sequences pH 5.5 than pH
7.5
FBA: ~37 Higher
thermostability at
5 mM CaCl2 than
1 mM CaCl2
Retained ~80%
residual activity
(80 °C/10 min) at
pH 5.5, 5 mM
CaCl2
(continued)
157
158
Phytase
Escherichia coli AppA NA NA NA Deletion of Deletion of 39% increased [39]
(HAP) flexible region seven C-terminal residual activity
residues at 80 °C/10 min
Escherichia coli AppA wt: 1541 NA Y311K, Structure- Selection of 30% increased [40]
(HAP) I427L based design flexible surface residual activity
residues based at 80 °C/10 min
Variant: 1274 on an MD Increased affinity
simulation (decreased Km)
(RMSF)
Incorporation of Epistatic effects
salt bridges (synergy)
between Y311K
and I427L
Yersinia APPA wt: 500 NA S51I Sequence Binding site S51I or S51T: [41]
frederiksenii (HAP) alignment analysis based on shift in the
a homology optimum pH from
model and 2.5 to 4.5
S51I: 3333 sequence Side chain
alignment structure (not the
charge) affected
shift in the
optimum pH
Bacillus sp. MD2 phytase wt: 32 NA E229V Binding site Decreasing E229V: 19% [42]
(BPP) charge negatively increased specific
modification charged residues activity
E229V: 38 K77R, near binding site, K77R-K179R:
K77R-K179R: 7.5 K179R widening ~20% increased
binding pocket residual activity
(pH stability) at
pH 2.6
Escherichia coli AppA wt: 37 NA C2 Altering Incorporation of 25% increased [43]
(HAP) glycosylation N-glycosylation residual activity
pattern sites at 80 °C/15 min
(no alteration in
glycosylation)
C200N/D207N/ Increased specific
S211N:65 activity may be
due to eliminating
a disulfide bond
(C200N) ➔
increased
flexibility of
α-domain
(continued)
159
160
Phytase
Bacillus PhyL wt: PhyLG-A: Restricting Incorporation of PhyLG-A: 2.5 fold [44]
licheniformis (BPP) G117A, conformational rigidifying improved free
G266A flexibility substitutions energy of
(Xaa ➔ Pro, Gly unfolding (ΔGu)
PhyLG-A: PhyLX-P: ➔ Ala) at Gly ➔ Ala
PhyLX-P: H32P, S256P, consensus substitution at
K304P, position consensus
K324P, positions was
S353P more promising
Synthetic Afp, Anp Anp: 110 Anp: 62.1 NA Structure- Swapping of Variant AB345F7 [45]
(Aspergillus niger (HAP) based fragment fragments based showed highest
and A. fumigatus) shuffling on structural specific activity
Afp: 40 Afp: 59 elements in two Variant A234567:
A234567: 40 A234567: homologous 7.8 °C increased
66.8 phytases (Anp: Tm compared to
AB345F7: 90
ABCDEFG and Afp
Afp: 1234567)
Bacillus Phytase wt: 15 D148E Sequence Increasing D148E: 27% [46]
amyloliquefaciens (BPP) alignment and glutamic acid increased residual
DSM 1061 structure content ➔ activity at
visualization improves 70 °C/10 min
D148E: 20 S197E thermostability D148E: increased
N156E specific activity
D52E
Tm melting temperature, HAP histidine acid phosphatase, BPP β-propeller phytase, NA not available/applicable, MD molecular dynamics
Fig. 6.3 A flowchart representing a design and discovery path of novel phytase by selection of a
“smart” phytase engineering strategy. Successful approaches employed in directed phytase evolu-
tion are illustrated. (Ts transition, Tv transversion-biased mutagenesis methods, 4-MUP
4-methylumbelliferyl phosphate assay, AMol ammonium molybdate assay)
thermally improved variants with a “stronger” hydrogen bonding network.

Thermostability of a phytase (AppA2) from E. coli was improved by 20%
(80 °C/10 min) via a strengthened hydrogen bonding network to adjacent secondary
structures [17]. A charged substitution (K46E) fostered intra-protein hydrogen
bonds resulting in 6 °C improved melting temperature (Tm). Additionally, in another
variant, a substitution K65E introduced hydrogen bonds that might stabilize the
interactions between α- and αβ-domain resulting in a 7 °C Tm improvement. A
protease-resistant phytase (phyA) from Penicillium was evolved using epPCR
employing Mn2+-dITP for higher thermal resistance. Thermal resistance was
improved by 74% (80 °C/5 min), and the substitutions G56E, L65F, Q144H, and
L151S introduced new hydrogen bonds to neighboring residues [35]. Recently, a
phytase from Yersinia mollaretii was evolved using a transversion-enriched muta-
tional pattern generated by the SeSaM (sequence saturation mutagenesis) random
mutagenesis method and a subsequent KeySIDE (iterative key-residues interroga-
tion of the wild type with substitutions identified in directed evolution) approach to
combine key beneficial substitutions [18]. A combination of substitutions yielded a
variant M6 with 54% improved thermostability (58 °C/20 min) and 3 °C higher
melting temperature. Molecular dynamics simulations on M6 variants revealed an
increased hydrogen bonding network in the phytase, stabilizing two adjacent helices
(K289Q) and a surface loop (G187S). Adopting a hydrogen bonding network and
ionic interactions from a thermostable A. fumigatus phytase into the A. niger, phy-
tase improved melting temperature by 7 °C. Substitutions A58E and P65S intro-
duced new hydrogen bonds that contributed to improved thermostability. These
results suggest employing an evolution strategy directed toward introducing hydro-
gen bonds in a phytase as a highly beneficial strategy for improving thermal resis-
tance. Almost all substitutions in the improved phytase variants were chemically
different ones (opposite charge or polar), suggesting to employ random mutagenesis
method which incorporates diverse amino acids. A transversion-biased mutagenesis
method able to introduce subsequent nucleotide mutations or a codon exchange
mutagenesis method offers a higher chance of incorporating chemically different
amino acids than routinely used epPCR method with a transition bias.
6.3.1.2 Phytase Flexibility

Phytase flexibility especially in loop regions (quantified via MD simulations) is an
important criterion to be considered during selection and evolution of a phytase for
higher thermal tolerance or activity [18, 40]. Simulations on the improved
Ymphytase variant revealed a reduced flexibility in loop regions compare to the
wild-type Ymphytase. Interestingly, these loops were located next to helices that
possess substitutions (e.g., reduced flexibility in a loop located after helix-K, sub-
stitution K289Q belongs to helix-K stabilizing helix-K and L). Similarly, T77K
(on helix-B) introduced a salt bridge that holds two helices (helix-B and C) together
resulting in a reduced flexibility of a loop located next to helix-B [18]. The active
site loop studies illustrated a vital role of the loop flexibility in highly active phy-
tases (specific activity 1000–3000 U/mg). This active site loop region was replaced
by a helix in phytases with lower specific activity (~100 U/mg; fungal and
Klebsiella phytase) [25], suggesting that the local flexibility in the active site is a
prerequisite for higher catalytic activity, and therefore the loops situated away
from the active site can be targeted for mutagenesis. Randomizing helices that are
located distantly from the active site loop and next to flexible loops may enrich the
library with thermostable phytase variants. A classical study of whole-gene ran-
domization, saturating each position in the E. coli phytase (AppA) and screening
(158,608 clones) with 384-well-based system, resulted in a phytase variant with
12 °C improved melting temperature and 3.5-fold increased gastric resistance; the
improved resistance toward gastric proteases might be due to reduced flexibility in
the loops and random coils [33]. In a semi-rational phytase design, a consensus
phytase designed using a sequence alignment of 13 fungal phytases showed
15–22 °C improved melting temperature. A loop that was unresolved (non-visible
electron density map) in the Aspergillus niger phytase was clearly seen in the con-
sensus phytase suggesting the substitution (V251D) in the improved variant stabi-
lized this loop by an increased number of hydrogen bonds with the neighboring
residues [24]. These studies suggest employing an evolution approach targeting
flexible loops may also enrich the library population and reduce the screening
efforts in the directed phytase evolution. Restricting the sequence space by sam-
pling only a part of the protein sequence might be a good strategy for rapid phytase
evolution.
6.3.1.3 Ionic and Hydrophobic Interactions

Ionic interactions also known as salt bridge interactions are associated with thermo-
stability; the correlation between the number of salt bridge and thermostability was
considered not to be as strong as hydrogen bonds [48]. However, recently several
studies have established a good correlation, showing that the increase in the number
of salt bridges improves protein thermostability (7 °C increased Tm) [49].
Incorporation of new salt bridge interactions (Q191R & T271R) in A. niger phytase
based on a homolog A. fumigatus structure as a template improved thermostability
(13% increased residual activity at 100 °C/10 min, for a triple mutant A58E, Q191R,
and T271R) [37]. Using a rational approach based on MD simulations, a phytase
from E. coli (AppA) was designed to introduce new salt bridge interactions. Variant
Q206E and Y311K showed 8–9% improved thermostability [40]. Screening of a
transversion-biased mutant library (SeSaM) of Ymphytase yielded a variant M6
with 3 °C increased melting temperature; substitution T77K introduced a salt bridge
interaction with adjacent helix and turned out to be unobtainable by epPCR [18, 22].
Therefore, employing a mutagenesis method that introduces chemically diverse
amino acid substitutions is crucial for phytase thermostabilization.
Incorporation of hydrophobic interactions has been a barely seen phenomenon in
phytase engineering for improving thermostability. Despite the low occurrences of
aromatic amino acid in protein sequences (may be due to fewer codons), aromatic
interactions showed a thermostabilizing tendency in 166 studied proteins [50]. A
phytase from Y. mollaretii evolved using OmniChange yielded incorporation of
aromatic-aromatic interactions (V298F) that might contribute to improved thermal
and pH resistance [26].
6.3.1.4 Surface Engineering

Role of surface charge-charge interactions of protein to solvent molecules has
received less attention; however, over the past decade, it became evident that the
surface modification plays a vital role in protein stabilization [51]. Recently, a phy-
tase from Y. mollaretii has been engineered with a computer-assisted method enti-
tled ProCoS (protein consensus-based surface engineering). Screening of only 1050
clones from an in vitro recombined library of synthetic genes (designed to incorpo-
rate 30–40 surface substitutions) yielded an Ymphytase variant (harboring 34 sub-
stitutions) with a 3.8-fold increased pH stability (pH 2.8/3 h). The improvement was
achieved by an increased fraction of polar surface, a significant increase in charged
surface (28%), and especially an increase in negative surface (20%) of the Ymphytase
variant. The neutral surface was decreased by 20% when compared to wild-type
Ymphytase [52]. In another study, a surface substitution (Q154H) improved ther-
mostability in the Ymphytase which might be attributed to increased surface expo-
sure to the solvent (decreased intra-protein hydrogen bonds in a MD simulation)
[18]. These studies suggest that phytase surface modifications and reengineering
may provide an alternate route for improving phytase stability.
6.3.2 Improving pH-Activity Profile and pH Stability
Application of phytase in animal feeding requires a decent gastric tolerance.

Stomach pH for many poultry animal ranges between pH 1.9 and pH 4.5 and a gas-
tric emptying time of 1–2 h depending on the diet components [53]. Optimum pH
of the first commercial phytase (phyA) from A. niger was shifted from pH 5.5 to the
more acidic side. Several variants targeting areas near the binding pocket were gen-
erated by swapping residue charges; the pH optima for these variants ranged from
2.5 to 5.5. The best variant E228K showed the pH optimum at 3.8 [54]. Another
phytase from A. niger, phyB, has a pH optimum at 2.5; swapping the charge near
binding pocket (E272K, from negative to positive) shifted the pH optimum to 3.2
that might be due to improved affinity for the negatively charged phytate [55]. These
studies indicate a convincing role of charged residues and ionization near binding
pocket of phytase that affects substrate binding at various pH values. However, a
bacterial phytase from Yersinia frederiksenii (pH optimum 2.5) was evolved by sub-
stituting a non-charged residue (Ser51) near the binding pocket. The residue Ser51
was found to be dissimilar to the close homologs (pH optimum 4.5) which were
identified using a sequence alignment. Substitutions S51T and S51I shifted the pH
optimum from 2.5 to 4.5 [41], suggesting the facts that a side chain structure can
also contribute to the phytase’s pH optimum. In a directed phytase evolution experi-
ment, substitutions far away from the active site yielded an Ymphytase variant with
twofold improved pH stability (pH 2.8/3 h). The latter novel combination of substi-
tutions was discovered by a multisite saturation mutagenesis (OmniChange) of pre-
viously identified sites from a thermally improved Ymphytase variant [26]. As
aforementioned, a surface engineering of Ymphytase using the ProCoS method
yielded a variant with 3.8-fold increased pH stability by increasing (20%) the
number of negatively charged amino acids on the phytase surface [52]. Therefore,
altering the overall charge by surface engineering and modifying the local environ-
ment near the binding pocket are promising strategies to consider in directed phy-
tase evolution, to alter pH optima, and to improve pH stability.
6.3.3 Enhancing Protease Resistance
Protease resistance in phytases has still not been a completely addressed challenge
in the field of directed phytase evolution. Surface exposed loops in A. fumigatus
phytase were targeted to remove protease cleavage sites; for instance, the substitu-
tion (S126N) yielded an ~eightfold (50 °C/90 min) improved protease resistance
against proteases present in the NW205 culture media used for phytase expression
[56]. A whole-gene randomization (GSSM) on E. coli phytase yielded a thermo-
stable variant that showed a 3.5-fold improved stability in the simulated gastric fluid
(pepsin 3.2 mg/ml, pH 1.2) [33]. Few patented phytases have been reported to pos-
sess a higher protease resistance. A phytase from Hafnia species was evolved for
higher tolerance against trypsin and pepsin [57]. A bacterial phytase (rAppA)
showed 30% increased residual activity after pepsin treatment [58]. Synthetic phy-
tase variants (from Yersinia mollaretii and Hafnia species) designed for higher sta-
bility retained 80–90% residual activity after pepsin treatment (pepsin 30 U/ml,
37 °C/30 min) [59]. Several wild-type phytases have been reported to possess desir-
able resistance against pepsin and trypsin [60]. Phytase from Penicillium species
[35] was found to be resistant to protease (pepsin) treatment; however, the specific
activity is low (32–136 U/mg) when compared to phytases from Enterobacteriaceae
family (1000–3000 U/mg) [60]. A bacterial phytase from Malaysian wastewater
was reported to retain ~95% of phytase activity after being treated with pepsin
(3000 U/37 °C/30 min, pH 2.0). This pepsin resistance was found to be equivalent
to the pepsin resistance of E. coli phytase [61]. Recently, phytases from Yersinia
species have been engineered for higher pepsin resistance by incorporating
N-glycosylation sites that obstructed pepsin accessibility to peptic cleavage sites
[62]. On the contrary, a phytase (Nov9X) was rationally designed to incorporate
protease cleavage sites (in the surface exposed loops) that generated protease sus-
ceptible phytase variants [63], suggesting a possibility to design phytases for
improved protease resistance by removing protease cleavage sites on the phytase
surface especially in loop regions.
6.3.4 C
ombining Key Substitutions Identified in Directed
Phytase Evolution
It has always been a major challenge to identify substitutions that contribute to the
desired property improvement in a directed evolution campaign. In the pool of
improved variants, the number of substitutions that will lead to the desired property
improvement is unknown. The latter unpredictability intensifies and renders it even
more challenging to identify cooperative effects (epistatic effects) between distantly

located substitutions. Therefore, implementing a strategy to combine substitutions
identified in iterative rounds of directed phytase evolution is a prerequisite for effi-
cient and rapid evolution of phytases. Presently, the additive or synergistic effects
are impossible to predict. Several substitutions identified in directed phytase evolu-
tion demonstrated a reliable proof on the second pathway of directed evolution in
which the epistatic effects play an important role (see Sect. 6.2.3). KeySIDE (itera-
tive key-residues interrogation of the wild type with substitutions identified in
directed evolution) was recently reported to iteratively combine substitutions that
were identified in directed Ymphytase evolution [18]. G187S mutation was identi-
fied as one of the substitution (out of five, D52N, T77K, K139E, G187S, V298M)
in the first round of directed evolution that improved thermostability of Ymphytase.
Incorporation of G187S in the wild-type phytase decreased activity (by ~50%) as
well as thermostability (20% less than wild type). However, when the G187S sub-
stitution was combined with T77K, the thermostability of the M2 variant (G187S-
T77K) was better than T77K alone, and the specific activity was restored to
wild-type phytase. Another substitution Q154H improved thermostability when
combined with T77K (20% improved residual activity) or incorporated into the wild
type (16% improved residual activity), without altering specific activity, whereas
combining Q154H to the M2 variant (G187S-T77K) improved specific activity by
~200 U/mg. These results suggest that cooperative effects play an important role in
modulating phytase properties. This is in evidence with the study employing site
saturation mutagenesis of all individual position in E. coli phytase and combining
substitutions that improved stability. Incorporation of two stabilizing substitutions
led to lower stability for the double mutant than that of single substitution. Therefore,
substitutions were combined by adding each substitution to the most stable variant
identified in the previous round of site-directed mutagenesis [33]. However, this
approach is limited by the possibility of missing combinations that act cooperatively
by ignoring detrimental substitution (e.g., G187S was detrimental substitution for
Ymphytase but improved thermostability in combination with T77K). Therefore,
combining substitutions iteratively by the KeySIDE approach to decipher the role of
each substitution and the role of possible combinations is a preferable approach to
rapidly and efficiently improve phytase properties.
6.3.5 Screening Methods for Directed Phytase Evolution
Successful directed evolution experiments rely on a robust high-throughput screen-

ing systems. Prescreening of a mutant library using bacterial colonies has been a
desirable option to reduce screening efforts in directed evolution experiments.
High-throughput screening methods developed for bacterial colonies were success-
ful to select only active or thermostable phytase variants for subsequent screening
in 96-well microtiter plates. A widely employed synthetic substrate in directed phy-
tase evolution, 4-methylumbelliferyl phosphate (4-MUP), has been one of the most
reliable substrates used in prescreening. A phytase from Y. mollaretii was evolved
by developing a 384-well microtiter plate-based prescreening method to select

thermostable variants (increased Tm by 3 °C). In this method, mutant libraries were
expressed in an auto-induction media [64] which omit steps involved in IPTG
induction. Further, bacterial colonies expressing Ymphytase variants were directly
subjected to heat treatment, omitting protein expression and lysis steps [22]. A sim-
ple agar plate assay employing 1-naphthyl phosphate and Fast Garnet salt was used
to identify the Klebsiella phytase from soil samples [65]. A filter paper-based assay
for “colony screening” was also developed for screening mutant libraries employing
the natural/application substrate (phytate) [22, 66]. The use of natural/application
substrate is always desirable to avoid evolving enzymes toward nonnatural substrate
[67]. Recently, an ultrahigh-throughput screening method (fur-shell technology)
employing fluorescence-activated cell sorting (FACS) was developed as a pre-
screening system for phytases. A coupled reaction of Ymphytase and glucose oxi-
dase produces a hydroxyl radicle by utilizing glucose-6-phosphate as a substrate.
The hydroxyl radicle initiates a poly(ethylene-glycol)-acrylate-based polymeriza-
tion incorporating a fluorescent molecule during polymerization and ultimately
forming a fluorescent hydrogel shell surrounding the E. coli that expresses an active
Ymphytase. Further, screening of ~10 million Ymphytase variants validated the fur-
shell technology by identification of an Ymphytase variant with improved specific
activity (increased by 97 U/mg) [68]. Screening conditions optimized for 96-well
microtiter plate-based screening platform employing either phytate or 4-MUP as a
substrate showed reliable true standard deviations (below 15%) [22]. These advances
in the prescreening and screening methods facilitate a rapid, efficient, and non-
laborious identification of novel phytase variants with desirable properties.
6.3.6 Expression Host for Phytase Engineering and Production
Host organism used for phytase expression exhibits an immense influence on phy-
tase activity and thermostability mainly due to posttranslational modifications.
Glycosylation has been considered an important feature for altering activity or ther-
mostability. Elimination of N-glycosylation sites (N204A or S206A) close to the
active site of E. coli phytase improved activity by reducing the steric hindrance for
phytate entry into the binding pocket [34]. A detailed study to alter the glycosyl-
ation pattern in E. coli phytase (AppA) resulted in an identification of a phytase
variant (C200N, D207N, S211N) with improved specific activity (65 U/mg) when
compared to the wild type (37 U/mg) [43]. Glycosylation has been shown to improve
stability of numerous proteins [69, 70]. However, the variant (C200N, D207N,
S211N) showed 25% higher thermostability (increased residual activity at
80 °C/15 min), despite having the same level of glycosylation. Whereas in another
study, an incorporation of two glycosylation sites in E. coli phytase showed >40%
enhanced thermostability (80 °C/10 min) and 4–5 °C higher melting temperature
[71]. These results suggest the posttranslational modifications to play a vital role in
phytase properties. The E. coli phytase expressed in yeast (Pichia pastoris) [60, 72]
showed ~threefold higher specific activity (~3100 U/mg) when compared to the
expression of the same phytase in E. coli (975 U/mg) [73]. Despite an advantageous

role of yeast in phytase expression, E. coli is a preferred host for directed evolution
due to its faster growth, nonchromosomal integration (as happens when using P.
pastoris), and high plasmid transformation efficiency, which shortens time required
for each round of directed phytase evolution. Saccharomyces cerevisiae as a host
offers an advantage of nonchromosomal integration when using episomal vectors.
However, P. pastoris is a widely adopted host for industrial production of phytases
due to correct posttranslational modifications, very high density cultures, and high
protein yields.
6.4 Structural Perspectives in Phytase Engineering
Rational phytase engineering has become an attractive strategy due to increased

knowledge on structural information and structural dynamics studies. Several stud-
ies have demonstrated a very good correlation between phytase flexibility and activ-
ity or thermostability (Tables 6.1 and 6.2). Rigidifying a protein in part or overall by
improving the intra-protein hydrogen bonding network has proven to be crucial in
improving phytase performance. A direct evidence on structural features or overall
fold of a phytase determining its activity or thermostability does not exist to date.
However, recent improvements in phytases by directed evolution and rational design
strategies have advanced our knowledge on factors affecting phytase properties and
could be employed in designing novel phytases for industrial demands. Several
approaches led to improved phytases ranging from simple sequence alignment to
MD simulation studies (Table 6.2). A fungal consensus phytase designed by
sequence alignment yielded a variant with 15–22 °C increased melting temperature
[24]. A structure-guided consensus sequence was designed using a homology model
for beta-propeller phytase resulted in a variant with ~80% residual activity (at
80 °C/10 min) [38]. Flexibility within the phytase enzyme is a critical parameter; in
detail a local flexibility in the binding pocket determines catalytic efficiency/spe-
cific activity; overall protein flexibility or surface loop flexibility governs thermo-
stability. Fungal and bacterial (Enterobacteriaceae) HAP phytases have a similar
overall fold (α-domain, α/β-domain, and binding pocket at the interface of these two
domains), but the Enterobacteriaceae phytases are ~30-fold more active than fungal
phytases that might be due to local flexibility in the binding pocket. The flexible
active site loop in the binding pocket of Enterobacteriaceae phytases has been
replaced by a less flexible helical structure in the fungal phytase, suggesting that a
local flexibility in the binding pocket is prerequisite for higher specific activity [25].
In addition, conformational flexibility in a Bacillus phytase was restricted by incor-
porating rigidifying substitutions and resulted in 2.5-fold improved free energy that
indicates a significantly more stable variant [44]. Deletion of a flexible C-terminal
part of E. coli phytase increased its residual activity (80 °C/10 min) by 39% [39].
MD simulation studies on a thermostable Y. mollaretii phytase variant discovered by
directed evolution revealed an overall decreased flexibility in the loop regions [18].
A structure-based design of E. coli phytase using MD simulation to target flexible
residues resulted in 30% increased residual activity (80 °C/10 min) [40]. These

studies suggest a strong correlation of phytase structural flexibility with its activity
and thermostability. Additionally, surface charge modifications of phytases are
emerging as a novel strategy for phytase engineering. For instance, surface
engineering of the Ymphytase resulted in a variant with an increased negative polar
surface (20%) that showed 3.8-fold improved pH resistance (pH 2.8/3 h) [52].
Whereas a local surface modification near to the binding site affects phytase activity
and pH optima. A substitution Q396R (positively charged) in A. niger phytase
improved affinity to phytate (negatively charged) [32], and substitution N354D in
Penicillium phytase affected the pKa of the catalytic center that shifted the pH
optima [35].
6.5 onclusions and Future Prospective for Phytases

C
Engineering
Although many phytases have been discovered from nature in the last century, the
potential to improve phytases for industrial applications is barely explored and
especially potentials of epistatic effects are barely realized. Nevertheless, directed
evolution and rational design methodologies provide an excellent tool to tailor-
made phytases with high specific activity, thermal and pH profile, and protease
resistance. Epistatic effects played an important role in the directed phytase evolu-
tion; interrogation of substitutions identified in directed evolution and rationally
combining these substitutions may provide an alternate and rapid route for the
uphill walk on the phytase fitness landscape. Phytase engineering for improved
thermostability revealed the following parameters for efficient improvements: (i)
the higher the chemical diversity in the mutant libraries the better, (ii) target flexible
loops away from the binding pocket, (iii) perform surface engineering, and (iv) alter
glycosylation pattern. Furthermore, surface charge modifications have a great influ-
ence on the pH optima and pH stability of phytases, suggesting an attractive route
for phytase stabilization. Optimal pH could be altered by charge modification near
the binding pocket, whereas overall charge modification may improve pH stability.
Elimination of potential protease-cleaving sites and stabilization of surface loops
represent additional strategies to improve gastric tolerance.
References
1. Bohn L, Meyer AS, Rasmussen SK (2008) Phytate: impact on environment and human nutri-
tion. A challenge for molecular breeding. J Zhejiang Univ Sci B 9(3):165–191
2. Harland BF, Oberleas D (1999) Phytase in animal nutrition and waste management. BASF Ref
Man 237–240.
3. Sebastian S, Touchburn SP, Chavez ER (1998) Implications of phytic acid and supplemental
microbial phytase in poultry nutrition: a review. World Peoult Sci J 54(01):27–47
4. Lei XG, Weaver JD, Mullaney E et al (2013) Phytase, a new life for an “old” enzyme. Ann Rev
Anim Biosci 1:283–309
5. Lei XG, Ku PK, Miller ER et al (1993) Supplementing corn-soybean meal diets with microbial
phytase maximizes phytate phosphorus utilization by weanling pigs. J Anim Sci
71(12):3368–3375
6. Lei XG, Ku PK, Miller ER et al (1994) Calcium level affects the efficacy of supplemental
microbial phytase in corn-soybean meal diets of weanling pigs. J Anim Sci 72(1):139–143
7. Lei X, Ku PK, Miller ER et al (1993) Supplemental microbial phytase improves bioavailability
of dietary zinc to weanling pigs. J Nutr 123(6):1117–1123
8. Gilbert N (2009) Environment: the disappearing nutrient. Nature 461(7265):716–718
9. Adeola O, Cowieson AJ (2011) BOARD-INVITED REVIEW: opportunities and challenges in
using exogenous enzymes to improve nonruminant animal production. J Anim Sci
89(10):3189–3218
10. Cowieson A, Cooper R (2010) Introduction to the event and overview of the phytase market.
In: International phytase summit. International Phytase Summit, Washington, DC
11. Meyer AS (2010) Enzyme technology for precision functional food ingredient processes. Ann
N Y Acad Sci 1190:126–132
12. Hubenova Y, Georgiev D, Mitov M (2014) Stable current outputs and phytate degradation by
yeast-based biofuel cell. Yeast 31(9):343–348
13. Fujita J, Fukuda H, Yamane Y-I et al (2001) Critical importance of phytase for yeast growth
and alcohol fermentation in Japanese sake brewing. Biotechnol Lett 23(11):867–871
14. Billington DC (1993) In: Billington DC (ed) The Inositols phosphates: chemical synthesis and
biological significance. VCH Verlagsgesellschaft, Weinheim
15. Fu S, Sun J, Qian L et al (2008) Bacillus phytases: present scenario and future perspectives.
Appl Biochem Biotechnol 151(1):1–8
16. Kim MS, Weaver JD, Lei XG (2008) Assembly of mutations for improving thermostability of
Escherichia coli AppA2 phytase. Appl Microbiol Biotechnol 79(5):751–758
17. Kim MS, Lei XG (2008) Enhancing thermostability of Escherichia coli phytase AppA2 by
error-prone PCR. Appl Microbiol Biotechnol 79(1):69–75
18. Shivange AV, Roccatano D, Schwaneberg U (2016) Iterative key-residues interrogation of a
phytase with thermostability increasing substitutions identified in directed evolution. Appl
19. Wong TS, Roccatano D, Zacharias M et al (2006) A statistical analysis of random mutagenesis
methods used for directed protein evolution. J Mol Biol 355(4):858–871
20. Di Giulio M (2005) The origin of the genetic code: theories and their relationships, a review.
Biosystems 80(2):175–184
21. Zhao J, Kardashliev T, Joelle Ruff A et al (2014) Lessons from diversity of directed evolution
experiments by an analysis of 3,000 mutations. Biotechnol Bioeng 111(12):2380–2389
22. Shivange AV, Serwe A, Dennig A et al (2012) Directed evolution of a highly active Yersinia
mollaretii phytase. Appl Microbiol Biotechnol 95(2):405–418
23. Singh B, Satyanarayana T (2009) Characterization of a HAP-phytase from a thermophilic
mould Sporotrichum thermophile. Bioresour Technol 100(6):2046–2051
24. Lehmann M, Kostrewa D, Wyss M et al (2000) From DNA sequence to improved functional-
ity: using protein sequence comparisons to rapidly design a thermostable consensus phytase.
Protein Eng 13(1):49–57
25. Shivange AV, Schwaneberg U, Roccatano D (2010) Conformational dynamics of active site
loop in Escherichia coli phytase. Biopolymers 93(11):994–1002
26. Shivange AV, Dennig A, Schwaneberg U (2014) Multi-site saturation by OmniChange yields
a pH- and thermally improved phytase. J Biotechnol 170:68–72
27. Salverda ML, Dellus E, Gorter FA et al (2011) Initial mutations direct alternative pathways of
protein evolution. PLoS Genet 7(3):e1001321
28. Chen W, Ye L, Guo F et al (2015) Enhanced activity of an alkaline phytase from Bacillus
subtilis 168 in acidic and neutral environments by directed evolution. Biochem Eng J 98:
137–143
29. Tracewell CA, Arnold FH (2009) Directed enzyme evolution: climbing fitness peaks one
amino acid at a time. Curr Opin Chem Biol 13(1):3–9
30. Chen C-C, Cheng K-J, Ko T-P et al (2015) Current progresses in phytase research: three-
dimensional structure and protein engineering. Chem Biol Eng Rev 2(2):76–86
31. Yao MZ, Zhang YH, Lu WL et al (2012) Phytases: crystal structures, protein engineering and
potential biotechnological applications. J Appl Microbiol 112(1):1–14
32. Liao Y, Zeng M, Wu ZF et al (2012) Improving phytase enzyme activity in a recombinant
phyA mutant phytase from Aspergillus niger N25 by error-prone PCR. Appl Biochem
33. Garrett JB, Kretz KA, O’Donoghue E et al (2004) Enhancing the thermal tolerance and gastric
performance of a microbial phytase for use as a phosphate-mobilizing monogastric-feed sup-
plement. Appl Environ Microbiol 70(5):3041–3046
34. Wu TH, Chen CC, Cheng YS et al (2014) Improving specific activity and thermostability of
Escherichia coli phytase by structure-based rational design. J Biotechnol 175:1–6
35. Zhao Q, Liu H, Zhang Y (2010) Engineering of protease-resistant phytase from Penicillium
sp.: high thermal stability, low optimal temperature and pH. J Biosci Bioeng 110(6):
638–645
36. Zhu W, Qiao D, Huang M et al (2010) Modifying thermostability of appA from Escherichia
coli. Curr Microbiol 61(4):267–273
37. Zhang W, Mullaney EJ, Lei XG (2007) Adopting selected hydrogen bonding and ionic interac-
tions from Aspergillus fumigatus phytase structure improves the thermostability of Aspergillus
niger PhyA phytase. Appl Environ Microbiol 73(9):3069–3076
38. Viader-Salvado JM, Gallegos-Lopez JA, Carreon-Trevino JG et al (2010) Design of thermo-
stable beta-propeller phytases with activity over a broad range of pHs and their overproduction
by Pichia pastoris. Appl Environ Microbiol 76(19):6423–6430
39. Fei B, Cao Y, Xu H et al (2013) AppA C-terminal plays an important role in its thermostability
in Escherichia coli. Curr Microbiol 66(4):374–378
40. Fei B, Xu H, Cao Y et al (2013) A multi-factors rational design strategy for enhancing the
thermostability of Escherichia coli AppA phytase. J Ind Microbiol Biotechnol 40(5):457–464
41. Fu D, Huang H, Meng K et al (2009) Improvement of Yersinia frederiksenii phytase perfor-
mance by a single amino acid substitution. Biotechnol Bioeng 103(5):857–864
42. Tran TT, Mamo G, Buxo L et al (2011) Site-directed mutagenesis of an alkaline phytase: influ-
encing specificity, activity and stability in acidic milieu. Enzym Microb Technol 49(2):
177–182
43. Rodriguez E, Wood ZA, Karplus PA et al (2000) Site-directed mutagenesis improves catalytic
efficiency and thermostability of Escherichia coli pH 2.5 acid phosphatase/phytase expressed
in Pichia pastoris. Arch Biochem Biophys 382(1):105–112
44. Tung ET, Ma HW, Cheng C et al (2008) Stabilization of beta-propeller phytase by introducing
Xaa-->Pro and Gly-->Ala substitutions at consensus positions. Protein Pept Lett
15(3):297–299
45. Bei J, Chen Z, Fu J et al (2009) Structure-based fragment shuffling of two fungal phytases for
combination of desirable properties. J Biotechnol 139(2):186–193
46. Xu W, Shao R, Wang Z et al (2015) Improving the neutral phytase activity from Bacillus amy-
loliquefaciens DSM 1061 by site-directed mutagenesis. Appl Biochem Biotechnol
175(6):3184–3194
47. Vogt G, Argos P (1997) Protein thermal stability: hydrogen bonds or internal packing? Fold
Des 2(4):S40–S46
48. Vogt G, Woell S, Argos P (1997) Protein thermal stability, hydrogen bonds, and ion pairs.
J Mol Biol 269(4):631–643
49. Karshikoff A, Nilsson L, Ladenstein R (2015) Rigidity versus flexibility: the dilemma of
understanding protein thermal stability. FEBS J 282(20):3899–3917
50. Folch B, Dehouck Y, Rooman M (2010) Thermo- and mesostabilizing protein interactions
identified by temperature-dependent statistical potentials. Biophys J 98(4):667–677
51. Strickler SS, Gribenko AV, Gribenko AV et al (2006) Protein stability and surface electrostat-
ics: a charged relationship. Biochemistry 45(9):2761–2766
52. Shivange AV, Hoeffken W, Haefner S, Schwaneberg U (2016). Protein consensus based sur-
face engineering (ProCoS): a computer-assisted method for directed protein evolution.
Biotechniques 61(6):305–314
53. Svihus B (2014) Function of the digestive system. J Appl Poult Res 23(2):306–314
54. Kim T, Mullaney EJ, Porres JM et al (2006) Shifting the pH profile of Aspergillus niger PhyA
phytase to match the stomach pH enhances its effectiveness as an animal feed additive. Appl
55. Weaver JD, Mullaney EJ, Lei XG (2007) Altering the substrate specificity site of Aspergillus
niger PhyB shifts the pH optimum to pH 3.2. Appl Microbiol Biotechnol 76(1):117–122
56. Wyss M, Pasamontes L, Friedlein A et al (1999) Biophysical characterization of fungal phy-
tases (myo-inositol hexakisphosphate phosphohydrolases): molecular size, glycosylation pat-
tern, and engineering of proteolytic resistance. Appl Environ Microbiol 65(2):359–366
57. Lassen SF, De Maria L, Friis EP et al. (2012) Hafnia phytase variants. (US20120225468)
Novozymes A/S
58. Lei X (2003) Enzymes with improved phytase activity. (US6511699 B1) Cornell Research
Foundation, Inc
59. Haefner S, Welzel A, and Thummer R (2014) Synthetic phytase variants. (US20140044835)
BASF SE
60. Huang H, Luo H, Wang Y et al (2008) A novel phytase from Yersinia rohdei with high phytate
hydrolysis activity under low pH and strong pepsin conditions. Appl Microbiol Biotechnol
80(3):417–426
61. Greiner R, Farouk AE (2007) Purification and characterization of a bacterial phytase whose
properties make it exceptionally useful as a feed supplement. Protein J 26(7):467–474
62. Niu C, Luo H, Shi P et al (2015) N-glycosylation improves the pepsin resistance of HAP phy-
tases by enhancing the stability at acidic pH and reducing the pepsin accessibility to peptic
cleavage sites. Appl Environ Microbiol 82:1004–1014
63. Basu SS and Zhang S (2010) Engineering enzymatically susceptible proteins. (US20100273198
A1) Syngenta Participations Ag
64. Studier FW (2005) Protein production by auto-induction in high density shaking cultures.
Protein Expr Purif 41(1):207–234
65. Sajidan A, Farouk A, Greiner R et al (2004) Molecular and physiological characterisation of a
3-phytase from soil bacterium Klebsiella sp. ASR1. Appl Microbiol Biotechnol 65(1):
110–118
66. Senn AM, Wolosiuk RA (2005) A high-throughput screening for phosphatases using specific
substrates. Anal Biochem 339(1):150–156
67. Aharoni A, Thieme K, Chiu CP et al (2006) High-throughput screening methodology for the
directed evolution of glycosyltransferases. Nat Methods 3(8):609–614
68. Pitzler C, Wirtz G, Vojcic L et al (2014) A fluorescent hydrogel-based flow cytometry high-
throughput screening platform for hydrolytic enzymes. Chem Biol 21(12):1733–1742
69. Mitra N, Sinha S, Ramya TN et al (2006) N-linked oligosaccharides as outfitters for glycopro-
tein folding, form and function. Trends Biochem Sci 31(3):156–163
70. Sola RJ, Griebenow K (2009) Effects of glycosylation on the stability of protein pharmaceuti-
cals. J Pharm Sci 98(4):1223–1245
71. Yao MZ, Wang X, Wang W et al (2013) Improving the thermostability of Escherichia coli
phytase, appA, by enhancement of glycosylation. Biotechnol Lett 35(10):1669–1676
72. Huang H, Luo H, Yang P et al (2006) A novel phytase with preferable characteristics from
Yersinia intermedia. Biochem Biophys Res Commun 350(4):884–889
73. Miksch G, Kleist S, Friehs K et al (2002) Overexpression of the phytase from Escherichia coli
and its extracellular production in bioreactors. Appl Microbiol Biotechnol 59(6):685–694
74. Suzuki U, Yoshimura K, Takaishi M (1907) Über ein enzym ‘Phytase’ das anhydro-oxy-
methylen diphosphorsaure’ spalter. Tokyo Imp Univ Coll Agric Bull 7:503–512
Strain Development by Whole-Cell
Directed Evolution 7
Tong Si, Jiazhang Lian, and Huimin Zhao
Abstract
Due to limited knowledge of complicated cellular networks, directed evolution
has played critical roles in strain improvement, especially for complex traits with
hundreds of genetic determinants and for organisms with few genetic tools.
Directed evolution mimics natural evolution in the laboratory via iterative cycles
of diversity generation and functional selection or screening to isolate evolved
mutants with desirable phenotypes. In this chapter, we summarize recent techno-
logical advances and applications of directed evolution in strain development,
focusing on the efforts for accelerating evolution workflows, expanding the range
of target phenotypes, and facilitating mechanistic understanding of evolved
mutations.
Tong Si and Jiazhang Lian contributed equally to this work

$
T. Si
Carl R Woese Institute for Genomic Biology, University of Illinois at Urbana-Champaign,
1206 W Gregory Dr, Urbana, IL 61801, USA
J. Lian
Department of Chemical and Biomolecular Engineering, University of Illinois at Urbana-
Champaign, 600 South Mathews Ave, Urbana, IL 61801, USA
H. Zhao (*)
Department of Chemical and Biomolecular Engineering, University of Illinois at Urbana-
Champaign, 600 South Mathews Ave, Urbana, IL 61801, USA
Departments of Chemistry, Biochemistry, and Bioengineering, University of Illinois at
Urbana-Champaign, 600 South Mathews Ave, Urbana, IL 61801, USA

DOI 10.1007/978-3-319-50413-1_7
174 T. Si et al.
7.1 Introduction
Metabolic engineering aims at “the improvement of cellular activities by manipula-

tion of enzymatic, transport, and regulatory functions of the cell” [8] and has been
successfully applied to create efficient microbial cell factories for producing fuels,
drugs, and value-added chemicals [59, 80, 83, 102, 117]. There are two fundamental
approaches for metabolic engineering: rational design and directed evolution. For
rational design, metabolic bottlenecks are identified and addressed via directed
genetic manipulations for improved production of a target molecule [8, 9]. However,
this approach is associated with four major limitations. First, effective rational
designs often require comprehensive understanding of metabolic pathways and cel-
lular metabolism, which is incomplete most of time. In particular, there is limited
knowledge of some complex phenotypes, such as stress tolerance and chemical
inhibitor resistance [122, 161]. Second, designed mutations may result in unex-
pected metabolic outcomes, due to the complicated nature of biological networks.
Third, genetic manipulation is generally difficult in many industrial strains. Finally,
there are regulatory issues that limit the use of genetically modified organisms
(GMOs) with rationally designed mutations in certain applications, such as the food
industry [18]. On the contrary, directed evolution mimics natural evolution in labo-
ratory settings, where iterative rounds of genetic diversity generation and pheno-
typic selection or screening are performed to isolate strains with improved traits
(Fig. 7.1) [18, 32, 68, 120, 150, 159]. Whole-cell directed evolution, also known as
evolutionary engineering, can overcome the aforementioned limitations of rational
design. It requires no prior knowledge on genotype-phenotype relationships and
achieves relatively defined phenotypic outcomes dependent on selection/screening
schemes. Mutant strains can be generated using spontaneous mutations without
genetic manipulations and hence considered as non-GMOs for higher public accep-
tance [18].
Adaptive evolution (AE) is the most widely used and well-established approach
for whole-cell directed evolution [32, 159]. In a typical AE experiment, a growing
asexual culture is maintained for a prolonged period of time via serial transfer in a
batch format or continuous dilution in a chemostat (see Sect. 7.2.2 for details).
Strain variants arise due to spontaneous mutations. Under a certain selection pres-
sure, variants with improved competitive fitness outgrow the parental and less-fit
sibling populations (Fig. 7.1). Genotypes of evolved strains are analyzed at indi-
vidual or population levels for understanding the molecular mechanisms conferring
improved phenotypes.
Although successfully applied in improving a wide range of industrially impor-
tant traits in microbial strains, AE suffers from various limitations, including low
rates and relatively restricted spectrums of mutations, prolonged cultivation, limited
ranges of target phenotypes, and challenges to differentiate beneficial mutants from
neutral to deleterious hitchhikers. In this chapter, we aim to provide an update on
the technical advances in the application of directed evolution for strain develop-
ment, and different methods are organized according to how they overcome various
limitations of the classical AE (Fig. 7.1). We focus on practical considerations on
7 Strain Development by Whole-Cell Directed Evolution 175
Fig. 7.1 Overview of directed evolution processes consisting of iterative cycles of diversity gen-
eration and functional screening or selection. For the classical AE, diversity is generated by spon-
taneous mutations (thin, solid arrow). During prolonged cultivation under selection pressures,
mutants with improved competitive fitness are enriched (thick, closed arrow). New methods are
developed to introduce systematic and multiplex mutations (thin, dashed arrow) for accelerating
evolution and to perform high-throughput screening (thick, open arrow) for engineering traits that
are not linked to growth. Genetic mutations are then analyzed using next-generation sequencing
(NGS)
the choice of appropriate techniques (Table 7.1). We then discuss recent examples
of directed evolution in improving various cellular properties and provide prospec-
tive on the future directions.
7.2 ecent Technology Advances in the Application

R
of Directed Evolution for Strain Development
7.2.1 Diversity Generation
Classical AE experiments depend on spontaneous mutations, primarily in the form

of single-base mutations and at a rate of 10−10–10−9 substitutions per base pair per
replication [76, 78, 88, 143]. Other types of mutations, such as insertion and dele-
tion of one or a few bases, gene duplication, and chromosomal rearrangement, occur
at even lower frequencies [93]. The rare and uncontrollable nature of naturally
occurring mutations renders AE as a very time-consuming process, often requiring
tens of thousands of replicative generations. Furthermore, the limited types of spon-
taneous mutations (mostly single-base substitutions) [93] exhibit modest impacts
on competitive fitness [45]. These relatively small steps of evolutionary paths on a
fitness landscape are prone to be trapped at local optimum [150]. To overcome these
obstacles, new methods are developed to increase random mutation rates, introduce
176
Table 7.1 Comparison of mutagenesis methods for directed evolution in strain development
Need Need
genetic genome Mutational Systematic Multiplex
Mutagenesis methods Host range tool sequence spectrum mapping mutation
Adaptive evolution Prokaryote/eukaryote No No Mostly single- No Yes
base substitution
Chemical/radiation mutagenesis Prokaryote/eukaryote No No Diverse No Yes
Mutator strain Prokaryote/eukaryote Yes No Diverse No Yes
Transposon insertion Prokaryote/eukaryote Yes No Mostly knockout Yes No
RNA interferencea Eukaryote Yes No Knockdown Yes Yes
Recombineering Prokaryoteb Yes Yes Diverse Yes Yes
CRISPR-Cas Prokaryote/eukaryote Yes Yes Diversec Yes Yes
Transcription/translation factor Prokaryote/eukaryote Yes No Diverse No Yes
engineering
Genome shuffling Prokaryote/eukaryote Yes No Diverse No Yes
a
Other regulatory RNA mechanisms, such as small RNAs and antisense RNAs, can also be used for directed evolution (see [133] for a summary)
b
The efficiency of oligonucleotide-mediated recombineering is not high enough for practical applications in yeast [28]
c
In addition to gene knockout, which was primarily discussed in this chapter, genetic activation and repression can also be enabled using CRISPR-Cas (see
[133] for a summary)
T. Si et al.
a b
c d
e
f
Fig. 7.2 Schemes of selected methods for diversity generation. (a) Recombineering. (b) CRISPR-
Cas. (c) RNA interference. (d) Transposon insertional mutagenesis. (e) Global transcription and
translation machinery engineering. (f) Genome shuffling
systematic perturbations, and create large-scale and multiplex genetic diversities

(Figs. 7.1 and 7.2). A comparison of various diversity generation methods is sum-
marized in Table 7.1.
Increasing random mutagenesis rates

Increasing random mutagenesis rates is proposed to improve the probability of gen-
erating beneficial mutations and accelerate the evolutionary processes [159].
Chemical mutagens and radiation are widely used to induce DNA damages that lead
to genetic mutagenesis [123]. In addition, mutator strains with impaired DNA rep-
lication or repair systems are employed to enhance genetic diversity [44, 125].
However, increased mutagenesis rates may also result in more frequent occurrence
of deleterious mutations in already adapted strains, rendering the fixation of benefi-
cial mutations difficult in an evolving population. To solve this dilemma, dynamic
control of mutagenesis rates is proposed, whereby mutagenesis rates are raised tran-
siently and then decreased to normal levels after adaptation is achieved. In one
example, GREACE (Genome Replication Engineering Assisted Continuous
Evolution) introduced defective components of a DNA polymerase complex on a
plasmid, converting the target strain into a mutator [92]. After evolution, the plas-
mid was cured to return to normal mutation rates, allowing for isolation of stable
strains for analysis. In another example, FREP (feedback-regulated evolution of
178 T. Si et al.
phenotype) inversely coupled the mutagenesis rates with the concentrations of a

target metabolite [23]. In particular, a genetic sensor activated expression of a proof-
reading exonuclease mutant at low target metabolite concentrations, and the activa-
tion effect diminished with increasing production. Application of this strategy
successfully identified mutant strains with improved production of tyrosine and iso-
pentenyl diphosphate [23]. Notably, the constant or dynamic use of mutator pheno-
types is limited to organisms with known DNA replication and repair mechanisms.
Controlling mutagenesis rates in less-characterized strains may be achieved by
varying the dosage of chemical and physical mutagens prior to or during
evolution.
Creating genome-scale diversity

Genome-wide strain libraries prove to be instrumental for studying genotype-
phenotype relationships [19, 134]. Compared with AE that generates random muta-
tions in limited numbers of genes, strain libraries allow systematic investigation on
nearly every gain-/reduction-/loss-of-function mutation within a target genome [7,
42, 50, 94]. Also, it is easier to interpret screening results using single-mutation
strain libraries, compared with hundreds of changes observed in evolved genomes
from AE experiments. However, conventional methods require each mutant strain to
be created individually in an arrayed manner [7, 42], which is prohibitively tedious
and costly for iterative screening in directed evolution. The main technical hurdle is
the low efficiency of homologous recombination (HR) for allelic replacement on a
genome.
To stimulate HR and accelerate library construction, two cellular mechanisms
have been utilized. Using bacteriophage-based recombination systems, recom-
bineering (recombination-based genetic engineering) enables efficient genome edit-
ing with single-stranded DNA oligonucleotides as homologous donors (Fig. 7.2a)
[34, 127]. For example, combined with microarray-based DNA synthesis, promoter
replacement of every gene in Escherichia coli was achieved using a pool of DNA
oligonucleotides in a single round of transformation [157]. Furthermore, HR effi-
ciency can be enhanced via introduction of double-stranded breaks at targeted
genomic loci, using programmable DNA nucleases derived from ZFPs (zinc finger
proteins), TALEs (transcription activator-like effectors), and CRISPR-Cas (clus-
tered regularly interspaced short palindromic repeats and CRISPR-associated pro-
teins) [36, 134]. In particular, recognition of a target DNA sequence by the
CRISPR-Cas nuclease Cas9 is mediated by a trans-acting guide RNA (gRNA) via
Watson-Crick base pairing (Fig. 7.2b) [25, 96]. Using gRNA expression libraries
synthesized by DNA microarray (Fig. 7.2b), genome-scale screening has been dem-
onstrated in human cells [154], and similar approaches should be readily applied in
engineering microbial systems.
HR-independent methods are also developed for constructing strain libraries.
First, trans-acting RNAs have been employed for facile introduction of genome-
wide perturbation (Fig. 7.2c) [133]. For example, the RNA interference (RNAi)
machinery has been reconstituted in Saccharomyces cerevisiae for genome-scale

knockdown screening. This RNAi-assisted genome evolution (RAGE) strategy
identified mutations conferring resistance towards fermentation inhibitors [133,
163]. Second, genome-scale insertion mutagenesis can be conducted using transpo-
sons, which are a class of mobile genetic elements (Fig. 7.2d) [2, 55]. For positive
mutants exhibiting improved fitness, the inserted transposon sequence can act as a
tag to identify insertion positions and affected genes. Most transposition events lead
to gene disruption, but may also enable gene activation if equipped with outward-
oriented promoters [123].
Notably, CRISPR-Cas and RNAi may be particularly useful for creating libraries
of industrial strains, many of which are polyploidy. CRISPR-Cas has been used for
simultaneous disruption of two alleles of a gene in several industrial yeast strains
[142], which is very challenging using traditional methods. In addition, RNAi-based
gene silencing targets mRNA transcripts and hence can regulate gene expression
without modifying multiple copies of the same gene.
Introducing large-scale and multiplex mutations

Given the positive correlation between the probability of finding improved mutants
and the degree of genetic diversity [66], it can be beneficial to introduce large-scale
and combinatorial mutations to a genome, especially when targeting complex phe-
notypes (e.g., tolerance) that have many genetic determinants [122, 161]. Although
AE is capable of accumulating hundreds of mutations, it requires prolonged cultiva-
tion and generates limited types of mutations. Therefore, it is desirable to develop
new methods for generating many mutations of diverse types in a short period of
time.
One strategy is to introduce mutations in cellular components that are involved
in transcription or translation processes (Fig. 7.2e). For example, gTME (global
transcriptional machinery engineering) introduced random mutations to master
transcription factors (TFs) via error-prone PCR [4, 5], altering expression levels of
hundreds of genes. Large perturbation in transcriptomic profiles can also be enabled
upon incorporation of artificial TFs containing tandem repeats of ZFPs [109, 110].
In addition, screening with lethal concentrations of ribosome-targeting antibiotics
can isolate genetic mutations in ribosomal components. Ribosomal mutations result
in perturbed proteomes for emergence of improved cellular phenotypes [53, 106,
130, 136].
Another strategy is to create multiplex mutations for combinatorial diversity.
Three aforementioned mechanisms—recombineering, CRISPR-Cas, and trans-
acting regulatory RNAs— are also widely applied for this purpose. Mediated by the
ssDNA-binding protein (β) from the λ-red bacteriophage, oligonucleotide pools
targeting the RBSs of 24 genes were transformed recursively into E. coli for allelic
replacement, creating over 4.3 billion combinatorial genomic variants per day [152].
Using CRISPR-Cas, multiplex gene disruption or integration can be achieved with
high efficiency, whereby multiple targeting gRNAs and HR donors are introduced
180 T. Si et al.
in a single cell [10, 57, 58, 129]. Moreover, iterative rounds of RNAi screening
result in the accumulation of beneficial knockdown mutations that act synergisti-
cally to improve acetic acid tolerance in S. cerevisiae [133].
Large-scale chromosomal rearrangement is another useful source for increasing
diversity. First, genome shuffling promotes HR between genomes following proto-
plast fusion, sexual mating, and transformation of whole-genome fragments
(Fig. 7.2f), generating combinations of mutations from a genetically diverse collec-
tion of parental genomes [12]. Genetic diversity among parents can be derived from
directed evolution or natural sources such as different strains of the same species or
even different species [12]. Second, targeted genome deletion can be achieved
following a generation of DNA breaks (DSBs or nicks) at two genomic loci by
CRISPR-Cas [24, 140]. Furthermore, during construction of a synthetic chromo-
some arm in yeast, a number of LoxPsym sequences were inserted after the stop
codon of every nonessential gene or near major genomic landmarks such as repeti-
tive sequences or telomeres [33]. With this SCRaMbLE (synthetic chromosome
rearrangement and modification by LoxP-mediated evolution) design, recombina-
tion occurred randomly between two LoxPsym sequences to produce inversions or
deletions, resulting in formation of structurally distinct genomes [33, 128].
7.2.2 Selection and Screening
High-throughput selection and screening are critical in isolating mutants with

improved phenotypes following diversity generation (Fig. 7.1). For classical AE
experiments, evolving populations are maintained via serial or continuous dilution
[45, 123]. For serial dilution in batch cultures, transfer is typically conducted during
the exponential phase prior to the onset of the stationary phase [148], and cells
experience fluctuating selection pressures due to ever-changing concentrations of
nutrients and cells. On the contrary, a steady-state cell culture is kept in a chemostat,
whereby fresh medium addition and culture removal are performed at defined rates
[105]. The cell density is determined by a limiting nutrient of a defined medium
[171], which acts as a constant selection pressure in a chemostat.
While readers are directed to a recent review for comprehensive comparison
between serial transfer and chemostat [45], we would like to discuss some practical
considerations based on the different types of selection pressures. It is arguably
easier to interpret the genetic basis of adaptation in a chemostat, thanks to the con-
sistency of selection pressures, whereas heterogeneous selection resulting from
dynamic environments in batch cultures renders explanation of functional causality
difficult. For example, competitive fitness in batch cultures may result from reduced
duration in the lag phase, increased growth rates during the exponential phase, or
enhanced capability to divide in the stationary phase [84]. As a result, entangled
mechanisms make interpretation of adaptive mutations very challenging for batch
selection [144]. However, constant selection pressures in a chemostat may be prob-
lematic, as isolated mutants may perform poorly in a different condition, a
phenomenon known as overfitting. For example, robust stress responses allow

microbes to enter a quiescent state for long-term survival upon starvation (e.g., a
limiting nutrient in a chemostat). While this capacity is beneficial in natural or
industrial environments that fluctuate in nutrient availability, activation of
quiescence stops proliferation and confers strong disadvantages in a chemostat.
Therefore, loss of stress response pathways is repeatedly observed in chemostat
selection [95, 104], rendering isolated mutants unsuitable in an industrial setting
due to reduced robustness. Taken together, serial transfer in batch cultures may be
preferred to isolate mutant strains for practical applications, whereas continuous
cultivation in a chemostat is more suitable for studying genetic determinants of a
target phenotype.
Both batch and continuous selections are based on competitive fitness and there-
fore useful to engineer traits that are related to growth or survival, such as utilization
of unnatural feedstock substrates, tolerance towards harsh industrial conditions, and
resistance to high concentrations of substrates, products, or inhibitors. However,
target molecule production is generally not linked to fitness. To screen for enhanced
productivity, three major strategies have been devised. First, product formation can
be coupled with growth advantages by improving redox balancing [39, 139],
enhancing resistance to toxic metabolite analogs [137], increasing tolerance to oxi-
dative stress [118], or rescuing engineered auxotrophy [6, 15]. Second, colorimetric
or fluorometric assays can be developed based on either the optical properties of
target compounds or chemical and enzymatic conversions linking concentrations to
spectrum signals [29]. Third, riboswitch or TF-based metabolite-sensing modules
can be used to link molecular concentrations to the abundance of a fluorescence
protein or an essential metabolite [29, 134], and strain libraries harboring these
genetic biosensors can then be subjected to screening or selection [30, 86, 98, 153].
In addition to expand the range of phenotypes for screening, it is also desirable
to accelerate directed evolution by automation. Currently, manual efforts are
required for culture maintenance, archival storage, contamination tests, selective
pressure adjustment, and phenotype analysis [159]. Inevitable human interventions
limit throughput and reproducibility and introduce subjective bias. To circumvent
these limitations, liquid handling robots were applied for automated batch evolution
in microtiter plates [52, 77]. A microfluidic platform was devised to maintain thou-
sands of microdroplet chemostat systems in parallel [56]. Feedback control systems
were also equipped to monitor cell growth and then dynamically regulate selection
pressures [147]. For these approaches, however, cautions need to be taken on how
transferable the improved phenotypes are during scale-up processes from a microw-
ell or a droplet to a real fermenter.
7.2.3 Mutation Analysis
Thanks to the advances in next-generation sequencing (NGS) (Fig. 7.1), acquired

genetic mutations during AE can be readily analyzed via whole-genome sequencing
(WGS) [11]. However, the main challenge is how to distinguish beneficial
182 T. Si et al.
mutations from neutral to deleterious hitchhikers [26]. Due to the lack of “golden
standard” workflows, it is advisable to integrate information from multiple analyti-
cal schemes.
First, both endpoint and time-course WGS analysis should be performed for indi-
vidual clones and the whole populations [11, 75]. Sequencing of an individual clone
can reveal all mutations in the specific evolved genome [11], but these mutations
represent only a random subset of all genetic variants in a population [65, 74].
Therefore, sequencing of several clones is recommended, but it requires proper
multiplexing techniques for reducing the analysis cost [72]. On the other hand, pop-
ulation sequencing provides a more comprehensive survey on mutation frequency
across different subpopulations, as well as information on evolutionary trajectory if
performed at different stages during evolution [11]. However, sequencing/align-
ment errors may occur during experimental and computational analysis of WGS. In
particular, it is more technically challenging to differentiate low-frequency muta-
tions from sequencing/alignment ambiguity for whole-population sequencing [75].
In general, WGS accuracy can be improved via higher sequencing coverage and
special sequencing techniques including paired-end sequencing [40], circular
sequencing [91], and long-read sequencing [71]. Moreover, time-course whole-
population sequencing can enhance detection of low-frequency alleles in evolving
populations [77]. Given the trade-offs in clonal and population sequencing, it may
be beneficial to combine both to assist mechanistic studies [75].
Second, statistical analysis can help distinguish between adaptive drivers and
neutral or deleterious hitchhikers. Multiple evolution experiments can be
performed in parallel, and mutations appearing in replicate populations are more
likely adaptive [77]. In addition, mutations can be grouped by their functions
(e.g., gene ontology (GO) enrichment analysis), and cellular processes that are
key to a specific trait can be revealed [119, 144]. Furthermore, expected ratios of
synonymous/non-synonymous mutations can be calculated under the assumption
of neutral selection, and underrepresentation of synonymous mutations is indicative
of adaption [11, 13].
Third, other genome-wide analyses should be performed in addition to WGS
[108, 160]. Transcriptomic analysis is necessary due to the difficulty of predicting
the impact of some genetic mutations, especially for noncoding sequences and regu-
latory proteins [47, 73]. For example, a global TF variant led to differential expres-
sion of hundreds of genes, which can only be revealed using transcriptional profiling
[4]. Moreover, proteomics [132] and metabolomics [145, 160] have been applied to
reveal molecular mechanisms of evolved traits, as proteins and metabolites are the
direct actuators of many cellular phenotypes. Given the challenges in interpreting
large-scale omics datasets, it is desirable to develop advanced computational tools
for data integration [38, 121].
Finally, reconstruction of mutations from evolved strains in isolation or in com-
bination in an unevolved ancestor background should provide the most direct obser-
vations on genotype-phenotype relationship. However, this approach is almost
impossible in strains lacking genetic tools. For genetically tractable organisms, it is
also very challenging to reconstitute the multitude of mutations obtained from
AE. To accelerate mutant creation, recombineering and CRISPR coupled with

microarray-based DNA synthesis may be helpful for large-scale and multiplex
introduction of mutations as discussed previously (Sect. 7.2.1). In addition, back-
crossing of isolated mutants with the ancestor via genome shuffling can separate
evolved mutations into a strain collection [11, 116], and progenies exhibiting
improved phenotypes are more likely to harbor causative mutations. To increase
throughput of phenotyping, the use of the molecular bar code technology allows
rapid profiling of competitive fitness of every mutant in a mixed population, whereby
population dynamics can be monitored via frequency quantification of mutation-
associated bar codes using NGS [53, 157]. Aforementioned automation technology
can also be applied to accelerate phenotyping of reconstructed mutants.
7.3 xamples of Directed Evolution for Construction

E
and Optimization of Cell Factories
Directed evolution approaches have been proven to be effective in creating indus-

trial microorganisms with extended substrate ranges, improved cellular properties,
and enhanced production [18, 83, 141, 159]. In this part of the chapter, examples in
using directed evolution approaches for strain improvement in recent years will be
discussed.
7.3.1 Extension of Substrate Utilization
Mostly driven by environmental and energy security considerations, there is a grow-

ing interest in engineering microorganisms for production of fuels and chemicals
from renewable feedstocks, such as lignocellulose and macroalgae. Notably, besides
glucose, these renewable feedstocks contain different sugar components, such as
xylose and arabinose from cellulosic biomass, galactose from red algae, and man-
nitol and 4-deoxy-L-erythro-5-hexoseulose urinate (DEHU) from brown algae. As
substrate utilization can be readily coupled to cellular growth, directed evolution
has been extensively applied to construct efficient fermentation strains with
expanded substrate scopes, especially of S. cerevisiae, which is a preferred cell fac-
tory for many industrial applications.
As the most abundant raw material, lignocellulose has attracted increasing atten-
tion for its conversion to fuels and chemicals. Although hexoses (such as glucose)
can be efficiently fermented by most microorganisms, the utilization of pentoses
(mainly xylose and arabinose), which constitute more than 30% of the total carbo-
hydrate, occurs with a much lower efficiency, even after extensive pathway and
strain engineering. For example, the fungal oxidoreductase pathway containing
xylose reductase (XR), xylitol dehydrogenase (XDH), and xylulose kinase (XKS)
has been introduced into S. cerevisiae to enable xylose fermentation. Unfortunately,
several bottlenecks including low xylose uptake, cofactor imbalance, and limited
metabolic fluxes of the pentose phosphate pathway result in non-optimal xylose
184 T. Si et al.
utilization and biofuel production (Fig. 7.3). Thus, directed evolution has been
applied to construct efficient xylose-fermenting yeast strains [63, 67, 124, 167].
One of the most successful examples was demonstrated by Kim et al. in which serial
transfer in xylose-containing media was performed to construct a yeast strain with
the highest xylose-fermenting capability reported to date [63]. Whole-genome
sequencing of the evolved strains indicated that the loss of function of PHO13
played a dominant role in efficient xylose utilization. Follow-up studies confirmed
that PHO13 deletion induced the activation of pentose phosphate pathway espe-
cially the TAL1 gene encoding the sedoheptulose-7-phosphate:D-glyceraldehyde-3-
phosphate transaldolase [64] and prevented the accumulation of sedoheptulose
[164] (Fig. 7.3). To bypass the cofactor imbalance issue of the fungal xylose utiliza-
tion pathway, researchers have switched their focus to the bacterial xylose isomer-
ase (XI) pathway [27, 82, 115, 151, 169]. While initial attempts were not very
successful, probably due to the low activity and/or poor expression of XI, directed
evolution strategies were adopted to construct yeast strains that could convert xylose
to ethanol at high yields. Besides S. cerevisiae, directed evolution has been applied
Fig. 7.3 Construction of an efficient xylose-fermenting S. cerevisiae strain using directed evolu-
tion. XR xylose reductase, XDH xylitol dehydrogenase, TAL1 sedoheptulose-7-phosphate:D-
glyceraldehyde-3-phosphate transaldolase, X-5-P xylulose-5-phosphate, R-5-P ribose-5-phosphate,
G-3-P glyceraldehyde-3-phosphate, S-7-P sedoheptulose-7-phosphate, E-4-P erythrose-4-
phosphate, F-6-P fructose-6-phosphate
to enable efficient xylose fermentation in other hosts, such as Pichia pastoris [85],
an industrially important cell factory for recombinant protein production.
Recently, the use of marine macroalgae as a renewable feedstock has attracted
increasing attention mainly because it does not require the arable lands and fresh
water, and the absence of lignin makes the depolymerization of seaweed rather sim-
ple and straightforward [156, 158]. Among several types of macroalgae, red algae
and brown algae are considered as ideal sustainable feedstocks for the production of
biofuels and chemicals. The major sugar components are glucose and galactose for
red algae and glucose, mannitol, and DEHU for brown algae. Although many micro-
organisms including S. cerevisiae can ferment galactose, its utilization and the cor-
responding biofuel production are still not efficient enough. Therefore, several
directed evolution efforts have been attempted to improve galactose fermentation
[79, 81]. Lee et al. introduced a genome-wide perturbation library into S. cerevisiae
and isolated fast galactose-fermenting strains. It was found that the overexpression
of the truncated TUP1 gene encoding a global transcriptional repressor resulted in
the most remarkable improvement of galactose fermentation [81]. Another directed
evolution study found that a mutation in the global carbon-sensing Ras/PKA path-
way led to significantly improved galactose fermentation [81]. Both studies high-
lighted the significance of the alteration of global regulatory networks for efficient
galactose fermentation in S. cerevisiae [81]. Recently, Lee et al. also applied directed
evolution to construct S. cerevisiae mutants with enhanced ability to produce bio-
ethanol from both galactose and red algae hydrolysate [79]. On the contrary, the
utilization of brown algae-derived sugars especially DEHU is not well explored. A
synthetic yeast platform for converting brown algae sugars into bioethanol was con-
structed by combining several strategies. The endogenous mannitol transporter and
mannitol-2-dehydrogenase were activated for mannitol utilization. A DEHU trans-
porter and a DEHU reductase were introduced to reduce DEHU to 2-keto-3-deoxy-
D-gluconate (KDG). Finally, a KDG kinase and a KDG-6-phosphate aldolase were
included to enable DEHU fermentation [35]. However, the utilization of DEHU was
poor and only possible under aerobic condition. Then this yeast platform was fur-
ther adapted to grow on mannitol and DEHU under anaerobic condition, which
yielded a yeast strain capable of producing ethanol from mannitol and DEHU with
a titer of 36.2 g/L and 83% of the theoretical yield [35].
Although directed evolution strategies have been extensively used to increase
the fermentation of a single sugar, it is desirable to engineer a platform strain
capable of consuming a mixture of sugars simultaneously to increase fermentation
productivity. Directed evolution of a xylose-fermenting S. cerevisiae strain lacking
the major hexose transporter genes yielded a mutant showing improved growth on
xylose, which was due to the expression of a normally silent HXT11 gene. Further
selection for growth on xylose based on a hexokinase deletion strain at high glu-
cose concentrations resulted in a mutation at N366 of Hxt11p, which reversed the
transporter specificity from glucose into xylose. The Hxt11p mutant was found to
enable efficient co-fermentation of xylose and glucose at industrially relevant
sugar concentrations [131]. Similarly, directed evolution was also performed in a
hexokinase-deficient xylose-fermenting S. cerevisiae strain for growth on xylose in
186 T. Si et al.
the presence of high glucose concentrations, which resulted in a mutation at

N367 in the endogenous chimeric Hxt36p transporter. Using the Hxt36pN367 vari-
ant, efficient co-consumption of glucose and xylose was achieved [103]. Notably,
co-consumption of glucose and xylose was only possible when the endogenous
hexose transporter genes were disrupted. Therefore, more engineering efforts,
including directed evolution, are needed to construct yeast strains capable of co-
fermenting glucose and xylose.
7.3.2 Improvement of Cellular Properties
Construction of robust cell factories with resistance to multiple stresses is highly

desirable due to the harsh conditions in industrial biotechnological processes. The
molecular basis of stress resistance is complicated, making it difficult to construct
multiple stress-resistant strains by rational approaches. On the contrary, directed
evolution has been proven successful in engineering the tolerance to inhibitors in
raw material hydrolysates, final products at high concentrations, and other industrial
harsh conditions.
After pretreatment and depolymerization of the sustainable raw materials, many
undesirable compounds arise in the hydrolysates, such as acetic acid and furfural,
whose concentrations are sufficient to dramatically inhibit the host growth. Tolerance
to these toxic compounds is generally engineered using the classical AE by serial
transfer or continuous culture [48, 99, 162]. The recently developed RAGE method
has been used to increase the tolerance to both acetic acid [133] and furfural [163]
(Fig. 7.4). By introducing a genome-wide RNAi library into a S. cerevisiae strain
followed by iterative rounds of screening under gradually increased stress condi-
tions, Si et al. identified three gene knockdown targets (PTC6, YPR086W, and
tRNAVal(AAC)) that acted synergistically to confer an engineered yeast strain with sub-
stantially improved acetic acid tolerance [133] (Fig. 7.4). Similarly, the same RNAi-
based directed evolution was applied to engineer furfural tolerance, and SIZ1, a
gene encoding the E3 SUMO-protein ligase, was identified as a novel determinant
of furfural tolerance [163] (Fig. 7.4). Besides the resistance to a single inhibitor,
directed evolution has been successfully used to construct cell factories with signifi-
cantly improved growth in the presence of a mixture of inhibitors [22] or even bio-
mass hydrolysates [3, 49, 70, 112–114].
For economically feasible industrial processes, final products are produced at
high concentrations, especially for biofuels and bulk chemicals, which may result in
slower or arrested fermentation. The tolerance of the producing host to the desired
product is one of the determinants in developing a successful biotechnological pro-
cess. Although S. cerevisiae has a long history as the host for ethanol fermentation
and shows the highest ethanol tolerance in nature, its ethanol tolerance can still be
further improved by directed evolution [138, 155]. Snoek et al. developed a large-
scale, robot-assisted genome shuffling strategy to increase the ethanol tolerance of
the industrial S. cerevisiae strains. In their work, a large collection of Saccharomyces
yeasts was characterized in detail, and eight parental strains were chosen for genome
Fragmented gDNA
Integration
Iterative rounds Verification
Cloning
Selection/
Screening
Transformation
Acetic acid Furfural
16 0.25
14 0.5% HAc
Fold Improvement in Biomass
0.7% HAc 0.2

Maximum specific
growth rate (h–1)
12
0.15
10
8 0.1
6
0.05
4
2 0
Ctrl SIZ1-kd GCN4-kd
0
Ctrl PTC6 YPR084W tRNA
Val(AAC)
R3
Fig. 7.4 Engineering acetic acid and furfural tolerance using RNAi-assisted genome evolution
(RAGE) (Reprinted from ACS Synthetic Biology 2015, 4, 283–291, Copyright 2014, with permis-
sion from American Chemical Society and Biotechnology for Biofuel 2014, 7:78)
shuffling, which yielded several novel hybrids outperforming the currently used
industrial yeast strains. To increase the n-butanol tolerance in E. coli, GREACE and
stress-induced mutagenesis-based AEs were developed, both of which yielded
n-butanol-tolerant strains in a short period of time [92, 170]. Ling et al. engineered
the transcription factors related to the pleiotropic drug resistance in S. cerevisiae to
improve the resistance to alkanes [89]. The construction of replacement jet fuel-
tolerant yeast strains using directed evolution has also been reported [14].
It is highly desirable for the industrial fermentation processes to be operated at
high temperatures, so as to reduce cooling costs and prevent contamination. Through
directed evolution, Caspeta et al. obtained several S. cerevisiae strains that demon-
strated much improved growth at a high temperature (>40 °C) [20]. Systematic
characterization of the evolved strains using system biology tools revealed that a
change in sterol composition, from ergosterol to fecosterol, caused by mutations in
188 T. Si et al.
the sterol desaturase gene and increased expression of genes involved in sterol bio-
synthesis, contributed the most to the thermotolerant phenotype. Shui et al. per-
formed a proteomic analysis of the evolved thermotolerant yeast strains, which led
to a comprehensive understanding of the molecular basis of thermotolerance, and
identified novel targets for further improvement [132].
Besides the application in industrial biotechnological processes, directed evolu-
tion approaches are also applied in food biotechnology. For example, in wine mak-
ing, directed evolution has been applied to reduce alcohol levels [146], to increase
the synthesis of aromas [16, 17], and to enhance the fermentation capability at low
temperatures [90].
7.3.3 Enhancement of Product Formation
The formation of a desired product at high titer and yield is the ultimate goal of most
biotechnological processes. Unfortunately, unlike substrate utilization and cellular
tolerance, product formation cannot be easily coupled to cellular growth and even
impairs cellular growth in many cases. In other words, a generally applicable high-
throughput screening strategy of improved production is not readily available.
Metabolic engineering has been proven effective in enhancing the yield of the
desired product, but often at the cost of cellular growth and fitness. In this case,
directed evolution and metabolic engineering can be combined to enhance both the
yield and productivity. The construction of a S. cerevisiae strain with abolished
ethanol production serves as one of such examples (Fig. 7.5). To eliminate ethanol
formation, pyruvate decarboxylases (PDCs) have to be inactivated. Unfortunately,
the Pdc- strain (pdc1Δ pdc5Δ pdc6Δ) is notorious for its inability to grow on glu-
cose as the sole carbon source, requiring the supplementation of a C2 compound
(acetate or ethanol) to synthesize cytosolic acetyl CoA [37] (Fig. 7.5). Several stud-
ies have reported to evolve C2-independent Pdc- yeast strains growing in glucose as
the sole carbon source [61, 149, 168]. Whole-genome sequencing of the evolved
strains revealed that an internal deletion [107] or point mutations (Ala81Pro [61] or
Ala81Asp [168]) in the MTH1 coding sequence enabled the growth of Pdc- strain on
glucose (Fig. 7.5). Alternatively, overexpression of MTH1 or the truncated MTH1
on a multi-copy plasmid resulted in a Pdc- strain with similar properties [87]. The
evolved Pdc- strain was able to accumulate pyruvate to a level as high as 135 g/L
[149], which can be further developed into an important platform cell factory to
produce a wide range of biofuels and value-added chemicals other than ethanol
(Fig. 7.5). Lactate with a titer up to 110 g/L could be obtained in a Pdc- strain
expressing an LDH gene from Lactobacillus casei in 1 L of fermenter under aerobic
conditions [1] (Fig. 7.5). High titer and yield production of 2,3-butanediol was also
reported using an evolved Pdc- strain overexpressing an acetolactate synthase, an
acetolactate decarboxylase, and a butanediol dehydrogenase from various carbon
sources, such as glucose [61], galactose [87], cellobiose [101], and xylose [62]. The
highest production was achieved using glucose- and galactose-fed-batch fermenta-
tion, with a titer around 100 g/L [87] (Fig. 7.5). Fermentative production of malate
Fig. 7.5 Construction of an ethanol nonproducing, C2-independent, and glucose-tolerant yeast

platform strain by combining metabolic engineering with directed evolution
using the Pdc- strain was also attempted by combined overexpression of a pyruvate
carboxylase (PYC2), a cytosolic malate dehydrogenase (MDH3ΔSKL), and a malate
transporter from Schizosaccharomyces pombe (SpMAE1). Malate titer of up to
59 g/L was reached with a yield of 0.42 mol/mol glucose in shake flask fermentation
[166] (Fig. 7.5). Recently, the malate producer was further engineered to produce
succinate. Under optimal conditions in a bioreactor, the engineered strain produced
around 13 g/L of succinate with a yield of 0.21 mol/mol glucose at low pH [165]
(Fig. 7.5). Although glycerol is not directly derived from pyruvate, the elimination
of ethanol formation in the Pdc- strain may redirect the metabolic fluxes to glycerol
formation, especially under low-oxygen or anaerobic conditions. By further engi-
neering cytosolic NADH availability and overexpressing GPD2, a titer of higher
than 50 g/L and a yield as high as 1.08 mol/mol glucose for glycerol production
were achieved in aerobic and glucose-limited chemostat cultures with formate co-
feeding [41] (Fig. 7.5).
Another example of combining directed evolution with metabolic engineering is
the effort to increase the yield of ethanol in S. cerevisiae by eliminating glycerol
formation [46]. Glycerol production is required for redox-cofactor balancing in
anaerobic cultures. Acetate reduction was found to replace glycerol formation under
anaerobic condition for NADH re-oxidation. However, the acetate-reducing (mhpF
from E. coli overexpression) and glycerol-nonproducing (GPD1 and GPD2 dele-
tion) yeast strain is sensitive to high sugar concentrations. Directed evolution
190 T. Si et al.
enabled the isolation of an evolved strain that grew anaerobically at 1 M of glucose,
and the ethanol yield on sugar increased from 79% of the theoretical maximum in
the reference strain to 92% in the evolved strain.
In some cases, it is possible to take advantage of the unique properties of the
final products to develop growth-based directed evolution strategies. For example,
glutathione is an antioxidant, and directed evolution can be performed by coupling
the enhanced glutathione accumulation phenotype with the acrolein resistance
phenotype [111]. The evolved strain accumulated glutathione in 3.3-fold higher
concentration compared to its parental strain and reached a particularly high gluta-
thione content of almost 6%. Similarly, by taking advantage of the antioxidative
properties of carotenoids, directed evolution was designed based on periodic
hydrogen peroxide shocking, and a threefold increase in carotenoids production
(from 6 mg/g dry cell weight to up to 18 mg/g dry cell weight) was achieved in the
evolved strain [118].
Increased production of isobutanol in E. coli was also attempted by directed
evolution [137]. The isobutanol production ability is closely related to the metabolic
flux through the valine biosynthetic pathway, which can be coupled to the cellular
resistance to the valine analog norvaline. Using this strategy, a final isobutanol titer
of 21.2 g/L was achieved in 99 h with a yield of 0.31 g isobutanol/g of glucose or
76% of theoretical maximum, in comparison with a production of 5.3 g/L obtained
with the wild-type strain.
7.4 Perspectives
Microorganisms are increasingly exploited to address some of the most challenging

global problems such as sustainability and energy security. In many cases, cell fac-
tories used for industrial applications require a combination of complex phenotypes
such as high tolerance to inhibitors in the raw materials, toxic products at high
concentrations, low pH, and high temperature. Directed evolution approaches have
been successful in coping with these challenges. Nevertheless, challenges and
opportunities still remain in strain development by directed evolution. First of all,
novel screening and selection methods should be developed and integrated into
directed evolution pipelines. Currently, the phenotypes that directed evolution can
cope are mainly limited to those closely related to cellular growth, such as substrate
utilization and tolerance to toxic compounds. The development of small molecule
biosensors based on TFs [153], G-protein-coupled receptors (GPCRs) [100], and
riboswitches [60] can be incorporated to expand the scope of directed evolution. For
example, a malonyl-CoA biosensor was developed by Li et al. and then used to
screen a cDNA library that increased the intracellular malonyl-CoA levels [86]. The
robotic platform [31, 138] and microfluidic system [54, 135] may also be used for
the screening of desired phenotypes. Another challenge is the trade-off of directed
evolution [21, 51, 97]. For example, the evolved galactose-fermenting strain dem-
onstrated decreased growth in glucose, and the evolved thermotolerant strain
showed trade-offs when growing at ancestral temperatures. Evolutionary trade-offs
may hinder its industrial applications, in which multiple and complex traits are
required. Systems biology tools [132] may help elucidate the molecular mecha-
nisms of evolved phenotypes and minimize the evolutionary trade-offs for cellular
factory development. Nevertheless, the development of novel genome engineering
tools such as the CRISPR-Cas system provides new dimensions in the generation of
strain libraries with improved diversity. Although genome-scale screening based on
CRISPR knockout [126], CRISPR interference [43], and CRISPR activation [43,
69] has been demonstrated in mammalian cells, their applications in the construc-
tion and optimization of cell factories have yet to be explored.
References
1. Abbott DA, Zelle RM, Pronk JT, Maris AJA (2009) Metabolic engineering of Saccharomyces
cerevisiae for production of carboxylic acids: current status and challenges. FEMS Yeast Res
9(8):1123–1136
2. Alexeyev MF, Shokolenko IN (1995) Mini-Tn10 transposon derivatives for insertion muta-
genesis and gene delivery into the chromosome of gram-negative bacteria. Gene 160(1):
59–62
3. Almario MP, Reyes LH, Kao KC (2013) Evolutionary engineering of Saccharomyces cerevi-
siae for enhanced tolerance to hydrolysates of lignocellulosic biomass. Biotechnol Bioeng
110(10):2616–2623
4. Alper H, Moxley J, Nevoigt E, Fink GR, Stephanopoulos G (2006) Engineering yeast tran-
scription machinery for improved ethanol tolerance and production. Science 314(5805):
1565–1568
5. Alper H, Stephanopoulos G (2007) Global transcription machinery engineering: a new
approach for improving cellular phenotype. Metab Eng 9(3):258–267
6. Atsumi S, Liao JC (2008) Directed evolution of Methanococcus jannaschii citramalate syn-
thase for biosynthesis of 1-propanol and 1-butanol by Escherichia coli. Appl Environ
Microbiol 74(24):7802–7808
7. Baba T, Ara T, Hasegawa M, Takai Y, Okumura Y, Baba M, Datsenko KA, Tomita M, Wanner
BL, Mori H (2006) Construction of Escherichia coli K-12 in-frame, single-gene knockout
mutants: the Keio collection. Mol Syst Biol 2:2006.0008
8. Bailey JE (1991) Toward a science of metabolic engineering. Science 252(5013):
1668–1675
9. Bailey JE, Sburlati A, Hatzimanikatis V, Lee K, Renner WA, Tsai PS (1996) Inverse meta-
bolic engineering: a strategy for directed genetic engineering of useful phenotypes. Biotechnol
Bioeng 52(1):109–121
10. Bao Z, Xiao H, Liang J, Zhang L, Xiong X, Sun N, Si T, Zhao H (2015) Homology-integrated
CRISPR-Cas (HI-CRISPR) system for one-step multigene disruption in Saccharomyces
cerevisiae. ACS Synth Biol 4(5):585–594
11. Barrick JE, Yu DS, Yoon SH, Jeong H, Oh TK, Schneider D, Lenski RE, Kim JF (2009)
Genome evolution and adaptation in a long-term experiment with Escherichia coli. Nature
461(7268):1243–1247
12. Biot-Pelletier D, Martin VJ (2014) Evolutionary engineering by genome shuffling. Appl
13. Blank D, Wolf L, Ackermann M, Silander OK (2014) The predictability of molecular evolu-
tion during functional innovation. Proc Natl Acad Sci U S A 111(8):3044–3049
14. Brennan TC, Williams TC, Schulz BL, Palfreyman RW, Kromer JO, Nielsen LK (2015)
Evolutionary engineering improves tolerance for replacement jet fuels in Saccharomyces
cerevisiae. Appl Environ Microbiol 81(10):3316–3325
192 T. Si et al.
15. Burgard AP, Pharkya P, Maranas CD (2003) Optknock: a bilevel programming framework for
identifying gene knockout strategies for microbial strain optimization. Biotechnol Bioeng
84(6):647–657
16. Cadiere A, Aguera E, Caille S, Ortiz-Julien A, Dequin S (2012) Pilot-scale evaluation the
enological traits of a novel, aromatic wine yeast strain obtained by adaptive evolution. Food
Microbiol 32(2):332–337
17. Cadiere A, Ortiz-Julien A, Camarasa C, Dequin S (2011) Evolutionary engineered
Saccharomyces cerevisiae wine yeast strains with increased in vivo flux through the pentose
phosphate pathway. Metab Eng 13(3):263–271
18. Cakar ZP, Turanli-Yildiz B, Alkim C, Yilmaz U (2012) Evolutionary engineering of
Saccharomyces cerevisiae for improved industrially important properties. FEMS Yeast Res
12(2):171–182
19. Carr PA, Church GM (2009) Genome engineering. Nat Biotechnol 27(12):1151–1162
20. Caspeta L, Chen Y, Ghiaci P, Feizi A, Buskov S, Hallstrom BM, Petranovic D, Nielsen
J (2014) Altered sterol composition renders yeast thermotolerant. Science 346(6205):75–78
21. Caspeta L, Nielsen J (2015) Thermotolerant yeast strains adapted by laboratory evolution
show trade-off at ancestral temperatures and preadaptation to other stresses. MBio 6(4):e00431
22. Chen Y, Sheng J, Jiang T, Stevens J, Feng X, Wei N (2016) Transcriptional profiling reveals
molecular basis and novel genetic targets for improved resistance to multiple fermentation
inhibitors in Saccharomyces cerevisiae. Biotechnol Biofuels 9:9
23. Chou HH, Keasling JD (2013) Programming adaptive control to evolve increased metabolite
production. Nat Commun 4:2595
24. Cobb RE, Wang Y, Zhao H (2015) High-efficiency multiplex genome editing of Streptomyces
species using an engineered CRISPR/Cas system. ACS Synth Biol 4(6):723–728
25. Cong L, Ran FA, Cox D, Lin SL, Barretto R, Habib N, Hsu PD, Wu XB, Jiang WY, Marraffini
LA, Zhang F (2013) Multiplex genome engineering using CRISPR/Cas systems. Science
339(6121):819–823
26. Dean AM, Thornton JW (2007) Mechanistic approaches to the study of evolution: the func-
tional synthesis. Nat Rev Genet 8(9):675–688
27. Demeke MM, Foulquie-Moreno MR, Dumortier F, Thevelein JM (2015) Rapid evolution of
recombinant Saccharomyces cerevisiae for Xylose fermentation through formation of extra-
chromosomal circular DNA. PLoS Genet 11(3):e1005010
28. DiCarlo JE, Conley AJ, Penttila M, Jantti J, Wang HH, Church GM (2013) Yeast oligo-
mediated genome engineering (YOGE). ACS Synth Biol 2(12):741–749
29. Dietrich JA, McKee AE, Keasling JD (2010) High-throughput metabolic engineering:
advances in small-molecule screening and selection. Annu Rev Biochem 79:563–590
30. Dietrich JA, Shis DL, Alikhani A, Keasling JD (2013) Transcription factor-based screens and
synthetic selections for microbial small-molecule biosynthesis. ACS Synth Biol 2(1):47–58
31. Dörr M, Fibinger MPC, Last D, Schmidt S, Santos-Aberturas J, Böttcher D, Hummel A,
Vickers C, Voss M, Bornscheuer UT (2016) Fully automatized high-throughput enzyme
library screening using a robotic platform. Biotechnol Bioeng. doi:10.1002/bit.25925
32. Dragosits M, Mattanovich D (2013) Adaptive laboratory evolution – principles and applica-
tions for biotechnology. Microb Cell Fact 12:64
33. Dymond JS, Richardson SM, Coombes CE, Babatz T, Muller H, Annaluru N, Blake WJ,
Schwerzmann JW, Dai J, Lindstrom DL, Boeke AC, Gottschling DE, Chandrasegaran S,
Bader JS, Boeke JD (2011) Synthetic chromosome arms function in yeast and generate phe-
notypic diversity by design. Nature 477(7365):471–476
34. Ellis HM, Yu D, Di Tizio T, Court DL (2001) High efficiency mutagenesis, repair, and engi-
neering of chromosomal DNA using single-stranded oligonucleotides. Proc Natl Acad Sci U
S A 98(12):6742–6746
35. Enquist-Newman M, Faust AM, Bravo DD, Santos CN, Raisner RM, Hanel A, Sarvabhowman
P, Le C, Regitsky DD, Cooper SR, Peereboom L, Clark A, Martinez Y, Goldsmith J, Cho MY,
Donohoue PD, Luo L, Lamberson B, Tamrakar P, Kim EJ, Villari JL, Gill A, Tripathi SA,
Karamchedu P, Paredes CJ, Rajgarhia V, Kotlar HK, Bailey RB, Miller DJ, Ohler NL,
Swimmer C, Yoshikuni Y (2014) Efficient ethanol production from brown macroalgae sugars
by a synthetic yeast platform. Nature 505(7482):239–243
36. Esvelt KM, Wang HH (2013) Genome-scale engineering for systems and synthetic biology.
Mol Syst Biol 9:641
37. Flikweert MT, Swaaf M, Dijken JP, Pronk JT (1999) Growth requirements of pyruvate-
decarboxylase-negative Saccharomyces cerevisiae. FEMS Microbiol Lett 174(1):73–79
38. Fondi M, Liò P (2015) Multi-omics and metabolic modelling pipelines: challenges and tools
for systems microbiology. Microbiol Res 171:52–64
39. Fong SS, Burgard AP, Herring CD, Knight EM, Blattner FR, Maranas CD, Palsson BO
(2005) In silico design and adaptive evolution of Escherichia coli for production of lactic
acid. Biotechnol Bioeng 91(5):643–648
40. Fullwood MJ, Wei CL, Liu ET, Ruan Y (2009) Next-generation DNA sequencing of paired-
end tags (PET) for transcriptome and genome analyses. Genome Res 19(4):521–532
41. Geertman JM, Maris AJ, Dijken JP, Pronk JT (2006) Physiological and genetic engineering
of cytosolic redox metabolism in Saccharomyces cerevisiae for improved glycerol produc-
tion. Metab Eng 8(6):532–542
42. Giaever G, Chu AM, Ni L, Connelly C, Riles L, Veronneau S, Dow S, Lucau-Danila A,
Anderson K, Andre B, Arkin AP, Astromoff A, Bakkoury M, Bangham R, Benito R, Brachat
S, Campanaro S, Curtiss M, Davis K, Deutschbauer A, Entian K-D, Flaherty P, Foury F,
Garfinkel DJ, Gerstein M, Gotte D, Guldener U, Hegemann JH, Hempel S, Herman Z,
Jaramillo DF, Kelly DE, Kelly SL, Kotter P, LaBonte D, Lamb DC, Lan N, Liang H, Liao H,
Liu L, Luo C, Lussier M, Mao R, Menard P, Ooi SL, Revuelta JL, Roberts CJ, Rose M, Ross-
Macdonald P, Scherens B, Schimmack G, Shafer B, Shoemaker DD, Sookhai-Mahadeo S,
Storms RK, Strathern JN, Valle G, Voet M, Volckaert G, Wang C-Y, Ward TR, Wilhelmy J,
Winzeler EA, Yang Y, Yen G, Youngman E, Yu K, Bussey H, Boeke JD, Snyder M, Philippsen
P, Davis RW, Johnston M (2002) Functional profiling of the Saccharomyces cerevisiae
genome. Nature 418(6896):387–391
43. Gilbert LA, Horlbeck MA, Adamson B, Villalta JE, Chen Y, Whitehead EH, Guimaraes C,
Panning B, Ploegh HL, Bassik MC, Qi LS, Kampmann M, Weissman JS (2014) Genome-
scale CRISPR-mediated control of gene repression and activation. Cell 159(3):647–661
44. Greener A, Callahan M, Jerpseth B (1997) An efficient random mutagenesis technique using
an E. coli mutator strain. Mol Biotechnol 7(2):189–195
45. Gresham D, Dunham MJ (2014) The enduring utility of continuous culturing in experimental
evolution. Genomics 104(6 Pt A):399–405
46. Guadalupe-Medina V, Metz B, Oud B, Graaf CM, Mans R, Pronk JT, Maris AJ (2014)
Evolutionary engineering of a glycerol-3-phosphate dehydrogenase-negative, acetate-
reducing Saccharomyces cerevisiae strain enables anaerobic growth at high glucose concen-
trations. J Microbial Biotechnol 7(1):44–53
47. Guimaraes PM, Berre V, Sokol S, Francois J, Teixeira JA, Domingues L (2008) Comparative
transcriptome analysis between original and evolved recombinant lactose-consuming
Saccharomyces cerevisiae strains. Biotechnol J 3(12):1591–1597
48. Hasunuma T, Sakamoto T, Kondo A (2016) Inverse metabolic engineering based on transient
acclimation of yeast improves acid-containing xylose fermentation and tolerance to formic
and acetic acids. Appl Microbiol Biotechnol 100(2):1027–1038
49. Hawkins GM, Doran-Peterson J (2011) A strain of Saccharomyces cerevisiae evolved for
fermentation of lignocellulosic biomass displays improved growth and fermentative ability in
high solids concentrations and in the presence of inhibitory compounds. Biotechnol Biofuels
4(1):49
50. Ho CH, Magtanong L, Barker SL, Gresham D, Nishimura S, Natarajan P, Koh JL, Porter J,
Gray CA, Andersen RJ, Giaever G, Nislow C, Andrews B, Botstein D, Graham TR, Yoshida
M, Boone C (2009) A molecular barcoded yeast ORF library enables mode-of-action analysis
of bioactive compounds. Nat Biotechnol 27(4):369–377
194 T. Si et al.
51. Hong KK, Nielsen J (2013) Adaptively evolved yeast mutants on galactose show trade-offs
in carbon utilization on glucose. Metab Eng 16:78–86
52. Horinouchi T, Minamoto T, Suzuki S, Shimizu H, Furusawa C (2014) Development of an
automated culture system for laboratory evolution. J Lab Autom 19(5):478–482
53. Hosaka T, Ohnishi-Kameyama M, Muramatsu H, Murakami K, Tsurumi Y, Kodani S,
Yoshida M, Fujie A, Ochi K (2009) Antibacterial discovery in actinomycetes strains with
mutations in RNA polymerase or ribosomal protein S12. Nat Biotechnol 27(5):462–464
54. Huang M, Bai Y, Sjostrom SL, Hallstrom BM, Liu Z, Petranovic D, Uhlen M, Joensson HN,
Andersson-Svahn H, Nielsen J (2015) Microfluidic screening and whole-genome sequencing
identifies mutations associated with improved protein secretion by yeast. Proc Natl Acad Sci
U S A 112(34):E4689–E4696
55. Hutchison CA, Peterson SN, Gill SR, Cline RT, White O, Fraser CM, Smith HO, Venter JC
(1999) Global transposon mutagenesis and a minimal Mycoplasma genome. Science
286(5447):2165–2169
56. Jakiela S, Kaminski TS, Cybulski O, Weibel DB, Garstecki P (2013) Bacterial growth and
adaptation in microdroplet chemostats. Angew Chem Int Ed 52(34):8908–8911
57. Jakociunas T, Bonde I, Herrgard M, Harrison SJ, Kristensen M, Pedersen LE, Jensen MK,
Keasling JD (2015) Multiplex metabolic pathway engineering using CRISPR/Cas9 in
Saccharomyces cerevisiae. Metab Eng 28:213–222
58. Jiang W, Bikard D, Cox D, Zhang F, Marraffini LA (2013) RNA-guided editing of bacterial
genomes using CRISPR-Cas systems. Nat Biotechnol 31(3):233–239
59. Keasling JD (2010) Manufacturing molecules through metabolic engineering. Science
330(6009):1355–1358
60. Kim HJ, Ha S, Lee HY, Lee KJ (2015) ROSics: chemistry and proteomics of cysteine modi-
fications in redox biology. Mass Spectrum Rev 34(2):184–208
61. Kim SJ, Seo SO, Jin YS, Seo JH (2013) Production of 2,3-butanediol by engineered
Saccharomyces cerevisiae. Bioresour Technol 146:274–281
62. Kim SJ, Seo SO, Park YC, Jin YS, Seo JH (2014) Production of 2,3-butanediol from xylose
by engineered Saccharomyces cerevisiae. J Biotechnol 192:376–382
63. Kim SR, Skerker JM, Kang W, Lesmana A, Wei N, Arkin AP, Jin YS (2013) Rational and
evolutionary engineering approaches uncover a small set of genetic changes efficient for
rapid xylose fermentation in Saccharomyces cerevisiae. PLoS One 8(2):e57048
64. Kim SR, Xu H, Lesmana A, Kuzmanovic U, Au M, Florencia C, Oh EJ, Zhang G, Kim KH,
Jin YS (2015) Deletion of PHO13, encoding haloacid dehalogenase type IIA phosphatase,
results in upregulation of the pentose phosphate pathway in Saccharomyces cerevisiae. Appl
65. Kinnersley M, Wenger J, Kroll E, Adams J, Sherlock G, Rosenzweig F (2014) Ex uno plures:
clonal reinforcement drives evolution of a simple microbial community. PLoS Genet
10(6):e1004430
66. Klein-Marcuschamer D, Stephanopoulos G (2008) Assessing the potential of mutational
strategies to elicit new phenotypes in industrial strains. Proc Natl Acad Sci U S A
105(7):2319–2324
67. Klimacek M, Kirl E, Krahulec S, Longus K, Novy V, Nidetzky B (2014) Stepwise metabolic
adaption from pure metabolization to balanced anaerobic growth on xylose explored for
recombinant Saccharomyces cerevisiae. Microb Cell Fact 13(1):37
68. Koffas M (2005) Evolutionary metabolic engineering. Metab Eng 7(1):1–3
69. Konermann S, Brigham MD, Trevino AE, Joung J, Abudayyeh OO, Barcena C, Hsu PD,
Habib N, Gootenberg JS, Nishimasu H, Nureki O, Zhang F (2015) Genome-scale transcrip-
tional activation by an engineered CRISPR-Cas9 complex. Nature 517(7536):583–588
70. Koppram R, Albers E, Olsson L (2012) Evolutionary engineering strategies to enhance toler-
ance of xylose utilizing recombinant yeast to inhibitors derived from spruce biomass.
Biotechnol Biofuels 5(1):32
71. Koren S, Phillippy AM (2015) One chromosome, one contig: complete microbial genomes
from long-read sequencing and assembly. Curr Opin Microbiol 23:110–120
72. Kryazhimskiy S, Rice DP, Jerison ER, Desai MM (2014) Global epistasis makes adaptation
predictable despite sequence-level stochasticity. Science 344(6191):1519–1522
73. Kucukgoze G, Alkim C, Yilmaz U, Kisakesen HI, Gunduz S, Akman S, Cakar ZP (2013)
Evolutionary engineering and transcriptomic analysis of nickel-resistant Saccharomyces
cerevisiae. FEMS Yeast Res 13(8):731–746
74. Lang GI, Botstein D, Desai MM (2011) Genetic variation and the fate of beneficial mutations
in asexual populations. Genetics 188(3):647–661
75. Lang GI, Desai MM (2014) The spectrum of adaptive mutations in experimental evolution.
Genomics 104(6):412–416
76. Lang GI, Murray AW (2008) Estimating the per-base-pair mutation rate in the yeast
Saccharomyces cerevisiae. Genetics 178(1):67–82
77. Lang GI, Rice DP, Hickman MJ, Sodergren E, Weinstock GM, Botstein D, Desai MM (2013)
Pervasive genetic hitchhiking and clonal interference in fourty evolving yeast populations.
Nature 500(7464):571–574
78. Lee H, Popodi E, Tang H, Foster PL (2012) Rate and molecular spectrum of spontaneous
mutations in the bacterium Escherichia coli as determined by whole-genome sequencing.
Proc Natl Acad Sci U S A 109(41):E2774–E2783
79. Lee HJ, Kim SJ, Yoon JJ, Kim KH, Seo JH, Park YC (2015) Evolutionary engineering of
Saccharomyces cerevisiae for efficient conversion of red algal biosugars to bioethanol.
Bioresour Technol 191:445–451
80. Lee JW, Na D, Park JM, Lee J, Choi S, Lee SY (2012) Systems metabolic engineering of
microorganisms for natural and non-natural chemicals. Nat Chem Biol 8(6):536–546
81. Lee KS, Hong ME, Jung SC, Ha SJ, Yu BJ, Koo HM, Park SM, Seo JH, Kweon DH, Park JC,
Jin YS (2011) Improved galactose fermentation of Saccharomyces cerevisiae through inverse
metabolic engineering. Biotechnol Bioeng 108(3):621–631
82. Lee SM, Jellison T, Alper HS (2014) Systematic and evolutionary engineering of a xylose
isomerase-based pathway in Saccharomyces cerevisiae for efficient conversion yields.
Biotechnol Biofuels 7(1):122
83. Lee SY, Kim HU (2015) Systems strategies for developing industrial microbial strains. Nat
Biotechnol 33(10):1061–1072
84. Lenski RE, Mongold JA, Sniegowski PD, Travisano M, Vasi F, Gerrish PJ, Schmidt TM
(1998) Evolution of competitive fitness in experimental populations of E. coli: what
makes one genotype a better competitor than another? Antonie Van Leeuwenhoek
73(1):35–47
85. Li P, Sun H, Chen Z, Li Y, Zhu T (2015) Construction of efficient xylose utilizing Pichia
pastoris for industrial enzyme production. Microb Cell Fact 14:22
86. Li S, Si T, Wang M, Zhao H (2015) Development of a synthetic Malonyl-CoA sensor in
Saccharomyces cerevisiae for intracellular metabolite monitoring and genetic screening. ACS
Synth Biol 4(12):1308–1315
87. Lian J, Chao R, Zhao H (2014) Metabolic engineering of a Saccharomyces cerevisiae strain
capable of simultaneously utilizing glucose and galactose to produce enantiopure (2R,3R)-
butanediol. Metab Eng 23:92–99
88. Lind PA, Andersson DI (2008) Whole-genome mutational biases in bacteria. Proc Natl Acad
Sci U S A 105(46):17878–17883
89. Ling H, Pratomo Juwono NK, Teo WS, Liu R, Leong SS, Chang MW (2015) Engineering
transcription factors to improve tolerance against alkane biofuels in Saccharomyces cerevi-
siae. Biotechnol Biofuels 8:231
90. Lopez-Malo M, Garcia-Rios E, Melgar B, Sanchez MR, Dunham MJ, Guillamon JM (2015)
Evolutionary engineering of a wine yeast strain revealed a key role of inositol and mannopro-
tein metabolism during low-temperature fermentation. BMC Genomics 16:537
196 T. Si et al.
91. Lou DI, Hussmann JA, McBee RM, Acevedo A, Andino R, Press WH, Sawyer SL (2013)
High-throughput DNA sequencing errors are reduced by orders of magnitude using circle
sequencing. Proc Natl Acad Sci U S A 110(49):19872–19877
92. Luan G, Cai Z, Li Y, Ma Y (2013) Genome replication engineering assisted continuous evo-
lution (GREACE) to improve microbial tolerance for biofuels production. Biotechnol
Biofuels 6(1):137
93. Lynch M, Sung W, Morris K, Coffey N, Landry CR, Dopman EB, Dickinson WJ, Okamoto
K, Kulkarni S, Hartl DL, Thomas WK (2008) A genome-wide view of the spectrum of spon-
taneous mutations in yeast. Proc Natl Acad Sci U S A 105(27):9272–9277
94. Lynch MD, Warnecke T, Gill RT (2007) SCALEs: multiscale analysis of library enrichment.
Nat Methods 4(1):87–93
95. Maharjan R, Seeto S, Notley-McRobb L, Ferenci T (2006) Clonal adaptive radiation in a
constant environment. Science 313(5786):514–517
96. Mali P, Yang L, Esvelt KM, Aach J, Guell M, DiCarlo JE, Norville JE, Church GM (2013)
RNA-guided human genome engineering via Cas9. Science 339(6121):823–826
97. Martinez JL, Bordel S, Hong KK, Nielsen J (2014) Gcn4p and the Crabtree effect of yeast:
drawing the causal model of the Crabtree effect in Saccharomyces cerevisiae and explaining
evolutionary trade-offs of adaptation to galactose through systems biology. FEMS Yeast Res
14(4):654–662
98. Michener JK, Smolke CD (2012) High-throughput enzyme evolution in Saccharomyces cere-
visiae using a synthetic RNA switch. Metab Eng 14(4):306–316
99. Mitsumasu K, Liu ZS, Tang YQ, Akamatsu T, Taguchi H, Kida K (2014) Development of
industrial yeast strain with improved acid- and thermo-tolerance through evolution under
continuous fermentation conditions followed by haploidization and mating. J Biosci Bioeng
118(6):689–695
100. Mukherjee K, Bhattacharyya S, Peralta-Yahya P (2015) GPCR-based chemical biosensors
for medium-chain fatty acids. ACS Synth Biol 4(12):1261–1269
101. Nan H, Seo SO, Oh EJ, Seo JH, Cate JH, Jin YS (2014) 2,3-butanediol production from cel-
lobiose by engineered Saccharomyces cerevisiae. Appl Microbiol Biotechnol 98(12):
5757–5764
102. Nielsen J, Keasling Jay D (2016) Engineering cellular metabolism. Cell 164(6):1185–1197
103. Nijland JG, Shin HY, Jong RM, Waal PP, Klaassen P, Driessen AJ (2014) Engineering of an
endogenous hexose transporter into a specific D-xylose transporter facilitates glucose-xylose
co-consumption in Saccharomyces cerevisiae. Biotechnol Biofuels 7(1):168
104. Notley-McRobb L, King T, Ferenci T (2002) rpoS mutations and loss of general stress resis-
tance in Escherichia coli populations as a consequence of conflict between competing stress
responses. J Bacteriol 184(3):806–811
105. Novick A, Szilard L (1950) Description of the chemostat. Science 112(2920):715–716
106. Ochi K (2007) From microbial differentiation to ribosome engineering. Biosci Biotechnol
Biochem 71(6):1373–1386
107. Oud B, Flores CL, Gancedo C, Zhang X, Trueheart J, Daran JM, Pronk JT, Maris AJ (2012)
An internal deletion in MTH1 enables growth on glucose of pyruvate-decarboxylase negative,
non-fermentative Saccharomyces cerevisiae. Microb Cell Fact 11:131
108. Oud B, Maris AJ, Daran JM, Pronk JT (2012) Genome-wide analytical approaches for
reverse metabolic engineering of industrially relevant phenotypes in yeast. FEMS Yeast Res
12(2):183–196
109. Park K-S, Lee D-K, Lee H, Lee Y, Jang Y-S, Kim YH, Yang H-Y, Lee S-I, Seol W, Kim J-S
(2003) Phenotypic alteration of eukaryotic cells using randomized libraries of artificial tran-
scription factors. Nat Biotechnol 21(10):1208–1214
110. Park KS, Jang YS, Lee H, Kim JS (2005) Phenotypic alteration and target gene identification
using combinatorial libraries of zinc finger proteins in prokaryotic cells. J Bacteriol
187(15):5496–5499
111. Patzschke A, Steiger MG, Holz C, Lang C, Mattanovich D, Sauer M (2015) Enhanced gluta-
thione production by evolutionary engineering of Saccharomyces cerevisiae strains.
Biotechnol J 10(11):1719–1726
112. Pereira SR, Sanchez INV, Frazao CJ, Serafim LS, Gorwa-Grauslund MF, Xavier AM (2015)
Adaptation of Scheffersomyces stipitis to hardwood spent sulfite liquor by evolutionary engi-
neering. Biotechnol Biofuels 8:50
113. Pinel D, Colatriano D, Jiang H, Lee H, Martin VJ (2015) Deconstructing the genetic basis of
spent sulphite liquor tolerance using deep sequencing of genome-shuffled yeast. Biotechnol
Biofuels 8:53
114. Pinel D, D’Aoust F, Cardayre SB, Bajwa PK, Lee H, Martin VJ (2011) Saccharomyces cere-
visiae genome shuffling through recursive population mating leads to improved tolerance to
spent sulfite liquor. Appl Environ Microbiol 77(14):4736–4743
115. Qi X, Zha J, Liu GG, Zhang W, Li BZ, Yuan YJ (2015) Heterologous xylose isomerase path-
way and evolutionary engineering improve xylose utilization in Saccharomyces cerevisiae.
Front Microbiol 6:1165
116. Quandt EM, Deatherage DE, Ellington AD, Georgiou G, Barrick JE (2014) Recursive
genome-wide recombination and sequencing reveals a key refinement step in the evolution
of a metabolic innovation in Escherichia coli. Proc Natl Acad Sci U S A 111(6):
2217–2222
117. Rabinovitch-Deere CA, Oliver JWK, Rodriguez GM, Atsumi S (2013) Synthetic biology
and metabolic engineering approaches to produce biofuels. Chem Rev 113(7):
4611–4632
118. Reyes LH, Gomez JM, Kao KC (2014) Improving carotenoids production in yeast via adap-
tive laboratory evolution. Metab Eng 21:26–33
119. Rodriguez-Verdugo A, Carrillo-Cisneros D, Gonzalez-Gonzalez A, Gaut BS, Bennett AF
(2014) Different tradeoffs result from alternate genetic adaptations to a common environ-
ment. Proc Natl Acad Sci U S A 111(33):12121–12126
120. Rosenzweig F, Sherlock G (2014) Experimental evolution: prospects and challenges.
Genomics 104(6, Part A):v–vi
121. Sanchez BJ, Nielsen J (2015) Genome scale models of yeast: towards standardized evaluation
and consistent omic integration. Integr Biol 7(8):846–858
122. Santos CNS, Stephanopoulos G (2008) Combinatorial engineering of microbes for optimiz-
ing cellular phenotype. Curr Opin Chem Biol 12(2):168–176
123. Sauer U (2001) Evolutionary engineering of industrially important microbial phenotypes.
Adv Biochem Eng Biotechnol 73:129–169
124. Scalcinati G, Otero JM, Vleet JR, Jeffries TW, Olsson L, Nielsen J (2012) Evolutionary engi-
neering of Saccharomyces cerevisiae for efficient aerobic xylose consumption. FEMS Yeast
Res 12(5):582–597
125. Serero A, Jubin C, Loeillet S, Legoix-Ne P, Nicolas AG (2014) Mutational landscape of yeast
mutator strains. Proc Natl Acad Sci U S A 111(5):1897–1902
126. Shalem O, Sanjana NE, Hartenian E, Shi X, Scott DA, Mikkelsen TS, Heckl D, Ebert BL,
Root DE, Doench JG, Zhang F (2014) Genome-scale CRISPR-Cas9 knockout screening in
human cells. Science 343(6166):84–87
127. Sharan SK, Thomason LC, Kuznetsov SG, Court DL (2009) Recombineering: a homologous
recombination-based method of genetic engineering. Nat Protoc 4(2):206–223
128. Shen Y, Stracquadanio G, Wang Y, Yang K, Mitchell LA, Xue Y, Cai Y, Chen T, Dymond JS,
Kang K, Gong J, Zeng X, Zhang Y, Li Y, Feng Q, Xu X, Wang J, Wang J, Yang H, Boeke JD,
Bader JS (2016) SCRaMbLE generates designed combinatorial stochastic diversity in syn-
thetic chromosomes. Genome Res 26(1):36–49
129. Shi S, Liang Y, Zhang MM, Ang EL, Zhao H (2016) A highly efficient single-step, marker-
less strategy for multi-copy chromosomal integration of large biochemical pathways in
Saccharomyces cerevisiae. Metab Eng 33:19–27
198 T. Si et al.
130. Shima J, Hesketh A, Okamoto S, Kawamoto S, Ochi K (1996) Induction of actinorhodin

production by rpsL (encoding ribosomal protein S12) mutations that confer streptomycin
resistance in Streptomyces lividans and Streptomyces coelicolor A3(2). J Bacteriol
178(24):7276–7284
131. Shin HY, Nijland JG, Waal PP, Jong RM, Klaassen P, Driessen AJ (2015) An engineered
cryptic Hxt11 sugar transporter facilitates glucose-xylose co-consumption in Saccharomyces
cerevisiae. Biotechnol Biofuels 8:176
132. Shui W, Xiong Y, Xiao W, Qi X, Zhang Y, Lin Y, Guo Y, Zhang Z, Wang Q, Ma Y (2015)
Understanding the mechanism of thermotolerance distinct from heat shock response through
proteomic analysis of industrial strains of Saccharomyces cerevisiae. Mol Cell Proteomics
14(7):1885–1897
133. Si T, Luo Y, Bao Z, Zhao H (2015) RNAi-assisted genome evolution in Saccharomyces cere-
visiae for complex phenotype engineering. ACS Synth Biol 4(3):283–291
134. Si T, Xiao H, Zhao H (2015) Rapid prototyping of microbial cell factories via genome-scale
engineering. Biotechnol Adv 33(7):1420–1432
135. Sjostrom SL, Bai Y, Huang M, Liu Z, Nielsen J, Joensson HN, Andersson Svahn H (2014)
High-throughput screening for industrial enzyme production hosts by droplet microfluidics.
Lab Chip 14(4):806–813
136. Skretas G, Kolisis FN (2012) Combinatorial approaches for inverse metabolic engineering
applications. Comput Struct Biotechnol J 3:e201210021
137. Smith KM, Liao JC (2011) An evolutionary strategy for isobutanol production strain devel-
opment in Escherichia coli. Metab Eng 13(6):674–681
138. Snoek T, Picca Nicolino M, Bremt S, Mertens S, Saels V, Verplaetse A, Steensels J, Verstrepen
KJ (2015) Large-scale robot-assisted genome shuffling yields industrial Saccharomyces cere-
visiae yeasts with increased ethanol tolerance. Biotechnol Biofuels 8:32
139. Sonderegger M, Sauer U (2003) Evolutionary engineering of Saccharomyces cerevisiae for
anaerobic growth on xylose. Appl Environ Microbiol 69(4):1990–1998
140. Standage-Beier K, Zhang Q, Wang X (2015) Targeted large-scale deletion of bacterial
genomes using CRISPR-nickases. ACS Synth Biol 4(11):1217–1225
141. Steensels J, Snoek T, Meersman E, Picca Nicolino M, Voordeckers K, Verstrepen KJ (2014)
Improving industrial yeast strains: exploiting natural and artificial diversity. FEMS Microbiol
Rev 38(5):947–995
142. Stovicek V, Borodina I, Forster J (2015) CRISPR–Cas system enables fast and simple genome
editing of industrial Saccharomyces cerevisiae strains. Metab Eng Commun 2:13–22
143. Sung W, Ackerman MS, Miller SF, Doak TG, Lynch M (2012) Drift-barrier hypothesis and
mutation-rate evolution. Proc Natl Acad Sci U S A 109(45):18488–18492
144. Tenaillon O, Rodriguez-Verdugo A, Gaut RL, McDonald P, Bennett AF, Long AD, Gaut BS
(2012) The molecular diversity of adaptive convergence. Science 335(6067):457–461
145. Teoh ST, Putri S, Mukai Y, Bamba T, Fukusaki E (2015) A metabolomics-based strategy for
identification of gene targets for phenotype improvement and its application to 1-butanol
tolerance in Saccharomyces cerevisiae. Biotechnol Biofuels 8:144
146. Tilloy V, Cadiere A, Ehsani M, Dequin S (2015) Reducing alcohol levels in wines through
rational and evolutionary engineering of Saccharomyces cerevisiae. Int J Food Microbiol
213:49–58
147. Toprak E, Veres A, Michel JB, Chait R, Hartl DL, Kishony R (2012) Evolutionary paths to
antibiotic resistance under dynamically sustained drug selection. Nat Genet 44(1):101–105
148. Torres EM, Dephoure N, Panneerselvam A, Tucker CM, Whittaker CA, Gygi SP, Dunham
MJ, Amon A (2010) Identification of aneuploidy-tolerating mutations. Cell 143(1):71–83
149. Maris AJ, Geertman JM, Vermeulen A, Groothuizen MK, Winkler AA, Piper MD, Dijken JP,
Pronk JT (2004) Directed evolution of pyruvate decarboxylase-negative Saccharomyces cere-
visiae, yielding a C2-independent, glucose-tolerant, and pyruvate-hyperproducing yeast. Appl
150. Vanee N, Fisher AB, Fong SS (2012) Evolutionary engineering for industrial microbiology.
Subcell Biochem 64:43–71
151. Vilela Lde F, de Araujo VP, Paredes Rde S, Bon EP, Torres FA, Neves BC, Eleutherio EC
(2015) Enhanced xylose fermentation and ethanol production by engineered Saccharomyces
cerevisiae strain. AMB Express 5:16
152. Wang HH, Isaacs FJ, Carr PA, Sun ZZ, Xu G, Forest CR, Church GM (2009) Programming cells
by multiplex genome engineering and accelerated evolution. Nature 460(7257):894–U133
153. Wang M, Li S, Zhao H (2016) Design and engineering of intracellular-metabolite-sensing/
regulation gene circuits in Saccharomyces cerevisiae. Biotechnol Bioeng 113(1):
206–215
154. Wang T, Wei JJ, Sabatini DM, Lander ES (2014) Genetic screens in human cells using the
CRISPR-Cas9 system. Science 343(6166):80–84
155. Wang Y, Zhang S, Liu H, Zhang L, Yi C, Li H (2015) Changes and roles of membrane com-
positions in the adaptation of Saccharomyces cerevisiae to ethanol. J Basic Microbiol
55(12):1417–1426
156. Wargacki AJ, Leonard E, Win MN, Regitsky DD, Santos CN, Kim PB, Cooper SR, Raisner
RM, Herman A, Sivitz AB, Lakshmanaswamy A, Kashiyama Y, Baker D, Yoshikuni Y (2012)
An engineered microbial platform for direct biofuel production from brown macroalgae.
Science 335(6066):308–313
157. Warner JR, Reeder PJ, Karimpour-Fard A, Woodruff LB, Gill RT (2010) Rapid profiling of a
microbial genome using mixtures of barcoded oligonucleotides. Nat Biotechnol 28(8):
856–862
158. Wei N, Quarterman J, Jin YS (2013) Marine macroalgae: an untapped resource for producing
fuels and chemicals. Trends Biotechnol 31(2):70–77
159. Winkler JD, Kao KC (2014) Recent advances in the evolutionary engineering of industrial
biocatalysts. Genomics 104(6 Pt A):406–411
160. Wisselink HW, Cipollina C, Oud B, Crimi B, Heijnen JJ, Pronk JT, Maris AJ (2010)
Metabolome, transcriptome and metabolic flux analysis of arabinose fermentation by engi-
neered Saccharomyces cerevisiae. Metab Eng 12(6):537–551
161. Woodruff LBA, Gill RT (2011) Engineering genomes in multiplex. Curr Opin Biotechnol
22(4):576–583
162. Wright J, Bellissimi E, Hulster E, Wagner A, Pronk JT, Maris AJ (2011) Batch and continu-
ous culture-based selection strategies for acetic acid tolerance in xylose-fermenting
Saccharomyces cerevisiae. FEMS Yeast Res 11(3):299–306
163. Xiao H, Zhao H (2014) Genome-wide RNAi screen reveals the E3 SUMO-protein ligase
gene SIZ1 as a novel determinant of furfural tolerance in Saccharomyces cerevisiae.
Biotechnol Biofuels 7:78
164. Xu H, Kim S, Sorek H, Lee Y, Jeong D, Kim J, Oh EJ, Yun EJ, Wemmer DE, Kim KH, Kim
SR, Jin YS (2016) PHO13 deletion-induced transcriptional activation prevents sedoheptulose
accumulation during xylose metabolism in engineered Saccharomyces cerevisiae. Metab Eng
34:88–96
165. Yan D, Wang C, Zhou J, Liu Y, Yang M, Xing J (2014) Construction of reductive pathway in
Saccharomyces cerevisiae for effective succinic acid fermentation at low pH value. Bioresour
Technol 156:232–239
166. Zelle RM, Hulster E, Winden WA, Waard P, Dijkema C, Winkler AA, Geertman JM, Dijken
JP, Pronk JT, Maris AJ (2008) Malic acid production by Saccharomyces cerevisiae: engineer-
ing of pyruvate carboxylation, oxaloacetate reduction, and malate export. Appl Environ
Microbiol 74(9):2766–2777
167. Zha J, Shen M, Hu M, Song H, Yuan Y (2014) Enhanced expression of genes involved in
initial xylose metabolism and the oxidative pentose phosphate pathway in the improved
xylose-utilizing Saccharomyces cerevisiae through evolutionary engineering. J Ind Microbiol
200 T. Si et al.
168. Zhang Y, Liu G, Engqvist MK, Krivoruchko A, Hallstrom BM, Chen Y, Siewers V, Nielsen
J (2015) Adaptive mutations in sugar metabolism restore growth on glucose in a pyruvate
decarboxylase negative yeast strain. Microb Cell Fact 14:116
169. Zhou H, Cheng JS, Wang BL, Fink GR, Stephanopoulos G (2012) Xylose isomerase overex-
pression along with engineering of the pentose phosphate pathway and evolutionary engi-
neering enable rapid xylose utilization and ethanol production by Saccharomyces cerevisiae.
Metab Eng 14(6):611–622
170. Zhu L, Li Y, Cai Z (2015) Development of a stress-induced mutagenesis module for autono-
mous adaptive evolution of Escherichia coli to improve its stress tolerance. Biotechnol
Biofuels 8:93
171. Ziv N, Brandt NJ, Gresham D (2013) The use of chemostats in microbial systems biology.
J Vis Exp 80:18
Back to Basics:
Creating Genetic Diversity 8
Kang Lan Tee and Tuck Seng Wong
Abstract
Directed evolution has emerged as a key enabling technology for improving the
properties of biomolecules, biochemical pathways, and microorganisms to sat-
isfy a wide range of biotechnological applications, from synthetic biology
through to industrial biocatalysis. Laboratory evolution is an iterative process,
alternating between creating genetic diversity and selection/screening to identify
improved variants. This book chapter focuses on genetic diversity only. We
describe and critically review recent advances in the methods for DNA assembly,
random mutagenesis, focused mutagenesis, and DNA recombination. We also
identify trends in these areas and highlight commercial kits that are developed to
streamline and expedite these molecular biology techniques.
8.1 Introduction
Charles Darwin has taught us an invaluable lesson on how to engineer a biological

system, using the principle of genetic diversity coupled with selection. Built upon
the concept of Darwinian evolution, directed evolution has transformed the field of
biological engineering. Not only is it an indispensable tool in the academic arena, it
is also a key enabling technology in the commercial sector. Directed evolution is
now a popular method of choice for tailoring the properties of nucleic acids, pro-
teins, pathways, and organisms to suit various applications, including biocatalysis,
bioremediation, biosensing, and synthetic biology.
K.L. Tee • T.S. Wong (*)

ChELSI Institute and Advanced Biomanufacturing Centre (ABC), Department of Chemical
and Biological Engineering, University of Sheffield,
Sir Robert Hadfield Building, Mappin Street, Sheffield S1 3JD, UK

DOI 10.1007/978-3-319-50413-1_8
202 K.L. Tee and T.S. Wong
Creating a high-quality mutant library is the first step in a successful directed

evolution campaign. The phrase Back to Basics in the title refers to this fundamental
step in directed evolution. It involves choosing the right mutagenesis strategy and an
efficient cloning method. In 2006, we wrote a comprehensive review on the meth-
ods for creating genetic diversity [1]. This survey was subsequently updated in 2013
and published in Biotechnology Advances [2]. Encouraged by the positive feedback
from peers and the attention received for these two reviews, we decide to extend our
previous work to survey methods published since 2013. In other words, this book
chapter is a sequel to the previous two papers [1, 2]; the organization and the method
categorization are identical to the systems used before to facilitate reading and
understanding. Those methods that have already been covered previously will not
be repeated here, and interested readers are kindly asked to refer to these papers [1,
2]. We have endeavored to be inclusive in this chapter, and we apologize if there is
a method that we miss.
The aim of this chapter is twofold: (1) to provide a method selection guide for
those wanting to create a gene library, particularly newcomers, and (2) to stimulate
the development of many more novel and exciting techniques to advance the field of
directed evolution. This chapter starts with a description of DNA assembly meth-
ods. This is followed by an update and critical review of the methodologies in ran-
dom mutagenesis, focused mutagenesis, and DNA recombination. We also provide
an overview of the commercial kits developed for each application and compare the
principles of these kits. This chapter is concluded with a perspective on the future
developments in the field of genetic diversity creation.
8.2 Molecular Cloning and DNA Assembly
Molecular cloning is a necessary step in most directed evolution experiments, where

the gene of interest (GOI) is introduced into a vector for subsequent mutagenesis or
the mutant library is subcloned for expression in an appropriate host organism.
Further to directed evolution of a single protein, we have seen increased application
of directed evolution in pathway engineering, where multiple genes in a pathway
are engineered to achieve a greater final output [3, 4]. Consequentially, more meth-
ods are now adapted for introducing multiple GOIs into a vector backbone simulta-
neously or sequentially, instead of just a single GOI. As molecular cloning is
traditionally used for introducing a GOI into a vector, the term DNA assembly is
used in this chapter to better encompass all the methods developed to insert a single
or multiple GOI(s) into a vector backbone.
We have previously categorized molecular cloning methods into four main strat-
egies back in the year 2013 [2]. Development of new methods, however, necessi-
tates the introduction of a new category. DNA assembly methods are now described
in the following categories: (1) complementary overhangs, (2) homologous
sequences, (3) overlapping polymerase chain reaction (PCR), (4) megaprimers, and
(5) bridged ligation. The principles of these five strategies are illustrated in Fig. 8.1,
and an overview is provided in Table 8.1.
8 Back to Basics: Creating Genetic Diversity 203
Fig. 8.1 Five strategies for DNA assembly: (a) complementary overhangs, (b) homologous
sequences, (c) overlapping PCR, (d) megaprimers, and (e) bridged ligation
8.2.1 DNA Assembly Based on Complementary Overhangs
DNA assembly based on complementary overhangs resembles the conventional

restriction–ligation cloning, but varies in the ways complementary overhangs
between the GOI(s) and vector are generated. Short complementary overhangs [2–4
nucleotides (nts)] are commonly created using Type II restriction enzymes (REs),
terminal transferase activity of some thermophilic DNA polymerases, asymmetric
PCRs, or zinc finger nucleases, before GOI(s) and vector are covalently linked using
a DNA ligase. Alternatively, long complementary overhangs (12–14 nts) allow for-
mation of a stable nicked plasmid that can be directly transformed without the need
of a DNA ligase. The nicks are then repaired by the host mechanisms. Examples of
the methods used to generate long overhangs include phosphorothioate bond cleav-
age with iodine/ethanol solution or the use of nicking DNA endonucleases. New
methods described since year 2013 continue to use the above methods to generate
complementary overhangs, with increasing effort dedicated to adapting these meth-
ods to assemble multiple DNA fragments.
In an attempt to increase the cloning efficiency, Gao et al. used an additional gel
extraction followed by a second ligation, after the first ligation of the linearized vec-
tor and the DNA insert. This additional gel purification step was aimed to isolate the
right-sized ligation product to reduce self-ligation of digested vectors [5]. Using the
Table 8.1 Methods for DNA assembly

Availability of
Category Strategy Mechanism cloning kit
A Complementary Type II RE ✗
overhangs Type IIS RE ✓ (Golden Gate)
Terminal transferase activity of ✓ (TA cloning)
DNA polymerase
Asymmetric PCR ✗
Zinc finger nuclease ✗
Chemical method (e.g., ✗
phosphorothioate + iodine/
ethanol)
Nicking endonuclease ✗
Uracil excision-based technique ✓ (USER)
Exonuclease ✓ (In-Fusion)
Gibson Assembly ✓ (Gibson
Assembly)
B Homologous sequences In vitro recombination ✓ (Gateway,
GeneArt Seamless)
In vivo recombination ✗
Yeast homologous recombination ✗
C Overlapping PCR PCR-based assembly ✗
D Megaprimers GOI as primer for whole-plasmid ✗
amplification
E Bridged ligation Ligase cycling reaction ✗
same principle, a sticky-/blunt-end ligation was used to clone two DNA fragments
into a vector [6]. simpleUSER cloning was developed to simplify the USER cloning
procedure by modifying the plasmid preparation step [7]. Instead of using a combi-
nation of one Type II RE (e.g., XbaI) and one nicking endonuclease (e.g., Nt.BbvCI)
to create a linearized vector with two different overhangs [8], simpleUSER cloning
made use of a single nicking endonuclease (i.e., Nb.BtsI). The same authors also
reported nicking cloning [7], in which the overhangs on the DNA insert were cre-
ated using a single nicking endonuclease (i.e., Nb.BbvCI), thereby abolishing the
need of a USER enzyme mix [uracil-DNA glycosylase (UDG) and the DNA
glycosylase-lyase endonuclease VIII] and uracil-containing primers. Nicking clon-
ing shares identical principle with the Nicking DNA Endonuclease (NiDE) method
[9]. In comparison, NiDE is more elegant as the same nicking endonuclease was
used to prepare both the plasmid and the DNA insert.
Complementary Annealing Mediated by Exonuclease (CAME) uses exonuclease
activity to expose the complementary overhangs on the DNA insert and the vector
[10]. Enzymes that could be used to recess the DNA ends include Pfu DNA poly-
merase (3′→5′ exonuclease activity), T4 DNA polymerase (3′→5′ exonuclease
activity), and λ exonuclease (5′→3′ exonuclease activity). This method is, in prin-
ciple, identical to In-Fusion cloning.
There are an increasing number of publications in the area of multiple DNA frag-
ment assembly. Gibson Assembly, developed in 2009, is an isothermal reaction
(50 °C) that uses the concerted action of three enzymes. T5 exonuclease generates
long overhangs, Phusion/Taq DNA polymerase fills in the gaps of the annealed
single-strand regions, and Taq DNA ligase seals the nicks of the annealed DNA
[11]. The simplicity of this method has made it the method of choice for many
researchers [12]. Variations of Gibson Assembly mainly revolve around the enzyme
mix used. Hot Fusion omits the Taq DNA ligase in its enzyme mix to achieve a
lower self-ligation of the base vector and, thus, a higher assembly efficiency com-
pared to Gibson Assembly (92% vs 70%) for a seven-DNA fragment assembly [13].
The Multiple Patch Cloning (MUPAC) uses T5 exonuclease, Klenow fragment
(exo-), and T4 DNA ligase to lower the isothermal reaction temperature to 37 °C,
which facilitates the assembly of fragments with shorter overhang (16 nts) [14].
Single Strand Assembly (SSA) is another variation of Gibson Assembly. Instead
of assembling double-stranded DNA fragments, the authors used long single-
stranded oligonucleotides with a homology sequence of 20 bp to enable Gibson
Assembly. This approach is particularly useful when investigating transcriptional
element (e.g., promoter) and translational element [e.g., ribosome-binding site
(RBS)]. Synthetic libraries of these elements can be easily synthesized as long
single-stranded oligonucleotides.
Further to Gibson Assembly, USER cloning has also been used for multiple DNA
fragment assembly to create expression vector for mammalian cell engineering [15],
integrative vector set for Saccharomyces cerevisiae [16], and vector for insect cell-
baculovirus expression system [17].
Biopart Assembly Standard for Idempotent Cloning (BASIC) relies on multiple
restrictions/ligations of the DNA fragments and linkers with BsaI (a Type IIS RE)
and T4 DNA ligase to assemble up to seven fragments [18]. The authors were able
to recapitulate the BsaI sites for subsequent rounds of assembly by methylation of
the specified linker oligonucleotides to avoid cleavage.
8.2.2 DNA Assembly Based on Homologous Sequences
Methods that use homologous sequences eliminate the need of complementary

overhangs between the GOI(s) and vector. This group of methods does not use REs
or ligases but instead utilizes DNA recombination of homologous sequences.
Recombination was demonstrated in vitro using E. coli cell extract or recombinases
and in vivo in E. coli or yeast. Recombination efficiency can further be improved by
including the λ prophage Red recombination for both in vivo and in vitro systems.
Advanced Quick Assembly (AQUA) is an in vivo homologous recombination
method for DNA assembly of up to four fragments using a 32 bp homologous
sequence [19]. The authors demonstrated the recombination using chemically
prepared E. coli TOP10 cells and compared its efficiency to that of five different
commercially available competent cells [i.e., TOP10, NEB5α, NEB10β,
BL21(DE3), and JM109]. The efficiencies across all E. coli strains for two DNA
fragment assembly were between 83–100%. Jacobus et al. investigated the opti-
mal conditions for in vivo homologous recombination of two DNA fragments in
E. coli DH5α [20]. When changing stoichiometry of vector to insert (from 1:1 to
1:4), the authors did not observe significant difference in the number of colonies
obtained. The authors did, however, observe an optimal number of colonies were
obtained when they used 150 ng of linearized pUC19 vector at a vector to insert
ratio of 1:2. The cloning efficiency increases from 25% to 70% and to 100%,
when the homology sequence increases from 10 bp to 20 bp and 30 bp, respec-
tively. In vivo recombination was also tested for E. coli XL10-Gold [21], pointing
to the fact that most of the laboratory E. coli strains can in fact be used for DNA
assembly via recombination.
In an attempt to make Seamless Ligation Cloning Extract (SLiCE) [22], an in vitro
homologous recombination method, more economically accessible, Motohashi and
coworkers investigated a homemade bacterial cell extract prepared from E. coli
JM109 using a Tris-HCl/Triton X-100 cell lytic buffer [23] instead of a commercial
cell lytic buffer (CelLytic B Cell Lysis Reagent from Sigma) as in the original proto-
col [22]. Both the homemade extract and the commercial In-Fusion HD Cloning Kit
provided comparable cloning efficiency, but the commercial kit generated about
twice the number of clones compared to the homemade extract [24].
HomeRun Vector Assembly System (HVAS) utilized both Gateway cloning (a
recombination system) and homing endonucleases for DNA assembly [25]. Despite
having the potential for high-throughput automation in cloning, this approach is
limited by the number of homing endonucleases available. Homing endonucleases
are double-stranded DNases that have large, asymmetric recognition sites (12–
40 bps). Currently, there are only four commercialized by New England Biolabs,
i.e., I-CeuI, I-SceI, PI-PspI, and PI-SceI.
Yeast-based homology recombination has been used since 1987 for DNA assem-
bly [26]. This powerful cloning method is not more widely adopted as it requires a
yeast-compatible shuttle vector. To increase the versatility of yeast homologous
recombination, the groups of Belden [27] and Andréasson [28] have both adapted
yeast homologous recombination cloning for use with bacteria plasmids. Belden
and coworkers achieved this by including a DNA fragment that contains both the
2-micron origin of replication (2 μm ori) and the ura3 gene for selection in yeast,
during a recombination of up to four DNA fragments [27]. Alternatively, Andréasson
and coworkers overcame the problem by adding to the bacteria plasmid a short
autonomous replication sequence (ARS) and a centromere sequence (CEN) for sta-
ble single-copy replication and mitotic segregation in yeast. Further, the authors
cloned the yeast TEF promoter of the kanMX cassette (a resistance marker cassette
for G418) upstream of the kanamycin resistance gene and were able to use the kan
gene for selection in both yeast and E. coli. The final plasmid became a shuttle vec-
tor that the authors used to demonstrate yeast homologous recombination of up to
seven DNA fragments. Yeast-based cloning method is proven an efficient way of
making large or complex DNA constructs, as demonstrated by the assembly of the
entire Mycoplasma genitalium genome [29]. Kilaru et al. also applied yeast-based
recombination as a convenient way to construct vectors for fungus Zymoseptoria
tritici [30].
8.2.3 DNA Assembly Based on Overlapping PCR
In overlapping PCR, DNA fragments that share overlapping regions at both ends are
assembled using PCR. In other words, only DNA polymerase is required. Cao et al.
demonstrated assembly of three PCR fragments and a linear vector in a single PCR
using PrimeSTAR HS DNA polymerase [31]. Prior to assembly, all fragments and
vector were amplified in high-fidelity PCRs to incorporate 20–50 bp overlapping
regions at both termini.
8.2.4 DNA Assembly Based on Megaprimers
Cloning based on megaprimer strategy is, in principle, similar to QuikChange muta-

genesis. Following the initial amplification of GOI, the generated PCR products
served as megaprimers for the subsequent linear or exponential amplification of the
whole plasmid.
QuickStep-Cloning, developed in our group, is a sequence-independent ligation-
free method for directional cloning of a GOI into any plasmid at any position [32]. The
method uses megaprimers with 3′ overhangs generated from asymmetric PCRs, thus
allowing exponential amplification of the plasmid. The generated nicked plasmids can
be transformed directly into expression host without DNA ligation. Based on the time
requirement and the cloning efficiency, we are confident to say that QuickStep-Cloning
fares exceptionally well in comparison to other megaprimer-based cloning methods.
Recombination-Assisted Megaprimer (RAM) [33] was designed to complement
Restriction-Free (RF) cloning [34] by revising the method into an exponential amplifi-
cation. RAM cloning involves three steps, i.e., megaprimer synthesis, integration of
megaprimer into the target vector, and in vivo homologous recombination for end join-
ing using E. coli [33].
8.2.5 DNA Assembly Based on Bridged Ligation
Ligase Cycling Reaction (LCR) is a one-step, scarless DNA assembly method [35].
It uses single-stranded bridging oligos complementary to the ends of adjoining
DNA fragments, thermostable ligase, and multiple cycles of denaturation–anneal-
ing–ligation to assemble complex DNA constructs. The authors demonstrated one-
step assembly of up to 20 DNA parts and up to 20 kb DNA constructs with very low
single-nucleotide polymorphisms (<1 per 25 kb) and insertion/deletion (<1 per
50 kb). Zhao group combined this automation-friendly LCR with in vivo yeast-
based DNA assembly for the construction of large biochemical pathways [36].
8.2.6 “De-cloning”
In the Sects. 8.2.1, 8.2.2, 8.2.3, 8.2.4, and 8.2.5, we have mainly focused on meth-
ods that “add” or “insert” a DNA into a construct. Equally important, we need the
capability to “remove” or “delete” a DNA sequence.
Krishnamurthy et al. described a multiplex gene removal protocol using a two-

step PCR [37]. They demonstrated the method through removing three genes from
a plasmid. In the first PCR, segments to be kept are amplified. These PCR fragments
are subsequently assembled in a PCR by incorporating single-stranded oligonucle-
otides of 40 bases in length, sharing a 20-nt complementarity with the two frag-
ments to be connected. In fact, most DNA assembly methods can be adapted for
gene removal. AQUA cloning (described in Sect. 8.2.2), for instance, was shown to
be a good tool for deleting a DNA sequence [19].
8.2.7 Trends in DNA Assembly Methods
Through following the development of methods for DNA assembly, we observe a

few trends in the development of DNA assembly methods: (1) The ability to synthe-
size long oligonucleotides with high accuracy and at a low cost has opened up many
possibilities for new method development. (2) DNA polymerases (e.g., Q5 and
Phusion) that display high fidelity, high processivity, and capability of amplifying
DNA with high GC content are increasingly used to enhance efficiency and avoid
unwanted mutations. (3) Many methods now involve whole-plasmid amplification or
preparing plasmid with a PCR, partly owing to the availability of high-fidelity DNA
polymerases. (4) Many variations of QuikChange protocol and megaprimer-based
integration are continuously being developed. (5) Intermediate DNA purification
steps (e.g., gel extraction, reaction cleanup) are undesirable. (6) Sequence-
independent and scarless insertion or assembly of DNA fragments (commonly
known as seamless) is emphasized. (7) Directional cloning or assembly in a desired
order is important. (8) Homologous recombination is gaining momentum as a main-
stream method. (9) Multiple fragment insertion or deletion in a single reaction (i.e.,
multiplexing) is preferred. (10) The use of combination of methods is getting more
common to create complex and large DNA constructs. (11) Many DNA assembly
methods have been adapted for focused mutagenesis, through incorporating primers
containing desired modifications. (12) Many methods for assembling DNA can, in
principle, be modified for “de-cloning.” (13) Automation-friendly methods (e.g.,
LCR) will continue to receive more attention.
8.3 Genetic Diversity
“Strategies and applications of in vitro mutagenesis,” written by Botstein and

Shortle, is perhaps one of the earliest comprehensive summaries of genetic diversity
creation methods [38]. Despite being published over 30 years ago, some of the
strategies described in this review (e.g., oligonucleotide-directed mutagenesis,

transposon mutagenesis) are still widely used in today’s laboratories.
Broadly, genetic diversity creation methods can be classified into three catego-
ries, i.e., random mutagenesis, focused mutagenesis, and DNA recombination. With
random mutagenesis, point mutations are introduced into the GOI at random
positions, typically through PCR employing an error-prone DNA polymerase.

Focused mutagenesis is the method of choice for altering a GOI at a selected posi-
tion. Point mutation, insertion, or deletion can be introduced by incorporating muta-
genic primer(s) containing the desired modification. Contrary to random mutagenesis
and focused mutagenesis in which the mutant library is prepared from a single
parental template, DNA recombination involves assembly of DNA fragments
derived from more than one parental sequence encoding proteins of identical/simi-
lar function.
Noteworthy, commercial kits are now available for each category to streamline
and standardize the experimental procedure. These kits contain all the components
required (e.g., buffers, chemicals, enzymes, control DNA, and competent cells) for
creating a gene library. They are also accompanied with a comprehensive manual,
describing the background/principle of the method, kit components, recipes for
media/buffers, step-by-step protocol, result expected, and troubleshooting guide.
This is particularly helpful for newcomers who are yet to develop their competency
in molecular biology and protein engineering. In this section, we discuss recent
development in each category focusing on those methods reported after 2013 and
provide a summary of commercial kits available in the market.
8.3.1 Random Mutagenesis
Ever since Leung described the protocol of error-prone polymerase chain reaction
(epPCR) in 1989, in which the fidelity of Taq DNA polymerase was intentionally
reduced by the use of suboptimal reaction conditions (e.g., high MgCl2 concentration
in the presence of MnCl2 and high number of PCR cycles starting with a low tem-
plate concentration) [39], this particular genetic diversity creation method remains
the front-runner in terms of its usage frequency. The popularity of epPCR is attrib-
uted to its technical simplicity and cost-effectiveness [2]. Not surprising, we continue
to see modifications being made to epPCR and diversification of its applications.
AXM mutagenesis was developed to eliminate the need of subcloning an epPCR
library [40]. GOI is amplified under error-prone conditions (70 mM MgCl2, 5 mM
MnCl2, and unbalanced nucleotide concentrations), using a reverse primer contain-
ing phosphorothioate linkages on its 5′ end. The double-stranded PCR product is
treated with T7 exonuclease to selectively degrade the unmodified strand. The
resulting ssDNA then acts as a megaprimer, which anneals to a uracilated, circular,
single-stranded phagemid DNA and primes DNA polymerization, similar to a
Kunkel mutagenesis [41]. The ligated heteroduplex product is then transformed into
E. coli, where the uracilated strand is cleaved in vivo by uracil-N-glycosylase.
Although the subcloning step is bypassed, the method necessitates a laborious prep-
aration of DNA template, which is less appealing. The same group also develops an
E. coli strain expressing Eco29kI restriction-methylation system to restrict incom-
ing parental DNA in the transformed cells [42].
Instead of accumulating mutations on the GOI, Schwaneberg group extended the
application of epPCR to mutagenize vector backbone to increase recombinant
protein expression, in a method called epMEGAWHOP [43]. GOI is first amplified

in a PCR. The PCR product serves as a megaprimer for the subsequent whole-
plasmid amplification that is conducted under error-prone condition (0.05 mM
Mn2+). After DpnI digestion to remove the methylated or hemimethylated parental
template, the resulting DNA is transformed into E. coli DH5α. Following plasmid
isolation and transformation into an expression host, the library is screened for
enhanced protein expression. The group demonstrated the applicability by enhanc-
ing the expression of cellulose (2× improvement), lipase (2×), and protease (4×),
which were cloned into pET28a(+) (5369 bp), pET22b(+) (5493 bp), and pHY-
300PLK (4870 bp), respectively. This strategy does not require a priori knowledge
of the bacterial transcription/translation machinery and is applicable to all enzymes,
expression vectors, and bacterial hosts. Potentially, the same strategy could be used
to increase the stability and the copy number of a plasmid.
Building on the previously reported TriNEx method [44], Jones group reported
an approach to introduce in-frame codon replacement with an amber stop codon
(TAG), as a means to incorporate noncanonical amino acid into protein [45]. The
method was verified by incorporating p-azido-L-phenylalanine or p-iodo-L-
phenylalanine into cytochrome b562 and keratinocyte growth factor. The possibility
to expand the genetic code would mean that the protein sequence space can further
be broadened from 20N to (20 + AANC)N, in which N represents the number of amino
acids within a protein and AANC is the number of noncanonical amino acids added
to the pool of 20 naturally occurring amino acids.
8.3.1.1 Commercial Kits for Random Mutagenesis

To provide a selection guide to readers, particularly newcomers, we subdivide the
commercial kits for random mutagenesis into four categories (Table 8.2). Kits in RI
category are PCR kits to conduct epPCR. Briefly, GOI is amplified under error-
prone conditions (e.g., high Mg2+ concentration, addition of Mn2+, unbalanced
dNTP concentration, or combination of these factors) or using an error-prone DNA
polymerase (e.g., Mutazyme I and/or Taq). In RII category, mutations are intro-
duced via the use of nucleotide analogues (e.g., dPTP, 8-oxo-dGTP) or DNA-
modifying chemicals. Kit in RIII contains E. coli XL1-Red competent cells ready
for plasmid transformation. This is a mutator strain that is deficient in three DNA
repair pathways: mutS (error-prone mismatch repair), mutD (deficient in 3′→5′ exo-
nuclease of DNA polymerase III), and mutT (unable to hydrolyze 8-oxo-dGTP).
Kits in category RIV are more recent compared to the other three categories, and
they involve the use of a transposase (e.g., MuA or Tn5) for in vitro transposition.
Through proper design of a transposon cassette, random insertion or deletion is pos-
sible with these kits.
8.3.1.2 Mutational Spectrum and Quality of a Random Mutagenesis

Library
While applying a random mutagenesis method or developing a new one for directed
evolution, key considerations include mutation frequency (i.e., mutations/kb), dis-
tribution of mutations, mutational spectrum (e.g., transition, transversion, insertion,
Table 8.2 Commercially available random mutagenesis kits
Mutation Mutational
Category Kit Company Principle Type of mutations frequency spectrum
RI JBS Error-Prone Kit Jena Addition of Mn2+ and unbalanced dNTP Point mutation 6–20 nt/kb Not reported
Bioscience concentration
Diversify PCR Clontech Addition of Mn2+ and unbalanced dNTP Point mutation 2–8 nt/kb Ts, 47–80%
Random concentration Tv, 20–53%
Mutagenesis Kit
GeneMorph II Agilent Blend of Mutazyme I and Taq DNA Point mutation 3–16 nt/kb Ts, 43%
polymerase mutant Tv, 51.4%
Indel, 5.5%
RII AquaMutant MoBiTec Chemical mutagen Point mutation 50 nt/kb Ts, 50–60%
Random Tv, 30–40%
Mutagenesis
8 Back to Basics: Creating Genetic Diversity
Reagent Kit
JBS dNTP- Jena Nucleotide analogues (dPTP, 8-oxo-dGTP) Point mutation up to 200 nt/bp Ts if dPTP is used
Mutagenesis Kit Bioscience Tv if 8-oxodGTP
is used
RIII XL1-Red Agilent E. coli strain deficient in three DNA repair Point mutation Not reported Not reported
pathways: mutS (error-prone mismatch
repair), mutD (deficient in 3′–5′
exonuclease of DNA polymerase III), and
mutT (unable to hydrolyze 8-oxo-dGTP)
RIV Mutation Generation Thermo In vitro transposition using MuA 15 bp insertion Transposition N/A
System Kit Scientific transposase frequency =
0.5–20%
EZ-Tn5 In-Frame Epicentre In vitro transposition using Tn5 transposase In-frame 19 codon Not reported N/A
Insertion Kit insertion
Stop Generation Thermo In vitro transposition using MuA C-terminal deletion Not reported N/A
System Kit Scientific transposase
211
deletion, consecutive nucleotide substitution), amino acid substitution pattern, per-

centage of unique sequences (or percentage of duplicated sequences), and percent-
age of wild-type sequences. These parameters are often used to assess the quality of
the resultant mutant library and have direct impact on the screening effort one needs
to invest. These are also the factors one needs to contemplate when choosing a com-
mercial kit to do the job (Table 8.2), besides its price and the reliability of the kit.
Quantifying the quality of a mutant library requires sequencing large number of
clones to obtain statistically significant data. Schwaneberg group attempted this by
preparing three random mutagenesis libraries of lipase (549 bp) and sequencing
1000 mutations per library [46]. These three libraries (epPCR-low, epPCR-high,
and SeSaM-Tv P/P) were prepared with standard epPCR protocols with a low muta-
tion frequency (0.1 mM MnCl2) or a high mutation frequency (0.3 mM MnCl2), as
well as with the Sequence Saturation Mutagenesis (SeSaM) [47]. Transitions were
predominant in both epPCR libraries (>72%), while transversions were enriched in
the SeSaM library (43%). Further, consecutive nucleotide substitutions occurred at
a higher frequency (30%) in the SeSaM library, while retaining high fraction of
clones expressing active lipase (52%). Interestingly, the amino acid substitution pat-
tern of both epPCR libraries complements that of the SeSaM library in terms of the
amino acid’s side-chain property. This important piece of work has conveyed a few
key messages: (1) the field of directed evolution can accommodate more genetic
diversity creation methods and should encourage their developments, (2) combina-
tion of methods allows exploring larger diversity, and (3) libraries created by afford-
able methodology (e.g., epPCR, SeSaM) are often sufficient to identify improved
variants.
8.3.2 Focused Mutagenesis
First described by Michael Smith in 1978 at the University of British Columbia,

Site-Directed Mutagenesis (SDM) has quickly become an invaluable tool for pro-
tein engineering and for studying structure-function relationships [48]. In Smith’s
oligonucleotide-directed mutagenesis, a short synthetic oligonucleotide of 12 bases
containing a mismatch was used to amplify a single-stranded phage DNA, using
combination of E. coli DNA polymerase I and T4 DNA ligase. Smith and his team
demonstrated that it was possible to introduce a permanent mutation in the circular
phage DNA, resulting in a phenotypic change. This seminar work has inspired the
development of QuikChange [49] and related techniques that are widely used for
focused mutagenesis [2].
Important to note, a lot of the techniques described in Sect. 8.2 for molecular
cloning or DNA assembly have now been adapted for focused mutagenesis. This
further supports the presentation of both DNA assembly and mutagenesis together
in this chapter.
QuikChange is a straightforward method for focused mutagenesis, using a pair
of overlapping mutagenic primers to linearly amplify the whole plasmid harboring
the GOI. Further to using Pfu DNA polymerase or its derivatives for QuikChange
mutagenesis, Q5 DNA polymerase (New England Biolabs) has been demonstrated

to be a suitable alternative [50]. Q5 is a high-fidelity, thermostable DNA polymerase
with 3´→5´ exonuclease activity, fused to a processivity-enhancing Sso7d domain
to support robust DNA amplification. Q5 is reported to have an error rate >100-fold
lower than that of Taq and 12-fold lower than that of Pfu. The other attractive feature
is its high polymerization rate of 10–40 s/kb depending on the template used. Xia
et al. attempted to revise the QuikChange protocol by using partially overlapping
primer and Phusion DNA polymerase [51], which has an error rate of >50-fold
lower than that of Taq and 6-fold lower than that of Pfu. The authors provided new
insights into the molecular mechanism of QuikChange. They proposed that their
revised protocol is an exponential amplification process. The resultant linear DNA
products with homologous ends are joined to generate circular plasmids within E.
coli via homologous recombination. From what we have gathered so far, Q5 and
Phusion have been popular DNA polymerases for development of molecular biol-
ogy techniques.
In our experience, the efficacy of QuikChange mutagenesis also depends on the
plasmid itself. The method does occasionally fail in cases involving amplifying a
large plasmid or a plasmid with high GC content. To overcome this, a Cut-and-Paste-
based Cloning Strategy (CPCS) was described for focused mutagenesis of a large
gene [52]. Target gene is split into two segments at the site for mutagenesis. Each
segment is amplified with one flanking primer and one mutagenic primer. The two
PCR fragments are individually cloned into vectors using TA cloning. Subsequently,
one fragment is cut out with a pair of restriction enzymes and inserted into the vector
containing the other fragment. Comparatively, the method employing Type IIS RE
for mutagenesis at multiple sites is far more elegant [53]. This approach is essentially
identical to Golden Gate Assembly, which exploits the ability of Type IIS RE [e.g.,
BsaI, BbsI, BsmBI/Esp3I] to cleave DNA outside of its recognition sequence. The
fragments to be assembled are designed in such a way that the Type IIS recognition
site is distal to the cleavage site, such that the Type IIS RE can remove its own rec-
ognition sequence completely from the assembly. Key advantages of such an arrange-
ment are threefold: (1) the overhang sequence created is not dictated by the RE, and
therefore, no scar sequence is introduced, (2) the fragment-specific sequence of the
overhangs allows assembly of multiple fragments simultaneously in the desired
order, and (3) the restriction site is eliminated from the ligated product. The net result
is an ordered and seamless assembly of DNA fragments in a one-pot reaction.
MUPAC, a Gibson Assembly derivative described in Sect. 8.2.1, is another technique
that can be used for molecular cloning, DNA assembly, and focused mutagenesis
[14]. As the overlapping sequence of adjoining fragments is much longer in Gibson
Assembly than those used in Golden Gate Assembly (typically an overhang of 4 nts),
higher percentage of correct assemblies is expected with a Gibson Assembly
approach. The authors of MUPAC reported >90% efficiency in assembly of six frag-
ments to introduce five mutations. Overlapping Extension Polymerase Chain
Reaction (OE-PCR) is perhaps a more economical method for introducing multiple
changes in a DNA sequence [54], compared to Gibson Assembly and Golden Gate
Assembly. The principle of OE-PCR is essentially the same, which involves
assembly of PCR fragments sharing overlapping regions, using DNA polymerase

such as Phusion. Another alternative is in vitro or in vivo DNA recombination.
Motohashi extended the use of SLiCE from E. coli to site-directed mutagenesis, in a
process termed SLiCE-mediated PCR-based site-directed mutagenesis (SLiP) [55].
It is also possible to carry out focused mutagenesis with in vivo recombination using
E. coli [e.g., AQUA cloning (19)] or yeast [e.g., plasmid constructed by Belden’s
group (27)]. In all the seven methods described above (CPCS, Type IIS assisted,
MUPAC, OE-PCR, SLiP, AQUA, and yeast-based recombination), mutations are
introduced into the primers used for generating the PCR fragments.
Protein fusion with variable linker insertion (P-Link) is classified as a focused
mutagenesis method here, as it was designed to create fusion protein library with link-
ers of variable length [56]. In other words, the linker is the localized variable region.
The method capitalizes on creating complementary ends using iodine cleavage of
phosphorothioate linkages within the primers. In P-Link, the two genes encoding the
proteins to be fused are first cloned into a vector, and this construct is served as the
template DNA for subsequent PCR to insert linker region. Two partially overlapping
primers, containing the linker sequence, are used to amplify the entire plasmid. Upon
DpnI digestion to remove the template DNA and iodine cleavage to expose comple-
mentary ends and enable annealing, the product is transformed into E. coli for protein
expression. Recombineering of Ends of Linearised Plasmids after PCR (REPLACR)
is a method for creating substitution, insertion, and deletion, which relies on in vivo
recombineering [57]. Partially overlapping primers containing the desired mutation
are used to amplify the whole plasmid. This generates a linear PCR product with both
the ends contain overlapping sequences for recombination. Following DpnI digestion
of template DNA, the product is transformed into E. coli expressing the recombineer-
ing proteins (Redγ, β, α, and RecA). P-Link and REPLACR are conceptually similar,
and mutations are designed into the primers like methods in the previous paragraph.
Their main difference is how the ends of the linear DNA product are joined.
Contrary to all the abovementioned methods that are developed for creating genetic
diversity using E. coli, Mutagenic Organized Recombination Process by Homologous
In Vivo Grouping (MORPHING) is a method to generate diversity at specific regions
by taking advantage of the efficient DNA recombination in S. cerevisiae [58]. GOI is
divided into segments, and each segment is amplified individually using a PCR. Only
the segments targeted for mutagenesis are amplified under error-prone conditions. As
all the adjoining fragments share homologous sequences, this pool of fragments can
be co-transformed into S. cerevisiae along with a linearized plasmid.
8.3.2.1 Randomization Scheme

Having decided on which residues to target for mutation, the next immediate task is
to define a randomization scheme for primer design. Through the use of a specific
codon or a degenerate codon, a residue can be substituted to a specific amino acid, a
set of amino acids, or all 20 canonical amino acids. The choice of a degenerate codon
is important as it determines the library size and, therefore, the screening effort.
In Table 8.3, we summarize the most commonly used degenerate codons for
focused mutagenesis [60–63]. Traditionally, one would use the codon NNN for
Table 8.3 Comparison of randomization schemes
Number of
Randomization oligonucleotide Mixing ratio of Number of Probability of Amino acids represented Redundant Amino acid
schemea,b syntheses requiredc oligonucleotides codons stop codons in the poold codons occurrence
NNN (64) 1 1 64 3/64 All (20] 41 Biased
NNK (32) 1 1 32 1/32 All (20) 11 Biased
NNS (32) 1 1 32 1/32 All (20) 11 Biased
NDT (12) 1 1 12 0 Mixture of polar, 0 Unbiased
nonpolar, positive, and
negative: G, V, L, I, C, N,
H, S, D, R, F, Y (12)
NTN (16) 1 1 16 0 Nonpolar: M, F, V, L, I 11 Biased
(5)
NAN (16) 1 1 16 2/16 Charged, large side 7 Unbiased
chains: Y, H, Q, N, K, D,
E (7)
NCN (16) 1 1 16 0 Smaller side chains, 12 Unbiased
polar and nonpolar: S, P,
T, A (4)
RST (4) 1 1 4 0 Small side chains: G, A, 0 Unbiased
S, T (4)
NDT (12) 4 12/20 20 0 All (20) 0 Unbiased
VMA (6) 6/20
ATG (1) 1/20
TGG (1) 1/20
NDT (12) 3 12/22 22 0 All (20) 2 Almost
9/22 unbiased
VHG (9)
TGG (1) 1/22
215
(continued)
216
Number of
Randomization oligonucleotide Mixing ratio of Number of Probability of Amino acids represented Redundant Amino acid
schemea,b syntheses requiredc oligonucleotides codons stop codons in the poold codons occurrence
HAT (3) 4 3/20 29 0 All (20) 9 Unbiasede
6/20
VMR (12) 8/20
WKK (8) 3/20
GDY (6)
SYN (16) 2 4/5 17 0 Aliphatic: A, V, L, I, P 12 Unbiasede
1/5 (5)
ATA (1)
YAY (4) 3 2/4 6 0 Aromatic: F, Y, W, H (4) 2 Unbiasede
TGG (1) 1/4
TTT (1) 1/4
RAM (4) 2 4/6 6 0 Polar: R, K, D, E, N, Q 0 Unbiased
CRA (2) 2/6 (6)
a
IUPAC nomenclatures: N = A/T/G/C, V = A/C/G, H = A/C/T, D = A/G/T, B = C/G/T, M = A/C, K = G/T, W = A/T, S = G/C, Y = C/T, R = A/G
b
Number in the bracket indicates number of codons
c
Number of oligonucleotide syntheses required for mutating a single strand
d
Number in the bracket indicates number of amino acids
e
Bias is removed through adjusting the ratio of oligonucleotides
K.L. Tee and T.S. Wong
saturation mutagenesis. Despite encoding all 20 possible amino acids, this set of 64
(4 × 4 × 4) codons also contains 3 stop codons and 41 redundant codons. Moreover,
the occurrence of each amino acid is not identical. Amino acids encoded by six
codons (e.g., S, L, R) occur at a higher probability, compared to those encoded by
one codon (e.g., M, W) or two codons (e.g., F, Y, H, Q, N, K, D, E, C). NNK and
NNS (each 32 codons) are proposed to reduce the redundancy of the mutant library.
Instead of using a single degenerate oligonucleotide, one can consider synthesizing
a set of oligonucleotides and mix them in a specific ratio to achieve an unbiased (or
less biased) occurrence of the desired amino acids. As an example, if NDT, VMA,
ATG, and TGG are mixed in the ratio of 12:6:1:1, all 20 amino acids are expected
to occur with the same probability. That being said, this ideal situation is only pos-
sible if (1) all oligonucleotides are synthesized perfectly (e.g., purify, sequence
accuracy, mixing of phosphoramidite building blocks), (2) the concentration of each
oligonucleotide is measured accurately, and (3) each oligonucleotide binds to its
template with the same affinity and is elongated by DNA polymerase in an unbiased
manner. In other words, the quality of the oligonucleotide is essential. Acevedo-
Rocha et al. explored the economics of focused mutagenesis libraries, by consider-
ing various randomization schemes, purity of oligonucleotide, and suppliers of
oligonucleotide [64].
Recently, our group reported OneClick, a user-friendly web-based program,
developed specifically for quick-and-easy design of focused mutagenesis experi-
ments [59]. To our best knowledge, OneClick is the only tool that offers a step-by-
step experimental design, from mutagenic primer design through to analysis of a
mutant library. Upon input of GOI sequence, OneClick designs the mutagenic prim-
ers according to user input, e.g., amino acid position to mutate, type of amino acid
substitutions (e.g., substitution to a group of amino acids with similar chemical
property), and type of mutagenic primers. OneClick has incorporated libraries of
commercially available plasmids and of DNA polymerases suitable for focused
mutagenesis. Therefore, OneClick also provides information such as PCR mixture
preparation, thermal cycling condition, the expected size of PCR product, and the
type of agar plate to use during bacterial transformation. Importantly, OneClick also
carries out a statistical analysis of the resultant mutant library, information of which
is important for selection/screening.
8.3.2.2 Commercially Available Kits

Similar to random mutagenesis, a host of commercial kits are available for focused
mutagenesis (Table 8.4). Again, to facilitate the selection of a kit, we classify these
kits into three subcategories (FI, FII, and FIII; Fig. 8.2).
Eighty percent of the commercial kits fall under category FI. This is perhaps not
a surprising figure, considering the ease of using a pair of primers to amplify the
whole plasmid. This primer pair could either be (FIA) two overlapping mutagenic
primers, (FIB) one non-mutagenic primer and one mutagenic primer that are non-
overlapping, and (FIC) one mutagenic primer and one selection primer. Among
these kits, the only subtle difference is the way of removing the parental DNA car-
rying the wild-type sequence. In Phusion Site-Directed Mutagenesis Kit, there is no
Table 8.4 Commercially available site-directed mutagenesis kits
218
Requirement
Primer of special
Category Kit Supplier Principle Primer designa modification Efficiency Enzyme strain Template
FI QuikChange Agilent Whole plasmid Two overlapping primers Nil >80% (one site) Derivative of PfuUltra, No dsDNA
Lightning amplification >55% (three sites) DpnI
ATB Ameritech Whole plasmid Two overlapping primers Nil Not reported Pfu Ultimate, DpnI No dsDNA
Biomedicines amplification
Muta-Direct SBS Gentech Whole plasmid Two overlapping primers Nil 100% (one site) Muta-direct, No dsDNA
amplification Mutazyme
TagMaster GM Whole plasmid Two overlapping primers Nil >95% (one site) Blend of Pfu and TagMaster dsDNA
Biosciences amplification 50% (two sites) Phusion cells
AMAP MBL Whole plasmid One mutagenic primer 5′-P >80% (one site) Pfu, Taq DNA ligase, No dsDNA
amplification >80% (two sites) DpnI
>60% (three sites)
40% (four/five sites)
Q5 New England Whole plasmid Two non-overlapping Nil >90% (one site) Q5, kinase, ligase, No dsDNA
Biolabs amplification primers DpnI
Phusion Thermo Whole plasmid Two non-overlapping 5′-P >80% (one site) Phusion Hot Start II, No dsDNA
Scientific amplification primers T4 DNA ligase
KOD Plus Toyobo Whole plasmid Two non-overlapping Nil 95% (one site) KOD Plus, T4 No dsDNA
amplification primers polynucleotide kinase,
T4 DNA ligase, DpnI
Change-IT Affymetrix Whole plasmid Two non-overlapping 5′-P 90% (one site) FideliTaq, DNA ligase, No dsDNA
amplification primers 90% (two sites) DpnI
80% (three sites)
EZchange Enzynomics Whole plasmid Two non-overlapping Nil Not reported nPfu-Forte, kinase, No dsDNA
amplification primers ligase, DpnI
Transformer Clontech Whole plasmid One mutagenic primer and 5′-P >70% (one site) T4 DNA polymerase, E. coli mutS dsDNA
amplification one selection primer T4 DNA ligase,
restriction enzyme
Mutan-Super TaKaRa Whole plasmid One mutagenic primer and 5′-P >80% (one site) TaKaRa LA Taq E. coli sup0 dsDNA
Express Km amplification one selection primer
K.L. Tee and T.S. Wong
FII Gibson Synthetic Assembly of Four partially overlapping Nil >90% (five sites) Non-disclosed enzyme No dsDNA
Assembly Genomics PCR fragments primers mix, DpnI
PickMutant Canvax Assembly of Four partially overlapping Nil Not reported Non-disclosed enzyme No dsDNA
PCR fragments primers mix
and cloning
vector
FIII GeneArt Thermo Assembly of Two overlapping primers Nil >90% (one site) DNA methylase, DH5α™- dsDNA
(Plus) Scientific PCR fragments >90% (two sites) non-disclosed enzyme T1R
>90% (three sites) mix
a
Primer design for mutagenesis at one site
219
a b c
Fig. 8.2 Principles of commercial kits for focused mutagenesis: (FI) whole-plasmid amplification
using (a) two overlapping mutagenic primers, (b) one non-mutagenic primer and one mutagenic
primer that are non-overlapping, and (c) one mutagenic primer and one selection primer, (FII)
Gibson Assembly, and (FIII) in vitro recombination
template removal step recommended. The supplier has argued that minute amount
of parental DNA is exponentially amplified in this method, and therefore, the frac-
tion of non-mutated template is minimal. Most kits apply DpnI to degrade methyl-
ated or hemimethylated parent DNA, which are usually plasmids isolated from
dam+ E. coli strains. Instead of relying on a RE, TagMaster cell can discriminate
parent DNA and novel synthesized daughter DNA. Therefore, parent plasmid tem-
plate DNA would be eliminated, and only daughter DNA synthesized in the muta-
genesis reaction can be enriched. In Transformer Site-Directed Mutagenesis Kit,
one mutagenic primer is used to introduce the desired mutation. It also employs an
additional selection primer containing a mutation in the recognition sequence for a
unique restriction enzyme site. The two primers simultaneously anneal to one strand
of the denatured double-stranded plasmid. After standard DNA elongation, ligation,
and a primary selection by restriction digest, the mixture of mutated and unmutated
plasmids is transformed into a mutS E. coli strain defective in mismatch repair.
Transformants are pooled, and plasmid DNA is prepared from the mixed bacterial
population. The isolated DNA is then subjected to a second selective restriction
enzyme digestion. Since the mutated DNA lacks the restriction enzyme recognition
site, it is resistant to digestion. A second transformation is then used to isolate the

individual transformant carrying the mutated DNA. Mutan-Super Express Km Site-
Directed Mutagenesis System also depends on a selection system. PCR is performed
using an oligonucleotide containing the desired mutation and a selection primer to
revert the amber mutations on the kanamycin-resistant gene present in the plasmid.
When the nicked DNA is transformed into supo E. coli strain, the nick is repaired,
and then transformants containing a site-specific mutation can be obtained after
culturing in the media containing kanamycin.
Kits in category FII and FIII rely on assembling PCR fragments, using either
Gibson Assembly (FII) or in vitro recombination (FIII). These kits are more effi-
cient for performing mutagenesis at multiple sites.
Recently, Agilent Technologies introduced a new product termed QuikChange
HT Protein Engineering System. This system capitalizes on Agilent’s capability of
synthesizing long oligonucleotides (~150 bases). The workflow consists of three
steps: (1) select one or more 50-amino acid mutational region in a protein sequence,
design a set of oligonucleotides containing desired modifications (random muta-
tions, site-specific mutations, or combinatorial mutations), and place the oligo
order; (2) amplify each oligo set with a pair of primers; and (3) perform QuikChange
mutagenesis with each amplified oligo set. This system does offer multiple advan-
tages, including elimination of codon redundancy and bias. Further, this system is
more economical compared to a synthetic library.
8.3.3 DNA Recombination
Invented by Stemmer, preparation of a shuffled DNA library involves DNA

fragmentation by DNase I followed by fragment assembly via PCR [65]. Besides
Stemmer shuffling (or DNA shuffling) [65], there is a host of other DNA recombina-
tion methods, which include Staggered Extension Process (StEP) [66], Incremental
Truncation for the Creation of Hybrid Enzymes (ITCHY) [67], Sequence Homology-
Independent Protein Recombination (SHIPREC) [68], In Vivo Shuffling [69],
Combinatorial Libraries Enhanced by Recombination in Yeast (CLERY) [70],
Nonhomologous Random Recombination (NRR) [71] etc. StEP is frequently used to
combine beneficial mutations found in random mutagenesis experiments. ITCHY
and SHIPREC are, in principle, identical. Both methods allow creating chimeras of
one crossover from two nonhomologous sequences. Contrary to most DNA recom-
bination methods, In Vivo Shuffling and CLERY rely on recombination within E. coli
and yeast, respectively. NRR, even though a more laborious process, is homology
independent and enables creating chimeras of more than one crossovers.
In comparison with random mutagenesis and focused mutagenesis, we see far
less molecular techniques being developed for DNA recombination. Lehtonen et al.
reported a minor change to Stemmer protocol [71] by using the Gateway cloning
procedure directly after the gene reassembly reaction, without additional purifica-
tion and amplification.
8.3.3.1 Commercial Kits for DNA Recombination

The only commercial kit available for DNA recombination is the JBS DNA-
Shuffling Kit (Jena Biosciences), which in essence adopted the principle of the
Stemmer protocol [65]. It can be applied for recombination of gene fragments origi-
nating from one or several related genes. When used for single-gene shuffling, only
one gene is digested and subsequently reassembled resulting in point mutations at a
mutation frequency of 7 nt/kb.
8.3.4 K
ey Milestones and the Trends in the Development
of Genetic Diversity Creation Methods
We consider it appropriate to end this book chapter with a reflection on the key
milestones in the field of genetic diversity creation. In Fig. 8.3, we cherry-pick some
methods that, in our opinion, are transformative or disruptive technologies for creat-
ing gene libraries. In the 1960s, Sol Spiegelman and coworkers demonstrated how
RNA molecules could be evolved in the test tube [72]. This seminar work is widely
acknowledged as the first in vitro Darwinian evolution experiment that paved the
way for the many directed evolution experiments that followed [73]. In the 1970s,
Michael Smith demonstrated oligonucleotide-directed mutagenesis, laying the
groundwork for focused mutagenesis [48]. It is fair to say that all DNA assembly
methods and methods for creating genetic diversity, applied today, rely on PCR and
Kary Mullis’s contribution shall not be forgotten [74]. The Nobel Prize in Chemistry
1993 was awarded “for contributions to the developments of methods within DNA-
based chemistry” jointly with one half to Kary Mullis “for his invention of the PCR
method” and with one half to Michael Smith “for his fundamental contributions to
the establishment of oligonucleotide-based, site-directed mutagenesis and its devel-
opment for protein studies.” Leung’s epPCR protocol [39] and Stemmer’s DNA
shuffling [65] are still popular and widely used in many laboratories, despite being
reported more than 20 years ago. In 1996, QuikChange was published [49], a highly
influential technique that has inspired the development of many focused mutagen-
esis methods and commercial kits [2]. Huimin Zhao and Frances Arnold reported
StEP in 1998 [66], an extremely convenient approach to recombine point mutations.
Stemmer and Arnold were awarded the Draper Prize 2011 for their contribution to
directed evolution. ITCHY was reported in 1999 to create combinatorial fusion
libraries between genes in a manner that is independent of DNA homology [67]. In
2000, the Mutazyme kit was commercialized, and it remains a popular kit to com-
plement epPCR [75]. Two years later, David Liu and coworkers published the NRR
that enabled DNA fragments to be randomly recombined in a length-controlled
manner without the need for sequence homology [76]. In 2004, we reported SeSaM
method to enrich the occurrence of transversion mutations and consecutive nucleo-
tide substitutions in a random mutagenesis experiment [47]. TriNEx was published
4 years later for trinucleotide exchange using an in vitro transposition approach
[44]. In 2015, Agilent released its QuikChange HT Protein Engineering System.
Fig. 8.3 Key milestones in the development of genetic diversity creation methods. Methods for
random mutagenesis are colored in light green, focused mutagenesis in light blue, and DNA
recombination in light orange
The timeline in Fig. 8.3 and the survey we have conducted allow us to spot a
few trends: (1) The pace of new method development will continue to be dictated
by the capability of oligonucleotide synthesis (length, accuracy, trimer phosphor-
amidite), the discovery of new enzymes for molecular biology, and the creativity
of researchers. (2) Genetic diversity creation methods will constantly be modified
to be compatible with more efficient cloning methods (e.g., Gibson Assembly,
Gateway cloning, In-Fusion). (3) Methods for single-gene mutagenesis will be
replaced by those for mutating multiple genes or DNA elements (e.g., promoter,
RBS). (4) E. coli-based methods will continue to dominate, but yeast-based meth-
ods will emerge to be equally powerful. (5) High-throughput mutagenesis meth-
ods (e.g., QuikChange HT Protein Engineering System) will eventually catch up
in terms of its usage frequency. (6) We envisage to see more automation in the
preparation of a gene library.
Conclusion
Genetic diversity creation is a rigorous field of research. It has never been stag-
nant, and we are flooded with a constant stream of new developments, making it
extremely difficult to keep up to date. This chapter has, hopefully, provided some
convenience and useful guides for readers.
References
1. Wong TS, Zhurina D, Schwaneberg U (2006) The diversity challenge in directed protein evo-
lution. Comb Chem High Throughput Screen 9:271–288
2. Tee KL, Wong TS (2013) Polishing the craft of genetic diversity creation in directed evolution.
Biotechnol Adv 31:1707–1721
3. Eriksen DT, Hsieh PC, Lynn P, Zhao H (2013) Directed evolution of a cellobiose utiliza-
tion pathway in Saccharomyces cerevisiae by simultaneously engineering multiple proteins.
Microb Cell Factories 12:61
4. Yuan Y, Zhao H (2013) Directed evolution of a highly efficient cellobiose utilizing pathway in
an industrial Saccharomyces cerevisiae strain. Biotechnol Bioeng 110:2874–2881
5. Gao S, Li Y, Zhang J, Chen H, Ren D, Zhang L, An Y (2014) A modified version of the
digestion-ligation cloning method for more efficient molecular cloning. Anal Biochem 453:55–57
6. Gao S, Zhang J, Miao T, Ma D, Su Y, An Y, Zhang Q (2015) A simple and convenient sticky/
blunt-end ligation method for fusion gene construction. Biochem Genet 53:42–48
7. Hansen NB, Lubeck M, Lubeck PS (2014) Advancing USER cloning into simpleUSER and
nicking cloning. J Microbiol Methods 96:42–49
8. Bitinaite J, Rubino M, Varma KH, Schildkraut I, Vaisvila R, Vaiskunaite R (2007) USER
friendly DNA engineering and cloning method by uracil excision. Nucleic Acids Res
35:1992–2002
9. Yang J, Zhang Z, Zhang XA, Luo Q (2010) A ligation-independent cloning method using nick-
ing DNA endonuclease. BioTechniques 49:817–821
10. Sun S, Huang H, Qi YB, Qiu M, Dai ZM (2015) Complementary annealing mediated by
exonuclease: a method for seamless cloning and conditioning site-directed mutagenesis.
Biotechnol Biotechnol Equip 29:105–110
11. Gibson DG, Young L, Chuang RY, Venter JC, Hutchison CA 3rd, Smith HO (2009) Enzymatic
assembly of DNA molecules up to several hundred kilobases. Nat Methods 6:343–345
12. Kahl LJ, Endy D (2013) A survey of enabling technologies in synthetic biology. J Biol Eng
7:13
13. Fu C, Donovan WP, Shikapwashya-Hasser O, Ye X, Cole RH (2014) Hot Fusion: an efficient
method to clone multiple DNA fragments as well as inverted repeats without ligase. PLoS
ONE 9:e115318
14. Taniguchi N, Nakayama S, Kawakami T, Murakami H (2013) Patch cloning method for mul-
tiple site-directed and saturation mutagenesis. BMC Biotechnol 13:91
15. Lund AM, Kildegaard HF, Petersen MB, Rank J, Hansen BG, Andersen MR, Mortensen UH
(2014) A versatile system for USER cloning-based assembly of expression vectors for mam-
malian cell engineering. PLoS ONE 9:e96693
16. Jensen NB, Strucko T, Kildegaard KR, David F, Maury J, Mortensen UH, Forster J, Nielsen J,
Borodina I (2014) EasyClone: method for iterative chromosomal integration of multiple genes
in Saccharomyces cerevisiae. FEMS Yeast Res 14:238–248
17. Zhang Z, Yang J, Barford D (2015) Recombinant expression and reconstitution of multiprotein
complexes by the USER cloning method in the insect cell-baculovirus expression system.
Methods 95:13–25
18. Storch M, Casini A, Mackrow B, Fleming T, Trewhitt H, Ellis T, Baldwin GS (2015) BASIC:
a new biopart assembly standard for idempotent cloning provides accurate, single-tier DNA
assembly for synthetic biology. ACS Synth Biol 4:781–787
19. Beyer HM, Gonschorek P, Samodelov SL, Meier M, Weber W, Zurbriggen MD (2015) AQUA
cloning: a versatile and simple enzyme-free cloning approach. PLoS ONE 10:e0137652
20. Jacobus AP, Gross J (2015) Optimal cloning of PCR fragments by homologous recombination
in Escherichia coli. PLoS ONE 10:e0119221
21. Wang Y, Liu Y, Chen J, Tang MJ, Zhang SL, Wei LN, Li CH, Wei DB (2015) Restriction-
ligation-free (RLF) cloning: a high-throughput cloning method by in vivo homologous recom-
bination of PCR products. Genet Mol Res: GMR 14:12306–12315
22. Zhang Y, Werling U, Edelmann W (2012) SLiCE: a novel bacterial cell extract-based DNA
cloning method. Nucleic Acids Res 40:e55
23. Okegawa Y, Motohashi K (2015) Evaluation of seamless ligation cloning extract preparation
methods from an Escherichia coli laboratory strain. Anal Biochem 486:51–53
24. Okegawa Y, Motohashi K (2015) A simple and ultra-low cost homemade seamless ligation
cloning extract (SLiCE) as an alternative to a commercially available seamless DNA cloning
kit. Biochem Biophys Rep 4:148–151
25. Li MV, Shukla D, Rhodes BH, Lall A, Shu J, Moriarity BS, Largaespada DA (2014) HomeRun
Vector Assembly System: a flexible and standardized cloning system for assembly of multi-
modular DNA constructs. PLoS ONE 9:e100948
26. Ma H, Kunes S, Schatz PJ, Botstein D (1987) Plasmid construction by homologous recombi-
nation in yeast. Gene 58:201–216
27. Joska TM, Mashruwala A, Boyd JM, Belden WJ (2014) A universal cloning method based on
yeast homologous recombination that is simple, efficient, and versatile. J Microbiol Methods
100:46–51
28. Holmberg MA, Gowda NKC, Andreasson C (2014) A versatile bacterial expression vector
designed for single-step cloning of multiple DNA fragments using homologous recombina-
tion. Protein Expr Purif 98:38–45
29. Gibson DG, Benders GA, Andrews-Pfannkoch C, Denisova EA, Baden-Tillson H, Zaveri J,
Stockwell TB, Brownley A, Thomas DW, Algire MA et al (2008) Complete chemical synthe-
sis, assembly, and cloning of a Mycoplasma genitalium genome. Science 319:1215–1220
30. Kilaru S, Steinberg G (2015) Yeast recombination-based cloning as an efficient way of con-
structing vectors for Zymoseptoria tritici. Fungal Genet Biol: FG & B 79:76–83
31. Cao P, Wang L, Zhou G, Wang Y, Chen Y (2014) Rapid assembly of multiple DNA frag-
ments through direct transformation of PCR products into E. coli and Lactobacillus. Plasmid
76C:40–46
32. Jajesniak P, Wong TS (2015) QuickStep-Cloning: a sequence-independent, ligation-free

method for rapid construction of recombinant plasmids. J Biol Eng 9
33. Mathieu J, Alvarez E, Alvarez PJ (2014) Recombination-assisted megaprimer (RAM) cloning.
MethodsX 1:23–29
34. van den Ent F, Lowe J (2006) RF cloning: a restriction-free method for inserting target genes
into plasmids. J Biochem Biophys Methods 67:67–74
35. de Kok S, Stanton LH, Slaby T, Durot M, Holmes VF, Patel KG, Platt D, Shapland EB, Serber
Z, Dean J et al (2014) Rapid and reliable DNA assembly via ligase cycling reaction. ACS
Synth Biol 3:97–106
36. Yuan Y, Andersen E, Zhao H (2016) Flexible and versatile strategy for the construction of large
biochemical pathways. ACS Synth Biol 5:46–52
37. Krishnamurthy VV, Khamo JS, Cho E, Schornak C, Zhang K (2015) Multiplex gene removal
by two-step polymerase chain reactions. Anal Biochem 481:7–9
38. Botstein D, Shortle D (1985) Strategies and applications of in vitro mutagenesis. Science
229:1193–1201
39. Leung DW, Chen EY, Goeddel DV (1989) Techniques 1:11–15
40. Holland EG, Buhr DL, Acca FE, Alderman D, Bovat K, Busygina V, Kay BK, Weiner MP,
Kiss MM (2013) AXM mutagenesis: an efficient means for the production of libraries for
directed evolution of proteins. J Immunol Methods 394:55–61
41. Kunkel TA (1985) Rapid and efficient site-specific mutagenesis without phenotypic selection.
Proc Natl Acad Sci U S A 82:488–492
42. Holland EG, Acca FE, Belanger KM, Bylo ME, Kay BK, Weiner MP, Kiss MM (2015) In vivo
elimination of parental clones in general and site-directed mutagenesis. J Immunol Methods
417:67–75
43. Jakob F, Lehmann C, Martinez R, Schwaneberg U (2013) Increasing protein production by
directed vector backbone evolution. AMB Express 3:39
44. Baldwin AJ, Busse K, Simm AM, Jones DD (2008) Expanded molecular diversity generation
during directed evolution by trinucleotide exchange (TriNEx). Nucleic Acids Res 36:e77
45. Arpino JA, Baldwin AJ, McGarrity AR, Tippmann EM, Jones DD (2015) In-frame amber
stop codon replacement mutagenesis for the directed evolution of proteins containing non-
canonical amino acids: identification of residues open to bio-orthogonal modification. PLoS
ONE 10:e0127504
46. Zhao J, Kardashliev T, Joelle Ruff A, Bocola M, Schwaneberg U (2014) Lessons from diver-
sity of directed evolution experiments by an analysis of 3000 mutations. Biotechnol Bioeng
111:2380–2389
47. Wong TS, Tee KL, Hauer B, Schwaneberg U (2004) Sequence saturation mutagenesis

(SeSaM): a novel method for directed evolution. Nucleic Acids Res 32:e26
48. Hutchison CA 3rd, Phillips S, Edgell MH, Gillam S, Jahnke P, Smith M (1978) Mutagenesis
at a specific position in a DNA sequence. J Biol Chem 253:6551–6560
49. Papworth C, Bauer JC, Braman J, Wright DA (1996) Site-directed mutagenesis in one day
with >80% efficiency. Strategies 9:3–4
50. Liu H, Ye R, Wang YY (2015) Highly efficient one-step PCR-based mutagenesis technique for
large plasmids using high-fidelity DNA polymerase. Genet Mol Res: GMR 14:3466–3473
51. Xia Y, Chu W, Qi Q, Xun L (2015) New insights into the QuikChange process guide the use of
Phusion DNA polymerase for site-directed mutagenesis. Nucleic Acids Res 43:e12
52. Wang C, Wang TY, Zhang LY, Gao XJ, Wang XW, Jin CJ (2015) Cut-and-paste-based cloning
strategy for large gene site-directed mutagenesis. Genet Mol Res: GMR 14:5585–5591
53. Zhang Z, Xu K, Xin Y, Zhang Z (2015) An efficient method for multiple site-directed muta-
genesis using type IIs restriction enzymes. Anal Biochem 476:26–28
54. Waneskog M, Bjerling P (2014) Multi-fragment site-directed mutagenic overlap extension
polymerase chain reaction as a competitive alternative to the enzymatic assembly method.
Anal Biochem 444:32–37
55. Motohashi K (2015) A simple and efficient seamless DNA cloning method using SLiCE from
Escherichia coli laboratory strains and its application to SLiP site-directed mutagenesis. BMC
Biotechnol 15:47
56. Belsare KD, Ruff AJ, Martinez R, Shivange AV, Mundhada H, Holtmann D, Schrader J,
Schwaneberg U (2014) P-LinK: a method for generating multicomponent cytochrome P450
fusions with variable linker length. BioTechniques 57:13–20
57. Trehan A, Kielbus M, Czapinski J, Stepulak A, Huhtaniemi I, Rivero-Muller A (2016)
REPLACR-mutagenesis, a one-step method for site-directed mutagenesis by recombineering.
Sci Rep 6:19121
58. Gonzalez-Perez D, Molina-Espeja P, Garcia-Ruiz E, Alcalde M (2014) Mutagenic Organized
Recombination Process by Homologous IN vivo Grouping (MORPHING) for directed enzyme
evolution. PLoS ONE 9:e90919
59. Warburton M, Omar Ali H, Liong WC, Othusitse AM, Abdullah Zubir AZ, Maddock S,
Wong TS (2015) OneClick: a program for designing focused mutagenesis experiments. AIMS
Bioeng 2:126–143
60. Currin A, Swainston N, Day PJ, Kell DB (2015) Synthetic biology for the directed evolution
of protein biocatalysts: navigating sequence space intelligently. Chem Soc Rev 44:1172–1239
61. Kille S, Acevedo-Rocha CG, Parra LP, Zhang ZG, Opperman DJ, Reetz MT, Acevedo JP
(2013) Reducing codon redundancy and screening effort of combinatorial protein libraries
created by saturation mutagenesis. ACS Synth Biol 2:83–92
62. Nov Y, Segev D (2013) Optimal codon randomization via mathematical programming. J Theor
Biol 335:147–152
63. Tang L, Gao H, Zhu X, Wang X, Zhou M, Jiang R (2012) Construction of “small-intelli-
gent” focused mutagenesis libraries using well-designed combinatorial degenerate primers.
BioTechniques 52:149–158
64. Acevedo-Rocha CG, Reetz MT, Nov Y (2015) Economical analysis of saturation mutagenesis
experiments. Sci Rep 5:10654
65. Stemmer WP (1994) DNA shuffling by random fragmentation and reassembly: in vitro recom-
bination for molecular evolution. Proc Natl Acad Sci U S A 91:10747–10751
66. Zhao H, Giver L, Shao Z, Affholter JA, Arnold FH (1998) Molecular evolution by staggered
extension process (StEP) in vitro recombination. Nat Biotechnol 16:258–261
67. Ostermeier M, Shim JH, Benkovic SJ (1999) A combinatorial approach to hybrid enzymes
independent of DNA homology. Nat Biotechnol 17:1205–1209
68. Sieber V, Martinez CA, Arnold FH (2001) Libraries of hybrid proteins from distantly related
sequences. Nat Biotechnol 19:456–460
69. Xu S, Ju J, Misono H, Ohnishi K (2006) Directed evolution of extradiol dioxygenase by a

novel in vivo DNA shuffling. Gene 368:126–137
70. Abecassis V, Pompon D, Truan G (2000) High efficiency family shuffling based on multi-step
PCR and in vivo DNA recombination in yeast: statistical and functional analysis of a combi-
natorial library between human cytochrome P450 1A1 and 1A2. Nucleic Acids Res 28:E88
71. Lehtonen SI, Taskinen B, Ojala E, Kukkurainen S, Rahikainen R, Riihimaki TA, Laitinen OH,
Kulomaa MS, Hytonen VP (2015) Efficient preparation of shuffled DNA libraries through
recombination (Gateway) cloning. Protein Eng Des Select: PEDS 28:23–28
72. Mills DR, Peterson RL, Spiegelman S (1967) An extracellular Darwinian experiment with a
self-duplicating nucleic acid molecule. Proc Natl Acad Sci U S A 58:217–224
73. Joyce GF (2007) Forty years of in vitro evolution. Angew Chem 46:6420–6436
74. Saiki RK, Scharf S, Faloona F, Mullis KB, Horn GT, Erlich HA, Arnheim N (1985) Enzymatic
amplification of beta-globin genomic sequences and restriction site analysis for diagnosis of
sickle cell anemia. Science 230:1350–1354
75. Cline J, Hogrefe H (2000) Randomize gene sequences with new PCR mutagenesis kit.
Strategies 13:157–162
76. Bittker JA, Le BV, Liu DR (2002) Nucleic acid evolution and minimization by nonhomologous
random recombination. Nat Biotechnol 20:1024–1029
Resurrected Ancestral Proteins
as Scaffolds for Protein Engineering 9
Valeria A. Risso and Jose M. Sanchez-Ruiz
Abstract
High stability and enhanced promiscuity (likely linked to conformational flex-
ibility/diversity) contribute to evolvability and are advantageous features in
protein scaffolds for laboratory-directed evolution and molecular design.
Furthermore, the two features are not necessarily incompatible, and proteins
may simultaneously be promiscuous/flexible and highly stable. In fact, it
appears plausible that the combination of the two features was not uncommon
among the most ancient proteins because (i) ancient life was likely thermo-
philic and (ii) ancient proteins were likely promiscuous generalists with broad
functionalities. Phylogenetic analyses allow the reconstruction of ancestral
sequences and provide an approach to explore the properties of ancient pro-
teins. High stability and promiscuity have been often found for proteins
encoded by reconstructed ancestral sequences, i.e., for “resurrected” ancestral
proteins. The combination of the two features, i.e., the ancestral hyperstable
generalist phenotype, has actually been obtained in recent studies. Ancestral
protein resurrection thus emerges as a useful source of scaffolds for protein
engineering.
V.A. Risso • J.M. Sanchez-Ruiz (*)

Facultad de Ciencias, Departamento de Química Física, Universidad de Granada,
18071 Granada, Spain

DOI 10.1007/978-3-319-50413-1_9
230 V.A. Risso and J.M. Sanchez-Ruiz
9.1 Very Brief Introduction to Ancestral Protein

A
Resurrection
The possibility of deriving plausible approximations to ancient protein sequences

from the known sequences of modern proteins was originally proposed by Pauling
and Zuckerkandl in the 1960s [106]. It remained, however, a theoretical idea until
the 1990s, when (i) advances in bioinformatics and the increasing availability of
protein sequences facilitated the reliable reconstruction of ancestral sequences and
(ii) advances in molecular biology methodologies allowed the preparation in the
laboratory (i.e., the “resurrection”) of the corresponding encoded proteins.
Derivation of ancestral sequences involves a phylogenetic analysis based on an
alignment of modern protein sequences and maximum likelihood or Bayesian esti-
mation of the sequences at the nodes of the phylogenetic tree. The reader is referred
to recent excellent reviews for detailed accounts of the procedures involved [13, 27,
91, 94, 127]. Nevertheless, an intuitive feeling about the process can be gained by
drawing an analogy with the reconstructions of extinct languages performed by
historical linguists since the late eighteenth century (Fig. 9.1). Knowledge of a
given word in several modern related languages can be used, together with plausible
models of word evolution, to derive a plausible reconstruction of the corresponding
word in the common ancestor languages [6, 21]. Extinct languages (known as pro-
tolanguages) have been thus reconstructed, including Proto-Germanic and Proto-
Indo-European (the common ancestor of Indo-European languages). Likewise, a
protein sequence can be viewed as a word written using an alphabet of 20 words,
and, therefore, the sequences (“words”) for a set of homologous proteins, together
with a model of protein evolution, can be used to reconstruct the sequence of the
common ancestor protein (i.e., the “ancestral word”).
9.2 xperimental Validation of Ancestral Protein

E
Resurrection: Phenotypic Robustness and Evolutionary
Narratives
Ancestral sequence reconstruction is an unavoidably uncertain task due to a number

of reasons: the evolutionary models used are necessarily simple (for instance, time-
invariant amino acid substitution matrices are typically used and site coevolution is
typically neglected), the set of modern sequences used as input for the analysis
might not be a fair representation of the extant (modern) sequence diversity, hori-
zontal gene transfer processes may complicate phylogenetic analysis, gene duplica-
tion events may add uncertainty by preventing a direct comparison of the tree
derived from the sequence alignment with known organisms phylogenies, etc.
It appears reasonable, therefore, to ask how reconstructed ancestral sequences
and the corresponding “resurrected” ancestral proteins (i.e., the proteins encoded by
the reconstructed sequences) can be validated. In this context, two important points
must be noted: (1) bioinformatics procedures used for ancestral sequence recon-
struction do not provide just a single reconstructed sequence for each phylogenetic
node. Rather, they return for each position a vector with the probabilities of being
9 Resurrected Ancestral Proteins as Scaffolds for Protein Engineering 231
a Language ‘sky’ ‘five’
Saaroa laŋica ulima
Proto-Austronesian Ilonggo laŋit lima

*laŋiC
Kilokaka kh laŋa gaha
*lima
Zabana k laŋa gaha u
Proto-MP Hawaiian lani lima

*laŋit Maori raŋi rima
Proto-EP
*lima
*laŋi Rarotongan raŋi rima
*lima
b Organism Sequence
Escherichia coli ...TMPAA...

...TTPAA... Salmonella enterica ...TMPAA...
ENCA Hamiltonella defensa ...TTPKI...

Klebsiella pneumoniae ...TTPAS...
...TTPAA...
Klebsiella variicola ...TTPAS...
...TTPAS... Yersinia pestis biovar ...TTPAS...
Fig. 9.1 Reconstructing words from extinct languages versus reconstructing sequences of pro-
teins from extinct organisms. (a) Language tree for Austronesian languages. The words for “sky”
and “five” in modern Austronesian languages were used [21] to reconstruct the corresponding
words in the extinct languages Proto-Austronesian, Proto-Malayo-Polynesian (Proto-MP), and
Proto-Eastern-Polynesian (Proto-EP) (Reprinted, with permission, from Atkinson [6]). (b) Section
of a tree derived from the phylogenetic analysis of an alignment of sequences of extant
(i.e., modern) class A β-lactamases. The complete tree is shown in Fig. 9.2, while only the section
corresponding to enterobacteria is shown here. Sequences of lactamases from modern organisms
were used [109] to reconstruct the sequences of the lactamases at the internal nodes. These internal
(ancestral) nodes correspond to extinct organisms (ENCA stands for the common ancestor of
enterobacteria). For illustration, only the sequence for residues 181–185 is shown (the complete
sequences of 262 residues were reconstructed at all the internal nodes). Note that ancestral
sequence reconstruction is a broad-scale phylogenetic procedure that simultaneously and consis-
tently reconstructs the sequences of all the nodes in a phylogenetic tree from the information
contained in all the extant sequences. Therefore, the reconstructed sequence at the ENCA node is
“affected” by the sequences at all the nodes in the tree (not only by the sequences “below” the
node). Note also that reconstructed ancestral sequences may substantially differ from consensus
sequences [110]. The residue at position 185 is reconstructed as A in ENCA lactamase, while the
consensus of the six modern lactamase sequences below ENCA is S at position 185
the ancestral for the 20 amino acids. While for each node a “most probabilistic”
sequence can thus be defined as the sequence with the most probable amino acid at
each position, alternative representations can be easily constructed by, for instance,
Monte Carlo sampling of the amino acid probabilities. In a certain sense, ancestral
sequence reconstruction procedures lead to an ensemble of statistically plausible
sequences at each phylogenetic node. (2) Obviously, the claim in the field is not that
the existed sequences of the proteins that existed millions or billions years ago can
be recovered but, rather, that the properties of the proteins encoded by the recon-
structed sequences (the laboratory resurrected proteins) may provide a useful
approximation to the ancestral protein properties. In short, the claim in the field is
that the ancestral protein phenotype (the protein properties) may be to some extent
recovered. Therefore, as elaborated below in some detail, validation is performed at
the protein phenotype level.
The outcomes of ancestral sequence reconstruction are routinely tested for phe-
notypic robustness. Briefly, several sequence representations of the protein at a
given node are “resurrected” in the laboratory and studied in terms of relevant prop-
erties (stability, catalysis, etc.). The most probabilistic sequence and several alterna-
tive sequences are typically studied. The desired result is, of course, that the same
or similar properties are obtained for different members of the ancestral sequence
ensemble. At a more general level, the properties of the resurrected proteins for dif-
ferent phylogenetic nodes must “tell” a convincing evolutionary story or, to use the
common jargon in the field, a convincing evolutionary narrative. A specific example
will suffice to illustrate the point. In one of the first experimental ancestral resurrec-
tion studies [69], Benner and coworkers addressed the laboratory resurrection of
ancestral ribonucleases for the artiodactyl lineage (the order of mammals to which
sheep, deer, camel, pig, and ox belong). They found the resurrected proteins to dis-
play the substrate scope expected for digestive ribonucleases, but only up to the
node corresponding to the common ancestors of ruminants. “Older” ribonucleases,
in fact, showed enhanced activity toward non-digestive substrates. The ancestral
resurrection exercise of Jermann et al. [69] thus supported that digestive ribonucle-
ases originated at about 40 million years when a non-digestive ribonuclease was
recruited for ruminant digestion, a scenario fully consistent with the lowering of
temperatures at the end of the Eocene, the concomitant widespread emergence of
grasses as a source of food, and the appearance of ruminant herbivores. This corre-
lation between the molecular and paleontological records provides clear support for
the ancestral ribonuclease resurrection exercise at the phenotype level [13, 69].
In view of the above, one ideal scenario for ancestral sequence reconstruction
would involve that the following results hold: (1) the tree derived from the analysis of
the alignment of modern sequences is reasonably close to accepted organismal phy-
logenies, and, in case that paralogous proteins are included, the tree is also consistent
with reasonable hypotheses regarding the corresponding gene duplication events; (2)
phenotypic robustness is demonstrated by the congruence of the biophysical proper-
ties of the proteins encoded by several alternative resurrections at key nodes (obtained
from random sampling of the posterior probability, alternative tree topologies below
the node, etc.); (3) the biophysical properties determined for the resurrected proteins
provide convincing evolutionary narratives that correlate the paleontological and
molecular records of life and reveal relevant evolutionary adaptations.
A recently reported [109] laboratory resurrection of ancestral forms of the anti-
biotic resistance enzyme β-lactamase approaches to some extent the ideal scenario
described in the preceding paragraph. A set of sequences providing a uniform cover-
age of the phyla in Bacteria was used as starting point of the ancestral sequence
reconstruction exercise (Fig. 9.2). Sequences from clinical isolates were excluded
a
Firmicutes
Actinobacteria
PNCA
GPBCA
ENCA TEM-1
Proteobacteria
GNCA
CFB Group
b PNCA GNCA
Denaturation temperature (°C)

90
(Third-generation/Penicillin)
100 GPBCA
Catalytic efficiency
10–1 80
Promiscuity
10–2
Stability
ENCA
70
10–3
10–4 60
10–5 TEM-1
3 2 1 Today
Time before present (billion years)
Fig. 9.2 Hyperstability and promiscuity in laboratory resurrections of Precambrian β-lactamases

[109]. (a) Tree derived from the phylogenetic analysis of an alignment of 75 sequences of modern
lactamases. The tree is close to an accepted phylogeny of the organisms and, therefore, sequence
reconstruction targeted well-defined Precambrian nodes: ENCA (common ancestor of enterobac-
teria), GPBCA (common ancestor of gammaproteobacteria), GNCA (common ancestor of various
Gram-negative bacteria), and PNCA (common ancestor of various Gram-positive and Gram-
negative bacteria). (b) Biophysical properties of laboratory resurrection of Precambrian lacta-
mases. Denaturation temperature and the ratio of catalytic efficiencies for the degradation of
cefotaxime (a third-generation antibiotic) and penicillin are plotted against the age of the targeted
Precambrian nodes. Note that the ancestral PNCA and GNCA lactamases display efficient cataly-
sis of both antibiotics (kcat/KM around 105 M−1 s−1 for penicillin and cefotaxime), while the modern
TEM-1 lactamase is a penicillin specialist
from the set to avoid interference in the reconstruction process from recent evolu-
tion during the antibiotic era. The inferred phylogenetic tree was reasonably close
to an accepted organismal phylogeny, indicating limited interference from horizon-
tal gene transfer. Phenotypic robustness was supported by stability and antibiotic
degradation determinations on the proteins encoded by several alternative recon-
structions of the lactamase corresponding to the last common ancestor of Gram-
negative bacteria (an organism that inhabited Earth about 2 billion years ago).
Finally, the biophysical properties of the resurrected ancestral lactamases provided
convincing evolutionary narratives: (1) lactamase stability, as probed by the value of
the denaturation temperature, increased by many degrees (30–35 °C) upon “travel-
ing back in time” 2–3 billion years. Furthermore, such change in stability was found
to correlate with estimates of ancestral ocean temperatures derived from isotopic
composition of cherts, supporting lactamase adaptation to environmental tempera-
tures over geological time scales. (2) Patterns of antibiotic degradation activities for
the resurrected ancestral lactamases conformed to the expected evolutionary transi-
tion from promiscuous generalists to specialist enzymes, and the estimated time for
the emergence of penicillin specialists roughly matched the divergence time of
fungi (about 1.2 billion years before present).
Certainly, ancestral sequence reconstruction does not need to conform to “ideal
scenarios” to be useful. This would be particularly true if the purpose of the ances-
tral reconstruction exercise is to obtain scaffolds for protein engineering. The goal
in this case would be the preparation of proteins with useful properties from an
engineering point of view (such as high stability or catalytic promiscuity), and,
therefore, the degree of phenotypic correspondence of the laboratory resurrected
proteins with the proteins that actually existed millions or billions of years ago
would be a secondary concern. We will return to this issue in several sections of this
chapter.
9.3 ncestral Protein Resurrection Has Been Used

A
in the Last ~20 Years to Address Important Issues
in Molecular Evolution
Resurrected ancestral proteins have been used in the last 20 years or so mostly as
“tools” to explore evolutionary hypotheses and processes that would have been
difficult to address otherwise. Early work has been summarized in excellent reviews
[13, 52, 53, 94]. Here we will only provide a list of recent applications to convey to
the reader the “flavor” of the field. Ancestral protein resurrection has been recently
used to study molecular adaptations to changing environmental conditions over
planetary time scales [2, 42, 107, 109]; the origin of thermophily [57]; the evolution
of complexity in biomolecular machines [36]; the evolution of divergent DNA spec-
ificities in paralogous transcription factors [60]; the evolutionary effects of epistatic
interactions across molecular interfaces [4]; the molecular mechanisms of evolution
of new functions [105]; the degree of conservation of protein structure over long
geological time scales [64]; the origin of detoxifying enzymes [11]; the role of gene
duplication in evolutionary innovation [130]; the characterization of the evolution-

ary events leading to gene silencing [85]; the determination of the time at which our
ancestors were exposed to alcohol in the diet [23]; the degree of conservation of
site-specific amino acid preferences over evolutionary history [111]; the evolution
of protein conformational dynamics [145]; the historical trajectory, timing, and
mechanisms of evolution of a new protein function important to organized multicel-
lularity in diverse animal phyla [5]; the mechanism for the de novo emergence of a
functional lectin β-propeller from short motifs [122]; the molecular origin and adap-
tive changes in the visual system of mammals [16]; and the adaptive evolution of
binding specificity in solute-binding proteins [26].
9.4 ecent Work Suggests the Biotechnological Potential

R
of Resurrected Ancestral Proteins
As described in the preceding section, resurrected ancestral proteins have been

mostly used as tools in molecular evolution studies. It has emerged from very recent
studies, however, that resurrected ancestral may often display the biophysical fea-
tures (high stability, substrate and catalytic promiscuity linked to enhanced confor-
mational flexibility/diversity) that are likely to contribute to protein evolvability and
that are, therefore, advantageous in scaffolds for molecular design and laboratory-
directed evolution. These useful ancestral properties should not come as a surprise.
High stability in resurrected ancestral proteins is likely a consequence of the prob-
able thermophilic character of ancient life. Likewise, promiscuity upon ancestral
resurrection is to be expected if ancient proteins, unlike many modern proteins,
were generalists with broad specificities. In the following sections, we expound on
the evolutionary origin and usefulness in engineering of the typical ancestral
phenotype.
9.5 High Stability Should Contribute to Protein Evolvability
Protein stability is often described as being marginal. The prevalent view in mod-
ern literature is that marginal protein stability is not adaptive. That is, high stabil-
ity does not necessarily impair function (by, for instance, making the protein more
“rigid”). In fact, the simplest and widely proposed explanation of marginal pro-
tein stability involves two straightforward assumptions: (1) evolution of protein
stability can be described in terms of an evolutionary stability threshold, in the
sense that mutations that bring stability below the threshold compromise organ-
ism survival are rejected (purifying selection), while mutations that keep stability
above the threshold are essentially neutral. (2) Most mutations in a protein are
destabilizing, and, consequently, available protein sequences become scarcer as
stability is increased. The combined effect of these two factors should make pro-
tein stability to fluctuate slightly above the threshold during evolution [15, 18, 47,
120, 126].
In a protein engineering scenario, marginal stability obviously limits protein

evolvability because many functionally useful mutations are destabilizing (since
most mutations are destabilizing anyway) and would compromise proper folding.
On the other hand, a substantial number of those mutations would become available
to molecular design or laboratory-directed evolution if the background scaffold has
been stabilized [17, 77, 90, 95].
9.6 hy We Should Expect to Obtain Proteins

W
with Enhanced Stability upon Ancestral Resurrection
Targeting Very Ancient Phylogenetic Nodes
The reason is simply that ancient life was likely thermophilic and proteins from
thermophilic organisms are expected to display enhanced stability. The several pro-
posed scenarios and experimental analyses that are consistent with ancient life
being thermophilic are summarized below:
1. Evidence from ribosomal RNA sequences suggested the thermophilic character

of the earliest branches of the tree of life [135].
2. Estimates of ancestral ocean temperatures derived from the isotopic composition
of cherts suggest that late Hadean/Archaean ancestral oceans that hosted life
were hot [79–81, 112].
3. High levels of CO2 may have contributed to make the Archean climate much
warmer than today [72]. See, however, point 4 below.
4. While some analyses [123] disfavor high levels of traditional greenhouse gasses
in the primitive Earth, recent work [73, 140] has demonstrated that nitrogen and
hydrogen may have served as greenhouse gasses under particular conditions and
would have helped to maintain high Archaean temperatures.
5. Primordial life may have flourished in hydrothermal vents on the ocean floor [86,
93].
6. Perhaps, only the “tougher,” thermophilic organisms could survive catastrophic
impact events in the young planet (“impact bottleneck” scenarios) [100, 121].
7. Experimental studies on the temperature dependence of the rates of nonenzy-
matic reactions suggest that a massive thermal acceleration of the emergence of
primordial chemistry was required to set the stage for enzyme evolution to get
started [125, 138, 139].
9.7 ncestral Protein Resurrection Targeting Precambrian

A
Phylogenetic Nodes Often Leads in Fact to Highly
Stable Proteins
If ancient life was thermophilic, as suggested by the evidence summarized in the

preceding section, ancient proteins should have been highly stable. Consequently,
ancestral sequence reconstruction targeting “old” Precambrian nodes (in particular
“very old” Archaean nodes) could be expected to lead to proteins with substantially
enhanced stability as compared with their modern counterparts. Indeed, several
recent ancestral resurrection exercises are consistent with this expectation.
Denaturation temperatures for laboratory resurrections of Precambrian elongation
factors [42], thioredoxins [107], β-lactamases [109], and nucleoside diphosphate
kinases [2] are up to about 30–35° higher than their modern homologs. Furthermore,
for these four different families, plots of denaturation temperature versus age of the
targeted phylogenetic nodes define a trend that mirrors the cooling of the oceans as
inferred from the isotopic composition of cherts [80, 112]. These results are clearly
consistent with the proposal that Archaean temperatures were higher than present
average temperatures, that oceans cooled over geological time scales, and that pro-
tein stability, at least for the four systems mentioned above, followed the cooling
trend. The overall congruence is, in fact, remarkable, given that the four protein
systems included display different sizes, functions, and structures.
It must be recognized, of course, that some ancestral resurrection studies target-
ing Precambrian nodes have failed to achieve large increments in denaturation tem-
perature [54]. This is not too surprising, however, because, as we have elaborated in
detail in Sect. 9.2, ancestral sequence reconstruction is an unavoidably uncertain
task that should be convincingly validated at the phenotypic level (as, for instance,
by the congruence of denaturation temperatures with estimated ancestral ocean tem-
peratures). Also, in some cases, denaturation temperatures for resurrected enzymes
may correspond to local environments that do not always reflect global environmen-
tal temperatures. Furthermore, it is plausible that natural selection operates in many
cases on the basis of protein kinetic stability [116], and enhanced kinetic stability
may not necessarily translate into increased values of the equilibrium denaturation
temperature or other metrics of thermodynamic stability (for a recent example, see
[101]). In view of all this, it would perhaps be unrealistic to expect that all ancestral
resurrection efforts lead to proteins with very high denaturation temperatures. On
the other hand, if ancient life was thermophilic, we should expect a statistically
significant trend toward substantially stabilized proteins when targeting Precambrian
nodes in ancestral sequence reconstruction. This is in fact supported by the several
experimental studies summarized in the preceding paragraph.
9.8 I s the Enhanced Stability of Resurrected Precambrian

Proteins an Artifact of Ancestral Sequence
Reconstruction?
About 10 years ago, Goldstein and coworkers reported an insightful computational

analysis into the accuracy of ancestral reconstruction methods [132]. They per-
formed population evolution simulations using an off-lattice protein model with a
fitness function related to protein stability. Their results supported that ancestral
reconstruction methodologies (maximum parsimony and maximum likelihood, in
particular) that reconstruct the “best guess” amino acid at each position may sub-
stantially overestimate ancestral protein stability (the problem appeared to be less
acute with Bayesian methodologies). The work of Goldstein and coworkers [132]
has been sometimes used to argue that the stability enhancements often observed for
resurrected proteins are artifacts inherent to ancestral sequence reconstruction pro-
cedures. This interpretation, however, hardly applies to the large stabilizations
observed upon Precambrian protein resurrection [2, 42, 107, 109], as the corre-
sponding increments in denaturation temperature are very large. They are, in fact,
much larger than (1) the increments in denaturation temperature typically obtained
in protein engineering studies aimed at preparing stabilized variants and (2) the
increments in denaturation temperature predicted by energetic biases reported by
Williams et al. [132]. We elaborate these two points below in some detail:
1. Figure 9.1 in the review paper of Wijma et al. [133] shows a statistics of the dena-
turation temperature increments obtained through protein engineering of enzymes.
Original data are taken from literature reports for the 2007–2012 period. It is clear
from the inspection of the figure that Tm enhancements by mutagenesis are typi-
cally in the 2–15° range. Only two specific studies, with ΔTm = 35° and ΔTm = 45°,
report increments comparable to those obtained by Precambrian protein resurrec-
tion. Actually, the ΔTm = 35° data point in Fig. 9.1c corresponds to resurrected
Precambrian thioredoxins [107]. The ΔTm = 45° increment was reported [31] for
an engineered variant of terpene synthase that originally displayed a very low
denaturation temperature, a fact that may have facilitated stabilization by muta-
tion. Overall, the ΔTm values often obtained through Precambrian protein resur-
rection are much larger than those typically achieved through engineering. It is
difficult to envision how simple methodology biases could produce stability
enhancements that are larger than those obtained on the basis of sophisticated
computational approaches and/or efficient directed evolution.
2. In their computational analyses into the accuracy of ancestral reconstruction
methods, Williams et al. [132] reported a bias of 1.5 kcal/mol in unfolding free
energy for reconstructions based on maximum likelihood (biases for reconstruc-
tions based on maximum parsimony and, in particular, Bayesian inference were
reported to be smaller). As a first approximation, changes in unfolding free
energy (ΔΔG) can be translated into changes in denaturation temperature by
using the well-known Shellman equation [117], ΔTm = ΔΔG/ΔS, where ΔS is
the unfolding entropy change. ΔS values can be estimated from the size of the
protein from known structure-energetic correlations [113]. Specifically, for a
temperature of 60 °C, the unfolding entropy change scales with protein size
according to ΔS≈2.1·NRES cal·K−1·mol−1, where NRES is the number of amino
acid residues. This gives ΔS values of 0.21, 0.42, and 0.63 kcal K−1·mol−1 for
proteins of 100, 200, and 300 residues, respectively. Application of Shellman
equation with ΔΔG = 1.5 kcal/mol then gives denaturation temperature incre-
ments of 7.1, 3.6, and 2.4 for 100, 200, and 300 residues. These values are con-
sistent with the estimate (about 6°) given by Williams et al. [132] using the
thermodynamic data for T4 lysozyme unfolding. While these are obviously
“back-of-the-envelope” calculations, it is clear that the stability overestimation
in ancestral reconstruction discussed by Williams et al. [132] is expected to lead
to increments in denaturation temperature of a few degrees, substantially smaller
than the increments of 30–35° found in several Precambrian resurrection studies

[2, 42, 107, 109].
9.9 nzyme Promiscuity Is a Convenient Feature

E
in Biotechnological Application Scenarios
Textbooks sometimes picture enzymes as being highly specific and enormously effi-
cient catalysts. This description, while valid in general terms, is perhaps an excessive
simplification. Certainly, enzymes can achieve rate enhancements of many orders of
magnitude with respect to the non-catalyzed reactions [136, 137]. Still, the catalytic
efficiency of most enzymes is substantially lower than the maximum attainable rate,
the so-called diffusion limit [10]. Regarding specificity, many proteins can actually
carry out several functions [7, 20, 28, 35, 74–76, 78, 102, 142]. Often, an enzyme
shows one clear primary activity, together with some other low-level “secondary”
activities involving substrates and/or chemical transformations similar to those
involved in the primary activity. Furthermore, some enzymes can efficiently perform
a variety of related functions (the detoxification enzyme glutathione transferase is a
paradigmatic example: [144]) or even clearly differentiated functions linked to dif-
ferent active sites, a type of promiscuity often referred to as “moonlighting.”
However, even the most common low-level, secondary promiscuous activities
are interesting from a biotechnological point of view [62, 74, 102]. Enzymes are
being increasingly employed in the industry (Chap. 19 in [48, 114]) since their use
as technological catalysts often replaces traditional processes based on chemicals
that are typically inefficient in terms of the use of energy and natural resources.
Unfortunately, biotechnological applications of enzymes are, in principle, severely
limited to their natural activities, while many technological applications require
catalysis of nonnatural reactions. However, some “nonnatural” reactions can actu-
ally be subject to enzyme catalysis, albeit as promiscuous activities. Certainly, in
most cases, these secondary catalysis levels are very low. Still they can be used as
“seeds” or backgrounds for the preparation of catalytically efficient engineered
variants through rational design or, more often, laboratory-directed evolution. In
this way, enzymes can be engineered to catalyze chemical reactions that, while
related to their primary activities, differ from them in terms of the substrate degraded,
the product obtained, and the chemical transformation achieved. In short, promiscu-
ity expands the scope of technological applications of enzymes.
9.10 W
hy We Should Expect Many Proteins Encoded
by Reconstructed Ancestral Sequences
to Be Promiscuous
Two lines of reasoning suggest that promiscuity should be a common outcome of

ancestral protein resurrection: (1) primordial enzymes were likely promiscuous
generalists, and such ancient promiscuity is likely to be recovered upon sequence
reconstruction targeting very ancient phylogenetic nodes; (2) most models of gene
duplication evolution predict that the ancestral single-copy gene had more functions
than the evolved daughter copies, and, therefore, sequence reconstruction targeting
pre-duplication nodes is expected to often yield promiscuous proteins. We elaborate
on these two scenarios below:
1. A substantial number of modern enzymes are highly efficient specialists, capa-

ble to enhance the rate of a very specific chemical reaction by many orders of
magnitude [136, 137]. Such high degree of specialization is clearly the product
of natural selection operating over evolutionary time scales. There is, therefore,
wide agreement [9, 30, 32, 41, 67, 68, 75, 76, 104, 145] that many ancient
enzymes were promiscuous generalists with broad functionalities and that many
proteins have evolved from generalists into specialist during evolution. The
fundamental idea was most clearly expressed by Roy Jensen in a highly influ-
ential article [68]. He pointed out that primordial life was likely restricted to
limited genetic information that encoded a small number of proteins. Therefore,
primitive enzymes likely possessed broad functionalities, thus maximizing the
catalytic versatility of the cell under limited enzyme resources. Subsequently,
gene duplication (see below) provided the opportunity for the “luxury of
increased specialization and the improved metabolic efficiency” [68]. According
to this view, the low-level, secondary promiscuous activities detected for many
modern enzymes would be a vestige of the broad functionality of their protein
ancestors, and sequence reconstruction targeting very old nodes should often
recover enzymes with enhanced promiscuity.
2. Gene duplication is the origin of most new genes and likely underlies the origin
of most new functions. Not unexpectedly, many models of evolution of gene
duplication, when applied to enzymes, suggest multifunctional (promiscuous)
ancestors. As famously discussed by Ohno [103], duplication generates redun-
dant gene copies, and if gene dosage is not crucial, one copy can evolve free
from functional constraints, while the ancestral function is maintained in the
other copy through purifying selection. The copy that is free to evolve could
occasionally accept mutations that generate a new function. This simple neo-
functionalization model faces, however, several problems. The most obvious one
is that accumulation of neutral mutations is likely to render nonfunctional the
gene copy that is freed from constraint before it can acquire a new function with
adaptive value. That is, in the simple model, nonfunctionalization (pseudoge-
nization) would be far more likely than neo-functionalization, and it would not
be clear how the large numbers of functional paralogous genes found in living
organisms could have originated. Several alternative models have been proposed
to solve this conundrum. In the escape from adaptive conflict (EAC) model [29,
56], the single-copy gene is supposed to be selected to perform a new function
while maintaining the ancestral function. While moderate increments in the new
function might perhaps occur without seriously compromising the ancestral
function [1, 41], large improvements may not be possible, however, because of
the concomitant detrimental effects on the ancestral function. Gene duplication
provides an escape from this conflict, as one copy is free to improve the new
function while the other can improve the ancestral function. In this way, each
copy specializes in one of the functions, while the ancestral gene prior to dupli-
cation was bifunctional. In the duplication-degeneration-complementation
(DDC) model [39], both genes resulting from duplication experience degenerat-
ing mutations that cause loss of function, i.e., sub-functionalization. In the quan-
titative sub-functionalization version of the model [50], there is one single
function, and the duplicates are maintained because their lower function levels
impose that the two copies are required to reach the level of the ancestral gene.
In the qualitative sub-functionalization version of the DDC model [50, 61], on
the other hand, the ancestral gene has two different functions that are indepen-
dently mutable. Degeneration after duplication causes complementary loss of
function in the duplicate genes. Each copy thus becomes specialized in one func-
tion through degeneration of the other function, while the pre-duplication ances-
tral gene had the two functions (i.e., the functions present in the ancestral gene
are partitioned between the two descendant duplicates). By contrast with the
EAC and DDC models, the innovation-amplification-divergence (IAD) model
(or adaptive radiation model) [14, 40] proposes that the duplication step is not
neutral. This model still assumes more than one function in the parent gene with,
perhaps, one major original function and one (or several) minor side activities.
At a certain point one of the side functions may become valuable (due to a
change in ecological niche, for instance). Since duplication and amplification
events are far more common than improvement by point mutations [14], the
requirement for increased levels of the side activity will be more likely met in the
first place by amplification of the original gene. Subsequently, however, the pos-
sibility of improvement by point mutation is increased because of the presence
of several extra copies of the gene that act as mutational targets. Improvement in
one copy brings about relaxation on the other copies, and these could then be
removed from the population in such a way that the required level of the new
function is eventually provided by one copy of the gene. Since optimization of
the new function is likely detrimental for the original one, selection will maintain
one copy with the original major function.
The three well-known models we have described above (EAC, DDC with quali-
tative sub-functionalization, and IAD) display one common feature: the ancestral
pre-duplication gene has more functions than the evolved daughter copies. Certainly,
we have not discussed all possible models of gene duplication evolution (for a more
exhaustive catalog, see Table 1 in [65]). It is clear from the above discussion, how-
ever, that ancestral sequence reconstruction targeting pre-duplication nodes is likely
to yield promiscuous proteins.
9.11 A
ncestral Protein Resurrection Has Been Shown to Lead
to Substrate Promiscuous Enzymes
In congruence with the proposal summarized in the preceding section, a number of

ancestral reconstruction exercises have indeed led to substrate promiscuity in resur-
rected ancestral enzymes and/or provide evidence supporting the relevance of
ancestral promiscuous activities for enzyme evolution. Some of these studies are
briefly described below.
Verstrepen and coworkers [130] noted the contrast between the large number of
available theoretical models of enzyme evolution upon gene duplication (see Sect.
9.10) and the scarcity of experimental evidence that could be used to decide between
the different models. To alleviate this situation, they used ancestral sequence recon-
struction to gain experimental insight into the functional properties of the ancestral
proteins encoded by pre-duplication genes. As a model system they chose the MALS
gene family, which encodes modern α-glycosidases with diversified substrate speci-
ficities (including a variety of maltose and isomaltose analogues). They indeed
found the protein encoded by the sequence reconstructed for the first pre-duplication
enzyme to be promiscuous, being primarily active on maltose sugars but also show-
ing significant levels of activity for isomaltose substrates. Duplication events then
produced enzymes in which subsequent mutations led to optimization of either
maltose or isomaltose activity.
Barkman and coworkers [59] also addressed the fate of ancestral protein activi-
ties during evolution, but used as a model system the plant methyltransferases of the
SABATH gene family. They used ancestral sequence reconstruction to explore
functional divergence after gene duplication with respect to salicylic acid, benzoic
acid, and nicotinic acid, the three substrates used by the modern enzymes for pro-
duction of substances involved in pathogen/herbivore defense and floral scent. Their
results provided evidence for evolution through improved catalysis of ancestrally
non-preferred substrates and overall supported the relevance of ancestral promiscu-
ous activities for enzyme evolution.
Risso et al. [109] reported a detailed biophysical and functional characterization
of proteins encoded by reconstructed sequences corresponding to Precambrian
nodes in the evolution of the antibiotic resistance enzyme β-lactamase (Fig. 9.2).
While a modern TEM-1 β-lactamase is a penicillin specialist and shows rather low
activity levels toward, for instance, third-generation antibiotics, resurrected ances-
tral β-lactamases for 2–3 billion years nodes were found to be moderately efficient
promiscuous generalists, able to degrade both penicillin and third-generation antibi-
otics. This result does not imply, of course, the existence of third-generation lactam
antibiotics (a human invention) in the Precambrian. It rather suggests, however, that
ancestral lactamases had to hydrolyze a variety of substances, perhaps because
ancient bacteria produced a diversity of antibiotics as a mechanism to obtain nutri-
ents by killing competitors [51] and lactamases arose as a defense device or perhaps
because the older Precambrian lactamases were at Jensen’s generalist stage [68].
Sterner, Merki, and coworkers [108] have recently used ancestral protein resur-
rection to study the history of bifunctionality of a sugar isomerase form histidine
biosynthesis (see, also, third paragraph in Sect. 9.13). They found evidence that the
substrate promiscuity of the ancient HisA enzymes had persisted over 2 billion
years at least, apparently without evolutionary pressure. This is a very interesting
result because, in our view, it suggests that, for many protein systems, vestiges of
Jensen’s generalist stage [68] may have “survived” over very long evolutionary
periods.
The issue of the persistence of ancestral promiscuity is also germane to the

interpretation of a very recent study into the evolution of polar amino acid-bind-
ing proteins (AABPs). Clifton and Jackson [26] have characterized proteins
encoded by reconstructed sequences corresponding to four nodes at which AABP
subfamilies diverged. One of the resurrected proteins appeared to be an ineffi-
cient specialist that was only capable of low-affinity binding to one specific
amino acid. The authors did not rule out the possibility that this result was due to
errors in the sequence reconstruction process. The other three resurrected ances-
tral AABPs displayed a promiscuous capability to bind several amino acids.
Comparison with the binding scope of the modern AABPs revealed two scenar-
ios. In one case, the evolutionary conversion of the ancestral promiscuous pro-
tein into a more specialized binder was apparent. In the two other cases, the
modern AABP proteins displayed a binding scope similar to that of their resur-
rected ancestors. In view of these results, the authors [26] concluded that “the
evidence that the ancient progenitors of modern proteins had a larger range of
physiologically relevant functions in comparison with their descendants remains
limited.” However, it is plausible (see preceding section) that many ancient pro-
teins were generalists, and this is, in fact, supported by the outcomes of the sev-
eral ancestral resurrection studies summarized in this section, including the work
of Clifton and Jackson [26] on ancestral AABPs. Furthermore, since many mod-
ern proteins are specialists, it is reasonable to expect that the generalist to spe-
cialist conversion has occurred often during evolution. However, there is no
reason to believe that such conversion has occurred in all instances. It is of course
conceivable that ancestral promiscuity has been retained (or modulated) in some
cases, provided that such retention confers an adaptive advantage. Regardless of
interpretation subtleties, however, it is relevant in the context of this chapter that
a resurrected ancestral AABP has been used to make a robust genetically encoded
sensor for arginine [134].
9.12 A
ncestral Protein Resurrection Has Been Recently
Shown to Lead to Catalytically Promiscuous Enzymes
The studies summarized in the preceding section provide examples of substrate

promiscuity in resurrected ancestral enzymes. Substrate promiscuity in the catalysis
of a given chemical reaction involves the capability of accepting substrates of dif-
ferent shape and size while maintaining the same basic structure for the transition
state of the process. In contrast, catalytic promiscuity (that is, the catalysis of differ-
ent reactions by a given enzyme) requires the capability to accept different transi-
tion states.
Kazlauskas and coworkers have recently addressed the catalytic promiscuity of
resurrected ancestral enzymes [30] using hydroxynitrile lyases as a model system.
These enzymes catalyze the elimination of hydrogen cyanide from cyanohydrins
and evolved from esterases about 100 million years ago likely as a mechanism of
defense against herbivorous insects. The authors thus resurrected ancestral enzymes
corresponding to branch nodes in the divergence of hydroxynitrile lyases from

esterases. While modern esterases and hydroxynitrile lyases show little catalytic
promiscuity, almost all the resurrected ancestral proteins catalyzed both ester hydro-
lysis and cyanohydrin cleavage. This is a rather remarkable result because, although
esterases and hydroxynitrile lyases share the α-/β-hydrolase fold and the catalytic
serine-histidine-aspartate triad, the reaction mechanisms for the two catalyzed reac-
tions show substantial differences [30]: (i) ester hydrolysis involves an acyl inter-
mediate, while the elimination reaction catalyzed by hydroxynitrile lyases has no
acyl intermediate; (ii) different leaving groups (apolar an polar, respectively) are
generated in the hydrolysis and elimination reactions; (iii) there are different
requirements regarding the binding of the carbonyl oxygen to the oxyanion hole.
The astounding catalytic promiscuity of the resurrected ancestral enzymes was fur-
ther highlighted by their capability to catalyze, albeit slowly, reactions of enzymes
in the α-/β-hydrolase family that are outside the phylogenetic nodes targeted for
reconstruction (decarboxylation, Michael addition, γ-lactam hydrolysis, and
1,5-diketone hydrolysis).
9.13 C
onformational Flexibility/Diversity Provides a Likely
Explanation for Modern and Ancestral Enzyme
Promiscuity
Protein promiscuity involves their capability to interact with different binding tar-
gets, substrates, or transition states (in the case of catalytic promiscuity). Clearly,
the protein needs to be able to “adapt” to a variety of molecules that may differ in
size, shape, and interacting groups. A link between promiscuity and conformational
flexibility is apparent [12, 67, 98, 145] and often interpreted in terms of induced-fit
or conformational diversity models ([19, 131]; see also following section). A few
illustrative examples of the relation between promiscuity and flexibility are pro-
vided below.
Ubiquitin is recognized by a diversity of proteins, and its flexibility is demon-
strated by the conformational heterogeneity in structures of different complexes and
by conformational ensemble derived from the analysis of residual dipolar couplings
[87]. Likewise, flexibility has been shown to allow the regulatory protein calmodu-
lin to bind and activate a large number of target enzymes [143].
Two related isomerization reactions in histidine and tryptophan biosynthesis are
catalyzed in some actinobacteria by a single bifunctional protein, despite a differ-
ence of a factor of two between the sizes of the two substrates involved. Structural
analysis shows the active site to be highly flexible, in such a way that two different
and substrate-specific active-site environments become available [33].
Detoxification enzymes tend to be notoriously promiscuous, as they typically
need to degrade a variety of toxic substances. Still, there may be differences in pro-
miscuity profile for related detoxification enzymes. The A1-1 isoform of cytosolic
human glutathione S-transferase is highly promiscuous and can degrade a variety of
structurally unrelated toxins. By contrast, the A4-4 isoform displays a preference
for lipid peroxidation products. X-ray crystallography, hydrogen/deuterium

exchange, and fluorescence lifetime distribution analysis support that the local and
global conformational flexibility/diversity is responsible for the much higher sub-
strate promiscuity of the A1-1 isoform [58].
The examples summarized above describe modern proteins that display promis-
cuity linked to conformational flexibility/diversity. Recent computational studies
[145] support that conformational flexibility/diversity is also responsible for the
fact that resurrected Precambrian β-lactamases are substrate promiscuous [109], as
we have already mentioned in a preceding section. Specifically, perturbation-
response-scanning analyses based on replica-exchange MD simulations [145] sup-
ported that the modern TEM-1 β-lactamase has a rigid active site, plausibly
reflecting an adaptation for efficient degradation of a specific antibiotic (penicil-
lin), while enhanced conformational flexibility accounts for the capability of the
ancestral resurrected lactamases to bind and degrade antibiotic molecules of differ-
ent size and shape.
9.14 C
onformational Flexibility/Diversity Should Contribute
to Protein Evolvability
It is widely accepted that protein promiscuity and the concomitant conformational

flexibility contribute to high protein evolvability, i.e., to the capability of the pro-
tein to evolve new functions. This point was eloquently made by James and
Tawfik [67] on the basis of a conformational diversity description of protein flex-
ibility and will follow here the general argument provided by these authors. There
is ample experimental and computational evidence that proteins in solution are
best envisioned as ensembles of more or less diverse conformations [18, 19, 24,
43, 44, 67, 71, 98, 128, 131, 145]. Such conformational diversity explains func-
tional diversity, as different conformations could perform different functions
(related to the binding of different partners, substrates, or transition states) and
furthermore suggest a simple molecular mechanism for the evolution of new func-
tions. For instance, while the major (highly populated) conformation of an enzyme
catalyzes the conversion of the “natural” substrate, a minor (scarcely populated)
conformation might catalyze the conversion of an alternative substrate. If, at some
time, the latter process becomes consequential for organismal fitness, natural
selection will bring about an improvement of the new function, likely through
mutations that shift the conformational equilibria toward the conformation
responsible for the new activity. The process will likely involve gene duplication
as necessary step (see Sect. 9.10) to bypass the possible trade-off between the
primary function and the new function.
Clearly, conformational flexibility/diversity contributes to protein evolvability in
the sense that it provides the molecular basis for the existence of low-level promis-
cuous activities that can be enhanced during natural evolution and, in biotechnologi-
cal application scenarios, through laboratory-directed evolution. Furthermore,
recent work supports that that conformational flexibility/diversity may facilitate the
emergence of totally new enzyme activities (i.e., linked to new active sties) in non-
catalytic scaffolds. The only (to our knowledge) reported example [83, 97] of sim-
ple, single-mutation generation of de novo activities in non-catalytic scaffolds used
calmodulin, a conformationally flexible protein ([143]; see also Sect. 9.13) as back-
ground. Screening of combinatorial libraries of proteins designed to fold into four-
helix bundles led to proteins able to rescue various E. coli auxotrophs [37], and
these de novo proteins were shown to be structurally dynamic [99]. An artificial
RNA ligase obtained through screening of a very large naive library [119] displays
enhanced conformational dynamics [25].
9.15 H
igh Stability and Enhanced Conformational Flexibility,
Likely Contributors to Protein Evolvability, Are Not
Necessarily Incompatible Features
Both high stability and promiscuity/flexibility should contribute to evolvability

(Sects. 9.5 and 9.14). It is important to note, therefore, that these two features are
not necessarily incompatible, a statement that is illustrated by the several examples
provided below.
As mentioned in Sect. 9.13, ubiquitin has been shown experimentally to display
conformational diversity, likely a reflection of its promiscuous capability of being
recognized by a large number of different proteins. Still, it is a highly stable protein
with a denaturation temperature of about 90 °C at neutral pH [63].
Hydrogen exchange experiments demonstrated substantial conformational flex-
ibility in a highly stable rubredoxin from the hyperthermophile Pyrococcus furio-
sus, as conformational opening for solvent access was shown to occur in the
milliseconds or faster time scale [55, 66]. Likewise, hydrogen exchange studies
supported a higher degree of flexibility in a thermophilic α-amylase as compared
with a mesophilic homolog [38]. Also, NMR studies on homologous mesophilic
and thermophilic ribonuclease HI enzymes do not support that stable thermophilic
proteins are more rigid [22], a conclusion that is reinforced by a number of compu-
tational studies (see [70] and references quoted therein).
Of course, the examples provided above should not be taken to imply that pro-
teins cannot be highly stable and conformationally rigid at the same time. They do
support, however, that some highly stable proteins can also be conformationally
flexible. This may perhaps come as a surprise to some readers because the notion
that protein marginal stability is adaptive can still be found occasionally in the lit-
erature (for instance, [118]). According to this interpretation, low protein stability
guarantees the degree of flexibility required for function. However, as we have elab-
orated in Sect. 9.5, marginal protein stability is not likely adaptive, but the outcome
of natural purifying selection linked to an evolutionary stability threshold combined
with the well-known fact that the number of available sequences decreases with
increasing stability. Accordingly, it should be possible to enhance protein stability
through laboratory evolution or engineering without compromising function. This,
in fact, has been experimentally demonstrated in several studies [45, 96, 129].
9.16 P
lausibly, the Combination of High Stability
and Promiscuity/Flexibility Was Not Uncommon
Among the Most Ancient Proteins, and It Is, in Fact,
Found in Several Resurrected Ancestral Proteins
As we have elaborated in preceding sections, enhanced stability and promiscuity/

flexibility contribute to protein evolvability and are not incompatible features.
Ancient proteins may, therefore, have displayed both features simultaneously. This
is, in fact, supported by the recent ancestral resurrection studies summarized below.
Some laboratory resurrections of several Precambrian thioredoxins [107] display
denaturation temperatures 30–35° higher than their modern mesophilic counterparts
and, at the same time, show enhanced activity in single-molecule assays of disulfide
reduction. This likely reflects promiscuity because a nonnatural substrate is used in
these single-molecule assays.
As previously noted (Fig. 9.2), laboratory resurrections of 2–3-billion-year-old
β-lactamases [109] display denaturation temperatures 30–35° higher than their
modern mesophilic counterparts and, at the same time, are moderately efficient pro-
miscuous enzymes capable to degrade a variety of antibiotics, including penicillin
and third-generation antibiotics (by contrast, the modern TEM-1 β-lactamase is a
penicillin specialist). Recent computational studies [145] support that conforma-
tional flexibility/diversity is responsible for the substrate promiscuity of these resur-
rected ancestral enzymes.
Very recently [30], a resurrected ancestral enzyme along the path from esterases
to hydroxynitrile lyases has been shown to be able to catalyze both ester hydrolysis
and lyase reactions. This catalytic promiscuity (attributed to enhanced conforma-
tional flexibility by [30]) did not imply a lower stability, however. Quite the con-
trary, the promiscuous ancestral protein was reported to display a denaturation
temperature (about 80 °C) substantially higher than the denaturation temperatures
for its modern descendants (reported as 54–70 °C).
9.17 A
ncestral Protein Resurrection Versus Consensus
Engineering
It emerges from the models, theoretical scenarios, and experimental results we have
discussed so far that ancestral sequence reconstruction should provide a convenient
and efficient method to search sequence space for the protein biophysical features
that are desirable in protein engineering scenarios.
Certainly, the idea of using sequence analyses as a basis for modulating protein
biophysical properties has been around for quite some time, and it is, in fact, readily
illustrated by the well-known consensus approach [3, 34, 46, 82, 89, 92, 124]. The
most frequent amino acid residue at a given position in a sequence alignment is
referred to as the consensus amino acid at that position. Back-to-the-consensus
mutations are often found to lead to stability enhancements and function modula-
tions qualitatively similar to those obtained through ancestral sequence
reconstruction (see [110] for a recent discussion). This is perhaps not too surprising
because the consensus amino acid at a given position may in many cases be the
ancestral amino acid. That is, some back-to-the-consensus mutations may actually
be back-to-the-ancestor mutations, and their stabilizing effect is consistent with the
conservation of energetic preferences over evolutionary history [111]. Still, as elab-
orated below, ancestral protein resurrection appears to have some clear advantages
over consensus engineering.
While comparative analyses are scarce, it appears that consensus engineering
captures the useful ancestral properties only to a limited extent [110]. This is no too
surprising because consensus is a simple counting procedure, while ancestral
sequence reconstruction is a broad-scale phylogenetic procedure that simultane-
ously and consistently reconstructs the sequences of all the nodes in a phylogenetic
tree from the information contained in all the extant sequences. As a result, differ-
ences between the ancestral amino acid and the consensus amino acid are likely to
occur in particular at sites with a high amount of diversity (see [110] for a more
detailed discussion).
While the alignment of a given set of modern protein sequences leads to essen-
tially one consensus sequence, ancestral sequence reconstruction leads to a large
number of sequences, a feature that allows for an efficient search of sequence space.
For instance, the trend observed in the properties determined for the resurrected
proteins corresponding to a given set of phylogenetic nodes may immediately sug-
gest additional nodes at which the properties targeted are likely to be optimized.
Likewise, several alternative representations of the protein sequence at a given
promising node (from a random sampling of the amino acid probabilities: see Sect.
9.2) can be subjected to laboratory resurrection for property optimization.
It must be recognized, of course, that, compared with ancestral sequence recon-
struction, the determination of a consensus sequence is a simple and computation-
ally undemanding procedure. This said, however, it is important to note that there
are currently several software packages and online services (for instance, [49, 88,
115, 141]) that can perform ancestral sequence reconstruction. Furthermore, the
time-consuming experimental validation procedures required for applications in
molecular evolution (see Sect. 9.2) can be substantially relaxed (or even omitted) if
the purpose of the ancestral reconstruction exercise is simply to obtain scaffolds
with extreme and useful properties in a protein engineering scenario.
9.18 Crystal Ball
“Prediction is very difficult, especially about the future” (commonly attributed to

Niels Bohr). Still, we will venture to make two specific predictions about future
uses of ancestral resurrection in protein engineering.
Promiscuity is a useful protein property in biotechnological application scenar-
ios (Sect. 9.9). However, it is an accidental property in most modern proteins, in the
sense that it is not adaptive (with obvious exceptions, such as some detoxification
enzymes). Search for promiscuity in modern proteins is, therefore, inefficient [30].
Ancestral protein resurrection targeting branch points in the divergence of new

activities will be used as an efficient method to search for protein promiscuity. This
prediction is supported by recent work by Kazlauskas and coworkers [30].
The generation of totally new catalytic sites is, arguably, one of the most impor-
tant unsolved problems in protein engineering. The development of procedures to
engineer new active sites capable to efficiently catalyze reactions totally unrelated
with those catalyzed by natural enzymes will have an enormous impact in biotech-
nology and our understanding of enzyme catalysis. However, previous attempts at
the rational design of new catalytic sites have met with limited success [84], as
comparatively low catalysis levels are typically achieved despite targeting simple
model reactions. The high evolvability of resurrected ancestral proteins will make
them the scaffolds of choice for the development of more efficient approaches to
engineer new catalytic sites. This prediction is supported by recent (unpublished)
work from our group that uses a very simple design approach to generate substantial
levels of a nonnatural activity in protein scaffolds derived from ancestral
resurrection.
Acknowledgments Work in the authors’ lab is supported by FEDER Funds and Grants,
CSD2009-00088, and BIO2015-66426-R from the Spanish Ministry of Economy and
Competitiveness.
References
1. Aharoni A, Gaidukov L, Khersonsky O et al (2005) The ‘evolvability’ of promiscuous pro-
tein functions. Nat Genet 37(1):73–76
2. Akanuma S, Nakajima Y, Yokobori S et al (2013) Experimental evidence for the thermophi-
licity of ancestral life. Proc Natl Acad Sci U S A 110:11067–11072
3. Amin N, Liu AD, Ramer S et al (2004) Construction of stabilized proteins by combinatorial
consensus mutagenesis. Protein Eng Des Sel 17:787–793
4. Anderson DW, McKeown AN, Thornton JW (2015) Intermolecular epistasis shaped the func-
tion and evolution of an ancient transcription factor and its DNA binding sites. Elife 4:e07864
5. Anderson DP, Whitney DS, Hanson-Smith V et al (2016) Evolution of an ancient protein
function involved in organized multicellularity in animals. Elife 5:e10147
6. Atkinson QD (2013) The descent of words. Proc Natl Acad Sci U S A 110(11):4159–4160
7. Babtie A, Tokuriki N, Hollfelder F (2010) What makes an enzyme promiscuous? Curr Opin
Chem Biol 14:200–207
8. Bahar I, Lezon TR, Yang LW et al (2010) Global dynamics of proteins: bridging between
structure and function. Annu Rev Biophys 39:23–42
9. Baier F, Tokuriki N (2014) Connectivity between catalytic landscapes of the metallo-β-
lactamase superfamily. J Mol Biol 426:2442–2456
10. Bar-Even A, Noor E, Savir Y et al (2011) The moderately efficient enzyme: evolutionary and
physicochemical trends shaping enzyme parameters. Biochemistry 50:4402–4410
11. Bar-Rogovsky H, Hugenmatter A, Tawfik DS (2013) The evolutionary origins of detoxifying
enzymes: the mammalian serum paraoxonases (PONs) relate to bacterial homoserine lac-
tonases. J Biol Chem 288(33):23914–23927
12. Ben-David M, Elias M, Filippi JJ et al (2012) Catalytic versatility and backups in enzyme
active sites: the case of serum paraoxonase 1. J Mol Biol 418:181–196
13. Benner SA, Sassi SO, Gaucher EA (2007) Molecular paleoscience: systems biology from the
past. Adv Enzymol Relat Areas Mol Biol 75:1–132
14. Bergthorsson U, Andersson DI, Roth JR (2007) Ohno’s dilemma: evolution of new genes
under continuous selection. Proc Natl Acad Sci U S A 104(43):17004–17009
15. Bershtein S, Segal M, Bekerman R et al (2006) Robustness-epistasis link shapes the fitness
landscape of a randomly drifting protein. Nature 444:929–932
16. Bickelmann C, Morrow JM, Du J et al (2015) The molecular origin and evolution of dim-
light vision in mammals. Evolution 69(11):2995–3003
17. Bloom JD, Labthavikul ST, Otey CR et al (2006) Protein stability promotes evolvability. Proc
Natl Acad Sci U S A 103:5869–5874
18. Bloom JD, Arnold FH, Wilke CO (2007) Breaking proteins with mutations: threads and
thresholds in evolution. Mol Syst Biol 3:76
19. Boehr DD, Nussinov R, Wright PE (2009) The role of dynamic conformational ensembles in
biomolecular recognition. Nat Chem Biol 5:789–796
20. Bornscheuer UT, Kazlauskas RJ (2004) Catalytic promiscuity in biocatalysis: using old
enzymes to form new bonds and follow new pathways. Angew Chem Int Ed
43:6032–6040
21. Bouchard-Côte A, Hall D, Griffiths TL et al (2013) Automated reconstruction of ancient
languages using probabilistic models of sound change. Proc Natl Acad Sci U S A
110:4224–4229
22. Butterwick JA, Loria JP, Astrof NS et al (2004) Multiple time scale backbone dynamics of
homologous thermophilic and mesophilic ribonuclease HI enzymes. J Mol Biol
339(4):855–871
23. Carrigan MA, Uryasev O, Frye CB et al (2015) Hominids adapted to metabolize ethanol long
before human-directed fermentation. Proc Natl Acad Sci U S A 112(2):458–463
24. Changeux JP, Edelstein A (2011) Conformational selection or induced fit? 50 years of debate
resolved. F1000 Biol Rep 3:19
25. Chao FA, Morelli A, Haugner JC 3rd et al (2013) Structure and dynamics of a primordial
catalytic fold generated by in vitro evolution. Nat Chem Biol 9(2):81–83
26. Clifton BE, Jackson CJ (2016) Ancestral protein reconstruction yields insight into adaptive
evolution of binding specifity in solute-binding proteins. Cell Chem Biol 23:236–245
27. Cole MF, Gaucher EA (2011) Utilizing natural diversity to evolve protein function: applica-
tions towards thermostability. Curr Opin Chem Biol 15(3):399–406
28. Copley SD (2003) Enzymes with extra talents: moonlighting functions and catalytic promis-
cuity. Curr Opin Chem Biol 7:265–272
29. Des Marais DL, Rausher MD (2008) Escape from adaptive conflict after duplication in an
anthocyanin pathway gene. Nature 454:762–765
30. Devamani T, Rauwerdink AM, Lunzer M et al (2016) Catalytic promiscuity of ancestral
esterases and hydroxynitrile lyases. J Am Chem Soc 138(3):1046–1056
31. Diaz JE, Lin CS, Kunishiro K et al (2011) Computational design and selections for an engi-
neered, thermostable terpene synthase. Protein Sci 9:1597–1606
32. Duarte F, Amrein BA, Kamerlin SC (2013) Modeling catalytic promiscuity in the alkaline
phosphatase superfamily. Phys Chem Chem Phys 15(27):11160–11177
33. Due AV, Kuper J, Geerlof A et al (2011) Bisubstrate specificity in histidine/tryptophan bio-
synthesis isomerase from Mycobacterium tuberculosis by active site metamorphosis. Proc
Natl Acad Sci U S A 108(9):3554–3559
34. Durani V, Magliery TJ (2013) Protein engineering and stabilization from sequence statistics:
variation and covariation analysis. Methods Enzymol 523:237–256
35. Erijman A, Aizner Y, Shifman JM (2011) Multispecific recognition: mechanism, evolution,
and design. Biochemistry 50:602–611
36. Finnigan GC, Hanson-Smith V, Stevens TH et al (2012) Evolution of increased complexity in
a molecular machine. Nature 481:360–364
37. Fisher MA, McKinley KL, Bradley LH et al (2011) De novo designed proteins from a library
of artificial sequences function in Escherichia coli and enable cell growth. PLoS ONE
6(1):e15364
38. Fitter J, Heberle J (2000) Structural equilibrium fluctuations in mesophilic and thermophilic
alpha-amylase. Biophys J 79(3):1629–1636
39. Force A, Lynch M, Pickett FB et al (1999) Preservation of duplicate genes by complemen-
tary, degenerative mutations. Genetics 151(4):1531–1545
40. Francino MP (2005) An adaptive radiation model for the origin of new gene functions. Nat
Genet 37:573–577
41. Garcia-Seisdedos H, Ibarra-Molero B, Sanchez-Ruiz JM (2012) How many ionizable groups
can sit on a protein hydrophobic core? Proteins 80:1–7
42. Gaucher EA, Govindarajan S, Ganesh OK (2008) Palaeotemperature trend for Precambrian
life inferred from resurrected proteins. Nature 451:704–707
43. Gerek ZN, Keskin O, Ozkan SB (2009) Identification of specificity and promiscuity of PDZ
domain interactions through their dynamic behavior. Proteins 77:796–781
44. Gerek ZN, Ozkan SB (2010) A flexible docking scheme to explore the binding selectivity of
PDZ domains. Protein Sci 19:914–928
45. Giver L, Gershenson A, Freskgard PO et al (1998) Directed evolution of a thermostable
esterase. Proc Natl Acad Sci U S A 95:12809–12813
46. Godoy-Ruiz R, Perez-Jimenez R, Ibarra-Molero B et al (2004) Relation between protein
stability, evolution and structure, as probed by carboxylic acid mutations. J Mol Biol
336:313–318
47. Godoy-Ruiz R, Ariza F, Rodriguez-Larrea D et al (2006) Natural selection for kinetic stabil-
ity is a likely origin of correlations between mutational effects on protein energetics and fre-
quencies of amino acid occurrences in sequence alignments. J Mol Biol 362:966–997
48. Grunwald P (2009) Use of enzymes in industry. In: Biocatalysis: biochemical fundamentals
and applications. Imperial College Press, London, pp 968–992
49. Guindon S, Lethiec F, Duroux P et al (2005) PHYML Online–a web server for fast maximum
likelihood-based phylogenetic inference. Nucleic Acids Res 33:W557–W559
50. Hahn MW (2009) Distinguishing among evolutionary models for the maintenance of gene
duplicates. J Herpetol 100(5):605–617
51. Hall BG, Barlow M (2004) Evolution of the serine beta-lactamases: past, present and future.
Drug Resist Updat 7(2):111–123
52. Harms MJ, Thornton JW (2010) Analyzing protein structure and function using ancestral
gene resurrection. Curr Opin Struct Biol 20:260–236
53. Harms MJ, Thornton JW (2013) Evolutionary biochemistry: revealing the historical and
physical causes of protein properties. Nat Rev Genet 14:559–571
54. Hart KM, Harms MJ, Schmidt BH et al (2014) Thermodynamic system drift in protein evolu-
tion. PLoS Biol 12(11):e1001994
55. Hernandez G, Jenney FE Jr, Adams MW et al (2000) Millisecond time scale conformational
flexibility in a hyperthermophile protein at ambient temperature. Proc Natl Acad Sci U S A
97(7):3166–3170
56. Hittinger CT, Carroll SB (2007) Gene duplication and the adaptive evolution of a classic
genetic switch. Nature 449:677–681
57. Hobbs JK, Shepherd C, Saul DJ et al (2012) On the origin and evolution of thermophily:
reconstruction of functional precambrian enzymes from ancestors of Bacillus. Mol Biol Evol
29:825–835
58. Hou L, Honaker MT, Shireman LM et al (2007) Functional promiscuity correlates with con-
formational heterogeneity in A-class glutathione S-transferases. J Biol Chem
282(32):23264–23274
59. Huang R, Hippauf F, Rohrbeck D et al (2012) Enzyme functional evolution through
improved catalysis of ancestrally nonpreferred substrates. Proc Natl Acad Sci USA
109(8):2966–2971
60. Hudson WH, Kossmann BR, de Vera IM et al (2016) Distal substitutions drive divergent
DNA specificity among paralogous transcription factors through subdivision of conforma-
tional space. Proc Natl Acad Sci U S A 113(2):326–331
61. Hughes AL (2005) Gene duplication and the origin of novel proteins. Proc Natl Acad Sci U
S A 102:8791–8792
62. Hult K, Berglund P (2007) Enzyme promiscuity: mechanism and applications. Trends
63. Ibarra-Molero B, Loladze VV, Makhatadze GI et al (1999) Thermal versus guanidine-induced
unfolding of ubiquitin. An analysis in terms of the contributions from charge-charge interac-
tions to protein stability. Biochemistry 38(25):8138–8149
64. Ingles-Prieto A, Ibarra-Molero B, Delgado-Delgado A et al (2013) Conservation of protein
structure over four billion years. Structure 21:1690–1697
65. Innan H, Kondrashov F (2010) The evolution of gene duplications: classifying and distin-
guishing between models. Nat Rev Genet 11(2):97–108
66. Jaenicke R (2000) Do ultrastable proteins from hyperthermophiles have high or low confor-
mational rigidity? Proc Natl Acad Sci U S A 97:2962–2964
67. James LC, Tawfik DS (2003) Conformational diversity and protein evolution -a sixty-years-
old hypothesis revisited. Trends Biochem Sci 28:361–368
68. Jensen RA (1976) Enzyme recruitment in evolution of new function. Annu Rev Microbiol
30:409–425
69. Jermann TM, Opotz JG, Stackhouse J et al (1995) Reconstructing the evolutionary history of
the artiodactyl ribonuclease superfamily. Nature 374:57–59
70. Kalimeri M, Rahaman O, Melchionna S et al (2013) How conformational flexibility stabi-
lizes the hyperthermophilic elongation factor g-domain. J Phys Chem B 117:13775–13785
71. Kar G, Keskin O, Gursoy A et al (2010) Allostery and population shift in drug discovery.
Curr Opin Pharmacol 10:715–722
72. Kasting JF (1987) Theoretical constraints on oxygen and carbon dioxide concentrations in
the Precambrian atmosphere. Precambrian Res 34:205–229
73. Kasting JF (2013) Atmospheric science. How was early Earth kept warm? Science
339(6115):44–45
74. Kazlauskas RJ (2005) Enhancing catalytic promiscuity for biocatalysis. Curr Opin Chem
Biol 2:195–201
75. Khersonsky O, Roodveldt C, Tawfik DS (2006) Enzyme promiscuity: evolutionary and
mechanistic aspects. Curr Opin Chem Biol 10:498–508
76. Khersonsky O, Tawfik DS (2010) Enzyme promiscuity: a mechanistic and evolutionary per-
spective. Annu Rev Biochem 79:471–505
77. Khersonsky O, Kiss G, Röthlisberger D et al (2012) Bridging the gaps in design methodolo-
gies by evolutionary optimization of the stability and proficiency of designed Kemp elimi-
nase KE59. Proc Natl Acad Sci U S A 109(26):10358–10363
78. Kim J, Copley SD (2007) Why metabolic enzymes are essential or nonessential for growth of
Escherichia coli K12 on glucose. Biochemistry 46:12501–12511
79. Knauth LP, Lowe DR (1978) Oxygen Isotope Geochemistry of Cherts from Onverwacht
Group (3.4 billion years), Transvaal, South Africa, with implications for secular variations in
isotopic composition of cherts. Earth Planet Sci Lett 41:209–222
80. Knauth LP, Lowe DR (2003) High Archean climatic temperature inferred from oxygen iso-
tope geochemistry of cherts in the 3.5 Ga Swaziland Supergroup, South Africa. Geol Soc Am
Bull 115:566–580
81. Knauth LP (2005) Temperature and salinity history of the Precambrian ocean: implications
for the course of microbial evolution. Palaeogeogr Palaeoclimatol Palaeoecol 219:53–69
82. Kohn A, Binz HK, Forrer P et al (2003) Designed to be stable: crystal structure of a consensus
ankyrin repeat protein. Proc Natl Acad Sci U S A 100:1700–1705
83. Korendovych IV, Kulp DW, Wu Y et al (2011) Design of a switchable eliminase. Proc Natl
Acad Sci U S A 108:6823–6827
84. Korendovych IV, DeGrado WF (2014) Catalytic efficiency of designed catalytic proteins.
Curr Opin Struct Biol 27:113–121
85. Kratzer JT, Lanaspa MA, Murphy MN et al (2014) Evolutionary history and metabolic
insights of ancient mammalian uricases. Proc Natl Acad Sci U S A 111(10):3763–3768
86. Lane N, Martin WF (2012) The origin of membrane energetics. Cell 151:1406–1416
87. Lange OF, Lakomek NA, Farès C et al (2008) Recognition dynamics up to microseconds
revealed from an RDC-derived ubiquitin ensemble in solution. Science
320(5882):1471–1475
88. Lartillot N, Lepage T, Blanquart S (2009) PhyloBayes 3: a Bayesian software package for
phylogenetic reconstruction and molecular dating. Bioinformatics 25:2286–2288
89. Lehman M, Pasamontes L, Lassen SF et al (2000) The consensus concept for thermostability
engineering of proteins. Biochim Biophys Acta 1543:408–415
90. Li Y, Drummond DA, Sawayama AM et al (2007) A diverse family of thermostable cyto-
chrome P450s created by recombination of stabilizing fragments. Nat Biotechnol 25(9):
1051–1056
91. Liberles D (2007) Ancestral sequence reconstruction. Oxford University Press, USA
92. Magliery TJ (2015) Protein stability: computation, sequence statistics, and new experimental
methods. Curr Opin Struct Biol 33:161–168
93. Martin W, Baross J, Kelley D et al (2008) Hydrothermal vents and the origin of life. Nat Rev
Microbiol 6(11):805–814
94. Merkl R, Sterner R (2016) Ancestral protein reconstruction: techniques and applications.
Biol Chem 397(1):1–21
95. Merski M, Shoichet BK (2012) Engineering a model protein cavity to catalyze the Kemp
elimination. Proc Natl Acad Sci U S A 109:16179–16183
96. Miyazaki K, Wintrode PL, Grayling RA et al (2000) Directed evolution study of temperature
adaptation in a psychrophilic enzyme. J Mol Biol 297(4):1015–1026
97. Moroz YS, Dunston TT, Makhlynets OV et al (2015) New tricks for old proteins: single
mutations in a nonenzymatic protein give rise to various enzymatic activities. J Am Chem
Soc 137(47):14905–14911
98. Münz M, Hein J, Biggin PC (2012) The role of flexibility and conformational selection in the
binding promiscuity of PDZ domains. PLoS Comput Biol 8:e1002749
99. Murphy GS, Greisman JB, Hecht MH (2016) De novo proteins with life-sustaining functions
are structurally dynamic. J Mol Biol 428(2 Pt A):399–411
100. Nisbet EG, Sleep NH (2001) The habitat and nature of early life. Nature
409(6823):1083–1091
101. Novak MJ, Pattammattel A, Koshmerl B et al (2016) “Stable-on-the-Table” enzymes: engi-
neering the enzyme–graphene oxide interface for unprecedented kinetic stability of the bio-
catalyst. ACS Catal 6(1):339–347
102. Nobeli I, Favia AD, Thornton JM (2009) Protein promiscuity and its implications for biotech-
nology. Nat Biotechnol 27:157–167
103. Ohno S (1970) Evolution by gene duplication. Springer, Berlin
104. O’Brien PJ, Herschlag D (1999) Catalytic promiscuity and the evolution of new enzymatic
activities. Chem Biol 6:R91–R105
105. Ortlund EA, Bridgham JT, Redimbo MR et al (2007) Crystal structure of an ancient protein:
evolution by conformational epistasis. Science 317:1544–1548
106. Pauling L, Zuckerkandl E (1963) Chemical paleogenetics. Molecular ‘restoration studies’ of
extinct forms of life. Acta Chem Scand 17:S9–S16
107. Perez-Jimenez R, Ingles-Prieto A, Zhao ZM et al (2011) Single-molecule paleoenzymology
probes the chemistry of resurrected enzymes. Nat Struct Mol Biol 18:592–596
108. Plach MG, Reisinger B, Sterner R et al (2016) Long-term persistence of bi-functionality
contributes to the robustness of microbial life through exaptation. PLoS Genet 12(1):e1005836
109. Risso VA, Gavira JA, Mejia-Carmona DF et al (2013) Hyperstability and substrate promiscu-
ity in laboratory resurrections of Precambrian b-lactamases. J Am Chem Soc 135:2899–2902
110. Risso VA, Gavira JA, Gaucher EA et al (2014) Phenotypic comparisons of consensus variants
versus laboratory resurrections of Precambrian proteins. Proteins 82(6):887–896
111. Risso VA, Manssour-Triedo F, Delgado-Delgado A et al (2015) Mutational studies on resur-
rected ancestral proteins reveal conservation of site-specific amino acid preferences through-
out evolutionary history. Mol Biol Evol 32(2):440–455
112. Robert F, Chaussidon M (2006) A palaeotemperature curve for the Precambrian oceans based
on silicon isotopes in cherts. Nature 443:969–972
113. Robertson AD, Murphy KP (1997) Protein structure and the energetics of protein stability.
Chem Rev 97(5):1251–1268
114. Robinson PK (2015) Enzymes: principles and biotechnological applications. Essays Biochem
59:1–41
115. Ronquist F, Huelsenbeck JP (2003) MrBayes 3: Bayesian phylogenetic inference under
mixed models. Bioinformatics 19:1572–1574
116. Sanchez-Ruiz JM (2010) Protein kinetic stability. Biophys Chem 148(1–3):1–15
117. Schellman JA (1987) The thermodynamic stability of proteins. Annu Rev Biophys Biophys
Chem 16:115–137
118. Schulenburg C, Miller BG (2014) Enzyme recruitment and its role in metabolic expansion.
119. Seelig B, Szostak JW (2007) Selection and evolution of enzymes from a partially randomized
non-catalytic scaffold. Nature 448(7155):828–831
120. Sikosek T, Chan HS (2014) Biophysics of protein evolution and evolutionary protein bio-
physics. J R Soc Interface 11:20140419
121. Sleep NH (2010) The Hadean-Archaean environment. Cold Spring Harb Perspect Biol
2:a002527
122. Smock RG, Yadid I, Dym O et al (2016) De novo evolutionary emergence of a symmetrical
protein is shaped by folding constraints. Cell 164(3):476–486
123. Som SM, Catling DC, Harnmeijer JP et al (2012) Air density 2.7 billion years ago limited to
less than twice modern levels by fossil raindrop imprints. Nature 484(7394):359–362
124. Steipe B, Schiller B, Pluckthun A et al (1994) Sequence statistics reliably predict stabilizing
mutations in a protein domain. J Mol Biol 240:188–192
125. Stockbridge RB, Lewis CA Jr, Yuan Y et al (2010) Impact of temperature on the time required
for the establishment of primordial biochemistry, and for the evolution of enzymes. Proc Natl
Acad Sci U S A 107(51):22102–22105
126. Taverna DR, Goldstein RA (2002) Why are proteins marginally stable? Proteins
46:105–109
127. Thornton JW (2004) Resurrecting ancient genes: experimental analysis of extinct molecules.
Nat Rev Genet 5:366–375
128. Tobi D, Bahar I (2005) Structural changes involved in protein binding correlate with intrinsic
motions of proteins in the unbound state. Proc Natl Acad Sci U S A 102:18908–18913
129. Van den Burg B, Vriend G, Veltman OR et al (1998) Engineering an enzyme to resist boiling.
130. Voordeckers K, Brown CA, Vanneste K et al (2012) Reconstruction of ancestral metabolic
enzymes reveals molecular mechanisms underlying evolutionary innovation through gene
duplication. PLoS Biol 10:e1001446
131. Vogt AD, Di Cera E (2012) Conformational selection or induced fit? A critical appraisal of
the kinetic mechanism. Biochemistry 51:5894–5902
132. Williams PD, Pollock DD, Blackburne BP et al (2006) Assessing the accuracy of ancestral
protein reconstruction methods. PLoS Comput Biol 2(6):e69
133. Wijma HJ, Floor RJ, Janssen DB (2013) Structure- and sequence-analysis inspired engineer-
ing of proteins for enhanced thermostability. Curr Opin Struct Biol 23(4):588–594
134. Whitfield JH, Zhang W, Herde MK et al (2015) Construction of a robust and sensitive argi-
nine biosensor through ancestral protein reconstruction. Protein Sci 24:1412–1422
135. Woese CR (1987) Bacterial evolution. Microbiol Rev 51:221–271
136. Wolfenden R (2006) Degrees of difficulty of water-consuming reactions in the absence of
enzymes. Chem Rev 106:3379–3396
137. Wolfenden R (2011) Benchmark reaction rates, the stability of biological molecules in
water, and the evolution of the catalytic power in enzymes. Annu Rev Biochem
80:645–647
138. Wolfenden R (2014a) Massive thermal acceleration of the emergence of primordial chemis-
try, the incidence of spontaneous mutation, and the evolution of enzymes. J Biol Chem
289(44):30198–30204
139. Wolfenden R (2014b) Primordial chemistry and enzyme evolution in a hot environment. Cell
Mol Life Sci 71(15):2909–2915
140. Wordsworth R, Pierrehumbert R (2013) Hydrogen-nitrogen greenhouse warming in Earth’s
early atmosphere. Science 339(6115):64–67
141. Yang Z (1997) PAML: a program package for phylogenetic analysis by maximum likelihood.
Comput Appl Biosci 13:555–556
142. Yip SH, Matsumura I (2013) Substrate ambiguous enzymes within the Escherichia coli pro-
teome offer different evolutionary solutions to the same problem. Mol Biol Evol
30(9):2001–2012
143. Yamniuk AP, Vogel HJ (2004) Calmodulin’s flexibility allows for promiscuity in its interac-
tions with target proteins and peptides. Mol Biotechnol 27(1):33–57
144. Zhang W, Dourado DF, Fernandes PA et al (2012) Multidimensional epistasis and fitness
landscapes in enzyme evolution. Biochem J 445(1):39–46
145. Zou T, Risso VA, Gavira JA et al (2015) Evolution of conformational dynamics determines
the conversion of a promiscuous generalist into a specialist enzyme. Mol Biol Evol
32:142–143
Molecular Modeling in Enzyme Design,
Toward In Silico Guided Directed 10
Evolution
Emanuele Monza, Sandra Acebes, M. Fátima Lucas,

and Victor Guallar
Abstract
Directed evolution (DE) creates diversity in subsequent rounds of mutagenesis in
the quest of increased protein stability, substrate binding, and catalysis. Although
this technique does not require any structural/mechanistic knowledge of the sys-
tem, the frequency of improved mutations is usually low. For this reason, com-
putational tools are increasingly used to focus the search in sequence space,
enhancing the efficiency of laboratory evolution. In particular, molecular model-
ing methods provide a unique tool to grasp the sequence/structure/function rela-
tionship of the protein to evolve, with the only condition that a structural model
is provided. With this book chapter, we tried to guide the reader through the state
of the art of molecular modeling, discussing their strengths, limitations, and
directions. In addition, we suggest a possible future template for in silico directed
evolution where we underline two main points: a hierarchical computational pro-
tocol combining several different techniques and a synergic effort between simu-
lations and experimental validation.
E. Monza • S. Acebes
Joint BSC-CRG-IRB Research Program in Computational Biology, Barcelona
Supercomputing Center, Jordi Girona 29, 08034 Barcelona, Spain
M.F. Lucas
Anaxomics Biotech, Balmes 89, 08008 Barcelona, Spain
V. Guallar, PhD (*)
ICREA, Passeig Lluís Companys 23, 08010 Barcelona, Spain

DOI 10.1007/978-3-319-50413-1_10
258 E. Monza et al.
10.1 Introduction
Biotechnology needs catalysts that can work under harsh conditions, catalyze a
broad range of substrates, generate maximum amount of product, and tolerate
changes in the environment. Enzymes, which are biodegradable and reusable cata-
lysts [1], in addition to remarkable reaction rates, can work in environmentally
friendly pH and temperature ranges and display control over stereochemistry and
regioselectivity which makes them ideal for many applications [2, 3]. When think-
ing about enzymes, people normally associate them to expressions such as “perfect
catalysts” or “outstanding reaction rate.” In fact, there are examples of enzymes that
catalyze reactions at extremely high rates such as triose phosphate isomerase, super-
oxide dismutase, or carbonic anhydrase [4]. These are often limited only by the rate
of ligand diffusion into the active site (diffusion-controlled rate). Nevertheless, an
extensive analysis by Bar-Even et al., of nearly 2,000 enzymes, showed that the
median maximal turnover rate value over all measured enzymes is about 10 s−1
nowhere close to the values of 105 or 106 normally associated with catalysts [5, 6].
So, it would appear that natural enzymes are “just good enough” for the function
they must perform in a given organism [7]. One might conclude that if they had
evolved to their optimum performance, then trying to improve them (from a kinetic
point of view) would be attempting the impossible. On the contrary, as seen by the
distribution of reaction rates, kcat, most enzymes function at a lower rate than the
diffusion limit, and thus, there is space to further increase their kinetic properties to
meet industrial needs. Additionally, we need enzymes capable of catalyzing reac-
tions for which no known enzymes exist, to work with different substrates and for
particular conditions that are industrially convenient and economically advanta-
geous. For all these reasons, in most cases, we cannot just use enzymes as they are
found, but instead we need to change their physical-chemical and functional proper-
ties. This is one of the reasons why engineering enzymes for biocatalysis is an
incessantly growing field [8–11].
In nature, enzymes have evolved over millions of years to meet specific
demands and operate under tight in vivo regulation. Their degree of adeptness
includes diverse criteria such as which substrates they accept, the effective reac-
tion rate, the environment in which they function and how well they tolerate
changes in it, inactivation by their own products, etc. These characteristics are
precisely the ones that scientists wish to control to their own advantage. Some
of the earliest attempts to modify enzymes required a deep knowledge of com-
plex structure/function relationships, and (to the authors’ contentment) com-
puter simulations have played an important part in it [12, 13]. Since the
pioneering work [10, 14, 15] in computationally designed protein sequences
(with experimental validation), many remarkable achievements have been
obtained. Interesting work includes predicting sequence changes that alter
atomic packing arrangements in buried protein regions or the creation of new
metal-binding sites which may have many applications along with potential
improvement in protein stability [16–18]. In addition to being able to correctly
predict changes in protein structure, there has been, of course, a large interest in
10 Molecular Modeling in Enzyme Design, Toward In Silico Guided Directed Evolution 259
altering proteins, through computational techniques, to create new function or

adapt them to particular conditions. Rational protein design, which involves
modification of specific amino acids in the protein’s three-dimensional (3D)
structure with previous structural/mechanistic knowledge, can be used to alter
specificity, stability, selectivity, and activity. Literature contains a vastness of
examples of rationally designed proteins (which we do not presume to cover
here) including creating new recognition [19–26], improving protein stability
[27–29], and protein-protein [30–34] or protein-DNA interactions [35, 36]. We
can find procedures to engineer a protein that binds a specific cofactor [37] or a
calcium-binding site [38, 39], redesign an enzyme by stabilizing the transition
state [40], or create new activity from scratch [41].
A special mention involves the design of new proteins from scratch, commonly
known as de novo design, and literature displays many truly interesting examples
of new proteins [42–45]. Currently one of the most common strategies to design
new enzymes is based on encountering complementary active sites for the transi-
tion states of interest [46, 47]. Despite the success of de novo design in providing
novel structures and activity, its difficulties in achieving fast kinetics make it still
preferable to modify templates available in nature for the desired chemistry.
Indeed, a recent computational study pointed out how target reactivity can be one
mutation away from a nonenzymatic protein (if well picked) [48]. Due to the scope
of this book, we refer the reader interested in de novo design to recent studies on
this topic [49].
Despite many promising studies, rational computational protein redesign has its
limitations: it requires a reliable 3D structure of the system of interest and an in-
depth comprehension of the catalytic mechanism; understanding the relationships
between a protein’s primary sequence, its three-dimensional structure, and its func-
tion is therefore a fundamental goal. Regrettably, our knowledge of enzyme activity
is still incomplete which makes our attempts to modifying them often limited.
Detailed understanding of the enzymatic structure/function relation is, however, not
necessary in directed evolution, an alternative engineering technique based on mas-
sive mutations and selective evolution.
Directed evolution (DE) has proven to be a powerful tool for adapting enzymes
to wider applications [50–53]. Briefly, in DE diversity is first created through muta-
genesis or recombination, followed by screening for improvements in desired prop-
erties. One of the main advantages of DE is most certainly that it does not require a
thorough understanding of structure/function relationships, unlike rational or de
novo design. The introduction of random mutations throughout the gene allows the
discovery of mutations that could be difficult to predict with studies based on
structure-function knowledge (mostly focused at the active site region). However,
the low frequency of improved mutations, some experimental bias, and the combi-
natorial explosion of possibilities limit this technique. Furthermore, DE requires the
development of high-throughput screening and not all processes can be adapted.
The methodologies and achievements of directed evolution were already discussed
in other sections of this book and will not be included here. Also we refer the reader
to interesting reviews [54–59].
260 E. Monza et al.
A remarkable observation of many DE experiments is that the location of the

beneficial mutations varies considerably. For example, most modifications in enan-
tioselectivity or substrate specificity are located in the vicinity of the active site or
in the access/exit of reactants/products [58, 60, 61]. Stability and activity however
can be affected by mutations in any part of the protein, close or far from the active
site [62], increasing significantly the number of possible mutations. To avoid screen-
ing massive number of mutations, one can reduce the region to explore by using
functional information (from point mutations, random mutagenesis, or deduction
from sequence alignments), or when structural information exists (by visual inspec-
tion, analysis, etc.), it would be advantageous to exploit this by concentrating muta-
tions where they might be the most effective [62]. Methods such as saturation
mutagenesis (where all other 19 amino acids are tested) on specific positions, gener-
ally near the active site, can increase the probability of finding beneficial mutations
[63–65]. This approach is particularly advantageous when a high-throughput
screening method is not available. Generally known as semi-rational approaches,
these are based on “smart” libraries that, in principle, should have a higher success
rate and try to overcome the limitations of the directed evolution and rational design
[66–69].
Although it is true that many computational approaches exist to complement DE
experiments [69–71], the scope of this chapter is to center on how physics-based
molecular modeling can aid in laboratory DE. For this reason, sequence-based strat-
egies that use evolutionary information or statistical data from previous DE rounds
will not be explored. These often use phylogenetic analyses and multiple sequence
alignments for exploring the amino acid conservation and relationships between
homologues of protein sequences [67, 72–78]. Instead we will center our attention
on how computations and structural information may aid and focus mainly in
physics-based methods to assist in the improvement of three major aspects in
enzyme design: catalytic rate constants, protein stability, and protein-ligand binding
processes.
The atomic/molecular detailed computational exploration of a protein’s amino
acid sequence space is a complex problem. As in most simulation fields, a com-
promise between sampling quality and quantity is necessary. Sampling quality
involves construction of the models together with the energy and scoring func-
tions necessary to rank them and evaluate molecular interactions, topics exten-
sively reviewed previously [63, 79–92]. An energy function describes the internal
energy of the protein and its interactions with the environment such as other pro-
teins, substrates, and solvent, aiming at reproducing the features of the folded
protein [34, 84, 93]. The level of theory used in these and their parameters vary
considerably, but most implementations include bonded (bonds, angles, and tor-
sions) and nonbonded terms (van der Waals and electrostatics) and solvent com-
ponents. Associated to the energy functions is the ability to efficiently score a
large number of protein structures and protein-ligand interactions. Scoring func-
tions are used to assess the quality of the designed protein and help select the
preferred sequence and the lowest-energy protein-substrate complex. Just as the

energy functions, also these vary considerably and can be statistics-based or
empirically based methods such as DMutant or PopMuSiC or physics-based and
rely on the derivation of energy terms from basic principles to calculate free
energy changes [94–97].
The second key aspect involves the system (model) sampling. Given the large
number of degrees of freedom, including possible mutations and structural changes
associated with them, sampling near-native protein conformations is difficult.
Moreover, in situations where protein-ligand interactions exist, sampling must also
extend to all (relevant) protein-ligand conformations. And if we don’t consider these
issues to be enough of a headache, scoring and sampling are not independent! So, to
overcome these limitations, it is essential to introduce many approximations.
Strategies to limit the sampling include restricting the backbone and side-chain
degrees of freedom [34, 82]. In most protein design strategies, sampling is simpli-
fied by using a fixed backbone which is normally obtained from an experimentally
determined protein structure [98] or a high-quality homology model. Although con-
troversy exists on the importance of dynamics in catalysis [99–101], currently we
see more and more cases where backbone flexibility is being taken into account [18,
102–106]. As shown below, development in molecular dynamics (such as high-
throughput molecular dynamics (HTMD), steered MD, etc.) and Monte Carlo tech-
niques is gaining importance in enzyme engineering.
As mentioned, this book chapter will center on physics-based methods to assist
in the improvement of catalytic rate constants, protein stability, and protein-ligand
binding processes. Before entering in these three topics, we refer to Fig. 10.1 and
Table 10.1 for a quick guide describing the main computational methods and mod-
els being used for these purposes (a guide for nonexperts in theoretical modeling).
Finally, we conclude this book chapter by introducing our perspective on how we
believe these techniques will aid in future enzymatic directed evolution.
Fig. 10.1 Scheme showing three different levels of granularity used in molecular modeling:
coarse grained, all atom, and electronic. In the coarse-grained model, the smallest particle is a bead
that includes condensed information on a set of atoms. All atom, as indicated by the name, uses the
atom as the smallest unit, while in an electronic treatment, electrons and nuclei are explicitly
included. Here we show the highest-energy molecular orbital (HOMO), only possible in an elec-
tronic treatment of the system
262 E. Monza et al.
Table 10.1 A quick guide to simulation methods and theoretical concepts

Methodology guide
Length-size-based models (see also Fig. 10.1)
Coarse-grained (CG) model. A group of atoms is described by a bead enclosing the
properties of the aggregation. For example, according to the MARTINI model [107], amino
acids are represented with one to four beads, classified as charged, polar, nonpolar, or
apolar, and also subdivided depending on their hydrogen bonding capacity. Reduction in the
number of beads decreases the number of pairwise interactions thus increasing the speed of
the simulation
All-atom model. All the atoms of the system are included in the model, where the energy
function used (see below) must describe their interaction. Electrons and the nuclei
information are condensed to a single particle that must contain an averaged description of
those properties
Electronic treatment model. Each atom is described as a nucleus with its electrons,
requiring, for its description, approximate solution of the Schrödinger equation
Physical theoretical methods
Molecular mechanics (MM) [108]. These methods perform a classical description where
atoms (or beads in a coarse-grained model) are represented as spheres (or spheroids)
connected by bonds, behaving as springs. Based on MM there arise several computing
simulations such as molecular dynamics, Monte Carlo, and docking methods
Force field. It is a set of parameters that define the property of atoms or beads (predefined
with a partial charge and radius) and the energy function describing their interactions in
MM methods. Typically they include bonding (bond, angle and torsion) and nonbonding
(electrostatic and van de Waals) terms
Elastic network model (ENM) [109]. It describes the collective dynamics of proteins by an
elastic network, typically using a reduced set of nodes, such as alpha carbons
Molecular dynamics (MD) [110]. It simulates the motion of a model accordingly to the
classical Newton’s equation. Most MD software uses force fields to describe the properties
of atoms and its interactions. With current high-performance computing (HPC), simulations
can be expanded up to the millisecond timescale [111] and few millions of atoms; typical
values, however, involve thousands of atoms and hundreds of nanoseconds
Monte Carlo (MC) simulations. The dynamics of the system are obtained by random
(stochastic) motion of the system to assemble a non-time-dependent trajectory [112]. As in
MD, it is mostly based on a force field description of the model
PELE [113]. The protein energy landscape exploration (PELE) software is a Monte
Carlo-based technique including protein structure prediction techniques (such as ENM)
capable of quickly modeling protein dynamics and protein/DNA-ligand interactions [114,
115]
Docking simulations. These propose the preferred relative bound orientation between
molecules, mostly used in protein-ligand (substrate) or protein-protein interactions. Usually,
docking methods first provide several conformations which are then classified by scoring
functions
Scoring functions. These are mathematical functions that predict the strength of
intermolecular interactions [116]. Scoring functions are mostly parameterized from MM
force fields, empirical data, or knowledge-based functions
Methodology guide

Rotamer library. It contains a restricted number of the most probable conformations (torsion
angle values) for a molecule, mostly applied to amino acid side chains, protein backbone,
and ligands. They are built from experimental structural data or from accurate quantum
simulations (e.g., in ligands). When used with sampling methods, they accelerate the
exploration by adopting discrete states instead of continuous values

Quantum mechanics (QM). These methods are based on solving the Schrödinger equation
(normally using approximations) under an electronic model description of the system. The
solution provides the wave function which fully describes the system: the electronic
distribution, the energy, and the gradients to describe the motion of the system. The main
limitation of QM methods is their high computational cost, limiting the system’s size and
simulation speed

Ab initio methods. These are quantum mechanics methods which parameters are obtained
exclusively from first principles solution of the equations (still under approximations) but
without any usage of parameterized data (see semiempirical methods)

Semiempirical methods. These referred to quantum mechanics methods that use parameters
derived from experimental data (or ab initio calculation), typically for the parameterization
of the electron-electron interaction terms (the most expensive to compute). Thus, they are
less computationally expensive and faster than ab initio ones, capable of dealing with large
systems. Their lack of accuracy, especially when fragments are not in the parameterized
data set, is their main limitation

QM/MM. This methodology is a combination of QM and MM methods to handle (large)
biological all-atom systems [117]. One part of the system where we require an electronic
description, such as the active site in an enzyme, is treated at the QM level, and the rest of
the model (remainder of the protein, solvent, etc.) is treated at the MM level.
10.2 S
tate of the Art of Molecular Modeling in Protein
Design
We provide here a general view of recent computational work on protein design. We

do not aim to review all studies produced in the field but to underline several ones
which we believe to be important for future developments of in silico DE approaches.
10.2.1 Protein Stability Improvement
Understanding and quantifying the effect of mutations on the thermodynamical

stability of a protein is of paramount importance for industrial applications. Two of
the most popular tools to prepare and score mutated proteins are Rosetta [118, 119]
and FoldX [27]. After introducing a mutation, the protein’s torsional degrees of
freedom (usually side-chain rotamers) are optimized using an energy function that
estimates the folding free energy for the created variant. Such energy functions
264 E. Monza et al.
depend on (i) physics-based terms, which account for van der Waals, hydrogen
bond, and solvation and electrostatic energies, and (ii) knowledge-based contribu-
tions, which determine the probability of a given rotamer according to the Protein
Data Bank (PDB) statistics [120]. Apart from these common energy terms, these
functions have unique features. For example, Rosetta approximates the free energy
change in the unfolded state due to a mutation with context-independent reference
energies for each residue [121]. On the other hand, FoldX explicitly estimates the
entropic cost to restrict a rotamer in the native state [27]. The relatively low com-
putational cost of these protocols permits to generate and score a large number of
mutations in a short time. As shown by Potapov et al., the accuracy/cost trade-off
is such that these tools can reproduce overall trends and therefore suggest stabiliz-
ing mutations with acceptable probabilities, but they are not good enough to pro-
vide detailed results [122].
Following Potapov’s accuracy assessment, Kellogg et al. tested the ability of
Rosetta to score mutations combining several energy functions and sampling meth-
ods with variable resolution [119]. As a main result, the authors concluded that the
choice of the sampling algorithm should be tuned with the resolution of the energy
function adopted. In other words, an accurate energy function performs better on a
finer sampling; likewise, roughly sampled structures should be scored by smoother
functions which can tolerate steric clashes better. Still, flexible backbone protocols
improved small to large residue mutations, where significant structural changes can
occur. In addition, the authors found that conformational sampling was still insuf-
ficient to recover the crystal conformation when a large to small hydrophobic resi-
due mutation was introduced, due to poor packing. Larger errors were found when
the polarity of the residue drastically changed upon mutation, which suggests poor
trade-off between polar desolvation and buried polar interactions. The lack of
explicit water molecules and ligand contacts was another factor in some failed pre-
dictions. Finally, the lack of a context-dependent unfolded state modeling (a given
mutation was considered to have the same effect on the unfolded state indepen-
dently of the environment [121]) was considered as a source of error, although not
a major one. In fact, a free energy variation of the unfolded state upon mutation
might change protein stability as well as a variation in the folded state. However, a
recent paper shows that an accurate conformational and energetic characterization
of the unfolded protein is not trivial, and its inclusion in protein stability scoring
significantly worsened the prediction [123].
The entropic scoring in FoldX [27, 124] only takes into account the change in
conformational entropy, which depends on the number of accessible conformers in
the unfolded state and their probabilities [124]. Although this entropic variation
dominates folding [125], large discrepancies in vibrational entropy (the intrinsic
entropy of a given protein conformer [124]) have been calculated between thermo-
philic and mesophilic proteins [126]. Therefore, the thoughtful inclusion of a vibra-
tional entropy contribution in protein design free energy functions might pay off.
Najmanovich and coworkers implemented this strategy in the ENcoM server [126],
where they combine FoldX [27] with their ENcoM protocol to rapidly estimate
vibrational entropy. ENcoM combines ENM techniques with a pairwise atom-type

nonbonded interaction term to include the specific nature of amino acids [127].
In an attempt to quantify free energies more rigorously, de Groot and coworkers
employed alchemical free energy MD simulations to score 109 mutants of ribonu-
clease barnase [128]. In this technique, sampling a convenient number of unphysi-
cal (“alchemical”) intermediates renders a rigorous evaluation of the free energy
difference (ΔG) between two states (e.g., wild-type and mutant protein). Unfolded
state’s free energy differences were calculated using a generic Gly-XXX-Gly pep-
tide with capped termini. This choice provides a universal, albeit less accurate and
context-independent, reference state whose values need to be calculated only once
and then are stored as a database. The overall Pearson’s correlation coefficient with
experimental values was 0.86, providing ~72% of the predicted values within
1 kcal/mol of the experimental one when using 30 ns of simulation time. Notably,
most of this accuracy (65%) is retained with only 5 ns. The generality of this accu-
racy/cost ratio will need to be tested against a wider benchmark of mutations. Larger
errors were detected for mutations that introduced changes in the electrostatics of
buried residues or large structure fluctuation: mutations to glycine, involving bulky
and/or well-packed residues, etc.
Due to the impossibility of scoring the entire sequence (mutation) space, several
strategies have been developed to focus the search in smaller regions. These include
(i) the identification of flexible backbone sites which can be rigidified [129, 130]
introducing salt bridges [131] and/or disulfide bonds [132], (ii) the optimization of
surface charge-charge interactions [133–135], (iii) the optimization of core packing
[136], (iv) the removal of unsatisfied buried polar groups [137], and (v) the localiza-
tion of critical residues in the active site entry tunnels, especially for co-solute toler-
ance, with MD [138] or other algorithms like our in-house software PELE [113].
Recently, Wijma and coworkers developed, and applied with success, a mixed
approach which aims to obtain highly thermostable protein variants in a short time
with minimum experimental screening [139]. In their computational workflow,
potentially stabilizing mutations were firstly produced and scored with Rosetta
[119] and FoldX [27]. To minimize the risk of affecting catalysis, only residues
beyond 10 Å of the substrate were mutated. Mutations were considered potentially
stabilizing if ΔΔGfold ≤−5 kJ/mol or if |ΔΔGfold| <5 kJ/mol, and the mutation type
was contained in the set XXX→Arg, XXX→Pro, and Gly→XXX. These were then
filtered to avoid undesired, typically destabilizing features such as increased unsat-
isfied hydrogen bond donors and acceptors or hydrophobic surface exposure to
water. Then, multiple short MD simulations were used to discard variants with
increased backbone flexibility. Finally, variants with experimentally confirmed
higher thermostability and preserved activity were combined in the lab. This com-
putational hierarchical workflow helps to unmask false positives (~50% of the
potentially stabilizing mutations), aiding to focus on reliable mutations; it is, there-
fore, a plausible strategy for future computer-aided directed evolution of thermo-
stable proteins. The main drawback is the exclusion of mildly damaging mutations
that could be coupled synergically to others to improve thermostability.
266 E. Monza et al.
As reported in a recent review [140], there is still substantial room for improve-
ment of structure- and physics-based (thermo)stability design. This will likely pass
through a strong synergy of computational and experimental efforts to improve our
understanding of protein stability. In addition, significant work will have to center
on developing more accurate energy functions, including polarization, solvation,
and vibrational entropy terms. These methodological developments will necessarily
have to couple with improvement of sampling algorithms, including a more effec-
tive modeling of unfolded state changes.
10.2.2 Protein-Ligand Binding Redesign
Whether we are talking about enzymes or receptors, they all share a common fea-
ture: at some stage a protein-ligand recognition process must occur. These are, how-
ever, notoriously slow and complex processes that require extensive sampling of the
protein-ligand dynamics which in many cases includes induced-fit protein confor-
mational changes. The accurate in silico design of protein-ligand interactions is thus
a challenging step [141] toward the engineering of proteins for therapeutic [85] and
enzymatic purposes [140]. Its difficulty roots in the low tolerance to error due to the
reduced number of protein-ligand interactions. In addition, these are largely domi-
nated by polar interactions, which are very sensitive to small changes in geometry
[142]. It is worth noting that, despite the small size of the protein-ligand interface,
we still face a huge number of possible combinations in sequence space (for 10
positions there are ~1013 sequences).
In a recent attempt to benchmark the state of the art of computational protein-
ligand interaction design, Allison and coworkers tested Rosetta’s [12] sequence
recovery (with respect to the wild type) capability over a set of 43 protein-ligand
complexes [142]. The Rosetta protocol involved simultaneous ligand motion and
side-chain rotamer discrete optimization. Overall, sequence recovery was more suc-
cessful when (i) a near-optimal pose was inputted and subjected to limited sampling
instead of blindly searched; (ii) the ligand was small, nonpolar, and rigid; and (iii)
the binding pocket packing was neither overcrowded nor poor. Another interesting
result was the significantly higher recovery for nonpolar residues. The authors sug-
gested that new terms should be added to the energy function to correct this bias
toward nonpolar interactions [142]. However, this bias could be an artifact of poor
sampling, which might limit the accuracy of polar interaction estimation (see
above). In fact, other suggested areas of improvement were the use of continuous,
instead of discrete, sampling of backbone and side-chain rotamers [143], the con-
certed rotation of the backbone of two adjacent residues allowing larger side-chain
motion (the so-called backrub motion) [144, 145], and the calculations of partition
functions providing a link between molecular behavior and bulk thermodynamic
quantities over structural ensembles [102, 146, 147]. All these features are grasped
by OSPREY [143], a recent open-source solution to protein design which includes
graphic processing units (GPU) acceleration [148], dead-end elimination algorithm
[149, 150], and the K* method [150]. K* aims to approximate the partition function
of the bound and unbound states over an ensemble of structures; their ratio provides
an estimation of the binding constant. The conceptual advantage of this methodol-
ogy is a mathematically rigorous, albeit approximate, free energy difference calcu-
lation that explicitly simulates the free ligand and protein. Consequently, ligand and
binding site pre-organizations are, in principle, included in the calculation. On the
other hand, this absolute free energy calculation is neither accurate nor efficient for
systems with a large number of energy minima [151], requiring extensive sampling
to reduce errors. However, the error of the method most likely compensates between
complex and free monomer calculations, making this strategy a valuable tool for
fast free binding energy simulations. Regardless of the methodology chosen, an
effort to produce new experimental data will be fundamental to benchmark these
high-throughput computational protocols and improve their predictive power [86].
An inaccurate description of the binding site is yet another possible source of
error. Indeed, some mutations could shift the pKa of ligand’s and protein’s titratable
sites or introduce a new titratable residue. Therefore, the system should be prepared
thoroughly before computational mutagenesis, and quick pKa predictors [152]
should be used to treat critical mutations. On the other hand, for situations where
pKa is close to the experimental pH conditions, simulation of all the possible com-
binations for the ambiguous titratable sites is required. For instance, a recent laccase
design effort required the simulation of sinapic acid in both protonation states [153]:
if one of the two protonation would have been picked, activity changes would have
been missed since they mostly involved one of the two accessible protonation states.
Finally, waters in the binding site might have an important role in binding, and their
neglect could affect the quality of the calculation [154].
A way to filter and correct designs is based on MD simulations [155]. Many
features of designed protein-ligand complexes can be inspected with this technique,
including hydrogen bond geometries, binding site structural integrity, solvent expo-
sure, and binding site pre-organization. In particular HTMD [156] was used by
Baker and coworkers to filter computationally designed candidates according to the
fraction of near-attack conformations (NAC), structures that resemble the transition
state (TS) [157]. Moreover, MD can help in the future to discern long-range effects.
In fact, it has been recently used to highlight the impact of distant mutations on
active site pre-organization in evolved enzymes [158, 159]. Furthermore, proteins
are dynamical entities organized in a network of correlated fluctuations whose
changes can significantly affect binding at large distances [160]. Importantly, such
network can be identified through a correlation matrix (which quantifies the correla-
tion degree of a pair of amino acids) and partitioned in communities of highly cor-
related residues, giving insights on allosteric interactions [161]. These analyses,
with contact and hydrogen bond maps, might be used in the future to identify
regions whose motion influences the binding site’s dynamics (e.g., making the side
chain of a catalytic residue too flexible), which can then be subjected to mutagenesis
in the lab.
A possible error when studying protein-ligand association arises when focusing
mostly on the binding site, as some mutations along a possible entrance channel
could hinder the ligand entrance/exit process. Sampling algorithms like PELE
268 E. Monza et al.
[113, 162] can help to recognize such mutations. Its combination of ENM, side-
chain prediction, ligand perturbation (translations and rotations), all-atom minimi-
zation, and implicit solvation makes it a suitable tool to quickly map the whole
ligand migration process with good accuracy, taking protein flexibility into account
[163–165]. Analogous MD-based techniques, such as HTMD [156], RAMD [166],
and steered MD [167], have addressed this problem. These provide more accurate
simulations, as it explicitly models water molecules, but also significantly more
expensive ones, difficult to apply to massive mutation studies. Additional tools
such as Fpocket [168] or CAVER [169] are widely used to quickly identify tunnels
and cavities. These techniques, however, do not explicitly simulate ligand or pro-
tein dynamics.
Finally, quantum chemical calculations can be used to validate promising muta-
tions, especially when charge transfer and polarization have an important role in the
binding process. Mixed QM/MM schemes [170], widely used to model large sys-
tems, can significantly improve protein-ligand binding prediction directly, through
explicit energy calculations [171], or indirectly [172] by recalculating ligand’s
atomic charges in an attempt to model ligand polarization effects. An alternative,
more accurate but slower approach to large systems is the fragment molecular
orbital (FMO) method [173]. FMO divides a system in N nonoverlapping fragments
(e.g., one for each protein residue and ligand) and calculates the total energy as the
sum of one-body fragment energies and two-body interaction energy corrections,
providing a ~N2 scalable fully parallelizable QM calculation. Jensen and coworkers
used this methodology to energetically score the cleavability of peptides by HIV-1
protease [174] by looking at the protein-peptide interaction energy.
10.2.3 Catalytic Rate Constant Enhancement
The improvement of catalytic activity of bond breaking/formation passes through

the modeling of the TS of the slowest chemical step of the targeted reaction; see
below for electron transfer (ET) processes. The problems with the design of optimal
activation energies are multiple: (i) the energy function should be sensitive enough
to effectively discriminate between the reactant (substrate) and the TS, whose
charges and geometries might be similar; (ii) the nature of the TS can change upon
mutation; and (iii) activation energies are very sensitive to molecular geometry
changes.
In OptZyme the TS is approximated by a transition state analogue (TSA), a sta-
ble molecule which electronically and sterically resembles the TS [175]. Once the
TSA and the substrate are docked in the active site, two parameters drive mutant
selection: the enzyme-substrate (ES) and the enzyme-TSA interaction energies.
These last two quantities are obtained using classical force fields. Through a num-
ber of conceptual and mathematical simplifications, the authors show that the for-
mer energy correlates with the Michaelis constant (KM) while the latter with the
specificity constant (defined as the ration between kcat and KM). Although it yielded
satisfactory correlations for their specific case, it is worth noting that these have no
general validity. If the rate constant is comparable or much higher than the ES dis-
sociation constant, the pre-equilibrium approximation is no longer valid. Then, the
Michaelis constant cannot be approximated by the ES dissociation equilibrium con-
stant (KD) [176]. Notwithstanding, ES and enzyme-TSA interaction energies are
still valuable tools for a fast semiquantitative evaluation of enzyme variants.
Khersonsky et al. combined computational design and directed evolution to opti-
mize a previously designed Kemp eliminase [177]. As in the previous example, the
classical (force field) interaction energy between the enzyme and an explicit TS
model was the parameter to be optimized during the sequence exploration. The TS
model, however, was derived from QM calculations in solution including key cata-
lytic residues. The authors individuated three main factors for the improved activity:
a more favorable electrostatic environment, a better packed active site, and a higher
degree of active site pre-organization.
Although the last two methods, based on classical interaction energies, allow to
test a big number of mutants, they both present a conceptual limitation: the use of
the enzyme-TS (model) interaction energy, which is size-dependent (extensive
property), to score the activation energy. To correct for this approximation, the ES
interaction energy should be taken into account, providing a relative value. However,
poor sampling and inaccurate energy functions might introduce uncertainties that
could make its introduction useless (as it often happens in molecular mechanics/
generalized born surface area (MM/GBSA) free energy calculations [178]). Still,
they are currently the best approach to test a large number of mutations and find
promising protein variants which can then be filtered with MD and quantum chemi-
cal methods [155].
The only way to properly calculate the activation energy barriers is the
introduction of a QM methodology, capable of describing the electronic effects
associated with TS formation. QM/MM schemes, which have been widely applied
to elucidate enzymatic reaction mechanisms [179], have been employed to rescore
promising candidates [155, 179]. A remarkable example is the hierarchical approach
of Zheng et al. used to design a cocaine hydrolase [180]. Firstly, the r eaction coor-
dinate and the TS of the rate-determining limiting step are determined in the wild
type. Then, many mutations are scored according to their protein-TS interaction
energy; if this is lower than the wild type, a QM/MM calculation along the reaction
coordinate is used to estimate the energy barrier. To allow fast computation, the
authors use a reaction coordinate approach, freezing at each step the reactive coor-
dinate and minimizing all the other degrees of freedom. A main drawback is still the
need of extensive sampling, which makes the presented methodology too expensive
for a general use.
A cheaper alternative to QM/MM calculations is empirical valence bond (EVB)
[181]. EVB is based on a semiempirical Hamiltonian which describes reactants and
products with their resonance structure (explicitly defining atom connectivity).
270 E. Monza et al.
Although EVB energies are less accurate than ab initio and DFT QM/MM methods,
free energy calculations are orders of magnitude quicker and still can achieve accu-
rate results [182, 183], making EVB a suitable tool to score a bigger number of
mutants [184].
In the attempt to describe the entire enzyme or a large part of it with QM calcula-
tions, Jensen and coworkers approximated the reaction coordinate with the linear
interpolation between reactant and product optimized geometries and calculated
each point with semiempirical methods [185–187]. These fast electronic calcula-
tions, united with algorithms for large-scale systems such as FMO [173, 188] or the
much faster effective fragment molecular orbital (EFMO) [189], make “ab initio
biochemistry” [190] closer, albeit still far away for design purposes.
In the particular case of oxidoreductases, where charge transfer processes domi-
nate, additional complexity is added to the protein design problem. Electrons must
move from a donor to an acceptor, sometimes through a long-range electron trans-
fer. According to Marcus’ theory, the ET rate constant [191] depends on three
parameters: (i) electronic coupling, the probability to jump from the reactant to the
product’s diabatic state, which exponentially depends on the donor-acceptor dis-
tance; (ii) reorganization energy, which is the energy penalty that accompanies
electron transfer; and (iii) the free energy difference between product and reactant
(driving force). The ET rate constant has a maximum when the sum of reorganiza-
tion energy and driving force equals zero. Although accurate QM/MM methodolo-
gies have been developed to study electron transfer rate in proteins [192, 193], their
use in enzyme design is limited by their computational cost. To overcome this
barrier, we have recently developed a new methodology to approximately evaluate
ET rates, which combines fast conformational sampling [162] and quick QM/MM
spin density calculations and has been used to evaluate the activity of laccase vari-
ants [153, 194]. While PELE provides a thorough and quick mapping of enzyme’s
and substrate’s dynamics, substrate’s spin density permits to promptly score the
relative changes in driving force upon mutation (the higher the spin density, in
principle, the higher the driving force). This protocol was successfully applied to
the rational design of a laccase toward the production of conductive polyaniline at
low pH [195]. In the same spirit, we rationally improved the oxidation rate of
2,2′-azino-bis(3-ethylbenzothiazoline-6-sulfonic acid) (ABTS) by a highly stable
manganese peroxidase (Fig. 10.2) relying on electron coupling calculations, after
the entire protein-ligand migration studies were performed [196]. In this case, it
was assumed that the driving force and the reorganization energies did not change
upon mutation, which can be a reasonable approximation in surface ET.
Since long-range ET is often the rate-limiting step in catalysis, engineering efforts
have also centered on mutating residues along the ET pathway. By using the QM/
MM e-pathway method [197], Vidal-Limon et al. studied P450BM3’s suicide inacti-
vation [198], a common process in heme peroxidases. From the QM-MM calcula-
tions, they identified key residues in the second heme coordination sphere, aiming at
reducing electron delocalization and obtaining a more stable enzyme against H2O2.
After mass spectrometry assays confirmed the oxidized sites predicted by QM/MM,
they generated a variant 260 times more stable against H2O2 inactivation.
Fig. 10.2 General scheme for the rational MnP6 engineering [196]. The study was divided into
two steps: (1) ligand diffusion and (2) electronic transfer process. First, we compare the ligand
diffusion in the active and inactive enzyme by computing the interaction energy (red and green
dots, respectively) and the distance between ligand and receptor. From the active enzyme, we
extract information about the active site environment that help us to redesign the inactive enzyme
by introducing two specific surface mutations (blue dots). Importantly, these mutations involved
solvent-exposed residues with low conservation and mutability score provided by bioinformatics
techniques. The activation was confirmed in silico by electronic coupling calculations in the sec-
ond step. Site-directed experimental mutagenesis validated the success of the new mutant, which
combines both stability and activity
10.3 Computer-Aided Directed Evolution, a Perspective
In the previous sections, we have seen multiple examples of computationally driven

enzyme engineering. While they use, to more or less degree, structure/function
knowledge for the modeling, we observe a tendency toward more random massive
sequence sampling; we expect to see in the near future full in silico directed evolu-
tion studies. By full we obviously do not refer to a complete study of the sequence
space (all residues and all possible mutations), but to an exhaustive random muta-
genesis combination on a large subset of selected mutants. For 100 residues, for
example, we have a sequence space of ~10130, which would take several lifetimes to
be evaluated even if using current supercomputers. If we restrict the exploration to
single, double, and triple mutants, we have now ~109 possible variants to model.
While this is still a huge number, one can think in a hierarchical scheme where this
sequence space can be explored in days. This is particularly true with the current
(and future) developments in lower-cost multicore servers based on mobile technol-
ogy (see, e.g., the MontBlanc project at https://www.montblanc-project.eu/).
We find a promising example in the work by Wijma et al. where a hierarchical
protocol is used to increase thermostability [139]. In this line, we expect the develop-
ment of additional techniques combining quick bioinformatics (or knowledge-based)
272 E. Monza et al.
screening of a large sequence space, with a molecular modeling refinement of

selected mutants. This last step could be further hierarchically broken down into a
first classical molecular modeling screening followed by selected quantum chemical
reevaluations, in a similar manner to the previously described study by Zheng et al.
[180]. Even though computational techniques are becoming more precise and easy to
implement, a synergic effort between in silico predictions and experimental valida-
tion will be, in our opinion, the preferred solution. Figure 10.3 shows a possible
workflow combining these ideas. In order to apply such a combined effort, we should
keep in mind that molecular modeling will require an accurate 3D structural model,
a possible limiting factor.
The workflow must start with a careful preparation of the wild-type structure, a
fundamental step as it determines the outcome of the computational design. This
preparation should include some sampling, aimed at generating conformational
diversity and providing useful information for design (such as the protein regions
where to look for improved variants). We should emphasize that most molecular
modeling predictions are based on relative values (rather than absolute ones), in
which case a wild-type reference value is needed. In the next step, high-throughput
screening of mutations is carried out with quick methods, such as bioinformatics
and knowledge-based methods. This step will have to rank the initial sequence
space, similar to a high-throughput screening in drug design. Taking into account
that each bioinformatics score can be accomplished in less than a second, we can
aim for several millions of mutants in a “doable” time; we are still looking at several
days of hundreds/thousands of core dedication, a feasible effort, however, in near
future multicore and accelerated computers (or cloud computing). Bioinformatics
screening of millions (billions) of mutants will benefit from new sequencing and
storage of mutational data in the years to come and, in particular, from its process-
ing using machine learning techniques [199]. At the present stage, such techniques
are mostly used to restrict mutagenesis to relevant protein regions [72–75, 77, 78]
System (wild type) Screening of ~107–9 First molecular

preparation and Mutants by modeling screening
sampling (docking, bioinformatics of ~104–6 mutants
MD, PELE, etc.) techniques (machine (Rosetta, ENMcoM,
learning based, etc.) FoldX, etc.)
Experimental test of Second molecular

the ~101–2 most modeling screening
promising in silico of ~102–3 mutants
designed mutants (MD, PELE, QM/MM,
(point mutations) EFMO, EVB, etc.)
Fig. 10.3 Proposed computer-aided directed evolution workflow

TOP 5
Identify all residues within 10 Å of
the substrate
Automatic mutation of residues by

all 19 possible AA
Relax system with PELE and

estimate interaction energy
Rank
positions
W
M
G
Q
R
H
P
A
S
Y
V
K
T
F
L
I
Fig. 10.4 Left panel: scheme of the used computational protocol. It includes identification of all
residues close to the bound substrate, mutation of these amino acids (AA), system relaxation, and
energy scoring. The image in the center is a heat map of the tested 43 mutations. The pink color
corresponds to an equal or worse value than the wild type, while increasing darker color corre-
sponds to improved classical scoring (protein-substrate interaction energy). From this heat map,
the best positions (with the largest number of improved variants relative to the wild type) are
identified, in this example the top five. On the right panel, we find the five amino acid positions (in
van der Waals representation) suggested to be tested experimentally
or to guide directed evolution “on the fly” [200]. A second filtering, for example,
could then be performed with fast molecular modeling techniques such as FoldX or
Rosetta. These techniques could be applied to (the best ranked) several tens of thou-
sands of compounds. The final goal is to provide a reduced set of candidates, few
hundreds/thousands, where we can apply a more accurate molecular modeling
refinement. The simulation time required for this last step will highly depend on two
factors: (i) the exhaustiveness of conformational sampling and (ii) the nature of the
property to improve. Conformational sampling aims to determine possible changes
in the structure produced by the mutation. Quick assessments, in the order of min-
utes to hours, can be currently performed by Monte Carlo techniques [153]; using
MD will require significantly more computational time, limiting the study to only
few hundred mutants. Another important aspect is how to reevaluate the desired
property to engineer. Substrate-binding energies and thermal stability could be
quickly evaluated, for example, with alchemical MM free energy calculations (with
respect to the wild type) [128]. Catalytic design, on the other hand, will require
expensive QM calculations. Due to their very high cost (hours to days), future work
will have to center in designing cheaper methods [201–203] and/or property evalu-
ations. For example, in our current efforts in oxidoreductases’ engineering, the driv-
ing force is approximated with the amount of spin density, calculated after five steps
of QM/MM geometry optimization, localized on the substrate (with respect to the
wild type) [194].
The proposed workflow includes an iterative computational-experimental
approach, where several schemes could be imagined. Currently we find very few
studies following such an approach, where we can underline, for example, Privett
et al. [204]. When thinking of future implementation, an initial less accurate in
274 E. Monza et al.
silico evolution could be tested in the lab and a more accurate second one per-
formed only on those regions that show more promising experimental results.
Similarly, more accurate simulations could be performed only in single mutants,
followed by experimental site-directed mutagenesis, and expanded then to a sec-
ond in silico round involving double mutants and so on. In this way, synergic
mutations can be partially recovered. An alternative strategy to retrieve coopera-
tive mutations, while computing single mutants only, is to rank sequence posi-
tions instead of point mutations. Positions are ordered according to their
frequency of beneficial mutations, following a fast computational saturated
mutagenesis protocol, and the most promising are communicated to the experi-
mental laboratory for (iterative) combinatorial saturated mutagenesis. Contrary
to the previous strategy, false positives are not filtered out since a position,
instead of a given mutation, is chosen. On the other hand, true positives can be
recovered. We are currently employing this strategy to improve oxidoreductases’
activity. In our initial test on a high redox potential fungal laccase, experimental
and theoretical evolutions were run in parallel. In the first DE experimental gen-
eration, one improved variant was identified. In the in silico round, over 40 posi-
tions were screened with PELE, using the protein-substrate interaction energy
after an induced-fit procedure, and the best five were identified (Fig. 10.4, central
panel). Interestingly the improved variant found experimentally was within these
five top positions. Then through combinatorial saturation mutagenesis using
NDT degenerated codons, three new variants were identified recovering syner-
getic effect of two of the suggested in silico positions. This approach allows
quickly guiding experimental mutagenesis: using 100 CPUs ~200 positions can
be scored in 1 day. Although this protocol requires testing more mutants in the
lab, it permits to go from the computer to the lab in 24 h with a focused library
of mutants, an appealing feature for industrial purposes where large number of
mutants can be assayed.
Conclusion
Biotechnology needs accurate enzyme evolution techniques, capable of
designing new catalysts able to work in environmentally friendly conditions
and, importantly, under industrial requirements. In this line, we find great
efforts in developing (and improving) site-directed mutagenesis and directed
evolution techniques, with computer simulations increasingly being used for
this purpose. Different methodologies, from a quick bioinformatics sequence
analysis to a robust solution of the Schrödinger equation, seek to aid the
experimental efforts. In this book chapter, we overviewed recent develop-
ments in molecular modeling for three different engineering tasks: protein
stability, protein-substrate binding, and catalytic rate, with the goal of illus-
trating how these techniques can influence directed evolution in the near
future. We underline two key factors in future implementations: (i) hierarchi-
cal combination of different computational solutions (with increasing accu-
racy but also computational cost), and (ii) close iterative efforts between in
silico and in vitro approaches.
References
1. Schmid A, Dordick JS, Hauer B, Kiener A, Wubbolts M, Witholt B (2001) Industrial bioca-
talysis today and tomorrow. Nature 409(6817):258–268
2. Patel RN (2008) Synthesis of chiral pharmaceutical intermediates by biocatalysis. Coord
Chem Rev 252(5–7):659–701
3. Sukumaran J, Hanefeld U (2005) Enantioselective C-C bond synthesis catalysed by enzymes.
Chem Soc Rev 34(6):530–542
4. Koenig SH, Brown RD (1972) H(2)CO(3) as substrate for carbonic anhydrase in the dehydra-
tion of HCO(3)(−). Proc Natl Acad Sci U S A 69(9):2422–2425
5. Bar-Even A, Noor E, Savir Y, Liebermeister W, Davidi D, Tawfik DS, Milo R (2011) The
moderately efficient enzyme: evolutionary and physicochemical trends shaping enzyme
parameters. Biochemistry 50(21):4402–4410
6. Milo R, Last RL (2012) Achieving diversity in the face of constraints: lessons from metabo-
lism. Science 336(6089):1663–1667
7. Currin A, Swainston N, Day PJ, Kell DB (2015) Synthetic biology for the directed evolution
of protein biocatalysts: navigating sequence space intelligently. Chem Soc Rev
44(5):1172–1239
8. Gutte B, Däumigen M, Wittschieber E (1979) Design, synthesis and characterisation of a
34-residue polypeptide that interacts with nucleic acids. Nature 281(5733):650–655
9. Russell AJ, Fersht AR (1987) Rational modification of enzyme catalysis by engineering sur-
face charge. Nature 328(6130):496–500
10. Hellinga HW, Caradonna JP, Richards FM (1991) Construction of new ligand binding sites in
proteins of known structure: II. Grafting of a buried transition metal binding site into
Escherichia coli thioredoxin. J Mol Biol 222(3):787–803
11. Jemli S, Ayadi-Zouari D, Hlima HB, Bejar S (2016) Biocatalysts: application and engineer-
ing for industrial purposes. Crit Rev Biotechnol 36(2):246–258
12. Schueler-Furman O, Wang C, Bradley P, Misura K, Baker D (2005) Progress in modeling of
protein structures and interactions. Science 310(5748):638–642
13. Steiner K, Schwab H (2012) Recent advances in rational approaches for enzyme engineering.
Comput Struct Biotechnol J 2:e201209010
14. Richardson JS, Richardson DC (1989) The de novo design of protein structures. Trends
Biochem Sci 14(7):304–309
15. Ponder JW, Richards FM (1987) Tertiary templates for proteins: use of packing criteria in the
enumeration of allowed sequences for different structural classes. J Mol Biol
193(4):775–791
16. Bolon DN, Marcus JS, Ross SA, Mayo SL (2003) Prudent modeling of core polar residues in
computational protein design. J Mol Biol 329(3):611–622
17. Dahiyat BI, Mayo SL (1997) Probing the role of packing specificity in protein design. Proc
Natl Acad Sci U S A 94(19):10172–10177
18. Desjarlais JR, Handel TM (1999) Side-chain and backbone flexibility in protein core design1.
J Mol Biol 290(1):305–318
19. Scrutton NS, Berry A, Perham RN (1990) Redesign of the coenzyme specificity of a dehy-
drogenase by protein engineering. Nature 343(6253):38–43
20. Carter P, Nilsson B, Burnier JP, Burdick D, Wells JA (1989) Engineering subtilisin BPN’ for
site-specific proteolysis. Proteins Struct Funct Bioinforma 6(3):240–248
21. Wells JA, Powers DB, Bott RR, Graycar TP, Estell DA (1987) Designing substrate specificity
by protein engineering of electrostatic interactions. Proc Natl Acad Sci U S A
84(5):1219–1223
22. Cedrone F, Ménez A, Quéméneur E (2000) Tailoring new enzyme functions by rational rede-
sign. Curr Opin Struct Biol 10(4):405–410
23. Looger LL, Dwyer MA, Smith JJ, Hellinga HW (2003) Computational design of receptor and
sensor proteins with novel functions. Nature 423(6936):185–190
276 E. Monza et al.
24. Craik C, Largman C, Fletcher T, Roczniak S, Barr P, Fletterick R, Rutter W (1985)

Redesigning trypsin: alteration of substrate specificity. Science 228(4697):291–297
25. Bastianelli G, Bouillon A, Nguyen C, Crublet E, Pêtres S, Gorgette O, Le-Nguyen D, Barale
J-C, Nilges M (2011) Computational reverse-engineering of a spider-venom derived peptide
active against Plasmodium falciparum SUB1. PLoS ONE 6(7):e21812
26. Oelschlaeger P, Mayo SL (2005) Hydroxyl groups in the ββ sandwich of metallo-β-lactamases
favor enzyme activity: a computational protein design study. J Mol Biol 350(3):395–401
27. Guerois R, Nielsen JE, Serrano L (2002) Predicting changes in the stability of proteins and
protein complexes: a study of more than 1000 mutations. J Mol Biol 320(2):369–387
28. Yu H, Huang H (2014) Engineering proteins for thermostability through rigidifying flexible
sites. Biotechnol Adv 32(2):308–315
29. Kuhlman B, Baker D (2004) Exploring folding free energy landscapes using computational
protein design. Curr Opin Struct Biol 14(1):89–95
30. Kortemme T, Baker D (2004) Computational design of protein–protein interactions. Curr
Opin Chem Biol 8(1):91–97
31. Kortemme T, Joachimiak LA, Bullock AN, Schuler AD, Stoddard BL, Baker D (2004)
Computational redesign of protein-protein interaction specificity. Nat Struct Mol Biol
11(4):371–379
32. Reina J, Lacroix E, Hobson SD, Fernandez-Ballester G, Rybin V, Schwab MS, Serrano L,
Gonzalez C (2002) Computer-aided design of a PDZ domain to recognize new target
sequences. Nat Struct Mol Biol 9(8):621–627
33. Shifman JM, Mayo SL (2002) Modulating calmodulin binding specificity through computa-
tional protein design. J Mol Biol 323(3):417–423
34. Lippow SM, Tidor B (2007) Progress in computational protein design. Curr Opin Biotechnol
18(4):305–311
35. Ashworth J, Havranek JJ, Duarte CM, Sussman D, Monnat RJ, Stoddard BL, Baker D (2006)
Computational redesign of endonuclease DNA binding and cleavage specificity. Nature
441(7093):656–659
36. Chevalier BS, Kortemme T, Chadsey MS, Baker D, Monnat RJ Jr, Stoddard BL (2002)
Design, activity, and structure of a highly specific artificial endonuclease. Mol Cell
10(4):895–905
37. Cochran FV, Wu SP, Wang W, Nanda V, Saven JG, Therien MJ, DeGrado WF (2005)
Computational de novo design and characterization of a four-helix bundle protein that selec-
tively binds a nonbiological cofactor. J Am Chem Soc 127(5):1346–1347
38. Yang W, Wilkins AL, Ye Y, Z-r L, S-y L, Urbauer JL, Hellinga HW, Kearney A, van der
Merwe PA, Yang JJ (2005) Design of a calcium-binding protein with desired structure in a
cell adhesion molecule. J Am Chem Soc 127(7):2085–2093
39. Palmer AE, Giacomello M, Kortemme T, Hires SA, Lev-Ram V, Baker D, Tsien RY (2006)
Ca2+ indicators based on computationally redesigned calmodulin-peptide pairs. Chem Biol
13(5):521–530
40. Lassila JK, Keeffe JR, Oelschlaeger P, Mayo SL (2005) Computationally designed variants
of Escherichia coli chorismate mutase show altered catalytic activity. Protein Eng Des Sel
18(4):161–163
41. Bornscheuer UT, Pohl M (2001) Improved biocatalysts by directed evolution and rational
protein design. Curr Opin Chem Biol 5(2):137–143
42. Faiella M, Andreozzi C, de Rosales RTM, Pavone V, Maglio O, Nastri F, DeGrado WF,
Lombardi A (2009) An artificial di-iron oxo-protein with phenol oxidase activity. Nat Chem
Biol 5(12):882–884
43. Tinberg CE, Khare SD, Dou J, Doyle L, Nelson JW, Schena A, Jankowski W, Kalodimos CG,
Johnsson K, Stoddard BL, Baker D (2013) Computational design of ligand binding proteins
with high affinity and selectivity. Nature 501(7466):212–216
44. Kaplan J, DeGrado WF (2004) De novo design of catalytic proteins. Proc Natl Acad Sci U S
A 101(32):11566–11570
45. Dahiyat BI, Mayo SL (1997) De novo protein design: fully automated sequence selection.
Science 278(5335):82–87
46. Jiang L, Althoff EA, Clemente FR, Doyle L, Röthlisberger D, Zanghellini A, Gallaher JL,
Betker JL, Tanaka F, Barbas CF, Hilvert D, Houk KN, Stoddard BL, Baker D (2008) De novo
computational design of retro-aldol enzymes. Science (New York, NY)
319(5868):1387–1391
47. Rothlisberger D, Khersonsky O, Wollacott AM, Jiang L, DeChancie J, Betker J, Gallaher JL,
Althoff EA, Zanghellini A, Dym O, Albeck S, Houk KN, Tawfik DS, Baker D (2008) Kemp
elimination catalysts by computational enzyme design. Nature 453(7192):190–195
48. Moroz YS, Dunston TT, Makhlynets OV, Moroz OV, Wu Y, Yoon JH, Olsen AB, McLaughlin
JM, Mack KL, Gosavi PM, van Nuland NAJ, Korendovych IV (2015) New tricks for old
proteins: single mutations in a nonenzymatic protein give rise to various enzymatic activities.
J Am Chem Soc 137(47):14905–14911
49. Zanghellini A (2014) De novo computational enzyme design. Curr Opin Biotechnol
29:132–138
50. Petrounia IP, Arnold FH (2000) Designed evolution of enzymatic properties. Curr Opin
51. Arnold FH (2001) Combinatorial and computational challenges for biocatalyst design.
Nature 409(6817):253–257
52. Minshull J, Willem Stemmer PC (1999) Protein evolution by molecular breeding. Curr Opin
Chem Biol 3(3):284–290
53. Packer MS, Liu DR (2015) Methods for the directed evolution of proteins. Nat Rev Genet
16(7):379–394
54. Jaeger K-E, Eggert T (2004) Enantioselective biocatalysis optimized by directed evolution.
Curr Opin Biotechnol 15(4):305–313
55. Jestin J-L, Kaminski PA (2004) Directed enzyme evolution and selections for catalysis based
on product formation. J Biotechnol 113(103):85
56. Tao H, Cornish VW (2002) Milestones in directed enzyme evolution. Curr Opin Chem Biol
6(6):858–864
57. Williams GJ, Nelson AS, Berry A (2004) Directed evolution of enzymes for biocatalysis and
the life sciences. Cell Mol Life Sci CMLS 61(24):3034–3046
58. Dalby PA (2003) Optimising enzyme function by directed evolution. Curr Opin Struct Biol
13(4):500–505
59. Bershtein S, Tawfik DS (2008) Advances in laboratory evolution of enzymes. Curr Opin
Chem Biol 12(2):151–158
60. Park S, Morley KL, Horsman GP, Holmquist M, Hult K, Kazlauskas RJ (2005) Focusing
mutations into the P. fluorescens esterase binding site increases enantioselectivity more effec-
tively than distant mutations. Chem Biol 12 (1):45–54
61. Strausberg SL, Ruan B, Fisher KE, Alexander PA, Bryan PN (2005) Directed coevolution of
stability and catalytic activity in calcium-free subtilisin. Biochemistry 44(9):3272–3279
62. Chockalingam K, Chen Z, Katzenellenbogen JA, Zhao H (2005) Directed evolution of spe-
cific receptor–ligand pairs for use in the creation of gene switches. Proc Natl Acad Sci U S A
102(16):5691–5696
63. Chica RA, Doucet N, Pelletier JN (2005) Semi-rational approaches to engineering enzyme
activity: combining the benefits of directed evolution and rational design. Curr Opin
64. Hill CM, Li W-S, Thoden JB, Holden HM, Raushel FM (2003) Enhanced degradation of
chemical warfare agents through molecular engineering of the phosphotriesterase active site.
J Am Chem Soc 125(30):8990–8991
65. Reetz MT, Carballeira JD (2007) Iterative saturation mutagenesis (ISM) for rapid directed
evolution of functional enzymes. Nat Protocol 2(4):891–903
66. Lutz S, Patrick WM (2004) Novel methods for directed evolution of enzymes: quality, not
quantity. Curr Opin Biotechnol 15(4):291–297
278 E. Monza et al.
67. Lutz S (2010) Beyond directed evolution–semi-rational protein engineering and design. Curr
Opin Biotechnol 21(6):734–743
68. Lippow SM, Moon TS, Basu S, Yoon S-H, Li X, Chapman BA, Robison K, Lipovšek D,
Prather KLJ (2010) Engineering enzyme specificity using computational design of a defined-
sequence library. Chem Biol 17(12):1306–1315
69. Sebestova E, Bendl J, Brezovsky J, Damborský J (2014) Computational tools for designing
smart libraries. Methods Mol Biol 1179:291–314
70. Voigt CA, Mayo SL, Arnold FH, Wang Z-G (2001) Computationally focusing the directed
evolution of proteins. J Cell Biochem 84(S37):58–63
71. Zaugg J, Gumulya Y, Gillam EM, Boden M (2014) Computational tools for directed evolu-
tion: a comparison of prospective and retrospective strategies. Methods Mol Biol
1179:315–333
72. Damborsky J, Brezovsky J (2009) Computational tools for designing and engineering bio-
catalysts. Curr Opin Chem Biol 13(1):26–34
73. Pei J (2008) Multiple protein sequence alignment. Curr Opin Struct Biol 18(3):382–386
74. Pavelka A, Chovancova E, Damborsky J (2009) HotSpot Wizard: a web server for identifica-
tion of hot spots in protein engineering. Nucleic Acids Res 37(suppl 2):W376–W383
75. Kuipers RK, Joosten H-J, van Berkel WJH, Leferink NGH, Rooijen E, Ittmann E, van
Zimmeren F, Jochens H, Bornscheuer U, Vriend G, Martins dos Santos VAP, Schaap PJ
(2010) 3DM: systematic analysis of heterogeneous superfamily data to discover protein func-
tionalities. Proteins Struct Funct Bioinforma 78(9):2101–2113
76. Jochens H, Bornscheuer UT (2010) Natural diversity to guide focused directed evolution.
ChemBioChem 11(13):1861–1866
77. Goldsmith M, Tawfik DS (2012) Directed enzyme evolution: beyond the low-hanging fruit.
Curr Opin Struct Biol 22(4):406–412
78. Barak Y, Nov Y, Ackerley DF, Matin A (2007) Enzyme improvement in the absence of struc-
tural knowledge: a novel statistical approach. ISME J 2(2):171–179
79. Rosenberg M, Goldblum A (2006) Computational protein design: a novel path to future pro-
tein drugs. Curr Pharm Des 12(31):3973–3997
80. Poole AM, Ranganathan R (2006) Knowledge-based potentials in protein design. Curr Opin
Struct Biol 16(4):508–513
81. Koder RL, Dutton PL (2006) Intelligent design: the de novo engineering of proteins with
specified functions. Dalton Trans 25:3045–3051
82. Butterfoss GL, Kuhlman B (2006) Computer-based design of novel protein structures. Annu
Rev Biophys Biomol Struct 35(1):49–65
83. Ambroggio XI, Kuhlman B (2006) Design of protein conformational switches. Curr Opin
Struct Biol 16(4):525–530
84. Vizcarra CL, Mayo SL (2005) Electrostatics in computational protein design. Curr Opin
Chem Biol 9(6):622–626
85. Morin A, Meiler J, Mizoue LS (2011) Computational design of protein-ligand interfaces:
potential in therapeutic development. Trends Biotechnol 29(4):159–166
86. Malisi C, Schumann M, Toussaint NC, Kageyama J, Kohlbacher O, Höcker B (2012) Binding
pocket optimization by computational protein design. PLoS ONE 7(12):e52505
87. Saven JG (2011) Computational protein design: engineering molecular diversity, nonnatural
enzymes, nonbiological cofactor complexes, and membrane proteins. Curr Opin Chem Biol
15(3):452–457
88. Ollikainen N, Smith CA, Fraser JS, Kortemme T (2013) Methods in enzymology: “Flexible
backbone sampling methods to model and design protein alternative conformations”.
Methods Enzymol 523:61–85
89. Park S, Yang X, Saven JG (2004) Advances in computational protein design. Curr Opin Struct
Biol 14(4):487–494
90. Samish I, MacDermaid CM, Perez-Aguilar JM, Saven JG (2011) Theoretical and computa-
tional protein design. Annu Rev Phys Chem 62(1):129–149
91. Smith RD, Damm-Ganamet KL, Dunbar JB, Ahmed A, Chinnaswamy K, Delproposto JE,
Kubish GM, Tinberg CE, Khare SD, Dou J, Doyle L, Stuckey JA, Baker D, Carlson HA
(2015) CSAR benchmark exercise 2013: evaluation of results from a combined computa-
tional protein design, docking, and scoring/ranking challenge. J Chem Inf Model ASAP
56:1022
92. Wijma HJ, Janssen DB (2013) Computational design gains momentum in enzyme catalysis
engineering. FEBS J 280(13):2948–2960
93. Boas FE, Harbury PB (2007) Potential energy functions for protein design. Curr Opin Struct
Biol 17(2):199–204
94. Boas FE, Harbury PB (2008) Design of protein-ligand binding based on the molecular-
mechanics energy model. J Mol Biol 380(2):415–424
95. Sirin S, Pearlman DA, Sherman W (2014) Physics-based enzyme design: predicting binding
affinity and catalytic activity. Proteins Struct Funct Bioinforma 82(12):3397–3409
96. Wickstrom L, Gallicchio E, Levy RM (2012) The linear interaction energy method for the
prediction of protein stability changes upon mutation. Proteins 80(1):111–125
97. Mendes J, Guerois R, Serrano L (2002) Energy estimation in protein design. Curr Opin Struct
Biol 12(4):441–446
98. Schneider M, Fu X, Keating AE (2009) X-ray vs. NMR structures as templates for computa-
tional protein design. Proteins 77(1):97–110
99. Adamczyk AJ, Cao J, Kamerlin SCL, Warshel A (2011) Catalysis by dihydrofolate reductase
and other enzymes arises from electrostatic preorganization, not conformational motions.
100. Gagné D, French Rachel L, Narayanan C, Simonović M, Agarwal Pratul K, Doucet N (2015)
Perturbation of the conformational dynamics of an active-site loop alters enzyme activity.
Structure 23(12):2256–2266
101. Bhabha G, Lee J, Ekiert DC, Gam J, Wilson IA, Dyson HJ, Benkovic SJ, Wright PE (2011)
A dynamic knockout reveals that conformational fluctuations influence the chemical step of
enzyme catalysis. Science 332(6026):234–238
102. Allen BD, Nisthal A, Mayo SL (2010) Experimental library screening demonstrates the suc-
cessful application of computational protein design to large structural ensembles. Proc Natl
Acad Sci 107(46):19838–19843
103. Fu X, Apgar JR, Keating AE (2007) Modeling backbone flexibility to achieve sequence diver-
sity: the design of novel α-helical ligands for Bcl-xL. J Mol Biol 371(4):1099–1117
104. Smith CA, Kortemme T (2008) Backrub-like backbone simulation recapitulates natural pro-
tein conformational variability and improves mutant side-chain prediction. J Mol Biol
380(4):742–756
105. Lassila JK (2010) Conformational diversity and computational enzyme design. Curr Opin
Chem Biol 14(5):676–682
106. Mandell DJ, Kortemme T (2009) Backbone flexibility in computational protein design. Curr
Opin Biotechnol 20(4):420–428
107. Marrink SJ, Risselada HJ, Yefimov S, Tieleman DP, De Vries AH (2007) The MARTINI
force field: coarse grained model for biomolecular simulations. J Phys Chem B
111(27):7812–7824
108. Bowen JP, Allinger NL (2007) Molecular mechanics: the art and science of parameterization.
Rev Comput Chem 2:81–97
109. Doruker P, Atilgan AR, Bahar I (2000) Dynamics of proteins predicted by molecular dynam-
ics simulations and analytical approaches: application to α-amylase inhibitor. Proteins Struct
Funct Bioinforma 40(3):512–524
110. Berendsen H (1988) Dynamic simulation as an essential tool in molecular modeling.
J Comput Aided Mol Des 2(3):217–221
111. Grossman J, Towles B, Greskamp B, Shaw DE (2015) Filtering, reductions and synchroniza-
tion in the anton 2 network. In: Parallel and Distributed Processing Symposium (IPDPS),
2015 IEEE International. IEEE, pp 860–870
280 E. Monza et al.
112. Rathore N, de Pablo JJ (2002) Monte Carlo simulation of proteins through a random walk in
energy space. J Chem Phys 116(16):7225–7230
113. Borrelli KW, Vitalis A, Alcantara R, Guallar V (2005) PELE: protein energy landscape
exploration. A novel Monte Carlo based technique. J Chem Theory Comput 1(6):1304–1311
114. Cabeza de Vaca I, Lucas MF, Guallar V (2015) New Monte Carlo based technique to study
DNA–ligand interactions. J Chem Theory Comput 11(12):5598–5605
115. Borrelli KW, Cossins B, Guallar V (2010) Exploring hierarchical refinement techniques for
induced fit docking with protein and ligand flexibility. J Comput Chem 31(6):1224–1235
116. Halperin I, Ma B, Wolfson H, Nussinov R (2002) Principles of docking: an overview of search
algorithms and a guide to scoring functions. Proteins Struct Funct Bioinforma 47(4):409–443
117. Gao J, Truhlar DG (2002) Quantum mechanical methods for enzyme kinetics. Annu Rev
Phys Chem 53(1):467–505
118. Korkegian A (2005) Computational thermostabilization of an enzyme. Science
308(5723):857–860
119. Kellogg EH, Leaver-Fay A, Baker D (2011) Role of conformational sampling in computing
mutation-induced changes in protein structure and stability. Proteins 79(3):830–838
120. Dunbrack RL Jr (2002) Rotamer libraries in the 21st century. Curr Opin Struct Biol
12(4):431–440
121. Kuhlman B, Baker D (2000) Native protein sequences are close to optimal for their struc-
tures. Proc Natl Acad Sci 97(19):10383–10388
122. Potapov V, Cohen M, Schreiber G (2009) Assessing computational methods for predicting
protein stability upon mutation: good on average but not in the details. Protein Eng Des Sel
22(9):553–560
123. Estrada J, Echenique P, Sancho J (2015) Predicting stabilizing mutations in proteins using
Poisson-Boltzmann based models: study of unfolded state ensemble models and develop-
ment of a successful binary classifier based on residue interaction energies. Phys Chem Chem
Phys 17(46):31044–31054
124. Karplus M, Ichiye T, Pettitt BM (1987) Configurational entropy of native proteins. Biophys
J 52(6):1083–1085
125. Chong S-H, Ham S (2015) Dissecting protein configurational entropy into conformational
and vibrational contributions. J Phys Chem B 119(39):12623–12631
126. Frappier V, Chartier M, Najmanovich RJ (2015) ENCoM server: exploring protein conforma-
tional space and the effect of mutations on protein function and stability. Nucleic Acids Res
43(W1):W395–W400
127. Frappier V, Najmanovich RJ (2014) A coarse-grained elastic network atom contact model
and its use in the simulation of protein dynamics and the prediction of the effect of mutations.
PLoS Comput Biol 10(4):e1003569
128. Seeliger D, Daniel S, de Groot BL (2010) Protein thermostability calculations using alchemi-
cal free energy simulations. Biophys J 98(10):2309–2316
129. Huang X, Gao D, Zhan C-G (2011) Computational design of a thermostable mutant of
cocaine esterase via molecular dynamics simulations. Org Biomol Chem 9(11):4138–4143
130. Joo JC, Pack SP, Kim YH, Yoo YJ (2011) Thermostabilization of Bacillus circulans xyla-
nase: computational optimization of unstable residues based on thermal fluctuation analysis.
J Biotechnol 151(1):56–65
131. Lee C-W, Wang H-J, Hwang J-K, Tseng C-P (2014) Protein thermal stability enhancement
by designing salt bridges: a combined computational and experimental study. PLoS ONE
9(11):e112751
132. Pikkemaat MG, Linssen ABM, Berendsen HJC, Janssen DB (2002) Molecular dynamics
simulations as a tool for improving protein stability. Protein Eng 15(3):185–192
133. Gribenko AV, Patel MM, Liu J, McCallum SA, Wang C, Makhatadze GI (2009) Rational
stabilization of enzymes by computational redesign of surface charge-charge interactions.
Proc Natl Acad Sci 106(8):2601–2606
134. Spector S, Wang M, Carp SA, Robblee J, Hendsch ZS, Fairman R, Tidor B, Raleigh DP
(2000) Rational modification of protein stability by the mutation of charged surface residues.
135. Schweiker KL, Arash Z-A, Davidson AR, Makhatadze GI (2007) Computational design of
the Fyn SH3 domain with increased stability through optimization of surface charge-charge
interactions. Protein Sci 16(12):2694–2702
136. Borgo B, Havranek JJ (2012) Automated selection of stabilizing mutations in designed and
natural proteins. Proc Natl Acad Sci U S A 109(5):1494–1499
137. Hendsch ZS, Thorlakur J, Sauer RT, Bruce T (1996) Protein stabilization by removal of
unsatisfied polar groups: computational approaches and experimental tests. Biochemistry
35(24):7621–7625
138. Koudelakova T, Chaloupkova R, Brezovsky J, Prokop Z, Sebestova E, Hesseler M, Khabiri M,
Plevaka M, Kulik D, Kuta Smatanova I, Rezacova P, Ettrich R, Bornscheuer UT, Damborsky
J (2013) Engineering enzyme stability and resistance to an organic cosolvent by modification
of residues in the access tunnel. Angew Chem Int Ed Engl 52(7):1959–1963
139. Wijma HJ, Floor RJ, Jekel PA, Baker D, Marrink SJ, Janssen DB (2014) Computationally
designed libraries for rapid enzyme stabilization. Protein Eng Des Sel 27(2):49–58
140. Wijma HJ, Floor RJ, Janssen DB (2013) Structure- and sequence-analysis inspired engineer-
ing of proteins for enhanced thermostability. Curr Opin Struct Biol 23(4):588–594
141. Schreier B, Stumpp C, Wiesner S, Hocker B (2009) Computational design of ligand binding
is not a solved problem. Proc Natl Acad Sci 106(44):18491–18496
142. Allison B, Combs S, DeLuca S, Lemmon G, Mizoue L, Meiler J (2014) Computational
design of protein-small molecule interfaces. J Struct Biol 185(2):193–202
143. Gainza P, Roberts KE, Georgiev I, Lilien RH, Keedy DA, Chen C-Y, Reza F, Anderson AC,
Richardson DC, Richardson JS, Donald BR (2013) OSPREY: protein design with ensembles,
flexibility, and provable algorithms. Methods Enzymol 523:87–107
144. Keedy DA, Georgiev I, Triplett EB, Donald BR, Richardson DC, Richardson JS (2012)
The role of local backrub motions in evolved and designed mutations. PLoS Comput Biol
8(8):e1002629
145. Davis IW, Bryan Arendall W, Richardson DC, Richardson JS (2006) The backrub motion:
how protein backbone shrugs when a sidechain dances. Structure 14(2):265–274
146. Chen C-Y, Georgiev I, Anderson AC, Donald BR (2009) Computational structure-based
redesign of enzyme activity. Proc Natl Acad Sci U S A 106(10):3764–3769
147. Frey KM, Georgiev I, Donald BR, Anderson AC (2010) Predicting resistance mutations
using protein design algorithms. Proc Natl Acad Sci U S A 107(31):13707–13712
148. Zhou Y, Xu W, Donald BR, Zeng J (2014) An efficient parallel algorithm for accelerating
computational protein design. Bioinformatics 30(12):i255–i263
149. Hallen MA, Keedy DA, Donald BR (2013) Dead-end elimination with perturbations
(DEEPer): a provable protein design algorithm with continuous sidechain and backbone flex-
ibility. Proteins 81(1):18–39
150. Lilien RH, Stevens BW, Anderson AC, Donald BR (2005) A novel ensemble-based scor-
ing and search algorithm for protein redesign and its application to modify the substrate
specificity of the gramicidin synthetase a phenylalanine adenylation enzyme. J Comput Biol
12(6):740–761
151. Leach AR (2001) Molecular modelling: principles and applications. Pearson Education,
New York
152. Shields GC, Seybold PG (2013) Computational approaches for the prediction of pKa values.
CRC Press, Boca Raton
153. Pardo I, Santiago G, Gentili P, Lucas F, Monza E, Medrano F, Galli C, Martínez A, Guallar V,
Camarero S (2016) Re-designing the substrate binding pocket of laccase for enhanced oxida-
tion of sinapic acid. Catal Sci Technol ASAP 6:3900
154. Young T, Abel R, Kim B, Berne BJ, Friesner RA (2007) Motifs for molecular recogni-
tion exploiting hydrophobic enclosure in protein–ligand binding. Proc Natl Acad Sci
104(3):808–813
155. Kiss G, Çelebi-Ölçüm N, Moretti R, Baker D, Houk KN (2013) Computational enzyme
design. Angew Chem Int Ed 52(22):5700–5725
156. Doerr S, De Fabritiis G (2014) On-the-fly learning and sampling of ligand binding by high-
throughput molecular simulations. J Chem Theory Comput 10(5):2064–2069
282 E. Monza et al.
157. Wijma HJ, Floor RJ, Bjelic S, Marrink SJ, Baker D, Janssen DB (2015) Enantioselective
enzymes by computational design and in silico screening. Angew Chem Int Ed Engl
54(12):3726–3730
158. Jiménez-Osés G, Osuna S, Gao X, Sawaya MR, Gilson L, Collier SJ, Huisman GW, Yeates
TO, Tang Y, Houk KN (2014) The role of distant mutations and allosteric regulation on LovD
active site dynamics. Nat Chem Biol 10(6):431–436
159. Osuna S, Jiménez-Osés G, Noey EL, Houk KN (2015) Molecular dynamics explorations of
active site structure in designed and evolved enzymes. Acc Chem Res 48(4):1080–1089
160. DuBay KH, Bowman GR, Geissler PL (2015) Fluctuations within folded proteins: implica-
tions for thermodynamic and allosteric regulation. Acc Chem Res 48(4):1098–1105
161. Sethi A, Eargle J, Black AA, Luthey-Schulten Z (2009) Dynamical networks in tRNA:protein
complexes. Proc Natl Acad Sci U S A 106(16):6620–6625
162. Madadkar-Sobhani A, Guallar V (2013) PELE web server: atomistic study of biomolecular
systems at your fingertips. Nucleic Acids Res 41(Web Server issue):W322–W328
163. Lucas MF, Guallar V (2012) An atomistic view on human hemoglobin carbon monoxide
migration processes. Biophys J 102(4):887–896
164. Takahashi R, Gil VA, Guallar V (2014) Monte Carlo free ligand diffusion with Markov
state model analysis and absolute binding free energy calculations. J Chem Theory Comput
10(1):282–288
165. Hosseini A, Brouk M, Lucas MF, Glaser F, Fishman A, Guallar V (2015) Atomic picture of
ligand migration in toluene 4-monooxygenase. J Phys Chem B 119(3):671–678
166. Lüdemann SK, Lounnas V, Wade RC (2000) How do substrates enter and products exit the
buried active site of cytochrome P450cam? 1. Random expulsion molecular dynamics inves-
tigation of ligand access channels and mechanisms. J Mol Biol 303(5):797–811
167. Grubmüller H, Heymann B, Tavan P (1996) Ligand binding: molecular mechanics calcula-
tion of the streptavidin-biotin rupture force. Science 271(5251):997–999
168. Le Guilloux V, Schmidtke P, Tuffery P (2009) Fpocket: an open source platform for ligand
pocket detection. BMC Bioinforma 10(1):168
169. Chovancova E, Eva C, Antonin P, Petr B, Ondrej S, Jan B, Barbora K, Artur G, Vilem S,
Martin K, Petr M, Lada B, Jiri S, Jiri D (2012) CAVER 3.0: a tool for the analysis of transport
pathways in dynamic protein structures. PLoS Comput Biol 8(10):e1002708
170. Senn HM, Walter T (2009) QM/MM methods for biomolecular systems. Angew Chem Int Ed
48(7):1198–1229
171. Chaskar P, Prasad C, Vincent Z, Röhrig UF (2014) Toward on-the-fly quantum mechanical/
molecular mechanical (QM/MM) docking: development and benchmark of a scoring func-
tion. J Chem Inf Model 54(11):3137–3152
172. Cho AE, Victor G, Berne BJ, Richard F (2005) Importance of accurate charges in molecular
docking: quantum mechanical/molecular mechanical (QM/MM) approach. J Comput Chem
26(9):915–931
173. Fedorov DG, Nagata T, Kitaura K (2012) Exploring chemistry with the fragment molecular
orbital method. Phys Chem Chem Phys 14(21):7562–7577
174. Jensen JH, Willemoës M, Winther JR, De Vico L (2014) In silico prediction of mutant HIV-1
proteases cleaving a target sequence. PLoS ONE 9(5):e95833
175. Grisewood MJ, Gifford NP, Pantazes RJ, Li Y, Cirino PC, Janik MJ, Maranas CD (2013)
OptZyme: computational enzyme redesign using transition state analogues. PLoS ONE
8(10):e75358
176. Atkins PW (1998) Physical chemistry. W H Freeman & Company, New York
177. Khersonsky O, Rothlisberge D, Wollacott AM, Dym O, Baker D, Tawfik DS (2011)
Optimization of the in silico designed Kemp eliminase KE70 by computational design and
directed evolution. J Mol Biol 407(3):391–412
178. Genheden S, Samuel G, Ulf R (2015) The MM/PBSA and MM/GBSA methods to estimate
ligand-binding affinities. Expert Opin Drug Discov 10(5):449–461
179. van der Kamp MW, Mulholland AJ (2013) Combined quantum mechanics/molecular mechan-
ics (QM/MM) methods in computational enzymology. Biochemistry 52(16):2708–2728
180. Zheng F, Fang Z, Wenchao Y, Mei-Chuan K, Junjun L, Hoon C, Daquan G, Min T, Hsin-
Hsiung T, Woods JH, Chang-Guo Z (2008) Most efficient cocaine hydrolase designed by
virtual screening of transition states. J Am Chem Soc 130(36):12148–12155
181. Kamerlin SCL, Arieh W (2011) The empirical valence bond model: theory and applications.
Wiley Interdiscip Rev Comput Mol Sci 1(1):30–45
182. Frushicheva MP, Cao J, Chu ZT, Warshel A (2010) Exploring challenges in rational enzyme
design by simulating the catalysis in artificial kemp eliminase. Proc Natl Acad Sci U S A
107(39):16869–16874
183. Frushicheva MP, Cao J, Warshel A (2011) Challenges and advances in validating enzyme
design proposals: the case of kemp eliminase catalysis. Biochemistry 50(18):3849–3858
184. Amrein BA, Ireneusz Szeler F, Purg M, Kulkarni Y, Kamerlin SCL (2017) CADEE:
Computer-aided directed evolution of enzymes. IUCrJ 4:50–64.
185. Hediger MR, De Vico L, Svendsen A, Besenmatter W, Jensen JH (2012) A computational
methodology to screen activities of enzyme variants. PLoS ONE 7(12):e49849
186. Hediger MR, Casper S, De Vico L, Jensen JH (2013) A computational method for the sys-
tematic screening of reaction barriers in enzymes: searching for Bacillus circulans xylanase
mutants with greater activity towards a synthetic substrate. PeerJ 1:e111
187. Hediger MR, De Vico L, Rannes JB, Christian J, Werner B, Allan S, Jensen JH (2013) In silico
screening of 393 mutants facilitates enzyme engineering of amidase activity in CalB. PeerJ
1:e145
188. Ito M, Mika I, Tore B (2014) Novel approach for identifying key residues in enzymatic reac-
tions: proton abstraction in ketosteroid isomerase. J Phys Chem B 118(46):13050–13058
189. Steinmann C, Fedorov DG, Jensen JH (2012) The effective fragment molecular orbital
method for fragments connected by covalent bonds. PLoS ONE 7(7):e41117
190. Steinmann C, Casper S, Fedorov DG, Jensen JH (2013) Mapping enzymatic catalysis using
the effective fragment molecular orbital method: towards all ab initio biochemistry. PLoS
ONE 8(4):e60602
191. Marcus RA (1993) Electron transfer reactions in chemistry. Theory and experiment. Rev
Mod Phys 65(3):599–610
192. Blumberger J, Jochen B (2008) Free energies for biological electron transfer from QM/
MM calculation: method, application and critical assessment. Phys Chem Chem Phys
10(37):5651
193. Wallrapp FH, Voityuk AA, Guallar V (2013) In-silico assessment of protein-protein elec-
tron transfer. A case study: cytochrome c peroxidase–cytochrome c. PLoS Comput Biol
9(3):e1002990
194. Monza E, Lucas MF, Camarero S, Alejaldre LC, Martínez AT, Guallar V (2015) Insights into
laccase engineering from molecular simulations: toward a binding-focused strategy. J Phys
Chem Lett 6(8):1447–1453
195. Gerard S, Felipe de S, Fátima Lucas M, Emanuele M, Sandra A, Ángel TM, Susana Camarero,
VG (2016) Computer-aided laccase engineering: toward biological oxidation of arylamines.
ACS Catalysis, 6:5415–5423
196. Acebes S, Fernandez-Fueyo E, Monza E, Lucas M, Almendral D, Ruiz-Dueñas FJ, Lund H,
Martinez AT, Guallar V (2016) Rational enzyme engineering through biophysical and bio-
chemical modeling. ACS Catal ACS Catalysis 6(3):1624–1629
197. Guallar V, Wallrapp F (2008) Mapping protein electron transfer pathways with QM/MM
methods. J R Soc Interface 5(0):S233
198. Vidal-Limón A, Águila S, Ayala M, Batista CV, Vazquez-Duhalt R (2013) Peroxidase activ-
ity stabilization of cytochrome P450 BM3 by rational analysis of intramolecular electron
transfer. J Inorg Biochem 122:18–26
199. Fox RJ, Huisman GW (2008) Enzyme optimization: moving from blind evolution to statisti-
cal exploration of sequence–function space. Trends Biotechnol 26(3):132–138
200. Feng X, Sanchis J, Reetz MT, Rabitz H (2012) Enhancing the efficiency of directed evolu-
tion in focused enzyme libraries by the adaptive substituent reordering algorithm. Chem Eur
J 18(18):5646–5654
284 E. Monza et al.
201. Cui Q, Elstner M (2014) Density functional tight binding: values of semi-empirical methods
in an ab initio era. Phys Chem Chem Phys 16(28):14368–14377
202. Christensen AS, Elstner M, Cui Q (2015) Improving intermolecular interactions in
DFTB3 using extended polarization from chemical-potential equalization. J Chem Phys
143(8):084123
203. Yilmazer ND, Korth M (2015) Enhanced semiempirical QM methods for biomolecular inter-
actions. Comput Struct Biotechnol J 13:169–175
204. Privett HK, Kiss G, Lee TM, Blomberg R, Chica RA, Thomas LM, Hilvert D, Houk KN,
Mayo SL (2012) Iterative approach to computational enzyme design. Proc Natl Acad Sci
109(10):3790–3795

Book - Directed Enzyme Evolution Advancees and Applications PDF

Uploaded by

Copyright:

Available Formats

Book - Directed Enzyme Evolution Advancees and Applications PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Book - Directed Enzyme Evolution Advancees and Applications PDF

Uploaded by

Copyright:

Available Formats

Miguel Alcalde Editor

ISBN 978-3-319-50411-7 ISBN 978-3-319-50413-1 (eBook)

Library of Congress Control Number: 2017932555

© Springer International Publishing AG 2017

Printed on acid-free paper

This Springer imprint is published by Springer Nature

on several evolutionary success stories, involving tryptophan synthases, therapeutic

Madrid, Spain Miguel Alcalde

1 Directed Evolution of an Allosteric Tryptophan Synthase to Create a

J. Murciano-Calles • A.R. Buller • F.H. Arnold (*)

© Springer International Publishing AG 2017 1

In addition to being one of the standard 20 proteinogenic α-amino acids, tryptophan

1.2  ctivation of TrpB from Pyrococcus furiosus by Directed

We constructed a random mutagenesis library of PfTrpB by error-prone PCR and

Fig. 1.4 Sites where

mutations distributed across 20 improved clones. Although the effect of each

1.3  ctivation of TrpB from Other Species by Transfer

change specificity from NADPH to NADH. However, engineering allostery

residue in the equivalent position of EcTrpB. We made a saturation mutagenesis

1.4  irected Evolution of PfTrpB for the Synthesis

L-Threonine Indole (2S,3S)-β-MeTrp

reaction proceeds with >99% enantiomeric and diastereomeric excess, highlighting

1.5 Summary and Outlook

Acknowledgments We gratefully thank Sabine Brinkmann-Chen for a critical reading of this

S. Lutz (*) • E. Williams • P. Muthu

© Springer International Publishing AG 2017 17

minimal immunogenic response from patient. These additional constraints signifi-

2.1.1 Pharmaceutical Enzymes

The subcategory of pharmaceutical enzymes represents proteins that themselves

2.1.2 Enzymes for Suicide Gene Therapy

A more recent application of therapeutic enzymes is toward the development of

decades, chemotherapy has been a component of standard of treatment for can-

2.1.3 Diagnostic Enzymes

Proteases play key roles in regulating fundamental biological processes, approxi-

2.2.1.1 Urokinase Plasminogen Activator and Tissue Plasminogen

2.2.1.2 Engineered Thrombin

2.2.2 Procoagulant Protease Engineering

Factor VII (FVII) is a protease in the procoagulant protease cascade required to

The clinical outcome of the modified FVIIa variants highlights a difficulty of

2.2.3 Alzheimer’s Disease

A role for therapeutic proteases in the treatment of multiple neurological disorders

In 1989, botulinum neurotoxin serotype A (BoNT/A) from Clostridium botulinum

Thrombin W215A/E217A Decreased procoagulant fibrinogen Cantwell (2000)

2.2.5 Digestive Proteases

physiologically relevant, acidic environment (pH 4.5) and to increase resistance to

2.2.6 Protease Engineering Strategies

In regard to protease engineering in general, a recent review highlights a range of

of engineering proteases as therapeutics. Equally interesting is the work of research-

Ribonucleases (RNases) have been investigated as anticancer therapeutics for over

2.4 Amino Acid-Degrading Enzymes

A number of enzymes have been investigated as potential anticancer agents due to

2.4.1 Asparagine-Degrading Enzymes

Asparaginase II from E. coli (EcAII) is currently a key component in the treatment

2.4.2 Arginine-Degrading Enzymes

performance and thermostability, as well as shifting the enzyme’s pH optimum from

2.5 Deoxyribonucleoside Kinases

Beyond applications of engineered enzymes as therapeutic agents themselves, their

2.5.1 Thymidine Kinase from Herpes Simplex Virus

mammalian dNKs [117]. In fact, the phosphorylative activation of aciclovir (ACV,

Madrid, Spain Miguel Alcalde

1.2 ctivation of TrpB from Pyrococcus furiosus by Directed

1.3 ctivation of TrpB from Other Species by Transfer

1.4 irected Evolution of PfTrpB for the Synthesis

1.5 Summary and Outlook

2.1.1 Pharmaceutical Enzymes

2.1.2 Enzymes for Suicide Gene Therapy

2.1.3 Diagnostic Enzymes

2.2.1.1 Urokinase Plasminogen Activator and Tissue Plasminogen

2.2.1.2 Engineered Thrombin

2.2.2 Procoagulant Protease Engineering

2.2.3 Alzheimer’s Disease

2.2.5 Digestive Proteases

2.2.6 Protease Engineering Strategies

2.4 Amino Acid-Degrading Enzymes

2.4.1 Asparagine-Degrading Enzymes

2.4.2 Arginine-Degrading Enzymes

2.5 Deoxyribonucleoside Kinases

2.5.1 Thymidine Kinase from Herpes Simplex Virus

2.5.2 2′-Deoxyribonucleoside Kinase from Drosophila

2.5.3 2′-Deoxycytidine Kinase from Homo sapiens

2.5.4 From Suicide Gene Therapy to PET Reporter Systems

2′-fluoro-2′-deoxy-1-β-L-arabinofuranosyl-5-methyluracil (L-FMAU, 6) and found

2.6 Cytosine Deaminases

2.7 Purine Nucleoside Phosphorylases

2.9 holinesterases: Bridging Protein Therapeutics

2.9.3 Other Esterases

3.2 Cytochrome P450 Monooxygenases

3.3 Baeyer-Villiger Monooxygenases

3.4 Lipases and Esterases

3.5 Epoxide Hydrolases

3.7 Alternative Mutagenesis Approaches