Journal of Chromatography A, 849 (2015) 421431

Isolation of indole alkaloids from Catharanthus roseus by

centrifugal partition chromatography in the pH-zone rening mode
Jean-Hugues Renault* , Jean-Marc Nuzillard, Gaelle
Le Crouerour,
Philippe Thepenier,

Monique Zeches-Hanrot,
` Louisette Le Men-Olivier
Laboratoire de Pharmacognosie, UPRES-A CNRS 6013, CPCBAI, B.P. 1039, Moulin de la Housse, 51097 Reims Cedex, France

Received 19 February 2015; received in revised form 6 April 2015; accepted 7 April 2015

Abstract

Centrifugal partition chromatography (CPC) in the pH-zone rening mode allowed a preparative and efcient isolation of
vindoline, vindolinine, catharanthine and vincaleukoblastine from a crude mixture of Catharanthus roseus alkaloids. The
separation protocol was tested with a synthetic mixture of vindoline, catharanthine and vincaleukoblastine. The fraction
content was analyzed and the results compared with theoretical chromatograms obtained by numerical simulation. The
increase in injected sample mass results in an improvement of the purity of the isolated compounds. This observation,
conrmed by theory, is of prime importance for the development of preparative pH-zone rening CPC as a preparative
separation method. 1999 Published by Elsevier Science B.V. All rights reserved.
Keywords: Counter-current chromatography; Catharanthus roseus; Centrifugal partition chromatography; pH-Zone-rening
counter-current chromatography; Alkaloids; Indole alkaloids

1. Introduction this category (Fig. 1). Industrial-scale production of

these molecules proceeds in two different ways. The
Catharanthus roseus G. Don (Apocynaceae) is a rst one follows the classical extraction and isolation
plant that produces indole alkaloids used in anti- scheme. This approach is hampered by the complexi-
cancer chemotherapy. The preparation and the puri- ty of crude extracts and by the low proportion of
cation of these compounds have been intensively valuable dimeric compounds. The other way resorts
studied [1]. Aerial parts of the plant contain between to the hemisynthesis or the biomimetic coupling
0.2 and 1% of a mixture of about 90 different reaction that uses monomers as starting material.
alkaloids. The most abundant ones are the monomers Both methods require the development of an ef-
like catharanthine and vindoline, bearing the indole cient, selective, and preparative chromatographic
and the indoline chromophore respectively (Fig. 1). technique able to isolate dimers and monomers from
The dimers resulting from the coupling of two such complex extraction mixtures. This article shows that
compounds present interesting pharmaceutical ac- this goal is achieved by means of Centrifugal
tivities. Anticancer drugs like leurocristine and vin- Partition Chromatography (CPC) when used in the
caleukoblastine (or vinblastine or VLB) belong to pH-zone rening mode [2].
CPC is a liquidliquid partition chromatographic
*Corresponding author. method generally referred to as counter-current

0021-9673/99/$ see front matter 1999 Published by Elsevier Science B.V. All rights reserved.
PII: S0021-9673( 99 )00495-1
422 J.-H. Renault et al. / J. Chromatogr. A 849 (2015) 421 431

stationary phase along with a base stronger than all UV/ Vis detector ISCO type V (Lincoln, NB, USA),

Fig. 1. Structures of interesting alkaloids found in the aerial parts of Catharanthus roseus.

chromatography (CCC) [3], even though there is, no problem due to saturation of the solid-state phase,
strictly speaking, no counter-current between the allowing CPC to be efcient in preparative sepa-
liquid phases. One is mobile and the other one is rations [5]. A high selectivity is obtained by a careful
kept stationary inside of the column by a constant choice of the biphasic solvent system, thus allowing
centrifugal eld. The column itself is a rotor built as the separation of compounds with very similar
a disk stack in which partition cells are engraved. structures [6].
The CPC technique has the ability to separate fragile Ito introduced the pH-zone-rening mode in CPC
compounds because there is no irreversible adsorp- as a variant of displacement chromatography [7]. It
tion or chemical reaction of the injected material is devoted to the purication of compounds whose
with a solid support [4]. Moreover, there is generally electric charge depends on pH value. For example, a

J.-H. Renault et al. / J. Chromatogr. A 849 (2015) 421 431 423

mixture of free bases is injected in the organic gradient pump. Detection was performed with an
4

the compounds to separate. This stronger base is set at 254 nm and a micro ow pH electrode
called the retainer base. The bases are moved along (Broadley James, Irvine, CA, USA) connected to a
the column by pumping through an acidic aqueous pH meter type PHM240 (Radiometer, Copenhagen,
solution of the displacer. Pure products are then Denmark). Fractions were collected with a collector
isolated in the efuent as salts. The column is used model Superfrac manufactured by Pharmacia (Upp-
in the descending mode, assuming that the density of sala, Sweden). Sample injections were carried out by
the aqueous mobile phase is greater than the density a Rheodyne valve type 7125 (Altech Associates,
of the organic stationary phase. In the ascending Deereld, IL, USA), through a 10-, 15- or 20-ml
mode, base salts and an acidic retainer introduced in loop.
the aqueous stationary phase can be displaced by a
base mixed to the mobile organic phase. Separation 2.2. Reagents
of organic acids or of their salts proceeds in a similar
way. Methyl tert.-butyl ether (MtBE, puro), acetonitrile
A characteristic of displacement chromatography (CH3CN, puro) chloroform (CHCl3, puro), methanol
is the liberation of the products from the column by (MeOH, puro) and hydrochloric acid (HCl 37%)
contiguous blocks arranged according their pKa were purchased from Carlo Erba, (Rodano, Italy);
values and also partition coefcients. In a block, the triethylamine (TEA) from Aldrich (SigmaAldrich,
concentration of the different species (and so the pH) Steinheim, Germany). The pure alkaloids (vindoline,
are constant and xed by the acidbase and reparti- catharanthine, vindolinine and vincaleukoblastine)
tion equilibrium constants. The overlap between came from our laboratory. The tartaric solution of a
blocks is minimum and does not vary with the crude alkaloid extract from Catharanthus roseus was
amount of injected sample. a kind gift from Omnichem (Louvain la Neuve,
The present study rst shows simulated and then Belgium).
real separations of a mixture of two monomers and a
dimer from C. roseus. The tuned protocol is then 2.3. Preparation of solvent phases and sample
applied to the separation of a crude alkaloid extract. solutions
The preparative aspect of CPC in pH-zone rening
mode is evaluated by increasing the mass of injected MtBE, CH3CN and water were thoroughly equili-
sample. The agreement between numerical simula- brated in suitable proportions (4:1:5, v / v) and the
tions and experimental chromatograms is discussed. two phases separated. The upper organic phase was
made basic with TEA at the concentrations of 8 mM
when it was used as the mobile phase (ascending
2. Experimental mode) and 10 mM when it was used as the stationary
phase (descending mode). The lower aqueous phase
2.1. CPC apparatus was acidied by HCl used as a retainer at a con-
centration of 10 mM (ascending mode) or used as a
The separations were performed using a HPCPC displacer at 8 mM (descending mode).
Sanki Series 1000 column (Tokyo, Japan). The In the case of the separation of a synthetic mixture
column is a stacked circular partition disk rotor of three pure alkaloids; vindoline, catharanthine and
which contains 2136 channels with a total internal vincaleukoblastine were dissolved in 10 ml of
volume of around 250 ml. The column is connected stationary phase in their salt or basic form following
to the injector and the detector through two high- which phase was chosen as stationary.
pressure rotary seals. A four-port valve, included In the case of the purication of the crude alkaloid
with the CPC, allows it to be operated in either the extract, the aqueous solution of tartrate alkaloids was
ascending or the descending mode. The HPCPC was lyophilized and dissolved in the mobile phase to be
connected to a Techlab (Erkerode, Germany) TIP50 injected.

424 J.-H. Renault et al. / J. Chromatogr. A 849 (2015) 421 431

2.4. Separation procedure 3. Results and discussion

The column was rst lled with the stationary The separation of alkaloids by means of pH-zone-
phase. Then, the rotation was brought to 800 rpm, rening CPC concerns, to date, only mixtures with a
the sample was injected and nally the mobile phase limited number of components [10]. The complexity
was pumped into the column in ascending or de- of the crude alkaloid extract from C. roseus merited
scending mode at a ow-rate of 3 ml / min, resulting some preliminary studies. An articial mixture of
in 3540 bar backpressure and 7585% of stationary two monomers, vindoline and catharanthine, and of
phase retention. The beginning and the end of the the dimer vincaleukoblastine was rst considered in
separation were checked by UV absorbance measure- computer simulations. Calculation requires the
ment at 254 nm and the evolution of the pH was knowledge of the thermodynamic constants of the
continuously monitored using an on-ow pH elec- partition and acidbase equilibria (Table 1). The
trode connected to a pH meter. former are determined by standard UV methods
[11,12] while the latter were found in literature data
2.5. Characterization of the alkaloids [13]. The monomers are potential dibases, while the
dimer is a potential tetrabase. However, the low
basicity of nitrogen atoms conjugated to the aromatic
All fractions were checked by TLC on Whatman
systems leads to the consideration that only the
K6F plates (Whatman, Maidstone, UK) developed
unconjugated tertiary amino sites are basic. The pKa
with CHCl3 MeOH (97:3). The fraction of interest
1 values reported in Table 1 concern these sites. The
13 2 volumes and species concentration required as pa-
recorded at 300 MHz on a a.c. 300 Bruker spec- rameters for the simulation are selected according to
the real separation conditions. In Fig. 2A, compound
trometer (Wissembourg, France).
amounts and pH are drawn as a function of the
number of collected fractions. The volume of each
2.6. Simulated separations fraction is equal to the volume of the mobile phase
within a single column section: 0.25 ml. Purity
Software was developed in order to simulate pH- proles are drawn in Fig. 2B. The purity is dened
zone-rening separations. The column is articially as the molar fraction of the currently leaving com-
divided into sections that mimic the theoretical pound in the efuent. The current compound is the
plates. Each section contains a fraction of the mobile most concentrated in the efuent. The result of the
and stationary phases. The chromatographic develop- simulation shows that a separation of the three
ment is decomposed into the repetition of two molecules of interest can be achieved with high
elementary processes: the equilibration of all the resolution. The steady-state concentration of each
chemical species present in each section and the compound should reect the number of protonation
pumping of the mobile phase. The latter stage sites. The dibase should be two times less concen-
requires the introduction of fresh mobile phase into trated than monobases because all the substances are
the rstisolated
and all section of the column,
compounds the simultaneous
were analyzed by H and
migration
C-NMR spectroscopy in C HClof
of all the mobile phases 3. each section
Spectra were Table 1
toward the next one, except for the last section Partition coefcients and pKa values of vindoline, catharanthine
whose content is analyzed to determine pH, con- and vinblastine
centration and purity proles. The most demanding KP5[AlcN]org / pKa value
part of the calculation is the equation resolution [AlcN]aq in 66% DMF
a

a
ships. The model proposed by Ito [8] is generalized
a

number of ionization sites of each compound [9]. method [12].

J.-H. Renault et al. / J. Chromatogr. A 849 (2015) 421 431 425

imposed by the partition and acidbase equilibria, Vindoline (l, nm) 37.4 (316 ) 5.5
constrained by charge and mass balance relation- Catharanthine (l, nm) 66.5 (284.6 ) 6.8
Vinblastine (l, nm) 17.0 (260 ) 7.4
to the separation of acids or bases, whatever the l used for partition coefcients calculation using the UV
Fig. 2. (A) Simulated fractogram of a mixture of vindoline, catharanthine and vincaleukoblastine plus retaining base (TEA) in the organic
stationary phase. The sample was in the rst theoretical plate. [TEA]510 mM, [HCl]58 mM, [vindoline]5200 mM, [catharanthine]5200
mM, [VLB]5100 mM (concentrations in the rst theoretical plate), number of theoretical plates5150, volume of stationary phase51.5 ml,
volume of aqueous mobile phase50.25 ml. (B) Purity and pH prole obtained for the same simulated separation.

426 J.-H. Renault et al. / J. Chromatogr. A 849 (2015) 421 431

isolated as salts of a displacer whose concentration is displacer (hydrochloric acid and triethylamine) previ-
constant. The calculated concentration of 0.5 is ously used by Ito (Table 2). The ascending and the
experimentally determined as being 0.85. This type descending modes were tested and led to approxi-
of discrepancy was already observed in real sepa- mately the same resolution. UV detection at 254 nm
rations involving the mixture of a diacid and mono- was completed with on-line pH monitoring to ob-
acids [14]. In the present case, the dibase is actually serve pH plates and thus the transitions between the
eluted as a mixture of the mono and the diprotonated different zones of elution. Fig. 3 presents the two
forms. Computer simulations show that the fully chromatograms and Table 3 the results. Only a few
diprotonated species is eluted only when both pKa fractions containing mixed products were collected,
values are identical. The monobasic behavior is and yields or recoverability are high (.80%).
enhanced by the increase of the difference between The graph in Fig. 4 shows a global similarity
the pKa values. between concentrations obtained by computer simu-
After these theoretical results, the separation of a lation (calculated from data in Fig. 2) and from the
purposely prepared mixture of vindoline, catharan- true separation. Nevertheless, it appears that the
thine and vinblastine was performed using the calculation overestimates concentrations of vindoline
biphasic solvent system methyl tert.-butyl ether and vincaleukoblastine. This may be due to the weak
acetonitrilewater 4:1:5 and classical retainer and dibasic character of vindoline and to the weak

Table 2
Experimental conditions for the separation of vindoline, catharanthine and vincaleukoblastine
Ascending mode Descending mode
Apparatus HPCPC Sanki Series 1000
Solvent system MtBECH3CNwater
Stationary phase (retainer) Aqueous (HCl 10 mM) Organic (TEA 10 mM)
Mobile phase (displacer) Aqueous (HCl 8 mM) Organic (TEA 8 mM)
Rotation (rpm) 800
Flow-rate (ml/min) 3
Back pressure (bar) 39 37
Detection UV (254 nm) and online pH detection
Fraction time (min) 1 1

Table 3
Separations results for the ascending and descending modes
Fractions Volume Alkaloid(s) Mass mMol Yield Concentration
(ml) (mg) (%/injected (mM)
products)
Ascending mode
6270 27 Vind. 96 0.21 95 7.77
71 3 Vind1Cathar. 8
7279 24 Cathar. 67 0.20 83.3 8.33
80 3 Cathar.1VLB 6
8186 18 VLB 89 0.11 84.6 6.11

Descending mode
5863 18 VLB 96 0.11 84.6 5.96
64 3 VLB.1Cathar. 7
6572 24 Cathar. 74.5 0.20 80 8.33
73 3 Cathar.1Vind. 9
7381 27 Vind. 98.5 0.20 91 7.4

J.-H. Renault et al. / J. Chromatogr. A 849 (2015) 421 431 427

Fig. 3. UV chromatogram and pH prole for the separation of vindoline, catharanthine and vincaleukoblastine. Sample: ascending mode:
110 mg (0.22 mmol) of vindoline chlorhydrate, 90 mg (0.24 mmol) of catharanthine chlorhydrate and 118 mg (0.13 mmol) of
vincaleukoblastine sulfate in 10 ml of aqueous stationary phase; descending mode: 100 mg (0.22 mmol) of vindoline, 84 mg (0.25 mmol) of
catharanthine and 105 mg (0.13 mmol) of vincaleukoblastine in 10 ml of organic stationary phase. For other experimental conditions, see
Table 2.

tribasic character of vincaleukoblastine. The extra The preliminary work shows that the desired
basic site would be located on the indoline nucleus, separation is workable and that the conclusions that
considering that indole itself is not basic enough to may be drawn from the results of computer simula-
be signicantly protonated. This hypothesis is con- tions make sense. In the second part of this work, the
rmed by the good agreement observed between real purication of a crude extract of C. roseus alkaloids
and computed steady-state concentration for (as salts of tartaric acid) was performed using the
catharanthine, a compound that bears two nitrogen previously described operating conditions. Four
atoms, a basic one, and another one which is part of separations were achieved with an increasing mass of
an indole nucleus. injected sample, in order to investigate the prepara-

428 J.-H. Renault et al. / J. Chromatogr. A 849 (2015) 421 431

Fig. 4. Comparison of the different alkaloid concentrations in the efuent between the simulation and the real separation in the ascending
mode.

Table 4
Experimental conditions for four separations (800 mg, 1.6 g, 2.4 g and 7 g) of tartrate alkaloids from Catharanthus roseus
Separation 1 Separation 2 Separation 3 Separation 4
Sample (injection volume) 800 mg (10 ml) 1.6 g (15 ml) 2.4 g (20 ml) 7 g (20 ml)
Apparatus HPCPC SANKI Series 1000
Elution mode Ascending
Biphasic solvents system MtBECH3CNwater (4:1:5)
Stationary phase (retainer) Aqueous (HCl 10 mM)
Mobile phase (displacer) Organic (TEA 8 mM)
Rotation (rpm) 800
Flow-rate (ml/min) 3
Back pressure 3740
Detection UV 254 nm/on line pH monitoring
Fraction time (min) 2
Fig. 5. UV chromatogram and pH prole for the separation with 2.4 g of tartrate alkaloids from Catharanthus roseus. For experimental
conditions, see Table 4.

J.-H. Renault et al. / J. Chromatogr. A 849 (2015) 421 431 429

Fig. 6. Results and yields of pure products for the four separations of tartrate alkaloids from Catharanthus roseus.
tive aspects of pH-zone rening. Experimental de- phase is easy and the alkaloids are recovered as free
tails are listed in Table 4. The ascending mode was bases.
selected, because the removal of an organic mobile In a rst experiment the injected sample mass was

430 J.-H. Renault et al. / J. Chromatogr. A 849 (2015) 421 431

Fig. 7. Simulated purity prole obtained for three increasing quantities of vindoline, catharanthine and vincaleukoblastine. A: [vindoline]5
100 mM, [catharanthine]5100 mM, [VLB]550 mM; B: [vindoline]5200 mM, [catharanthine]5200 mM, [VLB]5100 mM; [vindoline]5
300 mM, [catharanthine]5300 mM, [VLB]5150 mM. Other parameters are identical to Fig. 2.
 J.-H. Renault et al. / J. Chromatogr. A 849 (2015) 421 431 431

0.8 g. Only two pure products were isolated: matograms. It predicts the positive effect of mass
catharanthine and vindoline, the major monomer overloading on the purity of the separated products.
alkaloids produced by C. roseus. Upon increase of This effect is experimentally observed, and is of
the injected mass (1.6 g, 2.4 g, and 7.0 g), four prime importance for the development of pH-zone-
compounds were isolated: three monomers (vin- rening CPC as a preparative purication method.
doline, catharanthine, and vindolinine) and one dimer
(vincaleukoblastine). They were identied by co-tlc
1 13

those of reference compounds. The chromatogram Acknowledgements

and the elution proles for one of these separations is
presented in Fig. 5. We are grateful to Dr. K. Ple for linguistic advice
andYield evolution
comparison in separated
of their H and compounds
C NMR dataiswithre- and Dr. R. Margraff for interesting discussions and
ported in Fig. 6. It shows that the yields increase energetic encouragement. We thank the Region
with the amount of injected sample. This is a Champagne-Ardenne (France) for nancial support.
surprising situation in chromatography, in which
mass overloading has a positive effect. Even the
increase in injected volume, required for the solubili-
zation of the sample, has no noticeable impact on the References
efuent volume during transitions between the
[1] J. Sapi, G. Massiot, Heterocyclic Compounds, in: J.E. Saxton
(Ed.), Monoterpenoid Indole Alkaloids, Vol. 25, Part 4,
improvement brought by mass overloading was Wiley, Chichester, 1994, pp. 523646.
gained by computer simulations. Typical separation [2] Y. Ito, K. Shinomiya, H.M. Fales, A. Weisz, A.L. Sher, in:
conditions are used but with varying injected sample W.D. Conway, R.J. Petroski (Eds.), ACS Symposium Series,
steady-state periods. Some insight in the origin of the Vol. 593, American Chemical Society, Washington, DC,
1995, pp. 156183.
[3] Centrifugal Partition Chromatography, A.P. Foucault (Ed.),
the sample mass causes a lengthening of the elution Chromatographic Science Series, Vol. 68, Marcel Dekker,
time for each component of the mixture, without New York, 1994.
affectingAs
masses. thethetransition zones.ofTherefore,
concentration the efuentthe period
is driven [4] A.P. Foucault, L. Chevolot, J. Chromatogr. A 808 (1998) 3.
of time
by the during which pure
concentration compounds
of the arean
displacer, eluted are of
increase [5] J.-H. Renault, P. Thepenier, M. Zeches-Hanrot, L. Le Men-

Olivier, A. Durand, A. Foucault, R. Margraff, J. Chromatogr.

A 763 (1997) 345.
[6]
J.-H. Renault, K. Ghedira, `
P. Thepenier, C. Lavaud, M.
agreement
longer, with to
relative experimental
the overall data. The purity
separation en-
time. This `Zeches-Hanrot, L. Le Men-Olivier, Phytochemistry 44
hancement mayvisible
result is clearly also be understood
in the graphs inbyFig.
considering
7, in full (1997) 1321.
that each compound in the column behaves as a [7] A. Weisz, A.L. Sher, K. Shinomiya, H.M. Fales, Y. Ito, J.
Am. Chem. Soc. 116 (1994) 704.
[8] A.L. Sher, Y. Ito, Modern Countercurrent Chromatography,
in: W.D. Conway, R.J. Petroski (Eds.), ACS Symposium
the repartition
retainer of the
for the moleculesjust
compound in the column.
behind. AnFrom
increase Series, Vol. 593, American Chemical Society, Washington,
aofpragmatic point thus
sample mass of view, the observed
allows a better phenomenon
organization of DC, 1995, pp. 184202.
has highly interesting applications for the design of [9] ] J.-H. Renault, Thesis, University of Reims Champagne-
preparative and pilot-scale separations. Ardenne, 1997, pp. 174219.
[10] Y. Ma, Y. Ito, E. Sokolovsky, H.M. Fales, J. Chromatogr. A
 685 (1994) 259.
[11] Y. Shibusawa, Y. Hagiwara, Z. Chao, Y. Ma, Y. Ito, J.

4. Conclusion Chromatogr. A 759 (1997) 47.

[12] W.D. Conway (Ed.), Countercurrent Chromatography
Apparatus, Theory and Applications, VCH, Weinheim, New
York, 1990, pp. 302316.
[13] G.H. Svoboda, N. Neuss, M. Gorman, J. Am. Pharm. Assoc.,
of C. roseus alkaloids. A computer program de- Sci. Ed. 48 (1959) 659.

signed for the simulation of the chromatographic [14] M. Spraul, U. Brauma

nn, J.-H. Renault, P. Thepenier, J.-M.

process correctly interprets the experimental chro- Nuzillard, J. Chromatogr. A 766 (1997) 255.

This work presents a successful application of

pH-zone-rening CPC to the preparative purication