RESEARCH ARTICLE

Sensitive electrochemiluminescence (ECL)

immunoassays for detecting
lipoarabinomannan (LAM) and ESAT-6 in urine
and serum from tuberculosis patients
Tobias Broger ID1*, Michael Tsionksy2, Anu Mathew ID2, Todd L. Lowary3,
Abraham Pinter4, Tatiana Plisova2, Daniel Bartlett ID2, Simone Barbero2, Claudia
M. Denkinger1, Emmanuel Moreau1, Kiyonori Katsuragi5, Masanori Kawasaki ID5,
a1111111111 Payam Nahid6, George B. Sigal ID2*
a1111111111
1 FIND, Geneva, Switzerland, 2 Meso Scale Diagnostics, LLC., Rockville, Maryland, United States of
a1111111111 America, 3 Department of Chemistry and Alberta Glycomics Centre, University of Alberta, Edmonton,
a1111111111 Alberta, Canada, 4 Public Health Research Institute Center, New Jersey Medical School, Rutgers University,
a1111111111 Newark, New Jersey, United States of America, 5 Otsuka Pharmaceutical Co., Ltd., Tokyo, Japan,
6 University of California, San Francisco, California, United States of America

* [email protected] (GBS); [email protected] (TB)

OPEN ACCESS

Citation: Broger T, Tsionksy M, Mathew A, Lowary Abstract

TL, Pinter A, Plisova T, et al. (2019) Sensitive
electrochemiluminescence (ECL) immunoassays Background
for detecting lipoarabinomannan (LAM) and ESAT-
6 in urine and serum from tuberculosis patients.
Tuberculosis (TB) infection was responsible for an estimated 1.3 million deaths in 2017. Bet-
PLoS ONE 14(4): e0215443. https://doi.org/ ter diagnostic tools are urgently needed. We sought to determine whether accurate TB anti-
10.1371/journal.pone.0215443 gen detection in blood or urine has the potential to meet the WHO target product profiles for
Editor: Katalin Andrea Wilkinson, University of detection of active TB.
Cape Town, SOUTH AFRICA

Received: January 30, 2019 Materials and methods

Accepted: April 2, 2019 We developed Electrochemiluminescence (ECL) immunoassays for Lipoarabinomannan
(LAM) and ESAT-6 detection with detection limits in the pg/ml range and used them to com-
Published: April 18, 2019
pare the concentrations of the two antigens in the urine and serum of 81 HIV-negative and
Copyright: © 2019 Broger et al. This is an open
-positive individuals with presumptive TB enrolled across diverse geographic sites.
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and Results
reproduction in any medium, provided the original
LAM and ESAT-6 overall sensitivities in urine were 93% and 65% respectively. LAM and
author and source are credited.
ESAT-6 overall sensitivities in serum were 55% and 46% respectively. Overall specificity
Data Availability Statement: All relevant data are
was �97% in all assays. Sensitivities were higher in HIV-positive compared to HIV-negative
within the manuscript and its Supporting
Information files. patients for both antigens and both sample types, with signals roughly 10-fold higher on
average in urine than in serum. The two antigens showed similar concentration ranges
Funding: This work was funded by the Global
Health Innovative Technology (GHIT) Fund grant within the same sample type and correlated.
number G2015- 201, the UK Department for
International Development (DFID) grant number Conclusions
300341-102, the Dutch Ministry of Foreign Affairs
grant number PDP15CH14, the Bill and Melinda LAM and ESAT-6 can be detected in the urine and serum of TB patients, regardless of the
Gates Foundation grant number OPP1151258 and HIV status and further gains in clinical sensitivity may be achievable through assay and

PLOS ONE | https://doi.org/10.1371/journal.pone.0215443 April 18, 2019 1 / 19

 Sensitive tuberculosis LAM and ESAT-6 assays

National Institutes of Health and National Institute reagent optimization. Accuracy in urine was higher with current methods and has the poten-
of Allergy and Infectious Diseases (NIH/NIAID) tial to meet the WHO accuracy target if the findings can be transferred to a point-of-care TB
grant number 1R01AI104589. The authors include
employees of commercial organizations – GBS,
test.
MT, AM, TP, DB and SB are employees of Meso
Scale Diagnostics LLC (Rockville, USA) and MK
and KK are employed by Otsuka Pharmaceutical
Co., Ltd (Tokyo, Japan) – and received salaries
from their respective employers. The funders
provided support in the form of salaries and/or Introduction
research materials for authors but did not have any
additional role in the study design, data collection Tuberculosis (TB), an infection caused by Mycobacterium tuberculosis (Mtb), was responsible
and analysis, decision to publish, or preparation of for an estimated 1.3 million deaths in 2017 [1]. As most cases of TB can be safely and effectively
the manuscript. The specific roles of these authors treated after early diagnosis, the high observed mortality rate is associated with poor efficiency
are articulated in the ‘author contributions’ section. in identifying affected individuals and delays in treatment initiation [2]. Traditional methods
The corresponding author had full access to all the
like smear microscopy and nucleic acid tests like Xpert MTB/RIF (Xpert) rely on sputum as
data and had final responsibility for the decision to
submit for publication. the sample matrix, which leads to poorer test sensitivity in certain populations, like people liv-
ing with HIV and children, who are more likely to have disease that is extra-pulmonary or
Competing interests: TB, CMD and EM are
paucibacillary in nature and less likely to be able to produce sputum [3,4]. Data show that 20–
employed by FIND (Geneva, Switzerland), a
nonprofit organization that collaborates with 60% of HIV-positive patients presenting for TB diagnosis are unable to produce a sputum
industry partners. GBS, MT, AM, TP, DB, and SB sample [5,6].
are employed by Meso Scale Diagnostics LLC Diagnosing TB by detection of Mtb antigens in blood or urine provides an attractive alter-
(Rockville, USA) and received funding from FIND native approach. Antigen detection is also compatible with the use of simple and low-cost
and NIH. MK and KK are employed by Otsuka
immunoassay techniques, like rapid strip tests, that are better suited to the primary care setting
Pharmaceutical Co., Ltd (Tokyo, Japan). FIND,
MSD, and Otsuka provided support in the form of
in low and middle income countries. Most work to date on antigen-based assays for TB has
salaries for authors [TB, CMD, EM, GBS, MT, AM, focused on detection of lipoarabinomannan (LAM), a Mtb cell wall lipopolysaccharide. It is
TP, DB, SB, MK and KK], but did not have any well established that LAM can be found in the sputum [7–9] and urine [10–13] of some TB
additional role in the study design, data collection patients. There are also early-stage studies indicating that LAM can also be found in serum
and analysis, decision to publish, or preparation of [14–19]. Serological approaches based on the detection of serum antibodies to LAM have also
the manuscript. The specific roles of these authors
been investigated and have found anti-LAM antibodies in many patients, although the serolog-
are articulated in the ‘author contributions’ section.
TB and AP report patents in the field of LAM ical assays have not provided sufficient accuracy for clinical utility [20].
detection. The interests of authors do not alter the Commercial implementations of LAM tests have so far been limited to measuring LAM in
adherence to PLOS ONE policies on sharing data urine. A commercial lateral flow LAM (LF-LAM) test measuring urinary LAM to diagnose TB
and materials. (as detailed online in our guide for is available–the Alere Determine TB LAM Ag test from Abbott Diagnostics–but its adoption
authors http://journals.plos.org/plosone/s/
has been limited by its poor clinical sensitivity, which a recent meta-analysis estimated at 45%
competing-interests) All other authors declare no
competing interests.
with a 95% confidence range of 29% to 63% [21]. Clinical sensitivity is higher for HIV-positive
(HIV+) TB patients with low CD4 counts, so the Alere LF-LAM has been proposed as a com-
plement to the Xpert for immunocompromised HIV+ individuals, potentially compensating
for the poorer performance of Xpert in this population [22]. Recently, there has been encour-
aging evidence that the clinical sensitivity of urinary LAM for diagnosing TB in both HIV- and
HIV+ patients can be improved by using advanced assay methods with lower detection limits
[23–25] including a high sensitivity test suitable for use at the point-of-care—the SILVAMP
TB LAM Ag test—developed by Fujifilm [26].
Relative to LAM, considerably less attention has been given to assays for protein antigens of
Mtb. A 2011 review of antigen detection assays for TB lists a number of published reports dat-
ing back to the 1990s where research assays for undefined TB proteins were used to test a vari-
ety of clinical samples [27], but the lack of follow-up work makes it hard to judge the clinical
value of these assays. More recently a number of papers have provided evidence that proteins
secreted by Mtb, or peptides derived from these proteins, may be found in the serum of TB
patients when using new sensitive assay techniques. Specific Mtb proteins that have been mea-
sured in serum or plasma include ESAT-6 [28–33], CFP-10 [28–30,33], Ag85 complex [34–

PLOS ONE | https://doi.org/10.1371/journal.pone.0215443 April 18, 2019 2 / 19

 Sensitive tuberculosis LAM and ESAT-6 assays

36], Mpt32 and Mpt64 [36] and mshD [37]. Surprisingly, while the preponderance of LAM
assays studies use urine as the sample matrix, most studies of Mtb protein have focused on
blood-derived matrices. One study provided evidence that ESAT-6 can be found in the urine
of TB patients, but this study was limited to urine from four subjects [24]. Another study used
slow off-rate modified aptamers to assess the diagnostic potential of 11 antigens in serum and
urine but found limited diagnostic utility [38]. Finally, an ELISA developed for MoeX detec-
tion in urine detected the protein in 44% (11/25) of TB patients [39].
In an earlier study [25], we demonstrated that the application of sensitive electrochemilu-
minescence (ECL)-based assays and novel antibodies [8,40] to the measurement of LAM could
significantly improve the clinical performance of urinary LAM as a diagnostic biomarker for
TB relative to the Alere LF-LAM. In a small case-control study including HIV-positive and
HIV- negative individuals presenting at care centers with symptoms of TB, the sensitivity for
TB detection was 93% over all subjects and 80% when limited to HIV- subjects (in contrast the
analogous values for the Alere LF-LAM were 33% and 13%, respectively). In the work reported
here, we extend our earlier studies by applying the same approach and clinical cohort to (i) the
measurement of LAM in serum and (ii) the measurement of ESAT-6 in both urine and serum.
The work is designed to provide a preliminary assessment of the relative clinical utilities of
these biomarkers and sample matrices, as well as to provide data that can be used to identify
the analytical performance characteristics that will be required for the development of new
point-of-care diagnostics targeting these biomarkers.

Materials and methods

Antibodies and control materials
Purified LAM from Mtb strain Aoyama-B was obtained from Nacalai Tesque Inc. (Japan,
Product Number 02449–61). The Aoyama B strain has been commonly used in Mtb research
and has also been used in Japan for tuberculin production [41]. Recombinant ESAT-6 was
obtained from Alpha Diagnostics (San Antonio, TX, USA). Mtb culture filtrates were provided
by the Biodefense and Emerging Infections Research Resources Repository (BEI). The mono-
clonal antibodies used in the LAM assays have been previously described [8,25,40,42]. The
ESAT-6 assay used monoclonal antibodies from commercial sources (Bioporto, Hellerup,
Denmark and MSD, Rockville, MD, USA).

Immunoassays
Immunoassays for LAM employing a multiplexed sandwich immunoassay format and ECL
detection were carried out on commercial instrumentation and multi-well plate consumables
from Meso Scale Diagnostics, LLC. (MSD) [43]. The assays were run in MSD’s multiplexed
U-PLEX format. The U-PEX format employs multi-well plate consumables, in which each well
comprises a screen-printed carbon ink electrode supporting a generic 10-plex array of binding
reagents selected for their specificity to a set of 10 U-PLEX “linkers”. Capture antibodies are
coupled to these linkers (using reagents available from MSD) to generate reagents that are tar-
geted to specific elements of the generic arrays, enabling the solution phase self-assembly of
capture antibody arrays. Running assays in the U-PLEX format involves the steps of (i) incu-
bating a mixture of up to 10 capture antibody-linker conjugates in the U-PLEX wells to self-
assemble a capture antibody array and washing to remove excess unbound capture antibody;
(ii) adding the sample and a sample diluent comprising blocking components (including
blockers of human anti-mouse antibodies), incubating to bind analyte, and then washing to
remove unbound sample; (iii) adding labeled detection antibodies, incubating to bind cap-
tured analyte and form sandwich complexes, and then washing to remove unbound detection

PLOS ONE | https://doi.org/10.1371/journal.pone.0215443 April 18, 2019 3 / 19

 Sensitive tuberculosis LAM and ESAT-6 assays

antibody; and (iv) adding an ECL read buffer (MSD Read Buffer Gold) and analyzing the plate
on a MSD plate reader. The plate reader applies an electrical potential to the electrode in each
well and measures the resulting ECL emission from each array element on the electrode. The
LAM assay was run in the U-PLEX format using an array of anti-LAM capture antibodies rec-
ognizing a range of different LAM epitopes and a single anti-LAM detection antibody (A194-
01). The assay format and instrumentation allowed the assay signal from each capture anti-
body to be measured independently, although only the results from the two most sensitive cap-
ture antibodies (FIND 28 and S4-20) will be presented here. The assay procedure and
analytical performance are described in detail in a separate paper [25].
The Immunoassay for ESAT-6 was also conducted using MSD instrumentation and con-
sumables but was carried in a singleplex configuration using MSD’s ultra-sensitive S-PLEX
ECL format. The S-PLEX procedure was similar to that described for LAM above, except that
instead of using the U-PLEX plates, the capture antibody (Bioporto) was directly immobilized
on the carbon ink electrode. In addition, after completion of the detection antibody (MSD)
binding step, the S-PLEX format [44–49] includes additional signal enhancement steps using
proprietary enhancement reagents to increase assay signal and improve sensitivity.
Quantitation of the target analytes was based on 8-point calibration curves using purified
LAM (LAM assays) or Mtb culture filtrate (ESAT-6 assay) diluted into a serum-free calibration
diluent (2% bovine serum albumin in phosphate buffered saline). Each calibrator concentra-
tion was run in duplicate in each assay plate. ESAT-6 concentrations in the Mtb culture filtrate
were assigned based on comparison to a recombinant ESAT-6 reference material prepared in
E.coli. The relationship of ECL signal to analyte concentration for the calibration curves was fit
to a 4-parameter logistic (4-PL) function. Analyte concentrations for test samples were calcu-
lated by back-fitting ECL signals to the 4-PL fit and were corrected for sample dilution.

Sample preparation
To inactivate any host antibodies in urine and serum samples that could interfere with the
assay measurement, the samples were heat-inactivated prior to analysis with the LAM or
ESAT-6 assays. Serum samples were diluted 1:4 in PBS prior to heat-inactivation to avoid the
formation of clots; urine was heat-inactivated without dilution. For heat-inactivation, both,
urine and serum samples were heated to 85˚C for 10 minutes or 95˚C for 5 minutes.

Clinical samples
For this retrospective case-control study, a total of 75 urine samples and 75 serum samples
were selected from FIND’s biobank. For all but six of the urine samples, matching serum sam-
ples were available that were collected from the same study subjects on the same day. For the
remaining six urine samples (all of which were from TB- HIV+ subjects), serum samples were
identified from six additional TB- HIV+ subjects and these were not included in the compari-
sons that required matched samples. The samples were collected in studies from adults pre-
senting at primary care sites in Bangladesh, Peru, South Africa and Vietnam with clinical
symptoms of TB, but not receiving TB treatment at the time of sample collection. Approval by
local Ethics Committees and written informed patient consent was obtained before enrolling
patients and no personally identifiable information was available to FIND or to the research-
ers. For use in patient classification, sputum samples (typically two in the first 24 h) were col-
lected from all participants, decontaminated, and tested in up to six independent liquid
cultures (MGIT; BD, Franklin Lakes, NJ, USA) and solid cultures (Lowenstein-Jensen
medium). The presence of the M. tuberculosis complex in cultures was confirmed by Ziehl-
Neelsen staining or auramine O fluorescence microscopy to identify acid-fast bacilli, MPT64

PLOS ONE | https://doi.org/10.1371/journal.pone.0215443 April 18, 2019 4 / 19

 Sensitive tuberculosis LAM and ESAT-6 assays

antigen detection using rapid speciation assays (such as the Capilia TB test; TAUNS, Japan), or
molecular methods. TB-positive individuals were patients with at least one positive culture. All
TB-positive patients had positive microscopy results. Participants who were smear negative
and culture negative on cultures from all sputum samples and who exhibited symptom resolu-
tion in the absence of tuberculosis treatment and negative sputum culture results at a 2-month
follow-up visit were classified as TB negative. The subjects were further classified as HIV posi-
tive or HIV negative on the basis of HIV rapid tests. More details on sample collection and
classification of subjects can be found in the description of our previous urinary LAM study
[25]. LAM assay results for the urine samples (using the ECL assay and Alere LF-LAM) were
available from this previous study. New aliquots of the same urine samples were used for uri-
nary ESAT-6 measurements.

Data analysis and statistical methods

Data analysis was conducted using the R statistical programming language (version 3.5.1).
Confidence intervals for proportions were calculated as the Wilson interval using the “bin-
conf” function from the Hmisc package in R. Comparison of groups was carried out by the
Mann-Whitney test using the “wilcox.test” function in R. Correlation constants and p values
for the null hypothesis of no correlation were calculated using Pearson’s method and the “cor.
test” function in R. Correlations of concentration values were determined using log10 trans-
formed concentrations and only included data points where both methods being compared
had detectable concentrations. The Cohen’s kappa statistic for categorical agreement was cal-
culated using the “Kappa” function from the vcd package in R. All relevant data are within the
manuscript and its Supporting Information files.

Results
Analytical assay performance
The analytical sensitivity of the LAM and ESAT-6 ECL immunoassays was established by test-
ing the LAM and ESAT-6 calibration standards diluted in a serum-free calibration buffer. Fig
1 shows 8 point calibration curves plotting the measured LAM and ESAT-6 assay signals as a
function of the concentration of the calibration standards for each assay. For the LAM assay,
signals are shown for both the FIND 28 and S4-20 capture antibodies. Based on the Coeffi-
cients of variation (CVs) of the blanks and targeting detection thresholds at least 2.5 standard
deviations above the blank, we selected a threshold signal-to-blank (S/B) ratio of 1.375 for the
LAM assays and 1.425 for the ESAT-6 assay. Average CVs for calibrators above the selected
threshold ranged between 3 to 5% for the LAM assays and 9% for the ESAT-6 assay. The calcu-
lated analyte concentrations at the selected detection thresholds resulted in limits of detection
(LODs) of 6 pg/mL for the ESAT-6 assay, 6 pg/mL for the LAM assay using the FIND 28 cap-
ture antibody and 11 pg/mL for the LAM assay using the S4-20 capture antibody (S1 Table).
Optimization of sample preparation for urine and serum samples. We examined
whether optimal assay performance required sample preparation steps to inactivate proteins in
urine or serum that could bind the antigen targets and interfere with antibody recognition. In
our previous study of LAM in urine, we found no evidence of inhibitory proteins in urine, but
we also found no negative impact on measured LAM concentrations if we heated urine sam-
ples under conditions that inactivate host antibodies (85˚C for 10 min.). In the current study,
we investigated the effect of heat inactivation on detection of LAM in serum and ESAT-6 in
serum and urine. S2a Table shows the recovery of LAM spiked at ~3 ng/mL into normal urine
or serum samples relative to LAM spiked into a simple buffer (the assay calibrator diluent).
The table also shows the improvement in recovery (if any) if the sample was heat inactivated

PLOS ONE | https://doi.org/10.1371/journal.pone.0215443 April 18, 2019 5 / 19

 Sensitive tuberculosis LAM and ESAT-6 assays

Fig 1. Calibration curves (ECL signal as a function of the concentration) for the (a) LAM and (b) ESAT-6 assays. Each panel also provides the
measured signal for a blank sample containing no analyte. Solid lines show the 4PL fit to the data. The horizontal and vertical dashed lines represent,
respectively, the lowest detectable signals—based on the threshold signal to background (S/B) value as described in the text—and the limit of detections
(LOD) calculated from the lowest detectable signal by back-fitting to the 4PL curve.
https://doi.org/10.1371/journal.pone.0215443.g001

prior to conducting the immunoassay. The measured recoveries in urine in the absence of heat
treatment were 63–92% and sometimes higher than after heat treatment confirming our prior
observation that matrix inhibition in the urine matrix did not appear to be a significant issue.
In contrast, the LAM assays were strongly inhibited by the serum matrix. Heat inactivation of
the spiked serum samples largely reversed the matrix inhibition for signals measured with the
FIND 28 capture and brought the measured recoveries to 69–100%. Heat inactivation also
improved recognition by the S4-20 capture, but to a lesser extent. The effect of heat inactiva-
tion was also tested on a small set of urine and serum samples from TB+ individuals (S2b
Table). Heat inactivation had minimal effects on the assay signals from urine samples but pro-
duced large increases for serum samples. The effect was strongest with the FIND 28 capture,
with the signal from one serum sample increasing 45-fold with heat treatment.
In contrast to the LAM assay, significant inhibition of the ESAT-6 assay was not observed
when Mtb cell culture filtrate containing native ESAT-6 was spiked into either urine or serum
to provide a final ESAT-6 concentration of 1 ng/mL (S3a Table). Heat treatment of the spiked
samples increased the recoveries to concentrations two to five-fold above the expected values.
This increase occurred not only in spiked urine and serum, but also for spiked diluent when
compared to an unspiked control. This result suggests that the signal increase is not due to
inactivation of interfering species, and is more likely an effect of heating on the ESAT-6 struc-
ture likely leading to exposure of epitopes by conformation change such as protein unfolding
or by breaking up the association of ESAT-6 with another protein. An increase in signal was
also observed after heating of unspiked serum and urine samples from TB+ subjects (S3b
Table), although the magnitude of the increase was smaller (1.2-fold to 2-fold). Given the
observed stability of ESAT-6 to the heat treatment step, and to protect against the possibility
that some fraction of serum samples could contain inhibitory host antibodies, the heat pre-
treatment step was made part of the assay protocol for testing clinical samples.

PLOS ONE | https://doi.org/10.1371/journal.pone.0215443 April 18, 2019 6 / 19

 Sensitive tuberculosis LAM and ESAT-6 assays

LAM and ESAT-6 concentrations in clinical samples

Table 1 provides the characteristics of the study population. We selected the samples from
FIND’s repository of TB clinical samples to include a range of geographical locations (Asia,
Africa and South America), and to cover the different combinations of TB and HIV status.
CD4 counts were available for most of the TB+/HIV+ subjects and included subjects above (8
subjects) and below (14 subjects) the 100 cells/uL threshold used in the WHO recommenda-
tion for the Alere LF-LAM [22]. Matched serum and urine was available for 69 of the 81 sub-
jects in the study; matched samples were not available for 6 TB-/HIV+ subjects and samples
from other TB-/HIV+ subjects was tested to bring the number of serum samples up to 15 in
this negative group. Alere LF-LAM was run on all the urine samples and the results were
reported with our previous study [25]. The sensitivity of the Alere LF-LAM for the panel of
urine samples was 44% (11/25) for HIV+ subjects, but only 13% (2/15) for HIV- subjects, sug-
gesting that the patient population that was tested behaved similarly to the populations
described in the Cochrane meta-analysis by Shah et al. [21].
Fig 2 shows the measured assay signals and measured analyte concentrations for the two
LAM assays and the ESAT-6 assay. The results are presented for each of the two sample

Table 1. Characteristics of the study population broken down by TB and HIV status. NA indicates that information for the specified characteristic was not available
for a study subject. CD4 cell counts were only available for TB/HIV+ subjects. Alere LF-LAM results were only available for subjects with urine samples. Matched urine
and serum samples were available for 69 of the 81 subjects. The remaining 12 subjects consisted of 6 TB+/HIV- subjects with only urine samples and 6 TB+/HIV- subjects
with only serum samples.
All Subjects TB Negative TB Positive
HIV Negative HIV Positive HIV Negative HIV Positive
Category No. of Subjects (% of Total)
All subjects 81 (100%) 20 (25%) 21 (26%) 15 (19%) 25 (31%)
Gender
Female 23 (28%) 6 (7%) 5 (6%) 5 (6%) 7 (9%)
Male 53 (65%) 9 (11%) 16 (20%) 10 (12%) 18 (22%)
NA 5 (6%) 5 (6%) 0 (0%) 0 (0%) 0 (0%)
Age
18 to 40 51 (63%) 6 (8%) 14 (17%) 12 (15%) 19 (23%)
41 to 60 25 (31%) 13 (16%) 6 (7%) 2 (2%) 4 (5%)
61+ 2 (2%) 1 (1%) 0 (0%) 1 (1%) 0 (0%)
NA 3 (4%) 0 (0%) 1 (1%) 0 (0%) 2 (2%)
Location
Bangladesh 5 (6%) 5 (6%) 0 (0%) 0 (0%) 0 (0%)
Peru 19 (23%) 3 (4%) 14 (17%) 2 (2%) 0 (0%)
South Africa 20 (25%) 2 (3%) 5 (6%) 5 (6%) 8 (10%)
Vietnam 37 (46%) 10 (12%) 2 (2%) 8 (10%) 17 (21%)
CD4 Count
<= 100 cells/μL 14 (17%) 0 (0%) 0 (0%) 0 (0%) 14 (17%)
> 100 cells/μL 8 (10%) 0 (0%) 0 (0%) 0 (0%) 8 (10%)
NA 59 (73%) 20 (25%) 21 (26%) 15 (19%) 3 (4%)
Alere
Negative 62 (77%) 20 (25%) 15 (19%) 13 (16%) 14 (17%)
Positive 13 (16%) 0 (0%) 0 (0%) 2 (2%) 11 (14%)
NA 6 (7%) 0 (0%) 6 (7%) 0 (0%) 0 (0%)
Serum and Urine
Matched 69 (85%) 20 (25%) 9 (11%) 15 (19%) 25 (31%)
https://doi.org/10.1371/journal.pone.0215443.t001

PLOS ONE | https://doi.org/10.1371/journal.pone.0215443 April 18, 2019 7 / 19

 Sensitive tuberculosis LAM and ESAT-6 assays

Fig 2. LAM and ESAT-6 concentrations in urine (left column) and serum samples (right column). The plots show LAM (panels a-d) and
ESAT-6 (panels e-f) assay signals (normalized to the blank signal) for a set of urine (panels a, c and e) and serum samples (panels b, d and f)
as a function of the TB and HIV status of the donor. LAM assay results are provided for the two most sensitive anti-LAM capture antibodies:
FIND 28 (panels a-b) and S4-20 (panels c-d). The gray horizontal shows the blank signal (S/B = 1). The dashed orange line shows the assay
threshold (S/B = 1.375 for the LAM assays, S/B = 1.425 for the ESAT-6 assay) which is equal to the limit of detection (LOD). The second y
axis provides estimated analyte concentrations for concentrations above the LOD. The points are colored by the Alere LF-LAM results for
urine from the same subjects (0 = LF-LAM negative, 1 to 4 represents increasing LF-LAM test color intensity as determined by comparison
to the Alere LF-LAM reference card).
https://doi.org/10.1371/journal.pone.0215443.g002

PLOS ONE | https://doi.org/10.1371/journal.pone.0215443 April 18, 2019 8 / 19

 Sensitive tuberculosis LAM and ESAT-6 assays

matrices as a function of TB status and HIV status and are color coded based on urinary Alere
LF-LAM test grade. As presented in a previous report [25], the data for LAM concentrations
in urine of TB+ subjects ranged over many orders of magnitude, from below the detection lim-
its for the assays (~10 pg/mL) to concentrations greater than 100 ng/mL, with higher concen-
trations generally being observed for the HIV+ subjects relative to the HIV- subjects. The
improved detection limit of the assays led to the detection of LAM in many Alere LF-LAM
negative, TB+ urine samples. While the measured signals and LAM concentrations for the
assay using FIND 28 as the capture antibody tended to give higher signals than the assay using
S4-20 as the capture antibody, the FIND 28 capture also generated a number of false positive
results, presumably because of its lower specificity for LAM from Mtb [25].
Comparing the previous urinary LAM measurements from our previous work [25] to new
results using the two LAM assays to measure LAM in serum, Fig 2 shows that LAM was mea-
surable in serum from at least some TB+ subjects, although the measured concentrations
tended to be lower than those in urine. Using the FIND28 capture, the highest concentrations
measured in urine and serum were 312 ng/mL and 32 ng/mL, respectively. Similarly to the
urine results, concentrations of LAM in serum levels measured with the FIND 28 capture
tended to be higher than with the S4-20 capture. The highest concentrations levels measured
with FIND 28 and S4-20 were 32 ng/mL and 1.5 ng/mL, respectively. However, in contrast to
the urine results, use of the FIND 28 capture did not lead to elevated signals and false positive
results for TB- subjects.
Fig 2 also shows that ESAT-6 is directly measurable in urine or serum in several TB+ sub-
jects. Qualitatively, the behavior of ESAT-6 appears similar to LAM, with ESAT-6 concentra-
tions covering roughly the same range as LAM. As observed for LAM, ESAT-6 concentrations
tend to be higher in urine vs. serum, and in samples from HIV+ subjects vs. HIV- subjects.
Fig 3 shows the correlations between the measured serum and urine concentrations for the
subjects with matched serum and urine samples. For the ESAT-6 (Fig 3c) and the FIND 28
LAM assays (Fig 3a), the urine and serum concentrations generally correlated with correlation
coefficients of 0.76 (p = 0.0004) and 0.66 (p = 0.001), respectively. Categorical agreement
between urine and serum results was 84% and 72% for the ESAT-6 and LAM assays respec-
tively (S4 Table). The concentrations in serum tended to be lower than in urine; the highest
ESAT-6 concentrations in urine and serum were 45 ng/mL and 12 ng/ml.

Clinical assay performance

Table 2 provides the measured clinical sensitivity and specificity across the test sample set for
each of the assays, using each of the two sample matrices. The LAM assays using the S4-20 and
FIND 28 capture antibodies displayed different relative performance in urine vs. serum. The
assay using the S4-20 capture antibody provided better performance for urine samples, pri-
marily due to better sensitivity for the TB+HIV- subjects (80% for S4-20 vs. 40% for FIND 28)
and better specificity across all TB- subjects (97% for S4-20 vs. 66% for FIND 28). In contrast,
the assay using the FIND 28 capture provided better overall sensitivity for detecting LAM in
serum samples from the TB+HIV+ subjects (76% for FIND 28 vs. 48% for S4-20) and TB
+HIV- subjects (20% for FIND 28 vs. 0% for S4-20). In addition, the poor specificity of the
FIND 28 capture antibody for urine assays was not observed when testing serum samples:
both the S4-20 and FIND 28 capture antibodies provided perfect (100%) specificities for test-
ing serum samples regardless of HIV status. As might be expected, given the lowered measured
LAM concentrations in serum vs. urine (Figs 2 and 3), even when selecting the optimal capture
antibody for the sample matrix, the observed clinical sensitivities for measuring LAM in
serum were lower than for LAM in urine.

PLOS ONE | https://doi.org/10.1371/journal.pone.0215443 April 18, 2019 9 / 19

 Sensitive tuberculosis LAM and ESAT-6 assays

Fig 3. Correlation of LAM and ESAT-6 concentrations in urine and serum. The plots show the results for LAM measurements
using the FIND 28 (panel a) and S4-20 (panel b) capture antibodies, and the results for the ESAT-6 assay (panel c). Concentrations
in serum are corrected for the additional 4-fold dilution of serum samples relative to urine samples. The dashed orange lines show
the dilution-corrected detection limits for the assays. Data points with signals below the detection thresholds (categorized as “ND” or
not detectable) are shown for completeness, but do not have assigned concentrations. As indicated with the orange numbering in
panel a, the plots can be divided into four regions: (1) concentrations measurable in serum and urine; (2) concentrations only
measurable in serum; (3) concentrations only measurable in urine and (4) concentrations not measurable in either matrix.
Percentages of categorical agreement are shown in S4 Table. Data points are colored based on TB status and filled based on HIV
status as indicated in the key.
https://doi.org/10.1371/journal.pone.0215443.g003

The observed clinical sensitivity for the ESAT-6 assay using urine samples (76% for TB
+HIV+, 47% for TB+HIV-) was considerably lower than that of the LAM assay using the S4-
20 capture antibody (100% for TB+HIV+, 80% for TB+HIV-). The clinical sensitivity for mea-
suring ESAT-6 in serum (63% for TB+HIV+, 20% for TB+HIV-) was lower than that observed
using urine samples, but was comparable to the sensitivity of the best serum LAM assay using
the FIND 28 capture antibody (76% for TB+HIV+, 20% for TB+HIV-). Good specificity was

PLOS ONE | https://doi.org/10.1371/journal.pone.0215443 April 18, 2019 10 / 19

 Sensitive tuberculosis LAM and ESAT-6 assays

Table 2. Accuracy of LAM and ESAT-6 assays for measurements in urine and serum. LAM performance is provided for the FIND 28 and S4-20 capture antibodies.
The table provides the measured sensitivity (correctly classified TB+ samples / total number of TB+ samples) and specificity (correctly classified TB- samples / total number
of TB- samples). The values were calculated for the full sample set (All) or for the subsets of samples from HIV- and HIV+ subjects. 95% confidence intervals for the pro-
portion were calculated using Wilson’s method.
Matrix HIV Status Assay Sensitivity Specificity
Correct / Total % (95% CI) Correct / Total % (95% CI)
Urine All LAM (FIND 28) 30 / 40 75% (60%-86%) 23 / 35 66% (49%-79%)
All LAM (S4-20) 37 / 40 93% (80%-97%) 34 / 35 97% (85%-100%)
All ESAT-6 26 / 40 65% (50%-78%) 34 / 35 97% (85%-100%)
Neg LAM (FIND 28) 6 / 15 40% (20%-64%) 14 / 20 70% (48%-85%)
Neg LAM (S4-20) 12 / 15 80% (55%-93%) 20 / 20 100% (84%-100%)
Neg ESAT-6 7 / 15 47% (25%-70%) 19 / 20 95% (76%-100%)
Pos LAM (FIND 28) 24 / 25 96% (80%-100%) 9 / 15 60% (36%-80%)
Pos LAM (S4-20) 25 / 25 100% (87%-100%) 14 / 15 93% (70%-100%)
Pos ESAT-6 19 / 25 76% (57%-89%) 15 / 15 100% (80%-100%)
Serum All LAM (FIND 28) 22 / 40 55% (40%-69%) 35 / 35 100% (90%-100%)
All LAM (S4-20) 12 / 40 30% (18%-45%) 35 / 35 100% (90%-100%)
All ESAT-6 18 / 39 46% (32%-61%) 35 / 35 100% (90%-100%)
Neg LAM (FIND 28) 3 / 15 20% (7%-45%) 20 / 20 100% (84%-100%)
Neg LAM (S4-20) 0 / 15 0% (0%-20%) 20 / 20 100% (84%-100%)
Neg ESAT-6 3 / 15 20% (7%-45%) 20 / 20 100% (84%-100%)
Pos LAM (FIND 28) 19 / 25 76% (57%-89%) 15 / 15 100% (80%-100%)
Pos LAM (S4-20) 12 / 25 48% (30%-67%) 15 / 15 100% (80%-100%)
Pos ESAT-6 15 / 24 63% (43%-79%) 15 / 15 100% (80%-100%)
https://doi.org/10.1371/journal.pone.0215443.t002

observed for the ESAT-6 assay in both urine (97%) and serum (100%) when calculated across
all subjects regardless of HIV status.

Comparing and combining LAM and ESAT-6

Fig 4 shows the correlation of LAM and ESAT-6 concentrations in urine and serum. The plot-
ted LAM concentrations were determined using the LAM capture antibody with the highest
diagnostic accuracy for the matrix: S4-20 for urine and FIND 28 for serum. The correlation
coefficients were 0.70 (p < 0.0001) and 0.83 (p < 0.0001) for urine and serum, respectively.
The categorical agreement of LAM and ESAT-6 in urine was 80% (S5 Table), the correlation
plot for urine samples shows that when ESAT-6 was measurable in urine it tended to track
with LAM, and the highest ESAT-6 concentrations were associated with high LAM concentra-
tions. The categorical agreement of LAM and ESAT-6 in serum was 82% (S5 Table). We
assessed if sensitivity could be improved by running both assays and using a positive result on
either assay to judge a sample as positive (Table 3). Relative to the LAM assay by itself, includ-
ing the result of the ESAT-6 assay raised the serum sensitivity from 76% to 88% for the TB
+HIV+ group and from 20% to 33% for the TB+HIV- group, with no loss in specificity.
Effect of immunosuppression on LAM and ESAT-6 concentrations. We previously
reported that the distribution of LAM concentrations in the urine of immunocompetent TB
+/HIV+ subjects was not significantly different from the LAM concentrations of TB+/HIV-
subjects but that LAM concentrations were higher in immunosuppressed subjects with CD4
cell counts below 100 cells/uL [25]. We extended this analysis to LAM concentrations in
serum and ESAT-6 concentrations in urine and serum (Fig 5). Similar behavior was observed
for both bacterial antigens in both matrices. In all cases, the antigen concentrations measured

PLOS ONE | https://doi.org/10.1371/journal.pone.0215443 April 18, 2019 11 / 19

 Sensitive tuberculosis LAM and ESAT-6 assays

Fig 4. Correlation of LAM and ESAT-6 concentrations. The two plots compare the LAM and ESAT-6 concentrations in urine
(panel a) and serum (panel b). The plots display the LAM concentration measured with the most sensitive LAM capture antibody for
each matrix: S4-20 for urine and FIND 28 for serum. Concentrations in serum are corrected for the additional 4-fold dilution of
serum samples relative to urine samples. The dashed orange lines show the dilution-corrected detection limits for the assays. Data
points with signals below the detection thresholds (categorized as “ND” or not detectable) are shown for completeness, but do not
have assigned concentrations (see, the discussion in the Fig 3 legend). Percentages of categorical agreement are shown in S5 Table.
Data points are colored based on TB status and filled based on HIV status as indicated in the key.
https://doi.org/10.1371/journal.pone.0215443.g004

in TB+/HIV- subjects and TB+/HIV+ subjects with CD4 counts greater than 100 cells/uL were
not statistically distinguishable, while the distribution of concentrations in TB+/HIV+ subjects
with low CD4 counts was shifted significantly to higher values.

Discussion
In this study, we have shown that LAM and ESAT-6 can be detected in both urine and serum
samples of TB patients, using newly developed sensitive ECL immunoassays. The two antigens
assayed showed similar concentration ranges within the same sample type. The concentrations
were on average roughly 10-fold higher in urine than in serum, but correlated. The measured
concentrations of LAM in urine for HIV- patients (from undetectable to about 1,000 pg/mL
using the S4-20 capture antibody) were consistent with values reported for this population
using a technique involving pre-concentration of LAM from urine using nanocages followed
by quantitation by Western blot [24]. It is further known that the Alere LF-LAM threshold is
in the range of 1,000 pg/ml which is also consistent with our results.

Table 3. Accuracy of combining ESAT-6 result and LAM result. A sample was characterized as positive if either the LAM or ESAT-6 assays gave signals above their
respective thresholds. Sensitivity, specificity and confidence limits were determined as described in the Table 2 legend.
Matrix HIV Status Assay Sensitivity Specificity
Correct / Total % (95% CI) Correct / Total % (95% CI)
Serum All LAM (FIND 28) + ESAT-6 26 / 39 67% (51%-79%) 35 / 35 100% (90%-100%)
Neg LAM (FIND 28) + ESAT-6 5 / 15 33% (15%-58%) 20 / 20 100% (84%-100%)
Pos LAM (FIND 28) + ESAT-6 21 / 24 88% (69%-96%) 15 / 15 100% (80%-100%)
https://doi.org/10.1371/journal.pone.0215443.t003

PLOS ONE | https://doi.org/10.1371/journal.pone.0215443 April 18, 2019 12 / 19

 Sensitive tuberculosis LAM and ESAT-6 assays

Fig 5. Association of LAM and ESAT-6 concentrations with HIV status and CD4 counts. The plots show the assay signals (left axis) and estimated
concentrations (right axis) for (a) LAM measured in urine using the S4-20 capture antibody, (b) LAM measured in serum using the FIND 28 capture
antibody, (c-d) ESAT-6 measured in urine and serum. The plots are formatted as described for Fig 2, except only the TB+ subjects are shown, points are
colored by HIV status, and the TB+/HIV+ subjects are separated into two groups based on CD4 cell count (> 100 cells/uL and � 100 cells/uL). An “� ”
above the data points for a TB+HIV+ group indicates the distribution is significantly different than the distribution of points for the TB+HIV- subjects
in the same plot (Mann-Whitney test, p < 0.05).
https://doi.org/10.1371/journal.pone.0215443.g005

The measured concentrations of ESAT-6 in serum (0 to about 100 pg/mL for HIV- patients
and 0 to about 10,000 pg/mL for HIV+ patients), however, were considerably lower than the
concentrations reported by a group using a technique involving trypsin digest of serum sam-
ples, concentration of ESAT-6 associated peptides using nanodisks and quantitation by mass
spectrometry [30], who reported average concentrations of about 11,000 pg/mL (1.2 nM) in
HIV negative TB patients. The reason for this discrepancy is not clear. Such high antigen con-
centrations are possible in blood-borne diseases; for example, the median concentration of his-
tidine-rich protein II (HRP2) for patients with 1000 to 10,000 malaria parasites/μl blood is
reported to be about about 12,000 pg/mL (0.36 nM) [50]. The high concentrations reported by
Liu and colleagues seem unlikely for TB patients with pulmonary disease. On the other hand,
it is possible that the mass spectrometry technique is picking up ESAT-6 fragments that are
not recognized by antibody targeting native ESAT-6. It is also possible that the differences
reflect differences in how the assays were calibrated. Standardized control materials and sam-
ple panels like those recommended for other disease antigens by the WHO Expert Committee
on Biological Standardization [51] should also be developed for TB and would help to resolve
these discrepancies.
The correlation between LAM and ESAT-6 suggest that the two antigens enter urine
and blood by similar mechanisms in a large fraction of TB patients. Renal infection has been
proposed as a mechanism leading to LAM antigenuria [52]. Renal TB is more frequent in
immunosuppressed HIV+ and could explain the higher LAM concentrations and higher per-
formance of the Alere LF-LAM in patients with CD4 counts [52]. The authors of this report
also argued that the molecular weight of LAM, i.e. when caught in immune-complexes would
be too large to pass an intact glomerular basement membrane. We did observe higher concen-
trations of both LAM and ESAT-6 in the urine of immunosuppressed (CD4 cell counts � 100
cells/uL) TB+HIV+ subjects, relative to TB+HIV- subjects and TB+HIV+ subjects with higher
CD4 counts, and renal involvement could play a role for some of those patients. However, by

PLOS ONE | https://doi.org/10.1371/journal.pone.0215443 April 18, 2019 13 / 19

 Sensitive tuberculosis LAM and ESAT-6 assays

using an assay with about 100-times improved detection limit relative to the Alere LF-LAM,
we were able to detect LAM in 80% and ESAT-6 in 47% of urine samples from TB+HIV- sub-
jects (Table 2). These results suggest that the clearance of LAM (or LAM fragments) and
ESAT-6 out of blood by the kidneys might play a more important role than originally pro-
posed. Further support for a mechanisms beyond renal infection alone is the finding that both
LAM and ESAT-6 can also be detected in serum, and that a large proportion of patients (54%
(21/39) for LAM and 63% (17/27) for ESAT-6, S4 Table) with detectable urinary antigen con-
centrations also have detectable blood antigen concentrations. A limitation of our study was
the lack of mycobacterial blood culture and urinary Xpert results, which could have been used
to determine if the presence of antigen in blood or urine was associated with blood infections
or renal TB.
Binding of LAM to other antibodies and other proteins in clinical sample matrices can
potentially interfere with its detection [17]. We observed only minimal urine matrix interfer-
ence of our LAM assays when measuring LAM spiked into urine and pre-treatment to disrupt
potential sequestration of LAM in the urine had little effect on assay signals. In marked con-
trast, the measurement of LAM spike into serum was almost completely inhibited as has been
observed in other LAM studies [15]. We found, however, that this inhibition was almost
completely reversed by a simple heat inactivation step using conditions similar to those that
have been used to inactivate antibodies in other antigen detection tests [53], and that treatment
with acids or proteases pretreatment was not required. We noted that a highly Mtb specific
capture antibody (S4-20) provided the best clinical performance in urine but had relatively
poor sensitivity in serum relative to a more generic anti-LAM antibody (FIND 28). The S4-20
antibody targets a unique capping epitope (MTX-Man, which refers to mannose caps further
modified with a 5-methylthio-D-xylofuranose residue) that is almost exclusively found in TB-
causing mycobacteria, while FIND 28 targets branched arabinose structures that can be found
in LAM from a variety of mycobacteria and related species [8,25,40]. This result could indicate
that there are structural differences in the LAM epitopes detected in the two matrices. Interest-
ingly, while the FIND 28 provided poor specificity for urine measurements, it did not provide
any false positives in serum, possibly because urine is much more likely to be contaminated
with non-Mtb commensal organisms that can produce LAM and urine collection is more
prone to contamination during sampling. We saw no evidence of matrix inhibition of the
ESAT-6 assay in either matrix, although the heat inactivation step appeared to improve the sig-
nal generated by ESAT-6, even in the absence of urine or serum. While a mechanism for this
effect was not determined, potential explanations could be the exposure of epitopes by confor-
mation change (like protein unfolding) on heating or by disruption of protein complexes.
The preliminary clinical performance results from our case-control study indicate that the
ECL LAM and ESAT-6 assays for urine samples have the potential to meet the WHO sensitiv-
ity (>65%) and specificity (>98%) targets for a POC TB diagnostic test, providing support for
further research and development of improved POC tests targeting these antigens. LAM in
urine remains the most promising target. Although the observed clinical sensitivity for detec-
tion of LAM and ESAT-6 detection in serum were lower, blood-based tests could play an
important role for TB diagnosis in young children, where urine and sputum collection are less
feasible.
The clinical performance results presented here provide preliminary evidence for the
potential clinical validity of the ECL assays for detecting LAM and ESAT-6 but the generality
of the results is limited by the relatively small sample size, the use of a case-control design and
the limitation of cases to subjects with smear-positive pulmonary TB. Larger follow-up studies
are in progress to better characterize clinical performance through prospective recruitment of
relevant and unbiased populations of patients seeking care at TB centers and HIV clinics with

PLOS ONE | https://doi.org/10.1371/journal.pone.0215443 April 18, 2019 14 / 19

 Sensitive tuberculosis LAM and ESAT-6 assays

symptoms of TB and first results of the Fujifilm SILVAMP TB LAM test are available [26]. It is
also important to note that the study was limited to adult subjects, and characterization of per-
formance in a pediatric population is also needed.
Overall, our results suggest that LAM and ESAT-6 are present in urine and serum of TB
patients regardless of HIV status. It is difficult to rule out the possibility, particularly for LAM
in serum, that there may be forms of LAM present in samples that have epitopes that are either
not exposed to or not recognized by our LAM antibodies; further gains in clinical sensitivity
may be achievable through assay and/or reagent optimization. The accuracy of blood-based
TB tests may also be improved by combining the results of LAM and ESAT-6 measurements,
as our results identified subpopulations of TB+ subjects that had detectable concentrations of
only one of the antigens. Another option might be the simultaneous detection of the antigen
(s) in combination with a host blood marker to come up with a triage test. A recent systematic
review indicates that biomarker combinations are more likely to reach the required TPP per-
formance compared to single markers [54]. Research and development in the area of TB anti-
gen detection should be accelerated.

Supporting information
S1 Table. Analytical performance of LAM and ESAT-6 assays. The table summarizes the
analytical performance of LAM assays using the 6 different anti-LAM capture antibodies, as
well as the performance of the ESAT-6 assay. The results were determined from 8 point cali-
bration curves as described in Fig 1. The first two data columns show signal and CV for the
blank sample (n = 10). The third column shows the average CV for calibration standards
(n = 4 per level) giving signals above the selected detection signal to blank (S/B) threshold
value for each assay (column 4). The last column provides the limit of detection (LOD) calcu-
lated as the expected analyte concentration at the threshold as calculated from a 4-PL fit to the
calibration curve.
(DOCX)
S2 Table. Effect of heat inactivation as a sample pre-treatment step for the LAM assay. (a)
Effect of heat inactivation on spike recovery. LAM was spiked into three normal urine samples
(Neg Urine), three normal serum samples (Neg Serum) or a simple buffer (Diluent). The con-
centrations of LAM were measured in each of these samples with each of two LAM capture
antibodies (FIND 28 or S4-20), with (Heat) or without (No Heat) pre-treatment of the spiked
sample by heat inactivation. The table provides the concentrations normalized to the measured
level in diluent without pre-treatment (% Recovery). (b) Effect of heat inactivation on assay
signals for samples from TB+ individuals. LAM was measured in three urine samples and
three serum samples from TB+ individuals (Pos Urine and Pos Serum). The table provides the
assay signals with and without pre-treating the samples with heat inactivation, and also pro-
vides the fold-increase in signal with pretreatment (Ratio).
(DOCX)
S3 Table. Effect of heat inactivation as a sample pre-treatment step for the ESAT-6 assay.
(a) Effect of heat inactivation on spike recovery. ESAT-6 (~ 1 pg.mL) was spiked into three
normal urine samples (Neg Urine), three normal serum samples (Neg Serum) or a simple
buffer (Diluent). The concentrations of ESAT-6 were measured in each of these samples with
(Heat) or without (No Heat) pre-treatment of the spiked sample by heat inactivation. The
table provides the concentrations normalized to the measured level in diluent without pre-
treatment (% Recovery). (b) Effect of heat inactivation on assay signals for samples from TB+
individuals. ESAT-6 was measured in three urine samples and two serum samples from TB+

PLOS ONE | https://doi.org/10.1371/journal.pone.0215443 April 18, 2019 15 / 19

 Sensitive tuberculosis LAM and ESAT-6 assays

individuals (Pos Urine and Pos Serum). The table provides the assay signals with and without
pre-treating the samples with heat inactivation, and also provides the fold-increase in signal
with pretreatment (Ratio). Note that the samples used to generate this data are not the same
samples used to generate the data in S2 Table.
(DOCX)
S4 Table. Categorical agreement of urine and serum results. Results are shown for (a) LAM
measurements using the FIND 28 capture antibody, (b) LAM measurements using the S4-20
capture antibody, and (c) ESAT-6 measurements. Below each table are point estimates and
95% confidence intervals for the categorical agreement and Cohen’s kappa statistic.
(DOCX)
S5 Table. Categorical agreement of LAM and ESAT-6 results for (a) urine and (b) serum.
Below each table are point estimates and 95% confidence intervals for the categorical agree-
ment and Cohen’s kappa statistic.
(DOCX)

Author Contributions
Conceptualization: Tobias Broger, Payam Nahid, George B. Sigal.
Data curation: Tobias Broger, George B. Sigal.
Formal analysis: Tobias Broger, Michael Tsionksy, Anu Mathew, George B. Sigal.
Funding acquisition: Tobias Broger, George B. Sigal.
Investigation: Tobias Broger, Michael Tsionksy, Anu Mathew, Tatiana Plisova, Daniel Bart-
lett, Simone Barbero, Kiyonori Katsuragi, Masanori Kawasaki, George B. Sigal.
Methodology: Tobias Broger, Michael Tsionksy, Anu Mathew, Tatiana Plisova, Daniel Bart-
lett, Simone Barbero, Kiyonori Katsuragi, Masanori Kawasaki, George B. Sigal.
Project administration: Tobias Broger, George B. Sigal.
Resources: Kiyonori Katsuragi, Masanori Kawasaki, George B. Sigal.
Supervision: Tobias Broger, Payam Nahid, George B. Sigal.
Validation: Tobias Broger, George B. Sigal.
Visualization: Tobias Broger, George B. Sigal.
Writing – original draft: Tobias Broger, George B. Sigal.
Writing – review & editing: Tobias Broger, Todd L. Lowary, Abraham Pinter, Tatiana Plisova,
Daniel Bartlett, Simone Barbero, Claudia M. Denkinger, Emmanuel Moreau, Kiyonori Kat-
suragi, Masanori Kawasaki, Payam Nahid, George B. Sigal.

References
1. World Health Organization. Global Tuberculosis Report 2018 [Internet]. 2018. http://apps.who.int/iris/
bitstream/handle/10665/274453/9789241565646-eng.pdf?ua=1
2. Cazabon D, Alsdurf H, Satyanarayana S, Nathavitharana R, Subbaraman R, Daftary A, et al. Quality of
tuberculosis care in high burden countries: the urgent need to address gaps in the care cascade. Int J
Infect Dis. 2017; 56: 111–116. https://doi.org/10.1016/j.ijid.2016.10.016 PMID: 27794468
3. Pohl C, Rutaihwa LK, Haraka F, Nsubuga M, Aloi F, Ntinginya NE, et al. Limited value of whole blood
Xpert MTB/RIF for diagnosing tuberculosis in children. J Infect. 2016; 73: 326–335. https://doi.org/10.
1016/j.jinf.2016.04.041 PMID: 27394403

PLOS ONE | https://doi.org/10.1371/journal.pone.0215443 April 18, 2019 16 / 19

 Sensitive tuberculosis LAM and ESAT-6 assays

4. Steingart KR, Schiller I, Horne DJ, Pai M, Boehme CC, Dendukuri N. Xpert Mtb/Rif assay for pulmonary
tuberculosis and rifampicin resistance in adults. Cochrane Database Syst Rev. 2014; 1–166. https://doi.
org/10.1002/14651858.CD009593.pub3 PMID: 24448973
5. Huerga H, Ferlazzo G, Bevilacqua P, Kirubi B, Ardizzoni E, Wanjala S, et al. Incremental Yield of Includ-
ing Determine-TB LAM Assay in Diagnostic Algorithms for Hospitalized and Ambulatory HIV-Positive
Patients in Kenya. PLoS ONE. 2017; 12: e0170976. https://doi.org/10.1371/journal.pone.0170976
PMID: 28125693
6. Lawn SD, Kerkhoff AD, Burton R, Schutz C, Boulle A, Vogt M, et al. Diagnostic accuracy, incremental
yield and prognostic value of Determine TB-LAM for routine diagnostic testing for tuberculosis in HIV-
infected patients requiring acute hospital admission in South Africa: a prospective cohort. BMC Med.
2017; 15: 67. https://doi.org/10.1186/s12916-017-0822-8 PMID: 28320384
7. Cho SN, Shin JS, Kim JD, Chong Y. Production of monoclonal antibodies to lipoarabinomannan-B and
use in the detection of mycobacterial antigens in sputum. Yonsei Med J. 1990; 31: 333–338. https://doi.
org/10.3349/ymj.1990.31.4.333 PMID: 2127645
8. Kawasaki M, Echiverri C, Gler MT, Baay, T., Oda, T., Higashiyama, R., et al. Lipoarabinomannan in
sputum as a potential biomarker for bacterial load and treatment response in adult pulmonary TB
patients. Presented at the Colorado Mycobacterial Conference, Fort Collins, US; 2016.
9. Nguyen LN, Ho LM, Kuijper S, Jansen HM, Kolk AHJ. Development of Antigen Detection Assay for
Diagnosis of Tuberculosis Using Sputum Samples. J Clin Microbiol. 2000; 38: 2278–2283. PMID:
10834989
10. Minion J, Leung E, Talbot E, Dheda K, Pai M, Menzies D. Diagnosing tuberculosis with urine lipoarabi-
nomannan: systematic review and meta-analysis. Eur Respir J. 2011; 38: 1398–1405. https://doi.org/
10.1183/09031936.00025711 PMID: 21700601
11. Peter J, Green C, Hoelscher M, Mwaba P, Zumla A, Dheda K. Urine for the diagnosis of tuberculosis:
current approaches, clinical applicability, and new developments. Curr Opin Pulm Med. 2010; 16: 262–
270. https://doi.org/10.1097/MCP.0b013e328337f23a PMID: 20375787
12. Sarkar P, Biswas D, Sindhwani G, Rawat J, Kotwal A, Kakati B. Application of lipoarabinomannan anti-
gen in tuberculosis diagnostics: current evidence. Postgrad Med J. 2014; 90: 155–163. https://doi.org/
10.1136/postgradmedj-2013-132053 PMID: 24429376
13. Svenson S. Published PCT Application WO1997034149: Method of diagnosing a mycobacterial dis-
ease and immunoassay kit. 1997.
14. Amin AG, De P, Spencer JS, Brennan PJ, Daum J, Andre BG, et al. Detection of lipoarabinomannan in
urine and serum of HIV-positive and HIV-negative TB suspects using an improved capture-enzyme
linked immuno absorbent assay and gas chromatography/mass spectrometry. Tuberculosis. 2018;
111: 178–187. https://doi.org/10.1016/j.tube.2018.06.004 PMID: 30029905
15. Chan CE, Götze S, Seah GT, Seeberger PH, Tukvadze N, Wenk MR, et al. The diagnostic targeting of
a carbohydrate virulence factor from M.Tuberculosis. Sci Rep. 2015; 5: 10281. https://doi.org/10.1038/
srep10281 PMID: 25975873
16. Crawford AC, Laurentius LB, Mulvihill TS, Granger JH, Spencer JS, Chatterjee D, et al. Detection of the
tuberculosis antigenic marker mannose-capped lipoarabinomannan in pretreated serum by surface-
enhanced Raman scattering. The Analyst. 2017; 142: 186–196. https://doi.org/10.1039/C6AN02110G
PMID: 27924983
17. Laurentius LB, Crawford AC, Mulvihill TS, Granger JH, Robinson R, Spencer JS, et al. Importance of
specimen pretreatment for the low-level detection of mycobacterial lipoarabinomannan in human
serum. Analyst. 2016; 142: 177–185. https://doi.org/10.1039/c6an02109c PMID: 27934985
18. Sada E, Aguilar D, Torres M, Herrera T. Detection of lipoarabinomannan as a diagnostic test for tuber-
culosis. J Clin Microbiol. 1992; 30: 2415–2418. PMID: 1401008
19. Sarkar S, Tang XL, Das D, Spencer JS, Lowary TL, Suresh MR. A Bispecific Antibody Based Assay
Shows Potential for Detecting Tuberculosis in Resource Constrained Laboratory Settings. PLoS ONE.
2012; 7: e32340. https://doi.org/10.1371/journal.pone.0032340 PMID: 22363820
20. Steingart KR, Flores LL, Dendukuri N, Schiller I, Laal S, Ramsay A, et al. Commercial Serological Tests
for the Diagnosis of Active Pulmonary and Extrapulmonary Tuberculosis: An Updated Systematic
Review and Meta-Analysis. PLoS Med. 2011; 8. https://doi.org/10.1371/journal.pmed.1001062 PMID:
21857806
21. Shah M, Hanrahan C, Wang ZY, Dendukuri N, Lawn SD, Denkinger CM, et al. Lateral flow urine lipoara-
binomannan assay for detecting active tuberculosis in Hiv-positive adults. Cochrane Database Syst
Rev. 2016; 1–108.
22. World Health Organization. The use of lateral flow urine lipoarabinomannan assay (LF-LAM) for the
diagnosis and screening of active tuberculosis in people living with HIV [Internet]. 2015. http://www.
who.int/tb/publications/use-of-lf-lam-tb-hiv/en/

PLOS ONE | https://doi.org/10.1371/journal.pone.0215443 April 18, 2019 17 / 19

 Sensitive tuberculosis LAM and ESAT-6 assays

23. Hamasur B, Bruchfeld J, van Helden P, Källenius G, Svenson S. A Sensitive Urinary Lipoarabinoman-
nan Test for Tuberculosis. PLoS ONE. 2015; 10; e0123457. https://doi.org/10.1371/journal.pone.
0123457 PMID: 25905641
24. Paris L, Magni R, Zaidi F, Araujo R, Saini N, Harpole M, et al. Urine lipoarabinomannan glycan in HIV-
negative patients with pulmonary tuberculosis correlates with disease severity. Sci Transl Med. 2017; 9:
eaal2807. https://doi.org/10.1126/scitranslmed.aal2807 PMID: 29237757
25. Sigal GB, Pinter A, Lowary TL, Kawasaki M, Li A, Mathew A, et al. A novel sensitive immunoassay tar-
geting the MTX-Lipoarabinomannan epitope meets the WHO’s performance target for Tuberculosis
diagnosis. J Clin Microbiol. 2018; JCM.01338-18. https://doi.org/10.1128/JCM.01338-18 PMID:
30257899
26. Broger T, Sossen B, Du Toit E, Kerkhoff AD, Schutz C, Ivanova Reipold E, et al. Novel lipoarabinoman-
nan point-of-care tuberculosis test for people living with HIV with superior sensitivity: a diagnostic accu-
racy study. Lancet Infect Dis. 2019; in press.
27. Flores LL, Steingart KR, Dendukuri N, Schiller I, Minion J, Pai M, et al. Systematic Review and Meta-
Analysis of Antigen Detection Tests for the Diagnosis of Tuberculosis. Clin Vaccine Immunol. 2011; 18:
1616–1627. https://doi.org/10.1128/CVI.05205-11 PMID: 21832100
28. Fan J, Zhang H, Nguyen DT, Lyon CJ, Mitchell CD, Zhao Z, et al. Rapid diagnosis of new and relapse
tuberculosis by quantification of a circulating antigen in HIV-infected adults in the Greater Houston met-
ropolitan area. BMC Med. 2017; 15. https://doi.org/10.1186/s12916-017-0952-z PMID: 29089034
29. Liu C, Lyon CJ, Bu Y, Deng Z, Walters E, Li Y, et al. Clinical Evaluation of a Blood Assay to Diagnose
Paucibacillary Tuberculosis Via Bacterial Antigens. Clin Chem. 2018; clinchem.2017.273698. https://
doi.org/10.1373/clinchem.2017.273698 PMID: 29348166
30. Liu C, Zhao Z, Fan J, Lyon CJ, Wu H-J, Nedelkov D, et al. Quantification of circulating Mycobacterium
tuberculosis antigen peptides allows rapid diagnosis of active disease and treatment monitoring. Proc
Natl Acad Sci. 2017; 114: 3969–3974. https://doi.org/10.1073/pnas.1621360114 PMID: 28348223
31. Mukundan H, Kumar S, Price DN, Ray SM, Lee Y-J, Min S, et al. Rapid detection of Mycobacterium
tuberculosis biomarkers in a sandwich immunoassay format using a waveguide-based optical biosen-
sor. Tuberculosis. 2012; 92: 407–416. https://doi.org/10.1016/j.tube.2012.05.009 PMID: 22710249
32. Poulakis N, Gritzapis AD, Ploussi M, Leventopoulos M, Papageorgiou CV, Anastasopoulos A, et al.
Intracellular ESAT-6: A new biomarker for M ycobacterium tuberculosis infection: Intracellular Esat-6 as
Biomarker for TB. Cytometry B Clin Cytom. 2016; 90: 312–314.
33. Tang X-L, Zhou Y-X, Wu S-M, Pan Q, Xia B, Zhang X-L. CFP10 and ESAT6 aptamers as effective
Mycobacterial antigen diagnostic reagents. J Infect. 2014; 69: 569–580. https://doi.org/10.1016/j.jinf.
2014.05.015 PMID: 24968239
34. Bentley-Hibbert SI, Quan X, Newman T, Huygen K, Godfrey HP. Pathophysiology of antigen 85 in
patients with active tuberculosis: antigen 85 circulates as complexes with fibronectin and immunoglobu-
lin G. Infect Immun. 1999; 67: 581–588. PMID: 9916062
35. Kashyap RS, Rajan AN, Ramteke SS, Agrawal VS, Kelkar SS, Purohit HJ, et al. Diagnosis of tuberculo-
sis in an Indian population by an indirect ELISA protocol based on detection of Antigen 85 complex: a
prospective cohort study. BMC Infect Dis. 2007; 7: 74. https://doi.org/10.1186/1471-2334-7-74 PMID:
17620147
36. Kruh-Garcia NA, Wolfe LM, Chaisson LH, Worodria WO, Nahid P, Schorey JS, et al. Detection of Myco-
bacterium tuberculosis Peptides in the Exosomes of Patients with Active and Latent M. tuberculosis
Infection Using MRM-MS. PLoS ONE. 2014; 9: e103811. https://doi.org/10.1371/journal.pone.0103811
PMID: 25080351
37. Zeitoun H, Bahey-El-Din M, Kassem MA, Aboushleib HM. Mycothiol acetyltransferase (Rv0819) of
Mycobacterium tuberculosis is a potential biomarker for direct diagnosis of tuberculosis using patient
serum specimens. Lett Appl Microbiol. 2017; 65: 504–511. https://doi.org/10.1111/lam.12801 PMID:
28905401
38. Russell TM, Green LS, Rice T, Kruh-Garcia NA, Dobos K, De Groote MA, et al. Potential of High-Affin-
ity, Slow Off-Rate Modified Aptamer Reagents for Mycobacterium tuberculosis Proteins as Tools for
Infection Models and Diagnostic Applications. Land GA, editor. J Clin Microbiol. 2017; 55: 3072–3088.
https://doi.org/10.1128/JCM.00469-17 PMID: 28794178
39. Pollock NR, Macovei L, Kanunfre K, Dhiman R, Restrepo BI, Zarate I, et al. Validation of Mycobacterium
tuberculosis Rv1681 Protein as a Diagnostic Marker of Active Pulmonary Tuberculosis. J Clin Microbiol.
2013; 51: 1367–1373. https://doi.org/10.1128/JCM.03192-12 PMID: 23390284
40. Choudhary A, Patel D, Honnen W, Lai Z, Prattipati RS, Zheng RB, et al. Characterization of the Anti-
genic Heterogeneity of Lipoarabinomannan, the Major Surface Glycolipid of Mycobacterium tuberculo-
sis, and Complexity of Antibody Specificities toward This Antigen. J Immunol. 2018; 200: 3053–3066.
https://doi.org/10.4049/jimmunol.1701673 PMID: 29610143

PLOS ONE | https://doi.org/10.1371/journal.pone.0215443 April 18, 2019 18 / 19

 Sensitive tuberculosis LAM and ESAT-6 assays

41. Magnusson M, Kim HK, Bentzon MW. Tuberculin Production. Acta Pathol Microbiol Scand. 1963; 59:
357–368. https://doi.org/10.1111/j.1699-0463.1963.tb01807.x PMID: 14064817
42. Kaur D, Lowary TL, Vissa VD, Crick DC, Brennan PJ. Characterization of the epitope of anti-lipoarabi-
nomannan antibodies as the terminal hexaarabinofuranosyl motif of mycobacterial arabinans. Microbi-
ology. 2002; 148: 3049–3057. https://doi.org/10.1099/00221287-148-10-3049 PMID: 12368438
43. Debad JD, Glezer EN, Wohlstadter JN, Sigal GB. Clinical and Biological Applications of ECL. Electro-
generated Chemiluminescence. Marcel Dekker; 2004. pp. 43–78.
44. Diamandis EP, Stanczyk FZ, Wheeler S, Mathew A, Stengelin M, Nikolenko G, et al. Serum complexed
and free prostate-specific antigen (PSA) for the diagnosis of the polycystic ovarian syndrome (PCOS).
Clin Chem Lab Med CCLM. 2017; 55. https://doi.org/10.1515/cclm-2016-1124 PMID: 28361781
45. Eklund E, Diamandis EP, Muytjens C, Wheeler S, Mathew A, Stengelin M, et al. Serum complexed and
free prostate specific antigen levels are lower in female elite athletes in comparison to control women.
F1000Research. 2017; 6. https://doi.org/10.12688/f1000research.11821.1 PMID: 28815018
46. Glezer EN, Stengelin M, Aghvanyan A, Nikolenko GN, Roy D, Higgins M. Cytokine Immunoassays with
Sub-fg/mL Detection Limits [Internet]. AAPS Annual Meeting 2014; 2014. http://abstracts.aaps.org/
Verify/NBC14/PosterSubmissions/T2065.pdf
47. Nikolenko GN, Stengelin MK, Sardesai L, Glezer EN, Wohlstadter JN. Abstract 2012: Accurate mea-
surement of free and complexed PSA concentrations in serum of women using a novel technology with
fg/mL sensitivity. Cancer Res. 2015; 75: 2012–2012. https://doi.org/10.1158/1538-7445.AM2015-2012
48. Poorbaugh J, Samanta T, Bright SW, Sissons SE, Chang C-Y, Oberoi P, et al. Measurement of IL-21 in
human serum and plasma using ultrasensitive MSD S-PLEX and Quanterix SiMoA methodologies. J
Immunol Methods. 2019; 466: 9–16. https://doi.org/10.1016/j.jim.2018.12.005 PMID: 30590020
49. Todd JA, Evangelou M, Cutler AJ, Pekalski ML, Walker NM, Stevens HE, et al. Regulatory T Cell
Responses in Participants with Type 1 Diabetes after a Single Dose of Interleukin-2: A Non-Rando-
mised, Open Label, Adaptive Dose-Finding Trial. PLoS Med. 2016; 13: e1002139. https://doi.org/10.
1371/journal.pmed.1002139 PMID: 27727279
50. Das S, Jang IK, Barney B, Peck R, Rek JC, Arinaitwe E, et al. Performance of a High-Sensitivity Rapid
Diagnostic Test for Plasmodium falciparum Malaria in Asymptomatic Individuals from Uganda and
Myanmar and Naive Human Challenge Infections. Am J Trop Med Hyg. 2017; 97: 1540–1550. https://
doi.org/10.4269/ajtmh.17-0245 PMID: 28820709
51. World Health Organization. Main outcomes of the meeting of the WHO Expert Committee on Biological
Standardization held from 29 October to 2 November 2018 [Internet]. 2018. https://www.who.int/
biologicals/expert_committee/ECBS_Executive_Summary_final_20_NOV_2018.IK.pdf?ua=1
52. Cox JA, Lukande RL, Kalungi S, Marck EV, de Vijver KV, Kambugu A, et al. Is Urinary Lipoarabinoman-
nan the Result of Renal Tuberculosis? Assessment of the Renal Histology in an Autopsy Cohort of
Ugandan HIV-Infected Adults. PLoS ONE. 2015; 10: e0123323. https://doi.org/10.1371/journal.pone.
0123323 PMID: 25897661
53. DiGangi BA, Dworkin C, Stull JW, O’Quin J, Elser M, Marsh AE, et al. Impact of heat treatment on Dirofi-
laria immitis antigen detection in shelter dogs. Parasit Vectors. 2017; 10. https://doi.org/10.1186/
s13071-017-2443-7 PMID: 29143645
54. MacLean E, Broger T, Yerliyaka S, Fernandez-Carballo BL, Pai M, Denkinger CM. A systematic review
of biomarkers to detect active tuberculosis. Nat Microbiol. 2019; https://doi.org/10.1038/s41564-019-
0380-2 PMID: 30804546

PLOS ONE | https://doi.org/10.1371/journal.pone.0215443 April 18, 2019 19 / 19