Identification of highly conserved Trypanosoma cruzi antigens for the development of a universal serological diagnostic assay

ABSTRACT

Chagas Disease is an important neglected tropical disease caused by Trypanosoma cruzi. There is no gold standard for diagnosis and commercial serological tests perform poorly in certain locations. By aligning T. cruzi genomes covering parasite genetic and geographic diversity, we identified highly conserved proteins that could serve as universal antigens for improved diagnosis. Their antigenicity was tested in high-density peptide microarrays using well-characterized plasma samples, including samples presenting true infections but discordant serology. Individual and combination of epitopes were also evaluated in peptide-ELISAs. We identified >1400 highly conserved T. cruzi proteins evaluated in microarrays. Remarkably, T. cruzi positive controls had a different epitope recognition profile compared to serologically discordant samples. In particular, multiple T. cruzi antigens used in current tests and their strain-variants, and novel epitopes thought to be broadly antigenic failed to be recognized by discordant samples. Nonetheless, >2000 epitopes specifically recognized by IgGs from both positive controls and discordant samples were identified. Evaluation of selected peptides in ELISA further illustrated the extensive variation in antibody profiles among subjects and a peptide combination could outperform a commercial ELISA, increasing assay sensitivity from 52.3% to 72.7%. Individual variation in antibody profiles rather than T. cruzi diversity appears to be the main factor driving differences in serological diagnostic performance according to geography, which will be important to further elucidate. ELISA with a combination of peptides recognized by a greater number of individuals could better capture infections, and further development may lead to an optimal antigen mixture for a universal diagnostic assay.

KEYWORDS:

• Chagas disease
• enzyme-linked immunosorbent assay
• antibody
• diagnostic
• peptides

Introduction

Chagas Disease is an important neglected tropical disease, impacting an estimated 6–7 million people worldwide [1], with $627.46 million in annual health-care costs and a burden of 806,170 Disease Adjusted Life Years (DALYs) globally [2]. Trypanosoma cruzi, the etiological agent, is a genetically diverse protozoan that is currently divided into seven major genetic lineages, also referred to as Discrete Typing Units (DTUs), named TcI-TcVI and TcBat [3,4]. Diagnosis of T. cruzi infection relies mainly on serological tests; however, no gold standard T. cruzi diagnostic test currently exists. Indeed, current guidelines, require two positive tests based on different antigens or principles to confirm infection, leading to increased cost, patient loss, and delays in the start of treatment [5–7]. Limited diagnostic performance also creates a higher potential risk for both congenital transmission [8] and transfusion/transplant transmission. Therefore, the improvement of serological diagnostics for T. cruzi infection remains a priority.

Though manufacturers of currently used commercial serological tests for T. cruzi report sensitivities and specificities upwards of 99%, independant evaluations indicated much lower performance parameters [9–12], often associated with geographic location. For example, a sensitivity as low as 27.9% has been observed for Chagatest recombinant ELISA and 66–67% for Chagas Stat Pak and T-detect rapid tests when used with samples from Mexico, while the performance of these tests is much higher (>90% sensitivity) with samples from Argentina [8,9,11,12]. Similarly, a sensitivity of 30% has been reported for Chagas Stat Pak in samples from Peru [13]; but an overall sensitivity of 87% was found in Bolivian samples [13,14]. Chagas Detect Plus, the only rapid test currently approved by the FDA, was shown to have variable performance according to geographic region [15,16]. In a study performed on US blood donors, this rapid test showed the highest sensitivities for those donors born in South America, the lowest sensitivities for those born in Mexico, and intermediate sensitivities for those born in Central America [10,11].

Discordance among tests and poor diagnostic performance is thought to be due, at least in part, to genetic differences between parasite strains circulating in various regions. Indeed, TcI has a strong geographic structure across the Americas [17–19] and both TcIII and TcIV have been proposed to be genetically structured between North and South America [20,21]. A similar geographic structuration is also emerging among TcII, TcV and TcVI as these DTUs are increasingly detected in Central and North America [22,23]. On the other hand, most antigens used for T. cruzi diagnostics have derived from a limited number of strains, mostly from Brazil and Argentina [24,25] and their level of conservation may have been overestimated [19]. Thus, current tests do not reflect the broad antigenic diversity of the parasite across the Americas [9,10,26,27]. In fact, the idea that antigens derived from local T. cruzi strains can perform better than antigens from “reference” strains for serological testing has been circulating for some time [28,29].

New approaches, including the use of high-density peptide microarrays, are providing powerful opportunities to screen large number of antigens and have allowed the identification of many new parasite antigens recognized by well characterized sera [25,30–32]. However, the extent of conservation of many of these antigens across T. cruzi strains and DTUs and their potential recognition from samples with discordant serology is still unclear, although some have shown promise [30]. Indeed, most studies aiming at identifying T. cruzi antigens only rely on well-characterized consensus samples that are clearly negative or positive for T. cruzi antibodies based on several available tests, and do not include discordant samples. Consequently, these new antigens can only provide results in accordance with currently used diagnostics [31,33,34], failing to truly address the problem of test discordance and underdiagnosis.

Therefore, we aimed here at identifying highly conserved protein sequences among T. cruzi genomes from all DTUs and a broad geographic distribution, to identify truly universal parasite antigens for improved diagnosis, independently of parasite strains circulating in human populations. The antigenicity of these conserved parasite proteins was tested using panels of well-characterized plasma samples from broad geographic regions. Critically, these included samples presenting true infections (PCR positives for T. cruzi) but discordant serology with current tests, to assess potential differences in their antigenic recognition profile compared with samples unequivocally seropositive. Finally, a peptide ELISA was evaluated for the serological diagnosis of T. cruzi infection, to overcome the limitations of current tests. This work is critical for the development of a universal serological tests for T. cruzi diagnostic across the Americas.

Materials and methods

Identification of T. cruzi most conserved proteins

Fourteen T. cruzi genomes representing different geographies and DTUs TcI to TcVI (Supplementary Table 1) were aligned using progressiveMAUVE [35] using either current assemblies or newly assembled sequence reads as described before, and default alignment settings [36]. Available annotations were used to identify coding regions with good sequencing coverage across these genomes. Coding sequences with 80% nucleotide pairwise identity across all genomes were selected for further analysis.

Nucleotide sequences were translated in an appropriate reading frame and resulting protein sequences were aligned via MUSCLE using the standard PPP algorithm. Protein alignments with ≥90% pairwise identity across T. cruzi DTUs were selected and the consensus peptide sequence for each alignment was extracted. To prevent cross-reaction of any epitopes with other parasitic infections, all selected sequences were compared to other parasite genomes (Supplementary Table 2), including Leishmania spp., Trypanosoma rangeli, Trypanosoma brucei, Plasmodium vivax, and Toxoplasma gondii, using tBLASTn (amino acid inquiry of a DNA database). These species were selected for overlap of disease geography and/or relatedness to T. cruzi. Peptide sequences were also compared to a Homo sapiens genome.

Sequences with ≥90% pairwise identity in any of the non-T. cruzi genomes were discarded. All remaining unique protein consensus sequences were considered highly conserved and T. cruzi specific and selected for further analysis (Supplementary Table 3).

High-density peptide microarrays for antigenicity screening

The selected proteins were included in custom high-density peptide microarrays consisting of an overlapping peptide library of 15-mers, with an overlap of 13 amino acids covering all these proteins, to assess their antigenicity. The microarrays also included additional peptides derived from CMV large phosphoprotein antigen and the Herpes envelope epitope SHRANETIYNTTLKY as positive controls, as they are recognized by a majority of healthy adult humans. Similarly, two neo-proteins of random peptide sequence were included as negative controls [25], as well as Cy3 blank spots. Finally, peptides from well-characterized T. cruzi antigens, several of which used in diagnostic tests such as SAPA, CA-2/B13, or Ag30 were included as additional controls (Supplementary Table 4). For some of these antigens, sequences representing the different parasite DTUs were included to assess DTU-specific antibody responses. One copy of each unique 15-mer was synthesized on a C-terminal – βAla – Asp spacer by Schafer-N (Denmark) resulting in about 500,000 peptides in the arrays in a 20 × 20 µm layout with 10 µm separation. Microarrray slides were deprotected in TFA EDT H₂0 for 3 h at room temperature and blocked overnight in 0.1% BSA, 0.1% Tween-20 in PBS. After blocking, slides were incubated for 1 h at room temperature with one of the IgG probes described below (100 µg/ml IgG in 0. 1% BSA, 0.1% Tween-20 in PBS), washed 3 × 20 min with 0.1% BSA, 0.1% Tween-20 in PBS and incubated for 1 h at room temperature with Cy3- goat anti-Hu IgG (1 µg/ml in 0.1% B SA, 0.1% Tween-20 in PBS). Slides were washed 3 × 20 min with 0.1% BSA in PBS, dried, and scanned using a laser scanner with 1 µm resolution. Fluorescent signal intensity and Z scores were reported and microarray data is available in NCBI GEO database, accession #GSE235074.

Peptide microarray analysis

Antibody recognition of Herpes and CMV control peptides as well as Cy3 blank controls, expressed as binding intensity signal (arbitrary units) were first used for quality control of the microarrays. Next, peptide recognition profile of well-characterized antigens was assessed. Heatmaps were elaborated to compare antibody binding intensity signals to peptides among pools of well-characterized IgG/plasma samples. For the antigenicity analysis of conserved T. cruzi proteins, an initial global analysis was performed to compare binding intensity signals among groups. From this, a subset of the peptides showing the largest variance in binding intensity among sample pools were selected to assess their recognition profile. Again, heatmaps were elaborated to compare antibody binding intensity signals to peptides among pools of well-characterized IgG/plasma samples, using MORPHEUS (https://software.broadinstitute.org/morpheus) with clustering based on Euclidian distances or GraphPad Prism9.

Human plasma and IgG samples

All human plasma were derived from a biorepository of well characterized samples collected in a previous study [37]. Briefly, these samples had been tested with Chagas STAT PAK® (Chembio Diagnostic, Inc.), Trypanosoma Detect^TM (InBios), Chagatest ELISA recombinant v. 3.0 (Wiener), and Hemagen® Chagas’ Kit (Hemagen) as well as by quantitative PCR and two end-point PCR assays, one targeting nuclear satellite DNA (primers TcZ1-TcZ2) [38], and one targeting minicircle DNA (primers 121–122) [39], with extensive quality control and cross-validation among participating laboratories [40,41]. Samples from the biorepository were stored at −80°C, with constant temperature monitoring, and were retested with Chagas STAT PAK®, Trypanosoma Detect^TM and Chagatest ELISA recombinant v. 3.0 prior to use in this work to ensure that they had been preserved in optimal conditions.

Microarrays were initially screened using three pools of purified IgG to assess antigenicity of the conserved T. cruzi proteins. We selected a pool of unequivocally positive samples, i.e. positives for >3 serological tests and T. cruzi PCR (N = 9), and a pool of unequivocally negative samples, i.e. negative for all serological and PCR assays (N = 11). Importantly, we also used a pool of samples positive by PCR thus corresponding to true infections, but with discordant serology, typically reactive in only one of the serological assays (N = 11) (Supplementary Table 5). Each pool included samples from Argentina, Honduras and Mexico (N = 3–5 from each country). For each pool, two microliters of plasma per individual sample were combined and IgG antibodies were purified using Thermo Scientific™ Melon™ Gel IgG Spin Purification Kit, per kit instructions. Concentration and purity of resulting IgGs were measured on a Nanodrop2000 spectrophotometer.

For peptide-ELISA evaluation, the individual samples used in the screening pools described above were used for an initial evaluation. This panel included 11 confirmed negative samples and 20 confirmed positive samples. This later group included 9 positive control samples and 11 discordant positives. An extended set of 84 samples was used for a final evaluation of ELISA performance based on the best antigenic preparation. These included 40 confirmed negative samples, and 44 confirmed positive samples comprised of 23 positive controls and 21 discordant samples, based on the same assays described above (Supplementary Table 6). Peptide-ELISA performance was evaluated through sensitivity, specificity, and accuracy parameters, which were compared to those from the Chagatest ELISA recombinant v. 3.0 on the same samples.

Peptide-ELISA

Based on the microarray data, we selected peptides specifically recognized by positive samples for their evaluation in peptide-ELISA assays. Peptides were synthesized with an additional C- terminal cysteine residue to allow for conjugation to maleimide-activated BSA (Thermo- Fisher) following manufacturer’s protocols. Briefly, 0.5 mg of peptide was mixed with 0.5 mg of carrier protein and incubated at room temperature for 2 h. The mixture was then applied to a Zeba desalting spin column (Thermo Fisher) to remove excess unconjugated peptide and final BSA-conjugated peptide concentration was measured at 280 nm on a Nanodrop. Peptides were diluted in PBS, pH 7.4 for use in ELISA assays. Briefly, 100 ng/well of peptide-mBSA were coated onto 96-well polystyrene plates overnight at 4°C. Unbound peptide was removed by sequential washes with PBST wash buffer (PBS with 0.05% Tween) and plates blocked for 1 h at room temperature with 200 µl blocking buffer (PBS with 0.05% Tween, 1% BSA). After another series of washes, plates were incubated with 100 µl of diluted plasma samples in duplicates (1:500 in PBS with 0.05% Tween, 0.3% BSA) for 1 h at room temperature and washed. A horseradish peroxidase-labeled goat anti-human IgG secondary antibody (Jackson Immuno Research Labs) was added and incubated for 1 h at room temperature. Finally, 200 µl/well of substrate solution (Phosphate Citrate Buffer + TMB + 35% fresh hydrogen peroxide) was added and plates were incubated in the dark at room temperature for 30 min. The reaction was stopped using 2M H₂SO₄ and plates were read at 450 nm. OD₄₅₀ readings were adjusted by subtracting the average blank from all individual readings and adjusted values were averaged for each duplicate. Cut-off value of reactivity for each peptide were defined as the mean OD450 of negative samples plus two standard deviations (SD). For combinations of peptides, a total amount of peptide of 100–150 ng/well was maintained, so that 5–10 ng of each individual peptide was used. Sensitivity and specificity were calculated to compare diagnostic performance of peptide ELISAs.

Results

Identification of T. cruzi most conserved proteins

Sequence alignment of T. cruzi genomes covering TcI-TcVI allowed the identification of 3910 annotations shared among all DTUs, corresponding to approximately 6 Mbp of genome sequence and 28% of annotated genes. Of these, 1912 had >80% sequence identity among DTUs. After translation, these resulted in 1756 highly conserved protein sequences with >90% of pairwise sequence identity among T. cruzi DTUs. From these, 292 sequences showing some similarity with other pathogens or humans were discarded, and 1464 proteins were retained, corresponding to about 10% of T. cruzi proteome (Supplementary Table 3). These sequences had an average pairwise identity of 97.1% across T. cruzi DTUs, and low pairwise identity with other pathogens and humans (Supplementary Table 7). Thus, they represent a set of highly conserved and specific T. cruzi proteins that could be used for a universal diagnostic test, irrespective of parasite strains/DTUs infecting humans.

More than half (799) of these conserved proteins were hypothetical proteins and no single named protein group appeared to be over-represented, although three trans-sialidase genes were present. Only a few proteins corresponded to previously reported parasite antigens, including JL8, hypothetical protein TcCLB.508385.10, surface protein TolT, putative calpain-like cysteine peptidases or flagellum-associated proteins. Of note, we did not identify commonly used diagnostic antigens such as SAPA or TSSA, likely because these did not meet the sequence conservation threshold that we defined above (>90% identity).

Antigenic screening in peptide microarrays

The identified set of T. cruzi most conserved proteins was then screened for recognition by IgG from T. cruzi infected patients using high density arrays of overlapping peptides. These arrays also included Herpes and CMV epitopes for quality control since most adult humans have antibodies against these viruses. As expected, all pools showed a strong reactivity to epitopes from both viruses, while negligible binding was observed for negative Cy3 blank spots (Supplementary Figure 1), validating the arrays. Interestingly, IgG pools had a rather different peptide recognition profile of the CMV antigen, suggesting differences in the immune response against this common antigen among the pools.

We then examined IgG binding to well characterized T. cruzi antigens, including several of currently used diagnostic antigens such as SAPA, TSSA and CA-2/B13. As expected, IgG from the pool of negative samples showed negligible binding to these antigens, while there was a strong recognition of multiple peptide epitopes from all these antigens with IgG from T. cruzi positive control samples (Figure 1). On the other hand, IgG from serology discordant but T. cruzi positive samples presented very limited or no recognition of any of these antigens, in agreement with the failure of current commercial tests to identify these samples.

Figure 1. Heatmaps of IgG binding profile to well-characterized T. cruzi antigens. Binding intensity to overlapping peptides (numbered on the Y axis) covering previously characterized antigens was colour-coded according to the indicated scales (Low binding: green; high binding: red) for each antigen. These included CA-2/B13 (A), SAPA (B), Antigen 1 (C), TSSA (D), Ribosomal P2 (E), and Ribosomal L19 (F). IgGs from pools of negative (NEG), positive (POS), and discordant positive (DISC) samples were assessed.

To further explore if the lack of recognition of these antigens was due to sequence diversity among T. cruzi strains/DTUs used for diagnostic and/or infecting these patients, we evaluated IgG binding to sequence variants of some of the antigens. One of the best characterized epitopes for DTU-specific serology are derived from TSSA antigen, which performed well in a rapid-test format with human and animal samples from Argentina [42]. Analysis of these DTU-specific epitopes indicated that none were reactive with the negative pool, while the positive control pool was strongly reactive to epitopes from TcIII and TcIV TSSA, but not to epitopes from TcI and TcII/TcV/TcVI TSSA (Figure 2(A)). This was unexpected as infections with TcI and TcII/TcV/TcVI predominated in these samples and TcIII and TcIV have not been detected [43]. Furthermore, the pool of discordant positive samples did not react with any of the TSSA epitopes from the different DTUs. Analysis of the reactivity of additional DTU-specific variants for antigens such as SAPA, JL8 and Ag30 showed a similar pattern: the pool of positive controls showed high reactivity against several peptide epitopes from these antigens, with a strong cross-reactivity for peptides corresponding to variants from TcI, TcII and TcVI DTUs, while the pool of discordant positive samples presented weak or no reactivity against any of the sequence variants tested (Figure 2(B,D)). Together, these data suggested that contrary to our initial hypothesis, DTU-specificity of these well characterized antigens, or their insufficient sequence conservation among DTUs, did not explain the low/absent reactivity of the pool of discordant positive samples. Rather, these samples appeared to lack antibodies against these key parasite antigens.

Figure 2. Effect of antigen diversity on antibody recognition. Heatmaps of IgG binding profile to T. cruzi sequence variants of antigens is shown. Binding intensity to overlapping peptides (numbered on the Y axis) was colour-coded according to the indicated scales (Low binding: green; high binding: red) for each antigen and its strain/DTU variants. These included TSSA (A), SAPA (B), JL8 (C), and Antigen 30 (D). IgGs from pools of negative (NEG), positive (POS), and discordant positive (DISC) samples were assessed.

We then focused on the antibody recognition profiles of our set of most conserved T. cruzi proteins. Overall, of the over 400,000 overlapping peptides covering these proteins, the large majority presented no binding or non-specific also binding present in the pool of T. cruzi negative samples. Only 9870 peptides (2.5%) were specifically recognized by IgGs from the positive control pool (Figure 3(A)), 2965 (0.7%) by IgGs from the discordant positive pool (Figure 3(B)), and 1933 (0.5%) were recognized by both pools. We thus selected the top 4700 peptides with the largest variance in binding intensity across the three IgG pools to generate a heatmap of their recognition profile. After hierarchical clustering, three major groups of peptides could be observed (Figure 3(C)). One group consisted in peptides specific to the positive control pool, another to peptides specific to the discordant positive pool, and a group of 2176 peptides specifically recognized by both positive and discordant pools. These data highlight that while the discordant positive pool lacks IgGs against well characterized T. cruzi antigens, it has a distinct antibody profile against other T. cruzi conserved antigens compared to the positive control pool. Peptides reactive to both pools are good candidates for the development of a universal T. cruzi diagnostic test, independent of circulating parasite strains/DTUs.

Figure 3. Recognition profile of peptides from T. cruzi most conserved proteins. (A) Comparison of binding intensity of IgG from positive (Y axis) and negative pools (X axis) to individual peptides derived from T. cruzi most conserved proteins. Each dot represents a peptide, with red dots representing peptides specifically recognized by positive controls, orange dots representing peptides specifically recognized by discordant samples and/or positive controls, black dots are peptides not recognized by any sample pool, and grey circles are non-specific peptides recognized by the negative pool. (B) Comparison of binding intensity of IgG from discordant positive pool (Y axis) and negative controls (X axis) to individual peptides. (C) Heatmap of the binding profile from the top 4700 peptide following hierarchical clustering. Binding intensity to peptides was colour-coded according to the indicated scales (Low binding: green; high binding: red). Peptides specific to the positive control pool can be identified (left enlargement), as well as peptides specific to the discordant positive pool (right enlargement), and over 2000 peptides specifically recognized by both positive and discordant pools (centre enlargement).

Peptide-ELISA

We then selected 80 peptides for their evaluation in peptide-ELISAs using a panel of well-characterized individual plasma samples that again included negative and positive controls as well as discordant positive samples. A total of 24 peptides were insoluble and could not be tested, resulting in the evaluation of 56 T. cruzi peptides that included peptides from well-characterized antigens as control and peptides reactive to positive and discordant positive pools (Supplementary Table 8).

Control peptides derived from T. cruzi antigens CA-2/B13, PFK, SAPA, TSSA, Riboprotein P2, Riboprotein L19, P0, R27-2, Antigen 36 and Antigen 1 presented an overall good specificity as no or few negative samples reacted to these (Figure 4, top section). Also, most positive control samples were reactive to these peptides as expected, although not all samples reacted to all peptides, indicating some individual variation among samples. For example, only 1/9 positive controls reacted to Antigen 1 and TSSA peptides, 2/9 reacted to PFK peptide, and 3/9 reacted to B13b peptide. More importantly, very few discordant samples reacted to these peptides, in agreement with the peptide microarray screening and the failure of current commercial tests to detect these discordant but positive samples.

Figure 4. Peptide ELISA reactivity of individual plasma samples. The reactivity of individual peptides (horizontal lines) was colour coded as indicated (reactive: red; negative: green) based on peptide ELISAs with individual plasma samples (vertical columns). These included Negative and Positive controls, as well as Discordant positive samples, derived from Argentina (ARG), Honduras (HON), and Mexico (MEX). The top section corresponds to peptides from known antigens, and the bottom section to peptides from new conserved antigens.

Evaluation of the novel T. cruzi conserved peptides showed no or very limited binding from negative controls, indicating a high specificity (Figure 4, bottom section). On the other hand, several positive control samples presented a high binding above the cut-off value, indicating peptide-specific antibodies in these samples. Nonetheless, some positive control samples also failed to react to some of the peptides. Multiple discordant samples also reacted with several of the peptides, indicating potential to increase diagnostic sensitivity, but many samples also failed to recognize several of these peptides. Thus, there was extensive individual variation in their peptide recognition profile, as each sample had antibodies that recognized different peptides. There was also no particular clustering of these profiles according to the geographic origin of the subjects, and samples from Argentina, Honduras and Mexico presented a comparable variability.

We further assessed the proportion of samples from T. cruzi-infected individuals (including both positive and discordant positive individuals) that recognized peptides. Overall, most peptides (46/56, 82%) were recognized by less than 30% of infected individuals, and only 10 peptides (18%) were recognized by at most 33-62% of infected individuals (Figure 5(A)). Thus, most peptides represented private antigens recognized by a few subjects, and none of these peptides were truly universal antigens recognized by all subjects. This is in agreement with a recent study similarly based on a screen of the T. cruzi proteome, indicating that most of its antigenic regions were private, with a very limited set of antigens recognized by more than 50% of T. cruzi infected subjects from their panel [30]. Because some of these apparently shared antigens were sufficiently conserved among parasite strains to have been included in our microarray screen, we analysed their recognition by our sample pools. As expected, many peptides were strongly recognized by the positive control pool, but not by the negative pool, indicating excellent specificity (Figure 5(B)). However, several also failed to be recognized by our positive control pool and only three peptides were weakly recognized by the discordant positive pool, indicating that they are not as universal as initially thought and further emphasizing the uniqueness of antigen recognition profile of the infected subjects in our panels. Also, this very broad inter-individual variation in antibody profile implied that single peptide diagnostic is not feasible, but that a combination of peptides is needed for optimal diagnostic performance.

Figure 5. Variability in epitope recognition profiles. (A) The prevalence of samples from T. cruzi -infected individuals (including both positive and discordant positive individuals) that recognized each peptide was calculated. Ten peptides were recognized by a single individual (0.05% of infected), and only one peptide was recognized by 62% of infected individuals. (B) Recognition profile of previously identified “universal” epitopes that were included in the peptide microarray. The heatmap shows the binding profile to the indicated epitopes. Binding intensity was colour-coded according to the indicated scales (Low binding: green; high binding: red).

Therefore, we next tested mixtures ranging from 6 to 22 peptides in ELISAs using the same set of plasma samples. As expected, we observed an overall increase in the reactivity of positive plasma samples to peptide mixtures as the number of peptides increased (sensitivity of 30% to 65%), while the reactivity of negative samples remained low (specificity of 100%), suggesting improved performance of peptide mixtures compared to single peptides. The optimum sensitivity and specificity were obtained with Mixtures 12 and 14 (Table 1), which included 22 and 20 peptides, respectively (Supplementary Table 7). These reached a sensitivity of 65% with a specificity of 100%. Compared to the performance of a commercial ELISA, this resulted in a 20% increase in sensitivity (45% to 65%) while maintaining the high specificity (Table 1).

Table 1. Diagnostic performance of peptide mixtures and commercial ELISA.

Peptide Mixture	Performance*	95% Confidence Interval
Mix 1 (6 peptides)	Sensitivity: 30.0% Specificity: 90.9%	11.9% to 54.3% 58.7% to 99.8%
Mix 2 (8 peptides)	Sensitivity: 60.0% Specificity: 81.8%	36.0% to 80.9% 48.2% to 97.7%
Mix 3 (9 peptides)	Sensitivity: 65.0% Specificity: 100%	40.8% to 84.6% 71.5% to 100%
Mix 4 (14 peptides)	Sensitivity: 60.0% Specificity: 100%	36.0% to 80.9% 71.5% to 100%
Mix 9 (19 peptides)	Sensitivity: 60.0% Specificity: 100%	36.0% to 80.9% 69.1% to 100%
Mix 12 (22 peptides)	Sensitivity: 65.0% Specificity: 100%	40.8% to 84.6% 69.1% to 100%
Mix 14 (20 peptides)	Sensitivity: 65.0% Specificity: 100%	40.8% to 84.6% 69.1% to 100%
Chagatest ELISA	Sensitivity: 45.0% Specificity: 100%	23.0% to 68.5% 69.1% to 100%

*Assay performance was evaluated based on a panel of 11 confirmed negative samples and 20 confirmed positive samples. This later group included 9 positive control samples and 11 discordant positives.

To expand on the evaluation of the performance of one of the best peptide mixture (Mix 14), we then used a larger set of plasma samples (N = 84), again comparing with a commercial ELISA. As expected, negative control samples presented very low/no reactivity against the peptide mixture, while positive controls presented a high reactivity (Figure 6). In addition, many discordant samples were reactive to the peptides, indicating a clear improvement in the sensitivity of the assay that outperformed a commercial ELISA. Indeed, the peptide ELISA reached a sensitivity of 72.7%, although specificity was somewhat decreased to 87.5% (Table 2). Importantly, this improvement in sensitivity was observed in samples from all countries tested, i.e. Argentina, Honduras and Mexico (Figure 6). Indeed, the proportion of discordant samples identified as reactive was 1/3 (33.3%) for samples from Argentina, 3/4 (75%) for samples from Honduras, and 5/14 (35.7%) for samples from Mexico.

Figure 6. Peptide-ELISA readings for the best peptide combination. Peptide ELISA with mixture 14 was performed for 84 samples. OD₄₅₀ readings from individual plasma samples are shown, corresponding to negative (NEG) and positive (POS) samples, as well as discordant positive samples (DISC). Aggregated data (All samples) and country-specific data for Argentina (ARG), Honduras (HON) and Mexico (MEX) are presented. The horizontal dotted lines indicate the cut-off value to define reactive samples.

Table 2. Diagnostic performance of peptide and commercial ELISA.

Peptide Mixture 14		Chagatest ELISA
Parameter*	95% CI	Parameter*	95% CI
Sensitivity: 72.7%	57.2% to 85.0%	Sensitivity: 52.3%	36.7% to 67.5%
Specificity: 87.5%	73.2% to 95.8%	Specificity: 100%	91.2% to 100%
Accuracy: 78.6%	68.3% to 86.8%	Accuracy: 75%	64.3% to 83.8%

*Assay performance was evaluated based on an extended panel of 84 samples. These included 40 confirmed negative samples, and 44 confirmed positive samples. This later group included 23 positive controls and 21 discordant samples.

Discussion

Limited serological test performance and discordant results are making Chagas disease surveillance costly and complicated, leading to low access to diagnostics, patient loss, and delays in access to treatment [5–7]. Thus, improved diagnostics for T. cruzi infection are urgently needed as this is directly tied to improved health and economic outcomes. A major hypothesis is that test discordance is due in large part to the antigenic diversity of T. cruzi strains, as the parasite exhibits extensive diversity among and within its main lineages/DTUs and across geographic regions. Therefore, we proposed here to identify highly conserved parasite proteins, to search for universal antigens able to detect infections from any T. cruzi strain in broad human populations.

Alignments of T. cruzi genomes covering TcI-TcVI DTUs from a broad geographic distribution allowed identifying over 1500 of the most conserved parasite proteins, presenting over 97% of sequence identity within T. cruzi species. This set of conserved proteins included only about 10% of the parasite proteome, which is in agreement with its high level of genetic diversity [19]. These conserved proteins comprised a wide variety of functions, though not surprisingly, most were still described as hypothetical proteins. Importantly, none of the well characterized parasite antigens were included among this highly conserved set, highlighting their limited level of conservation among T. cruzi strains. Thus, this set of the most conserved T. cruzi proteins is an ideal source of diagnostic antigens to overcome issues of parasite diversity across the Americas.

The antigenicity of these highly conserved parasite proteins and previously identified antigens was then assessed using high-density peptide microarrays, using well-characterized plasma/IgG samples. While most studies focus on unambiguous negative and positive samples for such screenings, a key departure of our study was to include a set of discordant plasma samples for the evaluation the recognition profile of these parasite antigens. Thus, a first key observation from this screening was that samples from T. cruzi positive control subjects had a very different epitope recognition profile compared to samples from infected but serologically discordant subjects. Importantly, this applied to an extensively large collection of epitopes that included those from the CMV large phosphoprotein antigen, from conventional T. cruzi antigens such as several used in current tests (CA-2/B13, SAPA, TSSA, Antigen 1 or Antigen 30), from novel T. cruzi epitopes thought to be broadly recognized [30], and from many of the conserved T. cruzi proteins we evaluated.

Remarkably, while T. cruzi positive control samples reacted strongly with epitopes from all conventional antigens tested, serology discordant but T. cruzi positive samples showed very limited or no recognition of any of these antigens, which agrees with the failure of current serological tests to identify these samples. Even newly described epitopes thought to be recognized by 50–100% of T. cruzi infected subjects [30] failed to be reactive to our discordant positive pool, and several even did not react with our positive control pool, suggesting a limited usefulness to improve serological testing.

Importantly, the assumption that discordant serology may be due to infections with different parasite strains/DTUs between individuals from the discordant positive pool and those from the positive control pool was not supported by our data. Indeed, none of the strain/DTU specific antigens tested significantly altered the recognition profile from both pools, and the discordant positive pool failed to react to a broad range of antigen strain variants. On the other hand, the positive control pool presented a good cross-reactivity to antigen variants from multiple strains/DTUs. A notable exception was the TSSA DTU-specific epitopes, which unexpectedly showed a very specific reactivity to TcIII and TcIV DTUs, which have not been detected in these samples [43]. Together, these data suggests that while parasite diversity may be associated with difference in antibody profiles as noted before [30], it is not the main cause of poor diagnostic performance.

Indeed, microarray data clearly indicated that serology is discordant but T. cruzi positive samples do recognize to a broad set of conserved parasite epitopes, but these largely differed from those recognized by positive control samples, although a smaller set of over 2000 epitopes appeared to be shared. This later set of epitopes is a key resource for the development of improved diagnostic tests. However, evaluation of some of these epitopes in ELISAs further evidenced a great individual heterogeneity in specific anti-T. cruzi antibody responses. Thus, we identified only one epitope recognized by over 60% of our samples from infected individuals, while 10 epitopes were exclusively privates as each was recognized by a single subject. Again, this agrees with a comparable screen of T. cruzi proteome indicating that most of its antigenic regions were private, with a very limited set of antigens recognized by most infected subjects [30]. Comparable observations of variable antibody repertoires and private antigens have been made in infections with other parasites such as Plasmodium [44,45] and Schistosoma [46]. Thus, variable humoral responses may thus be a hallmark of parasitic infections as the immune system is faced with hundreds/thousands of potential antigen targets, as opposed to bacterial or viral infections, for which there is a much-reduced diversity of antigenic targets. This adds an important limitation when using pools of serum/plasma sample for antigen identification, as this prevents top detect individual variability in antibody repertoire. Human Leukocyte Antigen (HLA) gene variation has long been linked to associations with disease [47] and could be involved in individual differences in antibody repertoires and test discordance as suggested before [12]. Future studies should assess this aspect of host diversity on serological testing. Also, it is unclear how the antibody repertoire of individuals may change over time, particularly considering chronic T. cruzi infections. Indeed, these involve low and fluctuating levels of blood parasites over decades of infection, and antibodies against specific antigens may vary with bursts of parasitemia. Accordingly, the follow-up of drug-treated Chagasic patients using multiplex serology suggests different rates of seronegative conversion for the respective antigens used, in agreement with differential kinetics of specific antibody production [48,49].

In any case, a high inter-individual variability in antibody profile implies that multiple antigens need to be combined for an optimum performance of serological tests. A first attempt at evaluating some of the identified epitopes in a multipeptide-ELISA resulted in an assay that outperformed a commercial ELISA, with a sensitivity that was increased from 52.3% to 72.7%, and additional peptides may be needed to further increase sensitivity. Although the sample size remained limited, these results validated this approach to developing an improved serological test for T. cruzi infections and highlighted the critical need to include samples with discordant serology for the screening and evaluation of novel diagnostic antigens.

In conclusion, we identified here a set of the most conserved T. cruzi proteins towards the development of improved serological tests. Importantly, the discordance detected in current tests appears due to a lack of IgGs against diagnostic antigens in these subjects rather than to antigen sequence variation among parasite strains and DTUs. Thus, extensive inter-individual variation in antibody profiles results in the recognition of broadly different parasite antigens, with very limited overlap among individuals, making diagnosis challenging. Nonetheless, a multipeptide-ELISA including some of these shared epitopes was able to outperform a commercial ELISA, illustrating the usefulness of our approach and paving the way towards a universal T. cruzi diagnostic test.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the Eunice Kennedy Shriver National Institute of Child Health & Human Development [grant number 5R01HD94955] to CH and by the Louisiana Board of Regents through the Board of Regents Support Fund [# LESASF (2018-21)-RD-A-19] to ED.