This study on biobanked skin samples of the SSG-Allopurinol study (clinicaltrials.gov identifier: NCT04699383, study registered on 19 October 2020) and samples of the SpatialCL study (clinicaltrials.gov identifier: NCT05332093, study registered on 8 March 2022) was approved by the Ethiopian National Research Ethics Review Committee (NRERC; 17/292/973/22), the University of Gondar Institutional Review Board (VP/RTT/05/647/2022), and the Institute of Tropical Medicine Antwerp Institutional Review Board (1539/21). All participants provided written informed consent and the study was carried out in accordance with both national and international guidelines (Helsinki declaration, Good Clinical Practices, and local regulations).

Biobanked, in liquid nitrogen flash-frozen lesional punch biopsies from the index lesions were obtained from patients enrolled in two clinical studies on Ethiopian CL, namely the SSG-Allopurinol study and the SpatialCL study using a 3mm or 4mm (Kai Medical, Solingen, Germany) punch biopsy device, respectively. In total, 27 patients across the entire Ethiopian CL spectrum were included (of which 13 (48.1%) from SpatialCL and 14 (51.9%) from SSG-Allopurinol) for use in mass spectrometry-based immunopeptidomics (Figure 1). Patients from the SSG-Allopurinol study will be referred to as cohort 1, while patients from the SpatialCL will be referred to as cohort 2. The patients were selected based on a combination of availability and clinical presentation to select a representative subset of the complete Ethiopian CL spectrum. Based on clinical assessment by the treating clinicians, cohort 1 consisted of 3 (21.4%) LCL patients, 10 (71.4%) MCL patients, and 1 (7.2%) DCL patient. Cohort 2 consisted of 5 (38.5%) LCL patients, 5 (38.5%) MCL patients, and 3 (23%) DCL patients. It should, however, be noted that this classification (based on Ethiopian guidelines and borrowed from the South American setting) leads to many ambiguous cases and differential classification between dermatologists in practice (2).

To study epitope/antigen MHC-presentation early after infection, we studied L. aethiopica-infected in vitro THP-1 monocyte cell lines differentiated to macrophages. L. aethiopica was selected as it is the predominant Leishmania species in Ethiopia. Altogether, these immunopeptidomics experiments were performed in four batches (Table 1), including: the in vitro L. aethiopica-infected and HLA-restricted cell lines across early infection timepoints (5h and 24h post-infection for batch #1, and 72h post-infection for batch #2), and different human CL patient skin lesion biopsy samples across diverse HLA allele backgrounds ran on two different mass spectrometry machines (Batch #3 on TimsTOF-Pro (Bruker Daltonics, Bremen, Germany) and batch #4 on the more sensitive TimsTOF-SCP (Bruker Daltonics, Bremen, Germany)).

THP-1 monocytes (ATCC TIB-202) were cultured in T25, T75, and T175 culture flasks (Greiner Bio-one, Belgium) using a Roswell Park Memorial Institute (RPMI) (Biowest, U.S.A.) medium supplemented with 10% Fetal Calf Serum (Biowest, U.S.A.) and 50µM 2-mercapto-ethanol (Sigma-Aldrich, Belgium). These THP-1 monocytes were cultured until 5 million cells could be obtained. Next, the THP-1 monocytes were differentiated to macrophages by overnight stimulation with 0.1 µM Phorbol 12-Myristate Acetate (PMA, Thermo Fisher, Belgium). After differentiation to macrophages, the cells were infected with L. aethiopica promastigotes (MHOM/ET/72/L100, passage Rx+29). Infected macrophages were harvested at either 5 hours (batch 1), 24 hours (batch 1), or 72 hours (batch 2) post-infection for use in mass spectrometry-based immunopeptidomics.

The L. aethiopica-infected THP-1 macrophages and 18 CL patient lesion biopsy samples (batch 3) were processed by the Centre for Proteomics of the University of Antwerp according to the previously published protocol with minor modifications (20). In brief, the clinical samples were homogenized by sequential homogenization using a Next Advance BulletBlender Gold bead homogenizer, in a lysisbuffer containing 150mM NaCl, 50mM tris hydroxymethylaminomethane at pH 8, one tablet of cOmplete™ ethylenediaminetetraacetic acid (EDTA)-free anti-protease (Roche, Germany) per 50mL, and 0.1% Fos-choline-14 diluted ULC/MS-grade water, whereas the L. aethiopica-infected THP-1 macrophages were lysed using the same buffer without bead homogenization. The samples were then incubated for 1 hour at 4 °C. Next, the samples underwent several centrifugation rounds at 4 °C to pellet down the nuclei (see HLA genotyping section), and to remove unsolubilized membranes, purifying the MHC complexes. Next, lysates were pre-cleared on empty protein A/G-sepharose bead columns to remove non-specific binders. After this, MHC class I and class II complexes were subjected to sequential affinity purification by binding of the pre-cleared lysate to MHC immuno-affinity columns. These columns were generated by cross-liking protein 1mL of A sepharose beads to 5mg of anti-HLA I antibody (clone W6/32, IchorBio, UK) and anti-HLA II antibodies (clones L243, 1a3, and B7/21, IchorBio, UK) per sample. After crosslinking, uncrosslinked antibodies were stripped by incubating 5 minutes in 0.1M citrate buffer at pH3. After batch binding of the lysates on the antibody columns, beads were sequentially washed with a series of wash buffers to non-specifically bound material and residual salt and detergent, followed by a mild acid elution with 10% formic acid to elute MHC complexes from the column and dissociate immunopeptides. Subsequently, immune peptides were separated from the larger components of the MHC complexes by C18 solid phase extraction. Next, the peptides were loaded for measurement on a TimsTOF-Pro (Bruker Daltonics, Bremen, Germany) connected to an Evosep One chromatography system (Evosep Biosystems, Denmark). The Evosep was run with a 44 minutes default gradient and an EV1106 column. The Parallel Accumulation–Serial Fragmentation (PASEF) DDA method without pre-defined polygone was employed to select precursor ions for fragmentation involving 1 TIMS-MS scan and 10 PASEF MS/MS scans. The TIMS-MS survey scan spanned 0.70 to 1.40 Vs/cm2 and 100–1,700 m/z, with a ramp time of 166ms for MHC-I peptides, and 100ms for MHC-II peptides, while a linear collision energy ramp from 20 to 59 EV with an ion mobility range from 1.6 to 0.6 Vs/cm2 was used for peptide fragmentation.

Lesion biopsy samples from 9 CL patients (batch 4) were processed by ImmuneSpec (Niel, Belgium), using a more sensitive workflow and more sensitive mass spectrometry device. This workflow is similar to the above, with the following steps adapted: these samples were homogenized in ImmuneSpec’s propietary lysisbuffer by sequential homogenization steps using a Precellys bead homogenizer (Bertin Technologies, France) followed by 1 hour incubation (lysis). Lysates were precleared on empty protein A-sepharose beads (to remove non-specific binders) prior to binding on anti-HLA I-Sepharose beads (antibody clone W6/32 crosslinked to protein A Sepharose), and sequential binding on anti-HLA-II-Sepharose beads (antibody clones L243, 1a3, B7/21 crosslinked to protein A Sepharose). After C18 solid phase extraction, the eluted peptides were dried, resuspended in Mobile Phase A, and loaded on evotips prior to measurement on a Bruker Daltonics TimsTOF-SCP coupled to an Evosep One Nano-LC.

PEAKS Online 11 (Bioinformatics Solutions Inc., Canada) was used for MS data analysis. A database search was performed against the reviewed Uniprot Human Proteome (Swiss-Prot only) database (release 2024_03, 42.490 entries, downloaded March 2024), concatenated with the reference proteome of L. aethiopica strain L147 (version 67; 8.533 entries) acquired from the TriTryDB (21). Error tolerance was set at 20 ppm and 0.05 Da for precursor ions and fragmentation ions, respectively. Digestion was set as unspecific and no enzyme was specified. Oxidation (M) and Deamidation (NQ) were selected as variable modifications. A false discovery rate (FDR) of 1% was set at peptide level.

DNA was extracted from either patient whole blood collected in 10mL Lithium-Heparine tubes (BD Vacutainer, Erembodegem, Belgium; Cohort 2 study samples) or the nuclei byproducts of the mass spectrometry-based immunopeptidomics assay (Cohort 1 study samples) (Figure 1), using the Maxwell^® RSC 48 instrument (Promega, Wisconsin, U.S.A.) with the Maxwell^® RSC Whole Blood DNA kit (Promega, Wisconsin, U.S.A.) according to the manufacturer’s instructions. Next, the concentration of the DNA was measured using the Qubit 1X dsDNA BR assay (ThermoFisher, Waltham, U.S.A.) on a Qubit Fluorometer 4.0 device (ThermoFisher, Waltham, U.S.A.). A total of 200ng of participant’s DNA was used for HLA genotyping using the Oxford Nanopore Technologies (ONT)-sequencing-based NanoTYPE™ assay (Omixon, Budapest, Hungary) according to the manufacturer’s instructions. In brief, an enrichment PCR was performed using the HLA Multi Primer Mix and reagents included in the NanoTYPE™ kit. Next, amplicons were quantified using the Qubit 1X dsDNA BR assay (ThermoFisher, Waltham, U.S.A.) on a Qubit Fluorometer 4.0 device (ThermoFisher, Waltham, U.S.A.), and a total of 200ng of amplicon was transferred to a new tube. A barcoding step was then performed using the Rapid Barcoding Plate provided in the Rapid Barcoding 96 Kit (SQK-RBK110.96; ONT, Oxford, U.K.). After this, samples were pooled (between 8–12 samples per library), adding 8.5 µl of barcoded amplicons per sample to a library. Next, resulting pooled libraries were subjected to size selection and purification using the AMPure XP beads (Beckman Coulter, Brea, U.S.A.) also provided in the Rapid Barcoding 96 Kit. A total of 10 µl of purified library was then transferred to a new tube, and 1 µl of Rapid Adapter F (provided in the Rapid Barcoding 96 Kit) was added to the library. The resulting library with added adapters was then mixed with 37.5 µl of Sequencing Buffer II and 25.5 µl of Loading Beads II to load on to a R9.4.1 Flow Cell (ONT, Oxford, U.K.) for subsequent sequencing on a MinION Mk1B instrument (ONT, Oxford, U.K.) with MinKNOW version 22.05.5. Libraries were sequenced for at least 1 hour per sample in a library. The resulting FAST5 files were basecalled using Guppy version 6.1.5 using the high-accuracy model. Finally, basecalled FASTQ files were used for HLA genotype inference with the NanoTYPER v2.0.0 software (Omixon, Budapest, Hungary).

Peptides were assigned MHC restrictions using the MHCMotifDecon 1.2 software with default settings using the HLA genotyping assay data (22). All other downstream analyses on the PEAKS Online 11 output was performed in R version 4.3.3. Leishmania MHC-presented peptides were defined as those mapping exclusively to the Leishmania proteome and using MHC class-specific length thresholds of 8-14mer and 12-21mer peptides for MHC-I and MHC-II, respectively. To verify if our observed L. aethiopica peptides are similar in sequence to prior published Leishmania epitopes, we queried the IEDB and calculated Levenshtein distances using the stringdist package version 0.9.1. Next, heatmaps of antigen or peptide overlap between patients were created using the complexHeatmap package version 2.18.0. Next, we calculated the summed number of MHC-presented peptides derived from an antigen.

A gene ontology (GO) enrichment test was performed on all the MHC-presented Leishmania antigens using the enricher function (default settings) of the clusterProfiler package at version 4.10.1 in R version 4.3.3 with curated L. aethiopica gene ontologies acquired from the TriTrypDB (23). This was performed on all Leishmania antigens, and separately for MHC-I and MHC-II. Only Benjamini-Hochberg-adjusted p-values below 0.05 were considered significant.

Proteome-wide prediction of L. aethiopica epitopes was performed using the reference proteome of L. aethiopica strain L147 (version 67), acquired from the TriTrypDB, as input for NetMHCpan 4.1 and NetMHCIIpan 4.3. This was performed for the HLA genotypes of the THP-1 monocyte (ATCC TIB-202) cell line. The NetMHCpan peptide lengths were set to 8-14mer, which deviates from the default range of fragment lengths (between 8-11mer), and the NetMHCIIpan peptide lengths were set to 12-21. These fragment lengths were chosen to match the length thresholds applied in the immunopeptidomics data analysis. Subsequently, only predicted strong and weak binders, according to the default %Rank cut-offs, were included for further analysis.

Our study included a total of 27 patients, of which 23 (85%) were male with a median age of 22 years, matching prior observations that CL in Ethiopia affects mostly young adults (Table 2) (1). The CL episodes persisted for a median of 9 (6.5-15) months prior to sampling, and presented with a median of 1 (1-2) lesion of approximately 6 (4-7) cm predominantly located on the face (78%). Of the 27 CL patients, 4 (14.8%) had more than 5 lesions.

To identify MHC-presented Leishmania antigens and epitopes, we performed mass spectrometry-based immunopeptidomics across different experiments and conditions (see Figure 1).

We identified a total of 85.920 and 74.804 different unique MHC-I and MHC-II peptides, respectively Supplementary Figure S1A). Of these, 7.231 MHC-I and 3.452 MHC-II peptides were identified in the in vitro samples of batch #1 and batch #2, and 78.689 MHC-I and 21.003 MHC-II peptides were identified in the clinical samples of batch #3 and batch #4. The eluted peptide length distributions summed across experiments (Supplementary Figure S1B), or within each experiment (Supplementary Figure S2), were consistent with expected MHC-I and MHC-II peptide length distributions. Peptides of 9mer length were the predominant eluted peptide length for MHC-I, and MHC-II showed a peak around 15mer peptides.

In total, 511 MHC-I and 411 MHC-II presented peptides mapped exclusively to L. aethiopica (Figure 2A). As expected, a higher yield was reached in batch #4 compared to batch #3 due to the use of a more sensitive mass spectrometer. Further selection of high-confidence MHC-I peptides with a length restriction between 8-14mer and MHC-II peptides between 12-21mers (dashed lines on Figure 2B, Supplementary Figures S1B,S2–3) resulted in 379 MHC-I and 271 MHC-II L. aethiopica peptides across 443 L. aethiopica antigens included for further analyses (listed in Supplementary Excel). Of these, 333 MHC-I and 247 MHC-II Leishmania peptides across 398 Leishmania antigens were identified in clinical samples. Of note, 245 (37.7%) of the 650 MHC-presented Leishmania peptides were identified from a single DCL patient. None of these T cell epitopes were previously reported on the IEDB, and all identified peptides deviated by at least a Levenshtein distance of 4 (being largely dissimilar) from those epitopes previously described.

Next, we studied whether there was any overlap in epitope or antigen presentation between patients with different timings from infection, and with early in vitro infected samples. Those epitopes or antigens that are shared between individuals represent potential immunoprevalent targets, a measure for how prevalent an eventual immune response would be in a given population, which correlates with the immunodominance of a response (24–27).

Despite the HLA diversity between patients, and mass spectrometry detection limits revealing only a subset of all presented epitopes, 19 exact peptides (13 MHC-I, 6 MHC-II) were shared between two patients (Figure 3). In one case, a peptide was shared between two long-term CL patients and was also already presented as early as 24h post-infection in a L. aethiopica-infected THP-1 cell line, namely IINEPTAAAIAYGLN[Deamidated]K, derived from the antigen Heat-shock Protein (Hsp70; LAEL147_000494700). Six MHC-presented peptides, shared between two patients, were derived from antigens annotated as ‘hypothetical protein’. Patients sharing an exact peptide had at least one or more matching HLA allele with the other patient, except in the case of the peptides AAPAGAAAL and ASASASAAL where patients only carried the closely related HLA-C*07:01 and HLA-C*07:02, and C*03:03 and C*03:04 alleles, respectively, which likely share peptide specificity (Supplementary Excel 2). Due to limited HLA genotyping quality, HLA matches could not be determined and confirmed for all. For the peptide YLTAELLEL, no HLA genotype match or close match could be observed, suggesting that there is likely overlap in the binding peptide repertoires of two HLA alleles that are not closely related. Two peptides, M[Oxidation]PPQFGSGRPF and RLLDVVYNA, were presented in both the 5h and 24h post-infection in the L. aethiopica-infected THP-1 cell line samples, showing good concordance between samples of the same batch. The peptide IKSGTAHKSFVD was present at both 5h and 72h post-infection in these in vitro samples, also suggesting the detection of peptides is robust across batches.

As exact peptides might not be shared but still derived from the same protein, we subsequently analyzed which proteins were shared across patients. Here, 51 antigens were presented in a minimum of two patients and a maximum of 5 patients (Figure 4). Of these, 43 (84.3%) were shared across two patients, 7 (13.7%) were shared across three patients, and one (2%, Elongation Factor 2 (EF-2; LAEL147_0000067)) was shared between five patients (Figure 4). Of the five patients sharing EF-2 presentation, four had two matching HLA alleles, namely HLA-B*49:01, and HLA-C*07:01, while the other patient had the closely related HLA-C*07:02 allele.

When including the in vitro samples, 66 (14.9% of total antigens) antigens were shared between a minimum of two samples and maximum of 5 samples (Figure 4). Of these 66 shared antigens, 54 (81.8%) were shared between two samples, 6 (9.1%) were shared between three samples, 4 (6.1%) were shared between four samples, and 2 (3%, namely EF-2 and Hsp70) were shared between five samples. Three antigens, Clathrin heavy chain (LAEL147_000825100) and two copy number variants of 40S Ribosomal Protein S8 (LAEL147_000396600; LAEL147_000396700), were presented in the 5h and 24h post-infection in vitro samples, showing good concordance between samples of the same batch. Two copies of 40S Ribosomal Protein S3 were presented in the 5h and 72h post-infection in vitro samples, and a copy of Heat-shock Protein Hsp70 and Fusaric Acid Resistance Protein (LAEL147_000168900) were presented in the 24h and 72h post-infection in vitro samples, showing good concordance between all in vitro samples. As expected, the DCL patient where we identified the most MHC-presented peptides from also had the highest number of antigens shared between other patients with 27 (40.9%) of 66 shared antigens. Of the 66 antigens, 19 (24.2%) were annotated as ‘hypothetical protein’, for which there is no known function yet.

As immunodominance can be reflected by a high number of T cell epitopes across the protein (epitope-richness), we next looked at the number of MHC-presented peptides per identified L. aethiopica antigen.

EF-2, Alpha Tubulin (LAEL147_000015900), and glyceraldehyde 3-phosphate dehydrogenase (GAPDH; LAEL147_000578000) had 5 MHC-presented peptides across 5, 3, and 1, patients, respectively (Figure 5A). Next, Histone H2A (LAEL147_000533800; LAEL147_000533900) and 10kDa Heat Shock Protein (LAEL147_000431900) had 4 MHC-presented peptides across 3 and 1 patients, respectively. DNA Polymerase Zeta Catalytic Subunit (LAEL147_000368800), and ATPase Beta Subunit (LAEL147_000412300), Hsp70, Ankyrin Repeat Protein (LAEL147_000770300), all had 3 peptides across 3 patients each. Myosin Heavy Chain Kinase C-like Protein (LAEL147_000406700) also had 3 MHC-presented peptides but only in 2 patients.

When including the in vitro samples of early infection, the antigen with the highest number of eluted peptides was Histone H2A with 6 peptides across four samples (Figure 5B). In addition, the one peptide of this antigen that was shared between a patient and an in vitro sample was already presented as early as 72h post-infection (Figure 3). The antigens with the second highest number of eluted peptides were EF-2, Hsp70, Alpha Tubulin, 40S Ribosomal Protein S33 (LAEL147_000442300), and GAPDH with 5 epitopes per antigen (Figure 5B). While peptides of EF-2 and Hsp70 were presented in 5 different samples, and Alpha Tubulin in 3 samples, peptides of 40S Ribosomal Protein S33 and GAPDH were only presented in a single sample. A second gene copy number variant of Hsp70 (LAEL147_000511600) and Ankyrin Repeat Protein had 4 peptides presented in four samples, and 10kDa Heat Shock Protein also had 4 peptides but presented only in a single sample.

To further characterize the biological characteristics of the naturally presented L. aethiopica immunopeptidome, the cellular location and biological function of the MHC-presented Leishmania antigens was analyzed.

We first performed a GO enrichment test on all the MHC-presented Leishmania antigens, and only the terms “Ribosome”, “Structural constituent of ribosome”, and “Translation”, were significantly over-represented (Supplementary Figure S4). Performing this GO over-representation test for only MHC-I-presented or MHC-II-presented Leishmania antigens resulted in the terms “Nucleosome”, “DNA binding”, “Ribosome”, “Structural constituent of ribosome”, and “ATPase-coupled transmembrane transporter activity” for MHC-I, and “Ribosome”, “Structural constituent of ribosome”, and “Translation” for MHC-II. This suggests most of the observed MHC-presented Leishmania antigens tend to be involved in mRNA to protein translation as part of ribosomal machinery. Although for MHC-I in particular, nucleosomal and transmembrane proteins with DNA-binding, and ATPase activity, tend to be overrepresented as well.

All antigens have orthologues in four other common Leishmania species (L. major, L. donovani, L. infantum, and L. braziliensis) and share high degrees of sequence identity (68.95%-100%, Table 3), indicating a high level of protein conservation across species.

As a final step, we wanted to evaluate the overlap between our experimentally acquired results with the most widely used in silico peptide-MHC binding prediction tools (NetMHCpan and NetMHCIIpan). First, we compared L. aethiopica proteome-wide prediction outputs to the experimentally acquired immunopeptidomics data of the HLA-restricted L. aethiopica-infected THP-1 cell line (Figures 6A, B). In total, NetMHCpan and NetMHCIIpan predicted 1.587.865 MHC-I and 6.989.036 MHC-II peptides, respectively. For MHC-I, NetMHCpan correctly predicted 15 (36.6%) of the 41 epitopes observed to be MHC-presented in the in vitro immunopeptidomics experiments (Figure 6A). Similarly, NetMHCIIpan predicted 7 (29.2%) of the 24 MHC-II epitopes observed in the in vitro immunopeptidomics experiments (Figure 6B). Next, we compared L. aethiopica proteome-wide prediction outputs to the experimentally acquired immunopeptidomics data of the clinical samples for which we had full HLA genotyping data (Figures 6C, D). We predicted L. aethiopica MHC-binding peptides for 9 patients of experimental batch #3, and 4 patients of experimental batch #4. In total, NetMHCpan and NetMHCIIpan predicted 5.991.913 MHC-I and 14.607.989 MHC-II peptides for the clinical samples, respectively. For MHC-I, NetMHCpan correctly predicted 144 (71.6%) of the 201 epitopes observed to be MHC-presented in the immunopeptidomics experiments on clinical samples (Figure 6C). Similarly, NetMHCIIpan predicted 138 (80.7%) of the 171 MHC-II epitopes observed in the immunopeptidomics experiments on clinical samples (Figure 6D). Our findings indicate that, while immunopeptidomics only identifies the tip of the iceberg (likely the most abundant proteins), still ~60-70% and ~20-30% of these MHC-presented peptides were missed by the prediction tools in the in vitro and clinical experiments, respectively, despite the tools predicting millions of strong- and weak-binding peptides.