ERAP1 Allotypes 2 and 10 Differentially Regulate the Immunopeptidome of Melanocytes

ABSTRACT

Endoplasmic reticulum aminopeptidase 1 (ERAP1) is a polymorphic enzyme that shapes the peptide repertoire presented by MHC class I molecules and can regulate adaptive immune responses in cancer and autoimmunity. Common missense polymorphisms in ERAP1 modulate its activity and are found in specific allotypes in humans. ERAP1 allotypes are linked to predisposition to HLA‐associated inflammatory diseases such as psoriasis and Behçet's disease, through the generation of specific CD8+ T cell populations targeting disease‐specific HLAs. Given the established broad effects of ERAP1 activity on the cellular immunopeptidome, we hypothesised that ERAP1 allotypic variation may lead to broad immunopeptidome shifts capable of triggering the observed antigenic responses. To test this hypothesis, we generated two A375 melanoma cell lines, each one expressing one of the most common, disease‐associated ERAP1 allotypes, namely allotypes 2 or 10. Comparison of the immunopeptidome of these two cell lines showed only minor differences in peptide sequences presented but extensive changes in abundance that included alterations in length distribution, binding affinity, and sequence motifs. Our results suggest that enzymatic differences between ERAP1 allotypes are reflected primarily in the quantitative composition of the cellular immunopeptidome. These quantitative changes may constitute a mechanism that underlies ERAP1‐allotypic associations with HLA‐associated autoimmunity and variable immune responses.

ERAP1 is an intracellular aminopeptidase that edits antigenic peptides to be presented by MHC class I molecules and is found in distinct functional allotypes in individuals. To investigate how different disease‐associated ERAP1 allotypes affect the cellular immunopeptidome, we generated A375 melanoma cell clones carrying either allotype 2 or allotype 10 and isolated the immunopeptidome by affinity chromatography. Peptides were sequenced by LC–MS/MS, and sequences were assigned using the non‐specific HLA‐DIA Fragpipe workflow. This approach allowed the identification of significant shifts in the abundance of MHC class I presented peptides by cells carrying different ERAP1 allotypes, supporting the notion that ERAP1 allotypes contribute to individual immune variability. Image was created using material from Servier Medical Art (https://smart.servier.com/), licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/).

1. Introduction

Cellular adaptive immune responses are primarily mediated by specialised immune cell receptors that recognise Major Histocompatibility Class I molecules (MHC‐I; HLA‐I in humans) complexed with small peptides derived from the proteolytic degradation of intracellular proteins [1]. These peptides, collectively known as the immunopeptidome, are generated by complex intracellular proteolytic cascades and can encompass thousands of different amino acid sequences [2].

Peptides originating from self‐proteins do not normally elicit immune responses. This happens primarily due to the clonal deletion of T cell clones carrying receptors that can recognise them during the development of immunological tolerance [3]. On the contrary, peptides originating from infection or oncogenic transformation of the cell can be recognised by immune cells, such as CD8+ T cells, leading to the eradication of the antigen‐ presenting cell [4]. Indeed, presentation of cancer‐associated antigens is a major determinant of anti‐tumour immunity and often a target in immunotherapy efforts [5, 6].

Insufficient formation of tolerance, chronic inflammation, and pathogen antigenic mimicry can sometimes lead to the erroneous recognition of self‐peptides as pathogenic, initiating and/or sustaining autoimmune inflammatory responses and leading to disease [7, 8, 9]. Such a molecular mechanism has been suggested to underlie the pathogenesis of a group of diseases termed MHC‐I‐opathies, such as Ankylosing Spondylitis, Spondylarthritis, Behçet's disease, Birdshot Uveitis, and psoriasis [10]. Notably, while the HLA in humans is highly polymorphic with more than 20,000 different haplotypes described [11], MHC‐I‐opathies have a very strong genetic linkage with specific alleles. This suggests that possible pathogenic mechanisms may involve specific antigenic peptides or sets of antigenic peptides presented by the disease‐associated HLA [12]. Indeed, recent work has identified antigenic peptide recognition by specific T‐cell receptors on CD8+ cells, always in the context of disease‐associated HLA, providing further support for this pathogenic hypothesis [13, 14, 15].

Beyond HLA polymorphism, emerging evidence also highlights the pivotal role of intracellular antigen processing components, such as endoplasmic reticulum aminopeptidase 1 (ERAP1) [16] and its close homologue, endoplasmic reticulum aminopeptidase 2 (ERAP2) [17], in shaping the immunopeptidome and driving immune responses [18, 19, 20, 21, 22, 23]. ERAP1 and ERAP2 trim N‐terminal amino acids from extended precursors of mature antigenic peptides, optimising their length for binding to nascent HLA‐I [24, 25]. While ERAP2 preferentially trims shorter peptides and is generally considered to have a secondary role [25, 26], ERAP1 predominantly trims larger precursor peptides and acts as the principal determinant of antigen presentation diversity [27]. However, ERAP1 can often over‐trim mature antigenic epitopes so that they are no longer suitable for HLA‐I binding, essentially destroying them [28]. By this dual function, the activity of ERAP1 can regulate the immunopeptidome and adaptive immune responses [29, 30, 31]. Thus, ERAP1 is currently a pharmacological target aiming to regulate the antigenicity of cells in applications such as cancer immunotherapy [32, 33].

Like HLA, ERAP1 is also polymorphic, and several single nucleotide polymorphisms (SNPs) have been associated with predisposition to human disease, including viral infections, anti‐tumour responses, and most notably MHC‐I‐opathies [34, 35, 36]. Often, genetic predisposition between ERAP1 SNPs is found to be in epistasis with HLA alleles, consistent with the molecular role of the enzyme in generating the peptide repertoire that is available to bind to HLA‐I [37]. These SNPs can influence ERAP1 function in different ways. While some affect protein expression [38], many result in missense mutations generating variants with distinct enzymatic properties [39]. Within the human population, these coding SNPs organise in specific allotypes that present specific functional properties [40, 41]. Of note, ERAP1 allotype 10 was first reported to be a loss‐of‐function variant and exists in about a quarter of the European population [42]. Follow‐up studies, however, have suggested that this allotype rather presents a unique specificity for peptide substrates [41, 43]. Such differences in specificity are likely to influence antigen presentation and thereby drive autoimmune responses.

Indeed, a recent report identified a T‐cell receptor from psoriasis patients' T cells recognising an autoantigen presented by HLA‐C*06:02 in melanocytes [13], which required ERAP1 for its generation and was dependent on ERAP1 allotypic variation. Allotype 2 was highly efficient in generating the autoantigenic peptide and inducing a T cell response, while allotype 10 was ineffective and reduced immunogenicity. This finding was consistent with genetic predisposition findings which suggested that ERAP1 allotype 2 is predisposing to psoriasis, while allotype 10 is protective, introducing a molecular framework for understanding these effects. In addition, ERAP1 allotype 10 has been shown to change the immunogenicity of antigenic peptides presented by the Behçet's disease associated HLA‐B*51 allele [44, 45]. Since the enzymatic activity of ERAP1 has been demonstrated to also affect the immunogenicity of cancer cells and responses to immunotherapy, the differential activity of allotypes may also regulate anti‐tumour responses [19, 20, 23, 35].

Given the complex landscape of allotype 10 functional differences, we hypothesised that changes in antigen presentation between ERAP1 allotypes may not be limited to single antigens but rather extend to broad shifts in the immunopeptidome, which can have implications for both autoimmunity and cancer. To test this hypothesis, we generated monoallelic clones of the melanoma cell line A375 that carry either ERAP1 allotype 2 or allotype 10; these are the most common allotype worldwide and the most common allotype in European populations [40], respectively, and have been associated with autoimmunity [13, 44, 45].

Analysis of the immunopeptidomes of these two clones revealed broad differences in the abundance of presented peptides that comprise changes in length distribution, sequence motifs, and binding affinity, all of which can influence recognition by immune cell receptors. These results suggest that functional enzymatic differences between ERAP1 allotypes are reflected in the cellular immunopeptidome and may underlie mechanisms linking ERAP1‐mediated antigen processing to HLA‐associated autoimmunity and cancer.

2. Experimental Procedures

2.1. Plasmids

pcDNA3.1(+) vectors carrying the codon‐optimised coding sequence for ERAP1 allotype 2 and allotype 10 were obtained by custom gene synthesis (BioCat GmbH, Heidelberg). The ERAP1 gene was inserted in the pcDNA3.1(+) vector between BamHI and XhoI cloning sites. The amino acid sequences for both allotypes are shown in Figure S1.

2.2. Generation of Cell Lines

A375 ERAP1 Knock‐Out (KO) cells (clone 1B12, previously described [20]) were cultured in high glucose DMEM containing L‐glutamine (Biowest, L0104) with the addition of 10% Fetal Bovine Serum (Biowest, S1810) and 1% Penicillin‐Streptomycin (Biowest, L022) as usual at 37°C, 5% CO₂. For the generation of A375 clones, the pcDNA3.1(+) plasmids were linearised with SspI‐HF (New England Biolabs, R3132) and purified with the Monarch PCR & DNA Cleanup Kit (New England Biolabs, T1030) according to the manufacturer's instructions, prior to the transfection of A375 ERAP1 KO cells. For the transfection, cells were seeded in a 6‐well plate (500,000 cells/well) and 24 h later they were transfected with each vector, using the jetPRIME (Polyplus, 101 000 027) transfection reagent, according to the manufacturer's instructions. 48 h post‐transfection, both transfected and untransfected cells were exposed to 1.2 mg/mL G‐418 (InvivoGen, PHC4033) (established from an antibiotic killing curve on untransfected A375 KO cells) for 2 weeks, with frequent medium changes. Surviving cells were diluted in 96‐well plates (aiming for 0–1 cells/well) to obtain single clones, which were then allowed to grow. Western blots using a human aminopeptidase PILS/ARTS1 antibody (R&D Systems, AF2334) were performed to confirm the ERAP1 expressing clones and to identify clones with similar expression levels for the two allotypes. Positive clones were cultured in the complete medium stated above but also supplemented with 0.6 mg/mL G‐418.

2.3. Flow Cytometry

Selected positive clones (2B1 and 1H1) were seeded in a 24‐well plate (100,000 cells/well), and 24 h later, they were treated with human recombinant interferon‐γ (IFN‐γ, Gibco, PHC4033) at concentrations ranging from 0 to 50 ng/mL for another 24 h. Cells were subsequently detached and prepared for flow cytometry analysis after staining with anti‐human HLA‐ABC FITC labelled antibody (Biorad, MCA81F) (1:25). Analysis was performed with a FACSCelesta cell analyser (BD Biosciences, Heidelberg, Germany) using the FACSDiva Software (BD Biosciences).

2.4. Isolation of the Immunopeptidome

The immunopeptidome of A375 cells was isolated as previously described [18] with some modifications. Cells were grown in three separate biological replicates, treated with 20 ng/mL of IFN‐γ for 24 h, detached using Accutase (Merck Millipore, SCR005), pelleted, and stored at −80°C (2.5 × 10⁸ per sample). For isolation of MHC‐I, cell pellets were thawed on ice and cells lysed with 10 mL lysis buffer/sample (Tris–HCl, pH 7.5, 150 mM NaCl, 0.5% IGEPAL CA‐630, 0.25% sodium deoxycholate, 1 mM EDTA, pH 8.0, cOmplete EDTA‐free protease inhibitor cocktail tablets). The lysate was then cleared with ultracentrifugation at 100,000×g for 1 h at 4°C, prior to loading on CN‐Br pre‐columns followed by the W6/32 coupled columns (generated by coupling W6/32 antibody, grown in‐house, onto cyanogen bromide‐activated Sepharose 4B beads, as previously described [18]). The flow‐through from this procedure was loaded on the columns three more times prior to washing the columns with 20 bed volumes 20 mM Tris–HCl, pH 8.0, 150 mM NaCl, 20 bed volumes 20 mM Tris–HCl, pH 8.0, 400 mM NaCl, 20 bed volumes 20 mM Tris–HCl, pH 8.0, 150 mM NaCl, and finally with 40 bed volumes 20 mM Tris–HCl, pH 8.0. MHC‐I‐peptide complexes were eluted by washing with 1% trifluoroacetic acid (TFA), and the samples were stored at −80°C. Peptide eluates were subjected to SpeedVac and separated from MHC‐I molecules using reversed‐phase C18 disposable spin columns (Thermo Scientific, 89870). Adequate separation was evaluated with a western blot against β‐2 microglobulin (R&D Systems, MAB8248). Ultimately, the purified peptides were further processed by the Sp3 protocol for peptide clean‐up, as described previously [18], solubilised in the mobile phase A (0.1% FA in water) and sonicated.

2.5. Liquid Chromatography/Mass Spectrometry

For LC–MS/MS analysis, a setup consisting of a Dionex Ultimate 3000 nano RSLC online with a Thermo Q Exactive HF‐X Orbitrap mass spectrometer was used, and data were obtained both in a data‐dependent (DDA, 1 technical replicate/sample) and data‐independent acquisition (DIA, 2 technical replicates/sample) modes. Samples were initially injected in a 25 cm‐long analytical C18 column (PepSep, 1.9 μm³ beads, 75 μm ID) with a 30‐min gradient. At the start of the gradient, the mobile phase was composed of 7% Buffer B (0.1% FA in 80% ACN). The gradient was then increased to 35% B over 16.5 min, followed by a rise to 45% within 1.5 min, and then to 99% B over 0.5 min. Flow was held stable at 99% B for 1 min before returning to 7% over the next 10.5 min.

For DDA data, a range between 275 and 1100 m/z was scanned in the full MS, with 120 K resolving power, Automatic Gain Control (AGC) of 3 × 10⁶ and maximum Injection Time (max IT) of 100 ms. For the MS/MS, the 5 most abundant ions were selected, resolving power was set to 30 K, AGC to 1 × 10⁶, max IT to 50 ms and Normalised Collision Energy (NCE) to 30.

For the DIA data, a range between 375 and 1100 m/z was selected in the full MS with 120 K resolving power, AGC of 3 × 10⁶ and max IT of 60 ms. For the MS/MS, 8th windows (39 loop counts) were selected, resolving power was set to 15 K, AGC to 3 × 10⁶, max IT to 22 ms, and NCE to 26.

2.6. Data Analysis

MS/MS spectra were searched against the UniProt database (HUMAN_UP000005640_9606, 20597 entries, retrieved: 24.11.2023) with the non‐specific HLA‐DIA workflow in Fragpipe [46] v 22. In this workflow, which is specifically designed for immunopeptidomics, a hybrid spectral library was produced by DDA and DIA runs and used for a non‐specific search with MSFragger for peptides between 7 and 25 amino acids, accounting for N‐terminal acetylation, N‐terminal pyro‐glu of glutamine and glutamic acid, oxidation of methionine and cysteinylation as variable modifications. PSMs, ions, and peptides were filtered to 1% FDR at each level. Quantification was ultimately performed from the DIA files with DIA‐NN [47]. Data were analysed and visualised with FragPipe Analyst [48], without further normalisation and imputation. Peptides with log2 difference ≥ 1 and q‐value ≤ 0.05 were considered differentially expressed. MHCMotifDecon 1.0 [49] was used for motif extraction and binding affinity predictions via NetMHCPan 4.1 [50].

2.7. Nonmetric Multidimensional Scaling Analysis

Nonmetric multidimensional scaling analysis (NMDS) [51] was performed for the detection of condition‐specific peptide motifs for each HLA allele, as previously described [20] with some modifications. Briefly, 9‐mer binding peptides per HLA were subjected to a separate analysis using entropy‐weighted peptide distances in two‐dimensional space [MolecularEntropy() function from HDMD R package]. The NMDS function from the ecodist R package [52] was used with 10 separate ordinations of 500 iterations, and the configuration with the least stress was selected for the visualisation of the immunopeptidome. Density‐based spatial clustering for applications with noise (DBSCAN) [53] from the fpc package [54] was then applied to cluster peptides based on the elbow method for the estimation of the number of clusters (KNNdistplot function in dbscan [53]). The Eps parameter was customised for each HLA allele. Statistical evaluation of the differences in the proportion of H2 versus H10 peptides per cluster was then performed using Fisher's exact (counts < 5) or proportion test with Bonferroni correction for multiple testing. Gseqlogo R package [55] (ggplot2 R package [56]) was finally used to generate sequence logo plots.

3. Results

3.1. Generation of the Cellular Model System

Since we have previously shown that the A375 melanoma cell line is a good model system for studying the effects of ERAP1 on the immunopeptidome [18, 20, 29], we used A375 ERAP1 KO cells to generate mono‐allelic ERAP1 clones. To achieve this, A375 ERAP1 KO clone 1B12 [18] was transfected with linearised pcDNA3.1(+) vectors carrying either ERAP1 allotype 2 or allotype 10 and treated with G418 to eliminate untransfected cells. Single clones were isolated and expanded, and western blots were used to evaluate ERAP1 expression levels (Figure S2). To control for potential differences in enzymatic activity due to ERAP1 expression levels, we selected clones 2B1 (allotype 2) and 1H1 (allotype 10), which exhibited comparable ERAP1 expression, for further experiments (Figure 1B,C).

FIGURE 1.

Effects of IFN‐γ on cell‐surface HLA expression and ERAP1 expression. Panel A, flow cytometry analysis of A375 single ERAP1 allotype clones showing cell‐surface HLA‐I levels detected by the W6/32 antibody after incubation with increasing amounts of IFN‐γ. Shown are means ± SDs from 2 technical replicates. Panel B, western blot analysis of ERAP1 from whole cell lysates in the presence and absence of IFN‐γ. Panel C, quantification of ERAP1 expression from data shown in panel B, normalised for GAPDH.

Antigen presentation in cancer and autoimmunity often operates in the context of inflammation [13, 57]. In this context, cytokines such as interferon‐gamma (IFN‐γ) can upregulate several components of the antigen presentation machinery, including HLA and ERAP1 [58], thereby enhancing adaptive immune responses. To simulate inflammatory conditions, we treated the two A375 ERAP1 clones with IFN‐γ, optimising the concentration by exposing cells to 0, 10, 20, and 50 ng/mL and monitoring HLA‐I surface expression using flow cytometry (Figure 1A). IFN‐γ treatment led to the upregulation of cell surface HLA‐I that plateaued at concentrations over 20 ng/mL. As a result, we selected this concentration for immunopeptidome analysis. Cells were also tested for any changes in ERAP1 expression after IFN‐γ treatment, since differential upregulation of ERAP1 allotypes would bias the immunopeptidome analysis results. Western blot analysis indicated that ERAP1 expression was not affected by IFN‐γ treatment, as expected, given that the integration of the ERAP1 gene after transfection is unlikely to occur in genomic regions under IFN‐γ control (Figure 1B,C).

3.2. ERAP1 Allotypes Generate Distinct Immunopeptidomes

To analyse the immunopeptidome of the two A375 clones, cells were grown in three separate biological replicates and were treated with 20 ng/mL IFN‐γ for 24 h. HLA‐I‐peptide complexes were then isolated by affinity chromatography from 2.5 × 10⁸ cells using the W6/32 antibody, as previously described [18]. Eluted peptides from HLA‐I complexes were sequenced by LC–MS/MS using both data‐dependent and data‐independent acquisition. For each biological replicate, two technical DIA replicates were acquired, totaling 6 DIA replicates per clone.

Obtained MS spectra were then searched against the UniProt database (HUMAN_UP000005640_9606, 20597 entries, retrieved: 24.11.2023) using the Fragpipe platform [46]. Searching the DIA spectra with the hybrid spectral library built with both DDA and DIA runs resulted in the identification of 2408 peptides between 7 and 25 amino acids (Table S1), 2325 of which were in the range of 8–14 amino acids, which is normally expected for HLA‐I ligands (Figure S3).

Principle component analysis of the peptides between 8 and 14 amino acids long indicated that the immunopeptidomes of the two clones formed distinct clusters with wide separation across principal component 1 (Figure 2A). This observation was further supported by hierarchical clustering analysis, where the two experimental conditions formed distinct clusters due to both qualitative and quantitative differences (Figure 2B). Thus, this preliminary analysis suggested that the two ERAP1 allotypes produced distinct immunopeptidomes.

FIGURE 2.

Immunopeptidome analysis of A375 cells for peptides with length between 8 and 14 amino acids. Panel A, principal component analysis of the immunopeptidome isolated from the two clones carrying ERAP1 allotypes 2 and 10 (3 biological replicates, 2 technical replicates each). Panel B, heatmap of cluster analysis of identified peptides for both experimental conditions and replicates. Colour scale indicates peptide intensities (red = high, blue = low).

3.3. Immunopeptidome Shifts Are Primarily Quantitative in Nature

Comparison of the identified peptide sequences between 8 and 14 amino acids indicated that most peptides were shared between the two experimental conditions: 2162 peptides were common in both cell clones carrying the two ERAP1 allotypes, 130 were unique to allotype 2 and 33 to allotype 10 (Figure 3A). Notably, one of the unique allotype 2 peptides was KLGITMTV, which was mapped to ERAP1 and contains polymorphic position 349, where M is present in allotype 2 but replaced by V in allotype 10. The large overlap between the two experimental conditions was in stark contrast to previous analysis with the same cell line using ERAP1 KO or ERAP1 inhibitors that had revealed major changes in the sequences of detected peptides [20, 29]. Statistical analysis, however, revealed significant peptide abundance changes in the two conditions (Figure 3B). Specifically, 443 peptides were upregulated in allotype 2 and 405 peptides were upregulated in allotype 10. Combining the differentially upregulated with the unique peptides for each allotype, we calculated that of the total 2325 peptides identified, 573 peptides (24.6%) were upregulated in allotype 2 and 438 peptides (18.8%) were upregulated in allotype 10 (Figure 3C). These results indicate a potentially substantial quantitative immunopeptidome shift associated with different ERAP1 allotypes.

FIGURE 3.

Quantification of immunopeptidome shifts between cells carrying ERAP1 allotype 2 or 10 for peptides between 8 and 14 amino acids long. Panel A, Venn diagram depicting the overlap of identified peptide sequences between the two experimental conditions. Panel B, volcano plot showing differences in peptide abundance related to the statistical significance of the observation. Each dot corresponds to a unique peptide sequence. Peptide groups that show statistically significant differences between allotypes 2 and 10 are highlighted with coloured boxes. Panel C, Venn diagram depicting the immunopeptidome shift between cells carrying ERAP1 allotype 2 or 10, including both uniquely detected peptides and statistically significant abundance changes.

3.4. Presented Peptides Have Different Binding Affinity, Length Distribution and HLA Allele Preferences

To better understand the differences between peptides presented in each of the two immunopeptidomes, we used MHCMotifDecon 1.0 [49] to analyse all the identified peptides between 8 and 14 amino acids. 94.5% of the peptides (2197, Figure S3) were predicted to bind to one of the HLA alleles expressed in A375 cells (rank ≤ 2, NetMHCPan 4.1 [50]) and corresponded to expected motifs for the respective HLA, according to the MHC motif atlas database [59] (Figure 4). Surprisingly, we identified only a small number of peptides with motifs suitable for binding to HLA‐C*06:02, a major predisposition allele for psoriasis, and no peptides from the ADAMTSL5 autoantigen previously linked with CD8+ responses in other melanocyte cell lines [13]. The average predicted affinity was found to be better for the peptides that were unique or significantly upregulated in cells carrying ERAP1 allotype 10 (Figure 5A). Given that allotype 10 has been reported to have lower enzymatic activity, this finding may underscore the epitope‐destructive role of ERAP1 for the HLA alleles carried by this cell line.

FIGURE 4.

Motifs of identified peptides. Top, sequence motifs of identified peptides assigned to one of the HLA alleles carried by A375 cells. The number of peptides used to derive each motif is indicated in parenthesis. Bottom, expected sequence motifs for each HLA allele present in A375 cells, according to the MHC motif atlas database.

FIGURE 5.

Characteristics of identified 8–14 mer peptides from cell lines expressing different ERAP1 allotypes. Panel A, predicted binding affinity for at least one of the HLA alleles carried by A375 cells for peptides uniquely associated with or significantly upregulated in each of the two ERAP1 allotypes. Each dot corresponds to a different peptide sequence. Peptides below the dotted line are considered binders (Rank ≤ 2). The geometric mean of predicted rank is indicated, along with the 95% confidence interval. Statistical significance was evaluated with a Mann–Whitney test in GraphPad Prism v8. Panel B, length distribution of peptides uniquely associated with or differentially upregulated in the two ERAP1 allotypes. Statistical significance was evaluated with a proportion test with Bonferroni correction for multiple testing. Panel C, binding assignments to HLA alleles carried by A375 cells for peptides uniquely associated with or differentially upregulated in the two ERAP1 allotypes. Statistical significance was evaluated with a proportion test with Bonferroni correction for multiple testing. p < 0.001 (***), p < 0.0001 (****).

Length is another critical property for binding to MHC‐I, since the peptide binding groove can optimally accommodate only antigenic peptides of specific lengths, usually 9mers. Consistent with ERAP1's peptide trimming function, lack of its activity generally results in the presentation of longer peptides [18, 20, 29, 44]. In both experimental conditions, most peptides identified were 9mers, in line with the length preferences of all HLA alleles found in A375 cells (Figure 5B). Still, some notable differences were evident: (i) allotype 10 resulted in the presentation of more 10mers and 11mers and (ii) allotype 2 resulted in the presentation of more 8mers. While both of those observations suggest a lower enzymatic activity for allotype 10, comparison with previous studies (Figure S4) reveals that the length distribution of the KO cells was significantly more shifted to longer peptides [18, 20], suggesting that allotype 10 retains a considerable amount of trimming activity in the cell.

In addition to differences in length and predicted affinity, the two sets of peptides differed in the distribution per HLA. Allotype 10 specific peptides, composed both of unique and significantly upregulated ones, were mainly HLA‐B*44:03 binders, while allotype 2 specific peptides were mostly HLA‐A binders (Figure 5C), suggesting that ERAP1 allotypes can preferentially generate or destroy peptides that bind to specific HLA alleles.

Given the differences between HLA alleles, we further explored this phenomenon by analysing the predicted affinity and length distribution per HLA (Figure 6). While the tendency for larger peptides for allotype 10 appears to apply only for HLA‐A and HLA‐B alleles (Figure 6D–F,J), the smaller predicted affinity for allotype 2 was significant for A*01:01, A*02:01, B*44:03, C*06:02, and C*16:01 (Figures 6A and 6B,C,H,I). Interestingly, allotype 2 generated more 8mers and fewer 9mers than allotype 10 for HLA‐B*44:03 (Figure 6F) and HLA‐C alleles (Figure 6K,L). The same trend was observed for HLA‐A*01:01, although the results were non‐significant (Figure 6D). Peptides predicted to bind HLA‐C alleles also showed the largest difference in affinity between ERAP1 allotypes (Figure 6H,I), including HLA‐C*06:02, which has been previously associated with T cell responses in psoriasis [13, 60].

FIGURE 6.

Predicted affinity and length distribution per HLA allele for identified 8–14 mer peptides. Each identified peptide was assigned to one of the HLA alleles carried by the cells based on binding predictions. Panels A–C and G–I, comparison in predicted affinity for each HLA allele. Statistical significance was evaluated with a Mann–Whitney test in GraphPad Prism v8. Panels D–F and J–L, length distributions of peptides per HLA allele. Statistical significance was evaluated with Fisher's exact (counts < 5) or proportion test with Bonferroni correction for multiple testing. p < 0.05 (*), p < 0.01 (**), p < 0.001 (***), p < 0.0001 (****).

3.5. The ERAP1 Allotype 2 and 10 Derived Immunopeptidomes Have Differences in Sequence Motifs That May Underlie Differences in Immunogenicity

Changes in the sequences of presented peptides between cells expressing different ERAP1 allotypes can result in changes in immunogenicity. To explore this, we compared the sequence motifs per HLA allele for the peptides that were unique or upregulated for each ERAP1 allotype (Figure 7A). While, as expected, most motifs are dominated by the anchor residues that are recognised by each HLA allele, some differences in non‐anchor residue positions were evident. Specifically for A*02:01 and allotype 10, position 4 was enriched in amino acids with negatively charged side chains (D and E), which was statistically significant for E (Figure 7B). This was also evident, albeit less intense for C*06:02. In addition, for allotype 2 peptides presented by several alleles, positions 7 and 8 were enriched in positively charged amino acids (R and K) (Figure 7A,C–E). These observations suggest that functional differences between ERAP1 allotypes may shape the nature of non‐anchor residues of presented peptides. These non‐anchor residues are often key for interactions with the T‐cell receptor [61] and inhibitory NK receptors (KIRs) [62] and can determine the antigenicity of the presented peptides.

FIGURE 7.

Panel A, sequence motifs per HLA allele for peptides unique or upregulated for each ERAP1 allotype. The number of peptides per motif is indicated in parentheses. Observed differences in motifs are circled. Panels B–E, comparison of amino acid distributions between allotype 2 and allotype 10 cells in P4 for HLA‐A*02:01 (Panel B), P7 for HLA‐B*44:03 (Panel C), and P8 for HLA‐A*01:01 (Panel D) and HLA‐B*44:03 (Panel E). Statistical significance was evaluated with Fisher's exact (counts < 5) or proportion test with Bonferroni correction for multiple testing. Panel F, sequence motifs of suspected psoriasis autoantigens from the IEDB database.

Apart from the autoantigenic ADAMTSL5 epitope [13], several other suspected psoriasis autoantigens exist but were not discovered in our immunopeptidomic analysis. Nevertheless, to examine the potential role of ERAP1 allotypes in the generation of autoantigenic motifs, we downloaded suspected psoriasis autoantigens from the Immune Epitope Database (IEDB) [63] and used MHCMotifDecon [49] to extract sequence motifs. Most of the 63 peptide sequences tested were predicted to bind to HLA‐A*02:01 and HLA‐C*06:02. Consequently, MHCMotifDecon generated sequence logos for these alleles (Figure 7F). This was consistent with the well‐established association of HLA‐C*06:02 with psoriasis and the reported link between HLA‐A*02:01 and psoriasis in Caucasians [64]. In the motifs derived from the suspected psoriasis autoantigens, no D/E amino acid at position 4 was prominent. On the contrary, R/K amino acids at position 7 were common in the HLA‐C*06:02 allele. These results suggest that negatively charged residues at position 4 may be less immunogenic in psoriasis, whereas the presence of positively charged residues at positions 7 and 8 may enhance immunogenicity. These findings have very good correlation with the peptide motifs upregulated by allotype 10 and 2, respectively, suggesting that differences in psoriasis predisposition depending on ERAP1 allotypic variation are driven by the generation and presentation of sets of peptides with different immunogenicity.

3.6. Nonmetric Multidimensional Scaling Analysis Reveals Additional Differences in Anchor Residues due to ERAP1 Allotypic Variability

Given the observations in motifs described above and to discern motif differences in a more unbiased manner, we used nonmetric multidimensional scaling (NMDS) to project the 9mer peptides in two‐dimensional space based on sequence similarity, as shown before [20]. NMDS is a dimensionality‐reduction method that focuses on global repertoire similarity and is not biased by previous knowledge of allele specificity and known sequence motifs, thus allowing the discovery of unexpected relationships within the samples. The analysis was performed for all HLA alleles separately, but only the HLA alleles with the largest number of identified peptides, namely HLA‐A*02:01 and HLA‐B*44:03, resulted in significant differences between allotypes, as expected due to higher statistical power.

For HLA‐A*02:01 binders, analysis resulted in 9 sequence clusters (Figure 8A), three of which were significantly overrepresented by peptides derived from different ERAP1 allotypes (Figure 8B,C). Clusters 1 and 3 were mainly composed of peptides originating from the allotype 2‐carrying cells and featured hydrophobic residues at position 6. Moreover, cluster 3 featured primarily an aspartic acid at position 4 (Figure 8D). In contrast, cluster 8 was mainly composed of peptides derived from allotype 10‐carrying cells and featured an acidic glutamate residue at position 4. Moreover, allotype 2 appeared to favour alanine residues at position 9, while allotype 10 favoured valine. While valine is a typical anchor residue for HLA‐A*02:01, alanine is not. This observation is consistent with the notion that allotype 2 can over‐trim and destroy some optimal peptides, leading to lower overall binding affinity and shorter peptide lengths (8mers).

FIGURE 8.

Nonmetric multidimensional scaling analysis of peptides predicted to bind to HLA‐A*02:01 (Eps = 0.0066). Panel A, clustering of peptide sequences in two dimensions. Panel B, comparison on clustering distributions separated per allotype (allotype 2 in cyan, allotype 10 in orange). Panel C, comparison of peptide distribution per cluster between the two ERAP1 allotypes. Statistical significance was evaluated with Fisher's exact or proportion test with Bonferroni correction for multiple testing [p < 0.05 (*)]. Panel D, sequence motifs belonging to clusters found to be significantly enriched in peptides from one of the two different allotypes in panel C. The allotype in which each cluster was found to be more common is shown.

For HLA‐B*44:03 binders, analysis revealed 7 sequence clusters (Figure 9A), four of which were overpopulated by peptides derived from cells containing either allotype 2 or 10 (Figure 9B,C). Clusters 1 and 3 were significantly enriched in peptides from the allotype 2‐carrying cells and featured an aromatic residue at position 9 (Figure 9D). In contrast, clusters 2 and 5 were enriched in peptides from cells carrying allotype 10 and featured a small hydrophobic residue at position 9 (Figure 9D). Taken as a whole, NMDS analysis revealed that ERAP1 allotypes can also influence the nature of anchor residues.

FIGURE 9.

Nonmetric multidimensional scaling analysis of peptides predicted to bind to HLA‐B*44:03 (Eps = 0.0052). Panel A, clustering of peptide sequences in two‐dimensions. Panel B, comparison on clustering distributions separated per allotype (allotype 2 in cyan, allotype 10 in orange). Panel C, comparison of peptide distribution per cluster between the two ERAP1 allotypes. Statistical significance was evaluated with Fisher's exact or proportion test with Bonferroni correction for multiple testing [p < 0.05 (*)]. Panel D, sequence motifs belonging to clusters found to be significantly enriched in peptides from one of the two different allotypes in panel C. The allotype in which each cluster was found to be more common is indicated.

3.7. ERAP1 Allotypes Influence the Presentation of Melanoma Associated Antigens

Besides autoimmunity, enzymatic activity and polymorphic variability of ERAP1 have been shown to affect the antigenicity of cancer cells in multiple studies [23, 65]. In particular, immunogenicity towards melanoma cells is often attributed to the presentation of a set of melanoma‐associated antigens (MAGE). To examine if ERAP1 allotypes can affect the presentation of MAGE, we compared the abundance of known MAGE across the two A375 clones (Figure 10). We identified nine antigenic peptides to be upregulated by allotype 2 and four antigenic peptides to be upregulated by allotype 10. Interestingly, some of the peptides were markedly upregulated by allotype 2, such as MAGE A10, which was upregulated by 10‐fold. This finding further supports the notion that the two ERAP1 allotypes may induce a distinct immunogenic profile, which may be relevant also in the context of cancer.

FIGURE 10.

Effect of allotype 2 on the presentation of MAGE‐derived antigenic peptides. Log₂ fold changes in peptide abundance for identified MAGE‐derived antigenic peptides are shown. Each peptide sequence and the respective MAGE antigen are indicated.

4. Discussion

Several studies have established ERAP1 activity as a key regulator of the cellular immunopeptidome [27, 66], although the extent of the regulation appears to vary among the cellular systems examined. Functional ERAP1 SNPs have also been associated with immunopeptidome shifts, although comparisons between different cell lines have sometimes complicated interpretation [45, 67]. These SNPs mainly exist as specific allotypes in the human population, and their functional effects often synergise so that allotypes present significantly distinct enzymatic functions. This variability has been associated both with cancer and autoimmunity. Therefore, examining the effect of allotypic variability on immunopeptidome generation is of major importance [40, 41, 43].

Among the most common ERAP1 allotypes, allotype 10, the most common in Europeans [40], was recently demonstrated to downregulate autoreactive immune responses in psoriasis. This was attributed to the inefficient trimming of the ADAMTSL5 antigen, which was however efficiently produced by allotype 2 [13, 60]. At the same time, in the context of another MHC‐I‐opathy called Behçet's disease, allotype 10 enhances autoimmune responses, potentially due to the generation of HLA‐B51 restricted peptides that alter the immunodominance of subsequent CD8+ T‐cell responses [44]. Although initial functional assays marked allotype 10 as a loss‐of‐function variant [42], follow‐up in vitro analysis by our group suggested more complex changes in activity [41, 43], necessitating deeper analyses.

Our findings here demonstrate a distinct effect by allotypes 2 and 10 on the immunopeptidome that is not consistent with the perception of allotype 10 as a loss‐of‐function ERAP1 variant. The generation of optimal length 9mers is virtually identical for both allotypes, and most changes revolve around either the presentation of more 8mers by allotype 2 or more 10mers and 11mers by allotype 10. In addition, the shift in uniquely presented peptides is quite small since 93% of the detected peptides are common between the two allotypes, a pattern quite distinct from previous comparisons with inhibited or knocked‐out ERAP1 [20, 29]. Rather, the differences focus on the relative abundance of identified peptides, in which case the immunopeptidome shift affects 36.5% of the peptides. These findings are more consistent with two enzyme variants that differ in potency and specificity rather than comparing an active and an inactive variant.

Generation of shorter peptides and destruction of strong binders by over‐trimming, observed in allotype 2, may enhance immunogenicity by promoting the dissociation of weaker binding peptides from the cell surface [68], thereby either allowing the capture of unedited extracellular peptides or triggering the activation of Natural Killer cells [31]. In parallel, allotype 2‐carrying cells were more effective in presenting immunogenic MAGE peptides, some up to 10‐fold over allotype 10. Given that antigenic peptide abundance may be an important factor in determining adaptive immune responses [69], the findings of this study highlight the potential of ERAP1 allotypic variability to drive differences in immunogenicity. Moreover, the extensive changes in the presented peptide repertoire shift the focus from specific antigens to broader immunopeptidome changes. These immunopeptidome shifts result in specific changes in peptide motifs that include both anchor residues, which can affect peptide affinity and binding kinetics, as well as non‐anchor residues, which can affect interactions with T cell or Natural Killer cell receptors.

While the impact of these broad immunopeptidome shifts on adaptive immune responses needs to be addressed in future studies, it could be highly relevant in the context of established phenomena such as antigenic drift and T‐cell receptor polyspecificity [70, 71]. In addition, broad immunopeptidome shifts may affect Natural Killer cell activity, as these cells show a much broader specificity than T cells [62]. Indeed, an HLA tetramer carrying psoriasis‐associated antigenic peptides was recently reported to stain CD56+ natural killer cells or KIR2DL1/2DS1 receptors, providing initial evidence for the involvement of NK cells in this autoimmune condition [71]. Interestingly, the side chains at positions 7 and 8 of the presented peptides, and in particular residues R or K, can enhance Natural Killer cell activation. This pattern is similar to the allotype 2 motifs detected in our study. Moreover, ERAP1 has been reported to regulate NK cell responses in cancer [19, 23, 31] and to specifically regulate KIR receptors engaging HLA‐B alleles [72, 73]—which we found to be differentially affected by the two allotypes.

While our study defines a framework for understanding the implications of ERAP1 allotypic variation in autoimmunity and anti‐tumour responses, it carries some important limitations that should be weighed when interpreting results. Although this study was in part inspired by findings on ERAP1‐dependent processing of the ADAMTSL5 antigen in psoriasis, the lack of detection of this peptide and the identification of a low number of HLA‐C*06:02 binders, even in the presence of IFN‐γ, limit our ability to correlate our findings to that autoimmune condition. While this may be an inherent limitation of the A375 cell line, our findings can still contribute to our understanding of HLA‐associated autoimmunity and cancer. Further studies utilising cellular models with different HLA combinations may be necessary to clarify the roles of specific ERAP1 allotypes in different autoimmune conditions.

Overall, our results demonstrate a good level of translation of in vitro enzymatic properties of ERAP1 to the cellular immunopeptidome, despite the dominant motif‐filtering by HLA alleles. Future studies aiming to test the immune effects of these immunopeptidome shifts, including the role of ERAP1 allotypes in disease‐specific cells, will be necessary to establish the importance of the observations shown here on the pathogenesis of HLA‐associated autoimmunity as well as on anti‐tumour immunity.

Author Contributions

Martha Nikopaschou: conceptualisation, methodology, investigation, data curation, formal analysis, visualisation, writing – original draft, writing – review and editing. Martina Samiotaki: methodology, investigation, data curation (mass spectrometry). Anna Kannavou: investigation (monoclonal antibody production). Nikolaos V. Angelis: methodology, investigation (flow cytometry). Ourania Tsitsilonis: supervision, resources (flow cytometry). George Panayotou: resources, methodology (mass spectrometry). Efstratios Stratikos: conceptualisation, methodology, formal analysis, supervision, resources, project administration, writing – original draft, writing – review and editing, funding acquisition.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Data S1: Supporting Information.

Table S1: Overview of analysed samples, peptide identifications, and differential analysis results. Sheet 1 contains the sample identifiers used in the study. Sheet 2 lists all the identified peptides between 7 and 25 amino acids. Sheet 3 provides the results of the peptide‐level differential expression analysis for peptides between 8 and 14 amino acids.

Nikopaschou M., Samiotaki M., Kannavou A., et al., “ ERAP1 Allotypes 2 and 10 Differentially Regulate the Immunopeptidome of Melanocytes,” Immunology 177, no. 2 (2026): 398–412, 10.1111/imm.70056.

Data Availability Statement

All the data described are available in the article and associated Supporting Information. Numerical values used for the generation of graphs are available upon request to the corresponding author (Efstratios Stratikos; E‐mail: estratikos@chem.uoa.gr or stratos@rrp.demokritos.gr). The MS proteomics raw data have been deposited to the ProteomeXchange Consortium via the PRIDE [74] partner repository with the dataset identifier PXD066752 (http://www.ebi.ac.uk/pride/archive/).