Deciphering the landscape of phosphorylated HLA-II ligands

Abstract

CD4⁺ T cell activation in infectious diseases and cancer is governed by the recognition of peptides presented on class II human leukocyte antigen (HLA-II) molecules. Therefore, HLA-II ligands represent promising targets for vaccine design and personalized cancer immunotherapy. Much work has been done to identify and predict unmodified peptides presented on HLA-II molecules. However, little is known about the presentation of phosphorylated HLA-II ligands. Here, we analyzed Mass Spectrometry HLA-II peptidomics data and identified 1,943 unique phosphorylated HLA-II ligands. This enabled us to precisely define phosphorylated binding motifs for more than 30 common HLA-II alleles and to explore various molecular properties of phosphorylated peptides. Our data were further used to develop the first predictor of phosphorylated peptide presentation on HLA-II molecules.

Keywords: Cell biology; Immunology; Omics.

Graphical abstract

Figure 1

MS-based HLA-II peptidomics identifies multiple phosphorylated HLA-II ligands (A) Representative crystal structure of HLA-DRB1∗01:01 molecule in complex with a phosphorylated peptide (PDB identification code 3L6F (Li et al., 2010)). The binding core of the peptide is shown in turquoise, the peptide flanking regions (PFR) in dark gray, the phosphorylated residue in pink, and the HLA-DR in light gray. Anchor positions P1, P4, P6, and P9 are underlined in the peptide sequence and point toward the HLA-II binding site. (B) HLA-II peptidomics MS spectra were analyzed for each sample separately to identify HLA-II ligands, including phosphorylated peptides. Phosphorylated peptides were processed for each sample by applying the HLA-II ligand predictor MixMHC2pred. (C) Distribution of Andromeda search engine peptide spectrum match scores (Peptide Score) (top) and score differences to the second-best peptide spectrum match (Delta Score) (bottom) of phosphorylated peptides assigned to HLA-II alleles (blue) and phosphorylated peptides not assigned to any allele (orange). p-values were calculated using the Kolmogorov–Smirnov test. See also Figure S1. (D) Comparison of length distribution of phosphorylated and unmodified HLA-II ligands as well as predicted phosphorylated peptides not assigned to HLA-II alleles. (E) Amount of detected phosphorylated residues per phosphorylated HLA-II ligand in the HLA-II phosphopeptidome.

Figure 2

Phosphorylated peptides bind to HLA-II molecules with specific motifs Motifs of alleles with phosphorylated peptides. For each allele, the HLA-II motif based on unmodified ligands is shown on top, and the motif of phosphorylated HLA-II ligands determined in this work is shown below. Numbers correspond to the number of peptides (unmodified peptides/all phosphorylated peptides/only phosphorylated peptides with the phosphorylated residue in the core). Phosphorylated residues are shown in pink. Canonical anchor residues (P1, P4, P6, and P9) are highlighted in turquoise on the x-axis of binding motifs of unmodified ligands. See also Figure S2 and Table S3.

Figure 3

Phosphorylated residues show some positional specificity in HLA-II ligands (A) Distribution of phosphorylated residues and total residues in the binding core vs PFRs of phosphorylated HLA-II ligands. (B) Amount of phosphorylated residues found in the first three and last three residues in PFRs (top) and distribution of phosphorylated residues within the first three positions of the N-terminus (bottom) of phosphorylated HLA-II ligands. (C) Positional distribution of phosphorylated residues in the binding core of phosphorylated HLA-II ligands. Error bars represent the residue frequencies of individual alleles. (D–F) Competitor binding assays for peptides with a phosphorylated residue at each of the different core positions (turquoise box) and without phosphorylated residue (black box). The peptide initially found by MS is marked by a pink asterisk and the core predicted by MixMHC2pred is underlined. Error bars represent the two repetitions of the binding assays. HLA-II motifs of unmodified ligands of the respective allele are shown on the left.

Figure 4

The HLA-II phosphopeptidome improves prediction of phosphorylated HLA-II ligands (A) Frequency of kinase motifs that show significant enrichment between phosphorylated and unmodified HLA-II ligands (p ≤ 0.05) in phosphorylated HLA-II peptidome (1^st bar), in unmodified HLA-II ligands (2^nd bar) and in the human phosphoproteome (3^rd bar). Kinase motifs are sorted according to the frequency in the human phosphoproteome. (B) AUC values for the leave-one-sample-out cross-validation for HLA-DR samples from (Racle et al., 2019) and (Abelin et al., 2019) and with available HLA-DR in MARIA (29 in total) for MixMCH2pred v1.3, NetMHCIIpred-4.0, and MARIA. (C) AUC values for the leave-one-sample-out cross-validation for all HLA-II samples from (Racle et al., 2019) and (Abelin et al., 2019) (59 in total) for MixMHC2pred v1.3 and NetMHCIIpan-4.0. (D) AUC values from the external validation for samples from (Khodadoust et al., 2017) for MixMHC2pred v1.3, NetMHCIIpan-4.0, and MARIA. p-values between the different predictors in (B–D) were calculated using the paired two-sided Wilcoxon signed rank-test. See also Figure S4.