To investigate the peptides which are presented in the thirteen most frequently expressed HLA-DP molecules with more than 1% population frequency in the Netherlands, HLA class I and II–negative human immortalised myelogenous leukemia (K562) cell lines were transduced with individual HLA-DP alleles to obtain thirteen different K562 cell lines. The β-chain and the corresponding α-chain (Supplementary Table 1) were introduced using pLZRS vector HLA-DPA1-T2A-HLA-DPB1-IRES-DNGF-R combinations or as separate chains using MP71 DPA1-IRES-GFP and MP71 DPB1-IRES-DGFR vectors. The HLA-DP-expressing K562 were enriched by flow cytometric cell sorting based on expression of the mutated nerve growth factor receptor (DNGF-R) or green fluorescent protein (GFP) marker and subsequently the expression of the correct HLA-DP alleles was confirmed by sequencing (data not shown). The thirteen K562 cell lines were expanded in Iscove’s modified Dulbecco’s medium (IMDM; Lonza) supplemented with 10% heat-inactivated fetal calf serum (FCS; Gibco). The expanded K562 cell lines were cryopreserved for later use in functional T-cell reactivity assays and 2-8 x 10⁹ cells were stored as dry cell pellets at -80˚C for peptide elution experiments (37).

The hybridoma cell line B7/21 was expanded in Corning hybridigo SF medium to produce anti-HLA-DP antibodies (Abs). Protein A Sepharose beads (GE Healthcare) were used to purify the Abs and generate an immunoaffinity column (B7/21–Protein A Sepharose 2.5 mg/ml) (38). The cell pellets of the different HLA-DP-transduced K562 cell lines were lysed in 50 mM Tris–HCl (pH 8), 150 mM NaCl, 5 mM EDTA, and 0.5% ZWITTERGENT 3–12 (Merck) and supplemented with cOmplete Protease Inhibitor (Merck). The supernatant was precleared with CL4B beads (GE Healthcare) and applied to the immunoaffinity column with a flow rate of 2.5 ml/min. The bound peptide-HLA-DP complexes were eluted with 10% acetic acid (Merck) followed by the separation of peptides from the HLA-DP molecules using a 10-kDa membrane (Merck) (37).

The obtained filtrates containing the peptides were analyzed by online C18 nano-high-performance- liquid-chromatography (HPLC) tandem mass spectrometry (MS/MS) with a system consisting of an Easy-nLC 1200 gradient HPLC system (Thermo Fisher Scientific) and an Orbitrap Fusion Lumos Mass Spectrometer (Thermo Fisher Scientific) or an Orbitrap Exploris Mass Spectrometer (Thermo Fisher Scientific). The detailed procedure from injection to elution has been described previously (37). The raw data of the different HLA-DP peptidomes were first converted to peak lists using Proteome Discoverer version 2.2 (Thermo Electron) and submitted to the Uniprot Homo sapiens database (20,410 entries) using Mascot v. 2.2.07 (http://www.matrixscience.com) for protein identification with target decoy for false discovery rate (FDR). Further analysis was done on datasets with High Confident peptides (FDR 1%), Search Engine Rank = 1 and Mascot Ion Score above 10. Mass spectrometry (MS) contaminants such as keratins were excluded from the data. Peptide annotation, grouping and alignment was done in Access, Excel and R. The High Confident K562-HLA-DP datasets were compared for overlapping peptide sequences and analyzed at different levels without and with normalization for peptide amount and length variants of the same epitope presented in HLA-DP. Normalization for peptide amount was done on the High Confident Peptide to Sequence Matches (PSMs) from Proteome Discoverer and Mascot. High abundant peptides have high numbers of PSMs and low abundant only 1 or a few. The range was 1-169 PSMs on sequence level and expressed in % of total within one sample. To exclude the effect of the different length variants of ligands presented in HLA-DP between the samples, a Region Sequence was defined with a fixed start and end position on the 10^th amino acids within a protein containing the identified peptide (37). This dataset of sequences with mainly 10, 20 and 30 amino acid lengths was also normalized on numbers of PSMs and on length variants within a Region Sequence. The Network plot to visualize the % of overlapping peptides between the samples was made in R using ggraph.

Unmanipulated peripheral blood (without previous granulocyte colony-stimulating factor stimulation) was obtained from healthy donors after informed consent. Responder CD4 T cells were isolated by magnetic activated cell sorting (MACS) from peripheral blood mononuclear cells (PBMC) as previously described (39). The CD4 T cells were stimulated with irradiated (25 Gy) HLA-DP-mismatched antigen-presenting cells (APC) in a 10 to 1 responder to stimulator cell ratio in two different in vitro T-cell responses targeting HLA-DPB1*01:01 and HLA-DPB1*03:01 or HLA-DPB1*03:01 and HLA-DPB1*04:01, respectively (HLA-DP typing of the responder and stimulator combinations is shown in Supplementary Table 2), as described previously (39). The cells were cultured in IMDM containing 10% heat-inactivated human serum (ABOS; Sanquin) supplemented with IL-7 (10 ng/ml, Miltenyi Biotec), IL-15 (0.1 ng/ml, Miltenyi Biotec) and IL-2 (50 IU/ml, Novartis Sandoz). The cell cultures were restimulated at day 14 with irradiated HLA-DP-mismatched APC. At 24-36 hours after restimulation, reactive CD4 T cells were clonally isolated by single cell flow cytometric cell sorting based on CD137 expression using FACSAria (BD Biosciences), as described previously (39). The T-cell clones were expanded for two weeks using allogeneic feeder mixture consisting of IMDM containing 5% heat-inactivated ABOS and 5% FCS supplemented with 5x irradiated (35 Gy) allogeneic feeder cells, 0.5x irradiated (60 Gy) allogeneic Epstein-Barr virus lymphoblastoid cell lines (EBV-LCL), 100 IU/ml IL-2, and 800 ng/ml phytohemagglutinin (PHA-HA16, Oxoid) (39). The clonality of the generated CD4 T-cell clones was confirmed using ARTISAN polymerase chain reaction (PCR), as previously described (40, 41). The complementarity-determining region-3 (CDR3) of the TCRβ chains were identified to confirm the presence of one TCR clonotype in each CD4 T-cell clone.

CD4 T-cell clones recognizing only stimulator cells expressing the mismatched target HLA-DP allele(s) and not stimulator cells expressing the autologous HLA-DP allele(s) were selected as being allo-HLA-DP reactive, as described previously (39, 42). The reactivity of allo-HLA-DP-reactive T-cell clones against different HLA-DP molecules was analyzed using the panel of HLA-DP-transduced K562 cell lines as stimulator cells at a 1:5 responder to stimulator ratio in IMDM/10% ABOS supplemented with 25 IU/ml IL-2. The supernatants were harvested after overnight incubation and the amounts of interferon-gamma (IFNγ) and interleukin-4 (IL-4) in the supernatants were quantified using standard enzyme-linked immunosorbent assays (ELISA; Sanquin or Invitrogen). Reactivity was defined as at least 2 times higher cytokine production compared to the non-stimulated condition.

To identify individual peptide sequences presented by different HLA-DP molecules, K562 cell lines were transduced with various HLA-DPB1 and corresponding HLA-DPA1 chains (Supplementary Table 1) and the HLA-DP peptidomes were analyzed by immunoaffinity chromatography and mass spectrometry. Identical peptide sequences with various amino acid lengths presented in different HLA-DP molecules were unraveled. To correct for an effect of length variants, peptide sequences were aligned to identify tens of amino-acid long region sequences to include distinct peptides only once in the analyses. Peptides encoded by the gene itgam eluted from HLA-DPB1*09:01, HLA-DPB1*05:01 and HLA-DPB1*11:01 are shown as examples of correction for length variants by identifying the region sequences (Table 1). The peptidome analyses resulted in the detection of a median of 3131 peptide sequences per HLA-DP molecule (range 2172-4449), including length variants, and a median of 1566 distinct region sequences per HLA-DP molecule (range 1207-2168).

In Figure 1, the peptide sequences shared by the different HLA-DP molecules are shown. In the dark green boxes at the diagonal, the total number of sequences per HLA-DP molecule are depicted. The numbers of peptides shared between different HLA-DP molecules and the level of overlap in presented peptides is illustrated by the numbers in the heatmap and by coloring of the boxes, with white to yellow indicating no/limited overlap and green indicating relatively abundant overlap. The heatmap shows that the level of overlap in presented peptide sequences was most prominent within the two most distinct groups DPC-1 in the upper left corner and DPC-3 in the lower right corner. The differences between groups DPC-1 and DPC-3 are reflected by the minimal overlap in peptide sequences between these groups. Within group DPC-2, the HLA-DP molecules appear to segregate into two subgroups based on the overlap in presented peptides, with subgroup 2A being comprised of HLA-DPB1*13:01, HLA-DPB1*11:01, and HLA-DPB1*01:01 and subgroup 2B containing HLA-DPB1*03:01 and HLA-DPB1*06:01. In Supplementary Figure 1 we included the peptide-binding motifs of HLA-DPB1*11:01 and the other DPC-2 molecules to demonstrate that the DPC-2 molecules could indeed be separated into two subgroups. The heatmap also shows that HLA-DP molecules from DPC-1 and DPC-2 share substantial numbers of presented peptides. HLA-DPB1*05:01 shows minimal overlap with the HLA-DP molecules from the other DPC-groups. HLA-DP molecules from group DPC-3 shared the lowest numbers of peptides with other molecules. When the length variants were removed, and the data were normalized to region sequences, the same patterns were observed (Supplementary Figure 2).

To better visualize the similarities in peptides presented in the different HLA-DP molecules, percentages of shared peptides were plotted in a network graph (Figure 2). The level of overlap in presented peptides between the different HLA-DP molecules are illustrated by the thickness and color of the lines. Like the heatmap figure, the network plot indicates a distinction of four groups, with HLA-DPB1*05:01 clustered separate from the DPC groups due to very limited overlap in presented peptides with HLA-DP molecules from the DPC groups. The HLA-DP molecules within groups DPC-1 and DPC-3 share a high number of presented peptides as indicated by the thick and black/dark-blue lines. Within group DPC-1 >20% of presented peptides were shared between the HLA-DP molecules and within group DPC-3 the overlap in presented peptides was 10-40% between the HLA-DP molecules. Again, DPC-2 separated into groups DPC-2A and DPC-2B. All of the analyzed HLA-DP molecules could be attributed to one of the four groups in the plot, except for HLA-DPB1*05:01 that only showed limited overlap of peptides (10-20%) with HLA-DPB1*14:01 but minor overlap (<10%) with the other molecules from DPC-1 or from the other DPC groups. This network plot also shows that whereas DPC-3 has hardly any overlap with molecules from other DPC groups, DPC groups 1 and 2 share between 2-10% peptide sequences between the various HLA-DP molecules. The network plot based on region sequences demonstrated similar categorizations and connections (Supplementary Figure 3).

To search for biological relevance of the overlap in peptides presented in different HLA-DP molecules within or outside different DPC groups, the reactivity of several in vitro generated allo-HLA-DP-reactive T-cell clones was tested against the panel of HLA-DP-transduced K562 cell lines. The allo-HLA-DP-reactive T-cell clones were derived from two in vitro T-cell responses targeting HLA-DPB1*01:01 and HLA-DPB1*03:01 or HLA-DPB1*03:01 and HLA-DPB1*04:01, respectively, as described previously (39). HLA-DP typing of the responder and stimulator combinations is shown in Supplementary Table 2. Large numbers of allo-HLA-DP-reactive T-cell clones (170 from response 1 and 440 from response 2) were identified previously (39). In-depth analyses of allo-HLA-DP-directed cross-reactivity using the panel of HLA-DP-transduced K562 cell lines was performed with a selected number of allo-HLA-DP reactive T-cell clones (16 from response 1 and 19 from response 2) with consistent reactivity and sufficient proliferative capacity. The clonality and the diversity of the TCR usage of the allo-HLA-DP-reactive T-cell clones included in this study were confirmed by sequencing their CDR3-beta(β) regions. Based on this analysis 7 unique T-cell clones with their single CDR3-β sequence (3 from response 1 and 4 from response 2) were identified (Table 2).

In response to stimulation of these 7 unique HLA-DP-reactive T-cell clones with the panel of HLA-DP-transduced K562 cell lines, different reactivity patterns were observed. The levels of IFN-gamma/IL-4 production upon cross-reactivity against K562 cells expressing additional allogeneic HLA-DP-molecules were normalized for the level of IFN-gamma/IL-4 production by the T-cell clones in response to stimulation with reactivity against K562 cells expressing the target allogeneic HLA-DP molecule (Figure 3). The non-normalized data are shown in Supplementary Figure 4. Figure 3A shows the T-cell clones with restricted reactivity against only the targeted mismatched HLA-DP molecule (HLA-DPB1*03:01 or HLA-DPB1*01:01) (red symbols). Figure 3B shows the T-cell clones that exerted additional cross-reactive recognition against a different HLA-DP molecule within the same DPC group. The observation was made that targeting of HLA-DPB1*04:01 was often associated with cross-reactivity against HLA-DPB1*04:02 (data not shown). In Figure 3C, reactivity of three T-cell clones is shown that were found to be cross-reactive against HLA-DP molecules from different DPC groups.

To estimate the possibility that these cross-reactive T-cell clones recognized identical peptides in different HLA-DP molecules, we searched for shared peptides and shared regions in the peptidomes. As depicted in Table 2, indeed substantial numbers of shared peptides could be identified in the different HLA-DP molecules recognized by the T-cell clones, although our analyses did not include proof that the same peptide is recognized in these various HLA-DP molecules. The results illustrate that cross-reactivities are not only directed against HLA-DP molecules from the same DPC group, but also against HLA-DP molecules from different DPC groups.