Large-Scale Immunopeptidome Analysis Reveals Recurrent Posttranslational Splicing of Cancer- and Immune-Associated Genes

Abstract

Posttranslational spliced peptides (PTSPs) are a unique class of peptides that have been found to be presented by HLA class-I molecules in cancer. Thus far, no consensus has been reached on the proportion of PTSPs in the immunopeptidome, with estimates ranging from 2% to as high as 45% and stirring significant debate. Furthermore, the role of the HLA class-II pathway in PTSP presentation has been studied only in diabetes. Here, we exploit our large-scale cancer peptidomics database and our newly devised pipeline for filtering spliced peptide predictions to identify recurring spliced peptides, both for HLA class-I and class-II complexes. Our results indicate that HLA class-I-spliced peptides account for a low percentage of the immunopeptidome (less than 3.1%) yet are larger in number relative to other types of identified aberrant peptides. Therefore, spliced peptides significantly contribute to the repertoire of presented peptides in cancer cells. In addition, we identified HLA class-II-bound spliced peptides, but to a lower extent (less than 0.5%). The identified spliced peptides include cancer- and immune-associated genes, such as the MITF oncogene, DAPK1 tumor suppressor, and HLA-E, which were validated using synthetic peptides. The potential immunogenicity of the DAPK1- and HLA-E-derived PTSPs was also confirmed. In addition, a reanalysis of our published mouse single-cell clone immunopeptidome dataset showed that most of the spliced peptides were found repeatedly in a large number of the single-cell clones. Establishing a novel search-scheme for the discovery and evaluation of recurring PTSPs among cancer patients may assist in identifying potential novel targets for immunotherapy.

Keywords: HLA class-I; HLA class-II; immunopeptidomics; neoantigens; posttranslational spliced peptides.

Graphical abstract

Fig. 1

Pipeline design. The full list of peptidomic samples derived from cell lines and patients are inputted to Neo-Fusion (separately for HLA class-I and HLA class-II). Uniprot’s WT proteome-containing proteins and isoforms are used as a sequence database. PTSPs identified by Neo-Fusion are subsequently inputted as a sequence database, together with the Uniprot WT proteome, for a MaxQuant search against the same peptidomics samples. The pipeline stringently eliminates possible false-positives based on multiple criteria. A large-scale in-lab peptidomics database is exploited to identify recurrent PTSPs. PTSP, posttranslational spliced peptide.

Fig. 2

Comparison of PTSP versus WT cumulative filtration of PSMs in the pipeline. Both HLA class-I and class-II show a significant decrease in the NetMHCpan and Delta score filtration steps for PTSPs (dark-blue and light-blue lines, respectively). By contrast, WT peptides decrease only mildly (dashed lines). This is due to the significant peptide ambiguity observed only in PTSPs. Furthermore, filtration based on RNA-seq results in a large reduction mainly in WT peptides (consistently in class-I and class-II). This demonstrates that most of the peptides identified in the RNA-seq transcripts are WT (and not, e.g., PTSPs as a result of splice-junctions). HLA class-I and class-II stop declining significantly following the Delta score filtration step. This demonstrates that the final filtration steps remove mainly peptides already removed in previous steps. PSM, peptide spectrum match; PTSP, posttranslational spliced peptide.

Fig. 3

DAPK1- and HLA-E–derived peptides show potential immunogenicity. T cells derived from healthy donor PBMCs were stimulated and expanded for 24 days. Following coculturing with human B-LCL 721.221 expressing HLA-A∗02:01/HLA-A∗01:01 alleles that were pulsed with 10 μl/ml peptide (right) or no peptide (control; left), T cells were evaluated for activation markers by flow cytometry analysis. IFN-γ, TNF-α secretion, and 4-1BB expression by T cells stimulated with (A) DAPK1 (LLLDKLLSV) or (B) HLA-E (WSDSSGGKGGSY). The images are representative of ≥2 replicates. PBMC, peripheral blood mononuclear cell.

Fig. 4

HLA-I dextramer staining of DAPK1-specific CD8 T cells. T cells derived from healthy donor PBMCs were stimulated with 721.221 B-LCLs expressing the HLA-A∗02:01 allele pulsed with 10 μg/ml DAPK1 or 10 μg/ml viral peptide (control). Following expansion, T cells were stained with DAPK1 dextramers to evaluate the percentage of DAPK1-specific subpopulations. CD8 T cells were gated as single, live, CD3 positive and CD8 positive cells. The images are representative of two replicates. PBMC, peripheral blood mononuclear cell.

Supplemental Figure S3

Proportion of Spliced HLA-presented peptides in HLA-class-I and -II. The fraction of HLA-class-I spliced peptides detected (3.1%) is much smaller than that previously detected by Liepe et al. (Science 2016) (30% of HLA-class-I immunopeptidome). Both HLA-class-I and -II spliced peptides are identified, but to different extents.

Supplemental Figure S4

Peptide splicing occurs across multiple cancer types. Melanoma, breast and head and neck squamous cancer cells all present spliced peptides to the immune system.

Supplemental Figure S7

TIL reactivity toward spliced peptides. 4-1BB expression was detected in TILs following 24 h stimulation with 1 μg/ml DAPK1 (LLLDKLLSV), HLA-E (WSDSSGGKGGSY), NRAS Q61K (ILDTAGKEEY) or dimethyl sulfoxide (control), using flow cytometry analysis. T cell reactivity was validated by ≥2 independent experiments.

Supplemental Figure S8

Flow cytometry gating strategy of reactive CD8 T cells derived from TILs.

Supplemental Figure S9

Flow cytometry gating strategy of reactive CD8 T cells derived from healthy donor PBMCs.

Supplemental Figure S10

CD8 T cells derived from PBMCs demonstrated potential immunogenicity against a mixture of viral peptides and served as a positive control (A) IFN-γ (B) TNF-α (C) 4-1BB. CD8 T cell reactivity was analyzed by flow cytometry. The images are representative of ≥2 replicates.