Preprint Club
A cross-institutional Journal Club Initiative

Keywords:

Immunopeptidomics, CAR T cells, Tumor Immunotherapy

Main Findings

A critical challenge in T cell-based cancer immunotherapy is the scarcity of truly tumor-specific antigens that are both safely targetable and broadly expressed. In a recent preprint, Li et al. present a comprehensive antigen atlas constructed by integrating 7,473 bulk tumor RNA sequencing datasets, 1,564 tumor immunopeptidomes, and 208 single-cell RNA sequencing datasets spanning 21 cancer types. These datasets were compared against 17,384 normal tissue samples from the Genotype-Tissue Expression (GTEx) project to filter out non-tumor-specific events. The authors developed a multimodal pipeline to interrogate eleven molecular sources of tumor antigens including somatic mutations, gene fusions, alternative splicing events, transposable elements, cryptic open reading frames, and microbial sequences. Peptide identifications were refined using deep-learning-based rescoring and predictions of major histocompatibility complex class I (HLA-I) binding affinity.

Li et al. identify 15,079 tumor-specific peptides presented on HLA-I molecules dramatically expanding the known antigenic landscape. Most peptides originated from non-canonical sources including cryptic open reading frames and splicing variants while single-nucleotide variant-derived peptides accounted for just 0.009% of detections. The authors highlight a splicing-derived peptide from the PMEL gene that is more abundantly presented than canonical PMEL epitopes and engages a distinct T cell receptor (TCR) repertoire suggesting it may serve as a superior target for TCR-based or peptide-centric CAR (PC-CAR) therapies.The study also provides the first endogenous evidence of peptides derived from the LINE-1 open reading frame 2 (ORF2) protein, a key retrotransposon previously undetected by conventional proteomics. Beyond tumor cell-derived epitopes, the authors detect HLA-I-presented peptides from tumor-resident pathogens including cytomegalovirus and Niallia circulans, thereby opening new avenues for pathogen-targeted vaccine or TCR-based therapies. Finally, the authors show that differences in antigen processing machinery can shape tumor-specific immunopeptidomes offering insight into why peptide repertoires vary across cancer types.

Limitations

Although this study represents a major technical and conceptual advance, key limitations remain. The dataset reflects tumors of diverse clinical and biological contexts; however, metadata regarding patient treatments, tumor stage, immune status, or prior immunotherapy exposure were not analyzed, raising the possibility of confounding from unmeasured clinical heterogeneity. In addition, while mutation-derived peptides were catalogued, it remains unclear whether these antigens are clonal, subclonal, or bystander events, and whether they distinguish truly malignant cells from pre-malignant lesions. Importantly, the immunogenicity of the identified epitopes was not functionally validated in vitro or in vivo, nor did the study explore HLA class II (HLA-II) presentation, which is relevant for CD4+ T cell responses and tumor-antigen priming by antigen-presenting cells.

Significance / Novelty

This study provides the first pan-cancer atlas of non-canonical HLA-I-presented tumor antigens, unveiling a vastly expanded set of candidate targets for T cell-based therapies. By publicly releasing a web portal, immunopeptidomics search tool, and peptide libraries, Li et al. equip the field with a powerful foundation for rational design of next-generation TCR therapies, peptide-centric CAR T cells, and therapeutic cancer vaccines.

Credit

Reviewed by Benjamin Schrank as part of a cross-institutional journal club between the Icahn School of Medicine at Mount Sinai, the University of Oxford, the Karolinska Institute, the University of Texas MD Anderson Cancer Center and the University of Toronto.