IRIS: Discovery of cancer immunotherapy targets arising from pre-mRNA alternative splicing

Abstract

Alternative splicing (AS) is prevalent in cancer, generating an extensive but largely unexplored repertoire of novel immunotherapy targets. We describe Isoform peptides from RNA splicing for Immunotherapy target Screening (IRIS), a computational platform capable of discovering AS-derived tumor antigens (TAs) for T cell receptor (TCR) and chimeric antigen receptor T cell (CAR-T) therapies. IRIS leverages large-scale tumor and normal transcriptome data and incorporates multiple screening approaches to discover AS-derived TAs with tumor-associated or tumor-specific expression. In a proof-of-concept analysis integrating transcriptomics and immunopeptidomics data, we showed that hundreds of IRIS-predicted TCR targets are presented by human leukocyte antigen (HLA) molecules. We applied IRIS to RNA-seq data of neuroendocrine prostate cancer (NEPC). From 2,939 NEPC-associated AS events, IRIS predicted 1,651 epitopes from 808 events as potential TCR targets for two common HLA types (A*02:01 and A*03:01). A more stringent screening test prioritized 48 epitopes from 20 events with "neoantigen-like" NEPC-specific expression. Predicted epitopes are often encoded by microexons of ≤30 nucleotides. To validate the immunogenicity and T cell recognition of IRIS-predicted TCR epitopes, we performed in vitro T cell priming in combination with single-cell TCR sequencing. Seven TCRs transduced into human peripheral blood mononuclear cells (PBMCs) showed high activity against individual IRIS-predicted epitopes, providing strong evidence of isolated TCRs reactive to AS-derived peptides. One selected TCR showed efficient cytotoxicity against target cells expressing the target peptide. Our study illustrates the contribution of AS to the TA repertoire of cancer cells and demonstrates the utility of IRIS for discovering AS-derived TAs and expanding cancer immunotherapies.

Keywords: RNA splicing; T cell receptors; immunotherapy.

Fig. 1.

IRIS: A big-data informed computational platform for discovering AS-derived cancer immunotherapy targets. Overall workflow of IRIS, computational modules, large-scale reference database of AS profiles, and screening tests are illustrated. IRIS has three main computational modules: (A) RNA-seq data processing, (B) in silico screening, and (C) TCR/CAR-T target prediction. A flowchart illustrates the key components and analytical steps of IRIS. (D) Illustration of IRIS DB, a reference database of AS profiles across tumor and normal tissue samples, and two screening tests to assess tumor association and specificity. AS, alternative splicing; TCR, T cell receptor; CAR-T, chimeric antigen receptor T cell.

Fig. 2.

Proteotranscriptomics analysis of AS-derived peptides in cell line immunopeptidomes. (A) Proteotranscriptomics workflow adopted by IRIS for discovering AS-derived peptides in MS-based proteomics datasets. IRIS accepts various types of MS data (Right), such as whole-cell proteomics, surfaceomics, or immunopeptidomics (HLA peptidomics) data. Aided by RNA-seq data (Left), an RNA-seq-augmented custom proteome library is constructed and searched using MSGF+. (B) Summary of AS-derived epitopes in JeKo-1 and B-LCL cell lines. Peptide-spectrum matches (“PSMs”) and “Unique peptides” are provided by MSGF+ with a target-decoy FDR of 5%. “Predicted AS epitopes” are generated by the IRIS target prediction module, which utilizes IEDB predictors. “MS-validated AS epitopes” are defined as epitopes that are predicted by IRIS and detected in the immunopeptidomics data. (C) Percentage of IRIS-predicted AS-derived epitopes among all MS-detected epitopes for three cell lines. Graph shows the percentage of all MS-detected epitopes that are IRIS-predicted AS-derived epitopes (y-axis) as a function of the MSGF+ target-decoy FDR (x-axis). (D) Preferential detection of high-affinity AS-derived peptides in immunopeptidomics data. Graph shows the number of AS-derived peptides detected in JeKo-1 immunopeptidomics data (y-axis) as a function of the MSGF+ target-decoy FDR (x-axis). Peptides with high (IC₅₀ < 500 nM; orange) and low (IC₅₀ ≥ 500 nM; gray) predicted HLA binding affinities are shown. (E) Heatmap illustration of the distribution of AS-derived peptides detected in JeKo-1 immunopeptidomics data as a function of predicted HLA binding affinity and transcript expression level. AS-derived peptides are binned by their corresponding transcript expression levels and IEDB-predicted HLA binding affinities. Heatmap is colored from red (high) to blue (low), reflecting the proportion of IRIS-predicted AS-derived epitopes that are MS-detected in each bin.

Fig. 3.

IRIS discovery of AS-derived targets for NEPC. (A) Stepwise results of IRIS to identify AS-derived cancer immunotherapy targets from 23 NEPC samples. Skipped exon (SE) events identified by the IRIS RNA-seq data processing module were screened against 11 normal tissue types from the IRIS DB to identify tumor-associated events and predict corresponding TCR targets. (B) Heatmap of AS profiles of 2,939 NEPC-associated SE events across NEPC and 11 normal tissue types. (C) Summary of NEPC-associated/-specific targets. “Events w/ Specific SJs” are SE events that contain tumor-specific SJ(s) identified from the SJ count (SJC)-based tumor-specificity screen. (D) Bar plots showing the percentage of NEPC-associated and NEPC-specific SE events involving inclusion (Left) or skipping (Right) of microexons in NEPC. (E) Heatmap of gene expression levels of 220 splicing factors across NEPC and 11 normal tissue types. (F) Violin plots of log-transformed gene expression levels of serine/arginine repetitive matrix protein 4 (SRRM4) across NEPC, metastatic castration-resistant prostate cancer (CRPC), primary prostate adenocarcinoma (PRAD), and 11 normal tissue types. Two-sided Mann–Whitney U test was conducted to compare SRRM4 gene expression levels between NEPC and every other group. Red asterisks represent P-values (*: P ≤ 0.05, **: P ≤ 0.01, ***: P ≤ 0.001, ****: P ≤ 0.0001). Groups are colored by tumor (red) and normal (blue) tissue.

Fig. 4.

Evaluation and visualization of IRIS-predicted targets for NEPC. (A) The target evaluation process for NEPC. The three-dimensional scatterplot illustrates the three main criteria used to evaluate IRIS-predicted targets: degree of tumor association, FC of the tumor-enriched isoform between tumor and normal tissues, and gene expression level in tumor tissues. These and additional criteria to evaluate targets are listed below the scatterplot. Criteria illustrated in the scatterplot are bolded. (B) Representative examples of 10 IRIS-predicted TCR targets are visualized by IRIS in paired violin and bar plots. Each row shows one IRIS-predicted TCR target. Violin plots show the PSI values of each target in NEPC and the normal tissue panel. Bar plots show the fraction of samples expressing the SJ(s) of the tumor-enriched isoform in NEPC and the normal tissue panel. If the tumor-enriched isoform is the exon inclusion isoform, the bar plot displays the upstream and downstream inclusion SJ as two bars. If the tumor-enriched isoform is the exon skipping isoform, the bar plot displays the skipping SJ as one bar.

Fig. 5.

Isolation and characterization of TCRs reactive to IRIS-predicted NEPC epitopes. (A) IRIS-epitope priming using two APC systems: (1) conventional type 1 dendritic cell (cDC1)-like cells differentiated from autologous CD34+ hematopoietic stem cells (HSCs), and (2) existing APCs from PBMCs. (B) Example of reactive T cell populations primed with a DMSO negative control, IRIS epitope pool, or PMA/Ionomycin using the CLInt-seq TNFα/IFNγ intracellular marker staining strategy. (C) Example of reactive T cell populations primed with a DMSO negative control, IRIS epitope pool, or PMA/Ionomycin by the CD137 surface marker staining strategy. (D) Overview of the cloning strategy for TCRα/β chains in the pMAX system for Jurkat-NFAT-GFP screening. (E) Overview of the Jurkat-NFAT-GFP reporter system. (F) IFNγ ELISA of one specific TCR (JPTCR_238) targeting an IRIS-predicted AS-derived epitope in CLASP1, when co-cultured with K562-A2-GFP single-cell clones transduced to express a full-length or truncated CLASP1 protein isoform. Error bars indicate SD (n = 3). JPTCR_238: an isolated TCR targeting an IRIS-predicted epitope in CLASP1; F5: a clinically tested TCR targeting the MART1 melanoma antigen; NGFR: empty vector with no introduced TCR as a negative control; Untransduced: untransduced as a negative control. (G) Cytotoxicity analysis by live cell imaging of a K562-A2-GFP single-cell clone transduced with a full-length CLASP1 protein isoform containing an IRIS-predicted epitope targeted by JPTCR_238. F5 TCR, NGFR (no TCR introduced), and untransduced were used as negative controls.