Fig 1.
Tumor-specific splicing alterations are defined as significant variations with respect to the exon-intron structures expressed in normal samples and are classified into four different types: de novo exonization, new exon skipping (neoskipping), alternative (5’/3’) splice site, and intron retention. ISOTOPE calculates the modified open reading frame (ORF) from the reference ORF using the splicing alterations, and identifies the candidate splicing-derived neoepitopes and self-epitopes encoded by the reference transcript that would not be present in the modified ORF as a consequence of the splicing alteration. These candidate peptides are then tested for affinity with the MHC complexes (see Methods for details).
Fig 2.
(A) Validation of the HLA type prediction from tumor RNA-seq data. We show the predictions for MHC Class I (HLA-A, HLA-B, HLA-C) and II (HLA-DQA, HLA-DQB, HLA-DRB) with PHLAT (red) and SeqHLA (blue). Each bar corresponds to the proportion of samples (over a total of 24 small cell lung cancer samples) for which the prediction on the tumor sample coincides with the prediction on the matched normal sample (B) For each cell line, CA46, HL-60 and THP-1, we show the number of different splicing alterations measured (dark blue) and the number of cases leading to a change in the encoded open reading frame (light blue). Alterations shown are alternative (5’/3’) splice-site (A5_A3), de novo exonizations (Exonization), intron retentions (IR), and new exon skipping events (Neoskipping). (C) Number of splicing-derived neoepitopes (splicing-neoepitopes) (red) and splicing-affected self-epitopes (self-epitopes) (blue) detected for each of the splicing alterations in each of cell lines analyzed (CA46, HL-60 and THP-1). (D) as in (C) but separated by HLA-type. (E) Example of a splicing-neoepitope validated with MHC-I associated mass spectrometry data and derived from a neoskipping event in the gene ERF. The peptides are given in the same orientation as the 5’ to 3’ direction of the gene.
Fig 3.
Splicing epitopes in ten breast cancer cell lines.
(A) For each breast cancer cell line analyzed, the bar plots show the number of splicing alterations measured and the number of cases leading to a change in the reference protein. Alterations shown are alternative 5’ or 3’ splice-site (A5_A3), de novo exonizations (exonization), intron retentions (IR), and new exon skipping events (neoskipping). (B) Number of splicing-derived neoepitopes (splicing-neoepitopes) (red) and splicing-affected self-epitopes (self-epitopes) (blue) for each of the splicing alterations in each of the breast cancer cell lines tested. (C) as in (C) but separated by HLA-type. (D) Example of a splicing-derived neoepitope from a neoskipping event in the gene SIL1 validated with MHC-I associated mass spectrometry in the cell line BT549.
Fig 4.
Splicing epitopes in small cell lung cancer.
(A) Mutation burden (y axis) calculated separately for introns (INTRON), coding exons (CDS) and non-coding exonic regions in protein-coding genes (UTR) calculated from whole genome sequencing (WGS) data for several tumor types (x axis), including small cell lung cancer (SCLC). We indicate the pairs of distributions that were significantly different using a Wilcoxon test (* p-val <0.05, ** p-val <0.01, *** p-val<0.001, **** p-val<0.0001). (B) Number of splicing alterations (y axis) according to event type (x axis), indicating all alterations and the subset that impact the open reading frame (ORF). (C) Distribution of splicing-derived neoepitopes (splicing-neoepitopes) and splicing-affected self-epitopes (self-epitopes), separated by splicing alteration type. (D) Same as (C) but separated by HLA-type.
Fig 5.
Splicing epitopes in two melanoma cohorts.
(A) Total number of events and subset of protein-affecting events in the melanoma cohorts treated with anti-CTLA4 and with anti-PD1. (B) Distribution of the number of candidate tumor-specific splicing-derived neoepitopes (splicing-epitopes) and self-epitopes that would be depleted in the altered isoform (self-epitopes). (C) Distribution of the number of candidate epitopes from (B), separated by HLA-type.
Fig 6.
Splicing epitopes and response to immune therapy.
(A) Proportion of splicing-affected self-epitopes (self-epitopes) over the total of epitopes, i.e., splicing-derived neoepitopes (splicing-neoepitopes) plus self-epitopes, (y axis) for patients treated with anti-CTLA4, separated by type of splicing alteration (x axis) and by patient response: responder (green) or non-responder (red). (B) As in (A) but for a different cohort of melanoma patients treated with anti-PD1. (C) Cumulative plots of the binding affinities (x axis) of splicing-neoepitopes in melanoma tumors from exonization events separated in responders (green) and non-responders (red) to anti-PD1 therapy. Kolmogorov-Smirnov test p-value (KS) = 0.0465 (D) As in (C), for splicing-derived neoepitopes from neoskipping events, KS = 0.0274. (E) Cumulative plots of the affinities of splicing-neoepitopes in melanoma tumors from intron retention events separated in responders (green) and non-responders (red) to anti-CTLA4 therapy, KS = 0.0016. (F) Frequency of splicing-neoepitopes represented according to the total number of patients in which are predicted (x axis, total_patients_expressed) and to the absolute count-difference in responders and non-responders to anti-PD1 therapy (y axis, Difference number patients each class). Epitopes are indicated in green if they are more frequent in responders, and in red otherwise. The size of the point indicates the number of cases. (G) The same as in (F) but for responders and non-responders to anti-CTLA4 therapy.