NMD inhibition by 5-azacytidine augments presentation of immunogenic frameshift-derived neoepitopes

Abstract

Frameshifted protein sequences elicit tumor-specific T cell-mediated immune responses in microsatellite-unstable (MSI) cancers if presented by HLA class I molecules. However, their expression and presentation are limited by nonsense-mediated RNA decay (NMD). We employed an unbiased immunopeptidomics workflow to analyze MSI HCT-116 cells and identified >10,000 HLA class I-presented peptides including five frameshift-derived InDel neoepitopes. Notably, pharmacological NMD inhibition with 5-azacytidine stabilizes frameshift-bearing transcripts and increases the HLA class I-mediated presentation of InDel neoepitopes. The frameshift mutation underlying one of the identified InDel neoepitopes is highly recurrent in MSI colorectal cancer cell lines and primary patient samples, and immunization with the corresponding neoepitope induces strong CD8⁺ T cell responses in an HLA-A∗02:01 transgenic mouse model. Our data show directly that pharmacological NMD inhibition augments HLA class I-mediated presentation of immunogenic frameshift-derived InDel neoepitopes thus highlighting the clinical potential of NMD inhibition in anti-cancer immunotherapy strategies.

Keywords: Biological Sciences; Cancer; Immunology.

Graphical abstract

Figure 1

Immunopeptidomics workflow for the identification, validation, and quantification of InDel neoepitopes After high-throughput IP of HLA class I:peptide complexes, HLAp are separated (1; see also Figure S2) and subjected to LC-MS/MS using different fragmentation and precursor selection methodologies (HCD, EThcD, lowEThcD; 2). Data analysis of raw data is performed using de novo-assisted database search against the UniProt database to identify endogenous HLAp and against custom databases containing neoepitope sequences (3; see also Figure S3). All identified peptides are validated bioinformatically (binding prediction, sequence clustering, retention time prediction; 4). Neoepitope candidates are further validated by comparison with synthetic peptide spectra and validation of underlying mutations. In vivo processing and presentation are tested by immunization of an “HLA-humanized” mouse model (5). Neoepitopes are measured again using a targeted MS2 approach (6). Label-free quantification is performed for endogenous HLAp and validated neoepitopes (7). HT-IP, high-throughput immunoprecipitation; HLAp, HLA class I-presented peptides; MS, mass spectrometry; LC-MS/MS, liquid chromatography-tandem mass spectrometry; DB, database.

Figure 2

Quality control and characteristics of identified peptides (A) Number of peptides identified using different MS fragmentation methods and overlapping sets. Boxplot summary representing intensity distribution for subsets of peptides. Bars of boxplot summary represent 25th to 75th percentiles, middle line represents median. (B) Predicted hydrophobicity index (HI) against observed retention time of identified peptides for different MS fragmentation methods. (C) Typical length distribution of HLA class I-presented peptides. Colors represent the fraction of peptides with predicted binding affinity to a particular HLA allele determined by NetMHCpan 4.0. (D) Binding prediction of all identified peptides. Threshold for strong binders is top 0.5% ranked, for weak binders top 2%. (E) Sequence clustering of identified peptides to four distinct binding motifs matching HLA alleles expressed by HCT-116 cell line; 282 outliers (2.8%) were not clustered.

Figure 3

Validation of identified InDel neoepitopes (A) Overview of validation procedure. Candidates were filtered using BLASTp to exclude peptides matching endogenous proteins. Spectra of candidates were compared with spectra recorded from synthetic peptides, and underlying frameshift mutations were confirmed by Sanger sequencing. (B) Comparison of matched ions observed in candidate spectra (top) and synthetic peptide spectra (bottom). Top 10 most intense ions are labeled; retention time difference and correlation between experimental and synthetic peptide spectrum is reported. See also Table S2. (C) Base calls and Sanger traces of underlying frameshift mutations. Positions of InDel mutations are indicated by an arrow. m1, minus one base pair deletion. See also Figure S5 and Tables S1 and S4.

Figure 4

Treatment with 5AZA stabilizes NMD-targeted transcripts and augments HLA-mediated presentation of peptides originating from the encoded proteins (A) qPCR analysis of endogenous NMD targets (ATF3, ATF4, UPP1) and InDel-mutated transcripts (CKAP2, NFAT5, PSMC6, STK38, TUBGCP3) after treatment with 5 μM 5AZA for 24 h (red) or siRNA-mediated KD of UPF1 (orange). UPF1 mRNA levels were determined as control for siRNA-mediated knockdown (N.D. = not determined). Each bar represents the mean ± SD of 3 experiments, ∗p ≤ 0.0001 (two-sided, unpaired t test). See also Figure S1. (B) GO term enrichment for source genes of significantly upregulated hit peptides after 5AZA treatment for 24 h. (C) Volcano plot summarizing limma analysis of label-free quantification of the immunopeptidome isolated from 5AZA-treated versus DMSO-treated HCT-116 cells. Upregulated InDel neoepitopes (CKAP2, PSMC6) and peptides originating from putative endogenous NMD targets are labeled with the corresponding gene name. Color represents hit annotation; shape indicates if values were imputed (circle = no, triangle = yes). See also Figure S6 and Table S3. (D) Representative plots showing changes in intensity for InDel neoepitopes SLMEQIPHL (CKAP2), REKHSWHEP (PSMC6), and selected peptides originating from known NMD targets ATF3, ATF4, and UPP1 after treatment with 5AZA for 24 h. Bars represent 25th to 75th percentiles, middle line represents median, and points represent individual measurements of biological replicates. See also Table S3.

Figure 5

In vivo immunization of A2.DR1 mice with InDel neoepitopes induces CD8⁺ T cell responses; see also Figure S7 (A) Immunization scheme. See also Table S2. (B) Representative ELISpot assay results for isolated CD8⁺ T cells stimulated with ConA (assay positive control), no peptide control, and InDel neoepitopes SLMEQIPHL (CKAP2) and GVWEKPRRV (TUBGCP3). (C) Quantitative analysis of ELISpot assays. Bars represent mean ± SEM of N = 6 (CKAP2), N = 5 (TUBGCP3), or N = 3 (E7 11–19, control peptides) experiments, ∗p ≤ 0.005 (two-sided, unpaired t test).