Differential binding affinity of mutated peptides for MHC class I is a predictor of survival in advanced lung cancer and melanoma

Abstract

Background: Cancer mutations generate novel (neo-)peptides recognised by T cells, but the determinants of recognition are not well characterised. The difference in predicted class I major histocompatibility complex (MHC-I) binding affinity between wild-type and corresponding mutant peptides (differential agretopicity index; DAI) may reflect clinically relevant cancer peptide immunogenicity. Our aim was to explore the relationship between DAI, measures of immune infiltration and patient outcomes in advanced cancer.

Patients and methods: Cohorts of patients with advanced non-small-cell lung cancer (NSCLC; LUAD, n = 66) and melanoma (SKCM, n = 72) were obtained from The Cancer Genome Atlas. Three additional cohorts of immunotherapy treated patients with advanced melanoma (total n = 131) and NSCLC (n = 31) were analysed. Neopeptides and their clonal status were defined using genomic data. MHC-I binding affinity was predicted for each neopeptide and DAI values summarised as the sample mean DAI. Correlations between mean DAI and markers of immune activity were evaluated using measures of lymphocyte infiltration and immune gene expression.

Results: In univariate and multivariate analyses, mean DAI significantly correlated with overall survival in 3/5 cohorts, with evidence of superiority over nonsynonymous mutational and neoantigen burden. In these cohorts, the effect was seen for mean DAI of clonal but not subclonal peptides. In SKCM, the association between mean DAI and survival bordered significance (P = 0.068), reaching significance in an immunotherapy-treated melanoma cohort (P = 0.003). Mean DAI but not mutational nor neoantigen burden was positively correlated with independently derived markers of immune infiltration in both SKCM (P = 0.027) and LUAD (P = 0.024).

Conclusions: The association between mean DAI, survival and measures of immune activity support the hypothesis that DAI is a determinant of cancer peptide immunogenicity. Investigation of DAI as a marker of immunologically relevant peptides in further datasets and future clinical studies of neoantigen based immunotherapies is warranted.

Keywords: immunoinformatics; immunotherapy; neoantigen prediction; peptide immunogenicity.

Figure 1

(A) Distribution of DAI for all peptides in three LUAD samples (highest, average and lowest mean DAI, respectively). (B–D) Correlation between mean DAI and non-synonymous (NS) mutation load, proportion of peptides with a DAI >0 and maximum DAI across five cohorts was evaluated by linear regression.

Figure 2

. Density plots representing the distribution of mean DAI across cohorts, with dotted lines indicating the first quartile cut point used to stratify patients for subsequent survival analysis in LUAD and Rizvi lung cancer cohorts. One way ANOVA P-values are shown.

Figure 3

. Kaplan–Meier survival curves for patients with advanced lung cancer (A; TCGA LUAD and Rizvi cohorts) and melanoma (B; TCGA SKCM and Van Allen cohorts), stratified into high and low comparator groups for each variable (columns). (A) Mean DAI was calculated for all predicted neopeptides. For each variable, patients were stratified into high (>first quartile) and low (median) and low (P-values are shown. Mutational and neoantigen burden refers to the number of nonsynonymous mutations and neoantigens, respectively. OS, overall survival; PFS, progression free survival; NA, neoantigen.

Figure 4

. Multivariate Cox regression modelling of survival in advanced lung cancer (A) and melanoma (B). NS, non-synonymous; NA, neoantigen; HR, hazard ratio; CI, confidence interval. Data on n = 74/75 SKCM patients available for analysis.

Figure 5

. TCGA patients with advanced melanoma have previously been stratified into high and low immune-infiltrated groups based on unsupervised cluster analysis of transcriptomic data (RNAseq cluster) and histopathological assessment of lymphocyte density and distribution (lymphocyte score, LS). (A) Patients with immune-infiltrated tumours as defined by RNAseq cluster combined with a high LS have a significantly higher neoantigen mean DAI. (B, C) Mutational and neoantigen burden were not different between high- and low-infiltrated groups. Wilcoxon rank sum test P-values are shown.

Figure 6

. (A) A 13-gene MHC-II expression signature has previously been shown to correlate with immune infiltration in LUAD. Patients with a high (above the median) MHC-II expression score have higher mean DAI but no difference in mutational/neoantigen burden. (B) Patients in the Rizvi cohort were stratified into high and low PD-L1 expression groups based on previously published histopathological evaluation (n = 29 available for analysis). There is a non-statistically significant trend of association between PD-L1 expression and mutation/neoantigen burden and mean DAI. Wilcoxon rank sum test P-values are shown.