Melanoma-specific MHC-II expression represents a tumour-autonomous phenotype and predicts response to anti-PD-1/PD-L1 therapy

Abstract

Anti-PD-1 therapy yields objective clinical responses in 30-40% of advanced melanoma patients. Since most patients do not respond, predictive biomarkers to guide treatment selection are needed. We hypothesize that MHC-I/II expression is required for tumour antigen presentation and may predict anti-PD-1 therapy response. In this study, across 60 melanoma cell lines, we find bimodal expression patterns of MHC-II, while MHC-I expression was ubiquitous. A unique subset of melanomas are capable of expressing MHC-II under basal or IFNγ-stimulated conditions. Using pathway analysis, we show that MHC-II(+) cell lines demonstrate signatures of 'PD-1 signalling', 'allograft rejection' and 'T-cell receptor signalling', among others. In two independent cohorts of anti-PD-1-treated melanoma patients, MHC-II positivity on tumour cells is associated with therapeutic response, progression-free and overall survival, as well as CD4(+) and CD8(+) tumour infiltrate. MHC-II(+) tumours can be identified by melanoma-specific immunohistochemistry using commercially available antibodies for HLA-DR to improve anti-PD-1 patient selection.

Conflict of interest statement

D.B.J., J.M.B., V.S. and M.E.S. have filed provisionary patent on use of MHC-II to predict response to immunotherapy. The remaining authors declare no competing financial interests.

Figures

Figure 1. A unique subtype of melanoma expresses MHC-II.

(a) Microarray data from 60 melanoma cell lines in the CCLE were analysed for MHC-I (HLA-A/B/C and MHC-II (HLA-DRA) expression. Bars represent the mean±s.d. P value is the result of the Kolmogorov–Smirnov test comparing the distribution of MHC-I (HLA-A, HLA-B, HLA-C) expression with MHC-II expression (HLA-DRA). *represents the cutoff for defining MHC-II(+). (b) Gene-expression data from HLA-DRA(+) cell lines (Clusters Ia/Ib) were compared with HLA-DRA(−) cell lines (Clusters II and III) by an FDR-corrected row t-test. Significantly altered genes are shown on the y-axis and also listed in Supplementary Data 1. An ad hoc heat map is shown at the top, highlighting classical MHC-II genes. (c) Normalized microarray data were analysed by GSA using the curated Molecular Signatures Database, and the resulting gene set scores are presented as a hierarchical clustered heat map.

Figure 2. Characterization of MHC-II(+) melanoma cell lines.

Melanoma cell lines were treated with IFNγ for 24 h before collection and live-cell staining and flow cytometry analysis for MHC-I/HLA-A/B/C (a), MHC-II/HLA-DR (b) and PD-L1 (c). Bars represent mean±s.e.m. for at least three experiments (d) Representative flow plots from c. (e) Western blot analysis of melanoma cell lines after 24 or 48 h of IFNγ stimulation. (f) Phosphorylation of STAT1 (top row) and STAT5 (bottom row) in melanoma cell lines at 15 min after IFNγ stimulation. Histograms were coloured according to the arcsinh transformed ratio or MFI medians relative to the table minimum value.

Figure 3. MHC-II-positive melanoma cell lines associate with NRAS mutations.

(a) HLA-DRA mRNA expression in melanoma cell lines (n=60; one cell line lacked mRNA expression data) from the CCLE compared by genotype. P value (P<0.05) represents result of Tukey's post hoc analysis comparing pan-WT with NRAS-mutant cell lines, following a significant ANOVA (P=0.03) performed among all groups. Bars represent mean±s.e.m. (b) Representative IHC for HLA-DR (brown) and SOX10 (pink) in cases with isolated stromal positivity (top) and with tumour-specific staining (bottom). Both HLA-DR and SOX10 immunostaining is present in all four sections. Scale bar, 50 μm. (c) Analysis of HLA-DR IHC in a melanoma TMA (n=67 evaluable) by genotype. P value represents result of a χ2-test. (d) Overall survival of patients (n=58 evaluable) within the TMA by HLA-DR status (left censored at time of diagnosis). The remaining patient samples were included from outside institutions and follow-up data were not available from those institutions. P value is the result of the log-rank test.

Figure 4. Ex vivo culture of tumours derived from anti-PD-1-responding and non-responding patients identifies heterogeneity in interferon response.

(a) Patient tumour blocks stained for HLA-DR (brown) and SOX10 (pink) at low (scale bar, 500 um) and high magnification (scale bar, 200 μm); PT1: anti-PD-1 non-responder and PT2: anti-PD-1 responder. (b) Experimental schema. (c) Schema and images of PDX tissue sections (ex vivo organotypic culture). (d) Western blot analysis of tissue sections cultured in the presence or absence of IFNγ for 24–48 h.

Figure 5. MHC-II(+) melanomas have improved response rates and clinical benefit to PD-1/PD-L1 inhibition.

(a) HLA-DR positivity by IHC plotted versus response to PD-1/PD-L1-targeted therapy in the discovery set (n=30). Responders include partial and complete responders; non-responders include mixed responders and progressive disease patients. Mixed responders (n=3) are noted by a red triangle. P value is the result of the Wilcoxon's rank sum test. (b) HLA-DR positivity by IHC in the validation set (n=23) plotted versus response to PD-1/PD-L1-targeted therapy. P value is the result of the Wilcoxon's rank sum test (c) Representative images of scans from anti-PD-1 therapy-treated MHC-II(+) patients (d) Progression-free survival (left) and overall survival (right) in anti-PD-1/PD-L1-treated patients, stratified by HLA-DR/MHC-II positivity (5% total tumour cells staining on entire tissue section used as cutpoint). Data from both the initial and validation cohorts were included, when available. P value is the result of the log-rank test. (e) Correlation matrix of IHC markers. P values for the Pearson's correlation appear above the diagonal and correlation coefficients (r) appear below the diagonal.