Identification of tumor-intrinsic drivers of immune exclusion in acral melanoma

Abstract

Acral melanoma (AM) has distinct characteristics as compared with cutaneous melanoma and exhibits poor response to immune checkpoint inhibitors (ICIs). Tumor-intrinsic mechanisms of immune exclusion have been identified in many cancers but less studied in AM. We characterized clinically annotated tumors from patients diagnosed with AM at our institution in correlation with ICI response using whole transcriptome RNAseq, whole exome sequencing, CD8 immunohistochemistry, and multispectral immunofluorescence imaging. A defined interferon-γ-associated T cell-inflamed gene signature was used to categorize tumors into non-T cell-inflamed and T cell-inflamed phenotypes. In combination with AM tumors from two published studies, we systematically assessed the immune landscape of AM and detected differential gene expression and pathway activation in a non-T cell-inflamed tumor microenvironment (TME). Two single-cell(sc) RNAseq AM cohorts and 11 bulk RNAseq cohorts of various tumor types were used for independent validation on pathways associated with lack of ICI response. In total, 892 specimens were included in this study. 72.5% of AM tumors showed low expression of the T cell-inflamed gene signature, with 23.9% of total tumors categorized as the non-T cell-inflamed phenotype. Patients of low CD3⁺CD8⁺PD1⁺ intratumoral T cell density showed poor prognosis. We identified 11 oncogenic pathways significantly upregulated in non-T cell-inflamed relative to T cell-inflamed TME shared across all three acral cohorts (MYC, HGF, MITF, VEGF, EGFR, SP1, ERBB2, TFEB, SREBF1, SOX2, and CCND1). scRNAseq analysis revealed that tumor cell-expressing pathway scores were significantly higher in low versus high T cell-infiltrated AM tumors. We further demonstrated that the 11 pathways were enriched in ICI non-responders compared with responders across cancers, including AM, cutaneous melanoma, triple-negative breast cancer, and non-small cell lung cancer. Pathway activation was associated with low expression of interferon stimulated genes, suggesting suppression of antigen presentation. Across the 11 pathways, fatty acid synthase and CXCL8 were unifying downstream target molecules suggesting potential nodes for therapeutic intervention. A unique set of pathways is associated with immune exclusion and ICI resistance in AM. These data may inform immunotherapy combinations for immediate clinical translation.

Keywords: immune checkpoint inhibitors; immunotherapy; melanoma; tumor microenvironment.

Figure 1

Immune landscape of acral tumors across three cohorts (UPMC, NW, MIA). (A, B) n=109 samples shown with RNAseq data from all AM cohorts. Hierarchical clustering based on the expression of a defined T cell-inflamed gene signature identifies T cell-inflamed, intermediate, and non-T cell-inflamed tumor groups. Horizontal annotation bars above the heatmaps represent ICI response, CD8⁺ tumor lymphocyte infiltrates by IHC, CD8A gene expression and T cell-inflamed gene expression by RNAseq, and total TMB by WES, when data are available. (B) CD8⁺ T cell fraction in the three tumor groups via digital immune deconvolution from RNAseq data. (C–G) n=17 shown with tumor specimens available for imaging experiments from the UPMC cohort. (C) CD8⁺ IHC staining in representative images of a non-T cell-inflamed and T cell-inflamed tumor. Pink circle highlights heavy CD8 IHC staining. (D) Intratumoral CD3⁺CD8⁺PD1⁺ T cell infiltrates in the three tumor groups via multispectral immunofluorescence imaging. (E) Representative immunofluorescence images of non-T cell-inflamed and T cell-inflamed acral tumors. (F) Kaplan-Meier survival curves stratified by tumor groups, with p value shown for testing T cell-inflamed and intermediate tumor groups versus non-T cell-inflamed tumor group. P values were computed by two-sided Wilcoxon test in B, D, and by log-rank test in F. AM, acral melanoma; ICI, immune checkpoint inhibitor; IHC, immunohistochemistry; MIA, immune checkpoint inhibitor; TMB, tumor mutational burden.

Figure 2

Activation of 11 oncogenic pathways in non-T cell-inflamed versus T cell-inflamed acral tumors across three cohorts (UPMC, NW, MIA). (A) Pathways activated in non-T cell-inflamed versus inflamed UPMC tumors and one or both of the NW and MIA cohorts. A total of 102 pathways with a z-score of at least 0.9 at p<0.05 in at least one cohort are shown. Eleven shared pathways are identified across the three cohorts (shown on top, solid side bar). Each row represents one pathway (top to bottom). Pathways were sorted by z-score high to low in the UPMC cohort. (B) Correlation between pathway expression scores and the T cell-inflamed gene signature. All samples are shown; n=20 in UPMC, n=22 in NW, and n=67 in MIA. Each data point represents one tumor; color denotes the T cell-inflamed (red), non-T cell-inflamed (blue), and intermediate groups (gray). Spearman’s correlation coefficient ρ and FDR-adjusted p values are shown for each pathway. Linear regression was shown with 95% confidence bands. (C) The proportion of tumors harboring activation of each pathway in non-T cell-inflamed and T cell-inflamed groups. The x-axis shows the number of tumors with all three acral cohorts combined. (D) Correlation between the T cell-inflamed gene signature and activation of each pathway at a continuous scale. (Left to right) tumor samples were sorted by by higher to lower number of activated pathways within each group. n=109 tumors are shown. (E) Downstream target molecules shared among the 11 pathways. Molecules shared by at least four pathways are shown. (F) Protein-protein functional network of the 11 pathways and downstream target molecules from E. Nodes with at least one connection are shown, from STRING functional protein association networks (confidence score>0.4; active interaction sources as ‘experiments’, ‘databases’, and ‘coexpression’). Line thickness indicates the strength of data support. Nodes were clustered by graph-based Markov Cluster Algorithm (MCL) (inflation parameter=2.5), with color indicating each cluster and dotted lines indicating edges between clusters. P values were computed by Spearman’s correlation in B, with denotation: *P<0.05, **P<0.01, ***P<0.001 after FDR adjustment for multiple comparisons. MIA, Melanoma Institute of Australia; FDR, false discovery rate.

Figure 3

Single cell analysis of immunosuppressive oncogenic pathways in acral melanoma. Data from Li et al are shown after harmonization. (A) Cell populations visualized by UMAP. A total of 25 523 cells from 8 samples are shown (×10 Genomics 5’). One sample was excluded from analysis due to different ×10 chemistries. Color denotes cell populations annotated by SingleR followed by manual curation. (B) Distribution of pathway scores in the main cell populations. (Left to right) cell populations were sorted by median pathway scores higher to lower. Eight out of 11 pathways are shown consisting of 5–22 unique downstream target molecules per pathway (MYC, MITF, EGFR, SP1, ERBB2, SERBP1, SOX2, CCND1). Meta-pathway represents a conglomerate expression score of downstream target molecules from all pathways combined. (C) Comparison of tumor cell-expressing pathway scores between low and high T cell-infiltrated tumors. Each data point represents one malignant cell. n=1529 and 3146 malignant cells shown from low and high T cell-infiltrated tumors, respectively. (D) Differentially expressed genes (DEGs) in tumor cells comparing low versus high T cell-infiltrated samples. A total of 108 genes at FDR-adjusted p<0.05 shown. Cells are on the column, genes on the row. (E) Biological pathway enrichment (BioPlanet2019) on the tumor cell DEGs shown in D. P values were computed by two-sided two-sample t-test in C, two-sided Wilcoxon test in D, hypergeometric test in E. Denotation: ***P<0.001 after FDR adjustment for multiple comparisons. NK, neutral killer cells; FDR, false discovery rate.

Figure 4

Validation of immunosuppressive oncogenic pathways in ICI-treated cohorts. Baseline (pre-ICI treatment) tumors were used in all analyses. (A) Comparison of 11 pathways in acral melanoma (AM) samples from patients who showed clinical benefit (CB) or not (NCB) to ICI. n=14 shown, UPMC AM cohort. Patients in NW or MIA AM cohort did not receive ICI, hence not shown. (B) Comparison of 11 pathways in cutaneous melanoma (CM) samples from patients who received ICI (Riaz et al and Liu et al). (C) Pan-cancer analysis of pathways across ICI-treated patient cohorts. (left panel) The fold change of pathway scores in non-T cell-inflamed relative to T cell-inflamed tumors in each cohort; A label ‘X’ was added to the grids when significant upregulation (log₂(fold change) >0) in NCB versus CB or NR versus R was detected at FDR-adjusted P<0.10. (right panel) The correlation between pathway scores and T cell-inflamed gene expression; A label ‘X’ were added when significant inverse correlation (Spearman’s ρ<0) was detected at FDR-adjusted P<0.10. Eleven ICI datasets with gene expression data and deidentified clinical data publicly available were included in analysis, consisting of five cutaneous melanoma studies (Riaz et al, Liu et al, Hugo et al, Gide et al, and Checkmate-064 from Campbell et al), one triple negative breast cancer study (TNBC; Blenman et al), one head and neck squamous carcinoma study (HN; Foy et al), two non-small cell lung cancer studies (NSCLC; Ravi et al., Foy et al), one renal cell carcinoma study (RCC; Miao et al), and one urothelial carcinoma study (Rose et al). P-values were computed by two-sided two-sample t-test in A–C upper panel, Spearman’s correlation in C bottom panel. ICI, immune checkpoint inhibitor; MIA, Melanoma Institute of Australia; FDR, false discovery rate.