Background: The human leukocyte antigen (HLA) class II molecules are constitutively expressed in some melanoma, but the underlying molecular mechanisms have not yet been characterized. Methods: The expression of HLA class II antigen processing machinery (APM) components was determined in melanoma samples by qPCR, Western blot, flow cytometry and immunohistochemistry. Immunohistochemical and TCGA datasets were used for correlation of HLA class II expression to tumor grading, T-cell infiltration and patients’ survival. Results: The heterogeneous HLA class II expression in melanoma samples allowed us to characterize four distinct phenotypes. Phenotype I totally lacks constitutive HLA class II surface expression, which is inducible by interferon-gamma (IFN-γ); phenotype II expresses low basal surface HLA class II that is further upregulated by IFN-γ; phenotype III lacks constitutive and IFN-γ controlled HLA class II expression, but could be induced by epigenetic drugs; and in phenotype IV, lack of HLA class II expression is not recovered by any drug tested. High levels of HLA class II APM component expression were associated with an increased intra-tumoral CD4+ T-cell density and increased patients’ survival. Conclusions: The heterogeneous basal expression of HLA class II antigens and/or APM components in melanoma cells is caused by distinct molecular mechanisms and has clinical relevance.

Many patients with resected American Joint Committee on Cancer (AJCC) early-stage cutaneous melanoma nonetheless die of melanoma; additional risk stratification approaches are needed. Using prospectively-collected whole-tissue sections, we assessed in consecutive stage I-IIA patients (N = 439), a previously-validated, immunohistochemistry-based, 7-biomarker signature to prognosticate disease-free survival (DFS), melanoma-specific survival (MSS; primary end-point) and overall survival (OS), independent of AJCC classification. Seven-marker signature testing designated 25.1% of patients (110/439) as high-risk (stage IA, 13.3% [43/323], IB, 53.2% [42/79], and IIA, 67.6% [25/37]). A Kaplan-Meier analysis demonstrated high-risk patients to have significantly worse DFS, MSS and OS versus low-risk counterparts (P < 0.001). In multivariable Cox regression modelling also including key clinicopathological/demographic factors, 7-marker signature data independently prognosticated the studied end-points. Models with the 7-marker signature risk category plus clinicopathological/demographic covariates substantially outperformed models with clinicopathological/demographic variables alone in predicting all studied outcomes (areas under the receiver operator characteristic curve 74.1% versus 68.4% for DFS, 81.5% versus 71.2% for MSS, 80.9% versus 73.0% for OS; absolute differences 5.7%, 10.3% and 7.9%, respectively, favouring 7-marker signature risk category-containing models). In patients with AJCC early-stage disease, the 7-marker signature reliably prognosticates melanoma-related outcomes, independent of AJCC classification, and provides a valuable complement to clinicopathological/demographic factors.