DNA Methylation Classes of Stage II and III Primary Melanomas and Their Clinical and Prognostic Significance

Abstract

Purpose: Patients with stage II and III cutaneous primary melanoma vary considerably in their risk of melanoma-related death. We explore the ability of methylation profiling to distinguish primary melanoma methylation classes and their associations with clinicopathologic characteristics and survival.

Materials and methods: InterMEL is a retrospective case-control study that assembled primary cutaneous melanomas from American Joint Committee on Cancer (AJCC) 8th edition stage II and III patients diagnosed between 1998 and 2015 in the United States and Australia. Cases are patients who died of melanoma within 5 years from original diagnosis. Controls survived longer than 5 years without evidence of melanoma recurrence or relapse. Methylation classes, distinguished by consensus clustering of 850K methylation data, were evaluated for their clinicopathologic characteristics, 5-year survival status, and differentially methylated gene sets.

Results: Among 422 InterMEL melanomas, consensus clustering revealed three primary melanoma methylation classes (MethylClasses): a CpG island methylator phenotype (CIMP) class, an intermediate methylation (IM) class, and a low methylation (LM) class. CIMP and IM were associated with higher AJCC stage (both P = .002), Breslow thickness (CIMP P = .002; IM P = .006), and mitotic index (both P < .001) compared with LM, while IM had higher N stage than CIMP (P = .01) and LM (P = .007). CIMP and IM had a 2-fold higher likelihood of 5-year death from melanoma than LM (CIMP odds ratio [OR], 2.16 [95% CI, 1.18 to 3.96]; IM OR, 2.00 [95% CI, 1.12 to 3.58]) in a multivariable model adjusted for age, sex, log Breslow thickness, ulceration, mitotic index, and N stage. Despite more extensive CpG island hypermethylation in CIMP, CIMP and IM shared similar patterns of differential methylation and gene set enrichment compared with LM.

Conclusion: Melanoma MethylClasses may provide clinical value in predicting 5-year death from melanoma among patients with primary melanoma independent of other clinicopathologic factors.

Figure 1.

InterMEL melanoma sample workflow.

Figure 2.. MethylClasses identified by consensus clustering using the top 2% most variant probes (n=8,652) in 422 primary melanomas.

(A) Heatmap of methylation consensus clusters, with clinical variables indicated at the top. (B) CpG methylation in the three MethylClasses of probes used in clustering, by location relative to CpG island. Kruskal-Wallis p values were <0.001 for differences among the three MethylClasses for all genomic regions. All available probes (n=432,592) are included within the box plots. (C) Demographic and clinical variables of interest differing across the three MethylClasses, as determined by the Fisher’s exact test.

Figure 3.. Univariable and multivariable odds ratios for the associations of clinicopathologic characteristics or MethylClasses with 5-year melanoma-specific survival.

Odds ratios (ORs) for death due to melanoma within five years of diagnosis were calculated for patient or sample characteristics and displayed in Forest plots. Univariable ORs or multivariable ORs accounting for other prognostic features, together with 95% confidence intervals (CIs), were calculated using generalized linear models. The reference group in each comparison is indicated in parentheses along the y-axis. (A) Univariable ORs and 95% CIs for 5-year survival among all stage II/III patients. Note the three different scales for ORs and CIs. Multivariable ORs and 95% CIs for associations of MethylClasses with survival among (B) all stage II/III patients or (C) stratified on stage II or stage III. Multivariable models were adjusted for age (continuous), sex (male, female), natural log Breslow thickness, ulceration (absent, present), mitotic index (continuous) and N-stage (N0, N1, N2, N3).

Figure 4.. Patterns of differential methylation between MethylClasses.

Differentially methylated probes (DMPs) between MethylClasses were identified in pairwise comparisons (CIMP vs. IM, IM vs. LM, or CIMP vs, LM) using limma models with the full probe set (n=432,592) adjusted for age and sex (A-C). Hypermethylated DMPs are to the right of the vertical dashed line in each volcano plot while hypomethylated DMPs are to the left. DMPs significant at FDR p<0.05 are above the horizontal broken lines. Top DMPs are in red from (B) IM vs. LM (-log10 adjusted p≥20 for hypermethylated, -log10 adjusted p≥15 for hypomethylated) (B) are also highly significant in CIMP vs. LM (C). Pie charts in A-C summarize the numbers and proportions of DMPs from each comparison that are hypermethylated (red) or hypomethylated (blue). (D) Venn diagram indicating the overlap of DMPs between comparisons of CIMP or IM vs. the LM reference group.

Figure 5.. Top 50 gene sets differentially methylated between MethylClasses.

MethylRRA was used to evaluate enrichment for gene sets and pathways associated with significant (FDR p<0.05) promoter-associated, gene-annotated DMPs identified from the comparison of (A) CIMP vs. LM or (B) IM vs. LM. DMPs included both hypermethylated and hypomethylated. Gene sets queried were GO C5 biological processes, C6 oncogenic signatures and C2 curated genes sets obtained from the Broad Institute MSig Database (http://software.broadinstitute.org/gsea/msigdb/index.jsp). Gene sets are color coded according to biological or functional themes.