DNA Methylation Analysis of Commpass Reveals Distinct Epigenetic Programs of High-Risk Disease

Background. Identifying the biology of high-risk multiple myeloma (MM) is critical to improving outcomes. Current genetic alterations do not faithfully identify high-risk proliferative disease and there are limited data that integrate genetic, epigenetic, and transcriptional information with outcomes. We generated DNA methylation (DNAm) data on 415 MMRF CoMMpass samples and identified distinct epigenetic programs of high-risk disease.

Methods. Institutional IRB approval was obtained and DNA from CD138+ CoMMpass samples were analyzed by whole genome bisulfite sequencing on Illumina NovoSeq instruments. Reads were aligned to the GRCh38 genome with Bismark and DNAm values were obtained at 16,374,655 CpGs at an average coverage of 24x. WGBS from the BluePrint project were used for comparison to normal plasma cells. CoMMpass whole genome sequencing and RNA-seq were obtained from the Genomic Data Commons and outcome data (IA21) were provided by the MMRF.

Results. Newly diagnosed MM samples (N=370) exhibited DNA hypomethylation (median 45%, range 23-77%) as compared to normal plasma cells (median 69%), which occurred in late replicating regions of the genome (P=2.8e-135). Dimensionality reduction (t-SNE) of DNAm data clustered samples by subtype, most notably t(4;14) and t(14;16) samples. DNAm also better defined the high-risk Proliferation (PR) subtype as compared to unsupervised RNA analysis. Integration of DNAm and RNA data identified 1,494,744 loci predictive of gene expression (FDR≤0.01). The DNAm loci most tightly correlated with expression included genes highly relevant to MM pathology including cell cycle (e.g. CDKN2C, CCND2), proteasome function (e.g. PSMA8), immune signaling (e.g. CD28, CD27, IL5RA), and immune targets (e.g. GPRC5D).

Differential DNAm between MM subtypes indicated the high-risk t(4;14) subtype had the most distinct DNAm program with 2,075,489 Differentially Methylated Loci (DML; FDR≤0.01). The majority (92%) of t(4;14) DML had higher levels of DNAm as compared to other subtypes and this was specific to the late replicating regions of the genome (P=3.5e-38). Correlating t(4;14)-specific gene expression and DNAm identified several co-regulated genes including cell adhesion (ITGA7, JAM3) and frizzled (FZD8, FZD2) pathway components.

The PR subtype had the second most unique DNAm program with 418,474 DML (FDR ≤0.01). In contrast to the t(4;14) subtype, 99.95% of PR DML had less DNAm than other subtypes. PR hypomethylated loci were enriched for motifs of E2F transcription factors, which contain a CpG in their binding motif. Consistent with this, expression of E2F1 and E2F2 were significantly higher in the PR subtype. Analysis of DNAm and E2F1 binding in MM.1S cells showed that E2F1 exclusively bound unmethylated regions of the genome, corroborating the motif analysis of CoMMpass PR subtype DML.

Conclusion. MM exhibits a dramatically remodeled DNAm program with extensive DNA hypomethylation in late replicating regions of the genome, suggesting this hypomethylation reflects the proliferative history of MM. This phenomenon is consistent with normal plasma cell differentiation and other B cell malignancies. However, the t(4;14) subtype had a distinct DNAm program, which is likely connected to aberrant levels of H3K36me2 catalyzed by NSD2. H3K36me2 is recognized by the PWWP domains of DNA methyltransferases, potentially explaining the higher levels of DNAm in t(4;14). This suggests novel NSD2 inhibitors require time to not only deplete H3K36me2, but also reprogram t(4;14)-specific DNAm. Finally, consistent with DNA hypomethylation reflecting the proliferative history of MM, the PR subtype had the most pronounced DNA hypomethylation. Notably, the PR subtype had reduced DNAm at E2F1 motifs and E2F1 exclusively bound unmethylated regions of the genome, suggesting the DNAm restricts the proliferative program of high-risk MM.