Changes in Methylation and Chromatin Accessibility Underlie Subtype Classification and Disease Evolution in Waldenström’s Macroglobulinemia

Background: We previously identified three subtypes of Waldenstrom’s Macroglobulinemia (WM): B-cell like (BCL), Plasma cell like (PCL), and an intermediate group enriched for early/smoldering WM from which the other two subtypes evolved (Hunter et al, ASH 2022). Diffusion pseudo-time (DPT) analysis suggested all WM samples could be placed on a shared evolutionary path independent of subtype, and when samples were separated into Early and Late DPT values, it mirrored disease progression. The role of epigenetic dysregulation that underlies subtype classification and disease evolution remains poorly understood in WM.

Methods: We therefore performed chromatin availability analysis using ATACSeq and bisulfite sequencing of WM patients. ATAC-Seq libraries were constructed from 37 CD19+ BM cells from treatment-naive WM patients along with 6 paired CD19+CD27- and CD19+CD27+ selected healthy donor (HD) peripheral blood (PB) controls and 3 HD CD138+ plasma cell (PC) samples. Enhanced reduced representational bisulfite sequencing (ERRBS) was also performed on 65 MYD88 mutated and 13 MYD88 wild type (WT) WM patients using CD19-selected BM lymphoplasmacytic cells. For 11 WM MYD88 mutated patients, CD138+ cells were also selected. In addition ERRBS was also performed on paired CD19+ and CD138+ selected samples from 5 IgM multiple myeloma (MM) patients, 6 paired CD19+CD27- and CD19+CD27+ selected healthy donor (HD) peripheral blood cells; 8 HD CD138+ selected PC; as well as MYD88 mutated BCWM.1, BCWM.2, MWCL-1, RPCI-WM1, OCI-Ly3, TMD-8, HBL-1 and SU-DHL2 cells; MYD88 wild-type OCI-Ly7, OCI-Ly19, Ramos B-cells. To infer 5-hydroxymethylcytosine (5HMC) levels, oxidative bisulfite sequencing was also performed for 73 samples including the HD, MM, WT and WM CD19/138 pairs. All ATAC-Seq and 37 of the WM ERRBS samples were part of our multi-omic study that included 253 treatment naïve patients who underwent whole exome and RNASeq sequencing.

Results: A UMAP projection of the CpG ERRBS data clustered WM patients into two distinct groups (Figure 1). One group clustered directly with HD Memory B-cells (MB) and was greatly enriched for CXCR4 mutated WM patients. The other WM group clustered with IgM MM patients and was largely CXCR4 WT. The findings are consistent with the BCL and PCL WM subtypes, which were confirmed by a perfect assignment correlation with our larger multi-omic data set wherein these subtypes were defined. Notably, BCWM.1, BCWM.2 and MWCL-1 WM cell lines clustered with the WM samples. The 5HMC data displayed no clear clustering of the patient samples but clearly separated all HD samples from those of WM and IgM MM samples, consistent with a role for 5HMC dysregulation in oncogenesis. Differential methylation analysis showed differences by MYD88 and CXCR4 mutation status, deletions in Chr6q, familial history for WM, WM subtype, and stratification by early and late DPT values. Notably, the only comparison that did not yield any differences was WM CD19+ to WM CD138+, suggesting that this may not be a meaningful epigenetic distinction in WM. Similar comparisons were performed on the ATAC peaks. Differential chromatin accessibility was noted near many of the corresponding differentially expressed genes. ATAC peak coverage significantly correlated with the expression of top genes. This was particularly true for WM subtype and DPT comparisons (Table 1). DNA sequences from differentially accessible peaks between BCL and PCL subtypes were also evaluated for relative TF motif enrichment in open vs. closed chromatin revealing PAX5, FUBP1, and ZNF121 enrichment in BCL subtypes using PCL subtype peaks as a background control. TFs REL, SP1, SPIB, BC11A and JUN were enriched PCL subtype compared with BCL subtype. Correlation between DPT associated genes revealed two clear groups, one associated with stem cell and pre-B-cell genes (Group 1) and the other with inflammatory, myeloid, and T-cell genes (Group 2). Motif analysis of differential peaks correlating with genes in each group demonstrated comparative enrichment for MYC, MXI1, and E2F2 in group 1 while RUNX1, NFKB1, and ATF1 were in enriched in group 2.

Conclusions: This is the first independent validation of our previously reported multi-omic driven WM subtype classification. The studies underscore that epigenetic differences underlie the biology of WM subclassification and support a strong role for epigenetic changes driving WM evolution.