NSD2 expression remodels DNA methylation, creating a targetable epigenetic dependency in t(4;14) multiple myeloma

Abstract Introduction: The t(4;14) translocation subtype of multiple myeloma (MM) is characterized by overexpression of NSD2, a histone methyltransferase which catalyzes di-methylation of histone H3 lysine 36 (H3K36me2). About 12% of MM patients have t(4;14) and this is associated with poor outcome. If and how NSD2 drives unique epigenetic vulnerabilities in this subtype of MM is not well understood. Methods: Primary MM samples from the MMRF CoMMpass trial (NCT01454296) were analyzed by Whole Genome Bisulfite Sequencing (WGBS). CRISPR/Cas9 was used to genetically ablate the NSD2 translocated allele in KMS11 and KMS18 (hereafter referred to as NSD2-high and NSD2-low). Multiple single cell clones were isolated and confirmed by immunoblot analysis and sequencing. KMS11 NTKO and TKO models (Kuo et al. Molecular Cell 2011) were also used for CUT14) subtype compared to other subtypes (p=1.17x10-12). Consistent with DNAm being connected to NSD2 overexpression, KMS11 NSD2-high cells had higher DNAm than isogenic NSD2-low cells (p1x10-9). Elevated DNAm in NSD2-high cells co-occurred in genomic regions with higher levels of H3K36me2 (OR=1.85, p1x10-15). Treating KMS11 and KMS18 NSD2 isogenic cell lines with the DNA hypomethylating agent DAC showed NSD2-high cells are significantly more sensitive to DNAm inhibition (p0.01 at ≥100 nM for KMS11 and p0.001 at ≥500 nM for KMS18). This sensitivity to DNAm inhibition was seen again using the DNMT1 enzymatic inhibitor GSK-3685032 (p0.001 at ≥100 nM). DBP of KMS18 showed NSD2-high cells increased mitochondrial priming when treated with DAC (p0.001) whereas NSD2-low cells did not significantly change, consistent with the increased apoptosis observed in NSD2-high cells. RNAseq of KMS11 NSD2-high and NSD2-low clones treated with 0, 100, or 500 nM DAC showed 3,599 differentially expressed genes (FDR0.01, fold-change2). Genes downregulated with DAC treatment were common between NSD2-high and NSD2-low cells and included genes involved in cell cycle and MYC targets as indicated by Gene Set Enrichment Analysis (GSEA). Genes upregulated with DAC-treatment were more distinct between NSD2-high and NSD2-low cells, with GSEA indicating interferon response and TNFa signaling via NFKB more significantly induced in NSD2-high cells (FDR0.0023) than in the NSD2-low cells (FDR0.13). Further investigation of DAC-induced RNAseq shows an increase in transcription of endogenous retroviruses (ERVs), specifically in the NSD2-high KMS11 cells. This suggests that hypomethylating agents induce transcription of these ERVs, which then induce an intracellular interferon response, and this is specific to NSD2-high MM. Consistent with this, these ERVs had higher levels of DNAm in isogenic KMS11 NSD2-high cells as compared to NSD2-low cells (FDR0.01). This was also observed in CoMMpass samples where higher DNA methylation was found at these ERVs in t(4;14) MM. Conclusions: NSD2-mediated increases in H3K36me2 in t(4;14) MM correspond with higher DNAm as compared to other MM subtypes. This is consistent with data from other cell types suggesting H3K36me2 is recognized by the PWWP domains of the DNMT3 DNA methyltransferasaes. Multiple NSD2 isogenic models suggest this creates a preferential sensitivity to DNA hypomethylating agents in NSD2-high MM. This phenomenon was shown both with DAC, a cytosine analog that incorporates into DNA, as well as with GSK-3685032, a DNMT1 enzymatic inhibitor. Both Annexin V staining and DBP indicated increased apoptosis, specific to NSD2-high cells. In addition, RNAseq data suggests this increased apoptosis is due to re-expression of ERVs that are triggering a viral defense response and interferon signaling. Consistent with this, t(4;14) cell line models and patient samples had higher levels of DNAm at ERVs, suggesting these are silenced through DNAm in t(4;14), but by other mechanisms in non-t(4;14) MM. Together, these data indicate NSD2 remodels the epigenome of t(4;14) MM creating unique epigenetic dependencies that can be exploited with readily available DNA hypomethylating agents.