SMARCA2 Is a Novel Interactor of NSD2 and Regulates Prometastatic PTP4A3 through Chromatin Remodeling in t(4;14) Multiple Myeloma

Multiple myeloma is a hematologic malignancy arising from post-germinal center B cells characterized by clonal terminally differentiated plasma cell infiltration in the bone marrow. Multiple myeloma tends to affect older people and is becoming more common in today's aging population. Recurrent chromosomal translocations involving the immunoglobulin heavy chain locus (IgH) on chromosome 14q32 are central to the pathogenesis of multiple myeloma, accounting for approximately 50% of total multiple myeloma cases (1–3). Chromosomal translocation t(4;14)(p16;q32) is one of the most common translocations, affecting 15% to 20% of patients with multiple myeloma. The t(4;14) translocation subtype is generally associated with poor prognosis, and patients experience frequent relapse after autologous stem cell transplantation (4). The translocation leads to the aberrant expression of two genes, fibroblast growth factor receptor 3 (FGFR3) and nuclear receptor-binding SET domain 2 (NSD2; ref. 5). Although the transforming activity of FGFR3 has been shown both in vitro and in vivo, about 30% of patients with t(4;14) multiple myeloma do not express FGFR3. Moreover, poor prognosis is independent of FGFR3 expression. In contrast, a universal characteristic seen in patients with t(4;14) multiple myeloma is the overexpression of all three NSD2 spliced isoforms, indicating its critical role in the disease (5–7).

NSD2 is a large multidomain histone methyltransferase that contains the evolutionarily conserved SET domain responsible for its enzymatic activity (8–10). Dimethylation of H3K36 (H3K36me2) is the principal chromatin-regulatory activity of NSD2 (11). The functional consequence of the H3K36me2 mark results in a gene-expression profile shift from normal plasma cells to one that favors plasma cell transformation (11). Widespread distribution of H3K36me2 leads to the inappropriate activation of genes involved in cell cycle, apoptosis, and cell adhesion (12, 13). Studies have shown that NSD2 knockdown inhibits myeloma cell proliferation, induces apoptosis, and reduces tumor formation (11–13). NSD2 is also required for the adhesion of t(4;14) multiple myeloma cells to the bone marrow stroma and subsequently rely on the microenvironment for proliferation (14). Furthermore, reintroduction of full-length NSD2 could reinstate cell growth in NSD2-depleted cells. These findings suggest that NSD2 plays a pivotal role in the pathogenesis of t(4;14) multiple myeloma.

SMARCA2 and SMARCA4 are two mutually exclusive ATPase subunits of the evolutionarily conserved SWI/SNF chromatin remodeling complex that is fuelled by energy derived from ATP hydrolysis to reposition nucleosome (15, 16). This leads to an open and accessible chromatin conformation and global transcriptional activation that drives cancer-related gene networks. SMARCA2 or SMARCA4, together with SNF5, BAF155, and BAF170, represent the core subunits of SWI/SNF complex. Genes encoding the SWI/SNF complex subunits are frequently mutated in about 20% of total cancer cases and 16% in multiple myeloma, indicating the importance of this complex in cancer progression (17, 18).

Although extensive research on NSD2 has been performed, the detailed molecular mechanism of how dysregulation of NSD2 contributes to malignant transformation and myeloma progression remains unclear. Furthermore, there is an unmet need with NSD2 as a therapeutic target that can achieve tangible clinical endpoints. In this study, we employed SILAC-based mass spectrometry (MS) to identify relevant pathways downstream of NSD2 important for (t4;14) myelomagenesis. This approach allowed us to identify 74 high-confidence interacting partners, where we further uncovered the interplay between SMARCA2 and NSD2 in chromatin regulation and transcription.

KMS11, KMS18, KMS34, H929, OPM2, KMS12BM, U266, and RPMI8226 cells were purchased from ATCC and cultured in RPMI1640 medium supplemented with 10% FBS, 100 U/mL penicillin, and 100 μg/mL streptomycin for 1–2 weeks before experimental use. TKO cells are derived from t(4;14)+ KMS11 cells, where the translocated IGH-NSD2 allele has been inactivated by homologous recombination to retain one copy of the wild-type NSD2 allele (nontranslocated allele). HEK293T cells were cultured in DMEM supplemented with 10% FBS, 100 U/mL penicillin, and 100 μg/mL streptomycin. All of the listed cells above were cultured at 37°C and 5% CO₂. Cell lines were periodically tested for Mycoplasma (Lonza) and authenticated (Centre for Translational Research and Diagnostics, National University of Singapore, Singapore).

In the “forward” pull-down experiment, NSD2-interacting proteins were detected by incubation of “heavy” cell lysate with anti-NSD2 antibody while control IgG antibodies were incubated with the “light” cell lysate. The beads were combined at 1:1 ratio before the last wash and bound proteins were analyzed by MS. Specific NSD2-interacting proteins possess a high heavy-to-light SILAC ratio (H/L ratio), whereas background proteins display a 1:1 H/L ratio. Vice versa, in the “reverse” pull-down experiment, NSD2-interacting proteins were detected by incubation of “light” cell lysate with anti-NSD2 antibody while control IgG antibodies were incubated with the “heavy” cell lysate. More information can be found in Supplementary Methods.

Cells were washed in ice-cold PBS following by lysis with RIPA buffer. Equal amounts of protein samples were boiled in Laemmli buffer (Bio-Rad) for 5 minutes at 95°C. Samples were subjected to SDS-PAGE separation and transferred to polyvinylidene difluoride membrane (Bio-Rad). The membranes were blocked in 5% BSA in TBST (TBS + 0.1% Tween-20) at room temperature for 1 hour, followed by incubation of primary antibodies overnight at 1:2,000 dilution and 1-hour incubation at with respective horseradish peroxidase–labeled secondary antibody at 1:5,000 dilution. Bands were visualized with chemiluminescent detection reagent (GE Healthcare, RPN2106). Coimmunoprecipitation was performed with 500 μg of nuclear lysate incubated overnight with respective antibodies. Dynabeads Protein G (Invitrogen) was used to harvest the immunoprecipitates. Two-percent input was used as a loading control. Antibodies used are listed in Supplementary Methods.

Total RNA was extracted with miRNeasy Mini Kit (Qiagen) according to the manufacturer's instructions. Sequencing library was generated with 100 ng of total RNA using the TruSeq RNA Sample Prep Kit v2 (Illumina) according to manufacturer's protocol. Sequencing was performed with HiSeq2500 (Illumina). Trimming of adapter from the sequencing fragments (raw data) was performed with Trip Galore (version 0.3.7) with default settings. The trimmed data were mapped to the human reference genome GRCh37 using STAR aligner (19) version 2.5.3. Differentially expressed genes (DEG) between control and NSD2 knockdown were performed in biological duplicates and analyzed with DEseq2 (https://bioconductor.org/packages/release/bioc/html/DESeq2.html; ref. 20) package in R. Benjamini–Hochberg method was used for P_adj values. DEGs with P < 0.05 were selected and the top 120 DEGs were used to generate the heat map using Microsoft Excel.

5 × 10 (6) KMS11 cells were subcutaneously injected into female NSG mice (6–8 weeks old; InVivos) and monitored for tumor growth. Once tumors reached a size of 150 mm³ [calculated as TV = (length × width (2)/2], mice were randomized into DMSO control (1% final concentration) or 5 mg/kg PFI-3 group and intraperitoneal injection of drugs (200 μL) was performed once every two days. At the end of the experiment, mice were sacrificed and tumor weights were measured and plotted. Representative tumor from each group was subjected to Western blotting analysis. The tumor cell pellet was harvested through grinding frozen tumors in cold PBS using mortar and pestle, and clear lysate was extracted using sonication with protein lysis buffer. The responsible use of animals was approved in accordance with the Institutional Animal Care and Use Committee protocol (National University of Singapore, Singapore).

Significant differences between samples were evaluated by unpaired two-sample Student t test qRT-PCR assays were performed in duplicate with three biological repeats; CTG viability assays were performed in octuplicates with three biological repeats; apoptosis and cell-cycle assays were performed in triplicates. A representative plot was shown. Values are expressed as mean ± SEM. P values <0.05 were considered to be statistically significant. Survival analysis was performed using Kaplan–Meier method and compared using log-rank test. Full details can be found in Supplementary Methods.

The RNA-sequencing (RNA-seq) data from this study were submitted to NCBI GEO (http://www.ncbi.nlm.nih.gov/geo/) under accession number GSE140860.

To identify novel physiologic interactors of NSD2, we performed endogenous immunoprecipitation (IP) with anti-NSD2 antibody and control IgG antibody with heavy and light SILAC-labeled cell lysate from KMS11 cells. The MS analysis identified 308 proteins with H/L ratio of 1 or higher in the “forward” experiment and 1 or lower in the “reverse” experiment (Fig. 1A; Supplementary Fig. S1). Full-length NSD2 is localized predominantly in the nucleus (10). To eliminate the possibility of unspecific binding of NSD2 to cytoplasmic proteins after whole-cell lysis and to increase the accuracy in determining NSD2 interactors, we repeated the SILAC MS with nuclear extract. The second MS identified 221 NSD2-interacting proteins (Fig. 1A, right; Supplementary Table S1). We compared the identified NSD2-interacting proteins from the whole-cell lysate and nuclear fraction and discovered 74 common proteins (Fig. 1A).

To gain a functional understanding of the identified proteins, we performed Gene Ontology (GO) analysis with DAVID (Database for Annotation, Visualization and Integrated Discovery, v6.8; refs. 21–23) and biological pathways of similar functions were clustered together and illustrated with a network enrichment map. Three main clusters were identified, suggesting that NSD2-interacting proteins are implicated in pathways regulating chromatin organization, RNA processing, and translation (Fig. 1B). To determine whether there were curated or predicted interactions among the identified NSD2-interacting proteins, we next performed an in silico protein–protein interaction analysis using the String Database (string-db.org; ref. 24). SMARCA2 and SMARCA4 were among the identified proteins that link to NSD2 through direct protein–protein interaction. They were predicted to interact with NSD2 based on coexpression and biochemical data of putative homologs interacting with one another in other species (Fig. 1B).

To validate the findings from the MS analysis, we selected high potential targets from each of the enriched functional groups for IP/immunoblot (IB) assay in KMS11 cells. The results indicated that NSD2 interacted with SMARCA2 and topoisomerase IIa, both of which regulate chromatin structure. NSD2 also binds to RNA-processing proteins ADAR1, DDX3X, and DDX5 as well as ribosomal proteins such as RPL4, RPS3, and RPS6 (Fig. 1C). Importantly, we were able to confirm these interactions in another t(4;14) cell line, KMS34, which also harbored the longest isoform of NSD2 (Fig. 1C). In contrast, GTPBP4 and MYBBP1A are examples of identified NSD2-interacting proteins that do not interact with NSD2.

Among the NSD2-interacting proteins confirmed by IP/IB assay, we chose to further investigate SMARCA2 as NSD2 might play additional roles in the regulation of chromatin. To confirm the interaction between NSD2 and SMARCA2, we performed immunoprecipitation of SMARCA2, which clearly showed an interaction between endogenous NSD2 and SMARCA2 in KMS11 and KMS34 cells (Fig. 2A). Next, Flag-tagged NSD2 and EGFP-tagged SMARCA2 were coexpressed in t(4;14)-negative KMS12BM and demonstrated a clear interaction between both proteins (Fig. 2B). Given the high sequence similarity between SMARCA2 and SMARCA4, we also checked whether SMARCA4 could bind to NSD2. However, no interaction was observed between SMARCA4 and NSD2 (Supplementary Fig. S2A). Next, we sought to determine whether NSD2 binds to other core subunits of the SWI/SNF complex. A weak interaction was observed between NSD2 and SNF5, but not BAF155 or BAF170 (Supplementary Fig. S2B). This suggested that NSD2 interaction with SMARCA2 might be noncanonical and independent of the SWI/SNF complex.

Next, we examined the expression of SMARCA2 in t(4;14)⁺ and t(4;14)⁻ cell lines. While SMARCA2 protein expression was detected in four out of five t(4;14)⁺ cell lines, it was low in all three of the t(4;14)⁻ cells (Fig. 2C). The protein expression of SMARCA2 corresponded to the mRNA levels, suggesting that the regulation is at the transcriptional level. Interestingly, knockdown of NSD2 did not affect SMARCA2 expression and vice versa, indicating that NSD2 and SMARCA2 do not directly regulate one another (Supplementary Fig. 2C). Therefore, to account for this differential expression of SMARCA2, we referred to previous studies implicating epigenetic silencing as mechanisms for loss of SMARCA2 (25–27). To test this hypothesis, we treated the SMARCA2-low cell lines (OPM2, KMS12BM, U266, and RPMI-8226) with histone deacetylase inhibitor (LBH-589), EZH2 inhibitor (GSK126) and DNA methyltransferase inhibitor (5-azacytidine). Interestingly, treatment with 5-azacytidine upregulated the transcript levels of SMARCA2 in majority of the cell lines (Fig. 2D). Next, we analyzed publicly available multiple myeloma patient dataset (MM-Meta cohort; PMID: 25214461) to determine whether the expression of SMARCA2 has prognostic significance. High expression of SMARCA2 alone does not confer adverse survival outcome (Supplementary Fig. S2D). However, patients with concurrent high expression of SMARCA2 and NSD2 predict inferior overall survival in a large cohort of patients with multiple myeloma (Fig. 2E). We also asked whether SMARCA2 correlates with myeloma disease stage and progression. On the basis of gene expression microarray dataset from the Mayo Clinic (28), SMARCA2 expression is highest in patients diagnosed with multiple myeloma and lowest in normal subjects (Fig. 2F). Taken together, this suggested that SMARCA2 is specifically important for NSD2-driven multiple myeloma.

SWI/SNF complex and its subunits can act as tumor suppressors or oncogenes in a context-specific manner (29, 30). Hence, we sought to further characterize the role of SMARCA2 as an interactor of NSD2 in multiple myeloma. First, the effects of SMARCA2 silencing on cell viability were investigated in human myeloma cell lines (HMCL). While achieving approximately 50% abrogation of SMARCA2 levels, the effect on growth inhibition was mild (Supplementary Fig. S3A). According to the paralog dependency model of the SWI/SNF complex (29), it was suggested that SMARCA2 and SMARCA4 could compensate for the loss of each other. To evaluate this possibility, we examined the endogenous expression of SMARCA4 in HMCLs. Unlike in solid tumors where loss of SMARCA4 occurred at a high frequency (31, 32), SMARCA4 could be detected in all HMCLs tested (Fig. 3A). Importantly, we found that SMARCA2 knockdown upregulated mRNA and protein levels of SMARCA4 (Fig. 3B). This suggested that the effects of SMARCA2 inhibition might be masked due to compensation from SMARCA4. Hence, this prompted us to perform a combinational knockdown of SMARCA2 and NSD2. Knockdown was achieved at both the mRNA and protein levels (Fig. 3C, right). In comparison with single knockdown, we observed an additive effect on growth retardation upon dual inhibition (Fig. 3C, left). On the contrary, knockdown of SMARCA2 had no antigrowth effect on t(4;14)⁻ cells (Supplementary Fig. 3B).

Abrogation of NSD2 reduced t(4;14)⁺ myeloma cell long-term clonogenic assay, and led to cell-cycle arrest at G₁–S phase, indicated by a decreased number of cells in S-phase accompanied by accumulation of cells in G₁ phase (Supplementary Fig. S4A–S4F). A repeat of the cell-cycle assays indicated that SMARCA2 knockdown likewise led to cell-cycle arrest at G₁ phase, indicated by accumulation of cells in G₁ and sub-G₁ population (Fig. 3D) with lesser changes in cell death (Fig. 3E), consistent with NSD2 role in regulating cell cycle. Together, these indicated that silencing of SMARCA2 in NSD2-overexpressing cells inhibits myeloma cells in vitro.

To identify downstream gene targets of NSD2, we performed RNA-seq in KMS11-scrambled control versus KMS11-NSD2 knockdown cells (Supplementary Fig. S4A). Upon NSD2 abrogation, 921 genes were downregulated and 299 genes were upregulated (Fig. 4A and B). Functional analysis indicated that these genes are associated with key pathways such as cell growth, cell adhesion, and integrin-mediated signaling, which reinforced findings from previous studies. In addition, our analysis also identified enrichment of genes involved in IFNγ-mediated signaling pathway and regulation of Wnt signaling pathway (Supplementary Table S2). Among the downregulated targets, PRL-3 (PTP4A3) is an important gene frequently overexpressed in multiple myeloma (33–35). We have previously reported that the expression of PTP4A3 was highest in the t(4;14)⁺ subgroup and the expression of PTP4A3 was associated with multiple myeloma disease progression (35) (Supplementary Fig. S5A). Here, we further determined the correlation between the expression of NSD2 and PTP4A3 in patient datasets (namely CoMMpass, UAMS, and Mayo datasets), which showed a positive correlation in at least three multiple myeloma cohorts (Fig. 4C). To demonstrate that PTP4A3 expression is important for NSD2-myeloma cells, we knocked down PTP4A3 in KMS11 and KMS34 cells. Approximately 80% knockdown of PTP4A3 mRNA levels resulted in slower cell growth in these cells (Supplementary Fig. S5B). Till date, the mechanism of how NSD2 drives the aberrant expression of PTP4A3 remains unknown.

In other cancer models, literature mining suggested that PTP4A3 could be a potential downstream target of SMARCA2 as well (36, 37). Given the association, this prompted us to examine whether PTP4A3 might be a common downstream target of NSD2 and SMARCA2 in multiple myeloma. Using the lentiviral system, NSD2 or SMARCA2 targeting shRNAs were delivered into KMS11 cells. Subsequent qRT-PCR analysis and IB assay confirmed that ablation of either NSD2 or SMARCA2 (Fig. 4D; Supplementary Fig. S6) or dual knockdown of NSD2 and SMARCA2 (Fig. 4E) could downregulate the expression of PTP4A3. To complement the knockdown assay, we ectopically expressed NSD2 and SMARCA2 in t(4;14)⁻ KMS12BM (NSD2 and SMARCA2 low) cells. Concurrent overexpression of both NSD2 and SMARCA2 were required for the upregulation of PTP4A3 (Fig. 4F), suggesting that NSD2 and SMARCA2 are upstream regulators of PTP4A3.

NSD2 is a histone methyltransferase that regulates H3K36 dimethylation at gene promoter region (11) while SMARCA2 is part of the SWI/SNF complex that remodels chromatin through nucleosome repositioning (30). Given their proposed roles, we first determine whether SMARCA2 might affect H3K36 dimethylation levels through regulating NSD2 accessibility to the chromatin. We performed SMARCA2 knockdown and examined the global levels of H3K36me2. KMS11-TKO cells showed an almost complete abrogation of H3K36me2 levels, which was consistent to previous report (11). On the contrary, knockdown of SMARCA2 did not affect the global levels of H3K36me2 (Fig. 5A). Despite this, we reasoned that the regulation might be initiated at the focal level on specific gene promoters. Hence, we examined whether the regulation of PTP4A3 by NSD2 is associated with SMARCA2 through promoter binding. Chromatin immunoprecipitation assay (ChIP) was performed using primers targeting the promoter region upstream of transcriptional start site of PTP4A3 in the well-characterized isogenic cell line pair, KMS11 and TKO cells. ChIP assay revealed an enrichment of NSD2, H3K36me2 mark and SMARCA2 at PTP4A3 promoter in the KMS11 cells, and binding of these factors were reduced upon the loss of NSD2 in the TKO cells (Fig. 5B; Supplementary Fig. S6A). Furthermore, decrease in Pol II promoter occupancy justified the reduction in PTP4A3 gene expression in TKO cells. To determine whether SMARCA2 influenced the binding of NSD2 to PTP4A3 promoter, we repeated the ChIP assay using cells transfected with SMARCA2 shRNA. NSD2 and H3K36me2 levels at PTP4A3 promoter were reduced upon SMARCA2 knockdown (Fig. 5B). To further establish that the mechanism of NSD2 and SMARCA2 coregulating target genes is not restricted to PTP4A3, we repeated the experiments on CCND1 promoter. CCND1 is another gene we identified to be coregulated by NSD2 and SMARCA2 in our RNA-seq data (Fig. 4B; Supplementary Fig. S6B) and is highly correlated with MMSET (38). ChIP assays yielded similar results (Fig. 5C). These results suggested that NSD2 and SMARCA2 cooperatively activate gene expression through focal changes in promoters of myeloma-associated genes.

If SMARCA2 is important for NSD2 cancers, accordingly, t(4;14)⁺ cells will be selectively sensitive to SMARCA2 inhibition. Given the promising antitumor activities of BETi like JQ1 in preclinical models (39), we evaluated a bromodomain inhibitor, PFI-3, in our multiple myeloma cells. PFI-3 could selectively target SMARCA2 and SMARCA4 through the displacement of these proteins from the chromatin, but not other bromodomain-containing proteins (37). Dose-dependent treatment of myeloma cells with PFI-3 showed a striking difference in the sensitivity of the cells with t(4;14) translocation. While t(4;14)⁺ cells see a reduction of approximately 30% to 40% in viability, t(4;14)⁻ cells demonstrated resistance up to 50 μmol/L of PFI-3 in short-term viability assays (Fig. 6A; Supplementary Fig. S7). PFI-3 also inhibited the colony-forming ability of the myeloma cells in soft agar in terms of number and size of the colonies (Fig. 6B). Further analysis indicated that the reduction in cell viability observed in t(4;14)⁺ cells was attributed to an increase in apoptotic cell fraction as determined by flow cytometry (Fig. 6C). Given the high concentrations of drug treatment, we undertook rescue experiments to demonstrate the specificity of the drug. While KMS11 cells treated with PFI-3 showed approximately 35% decrease in cell viability and knockdown of SMARCA2 showed approximately 20% decrease in cell viability, SMARCA2 knockdown cells treated with PFI-3 as compared with its DMSO control did not demonstrate further reduction in cell viability (Fig. 6D and E). This indicated that knockdown of SMARCA2 could impart resistance to PFI-3 and hence the inhibitory effects were SMARCA2-dependent. In addition, we also checked the expression of BRD4, which is a prominent target of BETi (40). Notably, PFI-3 treatment did not alter the expression levels of BRD4 (Fig. 6F).

Next, we studied the consequence of PFI-3 treatment on PTP4A3 expression. PFI-3 treatment suppressed PTP4A3 with no changes in SMARCA2 and SMARCA4 transcript levels (Fig. 6G). This led us to hypothesize that although PFI-3 does not affect SMARCA2 expression, it might defer the occupancy of SMARCA2 on the promoter of PTP4A3. To explore this hypothesis, we performed ChIP assay after PFI-3 treatment. Notably, we observed a reduction in SMARCA2 that was bound to PTP4A3 promoter, with a concomitant decrease in NSD2 binding (Fig. 6H). Finally, we determined the in vivo effects of PFI-3 inhibitor on t(4;14) tumors. KMS11 cells were chosen among the t(4;14) cell lines due to its good engraftment ability in our mouse model. After tumors started developing following 2 weeks of inoculation, they were randomized into DMSO control group or treatment with 5 mg/kg PFI-3. Throughout the treatment process, no toxicity effects on the mice were observed. Tumor weight determination indicated that PFI-3 was efficacious in reducing tumor growth, and subsequent Western blot analysis of apoptotic proteins revealed a higher cleavage of caspase-3 and PARP in the PFI3-treated tumors (Fig. 6I). Altogether, these results suggested the feasibility of targeting NSD2 through SMARCA2 in vivo and in vitro.

Finally, we sought to determine the functional consequences of NSD2 and SMARCA2-mediated upregulation of PTP4A3 and how it contributes to a myeloma phenotype. Previously, we reported that PTP4A3 is an upstream regulator of MYC levels via various mechanisms such as the IL6–STAT3 pathway (in multiple myeloma) or β-catenin pathway (in acute myeloid leukemia; refs. 35, 41, 42). Notably, MYC activation signature is highly correlated with myeloma progression (43) and MYC is a downstream target of a 10-gene signature (which included PTP4A3) that predicts MMSET patient survival (38). These compelling evidences not only indicated that MYC is important for t(4;14)⁺ myelomagenesis, but also suggested that PTP4A3 could be the gene driving MYC expression in these translocated cells.

In the KMS11 and TKO isogenic cells, we found that KMS11-TKO cells harbored significantly lower levels of MYC (Fig. 7A). In addition, knockdown of PTP4A3 in KMS11 cells resulted in a perturbation of MYC mRNA and protein levels (Fig. 7B), which suggested that PTP4A3 was responsible for maintaining the high levels of MYC in NSD2^high cells. Notably, levels of MYC in t(4;14)⁻ KMS12BM could likewise be upregulated indirectly by enforced expression of NSD2 and SMARCA2, or directly by expression of PTP4A3 (Fig. 7C). Finally, we extended our analysis to multiple myeloma patient dataset for correlation between PTP4A3 with MYC levels and MYC activation signature. A comparison of the ρ values indicated that the correlation is highest in the t(4;14) group as compared with other disease classifier for myeloma (TC classification; Fig. 7D; Supplementary Fig. S8), which further exemplified our previous findings that PTP4A3 expression is highest in the t(4;14) subtype (35). This correlation is biologically significant given that MYC can be dysregulated via multiple mechanisms in cancer.

NSD2 is the primary oncogenic driver and represent a good druggable target due to its universal overexpression in t(4;14) multiple myeloma. However, given the challenges in the development of small-molecule inhibitors against NSD2, we took on a discovery approach to characterize NSD2 interactome as an alternative strategy to target NSD2. For a complete overview of NSD2-interacting proteins, we selected the cell line KMS11, which expresses the full-length MMSET protein arising from transcription at the MB 4–1 breakpoint (10). We performed bidirectional immunoprecipitation with SILAC-based proteomics, which was selected over other labeling methods for comparative accuracy and reproducibility. Finally, by overlapping the respective potential binding partners identified in the whole-cell lysate and nuclear extracts, we narrowed down to 74 high-confidence interacting proteins.

These 74 proteins are enriched in chromatin organization, RNA processing, and translation-related processes. Critically, these identified proteins are clearly distinct from the NSD2-interacting partners published in the curated database, BioGRID. Approximately 20% (15/74) of the proteins identified in our MS analysis overlap with those listed in BioGRID, while the remaining 80% (59/74) are novel interactors. We speculate that the novel NSD2 interactors specific to t(4;14) multiple myeloma might play important role(s) in the pathogenesis of myeloma, whereas the overlapping NSD2-interacting proteins regulate general downstream functions of NSD2, independent of cell types. We showed that NSD2 interacts with ADAR1, hnRNPA2B1, and DDX5, of which, they play important roles in RNA editing and alternative pre-mRNA splicing (44, 45). In addition, Mirabella and colleagues (46) reported that ReIIBP, one of the NSD2 isoforms, mediates RNA splicing through binding and methylating subunits of the SMN (survival of motor neuron) complex. Therefore, it is possible that full-length NSD2 may regulate other aspects of RNA processing via interaction with ADAR1, hnRNPA2B1, and DDX5. Furthermore, our study also indicated that NSD2 interacts with multiple ribosomal proteins, which reinforced two recently published data that multiple myeloma SET isoforms interacted with several ribosomal proteins (46, 47). multiple myeloma cells have high protein turnover rate and proteasome-mediated protein degradation plays an essential role in maintaining a homeostatic environment in multiple myeloma cells. Bortezomib inhibits proteasome activity, leading to accumulation of proteins that causes cell stress and cell death. Clinically, bortezomib has been shown to improve outcome in patients with t(4;14) multiple myeloma (48, 49). Our results suggested that NSD2 might also be involved in translational regulation and warrants further investigation in this area as well.

The role of NSD2 on chromatin regulation and transcription is complex and not fully understood. Several studies have indicated that through its canonical SET domain, NSD2 directly regulate genes through a global expansion of H3K36me2 mark in the intergenic regions, leading to widespread gene activation (11). Concomitantly, a reduction in the repressive mark H3K27me3 was reported (50), suggesting an interplay between NSD2 and another histone methyltransferase, EZH2. NSD2 is also reported to alter H3K27 acetylation, which leads to a more open chromatin conformation (51). In addition, it is known that histone modification marks could recruit chromatin remodeling factors (52). This highlights the importance of our findings that NSD2 and SMARCA2 were found to affect each other's recruitment to the promoter of their coregulated oncogenes, where NSD2 and SMARCA2 led to a coordinated effect that is permissible for transcription despite catalyzing fundamentally different biochemical processes. In particular, PTP4A3 was found to be upregulated by NSD2 and SMARCA2. PTP4A3 was highlighted as an important myeloma gene when Broyl and colleagues (33) discovered a novel cluster of patients with newly diagnosed multiple myeloma that showed elevated PTP4A3, and dataset analysis indicated that patients with t(4;14) translocation expressed the highest levels of PTP4A3 amongst the subgroups. We and others have demonstrated that PTP4A3 expression could be driven by mitogenic cytokines like IL6, and PTP4A3 promotes myeloma cell viability, clonogenicity, and tumorigenesis (35). Here, we identified a novel epigenetic mechanism that leads to PTP4A3 overexpression by NSD2 and SMARCA2. Furthermore, the elevated levels of PTP4A3 in t(4;14) cells were biologically important for maintaining the myeloma phenotype as it drives an upregulation of MYC. This is in agreement with our previous study where we identified MYC as one of the downstream targets of PTP4A3, through a feedforward loop involving STAT3 occupancy on MYC promoter and dependent on the phosphatase activity of PTP4A3 (41).

In solid tumors, SWI/SNF complex has a proposed tumor suppressor role due to the frequent loss-of-function mutations; however, SWI/SNF mutations do not emerge as recurrent genetic alterations in multiple myeloma (53). Furthermore, a paralog dependency model has been used to describe the SWI/SNF complex, whereby SMARCA2 and SMARCA4 are able to compensate for the loss of each other, as with the case for ARID1A and ARID1B (29). It appears that in SMARCA4 loss-of-function cells, SMARCA2 is upregulated and assembled into the residual SWI/SNF complex, which underlies the oncogenic switch (54). This is also the reason why multiple groups have independently identified SMARCA2 as a synthetic lethality gene in SMARCA4-mutant cases (31, 32, 54). These observations supported our hypothesis that SMARCA2 might be differentially oncogenic in the specific context of t(4;14) myeloma. Given the preclinical success of BETi such as JQ1, we chose to investigate PFI-3, with drug specificity against SMARCA2/4 bromodomain (37). Treatment of myeloma cell lines with PFI-3 showed a selective sensitivity of t(4;14)⁺ cells to PFI-3, while t(4;14)⁻ cells remained resistant. Importantly, PFI-3 treatment did not perturb the levels of NSD2 nor SMARCA2, rather, the targeting effect is achieved through their occupancy on oncogenes. This suggested that a subset of the genes coregulated by NSD and SMARCA2 played a pivotal role in myeloma cell survival.

Collectively, we demonstrated a proof-of-concept of how histone-modifying enzyme and chromatin-remodeling complex interplay to cause an altered epigenetic state in myelomagenesis through the regulation of an important myeloma gene. Targeting such protein–protein interactions on target genes instead of NSD2 expression is likely to be more specific and can be exploited for therapeutic intervention.

No disclosures were reported.

P.S.Y. Chong: Conceptualization, writing–original draft, project administration, writing–review and editing. J.Y. Chooi: Conceptualization, methodology, writing–original draft, project administration. J.S.L. Lim: Investigation, methodology. S.H.M. Toh: Investigation, methodology. T.Z. Tan: Data curation. W.-J. Chng: Conceptualization, supervision, writing–original draft, project administration, writing–review and editing.

The authors thank Dr. Xie Zhigang for the NSD2 materials and helpful discussions on the project, and Dr. Dennis Kappei for helpful advice on mass spectrometry results. The authors also thank Prof. Qi Zeng for providing the PRL3 antibody and Dr. Nicholas Grigoropoulos for providing the DDX3X antibody. The authors thank Dr. Zhou Jianbiao for advice on the in vivo mouse work. W.-J. Chng is supported by NMRC Singapore Translational Research (STaR) Investigatorship. This research is partly supported by the National Research Foundation Singapore and the Singapore Ministry of Education under the Research Centers of Excellence initiative as well as the RNA Biology Center at the Cancer Science Institute of Singapore, NUS, as part of funding under the Singapore Ministry of Education's Tier 3 grants (grant number MOE2014-T3-1-006). P.S.Y. Chong is supported by Open Fund Young Individual Research Grant awarded by National Medical Research Council (NMRC/OFYIRG/0039/2017).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.