Histone lysine methyltransferase NSD2 (WHSC1/MMSET) is overexpressed frequently in multiple myeloma due to the t(4;14) translocation associated with 15% to 20% of cases of this disease. NSD2 has been found to be involved in myelomagenesis, suggesting it may offer a novel therapeutic target. Here we show that NSD2 methyltransferase activity is crucial for clonogenicity, adherence, and proliferation of multiple myeloma cells on bone marrow stroma in vitro and that NSD2 is required for tumorigenesis of t(4;14)+ but not t(4;14)− multiple myeloma cells in vivo. The PHD domains in NSD2 were important for its cellular activity and biological function through recruiting NSD2 to its oncogenic target genes and driving their transcriptional activation. By strengthening its disease linkage and deepening insights into its mechanism of action, this study provides a strategy to therapeutically target NSD2 in multiple myeloma patients with a t(4;14) translocation. Cancer Res; 73(20); 6277–88. ©2013 AACR.
Compelling data has emerged that epigenetic mechanisms are linked with various pathologies such as cancer, infectious diseases, inflammation, degenerative diseases, and neurological disorders (1–4). Many epigenetic players possess genetic alterations in human diseases (5–8), suggesting important roles in driving deregulated signaling pathways.
The nuclear SET domain (NSD) containing protein family of histone methyltransferases includes NSD1, NSD2 (also named as WHSC1, Wolf–Hirschhorn syndrome candidate 1; or MMSET, multiple myeloma SET domain containing protein), and NSD3 (also known as WHSC1L1, WHSC1 like 1). Each family member has been linked to multiple diseases. NSD2 is involved in the t(4;14) translocation in 15% to 20% of multiple myeloma (9), which is the second most common translocation and associated with unfavorable prognosis (10–12). When the translocation occurs, the Eμ and 3′ enhancer of IgH locus are fused with NSD2 and FGFR3, respectively, resulting in overexpression of both genes (13). About 5% to 10% of t(4;14)+ myeloma further develop activating mutations in FGFR3 (14–16), indicating its important role in tumor progression. However, 30% of t(4;14)+ cases lack FGFR3 expression due to loss of der(14), whereas all overexpress NSD2 (17). Furthermore, poor prognosis of t(4;14)+ myeloma is independent of FGFR3 expression (18), suggesting that NSD2 dysregulation is the crucial oncogenic event.
NSD2 is a multidomain protein containing a catalytic SET (suppressor of variegation, enhancer of zeste and trithorax) domain, 2 PWWP (proline-tryptophan-tryptophan-prolin motif) domains, an HMG (high mobility group) box, 5 PHD (plant-homeodomain) zinc finger motifs, and a Cys-His rich C5HCH domain. The PWWP, HMG, and PHD domains have been shown to mediate chromatin interaction and recognition of histone marks (19–23). How different domains cross talk and contribute to the biological functions of NSD2 is not understood yet. In addition, at least 3 splicing isoforms of NSD2 have been reported: MMSET II (1-1365aa), MMSET I (1-647aa), and REIIBP (782-1365aa). Depending on the breakpoints in different t(4;14)+ myeloma cells, there can be N-terminal truncations up to the PWWP1 domain in MMSET I and MMSET II (24, 25). All variants of NSD2 are overexpressed under t(4;14) translocation whereas the outcome of t(4;14)+ myeloma is independent of the breakpoint types (26). Whether different NSD2 isoforms play differential roles remains to be determined.
In t(4;14)+ myeloma, NSD2 has been linked to cell-cycle progression, DNA damage response, cell adhesion, clonogenicity, and tumorigenesis (25, 27–31), presumably by reprogramming global histone methylation and gene expression. Two recent studies have shed light on the chromatin-regulatory functions of NSD2, and the connection between the imbalance of chromatin homeostasis and myelomagenesis (30, 31). Martinez-Garcia and colleagues showed that NSD2 overexpression in t(4;14)+ myeloma is correlated with a global increase of H3K36me2 and decrease of H3K27me3. Loss of NSD2 alters cell adhesion, inhibits cell growth, and induces apoptosis. Kuo and colleagues also identified H3K36me2 as the principal physiologic product of NSD2 but did not observe the reciprocal regulation of H3K27me3. They proposed that the dispersed H3K36me2 distribution throughout the chromatin leads to initiation of an oncogenic gene expression program, which drives myelomagenesis. These studies suggest that NSD2 inhibition may have therapeutic value for t(4;14)+ myeloma. However, questions remain such as how NSD2 activity influences myelomagenesis; how different domains within NSD2 contribute to its biological functions; and most importantly, whether targeting NSD2 can benefit myeloma patients stratified by t(4;14) translocation.
In this study we show that loss of NSD2 has minor growth-inhibitory effects on t(4;14)+ myeloma cells. However, depletion of NSD2 or inhibition of its activity dramatically impairs clonogenicity, adherence, and proliferation on bone marrow stroma, and tumorigenesis of t(4;14)+ but not t(4;14)− myeloma lines. NSD2 modulates histone methylation and chromatin remodeling, and mediates transcription activation of a t(4;14)+ specific myeloma gene set. Furthermore, PHD domains of NSD2 are important for its recruitment to target loci and transcription activation of oncogenic genes, and consequently biological functions in myeloma. Our findings suggest that NSD2 is a therapeutic target for t(4;14)+ myeloma and reveal potential ligand pockets for pharmacological intervene.
Materials and Methods
All myeloma lines were cultured in RPMI1640 (Gibco) with 12.5% FBS (Gibco), except for LP1, which was maintained in IMEM (Gibco) with 20% FBS. For cells infected with Tet-on inducible short hairpin RNAs (shRNA), Tet system approved FBS (Clontech) was used. KMS11-TKO and parental line were obtained from Horizon Discovery Ltd. 293T cells were purchased from American Type Culture Collection and kept in DMEM (Gibco) with 10% FBS. Bone marrow stromal cells (BMSC) were purchased from Millipore and cultured in mesenchymal stem cell culture media from the provider. The identity of cell lines was verified by STR genotyping (Promega).
NSD2 (Abnova, in-house); H3K4me3 (Millipore); H3K36me2 (Cell Signaling, Millipore); H3K36me3 (Abcam); H3K27me3 (Millipore, Cell Signaling, in-house); H3 (Cell Signaling, Abcam); Flag (Sigma); β-actin (Sigma); FGFR3 (Santa Cruz).
shRNAs were cloned into pLKO-H1-TO/puromycin for Tet-on inducible expression. Sequences of shRNAs: NSD2 (ACATGCTCACTATAGACAA); BACE2 (sh1–CCCAATACTTATGTTGTATTG, sh2–GCTTTCAAATCCTCCCTACTT, sh3–CTCCTACATAGACACGTACTT, sh4–CCCTGAGGTCGTCAATGATGA, sh5–GACATCGCTGGAAATGAATAG); CA2 (sh3–CCCTGGTTGCTTTGTGTCTAG, sh4–GCTCAGCCACTGAAGAACAGG, sh5–GCACTGGCATAAGGACTTCCC); IGF2BP2 (sh1–GGGCTTGACCATAAAGAACAT, sh3–CCCTCTCGGGTAAAGTGGAAT, sh5–GATCGGGAGCAAACCAAAGAC).
96-well plates were coated with 50 μL 0.6% low melting point agarose (Invitrogen), layered with 0.27% agarose containing myeloma cells, and topped with growth medium. Medium containing 1 μg/mL doxycycline (Sigma) was replenished every 3 to 4 days. Colonies were either stained with p-iodonitrotetrazolium violet (Sigma) for microscopic imaging and counting, or measured by AlamarBlue (Invitrogen) using fluorescence (545 nm excitation and 615 nm emission wavelength) on Envision multilabel reader (Perkin-Elmer) with background subtracted.
Bone marrow stromal adhesion/proliferation assays
A modified bone marrow stromal adhesion assay (32) was developed to measure the attachment of myeloma cells to BMSCs. Briefly, BMSCs were grown in 96-well plates to reach confluence. Luciferase-tagged myeloma cells was seeded and allowed to adhere for 2 hours. The plates were inverted and spun at 150 × g for 5 minutes to remove unattached cells and luminescence was measured to calculate the percentage of adhered myeloma cells using standard curve. To measure the proliferation of myeloma cells on BMSCs, the experimental procedure was modified in which all myeloma cells were kept and incubated for 3 days before luminescence read-out.
Six- to eight-week-old female SCID-beige mice (Vital River) were injected intravenously or subcutaneously with 107 myeloma cells tagged with luciferase, in 100 μL PBS with 50% Matrigel (BD Biosciences), 10 mice per group/treatment. To induce shRNA, mice were administrated with drinking water containing 5% sucrose and 0.2 mg/mL doxycycline. Luminescence images were taken weekly with d-luciferin (200 μL/mouse) intraperitoneal injection. Subcutaneous tumor volume (V = 1/2L × W2) and body weight were measured. Paralysis and death were monitored and recorded.
NSD2 is important for clonogenicity of t(4;14)+ myeloma cells
It has been reported that NSD2 knockdown in t(4;14)+ myeloma lines inhibited cell growth, caused cell-cycle arrest, and induced apoptosis (25, 28, 30). However, when we depleted NSD2 by an inducible shRNA (referring to shNSD2 below) across the t(4;14)+ and t(4;14)− myeloma cell panel, t(4;14)+ myeloma lines only showed none to mild growth inhibition, with no effects observed on t(4;14)− cells (Supplementary Fig. S1A and S1B). In addition, KMS11 cells with NSD2 knockout on the translocated allele (TKO; ref. 28) only showed a modest growth slowdown compared with the parental line (PAR; Fig. 1A; Supplementary Fig. S1C). Similarly, TKO cells reconstituted with wild-type MMSET II (WT) also showed only slight growth advantage over those with catalytic dead mutant (CDM, R1138A/C1144A; Fig. 1B; Supplementary Fig. S1D; ref. 33). Thus, loss of NSD2 protein or its activity is not lethal for t(4;14)+ myeloma cells under regular culture conditions, and proliferation advantage per se might not be the disease linked effect of NSD2 overexpression in t(4;14)+ cases.
One of the important features of myeloma cells is that they adhere to bone marrow stroma and colonize within the microenvironment. To investigate the role of NSD2 in a disease-relevant setting, we evaluated if it is important for clonogenicity of t(4;14)+ myeloma cells. Strikingly, loss of NSD2 dramatically impaired clonogenic growth of KMS11-TKO cells compared with PAR lines (Fig. 1C). WT MMSET II but not CDM rescued the deficiency of TKO to colonize (Fig. 1C), indicating that the catalytic activity is required. Next we asked whether NSD2 knockdown can differentially affect t(4;14)+ versus t(4;14)− myeloma cells on clonogenicity. Doxycycline induced shNSD2 expression inhibited colonization of t(4;14)+ KMS11, KMS18, KMS28BM, and LP1 cells, regardless of their FGFR3 expression and mutation status (Fig. 1D and E), suggesting that NSD2 rather than FGFR3 might be the major driver. In contrast, depletion of NSD2 in t(4;14)− RPMI822 and L363 cells had no effect on clonogenic growth (Fig. 1F). Thus, using clonogenicity assay, we showed that t(4;14)+ myeloma patients may benefit from inhibition of NSD2 over t(4;14)− population.
NSD2 mediates myeloma cell adherence and proliferation on bone marrow stroma
The results above suggested that NSD2 is required for the clonogenicity of t(4;14)+ myeloma cells, which may reflect their ability to attach to bone marrow stroma and subsequently colonize within the microenvironment. Therefore, we developed bone marrow stromal adhesion assay to measure the attachment of myeloma cells to BMSCs. NSD2 knockout impaired the adhesion of KMS11 to BMSCs (Fig. 2A). The deficiency of TKO to adhere to BMSCs can be restored by WT MMSET II but not CDM (Fig. 2A), suggesting that NSD2 activity is required. NSD2 knockdown in t(4;14)+ KMS11 and KMS28BM inhibited their adherence on BMSCs, whereas depletion of NSD2 in t(4;14)− RPMI8226, L363, and KMS12BM had no effect (Fig. 2B and C). Therefore, we for the first time showed that NSD2 is important for t(4;14)+ but not t(4;14)− myeloma interaction with bone marrow stroma.
The myeloma–stromal interaction can promote growth, survival, and migration of tumor cells (34). To determine whether NSD2-mediated adherence to stroma can influence myeloma cell growth, we measured the proliferation of KMS11-PAR, TKO, WT, and CDM cells on BMSCs. Notably, the growth of PAR cells was dramatically stimulated upon contact with BMSCs, whereas TKO cells did not show much response (Fig. 2E). These data suggest that NSD2-mediated myeloma–stromal interaction may promote the proliferation of tumor cells. WT MMSET II marginally but significantly increased cell growth of TKO on BMSCs, whereas CDM did not have any effects (Fig. 2E), suggesting the involvement of catalytic function of NSD2.
The reasons that WT MMSET II could only partially rescue myeloma cell proliferation on BMSCs are not fully understood. It could be due to the clonal difference between PAR and TKO cells. Another possibility is that TKO lost both MMSET II and MMSET I isoforms, whereas WT only restored MMSET II. Although MMSET I does not contain the SET domain and has no enzymatic activity, we cannot exclude the possibility that PWWP1 and HMG domains in MMSET I can facilitate MMSET II binding to chromatin by forming dimers (24) or recruit complexes to potentiate myelomagenesis. Therefore, we reconstituted both MMSET I and MMSET II in TKO cells (Fig. 2D). TKO rescued by both isoforms showed significant improvement of proliferation on BMSCs compared with those with MMSET II alone (Fig. 2E), suggesting a synergetic effect.
NSD2 is important for tumorigenesis of t(4;14)+ myeloma cells
To elucidate the role of NSD2 in myelomagenesis, we injected KMS11-luciferase cells stably transduced with inducible shNSD2 into SCID mice intravenously. KMS11 cells showed tumor homing to bone marrow, as determined by FACS analysis (Supplementary Fig. S2A). Following NSD2 knockdown, total luminescence signal was decreased, suggesting that depletion of NSD2 inhibited tumorigenesis (Fig. 3A and B). In addition, we monitored the development of paralysis in groups of mice with or without NSD2 knockdown. Although control mice exhibited severe myeloma-induced paralysis, induction of shNSD2 delayed the disease progression (Fig. 3C). Therefore, NSD2 is important for tumor initiation and progression of t(4;14)+ myeloma. In contrast, NSD2 knockdown in t(4;14)− L363 had no significant impact on tumorigenesis (Supplementary Fig. S2B).
To show a direct link between myelomagenesis and NSD2 activity, we inoculated KMS11-PAR, TKO, WT, and CDM cells tagged with luciferase into SCID mice. Consistent with the results obtained with shRNA-induced knockdown, NSD2 knockout in TKO led to loss of tumor formation, either in intravenous or subcutaneous models (Fig. 3D; Supplementary Fig. S2C). WT but not CDM MMSET II restored tumorigenesis of TKO, although with a lower tumor-take rate and smaller tumor sizes than PAR (Fig. 3D), suggesting that at least partial tumorigenic potential is dependent on NSD2 activity, and TKO cells rescued by both MMSET I and MMSET II showed enhanced tumorigenesis compared with those with MMSET II alone (Supplementary Fig. S2D), implicating that BMSC proliferation assay could well predict the in vivo response of tumor growth.
In summary, our data suggested that NSD2 is important for tumorigenesis of t(4;14)+ myeloma cells, and inhibition of NSD2 may have therapeutic value for this subtype of myeloma patients.
PHD domains are important for biological functions of NSD2
Our studies have showed that the catalytic activity of NSD2 is crucial for its oncogenic potential. However, NSD2 has multiple chromatin-binding motifs including 2 PWWP, 1 HMG, 5 PHD, and 1 PHD-like C5HCH domains (Fig. 4A). Therefore, we asked whether these reader domains contribute to its biological functions. We generated a series of truncations and scanned their cellular activities using “exo-H3” assay, in which coexpression of MMSET II with a H3-luciferase construct induced H3K36 dimethylation on the newly synthesized exogenous histone H3 in a methyltransferase activity-dependent manner (Fig. 4B). The methylation event happened only in the chromatin-bound fraction, suggesting that the interaction with chromatin is required (Supplementary Fig. S3B). We found that PHD1-3 motifs are important for NSD2 induction of H3K36me2, whereas PWWP1, HMG, and PHD5-C5HCH domains are dispensable (Fig. 4B). Specifically, removal of PHD1 decreased H3K36me2 activity, whereas further truncation of PHD2 and part of PHD3 (REIIBP isoform) caused NSD2 mislocalization into cytoplasm thus resulted in complete loss of activity on histones (Fig. 4B; Supplementary Fig. S3B).
It has been shown that 11 of 12 Sotos mutations in PHD4-6 of NSD1 disrupted binding to H3K4me2 and H3K9me2, and 8 of 9 mutations in PHD4 and PHD6 severely compromised binding to transcription cofactor Nizp1 (21). Therefore, we generated single-point mutations on C720, C743, L749, and H762 in NSD2 PHD2 domain, corresponding to the Sotos mutations in NSD1 (Supplementary Fig. S3A). Mutations in PHD2 except those on L749 abolished H3K36 dimethylation without causing NSD2 dissociation from chromatin (Fig. 4C; Supplementary Fig. S3C). Reconstitution of NSD2 truncates and mutant in TKO cells confirmed that H762Y mutation in PHD2 domain abolished the induction of H3K36me2–H3K27me3 switch by NSD2, whereas loss of PWWP1, HMG, and PHD5-C5HCH had no effect (Fig. 4D). Therefore, our data suggested that PHD2 domain is important for NSD2-mediated histone methylation in cells.
We further investigated the effects of reader domains on the oncogenic potential of NSD2. H762Y mutation in PHD2 domain decreased the clonogenic growth conferred by NSD2 (Fig. 4E), consistent with its loss of H3K36 dimethylation activity. Vice versa, PWWP1 and HMG domains are dispensable for the clonogenic potential of NSD2 (Fig. 4E), as expected from the intact H3K36me2 levels. However, to our surprise, loss of PHD5-C5HCH, which had no impact on histone methylation, impaired the clonogenic growth (Fig. 4E), implicating that H3K36me2 mediated events are necessary but not sufficient for the biological functions of NSD2.
It is noteworthy that all reader-domain truncates and H762Y mutant were active in biochemical assay (Supplementary Fig. S3D and S3E). Indeed, the fragment 941 to 1,240 containing AWS/SET/post-SET and the KR-rich region between post-SET and PHD5 is enough to render enzymatic activity on nucleosome in vitro (35). Thus, the PHD2 domain is critical only for H3K36 methylation in cells, probably through recognition of histone marks and recruitment of NSD2 to its target loci.
Together, our findings suggest that multiple domains including the catalytic SET domain, PHD2, and PHD5-C5HCH motifs are involved in the signaling pathways mediated by NSD2.
NSD2 regulates histone methylation and transcription of t(4;14)+-specific myeloma gene set
To elucidate the mechanisms underlying the oncogenic potential of NSD2 in myeloma, we conducted microarray analysis on KMS11-PAR, TKO, and 8 reconstituted lines. Based on whole-genome expression profile, the 10 samples fell into 4 clusters: (i) PAR; (ii) TKO; (iii) WT, WT+MMSET I, 526-1240 and 526-1365; and (iv) CDM, CDM+MMSET I, MMSET I, and H762Y (Supplementary Fig. S4A). Cluster 3 was the “active” group and cluster 4 the “inactive” group, correlated with their H3K36 dimethylation status. It also confirmed that H762Y mutation in NSD2 PHD2 domain abolished its activity-dependent function. The reasons that PAR did not merge with active group and TKO did not overlap with inactive group are probably due to NSD2 activity-independent functions and clonal difference.
Comparison of gene expression profiles of PAR versus TKO and active versus inactive group identified significant overlap in differentially expressed genes (Supplementary Fig. S4B and Tables S1 and S2). In addition, the majority of the overlapping genes showed consistent fold changes between the two comparisons (Supplementary Table S1), implying that these transcription modulations likely result from NSD2 activity. The 82.66% of overlapping target genes (P < 0.05, fold change > 1.5) were high in both PAR and active group compared with TKO and inactive group (Supplementary Table S1), consistent with the global elevation of H3K36me2 and suppression of H3K27me3 induced by NSD2 overexpression. Gene set enrichment analysis determined statistically significant enrichment of a previously identified t(4;14)+ myeloma patient gene set (12) to be upregulated in both PAR and active group (Supplementary Fig. S4C), confirming the role of NSD2 activity in driving gene signature of this subtype of myeloma. Multiple other cancer and stem cell gene sets were also found to be associated with NSD2 activity (Supplementary Table S3).
To link the histone modifications induced by NSD2 with downstream transcription events, we conducted CHIP-seq analysis of PAR and TKO cells on H3K4me3, H3K27me3, H3K36me2, and H3K36me3, as well as MMSET II. Based on the microarray profiles of PAR and TKO cells, we categorized all genes into high, medium, and low expression groups, and examined average distribution of different histone marks for each group. H3K4me3 and H3K36me3 showed clear correlation with gene expression in both cell lines: the most highly expressed genes have the highest H3K4me3 signal around translational start site (TSS) and H3K36me3 level in the transcribed bodies; whereas the lowest expressed genes have little enrichment for both marks (Supplementary Fig. S5). Interestingly, when we compared the distribution of H3K36me2 in PAR with that in TKO, we found significant differences: in TKO cells, H3K36me2 level at gene body is positively associated with transcription levels; however, in PAR cells, despite the globally elevated H3K36me2 signal, it is not correlated with gene expression (Supplementary Figs. S5 and S6), consistent with previous findings (31). In contrast, the repressive mark H3K27me3 is enriched for the lowest expressed genes in PAR cells whereas this correlation is partially diminished in TKO (Supplementary Fig. S5), and MMSET II binding is associated with highest gene expression in PAR cells (Supplementary Fig. S5).
We further analyzed the distribution of different histone marks within genes that are modulated by NSD2 activity, especially those upregulated in PAR and active group, because they are more likely to be direct targets of H3K36me2 induced by NSD2, and are enriched for cancer-related genes. In these genes (BACE2, CA2, and IGF2BP2 as examples), H3K4me3 signal around TSS and H3K36me2 levels in the gene body are higher in PAR than TKO, whereas H3K27me3 around the promoter is lower (Fig. 5A and B). These findings indicate that in t(4;14)+ disease setting, even though the genome-wide correlation of H3K36me2 with transcription levels is abolished, H3K36me2 elevation on NSD2-activated genes is correlated with H3K4me3 promoter enrichment, H3K27me3 decrease, and gene expression. NSD2 dimethylation of H3K36 therefore plays a role in chromatin remodeling into an open and permissive state that promotes transcription activation of cancer-related genes. In contrast, in genes suppressed by NSD2 activity (e.g., CBLN2), H3K4me3 signal around TSS and H3K36me2 levels in the gene body are higher in TKO than PAR, whereas H3K27me3 is low in both cell lines (Fig. 5A and B).
PHD2 recruits NSD2 to oncogenic gene loci, which drives t(4;14)+ myelomagenesis
Because H762Y mutation abolished oncogenic gene expression mediated by NSD2 but not its in vitro enzymatic activity, we asked whether PHD2 domain is required for NSD2 recruitment to its target loci. MMSET II binding to NSD2-activated genes BACE2, CA2, and IGF2BP2 was decreased by H726Y mutation, accompanied with lower H3K36me2 signal within transcribed bodies and downregulated gene expression levels, whereas CDM blocked H3K36me2 elevation and gene transcription without significantly affecting MMSET II occupancy (Fig. 6A).
To test whether NSD2-activated genes are important for myeloma clonogenicity, we conducted knockdown experiments on BACE2, CA2, and IGF2BP2. Depletion of the 3 target genes inhibited colonization of KMS11-PAR cells, implicating that they are involved in NSD2-mediated signaling pathways (Fig. 6B)
Therefore, PHD2 domain recruits NSD2 to its target loci, the SET domain then induces H3K36me2 and turns on oncogenic genes, which subsequently drives t(4;14)+myelomagenesis.
Histone methylation and transcriptional regulation of oncogenic genes by NSD2
Our data showed that in t(4;14)+ myeloma disease setting, the histone product for NSD2 is H3K36me2. This finding is consistent with 2 previous reports (30, 31). However, we and Martinez-Garcia and colleagues both observed the reciprocal regulation of H3K36me2 and H3K27me3, whereas Kuo and colleagues did not. The mechanisms underlying how H3K36me2 antagonizes H3K27me3 are still obscure. However, 3 hypotheses have been proposed. First, H3K36me2/3 may interfere with the recruitment of PRC2 (Polycomb Repressive Complex 2) to nucleosome, as in vitro PRC2 activity is greatly reduced on H3K36me2 preinstalled nucleosomes (36). Second, PRC2 activity is negatively correlated with nucleosome spacing (37). H3K36me2/3 can lead to an open chromatin conformation, which is unfavorable for PRC2-mediated H3K27 tri-methylation. Finally, H3K36 dimethylation can reduce rates of methylation on H3K27 (38). In any case, the outcome of NSD2 overexpression in t(4;14)+ myeloma is the shift of chromatin homeostasis to a globally open and active state. Interestingly, in prostate cancer and potentially several other solid tumor types, Ezh2 was shown to be coordinately expressed and function upstream of NSD2 through a microRNA network, whereas t(4;14)+ myeloma seems to bypass this Ezh2-NSD2 axis (39), implicating a context-specific regulation.
H3K36 methylation has been implicated in diverse processes of gene expression including transcriptional activation and repression, alternative splicing, and dosage compensation, depending on when and where H3K36 is methylated, the degree of methylation, and which reader protein decodes the mark (40). Previous studies have indicated that NSD2 can mediate both transcriptional activation and repression (24, 31, 33, 41). Herein, we found that majority of target genes are activated by NSD2. Among them, there are gene sets from t(4;14)+ myeloma patients, other cancer types, and stem cells. Upregulation of these genes is correlated with higher H3K4me3 and lower H3K27me3 around TSS, as well as elevated H3K36me2 within gene body. H3K36me2 induced by NSD2 overexpression may block the H3K27 tri-methylation on these gene loci and lead to an active chromatin state. Thus, in t(4;14)+ disease setting, NSD2 triggers transcription activation of cancer-related genes by concertedly regulating histone methylation and chromatin remodeling. Whether NSD2-mediated H3K36 methylation plays other roles needs further investigation.
Opportunities for pharmacologic intervene of NSD2
Because NSD2 is a methyltransferase and its catalytic function seems to be essential for t(4;14)+ myelomagenesis, it would seem straightforward to target its SAM pocket, substrate pocket, or allosteric pockets, which can induce conformational changes and inhibit the enzymatic activity. In addition, NSD2 contains several PHD motifs that seem to be important for its biological functions. Therefore, another opportunity to target NSD2 lies in the reader domains. Our data suggested that PHD2 domain is important for NSD2 recruitment to oncogenic gene loci and H3K36me2 activity in cells. Both PHD2 and PHD5-C5HCH are essential for clonogenic potential of NSD2, even though the latter has no impact on H3K36 dimethylation. Studies on other H3K36 methyltransferases such as Mes-4, NSD1, and NSD3 showed that PHD domains are important for binding to chromatin, interaction with methyl-lysines, and recruitment to target genes (19, 21, 22, 42). The PHD5-6 domain of NSD1, corresponding to PHD5-C5HCH in NSD2, has also been implicated in interaction with adaptor protein Nizp1 (21, 43). Therefore, we hypothesize a model for the multistep mechanisms underlying the oncogenic potential of NSD2 (Fig. 7). The PHD2 domain facilitates the recruitment of NSD2 to its target loci; the SET domain dimethylates H3K36 and turns on oncogenic genes; and the PHD5-C5HCH motif interacts with a currently unknown binding partner to initiate the signaling pathways involved in myelomagenesis.
However, gaps remain in our knowledge to design compounds that interfere with the interaction of PHD domains with chromatin. It is necessary to find out what substrates PHD domains bind to. It has been reported that single and tandem PHD fingers interact with different substrates (44, 45). Given that NSD2 contains multiple PHD domains, their substrates may be context dependent. To further increase the complexity, minor sequence variation despite the similar structural fold of PHD domains can result in different binding preferences (19). Because PHD family is one of the largest and most diverse of the methyl-lysine readers (46), whether the compounds can selectively target NSD2 over other PHD-containing proteins remains to be seen. Only with answers to these questions can we fully evaluate the compounds that target PHD domains for therapeutic applications.
Challenges of targeting epigenetic modulators for therapy
There are several challenges to target epigenetic players. Epigenetic modulators often form complexes with partners depending on the specific context, and they can conduct different tasks when partnering with different cofactors. Therefore, it is essential to elucidate whether the biological functions are mediated through enzymatic activity or protein–protein interaction. We have showed that NSD2 activity is critical for tumorigenesis of t(4;14)+ myeloma cells. However, we also found that MMSET I, the isoform without catalytic domain, plays a role by synergizing with MMSET II. There are activity-independent events mediated by PHD5-C5HCH domain that are important for clonogenicity of myeloma cells, likely through posttranscriptional regulation (data not shown).
Also, it is important to find out what are the disease-relevant downstream targets of epigenetic modulators. Methyltransferases can methylate not only histones but also nonhistone substrates, such as p53 and NF-κB (47–49), which play important roles in cancer and inflammation. Although we conclude that H3K36me2 is the histone product for NSD2 in t(4;14)+ myeloma, we cannot exclude the possibility that it may methylate nonhistone substrates as well. The reported C-terminal isoform REIIBP is localized in the cytosol fraction, suggesting that it may have cytoplasmic substrates. In addition, when we conducted in vitro methylation reaction on ProtoArray, we identified some RNA-binding proteins to be methylated by MMSET II (data not shown). Whether these nonhistone substrates are implicated in t(4;14)+ myelomagenesis needs further studies.
Moreover, because epigenetic alterations usually mediate transcription modulation of a large number of genes, it is difficult to elucidate which one is the key node. We have identified t(4;14)+-specific myeloma gene set as downstream target of NSD2. We also pin-pointed the role of NSD2 to regulation of myeloma–BMSC interaction, and found enrichment of cell adhesion pathways in NSD2-activated genes (data not shown). But we have not identified the exact adhesion molecule(s) nor cytokine(s)/chemokines(s) mediating the colonization process (data not shown). It is reported that NSD2 can promote myeloma proliferation by modulating c-MYC through microRNA (50). Whether microRNA network is involved in NSD2-mediated myelomagenesis requires further investigation. Also, because epigenetic alterations often induce global chromatin remodeling and reprogramming, and are involved in many physiologic processes, the liability to target such aberrations is unknown. More knowledge is needed to facilitate the translation of epigenetic drug discovery into clinical practice.
In summary, we have showed the dependency of t(4;14)+ myelomagenesis on NSD2 and identified domains that can potentially be targeted for pharmacologic intervene. NSD2 inhibitors will open a new avenue for intervention of t(4;14)+ myeloma and other cancer types in which NSD2 is dysregulated.
Disclosure of Potential Conflicts of Interest
J. Liu is employed (other than primary affiliation; e.g., consulting) as a group leader in Genomics Institute of the Novartis Research Foundation (GNF). H. Chan is employed (other than primary affiliation; e.g., consulting) as a research investigator in Novartis. No potential conflicts of interest were disclosed by the other authors.
Conception and design: Z. Huang, H. Wu, H. Zhang, Y. Wang, M. Zhang, T. Yang, J. Wu, J. Liu, X. Wu, H. Chan, C. Lu, P. Atadja, E. Li, Y. Wang, M. Hu
Development of methodology: Z. Huang, H. Wu, H. Zhang, M. Fang, Y. Wang, T. Yang, J. Dai, W. Yi, S. Zhou, Q. Li, J. Wu, P. Atadja, Y. Wang, M. Hu
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Z. Huang, H. Wu, F. Xu, F. Yan, N. Englund, Z. Wang, H. Zhang, Y. Wang, T. Yang, K. Zhao, Y. Yu, J. Dai, W. Yi, S. Zhou, Q. Li, M. Hu
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Z. Huang, H. Wu, S. Chuai, F. Xu, N. Englund, Z. Wang, H. Zhang, Y. Wang, K. Zhao, Y. Yu, J. Liu, X. Wu, C. Lu, P. Atadja, Y. Wang, M. Hu
Writing, review, and/or revision of the manuscript: S. Chuai, H. Zhang, T. Yang, Y. Yu, H. Chan, P. Atadja, Y. Wang, M. Hu
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Z. Huang, F. Xu, H. Zhang, T. Yang, Q. Li, J. Liu, Y. Wang, M. Hu
Study supervision: Y. Wang, J. Gu, T. Yang, X. Wu, H. Chan, P. Atadja, M. Hu
The authors thank Drs. J. Shou and Z. Chen, Novartis Institutes for BioMedical Research (China), for critical reading of the manuscript.
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.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
- Received April 5, 2013.
- Revision received July 30, 2013.
- Accepted August 7, 2013.
- ©2013 American Association for Cancer Research.