SMYD3-Mediated H2A.Z.1 Methylation Promotes Cell Cycle and Cancer Proliferation

Chromatin changes, such as gene promoter methylation and post-translational modifications (PTM) of histones, have been linked to cancers (1–3). As more than 50 human histone methyltransferases (HMT) have been identified (4), understanding the functions of histone methylation has become a crucial issue. Studies have found that overexpression of HMTs that are responsible for the transcription-repressive marks, such as H3K9 and H3K27 methylation, leads to the aberrant silencing of tumor suppressor genes. For example, EZH2, a H3K27 HMT, and G9a, a H3K9 HMT, are upregulated in various cancers (5–7). In addition to repression-associated histone marks, MLL, a HMT for H3K4, modulates expression of various homeotic (Hox) genes and plays a key role in leukemic progression (8). Dysregulation of HMTs such as copy number amplification or aberrant gene expression leads to imbalance in histone methylation pathways and thus the initiation/progression of cancer (9).

SMYD3 is a HMT that is overexpressed in breast, colorectal, prostate, and hepatocellular tumors (10–13). Overexpression of SMYD3 enhances cell growth and cancer invasion, while reduced SMYD3 expression results in growth suppression (11, 14). SMYD3 can methylate H3K4, H4K5, and H4K20 to promote oncogenes or repress tumor suppressor genes (11, 15–17). For example, SMYD3 promotes telomerase expression, which correlates with an increase in H3K4 methylation at the telomerase promoter (18). SMYD3 also methylates a nonhistone protein, MAP3K2, to facilitate Ras/Raf/MEK/ERK signaling (19). However, how SMYD3 promotes cancer cell proliferation is not clear.

H2A.Z is an oncogenic histone variant that is overexpressed in breast, colorectal, liver, and bladder cancers (20–23). Two isoforms of H2A.Z, H2A.Z.1 and H2A.Z.2, have been identified as products of two nonallelic genes (24, 25), and each can execute isoform-specific function. H2A.Z.1 is pivotal for hepatocarcinogenesis by selectively modulating cell cycle and EMT-regulatory proteins (22), whereas H2A.Z.2 is a mediator of cell proliferation and drug sensitivity in malignant melanoma (26). Moreover, the oncogenic function of H2A.Z may be mediated through PTMs on H2A.Z and/or different combinations of H2A.Z isoform with other epigenetic regulations (21, 27–32).

Here we report the identification of histone H2A.Z.1 as a new substrate of SMYD3. We found that SMYD3 methylated H2A.Z.1 at lysine 101 (H2A.Z.1K101), which is in the globular domain of H2A.Z.1. This modification increased H2A.Z.1 protein stability by interfering with its removal by ANP32E and strengthening its binding to histone H3. The microarray analysis indicated that cyclin A1 was an important target of SMYD3 and H2A.Z.1K101me2. We further demonstrated how H2A.Z.1K101me2 accelerates G₁–S transition and leads to breast cancer proliferation.

The breast cancer cell lines MCF7 and T47D were purchased from ATCC in 2012 and maintained in their respective media according to ATCC protocol. Cell line authentication was done by Food and Industry Research & Development Institute (Taiwan) using short random repeat DNA sequencing in 2016. Plasmid construction, virus preparation, and generation of transduced cell lines are described in the Supplementary Materials and Methods. The specific oligo sequences of shRNA are listed in Supplementary Table S2.

The in vitro HMTase assay was performed as described with some modifications (11). Briefly, GST-tagged human SMYD3 and mutant SMYD3^Y239F proteins were purified from E. coli, and an in vitro HMT assay was performed with 10 μg of calf thymus histones (Sigma), recombinant histones, or purified nucleosomes isolated from shSMYD3 MCF7 cells (33) as substrates. To determine the methylated site of human H2A.Z.1, recombinant His-tagged H2A.Z.1 and GST-tagged SMYD3 were coincubated with cold S-adenosyl-L-methionine (SAM, Sigma). The LC/MS-MS was employed to identify the methylated sites of histones.

Calorimetric titrations of GST-ANP32E with H2A.ZK101 and H2A.ZK101me2 peptides were performed on an iTC200 microcalorimeter (GE) at 20°C and 30°C for determination of binding enthalpy and affinity. Protein samples were extensively dialyzed against the isothermal titration calorimetry (ITC) buffer containing 25 mmol/L KPi (pH 8.0), 250 mmol/L KCl, and 1 mmol/L Tris(2-carboxyethyl)phosphine (TCEP). The 200-μL sample cell was filled with a 100 μmol/L solution of protein and the 40-μL injection syringe, with 1.6 mmol/L of the titrating peptides. The titration data were analyzed using the program Origin 7.0 and nonconstrained fitting was performed for the interaction of GST-ANP32E with peptides. For more details, see the Supplementary Materials and Methods.

Eviction assay was performed using immunoprecipitated Flag-H2A.Z–containing nucleosomes that were incubated with increasing amount of recombinant GST-ANP32E. Bound and unbound materials were blotted with either anti-H3 or anti-Flag antibodies. For more details, see the Supplementary Materials and Methods.

For each condition, 1 × 10⁶ MCF7 cells were harvested in 100-μL PBS and injected subcutaneously on both sides of the back in 4-week-old female nude mice (n = 6). Tumor size was measured weekly and calculated with the formula length × width²/2. Tumor incidence was defined when the tumor mass reached 100 mm³, and growth was monitored until it reached 500 mm³. After 10 weeks, the mice were sacrificed and the xenografts were excised and weighed.

All the experiments involving mice were performed in accordance with the protocols approved by the Institutional Animal Care and Use Committee (IACUC) of National Taiwan University College of Medicine and College of Public Health (IACUC no. 20120046). Detailed methods are provided in Supplementary Materials and Methods.

SMYD3 has been reported to methylate several histones including H3K4, H4K5, and H4K20 (11, 15, 16). To identify novel substrates of SMYD3, we used recombinant SMYD3 and catalytic dead mutant SMYD3^Y239F proteins (34) to perform an in vitro HMT assay on calf thymus histone extracts. The autoradiography analysis demonstrated that multiple histone proteins were detected as the potential substrates of SMYD3 (Fig. 1A). This result was consistent with our LC/MS-MS data, in which several well-known substrates of SMYD3 were identified (Supplementary Table S1). Interestingly, bovine H2A.Z (the human H2A.Z.1 homolog) methylation was identified in the mass data. To examine the specificity, we first confirmed that SMYD3^WT, but not the SMYD3^Y239F, also methylated recombinant human H2A.Z.1 (Fig. 1B). In addition, compared with H2A.Z.1, H4 was a better substrate, but H3 was not so preferable for SMYD3 (Fig. 1B). LC/MS-MS assay identified that lysine 101 of human H2A.Z.1 was monomethylated (H2A.Z.1K101me) and dimethylated (H2A.Z.1K101me2) by SMYD3 (Supplementary Fig. S1). However, the proportion of methylated protein in vivo was difficult to estimate. An in vitro HMT assay confirmed that H2A.Z.1^K101Q and H2A.Z.1^K101R mutants, in which the lysine residue was replaced with glutamine and arginine, respectively, were unable to be methylated by SMYD3 (Fig. 1C). Next, a lentiviral shRNA infection system was employed to knockdown SMYD3 (shSMYD3) in cells. When nucleosomes isolated from shSMYD3 MCF7 cells were used as substrates, a signal between H3 and H4 was detected (Supplementary Fig. S2A). We further generated an H2A.Z methylation–specific antibody. This antibody recognized only H2A.ZK101me2 peptide but not H2A.Z peptide without methylation (Supplementary Fig. S2B). The H2A.ZK101me2 antibody detected three signals from total lysates of MCF7 cells. Only the 14-kDa signal, the predicted molecular weight of H2A.Z, disappeared upon preincubation of H2A.ZK101me2 antibody with H2A.ZK101me2 peptide (Supplementary Fig. S2C). Furthermore, the antibody could detect exogenously expressed H2A.Z^WT but not H2A.Z^K101Q and H2A.Z^K101R (Supplementary Fig. S2D), suggesting that the detected signal was the H2A.ZK101me2 protein. When nucleosomes isolated from shSMYD3 MCF7 cells were used as substrates, SMYD3 increased the methylation level of K101 (Fig. 1D), suggesting that SMYD3 could methylate H2A.Z in the context of nucleosome. Interestingly, SMYD3 also increased H3K4 and H4K20 methylation but not that of H4K5 (Supplementary Fig. S2E). In addition, when examining H2A.ZK101me2 protein levels in shSMYD3 T47D and MCF7 breast cancer cell lines, the H2A.ZK101me2 level was reduced in SMYD3 knockdown cells, while the protein level of H2A.Z remained similar (Fig. 1E and Supplementary Fig. S2F). Consistently, overexpression of SMYD3^WT elevated the level of H2A.ZK101me2 by 60% in 293T cells (Fig. 1F). Together, these findings indicate that H2A.Z is very likely an in vivo substrate of SMYD3.

Previous studies demonstrated that deletion of the acidic patch, which contains K101, resulted in more dynamic association with chromatin and lower H2A.Z occupancy at target genes (35–37). To examine whether dimethylation of H2A.Z influences its association with chromatin, nuclei isolated from shLuc control or shSMYD3 cells were subjected to salt titration (Fig. 2A). Western blotting demonstrated that while H2A.Z in shLuc cells was stably associated with chromatin in up to 0.6 mol/L NaCl and partially depleted at 0.8 mol/L NaCl (dissociation of H2A.Z/H2B dimer usually occurs between 0.6 and 0.8 mol/L NaCl; refs. 38, 39), a fraction of H2A.Z in shSMYD3 cells dissociated at lower salt concentrations (0.4–0.6 mol/L NaCl; Fig. 2B). Consistently, a similar feature was observed in H2A.Z-mutant cells that H2A.Z^K101Q and H2A.Z^K101R were less tightly associated with chromatin than H2A.Z^WT (Fig. 2C and D). Finally, whether dimethylation of H2A.Z exerted any effect on chromatin structure remained unclear. SMYD3 knockdown cells or H2A.Z^WT, H2A.Z^K101Q, and H2A.Z^K101R-expressing MCF7 cells were subjected to micrococcal nuclease (MNase) digestion. No significant difference in the cleavage pattern was observed, implying that the absence of SMYD3 or the methylation status of H2A.Z does not affect global chromatin compaction (Fig. 2E and F).

The chromatin remodeler such as SRCAP (Snf2-related CREBBP activator protein) and TIP60–p400 complex mediates site-specific incorporation of H2A.Z in mammalian cells (27, 40, 41). Moreover, the H2A.Z-deposition chaperone, YL1 (30), and the H2A.Z-eviction chaperone, ANP32E (29, 42), are important for H2A.Z nucleosome association. In view of this, we tested the ability of H2A.Z^K101Q and H2A.Z^K101R to interact with these complexes and chaperones. A coimmunoprecipitation assay was performed with an anti-Flag antibody against Flag-tagged H2A.Z^WT, H2A.Z^K101Q, and H2A.Z^K101R proteins. Similar amounts of the two SRCAP subunits (SRCAP and RUVBL1), two TIP60-p400 subunits (TIP60 and p400), and YL1 were coprecipitated with H2A.Z^WT and H2A.Z^K101 mutants (Fig. 3A and Supplementary Fig. S3A), suggesting that the abilities of SRCAP-, TIP60-p400-, and YL1-mediated deposition of H2A.Z proteins were not altered in mutant cells. Conversely, H2A.Z^K101 mutants displayed a higher affinity to ANP32E compared with H2A.Z^WT (Fig. 3A). An in vitro GST-ANP32E pull-down experiment using biotinylated H2A.Z peptides containing amino acids 93–108 revealed that ANP32E bound directly to H2A.Z, with a preference for K101 to K101me2 (Fig. 3B).

To evaluate the interactions between ANP32E and peptides in quantitative terms, we measured the dissociation constants by ITC. ITC assay showed that H2A.ZK101 peptides bound to ANP32E with a K_d of 27.2 μmol/L (Fig. 3C, left), whereas H2A.ZK101me2 peptides displayed no detectable binding for GST-ANP32E (Fig. 3C, right and Table 1). These findings further confirmed that dimethylated H2A.Z inhibits the interaction between H2A.Z and ANP32E. Previous studies have shown that ANP32E was able to specifically remove H2A.Z from the nucleosome (29, 42). To test whether H2A.ZK101me2 could reduce histone displacement from chromatin, we performed a histone eviction assay (Fig. 3D). Flag-tagged H2A.Z^WT and H2A.Z^K101-mutant–containing nucleosomes were immunoprecipitated and then incubated with increasing amounts of recombinant GST-ANP32E or GST alone as a negative control. The bound histones and unbound material were subjected to Western blotting with anti-H3 or anti-Flag antibodies. Disassembled nucleosomes resulted in release of histones H3 and H4 while H2A.Z-H2B dimers remained bound to the beads. As expected, less histone H3 was detected in the evicted fraction from H2A.Z^WT than that from the H2A.Z^K101-mutant–containing nucleosomes (Fig. 3E), suggesting that the H2A.Z dimethylation decreases histone displacement/eviction on chromatin.

We further verified the interaction between H2A.Z and other histones, including histone H2B, which forms a dimer with H2A.Z, and histone H3, which directly interacts with H2A.Z via hydrogen bonds to stabilize nucleosomes (43). Interestingly, while the affinity between H2A.Z^K101 mutants and H2B remained the same (Fig. 3A), H2A.Z^K101 mutants exhibited decreased association with H3 both in vitro and in vivo (Fig. 3A and Supplementary Fig. S3B). Similarly, using K101- or K101me2-biotinylated peptides to pull down endogenous interacting proteins, we confirmed that the interaction between ANP32E and the K101 peptide was greatly increased compared with that between ANP32E and the K101me2 peptide, whereas histone H3 exhibited stronger affinity with K101me2 than with K101 (Supplementary Fig. S3C). In SMYD3 knockdown cells, the interaction between ANP32E and H2A.Z was increased. Conversely, the interaction between H3 and H2A.Z was decreased (Supplementary Fig. S3D). These data imply that the dimethylation of H2A.ZK101 may serve as a signal to prevent the binding of H2A.Z to its chaperone that removes H2A.Z from chromatin, thus stabilizing the interaction between H2A.Z and H3.

SMYD3 (10–13) and H2A.Z (19–22) are overexpressed in several tumors. Consistently, we observed a positive correlation of expression between SMYD3 and H2A.ZK101me2 in a series of cancer cell lines and human breast tissue samples (Supplementary Fig. S4A and S4B). To further investigate this correlation, we first knocked down endogenous H2A.Z.1 and the growth of H2A.Z.1 knockdown MCF7 cells was compromised. H2A.Z.1^WT expression was able to restore delayed cell growth, whereas H2A.Z.1^K101Q or H2A.Z.1^K101R failed to do that (Fig. 4A and Supplementary Fig. S4C). Consistent with MCF7 cells, H2A.Z^WT T47D cells grew much faster than cells expressing empty vectors or H2A.Z^K101Q (Fig. 4B). The effect of H2A.Z on cell proliferation was more prominent in T47D cells. Therefore, we further examined cell-cycle progression of T47D cells. H2A.Z^WT cells entered the S-phase rapidly at 8 hours and proceeded through the cell cycle, but H2A.Z^K101Q failed to accelerate cell-cycle progression (Fig. 4B). These results suggested that the methylation of H2A.Z at K101 is critical for S-phase entry and cell proliferation. To test whether K101 mutation affected H2A.Z-mediated tumorigenic activity, colony formation was analyzed in H2A.Z^WT, H2A.Z^K101Q, and H2A.Z^K101R-expressing MCF7 cells when endogenous H2A.Z expression was suppressed. Compared with H2A.Z^WT cells, H2A.Z^K101Q and H2A.Z^K101R cells could neither increase the number of colonies nor rescue the colony formation defect caused by H2A.Z knockdown (Fig. 4C). When H2A.Z^K101Q and H2A.Z^K101R cells were subcutaneously injected into nude mice, the tumor volumes were significantly smaller than those formed by H2A.Z^WT cells (Fig. 4D, right), even though H2A.Z^WT, H2A.Z^K101Q, and H2A.Z^K101R proteins were constitutively expressed at equal amounts in isolated tumors after 10 weeks (Fig. 4D, left). The soft agar assay revealed that knockdown of SMYD3 decreased the tumorigenic activity of H2A.Z^WT (Supplementary Fig. S4D). In addition, knockdown of SMYD3 also lowered the proliferation advantage of H2A.Z^WT cells (Fig. 4E). Furthermore, SMYD3 overexpression promoted the proliferation of MCF7 cells (Supplementary Fig. S4E). Compared with that of H2A.Z^WT cells, SMYD3 overexpression could not promote the proliferation of H2A.Z^K101Q cells in the endogenous H2A.Z knockdown MCF7 background (Fig. 4F, lane 2 vs. lane 4 and lane 6 vs. lane 8). Overexpression of both H2A.Z^WT and SMYD3 did not display cooperative effect on cell proliferation (Supplementary Fig. S4F, lane 4 vs. lane 2). These results suggest that hindered methylation on H2A.ZK101 disrupts the oncogenic potential of H2A.Z.

To identify the target genes of SMYD3-mediated H2A.ZK101me2, whole-genome microarray analysis of RNA isolated from two parallel experiments, including shLuc versus shSMYD3 cells (GEO accession number GSE58048) and H2A.Z^WT versus H2A.Z^K101Q in endogenous H2A.Z knockdown MCF7 cells (GEO accession number GSE58047), were conducted. We found that the expressions of 996 transcripts were altered upon SMYD3 knockdown (Supplementary Fig. S5A). The gene ontology (GO) analysis indicated that these 996 SMYD3-regulated transcripts were involved in cell cycle, organelle fission, and cell division (Supplementary Fig. S5B), which was consistent with previous reports (11, 44). In addition, 245 transcripts were altered by H2A.Z when K101 was mutated to Q (Supplementary Fig. S5A). In agreement with the multifunctional roles of H2A.Z (45), the GO analysis showed that the majority of H2A.ZK101-regulated genes were involved in nucleotide metabolism, regulation of blood vessel size, defense-related response, and vasodilation (Supplementary Fig. S5C). Most importantly, only 35 transcripts, which belong to 19 genes, were coregulated by SMYD3 and H2A.ZK101 (Supplementary Fig. S5A; Supplementary Table S2). Among these genes, 8 were downregulated and 11 were upregulated. To verify the accuracy of the microarray data, qRT-PCR was performed for these 19 genes. Most genes exhibited similar fold differences in mRNA expression as originally identified in the microarray analysis (Supplementary Fig. S5D and S5E).

As H2A.Z is known to promote cell growth (46), cyclin A1 (CCNA1), the only cell-cycle–controlled gene among the 19 genes coregulated by SMYD3 and H2A.ZK101me2, was further analyzed. Consistently, knockdown of SMYD3 (Fig. 5A) or H2A.Z (Fig. 5B) decreased the expression of cyclin A1. Moreover, expression of H2A.Z^WT or SMYD3^WT activated cyclin A1 protein level (Fig. 5C and Supplementary Fig. S4E, respectively). In contrast, the mRNA level of cyclin A2, another member in the cyclin A family, remained unchanged in H2A.Z knockdown cells (Supplementary Fig. S6A). Chromatin immunoprecipitation (ChIP) assays further indicated that the endogenous SMYD3 and dimethylated H2A.Z associated specifically with the cyclin A1 promoter from -623 to -532 bp (region b) relative to the transcription start site (TSS; Fig. 5D). The net amounts of dimethylated H2A.Z and H2A.Z were enriched at region b compared with those at other regions (Fig. 5D). Moreover, a re-ChIP assay confirmed the colocalization of H2A.Z and H2A.ZK101me2 at the same region (Supplementary Fig. S6B), demonstrating the specificity of the anti-H2A.ZK101me2 antibody. Furthermore, knockdown of SMYD3 also decreased SMYD3, H2A.ZK101me2, and H2A.Z binding to region b of the cyclin A1 promoter, but not to the control region (Fig. 5E, region a). SMYD3 knockdown did not affect the localization of histones H2A, H2B, H3, and H4, except that H2B had an increased level at the control region for a currently unknown reason (Supplementary Fig. S6C and S6D). To test whether H2A.Z^WT associates more closely with chromatin than mutant H2A.Z, ChIP assays using anti-Flag antibody were performed. The presence of H2A.Z^WT was 2.72- and 3.28-fold higher than that of H2A.Z^K101Q and H2A.Z^K101R, respectively, at the target region of the cyclin A1 promoter (Fig. 5F, region b), but not at the control region (Fig. 5F, region a). In addition, the increased level of H2A.Z in H2A.Z^WT cells compared with that in H2A.Z-mutant cells further supported the role of H2A.Z dimethylation for its chromatin association (Supplementary Fig. S6E). In contrast, other core histone subunits were equally distributed at the cyclin A1 promoter (Supplementary Fig. S6F and S6G). Taken together, these data indicate that SMYD3 may activate cyclin A1 gene expression by binding directly to the cyclin A1 promoter and dimethylating the local H2A.Z at K101.

To investigate the role of cyclin A1 in SMYD3- and H2A.ZK101me2-promoted cell proliferation, cyclin A1 was knocked down in empty vector, H2A.Z^WT, or H2A.Z^K101Q cells, followed by a cell count. The results showed that knocking down cyclin A1 reduced H2A.Z^WT-promoted cell proliferation (Supplementary Fig. S6G). Conversely, exogenous expression of cyclin A1, but not the control with expression of red fluorescent protein (RFP), restored cell proliferation in H2A.Z^K101Q cells (Fig. 5G, lanes 5 and 6). In endogenous H2A.Z knockdown cells, the overexpression of cyclin A1 could rescue the compromised cell growth in vector control and H2A.Z^K101Q cells (Fig. 5G, lanes 8 and 12). Finally, the complementation of cyclin A1 in H2A.Z^K101Q cells significantly increased tumor growth in a mouse xenograft model (Fig. 5H). Taken together, these findings suggest that SMYD3-mediated dimethylation of H2A.Z at K101 can directly activate cyclin A1 expression and contribute to tumorigenesis.

SMYD3 was originally identified as a methyltransferase that methylates histone H3K4 (11). Here, we identified a new substrate, histone H2A.Z.1, dimethylated at K101 by SMYD3 both in vitro and in vivo, but not by SMYD5 (Supplementary Fig. S6H). Although K101 is conserved in H2A and H2A.Z but not in H2A.X, H2A.Bbd and mH2A1.2, sequence diversity flanking K101 may alter the substrate specificity of SMYD3 (Supplementary Fig. S6I). In agreement with this, replacement of H2A amino acids 97-100 to H2A.Z decreased the ANP32E-H2A.Z association (28, 29). In addition, H2A.Z.1 can be methylated when it is in the free form (Fig. 1A–1C) or in the context of nucleosome (Fig. 1D). More importantly, H2A.Z.1 expression–stimulated cancer proliferation was prevented in H2A.Z.1^K101 mutant–expressing cells, suggesting that K101 dimethylation provides a major regulation for H2A.Z.1-driven cell proliferation (Fig. 6).

Our results showed that gene expression of several cell-cycle regulators, including CCNA1, CCNA2, CCNB1, and CCNE2 were decreased in SMYD3 knockdown MCF7 cells. Another study observed that mRNA levels of CcnA2, CcnD1, and CcnE1 were decreased in Smyd3-KO mice. Therefore, cyclin A2 seems to be regulated by SMYD3 in both human and mouse probably though H3K4me3 (13) but not H2A.ZK101me2 (Supplementary Fig. S6A). SMYD3 has been proposed to bind to specific DNA sequence, estrogen response elements, and conserved noncoding DNA sequence (11, 17, 47). Consistently, 11 of 19 qRT-PCR–confirmed genes contain these sequences/elements in their promoter regions (Supplementary Table S2). Even though SMYD3 usually performs as a transcriptional activator, 452 transcripts were upregulated upon SMYD3 knockdown. It is possible that SMYD3 may drive some other factors to suppress these genes indirectly, or SMYD3-mediated H4K20 trimethylation may invoke gene repression (16). Further analysis by ChIP-seq may clarify whether SMYD3 and H2A.ZK101me2 transcriptionally regulate these genes through localizing to their promoters.

Our data demonstrated that H2A.Z^K101 mutants and the nonmethylated H2A.Z peptide exhibited stronger association with ANP32E than H2A.Z^WT and the H2A.ZK101me2 peptide, respectively (Fig. 3). Interestingly, the net amounts of promoter-associated H2A.Z increased by 20% in ANP32E-knockout cells (29), which is consistent with our finding that the increased association between H2A.Z^K101 mutants and ANP32E decreases the presence of H2A.Z^K101Q and H2A.Z^K101R at the cyclin A1 promoter (Fig. 5F). Furthermore, stronger affinity between H2A.Z^WT and histone H3 than that between H2A.Z^K101 mutants and histone H3 supports the hypothesis that K101 dimethylation increases the stability of H2A.Z protein (Fig. 3). We propose that the methylation of H2A.Z may possess the ability to associate properly with chromatin. H2A.Z mutants that cannot be methylated may easily dissociate from chromatin. These data suggest that H2A.ZK101 dimethylation may hinder its interaction with ANP32E, reduce the ANP32E-facilitated removal of H2A.Z, and allow H2A.Z to remain on the chromatin to execute their function in cyclin A1 activation (Fig. 6).

Cyclin A1 promotes G₁–S cell-cycle progression in somatic cells and overexpressed cyclin A1 enhances S-phase entry to provide an oncogenic effect in leukemia cells (48, 49). A previous study attributed cyclin A1–mediated cancer progression to dimethylation of H3K36 in the coding region of cyclin A1 gene, which increased cyclin A1 transcription (50). Here we identified H2A.ZK101me2 as another important epigenetic factor that positively regulates cyclin A1 expression. The loss of H2A.ZK101me2 lowered H2A.Z protein association with chromatin (Fig. 2). In addition, the net amounts of H2A.ZK101me2 and H2A.Z, but not those of H2A, H2B, H3, and H4, were decreased at the cyclin A1 promoter in SMYD3 knockdown or in H2A.ZK101-mutant cells (Fig. 5 and Supplementary Fig. S6). Consistently, overexpression of H2A.Z^WT proteins promoted cyclin A1 gene expression (Fig. 5C). Therefore, we suggest that the enrichment of H2A.ZK101me2 at the cyclin A1 promoter is an active mark for gene expression.

No potential conflicts of interest were disclosed.

Conception and design: C.-H. Tsai, Y.-J. Chen, M.-C. Lin, K.-J. Wu, S.-C. Teng

Development of methodology: C.-H. Tsai, Y.-J. Chen, S.-C. Teng

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C.-H. Tsai, Y.-J. Chen, S.-R. Tzeng, W.-H. Kuo, M.-C. Lin

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C.-H. Tsai, Y.-J. Chen, C.-J. Yu, S.-R. Tzeng, M.-C. Lin, N.-L. Chan, K.-J. Wu, S.-C. Teng

Writing, review, and/or revision of the manuscript: C.-H. Tsai, Y.-J. Chen, M.-C. Lin, K.-J. Wu, S.-C. Teng

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y.-J. Chen

Study supervision: Y.-J. Chen, S.-C. Teng

Other (protein purification): I-C. Wu

We thank Drs. Luc Gaudreau, Cheng-Fu Kao, Ming-Shyue Lee, Tsai-Kun Li, Hsueh-Fen Juan, and Ali Hamiche for providing materials. We also thank Dr. Li-Jung Juan for her critical comments on the manuscript.

This work was supported by a grant from National Health Research Institute of Taiwan (NHRI-EX101-9727BI to S.-C. Teng).

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.