DNMT3A reads and connects histone H3K36me2 to DNA methylation
DNA methylation at the 5-position of cytosine (5mC) is a crucial epigenetic mark in regulating biological processes including gene silencing, gene imprinting, and X chromosome inactivation (Jaenisch and Bird, 2003; Smith and Meissner, 2013). Human genome encodes three DNA methyltransferases, DNMT1, DNMT3A and DNMT3B to catalyze 5mC. Although not tightly restricted, DNMT1 is thought to maintain the established pattern of 5mC throughout DNA replication, while DNMT3A and DNMT3B are largely responsible for the de novo establishment of 5mC. It has long been questioned how de novo DNA 5mC patterns are established in different genomic regions and whether histone modifications crosstalk to the process. Until recently, it was reported that through recognition of histone H3K36me3 mark, DNMT3B plays a dominant role in mediating DNA 5mC in the genic region undergoing active transcription (Baubec et al., 2015; Neri et al., 2017). However, 5mC occurs at both intergenic and genic regions, while H3K36me3 is largely absent in the intergenic regions, indicating that the intergenic 5mC may be mediated through different mechanisms.
Among the three DNA methyltransferases, both DNMT3A and DNMT3B possess a PWWP domain. As PWWP domain was reported to be capable in recognizing H3K36me2 or H3K36me3 (Sankaran et al., 2016), we hypothesized that the PWWP domain of DNMT3A may be able to read H3K36me2 and play a recruitment function for DNMT3A. Consistent with this idea, in vitro pull-down assay readily identified a specific interaction between H3K36me2 and the PWWP of DNMT3A (Fig. 1C). Interestingly, we found that the interaction was substantially stronger than that of H3K36me3 (Fig. 1C), and such preference was not affected by the sequence variation at the 31 position of the histone variants H3.1/2 and H3.3, despite the interactions with H3.3 peptides were generally weaker (Fig. 1C). Consistently, recombinant full length DNMT3A2 (isoform 2) purified from E.coli also preferentially binds H3K36me2 compared to H3K36me3 (Fig. 1C).
Various crosstalk mechanisms between histone modifications and 5mC have been documented, and it was known that the histone H3K4 methylations, especially H3K4me2/3, could strongly repress DNMT3A activity through ADD domain providing a mechanism to protect active promoters and enhancers from DNA 5mC methylation (Zhang et al., 2010; Li et al., 2011; Guo et al., 2015). We therefore wondered whether H3K36me2 binding by the PWWP domain could also affect the enzymatic activity of DNMT3A. To address this, we carried out in vitro methyltransferase assay using recombinant full length DNMT3A2 protein purified from insect cells and DNA substrate by spiking in H3K36me2 modified as well as several control histone peptides (Fig. 1D). In support of our hypothesis, we found that while H3K36me1 and H3K36me3 peptides only had minimal effects on DNMT3A activity, H3K36me2 peptides could significantly activate DNMT3A at a concentration as low as 3.2 μmol/L, using two independent, radioactivity and MBD domain-based, methylation detection methods (Fig. 1D).
Through MeDIP-seq and H3K36me2 ChIP-seq approaches in KMS11 and KMS11TKO cells, we again observed strong correlations between H3K36me2 and DNA methylation genome-wide (Fig. 2C). Interestingly, the correlation is significantly higher in KMS11TKO compared to KMS11 (Fig. 2C), indicating the abnormal hyper H3K36me2 significantly impaired proper 5mC establishment in KMS11. Close examination of mapped H3K36me2 and 5mC genome tracks found that the both marks are more dispersedly distributed in KMS11 genome due to genome-wide hyper H3K36me2 (Fig. 2D). While in KMS11TKO cells, we found that most intergenic H3K36me2 signals were lost as the result of loss of NSD2 overexpression, and the remaining H3K36me2 signals were generally retained in the genic and surrounding regions, exemplified by ARHGAP44 and COX10 in Fig. 2D. Consistent with our hypothesis, we also observed significant reduction of the intergenic 5mC in KMS11TKO (Fig. 2D, shadowed regions). Notably, genic 5mC signals were largely retained and in certain cases, such as COX10, even increased in KMS11TKO (Fig. 2D), likely due to the effect of relatively enriched focal H3K36me2. We also noticed that, in KMS11TKO, not all genic 5mC signals matched H3K36me2 patterns (for examples, HS3ST3A1 and HS3ST3B1 in Fig. 2D), and we speculated that H3K36me3 might play a recruitment role in these regions. Consistent with the observations in Fig. 2D, the genomic distributions of KMS11-specific H3K36me2 and 5mC peaks are similar as average genome proportion (genic 45.3% and intergenic 54.7%), while KMS11TKO-specific peaks are more enriched in genic regions (Fig. 2E).
To further explore the functional importance on gene expression by the H3K36me2 and DNA 5mC axis, we carried out RNA-seq analyses in KMS11 and KMS11TKO cells (Fig. S2A). As revealed by other studies (Kuo et al., 2011), we identified hundreds of differentially expressed genes (DEGs). Among these DEGs, we found that the up-regulated genes in KMS11 are significantly enriched for pathways in cancer (Fig. S2B), and the altered expression of several oncogenes in myeloma, such as HSPG2, TREML2 and NCAM1 were validated by RT-qPCR (Fig. S2C). We also found that the upregulated genes in KMS11 showed higher upstream DNA 5mC which tended to be lost in KMS11TKO, compared to the downregulated genes (Fig. S2D, left). While the 5mC levels of the CpG islands at the transcription start sites (TSSs) of the up- and down- regulated genes were generally low and similar, although statistically significant (Fig. S2D, right). This observation raises an intriguing possibility of the upstream hyper 5mC in KMS11 being functionally involved in the downstream gene transcription, which certainly needs future investigation.
As NSD2 overexpression was demonstrated as the driver mutation for multiple myeloma bearing T(4;14) translocation (Keats et al., 2003; Santra et al., 2003; Kuo et al., 2011), we therefore speculated that the gain of 5mC as the consequence of NSD2 overexpression may also functionally involved in the tumorigenic processes. Supporting this idea, we found that the application of DNA methylation inhibitor, 5-Aza, effectively suppressed the proliferation of KMS11 cells at 0.25 μmol/L in 4 days (Fig. 2F). While under same condition, even at 1 μmol/L concentration, 5-Aza showed no effect on the proliferation of KMS11TKO, indicating that the global hypermethylation of 5mC is required for cancerous growth driven by NSD2 overexpression in KMS11.
Our findings revealed a previously under-appreciated function of H3K36me2 in regulating DNMT3A mediated DNA 5mC. Although partially overlapped in genic regions with H3K36me3, H3K36me2 also demarcates many intergenic regions. Together with the fact that H3K36me3 directs genic DNA 5mC methylation through DNMT3B (Baubec et al., 2015), H3K36 methylations play critical role in guiding DNMT3A and DNMT3B to keep the proper level of DNA 5mC methylation genome-wide. Interestingly, both H3K36me2 and H3K36me3 are absent in the promoter regions, consistently with and perhaps also mechanistically contribute to the 5mC hypomethylation at the active promoters. Furthermore, as NSD family members frequently undergo gain-of-function mutations in MM, AML, lung and breast cancers (Rosati et al., 2002; Wang et al., 2007; Oyer et al., 2014), our findings also support therapeutic opportunities for DNA methylation inhibitors in treating these tumors.
We thank Dr. Yanhui Xu for sharing DNMT3A related reagents. This project was supported by the National Key Research and Development program of China (2016YFA0101800 to F.L., 2018YFC1005004 to F.W., 2018YFA0108700 and 2016YFC1000500), the Shanghai Municipal Science and Technology Major Project (2017SHZDZX01), the National Natural Science Foundation of China (81272295 to F.W., 81773014 to F.L. and 31900469 to W.X.).
Fei Lan is a consultant of Active Motif Inc. and shareholder of Constellation Pharmaceuticals. Wenqi Xu, Jiahui Li, Bowen Rong, Bin Zhao, Mei Wang, Ruofei Dai, Qilong Chen, Hang Liu, Zhongkai Gu, Shuxian Liu, Rui Guo, Hongjie Shen and Feizhen Wu declare that they have no conflict of interest.
This article does not contain any studies with human or animal subjects performed by the any of the authors.
- Oyer JA, Huang X, Zheng Y, Shim J, Ezponda T, Carpenter Z, Allegretta M, Okot-Kotber CI, Patel JP, Melnick A et al (2014) Point mutation E1099K in MMSET/NSD2 enhances its methyltranferase activity and leads to altered global chromatin methylation in lymphoid malignancies. Leukemia 28:198–201CrossRefGoogle Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.