Hbx Induces Hypomethylation of Distal Intragenic Cpg Islands Required for Active Expression of Developmental Regulators

HBx induces hypomethylation of distal intragenic CpG islands required for active expression of developmental regulators

Sun-Min Leea, Young-gun Leeb, Jae-Bum Baea,1, Jung Kyoon Choia,2, Chiharu Tayamac, Kenichiro Hatac, Yungdae Yund, Je-Kyung Seonge, and Young-Joon Kima,b,3

aDepartment of Biochemistry, College of Life Science and Technology, Yonsei University, Seoul 120-749, Korea; bDepartment of Integrated Omics for Biomedical Science, World Class University Program, Yonsei University, Seoul 120-749, Korea; cDepartment of Maternal-Fetal Biology, National Research Institute for Child Health and Development, Tokyo 157-8535, Japan; dDepartment of Life Science and the Center for Cellular Homeostasis, Ewha Woman’s University of Life Science, Seoul 120-750, Korea; and eLaboratory of Developmental Biology and Genomics, College of Veterinary Medicine, Seoul National University, Seoul 151-742, Korea

Edited* by Roger D. Kornberg, Stanford University School of Medicine, Stanford, CA, and approved May 19, 2014 (received for review January 13, 2014)

Epigenetic alterations caused by viral oncoproteins are strong Some unmethylated intragenic and intergenic CGIs contain distinct initiation factors for cancer development, but their mechanisms epigenetic markers and work as enhancers or alternative promoters, are largely unknown. To identify the epigenetic effects of viral demonstrating the regulatory potential of CGIs. Although most hepatitis B virus X (HBx) that lead to hepatocellular carcinoma (HCC), promoter associate CGIs are unmethylated, about 20–30% of CGIs we profiled the DNA methylomes of normal and HBx transgenic found in the gene body are highly methylated (13) and suggested as mouse liver. Intriguingly, severe hypomethylation of intragenic CpG marking actively transcribed regions (14) or as suppressing pro- islands (CGIs) was observed in HBx liver before the full development moters for antisense transcripts (15). of HCC. Normally, these CGIs were highly methylated (mCGIs) by the The HBV x (HBx) protein, which plays a critical role in the DNMT3L complex and marked with epigenetic signatures associated development of HCC, was shown to interact with several epi- with active expression, such as H3K36me3. Hypomethylation of genetic factors, such as DNMT3A and HDAC1 (16). However, mCGI was caused by the downregulation of Dnmt3L and Dnmt3a the physiological significance of these interactions and the reg- due to HBx bound to their promoters, along with HDAC1. These ulatory mechanism leading to gene silencing are not clear. HBx events lead to the downregulation of many developmental regu- was shown to recruit DNMT3A to certain promoters to repress lators that could facilitate tumorigenesis. Here we provide an their activities through de novo DNA methylation; however, intriguing epigenetic regulation mediated by mCGI that is required HBx also causes hypomethylation in certain promoters, possibly via the redistribution of DNMT3A to other promoters (16). In for cell differentiation and describe a previously unidentified epigenetic role for HBx in promoting HCC development. Significance DNA methylation | methylated CpG island | viral protein Epigenetic dysregulation by oncoviral protein plays a key role epatocellular carcinoma (HCC) is one of the most danger- in tumor development. DNA methylome analysis of hepatitis B Hous cancers that threaten many people, especially those with virus X (HBx)-induced hepatocellular carcinoma (HCC) revealed hepatitis B or C virus (HBV or HCV) (1–3). However, the exact drastic changes in host epigenome, but in an unconventional mechanisms underlying HCC are not clear because there are way: intragenic CpG islands (CGIs) were dramatically deme- multiple factors, including chronic inflammation (4), genetic al- thylated. We showed methylated intragenic CGIs as previously teration caused by viral integration into the host genome (5), and unidentified regulatory elements associated with active expression. The methylated CGIs are marked with distinct epi- the oncogenic actions of viral proteins (6). Although many genetic signatures and require DNA methyltransferase (DNMT) studies have previously demonstrated that these factors, alone or 3L complex for their high methylation levels. By directly sup- in combination with other factors, are able to initiate tumori- pressing Dnmt3L and Dnmt3a promoters, HBx induces hypo- genesis (1), the oncogenic potential of viral protein in cancer methylation of the intragenic CGIs and downregulation of the development is one of the most interesting exogenic factors that associated developmental regulators. We provide previously facilitates tumorigenesis, in particular, through the epigenetic unreported functional identification of intragenic CGIs that regulation of infected cells (7). may enhance our understanding of epigenetic regulation and Epigenetic alterations in cancer cells are now regarded as one a new epigenetic role for HBx in promoting HCC development. of the most important factors driving cancer initiation. Abnormal DNA methylation is the most frequently found change in many Author contributions: S.-M.L. and Y.-J.K. designed research; S.-M.L. and J.-B.B. performed cancers and is believed to control associated gene expression research; Y.-g.L., J.K.C., C.T., K.H., Y.Y., and J.-K.S. contributed new reagents/analytic without altering the DNA sequence itself (8). The transcriptional tools; S.-M.L. analyzed data; and S.-M.L. and Y.-J.K. wrote the paper. silencing of tumor suppressor genes associated with the hyper- The authors declare no conflict of interest. methylation of promoter regions is a typical epigenetic change *This Direct Submission article had a prearranged editor. in many cancers (9). The DNA hypomethylation in repeat se- Data deposition: The sequences reported in this paper have been deposited in the Gene quences and transposable elements is known to induce chromo- Expression Omnibus (GEO) database, www.ncbi.nlm.nih.gov/geo (accession no. somal instability and mutation events (10) that lead to cancer GSE48052). development and progression (11). In addition, various types 1Present address: Division of Structural and Functional Genomics, Center for Genome of cancer cells exhibit abnormal expression levels of DNA methyl- Science, Korea National Institute of Health, Osongsaemyeong 2-ro, Osong-eup, transferase (DNMT) families, which probably causes global Cheongwon-gun, Chungcheongbuk-do 363-700, Korea. changes in DNA methylation (12). 2Present address: Department of Bio and Brain Engineering, Korea Advanced Institute of In mammalian cells, large clusters of CpG dinucleotides Science and Technology, Daejeon 305-701, Korea. 3 known as CpG islands (CGIs) appear to act as a key epigenetic To whom correspondence should be addressed. E-mail: [email protected]. GENETICS element regulating gene expression. Most CGIs are found at the This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. 5′ end of transcripts and behave as functional promoters (13). 1073/pnas.1400604111/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1400604111 PNAS | July 1, 2014 | vol. 111 | no. 26 | 9555–9560 Downloaded by guest on October 1, 2021 addition, as a transcriptional activator, HBx modifies diverse HBx13 vs. HBx13). Significant methylation changes were already signal transduction pathways that mediate cellular transformation obvious in the early stage of transformation (Norm3 vs. HBx3) (17, 18). Most HBx transgenic (17) mice spontaneously develop and were maintained during cancer progression (Norm13 vs. HCC at about 1 y of age, providing genetic validation of the HBx13). The methylation levels did not change a lot between oncogenic potential of HBx even in the absence of viral in- different ages compared with the methylation changes caused by tegration and chronic inflammation (6). tumor development. The differentially methylated regions (DMRs) To address the role of HBx viral oncoprotein in the de- were highly enriched in exons, with the most striking changes at velopment of HCC, specifically the aspect of epigenetic regula- CGIs in both age groups (Fig. 1B). To understand the role of DNA tion of the host genome, we examined the differences in DNA methylation changes in the early stage of tumor development, we methylation patterns between normal and HBx-induced liver carried out additional MIRA-seq and RNA-seq analyses of the cancer cells. Intriguingly, DNA methylome analysis revealed 3-mo-old age group to expand the methylation sequence reads dramatic demethylation mainly in the intragenic CGIs with con- up to 170 million for each sample, which covers more than 85% current downregulation of the affected genes. Reanalysis of the of the mouse genome with >70× coverage of the CpG site (SI epigenome data from human cell lines revealed specific epige- Appendix,TableS1B and C). We used these high-density methyl- netic signatures associated with a distinct group of the intragenic omes of 3-mo-old normal and HBx liver to understand the detained CGIs, indicating that they are potential epigenetic regulatory nature of DNA methylation changes caused by HBx. elements for active gene expression. The suppression of Dnmt3L We found 10,553 tumor-associated DMRs (6,668 hyper- and Dnmt3a promoters by HBx appears to cause deficient DNA methylated and 3,885 hypomethylated regions) with differences methyltransferase activity in HBx transgenic (TG) mouse liver, of at least 7 RPKM in the 3-mo-old age group. For comparison, the expected frequency of variations was normalized by the length which renders the intragenic CGIs hypomethylated. Therefore, SI Appendix dysregulation of the epigenetic process at the intragenic CGIs or CpG counts of the mouse genomic structures ( , Table S2). Exonic regions occupy less than 2% of the mouse ge- appeared to contribute to tumorigenesis. Our findings reveal an ∼ intriguing epigenetic mechanism mediated by highly methylated nome, but 29 and 45% of the total hyper- and hypomethylated intragenic CGIs in the regulation of cell differentiation. regions, respectively, were within this region showing 14.9 (to length) and 34.1 (to CpG counts) times higher enrichment in SI Appendix Results exons compared with the genome-wide average ( , Table S2 A and B). Therefore, a portion of exons appeared to be Global Comparison of DNA Methylation in Normal and HBx TG Mouse highly deregulated in the HBx-induced liver cancer model at an Liver. To examine the DNA methylation changes leading to early stage of tumor development. cancer, we created DNA methylation maps for liver tissues from HBx TG mice and their wild-type littermates at early (3-mo-old) Drastic Hypomethylation of Intragenic CGIs in HBx-HCC. To examine and late (13-mo-old) stages of tumor development (6, 19). We whether DMR enrichment in exons was associated with any re- sequenced the methylated DNA from at least two independent petitive elements, various repeats and CpG islands were ana- preparations for each sample using the methylated CpG island lyzed for bias in DMR association. Except for demethylation of recovery assay (MIRA) (20). To measure enrichment of meth- the satellite elements, HBx TG liver showed no major methyla- ylation signals, we counted the number of overlapping reads at tion changes in most repeats. However, CpG islands were sig- every 50 bp across the whole genome and then calculated RPKM nificantly hypomethylated in HBx-induced tumor cells (SI Appendix, (reads per million base pairs mapped per kilobase) as normalized Table S2B). To further study the meaning of CGI-associated DNA methylation values (21–23). Each methylome contained demethylation, we subgrouped CGIs by their genomic locations: ∼20 million sequence reads, covering ∼40% of all genomic CpG promoter-linked, intragenic, and intergenic CGIs (13). About 70% sites, excluding centromeres. The average read depth for covered of CGIs are associated with promoters, and most of the promoter- each CpG was 10×, which is sufficient for reliable comparison associated CGIs are methylation-free. However, a considerable among the samples (SI Appendix, Table S1A). percentage of the intragenic CGIs (20–34%) were highly meth- Scatter plot analysis of the four methylomes revealed a de- ylated and showed variations in methylation levels among dif- crease in global DNA methylation both in 3- and 13-mo-old mice ferent tissues (15, 24). Comparison of the methylation levels at with HBx-induced HCC (Fig. 1A). The values in the scatter plots these CGIs revealed that only the intragenic CGIs were drasti- represented RPKM (at every 50 bp) of the four methylomes cally demethylated in the HBx TG liver (Fig. 1C), and this (Norm3 vs. HBx3, Norm13 vs. HBx13, Norm3 vs. Norm13, and was observed throughout the entire length of the CGIs without

Fig. 1. Global hypomethylation in the HBx-exposed liver. (A) Global comparisons of the four methylomes (Norm3 vs. HBx3, Norm13 vs. HBx13, Norm3 vs. Norm13, and HBx13 vs. HBx13). Scatter plots of RPKM at every 50 bp in the covered genome. The one dot represents methylation levels (RPKM) of the one site in windows of 50 bp of the mouse genome. The bold black lines show the Q-Q plots under different con- ditions for comparing two probability distributions by plotting their quantiles against each other. (B) The DMR in normal and HBx TG liver of 3- and 13-mo-old mice and the relative enrichment ratio to length of genomic elements. The expected frequency of variations was normalized by the length of the mouse genomic structures. (C) Differential methylation for supgrouped CGIs depends on their genomic locations in normal and HBx TG liver of 3-mo-old mice (HBx- Norm, RPKM). Differences were assessed with two- tailed t tests; *P < 0.01. (D) Average plot of DNA methylation on intragenic CGI and surrounding 4-kb regions (solid line, Norm; dashed line, HBx, RPKM) Each CGI was divided into 100 bins, and surrounding 4-kb regions were plotted at 40-bp resolution. (E) Intragenic CGIs consist of exonic and intronic regions. CGIs covering exons and introns were considered exonic. Box plot shows differential methylation for exonic vs. intronic CGIs (HBx-Norm, RPKM). Differences were assessed with a two-tailed t test; ***P < 0.0001.

9556 | www.pnas.org/cgi/doi/10.1073/pnas.1400604111 Lee et al. Downloaded by guest on October 1, 2021 affecting the methylation levels of their shore regions (Fig. 1D). Intriguingly, even for the intragenic CGIs, CGIs overlapping with exons were highly demethylated (Fig. 1E). Considering that most CGI promoters were not affected, this result indicates that CGIs overlapping with exons might be the major targets of the cancer-associated demethylation process.

Downregulation of Genes Containing Hypomethylated Intragenic CGIs. To understand the genetic structures associated with intragenic CGIs, we plotted the relative locations of CGIs and exons along the length of their corresponding gene structures for the 647 genes containing intragenic CGIs that were hypomethylated in HBx TG mouse liver. The DNA methylation levels in normal livers were displayed along with the levels of differential methylation and RNAs in the HBx TG mouse liver. The locations of CGIs revealed two distinct clusters, one at the promoter and the other near the 3′ end of the transcript (Fig. 2A). The promoter-associated CGIs were methylation-free, but the CGIs overlapping mostly with the distal exons near the 3′ end of the transcripts showed a significantly high level of DNA methylation. In general, distal exons with high levels of DNA methylation showed higher levels of RNA expression (Fig. 2B). In HBx liver, the DNA methylation level was drastically reduced, mainly at the distal exonic CGIs, along with the reduction of the transcript levels from the demethylated exons (Fig. 2A). In- triguingly, the negative effect of CGI hypomethylation was not limited to the associated exons. The RNA levels from the proximal exons of the hypomethylated CGIs were also concomitantly decreased (Fig. 2C). Therefore, the high levels of DNA methylation at the distal intragenic CGIs may play an important role in maintaining high RNA expression levels. Gene ontology analysis of the hypomethylated genes revealed a high enrichment of the genes involved in cell adhesion, embryonic morphogenesis, signaling, and transcription (SI Appendix, Table S3 and deregulated genes listed in Table S4), reflecting the regulatory changes associated with HBx-induced tumorigenesis. Downregulation of genes involved in epithelial-to-mesenchymal transition (E- and N-cadherins), the Smad-dependent TGF- β–signaling pathway (Smad6 and Kcp), and the Wnt-signaling pathway (Wnts) is known to play crucial roles in tumorigenesis. Indeed, these genes were downregulated in HBx-TG liver (SI Appendix, Fig. S1). Thus, the broad deregulation of these factors by CGI hypomethylation in HBx TG liver may contribute to Fig. 2. Hypomethylation of distal exonic CGIs downregulates gene ex- tumor development. pression. (A) Heat maps display 647 individual genes associated with hypomethylated exonic CGIs are indicated by a single line. Promoter regions [−1 Distinct Epigenetic Features of the Methylated Intragenic CGIs. To to +0.5 kb from transcription start site (TSS)] were divided into 30 bins, and explore the possibility of these methylated intragenic CGIs being gene-body regions (TSS + 0.5 kb to transcription end site) were divided into a distinct epigenetic element, we examined epigenetic signatures 100 bins (gray line indicates TSS + 0.5-kb site). Genes were arranged from specifically associated with them. Promoter-associated CGIs top to bottom by CGI location in their gene body. Exon and CGI localization found in most housekeeping genes are unmethylated and highly are represented as light blue and yellow colors, respectively (Exon, CGI). The enriched with trimethylation of H3 lysine 4 (H3K4me3) (25). red gradient indicates normal DNA methylation levels (Met, Norm). The heat On the other hand, intragenic CGIs marked with H3K4me3 or map shows relative methylation or RNA level differences as increased (red) and H3K4me1 act as “orphan promoters” and “enhancers,” respec- decreased (green) (Met, Diff; RNA, Diff). (B) RNA expression levels on distal exons associated with CGIs according to their methylation levels. (C) Gene tively (15, 26, 27). Because these CGIs with distinctive histone expression changes due to hypomethylation of distal exonic CGIs. Box plot of marks are generally unmethylated, the methylated CGIs at the RPKM changes (HBx-Norm) from RNA-seq data. (D)HeatmapbasedonK- distal exons do not appear to act as alternative promoters/ means clustering method (K = 3). Intragenic CGIs can be divided into three enhancers. To validate this idea, we analyzed intragenic CGIs subgroups based on the levels of H3K4me3 (ENCODE), H3K4me1 (ENCODE), in the mouse genome by K-means clustering with the levels of and DNA methylation. Average plot of H3 lysine 4 or DNA methylation on each H3K4me3, H3K4me1, and DNA methylation using mouse liver grouped CGI and the surrounding 4-kb regions. (black, H3K4me1; blue, ENCODE data (28). Intriguingly, these epigenetic markers can H3K4me3; red, DNA met, RPKM). Each CGI and the surrounding 4-kb regions cluster intragenic CGIs into three distinct classes: promoter-like were divided into 100 bins. (E) Box plot shows differential methylation for CGIs (pCGIs) with a high level of H3K4me3, enhancer-like CGIs mCGIs, pCGis, and eCGIs (HBx-Norm, RPKM). Differences were assessed (eCGIs) with intermediate levels of both H3K4me1 and H3K4me3, with a two-tailed t test; ***P < 0.0001. and the methylated CGIs (mCGIs) with high levels of only DNA methylation (Fig. 2D). Among them, only the mCGIs were aber- rantly methylated in HBx TG liver (Fig. 2E). into three groups (pCGI-, eCGI-, and mCGI-like) based on their To test whether the distinct types of CGIs were evolutionarily levels of DNA methylation, H3K4me1, and H3K4me3 (SI Ap- conserved features, epigenome data (29) obtained from three pendix, Fig. S2A). The clustering patterns were similar in all

different types of human cell lines (HepG2, hepatocellular car- three cell types in that most of the mCGIs in the three different GENETICS cinoma; GM12878, B-lymphocyte; and K562, leukemia) were cell lines overlapped (SI Appendix, Fig. S2B) and were associated similarly analyzed. Human intragenic CGIs were also clustered with exons (77%) (SI Appendix, Fig. S2C). The human genes

Lee et al. PNAS | July 1, 2014 | vol. 111 | no. 26 | 9557 Downloaded by guest on October 1, 2021 containing mCGIs were also enriched with developmental reg- levels were not limited to mCGIs but appeared to spread out to ulators as observed in mouse (SI Appendix, Table S5). These the entire gene body (Fig. 3A and SI Appendix, Fig. S2D). results indicate that mCGI methylation is an epigenetic marker that is strongly associated with active RNA expression from key Requirement of DNMT3L for mCGI Methylation. To examine whether developmental genes. the hypomethylation of mCGIs in HBx-induced HCC was caused by the loss of DNA methyltransferase activity, we examined total mCGI Methylation as an “Active” Epigenetic Mark. To further DNA methyltransferase activities of nuclear extracts from normal characterize the distinct classes of CGIs, we examined levels of liver and HBx TG liver. Indeed, HBx TG liver samples showed diverse epigenetic markers (H3K27me3, H3K36me3, H3K9me3, reduced DNA methyltransferase activity compared with their A MNase, and DNase I accessibility) associated with each CGI normal counterparts (Fig. 4 ). To identify the DNMT responsible using the human GM12878 cell line ENCODE epigenome data for the decreased DNA methyltransferase activity in HBx TG (29). In addition to the high level of H3K4me3, pCGIs were mouse liver, the levels of diverse DNMTs were measured by characterized by a highly open chromatin structure and a high quantitative RT-PCR (RT-qPCR). Although there were no major differences in Dnmt1 and Dnmt3b transcript levels, Dnmt3L ex- density of RNA polymerase II, whereas eCGIs were marked with A pression, along with a moderate decrease in Dnmt3a expression, a high level of H3K27me3 (Fig. 3 ). Conversely, mCGIs, found was significantly decreased in HBx TG mouse liver (Fig. 4B). mostly in exons, were highly enriched with H3K36me3 but DNMT3L and DNMT3A are known to play key roles in the ac- showed the lowest levels of H3K27me3, H3K9me3, and DNase I cumulation of high levels of gene-body–associated DNA methyl- accessibility among the intragenic CGIs. In addition, the analysis ation during embryo development (30). To test whether Dnmt3L- of the mouse ENCODE epigenome data revealed an identical dependent gene-body methylation is mediated through the result (SI Appendix, Fig. S2E). methylation of mCGIs, we analyzed DNA methylome changes in Therefore, mCGIs appear to have a chromatin structure dis- early embryos induced by partial deficiency of Dnmt3L. Genome- tinct from those found in pCGIs and eCGIs. The striking dif- wide comparison of the methylation levels between wild-type and − + ferences in DNA accessibility and histone modification marks Dnmt3L / mice revealed that the overall methylation defects between mCGIs and the other CGIs suggested that the levels of in exons were similar to those of the HBx-TG mouse (Fig. 4C); DNA methylation at mCGIs may be related to the levels of gene in particular, mGCIs were specifically hypomethylated in the −/+ expression. Alignment of mCGIs according to their methylation Dnmt3L mouse (Fig. 4D). In addition, adult mouse liver had levels demonstrated positive and negative correlation of meth- a low but significant level of Dnmt3L expression (Fig. 4E), and the ylation levels with those of H3K36me3 and H3K27me3, re- reduced levels of Dnmt3L expression correlated with the loss of spectively (Fig. 3B). Consistent with this observation, the highly DNA methylation at mCGIs of the target genes (Fig. 4F and SI methylated mCGIs showed higher expression levels (Fig. 3C). Appendix,Fig.S3). Therefore, the reduced expression of Dnmt3L In addition, reflecting the broad effect of mCGI methylation and Dnmt3a may synergistically lower the level of functional DNA on RNA expression, the changes in H3K27me3 and H3K36me3 methyltransferase activity in HBx-induced HCC, which results in the reduced DNA methylation of mCGIs during early development. To examine the recruitment of DNMT3L and its associated epigenetic effect on mCGIs, the levels of DNMT3L and histone modification markers (H3K4me36 and H3K4me27) in wild-type and HBx-TG mouse liver were monitored by chromatin immunoprecipitation (IP)-coupled quantitative PCR (ChIP-qPCR) analysis. The decrease of mCGI DNA methylation levels in the target genes appeared to be related to the loss of DNMT3L bound to the region along with the epigenetic deregulation of H3K27me3 and H3K36me3 (Fig. 4G and SI Appendix, Fig. S4). This result suggests that defective DNMT3L expression and recruitment of the mCGIs from key developmental genes in HBx TG liver may cause abnormal epigenetic regulation that con- tributes to the development of hepatocellular carcinoma.

HBx-Mediated Repression of the Dnmt3L Promoter by HBx. To understand how HBx downregulated Dnmt3L expression, epigenetic factors bound to the Dnmt3L promoters were examined by ChIP. In mice, Dnmt3L is regulated via three different promoters: Dnmt3L-S (stem cell, the canonical transcript), Dnmt3L-O (oocyte), and Dnmt3L-AT (adult testis). As previously shown (31), Dnmt3L-S appears to be the major promoter in liver; most of the transcription factors examined were detected only in the Dnmt3L-S region, and no significant binding of these factors was detected in either of the germ cell-specific promoters (Fig. 5A). The S promoter region, which contains four SP1-binding sites (32), was highly enriched with the SP1 transcription factor and Fig. 3. mCGIs: a distinct class of intragenic CGIs. (A) Average plot of active histone mark, H3K4me3, in normal liver. However, in H3K27me3, H3K36me3, MNase, DNase I, and H3K9me3 on intragenic CGI HBx TG liver, the levels of SP1 and H3K4me3 at the promoter and surrounding 4-kb regions in the Gm12878 cell line from Human ENCODE were greatly diminished. Instead, the promoter in HBx TG liver (solid line, mCGI; dash-dot line, pCGI; dot line, eCGI, RPKM). Each CGI and A surrounding 4-kb regions were divided into 100 bins. (B) H3K27me3 and was highly bound by HBx and HDAC1 (Fig. 5 ). These results H3K36me3 levels on mCGI antagonize each other and are dependent on indicate that the binding of HBx to the Dnmt3L-S promoter si- DNA methylation level. mCGIs were lined up according to their methylation lenced the promoter by recruiting HDAC1. Dnmt3a promoter level, and the black gradient indicates the levels of H3K27me3 and H3K36me3 was also similarly repressed by HBx-induced epigenetic changes in the Gm12878 cell line. (C) Expression levels (RPKM from RNA-seq) of genes (SI Appendix, Fig. S5). In addition, the transfection of exogenous associated with the 100 lowest methylated mCGIs (low, average: 16.3%) and FLAG epitope-tagged HBx into HepG2 caused downregulation highest methylated 100 mCGIs (high, average: 100%). Differences were of DNMT3L expression specifically (Fig. 5B), and ChIP-qPCR assessed with a t test; ***P < 0.0001. analysis revealed that the DNMT3L promoter was highly

9558 | www.pnas.org/cgi/doi/10.1073/pnas.1400604111 Lee et al. Downloaded by guest on October 1, 2021 on the entire gene region suggest their regulatory potential in the maintenance of active transcription. Third, mCGIs appear to harbor high nucleosome densities. CGIs found mostly at promoters are devoid of nucleosomes, and high DNA accessibility is an important feature required for transcription factor binding (34–36). Similarly, pCGIs and eCGIs also show high DNA accessibility. Although we do not know how mCGIs enhance gene expression, the distinct locations near the end of a gene and the high levels of nucleosome densities associated with mCGIs hinted that an active chromatin organi- zation mediated by specifically modified nucleosomes, rather than specific DNA-binding factors, may generate a favorable environment for active transcription. It would be intriguing to examine the distinct chromatin conformations associated with mCGIs. Unlike most CGIs that are maintained in methylation-free states, mCGIs are highly methylated, but how this targeted methylation is achieved is not well understood. De novo CGI Fig. 4. DNMT3L is required for mCGI methylation. (A) DNMT activity assay methylation is mediated by DNMT3A and DNMT3B, which are from nuclear extracts of normal and HBx TG mouse liver. The amount of stimulated by DNMT3L, a nonenzymatic subunit of DNA meth- methylated DNA, which is proportional to enzyme activity, was colorimet- – rically quantified. Each bar represents the mean + SEM of n = 3 pools. (B) yltransferase (37 39). Because the ADD domains in DNMT3A, DNMT family expression was analyzed by RT-qPCR, and the relative ex- DNMT3B, and DNMT3L bind to histone H3 only when H3K4 is pression was normalized to Gapdh. Each bar represents the mean + SEM of not methylated, the mCGIs that are devoid of methylated H3K4 n = 3 pools. (C and D) Knockdown effects of Dnmt3L. MeDIP-seq data from marks can be targeted specifically by the DNMT3 complex (39, − + embryonic day 8.5 wild-type and Dnmt3L / mice (46). (C) Hypomethylated 40). In addition, the high levels of H3K36me3 associated with DMR relative enrichment ratio to genomic element length (Dnmt3L MT-WT). mCGIs also contribute to the mCGI-specific recruitment of (D) Box plot shows differential methylation for grouped intragenic CGI DNMT3 complexes. The PWWP domain of DNMT3A, but not (Dnmt3L MT-WT, RPKM); Differences were assessed with a two-tailed t test; DNMT3B, can interact with H3K36me3, providing an additional ***P < 0.0001. (E and F) Knockdown effects of Dnmt3L. (E) Dnmt3L ex- link between DNA methyltransferase and mCGIs (41). In addi- − + − − pression levels form 8-wk-old wild-type (47), Dnmt3L / , and Dnmt3L / tion to the preferential recruitment of DNMT3 complexes to the mice. Dnmt3L expression in the liver was analyzed by RT-qPCR, and the relative mCGIs, the Setd2 complex (H3K36 methyltransferase) enriched expression was normalized to Gapdh. Each bar represents the mean + SEM of at the mCGIs prevents H3K4 methylation by recruiting KDM1b, n = 4pools.(F) Examples of plots are shown for DNA methylation and MIRA- a H3K4me1/2 demethylase (42). On the contrary, the loss of DNA ’ qPCR binding profiles. The MIRA data represent IP values for each region s methylation in mCGI has a part to play in recruitments of H3K4 relative ratio to the input. CGI locations are displayed in Wnt3 loci (solid line, + − − − methyltransferases (Mll1 and Cfp1/Set1) and H3K36me1/2 deme- Dnmt3L WT; dashed line, Dnmt3L / ; dotted line, Dnmt3L / ) The P5 indicates mCGI. (G) Examples of plots are shown for DNMT3L, H3K36me3, and H3K27me3 thylase (Kdm2a). Their CXXC domains bind to unmethylated ChIP-qPCR binding profiles in Wnt3 loci. The ChIP data represent IP values for each region’s relative ratio to the input (solid line, Norm; dashed line, HBx). The P5 indicates hypomethylated mCGI in HBx TG liver.

occupied with HBx along with HDAC1 (Fig. 5C). Therefore, HBx appears to directly regulate DNMT expressions via epigenetic modification. Discussion Here we present a previously unidentified epigenetic regulatory element, mCGI, required for active expression of many genes involved in development. Several lines of evidence support mCGIs as a previously unidentified type of epigenetic regulatory element that is different from the promoter/enhancer functions attributed to other types of CGIs. First, the high abundance and depletion of H3K36me3 and H3K27me3, respectively, are associated only with mCGI. This pattern is quite different from those of other intragenic CGIs that contain high H3K4 methylation or H3K27me3 marks required for the assembly of active or repressed transcription machineries, respectively. mCGIs also appear to differ from the recently reported partially methylated domains (PMDs) (33). Unlike the reverse correlation between methylation levels and the repressive histone modification marks H3K27me3 and H3K9me3, shown in PMDs, the hypomethylation of mCGIs did not cause higher levels of H3K9me3. Second, the locations of mCGIs show a strong bias toward the Fig. 5. Epigenetic repression of Dnmt3L by HBx. (A) Occupancies of HBx, 3′ end of the transcripts, whereas no such obvious structural HDAC1, H3Ace, H3K4me3, H3K27me3, and SP1 in Dnmt3L promoter (solid restrain was found for the other intragenic CGIs. Although line, Norm; dashed line, HBx). The ChIP data represent IP values for each ’ methylation is highly centered on mCGIs located near the end of region s relative ratio to the input. CGI locations are displayed in Dnmt3L loci (Dnmt3L -O, -S, and -AT). (B) HBx and DNMT3L expression levels in Mock and the transcripts, their hypomethylation caused a significant shift stably HBx-transfected HepG2 cells. Each bar represents the mean + SEM of

of histone modification from active to repressive patterns, not n = 3pools.(C) Occupancies of HBx and HDAC in the DNMT3L promoter (solid GENETICS just in the mCGIs but throughout the entire length of the mCGI- line, Mock; dashed line, HBx-flag). The ChIP data represent IP values for each containing genes. The specific location of mCGIs and their effect region relative to the input. The P4 indicates the promoter region.

Lee et al. PNAS | July 1, 2014 | vol. 111 | no. 26 | 9559 Downloaded by guest on October 1, 2021 CpGs, resulting in depletion of H3K36 methylation and increase of downregulation of Dnmt3L and suggests that a restoration of H3K4 methylation (43). Therefore, cooperative interplay between proper de novo DNA methylation activity may trigger pro- DNA methylation and histone modifications on mCGI may play an liferative tissues to differentiate. In addition, the functional important role in maintaining high RNA expression levels. identification of the mCGI element may enhance our under- We observed a severe decrease of mCGI methylation in HBx- standing of epigenetic regulation mediated by differential meth- HCC, which appeared to be induced by the loss of DNMT ylation during cell differentiation. expressions. HBx bound to the promoters of Dnmt3L and Dnmt3a and recruited HDAC1 to downregulate their expressions. HBx does Materials and Methods not bind directly to DNA but is recruited to promoters by indirectly HBx transgenic mice and nontransgenic mice (C57/BL/6) were maintained binding to other transcription factors (44). In addition to the tran- according to the Guide for Animal Experiments (edited by the Korean scriptional suppression of Dnmt3a and Dnmt3L, HBx can inhibit Academy of Medical Sciences) by the Institutional Animal Care and Use DNMT3A activity by directly binding to the enzyme (16). Therefore, Committee (IACUC) at Seoul National University. the multilayered inhibition of the DNMT3A:DNMT3L complex Protocols were reviewed and approved by the IACUC of the Yonsei Lab- by HBx would result in a dramatic reduction in mCGI methylation oratory Animal Research Center at Yonsei University (Permit 2008-0012). We even at the small decrease of individual subunits of DNMT complex. attempted to sequence methylated DNA from the normal liver and HBx- Indeed, we showed that the recruitment of DNMT3L at the hypo- induced HCC samples with the aid of next-generation sequencing and MIRA methylated mCGI is significantly decreased in HBx-induced HCC. It (20). The methylation levels of CGI, promoter, gene bodies, and repeat is intriguing that DNMT3L expression is also significantly suppressed elements were estimated by means of RPKM (or means of DNA methylation in human HBV-positive HCC (SI Appendix,Fig.S6)(45).There- percentage for bisulfite sequencing) overlapping each element. All public fore, the reduction of de novo DNA methylation activity caused by data were processed with the same procedures. Detailed descriptions of HBx may be responsible for the abnormal cell differentiation. methods are available in SI Appendix. Although HBx has been known to exert a pleotropic effect on diverse cellular mechanisms, the deregulation of a large number ACKNOWLEDGMENTS. This research was supported by the Global Research of development genes by mCGI hypomethylation appears to be Laboratory Program through the National Research Foundation of Korea one of the major factors driving tumoregenesis at least in liver. funded by the Ministry of Education, Science, and Technology (K20704000006- This study revealed a previously unidentified HBx function in 07A0500-00610).

1. Farazi PA, DePinho RA (2006) Hepatocellular carcinoma pathogenesis: From genes to 24. Illingworth RS, et al. (2010) Orphan CpG islands identify numerous conserved pro- environment. Nat Rev Cancer 6(9):674–687. moters in the mammalian genome. PLoS Genet 6(9):e1001134. 2. Lavanchy D (2004) Hepatitis B virus epidemiology, disease burden, treatment, and 25. Saxonov S, Berg P, Brutlag DL (2006) A genome-wide analysis of CpG dinucleotides in current and emerging prevention and control measures. J Viral Hepat 11(2):97–107. the human genome distinguishes two distinct classes of promoters. Proc Natl Acad Sci 3. Bowen DG, Walker CM (2005) Adaptive immune responses in acute and chronic USA 103(5):1412–1417. hepatitis C virus infection. Nature 436(7053):946–952. 26. Rada-Iglesias A, et al. (2011) A unique chromatin signature uncovers early de- 4. Rapicetta M, Ferrari C, Levrero M (2002) Viral determinants and host immune re- velopmental enhancers in humans. Nature 470(7333):279–283. sponses in the pathogenesis of HBV infection. J Med Virol 67(3):454–457. 27. Kowalczyk MS, et al. (2012) Intragenic enhancers act as alternative promoters. Mol 5. Sung WK, et al. (2012) Genome-wide survey of recurrent HBV integration in hepa- Cell 45(4):447–458. tocellular carcinoma. Nat Genet 44(7):765–769. 28. Stamatoyannopoulos JA, et al.; Mouse ENCODE Consortium (2012) An encyclopedia 6. Yu DY, et al. (1999) Incidence of hepatocellular carcinoma in transgenic mice ex- of mouse DNA elements (Mouse ENCODE). Genome Biol 13(8):418. – pressing the hepatitis B virus X-protein. J Hepatol 31(1):123 132. 29. Bernstein BE, et al.; ENCODE Project Consortium (2012) An integrated encyclopedia of 7. Dickinson M, Johnstone RW, Prince HM (2010) Histone deacetylase inhibitors: Potential DNA elements in the human genome. Nature 489(7414):57–74. – targets responsible for their anti-cancer effect. Invest New Drugs 28(Suppl 1):S3 S20. 30. Smallwood SA, et al. (2011) Dynamic CpG island methylation landscape in oocytes and – 8. Feinberg AP, Tycko B (2004) The history of cancer epigenetics. Nat Rev Cancer 4(2):143 153. preimplantation embryos. Nat Genet 43(8):811–814. 9. Dawson MA, Kouzarides T (2012) Cancer epigenetics: From mechanism to therapy. 31. O’Doherty AM, et al. (2011) DNA methylation plays an important role in promoter – Cell 150(1):12 27. choice and protein production at the mouse Dnmt3L locus. Dev Biol 356(2):411–420. 10. Wilson AS, Power BE, Molloy PL (2007) DNA hypomethylation and human diseases. 32. Aapola U, Mäenpää K, Kaipia A, Peterson P (2004) Epigenetic modifications affect – Biochim Biophys Acta 1775(1):138 162. Dnmt3L expression. Biochem J 380(Pt 3):705–713. 11. Badal V, et al. (2003) CpG methylation of human papillomavirus type 16 DNA in 33. Hon GC, et al. (2012) Global DNA hypomethylation coupled to repressive chromatin cervical cancer cell lines and in clinical specimens: Genomic hypomethylation corre- domain formation and gene silencing in breast cancer. Genome Res 22(2):246–258. lates with carcinogenic progression. J Virol 77(11):6227–6234. 34. Choi JK, Bae JB, Lyu J, Kim TY, Kim YJ (2009) Nucleosome deposition and DNA 12. Daniel FI, Cherubini K, Yurgel LS, de Figueiredo MA, Salum FG (2011) The role of methylation at coding region boundaries. Genome Biol 10(9):R89. epigenetic transcription repression and DNA methyltransferases in cancer. Cancer 35. Choi JK, Kim YJ (2009) Intrinsic variability of gene expression encoded in nucleosome 117(4):677–687. positioning sequences. Nat Genet 41(4):498–503. 13. Deaton AM, Bird A (2011) CpG islands and the regulation of transcription. Genes Dev 36. Choi JK, Kim YJ (2008) Epigenetic regulation and the variability of gene expression. 25(10):1010–1022. Nat Genet 40(2):141–147. 14. Lister R, et al. (2009) Human DNA methylomes at base resolution show widespread 37. Chen ZX, Mann JR, Hsieh CL, Riggs AD, Chédin F (2005) Physical and functional in- epigenomic differences. Nature 462(7271):315–322. teractions between the human DNMT3L protein and members of the de novo 15. Maunakea AK, et al. (2010) Conserved role of intragenic DNA methylation in regu- methyltransferase family. J Cell Biochem 95(5):902–917. lating alternative promoters. Nature 466(7303):253–257. 38. Jin B, et al. (2012) Linking DNA methyltransferases to epigenetic marks and nucleo- 16. Zheng DL, et al. (2009) Epigenetic modification induced by hepatitis B virus X protein – via interaction with de novo DNA methyltransferase DNMT3A. J Hepatol 50(2): some structure genome-wide in human tumor cells. Cell Reports 2(5):1411 1424. 39. Smallwood SA, Kelsey G (2012) De novo DNA methylation: A germ cell perspective. 377–387. – 17. Chin R, et al. (2007) Modulation of MAPK pathways and cell cycle by replicating Trends Genet 28(1):33 42. hepatitis B virus: Factors contributing to hepatocarcinogenesis. J Hepatol 47(3): 40. Zhang Y, et al. (2010) Chromatin methylation activity of Dnmt3a and Dnmt3a/3L is 325–337. guided by interaction of the ADD domain with the histone H3 tail. Nucleic Acids Res – 18. Lee YH, Yun Y (1998) HBx protein of hepatitis B virus activates Jak1-STAT signaling. 38(13):4246 4253. J Biol Chem 273(39):25510–25515. 41. Dhayalan A, et al. (2010) The Dnmt3a PWWP domain reads histone 3 lysine 36 tri- – 19. Koo JS, et al. (2005) Large liver cell dysplasia in hepatitis B virus × transgenic mouse methylation and guides DNA methylation. J Biol Chem 285(34):26114 26120. liver and human chronic hepatitis B virus-infected liver. Intervirology 48(1):16–22. 42. Fang R, et al. (2010) Human LSD2/KDM1b/AOF1 regulates gene transcription by 20. Rauch T, Li H, Wu X, Pfeifer GP (2006) MIRA-assisted microarray analysis, a new modulating intragenic H3K4me2 methylation. Mol Cell 39(2):222–233. technology for the determination of DNA methylation patterns, identifies frequent 43. Hashimoto H, Vertino PM, Cheng X (2010) Molecular coupling of DNA methylation methylation of homeodomain-containing genes in lung cancer cells. Cancer Res and histone methylation. Epigenomics 2(5):657–669. 66(16):7939–7947. 44. Murakami S (2001) Hepatitis B virus X protein: A multifunctional viral regulator. 21. Mortazavi A, Williams BA, McCue K, Schaeffer L, Wold B (2008) Mapping and J Gastroenterol 36(10):651–660. quantifying mammalian transcriptomes by RNA-Seq. Nat Methods 5(7):621–628. 45. Chen L, et al. (2013) Recoding RNA editing of AZIN1 predisposes to hepatocellular 22. Zhang L, et al. (2013) Tet-mediated covalent labelling of 5-methylcytosine for its carcinoma. Nat Med 19(2):209–216. genome-wide detection and sequencing. Nat Commun 4:1517. 46. Proudhon C, et al. (2012) Protection against de novo methylation is instrumental in main- 23. Vining KJ, et al. (2012) Dynamic DNA cytosine methylation in the Populus trichocarpa taining parent-of-origin methylation inherited from the gametes. Mol Cell 47(6):909–920. genome: Tissue-level variation and relationship to gene expression. BMC Genomics 47. Cowin PA, Anglesio M, Etemadmoghadam D, Bowtell DD (2010) Profiling the cancer 13:27. genome. Annu Rev Genomics Hum Genet 11:133–159.

9560 | www.pnas.org/cgi/doi/10.1073/pnas.1400604111 Lee et al. Downloaded by guest on October 1, 2021