Volume 60, Issue 4, Pages 584-596 (November 2015) H3K4/H3K9me3 Bivalent Chromatin Domains Targeted by Lineage-Specific DNA Methylation Pauses Adipocyte Differentiation  Yoshihiro Matsumura, Ryo Nakaki, Takeshi Inagaki, Ayano Yoshida, Yuka Kano, Hiroshi Kimura, Toshiya Tanaka, Shuichi Tsutsumi, Mitsuyoshi Nakao, Takefumi Doi, Kiyoko Fukami, Timothy F. Osborne, Tatsuhiko Kodama, Hiroyuki Aburatani, Juro Sakai  Molecular Cell  Volume 60, Issue 4, Pages 584-596 (November 2015) DOI: 10.1016/j.molcel.2015.10.025 Copyright © 2015 Elsevier Inc. Terms and Conditions

Molecular Cell 2015 60, 584-596DOI: (10.1016/j.molcel.2015.10.025) Copyright © 2015 Elsevier Inc. Terms and Conditions

Figure 1 H3K4/H3K9me3 Bivalent Chromatin Domains in Lineage-Committed Cells (A) Genome browser representation for H3K4me3 and H3K27me3 on Cebpa and Pparg in TT2 ESCs, 10T1/2 MSCs, and 3T3-L1 preadipocytes. (B) Schematic illustration of bivalent domains in ESCs (top). Even H3K27me3 disappear, numbers of developmental genes are still silent and poised for activation in lineage-committed progenitor cells (middle). In differentiated cells, bivalent domains are resolved to a monovalent mark for gene activation (bottom). (C and D) siRNA-mediated knockdown of SETDB1 promotes adipogenesis of 3T3-L1 preadipocytes (C) and 10T1/2 MSCs (D). ORO staining at day 8 and immunoblot analysis for 3T3-L1 preadipocytes ([C], left). (E) Genome-wide distribution of SETDB1 binding sites in 3T3-L1 preadipocytes. (F) ChIP-seq tag distribution of SETDB1, H3K9me3, and H3K4me3 near TSS in 3T3-L1 preadipocytes. (G) Venn diagrams representation of genes occupied by nucleosomes with H3K4me3 and H3K9me3 and occupied by SETDB1 within 2 kb from TSS in 3T3-L1 preadipocytes and GO analysis of these overlapping genes. (H) The sequence tag mapping shows SETDB1 binding and H3K9me3 on Cebpa and Pparg in preadipocytes but not in ESCs. Tracks 1, 2, 5, 6, 7, and 10 are the same data as Figure 1A. Molecular Cell 2015 60, 584-596DOI: (10.1016/j.molcel.2015.10.025) Copyright © 2015 Elsevier Inc. Terms and Conditions

Figure 2 SETDB1 Contributes to Formation of H3K4/H3K9me3 Bivalent Chromatin Domains in MSCs and Preadipocytes (A) Venn diagram representation of genes occupied by SETDB1 and nucleosomes with H3K4me3 and H3K9me3 and genes upregulated by SETDB1 knockdown (≥2-fold) in 3T3-L1 preadipocytes. (B and C) Upregulation of Cebpa and Pparg transcripts (B) and proteins (C) in SETDB1-knockdown preadipocytes. (D) siRNA mediated knockdown of C/EBPα reversed the effect of SETDB1 knockdown. ORO staining was performed at day 8 (left). Cebpa and Setdb1 transcripts were quantified by qPCR (right). (E) H3K4me3, H3K27me3, and H3K9me3 modification profiles in ESCs, 3T3-L1 preadipocytes, and adipocytes for Cebpa and Pparg genes. Tracks 1–5, 6–9, and 14 data are the same as Figure 1H. (F) Reduced H3K9me3 modifications on Cebpa gene in SETDB1-knockdown 3T3-L1 preadipocytes as assessed by ChIP-qPCR. See also Figure S4G. (G) Reduced SETDB1 recruitment and H3K9me3 modifications in differentiated adipocytes relative to preadipocytes as assessed by ChIP-qPCR. (H) SETDB1 protein levels during differentiation. Data are represented as mean ± SEM of at least three independent experiments. Student’s t test was performed for the indicated comparison in (G) (left), and ANOVA was performed followed by Tukey’s post hoc comparison in (B), (D), (F), and (G) (right). ∗p < 0.05; ∗∗p < 0.01. n.s., not significant. Molecular Cell 2015 60, 584-596DOI: (10.1016/j.molcel.2015.10.025) Copyright © 2015 Elsevier Inc. Terms and Conditions

Figure 3 Gene-Body H3K9me3 Prevents C/EBPβ Binding and H3K4me3 Deposition to Gene Body and Maintains Pol II in a Paused State (A) Genome browser representation for SETDB1, H3K4me3, H3K9me3, and Pol II on Cebpa and Pparg genes in 3T3-L1 preadipocytes or adipocytes and preadipocytes transfected with indicated siRNA. Tracks 1–3 and 11 are the same data as Figure 2E. (B) Increased C/EBPβ binding on Cebpa gene in SETDB1-knockdown 3T3-L1 preadipocytes as assessed by ChIP-qPCR. (C) Increased H3K27ac modification level in SETDB1 knockdown 3T3-L1 preadipocytes as assessed by ChIP-qPCR. (D) Hypothetical model for gene poising by H3K4/H3K9me3 bivalent chromatin domain. In ESCs, H3K4/H3K27me3 bivalent domain pauses Pol II at TSS on Cebpa gene (top). During adipocyte lineage commitment, developmental stage-specific gene body H3K9me3 modified by SETDB1 generates euchromatin/heterochromatin border to pause Pol II at TSS in MSCs and preadipocytes (middle). Loss of H3K9me3 in SETDB1 knockdown cells or adipocytes enables Pol II elongation (bottom). Data are represented as mean ± SEM of at least three independent experiments. Student’s t test was performed for the indicated comparison. ∗p < 0.05. n.s., not significant. Molecular Cell 2015 60, 584-596DOI: (10.1016/j.molcel.2015.10.025) Copyright © 2015 Elsevier Inc. Terms and Conditions

Figure 4 SETDB1 Recruitment Coincides with Lineage-Specific Gene-Body DNA Methylation (A) Sequence motif enriched in SETDB1 binding sites. (B) Genome wide distribution of m5CpG sites in 3T3-L1 preadipocytes. (C) Distribution of m5CpG near TSS in 3T3-L1 preadipocytes. (D) Venn diagrams showing number of sites within 2 kb from TSS overlapping with SETDB1-binding and m5CpG sites in 3T3-L1 preadipocytes. (E) Distribution of m5CpG sites and ChIP-seq tag of SETDB1 near TSS in ESCs. Data for preadipocytes are used in Figures 1F and 4C. (F) Genome browser shots of SETDB1 binding and m5CpG sites on Cebpa and Pparg genes in ESCs, MSCs, and 3T3-L1 preadipocytes. The arrow heads denote the regions of the sets of qPCR primers. See Figures S4K and S5A. Tracks 1, 3, 4, 6, 7, and 9 are the same as those in Figures 1H and S4D. Molecular Cell 2015 60, 584-596DOI: (10.1016/j.molcel.2015.10.025) Copyright © 2015 Elsevier Inc. Terms and Conditions

Figure 5 Gene-Body DNA Methylation Recruits SETDB1-MBD1-MCAF1 Complex (A) m5CpG, MBD1, and SETDB1 binding on Cebpa gene in 3T3-L1 preadipocytes. Tracks 1, 3, and 4 data are used in Figure 4F. (B) Venn diagram (left) and heatmap (right) representations of genes occupied by m5CpG, MBD1, and/or SETDB1. Heatmap analysis showing ChIP-seq data for m5CpG, MBD1, and SETDB1 at a 10-kb region centered over TSS. Sequence motifs enriched in SETDB1 binding sites (bottom). (C) ChIP-qPCR analyses of MBD1 and MCAF1 binding on Cebpa and Ppib in 3T3-L1 preadipocytes. (D) siRNA-mediated knockdown of MBD1 or MCAF1 promotes adipogenesis of 3T3-L1 preadipocytes. Cells were induced to differentiate by DEX alone. ORO staining at day 8 (top) and the mRNA levels of Cebpa quantified by qRT-PCR at day 0 (bottom). (E) ChIP-qPCR analysis of SETDB1 binding and H3K9me3 modification on the Cebpa and Ppib genes in 3T3-L1 preadipocytes transfected with indicated siRNA. (F) Genome browser representation showing SETDB1 binding on Cebpa in 3T3-L1 preadipocytes transfected with indicated siRNA. Each SETDB1 ChIP-seq was obtained by Illumina Genome Analyzer II in a single run. Track 4 and 5 are the same data as Figure 5A. (G) Microarray heatmap (left) depicting expression changes of genes co-localized SETDB1 with MBD1 and m5CpG in 3T3-L1 preadipocytes transfected with indicated siRNA. A fraction of genes co-localized with SETDB1, MBD1, and m5CpG were induced by knockdown of Setdb1, Mbd1, and Mcaf1. SETDB1 binding heatmap (right) depicting changes of SETDB1 binding scores on genes co-localized with MBD1 and m5CpG in MBD1 or MCAF1 knockdown 3T3-L1 preadipocytes. The numbers of SETDB1 binding sites co-localized with both MBD1 and m5CpG were severely reduced (less than 8-fold, shown in white color) by knockdown of either Mbd1 or Mcaf1. For reference, a color intensity scale is included at the right side of each heatmap. Data are represented as mean ± SEM of at least three independent experiments. Error bars in MCAF1 ChIP represent mean ± SEM of three technical replicates from the representative of three independent experiments. Student’s t test was performed for comparisons in (C), and ANOVA was performed followed by Tukey’s post hoc comparison in (D) and (E). ∗p < 0.05; ∗∗p < 0.01. n.s., not significant. Molecular Cell 2015 60, 584-596DOI: (10.1016/j.molcel.2015.10.025) Copyright © 2015 Elsevier Inc. Terms and Conditions

Figure 6 Both MBD1 and MCAF1 Are Required for SETDB1 Recruitment to Cebpa (A) 3T3-L1 preadipocytes transduced with empty vector, wild-type, or R22A mutant human MBD1 were treated indicated siRNAs, and Cebpa expression was quantified by qPCR (left). Immunoblot analysis using anti-MBD1 antibody (right). (B) 3T3-L1 preadipocytes transduced with empty control or Flag-tagged human MCAF1 were treated with indicated siRNAs and Cebpa expression was quantified by qPCR (left). Immunoblot analysis using anti-Flag antibody (right). (C and D) Retrovirally transduced 3T3-L1 preadipocytes treated with indicated siRNAs in (A) or (B) were analyzed for ChIP-qPCR using H3K9me3 or SETDB1 antibodies. H3K9me3 ChIP signals were presented as input percent (left). SETDB1 ChIP signal was presented as relative SETDB1 recruitment (right). (E) Model for formation of H3K4/H3K9me3 bivalent chromatin domain. In ESCs, H3K4/H3K27me3 bivalent domain poises expression of Cebpa gene (top). During lineage specification, lineage-specific gene-body DNA methylation recruits SETDB1-MBD1-MCAF1 complex to form H3K4/H3K9me3 bivalent chromatin domain to poise gene expression in MSCs and preadipocytes (middle). Loss of H3K9me3 by differentiation stimuli or SETDB1 knockdown enables, C/EBPβ binding, histone acetylation, Pol II elongation, and gene expression (bottom). The experiments were performed at least three times, and the representative one is shown. Error bars represent mean ± SEM of three technical replicates. Statistical test was performed for comparisons of groups (Tukey’s post hoc comparison or Student’s t test). ∗p < 0.05; ∗∗p < 0.01. n.s., not significant. Molecular Cell 2015 60, 584-596DOI: (10.1016/j.molcel.2015.10.025) Copyright © 2015 Elsevier Inc. Terms and Conditions