Cell culture

mESCs (E14Tg2A line) were kindly provided by Dr Fei Lan. mESCs were maintained in 90% Dulbecco's modified Eagle's medium (DMEM) (Life Technologies, Carlsbad), 10% fetal bovine serum (FBS) (Life Technologies, Carlsbad, CA, USA), 1% non-essential amino acid (NEAA), 1% penicillin and streptomycin (100× stock) (Gibco-Thermo Fisher, Waltham, MA, USA), 0.1 mM β-mercaptoethanol (Sigma-Aldrich, St. Louis, MO, USA), and 200 Units/ml murine leukemia inhibitory factor (LIF) (ESGRO; Merck-Millipore, Germany). mESC expansion was performed in gelatin-coated flasks (BBI Life Sciences, China). hESCs of H7 (from the WiCell Research Institute) were cultured on hESC-qualified Matrigel (BD Biosciences, San Diego, CA, USA) supplemented with mTeSR1 (Stemcell Technologies, Vancouver). Human embryonic kidney (HEK) 293T cells were originally from the American Type Culture Collection and were cultured in DMEM with 10% FBS.

Construction of vectors

Mouse Bach1 gene was generated and cloned into pcDNA3.1. Briefly, the full-length cDNA of Mouse Bach1 or DNA binding-defective cDNA of mouse Bach1 [Bach1-ΔbZIP (1–553 aa)] was generated by PCR and verified by DNA sequencing. HA-epitope tags were introduced to the C termini of the Bach1 gene. Glutathione S-transferase (GST)-BACH1, GST-BACH1-ΔbZIP or GST-BACH1-ΔBTB fusion protein was constructed by inserting PCR-generated DNA fragments encoding regions of full-length BACH1 (1–739 aa) or BACH1-ΔbZIP (1–553 aa) or BACH1-ΔBTB (128–739 aa) into PGEX-6P-1 vector. HIS-tagged NANOG was constructed by inserting PCR-generated DNA fragments encoding Nanog into PET-28A vector. Primers for cloning mouse Bach1 or Nanog cDNA are listed in Supplemental Table S2. All constructs were verified by sequencing.

Generation of Bach1-KO and DoxBach1 mESCs

Bach1-KO mESCs were generated by using CRISPR/Cas9 genome editing technology. sgRNAs (Supplemental Table S2) targeting the genomic regions of interest were designed with the CRISPR Design Tool (http://crispr.mit.edu/) and synthesized by Shanghai Sunny Biotechnology Co. Annealed sgRNA oligos were cloned into LentiCRISPR V2 (Addgene, Watertown, MA, USA), transiently transfected into mESCs with Lipofectamine 2000 (Thermo Fisher, Waltham, MA), and then selected with puromycin (Selleck Chemicals, Houston, TX, USA); single-cell clones were sorted 7 days later. Bach1-FLAG cDNA was cloned into the vector (pPB-TRE-EGFP-PGK-Neo). DoxBach1 mESCs were generated by using the piggyBac (PB) transposon system with the pCAG-T7-mPB, pCAG-rtTA-Puro and pPB-TRE-BACH1-PGK-Neo vectors (1:2:2 ratio) and puromycin selection (Selleck Chemicals, Houston, TX, USA). BACH1 expression was induced by treating DoxBach1 mESCs with 0.5 μg/μl doxycycline (Selleck Chemicals, Houston, TX, USA).

Viral transduction of mESCs

Lentiviral particles of Bach1 were generated by transfecting HEK293T cells with LvEP05- Bach1-FLAG vector and packing vectors psPAX2 and pMD2.G. mESCs were transduced with the concentrated lentivirus for 48 h.

RNA interference

mESCs were transfected with Bach1 siRNAs or control siRNAs by Lipofectamine 2000. The transfection was performed according to the instructions. siRNA sequences are listed in Supplementary Table S2.

Mouse embryonic fibroblast (MEF) reprogramming

Mouse embryonic fibroblast (MEF) reprogramming was performed as described previously (22). MEFs were isolated from 13.5 d.p.c. mouse embryos and cultured in DMEM containing 10% FBS. Lentiviral plasmids for the reverse tetracycline transactivator (FUW-M2rtTA, Addgene, Watertown, MA, USA) and the tetracycline-inducible expression of mouse OCT4, SOX2, Klf4 and Myc (TetO-FUW-OSKM, Addgene, Watertown, MA, USA) were packaged in HEK293T cells with psPAX2 and PMD2.G, respectively. MEFs (1 × 10⁴ cells/well) were plated into six-well plates and cultured for 24 h; then, concentrated lentivirus solutions containing mouse Yamanaka factors and Bach1 were added, and the cells were cultured for another 24 h before addition of the reprogramming medium (Knockout DMEM, 10% Knockout Serum Replacement [Thermo Fisher, Waltham, MA, USA], 1% NEAA, 1% PS, 0.01% CHIR98014, 0.01% PD0325901 and 1 × 10³ units/ml mouse LIF).

Alkaline phosphatase staining and colony assay

The alkaline phosphatase assay was conducted with an Alkaline Phosphatase Detection Kit (Sigma-Aldrich, St. Louis, MO, USA) as directed by the manufacturer's instructions. Cells (1 × 10⁴ cells/well) were cultured in 12-well plates for 2–3 days, fixed with 4% paraformaldehyde in phosphate-buffered saline (PBS) for 1–2 min, and then rinsed with TBST solution (20 mM Tris–HCl, pH 7.4, 0.15 M NaCl, 0.05% Tween 20). FRV-Alkaline solution (30 μl) and sodium nitrite solution (30 μl) were mixed for 2 min at room temperature and then added with naphthol AS-BI Alkaline solution (30 μl) into water (1350 μl); then, the mixture (1 ml/well) was added to the cells, and the cells were incubated in the dark at room temperature for 15 min, rinsed with TBST and covered with PBS. Colonies were viewed under a microscope. Colonies containing undifferentiated (i.e. alkaline phosphatase-positive, red) cells, differentiated (colorless) cells and mixed populations of cells were counted.

Cell proliferation assay

Cells (1.3 × 10⁴ cells/well) were seeded into gelatin-coated 12-well plates, cultured for the indicated periods, dissociated with 0.05% trypsin (Gibco-Thermo Fisher, Waltham, MA, USA), and then counted with a cell-counting chamber.

Embryoid body (EB) formation assay

mESCs were trypsinized and suspended in LIF-free EB medium at 1 × 10⁵ cells per well and plated onto six-well low-attachment plates. The EB culture medium contains DMEM supplemented with 10% FBS, 1 mM l-glutamine, 1% NEAA, 0.1 mM β-mercaptoethanol. The culture medium was changed every 2 days. EBs were cultured with EB medium until the indicated time points.

Quantitative real-time PCR and RNA- sequencing

Quantitative real-time PCR and RNA- sequencing were performed as described previously (19). Total RNA was isolated using Trizol reagent, and complementary DNA (cDNA) was synthesized with the ReverTra Ace qPCR RT Kit (TOYOBO, Osaka, Japan, #FSQ-101). Real-time quantitative PCR was performed with the qPCR SYBR^® Green Master Mix (Yeason, Shanghai, China), and the primer sequences are listed in Supplementary Table S2. The expression of each gene was analyzed by the Ct method and was normalized to the expression of Actb. Genome-wide gene expression analyses were performed using RNA-sequencing (RNA-seq). Non-ribosomal RNA was isolated from 1 μg total RNA by using a TrueLib Poly (A) mRNA Magnetic Isolation Module (Excell Bio, Shanghai, China), and sequencing libraries were prepared with the TrueLib mRNA Library Prep Kit for Illumina (ExCell Bio, Shanghai, China). Sequencing was performed on the NovaSeq 6000 and Hiseq XTen sequencer (Illumina Inc, San Diego, CA, USA) by Annoroad Company (Beijing, China) and Basepair Company (Suzhou, China).

Analysis of RNA-sequencing data

The FASTQ data were trimmed adaptor by trim_galore (bioinformatics.babraham.ac.uk), then mapped to mouse genome mm9 using HISAT2 (23), and assembled by HOMER(24) to get Transcripts per million (TPM) for each gene. The TPM value of genes was used to eliminate the statistical biases inherent in the RPKM measure (25). Differentially expressed genes were generated by limma (26) with fold change >2.0 and P <0.05. Gene Ontology analysis (27), KEGG Pathway analysis (28), and volcano plot were performed on www.omicsbean.cn.

Immunoprecipitation (IP) and western blotting

Cells were lysed in lysis buffer. For immunoprecipitation, cell lysates were incubated with the indicated antibodies and protein A/G plus agarose at 4°C overnight. After washing four times with PBS, protein SDS loading buffer was added and boiled for 5 min. Proteins were resolved on an SDS-PAGE gel, transferred onto a polyvinylidene fluoride (PVDF) membrane (Merck-Millipore, Germany), labeled with their respective antibodies. Blots were incubated with secondary antibodies and developed with ECL (Thermo Fisher Scientific, Waltham, MA, USA), and signals were detected via chemiluminescence on a Tanon-5500 Imaging System (Tanon Science & Technology Ltd, Shanghai, China). The intensity of the bands was measured via densitometric analysis with ImageJ software.

Immunofluorescence staining

Cells were fixed, permeabilized, incubated, stained and photographed as described previously (16). Briefly, cells were incubated with primary antibodies in blocking solution overnight and then with Cy3-conjugated donkey anti-rabbit/mouse IgG or FITC-conjugated donkey anti-goat secondary antibodies for 1 h (Jackson ImmunoResearch Laboratories, West Grove, PA, USA). Nuclei were counter-stained with DAPI, and then the cells were viewed under a fluorescence microscope (Olympus, Tokyo, Japan).

In vitro pull-down assay

The GST-tagged BACH1 or BACH1-ΔbZIP or BACH1-ΔBTB and HIS-tagged NANOG recombinant proteins or SETD1A complex recombinant proteins were expressed and purified with glutathione-Sepharose 4B beads. Purified GST-tagged recombinant proteins BACH1 or BACH1-ΔbZIP or BACH1-ΔBTB and purified HIS-tagged NANOG or SETD1A complex recombinant proteins were then incubated together for 2 h at 4°C, followed by three times of wash. Conjugated proteins were then eluted by SDS loading buffer, and examined by western blot analysis.

Chromatin immunoprecipitation (ChIP) and ChIP-seq

Chromatin immunoprecipitation (ChIP) and ChIP-seq were performed as described previously (29). Cells were cross-linked with 1% formaldehyde for 10 min, quenched with 0.125 M glycine for 10 min; then, the cell pellets were resuspended in high salt lysis buffer (50 mM HEPES/KOH, pH 7.5, 500 mM NaCl, 1% Trixon-X 100, 0.05% SDS, 10 mM EDTA and proteinase inhibitors), and sonicated with a Bioruptor (Diagenode) for 20 min in 30-s intervals to shear the chromatin into 200- to 400-bp lengths. Chromatin was immunoprecipitated with the indicated antibodies and Dynabeads protein A and protein G (Thermo Fisher, Waltham, MA, USA). For, BACH1-FLAG ChIP, DoxBach1 mESCs were treated with 0.5 μg/μl doxycycline (Selleck Chemicals, Houston, TX, USA) for 6 h. Then cells were cross-linked, quenched, and washed as described above. Cell nuclei were extracted before sonication. Furthermore, the sonicated chromatin was incubated with anti-FLAG affinity M2 agarose gel (Sigma-Aldrich, St. Louis, MO, USA) overnight at 4°C. Then beads were washed with high salt lysis buffer and low salt buffer (10 mM Tris–HCl pH 8.0, 250 mM LiCl, 1 mM EDTA, 0.5% NP40, 0.5% Na-Deoxycholate), and DNA was eluted in elution buffer (50 mM Tris–HCl pH 8.0, 10 mM EDTA, 1% SDS) with Proteinase K (Invitrogen-Thermo Fisher, Waltham, MA, USA) and RNase A (Thermo Fisher, Waltham, MA, USA) in 65°C for 4 h. Chromatin occupancy was quantified using qPCR and normalized to input DNA. The primers used in ChIP-qPCR assays are listed in Supplemental Table S2. For ChIP-seq experiments, ChIP DNA libraries and input DNA samples were prepared as directed by the protocol of KAPA Hyper Prep Kit for Illumina (Kapa Biosystems, Wilmington, MA, USA) and VAHTS Universal DNA Library Prep Kit for Illumina (Vazyme Biotech, Nanjing, China). Libraries were sequenced on Illumina Hiseq X Ten by Basepair Company (Suzhou, China).

Analysis of ChIP-sequencing data

The FASTQ data were trimmed adaptor by trim_galore (bioinformatics.babraham.ac.uk), and then they were mapped to mouse genome mm9 with Bowtie2 (30) in pair-end mode. Peaks were called by MACS1.4 (31) and MACS2 (32). Differential peaks were called by MACS2 -bdgdiff. ChIP-seq motif and ChIP-seq visualization were performed by HOMER tools (24) and visualized on UCSC genome browser. Heatmaps and signal plots were generated by deepTools (33) (3.3.0).

Chromatin loops and super-enhancers

Hi-C sequencing in mESCs was performed by Bonev et al. (34). Chromatin loops were generated from the 3D epigenome browser (35), and the loops were predicted by Peakachu (36). Chromatin loops combined with ChIP-seq data were visualized in WashU epigenome browser. Super-enhancers were defined by Benjamin Sabari, Alessandra Dall’Agnese et al. (37), in which ChIP-seq peaks of MED1 were used as input for ROSE to call super-enhancers.

Statistical analysis

Data are expressed as mean±SEM. Differences between two groups were evaluated for significance via two-tailed Student's t-test, and differences among three or more groups were evaluated for significance via one-way analysis of variance (ANOVA) followed by Tukey's multiple comparison test. Analyses were conducted with Prism software, and P <0.05 was considered significant.

BACH1 is required for the maintenance of mESC pluripotency

BACH1 colocalized with OCT4 in the inner cell mass of mouse blastocysts (Supplemental Figure S1A), and both BACH1 and OCT4 expression declined at day 6 and day 8 during embryoid body (EB) differentiation (Supplemental Figure S1B). Thus, we began to investigate whether BACH1 contributed to the maintenance of mESC pluripotency by generating Bach1-knockout (Bach1-KO) mESCs and doxycycline-inducible BACH1 overexpressing mESCs (DoxBach1 mESCs) (Supplemental Figure S2). Bach1 knockout induced dramatic morphological changes, as well as declines in alkaline phosphatase (AP) activity (Supplemental Figure S1C) and proliferation (Supplemental Figure S1D) in mESCs, and a greater proportion of Bach1-KO than WT mESC colonies were composed of primarily differentiated or mixed populations of cells (Supplemental Figure S1E). The expression level of pluripotency genes (SOX2, OCT4, NANOG, KLF4, stage-specific embryonic antigen 1 [SSEA1]) was also lower in Bach1-KO mESCs than in WT mESCs (Supplemental Figure S1F–G, Supplemental Figure S3A). When spontaneous differentiation was induced in mESCs via the withdrawal of leukemia inhibitory factor (LIF) for three days, Bach1-KO cells showed significantly higher levels of many differentiation genes, such as mesodermal (Brachyury [T], mesoderm posterior 1 [MESP1], MIXL1) and endodermal (GATA4, GATA6, SOX17) markers, but modestly lower expression levels of neuroectodermal-marker (SOX1 and NESTIN) (Supplemental Figure S3B). Furthermore, the effect of the Bach1-KO on pluripotency gene expression in mESCs could also be observed by transfecting the cells with Bach1 siRNA (Supplemental Figure S3C); whereas dox-induced BACH1 overexpression increased measures of AP activity (Supplemental Figure S3D) and the pluripotency gene expression (Supplemental Figure S3E); lentiviral-induced BACH1 expression largely reversed the effect of Bach1 knockout on mESC morphology (Supplemental Figure S1C) and pluripotency gene expression (Supplemental Figure S1G), and mesendodermal/neuroectodermal gene expression during spontaneous mESC differentiation (Supplemental Figure S3B).

RNA sequencing (RNA-seq) analysis indicated that 1883 genes were upregulated, and 2523 genes were downregulated in Bach1-KO mESCs than in WT mESCs (fold change > 2, P < 0.05) (Supplemental Figure S4A, Supplemental Table S1). Many of the downregulated genes were associated with signaling pathways regulating pluripotency (Supplemental Figure S4B). RT-qPCR assays further confirmed the decreased level of pluripotency genes (Supplemental Figure S1G). The upregulated genes were enriched for PI3K-AKT signaling and mitogen-activated protein kinase (MAPK) signaling pathways (Supplemental Figure S4C). Moreover, mouse embryonic fibroblasts (MEFs) transduced with lentivirus containing the BACH1 coding sequence together with four Yamanaka reprogramming factors had a significantly increased ability to form AP-positive mESC-like colonies (Supplemental Figure S1H–I), further demonstrating BACH1’s function in promoting pluripotency.

BACH1 interacts with NANOG, OCT4, SOX2 and MLL/SET1 complexes

Co-immunoprecipitation (co-IP) in mESCs (Figure 1A) and hESCs (Supplemental Figure S5A) indicated that BACH1 interacted with several components of MLL/SET1 complexes, including MLL1, WDR5, RBBP5, ASH2L and SETD1A. Consistent with our previous report in hESCs, BACH1 also interacted with OCT4, SOX2, and NANOG (OSN) in mESCs (Figure 1A and Supplemental Figure S5B–D). BACH1 colocalized with MLL/SET1 complexes subunits, RBBP5 and SETD1A, as well as OSN in the nucleus of mESCs (Supplemental Figure S5E). To elucidate whether the interaction of BACH1 and MLL/SET1 complexes depends on OSN, we conducted a co-IP assay in HEK293T cells that expressed FLAG-tagged BACH1. In the absence of OSN, exogenous BACH1 still interacted with MLL/SET1 complexes (Figure 1B). To further verify the direct protein-protein interaction, recombinant BACH1 protein was purified from bacteria and incubated with SET1 complexes or recombinant HIS-tagged NANOG protein. In vitro pulldown assays confirmed that GST-tagged BACH1 interacted directly with RBBP5, ASH2L, SETD1A or NANOG (Figure 1C–E). NANOG interacted directly with GST-tagged full-length BACH1 but not with BACH1 lacking BTB domain (BACH1-ΔBTB) (Figure 1D). SETD1A coimmunoprecipitated with the full-length BACH1 but not with BACH1-ΔBTB or bZIP domain (BACH1-ΔbZIP) (Figure 1E). To elucidate whether BACH1 affects the interaction between OSN and MLL/SET1 complexes, we compared the interaction between NANOG and RBBP5/SETD1A in WT and Bach1-KO mESCs. Notably, loss of BACH1 reduced the interaction between NANOG and RBBP5/SETD1A when a similar amount of NANOG protein was pulled down (Figure 1F). These findings indicate that BACH1 interacts with MLL/SET1 complexes and NANOG directly and individually, and that BACH1 bridges the interaction between NANOG and MLL/SET1 complexes. Moreover, we observed that only WT-BACH1 but not BACH1-ΔBTB or BACH1-ΔbZIP can rescue the effect of Bach1 knockout on mESC morphology and proportion of undifferentiated cells (Figures 1G and 6D).

Figure 1.

BACH1 interacts with NANOG, OCT4, SOX2 and MLL/SET1 complexes. (A) Co-immunoprecipitation of BACH1, MLL1, WDR5, RBBP5, ASH2L, SETD1A, NANOG, SOX2 and OCT4 in WT mESCs. (B) Co-immunoprecipitation of WDR5, RBBP5, ASH2L, SETD1A, BACH1-FLAG in BACH1-FLAG overexpression HEK293T cells. (C) Bacteria expressed GST-tagged BACH1 was incubated with SETD1A complex (SETD1A, ASH2L, RBBP5) recombinant proteins, and then GST-pulldown assay and immunoblot detected BACH1-GST and SETD1A complex (SETD1A, ASH2L, RBBP5). (D–E) Bacteria expressed NANOG-HIS (D) and/or HEK293T cells expressed SETD1A-FLAG (E) was incubated with GST-tagged full-length BACH1 protein (GST-BACH1-FULL) or GST-tagged BACH1 lacked bZIP domain (GST-BACH1-ΔbZIP) or BTB domain (GST-BACH1-ΔBTB); GST-pulldown assay and immunoblot detected the interaction between SETD1A-FLAG or NANOG-HIS and GST-BACH1-Full, GST-BACH1-ΔbZIP, or GST-BACH1-ΔBTB. (F) Co-immunoprecipitation of RBBP5, SETD1A and NANOG in WT and Bach1-KO mESCs. (G) Alkaline phosphatase (AP) staining of colonies (upper), the proportion of total cell clones in undifferentiated, partially differentiated, and fully differentiated cells; and BACH1 protein level in WT and Bach1-KO mESCs transfected with BACH1 FULL or BACH1-ΔBTB or empty plasmid (n = 3; ** P < 0.01 versus WT, ^##P < 0.01 versus KO, ΔΔP < 0.01 versus KO + BACH1) (lower).

Colocalization of BACH1 and NANOG at gene promoters and enhancers in mESCs.

Chromatin immunoprecipitation and sequencing (ChIP-seq) analyses of mESCs expressing a dox-inducible, FLAG-tagged version of BACH1 confirmed that BACH1 preferentially bound to active gene promoters and active enhancers (Figure 2A), and KEGG pathway analysis indicated that BACH1-bound genes are enriched in the pathways of proteasome and pluripotency (Supplemental Figure S6A). Motif analysis showed that BACH1 binding sites were enriched not only in the binding motif of itself, but also in the motif of pluripotency factors such as KLF4, NANOG, OCT4, SOX2 and ESRRB (Figure 2B), suggesting their colocalization. Notably, the ChIP-seq binding heatmaps showed that NANOG, ASH2L, H3K4me1, H3K4me3 and H3K27ac, but not H3K27me3, were highly co-occupied on BACH1 binding active enhancers and promoters (Figure 2C). Venn diagram showed that 1,909 genes were co-occupied by BACH1, NANOG, H3K4me1 and H3K4me3 (Figure 2D). Representative snapshots of ChIP-seq tracks show the colocalization of BACH1 and NANOG on the H3K4me1⁺ H3K27ac⁺ active enhancers neighboring Tns3 and Lefty2, and on the H3K4me3⁺ H3K27ac⁺ active promoter of Jarid2 (Figure 2E).

Figure 2.

Colocalization of BACH1 and NANOG on gene promoters and enhancers in mESCs. (A) BACH1-FLAG ChIP-seq data in mESCs were analyzed to identify the genomic distribution of BACH1 binding sites in genomic DNA. Promoter: ±2 kb of transcription start sites (TSS). Active enhancer: H3K4me1⁺; H3K27ac⁺ promoter negative distal region. Primed enhancer: H3K4me1⁺; H3K27ac⁻ promoter negative distal region. (B) The known transcription factors’ motifs at BACH1-FLAG enriched regions. (C) Heatmaps showing the genome-wide ChIP-seq binding profiles of BACH1-FLAG, ASH2L-GFP, NANOG, H3K4me1, H3K4me3, H3K27ac and H3K27me3 at BACH1-FLAG enriched regions on active enhancers (BACH1-FLAG⁺ H3K4me1⁺ H3K27ac⁺ distal regions) and BACH1-FLAG enriched regions on promoters (BACH1-FLAG⁺ regions within TSS±2 kb) (region center ±10 kb). (D) Venn diagram of genes enriched by BACH1-FLAG, NANOG, H3K4me1 and H3K4me3. (E) Representative snapshots of ChIP-seq tracks for BACH1-FLAG, H3K4me1, H3K4me3, H3K27ac, H3K27me3 and NANOG on the enhancers neighboring Tns3 and Lefty2 and on the promoter of Jarid2.

Next, we performed ChIP-seq analysis to evaluate the binding of NANOG in WT and Bach1-KO mESCs. Loss of BACH1 reduced the binding of NANOG at 2,365 regions (Supplemental Figure S6B–C, cluster 1), while 1,616 regions de novo built NANOG occupancy (Supplemental Figure S6B-C, cluster 3). The reason could be that knockout of Bach1 released NANOG from BACH1 recruited locus and rearranged it into novel locations. Furthermore, we found that NANOG enrichment was decreased in 1,361 genes in Bach1-KO mESCs, and these genes were involved in the development process and signaling pathways regulating pluripotency of stemness (Supplemental Figure S6D and E), consistent with our RNA-seq analyses.

Loss of BACH1 reduces the recruitment of NANOG and SETD1A to the targeting promoters and enhancers

We next overlapped ChIP-seq data of BACH1-binding genes with RNA-seq data from Bach1-KO versus WT mESCs. We identified 926 BACH1-targeting genes, which were directly bound by BACH1, and Bach1-KO led to their expression changes (Figure 3A). Among these 926 genes, we found that 538 genes were down-regulated and 388 genes were up-regulated upon Bach1 knockout (Figure 3A and B). Since BACH1 interacted with NANOG, MLL/SET1 complexes and Bach1-KO reduced their interaction (Figure 1A–F), we then asked whether loss of BACH1 affected their recruitment to chromatin. Indeed, we found that the chromatin occupancies of NANOG, RBBP5, and SETD1A near the BACH1-binding sites were lower in Bach1-KO than in WT mESCs (Figure 3C and D), especially on the enhancer and promoter of pluripotency-regulating genes.

Figure 3.

BACH1 recruits NANOG, RBBP5 and SETD1A to the BACH1-targeting enhancers and gene promoters, and maintains the tri-methylated state of H3K4. (A) Venn diagram showing the corresponding numbers of BACH1-bound genes analyzed from ChIP-seq, genes whose expression was affected by Bach1 knockout analyzed from RNA-seq. A total of 926 genes in the intersection were identified as BACH1-targeting genes, including 388 upregulated genes and 538 downregulated genes. (B) Box plot showing the expression of 926 BACH1-targeting genes and housekeeping genes in WT and Bach1-KO mESCs. (C) Representative snapshots of ChIP-seq tracks for BACH1-FLAG and NANOG in WT and Bach1-KO mESCs at the Cyp24a1 and Spic locus. (D) ChIP-qPCR analysis to validate the reduced recruitment of NANOG, RBBP5 and SETD1A in Bach1-KO mESCs vs WT mESCs on BACH1 target enhancers and gene promoters (n = 3; *P < 0.05, ** P < 0.01 versus WT). (E) Heatmaps of BACH1-FLAG, and H3K4me1, H3K4me3 and H3K27ac at BACH1 enriched regions on active enhancers or promoters in WT and Bach1-KO mESCs. Each horizontal line shows a separate BACH1-FLAG-bound peak (peak center ±10 kb). (F) ChIP-qPCR analysis to validate the increase of H3K4me1 and decrease of H3K4me3 in Bach1-KO mESCs vs WT mESCs on BACH1-targeting enhancers and gene promoters (n = 3; *P <0.05, **P < 0.01 versus WT). (G) RT-qPCR analysis of the genes expression in WT and Bach1-KO mESCs. Results were normalized to measurements in WT mESCs (n = 3; *P < 0.05, ** P < 0.01 versus WT).

Loss of BACH1 leads to a switch of H3K4me3 and H3K4me1 at both typical and super-enhancers

Since MLL/SET1 complexes catalyze the trimethylation of histone H3 lysine 4 (H3K4me3), the decrease of their chromatin localization may lead to changes on histone modifications. Although the global levels of histone modifications H3K4me1, H3K4me3, H3K27me3 and H3K27ac did not significantly change between WT and Bach1-KO mESCs (Supplemental Figure S7A), upon loss of Bach1 H3K4me1 enrichment increased, while H3K4me3 enrichment declined at BACH1 binding active enhancers and promoters, and H3K27ac enrichment had modest decrease on BACH1 binding active enhancers but largely unchanged (Figure 3E, Supplemental Figure S7B and C). These observations were validated by ChIP-qPCR at the promoters and enhancers of BACH1-targeting genes (Zfp710, Cyp24a1, Rybp, Spic and Tns3), especially stemness-related genes such as Nanog, Zfp42 and Lif (Figure 3F). Importantly, the expression level of these genes decreased in Bach1-KO mESCs (Figures 3G and 5D).

Notably, we found BACH1 was strongly enriched at super-enhancers in mESCs (called by Benjamin Sabari, Alessandra Dall’Agnese, et al. (37), through MED1 ChIP-seq in mESCs) (Figure 4A and B), and loss of BACH1 led to H3K4me3 partially switched to H3K4me1 at the super-enhancers (Figure 4C–E), consistent with the decreased chromatin occupancy of RBBP5 and SETD1A (Figure 3D). H3K4me3 partially switching to H3K4me1 suggested a reduction of super-enhancer activity. These BACH1 bound super-enhancers were neighboring the genes regulating embryo development and gene expression (Figure 4F), indicating their prominent role in regulating pluripotency. Collectively, these observations suggest that BACH1 is required for full level chromatin binding of NANOG and MLL/SET1 complexes in mESCs, and is important for the maintenance of H3K4me3 and activity of target promoters and (super-) enhancers to promote downstream gene expression, especially on pluripotency-regulating genes.

Figure 4.

BACH1 localizes at super-enhancers, and loss of BACH1 leads to H3K4me3 partially switched to H3K4me1 at super-enhancers. (A) Venn diagram of BACH1-FLAG enriched peaks and super-enhancers in mESCs. (B) Signal profiles of BACH1 and input at super-enhancers in mESCs. (C) Signal profiles of H3K4me1 (left) and H3K4me3 (right) in WT and Bach1-KO mESCs at super-enhancers. (D, E) Representative snapshots of H3K4me3 and H3K4me1 in WT and Bach1-KO mESCs at BACH1 bound super-enhancers. (F) The Gene Ontology analysis of genes neighboring BACH1 bound super-enhancers in mESCs (left) and the representative genes (right).

BACH1 and NANOG strongly co-bind at chromatin loop anchors, and BACH1 regulates the remote NANOG binding

In ChIP-seq tracks, we found that Bach1 knockout led to the decrease of NANOG binding not only at peaks colocalized with BACH1 (direct recruitment), but also at peaks close to BACH1 binding regions, possibly through the effect mediated by chromatin looping (Figure 5A and B, Supplemental Figure S7D). In order to test this hypothesis, we used published HiC data in mESCs (34) and found that these regions indeed formed chromatin loops with the BACH1 bound sites (Figure 5A and B, Supplemental Figure S7D). Loss of BACH1 led to the decrease of enhancer RNAs (eRNAs) at these chromatin loop anchors (Figure 5C), and reduced the expression of neighboring genes including stemness-regulating genes Lif and Rybp (Figure 5D), indicating the reduction of enhancer activity. Interestingly, we also found that NANOG occupancy was also higher in chromatin loop anchors which contained BACH1 (Figure 5E). These data suggested that BACH1 can also regulate the recruitment of NANOG on chromatin interaction sites, and further fine-tune the enhancer activity and gene expression, including stemness-regulating genes.

Figure 5.

BACH1 and NANOG strongly co-bind at chromatin loop anchors, and BACH1 regulates the remote NANOG binding and enhancer activity. (A, B) Representative snapshots of chromatin loops and ChIP-seq tracks for BACH1-FLAG and NANOG in WT and Bach1-KO mESCs. Genome regions to be explained were highlighted. Chromatin loops were generated by Hi-C sequencing in mESCs and predicted by Peakachu. (C) RT-qPCR analysis of enhancer RNAs (eRNAs) transcribed from highlighted chromatin loop anchors in (A, B and Supplemental Figure S7D) in WT and Bach1-KO mESCs. Normalized by internal control gene Actb and normalized to WT mESCs (n = 3; *P < 0.05, **P < 0.01 versus WT). (D) RT-qPCR analysis of expression of genes regulated by BACH1 and chromatin looping in (A, B) in WT and Bach1-KO mESCs. Normalized by internal control gene Actb and normalized to WT mESCs (n = 3; *P < 0.05, ** P < 0.01 versus WT). (E) Signal profiles of NANOG ChIP-seq signal at NANOG peaks within BACH1-positive or BACH1-negative chromatin loop anchors. BACH1⁺ loops: either one anchor of a chromatin loop contained BACH1; BACH1^– loops: none of the chromatin loop anchors contained BACH1.

BACH1 DNA binding domain is crucial for the recruitment of NANOG at chromatin and the enhancer activity

To determine if BACH1 directly affects NANOG binding to chromatin rather than merely reducing its expression, exogenously FLAG-tagged NANOG proteins were expressed at similar levels in WT and Bach1-KO mESCs. We found that the chromatin occupancy of exogenous FLAG-tagged NANOG was lower in Bach1-KO than in WT mESCs (Figure 6A). Moreover, only WT-BACH1 but not DNA binding-defective BACH1 (BACH1-ΔbZIP) rescued the decrease of the occupancy of NANOG, the level of enhancer RNAs (eRNAs) at the chromatin loop anchors, and the proportion of undifferentiated cells in Bach1-KO mESCs (Figure 6B–D), although BACH1-ΔbZIP could restore the NANOG expression in Bach1-KO mESCs as WT-BACH1 (Figure 6B, lower). Collectively, these data demonstrated that BACH1 directly affects the recruitment of NANOG at chromatin and maintains its enhancer activity, and the DNA binding domain of BACH1 is required.

Figure 6.

BACH1 DNA binding domain is crucial for the recruitment of NANOG at chromatin and the enhancer activity. (A) ChIP-qPCR analysis of exogenous FLAG-tagged NANOG in Bach1-KO mESCs overexpressing NANOG-FLAG vs WT mESCs overexpressing NANOG-FLAG on BACH1 target chromatin loop anchors (n = 3; *P < 0.05, ** P < 0.01 versus WT) (upper). Immunoblot verified the expression of exogenous FLAG-tagged NANOG and BACH1 (lower). (B) ChIP-qPCR analysis of endogenous NANOG in Bach1-KO mESCs rescued with full-length BACH1, BACH1-ΔbZIP or empty plasmid versus WT mESCs on BACH1 target chromatin loop anchors (n = 3; *P < 0.05, **P < 0.01 versus WT, ^#P <0.05, ^##P < 0.01 versus Bach1-KO) (upper). Immunoblot verified the expression of exogenous NANOG and rescued exogenous HA-tagged BACH1 (lower). (C) RT-qPCR analysis of enhancer RNAs (eRNAs) transcribed from highlighted chromatin loop anchors in (Figure 5A, B and Supplemental Figure S7D) in WT and Bach1-KO mESCs transfected with full-length BACH1, BACH1-ΔbZIP or empty plasmid. Normalized by internal control gene Actb and normalized to WT mESCs (n = 3; *P < 0.05, **P < 0.01 versus WT, ^#P < 0.05, ^##P < 0.01 versus Bach1-KO). (D) Alkaline phosphatase (AP) staining of colonies (upper), and the proportion of total cell clones of undifferentiated, partially differentiated and fully differentiated cells and BACH1 protein level in WT and Bach1-KO mESCs transfected with full-length BACH1, BACH1-ΔbZIP or empty plasmid (n = 3; ** P < 0.01 versus WT, ^##P < 0.01 versus KO, ΔΔP < 0.01 versus KO + BACH1) (lower). (E) Schematic review showing BACH1 recruits NANOG and histone H3 lysine 4 methyltransferase MLL/SET1 complexes to activate enhancer–promoters and maintains pluripotency in embryonic stem cells.