The ZBTB24-CDCA7-HELLS axis suppresses the totipotent 2C-like reprogramming by maintaining Dux methylation and repression

Abstract

Two-cell-like cells (2CLCs), a rare population (∼0.5%) in mouse embryonic stem cell (mESC) cultures, are in a transient totipotent-like state resembling that of 2C-stage embryos, and their discovery and characterization have greatly facilitated the study of early developmental events, such as zygotic genome activation. However, the molecular determinants governing 2C-like reprogramming remain to be elucidated. Here, we show that ZBTB24, CDCA7, and HELLS, components of a molecular pathway that is involved in the pathogenesis of immunodeficiency, centromeric instability, and facial anomalies (ICF) syndrome, function as negative regulators of 2C-like reprogramming by maintaining DNA methylation of the Dux cluster, a master inducer of the 2C-like state. Disruption of the ZBTB24-CDCA7-HELLS axis results in Dux hypomethylation and derepression, leading to dramatic upregulation of 2C-specific genes, which can be reversed by site-specific re-methylation in the Dux promoter. We also provide evidence that CDCA7 is enriched at the Dux cluster and recruits the CDCA7–HELLS chromatin remodeling complex to constitutive heterochromatin. Our study uncovers a key role for the ZBTB24-CDCA7-HELLS axis in safeguarding the mESC state by suppressing the 2C-like reprogramming.

Graphical Abstract

Graphical Abstract.

Introduction

Embryonic stem cells (ESCs), derived from the inner cell mass of blastocysts, are pluripotent, with the capacity to differentiate to all three germ layers of the embryo [1, 2]. Intriguingly, mouse ESCs (mESCs) sporadically transition to a transient totipotent-like state resembling that of blastomeres of two-cell (2C) stage embryos, capable of contributing to both embryonic and extraembryonic tissues [3, 4]. These 2C-like cells (2CLCs), a rare population (∼0.5%) in mESC cultures at any given time, are characterized by the burst expression of a subset of 2C-specific transcripts associated with zygotic genome activation (ZGA), including the Dux cluster, Zscan4 cluster, Usp17-like and Eif1α-like family members, as well as several families of endogenous transposable elements (TEs), most notably murine endogenous retrovirus with leucine transfer RNA primer (MERVL or MuERV-L) [3]. Additionally, 2CLCs share many molecular, epigenetic, and metabolic features of 2C embryos, including downregulation of pluripotency markers, slow DNA replication fork speed, decreased glycolytic and respiratory activity and increased glucose uptake, higher chromatin mobility and accessibility, global DNA demethylation, loss of chromocenters, disruption of nucleoli, and decondensation of pericentromeric heterochromatin [4–14]. Therefore, even though not equivalent to 2C embryos, 2CLCs have emerged as a powerful cellular model for investigating early developmental events, such as ZGA, totipotency acquisition, and exit. For instance, recent studies of 2CLCs have led to the identification of OBOX family members as master inducers of ZGA [15, 16], while DUXBL and ZFP352 as facilitators of ZGA exit [17, 18].

The homeobox transcription factor Dux (DUX4 in humans) is a key regulator of ZGA transcripts and 2CLC induction [19–21]. Dux binds and robustly activates many minor ZGA and 2C-specific genes in zygotes and early 2C embryos, including the Zscan4 cluster and MERVL, and is necessary and sufficient for converting mESCs to 2CLCs [20]. Importantly, Dux expression is decommissioned rapidly at the late 2C embryos to ensure the 2C-to-4C transition. Prolonged Dux expression in embryos induces developmental arrest of blastomeres with 2C signatures [22], while induced Dux expression in mESCs results in extensive DNA damage and cell death [23]. Recent studies have implicated the involvement of chromatin structure changes in the regulation of Dux expression. Specifically, the Dux locus, comprising a retrogene array of ∼31 copies in mouse, is normally localized in perinucleolar chromatin and repressed but mobilizes to nucleoplasm, leading to its derepression, in 2CLCs and 2C embryos [12, 24]. Nevertheless, the molecular mechanisms involved in regulating Dux expression and cell fate transitions between pluripotency and totipotency remain to be elucidated.

Immunodeficiency, centromeric instability, and facial anomalies (ICF) syndrome is a rare autosomal recessive immunological and neurological disease characterized by antibody deficiency, facial dysmorphism, intellectual disability, and mental retardation. A hallmark of ICF syndrome is hypomethylation of satellite DNA repeats in heterochromatin regions [25]. Four genes involved in the regulation of DNA methylation are known to be mutated in ICF syndrome: the de novo DNA methyltransferase gene DNMT3B (ICF1), the zinc finger (ZF) transcription factor gene ZBTB24 (ICF2), the cell cycle-regulated gene CDCA7 (ICF3), and the DNA helicase gene HELLS (also known as LSH, ICF4) [26–30]. Recently, compound heterozygous mutations in UHRF1, which encodes an accessory factor of the maintenance DNA methyltransferase DNMT1, were identified in a case with atypical ICF syndrome [31].

Comparative methylation profiling clearly distinguishes the genomic DNA methylation patterns in ICF1- and ICF2–4-deficient cell lines, with some heterochromatin regions being hypomethylated in ICF2–4-, but not ICF1-deficient, samples [25], suggesting distinct mechanisms underlying hypomethylation in different types of ICF syndrome. Notably, we and others have demonstrated that ZBTB24, CDCA7, and HELLS are components of a novel molecular pathway that regulates the specificity of DNA methylation. Specifically, ZBTB24 directly activates CDCA7 transcription, and CDCA7 recruits HELLS to heterochromatin regions to facilitate DNA methylation [32–39]. HELLS has been shown to accelerate replication-uncoupled maintenance DNA methylation at late-replicating heterochromatin regions by facilitating chromatin association of the UHRF1–DNMT1 complex in a adenosine triphosphatase (ATPase)-dependent manner [40]. Notably, ZBTB24 and CDCA7 play no major roles in regulating global DNA methylation [37], while knockout (KO) of HELLS results in substantial loss of global DNA methylation [40, 41], consistent with the notion that CDCA7 confers the specificity of the ZBTB24-CDCA7-HELLS axis in controlling DNA methylation [38].

Given the late-replicating perinucleolar heterochromatin localization of the Dux cluster, we investigated the possible involvement of the ZBTB24-CDCA7-HELLS axis in regulating Dux expression and 2C-like reprogramming. Here, we show that the ZBTB24-CDCA7-HELLS axis, but not DNMT3B, is a critical component of the molecular network that prevents cell fate transition from pluripotency to totipotent-like state. Deficiency in Zbtb24, Cdca7, or Hells in mESCs results in activation of 2C-specific genes and TEs in a Dux-dependent manner. Mechanistically, CDCA7 recruits the CDCA7–HELLS chromatin remodeling complex to heterochromatin, including the Dux cluster, to facilitate maintenance of DNA methylation, thus keeping Dux repressed. We also provide evidence that ICF mutations impair Dux promoter methylation, raising the possibility that abnormal epigenetic reprogramming during early embryogenesis contributes to the pathogenesis of ICF syndrome. Together, our results demonstrate that the ZBTB24-CDCA7-HELLS axis is required for heterochromatin silencing of the Dux cluster.

Materials and methods

Plasmid constructs

The plasmids expressing ZBTB24, CDCA7, HELLS, ICF2, or ICF3 missense mutations—CDCA7-ΔLZ, HELLS-K237A, HELLS-ΔDEAH, HELLS-ΔCC1, and HELLS-ΔC7BH—were generated by cloning the corresponding complementary DNAs (cDNAs) into pCAG-HA-IRESblast or pCAG-3xFLAG-IRESblast vector [42]. Mutations and deletions in ZBTB24, CDCA7, and HELLS were introduced by polymerase chain reaction (PCR)-based mutagenesis. The CRISPRme vector (Dnmt3A-3L-dCas9-P2A-TagBFP) for site-specific DNA re-methylation was generated by removing the KRAB domain in CRISPRoff-v2.1 vector (Addgene, #167981). Control guide RNA (gRNA) and Dux-a/b/c gRNAs for site-specific DNA re-methylation and demethylation were ligated into a modified pX458-puro plasmid (replacing dCas9-P2A-EGFP cassette in pX458 with puromycin). The primers and oligonucleotides used are listed in Supplementary Table S1. All vectors were verified by Sanger sequencing.

mESC culture

The MuERVL-LTR-tdTomato reporter cell line based on E14Tg2A (E14) mESCs (referred to as E14-2C::tdTomato hereafter) was used for all experiments. mESCs were maintained in Dulbecco’s modified Eagle’s medium (Hyclone, SH30243.01) supplemented with 15% fetal bovine serum (Gibco, 10099-141C), 0.1 mM nonessential amino acids (Gibco, 11140-050), 2 mM GlutaMAX (Gibco, 35050-061), 0.1 mM β-mercaptoethanol (Gibco, 21985-023), 50 U/ml penicillin, 50 μg/ml streptomycin (Gibco, 15140-122), and 10 ng/ml mouse leukemia inhibitory factor (ReproTech, 300-05). The cells were normally grown on gelatin-coated culture dishes without feeder cells. The medium was changed daily, and cells were passaged every 2 days. Routine testing revealed the absence of mycoplasma contamination.

Generation of gene KO mESC lines by CRISPR–Cas9 technology

For generating Zbtb24-, Cdca7-, or Hells-KO mESC lines by frameshift mutations, the gRNA sequences were based on our previous study [37]. For generating Dnmt3b-KO cell lines, the genomic region spanning exons 3–24, containing the entire coding sequences, was eliminated by two gRNAs based on a previous study [43]. Primers for each targeting site were annealed and ligated into Bbs I-linearized pSpCas9 (BB)-2A-EGFP plasmid (pX458; Addgene, #48138). All constructs were confirmed by Sanger sequencing. E14-2C::tdTomato mESCs were transfected with plasmid/paired plasmids using Lipofectamine 2000 (Invitrogen, #11668019) according to the manufacturer’s instructions. Twenty-four hours post-transfection, 5000 EGFP-positive cells were sorted by fluorescence-activated cell sorting (FACS) as single cells and plated into 60-mm culture dishes with low density for 7–10 days until individual clones became visible. The cells were then treated with lysis buffer [50 mM Tris–HCl (pH 8.0) plus 100 μg/ml Protease K] at 55°C for 2 h and 95°C for 10 min before genotyping. PCR primers flanking the targeted region were used to amplify the affected region for genotyping. The frameshift mutations of Zbtb24-, Cdca7-, or Hells-KO cell lines were identified by Sanger sequencing. For Dnmt3b-KO clones, the PCR products were analyzed by agarose gel electrophoresis to identify wild type (WT), homozygous KO, and heterozygous clones based on different sizes, and homozygous KO clones were further analyzed and confirmed by Sanger sequencing.

For deleting the entire Dux genome cluster in Zbtb24-, Cdca7-, or Hells-KO mESCs, two gRNAs were used based on previous studies [44, 45]. The 5′-end Dux-KO gRNA was ligated into pX458 and the 3′-end Dux-KO gRNA was ligated into a modified pX330-tdTomato plasmid (pSpCas9(BB)-P2A-tdTomato). Zbtb24-KO #1, Cdca7-KO #2, and Hells-KO #2 mESCs were transfected with the two Dux-KO gRNAs, and 5000 EGFP/tdTomato double positive cells were sorted by FACS and plated into 60-mm culture dishes for 7–10 days. Individual clones were picked for PCR amplification with two pairs of primers for WT allele and one pair of primers for Dux-KO allele. The Dux-KO clones were further verified by Sanger sequencing. All gRNA sequences are listed in Supplementary Table S1.

Western blotting

Cells were collected and washed twice in phosphate-buffered saline (PBS), lysed with cell lysis buffer (P0013, Beyotime) supplemented with phenylmethylsulfonyl fluoride and boiled in sodium dodecyl sulfate sample buffer at 99°C for 5 min. Equal amounts of proteins were resolved by 10%–12% Bis–Tris sodium dodecyl sulfate–polyacrylamide gel electrophoresis and transferred to 0.45 μm polyvinylidine difluoride membranes (Millipore, #IPVH00010). Nonspecific binding was blocked by incubation in 5% bovine serum albumin (BSA) in Tris-buffered saline with 0.1% Tween 20 (TBST) at room temperature (RT) for 1–2 h. Blots were then probed overnight at 4°C in 5% BSA in TBST with various primary antibodies—Zscan4 (AB4340, Millipore), MuERVL-Gag (A-2801-50, Epigentek), DUX (ER1901-52, HUABIO), HA-tag [3724S, Cell Signaling Technology (CST)], Dnmt1 (5032S, CST), Oct4 (sc-5279, Santa Cruz), ZBTB24 (PM085, MBL Life Science), CDCA7 (15249-1-AP, Proteintech), HELLS (11955-1-AP, Proteintech), histone H3 (ab1791, Abcam), H3K9me3 (ab8898, Abcam), HP1α (2616S, CST), HP1γ (2619S, CST), Kap1 (15202-1-AP, Proteintech), Setdb1 (11231-1-AP, Proteintech), Suv39h1 (10574-1-AP, Proteintech), Suv39h2 (11338-1-AP, Proteintech), Lamin B1 (12987-1-AP, Proteintech), and β-actin (P30002L, Abmart). The blots were then probed for 1–2 h at RT with appropriate horseradish peroxidase-conjugated secondary anti-Rabbit IgG (7074S, CST) or anti-mouse IgG (7076S, CST) at 1:5000. The protein bands were detected by the ultrasensitive ECL Chemiluminescence Detection Kit (PK10002, Proteintech). The antibodies used are listed in Supplementary Table S2. To ensure reproducibility, all western blotting experiments were done at least three times, and representative blots were shown.

Immunofluorescence microscopy

Cells were grown on gelatinized glass coverslips before usage as described previously [46]. Cells were washed twice in PBS, then fixed in freshly prepared 4.0% paraformaldehyde in PBS (pH 7.4) for 15 min on ice cube, permeabilized in 0.1% Triton X-100 in blocking solution (3% donkey serum plus 0.1% BSA in PBS) for 40 min at RT, washed three times (each for 15 min), and left in blocking solution at RT for 1 h. Cells were then incubated overnight at 4°C with primary antibodies against MuERVL-Gag (A-2801-50, Epigentek), 5mC (39649, Active Motif), HA-tag (3724S, CST), HA-tag (sc-7392, Santa Cruz), FLAG-tag (AE005, ABclonal), or HELLS (11955-1-AP, Proteintech), washed three times, and incubated at RT for 1 h with secondary antibodies, Alexa Fluor 488 donkey anti-mouse IgG (H + L) (ab150105, Abcam), Alexa Fluor 568 donkey anti-rabbit IgG (H + L) (ab175470, Abcam), or Alexa Fluor 488 donkey anti-rabbit IgG (H + L) (ab150073, Abcam) diluted 1:500 in blocking solution. Samples were washed three times and counterstained with the Prolong Gold antifade reagent with DAPI (P36935, Life Technologies). Digital images were captured using a CCD camera on a Leica confocal microscope equipped with Leica LAS X software. The primary and secondary antibodies used are listed in Supplementary Table S2.

Gene expression analysis by RT-qPCR

Total RNA was isolated from mESCs using the RNA Easy Fast Tissue/Cell kit (TIANGEN, #4992732) according to the manufacturer’s instructions. One microgram of RNA was subject to cDNA synthesis using the PrimeScript^TM RT Reagent kit with gDNA Eraser (RR047B, TAKARA) according to the manufacturer’s instructions. Real-time quantitative PCR (qPCR) reactions were set up in triplicate with the TB Green^® Premix Ex Taq™ II (Tli RNaseH Plus) (RR820B, TAKARA) and run on the CFX-Connect system (Bio-Rad) using primers. Most primers were designed using the IDT DNA website. The qPCR primers used are listed in Supplementary Table S1.

Site-specific DNA demethylation and re-methylation through CRISPRon and CRISPRme systems

For site-specific DNA demethylation at the Dux promoter, 1.5 μg CRISPRon_TETv4 (Addgene, #167983) plasmid plus 1.5 μg pX458-puro-Control gRNA or 1.5 μg CRISPRon_TETv4 plus each 0.5 μg pX458-puro-Dux gRNA-a/b/c were transfected into E14-2C::tdTomato cells by using Lipofectamine 2000 (Invitrogen, #11668019) according to the manufacturer’s instructions. Seventy-two hours post-transfection, mESCs were collected and genomic DNA was extracted and subjected to bisulfite sequencing. For site-specific DNA re-methylation at the Dux promoter, 1.5 μg CRISPRme plasmid plus 1.5 μg pX458-puro-Control gRNA or 1.5 μg CRISPRme plus each 0.5 μg pX458-puro-Dux gRNA-a/b/c were transfected into Zbtb24-, Cdca7-, or Hells-KO mESCs. Seventy-two hours post-transfection, mESCs were collected and genomic DNA was extracted and subjected to bisulfite sequencing.

Bisulfite sequencing

Genomic DNA was extracted from different experimental groups and subjected to sodium bisulfite conversion using the EZ DNA Methylation Kit (D5001, Zymo Research) according to the manufacturer’s instructions. Briefly, 0.5 μg genomic DNA was first mixed with 5 μl M-dilution buffer and incubated at 37°C for 15 min and then mixed with 100 μl CT Conversion Reagent prepared according to the manufacturer’s instructions. Mixtures were incubated at 50°C for 12–16 h. Bisulfite-converted DNA samples were loaded onto columns provided in the kit for desulfonation and purification. Concentration of eluted DNA was measured using a Nanodrop spectrometer.

Mouse Dux promoter, MuERVL-LTR, and MuERVL-ORF regions containing CpG sites were amplified using the following bisulfite primers. For Dux promoter, forward: 5′-TTTGTTAGGGATGAGGAGTT-3′; reverse: 5′-AAACCTCTAATAAACCTCTTTA-3′.

For MuERVL-LTR site, forward: 5′-AGGATATTGGAAGAAGGGAGTTTAG-3′; reverse: 5′-CAATTAAAACCCTTACTTTAACTAT-3′.

For MuERVL-ORF site, forward: 5′-ATATGAATAAAGTGGTTATGGTGGT-3′; reverse: 5′-AATTCCTAAACCCATAAATCCTAA-3′. The PCR products were cloned into the pEasy-T1 simple vector (CT111-01, Transgene) through TA cloning, and 8–15 colonies were randomly picked and sequenced and analyzed for each sample.

RNA-seq and data analysis

Total messenger RNAs (mRNAs) from WT, Zbtb24-KO, Cdca7-KO, Hells-KO, Zbtb24/Dux-DKO, Cdca7/Dux-DKO, and Hells/Dux-DKO mESCs were extracted according to the operating instructions of the VAHTS Universal V6 RNA-seq Library Prepkit for Illumina (NR 604-01/02). Briefly, 1 μg total RNA of each sample was used as the starting material to construct a transcriptome sequencing library, and the libraries were sequenced on the Illumina platform. Raw data were filtered using fastp software (version 0.23.2), and reads were mapped to GRCm39 using HISAT2 (version 2.1.0). featureCounts (version 2.0.1) was used to count the reads mapped to each gene. Differential expression analysis between two comparison sets was performed using DESeq2 software (version 1.40.2), and significantly differentially expressed genes (DEGs) were defined as P< .05 and log₂(FC) ≥ 1. Gene set enrichment analysis (GSEA) was performed using GSEA_4.3.2 software.

For the mapping and analysis of repetitive elements, fastp software (version 0.22.0) was used for quality control of sequencing data, and then HISAT2 (version 2.1.0) was used for comparison to generate sorted BAM files. Finally, SAMtools (version 1.6) was used to create an index for BAM files. After the BAM file was obtained by pre-processing, the gene index and TE index were generated with the reference of genome gtf and transposon gtf, respectively. The TEtranscripts (2.2.3) software was used with the following code: TEtranscripts –format BAM –mode multi -t *. group1.bam -c *. group2.bam –GTF Mus_musculus.GRCm38.84.gtf –TE GRCm38_GENCODE_rmsk_TE.gtf –project group1.txt –sortByPos. Among them, group 1 was the experimental group and group 2 was the control group.

ATAC-seq and data analysis

Approximately 50,000 mESCs were counted for ATAC sequencing (ATAC-seq) library preparation by using the Hyperactive ATAC-Seq Library Prep Kit for Illumina (TD711, Vazyme) according to the manufacturer’s instructions. Briefly, samples were lysed in 5 μl lysis buffer [10 mM Tris–HCl (pH 7.4), 10 mM NaCl, 3 mM MgCl₂, 0.1% NP-40] for 10 min on ice. Immediately after lysis, samples were incubated with the Tn5 transposase and tagmentation buffer at 37°C for 30 min (TD502, Vazyme). Then, 5 μl 5× Tagmentation Stop (TS) buffer was added directly into the reaction to stop the tagmentation process. PCR was then performed to amplify the library for 13 cycles using the following PCR conditions: 72°C for 3 min, 98°C for 30 s, then thermocycling at 98°C for 15 s, 60°C for 30 s, and 72°C for 3 min, followed by 5 min at 72°C. After the PCR reaction, libraries were purified with 0.5× and 1.2× AMPure beads (Beckman, A63880).

The samples were then processed by an Illumina HiSeq ×10 sequencer with 150-bp paired-end sequencing reactions. Clean reads were then mapped to the mouse mm10 genome assembly by Bowtie2 [47], using the parameter -k 20 (report up to 20 alignments per read). Reads with multiple genomic loci and mitochondria were discarded. Distributions of ATAC-seq read density around the transcription start site (TSS) regions (the upstream and downstream 3-kb regions, respectively) were plotted using deeptools with default settings [48].

H3K9me3 Cleavage Under Targets and Tagmentation Sequencing (CUT&Tag-seq) and data analysis

Approximately 50,000 mESCs from WT (three technical replicates), Zbtb24-KO (#1 and #2), Cdca7-KO (#2, #3, and #4), and Hells-KO (#1 and #2) mESCs were counted for CUT&Tag-seq library preparation by using the Hyperactive universal CUT&Tag Assay Kit for Illumina (Vazyme, TD904) according to the manufacturer’s instructions. Briefly, cells were harvested and washed with PBS. In total, 50,000 counted cells were incubated with Concanavalin A beads at RT for 10 min. Next, 2 μl H3K9me3 primary antibody (ab8898, Abcam) was incubated with cells at 4°C overnight. Secondary antibody incubation followed by pA/G-Tnp pro incubation was performed on a rotating platform at RT each for 1 h. The beads were washed using corresponding buffer. Then, transposable products were fragmented by TruePrep Tagment Buffer L (TTBL) at 37°C for 1 h. Libraries were purified with DNA Extract beads. Finally, PCR was performed to amplify the library for 12 cycles using the following PCR conditions: 72°C for 3 min, 95°C for 3 min, then thermocycling at 98°C for 10 s, 60°C for 5 s, followed by 1 min at 72°C. After the PCR reaction, libraries were purified with the VAHTS DNA Clean beads. High-throughput sequencing was performed by Illumina NovaSeq 6000 sequencing system.

For bioinformatics analysis, an initial quality assessment of the FASTQ files was conducted using FastQC (version 0.11.9). The raw sequenced reads were cleaned using the Trimmomatic (version 0.39) with the parameters SLIDINGWINDOW: 4:15 and MINLEN: 53 for paired-end read processing. Further adapter removal and trimming were performed using Cutadapt (version 3.4) and parameters. Cleaned reads from different samples were merged using the cat command. The merged clean reads were aligned to the reference genome using Bowtie2 (version 2.4.2), utilizing the –very-sensitive-local alignment mode. SAM files were converted to BAM format and sorted using SAMtools (version 1.13). Peak calling was performed using MACS2 (version 2.2.7.1) with the parameters: genome size (2677 331 709), –nomodel, –shift −50, –extsize 100, and a q-value cutoff of 0.1. BAM files were indexed using SAMtools. Genome-wide coverage was calculated and normalized using bamCoverage from deepTools (version 3.5.1) with the parameters: minimum fragment length of 80, normalization using CPM, and read extension. Genomic tracks were visualized using pyGenomeTracks (version 3.5).

ChIP-seq and data analysis

To identify genome-wide CDCA7-binding sites, chromatin immunoprecipitation sequencing (ChIP-seq) analysis was performed with HA antibody, using Cdca7^−/− mESCs stably expressing HA-tagged mCDCA7 (two biological replicates, HA-10 and HA-11). For each sample, ∼5 million mESCs were used for ChIP. ChIP-seq libraries were constructed using a KAPA HyperPrep kit (Roche) and sequenced in a 36-bp single-read run on an Illumina HiSeq2500 instrument (Illumina). Standard mapping and peak calling were performed to analyze the ChIP-seq data first. Briefly, the reads were mapped to the mouse genome (mm39) using Bowtie (version 1.1.2) [49] with the following parameters: “-v 2 -m 1 –best –strata”. To avoid PCR bias, only one copy of multiple reads mapped to the same genomic position was retained for further analysis. For each ChIP-seq sample, peaks were identified using MACS (version 1.4.2) [50] by comparing against the corresponding input sample. A window size of 300 bp was used, and a p-value cutoff of 1e−5 was applied. No peak was found within ±50 kb of the Duxf3 gene in mm39. Then the reads were mapped to the assembled contig containing all the Dux gene copies [24] using Bowtie with the parameters: “-v 2 -a –best –strata”. For each sample, the number of reads mapping to the Dux copies or ±50 bp of the CA repeats between the Dux copies was calculated, normalized to 1 million uniquely mapped reads to the whole genome, and presented in Fig. 4D. For each ChIP sample, the coverage of extended reads to 150 bp was normalized to 1 million uniquely mapped reads to the whole genome, normalized to the corresponding input sample, and presented in Fig. 4E.

Figure 4.

Disruption of the ZBTB24-CDCA7-HELLS axis leads to DNA demethylation at the Dux gene promoter. (A) Bisulfite sequencing analysis of CpG dinucleotides at the Dux gene promoter in WT, Zbtb24-, Cdca7-, or Hells-KO mESC lines (2–3 independent KO clones were analyzed for each disrupted gene). Solid black circles indicate methylated CpG dinucleotides, and open circles indicate unmethylated CpG sites. The percentage of methylated CpG dinucleotides in each sample is indicated. (B) Quantitative data (mean ± SEM) of panel (A). For each sample, 10–13 independent clones (n) were sequenced. ***P< .001; ****P< .0001. (C) Western blots (with CDCA7, HA, and β-actin antibodies) showing that the expression levels of the two stable clones (HA-10 and HA-11) were similar to the endogenous CDCA7 level in WT mESCs. (D) Normalized numbers of reads mapped to the Dux gene copies (left) or to ±50 bp of the CA repeats in-between the Dux copies (right). (E) Genome browser screenshots showing CDCA7 enrichment around the CA repeats. The CDCA7-ChIP signal was normalized to the input signal.

Statistical analysis

Data were analyzed by two-tailed unpaired Student’s t-test to compare two groups or one-way ANOVA to compare more than two groups using GraphPad Prism 8 software, and graphs were also prepared using GraphPad Prism 8 software. Statistically significant differences were defined as *P< .05, **P< .01, ***P< .001, ****P< .0001. The results were shown as mean ± standard error of the mean (SEM).

Results

The ZBTB24-CDCA7-HELLS axis negatively regulates 2CLC transition in mESCs

To characterize the roles of the ICF-related proteins—DNMT3B, ZBTB24, CDCA7, and HELLS—in mESCs, we generated individual KO clones with CRISPR/Cas9 gene editing using gRNAs published previously [37, 43]. Sanger sequencing and immunoblotting confirmed the successful generation of the KO mESC lines (Supplementary Fig. S1A–E). In agreement with our previous finding that ZBTB24 directly activates Cdca7 transcription [35, 37], the CDCA7 level is dramatically reduced in Zbtb24-KO mESCs (Supplementary Fig. S1E). To assess the effect of deficiency in Dnmt3b, Zbtb24, Cdca7, or Hells on the transcriptome, we performed RNA sequencing (RNA-seq) analysis of WT and the KO mESCs. Principal component analysis showed that Zbtb24-, Cdca7-, and Hells-KO mESC clones clustered together, which were separated from WT and Dnmt3b-KO mESC clones (Supplementary Fig. S2A), consistent with the notion that ZBTB24, CDCA7, and HELLS function in the same molecular pathway, with DNMT3B not being a major player. Indeed, Zbtb24-, Cdca7-, and Hells-KO mESCs had many common DEGs, both up- and downregulated ones (Supplementary Fig. S2B–D). Deficiency in Dnmt3b, Zbtb24, Cdca7, or Hells did not affect the expression of core pluripotency transcription factors Oct4, Nanog, and Sox2 and the typical dome-shaped morphology of mESC colonies, indicating that the gene expression changes were not consequences of mESC differentiation (Supplementary Fig. S2E and F).

GSEA of the RNA-seq data revealed significant upregulation of a 2C-specific gene set in Zbtb24-, Cdca7-, or Hells-KO mESCs, but not in Dnmt3b-KO mESCs, compared with WT mESCs (Fig. 1A). Specifically, Zbtb24, Cdca7, or Hells deficiency led to increased expression of signature transcripts of 2CLCs and 2C embryos, including the Dux and Zscan4 cluster genes, Usp17-like family members, and several TEs (Fig. 1B and C and Supplementary Fig. S2G–L). Of note, Gm4981, a truncated gene within the Dux locus, was also upregulated upon loss of Zbtb24, Cdca7, or Hells (Fig. 1C). Western blotting and reverse transcription-quantitative PCR (RT-qPCR) analyses of two to three independent mESC lines deficient for Zbtb24, Cdca7, or Hells KO confirmed the drastic upregulation of Zscan4 and MuERVL-Gag proteins and typical 2C-specific genes and MERVL (Fig. 1D and E). Immunostaining of MuERVL-Gag protein, a marker of 2CLCs [3], revealed significant increases in 2CLCs in independent Zbtb24-, Cdca7-, or Hells-KO mESC lines compared with WT mESCs (Fig. 1F and G). Consistent with ZBTB24 functioning upstream of CDCA7 and HELLS [32–35, 37], some of the changes were less severe in Zbtb24-KO mESCs than in Cdca7-KO and Hells-KO mESCs. Together, our results indicate that the ZBTB24-CDCA7-HELLS axis is involved in the negative regulation of 2C-like state transition in mESCs.

Figure 1.

Deficiency in ZBTB24, CDCA7, or HELLS induces a 2C-like gene signature. (A) GSEA using the 2C signature genes defined in GSE33923 [3] to compare Dnmt3b-, Zbtb24-, Cdca7-, and Hells-KO with WT mESCs. NES, normalized enrichment score. (B) Heatmap showing typical totipotent gene expression in WT, Dnmt3b-KO, Zbtb24-KO, Cdca7-KO, and Hells-KO mESCs. Technical replicates were applied for WT and Zbtb24-KO #1 mESCs. (C) Integrative Genomics Viewer screenshots of RNA-seq track peaks across all Zscan4 family members, Duxf3 and Gm4981 loci in WT and Zbtb24-, Cdca7-, and Hells-KO mESCs. (D) Western blotting analysis showing the upregulation of Zscan4 and MuERVL-Gag proteins in independent Zbtb24-, Cdca7-, and Hells-KO mESC lines compared with WT mESCs. (E) RT-qPCR analysis of representative totipotent 2C-specific genes and TEs in WT and Zbtb24-, Cdca7-, and Hells-KO mESCs. Two to three independent mESC cell lines for each disrupted gene were used. Error bars indicate SEM. (F and G) Immunofluorescence staining of MuERVL-Gag protein showing significantly higher percentages of 2CLCs in cultured Zbtb24-, Cdca7-, and Hells-KO mESCs compared with WT mESCs. Shown are representative images (F) and quantitative data (mean ± SEM) (G). Scale bar, 10 μm. n, number of cells counted from 5–8 fields for each mESC line for determining percentage of 2CLCs. **P< .01, ***P< .001, ****P< .0001.

Dux is required for activation of the 2C-like transcriptome related to disruption of the ZBTB24-CDCA7-HELLS axis

Dux is a master activator of 2C-like reprogramming. To assess its contribution to the upregulation of 2C-specific genes and TEs associated with disruption of the ZBTB24-CDCA7-HELLS axis, we biallelically deleted the entire ∼370-kb Dux cluster, using two gRNAs (Fig. 2A) [24, 44], in Zbtb24-, Cdca7-, or Hells-KO mESCs. Deletion of Dux was confirmed by PCR genotyping and Sanger sequencing (Fig. 2B and Supplementary Fig. S3). Consistent with the changes in Dux mRNA levels (Fig. 1), Dux protein was detected in Zbtb24-, Cdca7-, or Hells-KO, but not WT, mESCs, and it was eliminated in the Zbtb24/Dux, Cdca7/Dux, and Hells/Dux double KO (DKO) mESCs, as expected (Fig. 2C–E). In the DKO mESCs, the Dux targets Zscan4 and MuERVL-Gag were expressed at levels comparable to or even lower than those in WT cells, in contrast to their dramatic upregulation in Zbtb24, Cdca7, or Hells single KO mESCs (Fig. 2C–E). RNA-seq analysis revealed that ∼40% of the DEGs in Zbtb24-, Cdca7-, or Hells-single KO mESCs were reversed by deletion of the Dux cluster (Supplementary Fig. S4A). Among them, ∼30%–45% of the upregulated genes in the single KO cells became downregulated in the corresponding DKO cells (Supplementary Fig. S4B). Notably, the majority of the upregulated 2C-specific genes, including the Zscan4 cluster and Usp17-like family members, in the single KO mESCs were restored by deletion of the Dux cluster (Fig. 2F–H). These data demonstrate that the activation of 2C-like transcriptome induced by disruption of the ZBTB24-CDCA7-HELLS axis is Dux dependent.

Figure 2.

Activation of the 2C program induced by deficiency in ZBTB24, CDCA7, or HELLS is Dux dependent. (A) Schematic depiction of the Dux cluster, the deletion strategy using CRISPR/Cas9 technology with two gRNAs, and the locations of the primer pairs for genotyping. (B) Agarose gels illustrating the genotyping results for confirmation of clones with the Dux cluster deletion. Clone #18 in Zbtb24/Dux-DKO group is heterozygous. (C–E) Western blotting analysis showing the restoration of the Zscan4 and MuERVL-Gag protein levels in Zbtb24/Dux-DKO (C), Cdca7/Dux-DKO (D), or Hells/Dux-DKO (E) mESCs, compared with Zbtb24, Cdca7, or Hells single KO mESCs, and confirming Dux upregulation in the single KO cells and its elimination in the DKO cells. Note that the Dux antibody detected a strong nonspecific band in addition to Dux (denoted by *). (F–H) Heatmaps showing the restoration of totipotent gene expression in Zbtb24/Dux-DKO (F), Cdca7/Dux-DKO (G), or Hells/Dux-DKO (H) mESCs compared with single KO mESCs.

Expression of 2C-specific genes and TEs induced by Zbtb24, Cdca7, or Hells deficiency is correlated with increased chromatin accessibility

To decipher the mechanisms underlying elevated Dux expression and 2C-like transcriptome, we performed ATAC-seq to analyze the chromatin accessibility of 2C-specific genes and TEs. In general, Zbtb24-, Cdca7-, or Hells-KO mESCs had more ATAC-seq peaks than WT mESCs, indicating that the chromatin is more accessible in mESCs deficient for any of the genes (Fig. 3A and Supplementary Fig. S5A and B). While only small fractions (∼2.8%–3.0%) of the ATAC-seq peaks were mapped to the DEGs identified in the corresponding KO cells (Supplementary Fig. S5C), the vast majority of the DEGs [i.e. ∼80% (872/1119) in Zbtb24 KO cells, ∼84% (1114/1331) in Cdca7 KO cells, and ∼80% (942/1182) in Hells KO cells] gained peaks relative to WT cells (Supplementary Fig. S5D). Analysis of the transcription start sites (TSSs), as well as the proximal upstream and downstream regions, of all DEGs showed an overall more open chromatin state at TSSs in the KO mESCs compared with WT mESCs (Fig. 3B). Further analysis revealed that, consistent with their expression changes, the upregulated genes, including 116 common genes that were upregulated in all three KO groups, showed increased chromatin accessibility at their TSSs (Fig. 3C and D), while the downregulated genes, including the common ones (also 116 genes) shared by all three KO groups, displayed decreased chromatin accessibility at their TSSs (Fig. 3E and F). It is worth mentioning that genes that were not differentially expressed also showed stronger ATAC-seq signals at their TSSs in the KO cells (Supplementary Fig. S5E), suggesting that increased chromatin accessibility does not always correlate with changes in gene expression.

Figure 3.

Disruption of the ZBTB24-CDCA7-HELLS axis promotes chromatin accessibility at 2C-specific genes and TEs. (A) The numbers of ATAC-seq peaks in the genomes of WT, Zbtb24-, Cdca7-, or Hells-KO mESCs. Error bars indicat SEM. (B) The average profiles and heatmaps of the ATAC-seq enrichment around accessible promoters at all DEGs in WT and the indicated KO mESCs (Z24, Zbtb24; C7, Cdca7; Hs, Hells). (C,D) The average profiles and heatmaps of the ATAC-seq enrichment around accessible promoters at upregulated genes in Zbtb24-, Cdca7-, or Hells-KO mESCs compared with WT mESCs (C) and the 116 common upregulated genes in all three KO groups (D). (E,F) The average profiles and heatmaps of the ATAC-seq enrichment around accessible promoters at downregulated genes in Zbtb24-, Cdca7-, or Hells-KO mESCs compared with WT mESCs (E) and the 116 common downregulated genes in all three KO groups (F). (G) The average profiles and heatmaps of the ATAC-seq enrichment around accessible promoters at 2C genes that were upregulated in Zbtb24-, Cdca7-, or Hells-KO mESCs compared with WT mESCs. (H) The average profiles and heatmaps of the ATAC-seq enrichment showing increased chromatin accessibility at the MuERVL-int transposons in the indicated KO mESCs compared with WT mESCs. (I–K) The average profiles showing the increased ATAC-seq signals at the Dux cluster (I), Zscan4 cluster (J), and Usp17-like cluster (K) in the indicated KO mESCs compared with WT mESCs. TSS, transcription start site.

Prominent among the regions that exhibited increased chromatin accessibility in the KO cells were 2C-specific genes and the MuERVL-int transposons (Fig. 3G and H), in agreement with the activation of the 2C-like program in these cells (Fig. 1 and Supplementary Fig. S2). As examples, Fig. 3I–K showed the ATAC-seq signals in the Dux cluster and two of its targets, the Zscan4 and Usp17-like gene clusters, which were stronger in numerous regions, especially at their TSSs, in Zbtb24-, Cdca7-, or Hells-KO mESCs relative to WT mESCs.

In summary, the chromatin accessibility of many genes, including Dux and its target 2C-specific genes and TEs, is elevated upon disruption of the ZBTB24-CDCA7-HELLS axis. While Dux expression may influence the chromatin accessibility of its target genes, the general effect of Zbtb24, Cdca7, or Hells deficiency in inducing chromatin accessibility, including at many non-2C-specific upregulated genes (Supplementary Fig. S5F), indicates that the changes are mostly independent of Dux.

Disruption of the ZBTB24-CDCA7-HELLS axis results in DNA demethylation at the Dux locus

The Dux cluster is normally localized at the repressive perinucleolar heterochromatin regions and mobilizes to nucleoplasm in 2CLCs or following disruption of the nucleolar integrity, leading to its derepression and activation of the 2C-like transcriptome program [12, 51]. DNA methylation and repressive histone modifications (e.g. H3K9me3) are essential for heterochromatin formation and gene repression [52]. Thus, we attempted to elucidate the epigenetic barriers controlled by the ZBTB24-CDCA7-HELLS axis that prevent Dux derepression.

We first examined the repressive histone mark H3K9me3 and key proteins involved in “writing” or “reading” the mark. Western blotting and RNA-seq analyses showed that the total H3K9me3 level and the expression of components of the H3K9 methylation supercomplex, including Setdb1, Suv39h1/h2, Kap1, HP1α, HP1γ, and Lamin B1, were not altered in mESCs deficient for Zbtb24, Cdca7, or Hells (Supplementary Fig. S6A and B). More importantly, H3K9me3 CUT&Tag-seq analysis showed that disruption of the ZBTB24-CDCA7-HELLS axis had no effect on H3K9me3 occupancy at the Dux and Zscan4 clusters (Supplementary Fig. S6C and D). Our observation was consistent with a recent report that, in spleens from Cdca7^G305V ICF3 mutant homozygous mice, H3K9me3 enrichment was unchanged at known H3K9me3 sites, including endogenous retroviruses and the Zscan4 gene cluster [53].

We then analyzed the DNA methylation level at the Dux promoter. Bisulfite sequencing revealed significant demethylation of the Dux promoter in Zbtb24-, Cdca7-, or Hells-KO mESCs compared with WT mESCs (methylation levels dropped from ∼85% in WT to ∼42%–47% in KO mESCs) (Fig. 4A and B). In contrast, the DNA methylation levels at the Dux promoter in Dnmt3b-KO cell lines were comparable to those in WT mESCs (Supplementary Fig. S7). The results correlated with the dramatic upregulation of Dux and the 2C-like transcriptome in Zbtb24-, Cdca7-, or Hells-KO, but not in Dnmt3b-KO, mESCs (Fig. 1). Similar loss of DNA methylation was also observed in the long terminal repeat (LTR) and open reading frame (ORF) regions of MERVL retrotransposon in the KO mESCs (Supplementary Fig. S8), in agreement with the elevated expression of MERVL at both the transcriptional and protein levels (Figs 1 and 2 and Supplementary Fig. S2).

Recent studies have provided evidence that CDCA7 confers the specificity of the ZBTB24-CDCA7-HELLS axis in the regulation of DNA methylation [33, 37–39, 54]. Thus, we asked whether CDCA7 binds the Dux cluster. To this end, we performed ChIP-seq using Cdca7^−/− mESCs reconstituted with HA-tagged mCDCA7. Two stable clones (HA-CDCA7 #10 and #11) with expression levels comparable to that of endogenous mCDCA7 in WT mESCs (Fig. 4C) were selected for ChIP using an HA antibody. For each sample, we generated ∼18–29 million reads.

Bioinformatics analysis of the ChIP-seq data identified 1851 and 3801 CDCA7-binding peaks in the HA-CDCA7 #10 and HA-CDCA7 #11 clones, respectively. They were located mainly in upstream regions (−50k to −5k from TSS, 19% or 20%), gene bodies (introns and exons, 32% or 35%), and distant regions (+5k to +50k from transcription end site, 22% or 24%), and only a small portion was present in promoter regions (5% or 6%) (Supplementary Fig. S9A). Further analysis revealed strong enrichment of CDCA7 at simple repeats (Supplementary Fig. S9B), with short CA dinucleotide repeats (CA)n and the equivalent TG dinucleotide repeats (TG)n being dominant (Supplementary Fig. S9C). Indeed, simple CA repeats were identified as the most significant CDCA7-binding motif in both samples (Supplementary Fig. S9D).

The current reference genome (mm10 or mm39) is incomplete, containing only one or two of the many gene copies in the Dux cluster. Thus, we used an assembled contig containing all the gene copies from the literature [24]. The contig is a ∼3.7-Mb sequence with ∼31 Dux gene copies (2025 bp per copy except the last one, which is 1896 bp). An obvious feature among the gene copies is the presence of short CA repeats that locate in the highly similar duplicated Dux promoter regions, around 870-bp upstream of a Dux gene copy. Conventional peak calling identified no peak within ±50 kb of the Duxf3 gene in mm39. Mapping the reads to the assembled contig also revealed no CDCA7 enrichment (relative to input) in the Dux gene copies, but overt CDCA7 enrichment was observed around the CA repeats between the gene copies (Fig. 4D and E). The results suggest direct binding of CDCA7 to the Dux locus.

DNA demethylation of the Dux promoter induces its derepression

To establish the causal link between DNA demethylation of the Dux promoter and its derepression and subsequent activation of the 2C-like transcriptome, we first induced global DNA demethylation by using the DNMT1 degrader GSK-3484862 [55]. Transient treatment of mESCs with GSK-3484862 induced rapid and drastic loss of DNA methylation globally (Supplementary Fig. S10A and B), as well as at the Dux promoter (Supplementary Fig. S10C and D), resulting in strong activation of Dux expression (∼10-fold) and its target genes, including Zscan4, Tcstv1/3, Dub1, Gm2022, and Gm4340 and retrotransposons MuERV-L and MT2_Mm (Supplementary Fig. S10E). Western blotting analysis further verified the dramatic increases in Zscan4 and MuERVL-Gag proteins following GSK-3484862 treatment (Supplementary Fig. S10F). Immunostaining of MuERVL-Gag protein also indicated substantial increases in the fraction of 2CLCs in mESC population following GSK-3484862 treatment (Supplementary Fig. S10G and H). Our results were consistent with recent findings that hypomethylation facilitates, while hypermethylation impairs, the transition from the pluripotent to totipotent-like state in cultured mESCs [56–58]. Together, these observations suggest that DNA methylation is likely involved in the regulation of the transition between pluripotent and totipotent-like states in mESCs.

RNA-seq analysis identified 2389 upregulated genes and 965 downregulated genes (P< .05, fold change ≥ 2) following GSK-3484862 treatment for 3 days, consistent with a repressive effect of DNA methylation on gene expression. GSEA and Z-score analyses of the RNA-seq data revealed significant upregulation of a 2C-specific gene set following GSK-3484862 treatment (Supplementary Fig. S10I and J). Shown in Supplementary Fig. S10K–N are representative 2C-specific genes (such as Dux, Zscan4, Usp17-like, Tcstv1/3, Gm8300, and Gm2022) and retrotransposons (including MERVL-int, MT2-Mm, and multiple LINE-1s), which were dramatically upregulated following GSK-3484862-mediated global DNA hypomethylation.

To avoid complications associated with global DNA hypomethylation, we also utilized CRISPRon system to induce DNA demethylation specifically at the Dux promoter with three gRNAs (Fig. 5A) [59]. Dux gRNAs, relative to control gRNA, induced significant hypomethylation in the Dux promoter region (Fig. 5B and C), with DNA methylation reaching a level comparable to that in Zbtb24-, Cdca7-, or Hells-KO mESCs (Fig. 4A and B). As a result, Dux was strongly activated (∼10-fold) (Fig. 5D), and Dux targets (including Zscan4, Tcstv1/3, Gm2022, and Gm4340) and retrotransposon (MuERV-L) were even more strongly (>10-fold) upregulated, with Zscan4 exhibiting the highest induction (∼35-fold) (Fig. 5D). Western blotting analysis also verified the dramatic increases in Zscan4 and MuERVL-Gag proteins following CRISPRon treatment with the Dux gRNAs compared with control gRNA (Fig. 5E).

Figure 5.

Restoration of DNA methylation at the Dux promoter in mESCs deficient for ZBTB24, CDCA7, or HELLS reverses the elevation of Dux target genes and TEs. (A) Schematic representation of site-specific DNA demethylation and re-methylation at the Dux locus by CRISPRon or CRISPRme system. The locations of Dux-gRNA-a/b/c at the Dux promoter are shown. (B) Bisulfite sequencing analysis of CpG dinucleotides at the Dux promoter in WT mESCs transfected with CRISPRon plasmid and control gRNA or Dux gRNAs. Solid black circles indicate methylated CpG dinucleotides, and open circles indicate unmethylated CpG sites. The percentage of methylated CpG dinucleotides in each sample is indicated. (C) Quantification (mean ± SEM) of bisulfite sequencing results of 11–12 independent clones (n) at the Dux promoter as shown in panel (B). (D) RT-qPCR data showing expression of totipotent 2C-specific genes and TEs following site-specific DNA demethylation at the Dux promoter with the CRISPRon system. Error bars denote SEM. (E) Western blotting analysis showing the upregulation of 2C-specific proteins Zscan4 and MuERVL-Gag in WT mESCs after DNA demethylation at the Dux promoter with the CRISPRon system. (F, J, and N) Bisulfite sequencing analysis of CpG dinucleotides at the Dux promoter in Zbtb24-KO (F), Cdca7-KO (J), and Hells-KO (N) mESCs transfected with CRISPRme plasmid and control gRNA or Dux gRNAs. (G, K, and O) Quantification (mean ± SEM) of bisulfite sequencing results of ∼10 independent clones (n) at the Dux promoter using CRISPRme system in Zbtb24-KO (G), Cdca7-KO (K), and Hells-KO (O) mESCs. (H, L, and P) RT-qPCR data showing expression of totipotent 2C-specific genes and TEs following site-specific DNA re-methylation at the Dux promoter with the CRISPRme system in Zbtb24-KO (H), Cdca7-KO (L), and Hells-KO (P) mESCs. Error bars denote SEM. (I, M, and Q) Western blotting analysis showing the restoration of Dux targets Zscan4 and MuERVL-Gag in Zbtb24-KO (I), Cdca7-KO (M), and Hells-KO (Q) mESCs after DNA re-methylation at the Dux promoter with the CRISPRme plasmid and Dux gRNAs compared with control gRNA. In panels (B)–(E) and (F)–(Q), the cells were collected 72 h post-transfection for bisulfite sequencing, qPCR, and western blotting analyses. *P< .05, **P< .01.

To further establish the causal relationship between Dux promoter methylation and its repression, we asked whether site-specific re-methylation of the Dux promoter in Zbtb24-, Cdca7-, or Hells-KO mESCs would restore its repression state and subsequent restoration of 2C-specific genes and TEs. We utilized the CRISPRme system coupled with three Dux gRNAs to achieve site-specific DNA methylation at the Dux promoter (Fig. 5A). In all the KO mESC lines, CRISPRme treatment with Dux gRNAs, compared with control gRNA, led to significant increases in DNA methylation at the Dux promoter (Fig. 5F, G, J, K, N, and O), accompanied by varying degrees of decreases in Dux expression and complete or almost complete restoration of the expression of Dux targets Zscan4, Tcstv1/3, Dub1, Gm2022, and MuERV-L (Fig. 5H, L, and P). Western blotting analysis verified the restoration of Zscan4 and MuERVL-Gag at the protein levels in KO mESCs after CRISPRme treatment with Dux gRNAs (Fig. 5I, M, and Q).

Collectively, our data strongly suggest that DNA demethylation of the Dux promoter is responsible for its derepression and subsequent activation of numerous 2C-specific genes and TEs in Zbtb24-, Cdca7-, or Hells-KO mESCs.

CDCA7-mediated recruitment of HELLS to heterochromatin is required for maintaining the Dux promoter methylation

Recent studies have shown that CDCA7 forms a complex with HELLS, via its leucine zipper (LZ) motif, and recruits the complex to specific genomic regions by recognizing hemi-methylated CpG dinucleotides through its C-terminal cysteine-rich domain (CRD) (Fig. 6A) [33, 38, 39, 54]. Thus, CDCA7 is believed to confer the functional specificity of the ZBTB24-CDCA7-HELLS axis. We recently showed that CDCA7 is highly enriched at constitutive heterochromatin during DNA replication and that the ICF3 missense mutations—all occurring in the CRD—disrupt such enrichment [38] (Fig. 6A and Supplementary Figs S11 and S12A), suggesting a key role for CDCA7 in targeting the DNA methylation machinery to heterochromatin. Indeed, HELLS enrichment in constitutive heterochromatin is dependent on CDCA7, as “HELLS foci” at DAPI-bright spots were observed in WT, but not Cdca7-KO, mESCs (Fig. 6B, upper panel). Re-expression of HA-tagged WT mCDCA7 in Cdca7-KO mESCs restored HELLS enrichment at constitutive heterochromatin, whereas re-expression of mCDCA7 containing the ICF3 mutation R285H or the LZ deletion (mCDCA7-ΔLZ) did not (Fig. 6C), indicating that the abilities of CDCA7 to bind hemi-methylated DNA and interact with HELLS are both required for recruiting HELLS to constitutive heterochromatin. In contrast, the enrichment of CDCA7 at heterochromatin is not dependent on the presence of HELLS, as formation of HA-CDCA7 foci was unaffected in Hells-deficient mESCs (Fig. 6B, lower panel). Similar experiments revealed that the presence and absence of ZBTB24 and CDCA7 do not affect each other’s localization patterns (Supplementary Fig. S12B and C).

Figure 6.

The ZBTB24-CDCA7 axis maintains DNA methylation of the Dux promoter and suppresses 2C gene transcriptome. (A) Mouse CDCA7 protein with the known domains. The four missense mutations (R285H, R285C, G305V, and R315H) equivalent to ICF3 mutations in human patients are shown. (B) Representative immunofluorescence images showing that heterochromatin localization of HELLS is dependent on CDCA7 while heterochromatin localization of CDCA7 is independent on HELLS. Scale bar, 10 μm. (C) Representative immunofluorescence staining images showing that expression of HA-tagged WT CDCA7, but not the R285H or the ΔLZ mutant, in Cdca7-KO mESCs restores HELLS enrichment in heterochromatin foci. Note that CDCA7 foci formation is eliminated by the R285H mutation but unaffected by the LZ deletion. Cdca7-KO #2 mESCs were transiently transfected with indicated expression vectors and subjected to immunofluorescence staining 48 h post-transfection. Scale bar, 10 μm. (D–F) Western blotting analysis showing the restoration of Zscan4 and MuERVL-Gag levels in Cdca7-KO mESCs by stably expressing WT CDCA7 (D), but not CDCA7-R285H (D), CDCA7-ΔLZ (E), or WT ZBTB24 (F). (G) Bisulfite sequencing analysis showing methylation levels at the Dux promoter in Cdca7-KO mESCs stably expressing WT CDCA7, CDCA7-R285H, CDCA7-ΔLZ, or WT ZBTB24. Solid black circles indicate methylated CpG dinucleotides, and open circles indicate unmethylated CpG sites. The percentage of methylated CpG dinucleotides in each sample is indicated. (H) Quantification (mean ± SEM) of bisulfite sequencing results in panel (G). Note that only WT CDCA7 could rescue the DNA methylation level at the Dux promoter in Cdca7-KO mESCs. (I) Mouse ZBTB24 protein, including the BTB domain, AT-hook, and ZF domain with eight tandem ZFs. Two missense mutations (C382Y and C407G) equivalent to the ICF2 mutations in human patients are shown. (J, K) Western blotting analysis showing the restoration of Zscan4 and MuERVL-Gag protein levels in Zbtb24-KO mESCs by stably expressing WT ZBTB24 or CDCA7, but not ICF mutants (ZBTB24-C382Y, ZBTB24-C407G, and CDCA7-R285H). (L) Bisulfite sequencing analysis showing methylation levels at the Dux promoter in Zbtb24-KO mESCs stably expressing the indicated ZBTB24 or CDCA7 proteins. (M) Quantification (mean ± SEM) of bisulfite sequencing results in panel (L). Note that WT ZBTB24 or CDCA7, but not ICF mutants, could rescue the DNA methylation levels at the Dux locus in Zbtb24-KO mESCs. *P < .05, **P< .01, ***P < .001; n.s., not significant.

To assess the functional significance of CDCA7-mediated HELLS recruitment to chromatin, we established stable cell lines in Cdca7-KO mESCs expressing HA-tagged WT CDCA7, the R285H mutant, CDCA7-ΔLZ, or WT ZBTB24. Western blotting analyses revealed that the upregulation of the Dux targets Zscan4 and MuERVL-Gag in Cdca7-KO mESCs was fully prevented by WT CDCA7 but remained unchanged in CDCA7-R285H- or -ΔLZ-expressing cells (Fig. 6D and E). Overexpression of ZBTB24, an activator of Cdca7 transcription, in Cdca7-KO mESCs also showed no effect on the upregulated expression of Zscan4 and MuERVL-Gag (Fig. 6F), in agreement with the previous finding that the effect of ZBTB24 on DNA methylation (perhaps including methylation at the Dux promoter) is mediated by CDCA7 [37]. Consistent with the expression changes of the Dux targets, only WT CDCA7 was able to restore the DNA methylation level at the Dux promoter (Fig. 6G and H).

By stably expressing HA-tagged proteins in Zbtb24-KO mESCs, we also showed that, unlike WT ZBTB24, two ICF2 mutants (C382Y and C407G) in the ZF domain (Fig. 6I)—which lose the ability to induce Cdca7 transcription [35]—failed to restore the levels of Zscan4 and MuERVL-Gag proteins and DNA methylation at the Dux promoter (Fig. 6J, L, and M). Indeed, exogenously expressed CDCA7, but not the ICF3 mutant R285H, in Zbtb24-KO mESCs could reverse the changes in Dux methylation and expression of its targets (Fig. 6K–M), further suggesting that the inhibitory effect of ZBTB24 on 2C-like reprogramming is mediated mainly (if not solely) by CDCA7.

Taken together, our results demonstrate that CDCA7-mediated recruitment of HELLS to constitutive heterochromatin is critical for maintaining DNA methylation at the Dux promoter and preventing activation of the 2C program.

The ability to bind CDCA7 and chromatin remodeling activity of HELLS are required for maintaining the Dux promoter methylation

HELLS is an SNF2-family helicase harboring ATPase and chromatin remodeling activities when bound by CDCA7 [33]. Its ATPase and chromatin remodeling activities are required for promoting replication-uncoupled maintenance DNA methylation at late-replicating heterochromatin regions [40, 60, 61]. Two N-terminal coiled-coil alpha helices of HELLS also play critical roles in regulating DNA methylation, with CC1 showing autoinhibitory effect on the chromatin remodeling activity and CC2, also known as the CDCA7-binding helix (C7BH), acting as the CDCA7-binding interface (Fig. 7A) [39, 62].

Figure 7.

The formation of HELLS–CDCA7 complex and the chromatin remodeling activity of HELLS are required for maintaining DNA methylation at the Dux promoter. (A) Mouse HELLS protein with known domains. CC1 is a coiled-coil domain with autoinhibition effect on helicase activity; CC2 (C7BH) is required for interacting with CDCA7. (B) Representative immunofluorescence staining images showing the localization patterns of endogenous HELLS in WT mESCs and transiently expressed HA-HELLS proteins in Hells-KO mESCs. Note that heterochromatin enrichment of HELLS is abolished by C7BH deletion but unaffected by mutations in the other domains. Hells-KO #2 mESCs were transiently transfected with indicated expression vectors and subjected to immunofluorescence staining 48 h post-transfection. Scale bar, 10 μm. (C, D) Western blotting analysis showing that stable expression of WT HELLS or HELLS-ΔCC1, but not the other mutants (ΔC7BH, K237A, and ΔDEAH), in Hells-KO mESCs restored Zscan4 and MuERVL-Gag levels. (E) Bisulfite sequencing analysis showing methylation levels at the Dux gene promoter in Hells-KO mESCs stably expressing WT or mutant HELLS proteins. Solid black circles indicate methylated CpG dinucleotides, and open circles indicate unmethylated CpG sites. The percentage of methylated CpG dinucleotides in each sample is shown. (F) Quantification (mean ± SEM) of bisulfite sequencing results in panel (E). Note that WT HELLS or HELLS-ΔCC1, but not HELLS-ΔK237A, HELLS-ΔDEAH, or HELLS-ΔC7BH, could rescue the DNA methylation levels at the Dux locus in Hells-KO mESCs. *P< .05, ***P< .001, ****P< .0001; n.s., not significant.

We first investigated the contributions of different domains in HELLS to its enrichment at constitutive heterochromatin. Immunostaining of endogenous HELLS protein in WT mESCs and HA-tagged HELLS proteins transiently expressed in Hells-KO mESCs revealed that, except HELLS-ΔC7BH, which displayed a diffuse nuclear localization patterns, all mutant HELLS proteins, i.e. the K237A (ATPase-dead) mutation and ΔDEAH and ΔCC1 deletions, like WT HELLS, were enriched at constitutive heterochromatin foci (Fig. 7B), indicating that its ability to interact with CDCA7, but not its ATPase and chromatin remodeling activities, is required for HELLS to be recruited to constitutive heterochromatin.

To determine the functional significance of different HELLS activities in inhibiting 2C-like reprogramming, we opted for rescue experiments by stably expressing HA-tagged HELLS proteins in Hells-KO mESCs. WT HELLS and the ΔCC1 deletion mutant completely or almost completely reversed the changes in Zscan4 and MuERVL-Gag levels and DNA methylation at the Dux promoter, with the ΔCC1 mutant (especially clone #1) showing stronger effects, whereas the other mutants (ΔC7BH, K237A, and ΔDEAH) had no rescue activities (Fig. 7C–F). Thus, the chromatin remodeling activity and the ability to interact with CDCA7 are both required for HELLS to exert its role in maintaining DNA methylation at the Dux promoter and in preventing 2C-like reprogramming.

Discussion

mESCs cycle in and out of a transient totipotent 2C-like state and maintain a rare population (∼0.5%) of 2CLCs in cultures [3, 63, 64]. Dux is a master inducer of the 2C-like state and is necessary and sufficient for the conversion of mESCs to 2CLCs [19–21]. The Dux cluster is normally localized at the perinucleolar heterochromatin, a generally repressive environment [12]. Indeed, the repressive histone mark H3K9me3 has been implicated in repressing Dux expression, as the Dux locus is enriched with H3K9me3 [65–67] and genetic manipulations of H3K9me3 regulators result in activation of 2C-specific genes [3, 19, 67–70]. Nevertheless, the fact that Dux is silenced in the vast majority (∼99.5%) of mESCs at any given time and the observation that induced or prolonged Dux expression is detrimental to early embryos and mESCs [22, 23] suggest that Dux expression is tightly controlled, raising the possible involvement of multiple epigenetic mechanisms.

In this study, we demonstrate that DNA methylation, a highly stable repressive mark, is essential for Dux repression in mESCs and that the ZBTB24-CDCA7-HELLS axis, a newly identified molecular pathway that directs the DNA methylation machinery to heterochromatin regions [32–39], plays a key role in maintaining DNA methylation at the Dux locus. Our conclusions are supported by multiple lines of evidence. First, a CpG-rich region within the Dux promoter is highly methylated in mESCs, and induction of site-specific demethylation of the region attenuates Dux repression and activates 2C-specific transcripts and MERVL (Fig. 5A–E). Second, deficiency in Zbtb24, Cdca7, or Hells, but not Dnmt3b, in mESCs, results in Dux hypomethylation and derepression (Fig. 4A and B and Supplementary Fig. S7), leading to activation of the 2C-like program in a Dux-dependent manner (Figs 1 and 2), and these changes can be reversed by site-specific re-methylation at the Dux promoter (Fig. 5F–Q). Third, CDCA7, which confers the specificity of the ZBTB24-CDCA7-HELLS axis in the regulation of DNA methylation [33, 37–39, 54], is responsible for recruiting the CDCA7–HELLS chromatin remodeling complex to heterochromatin to facilitate the maintenance of DNA methylation at the Dux promoter (Fig. 6). In contrast, the CDCA7-ICF3 mutants fail to be concentrated in constitutive heterochromatin even though still capable of forming complexes with HELLS, thus exhibiting Dux hypomethylation and derepression (Fig. 6 and Supplementary Figs S11 and S12A). We and others recently uncovered that CDCA7 specifically recognizes a hemi-methylated non-B form DNA highly enriched in constitutive peri/centromeric heterochromatin regions that contributes to the concentration of CDCA7 in constitutive heterochromatin during DNA replication, while the ICF3 mutants lose the hemi-methylated DNA binding capability [38, 39, 54]. Given the perinucleolar heterochromatin localization of the Dux cluster [12], we speculate that CDCA7 binds the Dux cluster and then recruits HELLS for Dux promoter methylation maintenance. Indeed, our ChIP-seq analysis reveals enrichment of CDCA7 specifically at the CA repeats in the promoter of each Dux gene copy (Fig. 4C–E and Supplementary Fig. S9). Finally, the ability to bind CDCA7 and chromatin remodeling activity of HELLS are both required for maintaining Dux promoter methylation as well as its repression (Fig. 7). Both CDCA7-ΔLZ and HELLS-ΔC7BH mutants, which disrupt the CDCA7–HELLS interaction, fail to maintain Dux promoter methylation and its repression (Figs 6 and 7). In addition, the HELLS-K237A and HELLS-ΔDEAH mutants with abrogated ATPase and chromatin remodeling activities, respectively, also lose the ability to rescue Dux promoter methylation in Hells-KO mESCs (Fig. 7). Together, our results highlight the importance of DNA methylation in preventing 2C-like reprogramming in mESC cultures. Both DNA methylation and H3K9me3 are involved in the formation and maintenance of constitutive heterochromatin. Thus, they may function cooperatively to keep Dux repressed. Results from the current study demonstrate that the ZBTB24-CDCA7-HELLS axis controls Dux expression and 2C-like reprogramming by regulating DNA methylation with little, if any, involvement of H3K9me3.

As DNMT3B, a de novo DNA methyltransferase gene, is commonly mutated in ICF syndrome, accounting for ∼50% of cases [26, 29, 30], and HELLS has been reported to regulate de novo DNA methylation [60, 71], DNMT3B was initially thought to be the major downstream effector of the other ICF-related proteins. The assumption would predict more severe or at least similar DNA methylation changes in ICF1 compared with other types of ICF syndrome. However, comparative DNA methylation analysis turned out to be the opposite: despite unifying hypomethylation of pericentromeric repeats and a few common loci by mutations in the four ICF-related genes, some heterochromatin regions, including centromeric satellite repeats (α-satellite in humans and minor satellite in mouse), are hypomethylated in cells deficient for ZBTB24, CDCA7, or HELLS, but unaffected by DNMT3B deficiency [25, 28, 72, 73]. These observations suggest that additional components of the DNA methylation machinery act downstream of the ZBTB24-CDCA7-HELLS axis. Indeed, recent work suggests that HELLS primarily promotes replication-uncoupled DNA methylation maintenance by enhancing the chromatin association of the UHRF1–DNMT1 complex with late-replicating constitutive heterochromatin [40], which is normally wrapped in condensed nucleosome core particle (NCP) accompanied by restricted chromatin accessibility [74, 75]. The localization of hemi-methylated DNA within the NCP inhibits its detection by the SRA domain of UHRF1 [76]. Nevertheless, the CRD domain of CDCA7 could recognize the hemi-methylated DNA in NCP and recruit HELLS to the site, which, as a nucleosome remodeler, unwraps DNA from the NCP, revealing the hemimethylation site to UHRF1 [33, 39, 40]. Moreover, coevolution analysis reveals that CDCA7 and HELLS have stronger evolutionary links to the maintenance DNA methylation components DNMT1 and UHRF1 than to the de novo DNA methyltransferases DNMT3A and DNMT3B [77]. The recent identification of compound heterozygous UHRF1 mutations in an atypical ICF patient [31] provides further evidence that defective maintenance of DNA methylation is mainly responsible for the pathogenesis of ICF types other than ICF1. Our observation that deficiency in Zbtb24, Cdca7, or Hells, but not Dnmt3b, in mESCs leads to decreased Dux promoter methylation and activation of the 2C program supports the notion that the ZBTB24-CDCA7-HELLS axis is involved in maintaining, rather than establishing, methylation at the Dux locus. Likely, CDCA7 recruits the chromatin remodeler HELLS to late-replicating heterochromatin, including the Dux cluster, to facilitate the accessibility to the UHRF1–DNMT1 complex.

By repressing Dux expression, the ZBTB24-CDCA7-HELLS axis is perhaps a key component of the molecular network that maintains the mESC state. However, the presence of a rare population of 2CLCs in mESCs is essential for maintaining telomere lengths and safeguarding genomic stability, ensuring their long-term self-renewal in cultures [78, 79]. How is the mESC-to-2CLC transition triggered and regulated? It is possible that genetic lesions accumulated in individual mESCs eventually induce the expression of sufficient Dux activators, such as p53 [24]. Conceptually, the epigenetic barrier, including DNA methylation at the Dux locus, also needs to be breached for the transition to occur. DNA methylation and heterochromatin reinforce each other, and both are likely involved in keeping Dux silenced. Future work is warranted to determine their interplay and relative contributions in this context.

Dux is activated in late zygote and early 2C embryo during the minor ZGA stage and is quickly decommissioned in late 2C stage to allow development to proceed in mice, and prolonged expression of Dux in embryos significantly induced developmental arrest of blastomeres with 2C signatures [22]. Thus, the precise temporal regulation of Dux is important for embryonic development. It has been demonstrated recently that DUXBL controls exit from totipotency through competitive recruitment to Dux-bound regions to facilitate 2C-gene and MERVL silencing [18]. However, our knowledge about the factors involved in regulating cell states and transition during early embryogenesis is still limited. By reanalyzing the public expression dataset of early embryos [80], we found that the low expression of Cdca7 is coincident with the activation of Dux in late zygote and early 2C embryos, in contrast to the high expression of Hells during the whole preimplantation development stages (Supplementary Fig. S13). From late 2C embryo stage onward, Cdca7 is highly expressed (Supplementary Fig. S13). Although Hells is high in the minor ZGA stage, the low expression of Cdca7 may prevent its recruitment to the Dux cluster, allowing Dux to be induced. In later stages, the CDCA7–HELLS chromatin remodeling complex may facilitate Dux silencing. Future studies are warranted to explore the functional significance of the ZBTB24-CDCA7-HELLS axis in early embryos.

While loss of DNA methylation in specific genomic regions is considered the primary defect in ICF syndrome, the dysregulated genes, molecular pathways, and cellular and developmental processes that lead to major clinical features, such as antibody deficiency and intellectual disability, remain largely unknown. Our results suggest that ICF mutations are defective in maintaining DNA methylation at the Dux locus, raising the possibility that DUX4 (human homolog of Dux) derepression and the resultant defects in epigenetic reprogramming during early embryogenesis contribute to the pathogenesis of ICF syndrome.

Epigenetic derepression of DUX4 is also linked to facioscapulohumeral muscular dystrophy (FSHD), a form of muscular dystrophy with a distinctive pattern of skeletal muscle weakness [81]. The DUX4 cluster is localized within the CpG-rich D4Z4 subtelomeric repeat near the end of chromosome 4q in the nuclear periphery [82, 83], and its expression is regulated by telomere position effect [84]. Notably, the DUX4 cluster is highly methylated in most somatic tissues, including skeletal muscle [85]. In FSHD, upregulation of DUX4 correlates with short telomeres. FSHD patient myoblasts and myotubes with short telomeres exhibit over 10-fold upregulation of DUX4, and its expression is inversely proportional to telomere length [84]. Telomerase-deficient (Terc^−/−) murine embryonic fibroblast cells with short telomeres exhibit DNA hypomethylation at subtelomeric regions [86]. These observations suggest that DNA methylation is also a major epigenetic mechanism that represses DUX4 in humans. It is possible that CDCA7 also recognizes the DUX4 cluster in late-replicating subtelomeric heterochromatin and recruits HELLS to maintain DNA methylation for its repression and inhibiting the onset of FSHD. The demonstration of ZBTB24-CDCA7-HELLS axis in DUX4 repression may lead to more thorough understanding of the FSHD pathology and developing strategies for treating FSHD from a new angle. Given the demonstrated role of DUX4 in FSHD and its possible involvement in ICF syndrome, an intriguing question that warrants exploring in the future is, whether the pathogeneses of the two disorders are connected.

Supplementary data

Supplementary data is available at NAR online.

Conflict of interest statement. None declared.

Funding

This study was supported by grants from National Natural Science Foundation of China (32270845), Yunnan Fundamental Research Projects (202201BE070001-006 and 202301AT070304), “Xingdian Talent Support Program” of Yunnan Province (KKRD202273103), and Natural Science Foundation of Yunnan Province (202102AA100053). Funding to pay the Open Access publication charges for this article was provided by National Natural Science Foundation of China.

Data availability

The raw data of the RNA-seq, ATAC-seq, H3K9me3 CUT&Tag-seq, and HA-CDCA7 ChIP-seq analyses reported in this study are deposited in the NCBI GEO database, and the accession numbers are GSE280752, GSE279787, GSE290522, and GSE291998, respectively.