CDCA7 targets LSH to DNA maintenance methylation in S phase and transcription regulation in interphase via two distinct DNA-binding modes Authors: Fenghua Chen, Meilin Sun, Mingzhi Chang, Zhongye Dai, Zhaosu Chen, Jialun Li, Yu Duan, Jiwen Li, Xiongwen Cao, Yuanyong Huang, Jiemin Wong Published: 2026 ABSTRACT AbstractMutations or pathogenic variants in CDCA7 and lymphoid-specific helicase (LSH) lead to immunodeficiency, centromeric instability, and facial anomalies (ICF) syndrome, and CDCA7 has been recognized as a hemimethylated DNA sensor facilitating DNA methylation. Here, via genome profiling experiments we unravel that CDCA7 is actually highly enriched at the poorly methylated, CG-rich promoter regions and that an ICF syndrome-causing G294V mutation abolishes this activity. In vitro CDCA7 but not the G294V mutant can bind both CG-rich DNA probes and hemimethylated DNA. Consistent with these dual DNA-binding activities, CDCA7 exhibits two distinct subcellular localization patterns along the cell cycle: a diffuse nuclear distribution in interphase and a hemimethylated DNA-dependent pericentromeric heterochromatin domain localization in late S phase. Notably, LSH is recruited by CDCA7 to both the CG-rich promoter regions in interphase and heterochromatin domains in late S phase. Whole genome DNA methylation analysis confirmed a crucial and conserved role of CDCA7 and LSH in DNA maintenance methylation. Transcription analyses revealed that CDCA7 and LSH can interdependently regulate transcription independent of DNA methylation. Altogether, our study reveals that CDCA7 and LSH have a role beyond DNA methylation and can regulate gene expression via binding CG-rich DNA motifs, thus providing new insights into the function of CDCA7/LSH and the complexity of ICF syndrome. Introduction It has long been recognized that aberrant DNA methylation due to DNA methyltransferase (DNMT) 3B mutation is a causal factor contributing to the phenotype and pathogenesis of Immunodeficiency, centromeric instability, and facial anomalies (ICF) syndrome [1–4]. In recent years, new subtypes of ICF syndrome were reported, caused respectively by mutation of genes or pathogenic variants encoding the zinc finger (ZF) and BTB domain containing 24 (ZBTB24), cell division cycle associated 7 (CDCA7), lymphoid-specific helicase (LSH) [5–7], and ubiquitin-like with PHD and ring finger domains 1 (UHRF1) [8]. The ICF subtypes caused by mutations in ZBTB24, CDCA7, and LSH exhibit highly consistent pathological phenotypes with differences from the DNMT3B mutant subtype, suggesting that these three genes may all be involved in the same regulatory pathway [9–11]. Indeed, recent studies indicate that ZBTB24 functions as a transcription factor essential for CDCA7 expression [12, 13], whereas CDCA7 interacts with and forms a bipartite chromatin remodeler complex with LSH [14], a member of the SNF2 chromatin remodeling family helicases also known as HELLS that has long been known for function in DNA methylation [15–18], thus providing a molecular link for the role of CDCA7 in DNA methylation. Because DNA methylation is a central mechanism of epigenetic regulation with diverse biological functions [19–21], how CDCA7 and LSH regulate DNA methylation has been recently intensively investigated [13–15, 18, 22]. LSH is required for effective DNA methylation in mammals [15], most likely due to its ability to promote DNA maintenance methylation by the axis of DNMT1/UHRF1 in the context of chromatin [16, 17]. As a ZF domain containing protein, CDCA7 was first reported to form stable complex with LSH and stimulate its chromatin remodeling enzymatic activity [14]. Recently, CDCA7 was reported to specifically bind hemimethylated DNA [23, 24]. This finding suggests a working model that CDCA7 facilitates DNA maintenance methylation by targeting LSH to newly replicated DNA via its intrinsic hemimethylated DNA-binding activity, and LSH in turn promotes DNA methylation by DNMT1/UHRF1 in chromatin. Interestingly, CDCA7 has also been shown to bind a CpG dyad in unique non-B form DNA sequence motif that is highly enriched at centromeres of human chromosomes via the same ZF domain [25]. However, these findings are largely based on in vitro biochemical and structural studies, and how CDCA7 associates with genomic DNA in cells remains to be elucidated. CDCA7, also known as JPO1, was initially identified as a Myc target gene [26] and recognized as a transcription factor frequently overexpressed in human tumors [27]. However, it is unknown whether CDCA7 regulates transcription independent of its role in DNA methylation [28]. In addition, it is unknown whether the ICF-causing mutations affect only CDCA7 DNA methylation activity and/or transcription activity. To address these issues, we first aimed to define how CDCA7 associates with the genome. To this end, we first generated cell lines with an HA-tag knocked into the N-terminus of CDCA7 in human cells. We then knocked in an ICF disease-causing G294V point mutation into the CDCA7 gene. Via HA-tag-based chromatin immunoprecipitation DNA sequencing assay (ChIP-seq) and CUT&TAG assay and motif enrichment analysis, we surprisingly found that CDCA7 is highly enriched in the CpG-rich promoters and the G294V mutation impairs this activity. Furthermore, in vitro experiments validated that purified CDCA7 protein but not the G294V mutant can specifically bind either unmethylated CG-rich DNA or DNA probe with a single hemimethylated CpG. We present evidence that these two distinct DNA-binding modes allow CDCA7 to directly regulate transcription via binding of CG-rich promoters and facilitate DNA maintenance methylation in S phase via binding of hemimethylated DNA. Furthermore, we demonstrate that both activities are LSH dependent. Materials and methods Plasmid construction For CDCA7 knockout and G294V knockin mutations, the corresponding single guide RNA (sgRNA) sequences (Supplementary Table S1) were cloned into the pX459 Clustered regularly interspaced short palindromic repeat-associated protein Cas9 (CRISPR/Cas9) plasmid. For expression of human CDCA7 in mammalian cells, the CDCA7 coding sequence was cloned into the PCDH-CMV plasmid containing an HA tag. The resulting plasmid was used for site-directed mutagenesis to generate G294V, R274C, and R304H mutants. For expression of CDCA7 ZF domain (CDCA7-ZF) as a GST fusion, the complementary DNA sequence encoding CDCA7 amino acids 252–371 were cloned into the pGEX-4T-1 plasmid. The resulting plasmid was then used for generation of GST-ZF(G294V) mutant by site-directed mutagenesis. Cell culture and related experiments Human cell lines (HeLa, HEK293T) and a mouse cell line (NIH3T3) were cultured in Dulbecco’s modified Eagle’s medium (Gibco) supplemented with 10% Fetal bovine serum (FBS) (Gibco), 100 U/ml penicillin, and 100 μg/ml streptomycin (Gibco). Transfection was performed at 50%–70% cell confluency by using Lipofectamine 2000 (Invitrogen). Transiently transfected cells were selected with puromycin (MCE) for 48 h, whereas stably transfected cells were selected with puromycin for 7 days. Generation of HA-tag and mutation knockin CDCA7 cell lines To knock in an HA tag into the N-terminus of the CDCA7 gene, HeLa cells were co-transfected with pX459 plasmid harboring the target sgRNA and a single-stranded DNA donor (Supplementary Table S1). After selection by puromycin for 48 h, monoclonal cells were seeded to 96 wells and cultured. Once the cells had expanded to sufficient numbers after ~2 weeks, the clones with correct homozygous HA knockin were identified via western blotting followed by DNA sequencing. One of the HA-CDCA7 knockin cell line was then used for knockin of G294V mutation by the same approach and verified by western blotting and DNA sequencing. Western blotting Western blotting analysis was performed as described by Li et al. [29]. All antibodies used in this study are listed in Supplementary Table S2. Immunofluorescence staining assay Immunofluorescence staining was performed as described by Li et al. [29]. Images were visualized using a Leica TCS SP8 confocal microscope. For immunostaining detection of nascent DNA, cells were cultured with 10 μM EdU (5-ethynyl-2′-deoxyuridine) at 37°C for 15 min. The cells were then fixed and permeabilized, followed by Apollo staining of EdU newly incorporated into the genome using a RiBoBio Cell-Light™ EdU Fluorescent Detection Kit (#C10310) according to manufacturer’s instruction. Chromatin immunoprecipitation followed by DNA sequencing ChIP-seq analysis was performed as described by Huang et al. [30] using anti-HA antibody or LSH antibody (Supplementary Table S2.). The resulting DNA fragments were subjected to next-generation sequencing (NGS) and quantitative polymerase chain reaction (qPCR) analysis. The sequences of primers used for ChIP-qPCR (chromatin immunoprecipitation followed by quantitative PCR) analysis are listed in Supplementary Table S3. CUT&Tag CUT&Tag assays were performed using 50 000 cells and a Hyperactive Universal CUT&Tag Assay Kit (Vazyme #TD903-01) according to the manufacturer’s detailed procedure (https://bio.vazyme.com/). HPLC analysis To prepare genomic DNA for DNA methylation analysis by high-performance liquid chromatography (HPLC), cells were resuspended in cell lysis buffer [20 mM Tris–HCl (pH 7.5), 10 mM ethylenediaminetetraacetic acid (EDTA), 100 mM NaCl, 0.5% (w/v) sodium dodecyl sulfate] with 0.1 mg/ml RNase (TIANGEN) and incubated at 37°C overnight. Proteinase K (Merck) was then added to the cell lysate at a final concentration of 0.2 mg/ml and the mixture was incubated at 65°C for 12 h. Genomic DNA was extracted using phenol–chloroform method and dissolved in double-distilled water. To hydrolyze the genomic DNA, 25 μg of denatured genomic DNA was incubated in 133 μl hydrolysis solution [40 mM NaAc, pH 5.2, 1 mM ZnSO₄, 1.5 U/ml nuclease P1 (Wako, 145-08221)] at 37°C overnight. Subsequently, 5 U Fast AP (Thermo Fisher Scientific, EF0651) was added to the DNA and the mixture was incubated for an additional 8 h at 37°C. A 40-μl aliquot of hydrolysate of DNA was analyzed using an Agilent 1260 HPLC system with a Supersil AQ-C18 column (5 μm, 4.6 × 250 mm; Elite HPLC). Deoxycytidine (dC) and 5-methyl-2′-deoxycytidine (5mC) were detected at 280 nm using a UV detector. The DNA methylation level was calculated as 5mC/(C + 5mC). Protein purification To purify GST-tagged CDCA7-ZF and CDCA7-ZF(G294V) recombinant proteins, the corresponding pGEX-4T-1 expression constructs were transformed into Escherichia coli strain Rosetta 2 (DE3). The resulting positive colonies were inoculated into 10 ml LB medium containing ampicillin and cultured at 37°C overnight. The next day, the DE3 cultures were transferred to a 500-ml LB medium for cultivation, during which the OD600 value of the DE3 culture was measured every hour. When the OD600 of the DE3 culture reached 0.6, isopropyl beta-D-1-thiogalactopyranoside (IPTG) was added to a final concentration of 0.1 mM to induce protein expression and the culture was further incubated at 16°C overnight. The next day, DE3 were harvested by centrifugation at 12 000 rpm for 10 min and resuspended in lysis buffer [1× phosphate buffered saline, 1% (v/v) Triton X-100, 10% (v/v) glycerol] containing pre-added protease inhibitors (cocktail) and dithiothreitol (DTT) (1 mM). A high-pressure homogenizer (ATS #AH1500) was used to release the proteins of DE3 lysate at 4°C for 10 min. The supernatant of the lysate was collected after 12 000 rpm centrifugation. The GST magnetic beads were added to supernatant and rotated for 1 h to collect the GST-CDCA7 protein. After three times washing, the CDCA7 protein bound to the magnetic beads was eluted by glutathione, followed by ultrafiltration to obtain a pure aqueous protein solution according to manufacturer’s protocol. Electrophoresis mobility shift assay All DNA probes were chemically synthesized, with forward strand of DNA with a 5′ end 5-carboxyfluorescein (5-FAM) for chemiluminescence image detection (Biorad ChemiDoc MP). Forward and reverse single-stranded DNA probes (100 nM) are mixed with 2× annealing buffer [20 mM Tris–HCl (pH 8.0), 2 mM EDTA, 100 mM NaCl], followed by an annealing reaction on PCR machine (95°C for 5 min; decreasing by 0.1°C per second until reaching 25°C). Purified proteins (50 nM) were incubated with DNA probes (20 nM) in binding buffer [4% (v/v) glycerol, 1 mM MgCl2, 0.5 mM EDTA, 0.5 mM DTT, 0.1% (w/v) bovine serum albumin, 50 mM NaCl, 0.05 mg/ml poly(dI:dC), 0.05% (v/v) NP-40, and 10 mM Tris–HCl (pH 7.5)] for 30 min at room temperature. Then, the mixtures were subjected to electrophoresis (65 V, 90 min) on native polyacrylamide gel [6% (v/v) polyacrylamide, 2% (v/v) glycerol, 0.5× Tris–borate–EDTA (TBE)] in 0.5× TBE buffer. Finally, the signal of DNA probe separated on gel was detected by the Alexa 488 fluorescence channel of Biorad ChemiDoc MP. Knockout and knockdown of CDCA7 and LSH in Hela cells Transfections of small interfering RNAs (siRNAs) for knockdown or pX459 CRISPR–Cas9 vector for knockout into HeLa cells at the logarithmic growth phase were performed using liposome reagent according to manufacturer’s protocol (Hanbio #HB-LF-1000). For selection of CDCA7 knockout clones, the transfected single cells were picked to 96 wells to grow under selection by puromycin. Once cells proliferated to a sufficient number after ~2 weeks, the clones were splited and subjected to western blotting and DNA sequencing analyses. The homozygous knockout clones were selected for subsequent genome-wide assays or RNA-seq analysis. For knockdown experiments, 24 h after transfection, a portion of cells were harvested for qPCR and western blot (WB) analysis to detect knockdown efficiency. The remaining cells were used for genome-wide assays or RNA-seq experiments. The siRNA sequences used in this study are listed in Supplementary Table S4. Luciferase reporter assay The DNA fragment of CD24 promoter was amplified from HEK293T cell line genomic DNA by PCR and then inserted into the PGL3 reporter plasmid upstream of the luciferase gene. HEK293T cells (3 × 104) were cultured in triplicate in 24-well plates for 12–24 h. Then, the pGL3 recombinant reporter plasmids were transiently co-transfected with HA-CDCA7 and/or FLAG-LSH expression plasmid into HEK293T cells as indicated. Twenty-four hours after transfection, HEK293T cells were lysed to release cellular proteins. Subsequently, the luciferase reaction was performed using the Dual-Luciferase Reporter Assay Kit (Vazyme # DL101-01) and the signal intensity was immediately detected with a multifunctional microplate reader (biotek). Western blotting was performed simultaneously for HA and FLAG to examine the expression levels of exogenously transfected CDCA7 and LSH, respectively. RNA-seq analysis For RNA-seq analysis, total RNAs were extracted from each experimental group with three independent replicate samples and the equivalent number of cells. The construction of the mRNA (messenger RNA) libraries and NGS were performed by Berry Genomics. The RNA-seq reads were aligned to the hg38 reference human genome using the STAR aligner (version 2.7.8). The reference genome was obtained from the GENCODE website (https://www.gencodegenes.org/). Differential expression analysis between the two experimental groups was performed using DESeq2 (version 1.40.2). Differentially expressed genes were identified based on adjusted P-values (false discovery rate, FDR) <.05 and absolute log2 fold change >0.5. Data visualization was performed using the R packages ggplot2 (v3.4.1) and heatmap (v2.14.0). ChIP-seq and CUT&Tag data analysis Raw data were processed by TrimGalore (v.0.6.5) to remove adapter sequences and poor-quality sequences. For ChIP-seq analysis, the trimmed reads were mapped to the hg38 reference genome of human using Bowtie (v.2.4.5) with default parameters. For Cut&Tag analysis, genome mapping was run by Bowtie (v.2.4.5) with the following parameters: --no-mixed --no-discordant --dovetail --very-sensitive --local. The BAM files converted from output SAM files were sorted and indexed using Samtools v.1.7. Additionally, nonuniquely mapped reads and low-quality reads (q score <20) were also filtered out by Samtools. After removal of PCR duplicates using Picard Tools MarkDuplicates, peak calling was performed using MACS (v.2.2.7) with default parameters for ChIP-seq data. The peaks in ENCODE blacklist were filtered out, as were peaks that did not meet either the significance threshold of q-value <0.05 for narrow peak calling. CUT&Tag peaks were called on merged replicates and normalized to IgG using SEACR. Aggregate coverage plots and chromatin accessibility heatmaps were generated using the PlotHeatmap and PlotProfile functions from the DeepTools package. Whole genome bisulfite sequencing For whole genome bisulfite sequencing (WGBS) library construction and subsequent analysis, genomic DNA was fragmented to a size range of 200–400 bp using a Covaris S220 system. Single-stranded DNA fragments were subjected to bisulfite conversion, followed by adapter ligation, size selection, and PCR amplification, using the Accel-NGS Methyl-Seq DNA Library Kit. Library quality was evaluated by an Agilent 5400 system, and concentration was quantified by qPCR. Paired-end sequencing was conducted on an Illumina platform. Raw sequencing data were quality-filtered using TrimGalore with default parameters to remove low-quality reads and adapter sequence. Filtered reads were aligned to the human reference genome (hg38) using Bismark, with methylation calling performed by Bismark Methylation Extractor. Genomic annotations of highly methylated domains (HMDs) and partially methylated domains (PMDs) were as described by Zhou et al. [31]. Results CDCA7 is mainly enriched in the promoter regions containing CpG islands Due to the lack of a CDCA7-specific antibody suitable for chromatin immunoprecipitation assay of endogenous CDCA7 (data not shown), we knocked in an HA tag at the N-terminus of CDCA7 by CRISPR–Cas9-based gene editing in HeLa cells (Fig. 1A and Supplementary Fig. S1A). After puromycin selection and single-cell derived clone identification, we obtained two monoclonal cell lines with the HA tag correctly inserted into the N-terminus of CDCA7 (HA-CDCA7) as shown by DNA sequencing data in Supplementary Fig. S1B. To investigate the underlying mechanism of ICF syndrome-causing mutations, we also introduced the G294V point mutation into the CDCA7 gene in one of HA-CDCA7 cell line using the same CRISPR–Cas9 editing strategy (Fig. 1A and Supplementary Fig. S1A). G294V mutation was selected for study because an equivalent mutation has been studied in mouse cell lines and an ICF mouse model [22, 24]. Two monoclonal HA-CDCA7 cell lines with G294V point mutations [HA-CDCA7(G294V)] were obtained (Supplementary Fig. S1B). Additionally, we constructed CDCA7 knockout cell lines (CDCA7 KO) in HeLa (Fig. 1A and Supplementary Fig. S1A) and mouse NIH3T3 cells (Supplementary Fig. S1C) using the CRISPR–Cas9 system, which were confirmed by DNA sequencing analysis (Supplementary Fig. S1D and E). Endogenous CDCA7 associates predominantly at the CG-rich promoters. (A) Schematic diagram illustrating knockin of an HA tag into the CDCA7 N-terminus via CRISPR–Cas9 technology to generate HA-CDCA7 cell lines, and subsequent knockin of CDCA7 G294V mutation to generate HA-CDCA7(G294V) cell lines (right side). In addition, CDCA7 knockout (KO) cell lines were also generated via CRISPR–Cas9 technology (left side). (B) Western blotting analysis examining the protein levels of CDCA7 and other DNA methylation-related proteins in CDCA7 KO, HA-CDCA7, and HA-CDCA7(G294V) cell lines. GAPDH served as a loading control. Two independent monoclonal cell lines were examined for each type of cells. (C) Analyzing CDCA7 subcellular localization by subcellular fractionation assays. Control and two lines of CDCA7 KO cells were fractionated into whole cell extracts (WCE), cytosol, nuclear extract, and nuclear pellet, followed by western blotting analysis using antibodies as indicated. GAPDH, LAMIN A/C, and H3 were used as markers for the cytosol, nuclear supernatant, and nuclear pellet fractions, respectively. (D) Quantification of the levels of 5mC in genomic DNA derived from control, CDCA7 KO, HA-CDCA7, and HA-CDCA7(G294V) HeLa cells by HPLC. The levels of 5mC were shown as 5mC/(5mC + C)%. Data are representative of three independent experiments and are shown as the mean ± SEM *P < .05; **P < .01; ***P < .001; ****P < .0001, n = 3. (E) Pie chart showing the genomic annotation of HA-CDCA7 ChIP-seq peaks obtained from HA-CDCA7 cells using anti-HA antibody. (F) Heatmaps showing ChIP-seq signal of HA-CDCA7, HA-CDCA7(G294V), H3K4me3, and H3K9me3 around 2633 HA-CDCA7 peaks (±3 kb). (G) Heatmaps showing ChIP-seq signal of HA-CDCA7, HA-CDCA7(G294V), H3K4me3, and H3K9me3 around CpG islands (±3 kb). (H) Heatmaps of HA-CDCA7 ChIP-seq signals around the TSSs divided into CpG island and non-CpG island promoters. (I) A representative ChIP-seq IGV browser view of CDCA7 binding in promoter region. The highlighted regions correspond to CDCA7 peaks derived from the HA-CDCA7 cells. (J) A list of the top 5 CDCA7-binding motifs (ranked by P-value). The motifs were identified by HOMER software from the ChIP-seq data of HA-CDCA7 cells. We next examined whether CDCA7 knockout and G294V mutation affect the expression of CDCA7 and DNA methylation related proteins. We confirmed the absence of CDCA7 protein in both CDCA7 KO HeLa cell lines (Fig. 1B, lines 2 and 3) and the successful knockin of HA tag into the CDCA7 locus, because WB analysis using a CDCA7-specific antibody detected a slightly upshifted CDCA7 band that was also detected by WB analysis using anti-HA antibody in both HA-CDCA7 cell lines but not in parental HeLa cells (Fig. 1B, lines 4 and 5). The introduction of G294V mutation reduced the level of CDCA7 protein, as evident by WB using either anti-CDCA7 or HA antibody (Fig. 1B, lines 6 and 7). However, neither CDCA7 knockout nor G294V mutation in HeLa cells affected the protein levels of DNA methylation-related proteins DNMT1, UHRF1, LSH, DNMT3A, and DNMT3B (Fig. 1B), consistent with previous reports [22, 32]. Similarly, we observed that CDCA7 knockout in mouse NIH3T3 cells also did not affect the protein levels of DNMT1, UHRF1, and LSH (Supplementary Fig. S1F). G294V mutation downregulated CDCA7 protein level, most likely because this mutation impaired CDCA7 protein stability, as demonstrated by cycloheximide (CHX) chase assay (Supplementary Fig. S1G). To examine whether CDCA7 is associated with chromatin, we performed subcellular fractionation assay with parental and CDCA7 KO HeLa cells. CDCA7 was mainly detected in the nuclear pellet (Fig. 1C). A portion of LSH was also detected in the nuclear pellet, although like DNMT1 and UHRF1 more LSH was detected in the nuclear extract fraction (Fig. 1C). Loss of CDCA7 did not significantly affect the partition of LSH in soluble nuclear extract and nuclear pellet (Fig. 1C). Similarly, loss of CDCA7 did not affect DNMT1 and UHRF1 distribution in soluble nuclear extract and nuclear pellet (Fig. 1C). This analysis suggested that CDCA7 may bind tightly to DNA and/or chromatin. By co-immunoprecipitation (co-IP) assay, we confirmed that CDCA7 interacted with LSH, and the ICF-causing mutations of CDCA7, including R274C, G294V, and R304H, did not affect this interaction (Supplementary Fig. S2A), in agreement with previous studies [14]. We found that CDCA7 also interacts with UHRF1 (Supplementary Fig. S2B) and poorly, if any, with DNMT1 (Supplementary Fig. S2C). Since abnormal DNA hypomethylation is a characteristic phenotype of ICF syndrome [22, 33], we examined global DNA methylation of above cell lines we constructed by HPLC-based analysis. In HeLa cells, we detected an average of 17.7% and 15.1% decrease of DNA methylation in CDCA7 KO and HA-CDCA7(G294V) cells, respectively (Fig. 1D). However, no reduction of DNA methylation was detected in HA-CDCA7 cells, indicating that HA tag knockin did not affect CDCA7 function (Fig. 1D). mCdca7 KO NIH3T3 cells also exhibited a 11.8% decrease in global DNA methylation (Supplementary Fig. S2D), confirming a conserved role of CDCA7 in DNA methylation in mammalian cells. Having established that both CDCA7 KO and G294V mutation but not HA tag knockin impaired DNA methylation, we performed ChIP-seq experiments using anti-HA antibody with two biological replicates to profile the genome-wide occupancy of CDCA7 and the consequence of G294V mutation. To exclude potential non-specific signal of anti-HA antibody, ChIP-seq was also carried out with a control cell line that expresses HA tag (HA-vec). No significant HA peak was detected for control cells using the MACS2 peak calling algorithm, but 2633 peaks were detected for HA-CDCA7 cells (Fig. 1E), suggesting that these represent the authentic HA-CDCA7 binding peaks. Surprisingly, genomic element distribution analysis revealed that the detected HA-CDCA7 peaks were enriched overwhelmingly in the promoter regions (65.6%) (Fig. 1E). Notably, virtually all HA-CDCA7 peaks were lost in the G294V mutant cells (Fig. 1F), demonstrating that the ICF mutation abolished the CDCA7 DNA-binding activity. In agreement with its robust enrichment in the promoter region, HA-CDCA7 peaks strongly correlated with H3K4me3 but not H3K9me3 peaks (Fig. 1F). Further data mining revealed that CDCA7 peaks were highly enriched at CpG islands (Fig. 1G), and specifically enriched at the CpG island-containing promoters but not the non-CpG island promoters (Fig. 1H). A representative genome browser view clearly illustrates the strong enrichment of HA-CDCA7 at the CpG islands and correlation with H3K4me3 peaks (Fig. 1I). Motif enrichment analysis using HOMER [34] revealed that the top 5 CDCA7 binding motifs are strongly enriched with CG or GC dinucleotides (Fig. 1J). Altogether, our HA tag knockin-based genomic profiling experiments provided strong evidence that CDCA7 is highly enriched at the CpG island-containing promoters. Confirmation of CDCA7 enrichment at the CG-rich promoters Given that CDCA7 is expected to promote DNA methylation, the strong enrichment of CDCA7 at the CpG island promoters is unexpected, as these regions are typically unmethylated and/or hypomethylated. Since a previous similar effort failed to identify the association of CDCA7 with CG-rich promoters [25], it was imperative to rigorously confirm this finding. To this end, we introduced exogenous HA-CDCA7 into CDCA7 knockout cells through lentiviral infection (Fig. 2A) to establish cell mixture that overexpressed HA-CDCA7 at a level about two- to three-folds higher than the endogenous CDCA7 level (Fig. 2B). These cells were named HA-CDCA7oe for HA-CDCA7 overexpression. As a result of overexpression, subsequent ChIP-seq analysis identified 8466 HA-CDCA7 peaks from HA-CDCA7oe cells (Fig. 2C), much more than the 2633 peaks detected from HA-CDCA7 cells. Consistent with the result from HA-CDCA7 cells, the peaks were again strongly enriched in the promoter regions (accounting for 65.03% of total peaks) (Fig. 2C). In addition, HA-CDCA7 peaks from HA-CDCA7 and HA-CDCA7oe cells shared a high degree of overlap (Fig. 2D). Much like ChIP-seq data from HA-CDCA7 cells, CDCA7 peaks in HA-CDCA7oe cells also strongly co-localized with H3K4me3 but not with H3K9me3 peaks (Fig. 2E). Binding motif analysis again revealed that HA-CDCA7 is associated with CG-rich sequences (Fig. 2F). We confirmed by ChIP-qPCR analysis that HA-CDCA7 was highly enriched at IRX2 and CD24 gene promoters in both HA-CDCA7oe and HA-CDCA7 cells, but not the distal exon 4 of IRX2 (Fig. 2G). In addition, HA-CDCA7 was also highly enriched at the α-satellite sequences in the pericentromeric heterochromatin region (Fig. 2G), consistent with a previous study [13]. Notably, G294V mutation severely impaired the association of CDCA7 with IRX2 and CD24 gene promoters and the α-satellite sequences as shown by ChIP-qPCR analysis (Fig. 2G). The strong enrichment of CDCA7 at the CpG island promoters and satellite regions in both HA-CDCA7 and HA-CDCA7oe cells, impaired association of CDCA7 G294V mutant, and co-occupancy of CDCA7 with H3K4me3 were clearly illustrated in the representative genome browser track in Fig. 2H and I. Confirmation of CDCA7 enrichment at the CG-rich promoters. (A) Schematic diagram showing the strategy for stable introduction of exogenous wild-type HA-CDCA7 into CDCA7 KO cells lines. (B) Western blotting analysis showing that the resulting HA-CDCA7oe cells expressed CDCA7 at a level two- to three-folds higher than that in control cells. (C) Pie chart showing the genomic annotation of 8466 HA-CDCA7 peaks derived from ChIP-seq data of HA-CDCA7oe cells. (D) The Venn diagrams showing the significant overlap of CDCA7 peaks from ChIP-seq data of HA-CDCA7oe cells and HA-CDCA7 cells. (E) Heatmaps showing the 8466 HA-CDCA7 peaks were significantly co-occupied with H3K4me3 peaks but not H3K9me3 peaks. (F) The list of top 5 CDCA7-binding motifs (ranked by P-value) identified by HOMER software from the ChIP-seq data of HA-CDCA7oe cells. (G) Verification of CDCA7 promoter association by ChIP-qPCR. CDCA7 binding was confirmed at the promoters of IRX2 and CD24 genes and α-satellite, but not at the distal exon of IRX2. Data are representative of three independent experiments and are shown as the mean ± SEM, Two-tailed t-test. ns means no significance; *P < .05; **P < .01; ***P < .001; ****P < .0001. (H) A representative ChIP-seq IGV browser view of CDCA7 binding in multiple promoter regions. The highlighted regions correspond to CDCA7 peaks derived from the HA-CDCA7oe ChIP-seq data. (I) A representative ChIP-seq IGV browser view of CDCA7 binding in satellite region. The highlighted regions correspond to CDCA7 peaks derived from the HA-CDCA7oe ChIP-seq data. (J) Heatmaps showing CUT&Tag CDCA7 peaks from both HA-CDCA7 and HA-CDCA7oe cells. (K) The Venn diagrams showing the remarkable overlap of CUT&Tag CDCA7 peaks from HA-CDCA7oe and HA-CDCA7 cells. (L) The Venn diagrams showing the significant overlap of CDCA7 peaks from CUT&Tag assay and ChIP-seq assay of HA-CDCA7oe cells. CUT&Tag is a technique developed in recent years for detecting protein–chromatin interaction, and it has the advantages of higher sensitivity and lower cell input than regular ChIP-seq [35]. Thus, HA antibody-based CUT&Tag experiments were also performed to further assess the genome-wide distribution of HA-CDCA7. This analysis identified 16 959 HA-CDCA7 peaks in HA-CDCA7oe cells and 9262 peaks in HA-CDCA7, and both sets of peaks were primarily enriched at the promoter regions (Fig. 2J) and overlapped remarkably (Fig. 2K). In addition, the ChIP-seq peaks and CUT&Tag peaks from HA-CDCA7oe cells overlapped extensively (Fig. 2L). Furthermore, both sets of CDCA7 CUT&Tag peaks were highly enriched in the promoter regions (Supplementary Fig. S3A and B), especially the CpG island-containing promoters (Supplementary Fig. S3C). The representative IGV browser visualization of CUT&Tag data showed that the peaks of CDCA7 co-localized with CpG islands and H3K4me3 in promoters (Supplementary Fig. S3D) and also existed in the α-satellites (Supplementary Fig. S3E). Comparison of ChIP-seq and CUT&Tag data revealed that CDCA7 is consistently highly enriched at the CpG islands (Supplementary Fig. S3F). Taken together, using different expression systems (endogenous and ectopic overexpression) and different techniques (ChIP-seq and CUT&Tag), we uncovered that CDCA7 primarily localizes to CG-rich promoter regions. In addition, we demonstrated that the G294V ICF-causing mutation abrogates this activity. CDCA7 binds both double stranded CG-rich DNA sequences and hemimethylated DNA in vitro Our genomic profiling analyses indicated that CDCA7 is enriched in genomic regions with a high proportion of CG dinucleotides. We next carried out in vitro electrophoretic mobility shift assays (EMSA) to test the characteristics of DNA sequences bound by CDCA7. Previous structural and biochemical studies showed that the C-terminal ZF domain (also known as the zf-CXXC_R1 domain) of CDCA7 functions as the DNA-binding domain and can specifically bind either hemimethylated DNA [23, 24] or a CpG dyad in unique non-B form DNA sequence motif [25]. Therefore, we expressed in and purified from E.coli GST-tagged CDCA7 C-terminal ZF (amino acids 252–371) (GST-CDCA7-ZF) and G294V mutant [GST-CDCA7-ZF (G294V)] (Fig. 3A and B). We first synthesized complementary oligos with a single methylated and unmethylated CpG (single CpG probe) (Fig. 3C) and followed by annealing to generate unmethylated, hemimethylated, and full methylated dsDNA probes. Subsequent EMSA revealed that the wild-type GST-CDCA7-ZF protein, but not GST-CDCA7-ZF (G294V) mutant, bound specifically the hemimethylated probe (Fig. 3D), consistent with the recent report that CDCA7 is a hemimethylated DNA sensor. CDCA7, but not its ICF mutant, binds both CG-rich sequences and hemimethylated DNA in vitro. (A) Schematic diagram of GST-CDCA7-ZF and GST-CDCA7-ZF(G294V) fusion proteins. (B) Coomassie blue staining showing recombinant GST-CDCA7-ZF and GST-CDCA7-ZF (G294V) proteins expressed and purified from E. coli. (C) The probes used for EMSA assay. The single CpG probe had a single CpG in the middle, whereas the CpG-rich probe was derived from the enriched motifs of CDCA7 ChIP-seq. The CG sites marked in red are the positions where DNA methylation modifications were introduced. (D) EMSA showing that CDCA7 but not G294V mutant specifically bound hemimethylated DNA probe. (E) EMSA showing that CDCA7 but not G294V mutant specifically bound unmethylated, double-stranded CpG-rich probe (left panel). Note that addition of a hemimethylated CpG enhanced the binding (middle panel), whereas addition of a full methylated CpG reduced the binding. (F) EMSA showing CDCA7-bound unmethylated, double-stranded CpG-rich probe in a protein-dose dependent manner. (G) A list of probes with an increasing number of CG sites (1, 3, 5, 7 CG, respectively). (H) EMSA result showing elevated CDCA7 binding along with increased number of CG sites. We then tested whether CDCA7 binds in vitro an unmethylated multiple CG-containing DNA sequence (Fig. 3C) that is enriched in our CDCA7 binding motif analysis (Fig. 1J). For comparison, we also added methylation to one of the five CGs (marked in red) (Fig. 3C). Via combinatorial annealing of methylated and unmethylated synthetic oligos, we generated double-stranded DNA probes of unmethylated, hemimethylated, and full methylated states. Subsequent EMSA results clearly demonstrated that GST-CDCA7-ZF but not GST-CDCA7-ZF(G294V) mutant bound the unmethylated CG-rich probe (Fig. 3E). Furthermore, the presence of a hemimethylated CpG enhanced the binding of GST-CDCA7-ZF, whereas the presence of a full methylated CpG had an opposite effect (Fig. 3E). We confirmed that GST-CDCA7-ZF but not GST-CDCA7-ZF(G294V) mutant protein bound the unmethylated multiple CG probe in a dose-dependent manner (Fig. 3F). We next investigated whether the number of CGs is the key factor critical for CDCA7 binding, as motifs with multiple CGs are highly enriched in CDCA7 peaks (Fig. 1J). To this end, we designed a series of probes with a gradually increased number of CG dinucleotides (1CG, 3CG, 5CG, 7CG) at the middle of each probe (Fig. 3G). The representative EMSA results in Fig. 3H showed that CDCA7 exhibited no binding to the 1CG probe, weak binding to the 3CG probe, and strong binding to probes containing 5- and 7CGs (Fig. 3H). As a negative control, GST-CDCA7-ZF(G294V) mutant showed no binding to any of these probes (Fig. 3H). It is noteworthy that we did not observe any binding of GST-CDCA7-ZF to the single-stranded probes, regardless of their methylation status or CG numbers (Supplementary Fig. S4A–C), indicating that CDCA7 lacks the single-stranded DNA-binding activity. Although not further characterized, binding of CpG sequences in vitro was also observed in a previous study [23]. Taken together, these in vitro experiments demonstrated that CDCA7 possesses two distinct DNA-binding modes: hemimethylated DNA and DNA sequences with multiple CpGs. Preferential binding of hemimethylated DNA allows CDCA7 switch to DNA methylation mode in S phase Having demonstrated that CDCA7 binds both unmethylated CG-rich sequences and hemimethylated single CpG, we wondered their corresponding biological function and how CDCA7 selects between these two distinct DNA-binding modes. Since newly synthesized DNA is hemimethylated and then converts to full methylation by DNA maintenance methylation, the level of hemimethylated DNA is expected to be high in S phase cells and low in interphase. We thus hypothesized that CDCA7 may primarily bind to CG-rich promoter regions in interphase cells and to hemimethylated DNA in S phase cells. Consistent with this idea, immunofluorescent (IF) staining using anti-HA antibody revealed that HA-tagged endogenous CDCA7 in HeLa cells displayed two distinct patterns of subcellular localization: a nuclear speckle pattern in ~20% cells and a diffuse nuclear staining in the rest ∼80% cells (Fig. 4A). Notably, the nuclear speckle pattern was virtually not observed in HA-CDCA7(G294V) mutant cells (Fig. 4A). By using a 15-min 5-ethynyl deoxyuridine (EdU, 10 μM) pulse labeling that allowed us to distinguish interphase cells (no EdU labeling), early S phase cells (diffuse nuclear labeling except nucleolus) and late S phase cells (nuclear foci labeling representing the late replicating pericentromeric heterochromatin regions) according to the characteristic EdU staining patterns, we confirmed that the ∼20 percent cells with CDCA7 nuclear speckles corresponded to the late replicating cells in which CDCA7 nuclear speckles co-localized with EdU labeled late replicating pericentromeric heterochromatin regions (Fig. 4B). CDCA7 switches to binding of hemimethylated DNA in S phase of cell cycle. (A, B) Co-immunostaining of HA-CDCA7 and nascent DNA revealed two distinct CDCA7 subcellular localization. HA-CDCA7 and HA-CDCA7(G294V) cells were subjected to 15 min EdU labeling and then processed for double immunostaining of HA and EdU. Shown as multiple cell images in panel (A) or single cell images in panel (B) that cells were classified as the late S phase, early S phase, and interphase, respectively. In multiple cell images, in ∼ 20% late S phase cells, 81 out of 100 cells had HA-CDCA7 focal staining that overlapped with EdU foci. In the rest ∼80% interphase and early S phase cells, HA-CDCA7 had a diffuse nuclear staining pattern. Notably, virtually no focal staining cells was observed for HA-CDCA7(G294V) in late S phase cells (4 out of 100). Also shown in panel (B) are the intensity profiles of fluorescence at marked positions for representative single cells. Scale bars in panel (A), 10 μm; Scale bars in panel (B), 2.5 μm. (C) Representative immunostaining images of HA-CDCA7 and HA-CDCA7(G294V) ectopically expressed in NIH3T3 cells and classified based on interphase, early S phase, and late S phase. HA-CDCA7 was seen as focal staining only in late S phase, and HA-CDCA7(G294V) failed to form foci in late S phase. Also shown are the intensity profiles of fluorescence at marked positions for representative single cells. Scale bars, 2.5 μm. (D) Inhibition of DNA maintenance methylation markedly elevated the percentage of cells with HA-CDCA7 focal staining. HA-CDCA7 and HA-CDCA7(G294V) was transfected into NIH3T3 cell and then treated with the DNMT1 inhibitor GSK3685032 for 24 h. Cells were then pulse labeled with EdU for 15 min and processed for double immunostaining of HA and EdU. Representative immunostaining images and numbers of cells with focal HA-CDCA7 staining pattern in interphase, early S phase, and late S phase were shown. Also shown are the intensity profiles of fluorescence at marked positions for representative single cells. Scale bars, 2.5 μm. (E, F) Competitive EMSA assay revealed CDCA7 preferentially bound hemimethylated. The quantity of hemimethylated DNA probe was fixed and the binding of GST-CDCA7 was competed with an increasing amount of unlabeled CG-rich probe (E). The reverse experiment was performed in panel (F), where CG-rich probe was fixed and challenged with increasing amount of hemimethylated DNA probe. Taking the advantage that constitutive pericentromeric heterochromatin regions in mouse fibroblast cells display as bright DAPI staining foci, we also examined along cell cycle the dynamic subcellular localization of HA-CDCA7 and G294V mutant ectopically expressed in NIH3T3 cells. As shown in Fig. 4C, ectopically expressed HA-CDCA7 also displayed two distinct staining patterns in NIH3T3 cells: nuclear speckles in late S phase that were co-localized perfectly with both EdU and DAPI foci (Fig. 4C, upper panel) and diffuse nuclear localization in both early S phase and interphase cells (Fig. 4C, middle and lower panels). However, the nuclear speckle staining pattern was not observed for HA-CDCA7(G294V) mutant (Fig. 4C). This dynamic change from a diffuse nuclear localization in interphase and early S phase to pericentromeric heterochromatin association in late S phase most likely reflects the preferential binding of CDCA7 to hemimethylated DNA generated during DNA replication, because pre-treatment of transfected cells with GSK3685032, a potent and specific DNMT1 inhibitor [36], strongly increased the ratio of cells with HA-CDCA7 nuclear speckle staining pattern, even in early S phase and interphase cells (Fig. 4D). Inhibition of DNMT1 with this inhibitor is expected to markedly increase hemimethylated DNA levels as a consequence of DNA maintenance methylation blockage [36, 37]. Taken together with our genome profiling data, we conclude that CDCA7 dynamic subcellular localization along cell cycle we and others have observed is most likely a direct reflection of its two distinct DNA-binding modes: binding of the CG-rich sequences in interphase cells and hemimethylated DNA in late S phase. The accumulation of CDCA7 in late replicating pericentromeric heterochromatin regions also implies that CDCA7 binds hemimethylated DNA with a higher affinity than CG-rich sequences, as suggested by DNA-binding results in Fig. 3E. To test this, we performed in vitro binding competition assays via EMSA. As shown in Fig. 4E, the unmethylated CG-rich probe (five CGs) was relatively less effective in competing the binding of GST-CDCA7 to hemimethylated probe, whereas the hemimethylated probe was more effective in blocking the binding of GST-CDCA7 to the CG-rich probe (Fig. 4F), supporting the idea that CDCA7 binds hemimethylated DNA with a higher affinity than unmethylated CG-rich DNA sequences. Loss of CDCA7 downregulates DNA methylation across the whole genome The blood cells from CDCA7 mutant ICF patients show aberrant DNA hypomethylation of CpG-poor genomic regions and pericentromeric satellite repeat sequences. A recent study demonstrated that the human CDCA7 G294V equivalent Cdca7G305V mutation in mice impairs global DNA methylation in multiple tissues and affects primarily DNA methylation of heterochromatin compartment in the spleen [22]. However, it has also been reported that loss of CDCA7 in mouse ES cells impaired DNA methylation in minor satellites but without reduction in global DNA methylation [13], raising the possibility that the critical role of CDCA7 in DNA methylation observed in human patients and ICF mutant mice is developmental dependent. Although loss of CDCA7 has been shown to downregulate DNA methylation in human cell lines, this has only been examined by low coverage microarrays or targeted bisulfite sequencing [10, 33]. Thus, to better understand the role of and mechanism by which CDCA7 regulates DNA methylation, we performed WGBS of both parental and CDCA7 knockout HeLa cells. Analysis of differentially methylated regions using the criteria of ≥25% methylation difference and P < .05 identified 558 091 hypomethylated regions and 23 689 hypermethylated sites in CDCA7 KO cells (Fig. 5A). In addition, the overall mCG methylation level decreased from 0.5877 in parental cells to 0.4859 in CDCA7 KO cells, representing a 17.3% reduction of global DNA methylation (Fig. 5B), which is highly consistent with the 17.7% decrease of DNA methylation in CDCA7 KO cells measured by HPLC (Fig. 1D). DNA methylation status in CDCA7 KO and parental cells both exhibited a bimodal distribution, with a decreased full-methylated CpG and increased unmethylated CpG fraction in CDCA7 KO cells (Fig. 5C). Reduction in DNA methylation did not have an obvious chromosome bias, and all 23 chromosomes showed an overall reduction of ~15%–20% (Fig. 5D). Genomic element analysis revealed that hypomethylated regions tend to enrich in intergenic regions (42.3%) and intronic regions (42.9%) (Fig. 5E) and reduction of DNA methylation in CDCA7 KO cells occurred across different genomic regions, with the intergenic and promoter regions exhibited the relatively severer reduction (Fig. 5F). Loss of CDCA7 results in genome-wide reduction of DNA methylation. (A) A volcano plot illustrating the numbers of hypomethylated and hypermethylated regions in CDCA7 KO cells relative to control cells. (B) A bar graph showing the global DNA methylation levels in control and CDCA7 KO cells. (C) Bimodal mCpG distribution analysis showing the CpG methylation distribution patterns of control and CDCA7 KO cells. (D) A bar graph showing the changes of DNA methylation levels across different chromosomes. (E) Pie chart summarizing the distribution of hypomethylated CpGs across different genomic regions. (F) Boxplots showing the effect of CDCA7 KO on DNA methylation of different genomic features. (G) Profile plots showing methylation levels ± 3 kb around satellites region between control and CDCA7 KO cells. (H) Boxplots showing DNA methylation levels of different repetitive sequence types in control and CDCA7 KO cells. (I) DNA methylation levels of HMDs and PMDs between control and CDCA7 KO cells. The genome information of PMDs/HMDs derived from UCSC database. (J) Boxplots illustrating DNA methylation levels at CDCA7 binding regions versus other regions between control and CDCA7 KO cells. (K) Enrichment of CDCA7 at low DNA methylation CpG islands. The CpG islands were divided into high DNA methylation and low DNA methylation groups and the relative CDCA7 enrichment was profiled. (L) A representative WGBS IGV browser view of control and CDCA7 KO cells in the pericentromeric region. Satellite regions were highlighted. The HMD/PMD tracks were derived from the UCSC database. Previous studies have reported that pericentromeric satellite repeats exhibit significant hypomethylation in ICF patients [1, 10]. Therefore, we specifically analyzed the effect of CDCA7 KO on DNA methylation of various satellite repeats. Compared with the adjacent regions, satellite repeats showed a modestly greater reduction of DNA methylation (Fig. 5G). In addition, reduction of DNA methylation was observed across all repetitive sequences including LINEs, LTRs, retrotransposons, and SINEs (Fig. 5H). However, loss of CDCA7 resulted in a more substantial reduction of DNA methylation in PMDs than HMDs (Fig. 5I), consistent with a previous study [17]. Furthermore, although both CDCA7-enriched peak regions and non-enriched regions exhibited a similar level of DNA methylation reduction (Fig. 5J), CDCA7-enriched peak regions are strongly associated with hypomethylated CpG islands (Fig. 5K). A representative IGV browser visualization showed that DNA methylation was reduced more in PMD and pericentromeric satellite repeat regions and less affected in HMD regions (Fig. 5L), consistent with the previous study in mouse spleen tissue. It is noteworthy that the extent of DNA methylation loss (17.3%) in CDCA7 KO is highly consistent with that we reported previously in LSH KO cells [16], further supporting the notion that CDCA7 and LSH act in the same cell autonomous pathway to promote DNA methylation. CDCA7 is required for LSH chromatin localization LSH has been reported to regulate transcription and to be enriched in the promoter regions [38, 39]. Having established that CDCA7 is highly enriched at the promoter regions, we next tested whether CDCA7 targets LSH to genomic DNA, especially the promoter regions. To this end, we performed ChIP-seq for LSH in both parental and CDCA7 KO cells. We found that among 19 060 LSH peaks identified in parental HeLa cells, 57.5% was distributed within 1 kb of TSS, highlighting that, like CDCA7, LSH is also highly enriched in the promoter regions (Fig. 6A). More than 50% CDCA7 peaks detected by ChIP-seq or CUT&Tag were overlapped with LSH peaks (Supplementary Fig. S5A), indicating that LSH and CDCA7 extensively co-occupy genomic DNA. Notably, LSH peak intensity was drastically reduced in CDCA7 KO cells, demonstrating that CDCA7 is required for LSH genome-wide association (Fig. 6B). LSH peaks also showed strong co-occupancy with H3K4me3 but not H3K9me3 (Fig. 6B), consistent with previous studies for LSH [38–40]. Further analysis revealed that LSH peaks could be divided into three clusters, with 11 446 peaks significantly decreased, 6319 peaks modestly decreased and/or relatively unaltered, and 1295 peaks increased upon loss of CDCA7 (Fig. 6C). Notably, diminished LSH association not only occurred at the promoters with CDCA7 peaks based on CUT&Tag analysis in HA-CDCA7oe cells, but also observed at the promoters without obvious CDCA7 peaks (Fig. 6D). This suggested that the CDCA7 peak number is most likely underestimated even in our CUT&Tag assay with HA-CDCA7oe cells, presumably because HA antibody is not as efficient as LSH antibody in capturing the corresponding target protein–chromatin complexes. The correlation between CDCA7 and LSH peaks and strong CDCA7 dependency were clearly illustrated with a representative genome browser track in Fig. 6E. Similar to CDCA7, genome browser visualization revealed that LSH was also enriched at satellite regions and CDCA7 KO impaired LSH enrichment in these regions (Fig. 6F). We validated that, much like CDCA7 (Fig. 2G), LSH was strongly enriched at two promoters and one satellite region but not in the distal exon of IRX2 gene, and the preferential association of LSH with promoter and satellite sequences were severely reduced in CDCA7 KO cells by ChIP-qPCR assay (Fig. 6G). CDCA7 is required for LSH chromatin localization. (A) A Pie chart showing the genomic annotation of LSH peaks in control HeLa cells. (B) Heatmaps showing LSH ChIP-seq signals from control and CDCA7 KO cells. Also shown were H3K4me3 and H3K9me3 peaks around 19 060 LSH peaks (±3 kb). (C) Heatmaps showing LSH association-decreased, relatively unchanged, and increased peaks upon loss of CDCA7. (D) Heatmaps showing that LSH peaks were downregulated more extensively when co-occupied with CDCA7 peaks. (E) The representative ChIP-seq IGV browser view of the LSH peaks in control and CDCA7 KO cells. The promoter regions with LSH and CDCA7 were highlighted. (F) The representative satellite ChIP-seq IGV browser view of the LSH peaks in control and CDCA7 KO cells. The satellite regions were highlighted. (G) Confirmation of CDCA7-dependent LSH association by ChIP-qPCR. Data are representative of three independent experiments and are shown as the mean ± SEM, Two-tailed t-test. ns means no significance; *P < .05; **P < .01; ***P < .001; ****P < .0001. (H) The schematic diagram illustrating the LacO–LacI protein tethering system in U2OS cells. (I) Immunofluorescence assays showing the recruitment of LSH by both CDCA7 and CDCA7(G294V) mutant protein in cell through LacO–LacI protein tethering assay. To test further whether CDCA7 can target LSH to chromatin in cells, we employed a tethering assay based on the binding of LacI-GFP fusion of CDCA7 protein (LacI-GFP-CDCA7) to an integrated LacO array in U2OS cells (Fig. 6H). We observed that LacI-GFP-CDCA7 but not LacI-GFP could recruit co-expressed HA-tagged LSH (Fig. 6I). Consistent with our co-IP assay results (Supplementary Fig. S2A), we observed that the CDCA7 G294V mutant could also recruit LSH (Fig. 6I), further confirming the mutation did not affect the interaction between CDCA7 and LSH. Also consistent with our co-IP results (Supplementary Fig. S2B), we observed that CDCA7 can recruit UHRF1 but not DNMT1, DNMT3A, and DNMT3B (Supplementary Fig. S5B–E). Thus, the in vivo tethering assay nicely consolidated others [14, 23] and our observation that CDCA7 specifically interacts with and targets LSH to chromatin and demonstrate additionally that CDCA7 also interacts with UHRF1. We reported previously that LSH is associated with late-replicating pericentromeric chromatin domains in late S phase [16]. As reported by others [23], the association of LSH with late-replicating pericentromeric chromatin is also CDCA7 dependent, as it was severely impaired in CDCA7(G294V) mutant cells (Supplementary Fig. S6A and B). However, the association of CDCA7 with genomic DNA appeared to be LSH independent, because knockdown of LSH did not significantly affect the CDCA7 binding of IRX2 and CD24 (cluster of differentiation 24) promoters as well as α-satellite (Supplementary Fig. S7). CDCA7 and LSH regulate transcription beyond their function in DNA methylation Our findings that both CDCA7 and LSH are enriched at promoter regions raised the question whether CDCA7 and LSH may directly regulate transcription of their associated genes independent of their regulation of DNA methylation. To distinguish the direct role as a transcription factor and/or transcription regulator from their indirect regulation of gene expression via DNA methylation, we decided to compare the transcriptomes of long term CDCA7 or LSH KO cells with transient CDCA7 or LSH knockdown cells by RNA-seq. Two CDCA7 and LSH KO cell lines were generated from HeLa cells by CRISPR–Cas9 technique aforementioned. WB analysis confirmed the successful KO of CDCA7 or LSH in corresponding cell lines (Fig. 7A). Analysis of DNA methylation revealed 16.4%–17.8% reduction of DNA methylation in CDCA7 KO cell lines and 20.4%–22.5% reduction in LSH KO cell lines (Supplementary Fig. S8A). We then transiently knocked down CDCA7 or LSH by treatment with two distinct CDCA7 or LSH-specific siRNAs for 24 h. The successful knockdown of either CDCA7 or LSH protein levels was confirmed by WB analysis (Fig. 7B). However, HPLC analysis of global DNA methylation level revealed that, unlike the cases of long-term CDCA7 or LSH KO cells, transient knockdown of CDCA7 or LSH did not affect the level of global DNA methylation (Supplementary Fig. S8B). CDCA7 and LSH can directly regulate transcription via binding CG-rich sequences. (A) Western blotting verification of the absence of CDCA7 and LSH in respective long-term CDCA7 and LSH KO cells. (B) Western blotting analysis of HeLa cells treated transiently with siCDCA7 or siLSH for 24 h. “-”, blank control; “siNC”, scramble siRNA control. Two distinct siRNAs were used for each gene knockdown. The volcano plots illustrating the numbers of differentially expressed genes between control and CDCA7 KO cells (C) and LSH KO cells (D). (E) Scatterplots showing the correlation of differentially expressed genes between CDCA7 KO and LSH KO cells. Person’s r- and P-value of the correlations were indicated. The volcano plots illustrating the numbers of differentially expressed genes between control and transient CDCA7 knockdown cells (F) and LSH knockdown cells (G). (H) Scatterplots showing the correlation of differentially expressed genes between transient knockdown of CDCA7 and LSH cells. Person’s r- and P-value of the correlations were indicated. (I) Scatterplots showing the correlation of differentially expressed genes between CDCA7 KO and CDCA7 transient knockdown cells. Person’s r- and P-value of the correlations were indicated. (J) Heatmap of differentially expressed genes between CDCA7 KO and CDCA7 transient knockdown cells (upregulated, red; downregulated, blue). (K) Heatmap of differentially expressed neuron crest-related genes between CDCA7 KO and CDCA7 transient knockdown cells. (L) Pie charts show the numbers of CDCA7-bound and -unbound genes among the upregulated (right) and downregulated (left) genes induced by siCDCA7. Genes with CDCA7 peaks identified by CUT&Tag in HA-CDCA7 (e) cells were defined as CDCA7-bound genes. Conversely, they were defined as CDCA7-unbound genes. (M) Luciferase reporter assay showing that CDCA7 and LSH interdependently activated CD24 promoter. On the left panel, the reporter was co-transfected with an increasing amount of FLAG-LSH encoding plasmid and HA-CDCA7 or mutants as indicated. On the right panel, the reporter was co-transfected with an increasing amount of HA-CDCA7 encoding plasmid and FLAG-LSH mutants as indicated. The luciferase activities were representative of three independent experiments and shown as the mean ± SEM, Two-tailed t-test. ns means no significance; *P < .05; **P < .01; ***P < .001; ****P < .0001. The expression of transfected FLAG-LSH and HA-CDCA7 or mutants was verified by western blotting analysis (lower panel). RNAs were prepared from above four CDCA7 and LSH KO cell lines and transient knockdown cells, and subjected to RNA-seq analysis. Principal component analysis clustering confirmed that each group of RNA-seq samples exhibited excellent reproducibility and high sequencing quality (Supplementary Fig. S8C). Volcano plots of RNA-seq data showed 2299 upregulated and 1887 downregulated genes in CDCA7 KO cells (Fig. 7C), and 1556 upregulated and 1082 downregulated genes in LSH KO cells, with a cutoff of ≥1.414-fold change and P-value of .05 (Fig. 7D). Furthermore, a remarkable degree of correlation (R = 0.95 for upregulated genes and R = 0.70 for downregulated genes) was observed between differentially regulated genes in CDCA7 and LSH KO cells (Fig. 7E), and more genes were upregulated than downregulated (Fig. 7C and D), consistent with the observation that CDCA7 and LSH KO resulted in global DNA hypomethylation and that DNA methylation in general represses transcription. Interestingly, in the contrary to long-term CDCA7 and LSH KO, both CDCA7 and LSH transient knockdown resulted in more downregulated genes than upregulated genes, specifically with 1707 downregulated versus 955 upregulated genes for CDCA7 knockdown (Fig. 7F) and 1432 downregulated versus 820 upregulated genes for LSH knockdown cells (Fig. 7G). Both upregulated and downregulated gene were positively correlated between CDCA7 and LSH transient knockdown cells (Fig. 7H). This positive correlation in transcription is consistent with our ChIP-seq data that LSH associates with the gene promoters in a CDCA7-dependent manner. Taken together, these results indicate that CDCA7 and LSH can also directly regulate transcription independent of their function in DNA methylation. Notably, transient CDCA7 knockdown also share a common set of differentially regulated genes with long-term CDCA7 KO, especially for upregulated genes (Fig. 7I and J), suggesting that the effect of long-term loss of CDCA7 on transcription is likely a consequence of combinatorial loss of direct transcription regulation function and aberrant DNA hypomethylation. It has been reported that ICF3 patients exhibit significant neurodevelopmental defects [6]. We found that a set of neural crest genes showed consistent expression changes following short-term CDCA7 knockdown and long-term CDCA7 KO (Fig. 7K). This suggests that a subset of neurodevelopmental genes may be directly regulated by CDCA7 independent of DNA methylation. To investigate to what extent the differentially expressed genes upon transient knockdown of CDCA7 are likely directly regulated by CDCA7, we analyzed whether the affected genes contain CDCA7 peaks in their promoters based on the CUT&Tag data from HA-CDCA7 cells. The results showed that 907 downregulated genes and 562 upregulated genes had CDCA7 peaks in their promoters (Fig. 7L). Among the downregulated genes with CDCA7 peak, CD24 plays an important role in neurogenesis and immune checkpoints [41, 42]. We confirmed by RT-qPCR that CD24 transcriptional levels were significantly downregulated in LSH and CDCA7 KO cells and in cells with transient CDCA7/LSH knockdown, relative to WT controls (Supplementary Fig. S8D). We therefore chosen the promoter of CD24 to test whether it is directly regulated by CDCA7. We constructed a luciferase reporter by inserting a 3000-bp fragment containing the CDCA7 promoter upstream of the luciferase gene within the pGL-3 reporter vector. Subsequent reporter assays showed that expression of either wildtype CDCA7 or R274C and G294V mutants did not significantly elevate the luciferase activity (Supplementary Fig. S8E). However, co-expression of wildtype CDCA7 but not R274C or G294V mutant with LSH enhanced the luciferase activity in a dose-dependent manner (Fig. 7M, left panel), whereas co-expression of CDCA7 with the LSH mutants either deficient of ATPase activity [LSH(ΔATP)] or with the deletion of DEAH domain failed to activate the CD24 luciferase reporter (Fig. 7M, right panel). The expression of transfected wildtype and mutant CDCA7 and LSH was verified by WB analysis. These results suggest that CDCA7 and LSH act in an inter-dependent manner to activate CD24 gene transcription. Discussion In this study, we unravel two distinct DNA-binding modes for CDCA7: CG-rich motifs and hemimethylated CpG. By knocking in an HA tag or/and an ICF syndrome-causing G294V mutation into the endogenous CDCA7 gene (Fig. 1A), we were able to carry out high quality ChIP-seq experiments using anti-HA antibody to profile the DNA-binding landscape of both wildtype and mutant CDCA7. These analyses revealed that the wildtype but not G294V mutant CDCA7 associates mainly with the promoter regions (Fig. 1E–I) and CDCA7 peaks are highly enriched with CG-rich motifs (Fig. 1J). We confirmed this finding by establishing and analyzing cell lines that stably overexpressed HA-tagged wildtype or G294V mutant via ChIP-seq (Fig. 2A–H) and via more sensitive CUT&Tag assay (Fig. 2J–L). Our genome-wide profiling data also validated the previous observation that CDCA7 is associated with pericentromeric satellites [13] (Fig. 2G and I). Importantly, we confirmed by in vitro binding assay that CDCA7 can bind both CG-rich motifs and hemimethylated DNA (Fig. 3). The CDCA7(G294V) mutant is not only defective in binding hemimethylated DNA as reported previously [22–24] (Fig. 3D) but also is impaired in binding of unmethylated CG-rich sequences (Fig. 3E–H). Thus, our data indicate that CDCA7 is not only a hemimethylated DNA sensor but also possesses an intrinsic DNA-binding activity for CG-rich DNA sequences that are present mainly in the CpG island type of promoters. While a recent study reported that CDCA7 binds a CpG dyad in the context of a non-B form DNA [25], our data demonstrate for the first time that CDCA7 can bind CG-rich B form DNA with three or more CG sites (Fig. 3G and H). In this regard, a previous study also detected a weak binding of CG sequence by CDCA7 in vitro [23]. Furthermore, it was shown that the ATPase activity of CDCA7–LSH (HELLS) complex is enhanced in the presence of unmethylated CpG [23], thus supporting the binding of unmethylated CG-rich sequences is functionally relevant. Although the mechanism by which CDCA7 binds CG-rich sequences remains to be revealed by structural analysis, our data indicate that the same DNA-binding domain is involved in binding of hemimethylated CpG and multiple CGs, as G294V mutation impairs the binding activity for both types of DNA. Consistent with the aforementioned two distinct DNA-binding modes, we and others have observed that CDCA7 exhibits a notable change of subcellular localization along cell cycle [23, 24] (Fig. 4A–C). In interphase cells, CDCA7 is observed as a diffuse nuclear localization (Fig. 4B and C), whereas CDCA7 becomes focal pericentromeric heterochromatin localization in the late S phase (Fig. 4B and C). This change of subcellular localization along cell cycle most likely reflects a change of CDCA7 DNA-binding mode: from binding of CG-rich motifs in interphase to hemimethylated CpG in S phase. This switch is consistent with our in vitro binding results that CDCA7 binds with a relatively higher affinity to hemimethylated DNA than to unmethylated CG-rich sequences (Fig. 3E and Fig. 4E and F), and is supported by our observation that inhibition of DNMT1 activity by GSK 3685032 resulted in a persistent CDCA7 focal staining pattern even in the interphase (Fig. 4D). Thus, despite of the existence of two distinct DNA-binding modes, a higher affinity for hemimethylated DNA presumably allows CDCA7 and its associated LSH (see next) to switch to DNA maintenance methylation function during the S phase of cell cycle. Although the ICF syndrome-causing CDCA7 mutants display a diffuse nuclear staining indistinguishable from the wildtype CDCA7 in interphase and early S phase, they lost the characteristic focal pericentromeric heterochromatin localization in the late S phase due to the loss of DNA-binding activity for both hemimethylated DNA and CG-rich sequences. Our whole genome DNA methylation analysis supports CDCA7 as a positive regulator of DNA maintenance methylation. Because it was long observed that ICF syndrome patients have the characteristic DNA hypomethylation at (peri)centromeric repeats, much of the previous studies on CDCA7 and LSH have been focused on DNA hypomethylation in (peri)centromeric repeats and heterochromatin regions [33]. A recent CDCA7 mouse model study revealed that the equivalent G294V mutation in mice recapitulates the disease phenotype and leads to the formation of large, aberrant hypomethylated domains overlapping with the heterochromatin compartment [22]. However, as these studies were based on either patient-derived blood cells or mouse models that had underwent developmental process, some of the phenotype and/or DNA methylation alterations may not be the direct function of CDCA7 but the consequence of affected developmental process. Our knockout- and knockin-based cell models bypassed the development process and thus were ideal for assess of CDCA7 direct function in DNA methylation. We confirmed that loss of CDCA7 results in global DNA hypomethylation in an extent similar to loss of LSH (Fig. 5B and Supplementary Fig. S8A). Furthermore, inactivation of CDCA7 results in not only DNA hypomethylation in pericentromeric satellites and CpG poor regions but also many other sequences, although to a less extent (Fig. 5C–L). Our results are consistent with the recent reports [23, 24] that CDCA7 is a hemimethylated sensor and suggest that by recognizing and binding hemimethylated CpGs produced upon DNA replication, CDCA7 recruits LSH to facilitate DNA maintenance methylation by UHRF1/DNMT1 axis [16, 17]. The co-localization of CDCA7 and LSH with heterochromatin domains in the late S phase is a visual reflection of CDCA7 binding of hemimethylated DNA, as this co-localization can be persistent to interphase cells if DNA maintenance methylation is blocked (Fig. 4D). It is noteworthy that an interaction between CDCA7 and UHRF1 was detected by co-IP (Supplementary Fig. S2B) and tethering assay (Supplementary Fig. S5B), suggesting that the interaction with UHRF1 may also contribute to CDCA7/LSH involvement in DNA maintenance methylation. Our study also supports that CDCA7 and LSH have DNA methylation-independent transcription regulation function [27, 38, 43, 44]. By siRNA-based transient knockdown of either CDCA7 or LSH that had no observable effect on global DNA methylation (Supplementary Fig. S8B), we dissected the direct transcription regulation function of CDCA7 and LSH from their control of gene expression via DNA methylation (Fig. 7). In contrary to the observed prevailing transcriptional activation in long-term CDCA7 and LSH knockout cells (Fig. 7C and D), transient knockdown of CDCA7 and LSH results in more downregulated genes than upregulated genes (Fig. 7F and G). Thus, CDCA7 and LSH have transcriptional regulation function beyond their role in DNA methylation. In support of this idea, we demonstrated that CDCA7 and LSH can activate the CD24 gene promoter in a reporter assay (Fig. 7M). It is noteworthy that the robust activation was observed only when both wildtype CDCA7 and LSH were co-expressed, and it was not observed when the ICF-causing CDCA7 mutants or LSH deficient of ATPase activity were expressed (Fig. 7M). Thus, our data suggest that CDCA7 and LSH can extensively regulate transcription by binding of CG-rich promoters and activate transcription in a chromatin remodeler activity-dependent manner, consistent with the notion that CDCA7 and LSH function as a bipartite chromatin remodeler [14]. Although how CDCA7/LSH complex regulates transcription remains to be elucidated, a recent study reported that LSH regulates transcription in T-cell lymphomas by reducing unscheduled R-loops and by facilitating RNA polymerase II progression [45]. Taken together, we propose that CDCA7/LSH can regulate transcription in interphase cells by virtue of their ability to bind CG-rich motifs (see the “Graphic abstract” section). In interphase, CDCA7 binds predominantly the CG-rich promoters and recruits LSH to regulate target gene expression (see the “Graphic abstract” section, left panel) in an LSH chromatin remodeling activity-dependent manner. In S phase, CDCA7 binds primarily hemimethylated DNA and recruits LSH to facilitate DNA maintenance methylation by UHRF1/DNMT1 axis also via LSH chromatin remodeling activity (see the “Graphic abstract” section, right panel). CDCA7 and LSH ICF syndrome mutations cause not only DNA hypomethylation, but also result in dysregulation of their direct target genes. The direct transcription regulation function by CDCA7 and LSH may contribute to the phenotypic divergence between type I ICF syndrome caused by DNMT3B mutation and types Ⅲ and Ⅳ ICF syndrome caused by CDCA7 and LSH mutation, respectively. While currently identified ICF-causing mutations all lead to hypomethylated pericentromeric regions and centromeric instability, they vary in clinical phenotypes, with patients negative for DNMT3B mutations showing greater incidence of intellectual disability whereas patients with mutations in DNMT3B exhibit a more severe immunodeficiency. With our findings that CDCA7 and LSH share DNA methylation-independent transcriptional function and regulate the expression of a subset of neural crest genes, it is formally possible that the transcriptional defects caused by CDCA7 or LSH mutations also contribute to clinical presentation unique to ICF patients bearing CDCA7 or LSH mutation. Supplementary Material Acknowledgements We thank members of the Wong’s lab for valuable discussions and technical assistance. Author contributions: Fenghua Chen (Conceptualization [equal], Data curation [lead], Formal analysis [lead], Investigation [lead], Methodology [lead], Resources [lead], Validation [lead], Visualization [equal], Writing – original draft [equal], Writing – review & editing [equal]), Meilin Sun (Data curation [lead], Investigation [equal], Software [lead], Visualization [lead], Writing – review & editing [supporting]), Mingzhi Chang (Data curation [equal], Investigation [equal], Validation [equal], Writing – review & editing [equal]), Zhongye Dai (Investigation [equal], Methodology [equal], Supervision [equal], Validation [equal]), Zhaosu Chen (Formal analysis [supporting], Investigation [supporting], Methodology [equal], Validation [equal]), Jialun Li (Conceptualization [supporting], Investigation [supporting], Methodology [equal], Validation [equal]), Yu Duan (Methodology [equal], Validation [equal]), Jiwen Li (Conceptualization [equal], Funding acquisition [equal], Project administration [equal], Resources [equal], Supervision [equal]), Xiongwen Cao (Conceptualization [supporting], Methodology [equal], Resources [supporting], Supervision [equal]), Yuanyong Huang (Conceptualization [lead], Formal analysis [equal], Methodology [equal], Project administration [equal], Supervision [lead], Writing – review & editing [equal]), and Jiemin Wong (Conceptualization [lead], Funding acquisition [lead], Investigation [equal], Project administration [lead], Resources [lead], Writing – original draft [lead], Writing – review & editing [lead]) Contributor Information Fenghua Chen, Shanghai Key Laboratory of Regulatory Biology, Fengxian District Central Hospital-ECNU Joint Center of Translational Medicine, Institute of Biomedical Sciences and School of Life Sciences, East China Normal University, Shanghai 200241, China. Meilin Sun, Shanghai Key Laboratory of Regulatory Biology, Fengxian District Central Hospital-ECNU Joint Center of Translational Medicine, Institute of Biomedical Sciences and School of Life Sciences, East China Normal University, Shanghai 200241, China. Mingzhi Chang, Shanghai Key Laboratory of Regulatory Biology, Fengxian District Central Hospital-ECNU Joint Center of Translational Medicine, Institute of Biomedical Sciences and School of Life Sciences, East China Normal University, Shanghai 200241, China. Zhongye Dai, Shanghai Key Laboratory of Regulatory Biology, Fengxian District Central Hospital-ECNU Joint Center of Translational Medicine, Institute of Biomedical Sciences and School of Life Sciences, East China Normal University, Shanghai 200241, China. Zhaosu Chen, Shanghai Key Laboratory of Regulatory Biology, Fengxian District Central Hospital-ECNU Joint Center of Translational Medicine, Institute of Biomedical Sciences and School of Life Sciences, East China Normal University, Shanghai 200241, China. Jialun Li, Shanghai Key Laboratory of Regulatory Biology, Fengxian District Central Hospital-ECNU Joint Center of Translational Medicine, Institute of Biomedical Sciences and School of Life Sciences, East China Normal University, Shanghai 200241, China. Yu Duan, Shanghai Key Laboratory of Regulatory Biology, Fengxian District Central Hospital-ECNU Joint Center of Translational Medicine, Institute of Biomedical Sciences and School of Life Sciences, East China Normal University, Shanghai 200241, China. Jiwen Li, Shanghai Key Laboratory of Regulatory Biology, Fengxian District Central Hospital-ECNU Joint Center of Translational Medicine, Institute of Biomedical Sciences and School of Life Sciences, East China Normal University, Shanghai 200241, China. Xiongwen Cao, Shanghai Key Laboratory of Regulatory Biology, Fengxian District Central Hospital-ECNU Joint Center of Translational Medicine, Institute of Biomedical Sciences and School of Life Sciences, East China Normal University, Shanghai 200241, China. Yuanyong Huang, Shanghai Key Laboratory of Regulatory Biology, Fengxian District Central Hospital-ECNU Joint Center of Translational Medicine, Institute of Biomedical Sciences and School of Life Sciences, East China Normal University, Shanghai 200241, China. Jiemin Wong, Shanghai Key Laboratory of Regulatory Biology, Fengxian District Central Hospital-ECNU Joint Center of Translational Medicine, Institute of Biomedical Sciences and School of Life Sciences, East China Normal University, Shanghai 200241, China; Joint Center for Translational Medicine, Shanghai Fifth People’s Hospital, Fudan University and School of Life Science, East China Normal University, Shanghai 200241, China. Supplementary data Supplementary data is available at NAR online. Conflict of interest None declared. Funding This study was supported by grants from the National Natural Science Foundation of China (32330020 to J.W.) and National Key Research and Development Program of China (2023YFA1800402 to J.W. and Y.H., and 2023YFA1800401 to J.L.). This work was also supported by the Instruments Sharing Platform of School of Life Sciences, East China Normal University. Funding to pay the Open Access publication charges for this article was provided by grants from the National Natural Science Foundation of China (32330020 to J.W.) and National Key Research and Development Program of China (2023YFA1800402 to J.W. and Y.H., and 2023YFA1800401 to J.L.). Data availability The raw sequence data reported in this paper have been deposited in the Genome Sequence Archive in National Genomics Data Center, China National Center for Bioinformation/Beijing Institute of Genomics, Chinese Academy of Sciences (GSA-Human: HRA015656) that are publicly accessible at https://ngdc.cncb.ac.cn/gsa-human. References Associated Data Supplementary Materials Data Availability Statement The raw sequence data reported in this paper have been deposited in the Genome Sequence Archive in National Genomics Data Center, China National Center for Bioinformation/Beijing Institute of Genomics, Chinese Academy of Sciences (GSA-Human: HRA015656) that are publicly accessible at https://ngdc.cncb.ac.cn/gsa-human.