ZBTB24 is a conserved multifaceted transcription factor at genes and centromeres that governs the DNA methylation state and expression of satellite repeats

Abstract

Since its discovery as a causative gene of the Immunodeficiency with Centromeric instability and Facial anomalies syndrome, ZBTB24 has emerged as a key player in DNA methylation, immunity and development. By extensively analyzing ZBTB24 genomic functions in ICF-relevant mouse and human cellular models, we document here its multiple facets as a transcription factor, with key roles in immune response-related genes expression and also in early embryonic development. Using a constitutive Zbtb24 ICF-like mutant and an auxin-inducible degron system in mouse embryonic stem cells, we showed that ZBTB24 is recruited to centromeric satellite DNA where it is required to establish and maintain the correct DNA methylation patterns through the recruitment of DNMT3B. The ability of ZBTB24 to occupy centromeric satellite DNA is conserved in human cells. Together, our results unveiled an essential and underappreciated role for ZBTB24 at mouse and human centromeric satellite repeat arrays by controlling their DNA methylation and transcription status.

ICF syndrome, ZBTB24, DNA methylation, centromeric repeats

Introduction

The study of diseases with abnormal DNA methylation (DNAme) landscapes emphasized the key role of this epigenetic mark in the maintenance of genomic stability [1–3]. In particular, a growing body of evidence coming from studies of the etiology of these diseases, including but not limited to cancer, has suggested the existence of a functional link between CpG methylation and centromere integrity. At centromeres, being chromosomal domains specialized in faithful segregation of the genetic material during cell divisions, which assemble on large-scale arrays of tandem repeated sequences, this functional link is thought to have a major contribution to genome stability [4, 5].

The inherited disorder called the ICF syndrome (Immunodeficiency with Centromeric instability and Facial anomalies syndrome; MIM #242860) is one of the most emblematic pathological contexts that has underlined the connection between DNAme and centromere integrity. This syndrome is a very rare autosomal recessive disorder, genetically heterogeneous, which is mostly characterized by primary immunodeficiency [6, 7]. Although displaying varying degrees of severity, clinical manifestations also include mild facial anomalies, intellectual disability, congenital malformations and developmental delay [8]. The ICF syndrome has been described as a disease affecting DNAme at constitutive heterochromatin domains assembled on juxtacentromeric satellite repeats: Satellites type II (Sat II) of chromosomes 1 & 16 and, to a lesser extent, Satellites type III (Sat III) of chromosome 9. DNA hypomethylation of pericentromeric satellite DNA is an invariant hallmark of the ICF patients’ cells, which leads to juxtacentromeric heterochromatin decondensation, chromosomal breaks and rearrangements in the vicinity of centromeres, a molecular feature used to establish diagnosis [9]. In contrast, additional hypomethylation of centromeric alpha satellite DNA is a heterogeneous trait among ICF patients, thus reflecting the genetic heterogeneity of the syndrome [10].

The ICF syndrome is caused by homozygote or heterozygote composite mutations in the DNA methyltransferase 3B gene (DNMT3B, ICF1 MIM#242860) [11, 12], zinc finger and BTB domain containing 24 gene (ZBTB24, ICF2 MIM#614069) [13], cell division cycle-associated 7 gene (CDCA7, ICF3 MIM#616910) [14], or helicase lymphoid specific gene (HELLS, ICF4 MIM#616911) [14]. Strikingly, only DNMT3B has a DNA methyltransferase (DNMT) activity whereas the other factors are transcription factors (TFs) (ZBTB24 and CDCA7) with poorly known functions or chromatin remodeler (HELLS/LSH), questioning their role and respective interplay in DNAme pathways. Our comparative methylome analysis of ICF patients revealed a striking similarity between genome-wide DNAme profiles of ICF2-4 patients, which clearly distinguished them from ICF1 patients [15, 16]. Remarkably, ICF2-4 patient subtypes also exhibit hypomethylation of centromeric satellite repeats in addition to that of the juxtacentromeric satellites common to all ICF subtypes [10, 15]. Thus, the ICF syndrome represents a unique pathological context whose study has allowed to identify unanticipated molecular actors and pathways linking DNAme to centromere integrity.

In contrast to DNMT3B, the functional connections between ZBTB24, CDCA7 and HELLS with the DNA methylation machinery are not straightforward but begin to be clarified. Indeed, ZBTB24 was shown to be a transcriptional activator of the CDCA7 gene [17], whereas CDCA7 is required for HELLS to associate with chromatin and stimulates its remodeling activity [18]. As supported by recent studies, a CDCA7/HELLS complex contributes to DNAme maintenance by facilitating the recruitment of the DNMT1/UHRF1 complex [19, 20]. In light of these data, and combined to our previous study showing that ZBTB24 function is required to maintain DNAme at centromeric repeats in mouse embryonic fibroblasts [14], a molecular model has been proposed along which the ZBTB24-CDCA7 axis may regulate HELLS functions in DNAme maintenance of centromeric DNA repeats in murine cells [21]. However, the contribution of ZBTB24 during the DNAme establishment step, which occurs during early embryonic development, and specifically at these repeated sequences, remains untested.

ZBTB24 belongs to the broad-complex, tramtrack and bric-à-brac-zinc finger (BTB-ZF) family of transcription factors known for their essential role in the development of the immune system [22]. ZBTB24 is constitutively expressed, with the highest expression in B cells [13], suggesting an important role in the function of this cell lineage. ZBTB24 harbors a BTB domain thought to mediate protein oligomerization and/or dimerization, an AT-hook motif through which it can bind to AT-rich DNA sequences and eight Zinc-finger motifs in its C-terminal part, supporting the fact that ZBTB24 is a transcription factor [22]. Indeed, as shown by studies using a tagged version of the protein in the HCT116 colon cancer cell line or in mouse embryonic stem cells (mESCs), exogenous ZBTB24 acts as a transcriptional regulator, mostly as an activator, which localizes at gene regulatory regions and recognizes a specific DNA motif conserved between mouse and human [23, 24]. Interestingly, in murine NIH3T3 cells, exogenous ZBTB24 protein localizes to chromocenters, i.e. subnuclear clusters of (peri)centromeres from several chromosomes, suggesting that ZBTB24 would also play an important role at (peri)centromeric heterochromatin [25, 26]. As shown in zebrafish, ZBTB24 function would be essential for the methylated status of pericentromeric satellite DNA repeats and their transcriptional repression [27].

Despite increased knowledge of ZBTB24 functions, a comprehensive evaluation of its role in DNA methylation pathways, and more importantly, in mammalian cells relevant for the ICF syndrome, is still missing.

Considering that its loss of function is the cause of the ICF2 syndrome, we explored here the potential direct contribution of ZBTB24 functions to the immune response pathways, embryonic development, DNAme and centromere integrity. Taking advantage of mouse and human cellular systems and models relevant for the physiopathological features of the ICF syndrome, combined to multi-omics approaches, we addressed the multiple facets of the endogenous ZBTB24 protein. Notably, we document the genomic occupancy of endogenous ZBTB24 at regulatory sequences of genes and the transcriptomic consequences of its loss of function in human lymphocytes cells and in a dynamic cell-state transition during the differentiation process of mouse embryonic stem cells. In addition, we unveiled the striking and conserved occupancy of ZBTB24 at centromeric DNA repeats in both mouse and human, despite the poor conservation of centromeric sequences among species. We found that ZBTB24 has distinct roles at unique genes and tandem repeats with completely distinct genomic organization, mainly as a trans-activator of gene expression or a repressor of centromeric transcription. Importantly, our data indicate that ZBTB24 controls centromeric repeats transcription through the regulation of their DNAme status and the recruitment of the DNA methyltransferase DNMT3B.

As a whole, our study highlights the dual role played by ZBTB24 depending on genomic location. It places ZBTB24 at the crossroads between DNA methylation and maintenance of centromere integrity, two important facets of genome maintenance as evidenced by their alterations that characterize ICF syndrome, but also complex and multifactorial diseases like cancer.

Materials and methods

ICF2 mouse model

Mice mutant for Zbtb24 were generated at the Institut Clinique de la Souris (Illkirch, France, http://www.ics-mci.fr) by homologous recombination using standard gene targeting techniques. The targeting construct was designed to replace exon 2 by mutated sequences (deletion of two nucleotides TA), thus mimicking the human mutations of ICF2 patients of a Lebanese family described in [28] carrying the deletion c.396_397delTA (p.His132Glnfs^*19). Homozygous mice for Zbtb24 mutation (Zbtb24^mt/mt) were generated by intercrossing heterozygous mice (Zbtb24^wt/mt) in a C57BL/6J background. Mice were mated overnight and the presence of a vaginal plug the following morning was designated as E0.5. Generation and use of these mice has been approved by the Comité d’éthique en Expérimentation Animale Buffon (Paris, France) and received the agreement number 5314. The genotyping was performed using genomic DNA extracted either from cultured cells, yolk sac or tail of embryos with the REDExtract-N-Amp Tissue PCR kit (Merck) following the manufacturer’s instructions.

Cell lines and culture conditions

Mouse ES cells (mESCs) were derived from WT, Zbtb24 and Dnmt3b (Dnmt3b^m3/m3; [29]) mutant blastocysts as described in [30] at 3.5 dpc. ES cells deficient for all three active DNA methyltransferases, Dnmt1/Dnmt3a/Dnmt3b triple knockout (TKO), were kindly provided by Masaki Okano [31]. All mESCs used in this study were maintained in the absence of feeder cells, on plates coated with 1× Laminin (Merck) as described in [32], in serum-free 2i medium containing Dulbecco’s modified Eagle’s medium (DMEM) F12 Glutamax (Thermo Fisher Scientific) supplemented with 50% Neurobasal medium (Thermo Fisher Scientific), 1× N2 and B27 Supplements (both Thermo Fisher Scientific), 1% penicillin/streptomycin (Merck), 1× non-essential amino acids (PAA), 100 nM 2-mercaptoethanol (Thermo Fisher Scientific), 3 μM CHIR99201, 1 μM PD032552 inhibitors (both from Cell Guidance Systems) and 1000 U/ml units of Leukemia Inhibitory Factor (Merck). Cells were passaged with 0.05% accutase (Thermo Fisher Scientific) every two days and maintained at 37°C with 8% CO2 in a humidified chamber. For in vitro differentiation, 10⁶ mESCs were grown in 10 cm bacteriological Petri dish with differentiation medium containing 25% Advanced DMEM, 25% DMEM/F12 GlutaMAX, 50% Neurobasal medium, 1% of penicilline/streptavidin, 1× β-mercatoethanol and 11% of knock out serum, at 37°C with 5% CO2 (all products from Thermo Fisher Scientific). The differentiation medium was changed every day and embryoid bodies were collected by centrifugation for 4 min at 700 rpm and resuspended in 10 ml of fresh medium.

Lymphoblastoid cell lines (LCLs) from three healthy donors and two ICF2 patients were cultured in RPMI 1640 supplemented with 15% FCS and antibiotics (Thermo Fisher Scientific) as described in [15].

Generation of the auxin-induced-degradation system of ZBTB24 in mESCs

OsTIR1-9xMyc sequence linked via cleavable P2A sequence was introduced by CRISPR/Cas9-mediated homologous recombination at the C-terminus of the essential ubiquitously expressed RCC1 protein (regulator of chromosome condensation 1) into Zbtb24^wt/mt mESCs [guide RNAs (gRNAs)] are listed in Supplementary Table S12, all plasmids were a kind gift from by Dr Alexei Arnaoutov, NICHD, NIH). Transfections were all performed using Lipofectamin2000 (Thermo Fisher Scientific). ESCs were selected using 7 μg/ml blasticidin and single cell clones were isolated by limiting dilution and the osTir1 bi-allelic insertion was validated by PCR, Sanger sequencing, and western blotting. Sequences encoding a minimal functional auxin-inducible degron (mAID) and HA-tag were introduced by CRISPR/Cas9-mediated homologous recombination to the N-terminus of ZBTB24 protein (gRNAs are listed in Supplementary Table S12 were cloned in pSpCas9(BB)-2A-GFP (PX458) (Addgene #48138), the plasmids containing HA tag, a micro-AID tag (71–114 amino-acid), and puromycin resistance were kind gift from Dr Alexey Arnaoutov [33]. Following hygromycin selection at 150 μg/ml, single-cell clones were isolated by limiting dilution. These clones were then validated for the mAID insertion on WT allele, followed by the characterization of the AID system.

RNA extraction and RT-qPCR analysis

Total RNAs from cell lines were isolated using TRI Reagent® (Merck) according to manufacturer’s instructions. Contaminant genomic DNA was eliminated with TURBO DNA-free Kit (Thermo Fisher Scientific). Reverse transcription was carried out using 500 ng DNA-free RNA, 50 μM random hexamers (Thermo Fisher Scientific), 10 U of RNase inhibitor (New England Biolabs) and 200 U of ProtoScript II Reverse Transcriptase (New England Biolabs). Complementary DNA reactions were used as templates for real-time quantitative PCR reactions (qPCR). qPCR was performed using the SensiFAST SYBR NoROX Kit (Bioline) supplemented with 0.2 μM of specific primer pairs (Supplementary Table S12) and run on a light cycler rapid thermal system (LightCycler®480 2.0 Real time PCR system, Roche) with 20 s of denaturation at 95°C, 20 s of annealing at 60°C and 20 s of extension at 72°C for all primers and analyzed by the comparative CT(ΔCT) method using U6 RNA as an invariant RNA. Each data shown in RT-qPCR analysis is the result of at least three independent experiments.

Protein extraction and western blot analysis

Cells were washed in PBS and collected. After centrifugation, cell pellets were resuspended in RIPA buffer (300 mM NaCl, 1% NP-40, 0.5% Na-deoxycholate, 0.1% SDS, 50 mM Tris pH 8.0, complete protease inhibitor cocktail (Merck)). Samples were incubated for 30 min on ice and were then sonicated with a Bioruptor (Diagenode) at 4°C. Cell debris were removed by centrifugation for 10 min at 16 000 g and supernatants were collected. For Western blot analysis, proteins were resolved by SDS–PAGE, transferred onto PVDF membrane (Thermo Fischer Scientific), incubated with the appropriate primary antibody (Supplementary Table S13) and then with horseradish peroxidase-conjugated secondary antibodies. Protein detection was performed with ECL reagents (Thermo Fisher Scientific).

Immunofluorescence

Cells were grown on glass coverslips until reaching 60% of confluence. Cells were then fixed with 4% paraformaldehyde in PBS (Electron Microscopy Sciences). After permeabilization with 0.1% Triton X-100 in PBS, samples were blocked in 2% donkey serum (Jackson Immunoresearch) in PBS followed by incubation with primary antibodies (Supplementary Table S13) in blocking buffer overnight at 4°C. After incubation with secondary antibodies conjugated to Alexa-Fluor 488 or 594 (Jackson ImmunoResearch), coverslips were mounted in Vectashield medium containing 2 μg/ml DAPI (Vector Laboratories). Image acquisition was performed at room temperature on a fluorescence microscope (Axioplan 2; Zeiss) with a Plan-Neofluar 100X/1.3 NA oil immersion objective (Zeiss) using a digital cooled camera (CoolSNAP fx; Photometrics) and METAMORPH 7.04 software (Roper Scientific, Trenton, NJ). Images presented correspond to one focal plane.

Chromatin immunoprecipitation (ChIP)

Mouse ESCs and LCLs were crosslinked with 1% formaldehyde for 10 min at room temperature (RT) and the reaction was quenched by adding glycine at a final concentration of 0.125 M for 5 min at RT. Fixed cells were washed and adherent cells were harvested with phosphate-buffered saline (PBS). Cell nuclei were isolated using a cell lysis buffer (5 mM Pipes pH 8.0, 85 mM KCl, 0.5% NP40, Complete protease inhibitor cocktail (Merck)) for 20 min, at 4°C. Nuclei were then pelleted by centrifugation and lysed in a nuclei lysis buffer (50 mM Tris HCl pH 8.1, 150 mM NaCl, 10 mM EDTA, 1% SDS, Complete protease inhibitor cocktail (Merck)) for 20 min, 4°C. Chromatin was subjected to sonication with a Bioruptor (Diagenode) yielding genomic DNA fragments with a bulk size of 150–300 bp. Supernatant was diluted 10 times in IP dilution buffer (16.7 mM Tris pH 8, 167 mM NaCl, 1.2 mM EDTA, 1.1% Triton X- 100, 0.01% SDS, Complete protease inhibitor cocktail (Merck)). Immunoprecipitations were performed overnight at 4°C on 40 μg of chromatin with antibodies pre-bound to Dynabeads protein G (Thermo Fisher Scientific). The antibodies used are listed in Supplementary Table S13. The Beads were then washed once with Low salt buffer (0.1% SDS, 1% Triton, 2 mM EDTA, 20 mM Tris pH 8, 150 mM NaCl), once with High salt buffer (0.1% SDS, 1% Triton, 2 mM EDTA, 20 mM Tris pH 8; 500 mM NaCl), once with LiCl wash buffer (10 mM Tris pH 8.0, 1% Na- deoxycholate, 1% NP-40, 250 mM LiCl, 1 mM EDTA) and twice with TE (10 mM Tris pH 8.0, 1 mM EDTA pH 8.0). The elution of chromatin was performed in Elution buffer (50 mM Tris pH 8.0, 10 mM EDTA pH 8.0, 1% SDS) at 65°C for 45 min and crosslink was reversed O/N at 65°C after addition of NaCl to 0.2 M final concentration. The eluted material was digested with 40 μg of Proteinase K (Thermo Fisher Scientific) at 65°C for 2 h. DNA was purified by phenol–chlorofom extraction and ethanol precipitation, resuspended in TE. Quantitative PCRs were performed using the SensiFAST SYBR NoROX Kit (Bioline) and analyzed on a light cycler rapid thermal system (LightCycler®480 2.0 Real time PCR system, Roche). ChIP-qPCR results were normalized on input signal (% input). As a negative control, an additional mock ChIP using IgG was systematically added to each experiment. Sequences of primers are listed in Supplementary Table S12.

ChIP-seq experiment and analysis

ChIP-seq libraries were prepared either from ZBTB24 ChIP, IgG and input samples in LCLs derived from healthy subjects (n = 3) or from ZBTB24 ChIP and input samples in Zbtb24^wt/wt mESCs (n = 2) cultured in 2i conditions and at d3 of differentiation, using the MicroPlex Library Preparation Kit v2 (Diagenode) according to the manufacturer’s protocol. In brief, 1 to 25 ng were used as input material. After adapters ligation, fragments were amplified with Illumina primers for 10 cycles. Libraries were purified using AMPure XP beads protocol (Beckman Coulter, Indianapolis, IN) and quantified using the Qubit fluorometer (Thermo Fisher Scientific). Libraries’ size distribution was assessed using the Bioanalyzer High Sensitivity DNA chip (Agilent Technologies). Libraries were normalized and pooled at 2 nM and spike-in PhiX Control v3 (Illumina) was added. Sequencing was done using a NextSeq 500 instrument (Illumina) in paired-end mode (2 × 75 cycles). After sequencing, a primary analysis based on AOZAN software (ENS, Paris, France) [34] was applied to demultiplex and control the quality of the raw data (based on bcl2fastq v2.20 and FastQC modules/v0.11.5). Quality of raw data has been evaluated with FastQC (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Poor quality sequences have been trimmed or removed with Trimmomatic [35] software to retain only good quality paired reads. Reads were mapped to the reference genomes (mm10/hg38) using Bowtie2 v2.3.4.3 [36] with the option -k 2. Then, peaks were called with MACS2 v2.2.4 [37] using the input as control, an FDR of 0.01 and keeping all the duplicates. Peaks from multiple replicates were combined by taking regions covered by at least 2 replicates. Peak annotation was performed with ChiPseeker v1.18.0 [38] and Bedtools intersect v2.28.0 [39] was used to annotate peaks with repeated elements obtained on the repeatmasker website (http://www.repeatmasker.org), enhancers obtained on FANTOM5 [40] and genome segmentation from ENCODE data [41, 42]. Finally, motif detection was performed with RSAT [43]. The Gene ontology enrichment analysis was performed using the clusterProfiler R package. The significant GO categories were identified with Benjamini-Hochberg adjusted p-value (p.adjust) ≤ 0.05.

Bioinformatic analysis of ZBTB24 binding sites in DNA repeats and HORs

ZBTB24 occupancy on DNA repeats was determined using RepEnrich2 [44] with the following protocol available in https://github.com/nerettilab/RepEnrich2. Briefly, the reads were first uniquely aligned with Bowtie2 (−k 1) on the mm10 or hg38 genomes. Single and multiple reads were then separated, and multiple reads were assigned to a FASTQ file. In parallel, annotation of the repeats was performed using RepeatMasker annotations of the repeated elements. Overlap of single reads with repeated elements was tested. Multiple reads were aligned against the repeat assembly using Bowtie2. Finally, the number of reads aligning against families of repeated elements was determined by combining the counts of uniquely mapping reads and multi-mapping reads (fractional counts). Read counts were normalized by the library size and the input. Read mapping to HOR arrays was carried out described in [45]. The human ChIP_seq and RNA-seq FASTQ files were filtered against a library containing 6 455 351 sequences of 18 mers alpha satellite seed sequences [46]. Reads that contained at least one 18mers were retrieved and mapped with Bowtie2 on a reference composed of 64 centromeric HOR array consensus sequences [47].

RNA-seq experiment and analysis

RNA-seq was performed on independent clones of mESCs (2 clones of Zbtb24^wt/wt cells and of Zbtb24^mt/mt cells) and on LCLs derived from healthy (n = 2) and ICF2 (n = 2) subjects. RNA quality was verified using the Bioanalyzer RNA 6000 Nano kit (Agilent). The libraries were prepared following the TruSeq Stranded Total RNA protocol from Illumina, starting from 1 μg of high-quality total RNA (RIN > 7). Paired end (2 × 75) sequencing was performed on an Illumina NextSeq 500 platform, demultiplexing and quality control were performed with the AOZAN software (ENS, Paris, France) [34] (based on bcl2fastq v2.20 and FastQC modules/v0.11.5). The quality of raw data has been evaluated with FastQC (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Poor quality sequences have been trimmed or removed with Trimmomatic [35] software to retain only good quality paired reads. Star v2.5.3a [48] has been used to align reads on reference genome (mm10/hg38) using standard options. Quantification of gene and isoform abundances has been done with rsem 1.2.28, prior to normalisation on library size with DESEq2 [49] bioconductor R package. Finally, differential analysis has been conducted with edgeR [50] bioconductor R package. Multiple hypothesis adjusted p-values were calculated with the Benjamini-Hochberg procedure to control FDR.

DNA methylation of satellite repeats by southern blot

Genomic DNA was extracted from mESCs and LCLs using the Monarch Genomic DNA Purification Kit (New England Biolabs) according to the manufacturer’s instructions. Genomic DNA from mESCs and LCLs (500 ng) was digested respectively with 20 units of HpaII or HhaI enzymes (New England Biolabs) for 16 h to analyze the DNA methylation patterns of centromeric minor and alpha satellite repeats, respectively. Genomic DNA from Dnmt1/Dnmt3a/Dnmt3b triple-knockout (TKO) mESCs [31] and from Dnmt1 knock-out (Dnmt1−/−) mouse embryonic fibroblasts [51] were used as controls of low CpG methylated cellular contexts. The digested DNA fragments were separated by electrophoresis using 1% agarose gels and transferred overnight to Hybond-N+ membranes (GE Healthcare) in 20 × SSC. After ultraviolet crosslink, the membranes were pre-hybridized in 6 × SSC, 5 × Denhardt and 0.1% SDS and then hybridized with ³²P- labelled minor satellite (5′-ACATTCGTTGGAAACGGGATTTGTAGAACAGTGTATATCAATGAGTTACAATGAGAAACAT-3′) or alpha satellite (5′-ATGTGTGCATTCAACTCACAGAGTTGAAC-3′). The membranes were washed 3 times in 6X SSC and 0.1%SDS at 37°C and then subjected to phosphorimaging using FLA 7000 phosphorimager (Fuji). The quantification of radiolabeled signals intensity was determined using ImageJ software and normalized by the sum of the total signal per lane.

Methylome analysis by RRBS

The RRBS technique was originally described by Meissner et al. [52], and performed as previously described [53, 54]. RRBS libraries were prepared by MspI digestion from 50 to 100 ng genomic DNA44 and sequenced in paired-end 2 × 75 bp on an Illumina HiSeq4000 at Integragen SA (Evry, France). We trimmed RRBS reads to remove low-quality bases with Trim Galore v0.4.2 and aligned reads to the mm10 genome with BSMAP v2.74 (parameters -v 2 -w 100 -r 1 -x 400 -m 30 -D C-CGG -n 1). We calculated methylation scores using methratio.py in BSMAP v2.74 (parameters -z -u -g) and filtered CpGs covered by a minimum of eight reads. DMRs were identified using eDMR from the methylKit R package with a minimum of seven differentially methylated CpGs, a difference in methylation > 20% and a q-value < 0.001.

Visualization of sequencing data

For ChIP-seq data visualization, the plots of the distribution of reads density across all transcription start sites RefSeq genes were generated using deepTools (https://deeptools.readthedocs.io/). Bar plots, dot plots, Hierarchical clustering representations, Heatmaps, violin plots, were generated using functions in R software environment for statistical computing. Samples clustering and heatmap visualizations were performed using the ComplexHeatmap and the gplots R packages (https://www.bioconductor.org/). For hierarchical clustering, the hclust function were used with “euclidean” and “ward.D2” method parameters respectively for distance and agglomeration. Basic stacked bar plots and Violin were build using the ggplot2 R package.

Data availability

The ChIP-seq, RNA-seq and RRBS raw and processed data presented in this study have been deposited in the NCBI Gene Expression Omnibus (GEO) under the accession number GSE218867. The following published dataset was used: GSE111689 [23].

Results

ZBTB24 is a regulator of innate response-related genes in human lymphocytes

To explore the role of ZBTB24 as a transcription factor and the consequences of its pathological loss of function (LOF) on the transcriptome, we performed chromatin immunoprecipitation of endogenous ZBTB24 followed by DNA sequencing (ChIP-seq) in human lymphoblastoid cell lines (LCLs) from three healthy donors (HDs) and RNA-sequencing (RNA-seq) in LCLs derived from HDs and two ICF2 patients (Supplementary Fig. S1A). We chose this cellular context because i) it is particularly relevant to the pathophysiology of the ICF syndrome and ii) the ZBTB24 protein is expressed at high levels in LCLs compared to other somatic cell lines [13].

ChIP-seq data analysis revealed 183 genomic sites occupied by ZBTB24 in LCLs derived from HDs (Supplementary Table S1, merged peaks, q-value < 0.05), which predominantly localized at transcription start sites (TSS) or within proximal promoters (≤1Kb) and, to a lesser extent, at distal intergenic regions (Fig. 1A and B). The functional annotation of ZBTB24 sites, according to the chromHMM genome segmentation ENCODE data in LCLs (GM12878), showed that it mainly occupies active promoters (n = 106, “Active_Promoter” category) and enhancers (n = 32, “Strong_Enhancer” category) (Fig. 1A). ZBTB24 peaks showed a highly significant enrichment of the 5′-CAGGACCT-3′ DNA motif (E_value = 2.8e-34) (Fig. 1A, lower panel), identical to the one identified for a FLAG-tagged ZBTB24 protein in the colon cancer cell line HCT116 [23].

Figure 1

ZBTB24 acts as transcriptional regulator modulating innate immune response-related genes in patients-derived LCLs. (A) ChIP-seq analyses showing the distribution of genomic sites occupied by ZBTB24 (merged peaks from three biological replicates, n = 183 peaks) in lymphoblastoid cell lines (LCLs) across genomic (left) and functional annotations (right). The functional annotation was based on the chromatin states established in GM12878 LCLs (ChromHMM track from ENCODE/broad institute). Predicted binding motif of ZBTB24 was identified using regulatory sequence analysis tools (RSAT, http://rsat.sb-roscoff.fr/). Distal_int: Distal intergenic; prom: Promoter. (B) Heatmap of ZBTB24 ChIP-seq signal, i.e pile-ups of the ZBTB24 enrichment relative to chromatin input (subtraction mode, input from IP of bigWig files) in three LCLs on genomic loci defined as 3 kb upstream and downstream of annotated transcriptional start site (TSS, NCBI RefSeq) of genes. High and low density levels are shown. (C) Unsupervised clustering of genes TSS linked to ZBTB24 sites. The heatmap (ChIP ZBTB24) presents the enrichment of ZBTB24 at each TSS based on the peak score (mean log₂(MACS2 score)). The heatmap also shows the gene expression data, presented as a mean of the row Z-score of each gene in LCLs derived from the blood of two healthy (CTL) and two ICF2 subjects (ICF2). ZBTB24 genomic sites common to LCLs and HCT116 cell line (GSE111689, Thompson et al 2018) are notified in the column next to ``Gene Expression''. The motif columns refer to the occurrence of the predicted ZBTB24 binding motif occurrence (counts). (D) Gene set enrichment analysis (GSEA) of deregulated genes in LCLs derived from the blood of two ICF2 patients (pD and pV) (left panel). The expression status of genes belonging to the “Interferon_response” HALLMARK category is presented as the row Z-score and illustrated by a heatmap (right panel).

Based on the peak score (MACS2 score) at TSS region (defined from −1000 to +100), we identified the primary target genes of ZBTB24 which are CDCA7, RIOK2, CNPY4, TAF6, ARID5B, CAMKMT, PREPL, RNF187, RPL32, CDC40, WASF1, OSTC, ACO2, PHF5A, and DTNB (Fig. 1C and Supplementary Fig. S1B). Notably, this score did not consistently correlate with the presence of the predicted ZBTB24 consensus binding motifs (5′-CAGGACCT-3′ or 5′-AGGTCCTG-3′) within the targeted region. For example, OSTC, RIOK2 and RPL32 do not contain the predicted ZBTB24 consensus binding motif within the peak located in their proximal promoter region (−1000 to +100). Remarkably, several of these genes—TAF6/CNPY4, ACO2/PHF5A, CDC40/WASF1 and CAMKMT/PREPL—form bidirectional gene pairs and contain the predicted ZBTB24 binding motif in their proximal, and potentially bidirectional, promoter region.

The RNA-seq analysis revealed that CDCA7, RIOK2, ARID5B, RNF187 and OSTC were significantly downregulated (log2FC > 0.5, FDR < 0.01) in LCLs from ICF2 patients compared to HDs (Supplementary Table S2). While TAF6, PREPL, RPL32, CDC40, ACO2, PHF5A, and DTNB showed a downward trend in expression (according to the Z-score, Fig. 1C), this trend did not reach the threshold of significance (FDR < 0.01). In contrast, CAMKMT, WASF1 and CNPY4 were not downregulated in LCLs from ICF2 patients compared to HDs.

These results, combined with previous findings from gene reporter assays [17, 23, 26], strongly suggest that ZBTB24 functions as a transcriptional activator by binding to specific DNA motifs in target genes’ regulatory regions [24]. However, since not all the main target genes of ZBTB24 were downregulated in LCLs from ICF2 patients, it is likely that ZBTB24 functionally interacts with other context-specific TFs to modulate the expression of certain target genes.

The whole transcriptome analysis of ICF2 LCLs identified 356 up- and 50 down-regulated genes compared to normal LCLs (log2FC > 0.5, FDR < 0.01), suggesting a role of ZBTB24 as a negative gene regulator with a significant enrichment in a set of genes involved in the interferon response (adjusted p-value = 1.82e-03, Supplementary Table S3 and Fig. 1D) and, more broadly, in immune response pathways (Supplementary Fig. S2).

As a whole, these results show that, in human LCLs, ZBTB24 deficiency is associated with overexpression of genes involved in immune response pathways, more specifically of innate immune response-related genes, which is consistent with the autoimmune signs in the ICF patients [55, 56].

ZBTB24 is essential for mouse embryonic development

The molecular and phenotypic features of ICF2 patients clearly pinpoint a critical role for ZBTB24 during development since nearly all patients exhibit developmental delay and growth retardation [8]. To better define this role, we generated a mouse model for the ICF2 syndrome that reproduced the BTB domain mutations reported in Lebanese siblings [28] (Supplementary Fig. S1A and Supplementary Fig. S3A–D). Of note, the ZBTB24 protein sequence is highly conserved between mouse and human ranging from 76.9 to 96.9% of homology for the structural domains [26].

Crosses of mice heterozygous for Zbtb24 mutation (Zbtb24^wt/mt) did not produce newborns homozygous for the mutation (Zbtb24^mt/mt), indicating the embryonic lethality of Zbtb24^mt/mt embryos (Supplementary Table S4). We isolated embryos at different developmental stages [6.5, 9.5- and 11.5-days post coitum (dpc)] and noticed that Zbtb24^mt/mt embryos had a growth delay at all stages and presented morphological abnormalities (Fig. 2A). These observations highlighted an essential role for ZBTB24 during early mouse development, consistent with the growth delay that characterizes ICF patients.

Figure 2

ZBTB24 targets developmental genes in mESCs and is essential for embryonic development. (A) Pictures of Zbtb24^wt/wt and Zbtb24^mt/mt embryos at post-implantation stages 6.5, 9.5- and 11.5-days post coitum (dpc). The arrows point to the embryonic defects reported in Zbtb24^mt/mt embryos at 11.5 dpc. (B) Stacked bar chart of the distribution of the genomic sites occupied by ZBTB24 across genomic features. Peaks overlapping between two independent mESC clones derived from distinct embryos cultured in 2i conditions (2i) or at day3 of differentiation (d3) were merged and annotated to the mm10 genome. The predicted ZBTB24 binding motif identified by regulatory sequence analysis tools (RSAT, http://rsat.sb-roscoff.fr/) based on ChIP-seq data performed in 2i condition (2i) or at day3 of differentiation (d3) is reported below the plots. (C) Stacked bar chart of the distribution of ZBTB24 sites across functional regions of the mouse genome (mm10). ZBTB24 peaks overlapping between two independent ES clones cultured in 2i conditions (2i) or at day3 of differentiation (d3) were merged and annotated based on ChIP-seq assays of 8 histone modifications as part of the ENCODE 3 consortium (laboratory of Bing Ren). Pr: Promoter; Enh: Enhancer; Int,Inter: Intron, intergenic; ex: Exon; Hc: Heterochromatin. (D) Gene ontology of genes linked to ZBTB24 sites in mESCs at day3 of differentiation (ChIP ZBTB24 d3). The dot plot represents the enrichment in biological processes. GeneRatio on the x axis refers to the number of genes linked to ZBTB24 sites divided by the number of genes annotated in the corresponding process. The size and colors of the dots indicate gene counts and the adjusted p-value for the GO pathway listed on the y axis, respectively. Only the top 5 GO terms with the highest degree of enrichment are represented (GO:0007389 pattern specification process, GO:0003002 regionalization, GO:0060562 epithelial tube morphogenesis, GO:0048562 embryonic organ morphogenesis, GO:0019827 stem cell population maintenance). (E) Unsupervised clustering of genes linked to ZBTB24 merge peaks at TSS. The heatmap (ChIP ZBTB24) presents the enrichment of ZBTB24 at TSS of genes based on the peak score (mean log₂(MACS2 score)), in two independent ES clones cultured in 2i conditions (2i) or at day3 of differentiation (d3). For each locus, the gene status (“Bivalent” or “Dev.”: developmental genes) and the predicted ZBTB24 binding motif occurrence are indicated, and a color code is attributed to each category next right to the figure. The gene expression data are presented as the mean of the row Z-score in the two independent ES clones WT and mutant (mt) for Zbtb24 cultured in 2i condition (WT_2i, mt_2i) or at day3 of differentiation (WT_d3, mt_d3). (F) Heatmap of ZBTB24 ChIP-seq signal, i.e pile-ups of the ZBTB24 enrichment relative to chromatin input (subtraction mode, input from IP of bigWig files) in two independent mESCs cultured in 2i conditions (2i) or at day3 of differentiation (d3) on genomic loci defined as 3 kb upstream and downstream of annotated transcriptional start site (TSS, NCBI RefSeq) of genes. High and low density levels are shown.

To better explore this role, we first derived and characterized mouse embryonic stem cells (mESCs) from wild-type (Zbtb24^wt/wt) and mutant (Zbtb24^mt/mt) blastocysts, isolated at 3.5 dpc [30], in 2i medium (two inhibitors of the MAPK/ERK and GSK3β pathways) to maintain their naïve ground state [57] (Supplementary Fig. S3E–F). We took advantage of the dynamic cell-state transition during the differentiation process of mESCs to model the first molecular steps that occur in the early embryo. We performed ChIP-seq of endogenous ZBTB24 in Zbtb24^wt/wt mESCs and RNA-seq in Zbtb24^wt/wt and Zbtb24^mt/mt mESCs (two biological replicates of each genotype) maintained in naïve conditions (2i) or after 3 days of differentiation (d3). While the levels of ZBTB24 protein in Zbtb24^wt/wt (WT) mESCs were nearly identical in naïve and differentiated states (Supplementary Fig. S3G), we identified a four times greater number of ZBTB24 genomic binding sites in differentiated mESCs (191 sites, q-value < 0.05) compared to mESCs in 2i (47 sites, q-value < 0.05) (Fig. 2B and Supplementary Tables S5–S6). Nearly all the ZBTB24 sites identified in naïve conditions were maintained in differentiated mESCs (44 out of 47 peaks). At d3, ZBTB24 binding sites localized mainly at cis-regulatory elements, active promoters and enhancers (Fig. 2B and C).

The top 5 GO terms of genes linked to ZBTB24 peaks in differentiated mESCs, but not in naïve conditions, matched remarkably with the developmental defects reported in the Zbtb24^mt/mt embryos and are all linked to parent biological processes “Anatomical structure morphogenesis” and “Multicellular organismal process” (Fig. 2D and Supplementary Table S7). Nonetheless, most of the genes belonging to these GO terms and categorized as developmental or bivalent (Fig. 2E, based on ENCODE3 data), showed relatively weak ZBTB24 peaks consistent with the absence of the identified DNA binding motif, suggesting they are not primary targets of ZBTB24 at this early stage of differentiation. This could be a reason why we did not detect major transcriptional deregulation of this set of genes in Zbtb24^mt/mt mESCs cellular model at d3 of differentiation (Fig. 2E).

Together, these results showed that ZBTB24 is essential for mouse embryonic development possibly through its ability to target and regulate some key developmental genes.

ZBTB24 is a conserved transcriptional activator of a set of unique genes

We next evaluated the conservation of the role of ZBTB24 as a transcriptional regulator in mice and humans (Fig. 1 and Fig. 2). Compared to human cells, we identified the same functional and chromatin state features for ZBTB24 genomic sites in mESCs, an identical enriched DNA motif 5′-CAGGACCT-3′ within ZBTB24 peaks and a conserved genomic distribution, i.e. at promoters and at distal intergenic regions (Fig. 2B), at TSS (21/47 peaks in 2i and 55/191 peaks at d3, Fig. 2F) and enhancers, mostly at d3 (3 peaks in 2i compared to 35 peaks at d3) (Fig. 2C). Noteworthy, most of the genes showing the highest ZBTB24 enrichment at TSS were identical to those identified in human LCLs (Cdca7, Riok2, Cnpy4, Taf6, Arid5b, Camkmt, Prepl, Rnf187, Rpl32, Cdc40, Wasf1, Ostc, Axin2, Aco2, Phf5a and Dtnb) (Supplementary Figs S4 and S5), suggesting similar mechanisms of regulation of gene expression in humans and mice. As in human cells, several of these target genes exhibited reduced expression levels in Zbtb24^mt/mt mESCs, strongly supporting a conserved role of ZBTB24 as a transcriptional activator at this conserved set of unique genes (Supplementary Tables S8–S9, Fig. 2E, Supplementary Figs S4 and S5).

As a whole, these results revealed a discrete and conserved role of ZBTB24 as a transcriptional activator at unique genes, both in humans and mice. These results also validated the use of our ICF2 mouse model to further decipher the conserved genomic functions of ZBTB24.

ZBTB24 plays a role during the DNAme establishment step

Considering the central role played by DNAme during mammalian development [58] and the altered DNAme patterns that are typical of ICF2 patient cells [15], we further explored the role of ZBTB24 in shaping DNAme patterns during early embryonic developmental stages.

The first evidence of a role of ZBTB24 in DNAme came from the demonstration that ZBTB24 is required for the maintenance of the methylated status of tandem centromeric DNA repeats in the mouse [14], consistent with the additional DNAme loss at centromeric repeats in ICF2 patients compared to ICF1 [10]. To address the role of ZBTB24 at a genome-wide scale during the DNAme establishment step, which occurs in the early phases of mouse embryonic development, we profiled the DNA methylome of Zbtb24^wt/wt and Zbtb24^mt/mt mESCs, in naïve and differentiated state (d3), using Reduced Representation Bisulfite Sequencing (RRBS). Indeed, the dynamic mESCs transition from naïve to differentiated state is recognized as a powerful model that mimics in vitro the embryonic window where DNAme profiles are established de novo [59, 60]. As expected, the DNAme profiles obtained by RRBS allowed to clearly distinguish mESCs in 2i medium from those at d3 of differentiation, with a gain of both weakly and strongly methylated sequences in differentiated mESCs (Fig. 3A and B).

Figure 3

ZBTB24 moderately contributes to the establishment of DNAme landscape. (A) Dendrogram showing the clustering of the two mESCs clones wild-type (WT) and mutant (mt) for ZBTB24, cultured in 2i medium (2i) or at day3 of differentiation (d3), based on their DNAme levels across over 1 million CG sites analyzed by RRBS. (B) Violin plot of DNAme level in the two ES clones wild-type (WT) and mutant (mt) for ZBTB24, cultured in 2i medium or at day3 of differentiation. The number and the status (hypo: Hypomethylation; hyper: Hypermethylation) of differentially methylated regions (DMR) identified in Zbtb24^mt/mt compared to Zbtb24^wt/wt mESCs are shown next above to the plot. (C) Dot plots showing the distance in base pairs (bp) between DMR and TSS of genes in mESCs cultured in 2i medium or at day 3 of differentiation. (D) Dot plots presenting the distance in base pairs (bp) between ZBTB24 peak summits in Zbtb24^wt/wt mESCs and DMR in mESCs cultured in 2i medium and at day3 of differentiation (2i and d3). Dots which correspond to genes whose change in their level of expression is correlated to DMR position are indicated by gene symbol.

In naïve and differentiated mESCs, Zbtb24^mt/mt cells showed distinct DNAme profiles compared to Zbtb24^wt/wt mESCs, which indicates that ZBTB24 is required to ensure correct shaping of DNAme profiles at early development stages (Fig. 3A). Surprisingly, we observed reduced levels of DNAme in Zbtb24^mt/mt compared to Zbtb24^wt/wt mESCs in 2i conditions (10 974 hypomethylated methylated regions or hypoDMR) (Fig. 3B and Supplementary Table S10), a state at which the levels of DNAme are known to be low and minimal compared to mESCs cultured in serum conditions [60, 61]. Upon commitment to differentiation, the number of hypoDMR in Zbtb24^mt/mt cells was greatly reduced compared to Zbtb24^wt/wt cells (195 vs 10 974 hypoDMR), suggesting that the wave of de novo DNAme that occurs during this process compensated most of the methylation defects identified in the naive state despite the ZBTB24 LOF (Fig. 3B and Supplementary Table S11). HypoDMRs were mostly annotated away (> 1 kb) from TSS (Fig. 3C), i.e. to intergenic regions, within LINE (Long Interspersed Nuclear Elements) and ERV (Endogenous Retroviruses) retrotransposon elements (Supplementary Fig. S6A–B and Supplementary Tables S10 and S11). However, hypomethylation of these retrotransposon elements was not accompanied by a change in their expression (Supplementary Fig. S6C).

To assess whether the methylation status of the genomic sites occupied by ZBTB24 was directly impacted by ZBTB24 LOF, we computed the distance between ZBTB24 sites (peak summits) and the position of the DMRs identified in Zbtb24^mt/mt mESCs. We applied the same analyses to human cells using our previous methylome profiling in ICF2 patients [15]. In both cases, we found that hypoDMRs in ZBTB24 mutant cells were located far away (> 10Kb) from ZBTB24 binding sites both in mouse and human cells (Fig. 3D and Supplementary Fig. S6D). Notably, we identified a restricted number of hypermethylated (hyperMe) sites, mostly detectable in human LCLs, that were closer (< 1Kb) to ZBTB24 genomic targets compared to the hypoMe ones. Interestingly, these hyperMe sites resided within the promoter region of RNF187, CDC40, ZNF717, and ARID5B which are ZBTB24 target genes and whose expression levels were downregulated in ICF2 patient cells (Fig. 1C and Supplementary Fig. S4). Of note, the hyperMe of CpG within the promoter of ARID5B and its downregulation in ZBTB24 mutant cells was a conserved feature shared by human and mouse cells mutant for ZBTB24 (Fig. 3D).

Altogether, as inferred from ZBTB24 LOF, ZBTB24 does not seem to be directly involved in the DNAme establishment step for a large proportion of the genome. However, our methylome highlights potential role of ZBTB24 in the protection of some of its main genomic targets against DNAme gain during development [62].

ZBTB24 occupies minor satellite repeats and maintains their silent state

The discovery of ZBTB24 mutations as a genetic factor in ICF2 syndrome raised questions about its functions at centromeric satellite repeats, given that ICF2 patients display reduced methylation at these sequences, in addition to the reduced methylation at juxtacentromeric satellites characteristic of all ICF subtypes. Thus, we first analyzed whether ZBTB24 occupies these genomic regions in murine cells. We overcame the problem posed by the high sequence homology between satellite repeats units and their tandemly repeated nature, by using the Repenrich2 method [44] to quantify the enrichment of ZBTB24 ChIP-seq reads in satellite DNA sequences (see Materials and Methods section). Of note, mouse centromere DNA repeats are relatively homogeneous between all chromosomes and composed of AT-rich tandem repeats of 120 bp called minor satellite (MinSat), annotated as SYNREP_MM and CENSAT_MC in the Repeatmasker database (Supplementary Fig. S7A and B). Based on our ZBTB24 ChIP-seq data obtained from Zbtb24^wt/wt mESCs, we identified reads mapping to centromeric satellite DNA, although with a more modest enrichment in differentiated mESCs (d3) compared to the naïve condition (Supplementary Fig. S7A; d3/2i > 1). These analyses were further confirmed by ChIP-qPCR using primers specific for MinSat repeats (Fig. 4A). Because of its role as a transcription factor, we tested whether ZBTB24 LOF impacted the levels of centromeric transcripts (cenRNA) originating from these centromeric regions. In naïve mESCs, the levels of MinSat transcripts are low but almost 3-times higher in the context of ZBTB24 LOF (Fig. 4B and Supplementary Fig. S7C). These data support a key role of ZBTB24 as a transcriptional repressor of mouse centromeric repeats.

Figure 4

ZBTB24 occupies mouse centromeric satellite repeats and is required for their DNAme through DNMT3B recruitment. (A) ChIP assays performed on chromatin prepared from Zbtb24^wt/wt (Z_WT) and Zbtb24^mt/mt (Z_mt) mESCs cultured in 2i medium or at day 3 of differentiation, using ZBTB24 antibodies or IgG, and followed by quantitative PCR with primers specific of minor satellite sequences. Error bars represent standard error (n = 3 independent experiments), p: p-value; two-tailed t-test. (B) Expression levels of minor satellite RNA in Zbtb24^wt/wt (WT) and Zbtb24^mt/mt (mt) mESCs cells clones (1 and 2) cultured in 2i medium or at day 3 of differentiation were assessed by RT-qPCR, presented as a relative expression to U6 snRNA levels. Error bars represent standard error (n = 4 independent experiments), p: p-value; two-tailed t-test. (C) Southern blot of minor satellite repeats obtained from the digestion of genomic DNA of mouse embryonic fibroblasts (MEF) WT or knockout for Dnmt1 (D1−/−), of mESCs Zbtb24^wt/wt, Zbtb24^mt/mt and Dnmt3b^mt/mt cultured in 2i medium or at day 3 of differentiation, by the methylation-sensitive enzyme HpaII. (D) Bar plot showing the quantification of DNA methylation changes at minor satellite (unMe/Me signal; unmethylated signal: unMe; methylated signal: Me) in Zbtb24^wt/wt and Zbtb24^mt/mt cultured in 2i medium or at day 3 of differentiation. Error bars represent standard error (n = 3 independent southern blot experiments on two independent mESCs clones), p: p-value; two-tailed t-test. (E) (top) ChIP assays performed on chromatin prepared from Zbtb24^wt/wt, Zbtb24^mt/mt and Dnmt3b mutant (Dnmt3b^mt/mt) mESCs cultured in 2i medium or at day 3 of differentiation, using anti-DNMT3B or IgG antibodies, followed by quantitative PCR with primers specific of minor satellite sequence. Error bars represent standard error (n = 3 independent experiments on two different clones), p: p-value; two-tailed t-test. (bottom) western blot analysis showing the expression level of DNMT3B and γ-tubulin proteins in Zbtb24^wt/wt, Zbtb24^mt/mt and Dnmt3b^mt/mt mESCs cultured in 2i medium and at day3 of differentiation. Molecular weight is indicated on the left of the panel. (F) Southern blot of minor satellite repeats obtained from the digestion of genomic DNA of in AID_ZBTB24 mESCs cultured in 2i medium or at day 2 of differentiation (d2) in absence or presence of auxin for 48 h, by the methylation-sensitive enzyme HpaII. Bar plot shows the quantification of DNA methylation changes at minor satellite (unMe/Me signal). Error bars represent standard error (n = 3 independent southern blot experiments on two independent mESCs clones), p: p-value; two-tailed t-test; ns not significant.

Figure 5

ZBTB24 occupies alpha satellite repeats and maintains their DNAme and transcriptional silent states. (A) ChIP assays performed on chromatin prepared from WT LCLs, using ZBTB24 antibodies or IgG, and followed by quantitative PCR with primers specific of alpha satellite sequences. Error bars represent standard error (n = 3 independent experiments), p: p-value; two-tailed t-test. (B) Expression levels of alpha satellite RNA in LCLs derived from healthy (CTL1 and CTL2; WT for ZBTB24) and ICF2 subjects (pD, PV; mutant for ZBTB24) were assessed by RT-qPCR, presented as a relative expression to U6 snRNA levels. Error bars represent standard error (n = 3 independent experiments), p: p-value; two-tailed t-test. (C) Southern blot of alpha satellite repeats obtained from the digestion of genomic DNA of LCLs from healthy (CTL1 and CTL2) and ICF2 subjects (pD and pV) by the methylation-sensitive enzyme HpyCH4IV. (D) Unsupervised clustering of centromeric HORs occupied by ZBTB24. The heatmap (ChIP) presents the peak score computed from the mean difference between normalized reads count in ChIP_seq of ZBTB24 and of IgG at the considered HOR in LCLs derived from the blood of three healthy subjects. The bar chart on the right shows the log2 difference of the peak score between IP_ZBTB24 and IP_IgG. The heatmap (expression) shows the expression data of each HOR computed from RNA-seq data of two independent LCL clones WT (CTL) and mutant (ICF2) for ZBTB24 derived from the blood of subjects. The bar chart on the right presents the log2 fold change (FC) of HORs expression based on RNA-seq data, comparing the ICF2 to the WT subjects.

ZBTB24 is required for DNAme establishment at minor satellite repeats

Hypomethylation of centromeric satellite DNA being a molecular feature of ICF2 patient cells, we wanted to verify whether this feature was conserved in vivo in our ICF2 mouse model. We analyzed the DNAme status of centromeric satellite sequences in Zbtb24^wt/wt and Zbtb24^mt/mt embryos by Southern blot using DNAme-sensitive restriction enzymes, a classical and widely used technique to analyze DNAme at satellite repeats [14]. We found that, as in human, centromeric MinSat repeated sequences were hypomethylated in Zbtb24^mt/mt embryos compared to Zbtb24^wt/wt embryos (Supplementary Fig. S7D). These data showed that ZBTB24 is required for the methylated status of centromeric repeats in both mouse and human. Although we previously provided evidence that ZBTB24 is required for the maintenance of DNAme at centromeric MinSat repeats in murine somatic cells [14], the question of its involvement in the establishment of DNAme step at these satellite repeats during early development remained unaddressed. Hence, we followed the DNAme status of MinSat during the wave of de novo methylation that occurs as mESCs are induced to differentiation, using Southern blot. As controls, we used genomic DNA from mouse embryonic fibroblasts WT and KO for Dnmt1, which represents methylated and unmethylated cellular contexts, respectively (Fig. 4C). In line with our RRBS analysis (Fig. 3A and B), we observed that MinSat repeats were already strongly hypomethylated in Zbtb24^mt/mt compared to Zbtb24^wt/wt mESCs in naïve conditions (2i) (Fig. 4C and D). This hypomethylated state remained in differentiated Zbtb24^mt/mt cells (d3), although less pronounced than in naïve conditions, suggesting an incomplete or delayed establishment of DNAme patterns at MinSat repeats in the absence of a functional ZBTB24 (Fig. 4C and D). It is possible that the lower levels of DNAme at MinSat observed in differentiated mutant cells are a consequence or are aggravated by the pre-existing hypomethylated status of these repeats in Zbtb24^mt/mt compared to Zbtb24^wt/wt mESCs as observed in the naïve state.

The DNA methyltransferase DNMT3B is not only required but is also the sole DNMT shown to catalyze the de novo DNAme of MinSat sequences [63]. Mouse ESCs mutant for Dnmt3b (Dnmt3b^mt/mt) failed to properly establish DNAme at MinSat repeats during mESCs differentiation ([63] and Fig. 4C). Hence, we tested whether ZBTB24 was involved in DNAme establishment at MinSat repeats through the recruitment of DNMT3B. Whereas the expression levels of the DNMT3B protein was comparable between Zbtb24^wt/wt and Zbtb24^mt/mt mESCs (Fig. 4E), ChIP-qPCR experiments clearly showed a significant loss of DNMT3B occupancy at MinSat repeats in ZBTB24 mutant cells (Fig. 4E).

In order to formally address the role of ZBTB24 during the DNAme establishment step at MinSat in an unbiased manner, we generated and characterized an auxin-inducible degron (AID) by knock-in at the endogenous Zbtb24 locus in WT mESCs lines (Z-AID, Supplementary Fig. S7E). This system allowed us to specifically trigger a rapid and reversible degradation of ZBTB24 for kinetics experiments, at specific time points during mESCs differentiation, i.e. the time window when DNAme is established. To validate the degradation of ZBTB24 upon auxin treatment in these Z-AID mESCs lines, we performed western blot analysis at different time points (Supplementary Fig. S7F). We also assessed by RT-qPCR the expression of Cdca7, one of the major ZBTB24 target genes (Supplementary Fig. S7F). In 2i conditions, and after only 1 h of auxin treatment, the levels of ZBTB24 protein dropped drastically whereas Cdca7 expression significantly diminished 6 h after auxin addition (Supplementary Fig. S7F). We then tested the impact of ZBTB24 degradation on DNAme at MinSat after auxin treatment. Using the degron system, we observed cells showing signs of poor health after 3 days of differentiation in the presence of auxin. We thus limited our investigations of DNAme within 2 days of auxin treatment to avoid biases in our experiments related to cell viability. We verified that after 48 h of auxin treatment in 2i conditions, the levels of DNAme were unchanged compared to untreated cells (Fig. 4F). Hence, this system is suitable to further address the impact of ZBTB24 loss upon induction of differentiation and DNAme establishment. We found that ZBTB24 is required during the DNAme establishment step at centromeres since addition of Auxin and ZBTB24 degradation in differentiating mESCs (d2) impaired the correct establishment of DNAme patterns at MinSat (Fig. 4F and Supplementary Fig. S7G).

In sum, our findings highlight the key role of ZBTB24 in both the establishment and maintenance of DNA methylation at centromeric MinSat repeats, consistent with its occupancy at these loci. In addition, we showed that DNAme establishment involves the recruitment of DNMT3B.

ZBTB24 function at centromeric satellite DNA is conserved in human

The genomic organization of mouse and human centromeres are clearly distinct and not conserved. In contrast to mouse centromeres, human centromeres are composed of Alpha satellite DNA, monomers of 171 bp arranged in a head-to-tail manner and organized in repeated arrays called higher order repeats (HORs) that are chromosome-specific [64]. Using the same approach developed in murine cells, we were able to identify ZBTB24 ChIP-seq reads from human LCLs that matched satellite repeat sequences, and in particular ALR-Alpha satellite (Supplementary Fig. S8A). These analyses were further confirmed by ChIP-qPCR performed in LCLs using primers specific for alpha satellite sequences (Fig. 5A). We then compared the levels of alpha-satellite RNA and DNAme in LCLs derived from healthy (Control) and ICF2 subjects. We found an anti-correlation between the levels of DNAme at alpha-satellites and that of transcripts emanating from these sequences in ICF2 patient cells compared to control cells, with hypomethylation of alpha-satellites and higher levels of alpha-satellite RNAs (Fig. 5B and C, Supplementary Fig. S8B). Altogether, our data strongly support a model where ZBTB24 occupies centromeric satellite sequences and maintains their DNAme and transcriptional silent states, in both mouse and human cells.

We then performed deeper analyses of our sequencing reads mapping to human centromeres to investigate whether ZBTB24 occupies specific HORs and controls the levels of transcripts generated from these HORs. We found that ZBTB24 occupancy was enriched at specific centromeric HORs with a marked preference for cen1_1, cen6, cen7_1, cen8, cen9, cen10_1, cen11_1, cen16_1, cen18_1, cen22_1 and cenX (Supplementary Fig. S8C). Interestingly, targeted HORs were mainly the active HORs, i.e. HOR_1, defined as such based on their enrichment in CENP-A and binding sites for CENP-B (CENP-Box) [45], two essential proteins for centromere identity and function. Furthermore, the levels of transcripts originating from HORs occupied by ZBTB24 were higher in ICF2 patient cells (Fig. 5D).

Altogether, these results lend support to a model where ZBTB24 preferentially occupies active HORs and maintains their silent state, likely through its ability to ensure proper DNA methylation at these satellite sequences.

Discussion

ZBTB24 has long been considered as a ZBTB protein with an “obscure physiological function” [22], until the identification of mutations in ZBTB24 as a causative factor of the ICF2 syndrome [13] brought to light its importance for development, the immune system, centromere biology and DNAme. Here, using mouse and human cellular systems relevant to early developmental defects or to B-cells particularly affected in ICF patients, combined to multi-omics approaches, we validated a list of conserved and direct targets of endogenous ZBTB24 in relevant cellular systems, which includes regulatory regions of a set of unique genes. Importantly, we also added tandemly repeated sequences that underlie centromeric chromosomal regions as being occupied by ZBTB24. At murine and human centromeres occupied by ZBTB24, we revealed that ZBTB24 is required for their methylated status and their transcriptional repression, and, to our knowledge, the first of its kind to be identified. We demonstrated that ZBTB24 is also required for the establishment step of DNAme at mouse centromeric repeats, through the recruitment of the de novo DNA methyltransferase DNMT3B. In combination with our demonstration of its role in DNAme maintenance at these loci in somatic cells [14], our study lend support to a model where ZBTB24 plays pivotal roles in the maintenance of centromere integrity, which alterations invariably lead to chromosomal instability and disease.

ZBTB24 LOF affects innate immune response-related genes

We reported here the first transcriptomic profiling of ICF2 patients-derived lymphoblastoid cell lines, which highlighted the deregulation of genes involved in signaling pathways linked to the innate immune response, as evidenced by an activation of genes involved in the interferon response. This supports the implication of ZBTB24 in cellular defense mechanisms, a function of the protein that is also supported by other studies reporting an activation of interferon response-related genes, upon knockdown of ZBTB24 in the human B-cell line Raji and the colorectal cell line HCT116, or in a zebrafish ICF2 model [23, 27, 65]. Since these interferon response related genes are not directly bound by ZBTB24, it suggests that their up-regulation is most likely an indirect consequence of ZBTB24 LOF. Of note, autoimmune symptoms have been reported for some ICF2 patients [55, 56], and it is worth mentioning that the activation of the interferon response pathways could account for this clinical trait. Interestingly, ZBTB24 LOF in zebrafish leads to an aberrant de-repression of pericentromeric transcripts that triggers an interferon response in a zebrafish ICF2 model [27]. Combined to the abnormally high levels of centromeric transcripts that we documented in the LCLs derived from ICF2 patients, it is tempting to speculate that these transcripts could trigger RNA surveillance mechanisms and, in turn, an antiviral response-like mediated by an activation of interferon response-related genes identified in ICF2 patient cells.

ZBTB24 is essential for a healthy development in mouse and human

In mice, we showed that ZBTB24 is essential for early embryonic development. This result and the estimated developmental stage window of embryonic lethality are consistent with a previous study using a mouse model (called ΔBTB) carrying a large genetic deletion of the BTB domain, AT-Hook, and the first zinc fingers of ZBTB24 [17]. In contrast, our ICF2 mouse model is a knock-in model which recapitulates a deletion of TA dinucleotides identified in ICF2 siblings. This deletion generates a frameshift and a premature stop codon which theoretically produces a truncated version of ZBTB24 preserving the BTB domain but lacking the AT-hook and all the zinc finger domains. In both mouse models, the full-length ZBTB24 protein was undetectable in western blot and there was no proof of the existence of ZBTB24 truncated peptides. The comparison of RNA-seq data of mESCs cultured in 2i between the two models revealed a higher number of deregulated genes in the ΔBTB compared to our model (82 vs 26 genes downregulated and 57 vs 23 upregulated genes, respectively). Only 7 genes in common (Cdac7, Cdc40, Ostc, Taf6, Cox6b2, Rnf182 and Cnpy4) and 2 genes (Cfh and Lama4) were respectively down- or up-regulated between the two models. Except for Cdca7, also involved in ICF syndrome with convergence at the level of transcription, methylation and clinical signs [15, 17, 55], any relevance of the other genes to ICF syndrome is not clear. However, our omics analyses performed in differentiated mESCs allowed us to add a set of developmental genes occupied by ZBTB24. Among these genes, we identified Arid5b and Axin2 which have not been identified as ZBTB24 regulated genes in the previous ICF2 mouse model (ΔBTB) [17]. In our study, both genes are main targets of ZBTB24 in mouse and human cells and downregulated upon ZBTB24 LOF in differentiated mESCs (Fig. 1C, Fig. 2E and Supplementary Table S9). Loss of ARID5B functions in mice leads to smaller body size, leanness at birth, reduced postnatal growth rate, impact on hematopoiesis, decreased early T and B progenitors and altered chondrogenesis [66, 67]. Since ARID5B is considered as an epigenetic modulator of gene expression owing to its ability to recruit PHF2, a H3K9me2 demethylase [67], its downregulation could affect gene expression programs, and consequently, cell proliferation and differentiation. Given its significant role in development and immunity, it is plausible to speculate that ARID5B downregulation contributes to the ICF2 phenotype. On the other hand, AXIN2 negatively regulates the Wnt signaling pathway by promoting the degradation of β-catenin [68]. AXIN2 plays a crucial role in brain development [69]. During embryogenesis, Axin2 has a specific pattern of expression, with particularly high levels in neural crest precursors and migratory neural crest cells [70]. Given the facial anomalies and the intellectual disabilities observed in ICF2 patients carrying mutations in ZBTB24, the downregulation of AXIN2 could contribute to these clinical features.

In addition, in light of the pivotal role of DNAme in embryonic development [58] and given the deleterious impact of ZBTB24 LOF on the DNA methylome of patients [15] and murine cells (discussed below), it is likely that these alterations of DNAme landscapes could also account for the developmental defects observed in Zbtb24^mt/mt embryos (this study) and human patients [55, 56].

The requirement of ZBTB24 for normal embryonic development seems to be more widely conserved during evolution since ZBTB24 mutations in zebrafish lead to early developmental abnormalities and lethality in adulthood [27], or to a developmental disease in humans. In addition, a strong decrease in ZBTB24 expression is associated with recurrent spontaneous abortion in humans [71]. Overall, ZBTB24 LOF leads to major developmental alterations, strongly supporting a conserved role for this protein during the early stages of development.

Distinct roles of ZBTB24 in shaping DNA methylation patterns

We found that the genome of Zbtb24^mt/mt naïve mESCs was remarkably hypomethylated compared to their wild-type counterpart in which DNAme levels are already minimal [60, 61].

The significant difference in DNA methylation at minor satellites in naïve mESC is likely due to the impairment of DNA methylation maintenance activity. Specifically, the loss of ZBTB24 function impairs the maintenance of DNA methylation, leading to a progressive (passive) loss of DNA methylation during the culture of naïve mESCs in 2i. At day 3 of differentiation, these differences are likely mitigated by the massive increase of DNMT3B expression (Fig. 4E) and the global wave of DNA methylation establishment.

We cannot exclude the possibility of compensatory mechanisms involving different factors known to directly bind DNA and recruit DNMTs [72], such as zinc finger proteins, or a larger centromeric complex in which the centromeric protein MIS18Α reinforces centromere localization of de novo DNMTs [73].

Certain classes of DNA repeats remain methylated in pluripotent mESCs as they escape the wave of demethyation after fertilization [74]. Remarkably, among the sequences that loose DNA methylation upon ZBTB24 LOF, we found a number of these DNA repeats like LINE and ERVs. Hence, although we cannot exclude that ZBTB24 is required for DNA methylation establishment/maintenance in primordial germ cells, ZBTB24 appears as being required for DNAme maintenance at about 10 000 genomic loci. Since ZBTB24 binds regulatory sequences at motifs that do not contain CG dinucleotide, it is unlikely that it has the ability to “read” and then protect the DNAme mark from demethylation waves, unlike ZFP57, another zinc-finger protein implicated in DNAme maintenance at imprinted loci [75, 76]. In addition, although some other members of the BTB-ZF TF family are DNAme readers [77–79], the subnuclear localization of exogenous ZBTB24 at chromocenters in murine cells devoid of DNAme [25] suggest that more sophisticated mechanisms for DNA methylation maintenance are involved.

By mimicking “in-dish” the establishment step of DNAme patterns, which occurs while mESCs are induced to differentiation, we were able to show a limited contribution of ZBTB24 at a genome-wide scale, reinforcing the idea that ZBTB24 may play a role mainly in DNAme maintenance. This observation is in line with previous studies showing progressive loss of DNAme at (peri)centromeric DNA repeats in HEK293 cells upon ZBTB24 knock-out [80] or between larval and adult stages in Zbtb24 mutant zebrafish [27] or else, DNA hypomethylation in HCT116 cells upon ZBTB24 knockdown [23]. It is not known whether DNAme defects in ICF2 patients worsen over time and whether this phenomenon correlates with an exacerbation of clinical signs as reported and suggested in a case report [56].

Importantly, our study highlights the key role of ZBTB24 in DNAme of centromeric satellite sequences. On its own, the hypomethylated status of centromeric repeats in Zbtb24^mt/mt embryos and in ICF2 patient cells is not sufficient to conclude whether DNAme establishment, maintenance or both are altered upon loss of ZBTB24 function. Although we previously reported the requirement of ZBTB24 in DNAme maintenance at centromeric repeats in mouse embryonic fibroblasts [14], its contribution to DNAme establishment during early development has not yet been assessed. Taking advantage of two genetic mESCs knock-in ICF2 murine models (Zbtb24^mt/m ZBTB24-AID), we provided here evidence that ZBTB24 is required for the proper establishment of DNAme at centromeric MinSat repeats. This role of ZBTB24 is supported by its occupancy of MinSat repeats and its ability to recruit the de novo methytransferase DNMT3B, known to specifically establish MinSat DNAme patterns [63], during mESCs differentiation. Along the same lines, a functional link between ZBTB24 and DNMT3B proteins at gene bodies has been reported in HEK 293 cells, also suggesting an interplay between these two proteins in human cells [23]. However, since the physical interaction of ZBTB24 with DNMT3B has not been confirmed in HEK 293 nor in mESCs, we cannot rule out the possibility that the recruitment of DNMT3B to centromeric MinSat repeats is indirect [21]. Moreover, in a context of the crosstalk between DNA and histone methylation, as documented for H3K36me2/me3 and H3K9me2/me3 [81–83], DNMT3B recruitment may depend on a chromatin context shaped by ZBTB24 occupancy, which in turn would allow the recruitment of the DNAme machinery. In the particular case of centromeres, it will be interesting to investigate a possible interplay between ZBTB24 and centromeric proteins shown to recruit DNMT3B to centromeres like CENP-C [84], MIS18A [73] or HELLS [85] and the subsequent methylation of the underlying satellite DNA sequences. Nonetheless, our data showing the conserved occupancy and function of ZBTB24 at mouse and human centromeric satellite loci have added ZBTB24 to the very rare TFs as necessary for the localization of the DNAme machinery at specific places on the genome, and more specifically at centromeric satellite sequences [29, 72, 86].

ZBTB24 is a guardian of centromere integrity

We present here a novel facet of ZBTB24 function at centromeres. Our data provide strong evidence for a conserved and potentially direct role for ZBTB24 at murine and human centromeric satellite sequences. Since these A/T rich sequences lack the DNA recognition motif of ZBTB24 found at regulatory sequences, read through its 6th and 7th zinc finger motifs [24], it is likely that the binding of ZBTB24 at centromeric sequences involves its AT-hook domain, which is known to interact with the minor groove of A/T rich DNA. However, as the AT-hook domain alone has a low affinity for DNA, the affinity of ZBTB24 for centromeric satellite sequences would be potentiated by the joint action of both AT-hook and ZnF motifs [87]. In line with this, the presence of both motifs has been shown to be important for the localization of ZBTB24 to (peri)centromeric heterochromatin [26]. Considering the growing body of evidence that points to a role of DNAme in centromeric integrity [4] and the fundamental role of centromeres in ensuring faithful transmission of the genetic information [88], ZBTB24 emerges as a guardian of centromeric integrity and hence, of genomic stability. In this study, we showed that the loss of DNAme at centromeric repeats upon ZBTB24 LOF was associated with increased levels of centromeric transcripts, more prominently in human cells, which have been shown to lead to chromosomal instability [89–92]. Actually, the loss of DNAme at centromeric sequences could lead to their abnormal transcription, which poses a threat to genomic integrity through the formation of DNA:RNA hybrids and R-loops structures, known to cause replication stress or double strand breaks leading to genomic instability [80, 90, 91, 93–96]. Interestingly, our in-depth analysis of the ChIP-seq data of ZBTB24 in human LCLs highlighted its peculiar enrichment at specific centromeric HORs, mainly the active ones (HOR-1) according to their enrichment in CENP-A and binding sites for CENP-B (CENP-Box) [45]. Among the HORs mostly enriched in ZBTB24 occupancy and with deregulated expression in ICF2 cells, the presence of HOR_1 from chromosomes 1 and 16, which account for chromosomal anomalies in ICF patients, was striking. Whether the deregulation of specific HORs transcription in ICF cells is due to different degrees of DNAme loss or detectability due to a wide range of centromeres sizes in humans [97], and whether hypomethylation at pericentromeric regions of chromosomes 1 and 16 facilitates hypomethylation at nearby centromeric HORs are very exciting questions. These points should be addressed in a near future and be facilitated by the achievement of the complete sequence of the human genome, including centromeric satellite arrays, by the Telomere to Telomere (T2T) consortium [98].

Initially, the identification of three new ICFs factors lacking DNA methyltransferase activity, namely ZBTB24, CDCA7 and HELLS, raised the obvious question of their functional link with the DNAme machinery and centromere biology. Likely because of the known role of HELLS as “a guardian of heterochromatin at repeat elements” in mouse [99], and CDCA7 being downstream of ZBTB24 [17], most of the studies designed so far focused on the role of the CDCA7/HELLS complex [18, 19, 21, 80].

In sum, our work now provides a novel and previously unconsidered dimension of the role of ZBTB24 in maintaining genomic integrity.

Acknowledgements

The authors would like to thank Drs Florence Larminat, Daniele Fachinetti, Catalina Salinas for insightful discussions about this work, Dr Riccardo Gamba for support and advises in centromeric data analysis, Laure Ferry and the Epigenomic Core Facility of the UMR7216 for help with pyrosequencing, Franck Letourneur and Juliette Harmoune of the sequencing platform core facility Genom’IC at Cochin Institute in Paris, Damien Ulveling and Justine Guegan of the bioinformatics and biostatistics core facility ICONICS in Paris for their help and support with sequencing data analysis. We acknowledge all the patients and their family members for their participation in this study.

Author contributions

G.G., E.B., S.M., I.I., A.J., T.D. and G.V. performed the experiments. G.G., M.B, T.D., M.D., M.W. and G.V. performed bioinformatics analysis of sequencing data. G.V. and C.F. designed the study. G.V. and C.F. supervised the project. G.G., E.B. and G.V. wrote the original draft of the manuscript. G.V. and C.F. edited the manuscript.

Conflict of interest statement: The authors declare no conflict of interests. G.G. reports that at the time of publication, he was an employee of T-One Therapeutics srl. (https://tonetx.eu/).

Funding

Work in CF’s lab was funded by the Fondation pour la recherche sur le cancer (ARC), La Ligue contre le cancer (LNCC), the Agence Nationale pour la Recherche (grant ANR-19-CE12-0022), the Fondation Maladies Rares (GenOmics of Rare Diseases, call 2017; Mouse Models for Rare Diseases, call 2013), Fondation Jérôme Lejeune. G.G. was supported by the French Ministry of Research and Fondation ARC. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Author notes

Present address: T-One Therapeutics s.r.l., Via Francesco Sforza 35, 20122 Milan, Italy.

Present address: Institut de Biologie Paris Seine - CNRS UMR7622, INSERM U1156, Sorbonne Université, 75005 Paris, France.