RBBP4 is an epigenetic barrier for the induced transition of pluripotent stem cells into totipotent 2C-like cells

Abstract Cellular totipotency is critical for whole-organism generation, yet how totipotency is established remains poorly illustrated. Abundant transposable elements (TEs) are activated in totipotent cells, which is critical for embryonic totipotency. Here, we show that the histone chaperone RBBP4, but not its homolog RBBP7, is indispensable for maintaining the identity of mouse embryonic stem cells (mESCs). Auxin-induced degradation of RBBP4, but not RBBP7, reprograms mESCs to the totipotent 2C-like cells. Also, loss of RBBP4 enhances transition from mESCs to trophoblast cells. Mechanistically, RBBP4 binds to the endogenous retroviruses (ERVs) and functions as an upstream regulator by recruiting G9a to deposit H3K9me2 on ERVL elements, and recruiting KAP1 to deposit H3K9me3 on ERV1/ERVK elements, respectively. Moreover, RBBP4 facilitates the maintenance of nucleosome occupancy at the ERVK and ERVL sites within heterochromatin regions through the chromatin remodeler CHD4. RBBP4 depletion leads to the loss of the heterochromatin marks and activation of TEs and 2C genes. Together, our findings illustrate that RBBP4 is required for heterochromatin assembly and is a critical barrier for inducing cell fate transition from pluripotency to totipotency.


INTRODUCTION
In mice, zygotes and 2-cell (2C) blastomeres have the capacity to de v elop into all embryonic and extraembryonic cell types, exhibiting a totipotent state ( 1 , 2 ). The totipotency gradually decreases as de v elopment proceeds. The first lineage segregation occurs at the eight-cell stage, which separates the inner cell mass (ICM) from the trophectoderm (TrE) ( 3 , 4 ). Then second lineage segregation occurs at the blastocyst stage, when the ICM is divided into epiblast (EPI) and primitive endoderm (PE) ( 5 ). The EPI, PE and TrE are the precursors of the embryo proper, yolk sac and placenta, respecti v ely ( 6 , 7 ). In contrast, lineage restriction has not been formed in human blastocysts (8)(9)(10). Human na ïve pluripotent stem cells share properties with pre-implantation blastocysts and have been reported to have the potential to dif ferentia te into both embryonic and e xtraembryonic trophob last lineages (10)(11)(12)(13)(14)(15)(16). Pluripotent mESCs are isolated from the ICM ( 17 , 18 ) and contribute to all three germ layers of embryos but rarely to TrEderi v ed lineages ( 4 , 19 ). Ov ere xpression of the transcription factor (TF) genes of the TrE (such as Cdx2 , Elf5 and Eomes ) could lead to the direct conversion of mESCs into TrE-like cells (20)(21)(22)(23)(24), but lineage transition remains insufficient. In fact, the first cell lineage segregation during development is so refractory that even na ïve mESCs cannot fully overcome it ( 25 , 26 ). Therefore, capturing totipotent cells is vital for giving rise to both ICM-and TrE-derived lineages.
A rare subset of cells that resembles 2C stage embryos (known as 2C-like cells, 2CLCs) has been identified among cultured mESCs, and se v eral TF genes (such as Dux , Zs-can4 and Nelfa ) have been found to facilitate the generation of 2CLCs from mESCs (27)(28)(29)(30)(31). Extended or expanded pluripotent stem cells (EPSCs) wer e r eported to be another type of cells with both embryonic and extraembryonic dev elopmental potential ( 32 , 33 ). Howe v er, a recent study has brought into question whether EPSCs have the ability to dif ferentia te into the trophoblast lineages ( 4 ). Several laboratories hav e estab lished various types of totipotent-like stem cells by combining different chemical molecules (34)(35)(36)(37), yet the totipotency of these cells remains to be validated. Mor eover, curr ently, the epigenetic barrier in regulating the transition from pluripotency to totipotency remains unclear.
Her e, we r eport that the histone chaperone RBBP4, but not RBBP7, is necessary for maintaining the identity of mESCs. Acute degradation of RBBP4 in mESCs activates the 2C transcriptional program including transposable elements (TEs) and 2C genes, and enhances trophoblast formation. Remar kab ly, RBBP4 is co-enriched at all three classes of ER Vs (ER V1, ER VK and ER VL) with either H3K9me2 or H3K9me3. Intriguingly, RBBP4 could simultaneously recruit G9a and KAP1 to deposit H3K9me2 and H3K9me3 on the TEs, respecti v ely. Moreov er, RBBP4 increases nucleosome density at TE sites, and RBBP4 depletion reduces chromatin binding of both H3K9me2 and H3K9me3, and reprograms the pluripotent stem cells toward 2C-like totipotent cells.

Rbbp7 -AID stable mESCs
We used the cluster ed r egularly interspaced short palindromic repea ts (CRISPR) / CRISPR-associa ted 9 (Cas9) system to construct the gene-edited cell lines. To construct the Rbbp4 or Rbbp7 donor plasmid, a DNA fragment containing the stop site ( ∼1 kb), which was amplified by polymerase chain reaction (PCR) from the genomic DNA of 46C mESCs, was cloned into a pEASY-Blunt plasmid (TransGen Biotech, CB101). After creating a restriction enzyme BamHI site by inverse PCR, a DNA fragment containing the AID-mCherry tag linked to a SV40 promoter-dri v en neomy cin / kanamy cin resistance gene (AID-mCherry-NeoR) was cloned into the donor plasmid. The AID-mCherry-NeoR backbone was cut from the PTJ58 vector by BamHI. The specific guide RN A (gRN A) targeting the 3 -end sequences of the mouse Rbbp4 or Rbbp7 gene were designed and inserted into a pX330 plasmid (Addgene, 42230), respecti v ely. The single guide RN A (sgRN A) oligos are available in Supplementary  Table S1.
The parental mESCs expressing an OsTIR1 were established by co-transfection of the plasmids pEN396 (Addgene, 92142) and pX330-EN1201 (Addgene, 92144) expressing a gRNA targeting the Tigre locus using FuGENE6 Transfection Reagent (Promega, E2691) following the manufacturer's instructions. Clones were selected with 1 g / ml puromycin for 3 days. After validation of the OsTIR1 expression cassette integration into the Tiger locus, the donor plasmids and their corresponding sgRNA targeting Rbbp4 or Rbbp7 were co-transfected into the parental cell lines. The transfected cells were selected with 200 g / ml G418 for 5 days. Colonies were manually picked up and were further v alidated b y PCR for homozygous insertion of the sequences encoding AID-mCherry.

Plasmid construction and lentivirus production
For ov ere xpression, Rbbp4 and Rbbp7 cDNAs were cloned into the pSin-Fla g-Avita g vector with a puromycin resistance gene. For Rbbp4 knockdown, the shRNA oligos targeting Rbbp4 were inserted into the lentiviral vector pLKO.1. All sequences of the plasmids were validated by Sanger sequencing before using for further experiments. For lentivirus generation, HEK293T cells were plated, cultured overnight and then co-transfected with pSin or pLKO.1 vector containing target genes together with the packaging plasmids psPAX2 and pMD2.G by using polyethyleneimine (PEI; Polysciences, 24765-2). The culture medium was changed after 12 h. The viral supernatants were collected 48 h post-transfection. After filtration through a 0.45 m filter, the supernatants were used to infect mESCs. The sequences of shRNA oligos used in this study are listed in Supplementary Table S2.

Cell viability assay
A pproximatel y 5000 cells were plated per well of 96-well plates. The next day, the cells were cultured in 100 l of mES culture medium with or without IAA (this time point was recorded as day 0). The cell viability was detected at the indicated time points by using Cell Counting Kit-8 (CCK8, Beyotime, C0037) according to the manufacturer's instructions. In brief, 10 l of CCK8 solution was added to each well and incubated for 1.5 h, and the absorbance at 450 nm was measured using a spectrophotometer.

Apoptosis assay
Apoptosis assays were performed using the Annexin V-FITC Apoptosis Detection Kit (Beyotime, C1062) according to the manufacturer's instructions. In brief, 30000 cells were plated per well of 6-well plates and treated with 500 M IAA or PBS f or 24 h bef ore collection. The supernatants of the culture medium were collected. The adherent cells were washed once with PBS, digested with trypsin without EDTA and collected together with the supernatants. The cells were centrifuged and washed once with PBS. The cells were resuspended in 100 l of 1 × Annexin V binding buffer. Then 5 l of Annexin V-fluorescein isothiocyanate (FITC) was added to the cell suspension and incubated in the dark at RT for 10 min. Subsequently, 400 l of 1 × Annexin V binding buffer was added to each sample and the fluorescence intensity was detected using an LSR-Fortessa flow cytometer (BD Biosciences) within 1 h.

Cell cycle analysis
A total of 40000 cells were plated per well of 6-well plates and treated with IAA for 24 h before collection. The cells were washed and then fixed with pre-cooled 75% ethanol overnight a t 4 • C . The cells were washed once with PBS, resuspended with 100 l of PBS containing 0.25 g of 4 ,6diamidino-2-phenylindole (DAPI; Beyotime, C1002) and incubated in the dark at RT for 15 min. Finally, 500 l of PBS was added to each sample after washing and the fluorescence intensity was detected using an LSRFortessa flow cytometer (BD Biosciences).

Alkaline phosphatase (AP) staining
AP staining was performed following the protocol of the BCIP / NBT Alkaline Phosphatase Color De v elopment Kit (Beyotime, C3206). In brief, the cells wer e fix ed with 4% paraformaldehyde (PFA) solution for 5 min and gently washed, followed by staining with BCIP / NBT solution in the dark at RT for 30 min.

Biotin immunoprecipitation
Fla g-Avita g-or Fla g-Avita g-Rbbp4 -ov ere xpressing 46C mESCs were collected and lysed with cell lysis buffer [50 mM Tris-HCl (pH 7.9), 150 mM NaCl, 1% NP-40, 1 mM PMSF, 1 × protease inhibitors, 1 mM DTT]. The supernatants were obtained by centrifugation at 12000 rpm at 4 • C for 10 min and 1 mg of proteins were used for immunoprecipitation. A 15 l aliquot of M280 Streptavidin Dynabeads (Invitrogen, 11205D) was added to the samples and incubated overnight at 4 • C with gentle rotation. Beads were washed four times using 1 ml of cell lysis buffer at 4 • C for 10 min each time. Immunoprecipitated proteins were visualized by western blots.

RN A e xtraction and RT-qPCR analysis
To extract the total RNAs, the cells were lysed with TRIzol reagent (MRC, TR118), followed by addition of 1 / 5 volume of chloroform. After centrifuga tion a t 12000 rpm at 4 • C for 10 min, the total RNAs in the supernatants were precipitated with isopropanol. cDNAs were synthesized with HiScript ® III RT SuperMix for quantitati v e PCR (Vazyme, R323). Quantitati v e re v erse transciption-PCR (RT-qPCR) was performed using SYBR Green mix (Genstar, A301) on the CFX96 system (Bio-Rad). Gene e xpression le v els were normalized to Gapdh . The primers used in the RT-qPCR assays are listed in Supplementary Table S3.

Immunofluorescent staining
Cells were fixed with 4% PFA at RT for 10 min, permeabilized with 0.5% Triton X-100 at RT for 10 min and blocked with 3% bovine serum albumin (BSA) at RT for 1 h. Samples were then incubated with primary antibody overnight at 4 • C and subsequently with secondary antibody. A final concentration of 1 g / ml DAPI (Beyotime, C1002) was used to stain nuclei. Fluorescence images were captured with a ZEISS Axio Observer 7. The EOMES antibody (Abcam, ab183991) used in this study was diluted 1:500.

Bulk RNA-seq
The total RNAs were extracted as described above. RNA sequencing libraries were constructed by using a VAHTS mRNA-seq V3 Library Prep Kit (Vazyme, NR611) according to the manufacturer's instructions. In brief, mRN A ca ptur e beads wer e used to captur e poly(A)-enriched RNAs from 1.5 g of total RNAs and then the RNAs were fragmented at 85 • C for 6 min. Next, the first-strand and secondstrand cDNAs were synthesized. The cDNAs were purified with AMPure XP beads (Beckman Coulter, A63881), followed by end repair, adaptor-ligation, size selection and library amplification. Finally, the libraries were purified using AMPure XP beads and then sequenced on a NovaSeq 6000 sequencer.

Single-cell RNA sequencing (scRNA-seq)
scRNA-seq libraries were prepared by using a DNBelab C Series Single-Cell Library Prep set (MGI, 940-000047-00). In brief, the dissociated cell suspensions were converted to barcoded scRNA-seq libraries thr ough dr oplet encapsulation, em ulsion breakage, mRN A ca ptured bead collection, re v erse transcription, cDNA amplification and purification. The libraries were sequenced on a DNBSEQ-T7 sequencer.

Cleavage under targets and tagmentation (CUT&Tag)
CUT&Tag assays were performed as described previously with some modifications ( 50 ). Briefly, 1 × 10 5 cells were washed twice with wash buffer [20 mM HEPES (pH 7.5), 150 mM NaCl, 0.5 mM spermidine and 1 × protease inhibitors] by gentle pipetting. A 10 l aliquot of concanavalin A-coated magnetic beads (Bangs Laboratories, BP531) was activated and added per sample and incubated at RT for 10 min. The supernatants were removed and beadbound cells were resuspended in 100 l of wash buffer containing 0.01% digitonin and 2 mM EDTA. The H3K9me2 antibody (Cell Signaling Technology, 4658) was added and incubated overnight at 4 • C on a rotator. The beads were washed and resuspended in 300-wash buffer containing 300 mM NaCl with pG-Tn5 (Vazyme, S602) at RT for 1 h. The beads were washed and tagmentation was performed in 300-wash buffer supplemented with 10 mM MgCl 2 at 37 • C for 1 h. To stop the tagmentation reaction, 2.25 l of 0.5 M EDTA, 2.75 l of 10% SDS and 5 l of proteinase K (20 mg / ml) were added and further incubated at 55 • C for 2 h. Then the genomic DNAs were extracted by phenol-chloroform and subjected to PCR amplification using NEBNext High-Fidelity 2 × PCR Master Mix (NEB, M0541S) for 13 cycles. The libraries were size-selected with 1.2 × AMpure XP beads (Beckman Coulter, A63881) and sequenced on a NovaSeq 6000 sequencer.

Cleavage under targets and release using nuclease (CUT&RUN)
CUT&RUN assa ys were perf ormed as described previously with some modifications ( 51 ). Briefly, 1 × 10 5 cells were harvested, washed and bound to activated concanavalin Acoated magnetic beads as for CUT&Tag. The bead-bound cells were incubated with primary antibody in 100 l containing 0.01% digitonin wash buffer [20 mM HEPES (pH 7.5), 150 mM NaCl, 0.5 mM spermidine and 1 × protease inhibitors] on a rotator at 4 • C overnight. The primary antibodies used herein are as follows: H3K9me3 (Abcam, ab8898), KAP1 (Abcam, ab109545), CHD4 (Abcam, ab70469) and G9a (Cell Signaling Technology, 3306). After two washes, the beads wer e r esuspended in 100 l of wash buffer containing pAG-MNase and incubated at 4 • C for 1 h. After two washes in wash buffer and one wash in lowsalt rinse buffer [20 mM HEPES (pH 7.5), 0.5 mM spermidine and 1 × protease inhibitors], the tubes were chilled to 0 • C and the beads were resuspended in 100 l of icecold calcium incubation buffer [20 mM HEPES (pH 7.5), 2 mM CaCl 2 ]. The tubes were kept on ice 30 min, followed by immediate addition of STOP buffer [170 mM NaCl, 20 mM EDTA]. The beads were incubated at 37 • C for 30 min and the liquid was removed to a fresh tube on a magnet stand. Then SDS and proteinase K were added and incuba ted a t 55 • C for 2 h. The genomic DNAs were extracted by phenol-chloroform and the sequencing libraries were prepared through VAHTS Uni v ersal DNA Library Prep Kit (Vazyme, ND607), and sequenced on a NovaSeq 6000 sequencer.

Micrococcal nuclease sequencing (MNase-seq)
MNase-seq was performed as described previously with some modifications ( 52 , 53 ). In brief, 1 × 10 6 cells were harvested, washed in ice-cold PBS and resuspended in 1 ml of ice-cold lysis buffer [10 mM Tris-HCl (pH 7.4), 10 mM NaCl, 3 mM MgCl 2 , 0.5% IGEPAL CA-630, 1 × complete protease inhibitor cocktail] and rotated at 4 • C for 15 min. Nuclei were then pelleted at 300 g at 4 • C for 10 min and washed in 1 ml of digestion buffer [10 mM Tris-HCl (pH 7.4), 15 mM NaCl, 60 mM KCl]. MNase digestion was carried out in 100 l of digestion buffer containing 2 mM CaCl 2 and 2500 gel units of MNase (NEB, M0247S) with shaking at 37 • C for 5 min. The reaction was stopped using an equal volume of stop buffer (digestion buffer containing 20 mM EDTA, 2 mM EGTA) before RNase A and proteinase K treatment. The gDNAs were extracted by phenol-chloroform. Purified DNAs were loaded on a 1.5% agarose gel to determine digestion efficiency. The mononucleosome-DNAs were purified by AMPure XP beads (Beckman Coulter, A63881). The sequencing libraries were prepared from 10 ng of purified mononucleosome-DNAs using the VAHTS ® Uni v ersal DNA Library Prep Kit (Vazyme, ND607) with 10 cycles of PCR, and sequenced on a NovaSeq 6000 sequencer.
For principal component analysis (PCA) (Supplementary Figure S6C), the alignment and quantification processes were conducted as described above. Batch effects wer e r emoved using ComBat with the sva package ( 59 ). Gene set enrichment analysis was performed using GSEA software (v4.1.0) ( 60 ).
RBBP4, RBBP7, G9a and KAP1 peaks were called by MACS2 (v2.2.7.1) ( 66 ) with the parameters '-nomodelp 0.01'. For peak calling of broad histone modifications, MACS2 was applied with the parameters '-broad -q 0.01' for H3K9me2 and with the parameters '-nolambda -broad -q 0.01' for H3K9me3. RBBP4 peaks were filtered for fold enrichment > 2. Peaks were annotated using Homer ( 67 ). The Jaccard statistic r epr esenting the ratio of the intersection of two sets to the union of the two sets was calculated using bedtools (v2.30.0) ( 68 ). Differential peaks between distinct treatments were identified by the Diffbind pipeline and peaks with high confidence were chosen according to the thresholds of a P < 0.05 and an absolute FC > 1.5.
For analysis of TEs, repeat element annotations (Repeat-Masker) were downloaded from the UCSC genome browser database (mm10). TEs shorter than 300 bp or those TEs with < 50 copies were removed from the analysis. RBBP4 enrichment on TEs was calculated by dividing the observed v alue b y the input and was expressed as the log 2 (fold enrichment). Heatmaps and pile ups were generated using Deep-Tools (v3.5.1).

MNase-seq analysis
Raw reads were trimmed by Trim galore (v0.6.7) and mapped to the mouse genome (mm10) using Bowtie2. DANPOS3 ( 69 ) was used to count the length distributions of fragments and to analyze nucleosome position and occupancy with the default parameters. For visualization, the output files were used to create a profile with R. In addition, the meta profiles and boxplot (Figure 5 D) (e.g. around RBBP4-binding sites) were normalized for library size and other technical differences by dividing each profile by its median and then m ultipl ying it by the median over all profile medians in the same plot ( 70 ).
Nucleosome-fr ee r egions wer e evaluated as pr eviously described with some modifications ( 71 ). Regions spanning 1 kb upstream and downstream of the center of the curated RBBP4 peaks were divided into 25 bp bins. The curated genomic locations were ordered by ascending rank according to the tag counts from the central 200 bp. Sites were considered to be nucleosome depleted if the tag counts of their central 200 bp were smaller than that of the average 200 bp.

Single-cell RNA-seq analysis
Smart-Seq2 read files were mapped to the mouse genome (mm10 and Gencode gene annotation vM25) using STAR      on these genes using the CCA integration tool with 30 dimensions as implemented in the 'FindIntegrationAnchors' function. The data were then integrated using 'Integrate-Data' and scaled again using 'SCT'. PCA with 40 principal components was performed by 'RunPCA', and uniform manifold approximation and projection (UMAP) coordinates were computed using the first 26 principal components. Merged datasets were clustered using Seurat's shared nearest-neighbor algorithm implemented with the 'Find-Clusters' function.

Quantification and statistical analysis
Data ar e pr esented as mean values ± standard deviaition (SD) unless otherwise indicated in the figure leg-      ends. Sample numbers and experimental repeats are indicated in the figure legends. Statistical significance was determined by Student's t -test analysis (two-tailed) for two groups, unless otherwise indica ted. Dif ferences in means wer e consider ed sta tistically significant a t P < 0.05. Significance le v els were: * P < 0.05; ** P < 0.01; and *** P < 0.001.

RBBP4, but not RBBP7, is indispensable for maintaining the identity of mESCs
To identify the similarities and differences in the chromatin binding and transcriptional regulation of RBBP4 and RBBP7, we attempted to perform ChIP-seq for RBBP4 and RBBP7 by using commercial antibodies. Unfortunately, these commercial anti-RBBP4 and anti-RBBP7 antibodies did not work properly for ChIP experiments. Ther efor e, we generated Biotin-tagged Rbbp4 and Rbbp 7 mESC lines (Supplementary Figure S1A), respecti v ely, and performed Biotin-RBBP4 and Biotin-RBBP7 ChIP-seq experiments. Our data indicated that most of the binding sites for RBBP4 and RBBP7 were compatible (Figure 1 A, B; Supplementary Figure S1B). To explore whether RBBP4 and RBBP7 ar e functionally r edundant, we constructed an acute auxininduced degron (AID) system to deplete either RBBP4 or RBBP7 protein in mESCs (Figure 1 C). We integrated the AID-mCherry sequence at the C-terminus of the endogenous Rbbp4 or Rbbp7 gene locus by CRISPR / Cas9 genome editing (Supplementary Figure S1C Figure S1E). RBBP4 depletion slowed the proliferation of mESCs and caused apoptosis (Supplementary Figure  S1F, G). Moreover, RBBP4 depletion arrested cells at the G 1 and S phases, demonstrating that RBBP4 is necessary for normal self-renewal of mESCs (Supplementary Figure  S1H, I). Furthermore, RBBP4 depletion also resulted in a se v er e decr ease in AP staining (Supplementary Figur e S1J), leading to a dif ferentia ted phenotype and destruction of the normal mESC morphology (Supplementary Figure S1K). In contrast, RBBP7 depletion had little effect on cell proliferation, morphology and cell cycle of mESCs (Supplementary Figure S1F-K). Altogether, our data demonstrate that RBBP4 is necessary for maintaining the identity of mESCs, but that RBBP7 is dispensable.

RBBP4, but not RBBP7, r epr esses the tr anscriptional activity associated with the 2C-like cell program
To examine the effect of RBBP4 or RBBP7 on gene expression, we performed RNA-seq experiments after acute depletion of either RBBP4 or RBBP7, and PCA indicated high reproducibility between two biological replicates (Supplementary Figure S1L). A large number of genes were up-regulated after RBBP4 degradation, while only a few wer e alter ed after RBBP7 degradation ( Supplementary Figur e S1M). Mor eover, the numbers and le v els of differentially expressed genes increased with the duration of IAA tr eatment (Supplementary Figur e S1M). Additionally, we observ ed that v ery fe w differentially e xpr essed genes wer e shared between RBBP4-depleted and RBBP7-depleted cells (Supplementary Figure S1N). These results suggest that the degradation of RBBP4 and RBBP7 led to distinct gene expression patterns, which is consistent with the different phenotypes after depletion of RBBP4 and RBBP7 (Supplementary Figure S1F-K). Intriguingly, we found that RBBP4 degradation significantly up-regulated the expression of Zscan4 family members, Dux and the MERVL elements (including MT2 Mm and MERVL-int) (Figure 1 E), which are typical 2C markers ( 27 , 28 , 74 ). Consistently, by comparing our RNA-seq data with pre viously pub lished data on early-2C genes ( 75 ), we further observed that RBBP4 degradation strikingly activated the genes expressed at the 2C stage ( Figure 1 F; Supplementary Table S4). RT-qPCR confirmed the activation of 2C-specific genes ( Dub1 , Zscan4 and Dux ) and MERVL at 24 h after RBBP4 depletion (Figure 1 G). Conversely, RBBP7 degradation had little effect on the expression of 2C-specific genes, as re v ealed by RNA-seq and RT-qPCR analyses (Figure 1 E-G). To further demonstrate the role of RBBP4 in r epr essing the 2C gene program, MERVL-tdTomato reporter mESCs ( 76 ) were used to deplete Rbbp4 and Rbbp7 by using shRNA oligos (Supplementary Figure  S2A). Our data indicated that the expression of MERVL and 2C-like genes ( Zscan4 and Dub1 ) and the population of MERVL-tdTomato cells were significantly increased in the Rbbp4 -depleted cells (Supplementary Figure  S2B-D), but not in Rbbp7 -depleted cells (Supplementary Figure S2E-G).
To further examine whether RBBP4 degradation specifically promotes the transition of mESCs to 2CLCs, but not other embryonic stage-like cells, we performed scRNAseq by using Rbbp4 -AID mESCs cultured with or without IAA treatment, and integrated the data with published single-cell transcriptomes of mouse early embryos (77)(78)(79)(80). UMAP showed that a fraction of cells (7.38%, 622 / 8432) (named totipotent-like cells) among the RBBP4-depleted cells were clustered closer to 2C embryos, representing an ∼7-fold increase over the corresponding fraction of cells without RBBP4 degradation (1.06%, 97 / 9122) ( Figure 1 H;  Supplementary Figure S3A). The correlation between these totipotent-like cells and early embryos was recalculated and showed that these cells had a strong resemblance to midand late-2C embryos (Supplementary Figure S3B). Consistently, these totipotent-like cells expressed totipotency genes (such as Zscan4 , Usp17l , Gm13119 and Gm4027 ) at a high le v el and pluripotency genes (such as Oct4 and Nanog ) at a low le v el (Supplementary Figure S3C). Collecti v ely, our results strongly support that RBBP4, but not RBBP7, plays a vital role in regulating the stem cell fate transition from pluripotency to totipotency.

Rbbp4 depletion enhances trophoblast formation
Since totipotent-like cells, but not mESCs, have the capacity to de v elop into e xtraembryonic linea ges, we took advanta ge of Oct4 -EGFP (enhanced green fluorescent protein) and Cdx2 -tdTomato dual fluorescent reporter (OG-CT) mESCs ( 81 ) to explore whether trophoblast could be derived from the Rbbp4 -depleted cells in mouse TSC medium ( 82 ) (Figure 2 A). As expected, Rbbp4-depleted OG-CT mESCs activated the expression of 2C markers (MERVL and Zs-can4 ) (Supplementary Figure S4A). mESCs cultured in TSC medium gradually formed epithelial-like cells. The dif ferentia ted cells deri v ed by loss of Rbbp4 were mostly multila yered and f ormed cellular aggregates, while the differentiated control mESCs exhibited a flat layered cellular structur e (Figur e 2 B). We next analyzed the expression of pluripotency and trophoblast genes and observed that the expression of pluripotency genes (including Oct4 , Nanog , Sall4 and Tdgf1 ) was decreased in all differentiated cells (Supplementary Figure S4B) and the expression of trophoblast genes (including Cdx2 , Gata3 , Elf5 , Esx1 and Ascl2 ( 22 , 24 , 83-85 )) was dramatically increased only in Rbbp4 -depleted cells but not in control cells (Figure 2 C). Consistent with these findings, the expression of OCT4-EGFP was rarely observed in all dif ferentia ted cells, but the robust expression of CDX2-tdTomato and EOMES was observed only in Rbbp4 -depleted cells (Figure 2 D, E). Flow cytometry analysis showed that the proportion of Cdx2positi v e cells produced by Rbbp4-depleted mESCs was increased by > 4.5 times compared with that of the control cells ( Figure 2 F; Supplementary Figure S4C). Our results suggested that Rbbp4 depletion enhanced cell fate conversion from mESCs to trophoblast lineages.

RBBP4 regulates the expression of ERVs
To investigate how RBBP4 negati v ely regulates totipotency, we focused on the chromatin distribution of RBBP4 binding and observed that RBBP4 was mostly enriched at TEs (46.7%) and a small percentage at promoters (14%) (Figure 3 A). Intriguingly, we found that RBBP4 depletion upregulated 69 TE families (Supplementary Table S5 Figure S5A). TEs , especially ERVs , are abundantly activated during zygote genome activation (ZGA), which is critical for embryonic totipotency, and can function as cis -regulatory elements to regulate expression of their neighboring genes ( 86 , 87 ). Interestingly, we found that RBBP4 peaks were mostly enriched at ERVs and long interspersed nuclear elements (LINEs) (Supplementary Figure S5B, C). Furthermore, among these RBBP4-bound TEs, ERVs, such as MER VL (ER VL), RLTR10-int (ERVK) and RLTR6B Mm (ERV1), tended to exhibit much more increased expression after RBBP4 depletion (Supplementary Figure S5D,  E). Notably, RBBP4-bound MERVL elements were most highly expressed after RBBP4 degradation (Figure 3 C). Moreov er, we observ ed that the transcription start sites (TSSs) of the up-regulated genes were in closer proximity to the RBBP4-bound TEs than to other TEs ( Figure  3 D). Additionally, the distances between the strongly upregulated early-2C genes [log 2 FC > 5, false discovery rate (FDR) < 0.01] and RBBP4-bound TEs were < 1 kb (Supplementary Figure S5F), suggesting that RBBP4 might control the expression of these 2C genes by regulating TEs.
To explore how RBBP4 regulates TEs, we conducted hierarchical clustering of RBBP4 peaks with both acti v e (H3K4me1, H3K4me3 and H3K27ac) ( 88 , 89 ) and r epr essed (H3K9me1, H3K9me2, H3K9me3 and H3K27me3) ( 89-92 ) histone marks over genes and TE regions and observed that RBBP4 was mostly associated with two r epr essi v e histone mar ks H3K9me2 and H3K9me3 at TE sites, respecti v ely (Figure 3 E; Supplementary Figure  S5G , H). Remarka bly, strong enrichment of H3K9me2 and H3K9me3 at RBBP4-bound TEs relati v e to other TE sites was observed (Figure 3 F). The enrichment analyses showed that RBBP4 and H3K9me2 were mainly enriched at ERVL sites, and RBBP4 and H3K9me3 were mainly enriched on ER VK, ER V1 and LINE sites (Figure 3 G). Specifically, RBBP4 binding densities were positi v ely correlated with H3K9me2 on MER VL (ER VL) elements (Figure 3 H), and with H3K9me3 on IAPEz-int and RLTR10-int (ERVK) elements (Figure 3 I). Together, these data demonstrate that RBBP4 globally r epr esses the transcription of ERVs, and co-binds together with heter ochr omatin marks either H3K9me2 at ERVL sites or H3K9me3 at ERVK sites.

RBBP4 is r equir ed f or heter ochr omatin formation
To investigate whether heter ochr omatin formation is dependent of RBBP4, we identified the binding sites for H3K9me2 and H3K9me3 in Rbbp4 -AID mESCs with or without IAA trea tment. Our da ta indica ted tha t RBBP4 degradation resulted in the significant reduction of 5068 (12%) H3K9me2 peaks (42. At least fiv e H3K9-specific methyltr ansfer ases have been described in mammals: SUV39H1, SUV39H2, SETDB1, G9a and GLP ( 93 ). Next, we sought to determine whether RBBP4 establishes H3K9me2 / 3 on these TEs by regulating H3K9 methyltr ansfer ases or the associated factors. Using the ProteomicsDB database ( 94 ), we found that the protein e xpression le v el of RBBP4 showed a strong correlation with that of KAP1 ( R = 0.80, P = 2.95e-22), which recruits SETDB1 to the target sites ( 95 ), with that of G9a ( R = 0.70, P = 6.29e-15) and with that of GLP ( R = 0.70, P = 9.38e-15), but was relati v el y weakl y correlated with the e xpression le v els of other lysine methyltr ansfer ases ( R < 0.6) (Supplementary Figure S6A). Moreover, our data indica ted tha t KAP1-binding sites were highly correlated with RBBP4-dependent H3K9me3-marked TEs and that G9abinding sites were highly associated with RBBP4-dependent H3K9me2-marked TEs (Supplementary Figure S6B). Furthermor e, PCA r e v ealed that RBBP4 depletion in mESCs led to similar gene expression patterns to both Kap1 knockout ( 96 ) and G9a knockout ( 97 ), which resulted in a gene expr ession signatur e similar to that of pr eviously r eported 2CLCs ( 29 , 98 ) (Supplementary Figure S6C). These results suggest that both KAP1 and G9a might be crucial for RBBP4-mediated heter ochr omatin formation.
More importantly, RBBP4 depletion had no significant influence on the protein le v els of G9a and KAP1 (Supplementary Figure S6D), but diminished their binding to RBBP4-bound sites (Figure 4 D). Furthermore, RBBP4 degradation resulted in a reduction of binding not only of G9a at ERVL elements but also of KAP1 at ERVK elements (Supplementary Figure S6E-G). Specifically, the occupancy of both G9a on MERVL sites and KAP1 on IAPEz-int sites was significantly diminished (Figure 4 H, I; Supplementary Figure S6H). These data suggest that RBBP4 is an upstream regulator of both KAP1 and G9a in regulating TEs. Additionall y, Biotin-imm unoprecipitation analysis showed that RBBP4 interacted with both G9a and KAP1 (Figure 4 J). Together, these results indicate that RBBP4 could regulate heter ochr omatin establishment in mESCs by recruiting KAP1 and G9a.

AID-mediated RBBP4 degr adation r educes nucleosome occupancy at the regions of ERVK and ERVL within heter ochr omatin
RBBP4 is also a component of the NuRD complex ( 39 ), and CHD4 within this complex is responsible for remodeling chromatin structure in an ATP-dependent manner ( 99 , 100 ). Indeed, our Biotin-RBBP4 coimmunoprecipita tion experiments indica ted tha t RBBP4 also interacted with CHD4 (Figure 4 J). Next, we performed MNase-seq to determine the effects of RBBP4 on nucleosome organization and noticed that RBBP4 degradation a ttenua ted nucleosome occupancy around RBBP4-binding sites and decreased CHD4 binding ( Figure 5 A; Supplementary Figure S7A). Moreover, we found that 2C genes and 2C-associated TEs (including MERVL) were up-regulated in Chd4 -depleted mESCs by analyzing previously published RNA-seq data ( 101 ) (Supplementary Figure S7B, C). To understand the relationship between RBBP4 and local nucleosome density in more depth, we split the RBBP4 peaks into three classes (C1-C3) according to the RBBP4 enrichment str ength (Figur e 5 B). Our data showed that the strongest RBBP4-bound C1 sites had the highest intensity of nucleosome occupancy and CHD4 binding (Figure 5 C). Corr espondingly, the r eduction of nucleosome density and CHD4 binding was positi v ely correlated with the RBBP4 binding strength (Figure 5 D). Moreover, ∼75.3, 68.2 and 64.9% of the decreased nucleosome occupancy at C1, C2 and C3 sites, respecti v ely, was loca ted a t TE r egions (Figur e 5 E). In addition, our da ta showed tha t nucleosome occupancy was reduced at the RBBP4-bound ERVK (RLTR10-int and MMERVK10C-int) and ERVL (MERVL and MTA Mm-int) sites, whereas nucleosome occupancy at RBBP4-bound ERV1 and LINE sites was unaffected ( Figure 5 F, G; Supplementary Figure S7D). Together, these results demonstrate that RBBP4 could recruit CHD4 to remodel nucleosome occupancy at the ERVK and ERVL sites within heter ochr omatin regions.

DISCUSSION
Understanding how totipotent cells are generated will provide more insight for stem cell biology and promote the Nucleic Acids Research, 2023, Vol. 51, No. 11 5427 progr ess of r egenerati v e medicine. Here, we show that the histone chaperone RBBP4 acts as a powerful barrier by regulating heter ochr omatin assembly for reprogramming of the pluripotent state toward the totipotent 2C-like state. RBBP4 recruits different epigenetic factors to distinct TEs and modulates chromatin compaction through chromatin remodelers.
A pproximatel y 40% of the mouse genome is composed of TEs, of which approximately a quarter are ERVs ( 95 ). ERVs are grouped into three classes based on sequence similarity with different exogenous retroviruses ( 93 ), and their regulatory mechanisms are quite complicated. Distinct classes of ERVs are regulated by different chromatin marks and associated epigenetic factors. For instance, KAP1 has been identified as a key r epr essor of ERVK elements in mESCs and early embryonic de v elopment ( 95 , 102 ). KAP1 functions as a scaffold for multiple r epr essi v e partners, such as the KRAB-ZFP proteins, SETDB1 and METTL3 ( 96 , 103 , 104 ). In contrast, silencing of ERVL elements by chroma tin-associa ted factors has not been systematically characterized. Although G9a and H3K9me2 have been shown to r epr ess MERVL elements ( 105 , 106 ), how they ar e recruited to genomic targets and form a regulatory network is not fully understood.
Histone chaperones are responsible for histone deposition onto DNA to form chromatin during various processes, such as DNA replication, transcription and repair ( 107 , 108 ). Intriguingly, we found that the histone chaperone RBBP4 binds to ERVs and functions as an upstream factor to recruit G9a and KAP1, followed by deposition of H3K9me2 and H3K9me3 at ERVL and ERVK elements, respecti v ely. Furthermore, RBBP4 interacts with HP1 (Figure 4 J), which is important for heter ochr omatin spreading ( 109 ). Ther efor e, we conclude that RBBP4 acts as a critical regulator for heter ochr omatin establishment and possible spreading.
In addition to RBBP4, the two largest subunits (CHAF1A and CHAF1B) of the CAF-1 complex have also been shown to regulate the establishment of heter ochr omatin (110)(111)(112). CHAF1A / B could inhibit the expression of unintegrated HIV-1 DNA ( 112 ), but RBBP4 only r epr esses the transcription of integrated HIV-1 proviral DNA ( 113 ). Ther efor e, the biological functions of RBBP4 and CHAF1A / B in regulating heter ochr omatin formation might be independent or complementary. The relationships between RBBP4 and the CAF-1 complex still need further exploration.
ATP-dependent chromatin remodelers (including four families CHD, SWI / SNF, ISWI and INO80) can slide, eject, insert or replace histones within nucleosomes to alter chromatin structure and accessibility ( 70 ). In the totipotent cells, chromatin is highly r elax ed, especially at TE r egions ( 75 , 114 ), but it still remains unclear what chromatin remodelers do. Our results show that RBBP4 depletion leads to the changes in colony morphologies of mESCs, which is consistent with Chd4 deficiency ( 115 ). Moreover, RBBP4 depletion activates TEs, such as MERVL, which is similar to loss of CHD4 ( 101 ). Importantly, RBBP4 depletion reduces CHD4 enrichment and nucleosome occupancy around TEs ( Figure 5 ). RBBP4 and CHD4 co-exist in the NuRD complex, which also contains the deacety-lase HDAC1 / 2 and other accessory proteins ( 116 , 117 ). Surprisingly, we observed that the total H3 acetylation le v el was not changed after acute degradation of RBBP4 (data not shown), suggesting that histone acetylation is dispensable for transcriptional activation in mESCs. In addition, CHD4 is also a member of the ChAHP complex (comprising CHD4, ADNP and HP1), which plays an important role in heter ochr oma tin organiza tion ( 118 , 119 ). This direct and indirect evidence suggests that RBBP4 plays vital roles in r epr essi v e chromatin deposition and remodeling.
A recent study showed that mouse totipotent cells and 2CLCs exhibit much slower DNA replication speed than pluripotent stem cells ( 120 ). Our results showed that the cell prolifer ation r ate is retarded in RBBP4-depleted mESCs and the cell cycle is arrested at the G 1 and S phases along with a global decrease in the G 2 phase (Supplementary Figure S1I). These results also suggest that RBBP4-depleted mESCs acquire features similar to 2CLCs. RBBP4 depletion changes the cell cy cle, possib ly because RBBP4 also exists in DNA replica tion-modula ting complexes, including DREAM, CAF-1 and CRL4 (121)(122)(123), and RBBP4 has been found at the replication forks ( 124 ). In addition, the NuRD chromatin remodeling complex might be involved in heter ochr omatin assembly during S phase, and loss of CHD4 results in a slo w-gro wth phenotype with delayed S phase progression ( 125 , 126 ). These results indicate that RBBP4 and CHD4 might work in concert in heter ochr omatin assembly and epigenetic inheritance through DNA replication.
Our results show that RBBP4, but not its homolog RBBP7, is r equir ed for maintaining the identity of mESCs, which is consistent with a recent study that used a different RBBP4 / RBBP7 deletion strategy ( 46 ). We further found that RBBP4 but not RBBP7 is a r epr essor of 2C genes, which coincides with the different temporal and spa tial expression pa tterns of RBBP4 and RBBP7 during pre-implantation embryo de v elopment. The e xpression le v el of RBBP4 is lower at the mouse early 2C stage than at the other pre-implantation sta ges, b ut RBBP7 is highly expressed in oocytes and gradually decreased in later stages (data not shown). Our data suggest that RBBP4 and RBBP7 are not redundant in mESCs. The expression le v els of RBBP4 and RBBP7 also vary between different tissues in mice ( 40 ), and RBBP4 and RBBP7 only have one ortholog: p55 in Drosophila and RebL1 in Tetrahymena thermophila , respecti v ely ( 127 , 128 ). The e volutionary variation among different species and the different expression patterns of RBBP4 and RBBP7 might explain such functional di v ersity in higher eukaryotes.
Over all, this study illustr ates that RBBP4 plays a vital role in regulating heterochromatin assembly in mESCs and that loss of RBBP4 acti vates e xpression of a group of TEs and 2C genes to reprogram stem cell fate from pluripotency to totipotency ( Figure 6 ).

SUPPLEMENT ARY DA T A
Supplementary Data are available at NAR Online.