Relaxed 3D genome conformation facilitates the pluripotent to totipotent-like state transition in embryonic stem cells

Abstract The 3D genome organization is crucial for gene regulation. Although recent studies have revealed a uniquely relaxed genome conformation in totipotent early blastomeres of both fertilized and cloned embryos, how weakened higher-order chromatin structure is functionally linked to totipotency acquisition remains elusive. Using low-input Hi-C, ATAC-seq and ChIP-seq, we systematically examined the dynamics of 3D genome and epigenome during pluripotent to totipotent-like state transition in mouse embryonic stem cells (ESCs). The spontaneously converted 2-cell-embryo-like cells (2CLCs) exhibited more relaxed chromatin architecture compared to ESCs, including global weakening of both enhancer-promoter interactions and TAD insulation. While the former correlated with inactivation of ESC enhancers and down-regulation of pluripotent genes, the latter might facilitate contacts between the putative new enhancers arising in 2CLCs and neighboring 2C genes. Importantly, disruption of chromatin organization by depleting CTCF or the cohesin complex promoted the ESC to 2CLC transition. Our results thus establish a critical role of 3D genome organization in totipotency acquisition.


INTRODUCTION
In mammals, intense chromatin remodeling occurs shortly after fertilization, which involves both reprogramming of epigenetic modifications and reorganization of chromatin architecture (1,2). This process prepares both maternal and paternal genomes for the zygotic genome activation (ZGA), which takes place at the late 1-cell and 2-cell (2C) stages in mouse embryos and is characterized by transient expression of a group of 2C-specific genes and repeats, such as Zscan4 cluster genes and murine endogenous retrovirus with leucine tRNA (MERVL) repeats (3)(4)(5). It is noteworthy that ZGA is coupled with the acquisition of totipotency, the ability of a cell to generate all cell types in an organism, including both the embryo proper and extraembryonic tissues. In contrast, embryonic stem cells (ESCs), which are derived from the inner cell mass (ICM) of the blastocyst-stage embryos, are considered pluripotent, as they can differentiate into all three germ layers and germ cells of the embryo but rarely into extraembryonic tissues. Intriguingly, a rare dynamic subset of cells in mouse ESC cultures has been discovered to possess the ability to contribute to both embryonic and extraembryonic tissues as well, showing a totipotent-like developmental potency similar to the 2C-stage blastomeres and highly expressing 2Cspecific transcripts (6). These cells are thus known as 2C-like cells (2CLCs) and have been demonstrated to be a unique model for understanding totipotency and ZGA. Nevertheless, it remains elusive what defines the chromatin state of a totipotent cell.
Proper 3D genome architecture is crucial for gene regulation. It has been reported that the higher-order chromatin structures of mouse zygotes and 2C embryos are remarkably relaxed, with weakened topologically associating domains (TADs) compared to those of later-stage embryos and somatic cells (7,8). Similar relaxed chromatin ar-chitectures of early embryos were also noted in other organisms including fly, fish, and human (9)(10)(11). Moreover, when a somatic nucleus is reprogramed to the totipotent state during somatic cell nuclear transfer (SCNT), its chromatin architecture becomes markedly relaxed as well (12). It appears that the relaxed genome architecture is a unique feature of totipotency. Indeed, the totipotent 2CLCs have also been reported to exhibit increased histone mobility and chromatin accessibility as compared to ESCs (13,14). However, the higher-order chromatin structure in 2CLCs has not been studied genome wide, and how relaxed genome architecture is functionally linked to totipotency remains elusive. In this study, we profiled the dynamics of both 3D genome and epigenome during the spontaneous pluripotent to totipotent-like state transition in mouse ESCs and defined the role of chromatin architecture relaxation in establishing totipotency.

Analysis of mitotic cells and cell viability
Analysis of mitotic cells was performed as previously described (16). Briefly, cells were stained with the anti-H3S10ph antibody (a kind gift from Dr Fangwei Wang) and Hoechst 33342, and measured by flow cytometry. Cell viability was measured by 7-AAD and Annexin V staining followed by flow cytometry analysis.

sisHi-C
ESCs and 2CLCs were purified by fluorescence-activated cell sorting (FACS), and 100 000 cells were used for each sisHi-C library preparation. The sisHi-C libraries were prepared as previously reported with minor modifications (8). Briefly, the sorted cells were cross-linked with 1% formaldehyde for 10 min and quenched with 0.2 M glycine for 10 min, and then re-suspended in lysis buffer (10 mM Tris-HCl (pH 7.4), 10 mM NaCl, 0.5% NP-40, 0.1 mM EDTA and 1× proteinase inhibitor) for nucleus extraction. The extracted nuclei were further lysed in 10 l of 0.5% SDS solution for 10 min at 62 • C, followed by adding 28 l of 2% Triton X-100 solution and 15-min incubation at 37 • C. Genomic DNA was then digested overnight at 37 • C by adding 100 U of MboI and 5 l of 10× NEB buffer 2. After inactivation of MboI enzyme, cohesive ends were filled in by adding 4.5 l of dCTP/dGTP/dTTP mix (0.33 mM each), 3.75 l of biotin-14-dATP (0.4 mM) and 10 U of Klenow fragment, and incubation at 37 • C for 90 min. Subsequently, 39 l of water, 12 l of 10× T4 ligase reaction buffer, 7 l of 10% Triton X-100 solution, 1.2 l of 10 mg/ml BSA and 400 U of T4 DNA ligase were added for proximity ligation (5.5 h at room temperature). After crosslinking reversal, genomic DNA was purified and sheared to a length of ∼400 bp, and point ligation junctions were pulled down with Dynabeads MyOne Streptavidin C1 (65001, Thermo Fisher) for sequencing library preparation. The final libraries were sequenced on an Illumina HiSeq X Ten platform in the paired-end mode.

ULI-NChIP-seq and CUT&Tag
For ULI-NChIP-seq, 10 000 or 40 000 cells were used per reaction. The ULI-NChIP procedure was performed as previously described (17,18). Briefly, 2 g of H3K27ac antibody (39133, Active Motif), H3K4me1 antibody (39297, Active Motif), H3K4me3 antibody (9727, Cell Signaling Technology), H3K27me3 antibody (9733, Cell Signaling Technology) or H3K9me3 antibody (39161, Active Motif) was used for each immunoprecipitation reaction. The sequencing libraries were generated using the NEBNext Ultra II DNA Library Prep Kit for Illumina (E7645, New England Biolabs) following the manufacturer's instructions. The CUT&Tag assays for CTCF were performed using 50 000 sorted ESCs and 2CLCs as previously described (19). Two or three replicates were performed. Barcoded libraries were pooled and sequenced on an Illumina HiSeq X Ten platform in the paired-end mode.

ATAC-seq
The ATAC-seq libraries were prepared as previously described with minor modifications (20). Briefly, samples were lysed in lysis buffer (10 mM Tris-HCl (pH 7.4), 10 mM NaCl, 3 mM MgCl 2 and 0.1% Igepal CA-630) for 10 min on ice to prepare the nuclei. Immediately after lysis, nuclei were spun at 500 g for 5 min to remove the supernatant. Nuclei were then incubated with the Tn5 transposome and tagmentation buffer at 37 • C for 30 min (TD501, Vazyme Biotech). After the tagmentation, the products were purified with 2.0× VAHTS DNA Clean Beads (N411, Vazyme Biotech). PCR was performed to amplify the library for 14 cycles using the following PCR conditions: 72 • C for 3 min; 98 • C for 30 s; and thermocycling at 98 • C for 15 s, 60 • C for 30 s and 72 • C for 3 min; following by 72 • C 5 min. After the PCR reaction, libraries were purified with the 1.2× VAHTS DNA Clean Beads. Barcoded libraries were pooled and sequenced on an Illumina HiSeq X Ten platform in the paired-end mode.

RNA-seq and data analysis
RNA-seq libraries were prepared using the Smart-seq2 method as previously described (21). Barcoded libraries were pooled and sequenced on an Illumina HiSeq X Ten platform in the paired-end mode. Raw reads were trimmed to 50 bp and mapped to the mouse genome (mm9) using TopHat (version 2.1.1) with default parameters. Only uniquely mapped reads were kept for downstream analysis. Gene counts were generated using HTSeq-count (version 0.9.1). For each sample, the gene count matrices were merged together and then the 'Trimmed Mean of M values' normalization (TMM) method (edgeR package in Bioconductor, version 3.5.0) was used to calculate the normalized expression. P values were generated in edgeR (exactTest) and then adjusted to the false discovery rate (FDR). The significantly up-regulated and down-regulated genes were identified as genes with FDR < 0.05, log 2 (fold change) greater than 1.5 and less than −1.5, respectively.

Hi-C data analysis
Raw reads were first processed with HiC-Pro (version 2.11.1). Briefly, raw reads were trimmed to 50 bp and then aligned to the mouse genome (mm9) using the Bowtie 2 end-to-end algorithm and '-very-sensitive' option. To rescue the chimeric fragments spanning ligation junctions, the ligation site was detected and the 5 fraction of the reads was aligned back to the reference genome. Unmapped reads, multiple mapped reads and singletons were then discarded. Pairs of aligned reads were then assigned to MboI restriction fragments. Read pairs from uncut DNA, self-circle ligation and PCR artefacts were filtered out, and the valid read pairs involving two different restriction fragments were used to build the contact matrix. Valid read pairs were then binned at a specific resolution by dividing the genome into bins of equal size. The binned interaction matrices were then normalized using the iterative correction method in HiC-Pro with default parameters to correct for biases such as GC content, mappability and effective fragment length. For each chromosome, we obtained the expected Hi-C contact values by calculating the average contact intensity for all loci at a certain distance. We then transformed the normalized Hi-C matrix into an observed/expected (O/E) matrix by dividing each normalized observed value by its corresponding expected value. We identified A and B compartments and TADs from Hi-C data using cworld (Dekker lab, https://github.com/dekkerlab/cworld-dekker). To calculate PC1 values, we used the "matrix2compartment.pl" script using contact maps binned at 500 kb resolution as input.
Bins with positive or negative PC1 values were defined as A or B compartment, respectively. The compartment strength was then calculated as (AA + BB)/2AB as previously described (22). To calculate insulation scores, we used the 'ma-trix2insulation.pl' script with default parameters at 10 kb resolution. Insulation scores were further normalized by dividing each bin's insulation score by the average scores in the nearest 2 Mb region, log2-transformed, and multiplied by −1. The 'insulation2tads.pl' script was used with default parameters to identify TADs. Stable TADs between two samples were identified as previously described (7). Directionality index (DI) scores were calculated using a previously described pipeline (23). The relative frequency of inter-TAD interactions was calculated from the number of interactions between two TADs divided by the total number of cis-interactions in the corresponding chromosome. Aggregate Hi-C contact maps were generated by extracting subsets from the OE matrix (these can be either single regions along the diagonal, or region pairs corresponding the matrix segments off the diagonal) and averaging all resulting sub-matrices. If the sub-matrices were of different size, they were interpolated to a fixed size. TAD strength was calculated as the mean of values in the OE matrix in the TAD region. Loop strength was calculated as the mean of values in the OE matrix in the nearest 10 regions of both anchors. Global contact decay curves were plotted as previously described (24). In order to better visualize differences between conditions, we calculated the distribution of the Hi-C contacts as a simple contact probability (sum of the observed counts per log 2 bin, divided by total observed contacts, without normalizing for the bin size).

siRNA transfection
ESCs were transfected with siRNAs (Gene Pharma) using the DharmaFECT 1 (T-2001, Dharmacon) reagent following the manufacturer's instructions. The control siRNA (siNC) was derived from sequences in the Caenorhabditis elegans genome and does not target any mammalian genes. Cells were collected for RNA extraction 2 days after transfection.

RNA isolation, and RT-qPCR
Total RNA was isolated using RNeasy Plus Mini Kit (74134, Qiagen), cDNA was generated using HiScript II Q RT Super Mix with gDNA wiper (R223, Vazyme Biotech), and qPCR was performed using the AceQ qPCR SYBR Green Master Mix (Q111, Vazyme Biotech). Relative quantification was performed using the comparative Ct method with normalization to Actin. Primer sequences are available in Supplementary Table S2.

Statistical analysis
Statistical analyses were performed with the R software. Statistics were calculated using the Wilcoxon signed rank test unless indicated otherwise.

Global weakening of the 3D genome conformation during ESC to 2CLC transition
To investigate the 3D genome architecture in 2CLCs, we generated a mouse ESC line containing a td-Tomato transgene driven by the MERVL promoter (MERVL::tdTomato), which has been established as a faithful indicator of the 2C-like state (6). Consistent with previous reports, about 0.4% of tdTomato+ cells (i.e. 2CLCs) were observed in this ESC line by flow cytometry (Supplementary Figure S1A). These 2CLCs were viable and showed a similar percentage of mitotic cells as compared to ESCs (Supplementary Figure S1B and C). We then purified tdTomato+ 2CLCs and tdTomato-ESCs and performed RT-qPCR and RNA-seq analyses to validate the fidelity of the purified 2CLCs. As expected, 2C-specific transcripts such as Zscan4 family genes, Dux, Dub1, Zfp352 and MERVL repeats were highly expressed in 2CLCs (Supplementary Figure S1D-G), and RNA-seq revealed more upregulated genes than down regulated genes (3)(4)(5). Importantly, the 1044 up-regulated genes and 212 down-regulated genes respectively showed higher and lower expression in 2C embryos than in the ICM of blastocyst-stage embryos (Supplementary Figure S1H). Pathway analyses also revealed an enrichment of pluripotent genes in the down-regulated genes (Supplementary Figure S1I and J), consistent with a previous report that down-regulation of pluripotent genes occurred in the first step of 2C-like transition (25).
After validating the purified cells, we carried out smallscale in situ Hi-C (sisHi-C) (8) to determine the chromatin conformation landscapes in 2CLCs and ESCs. About 200 million valid read pairs were obtained for both 2CLCs and ESCs, and the biological replicates were highly correlated (Supplementary Table S1). We first examined largescale chromosome folding by examining active (A) and inactive (B) compartmentalization of the genome. As shown by contact maps and principal component 1 (PC1) values (26), A/B compartments are largely maintained during ESC to 2CLC transition ( Figure 1A and B). Nevertheless, we observed significant decrease in the strength of compartmentalization in 2CLCs compared to ESCs ( Figure 1C). Contacts between B compartments decreased and interaction frequency between A/B compartments increased (Supplementary Figure S2A and B). In addition, we identified 7.9% of genomic intervals with A-to-B (1.0%) or B-to-A (6.9%) compartment switching during ESC to 2CLC transition (Supplementary Figure S2C). As expected, total transcription within A-to-B switching compartments was downregulated, while B-to-A switching compartments showed increased transcription during ESC to 2CLC transition (Supplementary Figure S2D). Although we found a few 2Cspecific genes such as the Tdpoz family genes in the B-to-A switching compartments (Supplementary Figure S2E), it is worth noting that only a small fraction (85/1044, 8.1%) of 2C-specific genes manifested B-to-A switching (Supplementary Figure S2F), indicating that compartment switching is not a general requirement for the activation of 2C-specific genes in 2CLCs. Interestingly, genomic intervals showing Ato-B switching are slightly enriched for previously identified ESC enhancers (27) (Supplementary Figure S2G), implying inactivation of ESC enhancers in 2CLCs.
We next asked whether the local insulation of TADs was reprogrammed during ESC to 2CLC transition. Examination of insulation scores (28) at the previously reported ESC TAD boundaries (23) revealed significant attenuation of TAD insulation in 2CLCs as compared to ESCs ( Figure  1D and E; Supplementary Figure S2H), which is further supported by genome-wide decrease of absolute directionality index (23) in 2CLCs ( Figure 1F and G). In line with these observations, global contact decay curves showed reduced frequencies of short-to intermediate-range interactions (<500 kb) but increased frequencies of long-range interactions (>1 Mb) in 2CLCs as compared to ESCs (Figure 1H). Indeed, 2CLCs exhibited weaker TAD strength and higher frequencies of inter-TAD interactions ( Figure  1I-K). Despite the global reduction of TAD insulation, the numbers of TADs called in 2CLCs (n = 3674) and ESCs (n = 3625) were comparable, of which the vast majority (n = 2957) were stable (Supplementary Figure S2I), supporting the overall conserved TADs among different cell types (23,29). Taken together, our observations suggested a genome-wide weakening of TAD insulation in 2CLCs instead of specific gain or loss of TADs.

Transcription of 2C-specific genes and repeats did not establish local chromatin insulation in 2CLCs
Active transcription was reported to be frequently correlated with chromatin insulation (30,31). Given that 2CLCs highly express a set of 2C-specific genes and the MERVL repeats, we asked whether active transcription at these loci could establish novel insulation in 2CLCs. However, we did not observe any increase in insulation scores at the promoters of up-regulated genes in 2CLCs, indicating no significant changes in chromatin insulation around 2C genes (Supplementary Figure S2J). Similarly, the highly expressed MERVL repeats also did not cause increased local chromatin insulation in 2CLCs (Supplementary Figure S2K). Thus, our results support previous observations that transcription is not sufficient to cause chromatin insulation (30).

Loss of enhancer-promoter interactions and down-regulation of pluripotent genes during ESC to 2CLC transition
Reduced TAD insulation indicates loss of chromatin loops. Indeed, we observed in 2CLCs a mild reduction in the strength of previously identified chromatin loops, most of which are anchored at TAD boundaries (29) (Figure 2A and B). Given that chromatin loops not only link TAD boundaries but also tether enhancers and promoters, we specifically analyzed interactions between reported ESC enhancer-promoter pairs (32). This analysis revealed that enhancer-promoter loops were also weakened in 2CLCs ( Figure 2C and D). Furthermore, reanalysis of published Hi-C datasets showed reduced strength of TADs and both types of chromatin loops in 2C embryos as compared to ICM (Supplementary Figure S2L-N). These observations suggested that weakened chromatin loops, both between TAD boundaries and between enhancer-promoter pairs, is a unique feature of totipotency. Interestingly, the weak- ened enhancer-promoter interactions appeared to correlate with a reduction in the ESC enhancer activity. Indeed, we found that down-regulated genes in 2CLCs were located closer to ESC enhancers ( Figure 2E) and were enriched for ESC super-enhancer neighboring genes ( Figure  2F). In turn, super-enhancer neighboring genes also displayed significantly lower expression in 2CLCs ( Figure 2G).

Inactivation of ESC enhancers and formation of new enhancers in 2CLCs
Chromatin accessibility is a strong indicator of the regulatory activity of enhancers (34). To directly examine whether weakened enhancer-promoter interactions in 2CLCs are associated with a reduction in the ESC enhancer activity, we carried out ATAC-seq analysis (20) for both ESCs and 2CLCs. As expected, chromatin accessibility around MERVL repeats was strongly induced in 2CLCs ( ure 3A), consistent with the activation of MERVL during ESC to 2CLC transition. In contrast, meta-analysis at ESC enhancers and super-enhancers revealed a substantial decrease of chromatin accessibility in 2CLCs ( Figure  3B), indicating inactivation of ESC enhancers, particularly super-enhancers, during the pluripotent to totipotent state transition. In agreement with this observation, promoters of ESC super-enhancer neighboring genes and downregulated genes also displayed reduced chromatin accessibility ( Figure 3C). Interestingly, promoters of up-regulated genes did not show increased accessibility ( Figure 3C), suggesting that these genes may have already existed in a primed state in ESCs. To further confirm that ESC en-hancers were silenced during ESC to 2CLC transition, we performed ultra-low-input native ChIP-seq (ULI-NChIPseq) (17,18) and compared genome-wide histone modifications between ESCs and 2CLCs, including H3K27Ac, H3K4me1, H3K4me3, H3K27me3 and H3K9me3 ( are also inactive in 2C embryos and are established gradually during preimplantation development as revealed by our reanalysis of published ATAC-seq and ChIP-seq data of early embryos (Supplementary Figure S4A). Among the five histone modifications we profiled, H3K27Ac peaks displayed the most notable difference in genome-wide distribution between ESC and 2CLCs, with more peaks enriched in distal intergenic regions in 2CLCs (Supplementary Figure S3D). Examination of these 2CLCspecific distal H3K27Ac peaks (n = 11,133) not only revealed increases of chromatin accessibility, H3K27Ac, H3K4me1 and H3K4me3 during ESC to 2CLC transition ( Figure 3D), but also high levels of chromatin accessibility and H3K27Ac in 2C embryos (Supplementary Figure  S4A), suggesting the formation of new enhancers (i.e. putative 2C enhancers) at these regions during totipotency acquisition. Indeed, up-regulated genes in 2CLCs were located closer to these putative 2C enhancers in the genome ( Figure 3E). By comparing with published ChIP-seq data of DUX, an important transcription factor in 2CLCs (35)(36)(37), we found strong DUX signals at these putative 2C enhancers, suggesting that they may arise from DUX binding ( Figure 3F). Interestingly, the most up-regulated genes during the 2C-like transition tend to have ESC TAD boundaries separating them from putative 2C enhancers nearby ( Figure 3G), suggesting that weakened TAD insulation may potentially facilitate contacts between the putative 2C enhancers activated by sporadically expressed DUX in ESCs and the promoters of their neighboring 2C genes, as exemplified by Nelfa, an early driver of the 2C-like state (38) ( Figure 3H). However, the interaction between Nelfa and the DUX-occupied putative enhancer nearby could not be readily detected in our analysis, which may require Hi-C data of much higher resolution. Taken together, inactivation of ESC enhancers, formation of putative 2C enhancers, and weakening of TAD boundaries may function together to fully activate the totipotent-like transcriptional program in 2CLCs.

Depletion of CTCF or cohesin complex facilitates the ESC to 2CLC transition
Our results have demonstrated that 2CLCs possess a relaxed chromatin architecture, including weakened TAD boundary loops and enhancer-promoter loops. While reduced enhancer-promoter loops correlates with down-regulation of ESC enhancers and pluripotent genes, weakened TAD insulation may facilitate the activation of 2C genes by sporadically activated 2C enhancers (Supplementary Figure  S4B). We thus hypothesized that disruption of chromatin loops by depletion of CTCF or the cohesin complex would facilitate the ESC to 2CLC transition. To directly test this, we knocked down Ctcf and core cohesin complex genes including Smc1a, Smc3 and Rad21 individually with small interfering RNAs (siRNAs) in the MERVL::tdTomato ESCs (Supplementary Figure S5A). As a control, we also knocked down Yy1, a structural regulator that was reported to mediate only enhancer-promoter interactions but not TAD boundary loops (39). Supporting our hypothesis, knockdown of Ctcf, Smc1a, Smc3 and Rad21, but not Yy1, significantly increased the fraction of tdTomato+ cells (Fig-ure 4A, Supplementary Figure S5B), as well as the expression of MERVL and 2C-specific genes including Dux, Zscan4d, Zfp352 and Tdpoz4 ( Figure 4B), demonstrating an enhanced ESC to 2CLC transition upon disruption of all chromatin loops rather than enhancer-promoter loops alone. We further examined the effect of CTCF and cohesin removal on all 2C genes by reanalyzing published RNA-seq datasets of ESCs with acute protein degradation of CTCF or RAD21 (12,15). This analysis also showed that depletion of either CTCF or the cohesin complex could result in activation of 2C genes and repression of pluripotent genes ( Figure 4C, Supplementary Figure S5C).
We further constructed an ESC line (2C-CTCF-AID) containing both the MERVL::tdTomato reporter and an auxin-inducible degron (AID) system targeting CTCF. Endogenous CTCF in this cell line is also fused with an eGFP tag to indicate the protein level of CTCF (15). As shown by flow cytometry analysis, 2 days of auxin treatment led to a near complete depletion of the CTCF protein, which can be fully restored 2 days after auxin withdrawal ( Figure 4D). Reanalyzing published Hi-C data (15) confirmed weakened TAD insulations together with reduced chromatin loops upon 2 days of auxin treatment ( Figure 4E-G). Similar to Ctcf knockdown ( Figure 4A and B), the percentage of td-Tomato+ 2CLCs in the culture ( Figure 4D and H), as well as the expression of MERVL and 2C-specific genes ( Figure  4I), significantly increased upon CTCF degradation. It is noteworthy that CTCF-depletion-induced increase in ESC to 2CLC transition is reversible, as the percentage of td-Tomato+ 2CLCs returned to the initial state after auxin withdrawal and CTCF recovery ( Figure 4D and H). We next purified ESCs and 2CLCs in the cells treated with auxin for 2 days and performed RNA-seq analysis. These cells are both viable (Supplementary Figure S5D), comparable to untreated ESCs and 2CLCs (Supplementary Figure S1B). Our analysis showed similar transcriptomes of 2CLCs in the presence and absence of CTCF, both with up-regulation of 2C-specific genes and down-regulated pluripotent genes as compared to ESCs (Supplementary Figure S5E), demonstrating that 2CLCs converted in the absence of CTCF is indeed in a 2C-like state. Taken together, our results strongly suggested that the higher-order chromatin structure in ESCs is an impediment to the 2C-like state transition.

DISCUSSION
In addition to epigenetic modifications, 3D chromatin architecture has been increasingly recognized as another regulatory layer of gene expression, which is tightly linked to cell identity. ESC differentiation has been shown to coincide with formation or strengthening of TAD boundary loops, indicating a relatively relaxed chromatin architecture in ESCs compared to differentiated cells (30,40). Furthermore, mouse zygotes and 2C embryos also display more relaxed chromatin architectures compared to later-stage embryos, with much weaker TAD boundaries (7,8). Similar relaxed chromatin architectures of early embryos were also observed in other organisms including fly, fish and human (9)(10)(11). Therefore, high developmental potency appears to be associated with relaxed higher-order chromatin struc-  or Yy1. Data are presented as mean ± SD, n = 3. **P < 0.01, ***P < 0.001 (multiple t tests). (B) Relative expression levels of 2C-specific transcripts after knocking down Ctcf, Smc1a, Smc3, Rad21, or Yy1. Data are presented as mean ± SD, n = 3. *P < 0.05, **P < 0.01, ***P < 0.001 (multiple t tests). (C) Box plot showing log2 fold change of indicated groups of genes upon acute depletion of the CTCF protein in ESCs by auxin-inducible degron (AID). Analyses were performed using a published RNA-seq dataset (15). (D) Representative FACS analysis of 2C-CTCF-AID cells treated with or without auxin. Note that endogenous CTCF in this cell line is fused with an eGFP tag to indicate CTCF protein levels. (E) Average DI in a 0.6 Mb region centered at TAD boundaries. (F) Aggregate Hi-C contact maps between pairs of loop anchors. (G) Aggregate Hi-C contact maps between ESC enhancer-promoter pairs. (H) The percentages of 2CLCs at indicated time points after auxin treatment or withdrawal. Data are presented as mean ± SD, n = 3. *P < 0.05, ***P < 0.001 (multiple t tests). (I) Relative expression levels of 2C-specific transcripts upon auxin-induced CTCF degradation. Data are presented as mean ± SD, n = 3. *P < 0.05, **P < 0.01, ***P < 0.001 (multiple t tests). ture. In this study, using the pluripotent to totipotent-like transition in mouse ESCs as a model, we defined the role of genome architecture relaxation in establishing totipotency. This spontaneous transition represents a simplified reprogramming process in totipotency acquisition, which is valuable to investigate the functional contribution of one single factor among the complex developmental events. But the scarcity of spontaneous 2C-like transition in the ESC culture limits such investigations. By using low-input technologies, we presented genome-wide chromatin contact maps of totipotent-like 2CLCs and pluripotent ESCs, together with their transcriptome, chromatin accessibility, and histone modification profiles. This rich dataset has allowed us to investigate the dynamics of chromatin state and gene expression during the pluripotent to totipotent-like state transition.
We revealed that chromatin architecture of ESCs became more relaxed during the spontaneous 2C-like transition, with globally weakened chromatin loops. This included not only TAD boundary loops but also chromatin loops between ESC enhancers particularly super-enhancers and the promoters of their neighboring genes, correlating with ESC enhancer inactivation and transcriptional downregulation of many pluripotent genes controlled by ESC super-enhancers. The down-regulation of pluripotent genes in 2CLCs was previously reported to take place only at the protein level but not the mRNA level (6,14). Nevertheless, a recent single-cell analysis of DUX-induced ESC to 2CLC transition showed a two-step reprogramming process with transcriptional inactivation of pluripotent genes as the first step (25), in agreement with our findings. The downregulation of pluripotent genes at both mRNA and protein levels suggested a complex regulation of gene expression during ESC to 2CLC transition. Of note, our 20 kb resolution data did not allow us to perform robust de novo calling of loops, preventing detailed analysis of ESC-and 2CLCspecific loops. Also, it remains to be elucidated how chromatin relaxation occurs in spontaneously converted 2CLCs. As shown in Supplementary Figure S5F, none of the genes involved in genome organization that we examined were down-regulated in 2CLCs, at least at the transcriptional level. We further examined genome-wide CTCF binding for both ESCs and 2CLCs using CUT&Tag (19). This analysis showed that CTCF similarly bound to known ESC CTCF binding sites and insulator elements in both ESCs and 2CLCs, with 2CLCs showing a slightly decreased enrichment (Supplementary Figure S5G). While this observation may partially account for the weakened 3D genome conformation in 2CLCs, further studies are needed to understand the regulation and function of CTCF in the spontaneous ESC to 2CLC transition.
We further demonstrated a causal relationship between relaxed chromatin architecture and totipotency acquisition, by showing that disruption of 3D organization of chromatin in ESCs could facilitate 2C-like transition. We also observed formation of putative 2C enhancers during ESC to 2CLC transition. Importantly, up-regulated genes during the 2Clike transition tend to have ESC TAD boundaries separating them from putative 2C enhancers nearby. Therefore, disruption of TAD boundary loops in ESCs may facilitate the contacts between putative 2C enhancers that are likely spo-radically activated in ESCs and the promoters of nearby 2C genes, thus promoting their expression, which in turn fully activates the 2C-like transcriptional program. Consistently, a recent study showed that cohesin pre-depletion in donor cells could enhance SCNT efficiency at least partially through up-regulation of the minor ZGA genes in donor cells, many of which are also 2C-specific genes (12).
In sum, we demonstrated that the spontaneous pluripotent to totipotent-like state transition in mouse ESCs not only coincides with global weakening of 3D genome conformation, but also can be promoted by disruption of chromatin loops. Further investigations are still required to elucidate the molecular mechanisms underlying the global weakening of higher-order chromatin structure during totipotency acquisition, as well as to fully understand how 2C-specific genes are activated upon disruption of chromatin loops.

DATA AVAILABILITY
The RNA-seq, ATAC-seq, ChIP-seq, CUT&Tag, and Hi-C sequencing data reported in this paper have been deposited with the NCBI GEO under accession number GSE159623.