Epigenetic integrity of paternal imprints enhances the developmental potential of androgenetic haploid embryonic stem cells

The use of two inhibitors of Mek1/2 and Gsk3β (2i) promotes the generation of mouse diploid and haploid embryonic stem cells (ESCs) from the inner cell mass of biparental and uniparental blastocysts, respectively. However, a system enabling long-term maintenance of imprints in ESCs has proven challenging. Here, we report that the use of a two-step a2i (alternative two inhibitors of Src and Gsk3β, TSa2i) derivation/culture protocol results in the establishment of androgenetic haploid ESCs (AG-haESCs) with stable DNA methylation at paternal DMRs (differentially DNA methylated regions) up to passage 60 that can efficiently support generating mice upon oocyte injection. We also show coexistence of H3K9me3 marks and ZFP57 bindings with intact DMR methylations. Furthermore, we demonstrate that TSa2i-treated AG-haESCs are a heterogeneous cell population regarding paternal DMR methylation. Strikingly, AG-haESCs with late passages display increased paternal-DMR methylations and improved developmental potential compared to early-passage cells, in part through the enhanced proliferation of H19-DMR hypermethylated cells. Together, we establish AG-haESCs that can long-term maintain paternal imprints. Supplementary Information The online version contains supplementary material available at 10.1007/s13238-021-00890-3.


INTRODUCTION
Culture conditions with 2i (inhibitors of Mek1/2 and Gsk3β) supplemented with leukemia inhibitory factor (LIF) (2i/L) enhance the derivation of embryonic stem cells (ESCs) with naïve ground-state pluripotency (Ying et al., 2008). However, recent studies indicate that prolonged application of 2i results in a widespread loss of DNA methylation, including repetitive elements and imprinted genes, and impaired developmental potential in ESCs (Choi et al., 2017b;Yagi et al., 2017). An alternative 2i/L (Mek1/2 inhibitor is substituted with the Src inhibitor, known as a2i/L) has been developed (Shimizu et al., 2012) and used to preserve the epigenetic integrity and developmental potential of ESCs (Choi et al., 2017b;Yagi et al., 2017). Nevertheless, the longterm culture of female ESCs in a2i/L induced a reduction of ICR (imprinting control region) methylation (Yagi et al., 2017). Meanwhile, although male ESCs under a2i/L can stable maintain DMR methylation until passage 15 (Yagi et al., 2017), whether prolonged culture ensures intact DMR methylation is still unknown. Similarly, the conventional serum plus LIF (S/L) medium also induced loss of DNA methylation at ICRs in mouse ESCs upon prolonged culturing (Dean et al., 1998;Humpherys et al., 2001;Yagi et al., 2017). Therefore, there is an unmet need for the establishment or employment of appropriate derivation and culture conditions that would enable the long-term stable maintenance of imprinting marks in ESCs.
Application of 2i/L culture conditions has enabled the successes in deriving mouse haploid ESCs (haESCs) from parthenogenetic and androgenetic embryos (termed PG-haESCs and AG-haESCs, respectively) (Elling et al., 2011;Leeb and Wutz, 2011;Li et al., 2012;Yang et al., 2012). Both PG and AG-haESCs can be employed for high-throughput genetic analyses at a cellular level (Elling et al., 2011;Leeb et al., 2014;Leeb and Wutz, 2011;Sagi et al., 2016;Yang et al., 2013;Zhong et al., 2016b). Notably, AG-haESCs can be used as sperm replacement to support embryonic development upon injection into oocytes, leading to the production of semi-cloned (SC) embryos (Li et al., 2012;Yang et al., 2012;Zhang et al., 2020). However, 2i/L-treated PG-haESCs were globally hypomethylated (Choi et al., 2017a;Elling et al., 2011;Zhong et al., 2016a). Meanwhile, AG-haESCs lost DNA methylation at differentially DNA methylated regions (DMRs), i.e., H19-DMR and IG-DMR, that control two paternal imprinted loci, H19-Igf2 and Dlk1-Dio3, respectively, resulting in the low birth rate of SC mice (Yang et al., 2012). Interestingly, deleting both DMRs in AG-haESCs (termed DKO-AG-haESCs) could decrease the expression of H19 and Gtl2 and dramatically rescue the developmental potential of resultant SC embryos Zhong et al., 2015). We reasoned that, if AG-haESCs could be generated under a2i/L conditions, they might stably maintain paternal imprinting features at least at early passages, which allows for the better developmental potential of SC embryos without removal of paternal DMRs and thus provides a useful tool for an efficient combination of in vitro and in vivo studies of imprinting regulation.
In this study, we tested our hypothesis using a2i/L with serum, w/o serum or a two-step protocol (Yagi et al., 2017) (a2i/L with serum for derivation followed by a2i/L without serum for maintenance, TSa2i/L) and found that TSa2i/L enabled generating AG-haESCs with stable DNA methylation in paternal DMRs up to 60 passages. Furthermore, we showed that H3K9me3 deposition and ZFP57 binding are preserved at H19 and IG DMRs and highly correlated with DMR methylation, which are likely involved in the stable maintenance of paternal imprints in TSa2i/L-derived AG-haESCs. In contrast, H3K4me3 deposition may prevent DMRs from de novo methylation upon loss of imprints. Strikingly, we found that TSa2i/L-treated AG-haESCs with late passages exhibited higher paternal DMR methylations and increased developmental potential compared to cells of early passages. We further demonstrated that TSa2i/Ltreated AG-haESCs are a heterogeneous cell population regarding H19-DMR methylation levels and cells with hypermethylated H19-DMR display better proliferation potential compared to cells with hypomethylated H19-DMR, partially accounting for long-term maintenance of paternal imprints.

RESULTS
A two-step a2i/L derivation/culture condition enables generation of AG-haESCs with hypermethylated paternal DMRs To derive haploid ESCs, we reconstructed haploid embryos through sperm nuclear transfer, i.e., injection of sperm head into enucleated oocytes (Yang et al., 2012). The spermcloned embryos developed to the blastocyst stage in vitro, followed by ESC derivation in one of three culture conditions, S/L, S/L supplemented with 2i (2i/L) or S/L with a2i (a2i/L). Both 2i/L and a2i/L conditions promoted the derivation of ESCs from sperm-cloned blastocysts compared to S/L ( Fig. S1A and S1B). Among 12 2i/L-derived and 8 a2i/Lderived ESC lines, 7 and 4 of them respectively contained a subpopulation of haploid cells that can be enriched and maintained through regular fluorescence-activated cell sorting (FACS) . Consistent with our previous observations (Yang et al., 2012), 2i/L-derived AG-haESCs could give rise to SC mice upon intracytoplasmic AG-haESC injection (ICAHCI) at low efficiency at early passages, but lost the ability upon prolonged culturing, due to decreased DNA methylation at H19-DMR and IG-DMR Zhong et al., 2015) (Fig. S1E and S1F; Table S1). Expectedly, a2i/L-cultured haploid cells of early passages sustained DNA methylation at H19-DMR and IG-DMR and exhibited better developmental potential demonstrated by ICAHCI compared to 2i/L cells (Fig. S1G and S1H; Table S1). However, DNA methylation of paternal DMRs was progressively reduced in a2i/L-cultured haploid cells of later passages, leading to the decreased birth rate of SC mice ( Fig. S1H; Table S1), most likely due to the use of serum that causes loss of DNA methylation at DMRs in diploid ESCs upon prolonged culturing (Dean et al., 1998;Humpherys et al., 2001;Yagi et al., 2017).
We next adopted a well-defined serum-free ESC medium (DMEM/F12 + Nerobasal + N2 + B27 + LIF) supplemented with a2i (termed SF/a2i/L) for AG-haESC derivation. From a total of 25 sperm-cloned blastocysts, we obtained 16 outgrowths. However, none of them could be passaged, indicating that serum or its replacement is critical for ESC derivation (Martello and Smith, 2014). We next employed a two-step protocol (Yagi et al., 2017), in which, a2i/L medium was employed for formation of outgrowths and a few times of cell passaging before the use of SF/a2i/L conditions for longterm maintenance (termed TSa2i/L protocol) (Fig. 1A). We found that one time of expansion in a2i/L followed by transferring into SF/a2i/L could give rise to stable haploid cell lines from different genetic backgrounds through regular FACS-enrichment of haploid cells (Figs. 1B,1C,S2A,and S2B). Surprisingly, DNA methylations at both H19 and IG-DMRs were stably maintained at high levels upon prolonged culturing (up to passage 59, p59) in three tested lines (TSa2i-14, TSa2i-C57, and TSa2i-F1), although Rasgrf1-DMR was not stable in these cell lines (Figs. 1D and S2C). Consistently, ICAHCI analysis showed that all three tested TSa2i/L-treated haploid cells at early passages (<p30) efficiently supported the full-term embryonic development of SC embryos ( Fig. 1E; Table 1), reaching an efficiency similar to DKO-AG-haESCs cultured in 2i/L (Zhong et al., 2015). Meanwhile, while cesarean section (C-section) was required

RESEARCH ARTICLE
Hongling Zhang et al. to obtain SC pups when 2i/L or a2i/L-cultured cells were used for ICAHCI, the recipient females delivered TSa2i/Ltreated-cell-derived SC pups by themselves similar to those carrying SC embryos derived from DKO-AG-haESCs ( Fig. S2D) (Zhong et al., 2015).
Prolonged SF/a2i/L culture enhances the developmental potential of AG-haESCs We next examined whether genetic alterations can be introduced into these TSa2i/L-treated AG-haESCs, followed by ICAHCI to efficiently produce SC mice carrying expected genetic traits. To this end, we generated AG-haESCs carrying different genetic modifications, including a single mutant gene (Tet3), triple mutant genes (Tet1, 2, and 3), or a tagged gene (Dusp9-HA) (Choi et al., 2017a) by performing CRISPR-Cas9-based genetic manipulations ( Fig. S2E and S2F). Surprisingly, all gene-modified TSa2i/ L-treated haploid cells could give rise to corresponding SC mice via ICAHCI at a higher efficiency compared to haploid cells before manipulation (Table 1). Interestingly, wildtype TSa2i/L-cultured haploid cells, upon prolonged in vitro culture (p30-p62), also exhibited better ability to produce SC mice via ICAHCI ( Fig. 2A), thus excluding the possibility that genetic manipulations result in higher developmental potential in haploid cells and suggesting that prolonged period of culturing in SF/a2i/L per se may enhance the development of SC embryos. Postnatal growth profiling analysis showed that SC pups obtained from TSa2i/L-treated haploid cells displayed better growth compared to DKO-AG-haESCs (Fig. S2G), probably due to decreased H19 and increased Igf2 expression in

RESEARCH ARTICLE
Hongling Zhang et al. organs like liver during development ( Fig. S2H-K), consistent with the recent observations in 2i/L-cultured AG-haESCs with H19, H19-DMR and IG-DMR deletions .
Epigenetic integrity of paternal DMRs is preserved in TSa2i/L-cultured AG-haESCs Given the strong correlation of DNA methylation patterns and in vivo developmental potentials of ESCs (Choi et al., 2017b;Yagi et al., 2017;Zwaka, 2017) and previous studies have highlighted that H19 and IG DMR methylation state in AG-haESCs are related to their developmental potential through ICAHCI Zhong et al., 2015), we thus investigated DNA methylation state in cells with prolonged culture. Bisulfite-PCR analysis indicated increased DNA methylation at both of H19 and IG DMRs in TSa2i/L-treated haploid cells of late passages compared to early passages (Fig. 2B). To further assess their methylation patterns, we performed whole-genome bisulfite sequencing (WGBS) using haploid cells of early (<p30) and late passages (p30∼p60). The results indicated that TSa2i/ L-treated haploid cells sustained high levels of DNA methylation at DMRs of paternally imprinted genes at both early and late passages, even though the genome-wide DNA methylation level showed a slight decrease upon prolonged culture (Figs. 2C and S3A). In contrast, majority of tested canonical maternal DMRs (n = 20) were unmethylated in TSa2i/L-treated AG-haESCs during longterm culturing (Figs. 2C, 2D and S3B). We further analyzed secondary (somatic) DMRs in paternally imprinted loci which are known to depend on the presence of a germline DMR and established after fertilization (Barlow and Bartolomei, 2014;Ferguson-Smith, 2011) and found that two tested regions sustained hypermethylation in TSa2i/Ltreated cells (Fig. S3C). DNA methylation and H3K9 methylation are strongly associated (Du et al., 2015). In mouse ESCs cultured in 2i, while exhibiting globally hypomethylated DNA compared to ESCs maintained in conventional conditions involving serum, elevated DNA methylation correlates with the presence of H3K9me3 on imprinted loci (Habibi et al., 2013). Moreover, both parental genomes undergo large-scale H3K9me3 reestablishment after fertilization (Wang et al., 2018). We thus hypothesized that maintenance of hypermethylated germline DMRs in our sperm genome-bearing haploid cells may also correlate with the presence of H3K9me3, which is probably established during pre-implantation development of sperm-cloned embryos. To this end, we first analyzed the allelic H3K9me3 pattern on DMRs during mouse early embryo development based on singlenucleotide polymorphism (SNP) (Wang et al., 2018). As expected, both parental DMRs showed allelic-specific H3K9me3 patterns, which were established after fertilization and stably sustained from the zygote stage to the ICM stage ( Fig. S4A and S4B). We then investigated the distributions of H3K9me3 in our TSa2i/L and 2i/L haploid cells that were derived from ICMs by performing ChIP-seq analysis. Consistent with a published report in mouse diploid ESCs (Walter et al., 2016), H3K9me3 was also mainly enriched at repetitive elements, especially LTRs and LINEs in haploid ESCs ( Fig. S4C and S4D). Intriguingly, both H19 and IG DMRs were stably enriched with strong H3K9me3 signals in TSa2i/L-derived AG-haESCs during long-term culturing (Fig. 3A). In contrast, H3K9me3 of paternal DMRs was gradually lost in 2i/L-derived AG-haESCs upon prolonged culture. Furthermore, H3K9me3 specifically marked paternal DMRs rather than maternal DMRs in TSa2i/L AG-haESCs, while all DMRs were free of H3K9me3 in 2i/L AG-haESCs (Figs. 3B and S4E). We also noted the loss of H3K9me3 at Rasgrf1-DMR in TSa2i-C57 cells with late passages (Fig. S4F), in which, Rasgrf1-DMR methylation was absent ( Fig. S2C). Our results thus demonstrate that H3K9me3 of paternal DMRs is maintained in TSa2i/L AG-haESCs, which positively correlates with the DNA methylation state of DMRs.
ZFP57 is one of the Kruppel-associated box zinc finger proteins (KRAB-ZFPs) that can target and maintain genomic imprints after fertilization (Takahashi et al., 2015). Our sperm genome-carrying haploid cells were derived from androgenetic blastocysts, in which, paternal methylation imprints should be protected as those in normal diploid blastocysts (Li et al., 2008). Meanwhile, ZFP57 protein is highly expressed in ESCs, binds at all known imprinted DMRs in a parental origin-specific and methylation-sensitive manner and Zfp57null ESCs lost imprints at all tested imprinted DMRs (Quenneville et al., 2011;Strogantsev et al., 2015). Together, these observations suggest that ZFP57 may be critical for maintenance of the paternal DMR methylation in TSa2i/L cells. As expected, ChIP-seq analysis indicated that methylated DMRs (in TSa2i/L cells) rather than unmethylated DMRs (in 2i/L cells) were bound by ZFP57 (Figs. 3C and S4G). Similarly, maternal DMRs were free of ZFP57 bindings in both 2i/L and TSa2i/L AG-mESCs (Fig. 3D). We then deleted Zfp57 in TSa2i/L AG-haESCs and cultured them in SF/a2i/L for up to 20 passages (Fig. S4H). Expectedly, both H19 and IG DMRs gradually lost methylation (Fig. 3E). Furthermore, WGBS analysis indicated that Zfp57null cells were globally hypermethylated and sustained a reduced level of DNA methylations at paternal DMRs specifically (Fig. S4I). Consistently, H3K9me3 ChIP-seq analysis of Zfp57-null cells showed overt reduced signals on paternal DMRs (Fig. 3F). Taken together, TSa2i/L AG-haESCs preserve the epigenetic integrity of paternal Native imprints enhance the development of AG haESCs RESEARCH ARTICLE imprints, including hypermethylated DNA methylation, deposition of H3K9me3 marks, and ZFP57 bindings.
Deposition of H3K4me3 prevents de novo methylation of DMRs in 2i/L-SF/a2i/L-switched AG-haESCs Next, we sought to investigate whether imprints can be restored in 2i/L-cultured cells after being switched to SF/a2i/ L conditions. Interestingly, switching 2i/L-treated haploid cells to SF/a2i/L for 5-10 passages induced global hypermethylation shown by 5mC immunoblotting (Fig. S5A), which was probably caused by the increased protein levels of DNA methyltransferases including DNMT1 and DNMT3A/ 3B (Fig. S5B). WGBS analysis further confirmed that while the whole genome underwent extensive de novo methylation in 2i/L-SF/a2i/L-switched AG-haESCs, both maternal and

RESEARCH ARTICLE
Hongling Zhang et al. paternal DMRs maintained hypomethylated state (Fig. 4A). Bisulfite-PCR analysis further confirmed that paternal DMRs (H19 and IG DMRs) and maternal DMRs (Snrpn and Peg10 DMRs) did not undergo significant de novo methylation (Fig. 4B). Interestingly, when H19 and IG DMR deletions existed in 2i/L-cultured haploid cells, the somatic DMRs (Gtl2-DMR and Igf2-DMR) were significantly methylated after being switched to SF/a2i/L culture conditions (Fig. 4C). Given that Gtl2-DMR and Igf2-DMR showed higher DNA methylation levels in TSa2i/L AG-haESCs when compared to 2i/L AG-haESCs (Fig. S3C), our results together further confirmed that establishment of somatic DMRs relies on germline-derived DMRs and deletion of germline-derived DMRs can partially mimic DMR methylation state (Edwards and Ferguson-Smith, 2007;Tucker et al., 1996). Although the global H3K9me3 signals were rescued with the global increase of DNA methylation (Figs. 4D and S5C), the absence of H3K9me3 signals on paternal DMR loci of 2i/Lcultured cells could not be restored after being switched to SF/a2i/L conditions, which was consistent with DNA methylation changes (Fig. 4E). Of note, ICAHCI analysis showed that switched cells even with global Native imprints enhance the development of AG haESCs RESEARCH ARTICLE hypermethylation didn't restore the developmental potential when germline DMRs were not established (Table S2). Meanwhile, when H19 and IG DMR deletions existed in haploid cells, the switched cells displayed relatively high developmental potential similar to original 2i/L-treated DKO cells (Table S2). Collectively, these results indicate that imprints once lost can't be restored in cultured ESCs under SF/a2i/L conditions, consistent with previous observations in other culture conditions (Tucker et al., 1996;Yagi et al., 2017).
We next sought to reveal factors that are involved in preventing the de novo methylation at DMRs. Previous studies have shown that H3K4me3 is negatively correlated with DNA methylation that is associated with transcriptional

RESEARCH ARTICLE
Hongling Zhang et al. activation and removing DNA methylation may lead to the spread of H3K27me3 to DNA that is otherwise protected from methylation (Atlasi and Stunnenberg, 2017), we thus analyzed the distributions of H3K4me3 and H3K27me3 in our cells. The results showed that nearly all unmethylated maternal DMRs were marked by H3K4me3 but not by H3K27me3 in 2i/L and TSa2i/L AG-haESCs ( Fig. 5A and  5B). Meanwhile, paternal DMRs with hypomethylation also showed higher H3K4me3 enrichment in 2i/L-cultured AG-haESCs compared to TSa2i/L-cultured AG-haESCs with hypermethylated DMRs (Fig. 5C). Moreover, paternal DMRs showed enhanced H3K4me3 signals when AG-haESCs were switched from 2i/L to SF/a2i/L conditions ( Fig. 5B and  5C). Consistently, high-throughput analyses showed that the loci with H3K4me3 peak lost during switching (Group I) were highly methylated in 2i/L-SF/a2i/L-switched cells, while the loci with maintained (Group III) or newly generated (Group II) H3K4me3 peaks were free of DNA methylation (Fig. 5D). Taken together, these results indicate that de novo methylation rigidly excludes the regions marked with H3K4me3. Due to the direct antagonism of H3K4me3 to DNMTs (Guo et al., 2015;Ooi et al., 2007), we proposed that H3K4me3 deposition at DMRs upon loss of both DNA methylation and H3K9me3 may prevent DMRs from de novo DNA methylation.
Enhanced proliferation potential of AG-haESCs with hypermethylated H19-DMR accounts for increased DMR methylation in cells of late passages Previous studies have shown that 2i/L-cultured AG-haESCs are a heterogeneous cell population regarding paternal DMR methylations, in which cells with different DNA methylation levels at H19 and IG DMRs upon injection into oocytes result in different developmental potential of resulting SC embryos Yang et al., 2012;Zhong et al., 2015). Interestingly, two knockin haploid cell lines carrying Dusp9-HA established through single-cell expansion from the same experiment resulted in distinct differences in the birth rate of SC pups upon ICAHCI (32.5% vs. 8.3%, Table 1). As expected, higher levels of DNA methylation at H19 and IG DMRs were indeed relative to a better birth rate (Fig. 6A). These observations implied that TSa2i/L-cultured haploid cells even with a high level of DMR methylations are still a heterogeneous population containing cells with different DNA methylation levels at H19 and IG DMRs, leading to gene-modified cell lines with distinct DNA methylation and corresponding developmental potential. To confirm this, we first performed C-section at 18.5 days of gestation (E18.5) to recover a total of 111 SC pups originated from 111 single haploid cells of 3 cell lines, and found 3 growth-retarded pups (otherwise could not be obtained through natural delivery), which lost DNA methylation at H19-DMR as expected (Figs. 6B, S6A, and S6B). These results confirmed the heterogeneous methylome in TSa2i/L-cultured cells, excluding the roles of genome-editing manipulations in DMR methylation loss.
To further characterize the heterogeneity, we performed clonal expansion of single cells and bisulfite sequence analysis of H19 and IG DMRs. From a total of 21 single-cell clones (1st clones), while three (sc11, sc13, and sc16) retained H19-DMR methylation at a reduced level, the rest clones sustained relatively higher methylation levels of H19 and IG DMRs (Figs. 6C and S6C). These results suggested that most single cells carried hypermethylated H19-DMR and a small proportion of cells lost imprints during cell proliferation, consistent with a recent report that single-cell clonal expansion of diploid ESCs in S/L produces heterogeneous methylomes . In order to further determine the DNA methylation state of H19-DMR at single cell level of 1st clones, we performed secondary single-cell expansion using cells from 1st clones with hypermethylated (sc6), moderately-methylated (sc13), and hypomethylated (sc16) H19-DMR respectively. The results showed that the 2nd single-cell clones (2nd clones) from cells of 1st clones with hypermethylation or hypomethylation exhibited similar DNA methylation patterns at both DMRs (Fig. 6D), excluding the possibility of rigid DMR methylation changes in TSa2i/Ltreated cells.
Interestingly, 2nd clones from cells of clone 13 were separated into two groups: one with hypermethylated H19-DMR and the other with hypomethylated H19-DMR (Fig. 6D). These results imply that a small proportion of cells displays more dynamically and less reliably at paternal DMRs, probably due to that DNMT1 has imprecise activity and H3K9me3-marked-region-related de novo activity during ESC division . Consistent with this postulation, Dnmt1 mutation but not Dnmt3a/3b double knockout induced methylation loss at paternal germline DMRs in TSa2i/L AG-haESCs (Fig. S6D). We next investigated the changes of DNA methylation at paternal DMRs during prolonged culturing. To this, we expanded 1st clones with different methylation levels for 10 passages or more and analyzed DMR methylations. The results showed that clones with hypermethylation or hypomethylation overall maintained similar patterns (Fig. S6E). Consistently, in TSa2i/L-cultured cells that lost H19-DMR methylation, H3K9me3 signals were depleted, and as expected H3K4me3 marks were established (Fig. S6F). Interestingly, the sc13 with medium methylation of H19-DMR displayed increased methylation upon prolonged culture (Figs. 6E and S6E), consistent with our previous observations that late-passage cells sustained higher H19-DMR methylation level compared to early passages (Fig. 2B).

RESEARCH ARTICLE
Hongling Zhang et al. mRNA levels while cells with reduced H19-DMR methylation displayed increased H19 expression (Fig. 6G). However, the perturbation of Igf2 didn't affect cell proliferation in clones with intact H19-DMR methylation ( Fig. S6H and S6I). In contrast, deletion of H19 induced increased cell proliferation in the clones with reduced H19-DMR methylation (Figs. 6H and S6J). We also characterized the self-diploidization of H19 and IG DMR hyper-and hypo-methylated haploid cells and found that they displayed a similar diploidization rate (Fig. S6K). These results indicate that H19 overexpression impairs the cell proliferation, implying that a better growth rate of haploid cells with H19-DMR hypermethylation may account for increased methylation levels in haploid cells of late passages under SF/a2i/L conditions.
Considering that a small fraction of TSa2i/L-treated cells carried hypomethylated H19-DMR that are related to the growth-retarded SC pups, we thus investigated whether removal of H19-DMR or both H19 and IG DMRs in TSa2i/Ltreated haploid cells may further improve their birth rate of SC pups through ICAHCI. However, DMR deletions did not improve the developmental potential of SC embryos (Table 1). In contrast, these deletions led to a slight reduction in the birth rate (Figs. 6I, S6L, and S6M), implying that original germline DMRs may sustain better developmental potential than DMR deletions. Moreover, these results further suggest that epigenetic integrity is critical for in vivo development. Taken together, our study suggest that H19-and IG-DMR methylation and H3K9me3 marks sustained in ICM cells are generally maintained during haploid ESC derivation and long-term expansion under TSa2i/L conditions; and epigenetic balance may be achieved during cell proliferation through joint roles of enhanced division rate by cells with intact H19-DMR methylation and a small proportion of cell gradually losing paternal DMR methylation due to imprecise DNMT1 activity-induced epimutation .

DISCUSSION
The breakthrough of the derivation of mouse ESCs from embryos was made in 1981 under the conditions containing feeder layers and serum (Evans and Kaufman, 1981). Following studies identified that the cytokine LIF produced by feeder cells is the principal factor to support ESC self-renewal, which can replace feeders in both derivation and long-term culture of germline-competent ESCs (Martello and Smith, 2014). Nonetheless, S/L conditions yield overt cellular heterogeneity, consisting of a substantial number of differentiating cells. Meanwhile, there are nonpermissive strains of mice for ESC derivation using S/L. Moreover, while earlypassage ESCs under S/L can develop to all ESC-derived fetuses upon injection into tetraploid blastocysts, upon prolonged culturing the developmental potential of many ESC lines becomes impaired due to epigenetic alterations in imprinted genes (Dean et al., 1998;Nagy et al., 1990Nagy et al., , 1993Wang et al., 1997). The advent of 2i, however, not only enables efficient and reliable derivation of ESC lines from all mouse strains (Kiyonari et al., 2010;Nichols et al., 2009) but also establishes ground state pluripotency with robust selfrenewal of a biologically homogeneous population of cells (Ying et al., 2008). Nevertheless, two recent studies have shown that prolonged culture of ESCs in 2i conditions induces a widespread loss of DNA methylation, leading to the impaired developmental potential of ESCs (Choi et al., 2017b;Yagi et al., 2017). In both reports, a2i/L (Shimizu et al., 2012) was employed to preserve the epigenetic integrity as well as the developmental potential of ESCs. However, one study showed that prolonged culture of female ESCs in a2i/L resulted in a reduction of germline DMR methylation although with increased global methylation (Yagi et al., 2017). Meanwhile, stable maintenance of DMR methylation in XY ESCs has been reported only to passage 15, leaving an intriguing question whether germline DMR methylation can be stably maintained in mouse ESCs during long-term culturing. Although recent studies have been examining the maintenance of genomic imprints in diploid ESC under different media (Lee et al., 2018;Wu et al., 2020Wu et al., , 2021, an ESC line with stable imprints during long-term culture is still not established. In our study, we generated multiple haESC lines from androgenetic blastocysts of different strains with stably hypermethylated H19 and IG DMRs up to passage 60 using a reported two-step derivation/culture protocol with a2i/L (Yagi et al., 2017). These cells result in efficient generation of SC mice through ICAHCI. Strikingly, in contrast to previous observations in mouse ESCs, our TSa2i/L-treated AG-haESCs have increased paternal DMR methylation and improved developmental potential upon prolonged culture.
TSa2i/L conditions ensure the dynamic equilibrium of typical paternal epigenetic makeups and improved spermlike features without compromising genetic integrity in AG-haESCs. Multiple factors may be involved in this epigenetic balance, including: (1) the coexistence of DNA methylation and H3K9me3 marks and ZFP57 bindings ensures stable maintenance of paternal imprints; (2) the imprecise DNA methylation of DNMT1 may cause methylation loss at paternal DMRs in a small number of cells; (3) H3K4me3 deposition after loss of DNA methylation and H3K9me3 marks precludes de novo methylation at germline DMRs, and (4) cells with hypermethylated H19-DMR have enhanced proliferation potential compared to cells with hypomethylated H19-DMR (Fig. 7). In addition, haploid cells cultured in a2i/L with serum lost paternal imprints in late passages, which indicates that serum may result in loss of DMR methylation during prolonged culturing. Considering that our haploid cells were always maintained on feeder cells, we also wanted to know if feeder cells are an important factor for maintenance of imprints in our SF/a2i/L conditions. Although short-term culture without feeders didn't change the DMR methylations (Fig. S7), feeder cells may probably release important but unknown factors that contribute to paternal imprint maintenance during long-term culturing. Moreover, DNMT1 has H3K9me3-marked-region-related de Native imprints enhance the development of AG haESCs RESEARCH ARTICLE novo activity during ESC division , which may also play a role for the increased H19-DMR methylation in late-passage haploid cells. Future studies will be needed to understand the detailed mechanism underlying the longterm stable maintenance of paternal imprints through simplifying the culture conditions, such as using a feeder-free system.
Thus, TSa2i/L-treated AG-haESCs serve as a unique ESC system for imprinting control analysis in vitro. Importantly, in combination with ICAHCI technology, these analyses can be promoted to the organismal level in one step (Li and Li, 2019). Unlike mature sperm or round spermatids, a small proportion of TSa2i/L-treated AG-haESCs lost H19-DMR methylation during cell division. While DNMT1-mediated imprecise DNA methylation may partially account for this loss , the detailed mechanisms require further elucidation. Moreover, Rasgrf1-DMR methylation is stably maintained in two cell lines but dramatically lost in the other one at late passages. Further investigation into the underlying difference may help us to understand how Rasgrf1-DMR methylation is controlled. These comparative studies may also yield clues to further define conditions for stable maintenance of Rasgrf1 imprinting in vitro, which may facilitate the potential applications of these cells. One intriguing application is the derivation of PG-haESCs carrying oocyte genome that can long-term maintain methylation of the majority of maternal DMRs.

Animals
All mice were housed in individually ventilated cages (IVC) under specific pathogen-free conditions and a 12:12 h light/dark cycle. MII oocytes were all from B6D2F1 (C57BL/6♀ × DBA/2♂) female mice. All the pseudopregnant foster mothers were ICR females. All animal procedures were performed under the ethical guidelines of the Shanghai Institute of Biochemistry and Cell Biology, Chinese Academy of Sciences, Shanghai, China.

CRISPR-Cas9-mediated gene manipulation in haploid cells
To generate CRISPR-Cas9 plasmids for gene mutation, sgRNAs (Table S3) of target genes were synthesized, annealed, and ligated to the pX330-mCherry plasmids (Wu et al., 2013) that were digested with BpiI (Thermo Scientific). AG-haESCs were transfected with the corresponding pX330-mCherry plasmids using Lipofectamine 3000 (Life Technologies) following the manufacturer's instruction manual. 24 h after transfection, the haploid cells expressing red fluorescence protein were enriched by flow cytometry (FACS AriaII, BD Biosciences) and plated at low density. Six days after plating, single colonies were picked for the derivation of AG-haESCs.

Intracytoplasmic AG-haESC or sperm injection and embryo transfer
To generate SC embryos, AG-haESCs were treated with 0.05 μg/mL Demecolcine solution (Sigma, USA) for 10-12 h and synchronized to the M phase. Next, the M-phase clones were treated with trypsin and suspended in the HCZB medium. MII oocytes were collected from oviducts of superovulated B6D2F1 females (8 weeks old). AG-haESCs were injected into the cytoplasm of MII oocytes in a droplet of HCZB medium containing 5 μg/mL cytochalasin B (Sigma, USA) using a Piezo-drill micromanipulator. The injected oocytes were cultured in the CZB medium for 30 min and then activated for 5-6 h in Ca 2+ free CZB with SrCl 2 . For sperm injection, the procedures were the same as described above. Following activation, the reconstructed embryos were cultured in AA-KSOM (Merk, Germany) medium at 37°C under 5% CO 2 in the air. The embryos obtained through Intracytoplasmic AG-haESC Injection (ICAHCI) were cultured in KSOM medium for 24 h to reach the two-cell stage. Thereafter, 18-20 two-cell embryos were transferred into each oviduct of pseudopregnant ICR female mice at 0.5 dpc. Recipient mothers carrying SC embryos derived from 2i/L or a2i/L medium were euthanized at 18.5 or 19.5 days of gestation, and the pups were quickly removed from the uteri.

RNA extraction and Real-time quantitative PCR
Total RNA was isolated from the cells using TRIzol reagent (Invitrogen). The cDNA was obtained from about 1 μg RNA with a reverse transcription reaction by the ReverTra Ace qPCR RT Master Mix (TOYOBO). Real-time quantitative PCR reactions (RT-qPCR) were performed on a Bio-Rad CFX96 using the SYBR Green Mix (TOYOBO) in triplicate. All the gene expression levels were normalized to the internal standard gene Gapdh or β-actin. The primer sequences are listed in Table S3.

Bisulfite PCR
The genome extraction from ESCs used the TIANamp Genomic DNA Kit (TIANGEN, China). RNA was removed using RNaseA (Thermo Scientific) in a 37°C water bath for 1 h. DNA was recovered using a Universal DNA Purification Kit (TIANGEN, China). The bisulfite conversion was performed using the EZ DNA Methylation-Gold TM Kit (ZYMO research, USA) for 200-500 ng genomic DNA, following the manufacturer's instructions. The bisulfite DNA products were amplified by nested PCR. PCR primers are listed in Table S3. The amplified products were purified by gel electrophoresis using a Universal DNA Purification Kit and cloned into pMD™19-T Vector Cloning Kit (Takara, 6013). For each sample, more than 12 E. coli clones were picked for sequencing. The results were analyzed by the DNA methylation analysis platform. (http://services.ibc. unistuttgart.de/BDPC/BISMA/). To detect DNA methylation in multiple regions with high efficiency, we mixed all bisulfite PCR products from one sample and performed the high-throughput sequencing library preparation described later. The sequencing reads were mapped to mm9 using BSMAP with parameters -r 0 -w 100 -v 0.1 -A AGATCGGAAGAGC. Methylation level was calculated using methylated read number versus total read number for one CpG site. For each bisulfite PCR region, the average DNA methylation level of all CpG sites in it was calculated.

Whole-genome bisulfite sequence (WGBS)
A total amount of 5-10μg genomic DNA was mixed with 25ng lambda DNA and sonicated to 200-500bp, followed by end repair and dATP adding with the homemade Kit. Next, methylated adapters (synthesized by ThermoFisher Scientific) were ligated to the sonicated DNA. Ampure beads (Vazyme) were used to remove <200 bp fragments. Bisulfite treatment was performed using the EZ DNA Methylation-Gold TM Kit (ZYMO research, USA). After bisulfite conversion, the single-stranded, uracil-containing DNA was subjected to 10-12 cycles of PCR reaction with Illumina TruSeq PCR primers and 2.5U of Pfu Turbo C x Hotstart DNA polymerase (Agilent Technologies) to recover enough DNA for sequencing. The sequencing reads were aligned to mm9 using BSMAP with parameters -r 0 -w 100 -v 0.1 -A AGATCGGAAGAGC. Multiple mapped reads and PCR duplicates were removed. After mapping, those reads with total CG coverage less than 5 within 200 bp were removed. The methylation level was calculated using methylated CpG versus total CpG in each bin. The file with DNA methylation Native imprints enhance the development of AG haESCs RESEARCH ARTICLE levels for each 200 bp bin was used for further analysis. For the boxplots about DNA methylation, we calculated the average DNA methylation levels of each region for each kinds of gene elements. The genomic information of maternal DMRs (n = 20) and paternal DMRs (n = 3) used in the WGBS analysis were shown in Table S4.

ChIP-seq
Cells were collected and cross-linked with 1% formaldehyde. Cells were then suspended in cell lysis buffer (0.3% SDS, 50 mmol/LTris-HCl pH 8.0, 20 mmol/L EDTA, and freshly added protease inhibitors) and incubated for 10 min on ice followed by sonication. About 30-60 μg of fragmented chromatin was diluted in dilution buffer (16.7 mmol/L of Tris-HCl, pH 8.0, 1.1% Triton X-100, 1.2 mmol/L of EDTA, 167 mmol/L of NaCl, freshly added protease inhibitor) and then incubated with 5 μg antibody at 4°C for 6-8 h. Next, the pretreated Dynabeads (catalog no. 11201D; Invitrogen) were incubated with chromatin and antibody mixture overnight at 4°C. Then, Dynabeads were washed with washing buffer (50 mmol/L of HEPES, pH 8.0, 1% NP-40, 0.7% deoxycholate, 0.5 mol/L of LiCl, freshly added protease inhibitor) 5 times followed by washing with Tris-EDTA buffer (10 mmol/L of Tris-HCl, pH 8.0, 0.1 mmol/L of EDTA) once. Then, 100 μL of elution buffer (50 mmol/L of Tris-HCl, pH 8.0, 1 mmol/L of EDTA, 1% SDS) was added and Dynabeads were incubated in ThermoMixer at 65°C for 30 min at maximum speed. The supernatant was collected and treated with proteinase K at 55°C for 2 h, then purified with the TIANquick Mini Purification Kit (catalog no. DP203-02; Tiangen). RNA was removed by treatment with RNase at 37°C for 1 h. DNA was then purified by AMPure beads (Vazyme) and subjected to DNA library preparation as described later on. ChIP-seq reads were aligned to mm9 with Bowtie2 (version 2.2.2) with parameters -t -q -N 1 -L 25. All unmapped reads, multiply mapped reads and PCR duplicates were removed. To generate the ChIP-Seq signals for each histone modification shown in the UCSC genome browser, we normalized the read counts by computing the number of reads per kilobase of bin per million reads sequenced (RPKM). The RPKM file was used to generate genome browser snapshots and further analysis. The used antibodies included Histone H3K9me3 antibody (Active Motif, Cat. No. 39161, Lot. No. 09919003), Histone H3K4me3 antibody (MERCK,Lot. No. 3068440), Histone H3K27me3 antibody (Diagenode, Cat. No. C15410069, Lot. No. A1818P) and ZFP57 antibody (Abcam, Cat. No. ab45341, Lot. No. GR3270465-1).
High-throughput sequencing library preparation and sequencing DNA was end-repaired, adenylated, and ligated to TruSeq sequencing adapters. After purification with VAHTS DNA Clean Beads (Vazyme), DNA was amplified by Phusion High-Fidelity DNA Polymerase (catalog no. M0530L; New England Biolabs) or Pfu Turbo C x Hotstart DNA polymerase (for bisulfite-treated DNA libraries). The amplified DNA was size-selected using 2% agarose gel or VAHTS DNA Clean Beads for 200 to 500-bp DNA fragments. All libraries were sequenced using an Illumina HiSeq 2500 or X10 system according to the manufacturer's instructions.

RNA-seq
500-1000 haploid cells were calculated by trypan blue staining. Amplificated cDNAs were collected by Single-Cell Full Length mRNA-Amplification Kit (Vazyme) according to the manufacturer's instructions. RNA-seq libraries were prepared by TruePrep DNA Library Prep Kit V2 for Illumina (Vazyme) according to the manufacturer's instructions. All RNA-seq reads were mapped to mm9 with TopHat (version 2.2.1). The mapped reads were further analyzed by Cufflinks, and the expression levels for each transcript were quantified as Fragments Per Kilobase of transcript per Million mapped reads (FPKM).

5mC dot blot
DNA samples diluted to the defined concentration were treated with 0.4 mol/L of NaOH and 10 mmol/L of EDTA at 99°C for 10 min before being cooled on ice immediately. The denatured DNA was loaded onto a nylon transfer membrane (catalog no. RNP303B; GE Healthcare) followed by ultraviolet cross-linking. The membrane was air-dried and blocked in a blocking buffer (10% milk, 1% BSA in PBS with Tween 20 (PBST)) at room temperature for 1 h. 5mC antibody (1:1000 dilution; Eurogentec, BI-MECY-1000) was diluted in a blocking buffer and incubated with the membrane at room temperature for 3 h. The membrane was washed with 1× PBST and incubated with a secondary antibody at room temperature for another 1 h. Then, the membrane was incubated with enhanced chemiluminescence substrate for exposure after washing with 1× PBST.

OPEN ACCESS
This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creativecommons.org/licenses/by/4.0/.