Paternally inherited H3K27me3 affects chromatin accessibility in mouse embryos produced by round spermatid injection

Round spermatid injection (ROSI) is a valuable and the sole method for obtaining offspring from patients with male infertility due to spermatogenesis disruption, especially nonobstructive azoospermia (NOA) (Tekayev and Vuruskan, 2021; Tesarik et al., 1996). Embryos produced by ROSI (ROSI embryos) can develop to term in mice, rabbits, rats and humans (Hirabayashi et al., 2002; Kimura and Yanagimachi, 1995b; Ogura et al., 1994; Sofikitis et al., 1994; Tesarik et al., 1995). Nevertheless, the success rate of full-term development following ROSI is lower than intracytoplasmic sperm injection (ICSI) embryos. Furthermore, despite reports that healthy offspring were obtained using ROSI in humans, the therapeutic application of ROSI has not been widely adopted due to its uncertain safety (Aslam et al., 1998; Tanaka et al., 2015, 2018; Tekayev and Vuruskan, 2021). Therefore, toward its clinical application, it is important to understand at the molecular level why ROSI embryos show poor development. Current research implies transmitted histone proteins from round spermatid cause abnormalities in the epigenome, transcriptome and, consequently, embryonic development of ROSI embryos (Hayashi et al., 2003; Kishigami et al., 2006). Ectopic accumulation of a repressive mark H3K9me3 has been found in male pronuclei (PN) in one-cell ROSI embryos at the pericentromeric heterochromatin regions, suggesting the presence of histone marks inherited from round spermatids (Kishigami et al., 2006). However, the role of paternal chromatin and molecular underpinning of the abnormal transcriptome and poor embryonic development in ROSI embryos remains to be elucidated.

Male germ cells that reached the haploid round spermatid stage undergo dynamic morphological and physiological changes to form sperm capable of fertilization. This process is known as spermiogenesis and the chromatin structure changes dynamically: histones are replaced by protamine in order to pack the genome into a highly condensed state in sperm (Rathke et al., 2014). Thereby, histone content in sperm is dramatically reduced during spermiogenesis and, accordingly, the distribution of histone modifications and variants, which are tightly linked to gene expression, is dynamically altered (Erkek et al., 2013; Tatehana et al., 2020). Despite the global reduction in histone levels, histones in certain genomic regions are retained in sperm and are considered to be inherited into zygotes (Erkek et al., 2013; Hammoud et al., 2009; Jung et al., 2019; Yamaguchi et al., 2018; Yoshida et al., 2018). Furthermore, perturbation of an active histone methylation mark (H3K4me3) during spermatogenesis disturbs embryonic development (Lismer et al., 2020; Siklenka et al., 2015), highlighting the importance of inherited epigenome information for embryonic gene expression. In addition to H3K4me3, H3K27me3, a polycomb repressive complex 2 (PRC2)-mediated repressive mark, is also considered an inheritable epigenetic mark that regulates embryonic gene expression, and H3K4me3 and H3K27me3 together form a bivalent chromatin domain that represents germline developmental potential (Erkek et al., 2013; Lesch et al., 2013, 2016; Maezawa et al., 2018b; Sin et al., 2015; Teperek et al., 2016). Therefore, the fact that ROSI embryos receive more histones than ICSI embryos is likely to be linked to the presence of excessive and undesired histone marks, which may adversely influence the chromatin accessibility landscape and embryonic transcriptome.

We hypothesised that epigenetic information in round spermatids may be transferred to ROSI embryos, leading to detrimental effects in the regulation of gene expression or chromatin remodelling processes that are important for embryonic development. Therefore, in this study, we performed a comparative analysis between mouse embryos generated either by ROSI or ICSI using transposase-accessible chromatin sequencing (ATAC-seq) and RNA-seq. We identified differences in chromatin accessibility as well as gene expression, which were potentially triggered by carried-over epigenetic information. Notably, we found that chromatin regions marked by H3K27me3 in round spermatids, but not in mature sperm, are associated with lower accessibility and influence gene expression at zygotic genome activation (ZGA). Therefore, our results demonstrate that spermiogenesis completion is crucial for establishing a normal chromatin accessibility landscape as well as for proper embryonic gene expression.

To investigate the differences in chromatin accessibility and transcriptome between ROSI and ICSI embryos, we performed an assay using ATAC-seq and RNA-seq for ROSI and ICSI embryos at the one- and two-cell stages (Fig. 1A). We selected embryos with two PN with normal morphology and one-cell samples were used after 12-13 h post-activation (hpa) or injection (hpi) (for ROSI or ICSI, respectively) and 28-30 h for two-cell samples. In the ATAC-seq, as the number of available embryos upon ROSI and ICSI was limited, we employed Omni ATAC-seq, an improved ATAC-seq protocol (Gou et al., 2020; Corces et al., 2017). We used 400-450 zygotes and blastomeres per sample with the removal of the first and second polar bodies (PB) (Fig. 1B). We prepared two biological replicates and obtained at least 21 million reads uniquely mapped to the genome per sample. We also performed RNA-seq. We confirmed that reproducible data were obtained from biological replicates and that our ATAC-seq data for ICSI two-cell embryos showed high similarity with published ATAC-seq data for two-cell embryos (Wu et al., 2016) (Fig. 1C; Fig. S1A-E). Importantly, in contrast to a previous report in which ATAC-seq peaks (ATAC peaks) were undetectable at the one-cell stage (Wu et al., 2016), we detected clear ATAC signals using an improved method.

We first found that ATAC-seq signals in ROSI and ICSI embryos show globally similar profiles, suggesting that chromatin remodelling normally occurs in ROSI embryos (Fig. 1C,D). As expected, ATAC peaks that represent open chromatin regions (Table S1), were observed at active chromatin regions enriched in H3.3 and H3K4me3, but not at H3K27me3-marked repressive chromatin regions in ICSI or ROSI two-cell embryos (Ishiuchi et al., 2021; Liu et al., 2016; Wang et al., 2018; Wu et al., 2016) (Fig. 1C,D). H3K9me3 was enriched at the transcription start sites (TSSs) associated with relatively high accessibility at the two-cell stage. This may reflect a non-repressive function of H3K9me3 at these embryonic stages (Burton et al., 2020; Wang et al., 2018). Similar tendencies were also observed in ICSI or ROSI one-cell embryos, but under a weaker correlation with active marks. This is expected, as H3.3 and H3K4me3 show non-canonical distribution at the one-cell stage (Ishiuchi et al., 2021; Zhang et al., 2016). For a more detailed analysis of open chromatin regions, we classified gene promoter regions into four clusters by k-means clustering based on the strength of ATAC peaks of one-cell ROSI, one-cell ICSI, two-cell ROSI and two-cell ICSI embryos. Promoter ATAC signals were positively correlated with gene expression levels. Given the inheritance of maternal mRNA into one-cell embryos, the positive correlation between gene expression and ATAC-seq signals at the one-cell stage may indicate that genes actively transcribed in oocytes have high promoter accessibility, even at the one-cell stage (Fig. 1D). Notably, in clusters 1, 2 and 3, we observed gene expression upregulation during development to the two-cell stage (Fig. S1F), although ATAC signals at their promoters diminished (Fig. 1D). These findings suggest that promoter accessibility correlates with transcriptional output; however, is not the determinant. Therefore, other factors than promoter accessibility may cooperatively fine-tune transcriptional output during ZGA.

We next investigated how transcriptome is affected in ROSI embryos. We identified 186 upregulated and 16 downregulated genes in ROSI embryos at the one-cell stage [termed as one-cell ROSI-high and one-cell ROSI-low differentially expressed genes (DEGs), respectively] [false discovery rate (FDR)<0.01, fold change (FC)>1.5], and 649 upregulated and 513 downregulated genes were detected at the two-cell stage (Fig. 2A,B). At the one-cell stage, most DEGs were upregulated genes and their gene expression in ICSI embryos was extremely limited (Fig. 2A; Fig. S2A). As the transcription activity at the one-cell stage is low, we considered that a higher abundance of these transcripts in ROSI one-cell embryos may reflect carry-over of RNA from round spermatids. To test this hypothesis, we first performed gene ontology (GO) analysis of these genes and identified the terms ‘sperm motility’ and ‘spermatogenesis’ (Fig. 2C). Indeed, transition proteins Tp1 (also known as Tnp1) and Tp2 (Tnp2), which play a crucial role in spermiogenesis during histone-protamine replacement (Zhao et al., 2004), were contained, which is consistent with a previous study (Hayashi et al., 2003). We then analyzed published RNA-seq data for cells at different spermatogenesis stages to examine their gene expression levels (Hasegawa et al., 2015; Kobayashi et al., 2012). The expression levels of these genes peaked at the round spermatid stage and decreased during maturation into sperm (Fig. 2D). To further investigate this observation, we performed qRT-PCR. For this assay, we prepared one-cell ROSI embryos treated with or without 5,6-dichloro-1-β-D-ribofuranosyl-benzimidazole (DRB), which blocks zygotic transcription. Transcripts of one-cell ROSI-high DEG genes, Tp1, Tp2 and Smcp showed much higher abundance in ROSI embryos than ICSI embryos, and their higher abundance was not affected by DRB treatment (Fig. 2E). In contrast, zygotically expressed genes, such as Klf5, were significantly decreased by DRB treatment (Fig. S2B). These results strongly indicate that round spermatid-derived RNA is transferred into ROSI one-cell embryos.

At the two-cell stage, when embryos undergo major ZGA, DEGs were identified in ROSI embryos (Fig. 2B). We again performed GO analysis and found spermatogenesis-related genes were still enriched in the upregulated gene group in ROSI two-cell embryos (two-cell ROSI-high DEGs) (Fig. 2C). Accordingly, 54 genes of the two-cell ROSI-high DEGs overlapped with the one-cell ROSI-high DEGs (Fig. S2C) and the upregulated status of these 54 genes were maintained between ROSI one-cell and two-cell embryos (Fig. S2D). Therefore, RNA carry-over from round spermatids appears to still contribute to differential gene expression at the two-cell stage. However, its contribution to overall differential expression is small, suggesting ZGA may be affected in ROSI two-cell embryos. To gain further insight into the mechanisms underlying differential gene expression in ROSI two-cell embryos, we analyzed promoter chromatin accessibility. However, there were no obvious differences in chromatin accessibility at the promoters of DEGs between ICSI and ROSI embryos (Fig. 2F), suggesting chromatin accessibility at their promoter is not globally related to the altered transcriptome of ROSI embryos.

Although ATAC-seq signals were not altered at the DEG promoter regions, we speculated that differential chromatin accessibility outside of these promoters may be associated with abnormal development of ROSI embryos. We then first performed quantitative analysis of ATAC-seq data (see Materials and Methods) and identified ATAC peaks showing higher or lower accessibility in ROSI embryos compared with ICSI embryos (referred to as ROSI-high and ROSI-low ATAC peaks, respectively). Although the number of these peaks was limited and more than 90% of total ATAC peaks remained unchanged between ICSI and ROSI embryos (classified as common peaks) (Fig. 3A), we further investigated the potential cause of differential chromatin accessibility. The number of ROSI-high peaks was consistently higher than ROSI-low peaks in one- and two-cell ROSI embryos, and ROSI-high peaks at the one-cell stage were frequently observed within the intronic or intergenic region (Fig. 3B; Fig. S3A). Only a few cases of ROSI-high or -low peaks were commonly detected between one- and two-cell stages (Fig. S3B). Therefore, most of these differential peaks were developmental stage-specific. Because altered chromatin accessibility upon ROSI application is highly represented in one-cell ROSI embryos, we decided to focus on differential chromatin accessibility at the one-cell stage for further analysis.

To characterize one-cell ROSI-high and -low ATAC peaks, we analyzed published chromatin immunoprecipitation sequencing (ChIP-seq) data for epigenetic marks in normal embryos (H3.3, H3K4me3, H3K27me3 and H3K9me3) (Ishiuchi et al., 2021; Liu et al., 2016; Wang et al., 2018). One-cell ROSI-high ATAC peak regions were characterized by weak enrichment of H3K4me3, suggesting that accessible regions are excessively formed in regions where H3K4me3 is not normally enriched (Fig. S3C,D). On the other hand, one-cell ROSI-low ATAC peak regions were the regions in which H3K4me3 but not H3K27me3 is normally enriched. These features were similarly observed at randomly selected ATAC peak regions. Thus, this indicates that one-cell ROSI-low peak regions display a general feature of accessible regions in normal one-cell embryos and that accessibility of a small fraction of these H3K4me3-marked and H3K27me3-depleted regions become lower in one-cell ROSI embryos. Moreover, enrichment of H3K4me3 at these regions tended to be maintained during early development as well as in embryonic stem cells (ESCs), while H3K27me3 was not enriched (Fig. S3E). Therefore, these results indicate that chromatin accessibility decreased at generally active chromatin regions in one-cell ROSI embryos. We then speculated that undesirable inheritance of epigenetic information from round spermatids to ROSI embryos may result in differential chromatin accessibility. To explore this possibility, we analyzed published ATAC-seq and ChIP-seq data for round spermatids and sperm (Erkek et al., 2013; Jung et al., 2019; Maezawa et al., 2018a) (Fig. 3C). We observed higher chromatin accessibility in sperm compared with round spermatids irrespective of one-cell ROSI-high or one-cell ROSI-low ATAC peak regions (Fig. 3C; Fig. S4A,B). Thus, the differences in chromatin accessibility between round spermatids and sperm did not correlate with differential chromatin accessibility between one-cell ROSI and ICSI embryos. Histone H3.3, enrichment of which around the CpG-rich promoters in sperm reflects histone retention (Erkek et al., 2013), was enriched at the one-cell ROSI-low ATAC peak regions, but not at the one-cell ROSI-high ATAC peak regions in sperm, suggesting one-cell ROSI-low ATAC peak regions are the regions in which histones are preferentially retained (Fig. 3C). Interestingly, although H3K4me3 enrichment at one-cell ROSI-high or ROSI-low peak regions was similar between round spermatids and sperm, we observed higher H3K27me3 enrichment in round spermatids compared with sperm at the one-cell ROSI-low peak regions (Fig. 3C,D; Fig. S4A,B). This observation was confirmed by analyzing other publicly available datasets (Hammoud et al., 2014; Iwamori et al., 2016; Jung et al., 2017) (Fig. S4C). This indicates that H3K27me3-enriched chromatin regions in round spermatids, in which its enrichment becomes weaker during maturation into sperm, show lower chromatin accessibility in one-cell ROSI embryos. Notably, this lower chromatin accessibility was cancelled at the two-cell stage (Fig. 3C; Fig. S3B), thus these changes in accessibility appear to be transient. To explore our hypothesis that H3K27me3 in round spermatids affects the chromatin accessibility in ROSI embryos, we analyzed the enrichment of H3K27me3 and H3K4me3 at the one-cell ROSI-high and -low ATAC peak regions as well as at the randomly selected regions in round spermatids. Importantly, we observed significant enrichment of H3K27me3 at the one-cell ROSI-low ATAC peak regions (Fig. S4D,E). In addition, this region was also mildly enriched in H3K4me3 (Fig. S4D,E). Altogether, these data suggest that chromatin accessibility at regions generally marked by active H3K4me3 in early embryos is suppressed in one-cell ROSI embryos, possibly due to the inheritance of H3K27me3-marked repressive chromatin states from round spermatids.

To further test our hypothesis that repressive chromatin status is inherited from round spermatids, we performed immunostaining for H3K27me3 at the one- and two-cell stages. To distinguish paternal and maternal chromatin, we also performed immunostaining for H3K9me2, stronger signals of which were observed on the maternal chromatin at the one- and two-cell stages (Burton and Torres-Padilla, 2010). Similar to the previous observation for H3K9me3 (Kishigami et al., 2006), ectopic H3K9me2 signals were observed at the pericentromeric heterochromatin region in male PN of ROSI one-cell embryos (Fig. 4A,B). On the contrary, at the one-cell stage, H3K27me3 intensity in male PN in ROSI embryos was globally higher than that in ICSI embryos (Fig. 4A,B). Furthermore, consistent with this observation, asymmetric H3K27me3 signals between parental chromosomes in ICSI two-cell embryos was not observed in ROSI two-cell embryos (Fig. 4C). Round spermatids showed higher signals for H3K27me3, but not sperm (Fig. 4D), suggesting ROSI embryos may possess an abnormal landscape of H3K27me3 by inheriting this mark from round spermatids. We also asked whether there are any other histone marks that would alter chromatin accessibility in ROSI embryos. Among the histone marks analyzed (H3K4me1, H3K4me3, H3K27me3 and H3K36me3), remodelling of H3K4me1 during spermiogenesis was also observed, but this modification was similarly enriched in both one-cell ROSI-high and low ATAC peak regions (Fig. S4C). Thus, at this point, H3K27me3 is the only histone mark tightly associated with altered chromatin accessibility in ROSI embryos.

H3K27me3 represses gene expression (reviewed by Zhang et al., 2015). Therefore, we next investigated whether lower chromatin accessibility in ROSI one-cell embryos is associated with changes in gene expression. We classified ATAC peaks into distal ATAC peaks and promoter ATAC peaks (peaks at TSS±1 kb). We observed no changes in gene expression when focused on distal ATAC peaks or ROSI-high ATAC peaks (Fig. 5A). In contrast, gene expression associated with ROSI-low promoter ATAC peaks (146 genes) was lower than ICSI embryos (Fig. 5A; Fig. S5A-D). Lower expression was prominent at the two-cell stage, which suggests that promoter chromatin accessibility at the one-cell stage might be important for future gene activation. Furthermore, we found that the one-cell ROSI-low promoter ATAC peaks that show H3K27me3 enrichment in round spermatids (H3K27me3 RPKM>1 in round spermatid) are mainly associated with this lower gene expression (Fig. S6A). Notably, although the tendency of downregulation was observed in this gene group, only a few DEGs (FDR<0.01 and FC>1.5, as seen in Fig. 2B) were included (Fig. 5B), indicating that lower chromatin accessibility only mildly affects transcription.

Our immunostaining data revealed that the distribution of H3K27me3 in ROSI embryos differed largely from ICSI embryos, and therefore we hypothesised that inheritance of H3K27me3 from round spermatids may suppress gene expression without altering chromatin accessibility. We then analyzed H3K27me3 and H3K4me3 enrichment in round spermatids or sperm around the TSSs of DEGs at the two-cell stage. Interestingly, prominent enrichment of H3K27me3 as well as mild enrichment of H3K4me3 was observed at the ROSI-low DEG promoters in round spermatids, but H3K27me3 enrichment diminished in sperm (Fig. 5C,D; Fig. S6B). In contrast, weaker H3K27me3 enrichment as well as higher enrichment of H3K4me3 was observed at the ROSI-high DEG promoters in round spermatids, indicating that the promoters of two-cell high and low DEGs possess bivalent H3K4me3 and H3K27me3 modifications at different levels in round spermatids. Given the previous observation that bivalent promoters in round spermatids are prevalent for genes expressed after fertilization (Lesch et al., 2013, 2016; Maezawa et al., 2018b; Sin et al., 2015), our results suggest that removal of H3K27me3 during spermiogenesis might be prerequisite for proper embryonic transcription.

The results above indicated that the inheritance of repressive chromatin status from round spermatids might cause lower chromatin accessibility and altered gene expression in ROSI embryos. However, we did not find any possible involvement of histone modifications in acquisition of higher chromatin accessibility in ROSI embryos (Fig. 3C; Fig. S4C). Thereby, we suspected that an alternative mechanism such as a transcription factor (TF)-driven mechanism might be involved in this process. To explore this possibility, we performed a TF motif analysis on one-cell ROSI-high, ROSI-low and ROSI-common ATAC peaks. We obtained some TF motifs at these peaks (Fig. S7A). Among the identified motifs, we focused on SOX30, a member of the SOX TF family, as it was highly expressed in spermatocytes and round spermatids and the SOX30 motif was uniquely identified at the one-cell ROSI-high ATAC peaks (Bai et al., 2018; Osaki et al., 1999; Zhang et al., 2018) (Fig. S7B). To examine whether SOX30 was enriched at one-cell ROSI-high ATAC peak regions in round spermatids, we analyzed published SOX30 ChIP-seq data in round spermatids (Chen et al., 2018a). However, the enrichment of SOX30 was weak in these regions (Fig. S7C). Nevertheless, we consider that it would be possible that SOX30 carried over from round spermatids binds to genomic regions where SOX30 did not bind in round spermatids as large-scale chromatin remodelling occurs in early embryos. Next, we attempted to perform immunostaining for SOX30 at the one-cell stage. However, we were not able to detect reliable SOX30 signals in ROSI embryos (Fig. S7D). Although we could not identify factors which contribute to the higher chromatin accessibility in ROSI embryos, it would be important to identify RNAs or proteins transmitted from round spermatids in future studies to comprehensively understand the chromatin alteration in ROSI embryos.

To our best knowledge, this is the first report identifying the relationship between altered transcriptome and chromatin accessibility in ROSI embryos. Our transcriptome analysis revealed that ROSI embryos have RNA transmitted from round spermatids, as suggested previously (Hayashi et al., 2003). The number of detected DEGs was limited at the one-cell stage, which is consistent with limited zygotic transcriptional activity during this stage. In contrast, we detected a larger number of DEGs at the two-cell stage, possibly due to a transcriptional burst, called major ZGA, during this stage. Conversely, we found the number of differential accessible regions decreased during transition from one- to two-cell stages (Fig. 3B). Therefore, although defects in chromatin accessibility are corrected to some extent during development, ROSI embryos are likely to suffer from altered gene expression. However, we cannot exclude the possibility that there may be more defects in transcription in one-cell ROSI embryos than detected in this study as it is hard to see a relatively low amount of zygotic transcripts with a conventional RNA-seq at this stage.

By focusing on differentially accessible chromatin regions, we found that lower accessible regions in one-cell ROSI embryos are marked by H3K27me3 in round spermatids. Supporting this observation, we detected the presence of ectopic H3K27me3 in ROSI embryos. This finding prompted us to observe that genes significantly downregulated in ROSI embryos were associated with high levels of H3K27me3 in round spermatids. Therefore, we propose that transmission of H3K27me3 from round spermatids lowers chromatin accessibility at the one-cell stage and alters gene expression profiles at the two-cell stage in ROSI embryos (Fig. 6). This lower accessibility appeared to be corrected at the two-cell stage (Fig. 3C; Fig. S4B). These data suggest that chromatin accessibility at the one-cell stage is important to regulate gene expression at the two-cell stage, although there is still a possibility that ectopic H3K27me3 directly influences gene expression without altering chromatin accessibility at the two-cell stage. We also noticed that lower accessible regions in one-cell ROSI embryos showed high enrichment of H3.3 in sperm, but not in round spermatids. During spermiogenesis, histone H3 variant H3T is typically replaced by H3.3, another histone variant (Ueda et al., 2017). Therefore, H3K27me3 is likely to be inherited into ROSI embryos under the absence of this process. Thus, our results highlight the biological significance of chromatin remodelling during spermiogenesis for fine-tuning gene expression after fertilization.

In sperm, H3K27me3 at the one-cell ROSI-low ATAC peak regions decreased, but H3K4me3 was still enriched. Therefore, fertilized embryos inheriting H3.3 marked by H3K4me3 may be advantageous for embryonic gene activation (Fig. 6). These observations imply manipulating H3K27me3 inheritance may help improve low birth rates of ROSI embryos. Recently, a study reported that treating round spermatids with Trichostatin A (TSA) before round spermatids were injected into oocytes improved the development of ROSI embryos to the blastocyst stage (Hosseini and Salehi, 2022). Although exploring whether TSA treatment improves their full-term development is still essential, this method may be useful for identifying potential causes of low birth rates of ROSI embryos. Because previous studies reported that global levels of H3K27me3 decreased following TSA treatment in somatic cell nuclear transfer-derived embryos and ESCs (Karantzali et al., 2008; Selokar et al., 2016), it is possible that improved development of ROSI embryos by TSA is due to reduced H3K27me3 levels in round spermatid. Therefore, identifying the genomic distribution of H3K27me3 in ROSI-derived embryos in an allele-discriminative manner would be crucial given the recent progress in low-input epigenetic analytical techniques. These analyses will clearly pinpoint comprehensive knowledge on not only the causes of low birth rates but also the epigenetic events during spermiogenesis essential for embryonic development.

We used 8- to 12-week-old female ICR (for ATAC-seq; n=247) and C57BL/6 (for RNA-seq; n=72) mice as oocyte donors. Same strain male mice (ICR: n=12, C57BL/6: n=9) were also used as the sperm donor. All animal experiments were approved by the Ethics Committee of the University of Yamanashi (reference number: A29-24) and conducted following Guide for the Care and Use of Laboratory Animals and ARRIVE guidelines. All mice were housed under specific pathogen-free conditions at a constant temperature of 25°C, relative humidity of 50% and a 14 h light/10 h dark period with ad libitum access to a commercial diet and distilled water. In this study, body weight was not measured because the body weight of young mice does not affect embryo quality.

Cumulus cell and oocyte complexes (COCs) were obtained from super-ovulated ICR and C57BL/6 female mice by injection of 7.5 IU equine chorionic gonadotropin (ASKA Pharmaceutical) and human chorionic gonadotropin (ASKA Pharmaceutical) at 46-50 h intervals. Obtained COCs were treated with hyaluronidase for 10 min in HEPES-buffered CZB medium (CZB-HEPES) (Kimura and Yanagimachi, 1995a). For ICSI, sperm were collected from the cauda epididymis, then cultured in human tubal fluid (Quinn et al., 1995) for capacitation for 1 h. Before cytosolic injection of denuded oocytes, sperm tails were eliminated with a Piezo drive micromanipulator (Prime Tech) in CZB-HEPES medium supplemented with 10% PVP (10% PVP-CZB-HEPES). After cutting the tail, the zona pellucida and cytosolic membranes were also disrupted with a Piezo drive micromanipulator. For ROSI, the obtained testes were minced with scissors, sieved through a Mini Cell Strainer and then re-suspended in 10% PVP-CZB-HEPES. The nucleus of each round spermatid was collected with a narrow pipette (7-8 µm diameter). The zona pellucida and cytosolic membrane were disrupted as per the ICSI. The oocytes injected with round spermatid nuclei were activated by culturing in Ca²⁺-free CZB medium containing 5 mM SrCl₂ for 1-2 h.

Zygotes were treated with cytochalasin B (CB) before elimination for 10-20 min in KSOM, then moved to CB containing CZB-HEPES. The zona pellucida was broken by a Piezo drive micromanipulator. We then vacuumed the second PB with a pipette (10 µm diameter; Fig. 1B). The zona pellucida was solved by Acidic Tyrode. During the wash, the first PB was eliminated by repeated pipetting. Embryos were then cultured in KSOM until sample collection.

We collected 411-450 zygotes and 210-231 two-cell stage ICR embryos (corresponds to 420-462 blastomeres, respectively) for each sample. We then performed omni-ATAC following previously described methods (Corces et al., 2017), with small modifications. To avoid embryo loss, we collected live embryos without PB and zona pellucida, which were transferred into cold ATAC-resuspension buffer and mixed by tapping. After ATAC-reaction for 30 min at 37°C, DNA products were cleaned with a MinElute PCR purification kit (Qiagen). We stored 22 µl of the eluted samples at −20°C until subsequent PCR amplification. After five cycles of pre-amplification, we performed an additional PCR for nine cycles. Amplified DNA was purified by a MinElute PCR purification kit. The amplified ATAC-seq libraries were subjected to sequencing by Novaseq 6000 (150-nt paired-end sequencing).

RNA was extracted from 50 zygotes and 49-96 two-cell stage embryos (C57BL/6 background) with Isogen II (Nippon Gene), according to the manufacturer's instructions. We performed RNA quality analysis on the Bioanalyzer RNA Pico Chip (Agilent Technologies). RNA-seq libraries were constructed by SMART-seq stranded kit with 1 ng of extracted total RNA according to the manufacturer's instructions. RNA fragmentation was then conducted for 2 min. To amplify constructed cDNA, we performed PCR1 (94°C for 1 min, five cycles of 98°C for 15 s, 55°C for 15 s, 68°C for 30 s and 68°C for 2 min) and PCR2 (94°C for 1 min, 14 cycles of 98°C for 15 s, 55°C for 15 s and 68°C for 30 s). The amplified RNA-seq libraries were then subjected to sequencing using Novaseq 6000 (100-nt paired-end sequencing).

ICSI- and ROSI-derived-zygotes and two-cell stage embryos were collected at 10 and 28-30 hpi or hpa, respectively. Embryos were fixed by 4% paraformaldehyde (PFA) containing 0.2% Triton X-100 (Tx) in PBS, then washed in PBS containing 1% bovine serum albumin (BSA) and 0.2% Tween (Tw) (1% BSA-0.2% Tw). For Sox30 staining, the zona pellucida was removed using Acidic Tyrode after fixation with 4% PFA for 20 min, then the zygotes were permeabilized using PBS containing 0.2% Tx for 20 min. Embryos were then incubated in the following antibodies at 4°C overnight: H3K27me3 rabbit polyclonal antibody (Millipore 07-449, 1:500), H3K9me2 mouse monoclonal antibody (Abcam ab1220, 1:500), Sox30 mouse monoclonal antibody (Santa Cruz Biotechnology sc-390333, 1:500) and H3K9me3 rabbit polyclonal antibody (Abcam ab8898, 1:2000). Embryos were then washed by 1% BSA-0.2% Tw and incubated with secondary antibodies against rabbit or mouse IgG labelled with Alexa Fluor 488, 568 or Cy5 (Invitrogen A11034, A11011, A11001 and Abcam ab97037, 1:500). The immune-stained embryos were then mounted on an antifade mounting medium with DAPI (4′,6-diamidino-2-phenilindole, Vector Laboratories). Fluorescence signals were detected using confocal microscopes FV1000 or FV1200 (Olympus). For quantitative analysis of H3K27me3 levels, we measured fluorescence signal intensity using ImageJ software (Schindelin et al., 2012). We then divided the H3K27me3 signal in each nucleus using the DAPI signal for normalization. Statistical analysis was performed using Prism 9 software (GraphPad). Sperm and round spermatids obtained from minced seminiferous tubules were mounted on slides and fixed in 4% PFA containing 0.2% Tx in PBS for 20 min. Specimens were washed in PBS containing 1% BSA and immunostained for H3K27me3. The H3K27me3 signal in sperm and round spermatids was detected using confocal microscopes FV1200 (Olympus).

Ten ICSI or ROSI one-cell embryos at 10 hpi or hpa, respectively, were pooled per sample. To inhibit transcription in ROSI embryos, MII oocytes were cultured in KSOM containing 80 µM DRB (Sigma-Aldrich) before ROSI, and ROSI embryos were then cultured in the presence of DRB.

The samples were added with 5000 copies of H2B-eGFP mRNA (Ooga et al., 2016) as an external control. Quantitative real-time polymerase chain reaction (qRT-PCR) was performed using a CellAmp Direct RNA Prep Kit (TAKARA) and One Step TB Green PrimeScript RT-PCR Kit II (TAKARA) following the manufacturer's instructions. After treatment with DNase I to remove genomic DNA, cell lysates containing total RNA were used as a template for a 10 µl qRT-PCR reaction. The primers used were as follows: Tp1 (5′-GAGGAGAGGCAAGAACCGAG and 5′-CGGTAATTGCGACTTGCATCA); Tp2 (5′-GCACTCTCGACACTCACCTG and 5′-TGTATCTTCGCCCTGAGCTA); Smcp (5′-GTTGCCAGATCAGGAACTCCA and 5′-TGGACAGCACGGTTTAGGTG); eGFP (5′-GCTACCCCGACCACATGAAG and 5′-TCTTGTAGTTGCCGTCGTCC); Duxf3 (5′-CCTCAAAGAGGTCCATCAGAGG and 5′-TGAAACCATATGTGGATCGTGTC); MuERV-L (5′-CTCTACCACTTGGACCATATGAC and 5′-GAGGCTCCAAACAGCATCTCTA); Nid2 (5′-CACCGAGGACAGTTTCCATT and 5′-CCAGTTACCAGGTGCTGGAT); Klf5 (5′-ACGTACACCATGCCAAGTCA and 5′-GTGGGAGAGTTGGCGAATTA) (Abe et al., 2018).

Adaptor and low-quality sequences were removed by ‘fastp’ (version 0.20.1) (Chen et al., 2018b) using option ‘-l 30 -w 10’. Reads were aligned to the mouse genome (mm10) using Bowtie2 (version 2.4.3) (Langmead and Salzberg, 2012) with option ‘--no-mixed --no-discordant -X 2000’ and unmapped reads were discarded by Samtools (version 1.4.1) (Li et al., 2009) with option ‘-F 0×4’. Reads from PCR duplicates were removed using Picard MarkDuplicates (version 2.23.8) (https://broadinstitute.github.io/picard/). For suiting similar read numbers between replications, reads were divided using ‘divide_bam.py’ in RseQC (Wang et al., 2012). After confirming reproducibility between replicates, they were merged using Samtools ‘merge’. To identify differentially accessible chromatin regions and common peak regions between ROSI and ICSI embryos, each replication peak was first called by MACS2 (version 2.2.7.1) (Zhang et al., 2008) with the option ‘--nomodel --shift −90 --extsize 180 -f BAMPE -B -q 0.01 -g mm’. Overlapping regions from each replication were then calculated using BEDtools (Quinlan and Hall, 2010) intersect. Overlapping regions and merged read data were used to calculate normalized read density between samples using MAnorm (version 1.3.0) (Shao et al., 2012) and identify differentially accessible chromatin regions under the conditions of FC>1.5 (M-value in MAnorm output data). Published ATAC-seq data (GSE102954, GSE116854, GSE66390) (Jung et al., 2019; Maezawa et al., 2018a; Wu et al., 2016) were also processed similarly and aligned to the mouse genome, as described above. K-means clustering and heatmap depiction were conducted using ‘computeMatrix’ and ‘plotHeatmap’ in deepTools (version 3.5.1) (Ramírez et al., 2016). ATAC peaks were annotated with nearby genes using annotatePeaks.pl (used GENCODE vM12 database) from HOMER (Heinz et al., 2010) and defined at least ±1.0 kb away from the TSS as distal peaks. ‘findMotifsGenome.pl’ from HOMER with the default option was used for finding the sequence motif enriched in ATAC-seq peaks.

Adaptor and low-quality sequences were removed by ‘fastp’ with the option ‘-l 50 -w 10’. Reads corresponding to ribosomal RNA were discarded from analysis by extracting reads unmapped to ribosomal DNA using Bowtie2. Subsequently, reads were aligned to the mouse genome (mm10) using STAR (version 2.7.4a) (Dobin et al., 2013). Read counts were calculated using featureCounts (Rsubread version 2.4.3 package in R) (Liao et al., 2014, 2019). To identify DEGs between ROSI and ICSI embryos, the read count data were processed using EdgeR (version 3.32.1) (Robinson et al., 2010). Genes with FDR<0.01 and FC>1.5 or <0.67 were used to identify DEGs. We then performed GO enrichment analyses using DAVID (Huang et al., 2009a,b). Published RNA-seq data (GSE55060, DRA000484 and E-MTAB-2950) (Abe et al., 2015; Hasegawa et al., 2015; Kobayashi et al., 2012) were also processed, as described above. Transcripts per million (TPM) of each gene annotated by GENCODE vM12 was quantified using R (http://www.r-project.org/). The ‘cor’ and ‘hclust’ functions with option ‘method=ward.D2’ in R were used to calculate Spearman rank correlation coefficient and perform hierarchical clustering, respectively. We performed principal component analysis using the ‘prcomp’ function in R and ‘ggplot’ (package ggplot2 version 3.3.5) was used for plotting. We performed k-means clustering after calculating z-scores based on TPM values and heatmaps using the ‘pheatmap.type’ function (package Pigengene version 1.16.0). Kruskal–Wallis test (Fig. 5A) was performed using the R package ‘ggpubr’ version 0.4.0.

Published ChIP-seq data were obtained from GSE139527, GSE73952, GSE97778, GSE42629, GSE49624, GSE79230, DRA004778 and GSE107644 (Chen et al., 2018a; Erkek et al., 2013; Hammoud et al., 2014; Ishiuchi et al., 2021; Iwamori et al., 2016; Jung et al., 2017; Liu et al., 2016; Wang et al., 2018). Reads were aligned to the mouse genome (mm10) using Bowtie2 with the option ‘--no-mixed --no-discordant’ after removing adaptor and low-quality sequences using ‘fastp’. Unmapped reads were removed using Samtools with option ‘-F 0×4’ and PCR duplicates were removed using Picard MarkDuplicates. Heatmaps were plotted using ‘computeMatrix’ and ‘plotHeatmap’ in deepTools. The average read tag density at the promoter region of each DEG group was analyzed using ngs.plot.r (version 2.47.1) (Shen et al., 2014).

Statistical analyses were implemented with R or Prism 9 (GraphPad). Data for two experimental groups were analyzed using a two-tailed Wilcoxon rank-sum test.

We thank Drs Satoshi Kamimura, Jafar Sharif, So Maezawa, Akihiko Sakashita, Satoshi Kishigami, Mr Kousuke Kazama, Miss Mayu Tatara and members of the T.W. laboratory and University of Yamanashi, Advanced Biotechnology Centre, for constructive discussions and technical assistance with the NGS analysis.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.200696.reviewer-comments.pdf.