Introduction

Following fertilization, mouse embryos undergo a series of cell divisions and differentiation in the first few days. Zygotes and 2-cell blastomeres are totipotent and can form both embryonic and extra-embryonic supporting tissues. After several rounds of divisions, embryos undergo compaction at the morula stage and specifies the trophectoderm (TE) and inner cell mass (ICM). The ICM further segregates into the primitive endoderm (PE) and the epiblast (EPI) in the subsequent blastocyst stage. All cell types of the fetus can develop from the EPI, a feature known as pluripotency. During the establishment of pluripotency, dramatic molecular changes occur such as shifts in epigenetic modifications, chromatin accessibility and transcriptome (Deng et al, 2014; Guo et al, 2010; Wang et al, 2018; Wu et al, 2016; Zhang et al, 2016; Zhang et al, 2018). The precise mechanisms involved in establishing the transient pluripotent state during development remain elusive.

Oct4 (also known as Pou5f1) and Sox2 are prominent transcription factors (TFs) that regulate early embryonic development. In mice, Oct4 is expressed in all the cells of the compacted morulae, while Sox2 is first expressed in the inside cell and identified as one of the earliest markers to distinguish the inner from the outer cells (Guo et al., 2010; Palmieri et al, 1994; White et al, 2016). Both factors are confined to the pluripotent EPI at the late blastocyst stage (Avilion et al, 2003; Palmieri et al., 1994). Oct4 and Sox2 knockout (KO) leads to embryonic lethality around embryonic day (E) 4.5 and E6.0, respectively (Avilion et al., 2003; Nichols et al, 1998). Although Oct4- or Sox2-KO embryos develop apparently normal EPI, they fail to give rise to embryonic stem cells (ESCs) (Avilion et al., 2003; Nichols et al., 1998) and Oct4- KO embryos cannot contribute to embryonic tissues in chimera assays (Wu et al, 2013), suggesting that Oct4 and Sox2 are required for the EPI to acquire or maintain pluripotency. Surprisingly, Oct4- and Sox2-KO mouse embryos still express many pluripotency markers, such as Nanog, Klf4 and Zfp42 (Rex1) (Avilion et al., 2003; Frum et al, 2013; Le Bin et al, 2014; Stirparo et al, 2021; Wicklow et al, 2014; Wu et al., 2013). The controversy between the loss of pluripotency and the maintenance of the pluripotency-related genes in Oct4- and Sox2- KO EPI prompted us to investigate the role of Oct4 and Sox2 in early mouse embryos.

Recent studies suggested that Oct4 and Sox2 may function before the blastocyst stage. The long-lived presence of exogenous Oct4 and Sox2 proteins on DNA was observed in some of the 4-cell blastomeres, that tended toward the ICM rather than the TE (Plachta et al, 2011; White et al., 2016), suggesting that they may play a role in the initial lineage segregation between ICM and TE. Oct4 is also thought to contribute to the dramatic gain of chromatin accessibility at the 8-cell stage (Lu et al, 2016). Unlike in mouse embryos, the loss of OCT4 expression in human embryos compromises blastocyst development, leading to the downregulation of NANOG in the ICM and CDX2 in the TE (Fogarty et al, 2017). Oct4 controls zygotic genome activation (ZGA) in zebrafish and human, but not in mice (Gao et al, 2018; Leichsenring et al, 2013). Thus, when and how Oct4 and Sox2 function in the early embryogenesis varies from species to species.

Oct4 and Sox2 play preeminent roles in maintaining the pluripotency-related transcriptional network in ESCs (Masui et al, 2007; Niwa et al, 2000). However, it remains unclear how these two TFs regulate pluripotency-related genes in embryos. The obvious obstacles are the heterogeneity and the scarcity of the preimplantation embryos. Unlike ESCs, which comprise a homogenous cell population, individual cells within embryos gradually differentiate into three cell types, EPI, PE and TE. This inherent cellular heterogeneity within embryos, coupled with the temporal variability in development across different embryos, poses a significant challenge when investigating the molecular processes using conventional molecular techniques. For instance, using the entire blastocyst or ICM for qPCR and bulk RNA-seq analysis would inevitably lead to a dilution and blur of the molecular processes within EPI (Frum et al., 2013; Wu et al., 2013). Although single-cell genomics is powerful to dissect cellular heterogeneity, it suffers from high dropouts, which may lead to inaccuracies in genomic profiles and cell identities (Minow et al, 2023). Due to the scarcity of embryos, it is unfeasible to compensate for the inaccuracy by increasing the number of cells. For example, single-cell RNA-seq (scRNA-seq) still faces challenges in accurately distinguishing between ICM and TE progenitors in compacted morula, as well as between EPI and PE progenitors in the early ICM (Deng et al., 2014; Li et al, 2023; Yanagida et al, 2022).

MEK inhibitor (MEKi) promotes the ground state of ESCs by suppressing the PE-inducing ERK signaling pathway (Ying et al, 2008). In embryos, MEKi has been shown to activate Nanog in ICMs, effectively suppressing the development of PE and directing the entire ICM towards a pluripotent epiblast fate (Nichols et al, 2009). The naïve pluripotency and readiness for differentiation of MEKi-treated EPI has been confirmed by its contribution to chimaeras with germline transmission. MEKi may thus facilitate the study of molecular events within the EPI cells by reducing the heterogeneity in ICMs.

In this study, we analyzed the transcriptome and global chromatin accessibility in the Oct4- or Sox2-KO mouse embryos using low-input RNA-seq and ATAC-seq. The effects of Oct4 KO and Sox2 KO are relatively small in morulae, but become evident in ICMs at the blastocyst stage. Oct4 and Sox2 mainly activate pluripotency-related genes cooperatively through the enhancers containing the composite OCT-SOX motifs. We observed a substantial reorganization of open chromatin regions and transcriptome in the early ICM, which was disturbed in absence of either Oct4 or Sox2. These results indicate the critical roles of Oct4 and Sox2 in establishing the pluripotency network during early mouse development.

Results

Generation of maternal-zygotic KO embryos

To investigate the role of Oct4 and Sox2 in the development of preimplantation embryos, we produced four transgenic mouse lines: Oct4 KO labeled with mKO2 (monomeric kusabira- orange 2, Oct4^mKO2), Sox2 KO labeled with EGFP (Sox2^EGFP), floxed Oct4 (Oct4^flox) and floxed Sox2 (Sox2^flox) (Figure 1A, 1B; Materials and methods). Although the expression of mKO2 and EGFP is driven by the constitutive mouse PGK promoter, we observed that the expression of these reporters is weak until the 8-cell and blastocyst stages, respectively, when the strong zygotic expression of Oct4 and Sox2 is observed.

Figure 1.

Oct4 ^+/mKO2 and Sox2 ^+/EGFP heterozygous male mice were mated with Oct4 ^flox/flox; ZP3-Cre and Sox2 ^flox/flox; ZP3-Cre maternal KO female mice, respectively, to generate fluorescently labeled maternal-zygotic KO and unlabeled maternal KO (herein called control, Ctrl) embryos (Figure 1C, 1D). These reporter systems allow prospective identification of KO embryos using fluorescent microscopy (Figure 1-figure supplement 1), so we were able to pool embryos with the same genotype. Immunostaining results confirmed that the mKO2+ and EGFP+ embryos lost Oct4 and Sox2 proteins, respectively (Figure 1E, 1F).

Oct4 and Sox2 regulate the chromatin landscape and transcriptome in the ICM

To assess the genome-wide molecular impact of the loss of Oct4 and Sox2, we performed low-input ATAC-seq and RNA-seq using early (E2.75) and late (E3.25) compacted morulae, as well as early (E3.75) and late (E4.5) ICMs (Materials and methods). For ATAC-seq, we pooled embryos based on our reporter systems because single-embryo samples yielded sparse signals and too few peaks (data not shown). We opted to collect single morulae or ICMs for RNA-seq, as this approach enabled us to account for embryo-to-embryo variability and detect transcripts with greater sensitivity compared to scRNA-seq (Boroviak et al, 2015). Given that the ICM consists of the progenitors of EPI and PE cells, we treated the embryos with the MEKi (PD0325901) from E2.5 to suppress PE development and specify the entire ICM to the EPI (Nichols et al., 2009) (Figure 1-figure supplement 2).

Principal component analysis (PCA) of the ATAC-seq data reveals a clear clustering of samples based on the stages and genotypes (Figure 2A). The Oct4-KO early and late morulae clustered close to their Ctrl counterparts, while the Oct4- and Sox2-KO early and late ICM samples clustered separately from their Ctrl. In total, 152,393 peaks were identified across different stages and genotypes. Of these, 14,016 (9.2%) and 6,637 (4.4%) showed significant decreases, while 11,884 (7.8%) and 3,660 (2.4%) exhibited significant increases in Oct4- and Sox2-KO late ICM, respectively (Figure 2B; Data supplement 1). The number of decreased peaks exceeded that of increased peaks across all the KO samples. K-means clustering of all the 33,520 differential peaks shows that many peaks, which exhibited reduced accessibility in the KO ICMs, gained accessibility from the Ctrl morula to the Ctrl ICM (Figure 2C, clusters 1, 3, 8, 10, 11, 13 and 14). Notably, the peaks activated by Oct4 but not by Sox2 in the ICM tended to be already open at the morula stage (Figure 2B, clusters 1 and 11), whereas those dependent on both Oct4 and Sox2 became open in the ICM (Figure 2B, clusters 3, 8 and 14). Over 96.8% of the significantly changed peaks were identified as putative enhancers located distally to the transcription start sites (TSSs) (Figure 2-figure supplement 3A).

Figure 2.

To validate the authenticity of the Sox2-regulated downstream ATAC-seq peaks identified in this study, we compared the data to a recent study that did not employ MEKi treatment (Data ref: Li et al, 2023) (Li et al., 2023). The trends observed in our Sox2-KO late ICMs, including decreased, increased, and unchanged peaks, remained consistent in the absence of MEK inhibitor (Figure 2-figure supplement 3B). Remarkably, the decreased ATAC-seq peaks also showed higher enrichment with Sox2 CUT&RUN signals than the increased or unchanged peaks (Figure 2-figure supplement 3C). Analysis of published ChIP-seq data in ESCs (Data ref: Marson et al, 2008; Whyte et al, 2012) (Marson et al, 2008; Whyte et al, 2012) shows that the enhancers decreasing in the Oct4- or Sox2-KO late ICMs were frequently bound by Oct4, Sox2 and Nanog, and enriched with the active enhancer mark H3K27ac (Figure 2-figure supplement 3D). The Sox2-dependent peaks around the Sap30 gene are shown as examples (Figure 2-figure supplement 3E). These data suggest that the decreased peaks identified in our ATAC-seq could be the direct targets of Oct4 and Sox2 in the ICM.

To assess the impact of Oct4 and Sox2 deletion on the transcriptome, we performed RNA- seq on individual morulae or ICMs. Initially, we examined the consequences of maternal KO and zygotic heterozygous KO. PCA did not reveal significant distinctions between maternal flox and maternal KO, nor between WT/flox and flox/KO samples (Figure 2-figure supplement 4A). Furthermore, mice with heterozygous KO of Oct4 or Sox2 develop normally and exhibit fertility (Nichols et al., 1998). Therefore, we employed maternal-KO zygotic-WT/KO as our Ctrl group. In morulae, Oct4 KO had only a minor effect on the transcriptome (Figure 2D). We observed a progressively profound impact of Oct4 KO on the transcriptome from the morula to the late ICM, mirroring patterns at the chromatin level (Figure 2A-2D). As Sox2 is only expressed at very low levels until the blastocyst stage anyway, the transcriptome in Sox2-KO morulae was not significantly affected. Oct4 or Sox2 KO altered the expression of 2,485 (15.8%) and 967 (6.1%) genes in the late ICM, respectively (Figure 2D; Data supplement 2). Oct4 KO had a greater effect than Sox2 KO on both the chromatin accessibility and the transcriptome (Figure 2A-2D), suggesting Oct4 may play a more important role than Sox2. This is consistent with the earlier developmental arrest of Oct4-KO embryos at E4.5 compared to Sox2-KO embryos at E6.0. Down- and up-regulated genes identified in our Oct4- and Sox2- KO ICMs show enrichment with genes exhibiting down- and up-regulation, respectively, in the studies without MEKi (Data ref: Li et al., 2023; Stirparo et al., 2021) (Li et al., 2023; Stirparo et al., 2021) (Figure 2-figure supplement 4B-4D), confirming that MEKi did not alter the molecular functions of Oct4 and Sox2 in the ICM.

Next, we explored whether changes in chromatin accessibility affected the transcription of surrounding genes. Gene set enrichment analysis (GSEA) shows that the genes close to the decreased peaks were frequently downregulated, whereas those near the increased peaks were upregulated in the Oct4- and Sox2-KO embryos (Figure 2E; Figure 2-figure supplement 5). For example, the expression of Utf1, a known target of Oct4 and Sox2, and its enhancer accessibility decreased in the Oct4- and Sox2-KO ICMs (Figure 2F, 2G). On the other hand, the expression of Fgfr2 and its enhancer accessibility increased in the Oct4- and Sox2-KO ICM. In addition, integration of the ATAC-seq and RNA-seq data allowed us to identify previously unknown targets of Oct4 and Sox2, such as Sap30 and Uhrf1, which are essential for somatic cell reprogramming and embryonic development (Figure 2F, 2G) (Cao et al, 2019; Li et al, 2017; Maenohara et al, 2017).

Taken together, the data presented so far indicate that Oct4 and Sox2 play a crucial role in shaping the chromatin landscape and transcriptome in the ICM. Consequently, we observed apparently normal development of the Oct4- and Sox2-KO embryos up to the blastocyst stage.

Oct4 and Sox2 activate the pluripotency network in the ICM

To further explore the role of Oct4 and Sox2, we wanted to find out which genes they might influence in the ICM. As expected, many of the differential ATAC-seq peaks were consistently changed in Oct4- and Sox2-KO late ICMs (Figure 3A, Data supplement 1). However, there are a number of peaks exclusively changed in either Oct4- or Sox2-KO ICMs. The genes around the consistently decreased peaks were enriched for terms related to pluripotency and preimplantation embryonic development, such as cellular response to LIF signaling and stem cell population maintenance (Figure 3B, upper panel). The consistently elevated peaks preferentially located near the genes related to extraembryonic lineages and organogenesis, such as embryonic placenta, extraembryonic trophoblast, muscle and neural tube (Figure 3B, lower panel). Interestingly, the genes near the peaks which decreased only in the Oct4-KO but not in the Sox2-KO ICM were enriched with terms of LIF signaling and blastocyst formation (Figure 3-figure supplement 6).

Figure 3.

At the transcriptional level, many known pluripotency genes, such as Klf2, Etv5, Prdm14 and Pecam1, were downregulated in the Oct4- and Sox2-KO ICM (Figure 3C). In contrast, Gata3, Cdx2 and Eomes, which are important for TE development and differentiation (Ralston et al, 2010; Russ et al, 2000; Strumpf et al, 2005), were upregulated in the Oct4- and Sox2-KO ICM.

Interestingly, we observed that Nanog, Esrrb and Klf4 were significantly downregulated in the Oct4-KO ICM, whereas they were not or only slightly downregulated in the Sox2-KO ICM (Figure 3C, 3D). Downregulation of Pecam1 was confirmed at the protein level (Figure 3E). Chromatin accessibility at its putative enhancers also decreased accordingly (Figure 3F). Oct4 and Sox2 activated the components of several epigenetic modifiers, such as Ezh2 (PRC2 H3K27me2/3 methyltransferase) and Sap30 (a component of mSin3A histone deacetylase complex) (Figure 2F, 2G; Data supplement 2), suggesting their potential contribution to establishing the ICM-specific epigenetic status through regulation of the epigenetic modifiers.

Taken together, the above data show that Oct4 and Sox2 regulate a large number of TFs, epigenetic factors and signaling pathways in the ICM.

Oct4 and Sox2 co-activate their targets through OCT-SOX enhancers

Although there is a general consensus on the cooperative binding of Oct4 and Sox2 to the OCT-SOX composite motif, the principle of the cooperation still remains controversial in different scenarios (Biddle et al, 2019; Chen et al, 2014; Friman et al, 2019; Li et al, 2019; Michael et al, 2020). Therefore, we investigated how Oct4 and Sox2 regulate their target enhancers in the ICM. As expected, the OCT-SOX, OCT and SOX motifs were also among the most enriched motifs in the group of decreased enhancers in Oct4- or Sox2-KO ICMs (Figure 4A). The motifs of other known pluripotency-related TFs, such as Klf2/4 and Esrrb, were also enriched at the decreased enhancers. The elevated peaks were enriched for the GATA, TEAD, EOMES and KLF motifs, but not for the OCT-SOX, OCT or SOX motifs. Additionally, the gain of chromatin accessibility occurred at later stages compared to the loss of chromatin accessibility (Figure 2B). Thus, the increased peaks may not represent the direct targets of Oct4 or Sox2, but rather may be a secondary effect of the upregulated TE TFs, such as Gata3, Tead4, Eomes and Klf5/6, or the disruption of inhibitory interactions between Oct4/Sox2 and TE TFs (Niwa et al, 2005) (Figure 3C; Data supplement 2). As a control, the CTCF motif was not enriched in either the decreased or increased peaks. We confirmed the enrichment of those motifs at distal peaks and concluded that the motif analysis was not biased by the small number of the TSS peaks in the dataset (Figure 2-figure supplement 3A, Figure 4-figure supplement 7).

Figure 4.

To investigate the effect of Oct4 and Sox2 on the chromatin accessibility, we focused on the peaks in the ICM. It is worth noting that the top decreased peaks in the Oct4- and Sox2-KO early ICM were most enriched with the OCT-SOX, but not the OCT or SOX motifs (Figure 4B), which suggests that Oct4 and Sox2 could maintain open chromatin to a greater extent at peaks containing the OCT-SOX motif (hereafter referred to as OCT-SOX peaks). The KLF motif was more enriched at the moderately decreased peaks, indicating a possible secondary effect of down-regulated Klf2/4 (Figure 3C, 3D); alternatively, the depletion of Oct4 and Sox2 might lead to a decrease in DNA binding or the activity of Klf2/4 and other KLF TFs. In general, the accessibility of 8,993 OCT-SOX peaks decreased in both the Oct4- and Sox2-KO early and late ICM (Figure 4C; Figure 4-figure supplement 8). Furthermore, majority of the decreased OCT-SOX peaks were shared in Oct4- and Sox2-KO late ICM (Figure 4D), suggesting that both TFs participate in keeping these peaks open. The known OCT-SOX enhancers of Dppa3 and Klf4, which were enriched with the binding of Oct4 and Sox2, are shown as examples (Figure 4E).

The above results suggest that the enhancers containing the OCT-SOX motif are primary targets of Oct4 and Sox2 in the ICM.

Oct4 and Sox2 promote the developmental trajectory towards ICM

In early embryos, individual blastomeres are initially totipotent and indistinguishable before segregating into the ICM and TE. We subsequently investigated factors that influence the cellular trajectory toward the pluripotent ICM. Around 50% of the peaks and genes exhibiting decreased trend in Oct4- or Sox2-KO early ICMs were found to be activated in the Ctrl ICM during the natural development (Figure 5A). In alignment with earlier studies (Guo et al., 2010; Wu et al., 2016), a significant reorganization of open chromatin regions was observed upon the formation of the early ICM, resulting in the activation of 21,731 peaks (14.3%) (Figure 5B). The most prominently elevated peaks in the early ICM locate close to genes involved in cellular response to LIF, maintenance of stem cell population and blastocyst formation, in accordance with the developmental stage (Figure 5-figure supplement 9A). These peaks were enriched for OCT, SOX and OCT-SOX motifs (Figure 5-figure supplement 9B). However, when Oct4 and Sox2 were absent, the accessibility of the elevated peaks was decreased in the early ICM (Figure 5C). In alignment with the chromatin dynamics, the transcriptome also underwent substantial rearrangements in early ICMs, resulting in the upregulation of 1,115 genes (7.1%) (Figure 5D). These 1,115 genes were enriched with genes downregulated in the Oct4-KO and Sox2-KO early ICM (Figure 5E). These data indicate a critical role of Oct4 and Sox2 in initiating ICM-specific chromatin and transcriptional programs.

Figure 5.

The network of naïve pluripotency is governed by a core regulatory circuit of TFs, including Oct4, Sox2, Nanog, Esrrb and Klf4 (Adachi et al, 2018; Ng & Surani, 2011; Young, 2011). In particular, the activation of endogenous Sox2 represents a late-stage and deterministic event that triggers the fate towards pluripotency during reprogramming (Buganim et al, 2012). In mouse embryos, all the aforementioned TFs except Sox2 are highly expressed in the morula (Figure 5-figure supplement 9C); however, their expression does not seem sufficient to induce the mature pluripotent state at this stage. Therefore, we hypothesized that Sox2, in collaboration with Oct4, may activate the pluripotency-related genes in the early ICM. To investigate this hypothesis, we subjected the 1,115 genes and their log₂foldchange in Sox2- KO and Oct4-KO early ICMs to GSEA Wikipathway analyses (Liao et al, 2019). This analysis revealed an enriched downregulation of genes associated with PluriNetWork and ESC pluripotency pathways in Sox2-KO early ICMs (Figure 5F). In Oct4-KO early ICMs, although such enrichment was not observed, a significant downregulation of pluripotency-related genes was evident (Figure 5-figure supplement 9D, 9E). For example, in the early ICM, Oct4 and Sox2 activate Utf1 and Il6st (also known as gp130, receptor of the LIF/STAT pathway) through the OCT-SOX enhancers (Figure 2F, 2G, 5H). Notably, the OCT-SOX enhancer of Il6st is enriched with Sox2 CUT&RUN signals. Additionally, Oct4 activated enzymes regulating the pyruvate metabolism (Ldha, Me2, Pck2, etc.) and glutathione metabolism (Gstm1/2, Mgst2/3, Idh1), consistent with previous studies highlighting Oct4’s role in metabolism regulation (Frum et al., 2013; Stirparo et al., 2021). These findings indicate that Sox2 expression might function as a temporal regulator in the activation of OCT-SOX enhancers and a subset of pluripotency- related genes in the early ICM.

In summary, above data suggest that Oct4 and Sox2 play pivotal roles in promoting the ICM fate in early mouse embryos.

Discussion

Oct4 and Sox2 are key TFs for preimplantation embryonic development in several species (Daigneault et al, 2018; Fogarty et al., 2017; Gao et al, 2022; Lee et al, 2013; Leichsenring et al., 2013). However, because of the paucity of embryonic cells, it remains a matter of debate as when they start to function and how they regulate the development of the EPI lineage. In this study, we explored the roles of Oct4 and Sox2 in mouse preimplantation embryos using transgenic mouse models.

Previous reports suggested that Oct4 and Sox2 could direct 4-cell blastomeres towards ICM fate at the first lineage segregation (Plachta et al., 2011; White et al., 2016). siRNA-mediated knockdown experiments showed that Oct4 facilitates chromatin opening already at the 8-cell stage (Lu et al., 2016). Unexpectedly, our data showed that either Oct4-KO or Sox2-KO have minimal impact on both global chromatin accessibility and transcription until the early blastocyst stage (Figure 2A-2D). Oct4 is expressed at similar levels between morulae and ICMs; in contrast, Sox2 mRNA and protein levels are negligible in morulae, but are significantly upregulated in ICMs (Figure 5-figure supplement 9C) (Palmieri et al., 1994; Wicklow et al., 2014). Therefore, the function of Oct4 is likely limited by the insufficient levels of Sox2 and possibly other factors in morulae. Another plausible limiting aspect could be epigenetic constraints in morulae. The epigenetic state and 3-dimensional structure of the chromatin in the morula differ from those in the ICM, which may hinder Oct4 binding (Du et al, 2017; Wang et al., 2018; Wang et al, 2014; Zhang et al., 2016). However, considering that forced expression of Oct4 can sustain cell proliferation and pluripotency in Sox2-KO ESCs (Masui et al., 2007), it’s plausible that Oct4 and Sox2 may compensate each other in the morula as well. To delve deeper into this possibility, double KO embryos would be necessary for further investigation. A limitation of this study is that the whole morula was used for ATAC-seq and RNA-seq. Considering that morulae contain more TE progenitor cells than ICM progenitor cells, the signal from ICM progenitor cells might be diluted at the morula stage.

Oct4 is activated prior to Sox2 in both embryogenesis and reprogramming to iPSCs. In mouse embryos, Oct4 is initially highly expressed in all blastomeres from the 8-cell stage and later becomes restricted to the EPI, whereas Sox2 is repressed by the Hippo pathway in early stages and is specifically expressed in the EPI (Frum et al, 2019; Guo et al., 2010; Wicklow et al., 2014). In reprogramming, the activation of endogenous Sox2 occurs later than the endogenous Oct4 and signifies the final phase of iPSC generation (Buganim et al., 2012). In both scenarios, the expression of Sox2 coincides with the setup of the pluripotent state. Moreover, our data show that Sox2-KO ICMs fail to fully activate the pluripotent program (Figure 3C). Recently, we reported that the loss of Sox2 might contribute to decreased developmental potential of pluripotent cells upon priming (MacCarthy et al, 2024). This suggests that Sox2 plays a critical role in regulating the spatial-temporal onset of a subgroup of pluripotency-related genes.

Loss of Oct4 and Sox2 impairs embryonic pluripotency, as evidenced by the inability of KO embryos to give rise to ESCs or contribute to the embryonic part in chimera assays (Avilion et al., 2003; Nichols et al., 1998; Wu et al., 2013). To date, only a few of pluripotency-related genes have been found downstream of Oct4 and Sox2 in mouse embryos, whereas the majority, including Nanog, Klf4 and Zfp42 (Rex1), are reportedly independent of Oct4 and Sox2 (Frum et al., 2013; Stirparo et al., 2021; Wu et al., 2013). Here, we showed that both Oct4 and Sox2 are essential for establishing the pluripotent chromatin landscape and transcriptome (Figure 3A, 3C, 5F, 5G). We found that a large number of pluripotency-related genes were downregulated in Oct4-KO ICM, contrasting with previous reports which suggested that most pluripotency-related genes were maintained (Frum et al., 2013; Le Bin et al., 2014; Stirparo et al., 2021; Wu et al., 2013). This discrepancy may be due to differences in the experimental design. First, unlike previous study that used the whole blastocysts (Frum et al., 2013), we managed to isolate a pure EPI cell population performing immunosurgery and using short-term inhibition of MEK to reduce the heterogeneity of the ICMs while it did not alter the target enhancers and genes of Oct4 and/or Sox2 (Nichols et al., 2009) (Figure 1-figure supplement 2; Figure 2-figure supplement 3B-3D, 4B-4D). Second, we removed Oct4 exons 2 to 5 (Figure 1A), encompassing the whole POU domain (bipartite DNA-binding domain of Oct4), resulting in a complete ablation of Oct4 function. In contrast, some of the previous studies removed only exon 1 (Frum et al., 2013; Wu et al., 2013), which encodes for N-terminal transactivation domain. The remaining exons 2 to 5 might be translated into a truncated version of Oct4 that could retain partial function. Finally, unlike previous study that used maternal-WT zygotic-KO embryos (Stirparo et al., 2021), we used maternal-zygotic-KO embryos (Figure 1C). The residual levels of maternal Oct4 could have potentially rescued the KO.

To regulate the expression of developmental genes, preimplantation embryos undergo dynamic epigenomic reprogramming, governed by multiple epigenetic modifiers and complexes. For example, Ezh2 is required for the establishment and maintenance of global H3K27me3 in preimplantation embryos; Uhrf1 maternal-KO embryos fail to form healthy morula due to misregulation of DNA methylation and histone modifications (Cao et al., 2019; Maenohara et al., 2017; Puschendorf et al, 2008). Our data show that Oct4 and Sox2 activated the components of several epigenetic modifiers, including Ezh2, Uhrf1 and Sap30 (Figure 2F, 2G; Figure 2-figure supplement 3E; Data supplement 2), and thus may be involved this way in regulating the epigenetic status.

The core regulatory circuitry of pluripotency exhibits remarkable stability in embryos. Expression of the TFs within this circuitry, such as Oct4, Sox2, Nanog, Klf4 and Esrrb, is partially or fully maintained in the Oct4- or Sox2-KO ICMs (Le Bin et al., 2014; Wicklow et al., 2014; Wu et al., 2013) (Figure 3C, 3D). The independent and robust expression of these TFs likely ensures the faithful activation of the pluripotency regulatory circuit, safeguarding successful development.

While the cooperative role of Oct4 and Sox2 on the OCT-SOX enhancers in maintaining or inducing the pluripotent network is widely acknowledged, there are still debates regarding the precise mechanisms underlying their cooperation (Biddle et al., 2019; Chen et al., 2008; Han et al, 2022; Li et al., 2019; Michael et al., 2020; Velychko et al, 2019). In this study, most OCT- SOX peaks are more affected in Oct4-KO ICMs than in Sox2-KO ICMs, indicating that Oct4 plays a more important role than Sox2 in activating OCT-SOX enhancers. This observation aligns with a previous study showing that the forced expression of Oct4 can rescue pluripotency in Sox2-null ESCs (Masui et al., 2007). In addition to their cooperation on OCT- SOX enhancers, recent studies have unveiled independent functions of Oct4 and Sox2 (Friman et al., 2019; Gao et al., 2022). Consistently, we observed that Oct4 regulates a considerable number of peaks and genes independently of Sox2 in the ICMs (Figure 2C, 3A, 3C, 3D). We believe that this outcome is unlikely to be attributed to compensation from other members in the Sox family. Sox1, Sox3, Sox15 and Sox18 are promising candidates as they can either rescue the Sox2-KO phenotype in mESCs or replace Sox2 to generate iPSCs in reprogramming (Nakagawa et al, 2008; Niwa et al, 2016). However, our analysis reveals that Sox1, Sox3 and Sox18 are expressed at extremely low levels in both wildtype and Sox2-KO embryos (Data supplement 2), suggesting that they are unlikely to be able to fulfill Sox2’s role (Deng et al., 2014). Sox15 exhibits indistinguishable expression levels in all three cell types at the blastocyst stage. Furthermore, Sox15 KO mice displays normal health and fertility (Lee et al, 2004). Nonetheless, we cannot completely rule out the possibility of functional compensation by Sox15 in Sox2-KO ICMs.

Our results highlight the crucial role of Oct4 and Sox2 in establishing the transcriptome and chromatin state in the pluripotent EPI. At the blastocyst stage, Oct4 and Sox2 work together to open the OCT-SOX enhancers and activate the pluripotency-related genes (Figure 5I). The absence of Sox2 and other factors likely limits the function of Oct4 in morulae. However, the upstream factors driving the activation of the core pluripotency regulatory circuitry remain unknown. Further studies are needed to deepen our understanding of the molecular mechanisms governing the embryonic pluripotency program and the early lineage segregation.

Materials and methods

Mice

To generate ESCs carrying a floxed Oct4 allele (Oct4^flox), the construct containing floxed Oct4 exon 2-5 and a promoter-less FRT-IRES-βgeo-pA cassette was electroporated into germline- competent Acr-EGFP ESCs. Clones were screened for homologous recombination and transiently transfected with an FLP expression vector to remove the FRT cassette. To generate ESCs carrying a mutant Oct4 allele linked to mKO2 (Oct4^mKO2), Acr-EGFP ESCs were targeted with the construct containing a PGK-pacΔtk-P2A-mKO2-pA cassette 3.6 kb upstream of the Oct4 TSS and a promoter-less FRT-SA-IRES-hph-P2A-Venus-pA cassette in Oct4 intron 1. To generate ESCs carrying the floxed Sox2 allele (Sox2^flox), the construct containing the 5’ loxP site 1.9 kb upstream of Sox2 TSS and a promoter-less FRT-IRES-Neo- pA cassette followed by the 3’ loxP site in the 3’ UTR of Sox2 was electroporated into Acr- EGFP ESCs. Clones were screened for homologous recombination and transiently transfected with an FLP expression vector to remove the FRT cassette. To generate ESCs carrying a conditional allele of Sox2 linked to EGFP (Sox2^EGFP), Acr-EGFP ESCs were targeted with the construct containing the 5’ loxP site 1.9 kb upstream of Sox2 TSS, and a promoter-less FRT-IRES-Neo-pA cassette, the 3’ loxP site, and a PGK-EGFP-pA cassette in the 3’ UTR of Sox2. Chimeric mice were generated by morula aggregation and heterozygous mice were obtained through germline transmission. The Zp3-Cre transgenic mice were used to generate maternal KO alleles. Animal experiments and husbandry were performed according to the German Animal Welfare guidelines and approved by the Landesamt für Natur, Umwelt und Verbraucherschutz Nordrhein-Westfalen (State Agency for Nature, Environment and Consumer Protection of North RhineWestphalia). The work was funded by the Max Planck Society and the Max Planck Society’s White Paper-Project “Animal testing in the Max-Planck- Society”.

Embryo collection

All the embryos for RNA-seq and ATAC-seq were collected from female mice superovulated with PMSG and hCG. Early and late morula samples were collected at E2.75 and E3.25, respectively. Early and late ICM samples were collected at E2.5 and cultured in KSOM with MEKi (PD0325901, 1 μM) until early and late blastocyst stages, respectively.

ICM isolation

Zona pellucida was removed by brief incubation in prewarmed acidic Tyrode’s solution. ICM was obtained by immunosurgery. Briefly, blastocysts were incubated in 20% rabbit anti-mouse whole serum (Sigma-Aldrich) in KSOM at 37°C, 5% CO2 for 15 minutes, followed by three rinses in M2. Afterwards, embryos were incubated in 20% guinea pig complement serum (Sigma-Aldrich) in KSOM at 37°C, 5% CO2 for 15 minutes, followed by three rinses in M2. In the end, ICM was isolated by repetitive blowing in mouth pipette to remove the debris of dead trophectoderm cells.

Immunofluorescence

Immunofluorescence was performed using the following primary antibodies with dilutions: mouse monoclonal anti-Oct4 (sc-5279, Santa Cruz), 1:1000; goat anti-Sox2 (GT15098, Neuromics), 1:300; rabbit anti-mKO2 (PM051, MBL), 1:1000; rabbit anti-GFP (ab290, Abcam), 1:500; goat anti-Pecam1 (AF3628, R&D biosystems), 1:300; rat monoclonal anti-Nanog (14- 5761-80, Thermo Fisher Scientific), 1:100; goat anti-Sox17 (AF1924, R&D Systems), 1:1000 of 0.2 mg/ml.

ATAC-seq and data analysis

ATAC-seq libraries were prepared as previously described with some modifications. 20 to 40 morulae or ICMs were pooled and lysed in the lysis buffer (10 mM Tris·HCl, pH=7.4; 10 mM NaCl; 3 mM MgCl2; 0.15% Igepal CA-630; 0.1% Tween-20) for 10 min on ice. We added 0.1% Tween-20 to the lysis buffer as it greatly reduced reads mapped to mitochondrial DNA. Lyzed embryos were briefly washed, collected in 2.0 μl PBS containing 0.1 mg/ml polyvinyl alcohol (PVA) and transferred to 3.0 μl Tn5 transposome mixture (0.25 μl Tn5 transposome, 2.5 μl tagmentation buffer, Illumina, FC-121-1030; 0.25 μl H₂O). The samples were incubated in 37 °C water bath for 30 min. Tagmentation was stopped by adding 2.0 μl 175 mM EDTA and incubated at 50 °C for 30 min. Excess EDTA was quenched by 2.0 μl 160 mM MgCl₂. The libraries were amplified in the following reaction: 9.0 μl Transposed DNA, 10.0 μl NEBNext High-Fidelity 2×PCR Master Mix (New England Biolabs, M0541S), 0.25 μl 100 μM PCR Index 1, 0.25 μl 100 μM PCR Index 2 and 0.5 μl H₂O. The sequences of index 1 and 2 are in Table supplement 2. The PCR program is as the following: 72 °C for 5 min; 98 °C for 30s; 16 thermocycles at 98 °C for 10 s, 63 °C for 30 s and 72 °C for 1 min; followed by 72 °C 5 min. Amplified libraries were purified twice with 1.2 × AMPure XP beads. The sequencing was performed on the NextSeq 500 system with pair-end 75bp.

Sequence reads were trimmed for adapter sequences using SeqPurge and trimmed reads no shorter than 20 bases were aligned to the mm10 mouse reference genome using Bowtie2 with a maximum fragment size of 2000. Duplicated reads were removed using Picard MarkDuplicates (https://broadinstitute.github.io/picard/) and only reads uniquely mapped to the standard chromosomes except chrY and chrM (mitochondrial genome) with mapping quality of at least 30 were used for the following analysis. Reads on the positive and negative strands were shifted +4 bp and −5 bp, respectively. Peak calling was performed using MACS2 with the following parameters: --keep-dup all -g mm --nomodel --shift -50 --extsize 100 -B -- SPMR. Peaks that overlap with the blacklisted regions (https://sites.google.com/site/anshulkundaje/projects/blacklists; https://sites.google.com/site/atacseqpublic/atac-seq-analysismethods/mitochondrialblacklists-1) and satellite repeats (RepeatMasker) were removed.

Significant peak summits (q-value ≤0.001) from biological replicates were merged for each sample group using BEDTools with a maximum distance between two summits to be merged of 200 bp. Only merged peak summits that overlap with summits from at least two replicates were retained. Reads per million (RPM)-normalized pileup signals in the bedGraph format were converted into bigWig files using the UCSC bedGraphToBigWig tool. The bigWig files for the average signal of biological replicates were generated using the UCSC bigWigMerge tool. The heatmaps of RPM-normalized pileup signals around the TSSs of GENCODE (vM23) protein-coding genes were generated using deepTools.

For count-based analysis of transposon insertion events, merged peak regions were generated by combining the adjacent peak summits of all embryo groups and selecting 200- bp regions around the centers of each merged summit. 5’ ends of shifted read alignment on both the positive and negative strands were considered as transposon insertion sites. The number of transposon insertions in each merged peak region were counted for each sample using BEDTools and normalized and transformed to log₂ scale using the rlog function of the Bioconductor package DESeq2 and significantly differential peaks were identified with an adjusted p-value cutoff of 0.05. Peaks located within 100 bp from TSSs of GENCODE (vM23) genes were considered as TSS-proximal peaks. The presence of TF binding motifs in merged peaks was investigated using FIMO with the default cutoff (p-value <10^-4).

The following JASPAR motifs were used: MA0036.3 GATA2, MA0139.1 CTCF, MA0141.3 ESRRB, MA0142.1 Pou5f1-Sox2, MA0143.3 Sox2, MA0524.2 TFAP2C, MA0599.1 KLF5, MA0792.1 POU5F1B, MA0800.1 EOMES, MA0808.1 TEAD3 and MA0878.1 CDX1.

Significantly enriched motifs in each peak cluster were identified from using PscanChIP (Zambelli et al, 2013). Functional annotation of gene sets associated with ATAC-seq peaks was performed using GREAT (McLean et al, 2010) with default settings.

RNA-seq and data analysis

Single-embryo RNA-seq was performed as previously described with some modifications (Kurimoto et al. 2007, Nakamura et al. 2015). 2,493 copies of ERCC RNA Spike-In (Ambion) were added to each single-embryo sample. To remove PCR duplicates, we used R2SP-UMI- d(T)24 primer, instead of V1d(T)24 primer, for reverse transcription (RT). To reduce byproducts derived from the RT primer, the poly(A) tailing reaction was performed only for 1 min at 37°C. cDNA was amplified with V3d(T)24 and R2SP primers and Terra PCR Direct Polymerase (Clontech) for 16 cycles. The cDNA was further amplified for four cycles with NH2- V3d(T)24 and NH2-R2SP primers, and purified thrice with 0.6x AMPure XP beads. The 3’ end enriched libraries were constructed using KAPA HyperPlus Library Preparation Kit (VWR International, KK8513) with 6 cycles of PCR. All the oligos used in the RT and library amplification are in Table supplement 3. The sequencing was performed on the NextSeq 500 system with single-end 75bp.

UMI sequence of each read was extracted from FASTQ files for Read 2 using UMI-tools. Poly- A sequences were trimmed from the 3’ end of Read 1 using Cutadapt (https://cutadapt.readthedocs.org/). The trimmed Read 1 sequences were aligned to the mm10 mouse reference genome using TopHat2 with GENCODE (vM23) transcripts as a transcriptome reference and with default parameters. The reads were also mapped to ERCC reference sequences using Bowtie2 with default parameters. To confirm the genotype of Oct4- and Sox2-KO embryos, we also mapped the reads to the PGK-pacΔtk-P2A-mKO2-hGHpA or PGK-EGFP-hGHpA cassette. The reads mapped to the genome were assigned to GENCODE transcripts using featureCounts. The number of unique transcripts assigned to each GENCODE gene was calculated using TRUmiCount. The transcript counts were normalized and transformed to log₂ scale using the rlog function of the Bioconductor DESeq2 package. Differential expression analysis was performed on genes that were detected in at least half of the samples in at least one stage using DESeq2 and differentially expressed genes were identified with an adjusted p-value <0.05 and log₂ fold change ≥1.

GSEA

Gene set enrichment analysis (GSEA) (Subramanian et al, 2005) was performed to determine whether genes close to the selected ATAC-seq peaks are enriched in genes differentially expressed in Oct4- or Sox2-KO embryos. The ATAC-seq peaks were annotated to Ensembl protein-coding genes whose TSSs are located within 10 kb of the peak centers using the R package ChIPpeakAnno (Zhu et al, 2010). False discovery rate (FDR) was estimated by gene set permutation tests. Heatmaps were generated from normalized enrichment scores (NESs) for gene sets with FDR ≤0.1.

Acknowledgements

We appreciate the invaluable assistance from Ingrid Gelker, Claudia Ortmeier, David Obridge, Martina Sinn and the animal facility. Special thanks go to Dong Han, Rui Fan, Eva Kutejova and Erik Tolen for their insightful discussions. We acknowledge the generous support from Anika Witten, Christoph Bartenhagen and Carolin Walter of the Core Facility Genomics at the University of Muenster for their support with sequencing. Additionally, we would like to express our gratitude to Shixue Gou and Hui Zhang from the Guangzhou National Laboratory for their kind suggestions on RNA-seq data analysis and manuscript revisions, respectively. This work was supported by the Max Planck Society.

Data Availability

ATAC-seq and RNA-seq data have been deposited at GEO under GSE264614 and GSE264615, respectively. Published high-throughput sequencing datasets used in this manuscript are listed as follows: ATAC-seq and CUT&RUN of early embryos, GSE203194; scRNA-seq of early embryos, GSE203194 and GSE159030; Oct4/Sox2/Nanog ChIP of mESCs, GSE11724; H3K27ac ChIP of mESCs, GSE27844. This study includes no data deposited in external repositories.

Author contributions

K.A. and Y.H. designed the study. Y.H. performed the experiments, assembled and interpreted data, and wrote the manuscript. K.A. performed the bioinformatic analysis, interpreted the data and edited the manuscript. G.W. generated the transgenic mice and advised on the project.

Z.N. and S.H. assisted with the experiments. Y.H. and Q.J. contributed to the bioinformatic analysis. S.V. and I.B. advised on the project and edited the manuscript. H.R.S. supervised the project and edited the manuscript.

Figure supplements

Figure supplement 1.

Figure supplement 2.

Figure supplement 3.

Figure supplement 4.

Figure supplement 5.

Figure supplement 6.

Figure supplement 7.

Figure supplement 8.

Figure supplement 9.