Contribution of cohesins STAG1 and STAG2 to preimplantation development in mice

Abstract

Cohesin complexes carrying either STAG1 or STAG2 display functional specificities in both cohesion and genome organization. While any variant is sufficient for cell viability, both are essential for murine embryonic development, and mutations in STAG1 and STAG2 are associated to human syndromes. The requirements and regulation of the cohesin variants in the first stages of development remain unknown. Here, we show that Stag1 and Stag2 gene products are stored in oocytes. Maternal STAG1 persists beyond the 2-cell stage while accumulation of STAG2 depends on zygotic transcription. Despite clear differences in STAG1:STAG2 ratio in the two lineages emerging from the first cell-fate decision, both maternal and zygotic STAG1 and STAG2 are dispensable for viability in preimplantation stages if the other variant is present. However, the absence of STAG2 increases DNA damage in zygotes and affects transcription in blastocysts. These early-stage alterations may contribute to defects leading to embryo lethality by mid-gestation.

Imaging and transcriptomic analyses in murine embryos reveal distinct regulation of STAG1 and STAG2; while either cohesin complex is sufficient to complete preimplantation development, the absence of STAG2 affects both DNA repair and transcription.

Introduction

Cohesin mediates sister chromatid cohesion and also plays an important role in the spatial organization of the genome through loop extrusion¹. Two versions of cohesin that carry the STAG1 or STAG2 subunit associated with SMC1A-SMC3-RAD21 are present in all vertebrate somatic cells². Previous results from our group and others have revealed distinct roles for the two complexes, both in cohesion and 3D genome architecture. Cohesin-STAG1 and cohesin-STAG2 are more relevant for telomere and centromere cohesion, respectively^3–6. Cohesin-STAG2 acts preferentially in homologous recombination (HR)-mediated DNA repair and has been implicated in replication fork progression^7–9. The two cohesins display different chromatin association dynamics that likely explain why cohesin-STAG1 has a longer residence time on chromatin, makes longer loops, is present almost exclusively at CTCF-bound sites and is more important for delimitation of contact domains^10–14. Cohesin-STAG2 instead can be found at additional non-CTCF genomic positions, interacts with transcription factors and other transcriptional regulators and is responsible for local contacts that regulate tissue-specific gene expression^15–18. In mouse embryonic stem cells, cohesin-STAG1 preserves the integrity of topological associating domain (TAD) boundaries and prevents excessive compartmentalization, while cohesin-STAG2 contributes to local topologies such as superenhancers and Polycomb domains, which are important to define stem cell identity¹⁹. The specific functions of cohesin-STAG1 and cohesin-STAG2 are surprising given the extensive homology of the two paralogs, which display 79% identity along their central region (around 1000 amino acid (aa)-long) and only differ in the N-terminal (∼80 aa, 20% identity) and C-terminal (∼200 aa, 40% identity) regions².

Mutations in STAG1 and STAG2 have been identified in patients with developmental and intellectual impairment^20–24. Mouse embryos knock out (KO) for Stag1 and Stag2 are embryonic lethal but survive until mid-gestation. Lethality of Stag1 KO embryos can be observed after E11.5, although some exceptional embryos reach E18.5. No clear defects other than a smaller size have been described for these embryos⁵. In contrast, lethality is more acute for Stag2 KO embryos, and happens at E10.5. Abnormal heart development due to a combination of transcriptional and proliferation defects is the likely cause of death, although impairment of additional organs such as the brain has also been reported^6,25. Whether the defects causing embryonic lethality arise at mid-gestation or are the consequence of alterations at earlier stages is unclear.

After fertilization, the differentiated gametes that fuse to form the zygote undergo massive chromatin remodeling and epigenetic reprogramming to reach a totipotent state that then declines progressively from the zygote to the blastocyst²⁶. Initially, the embryo is transcriptionally silent and relies on maternal factors stored during oocyte growth^27,28. Just a few hours after fertilization, a wave of transcription occurs that is known as “minor” zygotic genome activation (ZGA)²⁹. At the 2-cell stage, a second wave of transcription or “major” ZGA is accompanied by active degradation of maternally-deposited RNA and protein^30,31. After two additional symmetric cell divisions, polarization and compaction at the 8-cell stage are followed by asymmetric divisions that segregate two lineages: the extraembryonic trophectoderm (TE), which gives rise to the placenta, and the pluripotent cells of the inner cell mass (ICM). The second lineage specification occurs in the blastocyst, when the ICM differentiates into the epiblast (EPI), which will give rise to fetal tissues, and the primitive endoderm (PE), which contributes mostly to the yolk sac³². These processes are accompanied by changes in chromosome organization, at least in part driven by cohesin and CTCF^33,34. TADs can be observed already in the zygote, but their boundaries are weak and intra-TAD interactions are low until the 8-cell stage, when they become “mature”^35–37. Germline deletion of Rad21 or Smc3 leads to unrepaired DNA lesions during the first cell cycle, causing persistent DNA damage that can delay entry into mitosis, accompanied by major segregation defects that cause 2-cell arrest^38,39. The essential role of cohesin for chromosome segregation makes it challenging to study its function in preimplantation development beyond the zygote⁴⁰. In contrast, a single cohesin variant complex is sufficient to achieve enough cohesion to allow cell viability, at least in tissue culture cells⁴¹. This gives us the opportunity to address the specific contributions of cohesin-STAG1 and cohesin-STAG2 to the early stages of embryonic development. The Stag1 KO and Stag2 KO embryos described above are zygotic KOs, and it is possible that the maternal contribution is preventing the observation of defects earlier in embryogenesis.

Here, we have analyzed the dynamics of cohesin-STAG1 and cohesin-STAG2 during mouse preimplantation development and found a clearly distinct behavior of the two complexes. While both paralogs have maternal contribution, this maternal supply appears to be the primary source of STAG1 protein, which persists beyond ZGA, whereas zygotic STAG2 becomes detectable at the 4-cell stage. Both maternal and zygotic STAG1 and STAG2 are dispensable for embryo viability; however, elimination of STAG2 results in defects in DNA repair and altered transcription.

Results

Distinct abundance of cohesin-STAG1 and cohesin-STAG2 throughout preimplantation stages

To assess the presence of the two cohesin variants in preimplantation stages, murine embryos were taken from pregnant wild-type females at different times after fertilization and stained simultaneously with antibodies against STAG1, STAG2 and RAD21. Nuclear STAG1 staining was clearly detected from the zygote to the 4-cell stage, decreased in 8-cell and morula stages and increased again in blastocysts (Fig. 1A; quantification in Fig. 1B). In contrast, STAG2 was detectable in the zygote and 2-cell stage but increased dramatically at 4-cell stage and was maintained in subsequent stages (Fig. 1A; quantification in Fig. 1C). RAD21 abundance changed accordingly to that of STAG paralogs, reflecting total amount of cohesin, and being highest in the blastocyst (Fig. 1A; quantification in Fig. 1D). Analysis of available transcriptomic and ribosome profiling data^42,43 showed that changes in STAG1 and STAG2 protein levels are paralleled by changes in transcript abundance and translation, while this is not the case for RAD21 (Supplementary Fig. 1A–D). Thus, cohesins STAG1 and STAG2 undergo distinct fluctuations in protein levels from the zygote to the blastocyst stage, probably driven by distinct regulation of their maternal components and/or zygotic transcription. This could respond to specific requirements for one or the other paralog at different stages of preimplantation development.

Fig. 1. Changes in cohesin-STAG1 and cohesin-STAG2 abundance in early embryos.

A Representative images of preimplantation embryos (a single confocal plane per embryo) stained with RAD21, STAG1 and STAG2 antibodies and DAPI. BF, brightfield. Scale bar, 20 µm. Quantification of STAG1 (B), STAG2 (C), and RAD21 (D) fluorescence across preimplantation development. Each dot represents the mean fluorescence signal of all nuclei from a single embryo. The line represents the mean of all embryos, and whiskers represent the standard deviation. At least three independent litters were analyzed for each stage.

Changes in the STAG1:STAG2 ratio upon the emergence of distinct cell types

The relative abundance of cohesin-STAG1 and cohesin-STAG2 is cell-type specific, with STAG2 being usually more abundant than STAG1¹¹. Moreover, we and others have previously observed increased STAG2 and decreased STAG1 levels when mouse embryonic stem cells progress from the more pluripotent naïve (2i) state to the primed epiblast-like state^19,44. The first cell-fate decision of the embryo takes place during the morula-to-blastocyst transition, and results in the specification of the outer layer of cells to the trophectoderm (TE) lineage while the ICM remains pluripotent (Fig. 2A). We therefore asked if changes in the relative abundance of cohesin variants could be observed upon TE-specification. Immunostaining of blastocysts with STAG1, STAG2 and RAD21 alongside CDX2, a TE lineage-specific transcription factor, revealed differences in the behavior of the two variants (Fig. 2B). STAG1 protein levels are higher in the ICM than in the TE, while the opposite is true for STAG2 (Fig. 2B–D). RAD21 behaves like STAG2, with its highest levels found in the TE, which implies that STAG2 is the more abundant variant at the blastocyst stage (Fig. 2B, E). Similar changes occur at the transcriptomic level⁴⁵ (Supplementary Fig. 1E). These results suggest that cohesin-STAG2 becomes more relevant as cells lose pluripotency, while cohesin-STAG1 is more critical in undifferentiated cells, consistent with the results observed in cultured cells.

Fig. 2. Changes in the STAG1:STAG2 ratio in cell types arising from the first cell-fate decision.

A Illustration of the first cell-fate decision of the embryo during the morula-to-blastocyst transition. B Representative images of a single confocal plane of blastocysts stained with antibodies against CDX2, STAG1 and STAG2 (top) or CDX2, STAG1 and RAD21 (bottom). Nuclei were stained with DAPI. BF, brightfield. Scale bar, 20 µm. Quantification of normalized fluorescence of STAG1 (C), STAG2 (D) and RAD21 (E) in CDX2-positive (TE) and CDX2-negative (ICM) nuclei. Each dot represents the mean normalized fluorescence from an individual embryo (see Methods). Three independent litters were analyzed for each set of stainings. P values were calculated with a paired non-parametric Wilcoxon test.

Depletion of maternal Stag1 mRNA does not affect preimplantation embryo development

Oocyte-stored gene products (maternal protein and RNA) sustain the embryo until zygotic transcription starts. For most genes, the maternal component is actively degraded upon ZGA, although in some cases it persists for longer periods. Our previously reported mouse model for Stag1 is a constitutive KO allele⁵. When crossing animals carrying this allele in heterozygosis, all resulting embryos are Maternal WT, as the oocytes harbor a wild-type copy of the gene (Stag1 WT), and 25% are expected to be Zygotic KO (Stag1 Z-KO) (Fig. 3A). At 2-cell stage, Stag1 WT and Z-KO embryos exhibit similar STAG1 staining, which confirms the storage of maternal Stag1 products (Fig. 3B, C) and is consistent with the detection of STAG1 protein (Supplementary Fig. 2A), and Stag1 mRNA and ribosome-protected fragments (Supplementary Fig. 1A, C) in oocytes. At the 4-cell stage, Stag1 Z-KO embryos still display STAG1 nuclear staining that is visually indistinguishable from that of Stag1 WT embryos, and clear differences between the two genotypes become evident only at the blastocyst stage (Fig. 3B, C). This result suggests that Stag1 maternal component persists beyond the 2-cell stage and that the gradual decrease in STAG1 levels described above (Fig. 1) may be due to gradual exhaustion of maternal Stag1 products and late activation of zygotic Stag1 transcription. Loss of STAG1 does not alter STAG2 or RAD21 protein levels at the blastocyst stage (Supplementary Fig. 2B–D).

Fig. 3. The maternal contribution of Stag1 gene products and their relevance for early development.

A Schematic representation of the cross used to obtain Stag1 WT or Stag1 Z-KO embryos. B Representative images, corresponding to the maximum projections of 6 (2-cell and 4-cell) or 3 (Blastocyst) confocal planes of the embryos obtained from the cross described in (A) at different stages, stained with STAG1 antibody and DAPI. Scale bar, 20 µm. C Quantification of STAG1 mean fluorescence from stainings like those shown in (B). Each dot is the mean fluorescence of all nuclei from an individual embryo. The line represents the mean of all embryos, and whiskers represent the standard deviation. Five litters were analyzed for the 2-cell and blastocyst stages, and six for the 4-cell stage. P values were calculated with an unpaired non-parametric Mann–Whitney test. D Representative brightfield images of embryos microinjected with ASOs at the zygote stage and cultured ex vivo. Images come from microinjection experiment #1. Scale bar, 75 µm. E Percentage of embryos found at blastocyst stage three days after ASO microinjection (not MI n = 16; Hprt n = 16; ASO1 n = 14; ASO2 n = 16). Data comes from a single experiment (#1). Replicates shown in Supplementary Fig. 2E, F (#2) and Supplementary Fig. 6C, D (#3). F Representative images of a single confocal plane of the different microinjection conditions performed, one day after the procedure, stained with STAG1. Nuclei were stained with DAPI. Images come from microinjection experiment #1. Scale bar, 20 µm. G Quantification of STAG1 mean fluorescence in 2-cell embryos after ASO microinjection from experiments #1 and #2. Each dot corresponds to an individual embryo. The line represents the mean, and whiskers represent the standard deviation. P values were calculated with an unpaired non-parametric Mann–Whitney test.

To address the functional relevance of Stag1 maternal component in early embryonic development, wild-type zygotes were microinjected with two different antisense oligonucleotides (ASO) against Stag1, or against Hprt as control⁴⁶. Zygotes were cultured ex vivo, and embryo progression was monitored every day until they reached the blastocyst stage. Embryos microinjected with Hprt ASO displayed slightly decreased viability in comparison to the non-microinjected group, indicating that embryo manipulation affected their progression (Fig. 3D, E). No differences in viability were observed between the zygotes microinjected with either Stag1 or Hprt ASO (Fig. 3D, E). Variability in embryo progression was observed between experiments; nonetheless, the lack of differences between the Hprt control and the two Stag1 ASOs was maintained (Supplementary Fig. 2E, F). Additionally, microinjection of two distinct concentrations of an siRNA pool targeting different regions of the Stag1 transcript confirmed that STAG1 loss does not alter embryo progression (Supplementary Fig. 2G, H). To evaluate the extent of protein depletion, a few embryos were fixed, immunostained and imaged one day after microinjection, at the 2-cell stage. A clear reduction in STAG1 levels was observed in embryos microinjected with Stag1 ASOs or siRNA in comparison to controls (Fig. 3F, G, Supplementary Fig. 2I, J). Taken together, these data suggest that maternal contribution of Stag1 consists mostly of mRNA and is dispensable for survival and progression throughout preimplantation development.

Maternal STAG2 is dispensable for embryonic viability

The Stag2 mouse model is a conditional KO. This allowed us to analyze the maternal contribution of Stag2 using the Zp3-Cre transgene, which drives Cre-recombinase expression during the oocyte growing phase, effectively preventing storage of maternal transcripts of target genes⁴⁷. As Stag2 is an X-linked gene, we crossed wild-type males with females carrying the Zp3-Cre transgene and the Stag2 floxed allele in heterozygosis (Fig. 4A) or in homozygosis (Fig. 4B). All embryos obtained in the former cross may carry maternal Stag2 gene products (Stag2 WT), and 25% of them are expected to be Zygotic KO (Stag2 Z-KO) (Fig. 4A). Embryos derived from Stag2^lox/lox females lack the maternal contribution (Stag2 M-KO) but half of them are both maternal and zygotic KO (Stag2 MZ-KO) (Fig. 4B).

Fig. 4. The maternal contribution of Stag2 gene products and their relevance for development.

A, B Schematic representation of the crosses used to obtain embryos of the indicated genotypes. C Representative images of a single confocal plane of 2-cell embryos obtained from these matings that were stained with STAG2 antibodies and DAPI. Scale bar, 20 µm. D Quantification of STAG2 mean fluorescence in 2-cell embryos. Each dot represents the mean fluorescence of all nuclei from an individual embryo. The line represents the mean of all embryos, and whiskers represent the standard deviation. Four litters were analyzed to obtain Stag2 WT or Z-KO and five for Stag2 M-KO or MZ-KO embryos. An unpaired non-parametric Mann–Whitney test was performed to test for significance; p values are indicated. E Representative images of a single confocal plane of Stag2 WT and Stag2 Z-KO embryos at the indicated stages, stained with STAG2 antibodies and DAPI. Scale bar, 20 µm. F Quantification of STAG2 mean fluorescence from (E). Each dot represents the mean fluorescence of all nuclei from an individual embryo. The line represents the mean of all embryos, and whiskers represent the standard deviation. Five litters were analyzed for the 4-cell and blastocyst stage, and six for the 8-cell and morula stage. An unpaired non-parametric Mann–Whitney test was performed to test for significance; p values are indicated. G Representative images of E9.5. embryos of the indicated genotypes. Scale bar, 500 µm. H Number of somites for the embryos obtained for (G). The Stag2 WT category comprises three different genotypes. Each dot corresponds to one embryo, and different litters are represented with different symbols. The line indicates the mean, and the whiskers indicate its standard deviation. P values were obtained with an unpaired non-parametric Mann–Whitney test. I Bar graph representing the average percentage of E9.5 embryos observed of each genotype (Obs.) in comparison to the expected (Exp.). Stag2 WT and Z-KO (6 litters) or Stag2 M-KO and MZ-KO (6 litters).

Immunofluorescent staining revealed, despite a high cytoplasmic background, that STAG2 is similarly detected in the nuclei of Stag2 WT and Stag2 Z-KO 2-cell embryos, and absent in those of Stag2 M-KO and Stag2 MZ-KO embryos (Fig. 4C, D). This result, confirmed with a second antibody (Supplementary Fig. 3A, B), indicates that there is a maternal contribution of Stag2 gene products. From the 4-cell stage onwards, STAG2 is detected in Stag2 WT embryos but is absent in Stag2 Z-KO embryos (compare with Stag2 WT littermates in Fig. 4E, F). Thus, zygotic Stag2 is readily expressed upon ZGA in the late 2-cell stage (Supplementary Fig. 1C) while maternal Stag2 gene products decline. Loss of STAG2 is at least partially compensated by the increase of STAG1, as changes in RAD21 levels in Stag2 Z-KO embryos are very mild (Supplementary Fig. 3C–E).

We next wondered if the maternal contribution of Stag2 is important for embryonic development. Previously, we reported that Stag2-null embryos die by E10.5. They are still found by E9.5 in the expected Mendelian ratios, although they are smaller and have fewer somites than their wild-type littermates, which implies a developmental delay⁶. Similar observations were made when comparing Stag2 WT and Stag2 Z-KO embryos at E9.5 (Fig. 4G–I). Importantly, Stag2 MZ-KO embryos are recovered at the expected Mendelian ratio at this stage and exhibit a delay similar to Stag2 Z-KO embryos (Fig. 4G–I). Stag2 M-KO embryos also have a smaller number of somites than Stag2 WT embryos by E9.5, but, unlike all other KOs, they survive past E10.5 and may be born. The delay observed in Stag2 M-KO embryos is not due to heterozygosity, as Stag2 WT embryos homozygous or heterozygous present a comparable number of somites (Fig. 4H). We also found no significant differences between Stag2 M-KO or MZ-KO embryos and Stag2 WT embryos by E3.5 (Supplementary Fig. 3F, G). These results indicate that the maternal contribution of Stag2 is dispensable for embryonic viability, but it contributes to proper embryonic development.

Cohesin-STAG2 is important for gene expression in the blastocyst

Simultaneous knockdown (KD) of both Stag1 and Stag2 by means of ASO microinjection in zygotes resulted in undetectable levels of STAG1, STAG2 and RAD21 after 3 days, and embryos failed to progress beyond the morula stage (Supplementary Fig. 4A, B, E, F). Co-injection of Stag1 siRNA and Stag2 ASO in zygotes, which reduced cohesin to low but detectable levels, resulted in a similar developmental arrest (Supplementary Fig. 4C, D, G, H). These findings suggest that a very small amount of cohesin can fulfill the essential cohesive function of the complex until the morula stage, while the presence of either STAG1 or STAG2 is required for further progression. Even though any cohesin variant appears to be sufficient for viability during preimplantation development, gene expression could be affected in these embryos. Considering the changes in the STAG1:STAG2 ratio during the first cell-fate decision (Fig. 2), we wondered if expression of key lineage-restricted factors was altered. Staining of Stag1 Z-KO or Stag2 Z-KO embryos with antibodies for CDX2 and NANOG showed proper allocation in the TE and ICM, respectively, as well as a similar percentage of double-positive cells in comparison to controls, indicating proper lineage segregation (Supplementary Fig. 5). While expression of these key factors seems undisturbed, we asked if changes in global gene expression could be observed. To reduce potential differences resulting from different genetic backgrounds, we compared the transcriptomes of wild-type blastocysts after STAG1 or STAG2 KD by ASO microinjection with those of embryos microinjected with the Hprt ASO or Buffer as a control. Immunofluorescent staining had confirmed efficient depletion of either protein at this stage and further showed that the decrease in total cohesin, as judged by RAD21 staining, is more apparent after STAG2 KD (Supplementary Fig. 6A, B). While most microinjected embryos were viable, there were some differences in their progression (Supplementary Fig. 6C, D). Therefore, blastocysts were selected by visual inspection to ensure that they were all in the same stage at the time of processing for low-input RNA-seq. Principal component analysis (PCA) does not separate well the different samples but suggests that depletion of STAG2 has the strongest effect on gene expression (Supplementary Fig. 7A). Indeed, we identified 963 differentially expressed genes (DEGs) in STAG2-deficient embryos, compared to control, but only 13 DEGs in STAG1-deficient embryos (Fig. 5A, B and Supplementary Data 1). Gene set enrichment analysis (GSEA) shows downregulation of pathways related to cell commitment and differentiation in STAG2-deficient blastocysts (Fig. 5C). Genes important at the blastocyst stage, both for the ICM and TE, are preferentially downregulated, while genes expressed in earlier stages, such as the zygote, 2-cell stage and minor ZGA, are upregulated after STAG2 KD (Fig. 5D, Supplementary Fig. 7B–D). Despite the much milder effect, STAG2 KD DEGs show a similar trend of deregulation after STAG1 KD and in CTCF-deficient blastocysts (Supplementary Fig. 7E, F). Finally, given the role of cohesin-STAG2 in superenhancer regulation in mESC, we took advantage of available capture Hi-C data to ask whether STAG2 KD DEGs were enriched for genes connected to superenhancers^19,48. This is indeed the case for downregulated genes in STAG2 KD blastocysts, while no enrichment is apparent for upregulated or downregulated genes in CTCF-deficient embryos (Fig. 5E). An example is Sox2 [log2FC = −2.29 STAG2 KD vs. control, FDR = 0.005; log2FC = 0.63 Ctcf MZ-KO vs. WT, FDR = 0.5³⁴] whose promoter interacts with a distal superenhancer bound by STAG2 (Fig. 5F). Taken together, these results suggest that cohesin-STAG2 is important for gene regulation in the blastocyst, to repress genes that are expressed in earlier stages and activate genes required for differentiation. While part of this function requires also CTCF and can be performed by cohesin-STAG1 as well, proper expression of superenhancer-regulated genes may be more specific to cohesin-STAG2.

Fig. 5. Transcriptional alterations in blastocysts lacking STAG2.

A, B Volcano plots of gene expression changes between controls (n = 9) and STAG1 KD (n = 10, two ASOs) (A) or STAG2 KD embryos (n = 9) (B). DEGs are colored in blue and pink for downregulated and upregulated genes, respectively (FDR < 0.05 and ∣log2FC∣ > 0). See also Supplementary Fig. 7A for PCA of individual embryos. C GSEA results for altered gene expression of STAG2 KD embryos (NES < 2, FDR < 0.05, gene set size < 50). D Enrichment score plots from GSEA of RNA-seq data of STAG2 KD embryos using gene sets generated from published transcriptomic data (Supplementary Data 2), as indicated in Methods. See also Supplementary Fig. 7B. E Percentage of expressed genes (categorized in “all,” “downregulated” or “upregulated”) in STAG2 KD embryos or Ctcf MZ-KO embryos at the blastocyst stage (GSE180307), whose promoters interact with superenhancers⁷² according to Capture Hi-C data in naïve mESC (GSE72164). Statistical significance was calculated with a Fisher’s exact test, and the resulting p values were corrected for multiple comparisons using the Bonferroni correction. F Snapshot from the browser showing tracks for STAG2 (GSE126659), H3K27ac (GSE23943) and p300 (GSE56098), as well as interactions detected in Promoter Capture Hi-C (GSE72164) around the Sox2 gene in naïve mESC.

Increased DNA damage in STAG2-deficient zygotes

Cohesin is required to repair DNA damage during the first cell cycle of the zygote^38,39. Replication stress and fork stalling have been observed in zygotes and proposed to be related to DNA demethylation^49,50. Since cohesin-STAG2 has been shown to be involved in replication fork progression and HR-mediated DNA repair, we asked whether its elimination results in increased spontaneous damage in the zygote during or after DNA replication^7,8. To answer this question, zygotes obtained from crosses of wild-type mice were compared to zygotes with Stag2 maternal deletion (Stag2 M-KO or MZ-KO) obtained from crosses of Stag2^lox/lox females carrying the Zp3-Cre transgene with wild-type males (as in Fig. 4B). Embryos were collected at different times after mating, cultured ex vivo and pulsed with EdU for 45 min before fixation (Fig. 6A). The combination of EdU and phosphohistone H3 (H3S10P) staining was used to classify embryos in S-phase (EdU positive/H3S10P negative) and G2-phase (EdU negative/H3S10P positive). Staining with an antibody against γH2AX allowed us to score DNA lesions generated in S phase (Fig. 6B, C) that persist in G2 (Fig. 6D, E). Consistent with previous results, we found that the paternal genome is more susceptible to these lesions and, more importantly, that lack of STAG2 increases their production and/or hinders their repair. Taken together with the observation that cohesin-STAG1 appears to be the dominant cohesin in the zygote stage (Fig. 1), this result supports the idea that cohesin composition, i.e., with STAG1 or STAG2, and not only overall cohesin amount, is important for the proper function of early embryos.

Fig. 6. Increased DNA damage in STAG2-deficient zygotes.

A Schematic illustration of the experimental design and duration of the first cell cycle after fertilization. B Representative maximum projection images from three confocal planes of S-phase zygotes stained with EdU and antibodies for H3S10P and γH2A.X and counterstained with DAPI. Stag2 WT embryos were obtained from crosses of wild-type mice, while Stag2 M or MZ-KO embryos were the offspring of crosses depicted in Fig. 4B. Scale bar, 20 µm. Maternal (♀) and paternal (♂) pronuclei are indicated. C Quantification of γH2A.X mean fluorescence from (B). Each dot represents the mean fluorescence of one pronucleus. The line represents the mean of all pronuclei, and whiskers represent the standard deviation. Two litters were analyzed for wildtype crosses and three for M/MZ-KO. P values were calculated with an unpaired non-parametric Mann–Whitney test. D, E As in (B) and (C), respectively. Three litters were analyzed for each condition.

Discussion

Here, we have addressed the roles of cohesin-STAG1 and cohesin-STAG2 in mouse preimplantation development. The abundance of STAG1 and STAG2 is distinctly regulated during this process, but the presence of either of the two cohesin variants appears to be sufficient for the embryo to progress beyond the blastocyst stage. Importantly, the lack of STAG2 has two major consequences: inefficient repair of DNA lesions that occur after the first round of replication in the zygote and altered gene expression in blastocysts. It is unlikely that the two phenotypes are directly linked, as genes related to DNA damage are not significantly deregulated in STAG2-deficient blastocysts.

STAG1 is clearly detected from the zygote to the 4-cell stage, diminishes at the 8-cell stage and recovers afterward. This, together with the detection of STAG1 protein in Stag1 Z-KO embryos beyond the 2-cell stage, suggests that the switch from maternal to zygotic transcription occurs between the morula and blastocyst stages for Stag1. The reason for this behavior remains unclear. Establishment of 3D chromatin organization in the early embryo is a gradual process, and TADs only become mature after the 8-cell stage^35–37. Despite the prominent role of cohesin-STAG1 in TAD boundary demarcation together with CTCF¹¹, STAG1 protein is barely detected in embryos at this stage. Moreover, elimination of most STAG1 from the zygote by means of ASO microinjection does not arrest development, and gene expression in STAG1-deficient blastocysts is not significantly altered. It is possible that 3D genome organization is not very relevant for gene expression regulation at this stage. Previous analyses in CTCF-deficient embryos show a strong effect on transcription that leads to embryonic lethality at the blastocyst stage, but which appears to be unrelated to genome architecture³⁴.

In the case of STAG2, maternally deposited products are present in zygotes and 2-cell embryos at low levels. A dramatic increase in abundance, likely driven by zygotic transcription, takes place at the 4-cell stage. Although it is not possible to estimate the amounts of cohesin-STAG1 and cohesin-STAG2 in early embryos from our staining results, comparison of the changes in abundance observed for STAG1, STAG2 and the common cohesin subunit RAD21 along preimplantation development allows us to make some educated guesses. STAG2 levels increase most dramatically in the transition from 2-cell to 4-cell embryos, coincidentally with a decrease in STAG1 and little change in RAD21. For RAD21, the increase is more notable in the blastocyst, when STAG1 reaccumulates after exhaustion of the maternal products and STAG2 levels continue to increase. In blastocysts, STAG2 KD impacts RAD21 abundance more than STAG1 KD, and a milder effect is similarly observed in Stag2 Z-KO blastocysts. We reckon that cohesin-STAG1 is the dominant cohesin variant in zygote and 2-cell embryos, while cohesin-STAG2 is more abundant from 4-cell onwards. Despite the seemingly complex regulation of STAG1 and STAG2 levels, strong reduction of STAG1 by ASO or siRNA microinjection, or genetic ablation of STAG2 from the oocyte stage onwards do not result in lethality or arrested preimplantation development. Of note, very low but detectable levels of STAG1 protein persist in the microinjected 2-cell embryos and are not further decreased when Stag1 siRNA concentration is doubled. As depletion efficiency is evidenced by the total absence of protein by the blastocyst stage, the small amount of STAG1 detected at the 2-cell stage must correspond to maternally deposited protein, not affected by transcript KD. While we cannot formally discard the possibility of an essential function of cohesin-STAG1 in these early embryos, we think this is unlikely. Double KD of STAG1 and STAG2 allows embryos to survive several rounds of cell division, until the morula stage, in the presence of cohesin amounts that are undetectable by immunostaining. Similar results have been reported after simultaneous KD of SMC3, RAD21 and NIPBL by siRNA⁵¹. Thus, a very low amount of either of the two cohesin variants provides sufficient sister chromatid cohesion to enable chromosome segregation and progression through the early stages of development. The role of cohesin in genome organization is probably not essential in these stages, but it seems to fine-tune gene expression in blastocysts. At the time of the first cell-fate decision in the early blastocyst stage, the committed TE cells have more STAG2 than the cells in the pluripotent ICM, while the opposite is true for STAG1 levels. These changes in abundance of STAG1 and STAG2 are reminiscent of those reported in the transition from naïve to primed mouse ES cells and may reflect the more prominent role of cohesin-STAG2 in tissue-specific transcription^19,44,52. Lack of cohesin-STAG2 does not prevent proper progression through the first cell-fate decision and proper lineage marker allocation (CDX2 in TE cells, Nanog in ICM cells). However, gene expression in the blastocysts is altered when cohesin-STAG2 is not present. On the one hand, at least a fraction of downregulated genes appears to be connected to superenhancers, consistent with a role for cohesin-STAG2 in these chromatin contacts¹⁹. On the other hand, genes important for early preimplantation development and ZGA become upregulated in STAG2 KD embryos. This is consistent with the finding that RAD21 depletion increases expression of minor-ZGA genes in mESC⁵³. Depletion of core cohesin subunits (RAD21, SMC1 and SMC3) or CTCF in mESC increases the fraction of 2-cell-like cells (2CLC) and expression of totipotency genes^54,55. In our hands, STAG1 KD in blastocysts did not result in significant upregulation of the 2CLC state, contrary to its effect in mESC⁴⁴. This discrepancy could be due to differences in STAG1 or STAG2 abundance between mESC and blastocysts, as well as changes in total cohesin levels after depletion. The existence of different isoforms of STAG1 with specific functions in early embryos, proposed by these authors, remains to be investigated, but at least one ASO used to deplete STAG1 is directed against the central region of the protein, present in all Stag1 mRNAs detected in mESC⁴⁴.

Germline mutations in STAG1 and STAG2 are associated with developmental syndromes that are considered cohesinopathies^20,21,23,24. Intellectual disability, a common symptom in cohesinopathy patients, has not been evaluated in STAG1 or STAG2-deficient mice. However, Stag2 deletion in embryonic brains results in premature death, growth retardation and neurological defects²⁵. Deletion of a single Smc3 allele also results in anxiety-related behavior, while anatomical defects in the brain are not apparent⁵⁶. Another of the most frequent symptoms in these patients is congenital heart defects. In mice, Stag2 deletion leads to embryonic lethality due to defective heart morphogenesis, and heart-specific Smc3 deletion results in perinatal lethality^6,57. While organogenesis occurs later during development, genes involved in cardiac and neural differentiation are already altered in STAG2 KD blastocysts. These changes are not sufficient to arrest embryo development, as Stag2 KO embryos survive until E10.5, but could contribute to defects that become lethal at later stages. Likewise, the increased DNA damage observed in STAG2-deficient zygotes could have negative consequences in the long term and contribute to the developmental delay scored by counting somites in E9.5 embryos with no maternal contribution of Stag2 (Fig. 4H). Moreover, spontaneous DNA damage and chromosome segregation errors caused by interference with DNA replication in early stages of development are more frequent in humans than in mice⁵⁰. Thus, germline mutations in STAG2 may be more detrimental for early development in human embryos.

Methods

Mouse housing, crossing and genotyping

We have complied with all relevant ethical regulations for animal use. Animal procedures were approved by our Institutional Animal Care and Use Committee (IAUC), the Ethics Committee for Research and Animal Welfare (CEIyBA) from CNIO-Instituto de Salud Carlos III and the regional government of Comunidad de Madrid (PROEX 337.8/21). Mice were housed in the pathogen–free Animal Facility of our institution. They were maintained within a 12-h light/dark cycle between 8:00 a.m. and 8:00 p.m. in a 22 °C ± 1 °C temperature-controlled room in individually ventilated units. Access to water and food was provided ad libitum. Euthanasia was performed using CO₂ inhalation in an automated gradual-fill chamber (displacement rate of 30% to 70% volume/minute) or by cervical dislocation performed by trained personnel. The following mouse lines were used in this work: Stag2^lox/lox6, Stag1^+/−5 and Zp3-Cre^tg/tg47. All mice were maintained in a predominantly C57BL/6 background. For embryo microinjection, experiments were carried out using hybrid F1 offspring females from mated C57BL6/J females with CBA males. Adult mice and embryos were genotyped by PCR using the primers described in Supplementary Table 1. The DNA of adult mice was extracted from ear tissue. DNA from preimplantation embryos was obtained from whole embryos after immunostaining and imaging, incubated with Lysis Buffer [50 mM Tris-HCl (pH 8.5), 1 mM EDTA, 0.5% Tween-20 and 50 µg/mL proteinase K] for 2 h at 55 °C.

Embryo collection and ex vivo culture

Preimplantation embryos were collected from pregnant females at different times, from the zygote (E0.5) to the blastocyst stage (E3.5). Zygotes were recovered from the ampulla, and cumulus cells were removed with hyaluronidase (Sigma, H3506). Embryos from the 2-cell to morula stage were recovered from the oviduct by flushing M2 medium (Sigma, M7167) using a 30-gauge needle. Blastocyst-stage embryos were collected from the uterus by flushing M2 medium using a 25-gauge needle. For cell cycle phase analysis (Fig. 6), collected zygotes were washed several times in M2 medium and placed in a drop of KSOM medium (Sigma, MR-101-D) with 10 µM EdU, covered with mineral oil (NO-100 NidOil) and incubated at 37 °C and 5% CO₂ for 45 min. For GV oocyte collection, females were injected intraperitoneally with 5 IU (100 µL) of pregnant mare serum gonadotropin (PMSG) and, 47 h after, oocytes were recovered by puncturing the ovary with a gauge in M2 medium with 10 µM 3-isobutyl-1- methylxanthine (IBMX). Post-implantation embryos were dissected in phosphate-buffered saline (PBS) at room temperature and imaged using a LEICA MZ10F microscope and LAS3.8 software.

ASO and siRNA microinjection of zygotes

Five-to-six-week-old hybrid females (C57BL/6.CBA.F1) were injected intraperitoneally between 1:00 and 3:00 p.m. with 5 IU (100 µL) of PMSG. After 46 to 48 h, mice were injected intraperitoneally with 5 IU (100 µL) of human chorionic gonadotropin (Merck, CG5) and were crossed with C57BL/6J males. Zygotes were recovered the next day as described above and then were microinjected in the cytoplasm with 1–2 pL from solutions of 20 μM ASO (Qiagen) for experiments #1-3, 40 μM total ASO for experiment #4, 25 µM or 50 µM siRNA (Dharmacon) for experiments #5-6 and a combination of 50 µM siRNA and 20 µM ASO for experiment #7 (Supplementary Table 1). All were diluted in microinjection buffer [10 mM Tris-HCl (pH 8.0) and 0.1 mM EDTA]. Microinjections were performed using a microscope equipped with differential interference contrast (Nomarski) optics, an electronic microinjector and micromanipulators. Additional controls for non-microinjected and buffer-microinjected zygotes were also performed. All embryos were then cultured in KSOM medium for up to 4 days.

Immunofluorescence

Preimplantation embryos were fixed for 10 min in 2% PFA-0.25% Triton X-100 in PBS, followed by 5 min in 4% PFA in PBS at room temperature. After further permeabilization with 0.5% Triton X-100 in PBS for 20 min, embryos were blocked for 1 h in 5% Normal Donkey Serum (Jackson ImmunoResearch, 017-000-121) in 0.1% Triton X-100 in PBS (blocking solution) at room temperature. Incubations with primary and secondary antibodies in blocking solution were carried out overnight at 4 °C and for 1 h at room temperature, respectively. The following antibodies were used: custom made rat monoclonal anti-STAG1¹¹; custom made rabbit polyclonal anti-STAG2¹⁸ at 2 µg/mL in all experiments except those in Supplementary Fig. 3A, which use rabbit polyclonal A302-580 (Bethyl) at 2 µg/mL; mouse monoclonal anti-RAD21 (Merck, 05-908) at 2.5 µg/mL; mouse monoclonal anti-CDX2 (Biogenex, MU392A-UC) at 1:200; rabbit monoclonal anti-CDX2 (abcam, ab76541) at 1:300; rat monoclonal anti-NANOG (eBioscience, 14-5761) at 1:300; H2A.X antibody (Merck, 05-636) at 1:200 and H3S10P antibody⁵⁸ at 1:10,000. Alexa Fluor 488-conjugated donkey anti-rat antibody (Invitrogen, A21208), Cy3-conjugated donkey anti-rabbit antibody (Jackson ImmunoResearch, 711-095-152) and Alexa Fluor 647-conjugated donkey anti-mouse antibody (Invitrogen, A32787) secondary antibodies were diluted 1:250 in blocking solution. Nuclei were visualized by incubating embryos in 1 µg/mL DAPI for 20 min. Single embryos were placed in an 18-well µ-Slide (Ibidi, 81826) and imaged using a Leica SP5 or Leica SP8 confocal microscope. To detect EdU (Fig. 6), Click-iT EdU reaction (Thermofisher, C10633) was performed after permeabilization. Embryos were then washed several times with 0.1% Triton X-100 in PBS, and blocking and antibody incubations proceeded as described above.

Image analysis and statistics

To quantify embryo staining, nuclear segmentation was performed on DAPI images using two pre-trained models, BlastoSPIM 1.0 for embryos from 2-cell to morula stages, and BlastoSPIM 2.0 for blastocysts⁵⁹. All masks generated with regions of interest were visually inspected and manually corrected. The mean intensity of every fluorophore was measured for each segmented nucleus. After correction, the mean intensity of all cells within the embryo was calculated and plotted. When cell-lineage markers were used, CDX2 or NANOG, individual nuclei were classified as positive or negative after setting a threshold for each experiment. In Fig. 2C–E, to compare intensities between ICM and TE, fluorescence levels of each nucleus were normalized to DAPI signal using linear regression. For each litter, a linear model was fitted to predict the intensity of each protein based on DAPI intensity. For normalization, the real fluorescence intensity was divided by its predicted value. In zygotes, low DAPI intensity precluded nuclear segmentation using pre-trained models. Thus, image analysis was performed using Fiji software (https://github.com/fiji/fiji), and mean intensity was measured within a defined nuclear area of each pronucleus. GraphPad Prism version 10.4.1 (GraphPad Software Inc) was used for all statistical analyses, with specific tests and parameters indicated in the figure legends.

Single embryo ultra-low input RNA-sequencing and analysis

Embryos microinjected with ASO and cultured ex vivo, as indicated above, were visually inspected and collected once they had reached the blastocyst stage. Single embryos were washed several times in PBS and placed in LoBind (Eppendorf) tubes containing 5 µL of Lysis buffer, snap-frozen in liquid nitrogen and stored at −80 °C until used. Samples (∼500 ng) were processed using the NEBNext Single Cell/Low Input RNA Library Prep kit (New England Biolabs, E6420). RNA Quality scores were 9.9 on average (range 9.1–10) when assayed on a PerkinElmer LabChip analyzer. Oligo(dT) primed reverse transcription with a template switching reaction was followed by cDNA production using limited-cycle PCR. Non-directional sequencing libraries were completed with the NEBNext Ultra II FS DNA Library Prep Kit for Illumina (New England Biolabs, E7508) and subsequently analyzed on an Illumina NextSeq 550 with v2.5 reagent kits following the manufacturer’s protocol.

RNA-sequencing data processing and analysis

For each sample, three FASTQ files were generated: paired-end reads (R1 and R3) and UMI barcodes (R2). Initial quality assessment of the FASTQ files was performed using FASTQC (version 0.11.9, https://www.bioinformatics.babraham.ac.uk/projects/fastqc/) and FastQ Screen (version 0.14.1, https://www.bioinformatics.babraham.ac.uk/projects/fastq_screen/). Subsequently, UMI sequences were appended to the read headers of the paired-end FASTQ files. Adapter trimming was conducted using Flexbar (version 3.5) with parameters specific to the NEBNext Single Cell/Low Input RNA Library Prep Kit⁶⁰. Reads were then aligned to the mouse genome (mm39) using STAR (version 2.7.11b)⁶¹. Gene assignment was performed using featureCounts (version 2.0.8) from the Subread package⁶². UMI-tools (version 1.1.6, https://umi-tools.readthedocs.io/en/latest/) was employed to quantify unique UMIs per gene. Differential gene expression analysis was carried out using DESeq2 with a false discovery rate FDR < 0.05⁶³.

Gene set enrichment analysis (GSEA)

GSEA (version 4.3.2) was used to identify statistically enriched gene sets⁶⁴. Enrichment was assessed using the Gene Ontology: Biological Process (GO:BP) as well as custom gene sets representing embryo developmental stages extracted from the original publications (Supplementary Data 2). For transcriptomic analyses of Ctcf-deficient embryos, raw data from Andreu et al.³⁴ were reanalyzed using DESeq2, and DEGs were selected based on an FDR < 0.05. For the single-cell RNA-sequencing data from Nowotschin et al.⁴⁵, the top 150 marker genes for embryonic stages 3.5 and 4.5 were selected. For the remaining datasets, stage-specific marker genes were identified using the following criteria: (1) Shannon entropy <2 (relative to all other stages), (2) mean fragments per kilobase of transcript per million mapped reads (mFPKM) > 3 in the stage of interest, and (3) no other stage exhibiting FPKM > 3^65–71. Finally, minor ZGA genes were defined as previously described by Zhang et al.⁷⁰: genes with FPKM > 3 in fertilized early 2-cell embryos and FPKM < 1 in both germinal vesicle oocytes and MII oocytes.

Statistics and reproducibility

For quantitative immunofluorescence, dots in graphs represent the mean fluorescence of all nuclei from an individual embryo and an unpaired non-parametric Mann–Whitney test was performed to test for significance. The only exception is Fig. 2, in which fluorescence staining of cohesin subunits is compared in two different populations within each embryo, and thus a paired non-parametric Wilcoxon test was applied instead. In all graphs, the line indicates the mean, and whiskers correspond to the standard deviation. To ensure reproducibility, observations were made in at least 5 embryos per genotype that had been derived from at least 3 different litters. Exceptions are: Supplementary Fig. 2 (oocyte staining), Supplementary Fig. 3B (testing a second STAG2 antibody in 2-cell embryos) and 3G (counting cell number in blastocysts). Regarding single ASO microinjections, three experiments are shown (#1, #2 and #3), but the variability in embryo progression among different replicates precludes plotting all experiments together. Data from experiment #1 are shown in Fig. 3D–G, while data from experiments #2 and #3 can be found in Supplementary Fig. 2D, E and Supplementary Fig. 6, respectively. Two experiments, #5 and #6, were performed using siRNA microinjection, and data are presented together in Supplementary Fig. 2G–J. A single experiment of Stag1/Stag2 depletion was performed for each of the approaches used: double ASO depletion (Supplementary Fig. 4A) from experiment #4 and combined ASO/siRNA (Supplementary Fig. 4C) from experiment #7.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

M.S.-F. designed, performed and analyzed all experiments. M.J.A. helped with experimental design and embryo stainings. D.G.-L. performed RNA-seq data and imaging analyses. J.M. and S.O. carried out embryo microinjection. A.L. obtained funding, supervised the study, and wrote the manuscript with M.S.-F. and contributions from all other authors.

Peer review

Peer review information

Communications Biology thanks Chih-Jen Lin and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editors: Eirini Trompouki and David Favero. A peer review file is available.

Funding

This research has been supported by grants PID2019-106499RB-I00 and PID2022-139333NB-I00 from the Agencia Estatal de Investigación (AEI/10.13039/501100011033), Ministerio de Ciencia, Innovación y Universidades and cofounded by the European Regional Development Fund (ERDF-EU). M.S.F. has been awarded predoctoral contracts from Comunidad de Madrid (PEJD-2019-PRE/BMD-15214) and AEI (PRE2020-091957). M.J.A. has been supported by a postdoctoral contract from “CNIO Friends.”

Data availability

The RNA-seq data generated for this study have been deposited in GEO, accession number GSE293021. All other datasets used are mentioned in the figure legends. Source data for all graphs can be found in Supplementary Data 3. Image data, mouse strains and custom-made antibodies are available upon reasonable request to A.L.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s42003-026-10665-3.