The methyltransferase Setdb1 is essential for meiosis and mitosis in mouse oocytes and early embryos

Mouse female germ cells enter meiosis during fetal development and arrest at the dictyate stage of meiotic prophase I shortly after birth until adulthood. During folliculogenesis, arrested oocytes grow in size and accumulate many transcripts and proteins, conferring the ability to progress through meiosis (meiotic competence) and to support early embryonic development (developmental competence) in absence of transcription (Hirao et al., 1993; Inoue et al., 2008; Wickramasinghe et al., 1991; Zuccotti et al., 2011, 2002).

In fully grown oocytes, also called germinal vesicle (GV) oocytes, chromatin undergoes an extensive remodeling, changing from a so-called non-surrounded nucleolus (NSN) to a surrounded nucleolus (SN) configuration, which normally coincides with a global shutdown of transcription (Bouniol-Baly et al., 1999; De La Fuente, 2006; Liu and Aoki, 2002; Moore et al., 1974). Meiotic arrest in GV oocytes is maintained by high levels of the ‘second messenger’ cyclic adenosine monophosphate (cAMP), which activates protein kinase A (PKA) to phosphorylate downstream-acting proteins, which in turn suppress activity of the so-called maturation promoting factor (MPF). MPF, consisting of Cdk1 and cyclin B, is the major driver of meiotic resumption (Dekel, 1996; Grieco et al., 1996; Han and Conti, 2006; Schultz et al., 1983). Meiotic resumption is triggered by a surge of the luteinizing hormone (LH), which activates the phosphodiesterase Pde3A to hydrolyze cAMP and to drive entry into meiotic maturation (Sun et al., 2009) as reflected, for example, by the breakdown of the nuclear envelope (also called GV breakdown, or GVBD). Alterations of this pathway often lead to infertility as exemplified by the arrest of oocytes at the GV stage in absence of Pde3a (Beall et al., 2010; Masciarelli et al., 2004). Following GVBD, oocytes condense their chromatin, align chromosomes on the metaphase I (MI) plate by a functional spindle, undergo chromosome segregation and extrude the first polar body (PB), to ultimately become arrested at metaphase II (MII). Spindle assembly in oocytes is driven by acentriolar centrosomes, called MTOCs (microtubule organization centers) that cluster together to form a bipolar spindle (Schuh and Ellenberg, 2007). Interactions between microtubules and kinetochores on chromosomes ensure proper chromosome segregation (Watanabe, 2012). Following fertilization, embryonic cleavage divisions share initially common mechanisms with meiosis such as a cAMP-PKA-dependent nuclear envelope breakdown (NEBD) and MTOC-driven bipolar spindle formation (Courtois et al., 2012; Cui et al., 2008; Poueymirou and Schultz, 1987; Yu et al., 2008). The organization of spindles gradually transitions over eight cleavage divisions from a large meiotic spindle into a short mitotic spindle harboring centrosomes (Courtois et al., 2012; Kubiak et al., 2008).

In mammals, several H3K9 methyltransferases suppress gene transcription and restrict differentiation potential during development (Nestorov et al., 2013; Shi et al., 2008). Setdb1 is able to mono-, di- and tri-methylate H3K9 (Loyola et al., 2009; Wang et al., 2003). Setdb1 is essential for pluripotency maintenance and repression of trophectodermal differentiation in embryonic stem cells (ESCs) (Lohmann et al., 2010; Yeap et al., 2009; Yuan et al., 2009), and for primordial germ cell (PGC) and neuronal progenitor development (Tan et al., 2012). Moreover, in these cells Setdb1 is required for H3K9me3 deposition at and transcriptional silencing of endogenous long terminal repeat (LTR)-containing retroviruses (ERVs) such as those belonging to the ERV1 (class I) and ERVK [class II; including intracisternal A-type particle (IAP) and early transposons (ETn/MusD)] repeat families. In contrast, ERVs belonging to class III families [ERVL elements (muERV-L/MERVL) and mammalian apparent LTR retrotransposons (ERVL-MaLR)] as well as non-LTR retrotransposons like long interspersed nuclear elements (LINEs) and short interspersed nuclear elements (SINEs) remained repressed (Karimi et al., 2011; Liu et al., 2014; Matsui et al., 2010; Tan et al., 2012). Setdb1-mediated repression of ERVs is required to prevent ERV-driven expression of nearby genes (Karimi et al., 2011; Tan et al., 2012).

In mouse oocytes and early pre-implantation embryos, however, many LTR and non-LTR retrotransposons are expressed. In oocytes, predominantly ERVL-MaLRs and, to a lesser extent, ERVK retrotransposons are transcribed. These elements also control gene expression through formation of chimearic transcripts (Fadloun et al., 2013; Peaston et al., 2004; Svoboda et al., 2004; Veselovska et al., 2015). In early embryos, ERVL, ERVK and ERV1 retrotransposons, as well as LINEs, are more highly expressed than ERVL-MaLRs (Fadloun et al., 2013; Ribet et al., 2008).

Setdb1 is essential for early development as mutant embryos die shortly after implantation (Dodge et al., 2004). It is highly expressed during oogenesis and is maternally provided to the embryo. Interestingly, Setdb1 is embryonically not transcribed prior to the blastocyst stage (Cho et al., 2011; Dodge et al., 2004). Maternal Setdb1 protein might therefore contribute to meiotic maturation and pre-implantation embryogenesis. Here, we address the function of Setdb1 for these processes by deleting Setdb1 in growing oocytes. We observed aberrant PKA signaling in mature Setdb1 mutant oocytes and a delay in meiotic resumption. In addition, kinetochore-spindle interactions, bipolar spindle organization and chromosome segregation are defective in Setdb1^−/− oocytes. These phenotypic defects relate to changes in gene expression measured in Setdb1^−/− oocytes. Maternal Setdb1 expression is also crucial for embryonic development as cell cycle progression and chromosome segregation are similarly impaired in maternally deficient embryos. We further measured upregulation of ERVL-MaLR and ERVK retrotransposons and an increase in frequency of splice junctions between these ERVs and gene exons in Setdb1^−/− oocytes. This study identifies Setdb1 as a maternal transcriptional co-regulator of genes implicated in cell cycle progression and chromosome segregation in oocytes and early embryos, and as a modulator of ERV repression.

To investigate the function of maternally provided Setdb1 in early embryogenesis, we deleted Setdb1 in growing oocytes and analyzed embryonic development of offspring. We generated animals that carried floxed alleles of Setdb1 (Setdb1^f/f) (Fig. S1A) and a transgenic allele of Cre recombinase that is specifically expressed in growing oocytes under the control of the Zona pellucida 3 promoter (Zp3-cre). By crossing Setdb1 mutant (Setdb1^f/f;Zp3-cre) females with Setdb1 floxed (Setdb1^f/f) control males we did not obtain any offspring, whereas crosses with Setdb1^f/f or heterozygous (Setdb1^f/+;Zp3-cre) control females produced offspring (Fig. 1A). To investigate when embryonic lethality occurred, we flushed embryos at day 3.5 of their development (E3.5) from the uteri of control and mutant females (Fig. 1B). Though many embryos from a Setdb1^f/f intercross had developed to the blastocyst stage, all embryos from Setdb1 mutant females and Setdb1^f/f males, being deficient for maternal (m−) but wild type for zygotic (z+) expression, died before the blastocyst stage (Fig. 1B; Fig. S1B).

To analyze in more detail the impairment of pre-implantation development we isolated zygotes and cultured them for 3 days in vitro. The development of Setdb1^m−z+ embryos was progressively delayed compared with that of control embryos (Fig. 1C,D). None of the Setdb1^m−z+ embryos developed into blastocysts, even when cultured for an additional day (Fig. S1C). Together, these results demonstrate that maternally provided Setdb1 is essential for pre-implantation development.

As Setdb1 deletion occurred during oogenesis, embryonic arrest of Setdb1^m−z+ embryos could reflect defects occurring during oogenesis. Histological analysis revealed that the ovarian structure and folliculogenesis were unaltered in Setdb1^f/−;Zp3-cre females (Fig. 1E,F). Moreover, the average number of fully grown GV oocytes per mouse was similar between control and Setdb1^f/−;Zp3-cre females (Fig. 1G) and the percentages of immature NSN and mature SN GV oocytes were comparable (Fig. S1D), arguing that the genome-wide chromatin reorganization and condensation towards the GV stage is unaffected in Setdb1^−/− oocytes. Together, these analyses did not reveal discernable differences in oocyte development in absence of Setdb1.

We next examined the levels of H3K9me1, H3K9me2 and H3K9me3 by immunofluorescence in GV oocytes (Fig. 1H,I). Setdb1 catalyzes mono-, di- and tri-methylation of H3K9 depending on interacting partners (Loyola et al., 2009; Wang et al., 2003). In Setdb1^−/− oocytes, we observed reduced H3K9me2 levels. In contrast, H3K9me1 and H3K9me3 levels were unaltered. Given that Setdb1 deficiency is induced in non-cycling oocytes, the differential effect on H3K9me1, H3K9me2 or H3K9me3 levels might reflect different degrees of modulation by histone demethylases, nucleosome turnover and/or compensatory re-methylation activities by other histone methyltransferases such as G9a (also known as Ehmt2), Glp1 (Ehmt1), Suv39h1 and Suv39h2. Indeed, we previously identified Suv39h2 as a maternal regulator of H3K9me3 at constitutive heterochromatin (Puschendorf et al., 2008). Despite the reduction in the repressive H3K9me2 mark, the characteristic loss of chromatin association of RNA polymerase II (representative of global transcriptional shutdown) and the typical nuclear remodeling occurred properly in GV oocytes deficient for Setdb1, suggesting normal chromatin maturation in absence of Setdb1 (Fig. S1E).

Meiotic maturation starts with GVBD of the oocyte, which is essential for female meiosis and prepares the oocyte for fertilization. We analyzed the efficiency of GVBD in control and Setdb1^−/− oocytes at different time points after removal of milrinone, an inhibitor of the phosphodiesterase Pde3A and GVBD (see Fig. 6B), and observed that Setdb1^−/− oocytes were delayed in undergoing GVBD. Moreover, more Setdb1^−/− oocytes remained at the GV stage compared with control oocytes (Fig. 2A). We then studied in detail the process of meiotic maturation in oocytes that had undergone GVBD within 2 h or between 2 and 18 h after milrinone removal (‘early’ versus ‘late’ GVBD). We first determined the capacity of oocytes to extrude the first polar body (PB). For controls, first PB extrusion efficiency was greatly reduced for those oocytes that had undergone their GVBD late, suggesting a reduced overall fitness. In contrast, we did not observe such a difference for early versus late GVBD Setdb1^−/− oocytes, arguing that the delay in GVBD in Setdb1^−/− oocytes does not necessarily prevent meiotic progression (Fig. 2B).

To assess the progression of oocytes from MI to MII, we stained meiotic spindles for alpha-tubulin and DNA with DAPI. We observed that meiotic progression was affected in 30% (24/79) of Setdb1^−/− oocytes (9/54 for early and 15/25 for late GVBDs), compared with only 1.3% (1/79) for control oocytes (Fig. 2C). Setdb1^−/− oocytes displayed various chromosome defects such as misaligned chromosomes, presence of multiple spindles, chromosome bridges in anaphase and, more surprisingly, chromosome decondensation and formation of pronuclei-like structures (Fig. 2D). Classification of all defects revealed that both metaphase stages were affected in Setdb1^−/− oocytes and that the defects were more pronounced in oocytes with delayed GVBD (Fig. 2C-E). In the early GVBD class of oocytes, MI and MII oocytes were equally affected, showing misaligned chromosomes or the presence of multiple spindles. In the late GVBD class of oocytes, the majority of abnormal oocytes displayed defects in the progression from MI to MII, at anaphase I, telophase I and cytokinesis. In addition, we observed the presence of pronuclei-like structures that form after the first meiotic division; an observation confirmed by live imaging (Fig. 2F; Movies 1 and 2). In Setdb1^−/− oocytes, extrusion of the first PB was followed by a phase of chromosome decondensation and recondensation, suggesting a short interphase-like period before re-entry in MII. We never observed this behavior in control oocytes.

Live imaging also revealed that misaligned chromosomes might even be a more common feature of Setdb1^−/− oocytes (Fig. 2G; Movies 1 and 2). Depending on the timing of fixation, chromosomes appear perfectly aligned on the metaphase plate, or not (Fig. 2G, compare T4 and T5, for instance). Interestingly, extrusion of the first PB in a Setdb1^−/− oocyte was followed by a meiotic spindle that moved erratically in the cytoplasm (Fig. 2G; Movies 1 and 2). In contrast, the spindle of the control oocyte was stably positioned at the periphery of the cell, near the extrusion site of the first PB, indicating stable anchorage to the cortex of the oocyte. The latter data suggest that Setdb1 might control asymmetrical positioning and anchoring of the meiotic spindle in oocytes. Together, these results demonstrate that Setdb1 expression during oocyte growth is required for proper meiotic maturation and various processes driving meiotic chromosome segregation.

To investigate in more detail whether the observed phenotypes were associated with defects in chromosome condensation and/or cohesion and would lead to actual chromosome mis-segregation, we generated chromosome preparations of MI and MII oocytes (Fig. 3A). We did not observe any defects in chromosome structure, nor in arm and centromeric cohesion at the first and secondary division of Setdb1^−/− oocytes, respectively. Nonetheless, 4/13 Setdb1^−/− MII oocytes were aneuploid, arguing for reasons other than defects in chromatin and cohesion dynamics underlying the chromosome segregation defects occurring in the first meiotic division.

Indeed, our previous analyses indicate that spindle organization is abnormal in Setdb1^−/− oocytes (Fig. 2D). In oocytes, multiple MTOC form and cluster together to organize a bipolar spindle (Schuh and Ellenberg, 2007). To study MTOC clustering in control and Setdb1^−/− oocytes, we used an antibody against pericentrin, which is a core component of MTOCs (Fig. 3B). We invariably detected pericentrin staining at two spindle poles in all control oocytes. However, 5/8 of Setdb1^−/− oocytes showed a defect in MTOC clustering, which might underlie the formation of multiple spindles described above.

We also tested the attachment of microtubules to kinetochores in Setdb1^−/− oocytes. In response to a short exposure to cold temperature, microtubules usually disassemble, with the exception of those spindle microtubules that are stably attached to kinetochores. We exposed control and Setdb1^−/− oocytes to cold and performed staining for alpha-tubulin (Fig. 3C). Cold treatment did not affect the spindle morphology in control oocytes, consistent with the presence of stable kinetochore-microtubule attachments. In contrast, 3/13 Setdb1^−/− oocytes displayed weak spindles with some chromosomes unattached. More strikingly, the spindle was undetectable in the remaining 10 Setdb1^−/− oocytes. This experiment suggests that spindle microtubules are not stably attached to kinetochores in absence of Setdb1.

Improperly attached kinetochores activate the spindle assembly checkpoint (SAC). To test whether the SAC was properly activated in Setdb1^−/− oocytes, we treated GV oocytes with nocodazole, a microtubule-depolymerizing agent. If the SAC was defective in Setdb1^−/− oocytes, they should be able to overcome the MI arrest induced by nocodazole. Both control and Setdb1^−/− oocytes were arrested in MI upon nocodazole treatment, suggesting that SAC is activated in absence of Setdb1 (data not shown). These results demonstrate the importance of Setdb1 expression during oogenesis for ensuring the formation of a bipolar spindle, stable kinetochore-microtubule attachment and accurate chromosome segregation during meiosis.

Given the known function of Setdb1 in gene regulation, we anticipated that its deficiency during oocyte growth would alter expression of genes controlling meiotic maturation and chromosome segregation. To test this hypothesis, we performed genome-wide expression profiling on mutant (Setdb1^f/−; Zp3-cre) oocytes and two different groups of control oocytes (Setdb1^f/− and Setdb1^f/+; Zp3-cre), enabling correction for transcriptional effects by Zp3-cre transgene expression during oogenesis (Fig. S2A). We found 480 and 372 genes to be up- and downregulated, respectively, upon disruption of the Setdb1 gene, with Setdb1 being one of the most significantly downregulated genes (Fig. 4A; Fig. S2B, Table S1).

Gene ontology (GO) analyses of mis-regulated genes revealed overrepresentation of several biological processes, including functions in cell cycle and cell division (Fig. S2C). Among upregulated genes, many have been implicated in cell cycle control (Fig. 4B,C). For example, Ube2c encodes a ubiquitin-conjugating enzyme known to regulate mitotic exit by promoting cyclin B degradation (Ben-Eliezer et al., 2015; Hao et al., 2012). Moreover, overexpression of the kinase Wee1 might delay cell cycle progression in oocytes and early embryos that is normally controlled by Wee1b and/or Wee2 (Han et al., 2005; Liu et al., 2013).

Other mis-regulated genes have been implicated in chromosome condensation, kinetochore-spindle attachment, heterochromatin structure, organization of a bipolar spindle, regulation of spindle orientation and cytokinesis (Fig. 4B,C). Their mis-regulation might contribute to the phenotypes observed in Setdb1^−/− oocytes. For instance, Kntc1 (kinetochore associated 1) is a component of the evolutionarily conserved Rod-Zwilch-Zw10 (RZZ) complex required for a stable kinetochore-spindle attachment, mutations of which are associated with chromosome segregation defects (Karess, 2005). Ckap2 (cytoskeleton associated protein 2) is required for the maintenance of microtubule nucleation sites and its depletion leads to multipolar mitosis, which could explain the presence of the multi-spindle and MTOC-clustering defects observed in Setdb1^−/− oocytes (Case et al., 2013). Importantly, as Setdb1 is known as a transcriptional repressor, the downregulation of genes suggests that Setdb1 might control meiotic maturation by orchestrating a network of gene interactions during oogenesis.

Setdb1 controls transcriptional repression of ERVK and ERV1 retrotransposons in ESCs and PGCs (Karimi et al., 2011; Liu et al., 2014). To assess a possible role of altered expression of repetitive elements in meiotic maturation and division, we analyzed their expression in Setdb1^−/− and control GV oocytes according to RefSeq annotation. Setdb1 deficiency during oocyte growth caused a 1.5-fold upregulation of repeats, with 18% of all reads mapping to repetitive elements in Setdb1^−/− oocytes, compared with 12% in controls (data not shown). Primarily LTR retrotransposons belonging to ERVL-MaLR and ERVK families were upregulated, in addition to some ERV1, ERVL, SINE B2 and GSAT_MM satellite sequences (FDR<0.05) (Fig. 5A,B; Tables S2 and S3).

In wild-type oocytes, ERVL-MaLRs contribute most highly to overall retrotransposon expression (Fig. 5B), as reported previously (Fadloun et al., 2013). Intriguingly, Setdb1 deficiency enables further upregulation of the already highly expressed MTA, MTB and MLT elements, as well as of more moderately expressed ORR1 elements. Likewise, expression of 46 families of ERVKs was upregulated upon Setdb1 depletion, including ETn, ERVB7 (MusD), IAPd, IAPEz, MMERVK10C and RLTR45, which are silenced in a Setdb1-dependent manner in ESC (Fig. 5C; Table S3) (Karimi et al., 2011). Thus, in contrast to ESCs, these data identify Setdb1 as a negative modulator of ERVK and ERVL-MaLR expression rather than a strong repressor of ERV transcription during oogenesis.

We next wondered whether induction of ERV transcription by Setdb1 depletion could promote transcription of nearby genes. Towards this, we identified and quantified sequence reads that consisted of the 5′ end of ERV and 3′ end of gene exon sequences, hereafter referred to as splice junctions. We classified such splice junctions into three groups, ‘upstream’, ‘within’ and ‘downstream’, according to the location of the ERV relative to the exon of an associated open reading frame (ORF). We also quantified splice junctions between ERVs and non-exonic sequences. We determined the levels of altered expression at such splice junctions upon Setdb1 deletion (Fig. 5D; Table S4). For ERVL-MaLR, over 48% of splice junctions involving an ERV occurred in the context of an annotated ORF, whereas for ERVK this was 24% (Fig. 5D). These data support a role of ERVL-MaLR and ERVK in driving gene expression in oocytes.

For both ERV classes, over 5% of splice junctions with the ERV localized upstream of the annotated exon were upregulated in Setdb1^−/− oocytes, indicating that Setdb1 suppresses ERV-driven gene transcription (Fig. 5D,E). In addition, ∼11% of splice junctions originating from ERVs within an ORF were upregulated in Setdb1 mutant oocytes, likely representing aberrant non-functional chimearic transcripts (Fig. 5D). Among the upregulated genes involved in cell cycle regulation and cell division, we measured for E2f7 (Fig. S3A) and Tpd52l1 increased expression of splice junctions, but only for those initiated from 3′ localized ERVs, linked to downstream exons.

To further enhance the likelihood of identifying ERV-induced gene expression impairing meiosis and mitosis, we related our RNA sequencing data to a recent annotation of the mouse oocyte transcriptome characterized by prominent ERV-driven gene expression (Veselovska et al., 2015). Indeed, we measured an increase of over tenfold in expression of ERV-induced transcripts in control and Setdb1 mutant oocytes, compared with RefSeq annotation (Fig. S3B). These analyses confirmed expression of ERV-initiated chimearic transcripts for E2f7 and Tpd52l1 and identified chimearic transcripts for 1700017B05Rik with unknown functions (Table S5). Finally, we did not observe increased levels of phosphorylated H2A.X, a marker of DNA damage, in Setdb1 mutant oocytes (Fig. S4A,B).

In summary, we conclude that Setdb1 restricts but does not prevent ERVK- and ERVL-MaLR-driven gene expression. We anticipate that the contribution of such expression to the meiotic maturation and division phenotype is limited as most relevant splice junction events initiated from ERVs are within genes, thereby limiting the potential for the generation of functional proteins.

Besides its role in meiotic maturation, Setdb1 regulated GVBD and thus the exit from meiotic arrest (Fig. 2A). Interestingly, using Ingenuity Pathway Analysis, we identified PKA signaling as a mis-regulated pathway upon Setdb1 deletion. Twenty-two genes belonging to the PKA signaling pathway were mis-regulated in Setdb1^−/− oocytes, including Prkaca, which encodes one catalytic subunit of PKA (Fig. 6A). It is well-established that PKA maintains meiotic arrest in response to high levels of cAMP (Fig. 6B). To test whether the PKA pathway is actually involved in the GVBD delay observed in Setdb1^−/− oocytes, we inhibited PKA using the cAMP antagonist 8-bromo-Rp-cAMP (Rp-cAMP). We first observed that Rp-cAMP overcame the milrinone-induced GV arrest in control and Setdb1^−/− oocytes, demonstrating activity of the compound (Fig. S5). After removal of milrinone, Rp-cAMP treatment did not further increase the high efficiency of GVBD in control oocytes (Fig. 6C). In Setdb1^−/− oocytes, however, we observed a complete alleviation of the delay in GVBD (Fig. 6C). These results suggest that GVBD delay observed in absence of Setdb1 is due to an impairment of the PKA signaling pathway.

We then wondered whether the maternal deficiency of Setdb1 would cause similar defects in early embryos as observed in meiosis. We analyzed the first embryonic cleavage by isolating Setdb1^m+z+ and Setdb1^m−z+ zygotes at E0.5 and culturing them in vitro for 24 h. Whereas all control zygotes developed into 2-cell embryos, only 25% (3/12) of Setdb1^m−z+ zygotes reached the 2-cell stage (Fig. 7A,B). Moreover, these 2-cell embryos were abnormal and contained micronuclei, suggesting defects in chromosome segregation. Interestingly, several Setdb1^m−z+ arrested zygotes contained pronuclei with aberrant morphologies, indicative of a failure in NEBD. By live imaging of Setdb1^m−z+ zygotes we saw that although the two parental pronuclei move into close proximity of each other, their nuclear envelopes never break down and the embryos arrest at the G2/M transition (Fig. 7C; Movies 3 and 4). In summary, we observed comparable defects in the G2/M transition and subsequent progression through meiosis and during the first cleavage division in response to Setdb1 maternal deficiency. It is likely that similar defects contribute to the arrest of Setdb1^m−z+ embryos observed at subsequent embryonic stages (Fig. 1).

Meiotic and embryonic developmental competencies rely on a proper maturation of the oocyte, which includes synthesis and accumulation of necessary transcripts and proteins. This study reveals that the lysine methyltransferase Setdb1 participates in the establishment of meiotic and embryonic developmental competencies. Transcriptional profiling shows that during oogenesis Setdb1 regulates the expression of a substantial number of genes with putative functions in meiotic maturation and early embryogenesis.

We show that the transition from prophase to metaphase, marked by GVBD, is delayed in Setdb1^−/− oocytes. Meiotic arrest in prophase is normally maintained by high levels of cAMP that stimulate PKA activity, which in turn suppresses the MPF activity. Meiotic resumption is triggered by cAMP hydrolysis resulting from increased PDE3A activity in response to LH-mediated reduction in paracrine cGMP signaling from the granulosa cells (Norris et al., 2009; Vaccari et al., 2009) (Fig. 6B). In Setdb1^−/− oocytes, Prkaca is upregulated and we demonstrate that inhibition of PKA by the cAMP antagonist Rp-cAMP fully rescues the GVBD delay, showing that Setdb1 controls GVBD by regulating the PKA pathway. Ingenuity Pathway Analysis of expression profiling data furthermore reveals that several other components of the PKA pathway are mis-expressed. For example, Prkcb (protein kinase C beta) is strongly downregulated in Setdb1^−/− oocytes. During mitosis, Prkcb mediates phosphorylation-dependent lamin B1 disassembly, and inhibition or depletion of Prkcb delays nuclear envelope breakdown (Mall et al., 2012). Possibly, reduced Prkcb expression might contribute to the delayed GVBD observed in Setdb1^−/− oocytes.

We frequently saw transient formation of pronuclei-like structures during the MI-to-MII transition of those Setdb1^−/− oocytes that displayed a delayed onset of GVBD. Interestingly, whereas completion of meiosis I requires a reduction in MPF activity, its re-establishment is necessary to maintain chromosome condensation in MII oocytes (Brunet and Maro, 2005). Inactivation of MPF in MII oocytes was shown to induce formation of pronuclei-like structures (Madgwick et al., 2006). Thus, the appearance of a short interphase between MI and MII might result from a sustained reduction of MPF activity upon completion of meiosis. Besides PKA signaling, MPF activity is controlled by ubiquitin-mediated degradation of cyclin B by the anaphase promoting complex/cyclosome (APC/C). Ube2c, encoding the ubiquitin-conjugating enzyme and partner of APC/C known to promote cyclin B degradation, is upregulated in Setdb1^−/− oocytes (Fig. 4B) (Xie et al., 2014). MPF activity might thus be impaired in Setdb1^−/− oocytes partly as a result of an increased degradation of cyclin B by Ube2c.

We observe that in the absence of Setdb1, the formation of meiotic spindle is compromised by defective MTOC clustering. Moreover, unstable kinetochore-microtubule attachments likely drive the chromosome segregation defects observed in Setdb1^−/− MII oocytes. In some mutant oocytes, the meiotic spindle moves dramatically throughout the cell, which likely results from a defect in anchorage of the spindle to the cortex, a process that normally ensures a proper asymmetric division (Azoury et al., 2008; Chaigne et al., 2012; Schuh and Ellenberg, 2008). We have shown that these later phenotypes correlate with changes in gene expression e.g. of Kntc1 and Ckap2. Intriguingly, several of the cell cycle and division genes are downregulated in the absence of Setdb1. Since Setdb1 is known to function as a transcriptional repressor, a secondary mechanism might be involved in which depletion of Setdb1 would induce the upregulation of transcriptional repressors.

Setdb1^−/− MII oocytes can be successfully fertilized; nevertheless, maternally deficient embryos arrest early during pre-implantation development, substantially prior to the lethality of Setdb1 zygotically null embryos occurring after implantation (Dodge et al., 2004). Thus, maternal Setdb1 is crucial for embryonic development. Indeed, similar to mutant oocytes, NEBD at the G2/M transition and chromosome segregation are severely impaired in Setdb1^m−z+ embryos. The presence of one large pronucleus in arrested Setdb1^m−z+ zygotes indicates that NEBD does ultimately occur, yet is followed by syngamy without chromosome segregation. Some Setdb1^m−z+ embryos went through the first cleavage division, but these embryos displayed chromosome segregation defects. Given that the meiosis-to-mitosis transition is mechanistically a gradual process lasting over many cleavage divisions (Courtois et al., 2012; Cui et al., 2008; Poueymirou and Schultz, 1987; Yu et al., 2008), we propose that the observed defects in G2/M transition and mitosis underlie the arrest of Setdb1^m−z+ embryos at successive stages of pre-implantation development.

Besides the transcriptional regulatory function during oogenesis, we cannot exclude a post-transcriptional role of Setdb1 in meiosis by modifying proteins directly (Fei et al., 2015; Kaustov et al., 2011; Van Duyne et al., 2008). Identifying such Setdb1 targets in oocytes, however, is a challenging prospect.

Finally, we identify Setdb1 as a modulator of ERVL-MaLR and ERVK expression during oogenesis. Setdb1-mediated H3K9me3 and DNA methylation have been implicated in the silencing of LTR elements in undifferentiated and differentiated cells, respectively (Hutnick et al., 2010; Matsui et al., 2010; Walsh et al., 1998). Setdb1 controls H3K9me3 deposition at and silencing of several class I ERV1 and class II ERVK retrotransposons in ESCs, PGCs and in fetal brain (Karimi et al., 2011; Liu et al., 2014; Tan et al., 2012). In contrast, ERVL-MaLR elements lack H3K9me3 occupancy in wild-type ESCs and PGCs, and fail to be upregulated upon Setdb1 deletion (Karimi et al., 2011; Liu et al., 2014; Matsui et al., 2010; Tan et al., 2012). During oogenesis, however, many ERVK elements that are typically silenced by Setdb1 in ESCs and PGCs are already expressed. Analysis of splice junctions involving individual elements shows that loss of Setdb1 further enhances the expression of already active elements while also enabling derepression of other ERVK elements. The mechanistic details underlying such cell type- and ERVK element- specific modulatory roles of Setdb1 in catalyzing H3K9me2 and/or H3K9me3 and in controlling repression are, however, not understood. We cannot exclude possible heterogeneity in H3K9 methylation levels at different elements leading potentially to ‘variegated’ expression levels of individual elements in different oocytes. In addition or alternatively, possible variegated or lack of ERV repression might relate to the rather low to moderate expression of the two KRAB-containing zinc finger proteins Zfp932 and Gm15446 in wild-type GV oocytes. These KRAB-ZFPs have recently been shown to control repression of various ERVK elements, known to also be silenced by Kap1 (also known as Trim28 or Tif1beta) and Setdb1 (Ecco et al., 2016).

We also identified Setdb1 as a negative regulator of ERVL-MaLR expression; decreasing, yet not silencing, the expression of such elements in oocytes. It remains to be determined whether its catalytic activity is required for such suppression. ERVL-MaLR and ERVK elements have been suggested to drive gene expression in oocytes and early mouse embryos (Peaston et al., 2004; Veselovska et al., 2015). For 55 out of 80 chimearic transcripts identified by Peaston and colleagues (2004) to be expressed in oocytes, we confirmed the presence of splice junctions in our expression data set, a few of which were upregulated upon Setdb1 deficiency (Table S1). Our findings thus characterize Setdb1 as a potential modulator of ERV-induced gene expression at the oocyte-to-embryo transition. Nonetheless, we did not obtain compelling evidence for a role of impaired ERV repression in driving gene expression associated with aberrant meiotic phenotypes. In early embryos, knockdown of Setdb1 increased expression of retrotransposons LINE1, SINE-B2 and IAP (Hatanaka et al., 2015). Nevertheless, a role for Setdb1 in controlling repeat-driven gene expression and development of early embryos remains to be shown.

In summary, our study reveals an essential function for the lysine methyltransferase Setdb1 in cell division. The data is in line with a transcriptional regulatory function for Setdb1 during oocyte growth that subsequently impacts on meiosis and mitosis in mouse oocytes and early embryos. Interestingly, Setdb1 deficiency in mouse neuronal progenitors was also associated with mis-regulation of genes involved in M-phase, suggesting that Setdb1 regulates expression of genes involved both in meiosis and mitosis, and even in somatic cells (Tan et al., 2012).

Mice maternally deficient for Setdb1 were generated by crossing Setdb1^f/f mice (Lohmann et al., 2010) with Zp3-cre transgenic mice to induce deletion in growing oocytes. Mice were maintained on a C57Bl/6J genetic background. All experiments were performed according to Swiss animal protection laws and institutional guidelines.

Zygotes and blastocysts were harvested from superovulated females mated to control males at 18 h and 90 h after hCG injection, respectively. We cultured embryos in vitro using standard conditions as detailed in the supplementary Materials and Methods. We performed piezo-driven intra-cytoplasmic sperm injection as described previously (Yoshida and Perry, 2007).

GV oocytes were isolated in M2 medium containing 2.5 mM milrinone (Sigma). After milrinone washout, oocytes were cultivated in M16 (Sigma) at 37°C in controlled atmosphere. When necessary, 5 mM Rp-cAMP (Sigma B2432) and 0.04 µg/ml nocodazole (Sigma M1404) were used. For the cold treatment experiment, the zona pellucida was removed with acidic tyrode and MII oocytes cultivated for 2 h before being incubated on ice for 20 min and fixed.

Immunofluorescence staining, preparation of chromosome spreads and histology were performed as described previously (Peters et al., 2001), with modifications as described below.

Oocytes and embryos were fixed in 4% paraformaldehyde, washed in PBT (PBS, 0.1% Tween 20), permeabilized (PBS, 0.5% Triton X-100), washed in washing solution (PBS, 2% BSA, 0.1% Tween 20), blocked (blocking solution: PBS, 2% BSA, 0.1% Tween 20, 5% goat serum) and incubated with primary antibody overnight at 4°C. After further washing in washing solution, oocytes and embryos were incubated for 1 h with secondary antibody, washed again in PBT and counterstained in Vectashield with DAPI. Antibodies used: rabbit anti-H3K9me1, H3K9me2, H3K9me3 (1:500; Peters et al., 2003), rabbit anti-H3 (1:800; Abcam, ab1791), mouse anti-nucleosome (1:1000; van der Heijden et al., 2007), mouse anti-γ-H2AX (1:1000; Millipore, 05636), mouse anti-alpha-tubulin (1:400; Sigma, T9026), human anti-crest (1:1000; Fitzgerald, 90C-C51058), mouse anti-pericentrin (1:200; BD Transduction Laboratories, 611814), mouse anti-RNA polymerase II (1:30; Covance, MMS 126R). Secondary antibodies (ThermoFisher; 1:500 unless stated otherwise): goat anti-rabbit Alexa Fluor 488 (A11034), goat anti-mouse Alexa Fluor 568 (A11031), donkey anti-rabbit Alexa Fluor 568 (A10042), donkey anti-mouse Alexa Fluor 488 (A21202), goat anti-human Alexa Fluor 633 (1:150; A21091). Acquisition of images was done with a Zeiss LSM700 confocal microscope. Intensity quantification and chromosome tracking were done using Ilastik (Interactive Learning and Segmentation Toolkit; Sommer et al., 2011) and MetaMorph (Molecular Devices) software. Intensities of H3K9me and γ-H2AX stainings were normalized to intensity of nucleosome staining.

Chromosome spreads were generated from MI and MII oocytes respectively isolated 8 h after milrinone removal and 18 h after hCG injection. Oocytes were incubated for 6 min in hypo-buffer (0.5% Na citrate, 15% FCS), 3 s in ice-cold 5-1-3 fixative (5 parts methanol, 1 part acid acetic, 3 parts dH₂O) and transferred onto a slide. A couple of drops of ice-cold 4-4-2 fixative (4 parts methanol, 4 parts acid acetic, 2 parts dH₂O) followed by a drop of 3-1 fixative (3 parts methanol, 1 part acid acetic) were carefully added to the oocyte. The slides were dried overnight and chromosomes stained with DAPI.

Ovaries were fixed in 4% paraformaldehyde overnight at 4°C and embedded in paraffin. Sectioning was done with an automatic microtome (Microm HM355S) and Hematoxylin and Eosin staining was performed according to standard procedures.

H2B-mCherry and tubulin-EGFP mRNA were transcribed in vitro (mMessage mMachine SP6/T7, Ambion) and purified (RNeasy Mini Kit, Qiagen). Cytoplasmic microinjection of 3-5 pl of mRNA (20 ng/µl H2B-mCherry and 100 ng/µl tubulin-EGFP mRNA) was performed on oocytes under a microscope (IX71, Olympus) equipped with a micromanipulator and a Femtojet (Eppendorf). For live imaging oocytes and embryos were placed in a micro-slide dish (Ibidi) with 50 µl of M16 covered by mineral oil. Time-lapse images were acquired with a spinning-disc confocal microscope (Olympus, 20× objective), at 37°C with 5% CO₂. Oocytes and embryos were imaged by multipoint acquisition function every 15 min. z-stacks were acquired with an interval of 3 µm.

We performed expression profiling on pools of 16 denuded GV oocytes isolated per mouse. We used oocytes from four Setdb1^f/+; Zp3-cre mice and two Setdb1^f/− mice as controls and oocytes from four Setdb1^f/−; Zp3-cre mice as mutant. RNA isolation and sequencing (50 cycles; single-end reads) were performed to standard procedures (see supplementary Materials and Methods). As a basis for all analyses we used the M. musculus genome assembly (GRCm38/mm10 December 2011), RepeatMasker repeat annotation (downloaded on 7 March 2012), RefSeq gene models (downloaded from UCSC on 4 February 2016), and the oocyte transcriptome annotation (downloaded on 29 February 2016 from the web page of the paper Veselovska et al., 2015). We performed differential gene expression and splice junction analyses according to standard operations further detailed in the supplementary Materials and Methods. Raw (fastq) and processed (bigWig) RNA sequencing data are available at GEO (GSE82002).

While this manuscript was under review, Kim and colleagues (Kim et al., 2016) reported largely comparable findings on the role of maternal Setdb1 for meiosis and embryogenesis. In addition, they observed increased mRNA and protein expression for the Cdc14b phosphatase, a known negative regulator of meiotic maturation (Schindler and Schultz, 2009). Using a siRNA-mediated knockdown approach, Kim et al. (2016) characterized upregulated Cdc14b expression as the major driver of impaired meiotic progression in Setdb1 mutant oocytes. In our study, however, Cdc14b was not differentially expressed between Setdb1-deficient and control GV oocytes as determined by RNA sequencing analyses. This finding is surprising as both laboratories studied the same conditional deletion allele of Setdb1 (Lohmann et al., 2010), yet on slightly different genetic backgrounds (C57Bl/6-129Sv hybrid versus C57Bl/6J). Moreover, in contrast to Kim et al. (2016), we did not perform hormonal superovulation in any of the experiments dealing with oocyte maturation. Future experiments will be required to resolve this discrepancy.

We thank T. Chen (currently at the University of Texas M.D. Anderson Cancer Center, Smithville, USA) and B. Knowles (the Jackson Laboratory, Bar Harbor, USA) for providing the Setdb1^F/F and Zp3-cre mice, respectively. We thank T. Roloff and S. Thiry for preparation and sequencing of RNA libraries, R. Thierry for image processing and the animal facility. We thank M. Gill and M. Tardat for critical reading of the manuscript.