Setd1b, encoding a histone 3 lysine 4 methyltransferase, is a maternal effect gene required for the oogenic gene expression program

Resetting the epigenome during germ cell development establishes the tabula rasa essential for early development. In the mouse, primordial germ cells originate from the base of the allantois and migrate through the hindgut into the genital ridge where they proliferate and form the gonads (Allegrucci et al., 2005). By the time primordial germ cells have migrated to the genital ridge, they appear to be largely devoid of DNA methylation (Reik, 2007). In the female gonad, primary oocytes are arrested in prophase of meiosis I at birth (Edson et al., 2009). Subsequently they enlarge, accumulate transcripts needed for early embryonic development and undergo notable changes in chromatin configuration accompanied by epigenetic changes including alterations of lysine methylation on the histone 3 tail and re-establishment of DNA methylation (Stewart et al., 2015; Dahl et al., 2016; Liu et al., 2016; Zhang et al., 2016).

The histone 3 (H3) tail serves a central role in epigenetic regulation primarily through alterations of lysine methylation (Zhang et al., 2015). Trimethylation of lysine 4 (H3K4me3) relates to trithorax-Group (trx-G) action (Roguev et al., 2001) and is a hallmark of active promoters (Santos-Rosa et al., 2002; Bannister and Kouzarides, 2011). H3K4me3 is opposed by repression through H3K27me3 mediated by Polycomb-Group (Pc-G) action (Voigt et al., 2012) and H3K9me2/3, which defines heterochromatin (Peters et al., 2001), whereas H3K36me3 is mainly found in transcribed regions due to the association of the major H3K36 methyltransferase with the elongating RNA Polymerase II through phosphorylation of serine 2 of its CTD tail (Li et al., 2003; Schaft et al., 2003; Dambacher et al., 2010). Recent data on growing oocytes indicates that local deposition of H3K36me3 and removal of H3K4me2/me3 sets the stage for de novo DNA methylation (Stewart et al., 2015). Development of the preovulatory follicle is finalized by global transcriptional silencing in the oocyte, which requires the H3K4 demethylases Kdm1a (Lsd1) and Kdm1b (Lsd2) (Stewart et al., 2015) and, paradoxically, also the H3K4 methyltransferase Mll2 (Kmt2b, Wbp7) (Andreu-Vieyra et al., 2010).

Shortly after fertilization, the male and female pronuclei have comparable levels of H3K4me1; however, H3K4me2/me3 is initially lacking in the male pronucleus (Lepikhov and Walter, 2004; van der Heijden et al., 2005). The acquisition of H3K4me2/me3 in the male pronucleus, pronuclear fusion and cell division precede zygotic genome activation (Aoshima et al., 2015), although in the mouse minor zygotic transcription activity has been detected at the pronuclear stage (Abe et al., 2015).

To elucidate the roles played by epigenetic mechanisms in oocyte and early embryo development, we are focusing on the six mammalian Set1/Trithorax-type H3K4 methyltransferases (Glaser et al., 2006). These six come in three pairs: Setd1a and Setd1b, which are orthologous to the sole yeast H3K4 methyltransferase Set1 (Roguev et al., 2001); Mll1 (Kmt2a) and Mll2, which are homologous to Drosophila Trithorax (Francis and Kingston, 2001), and Mll3 (Kmt2c) and Mll4 (Kmt2d), which are homologous to a fusion of Drosophila Cara Mitad (Lpt) and trithorax-related (Chauhan et al., 2012; Herz et al., 2012). Each of the six appears to reside in its own protein complex. Concordantly, all six are essential for mouse development for different reasons even though all six appear to be expressed ubiquitously (Ernst et al., 2004; Glaser et al., 2006; Lee et al., 2013; Bledau et al., 2014) (A.K., Q.Z., K.A. and A.F.S., unpublished). However, all six protein complexes share the same scaffold as the yeast Set1 complex, Set1C, based on four proteins, Wdr5, Rbbp5, Ash2l and Dpy30 (Wysocka et al., 2005; Dehe et al., 2006; Dou et al., 2006; Lee et al., 2007; Lee and Skalnik, 2012; Gu and Lee, 2013), termed WRAD (Ernst and Vakoc, 2012) or, less precisely, COMPASS (Shilatifard, 2012).

Recently, the Set1 H3K4 methyltransferases in Drosophila and Caenorhabditis elegans (named dSet1 and SET-2, respectively) have been shown to be essential for germline stem cells (GSCs) (Xiao et al., 2011; Robert et al., 2014; Yan et al., 2014). In mouse germ cell development, conditional mutagenesis identified Mll2 as essential for both oogenesis and spermatogenesis (Glaser et al., 2009; Andreu-Vieyra et al., 2010). Mll2 is the major di- and tri - but not mono-methyltransferase in developing oocytes and continues to be required during the first cell divisions after fertilization. However, by the blastocyst stage, Mll2 is neither the major H3K4 methyltransferase nor required (Andreu-Vieyra et al., 2010). These roles are acquired by Setd1a; however, Setd1a is not required for oogenesis or early mouse development (Bledau et al., 2014). Here, we report that Setd1b is required during oogenesis for the oocyte-to-embryo transition.

As previously reported, Setd1b^−/− embryos are growth retarded from embryonic day (E) 7.5 and die before E11.5. This result was obtained with both Setd1b^D/D and Setd1b^FDC/FDC genotypes (Fig. 1A; Bledau et al., 2014) indicating that both the targeted allele (D), which serves as a gene trap, and the Cre recombined allele (FDC), which lacks the frameshifting exon 5, are complete knockouts. During pre-implantation development, Setd1b is expressed in all stages from the oocyte to the blastocyst (Bledau et al., 2014). The first indication that Setd1b is required for oogenesis came from conditional knockout (cKO) of Setd1b in 2-month-old adults. Using Rosa26-Cre-ERT2 (RC; Seibler et al., 2003), tamoxifen induction of Setd1b^FD/FD;RC/+ females to generate Setd1b^FDC/FDC;RC/+ provoked infertility. After tamoxifen induction, Setd1b^FDC/+;RC/+ and Setd1b^FDC/FDC;RC/+ females were bred to wild-type males. On average, Setd1b^FDC/+;RC/+ females gave birth to 5.3 pups whereas Setd1b^FDC/FDC;RC/+ females never gave rise to offspring. To pursue this indication, we deleted Setd1b in oocytes using Gdf9-iCre (Fig. 1B; henceforth the Cre-deleted allele is called Setd1b^Gdf9 cKO) resulting in complete loss of Setd1b expression in oocytes as evaluated by qRT-PCR analysis (Fig. 1C), concordant with complete Cre recombination of genomic DNA (Fig. S1).

Breeding to males of verified fertility established that Setd1b^Gdf9 cKO females were sterile (Fig. 1D). The frequency of vaginal plugs and uterine wet weight were comparable to those of control females indicating that sex hormone levels and estrous cycle periodicity were normal in Setd1b^Gdf9 cKO females. Consequently, Setd1b^Gdf9 cKO ovaries were examined at 4, 8, 23 and 40 weeks.

At all analyzed time points, ovaries from Setd1b^Gdf9 cKO females were smaller compared with controls (Fig. 1E,F). At 4 weeks, Setd1b^Gdf9 cKO ovaries appeared otherwise normal and all expected follicular stages including small antral follicles were visible. At 8 weeks of age, the ovaries of control and Setd1b^Gdf9 cKO mice contained corpora lutea, which is evidence of cycling. At 23 weeks, Setd1b^Gdf9 cKO ovaries showed a reduction in follicle count in comparison with that of control mice whereas the number of atretic follicles and oocyte remnants was increased (Fig. 1G). These defects were even more apparent at 40 weeks.

Immunohistochemistry with major oocyte markers was carried out at selected stages; however, no striking differences between control and mutant were observed. Staining with Ki67 (Mki67) antibody to evaluate proliferative activity in the granulosa cells surrounding the oocyte was comparable in the Setd1b^Gdf9 cKO and control ovaries (Fig. 1H). Immunohistochemistry with antibodies against the prominent oocyte markers Zp2, ERα (Esr1), Sox9 and Ddx4 showed that oogenesis in Setd1b^Gdf9 cKO ovaries was apparently normal (Fig. 1H, Fig. S2A).

Gonadotropin-induced superovulation of pre-pubertal females yielded an average of 37.4 MII oocytes per control female, whereas Setd1b^Gdf9 cKO females yielded an average of 25.2 MII oocytes (Fig. 2A). Notably Setd1b^Gdf9 cKO periovulatory oocytes displayed the distinctive chromatin rearrangement from non-surrounding (NSN) to surrounding (SN) nucleolus configurations, which is the classical marker for oocyte maturation prior to quiescence and occurs concurrently with global transcriptional silencing (Fig. 2B, Fig. S2B).

In the final stages of oocyte maturation, the spindle assembles after nuclear envelope breakdown and segregates half of the chromosomes into the first polar body (Clift and Schuh, 2013). Then a second spindle assembles, relocates to the oocyte surface and the oocyte arrests in metaphase II awaiting follicular stimulation. As visualized by α-tubulin immunostaining, the meiotic spindles of Setd1b^Gdf9 cKO MII oocytes were considerably smaller than control spindles (Fig. 2C,D). Nevertheless, chromosome segregation (Fig. 2C) and metaphase II chromosome spreads (Fig. S2C) were normal indicating that all chromatids had been correctly attached to spindle fibers.

Setd1b^Gdf9 cKO oocytes and zygotes displayed increased cytoplasmic granularity (Fig. 2B,F) suggesting perturbed cytoplasmic organelles. Evaluation of the Golgi apparatus by immunostaining of GM130 (Golga2), a Golgi matrix protein associated with the cis-compartment (Wang et al., 2008), showed that Golgi structures were homogenously distributed and did not colocalize with the darkened cytoplasmic granules of the Setd1b^Gdf9 cKO MII oocytes (Fig. S2D). Furthermore, Tomm20 immunostaining revealed no disorganized mitochondrial distribution (Fig. 2E). Staining zygotes with Nile Red revealed that the granularity was caused by lipid droplets (Fig. 2F).

The impact of Setd1b removal on oocyte gene expression was evaluated by single-cell RNA profiling of four MII oocytes each from control and Setd1b^Gdf9 cKO superovulated mice (Figs 3 and 4). Using 75-base-pair reads, 16-23 million reads per sample with comparable mappability (80.7-86.4%) and 65.8-73.7% unique reads were obtained (Fig. 3A). All biological replicates clustered into two well-separated groups according to genotype thereby lending confidence regarding data quality (Fig. 3B). Our single-cell RNA-seq dataset is also in good agreement with a previously published dataset made from large numbers of MII oocytes (Park et al., 2013; data available at DNA Data Bank of Japan under accession number DRA001066) with a coverage of almost 80% of the transcripts found in their study (Fig. 3C). As expected, Setd1b itself was downregulated as was Rbm15, which has been reported to be a subunit of the Setd1b complex (Setd1bC; Lee and Skalnik, 2012). Similarly, another potential Set1C subunit, Bod1l (van Nuland et al., 2013), was downregulated but all the other known Setd1bC subunits were unaffected.

Unexpectedly for an H3K4 methyltransferase, we observed that more mRNAs were upregulated than downregulated. This observation was sustained in several ways. When we imposed a threshold of ten read counts per gene, about 10% more unique mRNAs (∼1100) were detected in each Setd1b^Gdf9 cKO MII oocyte than in controls (Fig. 3D). To analyze differentially expressed genes, we applied DESeq2 analysis to the RNA-seq dataset. Based on twofold changes in mRNA reads per kilobase million (RPKM), 1231 were upregulated, 650 were downregulated (and 8650 remained unchanged) in Setd1b^Gdf9 cKO oocytes (Fig. 3E). We also compared these differentially expressed mRNAs with a new transcriptome series generated from oocytes at different stages of oogenesis (Gahurova et al., 2017; data available at Gene Expression Omnibus under accession number GSE86297). This revealed that the 650 downregulated genes after deletion of Setd1b tend to be upregulated during normal oocyte development (Fig. 3F,G). Specifically, 375 of the 650 are normally upregulated, whereas only 43 of this group are downregulated and the remaining 214 show no change during oogenesis. This correlation accords with expectations for action by an H3K4 methyltransferase in transcriptional activity. Conversely, the 1231 upregulated mRNAs after deletion of Setd1b slightly correlated with downregulated mRNAs during normal oogenesis (357 normally downregulated, 243 upregulated and 574 unchanged) (Fig. 3F,G, Fig. S3, Table S3).

Remarkably, the most significant terms associated with downregulated mRNAs after Setd1b removal are ‘Regulation of transcription’, ‘Sexual reproduction’ and ‘Cell cycle processes’ (Fig. 4A). In contrast, we found an over-representation of terms related to ‘Redox’ and ‘Ribosome biogenesis’ for the significantly upregulated genes (Fig. 4B).

Notably, transcripts for the oocyte-specific homeobox genes (Obox1, Obox2, Obox5 and Obox7) were downregulated (Fig. 4C,G, Fig. S4A, Table S3). Concordantly, oocyte-specific genes such as Omt2a and Omt2b (oocyte maturation 2), Oosp1 (oocyte secreted protein 1) and Oog3 (oogenesin 3), were also downregulated (Fig. 4C,G). These genes are likely to be directly regulated by the Obox and other oocyte-specific transcription factors. Other notable transcription factors amongst the 90 that were downregulated more than twofold (Table S4) include Meis2 and Sall4. Additionally prominent were Klf17 and fifteen more Zfps (zinc finger proteins) with KRAB (Krüppel-associated boxes), which likely connect Setd1b regulation with the repression of transcription (Fig. 4D,H). In contrast, expression of oogenesis regulators such as Dppa5a and Nobox was sustained (Fig. 4I, Fig. S4B). Concordant with increased Nile Red staining (Fig. 2F), transcripts related to storage and binding of lipids were upregulated (Fig. 4E). Mature oocytes are characterized not only by transcriptional and translational silencing but also by respiratory quiescence (Sieber et al., 2016). Strikingly, transcripts of genes belonging to all five of the respiratory chain complexes were significantly upregulated (Fig. 4F, Fig. S4C). To validate the RNA-seq results, we performed quantitative RT-PCR on four selected genes (Sdhc, Sdhd, Cox7b, Ndufs2) and found that all of them were in good agreement (Fig. S4D). Transcripts encoding genes involved in ribosome biogenesis (Fig. S4E), purine and pyrimidine metabolism (Fig. S4F; data not shown) were also more prevalent in Setd1b^Gdf9 cKO oocytes. These transcriptional changes suggest a hyperactive metabolic state in Setd1b^Gdf9 cKO oocytes.

To examine fertility defects of Setd1b^Gdf9 cKO further, females were superovulated and mated with wild-type males. After verification of vaginal plugs, oocytes and zygotes were harvested for overnight culture. Most (69%) of the oocytes were fertilized and this proportion was comparable between control and Setd1b^Gdf9 cKO except for a few apoptotic mutant eggs (Fig. 5A).

Irregularities in the zona pellucida were consistently observed with Setd1b^Gdf9 cKO periovulatory oocytes and the irregularities persisted during the oocyte-to-embryo transition (Fig. 2B). Immunostaining of two-cell stage embryos with a Zp2 antibody revealed a disordered distribution in the zona matrix, which likely contributes to the irregularities and ruffling (Fig. 5B). Furthermore, we often observed multiple sperms (in more than half of the cases) within the Setd1b^Gdf9 cKO perivitelline space. Concordant with the polyspermy, polyploid embryos constituted approximately 10% of all fertilization events in Setd1b^Gdf9 cKO oocytes (Fig. 5C,D). Paradoxically, transcript levels for the major zona pellucida proteins were strongly elevated (Fig. 5E).

Almost all control zygotes advanced into the two-cell stage embryo compared with 26% for Setd1b^Gdf9 cKO zygotes (Fig. 6A). The two-cell Setd1b^Gdf9 cKO embryos displayed cellular and genetic asymmetries resulting in blastomeres and nuclei of different sizes (Fig. 2B, Fig. 5B, Fig. 6B) and none progressed to the four-cell stage.

Of the zygotes that did not engage the first cleavage, most (87%) were arrested in the pronuclear stages (PN2-5) (Fig. 6C). A further 13% arrested shortly after nuclear envelope breakdown and pronuclear fusion. Therefore, the majority arrested between the beginning of DNA replication and nuclear envelope breakdown. Together, these data indicate Setd1b is essential for progression through the first cell cycle and the predominant one-cell-stage arrest indicates that the requirement for Setd1b operates before zygotic genome activation.

No differences in global levels of H3K4me1, me2 or me3 were observed between control and Setd1b^Gdf9 cKO oocytes and zygotes (Fig. S5A-D). In Setd1b^Gdf9 cKO zygotes, both male and female pronuclei were established with apparently normal decondensed chromatin and extrusion of the second polar body (Fig. 2B), followed by successful migration from the cell periphery towards each other (Fig. S5B-E). Immunofluorescence showed comparable levels of H3K4me1 in female and male pronuclei of PN4 control and Setd1b^Gdf9 cKO zygotes. Also in both control and Setd1b^Gdf9 cKO zygotes, H3K4me2 and H3K4me3 immunofluorescence displayed the expected asymmetry between female and male pronuclei at PN4 (Fig. S5C,D). At PN5, the male pronucleus acquired H3K4me2 and me3 (data not shown). Hence, no perturbations of H3K4 methylation were observable in situ in Setd1b^Gdf9 cKO oocytes and zygotes. In both genotypes, only the female pronucleus was decorated with the silencing mark H3K9me2 (Fig. S5E). Also, DNA replication proceeded in both paternal and maternal nuclei (Fig. S5F).

This study began because near-ubiquitous conditional mutagenesis of Setd1b in 2-month-old females provoked infertility, which indicated that Setd1b is required for an essential aspect of fertility and/or embryonic development after normal MII oocytes had been established. However, our work with Gdf9-iCre did not deliver an explanation for this adult Setd1b requirement for fertility and further investigations are required. We also investigated whether Setd1b was required for oocyte development before deletion by Gdf9-iCre. Tamoxifen induced Rosa26-Cre-ERT2 deletion in utero at E6.5 and dissection at E10.5 revealed comparable numbers of alkaline phosphatase-positive cells migrating into the anlage of the female genital ridges of control and Setd1b^FDC/FDC;RC/+ embryos (Fig. S6) indicating that Setd1b is not required for specification or migration of primordial germ cells.

H3K4 methyltransferases are known to promote and/or sustain gene expression either through trimethylation of H3K4 on promoter nucleosomes, which presents binding epitopes for proteins involved in transcriptional activity (Ruthenburg et al., 2007; Vermeulen et al., 2007), action at enhancers (Guo et al., 2013; Hu et al., 2013) or by opposing Polycomb group repression, as exemplified by action of the paradigm, fly Trithorax (Klymenko and Müller, 2004). Consequently, loss of an H3K4 methyltransferase is expected to be accompanied by loss of target gene expression. This expectation was partly fulfilled here by the observation that mRNA levels for the oocyte-specific Hox genes Obox 1, 2, 5 and 7 were all decreased in the absence of Setd1b along with oocyte-specific target genes that are presumably regulated by the Obox factors. Indeed, along with the Oboxs, the prominence of Meis2 and Sall4 in the top ten most highly expressed transcription factors that were downregulated in the absence of Setd1b leads us to propose that they are also part of the oocyte-specific gene expression program (Fig. 7). Meis factors are known binding partners of other homeodomain transcription factors (Chang et al., 1997; Shen et al., 1999; Moens and Selleri, 2006) suggesting that Meis2 may be an Obox partner. Sall4 is central to early development and pluripotency (Elling et al., 2006), mutated in premature ovarian failure (Wang et al., 2009) and dependent on the early oocyte-specific transcription factor Nobox for expression in oocytes (Choi et al., 2007).

However, against expectations for H3K4 methyltransferases, we found that more mRNAs were upregulated than downregulated in Setd1b^Gdf9 cKO MII oocytes (Fig. 3). Potentially, the widespread mRNA upregulation was due to reactive compensation caused by loss of Setd1b-dependent mRNA expression and consequent functional deficits. This indirect explanation could be part of the reason but it is unsatisfactory because it does not explain why upregulation exceeded downregulation. Furthermore, many of the upregulated mRNAs were either related to metabolic processes, such as ribosome biogenesis, that should have been downregulated in MII oocytes after their peak levels earlier in oogenesis or were mRNAs that should not be expressed in oogenesis. Consequently, we suggest that Setd1b is not only required for appropriate expression of transcriptional activators, such as the Obox factors, but also transcriptional repressors required to maintain the later stages of the oocyte-specific expression program by reducing earlier and unwanted gene expression. Support for this proposition was found in the observation that the most significant term associated with the downregulated mRNAs is ‘Regulation of transcription’ (Fig. 4A). Further investigation revealed that amongst the list of transcription factors downregulated more than twofold in Setd1b^Gdf9 cKO MII oocytes (Table S4), a quarter (16/60) are Zfps with KRAB domains. The KRAB domain mediates gene repression through Trim28/KAP1, the NuRD (nucleosome remodeling deacetylase) complex, Setdb1 (the histone 3 lysine 9 methyltransferase) and HP1 (heterochromatin protein 1; Ayyanathan et al., 2003; Lupo et al., 2013). Potentially, these 16 Znf-KRAB factors mediate a substantial part of the Setd1b repressive capacity. Highest on this list of potential oocyte-specific repressors is the unnamed gene AI987944 (and we suggest the name Zaphod1 for zinc finger-associated protein highly implicated in oocyte development), followed by Klf17, Zfp979, Zfp735, Gli3, Zfp616 and Zfp820. To add further context, we examined the list of 1231 mRNAs that were upregulated more than twofold (excluding poorly expressed mRNAs not upregulated to above 100 RPKM) for transcription factors and nuclear proteins (Table S5). Notably, the key early regulator of oocyte development Nobox (Rajkovic et al., 2004) rated high on the list. Sustained expression of Nobox in MII oocytes implies a failure of appropriate timely downregulation, suggesting that downregulation of Nobox is mediated by Setd1b-associated repressors as a late part of the oocyte-specific gene expression program. In contrast to the 16 Zfp-KRAB factors downregulated more than twofold, only three were upregulated more than twofold thereby lending some support to our proposition regarding the mechanism of Setd1b repression. Of other potential regulators high on the upregulated list we note the presence of Gbx2 (gastrulation brain homeobox 2), Glis1, the CBP/p300 co-factor Cited2 and five subunits of the Mediator complex, Med26, 7, 17, 6 and 14. We also note Dppa5, which was upregulated more than 30-fold, has an atypical KH RNA-binding domain like Ooep (oocyte expressed protein) (Pierre et al., 2007). Its substantial overexpression might interfere with Ooep action.

The paradoxical overexpression of zona pellucida components was unexpected because the zona pellucida was obviously compromised. It is possible that an oversupply of components provoked faulty assembly or was provoked by faulty assembly and a zona pellucida quality control feedback mechanism. Alternatively, overexpression could indicate delayed downregulation and persistence of the peak Zp mRNA levels achieved in growing follicles. Notably, ovastacin (Astl) expression was unchanged. In normal conditions, zona-penetrated sperm fuse with the oocyte plasma membrane and initiate oocyte activation including the release of cortical granule contents leading to modification of the zona pellucida that prevents subsequent sperm penetration (Clift and Schuh, 2013). Thus, superfluous spermatozoa do not enter the perivitelline space of fertilized oocytes. In Setd1b^Gdf9 cKO zygotes, superfluous sperm accumulated in the perivitelline space thereby testing the second polyspermy block at the oolemma. In 10% of the Setd1b^Gdf9 cKO zygotes this polyspermy block was not sufficient to prevent triploid zygotes. Alternatively, the triploid zygotes might not have been due to breaches of the oolemma but to non-extrusion of the second polar body, which formed a pronucleus. Non-extrusion of the second polar body mechanism is the main cause for tripronuclear zygotes after intracytoplasmic sperm injection (Staessen and Van Steirteghem, 1997).

The Setd1b^Gdf9 cKO MII oocyte gene expression profiles also displayed perturbations in metabolic processes particularly related to mitochondria, which concords with the low developmental capacity and abnormal morphology. Mature oocytes enter a quiescent state in which transcription has ceased, translation is suppressed and mitochondrial activity is very low (Sieber et al., 2016). Despite displaying the transition from the NSN to the SN configuration, Setd1b^Gdf9 cKO MII oocytes had more pervasive transcription than normal and appeared to maintain a more active metabolic state.

Based on these observations, we propose that Setd1b plays a crucial role in the regulation of the late oocyte-specific gene expression program. A part of this relationship involves upregulation of oocyte-specific transcripts and another part involves downregulation of metabolic processes particularly after their peak during follicular growth (Fig. 7).

Notably, the sole ortholog of Set1 in C. elegans, SET-2, is required for normal germline development and fertility (Xiao et al., 2011) and soma-specific genes were ectopically upregulated in the germline of worms lacking SET-2 with loss of H3K4 methylation and germline transdifferentiation (Robert et al., 2014) again indicating a role for Set1 in repression of inappropriate germline gene expression. Similarly, removal of the Drosophila Set1, dSet1, leads to gradual GSC loss with disruption of H3K4 trimethylation in the ovary (Xuan et al., 2013). In an additional study (Yan et al., 2014), using transgenic RNAi in the Drosophila female germline, several candidate genes including dSet1 were identified with a specific requirement in GSCs. Potentially, these similarities indicate conserved aspects of H3K4 methylation action in germ lineages.

In mouse oocytes, mRNAs for all six Set1/Trithorax-type members, Setd1a, Setd1b and Mll1-4, are expressed (Kageyama et al., 2007; Andreu-Vieyra et al., 2010; Pfeiffer et al., 2011; Bledau et al., 2014; Aoshima et al., 2015). To date, we have conditionally mutated Mll1, Mll2, Setd1a and Setd1b using Gdf9-iCre to establish that only oocyte-specific deletion of Setd1b and Mll2 perturb fertility (Andreu-Vieyra et al., 2010; Bledau et al., 2014; this paper; and A.K., Q.Z., K.A. and A.F.S., unpublished). Mll2 is required during oogenesis and shortly after fertilization when it is the major H3K4 trimethyltransferase. By the blastocyst stage, the embryo and embryonic stem cells are independent of Mll2 but dependent on Setd1a (Bledau et al., 2014). Mll2 is again required after gastrulation; however, from the blastocyst stage onwards it is no longer the major H3K4 trimethyltransferase (Glaser et al., 2006). Setd1b is required for progression through the first cell cycle before zygotic genome activation but like Mll2 is not required in embryonic stem cells. Notably, Setd1b, like Mll2, is again required after gastrulation with the knockout of either gene presenting a very similar retarded and disorganized phenotype (Bledau et al., 2014). These functional shifts amongst the H3K4 methyltransferases suggest aspects of cell-type specificity that are difficult to explain. However, the role of Setd1b in the mouse oocyte potentially provides a new paradigm for action by an H3K4 methyltransferase as an intermediate coordinator in a cell type-specific gene expression program.

The Setd1b conditional strain (Fig. 1A) was generated previously (Bledau et al., 2014). Setd1b^D/+ mice were crossed to the CAGGs-Flpo (Kranz et al., 2010) line to generate Setd1b^FD/+ mice. Subsequently, those mice were crossed to the Rosa26-Cre-ERT2 (RC) line (Seibler et al., 2003) to generate conditional, tamoxifen-inducible Setd1b^FD/+;RC/+ mice. In order to achieve Setd1b deletion in utero, tamoxifen was administered to pregnant Setd1b^FD/+ females after crossing to Setd1b^FD/FD;RC/RC males. Pregnant females at E6.5 received one dose of 3 mg tamoxifen by oral gavage and embryos were dissected 4 days later. To obtain Setd1b-deficient oocytes by Cre deletion from postnatal day 3, Setd1b^FD/FD females were crossed with Gdf9-iCre males (Lan et al., 2004). For brevity, Setd1b^{FDC/FDC;Gdf9iCre/+} mice are termed Setd1b^Gdf9 cKO and Setd1b^FD/FD or Setd1b^{FDC/+;Gdf9iCre/+} are termed control (Fig. 1B). Primers for genotyping are provided in Table S1. All animal experiments were performed according to German law.

Oocytes from 24- to 28-day-old females were obtained by injection with 5 IU of pregnant mare serum gonadotropin (PMSG, Intravet Deutschland, 707165) followed by an injection of 5 IU human chorionic gonadotropin (hCG, Sigma-Aldrich, CG-10-1VL) 46 h later. Cumulus-oocyte complexes were isolated from the ampulla 16 h after hCG injection and collected in KSOM containing 1 mg/ml hyaluronidase (Sigma-Aldrich) to dissociate cumulus cells. In order to obtain periovulatory oocytes, ovaries were digested and fully grown oocytes (>60 µm) with a germinal vesicle were collected. To evaluate in vitro development, females were mated, embryos collected, cultured in KSOM and scored at the indicated time points.

Ovaries were fixed with 4% paraformaldehyde (PFA). The dehydrated ovaries were embedded in paraffin, and 5 μm-thick sections were prepared and stained with Periodic acid-Schiff (PAS) stain. Immunohistochemistry was performed as described (Bledau et al., 2014). Primary antibodies for immunohistochemistry were used as described in Table S2. Embryos were isolated, fixed in 4% PFA and stained with 25 mM Tris-maleate pH 9.0, 8 mM MgCl₂, 0.4% 1-naphthyl phosphate (Sigma-Aldrich) and 1% Fast Red (Sigma-Aldrich).

Oocytes were fixed in 4% PFA in PBS/0.2% bovine serum albumin (BSA) for 20 min at 37°C. After several washes in 0.2% BSA in PBS-Tween 20 (0.05% Tween 20 in PBS), the oocytes were permeabilized in 0.5% Triton X-100 in PBS for 15 min at room temperature (RT). Oocytes were incubated with the desired primary antibody according to Table S2. Secondary antibodies were used as followed: bovine anti-rabbit IgG-Rhodamine (1:50; Santa Cruz, sc-2367), goat anti-rabbit IgG-CFL 488 (1:200; Santa Cruz, sc-362262), fluorescein rabbit anti-rat IgG (1:200; Vector Laboratories, FI-4001), goat anti-mouse IgG-CFL 488 (1:100; Santa Cruz, sc-362257). For detection of lipid droplets, oocytes were incubated with Nile Red (Sigma-Aldrich, 5 mg/ml) in KSOM for 5 min. Oocyte chromosome spreads were performed as described (Hodges and Hunt, 2002).

Fertilized oocytes were labeled in vitro with 5-bromo-2'-deoxyuridine (BrdU) in order to evaluate DNA replication according to Wossidlo et al. (2010). Zygotes were incubated in KSOM medium supplemented with 1 mM BrdU (Sigma-Aldrich) until 44 h post-hCG injection. After incubation, embryos were fixed for 15 min in 4% PFA in PBS/0.2% BSA at RT, permeabilized with 0.2% Triton X-100 in PBS for 20 min RT, treated with 4 M HCl for 10 min and neutralized for 10 min with 100 mM Tris-HCl (pH 8.5). After a second fixation step, embryos were blocked with 3% BSA/0.1% Triton X-100 in PBS for 2 h at 4°C and then incubated for 2 h at RT with anti-BrdU FITC-conjugated rat monoclonal antibody (1:50; Abcam, ab74545). Embryos were washed and mounted on slides with a small drop of Vectashield mounting medium (Vector Laboratories) containing 5 µg/ml propidium iodide.

Images of the mounted oocytes/embryos were taken using a Zeiss Axiocam MRm camera on an Axiovert 200M microscope with ApoTome and HXP illumination using 63×/1.4 oil immersion objective operated by ZEN blue software. z-stacks of the spindles were acquired by laser-scanning confocal microscopy using a Zeiss LSM 780 with 40×/1.2 W objective. Volume reconstructions used Arivis Vision 4D software. Length measurements of the spindle tubulin filaments were carried out using the Fiji Simple Neurite Tracer plugin (Longair et al., 2011). Images of histological sections were collected using an Olympus WF upright microscope.

Total RNA from oocytes was isolated using Trizol reagent (Ambion) and reverse transcribed using the SuperScript III First-Strand Synthesis System (Invitrogen). Real-time quantitative PCR was performed with Go Taq qPCR Master Mix (Promega) by Mx3000P multiplex PCR (Agilent). Ct values were normalized against Rpl19. Primer sequences and the lengths of the amplified products are given in Table S1. Fold differences in expression levels were calculated according to the 2^–ΔCt method (Livak and Schmittgen, 2001).

For single oocyte RNA profiling Setd1b^Gdf9 cKO and control mice were superovulated and four oocytes collected for each genotype. Cells were lysed, RNA reverse transcribed and amplified according to the protocol of SMARTer Ultra Low RNA Kit for Illumina Sequencing (vers. 1, Clontech). Subsequently, samples were subjected to NEBNext Ultra DNA library preparation for Illumina using indexed adaptors (NEB). Resulting libraries were pooled in equimolar quantities for 75-bp single-read sequencing on Illumina HiSeq 2000, resulting in about 20 (16-23) million reads per sample.

The short reads were aligned to the mm10 transcriptome with GSNAP (2014-10-14; Wu and Watanabe, 2005; Wu and Nacu, 2010) and a table of read counts per gene was created based on the overlap of the uniquely mapped reads with the Ensembl Gene annotation version 69, using featureCounts (version 1.4.4; Liao et al., 2014). The raw read counts were further processed with the DESeq2 R package (version 1.3.79; Anders and Huber, 2010). Sample-to-sample correlation was computed by Euclidean distance between samples based on the normalized counts. Differential gene expression analysis was performed with DESeq2, for which a maximum of 1% false discoveries (1% FDR) was accepted. To identify enrichment for particular biological processes associated with the differentially expressed genes (DEGs), the DAVID GO/BP/FAT database (Huang et al., 2009) was used. Gene set enrichment analysis was performed using GSEA software from the Broad Institute (Subramanian et al., 2005).

We thank Stefanie Weidlich and Mandy Obst for excellent technical assistance. We thank the Biomedical Services of the Max Planck Institute of Molecular Cell Biology and Genetics, Dresden, for excellent service. We are also grateful to Konstantin Lepikhov for helpful advice on monitoring replication in zygotes by BrdU incorporation.