Histone variant H3.3-mediated chromatin remodeling is essential for paternal genome activation in mouse preimplantation embryos

Derepression of chromatin-mediated transcriptional repression of paternal and maternal genomes is considered the first major step that initiates zygotic gene expression after fertilization. The histone


Derepression
of chromatin-mediated transcriptional repression of paternal and maternal genomes is considered the first major step that initiates zygotic gene expression after fertilization. The histone variant H3.3 is present in both male and female gametes and is thought to be important for remodeling the paternal and maternal genomes for activation during both fertilization and embryogenesis. However, the underlying mechanisms remain poorly understood. Using our H3.3B-HA-tagged mouse model, engineered to report H3.3 expression in live animals and to distinguish different source of H3.3 protein in embryos, we show here that sperm-derived H3.3 protein (sH3.3) is removed from the sperm genome shortly after fertilization and extruded from the zygotes via the second polar bodies (PB II) during embryogenesis. We also found that the maternal H3.3 (mH3.3) protein is incorporated into the paternal genome as early as 2h postfertilization and is detectable in the paternal genome until the morula stage. Knockdown of maternal H3.3 resulted in compromised embryonic development both of fertilized embryos and of androgenetic haploid embryos. Furthermore, we report that mH3.3 depletion in oocytes impairs both activation of the Oct4 pluripotency marker gene and global de novo transcription from the paternal genome important for early embryonic development. Our results suggest that H3.3-mediated paternal chromatin remodeling is essential for the development of preimplantation embryos and the activation of the paternal genome during embryogenesis.
H3.3 is essential for paternal genome activation 2 _______________________________________ Embryogenesis begins when two haploid genomes fuse to form a diploid zygotic genome after the sperm entering into the oocyte [1,2]. At the time of fertilization, both the sperm and oocyte genomes are transcriptionally repressed, whereas the maternal stored factors in the oocyte support and control the process of early embryogenesis. Maternal-to-zygotic transition (MZT) is an embryonic development stage under the exclusive control of the newly formed zygotic genome. This developmental process requires the zygotic genome activation (ZGA) to allow for the transition from specified germ cells to a totipotent embryo, in which both paternal and maternal genomes undergo dramatic epigenetic reprogramming regulated by maternal factors [3]. Overcoming chromatin-mediated transcriptional repression of paternal and maternal genomes is thought to be the first major step to initiate zygotic gene expression after fertilization [4][5][6][7][8][9].
The mammalian sperm genome is packaged into highly condensed chromatin consisting primarily of protamine but 5-15% residual histones. Following fertilization, ZGA occurs first in the paternal genome (male pronucleus) at the 1-cell stage embryo, while activation of the maternal genome is usually delayed and occurs at the 2-cell stage in mice [10][11][12][13]. Soon after fertilization, protamine is removed from the sperm genome. The paternal genomes subsequently undergoes chromatin remodeling through a massive and highly regulated exchange of canonical and variant histones including H1FOO, H3.3, microH2A and H2A.Z. The incorporation of histone variants into paternal genome is suggested to be necessary for the acquisition of totipotency and ZGA [14][15][16][17][18], but the function and mechanism for each histone variant in this process is still poorly understood. The repressive H2A variant macroH2A is found preferentially in the mouse female pronucleus, and appears to contribute to its transcriptional silence [15], while H1FOO and H3.3 can incorporate into both the paternal and maternal genomes and are probably associated with the transcriptional activation of zygotic genomes [8,9,[16][17][18][19].
Chromatin is composed of histones and DNA. Each 147 base pairs of DNA wraps around 8 core histone proteins H2A, H2B, H3 and H4 to form the basic chromatin unit defined as nucleosome. Packaging of DNA into nucleosomes not only helps store genetic information but also creates diverse means for regulating DNA-templated processes [6]. Interplay between transcriptional machinery and chromatin in the early embryo likely regulates the timing of ZGA [20]. Chromatin accessibility by transcriptional machinery is regulated through nucleosome positioning and configuration, heavily influenced by histone variants as well as post-transcriptional modifications of histone tails [14]. Histone exchange is a general mechanism for chromatin remodeling during embryogenesis, as gametespecific variants are replaced by somatic versions to regulate chromatin accessibility [15][16][17][18][19]. It has been observed that the epigenetic reprogramming of the paternal and maternal genomes is asymmetric and involves different maternal factors [15,16,18,21,22], indicating distinct mechanisms for the activation of paternal and maternal genomes during early embryonic development.
Histone variant H3.3 in mammals is encoded by two different genes (h3f3a and h3f3b), in which give rise to an identical protein product [23,24]. H3.3 is constitutively expressed in cells and incorporated into the chromatin through a DNA synthesis-independent (DSI) pathway during and outside of S phase [25]. H3.3 has been a subject of increasing interest in the field of developmental biology due to its distinguishing roles in remodeling both the male and female genomes during fertilization and early embryonic development. H3.3 protein is enriched in both the genomes of sperm (sperm-derived H3.3, sH3.3) and oocyte (oocyte-derived H3.3, oH3.3). Notably, mature oocytes contain abundant H3.3 mRNA which would give rise to maternal H3.3 (mH3.3) upon activation (Fig. 1A) [26]. H3.3 has been found in the decondensing sperm nucleus at fertilization and is important for the male 3 pronucleus formation [6,8,9,[27][28][29]. However, the fate of these differently originated H3.3 and their possible functions in regulation of paternal and maternal genomes during development remain unclear. In this study, we tracked the sH3.3 and mH3.3 in the paternal genome during fertilization and embryogenesis by taking advantage of our H3.3B-HA-tagged mice model [26]. By applying H3.3-specific siRNA, in oocytes, we investigated the effects of mH3.3 knockdown on the transcriptional activation of paternal genome in mouse embryos.

Sperm-derived H3.3 is removed from the paternal genome and extruded from the zygote during fertilization
We have detected H3.3 in mature sperm genome using our H3.3B-HA-tagged mice [26]. However, the fate of sH3.3 in the embryo is not clear. In order to track sH3.3 during fertilization, we injected H3.3B-HA sperm into wild-type (WT) oocytes and performed immunofluorence (IF) staining with anti-HA antibody in the embryos collected 1h, 2h and 3h after sperm injection. The α-tubulin antibody was used to display the meiosis II spindles and the second polar bodies (PBII). As expected, sH3.3 was detected in the decondensing sperm genome in the embryo collected 1h postinjection (Fig. 1B). However, sH3.3 becomes undetectable in the male pronucleus in any of the embryos collected at 2h and 3h post-ICSI (Fig. 1C). Female pronuclei remain negative for HA staining throughout ( Fig. 1B -D).To our surprise, HA staining was found positive in meiosis II spindles and PBII in these embryos (Fig. 1C, D&F). To exclude the non-specific staining for the HA antibody in organelles, we used IgG as the control. We found that only the HA antibody staining was positive while the IgG staining was negative in these embryos (Fig. 1E), suggesting that the HA antibody is specific. To determine whether the protein recognized by HA antibody in the meiosis II spindle and the PBII is from H3.3B-HA sperm and is H3.3B-HA sperm-specific, we generated embryos using WT sperm for ICSI and the parthenogenetic embryos from artificially activated WT oocytes as the controls. Indeed, the staining was positive only in the embryos obtained by ICSI with H3.3B-HA tagged sperm (Fig. 1F); embryos obtained with WT sperm (Fig. 1G) or parthenogenetically activated embryos (Fig. 1H) were all negative. Our results showed that the HA staining in the meiosis II spindles and the PBII are H3.3B-HA sperm-specific, suggesting that the protein recognized by HA antibody is from the H3.3B-HA sperm. This result is further supportive of our observation that sH3.3 is removed from the sperm genome post-fertilization. Collectively, our results suggest that sH3.3 is removed from the sperm genome and excluded from the zygotes through PBII extrusion during embryogenesis.
In SCNT embryos, we have shown that donor nucleus-derived H3.3 is removed from the genome after activation, and the removal of donor nucleusderived H3.3 requires the incorporation of mH3.3 [19]. To determine whether the removal of sH3.3 requires the incorporation of mH3.3, we depleted mH3.3 by injection of siH3.3 into the MII oocytes. As shown in our previous study, over 80% of the mH3.3 is depleted 10h after oocyte activation [19]. To achieve the maximum depletion of mH3.3 in the oocytes, we injected H3.3B-HA sperm into the H3.3 knockdown (mH3.3KD) oocytes 10h after activation and collected the embryos for IF staining of H3.3B-HA 3h post-sperm injection. We found that sH3.3 was retained in the paternal genome in these embryos (Fig. 1I). This result suggests that mH3.3 is required for the removal of sH3.3 in the embryos during fertilization.

Maternal H3.3 is incorporated into the paternal genome and is detectable in the genome of preimplantation embryos until the morula stage
Maternal H3.3 can be deposited to the sperm genome after fertilization [22,28]. To track the deposition of mH3.3 into the paternal genome, we injected WT sperm into H3.3B-HA oocytes and monitored the incorporation of mH3.3 in the paternal genome in embryos recovered at different time points after sperm injection. At 1h postfertilization, mH3.3 was hardly detectable in the sperm genome while strong signal was detected in 4 the maternal genome (oH3.3) ( Fig. 2A). Then mH3.3 was clearly detected in the decondensing paternal genome in embryos 2h post-fertilization ( Fig. 2B), approximately the same time when sH3.3 was removed from the paternal genome (Fig.  1C). In order to track the deposition of mH3.3 in the paternal genome after the first embryonic division, we generated androgenetic diploid embryos by injecting two WT sperm into an enucleated H3.3B-HA oocyte (Fig. 2C) and monitored the incorporation of mH3.3 into the paternal genome. We could detect mH3.3 in the cytoplasm of the oocytes 1-2h post-sperm injection (Fig. 2D). Strong signals were detected in the paternal genome after the first cell division at the 2, 4-cell stage embryos and remained detectable in the early morula embryos ( Fig. 2E-2G). However, mH3.3 was not detected in androgenetic blastocysts ( Fig. 2H), suggesting that mH3.3 is either removed from the genome or diluted out at the blastocyst stage. Altogether, our results demonstrated that mH3.3 is deposited into the paternal genome shortly after fertilization and retained in the paternal genome until the early morula stage.

Maternal H3.3 is required for early embryonic development and the activation of paternal Oct4 pluripotent gene
We reported previously that mH3.3 is required for development of SCNT embryos and for the activation of pluripotency-associated genes during oocyte reprogramming [19]. We hypothesize that mH3.3 is also required for the development of fertilized embryos and for the activation of paternal pluripotent genes. To test this hypothesis, we knocked down mH3.3 in oocytes by injecting siH3.3 followed by ICSI fertilization. Untreated fertilized embryos and embryos injected with an siRNA against luciferase served as controls. In a total of 82 siH3.3 injected oocytes, 53.6%, 36.6%, 13.4% and 4.8% of them reached to 2-cell, 4-cell, morula and blastocyst stage, respectively ( Fig. 3A; Table 1). The developmental potential of H3.3KD embryos was significantly decreased compared to both control embryos (P<0.01, χ2 test), suggesting that mH3.3 is essential for early embryogenesis for fertilized embryos.
Previous studies have demonstrated that androgenetic haploid embryos generated by injection of one sperm into enucleated oocyte can develop to blastocysts and give rise to embryonic stem cell (ESC) lines [30,31], providing a simple model to study the reprogramming of the paternal genome by oocytes. To investigate the role of mH3.3 on remodeling the paternal genome without considering the effect of maternal genome on the developing embryos, we generated androgenetic haploid embryos by injecting one sperm head into an enucleated oocyte after mH3.3 was depleted by siH3.3. In rescue experiments, exogenous H3.3 mRNA was injected back (H3.3-addback) into the oocytes. Untreated and luciferase siRNA injected androgenetic haploid embryos were used as controls. Of luciferase-injected embryos, 25.6% and 16.3% developed to morula and blastocyst stages, respectively, similar to what was observed of untreated androgenetic haploid embryos (30.8%, and 19.2%). In contrast, only 3.7% of H3.3KD embryos developed to morula stage, and none of the 82 mH3.3KD embryos developed to blastocysts ( Fig. 3B; Table 2); H3.3-addback significantly increased the percentage of morula (23.2%) and blastocysts (12.5%) embryos (P<0.01, χ2 test). Our results demonstrated that knockdown of mH3.3 resulted in compromised embryonic development for both the fertilized embryos and androgenetic haploid embryos, suggesting that mH3.3 plays an essential role for reprogramming of the paternal genome during embryogenesis.
Expression of Oct4 (Pou5f1) is the hallmark for pluripotency establishment and the activation of zygotic genome during embryonic development [32]. In order to determine whether mH3.3 is involved in the activation of paternal pluripotent genes, we first check the effect of mH3.3 on expression of Oct4 in androgenetic haploid embryos using RT-qPCR. Since Oct4 mRNA is largely stockpiled in the MII oocyte cytoplasm, we used maternal Oct4 mRNA level in MII oocytes as the baseline to normalize our RT-qPCR data. Initially, Oct4 transcript levels from 1-cell (8h after fertilization) to 2-cell stages was decreased, 5 indicating degradation of maternal Oct4 mRNA after fertilization. Expression of Oct4 then increased from the 4-cell to 8-cell stages and peaked at the morula stage. In mH3.3KD embryos, Oct4 levels were significantly lower at the 8-cell and the morula stages (Fig. 3C). This result suggests that the expression of paternal Oct4 gene is significantly down-regulated when mH3.3 is depleted in the oocytes.
The expression of Oct4 can be visualized by GFP in morulae and blastocysts using the GFP transgene driven by the Oct4 promoter (Oct4-GFP) mice [33]. To validate if knockdown of mH3.3 in oocytes affects the expression of paternal Oct4 protein in fertilized embryos, we injected Oct4-GFP sperm into WT oocytes and monitored GFP expression with fluorescence microscopy (Fig.  3D). As expected, strong GFP fluorescence was observed in morula and blastocyst control embryos, while the majority of mH3.3KD embryos did not show GFP fluorescence at either stage (Fig.  3E&3F). Our study demonstrated that depletion of mH3.3 in oocytes resulted in down-regulation of Oct4 level in androgenetic haploid embryos and failure to activate the paternal Oct4-GFP gene in fertilized embryos. This result indeed suggests that mH3.3 is required for the activation of paternal Oct4 gene during embryogenesis.

Maternal H3.3 facilitates the paternal genome activation during embryogenesis
As we shown above, mH3.3 is required for the activation of paternal Oct4 gene during embryogenesis. We further investigated whether mH3.3 is involved in the global activation of paternal genome during development. The earliest zygotic transcription in mouse embryos begins at 1-cell stage as a minor wave of ZGA involving as many as 800 genes; the second wave of ZGA occurs at the 2-cell stage and involves about 3500 genes [11][12][13][14]. We thus used 5-EU (5-Ethynyl Uridine) to label newly synthesized transcripts in 2-cell androgenetic embryos and indeed observed robust transcriptional activity which was markedly reduced in H3.3KD embryos. Notably, 5-EU incorporation was rescued by H3.3-addback (H3.3-R) in all mH3.3 depleted embryos. Embryos treated with actinomycin D, which inhibits transcription in general, served as negative control (Fig. 4A). Our results suggest that mH3.3 is required for de novo transcription of the paternal genome in androgenetic haploid embryos, at least at the 2-cell stage.
We further performed high-throughput RNA sequencing (RNA-seq) on androgenetic haploid embryos. We collected embryos 50h after ICSI (8cell or 16-cell stage), which are expected to have robust zygotic transcription with maternal mRNAs being largely depleted. RNA-seq results confirmed excellent knockdown of both H3.3A and H3.3B mRNAs in mH3.3KD and H3.3-addback embryos (Fig. 4B). A total of 460 genes were found significantly (P<0.05) differentially expressed between WT and mH3.3KD embryos including Oct4 (pou5f1) gene, and 176 of them were downregulated at least 1.5 folds in mH3.3KD embryos ( Fig. 4C&4D; Table S1). From a total of 333 genes significantly differentially expressed between mH3.3KD and H3.3-addback (P<0.05), 165 genes were increased in expression in H3.3addback embryos (Table S2). We also identified 21 genes that were down-regulated in mH3.3KD embryos and significantly up-regulated by H3.3addback ( Fig. 4E&F; P<0.05). RT-qPCR analysis further confirmed the expression patterns of these genes (Fig. 4G). Pathway analysis revealed that many of these genes were involved in cell cycle, mitotic nuclear division, and cell division (Table  S3). Collectively, our study revealed that mH3.3 is essential for the expression of many paternal genes in androgenetic haploid embryos and plays a critical role in regulating the paternal genome activation during embryogenesis.

Discussion
Histone variant H3.3 has drawn particular interest in the fields of chromatin and developmental biology for its distinctive characteristics. First, unlike replication-coupled deposition of its canonical counterparts H3.1/2 proteins, H3.3 incorporation into chromatin is replication independent. Second, H3.3 has been consistently associated with an active state of 6 chromatin, which is expected to be causally involved in the regulation of gene expression or to have downstream consequences for the structure and function of chromatin [34][35][36][37][38][39]. Third, we and others have shown that H3.3 is enriched in both the genomes of male (sH3.3) and female gametes (oH3.3) and that H3.3 mRNA is abundant in the cytoplasm of mature oocytes (mH3.3) [19,26,28,29]. H3.3 plays a critical role in regulating the formation of pronucleus in zygotes and the development of early embryos [8,9,19,22,[27][28][29]. As H3.3 differs from its canonical counterparts H3.1/2 in only 4-5 amino acids, it has been difficult to generate highly specific H3.3 antibodies for in vivo studies. In order to monitor H3.3 deposition in vivo, we have generated H3.3B-HA-IRES-EYFP/mCherry knock-in mice [26], which allow us to track H3.3 protein in vivo and importantly, to distinguish the different origins of H3.3 proteins (sH3.3, mH3.3 and oH3.3) during fertilization and embryogenesis. In this study, we used H3.3B-HA sperm to fertilize WT oocytes to track sH3.3 in embryos by HA antibody staining. We found that sH3.3 is removed from the sperm genome shortly after fertilization, and that sH3.3 removal is tightly associated with the incorporation of mH3. 3. Notably, we further demonstrated that mH3.3 is essential for early embryonic development and for paternal genome transcriptional activation. Our results suggest a multifaceted role of H3.3 in paternal chromatin remodeling during fertilization and embryogenesis.
Mammalian spermatozoon is packaged mainly with protamine and remaining 5-15% histones [40,41]. Global analysis on both mouse and human spermatozoon shows that retained histones are mainly enriched at embryonic developmental promoters, coupled with transcription repressionassociated histone modifications. This pattern is suggested to serve as an epigenetic memory mark that poises the genes in sperm genome for activation upon fertilization [1,22,[40][41][42][43][44]. During the transition from transcriptional quiescence to ZGA, the sperm nucleus undergoes massive changes in chromatin composition, including protamine removal from the sperm genome and loading of maternal-derived histones [34,36,40,45,46]. While protamine has been known to be quickly removed from the sperm genome, the fate of retained histones in the sperm genome during fertilization remains unclear, partially due to the difficulty in distinguishing sperm-derived histones from maternal-derived ones in fertilized embryos. Using our H3.3B-HA tagged mouse sperm, we found that sH3.3 is removed from the paternal genome shortly after fertilization, a process that requires the presence of mH3.3. This finding is unexpected as H3.3 enrichment in the sperm genome is thought to serve as an epigenetic signature for gene activation and to be transmitted to the embryo through fertilization [1,[40][41][42][43][44]47]. Our observation underlines the need to further investigate the function of H3.3 in the sperm genome. Indeed, H3.3 is enriched at the promoter of many genes in the sperm genome whose transcriptional activation depends on H3.3 during embryogenesis (Fig. S1). We speculate that H3.3 enrichment at the promoter, together with repressive-associated histone modifications in the sperm genome marks developmentally important genes for activation during ZGA. H3.3 replacement (newly synthesized H3.3 replacing nucleosomal canonical H3.1/2 and H3.3) is a general mechanism for pluripotent gene activation during reprogramming [17,19,48]. Interestingly, H3.3 replacement is also closely associated with the removal of repressive histone modifications at the promoter such as histone H3 lysine 27 trimethylation (H3K27me3) (our unpublished data). We hypothesize that sH3.3 carrying repressive histone modifications is replaced by mH3.3 to create a permissive chromatin state for transcriptional activation during ZGA. While the results from this study support an extensive or even complete sH3.3 removal during fertilization, it remains possible that a fraction of sH3.3 with special modifications retains and is transmitted to the zygote [1,43,44]. Employing the recently established technology for ChIP-sequencing on embryos [49,50] and our H3.3B-HA tagged mice, identification of the dynamic location of mH3.3 and sH3.3 as well as histone modifications in the 7 paternal genome during ZGA would provide more important insights. While sH3.3 is found to be removed from the paternal genome during fertilization, our results also showed that mH3.3 is incorporated into the paternal genome approximately the time as sH3.3 is removed, and that incorporated mH3.3 remains detectable in the paternal genome in developing embryos until the morula stage. This result indicates that mH3.3 may play an important role in regulating the paternal genome activation during embryogenesis. Indeed, depletion of mH3.3 in oocytes resulted in significantly compromised embryonic development of both normal fertilized embryos and androgenetic haploid embryos. Mechanistically, we found that activation of pluripotent Oct4 gene in the paternal genome was impaired and the de novo transcription from the paternal genome was significantly affected when mH3.3 was depleted in oocytes. Furthermore, analysis of high-throughput RNA sequencing of androgenetic haploid embryo revealed that hundreds of embryonic expressed genes (at 8-cell and 16-cell stage) were transcriptionally altered by mH3.3 depletion in oocytes. Thus, mH3.3 supports early embryonic development at least partially through paternal genome activation. The next step will be to identify the exact location of mH3.3 in the paternal genome during fertilization and embryogenesis. Our system in combination with the recently established low input ChIP-seq technology on mouse embryos [49,50] would provide a practical model for such future studies.
Our findings may also be relevant to humanassisted reproduction technologies, as improper zygotic genome activation is one of the major reasons observed in fertilized embryos that fail to develop [51]. It may prove useful to determine whether in such cases the oocyte has proper level of H3.3 mRNA and whether the sperm genome has incorporated mH3.3 adequately after fertilization.

Experimental procedures Animals and oocytes
Animals were housed and prepared according to the protocol approved by the IACUC of Weill Cornell Medical College (Protocol number: 2014-0061). B6D2F1 and ICR mice were purchased from Taconic Farms (Germantown, NY). Females were superovulated at 6-8 weeks with 5 IU PMSG (Pregnant mare serum gonadotrophin, Sigma-Aldrich, St. Louis, MO) and 5 IU hCG (Human chorionic gonadotrophin, Sigma-Aldrich) at intervals of 48h. MII oocytes from superovulated female mice were recovered 14-16h of hCG injection.

Injection of siRNA and mRNA into MII oocytes
MII oocytes from superovulated B6D2F1 females with PMSG (Pregnant mare serum gonadotropin, Sigma, G4527) were recovered 14-16h of hCG injection. siRNAs or mRNAs were injected into the oocytes with a piezo-operated microcapillary pipette (3-5 µm inner diameter). After injection, oocytes were kept in the room temperature for 10 min and then moved into the incubator for at least another 30min before enucleation and ICSI or parthenogenetic activation. siRNA sequences for H3.3A (h3f3a) and H3.3B (h3f3b) and cDNA sequence for H3.3addback construct are provided previously [26].

Enucleation and intracytoplasmic sperm injection (ICSI)
Oocytes were transferred into a droplet of HEPES-CZB containing 5 µg ml -1 cytochalasin B, which had previously been placed in the operation chamber on the microscope stage. Oocytes undergoing micromanipulation were held with a holding pipette, the metaphase II chromosomespindle complex was aspirated into the pipette with a minimal volume of oocyte cytoplasm. After enucleation, oocytes were transferred into cytochalasin B-free KSOM and returned to the incubator for at least 1 hrs for recovery. The sperm head was picked up with the injection pipette and each enucleated oocyte was injected one sperm head. Following ICSI, oocytes were kept in the room temperature for 10 min and moved to the KSOM medium to culture in the incubator at 37°C under 5% (v/v) CO 2 in air. Fluorescence microscopy, Immunohistochemistry and confocal imaging 8 Expression of fluorescence was detected in live embryos using a fluorescence inverted microscope (Nikon, TE2000-U). Images were captured with a digital camera and merged in NIS-Elements D software (Nikon).
For detecting newly synthesized RNA of paternal genes in androgenetic embryos, iClick™ EU Andy Fluor 488 Imaging Kit (GeneCopoeia, A009) was used. Briefly, 2-cell stage embryos were incubated with 1mM EU, which can incorporate into nascent RNA, for 5 min, and then in iClick reaction cocktail for 30min. Embryos treated with 0.1ug/ml actinomycin D for 30 min were served as negative control.

RT-qPCR
Primers were designed to span an exon-exon junction (Table S4). RT-qPCR was performed by using an Applied Biosystems StepOnePlus system and Power SYBR Green PCR Master Mix. RNA from 3 to 10 embryos was isolated by using TRIzol, and cDNA was made by using SuperScript III (Invitrogen). cDNA was treated with RNase H and diluted 1:10 in H 2 O, with 8 μL used per PCR. Gapdh was used as a control. Experiments were performed in biological triplicate and technical duplicate, with data represented as the mean and error represented as SD.

RNA sequencing of embryos
Five to ten embryos were transferred into a PCR tube with 100 µl lysate buffer by mouth pipette and RNA was prepared using Arcturus PicoPure kit (Life technologies, KIT0204). Library preparation was done according to a published protocol [52. Briefly, purified RNA was used for 1st strand synthesis, 2nd strand synthesis, and PCR amplification (x2). 200 ng of resulting DNA was sonicated to ~100-300 bp and then used to construct a sequencing library according to standard Illumina protocols. Libraries were sequenced using the Illumina HiSeq2000 platform (single end, 51bp), multiplexed at 3 samples per lane.

Data analysis
All data are presented as mean ± SD (Standard deviation). Differences between groups were tested for statistical significance using Student's ttest or χ 2 -test. Statistical significance was set at p < 0.05 or p < 0.01. For RNA-seq analysis, RNAseq reads were aligned to the mouse genome (mm10) using TopHat. Expression levels were obtained using CuffLinks, using upper-quartile and GC-normalization. Genes were considered expressed if their FPKM value was greater than 5. Differentially expressed genes were identified using a 2-fold cutoff, and requiring that one of the expression values (FPKM) is equal or greater to 5. Pathway analysis was performed using mouse Gene Ontology pathways.