Dppa2 and Dppa4 directly regulate the Dux-driven zygotic transcriptional program

1Activation of transcription from the embryonic zygotic genome is a key concerted molecular and developmental event occurring in two waves at the one- to two-cell stage in mice and the four- to eight-cell stage in humans (for reviews, see;;; ). Despite its importance, the precise molecular regulation of zygotic genome activation (ZGA) remains poorly understood. In particular, we still know little of the transcription factors and chromatin regulators that drive ZGA transcription and of their coordination. Recently, the transcription factor Dux was shown to bind and activate many such ZGA transcripts in an embryonic stem cell (ESC) model of ZGA and be required for correct preimplantation development. However, Dux itself is expressed in only the first or minor wave of ZGA, and what regulates Dux remains unknown.

2Mouse ESCs represent an ideal system to study the molecular mechanism governing ZGA. Under serum or primed culture conditions, ESCs are heterogeneous and contain a small percentage of cells that not only transiently express ZGA transcripts but also share certain epigenetic characteristics with the two-cell embryo (for reviews, see; ). These so-called “2C-like” ESCs can be easily identified using fluorescent reporters driven by the promoters of ZGA transcripts, such as the endogenous retrovirus MERVL or Zscan4 cluster. To date, repressors of the 2C-like state and ZGA gene transcription have been identified, including Kap1/Trim28, the histone demethylase Lsd1/Kdm1a, the histone chaperone Caf-1, and the LINE1–nucleolin complex, among others. However, aside from Dux, positive regulators that activate ZGA transcripts remain elusive.

3Developmental pluripotency-associated 2 (Dppa2) and Dppa4 are small putative DNA-binding proteins expressed exclusively in preimplantation embryos, pluripotent cells, and the germline. These small proteins contain a DNA-binding SAP domain and a conserved histone-binding C-terminal domain and physically interact and localize to euchromatin. Both single- and double-knockout ESCs retain expression of pluripotency markers and self-renewal, suggesting that these proteins are dispensable for stem cell pluripotency. Intriguingly, both single- and double-knockout mice survive early embryonic development only to develop lung and skeletal defects and perinatal lethality at a time when these genes are no longer expressed. This has led to suggestions that the proteins may be involved in epigenetic priming in early development; however, a role in preimplantation development or in regulating ZGA transcription has not been investigated.

4In order to identify new positive regulators of ZGA transcription, we performed a screen in ESCs, identifying 12 chromatin and epigenetic factors that increase the percentage of 2C-like cells within a population. Among these were Dppa2 and Dppa4. We investigated the regulation of these two proteins, revealing that promoter DNA demethylation during the germline cycle coincides with their expression in vivo, including in the oocyte. Knockdown of either Dppa2 or Dppa4 reduces 2C-like cells as well as expression of ZGA transcripts. Furthermore, knockout of Dppa2 and/or Dppa4 is sufficient to completely abolish this cell population. Importantly, this phenotype can be restored upon re-expression of both Dppa2 and Dppa4 but not Zscan4c, confirming that these two proteins are necessary to activate expression of ZGA transcripts. Furthermore, we show that both Dppa2 and Dppa4 bind and activate Dux. Notably, Dux is required for Dppa2 and Dppa4 to activate the 2C-like state and ZGA transcription. Therefore, Dppa2 and Dppa4 act as master activators of a ZGA transcriptional program by directly regulating the ZGA transcription factor Dux.

5In mice, ZGA takes place in two waves: a minor wave that occurs predominantly at the paternal pronucleus in the zygote and a more substantial major wave that takes place in the two-cell embryo. Unfortunately, these stages of development are not easily manipulated on the scale required for high-throughput screens. To circumvent this, we took advantage of a spontaneously occurring rare subpopulation of primed mouse ESCs that express transcripts usually restricted to ZGA, including the MERVL endogenous retrovirus and Zscan4 cluster. These “2C-like” ESCs also share several epigenetic features with the two-cell embryo, including global DNA hypomethylation, decondensed chromatin, and increased histone mobility.

6We first performed an in silico screen for potential positive regulators by selecting epigenetic and chromatin regulators that are expressed in the oocyte and/or zygote (Fig. 1A,B; see the Materials and Methods). As a positive control, we included Zscan4c, which has been implicated previously in activating early embryonic genes in stem cells. Candidate genes were individually cloned as GFP fusions and transiently transfected into ESCs containing a tdTomato fluorescent reporter driven by the MERVL promoter (Fig. 1A;; ). The ability of the individual genes to promote an early embryonic gene signature was tested both by flow cytometry analysis of the MERVL::tdTomato reporter (Fig. 1C,D) and quantitative RT–PCR (qRT–PCR) of a panel of ZGA transcripts (Supplemental Fig. 1A). Of the 22 candidates investigated, 12 promoted a 2C-like state by both flow cytometry and qRT–PCR. Following Zscan4c, Dppa4 was the strongest-scoring screen candidate, with its closely related and interacting partner, Dppa2, also among the screen hits, leading us to investigate these two genes further. Importantly, analysis of an independent microarray data set in which the transcriptome of ESCs following transcription factor overexpression was assessed revealed that just two of the 50 factors investigated promoted an early embryonic transcriptome (Supplemental Fig. 1B). Of the two factors that did promote expression of ZGA transcripts, Gata3 was similarly identified in our candidate-based screen, indicating that our bioinformatic preselection of candidates enriched substantially for potential ZGA regulators.

7To validate the 12 screen hits, we performed RNA sequencing (RNA-seq) of the GFP-positive and GFP-negative sorted cells following transient transfection of the relevant GFP fusion construct. Transcriptome analysis confirmed an up-regulation of 2C-like transcripts in the GFP-positive sorted cells compared with GFP-negative sorted controls (Fig. 2A; Supplemental Tables 1, 2). Consistently, the 12 screen hits also up-regulated genes that are similarly up-regulated following Dux overexpression, Caf-1 knockdown, or LINE1 knockdown (Fig. 2B; Supplemental Fig. 2A; ), indicating that the up-regulation of 2C-like transcripts is independent of how they are defined. To accurately determine transcript levels of Dux, we remapped the RNA-seq data to the Dux repeat sequence (see the Materials and Methods). Importantly, all 12 screen hits, including Zscan4c, Dppa2, and Dppa4, resulted in a significant up-regulation of Dux transcript (Fig. 2C). Interestingly, overexpression of Dux using a docycycline-inducible transgene induced expression of several of the screen hits, including the 2C-like genes Zscan4c and Sp110 as well as Dppa2 but not Dppa4 (Supplemental Fig. 2B). As all three of these genes are also up-regulated in 2C-like ESCs (Supplemental Fig. 2C), this could suggest that positive feedback loops act to reinforce the 2C-like state. Supporting this, analysis of published ChIP-seq (chromatin immunoprecipitation [ChIP] combined with high-throughput sequencing) data revealed Dux binding to the Dppa2 promoter (Supplemental Fig. 2D).

8One of the hallmarks of 2C-like ESCs and preimplantation embryos is the up-regulation of repetitive elements, including the MERVL endogenous retrovirus. We therefore examined the repetitive portion of the transcriptome. Sequencing data were remapped to the consensus sequence of specific repeat families implicated in early embryonic development (see the Materials and Methods). There was a mild up-regulation of LINE L1 elements and a large up-regulation of MERVL elements in the GFP-positive sorted cells compared with GFP-negative sorted controls (Fig. 2D). Dppa2 and Dppa4 overexpression was able to induce an eightfold and 14-fold increase in MERVL expression, respectively, consistent with the MERVL::tdTomato reporter activation in these cells. Other repeat families, including IAP, MaLR, and major satellites, remained unchanged (Supplemental Fig. 2E), illustrating the specificity of repeat up-regulation.

9One-hundred-ninety-five genes, including the 2C-like transcripts, were similarly up-regulated by Dppa2, Dppa4, and Zscan4c (Supplemental Fig. 2F). In all cases, the transcripts up-regulated by Dppa2, Dppa4, Zscan4c, or other screen hits, including Bahd1, Eya1, Hdac9, and Sp110, were similarly up-regulated in the mid to late two-cell stage during embryogenesis (Fig. 2E; Supplemental Fig. 2G), confirming that these transcripts are activated during ZGA. Other screen candidates, such as Gata3, Irf1, Tox3, and Trp63, up-regulated not only 2C-like transcripts but also other non-ZGA transcriptional networks (Supplemental Fig. 2G; Supplemental Table 2).

10To further support the role of Dppa2 and Dppa4 in promoting expression of 2C-like ESCs, we looked at their expression patterns during induced pluripotent stem cell (iPSC) reprogramming. Up-regulation of 2C-like transcripts at intermediate stages of iPSC reprogramming has been reported previously. Consistently, both Dppa2 and Dppa4 are expressed when the ZGA transcripts are up-regulated (Fig. 2F). In summary, Dppa2 and Dppa4 up-regulate an early embryonic transcriptional program.

11Given the specific and restricted expression pattern of Dppa2 and Dppa4 in the germline and early embryo, we investigated the regulation of Dppa2 and Dppa4 in vivo. Primordial germ cells (PGCs) undergo a wave of DNA demethylation, which is then re-established in the mature gametes before a second wave of DNA demethylation takes place after fertilization in the preimplantation embryo (for reviews, see; ). Consistently, in both male and female PGCs, the Dppa2/4 locus is demethylated (Fig. 3A; Supplemental Fig. 3A,B), which coincides with their expression in the gonads and developing oocytes (Fig. 3B). In sperm and oocytes, there is a gain in DNA methylation across the locus; however, the promoters of both Dppa2 and Dppa4 remain hypomethylated (Fig. 3A; Supplemental Fig. 3A). This is in contrast to 2C-like gene promoters that are more highly methylated compared with all gene promoters in sperm (Supplemental Fig. 3C). During preimplantation, there is a second wave of DNA demethylation across the entire Dppa2/4 locus (Fig. 3A). After implantation, levels of DNA methylation, including at the promoter, increase dramatically, consistent with the rapid silencing of Dppa2 and Dppa4 (Fig. 3C). The promoters of Dppa2 and Dppa4 remain methylated across all somatic tissues in which Dppa2 and Dppa4 are not expressed (Supplemental Fig. 3D). To further investigate the link between promoter DNA methylation and Dppa2/4 expression, we investigated transcriptome data from embryonic day 8.5 (E8.5) embryos that lacked the de novo DNA methyltransferase Dnmt3b, which is primarily responsible for establishing DNA methylation at promoter regions. Importantly, there was an increase in both Dppa2 and Dppa4 expression in Dnmt3b−/− embryos at a time when they are usually completely silenced (Fig. 3D), supporting a role for promoter DNA methylation in repressing these two genes in vivo. In summary, the Dppa2 and Dppa4 genes are primarily regulated by global demethylation during germline and early embryo development, and their products are therefore present in the oocyte at fertilization.

12To test their necessity for ZGA transcripts expression, we first performed Dppa2 and Dppa4 knockdowns in MERVL::tdTomato/ Zscan4c::eGFP reporter serum ESCs. Expression of either reporter accurately labels 2C-like cells. Cells were transfected with either control, Dppa2, or Dppa4 targeting siRNAs for 4 d, achieving 92% and 74% knockdown efficiency at the mRNA level, respectively (Supplemental Fig. 4A). This corresponded to 94% and 85% reduction in protein levels for Dppa2 and Dppa4, respectively (Fig. 4A,B). Interestingly, depletion of Dppa4 protein also reduced Dppa2 protein levels by 73% (Fig. 4A,B) despite mRNA levels remaining the same (Supplemental Fig. 4A), potentially due to protein destabilization. Dppa2 siRNA also led to a 22% reduction in Dppa4 protein levels (Fig. 4A,B). Analysis of the MERVL::tdTomato/ Zscan4c::eGFP reporters by flow cytometry revealed a dramatic depletion of the 2C-like ESC population (Fig. 4C), which was consistently reflected in the expression of selected ZGA transcripts, including Dux, by qRT–PCR, while pluripotency markers remained unchanged (Supplemental Fig. 4B). To further investigate the transcriptional changes occurring after Dppa2 or Dppa4 knockdown, we performed RNA-seq. The majority of differentially expressed transcripts were down-regulated, overlapped with 2C-like transcripts (Fig. 4D), and were similarly deregulated between Dppa2 and Dppa4 knockdowns (Fig. 4E; Supplemental Tables 3, 4). In addition to ZGA transcripts, there was milder down-regulation of a second group of genes in the knockdown samples that contained many lineage markers such as the gametogenesis genes Syce1, Sohlh2, and Mael (Fig. 4D), consistent with knockout ESC studies. Thus, while Dppa2 and Dppa4 likely have additional roles, the largest changes in gene expression occurred at the 2C-like transcripts. Analysis of the repetitive proportion of the genome revealed a down-regulation of LINE L1 and near absence of MERVL element expression (Fig. 4F). The down-regulated gene transcripts were expressed at the time of ZGA in preimplantation embryos (Fig. 4G). Transcripts expressed in 2C-like ESCs as well as those that are up-regulated following Dppa2, Dppa4, or Dux overexpression or following CAF-1 or LINE1 knockdown were all down-regulated following Dppa2 or Dppa4 knockdown (Supplemental Fig. 4C), indicating that the same set of ZGA transcripts is being regulated by Dppa2 and/or Dppa4 irrespective of how they are defined. Importantly, transcript levels of the Dux transcription factor were barely detected following Dppa2 and Dppa4 knockdown (Fig. 4H). Therefore, Dppa2 and Dppa4 knockdown results in a decrease in Dux expression, 2C-like transcripts, and cells in the 2C-like state.

13To confirm that Dppa2 and Dppa4 are required for ZGA-like gene transcription, we generated single- and double-knockout ESCs deficient for Dppa2 and/or Dppa4 using CRISPR–Cas9 targeting in MERVL::tdTomato reporter cells. Knockout of either or both proteins was confirmed by Western blotting (Fig. 5A). Strikingly, flow cytometry analysis of the MERVL::tdTomato reporter revealed that both single- and double-knockout ESCs completely lacked the 2C-like subpopulation (Fig. 5B). Next, we performed RNA-seq to further investigate the transcriptional changes that occur following loss of Dppa2 and/or Dppa4 (see the Materials and Methods). We observed a dramatic depletion of 2C-like transcripts, including the Zscan4 cluster, Tcstv3, and Zfp352 (Fig. 5C; Supplemental Fig. 5A; Supplemental Tables 5, 6), consistent with the knockdown experiments. The absence of Zscan4c was validated at the protein level by Western blotting (Fig. 5A). Furthermore, expression of the ZGA transcription factor Dux (Fig. 5D) and MERVL endogenous retrovirus (Fig. 5E) was mostly absent in the knockout cells. Despite the loss of the 2C-like state, the cells retained mRNA and protein expression of the pluripotency markers Nanog and Oct4 (Supplemental Fig. 5A–C), grew at a similar rate, and were able to be passaged for >20 generations (data not shown), consistent with previous reports. Thus, these results indicate that the 2C-like state as well as Zscan4 protein expression are dispensable for sustained self-renewal in culture.

14The intersection of differentially expressed genes from either Dppa2−/− or Dppa4−/− ESCs was similarly deregulated in the Dppa2−/−Dppa4−/− ESCs (Fig. 5C; Supplemental Fig. 5D), suggesting that the three genotypes are largely indistinguishable transcriptionally. Similar to the knockdown experiments, we also observed a second group of misregulated genes in the knockout ESCs that included many lineage-specific genes, including Mael, Syce1, and Sohlh2, consistent with previous findings. However, unlike the differentially expressed 2C-like transcripts, this second group of differentially expressed genes is not normally up-regulated at the time of ZGA during preimplantation development (Fig. 5F) and therefore likely represents a separate function of Dppa2 and Dppa4 outside of regulating 2C-like transcription. Interestingly, there were significant transcriptional differences between siRNA knockdown and CRISPR knockout approaches (Supplemental Fig. 5E,F). While the majority of genes, including the 2C-like genes, Mael, Syce1, and Sohlh2, showed at least a trend toward down-regulation in both siRNA knockdown and CRISPR knockout, there were other genes that were up-regulated and down-regulated specifically in the knockout cells compared with siRNA knockdown and vice versa (Supplemental Fig. 5G–K). These genes may be sensitive to either the dosage of Dppa2/4 or the duration of Dppa2/4 depletion and thus may represent secondary adaptive effects of Dppa2/4 deletion. Despite these differences, the 2C-like genes were down-regulated by both approaches. Therefore, Dppa2 and Dppa4 are necessary for ZGA transcript expression in ESCs.

15Next, we performed rescue experiments in the double-knockout ESCs. Consistent with our initial screen, overexpression of Dppa4 and Dppa2 in wild-type cells (Supplemental Fig. 6A) up-regulated the 2C-like cell fraction by flow cytometry (Fig. 6A). Expression of 2C-like transcripts, including Dux (Fig. 6B), was increased. Overexpressing both Dppa2 and Dppa4 resulted in a larger up-regulation in 2C-like cells and associated transcripts than either one alone (Fig. 6A,B), consistent with them acting in a complex. Consistently, in Dppa2/Dppa4-null ESCs, Dppa2 was not able to induce the 2C-like state or associated transcripts, and Dppa4 alone led to only a modest increase. Moreover, overexpression of Zscan4c was not able to rescue the Dppa2/Dppa4 knockout phenotype (Supplemental Fig. 6B), suggesting that Zscan4c requires Dppa2 and Dppa4 to enhance the 2C-like cell state. Importantly, reintroduction of both Dppa2 and Dppa4 resulted in a substantial increase in the MERVL::tdTomato-positive cell fraction (Fig. 6A) and associated transcripts, including Dux (Fig. 6B). Therefore, Dppa2 and Dppa4 together drive the 2C-like cell state and ZGA-like transcriptional network.

16Our results so far have revealed a role for Dppa2 and Dppa4 in regulating the 2C-like state and ZGA-like transcripts. Additionally, Dppa2 and Dppa4 are necessary and sufficient to regulate expression of the ZGA transcription factor Dux, which itself has been shown recently to regulate a similar ZGA transcriptional program. To determine whether Dppa2 and Dppa4 act to regulate Dux directly or exert their effects through parallel pathways, we determined whether Dux is required for Dppa2 and Dppa4 to regulate the 2C-like transcripts. Wild-type and Dux knockout ESCs were cultured in serum conditions and transfected with constructs containing Dppa2, Dppa4, or both Dppa2 and Dppa4 constructs simultaneously and compared with those receiving an empty vector (Supplemental Fig. 7A,B). While Dppa2 and/or Dppa4 were able to induce expression of the 2C-like transcripts in wild-type cells, this ability was abolished in Dux knockout cells (Fig. 7A). Therefore, Dux is required for the transcriptional effects exerted by Dppa2 and Dppa4.

17Next, we analyzed published Dppa2 and Dppa4 ChIP-seq data to determine whether these proteins directly bind the Dux repeat (see the Materials and Methods). Importantly, we observed clear enrichment of Dppa2 and Dppa4 binding across the promoter and into the Dux transcript itself in E14 ESCs (Fig. 7B; Supplemental Fig. 7C). Furthermore, Dppa4 similarly bound to Dux in P19 embryonal carcinoma cells but not in 3T3 fibroblasts (Supplemental Fig. 7C). Thus, Dppa2 and Dppa4 directly bind to Dux in PSCs, consistent with when Dux is expressed. We next investigated whether Dppa2 and Dppa4 bind to other genes that are differentially expressed in Dppa2−/− or Dppa4−/− ESCs. There was a strong enrichment for both Dppa2 and Dppa4 at the promoters of the non-2C-like genes (Fig. 7C), including Syce1, Sohlh2, and Mael (Supplemental Fig. 7D–F), again confirming a separate role for Dppa2 and Dppa4 in regulating non-ZGA transcripts in ESCs. Importantly, Dppa2 and Dppa4 did not bind to the transcriptional start sites (TSSs) of other 2C-like transcripts (Fig. 7C), including the Zscan4 cluster, Gm428, and Dub1 (Supplemental Fig. 7G–I), which are direct targets of the Dux protein itself.

18Our results support a model in which Dppa2 and Dppa4 act by directly regulating levels of the Dux transcription factor, which in turns acts to bind and promote expression of a zygotic transcriptional program in ESCs. Therefore, expressing Dux in the absence of Dppa2 and Dppa4 should restore the 2C-like state and associated ZGA transcripts. To this end, we overexpressed Dux in the Dppa2−/− Dppa4−/− ESCs (Supplemental Fig. 7J). Indeed, flow cytometry analysis of the MERVL::tdTomato reporter revealed that Dux overexpression is able to induce the 2C-like state in both wild-type and Dppa2−/−Dppa4−/− ESCs (Fig. 7D). Furthermore, expression of 2C-like transcripts was induced following Dux overexpression (Fig. 7E). Together, this suggests that Dux acts downstream from Dppa2 and Dppa4. In summary, Dppa2 and Dppa4 induce the 2C-like state and associated ZGA transcripts by directly binding and activating the ZGA transcription factor Dux, which is then able to bind and activate downstream 2C-like target genes.

19Initiation of transcription of the zygotic genome is a critical step in embryogenesis. To understand its molecular regulation, we performed a screen for chromatin and epigenetic regulators of ZGA transcription using 2C-like ESCs as a model, identifying, among others, Dppa2 and Dppa4. Here, we propose a model in which promoter DNA demethylation during global epigenetic reprogramming enables expression of Dppa2 and Dppa4 in the germline and oocytes. Dppa2 and Dppa4 then directly bind and up-regulate both non-ZGA genes as well as, under permissive conditions such as chromatin decompaction, the ZGA transcription factor Dux (Fig. 7F). Dux is subsequently able to bind and activate downstream ZGA transcription, including the Zscan4 cluster. Several feedback loops reinforce this system, including Zscan4-induced stabilization as well as Dux-induced up-regulation of Dppa2. Our study provides crucial insights into the molecular hierarchy that triggers ZGA transcription and links it with epigenetic reprogramming in the germline.

20The existence of a 2C-like state in ESCs, while not the same as the two-cell embryo, represents a useful in vitro approximation for studying ZGA, making many molecular and screening-based studies possible. This state is characterized by an increase in chromatin mobility, decondensed chromocenters, and increased chromatin accessibility and global DNA hypomethylation. Consistently, depletion of factors involved in chromatin assembly or treatment with inhibitors that ultimately induce chromatin decompaction increases the proportion of these cells in culture. Furthermore, knockdown or knockout of many repressive epigenetic regulators, including the histone demethylase Kdm1a, histone methyltransferase Ehmt2, heterochromatin protein HP1, and components of the PRC1.6 subcomplex among others, have also been shown to enhance the 2C-like state.

21In PGCs, the promoters of Dppa2 and Dppa4 are demethylated, coinciding with their expression. They remain robustly expressed through preimplantation development, including at the time of ZGA, until the onset of gastrulation, when expression rapidly ceases and their promoters reacquire DNA methylation. Dppa2 is highly expressed in oocytes and zygotes prior to ZGA, and, while Dppa4 transcripts are less abundant, proteins for both Dppa2 and Dppa4 are readily detectable. Additionally, Dppa2 and Dppa4 are expressed at the time during iPSC reprogramming, when 2C-like transcripts are transiently expressed. It will be interesting to determine whether Dppa2/4-mediated passage through a 2C-like state is required for iPSC reprogramming. Importantly, Dppa2 and Dppa4 are homogeneously expressed across all ESCs, not just 2C-like cells (data reanalyzed from ). However, this raises the question of why the ZGA genes are expressed in only a small subset of, but not all, ESCs. Activation of the 2C-like state is likely a multifaceted process requiring the presence of not only the upstream activators Dppa2 and Dppa4 but also chromatin decompaction and/or reduced expression of repressors such as Kap1 or PRC1 and their modifications. Once activated, factors such as Zscan4c may act to stabilize and prolong expression of the transcriptional program, which then requires repressors such as the LINE-1/Nucleolin complex, NuRD, or Caf-1 to repress it once again. In this way, Dppa2 and Dppa4 regulate the entry into the 2C-like state during permissive conditions by directly activating the ZGA major transcription factor Dux, which subsequently activates the remainder of the ZGA transcripts in the cell.

22In this study, we used Zscan4c as a positive control in the candidate-based screen. Zscan4c is one of a tandemly encoded family of zinc finger and SCAN domain-containing proteins that are expressed in early embryos and in 1%–5% of ESCs, including the rarer MERVL-positive 2C-like cells. Zscan4c has been implicated as a positive regulator of 2C-like cells, which we confirmed here. However, Zscan4c is unable to induce the 2C-like state or ZGA-like transcription in the absence of Dppa2, Dppa4, or Dux (data not shown). While it remains to be determined whether Zscan4c or other members of the Zscan4 cluster are necessary for ZGA transcription, our results suggest that Zscan4c may act to stabilize or reinforce the 2C-like state rather than induce it. Interestingly, as well as Zscan4c, Sp110 is also up-regulated in 2C-like cells and following Dux overexpression and was also identified in our screen as a positive regulator of the 2C-like state. It will be interesting to see whether it may also act as a reinforcer of the 2C-like state in ESCs.

23By bioinformatically preselecting candidates based on their expression pattern and gene ontology, we were able to enrich for 2C-like regulators in our candidate-based screen. It will be exciting to follow up on the other screen hits to determine whether they may be part of a larger mechanism working either in parallel or together with Dppa2 and Dppa4 to regulate this crucial developmental progression. Interestingly, consistent with a previous report, the Dppa family member Dppa3 (also known as Stella or Pgc7) was not able to induce the 2C-like state despite activating MERVL elements in embryos. This may be explained by cofactors present in the oocyte but absent in ESCs and required for Dppa3 function and/or differences in chromatin structure.

24Dppa2 and Dppa4 physically interact and bind to euchromatin in PSCs. Here, we show that Dppa2 and Dppa4 are regulated by DNA methylation and are necessary to induce an early embryonic transcriptional network by directly binding and regulating the ZGA transcription factor Dux in ESCs. Given the high perinatal lethality in the knockout and maternal stores of Dppa2 and Dppa4, it has not yet been possible to investigate whether Dppa2 and/or Dppa4 are required for ZGA in vivo. However, rare surviving Dppa4 knockout females have impaired fertility, yet germ cell development appears unaltered, suggesting potential important roles for this protein in preimplantation development. Furthermore, Dux knockdown in vivo results in impaired early embryonic development and defective ZGA. Moreover, injection of a dominant-negative form of Dppa2 lacking the SAP domain into zygotes induces two-cell arrest, suggesting that our results showing that Dppa2 and Dppa4 regulate the ZGA transcriptional network in ESCs may also apply to embryos.

25In summary, in this study, we performed a candidate-based screen to identify epigenetic and chromatin regulators of the 2C-like state and ZGA transcriptional program. Among these were Dppa2 and Dppa4, which act together to bind and up-regulate the ZGA transcription factor Dux, among other non-ZGA targets. Depleting Dppa2 and Dppa4 levels reduces the 2C-like population and ZGA transcription, which can be restored by reintroducing either both Dppa2 and Dppa4 together or the downstream factor Dux. In conclusion, our findings reveal important insights into the molecular mechanisms regulating ZGA transcription.

26Sequence-verified cDNA sequences lacking stop codons were amplified from plasmids purchased from Thermo Fisher using primers containing AttB1 and AttB2 sequences and cloned into the pDONR221 vector. Gateway cloning was then used to transfer the cDNA sequences into an in-house-built pDEST vector containing a CAG promoter and an in-frame C-terminal eGFP-coding sequence and blasticidin resistance by IRES fusion. Expression plasmids were sequence-verified by Sanger sequencing prior to use and are available on request.

27E14 mouse ESCs were grown under standard serum/LIF conditions (DMEM, 4,500 mg/L glucose, 4 mM L-glutamine, 110 mg/L sodium pyruvate, 15% fetal bovine serum, 1 U/mL penicillin, 1 mg/mL streptomycin, 0.1 mM nonessential amino acids, 50 mM b-mercaptoethanol, 103 U/mL LIF). Single-MERVL::tdTomato and double-MERVL::tdTomato/Zscan4c::eGFP reporter cell lines were described in, and Dux knockout cells were described in. Transfections were performed using Lipofectamine on preplated cells in six-well or 10-cm plate formats. Flow cytometry analysis was performed using BD LSR Fortessa, and sorts were performed on a BD Aria III or BD Influx high-speed cell sorter. siRNA transfections were performed by transfecting Dharmacon siRNA ON-TARGETplus siRNA SMARTpool at a final concentration of 50 nM with Lipofectamine.

28CRISPR knockout ESCs were performed as described previously. Guide RNAs were designed against exons 2 and 3 of both Dppa2 and Dppa4 using CRISPR design (http://crispr.mit.edu). Cells were transfected with a single guide targeting Dppa2 and/or Dppa4 and FACS (fluorescence-activated cell sorting)-sorted after 24 h into single cells, and clones were screened by surveyor assay and genomic DNA PCR. Successfully targeted clones were validated by Western blotting. Dppa2 single-knockout clones (clone 5 and clone 12) used in this study were generated using a guide RNA targeting Dppa2 exon 2 (ACCTTAGACCACACACCACCAGG), Dppa4 single-knockout clones (clone 23 and clone 29) were generated with a guide RNA targeting Dppa4 exon 2 (CTGCAAAGGCTAAAGCAACGGGG), and Dppa2/Dppa4 double-knockout clone (clone 43) was generated using a guide targeting Dppa2 exon 3 (TAACTTGAGTACGGATGGCAAGG) together with the guide RNA targeting Dppa4 (as above).

29RNA was isolated using Qiagen RNA–DNA allprep columns or TriReagent (Sigma) and treated with DNase I (Ambion DNA-free DNA [1311027] or Thermo Fisher RNase-free [EN0525]) following the manufacturer's instructions. cDNA was generated using 0.5–1 µg of RNA (Thermo RevertAid, K1622), and qRT–PCR was performed using Brilliant III SYBR master mix (Agilent Technologies, 600882). Relative quantification was performed using the comparative CT method with normalization to CycloB1 levels. Primer sequences are available on request. Opposite strand-specific polyA RNA libraries were made using 1 µg of DNase-treated RNA at the Sanger Institute Illumina bespoke pipeline and sequenced as single-end 50-base-pair (bp) reads using the Illumina HiSeq 2500 Rapid Run platform.

30Candidate 2C-like ZGA regulators were selected based on the following criteria. First, genes expressed in oocytes/zygotes (reads per kilobase per million [RPKM] > 1 in data) were selected and further filtered based on gene ontology (AmiGO Ontology search for “chromatin”-associated genes). This gave a total list of 84 candidate genes that was manually curated to remove nonnuclear proteins and overlapping transcripts. Genes were finally filtered to include those that had sequence-verified cDNA clones commercially available.

31Raw FastQ data were trimmed with Trim Galore (version 0.4.3, default parameters) and mapped to the mouse GRCm38 genome assembly using Hisat2 version 2.0.5. Data were quantitated at the mRNA level using the RNA-seq quantitation pipeline in SeqMonk software (http://www.bioinformatics.babraham.ac.uk/projects/seqmonk). Strand-specific quantification was performed using mRNA probes and cumulative distributions matched across samples. Differentially expressed genes were determined using DESeq2 (P-value of 0.05 with multiple testing correction) and intensity difference filter (P-value of 0.05), with the high-confidence differentially expressed genes defined as the intersection between the two statistical tests. For Dppa2/Dppa4 single- and double-knockout transcriptome analysis, as there was only one double-knockout ESC clone, differentially expressed genes were defined independently in Dppa2-knockout and Dppa4 knockout ESCs, and the two lists were combined to get the “Dppa2−/− or Dppa4−/−” differentially expressed genes.

32Published data sets were analyzed using SeqMonk software. For DNA methylation wiggle plots, the percentage methylation of individual CpGs with at least four reads are shown. ChIP-seq wiggle plots show normalized read counts for 10-bp running windows. For quantification of at gene promoters, probes were generated over TSSs ±500 bp, and the percentage DNA methylation or normalized ChIP-seq read count of the entire probe was calculated.

33The mouse dux ORF is part of a 5-kb repeat unit that is organized as a large tandem array, with repeat numbers being polymorphic between strains and within outbred mice. In the current genome assembly (GRCm38.p6), the protein-coding duxf3 is located on a patch (CHR_MG4264_PATCH: 58,251, 173–58,259,474) with several homologs and pseudogenes present. To minimize confound by genomic multimapping in the analysis of this gene, we mapped RNA-seq, bisulfite sequencing (BS-seq), and ChIP-seq data directly against the mouse dux repeat (AM398147.1). Read counts were normalized to the total read count of the sample.

34For alignments to repetitive regions in the genome, we constructed artificial repeat genomes. Repeat annotations were downloaded from the University of California at Santa Cruz (UCSC) browser (RepeatMasker, mm10, November 2018) and filtered for long instances of MERVL, MaLR, IAP, and LINE1 elements as well as major satellites (see Table 1). The length cutoff was introduced to enrich for functional full-length elements and exclude fragmented/truncated instances. Sequences of the filtered list of repeat element instances were stitched together, separated by “NNNNN” to create repeat-specific genomes. Trimmed reads from each sample were aligned against all individual repeat genomes using Bowtie2 (version 2.3.2). Values given are cumulated reads mapping to a specific repeat group as the percentage of the total read count.

35Western blotting was performed using 50 µg of protein extracted using detergent buffer (10 mM Tris-HCl at pH 7.4, 150 mM NaCl, 10 mM KCl, 0.5% NP-40) containing a protease inhibitor cocktail (Sigma, P2714) and quantified by Bio-Rad protein assay dye reagent. Proteins were resolved using 4%–12% SDS-PAGE gels (Expedon, NBT41212) and blotted on PVDF membranes. Following blocking in 5% skim milk/0.01% Tween/PBS, membranes were incubated with primary antibodies for 3 h to overnight. Secondary horseradish peroxidase-conjugated secondary antibodies (1:3000; Santa Cruz Biotechnology) were incubated for 1 h, and detection was carried out with enhanced chemiluminescence (ECL) reaction (GE Healthcare, RPN2209). The primary antibodies used were anti-Zscan4 (1:500; Millipore, 2793611), anti-Dppa2 (1:500; Millipore MAB4356), Dppa4 (1:200; Santa Cruz Biotechnology, sc-74614), and anti-HSP90 (1:5000; Abcam, ab13492).