DPPA2 and DPPA4 are dispensable for mouse zygotic genome activation and pre-implantation development

Following fertilization, embryonic development relies initially on maternal factors deposited during oogenesis and subsequently on the newly generated embryonic product after its genome is activated. The awakening of the embryonic genome is known as zygotic or embryonic genome activation (ZGA/EGA), which is essential for an embryo to acquire totipotency and to undergo normal development. In mice, ZGA consists of two successive waves of transcription – a minor and a major wave occurring at the late one-cell and late two-cell stages, respectively (Schultz et al., 2018). Transcription at the late one-cell stage is largely promiscuous and splicing and 3′ processing of the transcripts are largely incomplete (Abe et al., 2015), whereas the major ZGA at the late two-cell stage involves the expression of thousands of translatable mRNAs (Hamatani et al., 2004; Wang et al., 2004; Zeng et al., 2004). Notably, both minor and major ZGA are essential for mouse pre-implantation development (Abe et al., 2018; Schultz et al., 2018; Liu et al., 2020).

Two-cell (2C)-like cells are a rare cell subpopulation of embryonic stem cells (ESCs) that is characterized by the expression of many transcripts that are specific to two-cell embryos, such as Zscan4 and Zfp352, and has the expanded potential to contribute to both embryonic and extra-embryonic lineages (Macfarlan et al., 2012). 2C-like cells have been widely used as an in vitro system in which to study totipotency and ZGA (Fu et al., 2020; Genet and Torres-Padilla, 2020). Studies in the past several years have identified DUX (human homolog DUX4), a double homeodomain protein, as a master regulator of the 2C-like state in ESCs (De Iaco et al., 2017; Hendrickson et al., 2017; Whiddon et al., 2017). Most of the mechanisms that promote the 2C-like state in ESCs identified so far directly or indirectly regulate Dux activation. Intriguingly, loss of Dux in embryos only mildly affects ZGA and Dux null embryos are viable with reduced litter sizes (Chen and Zhang, 2019; Guo et al., 2019; De Iaco et al., 2020; Bosnakovski et al., 2021). Thus, although DUX is important for synchronizing and enhancing the expression of some 2C embryo-specific genes, it is not essential for ZGA and embryogenesis.

Given that Dux is not expressed in oocytes and it gets activated only at the late one-cell stage, upstream factors must have already been expressed in oocytes in order to trigger Dux expression during the minor ZGA wave. Recent studies have identified developmental pluripotency associated 2 and 4 (DPPA2 and DPPA4) as essential factors for establishing the 2C-like state in ESCs by activating Dux (De Iaco et al., 2019; Eckersley-Maslin et al., 2019; Yan et al., 2019). In addition, DPPA2 and DPPA4 directly regulate young LINE-1 elements in ESCs in a DUX-independent manner (De Iaco et al., 2019). The young LINE-1 elements levels also increase during the major ZGA wave. Taken together, these findings suggest that DPPA2 and DPPA4, which are expressed in oocytes, may regulate ZGA as maternal factors through both DUX-dependent and -independent pathways. In support of this, overexpression of the dominant-negative forms of Dppa2 has been shown to impair mouse pre-implantation development (Hu et al., 2010; Yan et al., 2019). However, because overexpression of dominant-negative forms might cause unknown side effects, these two studies did not provide definite evidence for Dppa2/4 function in ZGA. In addition, the previous Dppa2 and Dppa4 zygotic knockout (KO) studies (Madan et al., 2009; Nakamura et al., 2011) did not address the potential maternal contributions of these two proteins in ZGA and pre-implantation development. Thus, whether oocyte-derived DPPA2 and DPPA4 activate Dux expression and regulate ZGA in mouse embryos remains to be formally confirmed.

In this study, we generated maternal KO and maternal-zygotic KO mouse embryos for both Dppa2 and Dppa4, and determined their functions in Dux activation, ZGA and pre-implantation development. Our results demonstrate that both DPPA2 and DPPA4 are dispensable for ZGA and pre-implantation development.

To test the possibility that DPPA2 and DPPA4 are involved in activating Dux expression and ZGA, we first determined the expression dynamics and cellular localization of these proteins in early embryos by RNA sequencing (RNA-seq) and immunostaining analyses. We found that low levels of Dppa2 and Dppa4 RNAs were detectable in both oocytes and zygotes, and their expression levels were dramatically increased during major ZGA and reached a peak at the eight-cell stage (Fig. 1A). However, both Dppa2 and Dppa4 RNA became undetectable soon after embryo implantation (Fig. 1A); this transcriptional silencing is presumably achieved by gain of DNA methylation at the promoters (Eckersley-Maslin et al., 2019).

In contrast to the RNA levels, DPPA2 and DPPA4 immunosignals were not detectable in either oocytes or zygotes (Fig. 1B,C). Signals were first detectable at the late two-cell stage and became stronger at subsequent stages (Fig. 1B,C). At the blastocyst stage, DPPA2 and DPPA4 were mostly located in the inner cell mass (i.e. NANOG positive) rather than trophectoderm (i.e. CDX2-positive cells) (Fig. 1D). This observation is consistent with previous RNA in situ hybridization experiments showing that Dppa2 and Dppa4 are restricted to the inner cell mass at this stage (Maldonado-Saldivia et al., 2007).

The immunostaining results are incompatible with a potential role of DPPA2 and DPPA4 in Dux activation and ZGA in embryos. However, it is also possible that the very low levels of maternal DPPA2 and DPPA4 proteins in oocytes (barely detected by immunostaining) may still play a role in activating Dux and ZGA. To test this possibility, we generated Gdf9-Cre-mediated oocyte-specific conditional KO (CKO) models for Dppa2 (exon 3-4 floxed) (Fig. 2A) and Dppa4 (exon 2-7 floxed) (Fig. 2B). The Gdf9-Cre is expressed in early growing oocytes around postnatal day 3 (Lan et al., 2004). As the Dppa4 flox (fl) allele has been previously established and described (Nakamura et al., 2011), we only characterized in detail the Dppa2 fl allele that was generated in this study using the two-cell homologous recombination (2C-HR)-CRISPR method (Gu et al., 2018) (Fig. S1A,B). Sanger sequencing analyses confirmed that exons 3 and 4 of Dppa2 were successfully depleted in the CKO oocytes (Fig. S1C), resulting in a frameshift disrupting both the SAP and C-terminal domains. Because Dppa2 and Dppa4 are closely linked on the same chromosome, it is not feasible to generate Dppa2 and Dppa4 double CKO mice by natural mating.

We next sought to confirm successful knock out of DPPA2 and DPPA4 at the protein level. Because of the weak DPPA2 and DPPA4 immunostaining signals before the four-cell stage (Fig. 1B), it was challenging to determine their reduction in CKO oocytes and maternal KO one-cell/two-cell embryos. To circumvent this issue, immunostaining analyses were performed at the four-cell stage for maternal KO (m−z+) and maternal-zygotic KO (m−z−) embryos generated by crossing CKO female mice with heterozygous male mice (Fig. 2C). DPPA2 and DPPA4 signal intensities in maternal KO four-cell embryos were largely comparable to that of wild-type (WT) embryos (Fig. 2C, Fig. 1B). This suggests that Dppa2 and Dppa4 were zygotically expressed from the WT paternal alleles, which compensated for the maternal loss. In contrast, DPPA2 and DPPA4 signals were not detectable in m−z−Dppa2 and m−z−Dppa4 four-cell embryos, respectively (Fig. 2C,D), confirming the successful knock out of these proteins. It is worth noting that loss of one protein caused reduced signal of the other in maternal-zygotic KO four-cell embryos, albeit to different extents (Fig. 2C,D). This is perhaps due to the fact that DPPA2 and DPPA4 function as a heterodimer (Nakamura et al., 2011; Hernandez et al., 2018) and loss of one may therefore affect the stability of the other, as has been observed in ESCs (Gretarsson and Hackett, 2020).

Having confirmed the successful knock out of DPPA2 and DPPA4 in embryos, we next examined the impact of DPPA2 and DPPA4 knock out on pre-implantation development. To this end, spermatozoa from WT male mice were used to fertilize oocytes from control and CKO female mice in vitro, generating WT (m+z+) and maternal KO (m−z+) embryos. To exclude the possibility of the WT paternal copy compensating for the maternal losses, CKO oocytes were also fertilized with heterozygous spermatozoa, which should generate maternal-zygotic KO (m−z−) in half of the embryos. Unexpectedly, none of the embryo groups showed apparent pre-implantation defects and m−z− blastocysts were identified by immunostaining at the expected Mendelian ratio (Fig. 3A,B). It should be noted that loss of both DPPA2 and DPPA4 in m−z−Dppa4 blastocysts makes these embryos equivalent to Dppa2/4 double KO (Fig. 3B). Importantly, analyses of in vivo blastocysts collected after natural mating also led to the same observations (Fig. 3C,D, Fig. S2A). In summary, these results indicate that DPPA2 and DPPA4 are not required for mouse pre-implantation development.

Having confirmed the dispensable role of DPPA2/4 in pre-implantation development, we next examined whether they are required for post-implantation development. Mating analyses revealed that, although maternal DPPA2 or DPPA4 is not required for development, few m−z−Dppa2 and m−z−Dppa4 developed to weaning stage (Fig. S2B). Lethality of m−z− mutants should occur around or after birth because a close-to-Mendelian ratio was observed at embryonic day (E) 18.5 (Fig. S2B). However, the m−z− mutants already showed some phenotypes, including smaller sizes and/or pale skins by E18.5 (Fig. S2C). These data suggest that loss of maternal-zygotic DPPA2 or DPPA4 causes perinatal lethality, which is very similar to the effect of zygotic knockouts of DPPA2 and/or DPPA4 previously reported (Madan et al., 2009; Nakamura et al., 2011). Thus, DPPA2 and/or DPPA4 are essential for post-implantation, but not pre-implantation, development.

Given that mouse pre-implantation development is largely normal without DPPA2 or DPPA4, minimal ZGA defects are expected in these mutants. To confirm this prediction, we performed RNA-seq experiments. To determine whether DPPA2 and DPPA4 initiate Dux transcription during minor ZGA and subsequently affect major ZGA, late one-cell and late two-cell embryos of control and maternal KO were collected for RNA-seq analyses (Fig. S3A). Maternal-zygotic KO single two-cell embryos were also analyzed to exclude the possibility of the WT paternal allele in m−z+ embryos compensating for the maternal loss (Fig. S3B). All RNA-seq biological replicates were highly reproducible (Fig. S3A,B), and the RNA-seq genome browser views confirmed the success of Cre-mediated depletion of Dppa2 and Dppa4 in the m−z+ late one-cell (Fig. S4A) and m−z− late two-cell embryos (Fig. 4A).

We next performed comparative analyses in m+z+, m−z+, and m−z− embryos to identify differentially expressed genes [fragments per kilobase of transcript per million mapped reads (FPKM) >1, fold change (FC) >2, and adjusted P-value<0.05]. As expected, minimal changes in gene/repeat expression were observed in both m−z+ and m−z− mutant embryos (Fig. S4B,C, Fig. 4B, Tables S1 and S2). Note that both Dux and its target genes/repeats, such as Zscan4, Zfp352, MERVL-int and MT2_Mm were normally activated during minor and major ZGA. In addition, the young LINE-1 elements, including L1Md_A and L1Md_T, which are regulated by DPPA2 and DPPA4 independently of DUX in ESCs (De Iaco et al., 2019), were also normally expressed in late two-cell embryos (Fig. S4C, Fig. 4B). Consistent with minimal transcriptome alterations, the major ZGA genes (n=2470, two-cell/one-cell: FC>5, FPKM>3, adjusted P-value<0.05), including those previously reported to be regulated by DPPA2 and DPPA4 in ESCs, also showed normal activation (Fig. 4C,D, Fig. S4D). Thus, our data support the suggestion that DPPA2 and DPPA4 are dispensable for Dux expression and ZGA in mouse early embryos.

Collectively, our data provide definite evidence that maternal DPPA2 and DPPA4 are not required to trigger the activation of Dux and other two-cell embryo-specific genes during mouse ZGA. Although generation of double KOs were not feasible by natural mating owing to their close genetic linkage, DPPA2 and DPPA4 should not compensate for each other for the following reasons. First, DPPA2 and DPPA4 function as a heterodimer (Nakamura et al., 2011; Hernandez et al., 2018). Loss of either protein causes comparable phenotypes to the double KOs during both ESC differentiation (i.e. failure of developmental gene activation) (Eckersley-Maslin et al., 2020; Gretarsson and Hackett, 2020) and embryogenesis (i.e. lung developmental defects and perinatal lethality) (Madan et al., 2009; Nakamura et al., 2011). Second, our immunostaining analyses indicated that DPPA2 became almost undetectable when DPPA4 was depleted in early embryos (Fig. 2C and Fig. 3B,D), suggesting that similar results are expected in the maternal-zygotic double KO. Therefore, compensation for each other should not explain the lack of apparent pre-implantation phenotype for the Dppa2 and Dppa4 mutants analyzed in this study.

Together with the evidence that DUX does not initiate ZGA in embryos (Chen and Zhang, 2019; Guo et al., 2019; De Iaco et al., 2020; Bosnakovski et al., 2021), this study further highlights the key differences between 2C-like cells and two-cell embryos. In ESCs, both DUX and the DPPA2/4 heterodimer are essential for establishing the 2C-like state (De Iaco et al., 2017; Hendrickson et al., 2017; De Iaco et al., 2019; Eckersley-Maslin et al., 2019; Yan et al., 2019). However, this is not the case in mouse embryos. Therefore, conclusions drawn from 2C-lilke cells should be carefully considered before being applied to the embryo scenario. Recently, TP53 (TRP53 in mouse) has been identified as a maternal factor that regulates DUX and 2C genes in both ESCs and embryos (Grow et al., 2021; Sun et al., 2021). Nonetheless, Dux also gets activated in Tp53 maternal-zygotic KO embryos, although to a lesser extent than in WT (Grow et al., 2021). Thus, multiple pioneer factors may exist to trigger Dux and/or other minor ZGA genes (Kobayashi and Tachibana, 2021).

Despite DPPA2 and DPPA4 being dispensable for ZGA, they may be required for maintaining a permissive chromatin state during gastrulation by counteracting DNA methylation, as suggested by the studies in ESCs (Eckersley-Maslin, 2020; Eckersley-Maslin et al., 2020; Gretarsson and Hackett, 2020). Indeed, our data revealed a perinatal lethality phenotype in Dppa2 and Dppa4 maternal-zygotic KOs, which is largely similar to the previously reported Dppa2 and/or Dppa4 zygotic KO mutants. It is likely that loss of DPPA2 and DPPA4 may cause epigenomic defects around gastrulation, which ultimately contribute to the perinatal lethality phenotype (Madan et al., 2009; Nakamura et al., 2011). This hypothesis warrants further examination.

All animal studies were performed in accordance with guidelines of the Institutional Animal Care and Use Committee at Harvard Medical School. The procedures of GV and MII oocytes collection and in vitro fertilization were described previously (Chen and Zhang, 2019; Zhang et al., 2020). For all experiments, 6- to 9-week-old mice were used. The in vitro-fertilized embryos were cultured in KSOM (Millipore) at 37°C under 5% CO₂ with air. The in vivo blastocysts were collected by flushing reproductive tracts at E3.5. The day of vaginal plug was counted as E0.5.

The Gdf9-Cre transgenic line and the Dppa4 fl line were described previously (Lan et al., 2004; Nakamura et al., 2011). The Dppa2 fl allele was generated by the 2C-HR-CRISPR method (Gu et al., 2018). Specifically, Cas9 mRNA (100 ng/μl), two sgRNAs (80 ng/μl each), and donor DNA PCR fragment (no biotin) (25 ng/μl) were co-injected into the cytoplasm of each two-cell blastomere using a Piezo impact-driven micromanipulator (Prime Tech). A PCR fragment with ∼0.9-1 kb homologous arms was used because high knock-in efficiency was reported for this donor DNA preparation method (Yao et al., 2018). After injection, the embryos were cultured for a few hours before being transferred into oviducts of surrogate ICR strain mothers. The synthesis of Cas9 mRNA and sgRNA was described previously (Wang et al., 2013). The donor DNA targeting vector was cloned using the Gibson Assembly method and the primers used for cloning are listed in Table S3. The Dppa2 fl F₀ mice (BDF1×BDF1) (The Jackson Laboratory, #100006) were crossed with B6 (The Jackson Laboratory, #000664) to confirm germline transmission.

The breeding schemes were the same for Dppa2 and Dppa4 lines. Specifically, the +/fl lines were crossed with Gdf9-Cre to obtain Gdf9-Cre, +/fl males. The Gdf9-Cre, +/fl males were then crossed with +/fl females to obtain Gdf9-Cre, fl/fl males and fl/fl females. They were then crossed to generate fl/fl (control) and Gdf-Cre, fl/fl females (CKO) for experiments. Male mice that were heterozygous for Cre-mediated depletions were obtained by crossing Gdf9-Cre, +/fl females with WT B6 males. For all mouse lines, tail tips were used for genotyping using the primers listed in Table S3.

Immunostaining, image acquisition, and analyses were performed as previously described (Inoue et al., 2018). Primary and secondary antibodies are listed in Table S3.

Reverse-stranded total RNA-seq libraries (Fig. S4, Fig. S3A) were prepared using the SMARTer-Seq Stranded kit (Takara Bio) following the manufacturer's instructions. For single-embryo RNA-seq (Fig. 4, Fig. S3B), cDNA was synthesized using the SMARTer Ultra low Input RNA cDNA preparation kit (Takara Bio). The cDNA was then used for genotyping by quantitative PCR (primers in Table S3) and three embryos for each genotype (i.e. m−z+ or m−z−) were selected for library construction using the Nextera XT DNA Library Preparation Kit (Illumina). The single-embryo RNA-seq libraries were non-stranded and only PolyA+ RNA was captured. For all RNA-seq libraries, paired-end 75-bp sequencing was performed on a NextSeq 550 sequencer (Illumina). A summary of the generated data sets is available in Table S2.

The total RNA-seq (Fig. S4) and polyA RNA-seq (Fig. 4) data were processed following the pipelines as previously described (Chen and Zhang, 2019; Chen et al., 2021). Briefly, RNA-seq reads were trimmed to remove low quality reads and adaptors using Trimgalore (v0.4.5) before being aligned to mm10 assembly using HISAT2 (v.2.1.0) (Kim et al., 2015). StringTie (v1.3.3b) was used to quantify gene FPKM values (Pertea et al., 2016). For differential expression analyses, DESeq2 (v1.24.0) (Love et al., 2014) was used to compute adjusted P-values on read counts generated by TEtranscripts (v. 2.1.4) (Jin et al., 2015). TEtranscripts summarizes only uniquely aligned reads for genes and both uniquely and ambiguously (i.e. due to multiple insertions of repeats) mapped reads for transposable elements. RNA-seq pipeline and data processing R-codes are available at Github (https://github.com/YiZhang-lab/Nonessential_role_of_Dppa2_4_in_ZGA).

The RNA-seq data generated in this study have been deposited in the Gene Expression Omnibus under accession number GSE181723. The RNA-seq data of mouse oocytes and early embryos (Fig. 1A) were from GSE66582 (Wu et al., 2016) and GSE76505 (Zhang et al., 2018). The RNA-seq data of ESCs presented in Fig. 4 and Fig. S4 were from GSE120952 (Eckersley-Maslin et al., 2019), GSE126621 (De Iaco et al., 2019) and GSE127811 (Yan et al., 2019).

Statistical analyses were performed in R (www.r-project.org/). All sequencing tracks were visualized using the UCSC genome browser (Kent et al., 2002).

We thank Dr Yota Hagihara for his help in some immunostaining experiments; Drs Chunxia Zhang, Wenhao Zhang, Yota Hagihara and Cheng-Jie Zhou for critical reading of the manuscript.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.200178.