Some of the earliest transcripts produced in fertilized human and mouse oocytes code for DUX, a double homeodomain protein that promotes embryonic genome activation (EGA). Deleting Dux by genome editing at the one- to two-cell stage in the mouse impairs EGA and blastocyst maturation. Here, we demonstrate that mice carrying homozygous Dux deletions display markedly reduced expression of DUX target genes and defects in both pre- and post-implantation development, with, notably, a disruption of the pace of the first few cell divisions and significant rates of late embryonic mortality. However, some Dux−/− embryos give rise to viable pups, indicating that DUX is important but not strictly essential for embryogenesis.
Fertilization of the vertebrate oocyte is followed by transcription of the parental genomes, a process known as zygotic or embryonic genome activation (ZGA or EGA) (Jukam et al., 2017). In zebrafish and Drosophila, maternally inherited transcription factors are responsible for this event (Lee et al., 2013; Liang et al., 2008). In placental mammals, the EGA transcriptional program is directly activated at or after the two-cell (2C) stage by a family of transcription factors expressed after fertilization, the DUX proteins (De Iaco et al., 2017; Hendrickson et al., 2017; Whiddon et al., 2017). Recent studies suggest that DPPA2 and DPPA4 are maternal factors responsible in the mouse for DUX expression and downstream target activation, although this model still needs to be validated in vivo (De Iaco et al., 2019; Eckersley-Maslin et al., 2019). Forced expression of DUX proteins in murine or human cell lines triggers the aberrant activation of EGA-restricted genes. Conversely, deleting Dux by CRISPR-mediated genome editing before the two-cell stage in murine embryos leads to reduced expression of DUX targets such as Zscan4 and the transposable element (TE) MERVL and severe defects in early development, with many embryos failing to reach the morula/blastocyst stage (De Iaco et al., 2017). However, this procedure also yields some viable mice carrying heterozygous Dux deletions. Here, we demonstrate that crossing these Dux+/− animals results in Dux−/− embryos with impaired EGA and severe but not uniformly fatal defects in early development.
RESULTS AND DISCUSSION
The murine Dux gene is found in tandem repeats of variable lengths in so-called macrosatellite repeats (Leidenroth et al., 2012). We injected zygotes collected from B6D2F1 mothers with sgRNAs directed at sequences flanking the Dux locus (Fig. 1A,B), and transferred the resulting products into pseudo-pregnant B6CBA mothers. One out of 42 pups carried a mono-allelic deletion of the targeted region (Dux+/−) validated by Sanger sequencing of the junction. This animal was backcrossed twice with wild-type (WT) B6D2F1 mice to ensure germline transmission of the mutation. The resulting Dux+/− mice were healthy and did not display any macroscopic phenotype.
Transcription of Dux normally starts in zygotes just after fertilization and stops a few hours later (De Iaco et al., 2017), suggesting that the presence of a functional Dux allele is not necessary in germ cells. In our previous work, we demonstrated that inhibition of DUX expression in zygotes impairs early embryonic development. To characterize further the role of DUX, Dux+/− mice were crossed and the frequency of Dux mono- and bi-allelic deletions was determined in the progeny (Table 1). There was only a minor deviation from a Mendelian distribution of genotypes, with a slightly lower than expected frequency of Dux−/− pups. Furthermore, adult Dux−/− mice were healthy and had a normal lifespan. To ensure that Dux was not expressed from some other genomic locus, the absence of its transcripts was verified in testis of Dux−/− mice, because this is an adult tissue where these RNAs are normally detected (Snider et al., 2010) (Fig. 1C).
To explore further the role of DUX in pre-implantation embryos, we compared the size of litters yielded by isogenic Dux+/+ or Dux−/− crossings (Table 2, Fig. 1D). Crosses between Dux−/− mice led to strong reductions in litter size and delayed delivery, and some of the rare pups were eaten by their mother after delivery, probably because they were either stillborn or exhibited physical impairments. Furthermore, some Dux−/− females failed to give any pup, even when crossed with Dux−/− males that had previously demonstrated their fertility when bred with other Dux−/− females (not shown). Because this subgroup of Dux−/− females visually appeared pregnant at the time of expected delivery, we bred them with the same Dux−/− males and examined their uterus at embryonic day (E) 18.5 (Fig. 1E). Surprisingly, we found a significant number of macroscopically normal embryos, suggesting that death occurred around birth. Interestingly, the perinatal lethality was rescued when Dux−/− females were crossed with wild-type males (Fig. 1F). To exclude a role of paternal Dux in embryonic development, we also bred wild-type females with Dux−/− males and found normal litter size (Fig. 1G).
We then analyzed whether the strong lethality observed after Dux−/−×Dux−/− crosses occurred before implantation. For this, we repeated isogenic crosses of WT or Dux−/− mice, retrieved the zygotes at E0.5 (27 embryos from three WT×WT and 42 embryos from five Dux−/−×Dux−/− crosses), and monitored their ex vivo development for 4 days (Fig. 2A). We found that starting at E1.5, Dux−/− embryos divided faster than their WT counterparts, yet sometimes unevenly, with formation of three-cell (3C) structures (Fig. 2B). At E2.0, WT embryos caught up whereas Dux−/− embryos seemed partially blocked, and exhibited a clear delay at E3.5 with significantly reduced blastocyst formation. By E4.5, only 65% Dux−/− embryos reached the blastocyst stage, compared with 100% for WT. Confirming these findings, examination of E3.5 embryos from WT×WT or Dux−/−×Dux−/− crosses revealed a strong delay in blastocyst formation and increased levels of lethality in the absence of DUX (Fig. 2C,D). In conclusion, a subset of embryos derived from Dux−/− crosses fails to implant, and the rest generally die around birth.
Finally, we tested the consequences of lack of zygotic DUX on the transcriptional program of 2C-stage embryos. We collected 15 zygotes from three heterozygous Dux+/−×Dux+/− crosses, incubated them in vitro and collected RNA 5 h after the formation of 2C embryos. The transcriptomic analysis revealed three embryos with undetectable levels of Dux transcripts, indicating that they most likely were Dux−/− (Fig. 3A). Interestingly, the three Dux RNA-depleted 2C embryos exhibited significant reductions in the expression of genes and TEs (MERVL-int) previously identified as DUX targets, but not of other ZGA-specific genes (Fig. 3B,C) (De Iaco et al., 2017). We then bred one WT and one Dux−/− female with males from the same genetic background, and compared the transcription of putative DUX target genes and TEs in the two resulting 2C embryos. Products of the Dux−/−×Dux−/− crosses displayed a strong decrease in the expression of Dux and candidate DUX target genes and TEs (Fig. 3D-F). A mild loss of expression of 2C genes that were previously shown to be independent of DUX in mouse embryonic stem cells (mESCs) was also detected.
We further validated these results by collecting 17 zygotes from three heterozygous Dux+/−×Dux+/− crosses, and analyzing gene expression by qPCR (Fig. 4A). All three Dux RNA-depleted 2C embryos exhibited significant reductions in the expression of some (MERVL, Zscan4, Eif1a, Usp17la, B020004J07Rik, Tdpoz4 and Cml2), but not all Duxbl, Sp110, Zfp352 genes previously suggested to represent DUX targets (De Iaco et al., 2017). We then analyzed RNA from seven WT and 11 Dux−/− 2C embryos derived from breeding two WT and three Dux−/− females with males from the same genetic background (Fig. 4B). Products of the Dux−/−×Dux −/− crosses displayed a clear decrease in the expression of a subset of candidate DUX targets (MERVL, Zscan4, Eif1a, Usp17la, B020004J07Rik), whereas others (Tdpoz4, Cml2, Duxbl, Sp110, Zfp352) were again unaffected.
In summary, the present work confirms that DUX promotes murine embryonic development. In spite of also surprisingly demonstrating that this factor is not absolutely essential for this process, it further reveals that DUX depletion results in a variable combination of pre- and post-implantation defects, the consequences of which also appear to be cumulative over generations. DUX-devoid embryos originated from homozygous knockout (KO) breeding displayed deregulations in the timing and the ordinance of the first few cell divisions, various degrees of impairments in their ability to become blastocysts, and, for those reaching that stage, high levels of perinatal mortality. Nevertheless, these defects became truly apparent only when DUX was absent already in the oocyte, as the frequency of Dux−/− pups derived from the crossing of heterozygous Dux+/− parents was only slightly below a Mendelian distribution whereas the resulting Dux−/− females yielded markedly reduced progenies, some even appearing sterile when crossed with Dux−/− males. However, this defect was completely rescued by zygotic expression of Dux, as breeding these Dux−/− females with WT males resulted in the production of normal-size litters of pups devoid of obvious defects. Thus, the presence of DUX during only a few hours after fertilization appears to condition not only the conduct of the first few embryonic cell divisions, but also to bear consequences that extend well beyond the pre-implantation period, long after Dux transcripts have become undetectable. Our data are in line with the results of two recent studies, both of which found that Dux is necessary for normal fertility but is not absolutely essential for mouse development (Chen and Zhang, 2019; Guo et al., 2019). However, while one of these studies (Guo et al., 2019) documented ZGA abnormalities closely resembling those observed in our work, the other (Chen and Zhang, 2019) detected only minimal transcriptional disturbances at this developmental stage. The bases for these differences are unknown, but our finding that the consequences of Dux depletion vary from pup to pup and are cumulative over generations indicate both stochasticity in the observed phenotype and its attenuation by compensatory factors that have yet to be identified. Deleting the Dux inducers Dppa2 or Dppa4 also results in perinatal lethality (Madan et al., 2009; Nakamura et al., 2011), but in this case defects in lung and skeletal development are observed, which correlate with the expression of these two genes later in embryogenesis. Future studies should therefore attempt to characterize better the molecular defects induced by DUX depletion, to explain how the full impact of the Dux KO phenotype is only expressed at the second generation, and how even at that point it can be fully rescued by paternally encoded Dux zygotic expression.
MATERIALS AND METHODS
Two single guide RNAs (sgRNAs) targeting sequences flanking the Dux macrosatellite repeat (Fig. 1A) were cloned into px330 using a standard protocol. The primers used to clone the sgRNAs have been previously described (De Iaco et al., 2017).
Generation of transgenic mice carrying Dux−/− alleles
Pronuclear injection was performed according to the standard protocol of the Transgenic Core Facility of EPFL. In summary, B6D2F1 mice were used as egg donors (6 weeks old). Mice were injected with pregnant mare serum gonadotropin (10 IU), and human chorionic gonadotropin (10 IU) 48 h after. After mating females overnight with B6D2F1 males, zygotes were collected and kept in KSOM medium pre-gassed in 5% CO2 at 37°C. Embryos were then transferred to M2 medium and microinjected with 10 ng/μg of px330 plasmids encoding for Cas9 and the appropriate sgRNAs diluted in injection buffer (10 mM Tris HCl pH 7.5, 0.1 mM EDTA pH 8, 100 mM NaCl). After microinjection, embryos were re-implanted in pseudopregnant B6CBA mothers. The pups delivered were genotyped for Dux null alleles using previously described primers (De Iaco et al., 2017). The mouse carrying the Dux null allele was then bred with B6D2F1 mice to ensure that the transgenic allele reached the germ line and to dilute out any randomly integrated Cas9 transgene. This process was repeated once again to obtain second filial generation (F2) Dux−/+ mice.
Mice between 6 and 25 weeks old were used for breeding experiments.
Monitoring of pre-implantation embryos
Mothers were superovulated and mated with males as described above. Zygotes were collected and cultured in KSOM medium at 37°C in 5% CO2 for 4 days. Each embryo was monitored every 12 h to determine the stage of development. When 2C were used for RNA-seq or qPCR, zygotes were monitored every hour until cell division and 5 h later were collected for further analysis. Time-lapse experiments of pre-implantation embryos were carried out in 96-well plates using Operetta CLS High-Content Analysis System for image acquisition. Embryos were checked from zygote to morula every hour.
Randomization and blind outcome assessment were not applied. All animal experiments were approved by the local veterinary office and carried out in accordance with the EU Directive (2010/63/EU) for the care and use of laboratory animals.
Standard PCR, RT-PCR and RNA sequencing
For genotyping the Dux null allele, genomic DNA was extracted with DNeasy Blood & Tissue Kits (Qiagen) and the specific PCR products were amplified using PCR Master Mix 2X (Thermo Scientific) combined with the appropriate primers (design in Fig. 1A, previously described; De Iaco et al., 2017). Ambion Single Cell-to-CT kit (Thermo Fisher) was used for RNA extraction, cDNA conversion and mRNA pre-amplification of 2C-stage embryos. Primers (previously listed) were used for SYBR green qPCR (Applied Biosystems) (De Iaco et al., 2017). RNA from 2C embryos was amplified using Smart Seq V4 Ultra Low Input RNA kit (Takara) and library were prepared using Nextera XT DNA Library Prep Kit (Illumina).
RNA-seq dataset processing
RNA-seq of mouse embryo samples was mapped to mm9 genome using hisat2 aligner (Kim et al., 2015) for unstranded and paired-end data with options -k 5 --seed 42 -p 4. Counts on genes and TEs were generated using featureCounts (Liao et al., 2014) with options -p -T 4 -t exon –g.gene_id -Q 10, using a gtf file containing both genes and TEs to avoid ambiguity when assigning reads. For repetitive sequences, an in-house curated version of the mm9 open-3.2.8 version of the Repeatmasker database was used (fragmented LTR and internal segments belonging to a single integrant were merged). Only uniquely mapped reads were used for counting on genes and TEs.
Normalization for sequencing depth and differential gene expression analysis was performed using the TMM method as implemented in the limma package of Bioconductor (Gentleman et al., 2004), using the counts on genes as library size. TEs overlapping exons or having fewer than one read per sample on average were removed from the analysis. To compute total number of reads per TE family/subfamily, counts on all integrants were summed using multi-mapping read counts with fractions (featureCounts with options -M --fraction -p -T 4 -t exon -g gene_id -Q 0) to compensate for potential bias in repetitive elements.
Differential gene expression analysis was performed using voom (Law et al., 2014) as implemented in the limma package of Bioconductor. A gene (or TE) was considered to be differentially expressed when the fold change between groups was greater than two and the P-value was less than 0.05. A moderated t-test (as implemented in the limma package of R) was used to test significance. P-values were corrected for multiple testing using the Benjamini–Hochberg method.
We thank the Transgenic Core Facility of EPFL for technical assistance.
Conceptualization: A.D.I., D.T.; Methodology: A.D.I., S.V., S.O., D.G.; Investigation: A.D.I., S.V., S.O.; Data curation: A.D.I., D.G.; Writing - original draft: A.D.I.; Writing - review & editing: D.T.; Supervision: D.T.; Funding acquisition: D.T.
This work was supported by grants from the European Research Council (KRABnKAP, No. 268721; Transpos-X, No. 694658) and the Swiss National Science Foundation (Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung).
RNA-seq data have been deposited in Gene Expression Omnibus under accession number GSE141321.
The authors declare no competing or financial interests.