Dear Editor,

Normal mammalian development requires participation of both maternal and paternal genomes because of the existence of genomic imprinting, whereas the gynogenetic and androgenetic embryos die shortly after implantation1,2,3. Generation of gynogenetic bimaternal mice containing two sets of maternal genomes was achieved using non-growing oocytes with imprinting modifications4; however, the approach was technically challenging and impractical for further applications. Recently, we and others have derived mammalian androgenetic and parthenogenetic haploid embryonic stem cells (ahESCs and phESCs), and showed that ahESCs can replace gametes to produce offspring5,6,7, which provided alternative resources for reproduction6,7. Here we report that after proper imprinting modifications, the mouse phESCs can efficiently produce viable fertile offspring upon intracytoplasmic injection into MII oocytes. We thus establish a novel strategy of generating bimaternal mammals, which is valuable to uncover the function of genomic imprinting, and to improve assisted reproduction in diverse mammalian species.

Since our previous study showed that the MII oocyte-derived phESCs lost maternal imprinting in differentially methylated regions (DMRs) during culture8, we hypothesized that bimaternal embryos generated by injecting the haploid phESCs into MII oocytes would have improved developmental potential compared with parthenogenetic embryos, which is known to die before embryonic day 9.5 (E9.5). To investigate the potential of phESCs to produce bimaternal mice, we first established phESC lines from mouse haploid parthenogenetic blastocysts carrying a constitutively expressed green fluorescent protein (Gfp) gene (Supplementary information, Figure S1A). The haploidy of phESC lines could be maintained by fluorescence-activated cell sorting (FACS)-based purification every 4-6 passages (Supplementary information, Figure S1B-S1D). The phESC lines expressed typical pluripotency marker genes (Supplementary information, Figure S1E), formed teratomas containing all three germ layers (Supplementary information, Figure S1F), and produced chimeric mice (Supplementary information, Figure S1G), indicating that they were pluripotent. Consistent with our hypothesis, the phESC-derived bimaternal embryos could develop beyond E10.5, but none developed beyond E13.5 (Supplementary information, Table S1A). One live E13.5 phESC-derived bimaternal fetus was generated (Figure 1A and Supplementary information, Figure S1H). The reduced representation bisulfite sequencing (RRBS) analysis revealed that the maternally methylated DMRs showed similar methylation levels between wild-type (WT) embryos and the phESC-derived bimaternal embryo, but all 3 paternally methylated DMRs (H19-, IG- and Rasgrf1-DMRs) were significantly hypomethylated in the bimaternal embryo compared with WT embryos (Figure 1B). Consistently, transcriptome analysis showed that only 8 of all imprinted genes had more than 2-fold difference of expression levels, including the upregulation of maternally expressed H19 and Rian, and the downregulation of paternally expressed Igf2 and Igf2as in the bimaternal embryo; although the expression level of Rasgrf1 was very low at E13.5, it was also downregulated in the bimaternal embryo (Figure 1C and 1D). The global gene expression profiles exhibited high correlation (R2 = 0.94) between the bimaternal embryo and WT control (Figure 1C). These results suggest that the aberrant expression of imprinted genes within the 3 paternally methylated imprinted clusters, H19-Igf2, Dlk1-Dio3 and Rasgrf1, likely contributed to the developmental defects of phESC-derived bimaternal embryos.

Figure 1
figure 1

Birth of fertile bimaternal offspring following intracytoplasmic injection of phESCs. (A) Morphology of a live E13.5 bimaternal embryo. Scale bar, 2 mm. (B) Heat map of the methylation level of ICRs in bimaternal and WT E13.5 fetuses. DMRs with significantly different methylation level between bimaternal and WT groups were marked with asterisk (P < 0.05). (C) Scatter plot comparison of transcriptomes of bimaternal and WT E13.5 fetuses. The raw FPKM for each gene was transformed to log2 value, and genes with more than 1 FPKM were shown. The red and blue points represent imprinted genes expressed from maternal and paternal alleles, respectively. Dashed lines depict 2-fold changes. The R2 was determined by Pearson's correlation. (D) Five imprinted genes with more than 2-fold changes in expression levels between bimaternal and WT embryos were shown. (E) A full-term bimaternal pup derived by oocyte injection of phESCΔIG. (F) The body and placenta weights of phESCΔIG pups, showing severe retardation compared with WT pups. ***P < 0.001. (G) Bisulfite analysis of H19-, Igf2r- and Snrpn-DMRs in phESCΔIG pups. (H) RRBS analysis of whole brain of phESCΔIG and WT pups. (I) Survived bimaternal pups derived by oocyte injection of phESCΔIG/Δ15H19. (J) The body and placenta weights of phESCΔIG/Δ5H19, phESCΔIG/Δ15H19 and WT pups. ***P < 0.001. (K) Gene expresson analysis showing reduced expression of Igf2 in phESCΔIG and phESCΔIG/Δ5H19 pups, and its normal expression in phESCΔIG/Δ15H19 pups. (L) Survived bimaternal pups produced by oocyte injection of phESCΔIG/Δ5H19/Igf2. (M) The body and placenta weights of phESCΔIG/Δ5H19/Igf2 and WT pups. (N) Body weight of the bimaternal mice 1-7 weeks after birth, showing 20% growth retardation compared with WT ones (WT female, red, n = 8; phESCΔIG/Δ15H19 mice, blue, n = 8; phESCΔIG/Δ5H19/Igf2 mice, black, n = 8). The values represent the means ± SEM. (O) An adult bimaternal mouse and its offspring. (P) Corrected litter size of bimaternal mouse (5.8 ± 0.9, n = 9) was comparable with the WT (6.2 ± 0.70, n = 9). P = 0.70. (Q) Summary of viability of bimaternal mice produced from phESCs carrying different imprinting modifications. Y, viable; N, non-viable.

To improve the development of phESC-produced gynogenetic embryos, we modified the imprinting status of the H19-Igf2 and Dlk1-Dio3 clusters that controlled fetal growth by knocking out the DMRs, which may cause bidirectional loss of imprinting of all genes within the clusters. According to previous studies9,10,11,12,13, three types of genetic deletions were designed, including a 5 kb deletion of the H19-Igf2 locus (from −4 kb to +1 kb of H19 transcription start site (TSS)), a 15 kb deletion (from −10 kb to +4 kb of H19 TSS) spanning the H19 gene, H19-ICR (from −4 kb to −2 kb of H19 TSS) and the upstream region of H19-ICR, and a 4 kb deletion spanning the IG-DMR of the Dlk1-Dio3 cluster (Supplementary information, Figure S2A and S2E). To boost the efficiency of homologous recombination-mediated genetic knockout, we used the CRISPR/Cas9 system to introduce a targeted DNA double strand break in the knockout region. The correctly targeted clones were identified by PCR with specific primer pairs that could distinguish the WT and knockout alleles (Supplementary information, Figure S2B, S2F and Table S1C). Consistent with the genetic deletions, H19 expression was abolished in phESCs carrying the 5 kb deletion (phESCΔ5H19) or the 15 kb deletion (phESCΔ15H19) of the H19-Igf2 locus, and Gtl2 expression was silenced in phESCs carrying the IG-DMR deletion (phESCΔIG) (Supplementary information, Figure S2C and S2F). Then we purified the haploid cells of these genetically modified phESC lines by FACS and generated bimaternal embryos by MII oocyte injection to investigate the gynogenetic development (Supplementary information, Figure S2D and S2G). A total of 624 phESCΔ5H19 embryos from 2 phESCΔ5H19 sublines, and 611 phESCΔ15H19 embryos from 2 phESCΔ15H19 sublines were transferred into pseudopreganant mothers. None developed to term (E19.5), though some embryos developed beyond E13.5 (Supplementary information, Table S1A). However, of the 568 phESCΔIG embryos transplanted, 13 (2.3%) full-term pups were recovered by cesarean section (Supplementary information, Table S1A). These phESCΔIG pups were severely growth-retarded (0.57 ± 0.02 g, n = 10), with about 46% of normal body weight of WT pups (1.25 ± 0.04 g, n = 12; Figure 1E and 1F), and all died within several hours after birth. Comparison of global imprinting status between phESCΔIG and WT pups showed that of 16 DMRs recovered by RRBS, 15 maternally methylated DMRs had similar methylation levels, but the H19-DMR was hypomethylated in the phESCΔIG pups (Figure 1H and Supplementary information, Figure S2L). Bisulfite sequencing analysis confirmed the normal methylation level of maternally methylated Igf2r- and Snrpn-DMRs, and hypomethylation of H19-DMR (Figure 1G).

Next, we deleted either the 5 kb or the 15 kb H19-Igf2 region in the phESCΔIG lines to generate double knockout (DKO) phESC lines (phESCΔIG/Δ5H19 or phESCΔIG/Δ15H19). Of the 568 transplanted phESCΔIG/Δ5H19 embryos, 21 (3.7%) live full-term pups were recovered (Supplementary information, Table S1A). The phESCΔIG/Δ5H19 pups had similar body weight (0.57 ± 0.01 g, n = 13; 46% of WT body weight) to the phESCΔIG pups, and died soon after birth (Figure 1J). However, of the 475 transplanted phESCΔIG/Δ15H19 embryos, 43 (9.1%) live full-term pups were recovered, which had similar body weight (1.34 ± 0.03 g, n = 30) to the WT pups (Figure 1I, 1J and Supplementary information, Table S1A). Thirty-one phESCΔIG/Δ15H19 pups survived to adulthood, while the rest 12 died within 3 days after birth due to lack of nursing from the surrogate mothers (Supplementary information, Table S1A). Gene expression analysis showed that Igf2 expression level was much lower in phESCΔIG/Δ5H19 pups, but was comparable between phESCΔIG/Δ15H19 and WT pups, indicating that Igf2 expression was not activated by the 5 kb deletion of the H19-Igf2 locus (Figure 1K). To confirm whether the defect in Igf2 activation caused the growth retardation of the phESCΔIG/Δ5H19 pups, we knocked the exogenous Igf2 gene into the endogenous H19 gene locus in phESCsΔIG/Δ5H19 (phESCsΔIG/Δ5H19/Igf2) by replacing the “PGK-neo” cassette of the 5 kb-targeting vector with an “EF1α-Igf2-2A-Rfp-IRES-neo” cassette (Supplementary information, Figure S2I). phESCsΔIG/Δ5H19/Igf2 carried the same 5 kb deletion of the H19-Igf2 locus as the phESCsΔIG/Δ5H19, but they expressed the exogenous Igf2 and Rfp genes (Supplementary information, Figure S2J and S2K). We then injected the haploid phESCsΔIG/Δ5H19/Igf2 into MII oocytes and transplanted the embryos into the psudopregnant mice to test their developmental ability (Supplementary information, Figure S2L). Of the 450 transplanted embryos, 32 live full-term pups were generated (Figure 1L and Supplementary information, Table S1A). Interestingly, the phESCΔIG/Δ5H19/Igf2 pups (1.55 ± 0.04 g, n = 24) had similar size and body weight to the WT pups (Figure 1M). Twenty-three pups survived to adulthood, and the rest 9 pups died due to lack of nursing from the surrogate mothers (Supplementary information, Table S1A).

We analyzed the postnatal growth of the survived bimaternal mice by recording their body weight every week. Though born with normal body weight, the bimaternal mice were consistently 20% lighter than the WT controls since the first week (Figure 1N). Except for the growth retardation, no obvious abnormalities were observed in the bimaternal mice. The female bimaternal mice were mated with WT males to examine their fertility. A total of 29 offspring were delivered in 9 litters (Figure 1O). Four neonatal pups carrying an IG-DMR-knockout allele died soon after birth, and the rest 25 mice grew into adults (Supplementary information, Table S1B), consistent with a previous finding that mice with a maternally inherited IG-DMR deletion died at later gestational stages12. According to the Mendelian ratio, the real litter size of bimaternal mice was calculated as 5.8 ± 0.9, comparable to the litter size of WT females (6.2 ± 0.7), indicating that the bimaternal mice had normal fertility (Figure 1P).

In summary, here we show that the MII oocyte-derived phESCs can efficiently produce bimaternal mice through injection into WT MII oocytes after proper imprinting modifications. Compared to previous reports that used non-growing oocytes to achieve gynogenetic development4, our approach provides a more robust platform to study the developmental roles of genomic imprinting. For example, we showed that the IG-DMR deletion, but not H19-DMR deletion, was suffcient to produce full-term bimaternal pups, indicating that the Dlk1-Dio3 region has a more profound effect on fetal growth. Though both the 15 kb and 5 kb regions contained the ICR of the H19-Igf2 cluster, only the 15 kb deletion could reactivate Igf2 expression and support the adult development of bimaternal mice, suggesting that the H19-Igf2 imprinting was regulated not only by the methylation of ICR, but also by the relative distance and spatial organizations of different genomic elements in this region (Figure 1Q). In general, we have established a convenient and efficient approach to generate bimaternal mammals from imprinting-modified phESCs, which can provide a novel platform for genomic imprinting studies, and may also shed new light on mammalian reproduction. Given the successful establishment of phESCs in the mouse, rat and monkey, we expect that our approach can also be applied in other mammalian species to produce bimaternal animals.