DNA methylation analysis and editing in single mammalian oocytes

Significance

Many mammalian nongenetic diseases and developmental disorders originate from oocyte DNA methylation abnormalities. However, prevention and correction of this maternally transmitted nongenetic disorder remains challenging because of the lack of strategy that can evaluate and manipulate specific methylation at single oocyte level. In this study, using several models of epigenetic inheritance via oocytes, we have shown that specific methylation in a single oocyte can be evaluated from its sibling PB1 and can be edited in a targeted manner. Our study provides a strategy for prevention and correction of maternally transmitted nongenetic diseases or disorders, and will also facilitate the investigation of maternally transmitted nongenetic information at the very beginning of life.

Abstract

Mammalian oocytes carry specific nongenetic information, including DNA methylation to the next generation, which is important for development and disease. However, evaluation and manipulation of specific methylation for both functional analysis and therapeutic purposes remains challenging. Here, we demonstrate evaluation of specific methylation in single oocytes from its sibling first polar body (PB1) and manipulation of specific methylation in single oocytes by microinjection-mediated dCas9-based targeted methylation editing. We optimized a single-cell bisulfite sequencing approach with high efficiency and demonstrate that the PB1 carries similar methylation profiles at specific regions to its sibling oocyte. By bisulfite sequencing of a single PB1, the methylation information regarding agouti viable yellow (A^vy)-related coat color, as well as imprinting linked parthenogenetic development competency, in a single oocyte can be efficiently evaluated. Microinjection-based dCas9-Tet/Dnmt–mediated methylation editing allows targeted manipulation of specific methylation in single oocytes. By targeted methylation editing, we were able to reverse A^vy-related coat color, generate full-term development of bimaternal mice, and correct familial Angelman syndrome in a mouse model. Our work will facilitate the investigation of specific methylation events in oocytes and provides a strategy for prevention and correction of maternally transmitted nongenetic disease or disorders.

Mammalian oocytes carry not only genetic but also certain epigenetic information to the next generation (1–6). DNA methylation is one of the most important layers of the epigenome and plays critical roles in many biological processes (7, 8). Many mammalian diseases, such as imprinting diseases, nongenetic cancer, and diabetes, as well as many developmental failures and disorders, are caused by oocyte methylation abnormalities (2–4, 9, 10). One way to address this issue is to evaluate the disease- or disorder-linked specific methylation in single oocytes and correct it to the normal state.

As the female gametic cell, the mammalian oocyte has its unique characteristics. A mature oocyte contains a tiny first polar body (PB1), which includes a counterpart of genomic material to its sibling oocyte but is dispensable for subsequent development (11, 12). DNA methylation in oocytes is established during growth (13, 14) during which the ooplasm provides a comparable microenvironment to the oocyte and its sibling PB1 genome. The PB1 thus potentially carries a similar DNA methylome as its sibling oocyte. Previous studies showing that PB1 transfer can generate normal healthy offspring without obvious epigenetic abnormality supports this idea (15, 16). Therefore, methylation information associated with development and disease in single oocytes can be potentially evaluated from its sibling PB1. Moreover, recent technique breakthroughs have achieved targeted DNA methylation editing in mammalian cells (17–19). The characteristic that individual oocytes can be easily manipulated may enable DNA methylation editing at the single oocyte level.

Many mammalian nongenetic phenotypes are correlated with DNA methylation variation at specific regions in oocytes (2, 10, 20, 21). For example, agouti variable yellow (A^vy), in which an intracisternal A particle (IAP) retrotransposon is inserted upstream of the agouti locus, determines coat color in mice (22, 23). A^vy is classified as an epigenetic metastable allele (3, 23). This means that the degree of IAP methylation varies among individuals, causing coat color ranging from yellow to pseudoagouti (brown), and this phenotype variation is largely heritable following transmission of A^vy through the female germline (2, 21, 23). Parthenogenesis, a way of producing offspring solely from maternal genomes, is restricted in mammals due to imprinting disorders (24, 25). Mouse parthenogenetic embryos can only live to day 10 of gestation due to imprinting dysfunction (26, 27). Modified parthenogenetic embryos constructed with nongrowing (ng) and fully grown (fg) oocytes can develop to 13.5 d by appropriate methylation of many imprinted loci (28, 29), but do not develop further due to failure of two imprinted loci: the H19 imprinting control region (ICR) and the Gtl2 ICR, which is also known as the Dlk-Dio3 intergenic germline (IG)-derived differentially methylated region (DMR) (30, 31). Many human nongenetic disease and developmental disorders also originate from oocyte methylation abnormality at specific regions (2, 4, 9). For example, there are rare cases of familial Angelman syndrome (AS) caused by failure of maternal-specific imprinting at SNRPN (32).

In this study, we examined whether methylation information associated with nongenetic phenotypes or diseases in single oocytes can be evaluated from its sibling PB1, and whether it can be edited in a target manner. Using bisulfite sequencing of specific regions in a single PB1, we were able to evaluate the methylation information regarding the nongenetic phenotype of its sibling oocyte. Furthermore, by targeted methylation editing in single oocytes, we were able to manipulate coat color-related phenotypes, improve parthenogenetic development, and correct familial imprinting disorders in a mouse model.

Results

An Optimized Protocol for DNA Methylation Analysis at Specific Region in Single Oocytes.

To investigate DNA methylation of a specific region in single oocytes, we optimized a single-oocyte bisulfite-sequencing method that can efficiently measure DNA methylation of specific loci at the single-nucleotide and single-oocyte levels. In commonly used bisulfite-sequencing protocols, bisulfite conversion is achieved at high temperature and low pH conditions, which results in a high degree of DNA degradation. To protect a minute amount of DNA from degradation, we used tetrahydrofurfuryl alcohol as a protector during bisulfite conversion. We subsequently ligated the fragmented DNA with a nine-nucleotide adaptor to further protect the postbisulfite DNA and amplified the fragments three times to generate sufficient templates for PCR amplification of the targeted region (see SI Appendix, Note S1 for a detailed procedure). To validate this approach, we performed our optimized protocol on the Pou5f1 promoter from a single MII oocyte nucleus and observed a high cytosine conversion rate of over 99% (SI Appendix, Fig. S1A). For further validation, we applied our optimized protocol on a variety of elements distributed widely in the genome and still observed an over 99% conversion rate (SI Appendix, Fig. S1 B–D), demonstrating the high fidelity of this protocol. In total, we measured DNA methylation on 1,691 single oocyte nuclei or PB1 and successfully generated methylation profiles from 1,451 of them, with an efficiency of ∼85.8% (detailed information is summarized in Dataset S1). Therefore, our optimized protocol provides an accurate and efficient tool for identification of DNA methylation information at the single-oocyte level.

Similar Methylation Status of Specific Regions in Single-Oocyte Nucleus and Its Sibling PB1.

Becausee the epigenetic marks of the oocyte nucleus and its sibling PB1 are established in the same ooplasm with a comparable microenvironment, they potentially carry the similar DNA methylation marks at specific regions. To test this, we examined the methylation status of a variety of regions classified as a series of genomic elements distributed genome-wide in single oocyte nuclei and its sibling PB1. Many of these regions mediate critical maternally induced nongenetic phenotype transmission between generations. Promoters of pluripotent markers, Pou5f1 and Nanog, were unmethylated in all oocytes and their sibling PB1 (12 of 12 for Pou5f1 and 12 of 12 for Nanog, respectively) (SI Appendix, Fig. S2 A and B). The repeat element Line1 also largely exhibited unmethylated patterns in all oocytes and their sibling PB1 (9 of 11 identical; 1 of 11 10% methylation difference; 1 of 11 20% methylation difference) (SI Appendix, Fig. S2C). The transposable element, IAP at the A^vy locus, whose methylation degree causes coat color variation ranging from yellow (unmethylated) to pseudoagouti (methylated), showed methylation differences between oocytes. Of nine oocytes derived from three yellow mothers, seven were unmethylated, one was moderately methylated, and one was heavily methylated (Fig. 1A). Of 10 oocytes derived from three pseudoagouti mothers, six were densely methylated, three were moderately methylated, and one was unmethylated (Fig. 1B). This result explains why a yellow mother delivers more yellow pups, but with a small percentage of pseudoagouti ones, and a pseudoagouti mother delivers more pseudoagouti pups, but with a small fraction of yellow ones. Notably, although methylation difference was observed between oocytes from the same mother, all oocytes largely carry a similar methylation mark at this locus compared with its sibling PB1 (10 of 19 identical; 7 of 19 9% methylation difference; 1 of 19 18% methylation difference; 1 of 19 27% methylation difference). Methylation imprinting is essential for normal mammalian development, and even partial erasure may be lethal to mammalian embryos (33, 34). Aberrant methylation imprinting is also linked to a number of diseases (35).

Fig. 1.

Similar methylation status of specific regions between a single oocyte nucleus and its sibling PB1. (A) Single-cell bisulfite sequencing for the methylation status of the transposable element: IAP at the A^vy locus from yellow mothers. (B) Single-cell bisulfite sequencing for the methylation status of the transposable element: IAP at the A^vy locus from pseudoagouti mothers. (C) Single-cell bisulfite sequencing for the methylation status of H19 ICR. (D) Single-cell bisulfite sequencing for the methylation status of IG-DMR, also known as the Gtl2 ICR. (E) Single-cell bisulfite sequencing for the methylation status of Snrpn ICR. (F) Single-cell bisulfite sequencing for the methylation status of Peg3 ICR. Each single oocyte sample was assigned an ID beginning with a letter and followed by a two-digit number. For each region, different letters represent different oocyte donors, and different numbers distinguish different oocytes from the same donor. For each single MII oocyte, the methylation status at a specific region of the oocyte nucleus (Oo) and its PB1 are shown on the top and bottom, respectively. White circles represent unmethylated CpGs; black circles represent methylated CpGs.

We then examined the methylation status of two paternally methylated imprinting control regions (H19 ICR and IG-DMR, also known as the Gtl2 ICR) and two maternally methylated imprinting control regions (Snrpn ICR and Peg3 ICR) in single oocytes and their sibling PB1. Of 10 ovulated MII oocytes from three independent mothers, the majority (8 of 10) was unmethylated at H19 ICR in both the oocyte and sibling PB1, whereas a few (2 of 10) were hypomethylated in both the oocyte and sibling PB1 (Fig. 1C). This observation supports previous studies showing that H19 methylation is susceptible to superovulation (36, 37). IG-DMR was unmethylated in all oocytes and its sibling PB1 (10 of 10; four mothers) (Fig. 1D). Snrpn ICR was fully methylated in the majority of oocytes and its sibling PB1 (eight of nine) and was partially methylated in the minority of oocytes and its sibling PB1 (one of nine) (Fig. 1E). As observed in H19, this result can be explained by superovulation-induced changes in Snrpn ICR methylation. Peg3 ICR was densely methylated in all oocytes and its sibling PB1 (nine of nine; three mothers) (Fig. 1F). Collectively, these data demonstrate that the oocyte nucleus and sibling PB1 carry similar methylation marks at specific regions, although there were differences among individual oocytes even from the same mother.

Evaluation of Specific Methylation in Single Oocyte from Its Sibling PB1.

We next investigated whether methylation information associated with nongenetic phenotypes within an oocyte can be evaluated from its sibling PB1 without destroying the oocyte itself. We first tested whether the A^vy-related coat color phenotype can be evaluated by methylation analysis of PB1. We defined oocyte methylation level at the A^vy locus with >90% methylation as hypermethylated (F), 30–70% methylation as intermediate methylated (I), and <10% methylation as hypomethylated (U), based on our previous single oocyte methylation analysis on the A^vy locus. PB1 from 25 MII oocytes from 4 yellow mothers and 23 MII oocytes from 4 pseudoagouti mothers were successfully sequenced at the IAP promoter and were used for subsequent in vitro fertilization. The methylation proportion of yellow mother derived 25 oocytes was: hypermethylated (F), 3 of 25; intermediate methylated (I), 8 of 25; and hypomethylated (U), 14 of 25, respectively. The methylation proportion of pseudoagouti mother derived 23 oocytes was: F, 16 of 23; I, 6 of 23; and U, 1 of 23, respectively. For in vitro fertilization, although the coat color in offspring is unaffected following transmission of A^vy through the male germline, we always used males with the same coat color as sperm donors to avoid any paternal effects. After in vitro fertilization, 23 of 25 of the yellow mother-derived oocytes (F: I: U, 2: 8: 13) and 21 of 23 pseudoagouti mother-derived oocytes (F: I: U, 14: 6: 1) developed to blastocyst. We transferred these blastocysts each to an independent pseudopregnant C57BL/6 background black mother, which provides an intrauterine environment equal to that of yellow or pseudoagouti mother (23). C57BL/6 background blastocysts were cotransferred together so that the offspring can easily be distinguished by coat color. Eighteen of 23 yellow mother oocyte-generated blastocysts (F: I: U, 1: 4: 13) and 13 of 21 pseudoagouti mother oocyte-generated blastocysts (F: I: U, 9: 3: 1) developed into normal offspring (Dataset S2). Of yellow mother oocyte-generated offspring, the only offspring derived from full methylation oocytes exhibited pseudoagouti coat color; the 4 offspring derived from intermediate methylation oocyte exhibited mottled coat color with variety among individuals; and the 13 offspring derived from unmethylated oocytes showed yellow coat color (11 of 13) or with slight mottle (2 of 13). Of pseudoagouti mother oocyte-generated offspring, the nine offspring derived from full methylated oocytes showed pseudoagouti coat color; the three offspring derived from intermediate methylated oocytes showed mottled coat color with two heavily mottled and one slightly mottled; and the only offspring derived from unmethylated oocytes exhibited yellow coat color (Dataset S2). This further explains why yellow mothers produce more yellow offspring but with a small fraction of pseudoagouti and mottled offspring, and vice versa. Bisulfite sequencing of two yellow mother oocyte-generated yellow offspring (randomly selected) showed a low methylation level, whereas the yellow mother oocyte-generated pseudoagouti offspring showed a high methylation level at the A^vy locus (Fig. 2A). Similarly, pseudoagouti mother oocyte-generated pseudoagouti offspring (randomly selected) exhibited high methylation, whereas the pseudoagouti mother oocyte-generated yellow offspring exhibited low methylation at the A^vy locus (Fig. 2B). Moreover, yellow offspring derived from PB1 analysis with low methylation at the A^vy locus displayed increased body weight at adulthood, whereas pseudoagouti offspring derived from PB1 analysis with high methylation at the A^vy locus maintain a lean phenotype (SI Appendix, Fig. S3). These results demonstrate that the coat color and physiological-related phenotype can be evaluated by methylation analysis of PB1 at the A^vy locus.

Fig. 2.

Evaluation of A^vy-related coat color by targeted methylation analysis of PB1. (A) Bisulfite sequencing on offspring generated from yellow mother oocytes. Two yellow offspring (Top and Middle) showed low methylation level, whereas the only pseudoagouti offspring (Bottom) showed high methylation level at the A^vy locus, respectively. (B) Bisulfite sequencing on offspring generated from pseudoagouti mother oocytes. Two pseudoagouti offspring randomly selected (Top and Middle) exhibited high methylation, whereas the only yellow offspring (Bottom) exhibited low methylation at the A^vy locus, respectively. For offspring, tail DNA was used for methylation analysis. For offspring phenotype classification, 1–5 represents coat color scoring as described in Methods. “1” represents completely yellow, whereas “5” represents completely pseudoagouti in this figure. White circles represent unmethylated CpGs; black circles represent methylated CpGs. Values on each bisulfite grouping indicate the percentage of CpG methylation, with number of clones analyzed in parentheses.

We then asked whether developmental potency can be evaluated by methylation analysis of PB1 using a classic imprinting disorder-induced developmental arrest bimaternal mouse model (30, 31). Mouse parthenogenetic embryos can only develop to day 10 of gestation due to twofold expression of maternal imprinted genes and loss of expression of paternal imprinted genes (26, 27). The modified bimaternal mouse constructed with ng oocyte and fg oocyte supports expression of many imprinted genes and therefore can extend development to 13.5 d. The ng oocyte from newborn mice can be considered as imprinting-free, allowing expression of many imprinted genes that were normally expressed solely from the paternal genome (28, 29). However, these bimaternal mice cannot finish full-term development due to the major barrier of two paternally methylated imprinting control regions: H19 ICR and IG-DMR (31). We investigated whether the developmental potency of these bimaternal embryos can be evaluated by methylation analysis of these two regions in PB1. Because the causal relationship between the phenotype of developmental potency and the molecular changes has been well established in previous studies, we assessed the molecular alterations in constructed embryos instead of growth-retarded related phenotypes. We constructed ng/fg oocytes by serial nuclear transfer, as described previously (30), and recovered ng-originated PB1 and fg-originated PB1 for single-cell methylation analysis, respectively. After artificial activation and in vitro culture, bimaternal embryos that successfully developed to blastocysts were subjected to single blastocyst methylation analysis. Two paternally methylated DMRs, H19 ICR and IG-DMR, showed hypomethylation in all pairs of PB1 (five of fivefor H19 and five of five for IG-DMR, respectively). Methylation analysis of single blastocysts derived from these oocytes revealed hypomethylated status on both DMRs in the majority of embryos (four of five for H19 and five of five for IG-DMR, respectively). One blastocyst was the exception and showed partial gain of methylation at 3′ region of H19 ICR (Fig. 3 A and B and SI Appendix, Fig. S4 A and B), which was likely due to effects of in vitro embryonic culture on methylation, as reported previously (37). Thus, this finding further supports why bimaternal embryos constructed from ng and fg oocytes cannot complete full-term development. Two maternal DMRs, Peg3 and Snrpn DMR, showed a normal maternal methylation pattern in all PB1 (five of five for Peg3 and five of five for Snrpn, respectively) from fg oocytes, and showed a paternal-like unmethylation pattern in all PB1 from ng-originated MII oocytes. Both DMRs showed a normal methylation pattern in all blastocysts (five of five for Peg3 and five of five for Snrpn DMR, respectively), with approximately half methylated and half unmethylated (Fig. 3 C and D and SI Appendix, Fig. S4 C and D). This explains why bimaternal embryos have a more extended development than traditional parthenogenetic embryos. Taken together, these data demonstrate that the imprinting-related developmental ability can be potentially evaluated by methylation analysis of PB1.

Fig. 3.

Evaluation of the developmental potency of bimaternal embryos by methylation analysis of specific DMRs in PB1. (A) Bisulfite sequencing on H19 ICR of a single bimaternal blastocyst after PB1 methylation evaluation. (B) Bisulfite sequencing on IG-DMR of a single bimaternal blastocyst after PB1 methylation evaluation. (C) Bisulfite sequencing on Snrpn DMR of a single bimaternal blastocyst after PB1 methylation evaluation. (D) Bisulfite sequencing on Peg3 DMR of a single bimaternal blastocyst after PB1 methylation evaluation. Each single blastocyst was constructed with one ng oocyte and one fg oocyte. The methylation status of specific regions of ng-originated PB1 and fg-originated PB1 was indicated at the top of each panel. White circles represent unmethylated CpGs, and black circles represent methylated CpGs. Values on each bisulfite grouping indicate the percentage of CpG methylation, with number of clones analyzed in parentheses. For each DMR, one single blastocyst was shown here. Four additional single blastocysts for each region are present in SI Appendix, Fig. S4.

Targeted DNA Methylation Editing in Single Oocytes.

By fusion of Tet or Dnmt3 to the catalytically inactive Cas9 (dCas9), recent approaches have enabled targeted DNA methylation editing in mammalian cells (17–19). The advantage of oocytes is that they can be easily handled individually, allowing many manipulations at a single-cell level. We investigated whether DNA methylation at a specific region can be targeted for precise editing in single oocytes. Establishment of DNA methylation in mammalian oocytes depends on DNMT3A, whereas erasure of methylation in female germ cells depends on TET1. Based on a previous study that the catalytic domain of Tet1 has a higher catalytic activity than the full-length form (19), we used the catalytic domain of Tet1. We fused Dnmt3a or the catalytic domain of Tet1 to dCas9, respectively, and generated mRNA by in vitro transcription. We employed a microinjection-based approach to edit the methylation status of a targeted region. We coinjected the dCas9-Dnmt3a or dCas9-Tet1 mRNA with sequence-specific guide RNA (gRNA) into germinal vesicle (GV) oocytes, and examined the effect of targeted editing on mature MII oocytes. The injected RNA does not integrate into the genome and only has a transient effect, thus providing a unique system for functional study of specific methylation events in oocytes.

Targeted methylation editing of A^vy locus reverses coat color-related phenotype.

We first asked whether phenotypes associated with coat color in A^vy mice can be altered by targeted methylation editing at the A^vy locus in oocytes (Fig. 4A). We designed three single-guide RNA (sgRNAs) targeting 11 CpGs of the IAP promoter at the A^vy locus, and coinjected the dCas9-Dnmt3a mRNA with sgRNA into GV oocytes derived from yellow mothers. After in vitro maturation, the majority of oocytes (10 of 11) showed full methylation at this locus, indicating successful de novo methylation (SI Appendix, Fig. S5A). In contrast, catalytically dead DNMT3A or unrelated gRNA resulted in no change in the frequency of DNA methylation patterns from that in Fig. 1 (SI Appendix, Fig. S5 B and C), indicating a lack of de novo methylation in these controls. This confirms that the observed methylation editing is not caused by introduction of gRNA or overexpression of Dnmt3a. Similarly, when we coinjected the dCas9-Tet1 mRNA and sgRNA into GV oocytes derived from pseudoagouti mothers, we observed substantial demethylation, as evidenced by the lack of methylation in the majority of oocytes (7 or 10) (SI Appendix, Fig. S5D). Replacing either the TET1 with a catalytically dead form or the gRNA with an unrelated gRNA did not show any change in the frequency of the DNA methylation patterns from that observed in Fig. 1 (SI Appendix, Fig. S5 E and F), indicating a lack of demethylation in these controls. We also assessed the off-target effects in this system by bisulfite sequencing of one potential off-target site of sgRNA3, but observed no significant off-target activity, suggesting the potential specificity of this editing system (SI Appendix, Fig. S6A). We then investigated the effect of methylation editing on the phenotype of offspring. We performed in vitro fertilization using oocytes modified via methylation editing and transferred each blastocyst into a foster mother along with C57BL/6 background embryos, as described above. Because a minority of oocytes was not successfully edited as indicated above, we performed PB1 analysis to include only blastocysts developed from successful edited oocytes for embryo transfer. All offspring (eight of eight) generated from yellow mother-derived oocytes that were successfully edited exhibited pseudoagouti coat color (SI Appendix, Fig. S7A and Dataset S3). Bisulfite sequencing on three randomly selected offspring showed high methylation at A^vy, which correlates with their coat color (Fig. 4B). On the other hand, all offspring (10 of 10) generated from pseudoagouti mother-derived oocytes that were successfully edited exhibited yellow coat color (SI Appendix, Fig. S7B and Dataset S3). Bisulfite sequencing on three randomly selected offspring showed low methylation at A^vy, and also correlates with their coat color (Fig. 4C). These results demonstrate that coat color-related phenotype can be reversed by targeted methylation editing of a specific locus in oocytes.

Fig. 4.

Targeted methylation editing of the A^vy locus reverses coat color-related phenotype. (A) Schematic representation of targeting the A^vy promoter region by dCas9-Dnmt3a or dCas9-Tet1 with specific sgRNAs to manipulate the coat color-related phenotype. Unmethylated IAP that lies upstream of the agouti gene allows ectopic expression of the gene, resulting in a yellow coat color. Methylated IAP induces expression of the gene under normal developmental control, leading to a pseudoagouti coat color (termed pseudoagouti as the mice are isogenic with fully yellow mice and not genetically agouti). (B) Bisulfite sequencing on three offspring generated from yellow mother oocytes that have been successfully de novo-methylated. (C) Bisulfite sequencing on three offspring generated from pseudoagouti mother oocytes that have been successfully demethylated. For offspring, tail DNA was used for methylation analysis. For offspring phenotype classification, 1–5 represents coat color scoring as described in Methods. “1” represents completely yellow, whereas “5” represents completely pseudoagouti in this figure, respectively. White circles represent unmethylated CpGs, and black circles represent methylated CpGs. Values on each bisulfite grouping indicate the percentage of CpG methylation, with number of clones analyzed in parentheses.

Targeted de novo methylation of H19 ICR and IG-DMR allows full-term development of bimaternal mice.

We then asked whether developmental competency can be improved by targeted methylation editing of specific regions using the bimaternal mouse model (Fig. 5A). Because previous evidence demonstrated that two paternally methylated imprinting control regions, H19 ICR and Dlk1-Dio3 IG-DMR, are the major barrier that prevents bimaternal mouse embryos from developing to term (30, 31), we asked whether reconstructing methylation of these two DMRs can generate full-term development bimaternal mice (Fig. 5B). We attempted to de novo methylate these two DMRs in ng oocytes to mimic the methylation state of the paternal genome. The H19 imprinting cluster contains a well-defined ∼2-kb paternal-specific methylated ICR, within which there are four methylation-sensitive CTCF binding sites that are critical for regulation of H19 domain imprinting (38). We designed 10 sgRNAs targeting this region (SI Appendix, Fig. S8 A and B). We coinjected these sgRNAs with dCas9-Dnmt3a mRNA into ng-derived GV oocytes and analyzed the methylation status of the targeted region in MII oocytes after in vitro maturation. For subsequent analysis, we designated three minor regions within the large ∼2-kb ICR, which overlap all four CTCF binding sites on the large ICR. This includes a region prevalently used for H19 methylation analysis, which we designated as region 1 (R1), and two other regions, designated as region 2 (R2) and region 3 (R3), respectively (SI Appendix, Fig. S8 A and B). We observed that the majority of oocytes (9 of 10) were de novo-methylated at R1, indicating successful methylation editing at this locus (SI Appendix, Fig. S8C). Similarly, all oocytes (eight of eight) at R2 and the majority of oocytes (eight of nine) at R3 also showed de novo methylation (SI Appendix, Fig. S8 D and E). Introducing unrelated sgRNAs or dCas9 with catalytically dead DNMT3A did not show any change in DNA methylation levels in oocyte–PB1 pairs compared with that observed in Fig. 1 (SI Appendix, Fig. S8C), demonstrating a lack of de novo methylation. Thus, this system not only efficiently edits methylation at the H19 ICR but that the induced de novo methylation is specific.

Fig. 5.

Targeted de novo methylation of H19 ICR and IG-DMR allows full-term development of bimaternal mice. (A) Diagram of generating bimaternal embryos, which contain two sets of haploid genome derived from an ng oocyte (blue) and an MII fg oocyte (dark red). The first nuclear transfer (NT) is conducted to resume the first meiotic division and to form haploid chromosomes. The second NT is conducted to form diploid bimaternal embryos. After first NT, dCas9-Dnmt3a mRNAs and sgRNAs are coinjected into the cytoplasm of GV oocytes to achieve targeted methylation editing. (B) Schematic representation of targeting the H19 ICR and IG-DMR by dCas9-Dnmt3a with specific sgRNAs to generate full-term development of bimaternal mice. “E” represents enhancer in this panel. (C) Bisulfite sequencing on H19 ICR and IG-DMR in three live bimaternal adults. (D) Bisulfite sequencing on H19 ICR and IG-DMR in three growth retarded bimaternal mice that were alive at term but died within 3 d. Tail DNA was used for methylation analysis. White circles represent unmethylated CpGs, and black circles represent methylated CpGs. Values on each bisulfite grouping indicate the percentage of CpG methylation, with number of clones analyzed in parentheses.

To investigate whether these edited DMRs can maintain methylation like naturally established ones postzygotically, we constructed ng^H19me+/fg^wt bimaternal embryos by serial nuclear transfer, as described above, and determined the methylation status of related regions in blastocysts after in vitro culture. Of eight blastocysts generated from successful de novo-methylated oocytes as PB1 R1 indicated, seven of them displayed correct methylation pattern at all three regions, suggesting that these edited DMRs can maintain methylation postzygotically similar to those of naturally established; one of them showed potentially partial loss of methylation (SI Appendix, Fig. S9), which can be explained by incomplete maintenance of imprinting caused by in vitro embryo culture (37). On the other hand, two blastocysts were not successfully edited for DNA methylation, because the PB1s (R1) and blastocysts (all three regions) were unmethylated (SI Appendix, Fig. S9). These results also suggest that R1 can indicate the success of methylation editing on the targeted large DMR. Similarly, we designed sgRNAs to target another critical DMR, IG-DMR. The IG-DMR contains a well-defined ∼4.15-kb paternally specific methylated region that regulates imprinting of an ∼1-Mb cluster of genes due to allele-specific methylation (39). We designed 12 sgRNAs targeting CpGs in this region (SI Appendix, Fig. S10 A and B). We coinjected these sgRNAs with dCas9-Dnmt3a into ng-derived GV oocytes and determined the effect of methylation editing on MII oocytes after in vitro maturation. For methylation analysis, we designated three small regions dispersed on the large DMR, including a region prevalently used for methylation analysis at this locus, R1, and two other regions we designated as R2 and R3, respectively (SI Appendix, Fig. S10 A and B). Most or all MII oocytes showed successful de novo methylation at three regions (R1: seven of nine; R2: eight of eight; R3: nine of nine), demonstrating that this system can efficiently induce methylation editing at this locus (SI Appendix, Fig. S10 C–E). Moreover, blastocysts generated from ng^IGme+/fg^wt oocytes exhibited a correct methylation pattern, with approximately half of the alleles methylated and approximately half of the alleles unmethylated, suggesting that these added methylation marks can be maintained postzygotically similar to natural ones (SI Appendix, Fig. S11). In addition, as seen in H19 ICR when the R1 showed successful de novo methylation, all three regions largely exhibited correct methylation patterns in blastocysts, and vice versa (SI Appendix, Fig. S11). Thus, the R1 can indicate the success of methylation editing at this locus.

Next, we coinjected sgRNAs targeting both regions with dCas9-Dnmt3a to de novo methylate both DMRs simultaneously. We constructed ng^H19me+IGme+/fg^wt embryos using ng oocytes of both DMRs edited, and determined the methylation status of both DMRs in blastocysts. Because we can analyze only one specific region each time for each oocyte with high efficiency, we used H19 and IG R1 interchangeably. Most blastocysts exhibited a correct methylation pattern at both regions, showing that this system can simultaneously target both regions (SI Appendix, Fig. S12). Notably, we observed that when either region of ng-derived oocytes shows successful editing, both regions displayed a correct methylation pattern in the vast majority of blastocysts. In contrast, when either region exhibited unsuccessful editing, both regions displayed incorrect methylation patterns in blastocysts (SI Appendix, Fig. S12). This suggests that either region might be served as a marker of successful editing on both regions. In total, we had a methylation-editing success rate of ∼82% of embryos. We also validated the off-target effects in this system by methylation analysis of potential off-target sites of the H19 R1-sgRNA4 (SI Appendix, Fig. S6B) and H19 R3-sgRNA1 (SI Appendix, Fig. S6C), but did not observe significant off-target activity. In our system, dCas9-Dnmt3a induces full or nearly full de novo methylation at ∼500 bp from the target, but only partial de novo methylation at ∼1,100 bp from the target (SI Appendix, Fig. S13), suggesting that the effective targeting does not extend past 1,100 bp. Thus, designing multiple sgRNAs at a distance of 500–1,000 bp within a target locus would provide high edited specificity that, furthermore, would not be expected to extend to promoters of neighbor genes or other ICRs. Moreover, construction of bimaternal embryos using polymorphisms to track alleles demonstrates that this system cannot edit fg genomes postzygotically (SI Appendix, Fig. S14).

We next investigated whether methylation reconstruction of both DMRs in ng oocytes supports full-term development of bimaternal embryos. Because H19 ICR has better amplification efficiency and is more prone to error in fg oocytes, we selected H19 ICR for single PB1 analysis and chose those qualified oocytes for subsequent development. In total, we constructed 247 ng^H19me+IGme+/fg^wt oocytes by serial nuclear transfer. After in vitro maturation and activation, we recovered 187 (75.7%) diploid one-cell bimaternal embryos that formed two pronuclei and second polar bodies. After in vitro embryo culture, 169 (90.4%) developed to blastocyst stage and were transferred to 12 recipient females (Table 1). A total of 44 (26.0%) live pups were recovered by autopsy at 19.5 d of gestation. Of these, 41 pups were successfully nursed by their foster mothers. Ten of 41 pups that had a mean birth weight of 0.89 ± 0.05 g (mean ± SEM) showed slight growth retardation at birth and died within 3 d. The remaining 31 pups had a mean body weight of 1.17 ± 0.04 g, which was similar to wild type (1.19 ± 0.02 g) and grew to adulthood. To gain molecular insight into the full-term development of ng^H19me+IGme+/fg^wt bimaternal embryos, we randomly selected three individuals at 8 wk and conducted methylation analysis on both DMRs on tail tissue. Bisulfite sequencing demonstrated that both H19 ICR and IG-DMR exhibited correct methylation status in these live adults (Fig. 5C). These results indicate that the reconstructed methylation of both DMRs can be inherited to the next generation and is critical for supporting full-term development of bimaternal embryo.

Table 1.

Development of ng^H19me+IGme+/fg^wt bimaternal embryos

Developmental progress	Number
No. of reconstructed eggs	187
No. of embryos developed to blastocyst	169 (90.4% of reconstructed eggs)
No. of embryos transferred	169 (100% of blastocysts)
No. of pregnants/recipients	12 of 12
No. of live pups	44 (26.0% of embryos transferred to recipients)
No. of survived pups	31 (18.3% of embryos transferred to recipients)

To understand why a minority of ng^H19me+IGme+/fg^wt bimaternal embryos did not grow to adult, we examined methylation of both DMRs as above and expression of imprinted genes regulated by both imprinting clusters. In three randomly selected growth-retarded bimaternal mice that were alive at term but died within 3 d, either H19 ICR or IG-DMR exhibited loss of imprinted methylation (Fig. 5D). Consistently, quantitative real-time PCR analysis on H19, Igf2, Dlk1, and Gtl2 revealed that at least one of the four critically imprinted genes showed abnormal (beyond twofold change) expression in each dead pup, which correlated with loss of imprinted methylation, compared with controls (SI Appendix, Fig. S15). These data further demonstrate that both DMRs are essential for normal development of bimaternal mouse embryos. Moreover, we observed that although 31 bimaternal mice survived beyond 3 d after birth grew to adult, they displayed postnatal growth retardation, with ∼15% reduction in body weight compared with controls (SI Appendix, Fig. S16A). The most likely explanation is due to the third critical paternally methylated imprinting control region besides H19 and Gtl2, which is located on chromosome 9 and regulates Rasgrf1 expression (40). Rasgrf1 plays an important role in postnatal growth by inducing growth hormone secretion from the pituitary (41). In our study, it was down-regulated in the brain of bimaternal mice compared with controls (SI Appendix, Fig. S16B). Additional editing of this locus may lead to normal postnatal growth in bimaternal mice. Taken together, these data suggest that developmental potency can be improved by targeted methylation editing of specific regions.

Targeted de novo methylation of Snrpn corrects familial AS disorder in a mouse model.

Finally, we tested whether familial AS disorder can be corrected by targeted methylation editing of Snrpn using a mouse model (Fig. 6A). In humans, familial AS is caused by imprinting center deletion and subsequent failure in establishment of maternal-specific methylation at SNRPN (32). We generated a mouse model to mimic human familial AS by CRISPR/Cas9-mediated engineering (42). We deleted an 80-kb region located 15-kb upstream of Snrpn exon1, which contains the AS imprinting control region and is essential for establishing maternal-specific DNA methylation at the neighboring DMR during oogenesis (43), by designing two sgRNAs targeting both ends of this region. The deletion was confirmed by PCR and sequence analysis (SI Appendix, Fig. S17A). Bisulfite sequencing on oocytes derived from these females showed loss of methylation at Snrpn, confirming that they were unable to establish maternal specific methylation (SI Appendix, Fig. S17B). Consistently, bisulfite sequencing on three in vitro blastocysts and three offspring derived from these oocytes showed loss of methylation at Snrpn, further demonstrating the maternally transmitted methylation disorder in these animals (SI Appendix, Fig. S17 C and D).

Fig. 6.

Targeted de novo methylation of Snrpn DMR corrects familial AS disorders. (A) Schematic representation of targeting the mouse Snrpn DMR locus by dCas9-Dnmt3a with specific sgRNAs to correct familial AS disorder. The imprinting control region for this locus has a bipartite structure, with an upstream component acting as AS-ICR, and a downstream region containing the promoter of Snrpn gene: Snrpn DMR. Naturally, maternal AS-ICR is unmethylated and in an open chromatin configuration, conferring DNA methylation and closed chromatin structure at the Snrpn DMR, resulting in silence of bidirectionally all paternally expressed genes (in yellow), and activation of maternally expressed genes, including Ube3a (in pink). When AS-ICR is deleted, Snrpn DMR remains in its default unmethylation status, resulting in activation of all paternally expressed genes and silence of Ube3a, further leading to AS. (B) GV oocytes were coinjected with dCas9-Dnmt3a mRNA and sgRNA targeting Snrpn DMR. After in vitro maturation, single MII oocyte nuclei were subjected to bisulfite sequencing analysis. Oocyte ID was shown at the left of the panel, with the first three digits following “S” representing the donor ID, and the following two digits representing different oocytes from this donor. (C) Bisulfite sequencing on five offspring generated from embryos using edited oocytes. (D) Bisulfite sequencing on three offspring generated from embryos using oocytes edited by dCas9-Dnmt3a and scrambled sgRNA as control. For offspring, brain DNA was used for methylation analysis. White circles represent unmethylated CpGs; black circles represent methylated CpGs. Values on each bisulfite grouping indicate the percentage of CpG methylation, with number of clones analyzed in parentheses.

Finally, we asked whether targeted de novo methylation of Snrpn in oocytes can correct this deletion-induced nongenetic disorder. We designed 4 sgRNAs targeting 16 CpGs in the Snrpn DMR (SI Appendix, Fig. S18A) and coinjected these sgRNAs with dCas9-Dnmt3a mRNA into GV oocytes derived from AS females. After in vitro maturation, we determined the editing effect by methylation analysis on Snrpn in single MII oocytes. Bisulfite sequencing on eight oocytes showed full or hypermethylation at Snrpn, indicating successful de novo methylation at this locus (Fig. 6B). In contrast, replacing the sgRNA with scrambled sgRNA resulted in no acquisition, confirming that DNA methylation acquisition is specific to the edited allele, as well as the ability to bypass the effects of the AS ICR deletion (SI Appendix, Fig. S18B). After in vitro fertilization, all blastocysts (three of three) derived from these edited oocytes exhibited normal methylation with ∼half of the alleles methylated and ∼half unmethylated, demonstrating that these edited marks can be inherited during postzygotic development (SI Appendix, Fig. S18C).

Next, we performed an embryo transfer experiment to allow generation of offspring. As a control, we transferred embryos developed from oocytes coinjected with dCas9-Dnmt3a mRNA and scrambled sgRNA. We performed qRT-PCR to determine the expression of Ube3a and Snrpn in offspring, both of which are critical for AS. qRT-PCR on brain tissues from six offspring (randomly selected) revealed significantly decreased expression of Snrpn (SI Appendix, Fig. S18D). In contrast, the expression of Ube3a in these animals was significantly increased (SI Appendix, Fig. S18E). Bisulfite sequencing on five animals showed correct methylation patterns at Snrpn, confirming that the edited marks transmitted to somatic tissues of offspring (Fig. 6 C and D). We determined phenotype alteration after editing in these animals by behavior tests. The offspring of methylation editing showed significantly higher activity level in an open-field test (SI Appendix, Fig. S18F). Moreover, these animals exhibited significant improved motor coordination under the accelerating rotarod test (SI Appendix, Fig. S18G). Thus, maternally transmitted familial AS disorder can be corrected by targeted methylation editing of Snrpn. Collectively, these data demonstrate that DNA methylation associated with mammalian phenotype, developmental competency, and nongenetic disease in single oocytes can be edited in a targeted manner.

Discussion

In the present study, we demonstrate that the specific methylation in single oocytes can be evaluated from its sibling PB1 and can be edited in a targeted manner. Because all differentiated somatic cells composing an organism are generated from fusion of two gametes, manipulation at the single-oocyte level allows prevention or correction of maternally inherited nongenetic disorders from origin. Although all manipulations of this approach do not involve genome sequence alterations, very extensive additional testing for safety will be required before there is any thought of translating it to the clinic.

Single-oocyte DNA methylation analysis has been achieved in previous studies (44, 45). These studies provide valuable tools for identify epigenetic errors at the single oocyte level. However, the efficiency is lower, although these protocols do not require removal of the nucleus before analysis. Using our optimized single-oocyte bisulfite sequencing method with an efficiency of ∼85.8%, we show that the methylation information associated with specific phenotypes or diseases within an oocyte can be evaluated from its sibling PB1. This provides a framework for PB epigenetic diagnosis and the exclusion of those oocytes with nongenetic defects. A recent study reported a genome-wide single-cell bisulfite-sequencing technique that can accurately measure DNA methylation overlapping 68 ± 8% genomic CpGs (46). Future developments may allow most genomic CpGs to be detectable, thus enabling better detection and evaluation of maternally inherited nongenetic disorders. While the PB1 methylation analysis has the advantage of detecting epigenetic disorders transmitted via oocytes, the disadvantage is that it cannot identify epigenetic errors inherited from the sperm or those arisen from early preimplantation development.

Single-oocyte methylation editing provides a simple system to manipulate DNA methylation at any locus of interest. One critical question is the specificity of the methylation-editing system. By testing three independent loci, we did not detect significant off-target activity. One explanation for the observed high specificity is that the injected short-lived RNA in the oocyte-editing system only has a transient effect, whereas the transfected plasmid DNA in the somatic cell-editing system has a long-lasting effect and increases the incidence of off-targets. Another critical question is whether the injected short-lived RNAs have an extended postfertilization effect. Although bisulfite sequencing on bimaternal blastocysts did not detect any de novo methylation in the fg genome, we cannot rule out this issue in fertilized embryos. Construction of in vitro-fertilized embryos using polymorphism to track alleles will help us to comprehensively understand this issue.

Previous evidence showed that transcription is required for natural acquisition of maternal-specific methylation at imprinted genes (47), whereas the dCas9-mediated sgRNA-guided methylation editing system employs an entirely new mechanism. We do not know whether these edited marks are epigenetically indistinguishable with those naturally established ones. Consistently, it has been demonstrated that IG-DMR can be contracted during early embryonic development (48), whereas we did not observe loss of methylation at IG R3 in embryonic day (E)3.5 blastocysts. It is possible that methylation editing might involve minor chromosomal alterations. Further study will be required to clarify this issue.

Epigenetic inheritance via gametes is a developing field, with new mechanisms and new questions constantly emerging (2). A previous study reported complete absence of cytosine methylation in blastocysts of pseudoagouti mothers (49), raising the question how epigenetic information could be transmitted using cytosine methylation. However, given that bisulfite sequencing cannot distinguish unmethylated cytosine with 5-formylcytosine (5fC) or 5-carboxylcytosine (5caC), it is possible that the unmethylated cytosine represent a pooled status of C and 5fC/5caC and methylation was gradually re-established based these modifications during subsequent development. Further study is required to completely understand this mechanism.

Taking these data together, our study facilitates the dissection of maternally transmitted specific nongenetic information regarding development and disease. Our work also provides a strategy for preventing and correcting maternally inherited nongenetic diseases and disorders.

Methods

Mice.

All animal care and use procedures were in accordance with guidelines of the Institutional Animal Care and Use Committee of the Ren Ji Hospital, School of Medicine, Shanghai Jiao Tong University. For coat color studies, A^vy mice were obtained from an isogenic C57BL/6 A^vy colony maintained at Oak Ridge National Laboratories. For bimaternal studies, mice used for GV and superovulated MII oocyte donors were B6D2F1 (C57BL/6 × DBA/2) background, whereas mice used for ng oocyte donors were C57BL/6J background. For familial AS models, mice used for zygote donors were B6D2F1 background. For embryo transfer experiments, all recipient females were CD1 background. For other studies, if not specified, all mice used were C57BL/6 background. Additional details are provided in SI Appendix, SI Methods.

Single-Oocyte Specific-Region Bisulfite Sequencing.

A single-oocyte nucleus or PB1 was collected by a piezo-mediated micromanipulation. Bisulfite modification was accomplished using the EpiTect Fast LyseAll Bisulfite Kit (Qiagen) with the following modifications: all volumes were quartered and DNA protect buffer (Qiagen) containing tetrahydrofurfuryl alcohol (Qiagen) was included. The converted DNA fragment was further ligated with a nine-nucleotide adaptor and amplified three rounds using a GenomePlex Single Cell Whole Genome Amplification Kit (Sigma) with the following modifications: the single-cell lysis and fragmentation step was omitted and amplification was changed to three cycles. The templates were then subjected to nested PCR using Hot Start Taq Polymerase (Takara), with primer sequences summarized in Dataset S4. The PCR products were purified using Wizard SV Gel and PCR Clean-Up System (Promega) and cloned into the pUC57 vector. Individual clones were grown and the plasmids were purified using QIAprep Spin Miniprep Kit (Qiagen). The positive clones were confirmed by electrophoresis and sequenced using an automatic sequencer (ABI PRISM-77). A step-by-step procedure is provided in SI Appendix, Note S1.

Mouse Oocyte Collection and Culture.

Superovulated mature oocytes at the MII stage were collected by ovarian stimulation. Pubertal mice were stimulated by intraperitoneal injection with five international units (IU) of equine CG (eCG) and 44 h later with 5 IU of human CG (hCG). After an additional 16 h, cumulus oocyte complexes were surgically removed from oviducts. Oocytes at the GV stage were used for microinjection. Ovaries were isolated from mice 46 h after intraperitoneal injection of 10 IU of eCG, and cumulus oocyte complexes were recovered from ovaries by repeatedly puncturing antral follicles with a fine steel needle under the visual field of a dissecting microscope. Additional details are provided in SI Appendix, SI Methods.

A^vy Phenotype Scoring.

Coat color in A^vy offspring was assessed at weaning (3-wk-old) by a trained observer. A numerical scale from 1 to 5 was used to represent the color classification, as previously described (21): 1 represents completely yellow; 2 represents mostly yellow with slight agouti; 3 represents approximately half yellow and half agouti; 4 represents mostly agouti with slight yellow; 5 represents completely agouti.

Generation of Bimaternal Embryo by Nuclear Transfer.

Fully grown GV oocytes were collected from ovarian follicles of B6D2F1 female mice 46 h after an eCG (10 IU) injection. Mature MII oocytes were collected from superovulated B6D2F1 female mice after 44 h of eCG (5 IU) and 16 h of hCG (5 IU) injection. Nongrowing oocytes were collected from the ovary of 1-d-old newborn C57BL/6J female mice. Nuclear transfer was performed as described previously (30, 31). The GV oocytes were enucleated in M2 medium containing 200 μM dbcAMP and 5% FBS. Nongrowing oocytes were fused with enucleated GV oocytes by an inactivated Sendai virus (HVJ, 2700 hemagglutinating activity unit per milliliter). The fused oocytes were then cultured in α-MEM medium containing 5% FBS for 14 h. A spindle-chromosome complex from the reconstructed oocytes was transferred into ovulated MII oocytes and MII oocytes were artificially activated by treatment in Ca²⁺-free M16 medium containing 10 mM SrCl₂ for 2 h. Reconstructed embryos were cultured in M16 medium for 3.5 d. The embryos that developed to blastocyst stage were transferred into the uterine horns of CD1 recipient females at 2.5 d of pseudopregnancy. Live pups were recovered by autopsy at 19.5 d of gestation. Additional details are provided in SI Appendix, SI Methods.

Generation of dCas9-Tet1/Dnmt3a mRNA and sgRNA.

We introduced the T7 promoter to the targeted sequence by PCR to generate templates used for in vitro transcription. For the dCas9-Tet1 catalytic domain, the T7 promoter was added by PCR using primer m dCas9-Tet1 forward and reverse (Dataset S5) and Addgene plasmid 84475 as template. For the dCas9-Tet1 inactive form, the T7 promoter was added by PCR using primer m dCas9-Tet1 IF forward and reverse and Addgene plasmid 84479 as template. For dCas9-Dnmt3a, the T7 promoter was added by PCR using primer m dCas9-Dnmt3a forward and reverse and Addgene plasmid 84476 as template. For dCas9-Dnmt3a inactive form, the T7 promoter was added by PCR using primer dCas9-Dnmt3a IF forward and reverse and Addgene plasmid 84478 as template. For sgRNA, the T7 promoter was added to specific sgRNAs by PCR using different primers and Addgene plasmid 42230 as template. Primers used are listed in Dataset S5. The full-length protein sequences of dCas9-Tet1CD/Dnmt3a and their mutants are listed in Dataset S6. Additional details are provided in SI Appendix, SI Methods.

GV Oocyte Microinjection.

Oocytes at the GV stage were used for microinjection. Microinjections were performed using a Nikon Diaphot ECLIPSE TE 300 (Nikon) inverted microscope equipped with Narishige MM0-202N hydraulic 3D micromanipulators (Narishige). dCas9-Tet1/Dnmt3a mRNA (20 ng/μL) and sgRNA (20 ng/μL) was injected into the cytoplasm of GV oocyte with a well-recognized GV in M2 medium. The injected volume was estimated to be 2 pL (estimated at 1% of the ooplasmic volume). After microinjection, the oocytes were cultured in M16 medium supplemented with 10% FBS under mineral oil at 37 °C in a humidified atmosphere of 5% CO₂ in air to MII stage by 14 h.

Quantitative Real-Time PCR.

We analyzed mRNA levels by qRT-PCR after reverse transcription, as described previously (50). Total RNA was extracted using TRIzol reagent (Invitrogen) and quantified by absorbance at 260 nm and 280 nm in a PerkinElmer spectrophotometer. It was then used as a template for cDNA synthesis by SuperScript III first strand synthesis (Invitrogen) with random hexamers. mRNA quantity was determined with the qRT-PCR 7500 system (Applied Biosystems), using primer sequences summarized in Dataset S7 and SYBR Green SuperMix UDG (Invitrogen). Additional details are provided in SI Appendix, SI Methods.

Bisulfite Sequencing on Blastocysts and Somatic Tissues.

Bisulfite genomic sequencing on blastocysts and somatic tissues was performed as previously described (50). Briefly, for blastocysts, a single E3.5 mouse blastocyst was directly subjected to EZ DNA Methylation-Direct Kit (Zymo Research) for bisulfite conversion. For somatic tissues, genomic DNA was purified using Wizard Genomic DNA Purification Kit (Promega), after which bisulfite modification was accomplished using the EZ DNA Methylation Kit (Zymo Research). The converted DNA was then amplified by PCR with primer sequences summarized in Dataset S4. The obtained PCR products were purified using Min Elute Gel Extraction Kit (Invitrogen) and cloned into the pUC57 vector. Individual clones were grown and the plasmids were purified using PureLink Miniprep Kit (Invitrogen). The positive clones were confirmed by PCR and no fewer than 10 clones for each subject were sequenced using an automatic sequencer (ABI PRISM-77). DNA methylation levels of blastocysts and tissues were calculated as the number of methylated CpGs divided by the total number of CpG dinucleotides in the region of analysis. Additional details are provided in SI Appendix, SI Methods.

Generation of Familial AS Mouse Model.

The familial AS mouse model was generated by CRISPR/Cas9-mediated engineering (42). We deleted an 80-kb region 15-kb upstream of Snrpn exon1, which contains imprinting center and is essential for establishing maternal-specific imprinting (43), by designing two sgRNAs targeting both ends of this region. B6D2F1 and CD1 mice were used as zygote donors and foster mothers, respectively. Cas9 mRNA and sgRNAs were injected into the cytoplasm of pronuclear stage fertilized eggs to produce F₀ founders. F₀ founders carrying the deletion were identified by PCR and sequencing analysis. F₀ founders were mated with wild-type mice to produce F₁ heterozygote. The F₁ heterozygote mice were intercrossed to generate F₂ homozygote. The successful deletion in F₂ homozygote females was identified by PCR and sequence analysis, and used for subsequent studies. Additional details are provided in SI Appendix, SI Methods.

Statistical Analyses.

Phenotype data were analyzed by SPSS 16.0 after log-transformation or square-root transformation unless raw data were normally distributed. Measurements at single time points were analyzed by ANOVA or, if appropriate, by two-tailed Student’s t test. Time courses were analyzed by repeated-measurements ANOVA. All data are shown as mean ± SEM; P < 0.05 was considered statistically significant.