Sv/129 mice heterozygous for the Dnmt1 P allele mutation (Shaffer et al., 2015) were obtained from Dr. Mellissa Mann (Magee-Womens Research Institute, Pittsburg, PA). Natural matings between heterozygous pairs were used to generate Dnmt1 ^+/+ , Dnmt1 ^P/+ and Dnmt1 ^P/P embryos, which were collected at 9.5, 12.5, 15.5 and 18.5 days post coitum (dpc). Natural matings were also used to generate offspring in order to maintain the Dnmt1 P allele in the colony. Ethical approval for procedures involving animals was granted by the Bryn Mawr College Institutional Animal Care and Use Committee, PHS Welfare Assurance Number A3920-01.

Genotypes were determined using a PCR-based assay described by Shaffer et al. (Shaffer et al., 2015). Briefly, DNA was extracted from embryo or 3–4 weeks mouse tails using proteinase K digestion and genomic DNA was purified using a Genomic DNA Clean & Concentrator kit (Zymo Research, Irvine, CA, cat#D4011). PCR using oligonucleotides flanking the P allele mutation was followed by restriction digestion with AvaI, and wild type vs. mutant P alleles were distinguished by agarose gel electrophoresis (wild type allele, 627 bp; P allele, 447 + 180 bp). Chi-square goodness of fit tests were conducted in Microsoft Excel, using the raw number of Dnmt1 ^+/+ , Dnmt1 ^P/+ and Dnmt1 ^P/P embryos or pups collected at each developmental stage, to determine whether the observed values deviated significantly from the Mendelian ratios expected from crosses between heterozygous pairs.

Genomic DNA was isolated from 9.5, 12.5, 15.5 and 18.5 dpc embryo heads following proteinase K digestion and a series of phenol/chloroform extractions as described previously (Davis et al., 1999). Purified DNA was subjected to bisulfite mutagenesis using an EZ DNA Methylation-Direct kit (Zymo Research, Irvine, CA, cat#D5020). Mutagenized DNA was subjected to nested or semi-nested PCR amplification; primers, PCR annealing temperatures and expected second round PCR product size for each locus analyzed are detailed in Supplementary Table S1. Resulting amplicons were purified from agarose gels using a Zymoclean Gel DNA Recovery kit (Zymo Research, Irvine, CA, cat#D4002) and quantified using a Qubit 3.0 Fluorometer (ThermoFisher Scientific, cat#Q33216). Equimolar amounts of PCR product from multiple loci were combined with a minimum of 50 ng from each amplicon and submitted to Azenta (South Plainfield, NJ) for NextGeneration-based amplicon sequencing. Sequence reads were uploaded to a Galaxy Instance hosted at Bryn Mawr College, paired and processed using fastp, mapped to known target sequences, and analyzed for non-CpG bisulfite conversion efficiency as well as for the presence of cytosines vs. thymines in a CpG context (Afgan et al., 2018). The bisulfite conversion efficiency was >99% for all datasets used in this analysis.

Downsampling of NGS data was performed using Galaxy tools to obtain 20–25 sequencing reads for each locus analyzed. Percent methylation for each strand was calculated and the raw data from each allele in Dnmt1 ^+/+ and Dnmt1 ^P/P embryos was ranked and assessed for statistically significant differences using a Mann-Whitney U test (http://vassarstats.net/utest.html).

The amino terminal intrinsically disordered domain of Dnmt1, residues 92–391, includes a 160 amino acid region unique to eutherian mammals proposed to play a role in mammalian-specific methylation processes such as genomic imprinting (Borowczyk et al., 2009; Liu et al., 2017). Previous research indicated that different portions of this domain may influence the catalytic activity of Dnmt1 at different sequences (Borowczyk et al., 2009). While the primary sequence of Dnmt1 is highly conserved across species (Supplementary Figures S1A–C), Shaffer et al. (Shaffer et al., 2015) identified a 10 amino acid region present in mouse and rat that is not present in humans and suggested that this region may be responsible for species specific methylation. Further analysis illustrated that the mouse-rat region is specific to Mus and Rattus genera, as it is not present in other rodents, including the closely related deer mouse (Peromyscus leucopus and Peromyscus maniculatus) (Figure 1; Supplementary Figures S1D). The Dnmt1 P allele, which substituted the mouse codons specifying LESHTV for the rat codons specifying PEPLSI, is embryonic lethal, displaying dramatically reduced levels of global DNA methylation, indicating that this region does not function similarly in Mus vs. Rattus (Shaffer et al., 2015).

Embryos homozygous for the Dnmt1 P allele have dramatically reduced levels of global methylation but methylation was observed to be better maintained at primary DMRs associated with imprinted genes (Shaffer et al., 2015). Since the methylation at secondary DMRs associated with imprinted genes is dependent on the methylation of the corresponding primary DMR (Saito et al., 2018; Hara et al., 2019), we wanted to determine whether Dnmt1 ^P/P mutant embryos would also retain most of their methylation at secondary DMRs as a consequence of the methylation profile at the associated primary DMR or whether the preferential retention of methylation at primary DMRs is unique in Dnmt1 ^P/P mutant embryos, suggesting either that Dnmt1 functions differently at these sequences or that the maintenance of methylation at primary DMRs can be achieved with other DNA methyltransferases.

We collected and genotyped embryos derived from natural matings between Dnmt1 ^P/+ mice at 9.5, 12.5, 15.5 and 18.5 dpc; at least four litters were collected at each developmental stage. While there was some deviation from the expected 1:2:1 Mendelian ratio at each developmental stage, none of the differences were significant (Table 1). Furthermore, we did not observe consistent differences in morphology between wild-type, heterozygous or homozygous mutant embryos, suggesting that Dnmt1 ^P/P embryos survive throughout gestation (Supplementary Figure S2). In contrast, no Dnmt1 ^P/P pups survived beyond 1 day after birth. Three of the 245 pups that were observed following natural matings between Dnmt1 ^P/+ mice were Dnmt1 ^P/P , and all three were deceased on postpartum day 1. Of the 242 pups that survived beyond postpartum day 1, all survived into adulthood: 84 of the surviving offspring were wild-type, and 158 were heterozygous for the Dnmt1 P allele mutation (Table 2). These data indicate that the Dnmt1 P allele is a neonatal lethal. Our observations of the three dead Dnmt1 ^P/P pups indicated that none of them had milk in their stomachs. We hypothesize that Dnmt1 ^P/P embryos may survive gestation but are unable to eat and/or breathe after birth and are therefore inviable, and that the majority of the Dnmt1 ^P/P pups that were born were consumed by their parents before they were observed on postpartum day 0 or day 1.

We analyzed methylation levels at primary and secondary DMRs associated with imprinted loci as well as at non-imprinted loci in DNA derived from 9.5, 12.5, 15.5 and 18.5 dpc wild-type and Dnmt1 ^P/P siblings (Table 3). Purified genomic embryo DNA was subjected to bisulfite mutagenesis and target loci were amplified by PCR (Supplementary Table S1). Purified amplicons were quantified and pooled at equimolar amounts prior to Next-Generation sequencing. NGS data was analyzed using a Galaxy instance to determine bisulfite mutagenesis efficiency based on non-CpG cytosine conversion to uracil (thymine) and the frequency of cytosine methylation at CpG dinucleotides. Data were obtained from one wild-type and one Dnmt1 ^P/P embryo at 9.5, 12.5 and 15.5 dpc, and from two wild-type and two Dnmt1 ^P/P embryos at 18.5 dpc.

DNA methylation levels were reduced in Dnmt1 ^P/P embryos at all loci examined and at all developmental stages analyzed relative to their wild-type siblings. At primary DMRs, the amount of methylation detected in Dnmt1 ^P/P DNA was between 80% and 100% of the wild-type value at 87% of the sequences analyzed (Figure 2A; Supplementary Table S2). While all primary DMRs except the Airn ICR consistently exhibited reduced levels of DNA methylation in Dnmt1 ^P/P mutant embryos, the difference in methylation between Dnmt1 ^+/+ and Dnmt1 ^P/P embryos was generally less than 16%, suggesting either that the P allele form of Dnmt1 functions reasonably well at these sequences or that an alternative mechanism for maintaining methylation at these sequences exists. Furthermore, the amount of methylation observed at primary DMRs in wild-type and Dnmt1 ^P/P embryos was consistent in biological replicates (Supplementary Figure S3A) and throughout the embryonic stages analyzed (Figure 2A; Supplementary Table S2). The observation that there is some loss of methylation at primary DMRs in Dnmt1 ^P/P embryos suggests that methylation is imperfectly maintained at these sequences during early embryonic development. However, as additional loss of methylation was not observed in Dnmt1 ^P/P mutant embryos as development progressed, whatever deficit the Dnmt1 P allele has in maintaining methylation occurs early and does not accumulate. The H19 ICR displayed more dramatic differences in DNA methylation between the wild-type and Dnmt1 ^P/P samples than the other primary DMRs analyzed, accounting for three of the four primary DMR samples where methylation maintenance was below 80% in Dnmt1 ^P/P mutant embryos (Supplementary Table S2). This could be attributed to the fact that the H19 ICR sample sizes were consistently very small and as a result the data obtained for this locus may not as accurately reflect the DNA methylation patterns present (Supplementary Table S3). Similarly, the observation that the amount of methylation at Airn is higher in Dnmt1 ^P/P samples as compared to their wild-type siblings is likely an artifact associated with the small sample size. We consistently detected methylation levels around 70% for two of the primary DMRs analyzed, Rasgrf1 and Kcnq1ot1, a higher value than would be expected based on their known parent of origin-specific methylation patterns (Figure 2A; Figure 3A). We believe this is likely due to biased amplification of the methylated allele at these loci, which appears to be occurring at the same frequency in the wild-type and Dnmt1 ^P/P embryos (Figure 3A). Methylation levels at the remaining DMRs associated with imprinted loci were detected at expected frequencies in wild-type embryos.

To assess methylation levels at additional primary DMRs in wild-type vs. Dnmt1 ^P/P embryos, we analyzed the 15.5 dpc RRBS data generated by Shaffer et al. (Shaffer et al., 2015). 17 of the 22 ICRs we analyzed were not represented in the tiles generated in their analysis, including all six of the primary DMRs targeted in our study. The RRBS data illustrated that methylation was well maintained at the primary DMRs associated with Nespas/GnasXL, Inpp5f and Peg13, with the percent methylation in Dnmt1 ^P/P embryos being 89, 87% and 86% the level detected in wild-type embryos, respectively. Two ICRs, Fkbp6 and Cdh15, showed more variation between wild-type and Dnmt1 ^P/P embryos, with methylation differences of 24%.

In contrast to what was observed at primary DMRs, DNA methylation levels were dramatically reduced at secondary DMRs in Dnmt1 ^P/P embryos (Figure 2B; Supplementary Figure S3B). At most secondary DMRs, the level of methylation in Dnmt1 ^P/P embryos varied from 13% to 85% of the amount observed in their wild-type siblings. One notable exception to this finding was observed at Igf2r, which displayed a minimal reduction in methylation detected in wild type vs. Dnmt1 ^P/P embryos. This difference could be attributed to the fact that while methylation is acquired at most secondary DMRs by 9.5 dpc, the secondary DMR associated with Igf2r acquires methylation during late gestation (Stöger et al., 1993; Bhogal et al., 2004; Gagne et al., 2014; Guntrum et al., 2017; Nechin et al., 2019). In support of this hypothesis, the average amount of methylation observed at Igf2r was approximately two-fold higher in 18.5 dpc embryos than in embryos collected at earlier developmental stages (Figure 2B; Supplementary Figure S3B). Excluding the Igf2r results, methylation levels in Dnmt1 ^P/P embryos were below 80% the value observed in wild-type embryos at 84% of the sequences analyzed and below 50% in 49% of the sequences analyzed (Supplementary Table S2), considerably less than what was observed at the corresponding primary DMRs. Similar to what was observed at primary DMRs, methylation levels were relatively consistent across development suggesting that for the most part, methylation levels did not change in wild type nor in Dnmt1 ^P/P embryos once it was acquired during early post-implantation development.

Shaffer et al. (Shaffer et al., 2015) illustrated a global loss of DNA methylation in Dnmt1 ^P/P mice by analyzing methylation levels using methylation-sensitive Southern blots to examine methylation levels at IAP elements as well as LUMA assays to examine methylation levels across the genome. We took a targeted approach to examine DNA methylation levels at non-imprinted, single copy sequences. We analyzed methylation at two loci reported to have tissue-specific DNA methylation patterns, Glut3 and Hnf4a (Yagi et al., 2008; Ganguly et al., 2014) as well as four ZFP57-bound loci displaying strain-specific methylation in embryonic stem cells: Zfp553, Qrsl1, Cmtm4 and Talpid3 (Strogantsev et al., 2015). All of the non-imprinted loci showed a reduction in the amount of DNA methylation present in Dnmt1 ^P/P embryos as compared to their wild type siblings (Figure 2C), but the extent to which DNA methylation was lost varied between loci. Methylation was reasonably well maintained at Zfp553 and Talpid3, but was dramatically reduced at Glut3, Hnf4a, Cmtm4 and Qrsl1. Examination of these sequences using the CpG Island Finder (Gardiner-Garden and Frommer, 1987; EMBOSS, 2023), and the UCSC Genome Browser (UCSC Genome Browser) illustrated that CpG density varies at the loci examined in our study. All six primary DMRs and all nine secondary DMRs contain CpG islands or are CpG-rich. In contrast, the regions of Zfp553, Glut3 and Hnf4a analyzed are CpG-rich, but the methylated regions of Cmtm4, Qrsl1 and Talpid3 examined in this study are CpG-poor (Supplementary Figure S4A). Therefore, CpG density does not correlate with the ability of the Dnmt1 P allele to maintain methylation.

We further investigated these loci to determine whether Zfp553 and Talpid3, non-imprinted sequences that retain methylation in Dnmt1 ^P/P mutant embryos, share any features with primary DMRs that might make them resistant to methylation loss. While Zfp553 and Talpid3 display methylation in embryonic stem cells that is presumably gametic in origin (Strogantsev et al., 2015), so do Qrsl1 and Cmtm4, which show dramatic loss of methylation in Dnmt1 ^P/P mutant embryos (Figure 2C). Furthermore, Shaffer and others (Shaffer et al., 2015) illustrated that methylation is lost at gametically methylated IAP elements in Dnmt1 ^P/P mutant embryos. Together, these data suggest that gametic inheritance of methylation does not predict a sequence’s ability to retain methylation in the presence of the P allele form of Dnmt1. We additionally assessed each locus for other chromatin features, including euchromatin vs. heterochromatin status, enhancer vs. promoter vs. transcription unit status, the presence of the transcriptionally permissive histone modifications H3K4me1, H3K4me3 and H3K27ac, and the presence of histone modifications associated with transcriptional repression and/or DNA methylation, H3K9me3, H3K36me3 and H3K27me3, in 12.5 dpc mouse midbrain using the UCSC Genome Browser (Supplementary Figure S4B). While both Zfp553 and Talpid3 are enriched for H3K36me3, the only primary DMR containing this modification is Airn. Based on these analyses, there was no apparent association between a particular chromatin signature and the ability of the corresponding sequence to retain methylation in Dnmt1 ^P/P mutant embryos. Of note, many of the DMRs associated with imprinted loci displayed both permissive and repressive modifications, consistent with the fact that the parental alleles have opposing epigenetic states.

The lower levels of DNA methylation observed in Dnmt1 ^P/P embryos could be due to loss of methylation across a subset of DNA strands, loss of methylation at specific sequences within the DNA, or non-specific loss of methylation across all DNA strands. To distinguish between these possibilities, we extracted a subset of the NGS sequences obtained and analyzed the DNA methylation profiles of individual sequences. Analysis of the extracted primary DMR sequences for Rasgrf1 and the Dlk1-Dio3 IG-DMR showed that the methylation profile amongst alleles obtained from Dnmt1 ^P/P embryos was not significantly different than it was in their wild-type siblings (p values = 0.1802, 0.4839), and further illustrated that methylation was generally well maintained across the sequences analyzed (Figure 3A; additional data not shown). In contrast, the methylation patterns observed at the secondary DMRs associated with Dlk1 and Peg12 in wild type vs. Dnmt1 ^P/P embryos showed significant loss of methylation across each locus (p values = 0.0041, 0.0088), with inconsistent methylation remaining (Figure 3B; additional data not shown). Similar trends were observed at non-imprinted loci that showed either modest or dramatic differences in methylation between wild type and Dnmt1 ^P/P embryos (Figure 3C). Methylation was better maintained at Zfp553 in Dnmt1 ^P/P embryos and was distributed evenly across the sequences analyzed, although the loss of methylation between wild type and Dnmt1 ^P/P embryos was significant (p-value = 0.0005). Cmtm4 displayed a dramatic and dispersed loss of methylation in Dnmt1 ^P/P embryos (p-value = <0.0001). Overall, these data suggest that methylation is poorly maintained across loci in Dnmt1 ^P/P embryos rather than lost entirely from specific sequences.