This research was regulated under the Animals (Scientific Procedures) Act 1986 Amendment Regulations 2012 following ethical review by the University of Cambridge Animal Welfare and Ethical Review Body.

Mtrr ^Gt(XG334)Byg (MGI:3526159) mouse line, referred to as the Mtrr ^gt mouse line, was generated when a β-geo gene-trap (gt) vector was inserted into intron 9 of the Mtrr gene in 129P2Ola/Hsd embryonic stem cells (ESCs) (Elmore et al., 2007; Padmanabhan et al., 2013). Mtrr ^gt ECSs were injected into C57Bl/6J blastocysts and upon germline transmission, the Mtrr ^gt allele was backcrossed into the C57Bl/6J genetic background for at least eight generations (Padmanabhan et al., 2013). Mtrr ^+/+ and Mtrr ^+/gt mice were generated from Mtrr ^+/gt intercrosses. Mtrr ^gt/gt mice were generated by Mtrr ^gt/gt intercrosses. Since the Mtrr ^gt allele has a multigenerational effect (Padmanabhan et al., 2013; Bertozzi et al., 2021; Blake et al., 2021), C57Bl/6J mice from The Jackson Laboratories (www.jaxmice.jax.org) were used as controls and were bred in-house and maintained separately from the Mtrr ^gt mouse line. The effects of the maternal grandpaternal Mtrr ^gt allele were determined by the following pedigree: F0 Mtrr ^+/gt males were mated to C57Bl/6J females. The resulting F1 Mtrr ^+/+ females were mated to C57Bl/6J males to generate F2 Mtrr ^+/+ conceptuses. Genotyping for Mtrr ⁺ and Mtrr ^gt alleles was performed using PCR on DNA extracted from ear tissue or yolk sac using a three-primer reaction resulting in a wildtype band at 252 bp and a mutant band at 383 bp (Padmanabhan et al., 2013). Primer sequences: primer a (5′-GAG​ATT​GGG​TCC​CTC​TTC​CAC), primer b (5′-GCT​GCG​CTT​CTG​AAT​CCA​CAG), and primer c (5′-CG ACT​TCC​GGA​GCG​GAT​CTC) (Padmanabhan et al., 2013). All mice were housed in a temperature-and humidity-controlled environment with a 12 h light-dark cycle. All mice were fed a normal chow diet (Rodent No. 3 chow, Special Diet Services) ad libitum from weaning, which included (per kg of diet): 1.6 g choline, 2.73 mg folic acid, 26.8 μg vitamin B₁₂, 3.4 g methionine, 51.3 mg zinc.

Noon of the day that the vaginal plug was detected was defined as embryonic (E) day 0.5. Mice were euthanized by cervical dislocation. Fetuses and placentas were dissected in cold 1x phosphate buffered saline (PBS) at E10.5 using a Zeiss SteReo Discovery V8 microscope, scored for phenotypes, and photographed. Fetuses and placentas were weighed and measured separately and snap frozen in liquid nitrogen (stored at −80°C). Both male and female placentas were assessed since no phenotypic sexual dimorphism was identified at E10.5 (Padmanabhan et al., 2017).

Conceptuses were rigorously scored for gross phenotypes during dissection and allocated to the phenotypic categories that were previously defined, including phenotypically normal (PN), fetal growth enhancement (FGE), fetal growth restriction (FGR), developmental delay, severe abnormalities (e.g., congenital heart defects, neural tube closure defects, hemorrhages, skewed conceptus orientation, twinning, etc.), and resorption (Padmanabhan et al., 2013; Wilkinson et al., 2021). Notably, conceptuses with >1 phenotype were counted once and classified by the most severe phenotype observed. Only PN, FGR, and FGE conceptuses were assessed in this study. Phenotype parameters are defined below.

Whole placentas at E10.5 were homogenized using a MagNA Lyser Instrument (Roche) and incubated on an Eppendorf ThermoMixer at 1,000 rpm at 56°C for 10 min. Genomic DNA was extracted using a QIAamp Fast DNA Tissue kit (Qiagen) following the manufacturer’s instructions. MeDIP-Seq was carried out as described previously (Ficz et al., 2011). Briefly, genomic DNA was sonicated to yield 150–600 bp fragments, and adaptors for paired-end sequencing (Illumina) were ligated using the NEBNext Ultra II DNA Library Prep Kit for Illumina (New England Biolabs). Immunoprecipitations were carried out using 500 ng DNA per sample, 1.25 μg anti-5mC antibody (Eurogentec Cat# BI-MECY-0100, RRID:AB_2616058) or mouse immunoglobulin G (IgG) control and 10 μL Dynabeads coupled with M-280 sheep anti-mouse antibody (Invitrogen). Pulled down DNA was amplified for 12 cycles (meDIP) or 15 cycles (IgG control) with adapter-specific indexed primers. Final clean-up and size selection was carried out with AMPure-XP SPRI beads (Beckman Coulter). Libraries were quantified and assessed using the Kapa Library Quantification Kit (Kapa Biosystems) and Bioanalyzer 2100 System (Agilent). Indexed libraries were sequenced (50-bp paired-end) on an Illumina HiSeq 2500 sequencer. Raw fastq data were trimmed with TrimGalore (v0.6.6), using default parameters, and unique reads mapped to the Mus musculus GRCm38 genome assembly using Bowtie2 (v2.4.1). Data analysis was carried out using SeqMonk software (www.bioinformatics.babraham.ac.uk).

Whole male placentas at E10.5 were homogenized using lysing matrix D beads. RNA library preparation and sequencing was performed by Cambridge Genomic Services, Department of Pathology, University of Cambridge. The concentration and purity of RNA was determined by a SpectroStar spectrophotometer (BMG LABTECH) and the RNA integrity was determined by an Agilent Tapestation Bioanalyzer (Aligent Technologies LDA United Kingdom Ltd.). Libraries were prepared using 200 ng of total RNA and TruSeq stranded mRNA Library Preparation kit (Illumina). A unique index sequence was added to each RNA library to allow for multiplex sequencing. Libraries were pooled and sequenced on the Illumina NextSeq500 platform with 75 bp single-end reads. Sequencing was performed in duplicate to provide >18 million reads per sample. To monitor sequencing quality control, 1% PhiX Control (Illumina) spike-in was used. Quality control of Fastq files was performed using FastQC and fastq_screen. Sequences were trimmed with Trim Galore! and aligned to GRCm38 mouse genome using STAR aligner. Alignments were processed using custom ClusterFlow (v0.5dev) pipelines and assessed using MultiQC (0.9. dev0). Gene quantification was determined with HTSeq- Counts (v0.6.1p1). Additional quality control was performed with rRNA and mtRNA counts script, feature counts (v 1.5.0- p2) and qualimap (v2.2). Differential gene expression was performed with DESeq2 package (v1.22.2, R v3.5.2). Read counts were normalised on the estimated size factors.

To include transposon-derived reads that do not map uniquely, the meDIP-seq datasets were re-aligned using the default settings of bowtie2 to assign reads with multiple equally best alignments to one of those locations at random. Average methylation levels over pro-viral, full-length elements were generated after merging Repeatmasker annotations for RLTR4_Mm and RLTR4_MM-int elements. RNA-seq data was analysed using SQuIRE (Yang et al., 2019), which assigns multimapping reads using an expectation-maximisation algorithm and provides both subfamily-level and single copy-level information. Differential expression analysis was performed using SQuIRE’s Call function.

First, we analysed the extent to which impaired one-carbon metabolism affected the placental methylome and ascertained whether there was an impact on fetal growth. Our initial focus was on Mtrr ^gt/gt placentas of conceptuses derived from Mtrr ^gt/gt intercrosses (Figure 1A). We carried out high-throughput sequencing of immunoprecipitated methylated DNA (meDIP-seq) from C57Bl/6J control and Mtrr ^gt/gt placentas at E10.5. Mtrr ^gt/gt placentas were divided into two phenotypic groups including those from fetuses that were phenotypically normal (PN) or were FGR based on crown-rump length (Padmanabhan et al., 2013; Wilkinson et al., 2021). Placentas from C57Bl/6J mice were controls since the Mtrr ^gt allele was backcrossed into the C57Bl/6J genetic background (Padmanabhan et al., 2013). However, we previously identified four regions of structural variation between C57Bl/6J and the Mtrr ^gt line (Blake et al., 2021). To avoid false discovery of changes in DNA methylation during the meDIP-seq data analysis, these regions were excluded bioinformatically along with the 20 Mb region of 129P2Ola/Hsd genomic sequence surrounding the gene-trapped Mtrr allele (Bertozzi et al., 2021; Blake et al., 2021), that remained after eight backcrosses (Padmanabhan et al., 2013).

At the global level, we found that the distribution of meDIP-seq reads across different genomic features were not significantly different between C57Bl/6J control and Mtrr ^gt/gt placentas even when phenotypic severity was considered (Supplementary Figure S1A). Furthermore, meDIP-seq datasets from individual placentas did not cluster by Mtrr genotype or fetal growth phenotype when data store similarity tools were implemented (Supplementary Figure S1B). As global DNA methylation patterns were similar between experimental groups, we next ascertained differences in DNA methylation at individual loci compared to control placentas. DMRs were defined using the EdgeR function embedded within Seqmonk software (www.bioinformatics.babraham.ac.uk) with default settings (p < 0.05, with multiple testing correction) assessing 500 bp contiguous regions. The resulting DMRs were further filtered for regions that displayed a log ₂ fold change (FC) > 1 in DNA methylation compared to controls. Only a few DMRs were present in Mtrr ^gt/gt placentas (i.e., PN: 13 DMRs; FGR: 9 DMRs), though both hyper- and hypomethylated regions were observed (Figures 1B, C; Supplementary File S1). The low number of DMRs caused by the Mtrr ^gt allele suggested that DNA methylation changes were subtle regardless of fetal growth phenotypes or were “hidden” by our analysis of whole placentas as individual cell types might be differently affected.

Despite the low number of total DMRs, three key findings emerged (explored further below). Firstly, only one placental DMR associated with the misexpression of a protein-coding gene (Cxcl1; Figures 1D, E). Secondly, we identified two hypermethylated DMRs located within the En2 gene in Mtrr ^gt/gt placentas that were common to PN and FGR conceptuses (Figures 1B, C). Since the En2 DMRs were also identified in mature spermatozoa from Mtrr ^gt/gt males and in Mtrr ^gt/gt embryos at E10.5 (Blake et al., 2021), they were flagged for further analysis in the context of development and epigenetic inheritance. Lastly, of the 12 hypomethylated DMRs that were identified (i.e., 3 shared DMRs in PN and FGR placentas, 5 DMRs in PN placentas only, 4 DMRs in FGR placentas only), ten were associated with endogenous retroviruses (ERVs; Supplementary File S1). Strikingly, the majority of these DMRs (7/10) overlapped with ERV1 elements of the RLTR4 subclass (Figures 1B, C) with implications for genetic stability. Further exploration into the importance of these findings was explored below.

To explore whether altered placental DNA methylation caused by the Mtrr ^gt/gt genotype had a gene regulatory effect, we carried out RNA-seq on Mtrr ^gt/gt placentas at E10.5 associated with PN and FGR fetuses. Using DESeq, the RNA-seq data was assessed for differentially expressed genes that were within 2 kb of a DMR (identified in Mtrr ^gt/gt placentas) and had transcript levels with a log ₂ FC > 0.6 compared to control placentas. The Cxcl1 gene [chemokine (C-X-C motif) ligand 1] was the only dysregulated gene identified in this context. We observed that hypomethylation at the DMR located in the promoter of Cxcl1 was associated with a modest upregulation of Cxcl1 transcripts (Figures 1D, E). This finding exemplifies canonical regulation of a gene by DNA methylation. Since Mtrr ^gt/gt placentas from both PN and FGR fetuses displayed hypomethylation at the Cxcl1 DMR and upregulation of Cxcl1 transcripts (Figures 1D, E), these molecular changes were likely insufficient to drive the fetal growth phenotype. Yet, CXCL1 is important for decidual angiogenesis to promote maternal blood flow into the implantation site (Ma et al., 2021). Therefore, dysregulation of Cxcl1 mRNA in Mtrr ^gt/gt placentas might have implications for fetoplacental development beyond fetal growth.

The only two hypermethylated DMRs identified in Mtrr ^gt/gt placentas were found within the En2 gene. Therefore, their functional importance was explored. These two 500 bp DMRs were in fact contiguous and represented one single 1 kb region in the single En2 intron (Figure 2A). The En2 gene encodes a homeobox transcription factor that, when knocked out in mice, leads to autism-spectrum disease-like behaviours (Cheh et al., 2006; Brielmaier et al., 2012; Provenzano et al., 2014) that are accompanied by cerebellar foliation defects (Joyner et al., 1991) and loss of GABAergic interneurons in somatosensory and visual cortical areas (Sgado et al., 2013; Allegra et al., 2014). Indeed, En2 mRNA is expressed in multiple regions of the developing brain (Davis et al., 1988) and is involved in neurogenesis (Lee et al., 1997). Low levels of En2 transcripts were reported by RNA-seq in the ectoplacental cone (Bastian et al., 2021) (a population of trophoblast progenitor cells in the mouse placenta). However, our RNA-seq data from whole C57Bl/6J control placentas at E10.5 showed that En2 transcripts were very lowly expressed (Figure 2B) and thus, En2 might be considered as an unexpressed gene in the placenta at this developmental stage. Importantly, hypermethylation of the En2 DMR in Mtrr ^gt/gt placentas was not associated with a change in En2 transcript levels (Figure 2B) indicating that the En2 DMR is an unlikely regulator of En2 gene expression in the placenta.

To investigate a broader regulatory role of the En2 DMR in the placenta, we explored histone methylation (e.g., H3K4me3 and H3K27me3) enrichment and potential interactions of the DMR with neighbouring genes. To do this, we analysed published H3K4me3 and H3K27me3 ChIP-seq datasets and promoter capture Hi-C datasets from wildtype mouse trophoblast stem cells (TSCs) (Schoenfelder et al., 2018). TSCs are an in vitro model of undifferentiated trophoblast cells of the placenta (Tanaka et al., 1998), and these datasets represent the most suitable available for analysis. No enrichment of H3K4me3 or H3K27me3 modifications was evident at the En2 DMR in TSCs and no complete DMR-promoter interactions were evident in TSCs within the genomic region assessed (Figures 2B, C). Accordingly, genes downstream of the En2 DMR were also expressed at normal levels in Mtrr ^gt/gt placentas (Figure 2B) indicating that hypermethylation of this region had little to no effect on cis regulation of gene expression in the placenta.

The hypermethylated En2 DMR is prevalent in different tissue types including mature spermatozoa of Mtrr ^gt/gt males and Mtrr ^gt/gt embryos and placentas at E10.5 (Blake et al., 2021; this study). Given that the En2 gene is important for the development of embryonic lineages [e.g., neurogenesis (Lee et al., 1997)], the potential regulatory importance of the En2 DMR was explored outside of the placenta. Additional ChIP-seq and promoter capture Hi-C datasets from wildtype mouse ESCs (Schoenfelder et al., 2018) were analysed in the region proximal to the En2 DMR. The data revealed that the En2 DMR had hallmarks of a regulatory locus in ESCs since it was bivalently marked by the enrichment of repressive H3K27me3 and active H3K4me3 modifications (Figure 2D) in a manner that poises this region for activation upon cell differentiation (Macrae et al., 2023). The genomic region defined by the En2 DMR in ESCs was also enriched for the DNA demethylating enzyme TET1 (Figure 2D), which typically co-localises with polycomb complexes and contributes to keeping unmethylated enhancers and promoters methylation-free (Parry et al., 2021). Furthermore, promoter capture Hi-C experiments in ESCs (Schoenfelder et al., 2018) revealed a potential interaction of the En2 DMR with the promoters of nearby genes, including Cnpy1 (canopy FGF signalling regulator 1), Rbm33 (RNA binding motif 33), and the developmental regulator Shh (sonic hedgehog) (Figure 2D). These data contrasted the TSCs data, which showed no such DMR-promoter interactions (Figure 2C). Therefore, we hypothesised that ectopic hypermethylation of the En2 DMR specifically within Mtrr ^gt/gt embryos might affect expression of surrounding genes with developmental consequences. Further analysis of the developmental role of the En2 DMR in the embryo is required, particularly in the context of abnormal one-carbon metabolism.

We identified ten hypomethylated DMRs in Mtrr ^gt/gt placentas at E10.5 that were associated with ERV elements (Supplementary File S1). Specifically, seven of these overlapped with RLTR4 elements of the ERV1 subfamily (separately annotated as RLTR4_Mm and RLTR4_MM-int for the LTRs and internal region, respectively; Figures 1B, C). RLTR4 elements are relatively young retrotransposons that are closely related to murine leukemia virus and that, at least in some mouse strains, remain transpositionally active (Maksakova et al., 2006). Some transposable elements (e.g., IAPs) are highly methylated and resistant to epigenetic reprogramming to avoid genomic transposition (Kobayashi et al., 2012). It is unclear whether this is the case for RLTR4 elements. The RLTR4 elements that associated with placental DMRs in this study typically displayed a pro-viral, full-length configuration, rather than being solo LTRs or other isolated fragments. Furthermore, six out of seven of the RLTR4-associated DMRs mapped to two discrete genomic loci on mouse chromosomes 11 or 18, including in regions that were intragenic (and antisense) to Camk2b or were upstream of the gene Pik3c3, respectively (Figures 3A, B). Remarkably, a loss of DNA methylation at these DMRs corresponded with ectopic expression of the RLTR4 element in Mtrr ^gt/gt placentas, independent of the fetal growth phenotype (Figures 3A, B). However, no expression changes in the associated protein-coding genes were observed (Figures 3A, B). Therefore, the genomic regions demarcated by these DMRs appear to require methylation to repress the ERV element activity and not to regulate cis gene expression in the placenta.

Due to their repetitive nature and evolutionary young age, the mapability of short sequencing reads to RLTR4 elements is low. Therefore, to fully appreciated the dysregulation of DNA methylation at RLTR4 elements in Mtrr ^gt/gt placentas, the placental meDIP-seq and RNA-seq datasets from C57Bl/6J and Mtrr ^gt/gt placentas at E10.5 were re-mapped to include non-unique reads by using random assignment (bowtie2) for meDIP-seq data and an expectation-maximisation algorithm [SQuIRE (Yang et al., 2019)] for RNA-seq data. When considered globally, the remapped data revealed consistent DNA hypomethylation at the 5′ end of RLTR4 full-length elements in Mtrr ^gt/gt placentas at E10.5 (Figure 3C). This pattern of DNA hypomethylation was associated with transcript enrichment of global RLTR4_Mm and RLTR4_MM-int elements in Mtrr ^gt/gt placentas compared to controls (Figure 3D). Differential expression analysis of individual elements uncovered significant upregulation of twenty-four RLTR4_Mm or RLTR4_MM-int elements that converged upon 15 full-length loci (Supplementary File S2). None of these methylation changes associated with altered expression of nearby protein-coding genes in Mtrr ^gt/gt placentas. Furthermore, methylation and transcriptional dysregulation at RLTR4_Mm and RLTR4_Mm-int elements was unlikely to regulate fetal growth since the RLTR4 elements were similarly affected in Mtrr ^gt/gt placentas associated with PN and FGR fetuses (Figure 3D). Overall, these data reinforced the hypothesis that epigenetic instability is inherent to the Mtrr ^gt mouse line (Padmanabhan et al., 2013; Blake et al., 2021) with implications for genetic stability and phenotype establishment beyond FGR.

In our previous study of epigenetic inheritance in the Mtrr ^gt mouse line (Blake et al., 2021), we found that a small number of candidate DMRs identified in mature spermatozoa (obtained from the cauda epididymis and vas deferens) were not recapitulated in embryos or placentas at E10.5 when interrogated by bisulfite pyrosequencing. Here, we aimed to validate this finding on a genome-wide scale by comparing our spermatozoa (Blake et al., 2021) and placenta meDIP-seq datasets from control and Mtrr ^gt/gt mice. First, we harmonised DMR calling between datasets by reanalysing the spermatozoa meDIP-seq datasets according to our analysis of the placenta meDIP-seq data. Hypermethylated spermatozoa DMRs in Mtrr ^gt/gt males that were previously identified in very highly methylated regions in control spermatozoa and described as false positives (Blake et al., 2021), were also clearly identifiable by the current analysis. Accordingly, we screened out these DMRs using whole genome bisulphite sequencing data to quantify absolute methylation levels across all DMRs (Sun et al., 2018). Similar to our candidate-based approach (Blake et al., 2021), DNA methylation patterns in nearly all genomic regions identified as spermatozoa DMRs in Mtrr ^gt/gt males were normal in Mtrr ^gt/gt placentas at E10.5 compared to control placentas (Figures 4A–C). This finding occurred regardless of the Mtrr ^gt/gt fetal growth phenotype. The only exception was the common hypermethylated En2 DMR that appeared in both spermatozoa and placentas (Figures 4B, C). Conversely, the placental DMRs that overlap with RLTR4 elements were normally methylated in sperm of Mtrr ^gt/gt males (relative to control spermatozoa) suggesting that germline transposon silencing is maintained and unlikely to play a key role in epigenetic inheritance mechanisms in the Mtrr ^gt mouse line. Overall, these results indicated that placenta and spermatozoa DMRs in Mtrr ^gt/gt mice were tissue-specific and that most of the spermatozoa DMRs were effectively reprogrammed in the pre-implantation embryo or during placental development, notwithstanding the shared Mtrr ^gt/gt genotype of the parental and offspring generations. These data might negate DNA methylation as a mechanistic factor in epigenetic inheritance within the Mtrr ^gt mouse line. Instead, altered heritability of other epigenetic factors, such as histone modifications or small non-coding RNA content in germ cells, might be an alternative or additional mechanism.

Previously, our locus-specific analysis showed that some genes associated with spermatozoa DMRs were misexpressed in somatic tissues despite reprogramming of DNA methylation at these sites (Blake et al., 2021). Therefore, we questioned the extent to which this association occurred in the wider placental genome. Using DESeq, the Mtrr ^gt/gt placental RNA-seq dataset was assessed for differentially expressed genes that were within 2 kb of a sperm DMR from Mtrr ^gt/gt males and had transcript levels with a log ₂ FC > 0.6 compared to control placentas. Five misexpressed genes (i.e., Stum (mechanosensory transducer mediator; membrane protein), Tshz3 (teashirt zinc finger family member 3; transcription factor), Ovol2 (ovo like zinc finger 2; transcription factor), Osm (oncostatin m; cytokine), and Ugt1a7c (UDP glucouronosyltransferase 1 family, polypeptide A7C; enzyme in glucouronidation pathway) met these criteria but only in Mtrr ^gt/gt placentas with FGR (Figures 4D–H). The occurrence of transcriptional disruption despite normal DNA methylation reinforced our hypothesis that abnormal one-carbon metabolism influences other epigenetic mechanisms.

Next, we explored whether histone modifications were present in the developing germline at regions demarcated by spermatozoa DMRs to better understand the broader epigenetic context of these regions. To do this, ChIP-seq datasets were analysed for histone mark enrichment in developing wildtype male germ cells (i.e., prospermatogonia) (Shirane et al., 2020) at specific genomic regions defined by spermatozoa DMRs from Mtrr ^gt/gt males. First, we found that 103 out of 252 DMRs (40.9%) overlapped with an H3K4me3 peak in wildtype prospermatogonia, the majority which were located within gene bodies (Supplementary Table S1). This value represented a 12-fold enrichment compared to the baseline genome (i.e., only 3.3% of 500 bp regions across the whole genome overlapped with H3K4me3 peaks) and was substantially more enriched than the other histone modifications at the same locations (Figure 4I). Since H3K4me3 is typically associated with active transcription (Howe et al., 2017), it was an ideal candidate to further explore as an underlying inherited epigenetic mark associated with transcriptional disruption in the placenta. Therefore, we assessed whether wildtype trophoblast progenitor cells at E6.5 (i.e., extraembryonic ectoderm) displayed H3K4me3 enrichment at genomic locations identified as spermatozoa DMRs using a published ChIP-seq dataset (Hanna et al., 2019). Indeed, a substantial subset of these genomic regions was also enriched for H3K4me3 in extraembryonic ectoderm (Figure 4J). Remarkably, four out of five dysregulated genes in Mtrr ^gt/gt placentas that were associated with a spermatozoa DMR in Mtrr ^gt/gt males (i.e., Stum, Tshz3, Ovol2, Ugta7c) were among those enriched for H3K4me3 in both prospermatogonia and extraembryonic ectoderm (Figure 4K–L; Supplementary Figure S2). We infer from these data that the Mtrr ^gt allele potentially disrupts histone marks, such as H3K4me3, in developing and/or mature germ cells leading to altered patterns of the same histone mark in the early conceptus with implications for gene regulation. Indeed, we found that 59.0% of H3K4me3 peaks identified in extraembryonic ectoderm were also found in prospermatogonia (using MACS peak calling function embedded in SeqMonk software). This finding more broadly supports a role for H3K4me3 in inheritance of epimutations from germ cells to the placenta. Future mechanistic experiments should focus on multigenerational patterns of H3K4me3 in the Mtrr ^gt mouse line.

Our previous locus-specific analyses in F2 Mtrr ^+/+ placentas indicated significant alteration of DNA methylation patterns caused by either a maternal grandfather or maternal grandmother Mtrr ^gt allele (Padmanabhan et al., 2013). Here, a transgenerational mechanism was explored in the Mtrr ^+gt maternal grandfather pedigree (Figure 5A) using a genome-wide approach to identify the locations of spermatozoa DMRs from F0 Mtrr ^+/gt males (Blake et al., 2021) and determine whether the placental methylome and transcriptome was altered in these regions two generations later in the F2 Mtrr ^+/+ grandprogeny. The following matings were performed to generate this pedigree (Figure 5A): F0 Mtrr ^+/gt males were mated with C57Bl/6J control females, and the resulting F1 Mtrr ^+/+ females were selected for mating with C57Bl/6J males to generate F2 Mtrr ^+/+ conceptuses. F2 Mtrr ^+/+ conceptuses were rigorously phenotyped at E10.5, and the placentas from PN and FGE fetuses were examined. First, the broader methylome of whole F2 Mtrr ^+/+ placentas was assessed via meDIP-seq. When compared with C57Bl/6J controls, there were no significant differences in the distribution of meDIP reads across genomic features (Supplementary Figure S1C) and no clustering of biological replicates according to phenotype or pedigree (Supplementary Figure S1D) indicating similar global methylation among experimental groups. Few DMRs were identified by the meDIP-seq analysis including one hypermethylated DMR in F2 Mtrr ^+/+ placentas from PN fetuses (Figure 5B) and 11 DMRs (10 hypomethylated, 1 hypermethylated) in FGE-associated F2 Mtrr ^+/+ placentas (Figure 5C). Closer analysis revealed that 9 out of 10 of the hypomethylated DMRs from F2 Mtrr ^+/+ FGE placentas were clustered in two locations on chromosomes 14 and 17, which are frequently susceptible to mapping artefacts in our datasets and so were excluded. The remaining three placental DMRs from F2 Mtrr ^+/+ placentas were in nondescript genomic regions (Supplementary File S1). Importantly, the RLTR4 elements identified in Mtrr ^gt/gt placentas (Figure 3) exhibited normal levels of DNA methylation and transcript expression in the F2 Mtrr ^+/+ placentas relative to control placentas (Supplementary Figure S3). This result suggested that the changes in DNA methylation described in Mtrr ^gt/gt placentas are intrinsically associated with the Mtrr ^gt allele and are unlikely to be transgenerationally inherited or caused by genetic differences between the C57Bl/6J control and Mtrr ^gt mouse lines.

When the meDIP-seq datasets from spermatozoa of F0 Mtrr ^+/gt males (Figure 5D) (Blake et al., 2021) were compared to placentas of F2 Mtrr ^+/+ conceptuses (PN and FGE) at E10.5, there was no DMR overlap (Figures 5E, F). This finding reinforces our hypothesis that specific DMRs in the Mtrr ^gt mouse line are not inherited from germline to somatic cells over multiple generations. This was even the case at the En2 DMR, which was present in spermatozoa from F0 Mtrr ^+/gt males and not in F2 Mtrr ^+/+ placentas (Figure 5G). In this context, the spermatozoa and placenta methylome data revealed that the dosage of the Mtrr ^gt allele in mice correlated with the degree of hypermethylation at the En2 DMR (Figure 5G). Therefore, the En2 locus was particularly responsive to Mtrr-driven disruption of one-carbon metabolism. Altogether, these data further separate the transmission of specific differential methylation patterns via the germline from fetal growth phenotype inheritance in the Mtrr ^gt mouse line.