Disruption of folate metabolism causes germline epigenetic instability and distinguishes HIRA as a biomarker of maternal transgenerational epigenetic inheritance

The mechanism behind transgenerational epigenetic inheritance (TEI) is unclear, particularly through the maternal line. We previously showed that disruption of folate metabolism in mice by the Mtrrgt hypomorphic mutation results in TEI of congenital malformations. Either maternal grandparent can initiate this phenomenon, which persists for at least four wildtype generations. Using genome-wide approaches, we reveal genetic stability in the Mtrrgt model and epigenome-wide differential DNA methylation in the germline of Mtrr+/gt maternal grandfathers. While epigenetic reprogramming occurs, wildtype grandprogeny and great grandprogeny exhibit memory of germline methylation defects. One such region is associated with misexpression of the Hira gene at least until the F3 generation in a manner that distinguishes the HIRA histone chaperone as a biomarker of maternal epigenetic inheritance.


MAIN TEXT
Environmental stressors can impact an individual's health and that of their progeny [1][2][3][4][5] . Phenotypic risk that persists for several generations once the stressor is removed is termed transgenerational epigenetic inheritance (TEI) 6 . While the mechanism is unclear, this non-conventional inheritance likely occurs independent of DNA base-sequence and involves the inheritance of an epigenetic factor(s) via the germline 6,7 . Candidates in mammals include: DNA methylation, histone modifications, and/or non-coding RNA 1,2,4,[8][9][10] . How an epigenetic message resists reprogramming and is transmitted between generations remains elusive.
We reported the Mtrr gt mouse line, a rare model of maternal TEI in which congenital malformations are transgenerationally inherited for at least four wildtype generations 3 . MTRR (methionine synthase reductase) is a key enzyme required for folate and methionine metabolism ( Supplementary Fig. 1) [12][13][14] . Folate is best known for its role in neural tube closure, yet its function in development is complex and remains poorly understood. Folate metabolism is required for thymidine synthesis 15 and cellular methylation. It provides one-carbon methyl groups for the methylation of homocysteine by methionine synthase (MTR) to form methionine and tetrahydrofolate 16 . Methionine acts as a precursor for S-adenosylmethionine (SAM), which serves as the sole methyl-donor for substrates involved in epigenetic regulation including DNA, RNA, and proteins [17][18][19] . MTRR activates MTR through the reductive methylation of its vitamin B 12 cofactor 14 .
Consequently, MTRR helps to maintain genetic and epigenetic stability during pregnancy.
The Mtrr gt mutation in mice knocks down Mtrr expression, reduces MTR enzymatic activity, and consequently disrupts folate metabolism 3,12 . Similar to humans with an MTRR mutation 13,[20][21][22] or dietary folate deficiency 23 , Mtrr gt/gt mice display hyperhomocysteinemia 3,12 , macrocytic anaemia 24 , and a wide spectrum of developmental phenotypes at midgestation (e.g., growth defects and congenital malformations including neural tube, heart and placenta defects) 3 . Therefore, Mtrr gt mice are relevant for studying the effects of abnormal folate metabolism.
Furthermore, the Mtrr gt mouse line is a model of maternal TEI. Through highly-controlled genetic pedigrees (Supplementary Fig. 2a-b), we determined that a male or female carrier of the Mtrr gt allele (i.e., Mtrr +/gt ) is sufficient to cause TEI of developmental phenotypes in wildtype (Mtrr +/+ ) descendants 3 . Phenotypic inheritance occurs only via wildtype daughters for at least four generations 3 in conditions of normal folate metabolism 24 . The spectrum and frequency of developmental phenotypes is largely comparable in each generation, regardless of whether an Mtrr +/gt maternal grandmother or grandfather initiated the effect 3 . An exception is the F1 generation where phenotypic risk occurs only when individuals are derived from an F0 Mtrr +/gt female ( Supplementary Fig. 2a) 3,25 . Though the mechanism is unclear, this phenomenon suggests a maternally inherited factor. Embryo transfer of F2 blastocysts demonstrated that congenital malformations (but not the growth phenotypes) occurred independent of a defective uterine environment 3 (Supplementary Fig. 2c) and emphasizes the mechanistic importance of germline epigenetic inheritance in the Mtrr gt model. Epigenetic instability occurs in the Mtrr gt model, particularly in placentas of wildtype F1 and F2 generations as indicated locus-specific dysregulation of DNA methylation associated with gene misexpression 3

. A genome-wide analysis
has not yet been performed.
Here, we investigate potential mechanism(s) of TEI in the Mtrr gt model. First, we demonstrate that Mtrr gt/gt mice are genetically stable and hence reassert focus on an epigenetic mechanism. Second, we show that the Mtrr gt mutation alters differential DNA methylation in the germline. Sperm were chosen for analysis because of their experimental tractability and because an F0 Mtrr +/gt male can initiate TEI through his wildtype F1 daughters. We observe that differentially methylated regions (DMRs) in sperm of F0 Mtrr +/gt males are reprogrammed in somatic tissue of wildtype progeny and grandprogeny. Yet, memory of germline epigenetic disruption persists at least until the F3 generation. This includes misexpression of Hira, a gene important for chromatin stability. We propose Hira as a biomarker of maternal TEI in the Mtrr gt model.

Genetic stability in Mtrr gt mice
Since folate metabolism is directly linked to DNA synthesis, we first addressed whether the Mtrr gt allele influences genetic stability. Whole genome sequencing (WGS) was performed on phenotypically normal C57Bl/6J control embryos (n=2) and Mtrr gt/gt embryos with congenital malformations (n=6) (Supplementary fig. 2b,e). DNA libraries were sequenced resulting in ~30x coverage per sample (~3.5 x 10 8 paired-end reads/genome). The sequenced genomes were compared to the C57Bl/6J reference genome to identify structural variants [SVs] and single nucleotide polymorphisms [SNPs].
The Mtrr gt mutation was generated by a genetrap (gt) insertion in the 129P2Ola/Hsd background before backcrossing for eight generations into the C57Bl/6J strain 3 . As expected, the majority of variants identified in Mtrr gt/gt embryos were located on Chr13 in the genomic region surrounding the Mtrr gene (Supplementary Fig. 3a-b). Many SNPs in this region showed sequence similarity to the 129P2Ola/Hsd genome and likely persisted due to Mtrr gt genotype selection and regional crossover frequency. Using these SNPs, we defined a 20 Mb region of 129P2Ola/Hsd sequence surrounding the Mtrr gt allele (Fig. 1a) Fig. 3c-d). Moreover, genetic variation within the masked region had minimal functional effect (beyond that of the gene-trap insertion) since expression of individual genes from this region was similar among C57Bl/6J, 129P2Ola/Hsd and Mtrr gt/gt mice (Fig. 1a,d). As further indication of genomic stability, we showed that expression of several transposable elements 26 was similar in C57Bl/6J and Mtrr gt/gt tissue (Fig. 1e) indicating that their repressive state was preserved. Overall, these data suggested that the genome of the Mtrr gt model was stable, and that phenotypic inheritance was unlikely caused by an increased frequency of de novo mutation. Therefore, focus shifted to an epigenetic mechanism.  Fig. 1 continued] ± standard deviation (sd). Independent t test. d, Graphs showing RT-qPCR analysis of selected genes (highlighted red in a) in embryos (E10.5), placentas (E10.5) and/or adult testes from C57Bl/6J (black bars), 129P2Ola/Hsd (grey bars) and phenotypically-normal Mtrr gt/gt (blue bars) mice. e, Data indicating RNA expression of specific groups of transposable elements as determined by RT-qPCR in C57Bl/6J tissue (black bars) and Mtrr gt/gt tissue (blue bars: phenotypically normal; white bars: severely affected). Adult liver and placenta at E10.5 were assessed. Data from RT-qPCR analyses in (d, e) are shown as mean ± sd and relative to C57Bl/6J tissue levels (normalized to 1). N=5-6 samples per group. One-way ANOVA with Dunnett's multiple comparison test, *p<0.05; **p<0.01.

Germline DNA methylation is altered in the Mtrr gt model
MTRR plays a direct role in the transmission of one-carbon methyl groups for DNA methylation 3,12,14 . Therefore, germline DNA methylation was considered as a potential mediator of phenotype inheritance. Since an Mtrr +/gt maternal grandmother (MGM) or maternal grandfather (MGF) initiates TEI (Supplementary Fig. 2a) 3,6 and due to the experimental tractability of male gametes, we chose to focus our analysis on sperm. Mature spermatozoa were collected from C57Bl/6J, Mtrr +/+ , Mtrr +/gt and Mtrr gt/gt mice (Supplementary Fig. 2b,d-e) and the purity was confirmed by assessing imprinted regions of known methylation status via bisulfite pyrosequencing ( Supplementary Fig. 4a). Global 5-methylcytosine (5mC) and 5-hydroxymethylcytosine (5hmC) levels were normal across all Mtrr genotypes relative to C57Bl/6J controls as determined by mass spectrometry (Fig. 2a).
derived from Mtrr +/gt intercrosses (Supplementary Fig. 2d) indicated a parental effect of the Mtrr gt allele. Hypomethylated and hypermethylated regions were identified in each Mtrr genotype when compared to C57Bl/6J controls (Fig. 2c), consistent with earlier findings in placentas 3 . These data suggested that the Mtrr gt allele was sufficient to dysregulate sperm DNA methylation.
To ensure the robustness and reliability of the MeDIP-seq data, we randomly selected hyper-and hypomethylated DMRs to validate using bisulfite pyrosequencing. Sperm DNA from C57Bl/6J, Mtrr +/+ , Mtrr +/gt and Mtrr gt/gt males was assessed (N=8 males/group: four sperm samples from MeDIP-seq experiment plus four independent samples). DMRs were validated in the Mtrr genotype in which they were identified (Figs. 2d, 3, Supplementary Fig. 5). The overall validation rate was 94.1% in hypomethylated DMRs and 58.3% in hypermethylated DMRs (Supplementary Table 1) and indicated a high degree of corroboration between techniques. The majority of DMRs that did not validate showed extensive methylation (>80% CpG methylation) in C57Bl/6J sperm and were identified as hypermethylated in the MeDIP-seq experiment (Supplementary Fig. 5). This might reflect some false positives in line with another study 4 .
For most DMRs assessed, methylation change was consistent across all CpG sites and the absolute change in CpG methylation ranged from 10 to 80% of control levels (Figs. 2d, 3,   Supplementary Fig. 5). Within each genotypic group, a high degree of inter-individual consistency of methylation change was also observed. Therefore, we conclude that the Mtrr gt mutation, or parental exposure to it as in Mtrr +/+ males, is sufficient to lead to distinct DNA methylation changes in sperm.

Most DMRs associated with metabolic dysregulation rather than genetic effects
A proportion of the DMRs was located within the region around the gene-trap insertion site in Mtrr +/gt and Mtrr gt/gt males (Fig. 1a, Supplementary Fig. 6), consistent with Mtrr gt/gt liver 29 and suggesting that the gene-trap or underlying 129P2Ola/Hsd sequence might epigenetically dysregulate the surrounding region. Comparison to the whole genome sequencing data revealed that genetic variation did not influence DMR calling to a great extent since only a small proportion (2.8-5.5%) of these DMRs contained one or more SNP. Outside of the Mtrr genomic region, 54 DMRs were common to Mtrr +/+ , Mtrr +/gt and Mtrr gt/gt males (Fig. 2e) and were primarily located in distinct chromosomal clusters (Supplementary Fig. 6). These data implicate epigenetic hotspots or underlying genetic effects. However, beyond a polymorphic duplication on Chr19 in the C57Bl/6J strain 30

Sperm DMR genomic distribution and potential regulatory function
DMR distribution was determined to explore regional susceptibility of the sperm methylome to the Mtrr gt allele. First, the sperm 'background methylome' was established to resolve the expected genome-wide distribution of CpG methylation (see Methods). By comparing the regional distribution of sperm DMRs to the background methylome, DMRs in all Mtrr genotypes were not significantly enriched in repetitive regions (Fig. 2f). However, sperm of Mtrr +/+ and Mtrr +/gt males had an over-representation of DMRs in introns and exons, and under-representation in intergenic regions (p<0.0003, Chi-squared test; Fig. 2g). This was not the case for Mtrr gt/gt males since DMRs became more proportionately distributed among most genomic regions (Fig. 2g). While the majority of sperm DMRs were located within CpG deserts, DMRs from Mtrr +/gt and Mtrr gt/gt males were over-represented in CpG islands (p<0.0014, Chi-squared; Fig. 2h). Altogether, these results might have implications for gene regulation.
During sperm maturation, histones are replaced by protamines 31 . However, nucleosome retention occurs primarily at promoters of developmentally-regulated genes and gene-poor repeat regions providing scope for epigenetic inheritance [32][33][34] . Using a published data set 34 , the genomewide nucleosome retention rate was estimated at 1.94% as determined by assessing 10,000 randomly-selected 500 bp windows. Some DMRs were significantly enriched at nucleosome retention regions (14.5-34.1% of DMRs, p<0.0001, binomial test; Table 1) and represented key candidate regions for epigenetic inheritance. Therefore, to explore the epigenetic signature of DMRs, mean enrichment for histone modifications and/or Tn5 transposase sensitive sites (THSS) in mouse spermatozoa 37 , epiblast 38 and extraembryonic ectoderm (ExE) 38 at E6.5 was determined using published ChIP-seq and ATAC-seq data sets. All DMRs, except those surrounding the Mtrr gene-trapped site, were analysed (N=379 DMRs, all Mtrr genotypes combined) alongside randomly selected regions (N=379; see Methods) representing the 'baseline genome'. In sperm 37 , DMRs were collectively enriched for the repressive histone mark H3K9me3 and not active histone marks (e.g., H3K4me1, H3K27ac) compared to the background methylome ( Supplementary Fig. 7a), and were more likely to associate with a closed chromatin state due to enrichment for protamine 1 (PRM1) and not THSS (Supplementary Fig. 7a-b). In contrast, the same regions in epiblast and ExE tissues 38 were more likely in an open chromatin conformation when compared to tissue-specific baseline methylomes (Supplementary Fig. 7c). Therefore, a regulatory role for the DMRs during development is possible and required further evaluation (see below).

DMRs were located in regions of reprogramming resistance
DNA methylation is largely reprogrammed during pre-implantation development and in the developing germline 39,40 . Recently, several loci were identified as 'reprogramming resistant' 35,36,41 .
Using published data sets 35,36 , we determined that 40.7-54.3% of sperm DMRs across all Mtrr genotypes fell within loci resistant to pre-implantation reprogramming ( Table 1). Sixteen of these DMRs were common among Mtrr +/+ , Mtrr +/gt and Mtrr gt/gt males. Fewer DMRs correlated with regions resistant to germline reprogramming or both pre-implantation and germline reprogramming (2.2-3.8% and 2.0-2.7% of DMRs/Mtrr genotype, respectively; Table 1). Only one DMR located in a region resistant to germline reprogramming was common to all genotypes. Interestingly, several DMRs in reprogramming resistant regions 35,36 also overlapped with regions of nucleosome retention 34 (Table 1). Overall, DMRs in these key regions might be important for epigenetic inheritance.

Sperm DMRs are reprogrammed in wildtype F1 and F2 generations
TEI in the Mtrr gt model occurs via the maternal lineage. However, an F0 Mtrr +/gt male can initiate the effect 3 (Supplementary Fig. 2a). To determine the heritability of sperm DMRs, bisulfite pyrosequencing was used to validate ten DMRs from F0 Mtrr +/gt males in tissue of wildtype F1 and F2 progeny. Candidate DMRs displayed features including reprogramming resistance, localization to intragenic or intergenic regions, and/or hyper-or hypomethylation (Supplementary Table 2). In general, all DMRs tested lost their differential methylation in wildtype F1 and F2 embryos and placentas at embryonic day (E) 10.5, and showed DNA methylation patterns similar to C57Bl/6J tissue (Fig. 3). This result occurred even when wildtype F2 conceptuses displayed congenital malformations (Fig. 3). DMRs were also assessed in Mtrr gt/gt conceptuses at E10.5 to determine whether these regions were capable of differential methylation outside of the germline. Seven out of 10 sperm DMRs were hypermethylated in Mtrr gt/gt embryos and/or placentas (Supplementary CpG methylation at specific sperm DMRs identified in F0 Mtrr +/gt males was assessed in wildtype embryos and placentas at E10.5 from F1 and F2 progeny. Pedigrees indicate specific mating scheme. Pedigree legend: circle, female; square, male; blue outline, C57Bl/6J control; black outline, Mtrr gt mouse line; white filled, Mtrr +/+ ; half-white/half-black filled, Mtrr +/gt . a-j, Schematic drawings of each DMR assessed indicate its relationship to the closest gene and are followed by graphs showing the average percentage of methylation at individual CpGs for the corresponding DMR as determined by bisulfite pyrosequencing. In each case, methylation was assessed in sperm from F0 Mtrr +/gt males (green line), phenotypically normal F1 wildtype (Mtrr +/+ ) embryos and [ Fig. 3 continued] placentas at E10.5 (orange lines), and phenotypically normal (purple line) or severely affected (pink line) F2 wildtype (Mtrr +/+ ) embryos and placentas at E10.5. C57Bl/6J (black lines) are shown as controls. N=4-8 individuals assessed per group. Data is shown as mean ± sd for each CpG site. Two-way ANOVA, with Sidak's multiple comparisons test, performed on mean methylation per CpG site per genotype group. *p<0.05, **p<0.01, ***p<0.001. Fig. 8a-j), a similar pattern to sperm from Mtrr +/gt males (Fig. 3). It was unclear whether these regions resisted reprogramming, or were erased and re-established or maintained due to intrinsic Mtrr gt/gt homozygosity. Overall, the altered DNA methylation in sperm of Mtrr +/gt males was not evident in somatic tissue of wildtype progeny and grandprogeny.

Epigenetic memory of sperm DMRs
Previous studies in an intergenerational model suggested that sperm DMRs might be associated with perturbed transcription in offspring even when DNA methylation was re-established to normal levels 4 . Therefore, expression of six genes located in or near sperm DMRs identified in F0 Mtrr +/gt males was assessed in F1 and F2 wildtype individuals. While these genes displayed normal expression in F1 tissues (Fig. 4a-c), Hira (histone chaperone), Cwc27 (spliceosome-associated protein), and Tshz3 (transcription factor) (Supplementary Table 2) were misexpressed in F2 wildtype embryos or adult livers compared to C57Bl/6J controls (Fig. 4d-f). This result might reflect an epigenetic memory of the associated sperm DMR or wider epigenetic dysregulation in sperm of the F0 Mtrr +/gt males.

HIRA as a potential mediator of maternal inheritance in the Mtrr gt model
The Hira DMR was further considered based on its resistance to germline reprogramming 36 and potential for epigenetic memory (Fig. 4d). HIRA is a histone H3.3 chaperone central to transcriptional regulation 48 , and to maintenance of chromatin structure during oogenesis 49 and in the male pronucleus after fertilization 50 . Hira -/mice 51 and the Mtrr gt mouse line 3 display similar phenotypes including growth defects, congenital malformations and embryonic lethality by E10.5.
Furthermore, Mtrr gt genotypic severity correlated with the degree of hypermethylation in the Hira DMR in sperm (Fig. 5b-c) suggesting that the Hira DMR is responsive to alterations in folate metabolism.
Next, the potential regulatory legacy of the Hira DMR in sperm was explored in the Mtrr +/gt MGF pedigree. We previously reported that nearly all MGF F1 conceptuses from this pedigree were phenotypically normal at E10.5 whereas F2-F3 wildtype conceptuses displayed a wide spectrum and frequency of developmental phenotypes 3 (Supplementary Fig. 2a). Here, we showed substantial hypermethylation of the Hira DMR (39.0 ± 4.1% average increase per CpG) was observed in sperm of F0 Mtrr +/gt males compared to controls (Fig. 5c), yet the Hira DMR was reprogrammed in MGF F1-F3 wildtype embryos and placentas (Fig. 3a, Supplementary fig. 11a).
Conversely, significant dysregulation of Hira lncRNA expression was apparent only in MGF F1 wildtype embryos (Fig. 6c-e). As phenotypic severity increased, so too did the degree of Hira mRNA dysregulation in the MGF F2 and F3 embryos (Fig. 6d-e). Similar to Mtrr gt/gt conceptuses, placental Hira RNA expression was unaffected (Fig. 6a,c-d).
To investigate a link between Hira expression and phenotypic inheritance, we analysed wildtype F1-F3 conceptuses at E10.5 derived from F0 Mtrr +/gt maternal grandmothers ( Supplementary Fig. 2a), of which all generations display a wide spectrum of developmental phenotypes 3 . Supporting our hypothesis, Hira mRNA expression was down-regulated in MGM F1-F3 wildtype embryos compared to C57Bl/6J controls (Fig. 6f). As expected, Hira lncRNA transcripts were unchanged in MGM F1 and F3 wildtype embryos, yet were down-regulated in MGM F2 embryos (Fig. 6f), which display the highest frequency of phenotypes among the three generations 3 . Overall, these data indicated that Hira RNA is a potential marker of maternal phenotypic inheritance.
Disruption of post-transcriptional regulation of HIRA was also apparent in F2 embryos and placentas of both Mtrr +/gt maternal grandparental pedigrees, yet in a manner not predicted by Hira mRNA expression. For example, down-regulated Hira mRNA (Fig. 6c-d,f) was associated with normalized or up-regulated HIRA protein ( Fig. 6g-h, Supplementary fig. 11c-d), and normal Hira mRNA was associated a down-regulation of HIRA protein (Fig. 6d,g, Supplementary fig. 11c-d).
Importantly, dysregulation of Hira RNA and HIRA protein correlated with a pattern of maternal phenotype inheritance.

DISCUSSION
We investigated potential mechanisms contributing to epigenetic inheritance in Mtrr gt mice, a unique model of mammalian TEI 3 . In the Mtrr gt model, inheritance of developmental phenotypes and epigenetic instability occurs via the maternal lineage, though the mechanism is complex because an Mtrr +/gt female or male can initiate the effect through their daughters and granddaughters 3 . Due to its experimental tractability, we assessed DNA methylation in spermatozoa to understand how the germline epigenome was affected by the Mtrr gt allele. We Dietary folate deficiency causes differential methylation in sperm 55 , though whether it causes TEI is unknown. Similar phenotypes appear in Mtrr gt mutation in mice and in folate-deficient humans 13,20-23 , but have not been reported in genetically wildtype mice fed folate-deficient diets 56,57 . There was no overlap between sperm DMRs in the diet model versus Mtrr +/gt males, which disputes the existence of folate-specific epigenomic hotspots in sperm. Severity of insult or technical differences (e.g., MeDIP-array 55 versus MeDIP-seq) might explain this discrepancy.
Whether DNA methylation patterns observed in F0 sperm are reconstructed 58,59

in the F1
germline is yet-to-be determined in the MGF pedigree. Exploring oocyte methylation has significant challenges and comparing DNA methylation patterns in F0 sperm to F1 oocytes might be incongruous. However, we showed that sperm of Mtrr +/+ males derived from Mtrr +/gt intercrosses exhibited several DMRs that were independent of genetic variation yet overlapped with sperm DMRs in Mtrr +/gt males (representing their fathers). Therefore, reconstruction of specific atypical F0 germline methylation patterns likely occurs in the F1 germline in the Mtrr gt model. In contrast, vinclozolin toxicant exposure of rats results in TEI and dissimilar DMRs in spermatozoa of F1 and F3 offspring 8 . However, epigenetic patterns might shift as generational distance from the F0 individual increases. The effects of vinclozolin on testis formation 60 , not observed in the Mtrr gt model 61 , might explain the discrepancies between TEI models.

Hira transcription emerged as a biomarker of maternal phenotypic inheritance even though
Hira was identified through an associated sperm DMR. Specifically, embryos derived from wildtype oocytes of Mtrr +/gt females (or of wildtype females with Mtrr gt ancestry) displayed a broad spectrum and frequency of developmental phenotypes alongside abnormal Hira mRNA expression at E10.5.
In contrast, embryos derived from wildtype sperm of Mtrr +/gt males displayed normal phenotypes and Hira mRNA levels, yet Hira lncRNA was misexpressed at E10.5. The HIRA histone chaperone complex is important for histone deposition/recycling to maintain chromatin integrity during transcription 62 . The function of Hira lncRNA is unknown, though lncRNA-based mechanisms often control cell fates during development by influencing nuclear organization and transcriptional regulation 63 . Regardless, Hira -/embryos 51 phenocopy embryos in the Mtrr gt model 3 suggesting general epigenetic instability is detrimental to embryo development, and that Hira transcription might be responsive to epigenetic instability caused by the Mtrr gt allele.
It is of interest that F1 progeny are phenotypically different when derived from an F0 Mtrr +/gt male versus F0 Mtrr +/gt female 3 (Supplementary fig. 2a). The difference might relate to a maternally-supplied factor present in ooplasm that is cytoplasmically-inherited by the F1 wildtype zygote (Fig. 7). Maternal Hira mRNA and protein are present in the oocyte and zygote 49,64 , where it is involved in protamine replacement by histones in the paternal pronucleus 50 . Therefore, HIRA is a maternal factor suitably placed to perpetuate epigenetic instability between generations in the  To determine the multigenerational effects of the Mtrr gt allele in the maternal grandfather (MGF), the following mouse pedigree was established (Supplementary Fig. 2a). For the F1 generation, F0 Mtrr +/gt males were mated with C57Bl/6J females and the resulting Mtrr +/+ progeny were analysed. For the F2 generation, F1 Mtrr +/+ females were mated with C57Bl/6J males and the resulting Mtrr +/+ progeny were analysed. For the F3 generation, F2 Mtrr +/+ females were mated with C57Bl/6J males and the resulting Mtrr +/+ progeny was analysed. A similar pedigree was established to assess the effects of the Mtrr gt allele in the maternal grandmother (MGM) with the exception of the F0 generation, which involved the mating of an Mtrr +/gt female with a C57Bl/6J male.
Tissue dissection and phenotyping. For embryo and placenta collection, timed matings were established and noon on the day that the vaginal plug was detected was considered embryonic day (E) 0.5. Embryos and placentas were dissected at E10.5 in cold 1x phosphate buffered saline and were scored for phenotypes (see below), photographed, weighed, and snap frozen in liquid nitrogen for storage at -80°C. Livers were collected from pregnant female mice (gestational day 10.5), weighed and snap frozen in liquid nitrogen for storage at -80°C.
A rigorous phenotyping regime was performed at E10.5 as previously described 3 . Briefly, all conceptuses were scored for one or more congenital malformation including failure of the neural tube to close in the cranial or spinal cord region, malformed branchial arches, pericardial edema, reversed heart looping, enlarged heart, and/or off-centered chorioallantoic attachment. Twinning or haemorrhaging was also scored as a severe abnormality. Embryos with <30 somite pairs were considered developmentally delayed. Embryos with 30-39 somite pairs but a crown-rump length more than two standard deviations (SD) from the mean crown-rump length of C57Bl/6J control embryos were considered growth restriction or growth enhanced. Conceptuses were considered phenotypically-normal if they were absent of congenital malformations, had 30-39 somite pairs and had crown-rump lengths within two SD of controls.
Spermatozoa collection. Spermatozoa from cauda epididymides and vas deferens were collected from 16-20 week-old fertile mice as previously described 67 with the following amendments.
Samples were centrifuged at 500 x g (21°C) for 10 minutes. The supernatant was transferred and centrifuged at 1,300 x g (4°C) for 15 minutes. After the majority of supernatant was discarded, the samples were centrifuged at 1,300 x g (4°C) for 5 minutes. Further supernatant was discarded and the remaining spermatozoa were centrifuged at 12,000 x g for 1 minute and stored at -80°C.
Nucleic Acid Extraction. For embryo, trophoblast and liver tissue, genomic DNA (gDNA) was extracted using DNeasy Blood and Tissue kit (Qiagen) according to the manufacturer's instructions. RNA was extracted from tissues using the AllPrep DNA/RNA Mini Kit (Qiagen). For sperm, Solution A (75 mM NaCl pH 8; 25 mM EDTA) and Solution B (10 mM Tris-HCl pH 8; 10 mM EDTA; 1% SDS; 80 mM DTT) were added to the samples followed by RNAse A incubation (37°C, 1 hour) and Proteinase K incubation (55°C, overnight) as was previously described (Radford et al. 2014). DNA was extracted using phenol/chloroform/isoamyl alcohol mix (25:24:1) (Sigma-Aldrich) as per the manufacturer's instructions. DNA was precipitated using 10 M ammonium acetate, 0.1 mg/ml glycogen, and 100% ethanol and incubated at -80°C for at least 30 minutes. DNA was collected by centrifugation (13,000 rpm, 30 minutes). The pellet was washed twice in 70% ethanol, air-dried, and resuspended in TE buffer. DNA quality and quantity was confirmed using gel electrophoresis and QuantiFluor dsDNA Sample kit (Promega) as per the manufacturer's instructions.
Structural variant analysis was performed using Manta 70 . Structural variants (SVs) were filtered using vcftools (version 0.1.15) 71 . In order to identify single nucleotide polymorphisms (SNPs), the data was remapped to the mm10 reference mouse genome using BWA (version 0.7.15-r1144-dirty) 72 . Reads were locally realigned and SNPs and short indels identified using GenomeAnalysisTK (GATK, version 3.7) 73 . Homozygous variants were called when more than 90% of reads at the locus supported the variant call, whereas variants with at least 30% of reads supporting the variant call were classified as heterozygous. Two rounds of filtering of variants were performed as follows. Firstly, low quality and biased variant calls were removed. Secondly, variants with: i) simple repeats with a periodicity <9 bp, ii) homopolymer repeats >8 bp, iii) dinucleotide repeats >14 bp, iv) low mapping quality (<40), v) overlapping annotated repeats or segmental duplications, and vi) >3 heterozygous variants fell within a 10 kb region were removed using vcftools (version 0.1.15) as was previously described 74  Quality assessment of the sequencing reads was performed using FastQC (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Adaptor trimming was performed using Trim Galore (http://www.bioinformatics.babraham.ac.uk/projects/trim galore/). Reads were mapped to the GRCm38 (mm10) reference genome using Bowtie2 (http://bowtiebio.sourceforge.net/bowtie2/index.shtml) 69 . All programmes were run with default settings unless otherwise stated. Sample clustering was assessed using principle component analysis (PCA), using the 500 most variable windows with respect to read coverage (as a proxy for methylation) for 5 kb window across all samples. Further data quality checks and differential methylation analysis was performed using the MEDIPS package in R 28 . The following key parameters were defined: BSgenome = BSgenome.Mmusculus.UCSC.mm10, uniq = 1e-3, extend = 300, ws = 500, shift=0.
Differentially methylated regions (DMRs) were defined as windows (500 bp) in which there was at least 1.5-fold difference in methylation (reads per kilobase million mapped reads (RPKM)) between Homogenates were then incubated on ice for 20 minutes with brief intervening vortexing steps occurring every 5 minutes. Samples were then centrifuged at 10,000 x g for 5 minutes.
Supernatant from each sample was transferred to a new tube and centrifuged again at 10,000 x g for 5 minutes to ensure that all residual tissue was removed. Protein concentration of tissue lysates was determined using bicinchoninic acid (Sigma-Aldrich). Proteins were denatured with gel loading buffer (50 mM Tris [pH 6.8], 100 mN DTT, 2% SDS, 10% glycerol and trace amount of Statistical analysis. Statistical analysis was performed using GraphPad Prism software (version 7). RT-qPCR data were analysed by independent unpaired t tests or ordinary one-way ANOVA with Dunnett's or Sidak's multiple comparison testing. SV and SNP data were analysed by independent unpaired t tests with Welch's correction. Bisulfite pyrosequencing data, SV chromosome frequency and DMR distribution at repetitive elements were analysed by two-way ANOVAs with Dunnett's, Sidak's or Tukey's multiple comparisons tests. Western blot data were analysed using a non-parametric Kruskal-Wallis test with Dunn's multiple comparisons test. p<0.05 was considered significant unless otherwise stated. Highly controlled genetic pedigrees demonstrated that the Mtrr gt mouse line is a model of transgenerational epigenetic inheritance 3 , many of which are used in this study. Grey shaded boxes indicated pedigrees that do not display phenotypes at E10.5 in the final generation shown 3 . Pink shaded boxes indicate pedigrees that display a wide spectrum of phenotypes at E10.5 in the final generation shown 3 . a Mtrr +/gt maternal grandfather pedigree (top panel) and Mtrr +/gt maternal grandmother pedigree (bottom panel). b Control C57Bl/6J pedigree. c Embryo transfer of specified F2 wildtype blastocysts into control B6D2F1 pseudopregnant females demonstrated that some of the developmental phenotypes at E10.5 were independent of the F1 uterine environment 3 . d Mtrr +/gt intercross pedigree. e Mtrr gt/gt intercross pedigree. Pedigree legend: circle, female; square, male; blue outline, C57Bl/6J mouse line; black outline, Mtrr gt mouse line; pink outline, B6D2F1 mouse line; white fill, Mtrr +/+ ; half black-half white fill, Mtrr +/gt ; black fill, Mtrr gt/gt . Supplementary Fig. 4. Confirmation of spermatozoa purity and validation of immunoprecipitation. a Bisulphite pyrosequencing of imprinting control regions in DNA from spermatozoa collected from cauda epididymides to determine sperm purity. Percentage methylation at specific CpG sites in the maternally imprinted Peg3 DMR and paternally imprinted H19 DMR were determined in sperm samples isolated from C57Bl/6J (black line), Mtrr +/+ (purple line), Mtrr +/gt (green line) and Mtrr gt/gt (blue line) mice. C57Bl/6J liver (grey line) was assessed as a control. Data is represented as mean ± sd for each CpG site. N=8 samples per group. b Percentage recovery of DNA input after MeDIP experiment as determined using qPCR to amplify known methylated (Nanog and H19) and unmethylated (H1t and TsH2B) regions. MeDIP samples from sperm of C57Bl/6J (black), Mtrr +/+ (purple), Mtrr +/gt (green), and Mtrr gt/gt (blue) males are shown. Each bar indicates one individual (N=8 males per genotype). Supplementary Fig. 7. General epigenetic signature of genomic region identified as sperm DMRs in Mtrr males. a, Using published ChIP-seq data sets in wildtype CD1 spermatozoa 37 , mean enrichment of selected histone modifications and DNA binding proteins in the genomic regions identified as differentially methylated in sperm of all Mtrr genotypes combined (N= 379 DMRs; red line) was determined and compared to the baseline genome (blue line). b-c, Using published ATAC-seq data sets in (b) wildtype CD1 spermatozoa 37 and (c) wildtype B6D2F1 mouse epiblast and extraembryonic ectoderm (ExE) at embryonic day 6.5 38 , mean enrichment of Tn5 transposase sensitive site (THSS) in the genomic regions identified as differentially methylated in sperm of all Mtrr genotypes combined (N= 379 DMRs; red line) compared to the baseline genome (blue line). DMRs in the region surrounding the Mtrr gene-trap insertion site were not included in either analysis. Dotted lines indicate the start and end of the DMR. Six kilobases (Kb) of DNA surrounding the DMR was also considered. Supplementary Fig. 8. Analysis of DNA methylation and gene expression at sperm DMRs in Mtrr gt/gt tissue. a-j, Schematic drawings of each DMR assessed indicate its relationship to the closest gene and are followed by graphs showing the average percentage of methylation at individual CpGs for the corresponding DMR as determined by bisulfite pyrosequencing. CpG methylation at several sperm DMRs from F0 Mtrr +/gt males was assessed in C57Bl/6J control (black lines) and Mtrr gt/gt (blue lines) embryos and placentas at E10.5. N=4 placentas and N=6-8 embryos assessed per genotypic group. Data is shown as mean ± sd for each CpG site. Two-way ANOVA, with Sidak's multiple comparisons test, performed on mean methylation per CpG site per genotype group. *p<0.05, **p<0.01, ***p<0.001. k-m, RT-qPCR analysis of mRNA expression of genes proximal to Supplementary Fig. 9. Epigenetic signature of the intragenic Cwc27 DMR in wildtype ESCs, TSCs and spermatozoa. Enrichment of DNA binding proteins (TET1, CTCF, PRM1, H3) and histone modifications (H3K27ac, H3K27me3, H3K4me1, H3K4me3, H3K9me3) in the Cwc27 locus on Chr13 (~37,000 kb downstream of Mtrr gene) using published ChIP-seq data sets in wildtype embryonic stem cells (ESCs), trophoblast stem cells (TSCs) 42 , and spermatozoa (Sp) 37 . Tn5 transposase sensitive sites (THSS) were also determined using published ATAC-seq data sets of normal B6D2F1 mouse epiblast and extraembryonic ectoderm (ExE) at embryonic day 6.5 38 or normal CD1 spermatozoa 37 . Grey box and shading indicate region of Cwc27 DMR identified in sperm of Mtrr +/gt males. Schematic of protein encoding Cwc27 partial transcript is shown at the bottom. Supplementary Fig. 10. Epigenetic signature of the intragenic Tshz3 DMR in wildtype ESCs, TSCs and spermatozoa. Enrichment of DNA binding proteins (TET1, CTCF, PRM1, H3) and histone modifications (H3K27ac, H3K27me3, H3K4me1, H3K4me3, H3K9me3) in the Tshz3 locus on Chr7 using published ChIP-seq data sets in wildtype embryonic stem cells (ESCs), trophoblast stem cells (TSCs) 42 , and CD1 spermatozoa (Sp) 37 . Tn5 transposase sensitive sites (THSS) were also determined using published ATAC-seq data sets of wildtype B6D2F1 mouse epiblast and extraembryonic ectoderm (ExE) at embryonic day 6.5 38 or CD1 spermatozoa 37 . Grey box and shading indicate region of Tshz3 DMR identified in sperm of Mtrr +/gt males. Schematic of protein encoding Tshz3 transcript is shown.