Iterative Oxidation by TET1 is Required for Reprogramming of Imprinting Control Regions and Patterning of Mouse Sperm Hypomethylated Regions

Summary

TET enzymes iteratively oxidize 5-methylcytosine to generate 5-hyroxymethylcytosine (5hmC), 5-formylcytosine, and 5-carboxycytosine to facilitate active genome demethylation. Whether these bases are required to promote replication-coupled dilution or activate base excision repair during mammalian germline reprogramming remains unresolved due to inability to decouple TET activities. Here, we generated two mouse lines expressing catalytically inactive TET1 (Tet1-HxD) and TET1 that stalls oxidation at 5hmC (Tet1-V). Tet1 knockout and catalytic mutant primordial germ cells fail to erase methylation at select imprinting control regions and promoters of meiosis associated genes, validating the requirement for iterative oxidation of 5-methylcytosine for complete germline reprogramming. TET1^V and TET1^HxD rescue most hypermethylation of Tet1^−/− sperm, suggesting the role of TET1 beyond its oxidative capability. We additionally identify a broader class of hypermethylated regions in Tet1 mutant mice sperm that depend on TET oxidation for reprogramming. Our study demonstrates the link between TET1-mediated germline reprogramming and sperm methylome patterning.

Graphical Abstract

Introduction

DNA methylation is a major conveyer of epigenetic information in the eukaryotic genome. 5-methylcytosine (5mC) bases are found primarily within CpG dinucleotides, with 70-80% of CpGs in the mammalian genome showing mitotically stable methylation¹. 5mC enrichment within enhancers and gene promoters is associated with transcriptional repression and shapes cell-specific gene expression profiles². DNA methylation is also essential for maintaining genomic stability by repressing repetitive elements³. In the germline, DNA methylation marks imprinted genes, where cis-regulatory elements known as imprinting control regions (ICRs) are methylated in a parent-of-origin specific manner to confer monoallelic expression of developmentally important genes⁴. ICR methylation is established during gametogenesis and uniquely escapes post-fertilization methylation erasure.

Mammalian development features two periods when DNA methylation is reprogrammed genome wide. The first occurs during post-fertilization embryonic development to achieve totipotency with subsequent establishment of tissue-specific methylation patterns². The second occurs in primordial germ cells (PGCs), germ cell precursors derived from cells of the most proximal epiblast^5–7. In both instances, DNA methylation erasure is achieved through a combination of two distinct mechanisms: 1) global replication-coupled passive dilution through the suppression of maintenance DNA methyltransferase (DNMT) activity, and 2) active demethylation facilitated by the family of ten-eleven translocation (TET) methylcytosine dioxygenases.

TET enzymes iteratively oxidize 5mC to 5-hydroxymethylcytosine (5hmC), 5-formylcytosine (5fC), and 5-carboxylcytosine (5caC)^8,9. All three oxidized 5mC bases (ox-mCs) are poorly recognized by the maintenance DNMT1 complex¹⁰, thus further promoting passive dilution. This active modification, passive dilution pathway is dominated by 5hmC which is the most prevalent ox-mC^11,12. Alternatively, 5fC and 5caC, but not 5hmC, can be excised by thymine DNA glycosylase (TDG), generating an abasic site further processed by base excision repair to recover unmodified cytosine in a pathway of active modification and active excision¹³. The three TET isoforms are major regulators of mammalian development. TET1 is expressed in PGCs, where it plays a role in the complete reprogramming of ICRs^14–16 and timely activation of meiosis-associated promoters^17,18. TET2 and TET3, by contrast, have tissue-specific roles in somatic development^19–23. Inactivation of all three Tet genes causes early embryonic lethality due to ectopic regulation of Lefty-Nodal signaling and failed gastrulation²⁴. Beyond physiological development, TET-mediated active demethylation is also important for successful somatic cell reprogramming to pluripotency^25–27.

The iterative modes of 5mC oxidation by TET enzymes prompt questions of whether specific ox-mCs have biological significance. Identifying functions for 5fC and 5caC has been challenging due to their low abundance and rapid removal by TDG⁹. Our recent work using a Tet2 mutant with reduced efficiency for oxidizing beyond 5hmC demonstrated that generation of higher order ox-mCs is required for a significant portion of DNA demethylation at enhancers during induced pluripotent stem cell formation²⁷. In PGCs, it has been shown that Tet1 deletion results in incomplete demethylation of ICRs and meiosis- and gametogenesis-associated gene promoters^15–18. It has been postulated that TET1 functions in germline reprogramming exclusively for the formation of 5hmC to promote replication-coupled passive dilution^12,28,29, a supposition largely based on the rapid rate of cellular division in PGCs and the scarcity of detectable 5fC/5caC^18,28–30. However, whether demethylation through 5fC/5caC contributes significantly to germline reprogramming has remained unclear. Moreover, a recent study challenged the role of TET1 in PGC methylation erasure altogether, instead proposing that TET1 protected regions from ectopic re-methylation following the completion of germline epigenetic reprogramming¹⁸. This model relied on a reduced representation bisulfite sequencing (RRBS) data set that suggested complete methylation erasure of well-established TET1-dependent loci in Tet1^−/− PGCs including ICRs and meiosis-associated gene promoters, contrary to previous studies^15–17. Instead, the authors reported a gain of methylation in Tet1^−/− PGCs during the period of mitotic or meiotic arrest in XY and XX PGCs, respectively¹⁸.

Functional studies of TET proteins have largely relied on conventional knockout mouse models, which, while highly informative, fail to disentangle the diverse catalytic activities of TET enzymes. To distinguish between TET1’s role in generating 5hmC to promote continued passive dilution or the requirement for 5fC/5caC generation during germline epigenetic reprogramming, we developed two mouse models with Tet1 catalytic mutations. The Tet1^T1642V (Tet1^v) mutant preserves 5hmC generation but has diminished oxidative activity to 5fC/5caC, and the Tet1^{H1654Y,D1656A} (TET1^HxD) mutant expresses catalytically-inactive TET1 (see Figure 1A)^27,31–33. By analyzing demethylation dynamics of previously described TET1-dependent ICRs and gene promoters in Tet1 mutant PGCs, we substantiate the necessity of TET1’s catalytic activity in germline epigenetic reprogramming. Our results show that TET1-mediated oxidation through 5fC and 5caC is required for complete ICR reprogramming in the germline. Additionally, methylation defects in Tet1 mutant sperm are more extensive than previously demonstrated, including rare but potentially important sperm hypomethylated regions that are not ICRs. This finding suggests that sperm-specific hypomethylated regions may originate at loci that require an active pathway for methylation erasure during germline reprogramming. Moreover, genome-wide methylation analysis of mutant sperm reveals that TET1^V and TET1^HxD can partially rescue Tet1^−/− sperm hypermethylation defects, supporting a role for TET1 beyond its oxidative capability. The use of Tet1 catalytic mutants to study germline epigenetic reprogramming reveals an added complexity to PGC genome regulation, in which the locus-specific modality of methylation erasure (active vs. replication-coupled) appears to contribute to germ cell methylome patterning.

Figure 1.

Generation and validation of 5hmC stalling Tet1-V and catalytically inactive Tet1-HxD mouse lines.

A) Schematic of WT and mutant TET1 proteins with the oxidative capabilities indicated⁵⁷. B) Sanger sequencing of Tet1^+/+, Tet1^V/V, and Tet1^HxD/HxD alleles. C) Western blot for full length TET1 protein in mutant testes with GAPDH as loading control. D) Expression of Tet isoforms (CPM – counts per million reads) in E14.5 XY PGCs as determined by RNAseq (n=3-4; FDR < 0.05; fold change > 1.5). E) Venn overlaps of DEGs in Tet1^−/−, Tet1^V/V, and Tet1^HxD/HxD PGCs compared to WT with enriched GO terms indicated for the commonly altered genes. F) Sparse BS-seq of Tet1^−/−, Tet1^V/V, and Tet1^HxD/HxD PGCs compared to WT at E12.5 and E14.5 (t-test vs WT, *p<0.05, ***p<0.0005). See also Figure S1, Figure S2 and Table S1.

Results

Characterization of 5hmC stalling Tet1^V and catalytically inactive Tet1^HxD mice

We previously showed that expression of 5hmC-dominant Tet1^V results in 5hmC generation without detectable 5caC in transfected HEK293T cells²⁷, while Tet1^HxD ablates all detectable catalytic activity^31,33. To test the catalytic requirements for TET1 during mammalian germline epigenetic reprogramming, we developed two mouse lines harboring these mutations using CRISPR/Cas9 genome editing (Figure 1A), with mutant alleles confirmed by Sanger sequencing and restriction fragment length polymorphism (RFLP) analysis (Figure 1B, Supplemental Figure 1A–B), and Southern analysis to ensure that the CRISPR/Cas9 mutagenesis did not cause chromosomal rearrangements in the Tet1 locus (Supplemental Figure 1C–E).

In adult mice, full-length TET1 protein can be detected in Tet1^V/V and Tet1^HxD/HxD testes at levels comparable to WT as assayed by Western blots (Figure 1C, Supplemental Figure 1F for full images), indicating that protein stability is unaffected in our Tet1 mutants. To measure expression, we designed an RNA pyrosequencing assay. Cerebral cortex samples from heterozygous mice, where Tet1 is robustly transcribed, showed that the mutant Tet1^V and Tet1^HxD alleles were expressed at equivalent levels to the WT Tet1 (Supplemental figure 1G–I).

Turning our focus to PGCs at embryonic day (E)14.5 in both sexes, we performed RNA sequencing (RNA-seq) to assess the transcriptome (Figure 1E, Supplemental Figure 1J). Tet1 was downregulated in Tet1^−/− PGCs, while Tet1 expression was unchanged in Tet1^V/V and Tet1^HxD/HxD PGCs (Figure 1D, XY PGCs). Tet2 and Tet3 are normally more lowly expressed at this developmental stage and their expression was unchanged in the Tet1 mutants (Figure 1D). Tet1^V/V and Tet1^HxD/HxD PGCs showed distinct transcriptomic changes compared to Tet1^−/− with a modest overlap of 10 downregulated and 31 upregulated genes among XY mutants and 24 commonly downregulated and 94 commonly upregulated genes among XX mutants (Figure 1E, Supplemental Figure 1J, Table S1). Commonly downregulated genes among Tet1^V/V, Tet1^HxD/HxD, and Tet1^−/− PGCs of both sexes include genes that are important for meiotic entry and germ cell identity (e.g. Mael, Dazl, Tex15, Rb1, and Dmc1) whose expression was previously reported to be regulated by DNA methylation erasure during germline epigenetic reprogramming^17,18. Consistently, gene ontology (GO) enrichment analysis identified meiotic nuclear division and meiotic cell cycle process as top altered categories amongst the commonly downregulated genes (Figure 1E, Supplemental Figure 1J). This result suggests that activation of promoters associated with meiotic entry (or germline reprogramming responsive (GRR) genes as defined by Hill et al.¹⁸) requires both catalytic activity of TET1 and the capacity for 5fC/5caC generation in PGCs (Supplemental Figure 2A–B, Table S1).

To evaluate global cytosine modification levels in reprogramming PGCs, we employed sparse-coverage whole genome bisulfite sequencing (sparse BS-seq) or whole genome bisulfite assisted APOBEC sequencing (sparse bACE-seq) to approximate total 5mC and 5hmC levels or 5hmC levels, respectively^27,34,35. This methodology for BS-seq was previously shown to accurately estimate global levels of genome modification (5mC + 5hmC) through sampling of at least 20,000 cytosines in the CpG context using next-generation sequencing and is amenable for low-input samples such as PGCs³⁴. We further benchmarked sparse bACE-seq for resolving 5hmC from 5mC by comparison to liquid-chromatography-tandem mass spectrometry (LC-MS/MS) using the adult mouse cortex, where 5hmC is relatively abundant (Supplemental Figure 2C)^36,37. Using this approach, in adult testes where post-meiotic germ cells predominate, global modified cytosine levels were unchanged across all mutant genotypes (sparse BS-seq, Supplemental Figure 2D), and all samples showed scant levels of 5hmC (sparse bACE-seq, Supplemental Figure 2E). In adult mouse cortex, where 5hmC is more prevalent, global levels of total modified cytosine and 5hmC were unchanged in mutants compared to WT (Supplemental Figure 2F, G). The unaltered levels of 5hmC in the catalytically dead Tet1^HxD/HxD or Tet1^−/− samples were not unexpected as TET2 actively maintains gene body 5hmC accumulation in adult mouse cortex³⁷.

In mice, germline epigenetic reprogramming occurs immediately following PGC specification (~E7.25) and continues as PGCs migrate from the proximal epiblast, to the base of the allantois, to the genital ridges³⁸. At E12.5 demethylation of the PGC genome is nearing completion and by E14.5, male PGCs mitotically arrest¹⁸. Sparse-BS-seq of E12.5 PGCs revealed global hypermodification of Tet1^V/V and Tet1^HxD/HxD PGCs compared to WT, but not in Tet1^−/− PGCs (Figure 1F)^15,17,18. Interestingly, by E14.5, hypermodification of cytosine resolved (Figure 1F) and global levels of modified cytosine in testes were similar across genotypes (Supplemental Figure 2D). This result is consistent with previous reports that compromised TET1 activity does not cause global hypermethylation of the PGC genome at the completion of reprogramming^15,17,18. The sparse BS-seq result suggested a subtle difference in the dynamics of global methylation erasure in Tet1 catalytic mutant PGCs compared to Tet1^−/− PGCs, perhaps due to modest mouse background strain differences. We assessed the expression of maintenance and de novo DNMTs in the RNA-seq data to ask whether the slight delay in methylation reprogramming in E12.5 Tet1^V/V and Tet1^HxD/HxD PGCs can be explained by their differential expression. Dnmt1, Dnmt3a, and Dnmt3b are unchanged in Tet1^V/V, Tet1^HxD/HxD, and Tet1^−/−, while DNMT1 binding partner Uhrf1 is downregulated in Tet1^HxD/HxD and Tet1^−/− PGCs (Supplemental Figure 2H). Overall, we thus generated two viable mouse Tet1 catalytic mutants, which express full length TET1 proteins, with potentially distinct phenotypes from the previously established Tet1^−/− mutants^15,39. We therefore next employed these mutants to test the requirement for TET1 catalytic function and 5fC/5caC generation with respect to germline reprogramming.

Tet1 catalytic activity is required for ICR and meiotic promoter reprograming

ICRs are the most well-characterized loci to require TET1 during germline epigenetic reprogramming^15,16. Germ cells of male and female Tet1^−/− mice exhibit ICR hypermethylation, despite unchanged genome-wide methylation^15,16,18. However, while TET1 and 5hmC are detected in migrating PGCs, it is unknown whether 5hmC generation is sufficient to promote replication-coupled passive dilution or whether TET1-dependent ICR reprogramming requires the generation of 5fC/5caC¹². As stated above, a recent report challenged the role of TET1 in methylation erasure, suggesting that Tet1^−/− PGCs undergo normal reprogramming but instead ectopically gain methylation subsequently¹⁸. We conducted locus specific analyses using pyrosequencing to elucidate the kinetics of ICR and meiosis-associated promoter methylation erasure in WT and Tet1 mutant PGCs. We first assessed methylation of representative ICRs at E12.5, a time at which Hill et al. suggested that PGC reprogramming would be complete (i.e. hypermethylation of ICRs was not observed in Tet1^−/− PGCs)¹⁸. The ICRs analyzed here are the most widely studied among the approximately dozen mouse ICRs. At E12.5, KvDMR and Peg3 showed hypermethylation in all mutant XY PGCs compared to WT, with IG-DMR and Peg1 ICRs exhibiting hypermethylation in at least one catalytic mutant (Figure 2A). At E14.5 when XY and XX PGCs are in mitotic and meiotic arrest, respectively, four out of the seven ICRS we analyzed (Rasgrf, KvDMR, Peg1, Peg3) showed hypermethylation in Tet1^V/V, Tet1^HxD/HxD, and Tet1^−/− XY PGCs compared to WT (Supplemental Figure 3A)²⁸. Hypermethylation of representative ICRs at E12.5 and E14.5 was also observed in Tet1^V/V, Tet1^HxD/HxD, and Tet1^−/− XX PGCs compared to WT (Supplemental Figure 3B,C). Notably, at E14.5 Tet1 mutant XX PGCs showed hypermethylation at H19, Rasgrf1, KvDMR, and Peg1, while hypermethylation at IG-DMR was exclusive to the Tet1 catalytic mutant PGCs. Peg3 hypermethylation in E14.5 XX PGCs was observed in Tet1^HxD/HxD and Tet1^−/− mutants. This result demonstrated that PGCs with altered TET1 functions do not complete reprogramming by E12.5 and remain hypermethylated at the time of cell-cycle arrest at E14.5 for both sexes.”

Figure 2.

Loss of TET1 or TET1 catalytic activity leads to incomplete reprogramming of ICRs and meiosis-associated promoters.

A) Methylation levels of representative ICRs as measured by pyrosequencing. Each data point is E12.5 XY PGC sample collected from one embryo. Time course analysis of methylation levels of KvDMR (B), Peg1 (C), and Peg3 (D) from E11.5 to E14.5 in XY PGCs. Theoretical passive dilution curve (maroon-dashed line) is a fitted one-phase logarithmic decay curve with PGC doubling time of 12.6 hours. Methylation levels of meiosis-associated promoters, Mael (E) and Sypc1 (F) at E13.5 and E14.5 in XY PGCs (mean methylation ± SEM; n=4-9, one-way ANOVA with Dunnett’s multiple comparisons test, distinct letters indicate statistical significance) . G) Expression of Mael and Sycp1 in E14.5 XY PGC RNAseq (n=3-4; FDR < 0.05; fold change > 1.5). See also Figure S3.

To examine the possibility of ectopic gain of methylation during germline reprogramming period, we assessed demethylation dynamics of ICRs from E11.5 to E14.5 in WT and Tet1 mutant PGCs. Three representative ICRs (KvDMR, Peg1, and Peg3) with distinct demethylation kinetics are shown in Figure 2B–D. First, we determined one-phase decay approximation of purely passive, replication coupled methylation erasure dilution using the previously reported PGC doubling time of 12.6 hour²⁸. Consistent with previous reports, methylation erasure of ICRs in WT PGCs (blue curve, Figure 2B–D) did not proceed faster than the predicted rate of passive dilution in the absence of maintenance DNA methyltransferase activity (DNMT1/UHRF1). With Tet1 deletion, ICR methylation erasure proceeded slower than in WT PGCs (Figure 2B–D), particularly between E12.5 and E14.5, where WT PGCs demonstrated further methylation loss to reach the E14.5 hypomethylated state. Like Tet1^−/− PGCs, Tet1 catalytic mutant PGCs resisted methylation erasure between E12.5 and E14.5 (Figure 2B–D), resulting in hypermethylation of KvDMR, Peg1, and Peg3 by E14.5. In this empirical determination of ICR reprogramming kinetics in Tet1 mutant PGCs, we did not observe ectopic gain of methylation in PGCs. The similar rates of ICR reprogramming for Tet1^V/V, Tet1^HxD/HxD, and Tet1^−/− PGCs supported the requirement for efficient oxidation to 5hmC and beyond to 5fC/5caC for the complete methylation erasure at ICRs.

Next, we examined methylation of two representative meiosis-associated promoters whose activation during germline reprogramming period is TET1-dependent^15,17,18. At E14.5, Mael was significantly hypermethylated in Tet1^V/V and Sycp1 was significantly hypermethylated in all Tet1 mutant PGCs compared to WT in XY (Figure 2E, F) and XX (Supplemental Figure 3D) PGCs. Comparing E13.5 and E14.5 WT and Tet1 mutant XY PGCs (Figure 2E, F), we did not observe ectopic gain of methylation at Mael and Sycp1 promoters. Hypermethylation at Mael and Sypc1 promoters was associated with suppressed gene expression in XY and XX Tet1 mutant PGCs at E14.5 as measured by RNA-seq (Figure 2G, Supplemental Figure 3E). The incomplete methylation erasure of these promoters in mutant PGCs suggested that reprogramming of meiosis-associated promoters, and possibly GRR gene promoters (Supplemental Figure 2A, B), require TET1 complete catalytic function during germline reprogramming in PGCs of both sexes.

Catalytic Tet1 mutations lead to incomplete ICR reprogramming in sperm and imprinting defects in F1 offspring

To assess the consequence of incomplete reprogramming for mature germ cells, we measured the methylation of representative ICRs in sperm (Figure 3A). Peg1, Peg3, KvDMR, and Snrpn are maternally methylated ICRs that are normally hypomethylated in sperm. Tet1^V/V, Tet1^HxD/HxD, and Tet1^−/− sperm showed Peg1, Peg3, and KvDMR hypermethylation compared to WT, confirming persistence of incomplete ICR reprogramming in mutant germ cells (Figure 3A). Consistent with our previous report, Snrpn demethylation was TET1-independent, with unaffected methylation observed in mutant sperm¹⁶. The paternally methylated H19/Igf2 ICR showed the expected hypermethylation pattern in sperm for all genotypes.

Figure 3.

Tet1^V/V and Tet1^HxD/HxD males exhibit methylation defects at ICRs that are inherited by offspring.

A) Methylation levels at representative maternally methylated ICRs measured by pyrosequencing. Each data point is a sperm sample from one adult mouse. H19/Igf2 ICR is a paternally methylated ICR that exhibits full methylation in sperm (mean methylation ± SEM; n=3-5). The number of live embryos (B) and resorbed embryos (C) per litter at E10.5 (mean number of pups/litter ± SEM, n=3-5 litters). The number of live pups (D) and dead pups (E) per litter at PND0 (mean number of pups/litter ± SEM, n=5-6 litters). For panel A-E: one-way ANOVA with Dunnett’s multiple comparisons test, distinct letters indicate statistical significance. F) Heatmap representation of DNA methylation levels measured by pyrosequencing at ICRs of all E10.5 embryos from Tet1^+/+, Tet1^V/V and Tet1^HxD/HxD males. Each row is an individual embryo of the indicated paternal genotype. The same offspring are depicted by locus for Peg1 (G) and Peg3 (H) ICRs. pWT n=22 embryos (3 litters), pVV n=31 embryos (4 litters), pHxD n=37 embryos (4 litters). Fisher’s exact test for frequency of hypermethylated embryo; *p<0.05, **p<0.01; shaded bars indicate average methylation of pWT embryos ± 1 STDEV. See also Figure S4.

To test whether hypermethylated sperm of Tet1 catalytic mutant males contributed to fertility, we mated Tet1^V/V, Tet1^HxD/HxD, or WT males to C57BL/6J females (see Supplemental Figure 4A for breeding strategy). At midgestation (E10.5), pregnancies sired by Tet1^V/V (pVV), Tet1^HxD/HxD (pHxD) or WT (pWT) males showed equivalent numbers of developing (Figure 3B) and resorbed (Figure 3C) embryos. At birth (PND0), pVV showed significantly decreased litter size compared to pWT, while the decrease in pHxD litter size was not statistically significant (Figure 3D). Because there was not a statistically increased number of dead pups in pVV or pHxD litters at birth, litter attrition likely occurred between E10.5 and birth (Figure 3E).

ICRs are protected from the post-fertilization global DNA demethylation^40,41, and thus incomplete erasure of ICRs during germline development is expected to be stably inherited by TET1 mutant offspring. Figure 3F depicts Peg1, Peg3, and KvDMR methylation levels in E10.5 embryos as a heatmap. Snrpn is included as a maternally methylated ICR control that is unaffected by TET1 mutations, and H19/Igf2 is included as paternally methylated ICR control (Supplemental Figure 4C, D). The number of hypermethylated offspring was significantly increased at Peg1 and Peg3 for pVV (19.35% for both) and pHxD (21.62% and 32.43%, respectively) compared to pWT (Figure 3G–H). Incidence of hypermethylation at KvDMR in pVV and pHxD litters was not statistically significant compared to pWT (Supplemental Figure 4B). Notably, the proportion of pVV and pHxD embryos exhibiting hypermethylation for a given ICR mirrored the degree of hypermethylation observed in Tet1 mutant sperm. While most affected pVV or pHxD embryos only showed hypermethylation at one ICR, a few exceptions demonstrated hypermethylation at multiple ICRs (Figure 3F). However, no correlation was observed between DNA methylation levels at Peg1, Peg3, and KvDMR in individual embryos, suggesting independent segregation of alleles with affected loci during meiosis. We similarly measured hypermethylation incidence in pVV or pHxD PND0 brain and observed lower frequencies of affected pups compared to E10.5 embryos (Supplemental Figure 4F–J), consistent with ICR hypermethylation as a driver for increased embryonic lethality in Tet1-mutant offspring.

In summary, the hypermethylation of representative ICRs in Tet1 catalytic mutant sperm and increased incidence of hypermethylated offspring of catalytic Tet1 mutant males phenocopies that of Tet1^−/− males that we previously reported¹⁶. Integrating across the variants and considering PGC and sperm methylation data, our findings demonstrate that, in contrast to the previously assumed role for 5hmC as a driver for replication-coupled passive dilution of 5mC, ICRs require efficient TET1-mediated iterative oxidation capacity to complete PGC reprogramming.

Global methylation analysis revealed partial rescue of Tet1^−/− sperm methylome by full length catalytic mutant proteins

During germline epigenetic reprogramming, DNA methylation is erased from epiblast cells to establish the germline fate. Previously, Tet1 deletion was reported to affect a relatively small number of late demethylating loci, including ICRs and gametogenesis-related genes^15,17,18. The methylome of Tet1^−/− sperm DNA, however, had only been analyzed using RRBS, which limits interrogation to regions enriched with CCGG motifs¹⁵. To globally determine the differential requirements of 5hmC versus 5fC/5caC generation during germline epigenetic reprogramming, we employed Illumina’s Mouse Infinium Methylation BeadChip to assess the methylome of Tet1^V/V, Tet1^HxD/HxD, Tet1^−/−, and WT sperm DNA. The BeadChip interrogates >285,000 CpGs representative of the mouse genome with manually curated coverage of gene promoters, enhancers, repetitive elements, and known CpG islands, including regions relevant for imprinting biology⁴². Methylation arrays have been shown to perform with two- to three-fold less technical variability than whole genome bisulfite sequencing (WGBS) conducted at 30x sequencing depth⁴³. Additionally, in contrast to WGBS approaches which limit the number of samples analyzed, array-based approaches allowed for coverage of a higher number of independent replicates (n=8-10) for each genotype, which is particularly valuable given the stochastic nature of methylation changes. With the high depth coverage at each CpG and the inclusion of a large number of biological replicates, the BeadChip array granted us unprecedented statistical power to identify differentially methylated regions (DMRs) with high confidence⁴⁴. A systematic comparison of TruSeq targeted bisulfite sequencing and EPIC methylation arrays reported that arrays outperform NGS methods for DMR identification⁴⁴. We identified 1411 DMRs (each DMR corresponds to a single probe on the array) in Tet1^V/V, 2488 DMRs in Tet1^HxD/HxD, and 6005 DMRs in Tet1^−/− sperm compared to WT (Figure 4A, Table S2). While most DMRs in Tet1^V/V and Tet1^−/− sperm were hypermethylated (Tet1^V/V: 1359 hypermethylated, 52 hypomethylated; Tet1^−/−: 5631 hypermethylated, 374 hypomethylated), consistent with the role of TET1 in active demethylation, a substantial number of DMRs in Tet1^HxD/HxD sperm were unexpectedly hypomethylated (1006 hypermethylated, 1482 hypomethylated). This result raises the possibility of a distinct function for the catalytically inactive TET1^HxD in DNA methylation, which may be unrelated to reprogramming (Supplemental Figure 5A, right). Global methylation levels in Tet1 mutant sperm are unchanged compared to WT (Supplemental Figure 5B).

Figure 4.

Global methylation analysis using Mouse Infinium Methylation BeadChip shows distinct methylome defects in catalytic mutant (Tet1^V/V, Tet1^HxD/HxD) sperm compared to Tet1^−/− sperm.

A) Flow chart showing differential methylation analysis of each Tet1 mutant sperm sample compared to Tet1^+/+ (WT), n=8-10. A DMR is a probe with FDR < 0.05 with minimum change in average methylation of greater than 10%. B) Venn overlap of significantly hypermethylated (top) and hypomethylated (bottom) DMRs in Tet1 mutant sperm vs. WT. Volcano plots comparing the methylation status of Tet1^−/− sperm DMRs to Tet1^V/V sperm (C) or Tet1^−/− sperm DMRs to Tet1^HxD/HxD sperm (D). KYCG analysis to identify enrichment for probe design groups of rescued (E) and not-rescued (F) DMRs. See also Figure S5, Table S2, and Table S3.

We next determined the overlap of hypermethylated and hypomethylated DMRs between the catalytic mutants (Tet1^V/V and Tet1^HxD/HxD) and Tet1^−/− sperm (Figure 4B). 556 DMRs were commonly hypermethylated in all three mutants. These loci were classified as requiring TET1 catalytic activity to generate 5fC/5caC for complete reprogramming in the male germline (Figure 4B, Table S3). Following NCBI reference sequence (RefSeq) annotation, we identified many imprinting-associated regions within these 556 DMRs (Table S2), consistent with our previous conclusion that ICRs require the full competency of TET1 catalytic activity for reprogramming (Figure 2, 3). While the majority of hypermethylated DMRs in Tet1^V/V and Tet1^HxD/HxD overlapped Tet1^−/− hypermethylated DMRs, Tet1^HxD/HxD showed unique hypomethylated DMRs in sperm (Figure 4B). Partially supervised hierarchical clustering of DMR methylation levels demonstrated the distinct severity of Tet1^−/− hypermethylation defects and the similarity of Tet1^V/V and Tet1^HxD/HxD hypermethylation signatures (Supplemental Figure 5A, left).

Significantly greater numbers of DMRs were identified in Tet1^−/− sperm compared to either catalytic mutant (6005 Tet1^−/− vs 1411 Tet1^V/V DMRs; 6005 Tet1^−/− vs 2488 Tet1^HxD/HxD DMRs), suggesting that TET1^V or TET1^HxD proteins can partially rescue the DNA methylation defects in the KO sperm. To clarify the degree of rescue that our catalytic mutants provided, we assessed methylation levels and statistical significance of the 6005 DMRs of Tet1^−/− sperm in Tet1^V/V or Tet1^HxD/HxD samples. Tet1^−/− DMRs were considered rescued (grey dots, Figure 4C, D) by full length TET1^V or TET^HxD if those DMRs no longer reached the threshold for statistical significance (FDR < 0.05), while partially rescued DMRs (blue dots, Figure 4C, D) were still significantly altered in Tet1^V/V or Tet1^HxD/HxD samples but differential methylation between catalytic mutant samples and WT was no longer greater than 10%. Approximately 75% of KO DMRs were fully rescued by the expression of TET1^V (4369/6005, Figure 4C) or TET1^HxD (4679/6005, Figure 4D). The similar degrees of complete rescue that were demonstrated by the 5hmC dominant TET1^V and the catalytically inactive TET1^HxD suggest that most methylome defects observed in Tet1^−/− sperm were due to the loss of TET1 non-catalytic function, as full length TET1 with diminished or ablated 5mC oxidation activity is sufficient to achieve the WT methylation state at these loci. Alternatively, non-enzymatic domains of TET1 protein may be sufficient to prevent aberrant de novo methylation at these loci during spermatogenesis, outside of the PGC reprogramming period. 3844 KO DMRs were rescued in both Tet1^V/V and Tet1^HxD/HxD sperm, further emphasizing the importance of TET1, distinct from its oxidative capacity, for normo-methylation at these regions. Enrichment analysis using knowYourCG (KYCG) tool of the SeSAMe pipeline identified most rescued DMRs as CpGs with probes designed to interrogate enhancers (Figure 4E, Supplemental Figure 5B–C, Table S3)^42,45.

By contrast, only a modest number of KO DMRs were partially rescued in Tet1^V/V and Tet1^HxD/HxD samples (blue dots in Figure 4C, D, 350 in Tet1^V/V and 249 in Tet1^HxD/HxD), perhaps representing a subset of loci where reprogramming is less efficient in the presence of TET1 catalytic mutants. Finally, Tet1^−/− DMRs that remained significantly altered in Tet1^V/V (“no rescue”: 1286/6005) and Tet1^HxD/HxD (“no rescue”: 1073/6005) sperm likely represent loci that require 5mC oxidation, but 5hmC generation alone is insufficient to achieve normal level of methylation. Non-rescued DMRs were enriched for probes designed to interrogate ICRs, monoallelically methylated regions in the genome, and regions that are known to be unmethylated in the sperm (Figure 4F, Supplemental Figure 5D–E). Taken together, the data indicate that the 5hmC-dominant TET1^V and catalytically inactive TET1^HxD rescue a significant proportion of the methylation defects observed in Tet1^−/− sperm, thus supporting a more expansive role for TET1’s non-enzymatic activity in patterning of the male germline methylome.

Tet1 DMRs are located in regions that are excluded from de novo methylation in the hypermethylated sperm genome

Upon the completion of epigenetic reprogramming in PGCs, the male germline undergoes de novo methylation late in gestation, which yields a highly methylated sperm genome compared to that of somatic cells or the oocyte genome^46,47. In mouse, ~90% of the sperm genome is methylated⁴⁷. While de novo methylation occurs indiscriminately in prospermatogonia, there are regions that are excluded from de novo DNMTs DNMT3A/3L and thereby protected from gaining methylation in the sperm genome^48,49. Among these sperm-specific hypomethylated regions (HMRs) are maternally methylated ICRs (Figure 5B)⁵⁰. Using Bernoulli’s distribution testing, we determined that imprinting probes, which only make up 0.2% of the array, were overrepresented in the Tet1 mutant DMRs, particularly in Tet1^V/V sperm (Figure 5A).

Figure 5.

Identification of TET1-dependent sperm-specific hypomethylated regions.

A) Distribution of DMRs classified as related to imprinting biology in the array annotation (*p-value < 0.05; two-sided Bernoulli distribution test vs. all array probes). B) Schematic of a sperm hypomethylated region (sperm HMR) that is excluded from de novo methylation through enrichment of H3K4me3. C) Distribution of DMRs that overlap sperm HMRs (DNMTools function “hmr” of sperm WGBS GEO: GSE56697⁵²)⁵³. D, F) Representative regions that are commonly hypermethylated in Tet1^−/−, Tet1^V/V, Tet1^HxD/HxD sperm, overlapping sperm HMRs (grey bars) and H3K4me3 enrichment during de novo methylation in E17.5 prospermatogonia. E) Heatmaps and metaplots of E17.5 prospermatogonia H3K4me3 enrichment centered on DMRs for each genotype or shared among three mutants, measured in counts per million (CPM). G) Genomic distribution of DMRs that overlap with H3K4me3 enrichment in E17.5 prospermatogonia. H) Methylation analysis of oxC base-dependent sperm HMRs in demethylating WT and Tet1 mutant E14.5 PGCs using targeted bisulfite sequencing (mean methylation ± SEM; n=3-4, one-way ANOVA with Dunnett’s multiple comparisons test, distinct letters indicate statistical significance). See also Figure S6 and Table S4.

We then asked whether DMRs of Tet1 mutant sperm are located in sperm-specific HMRs unrelated to imprinting. We identified sperm-specific HMRs using DNMTools “hmr” algorithm, which searches for methylation canyons in WGBS datasets⁵¹. Using previously published sperm WGBS (GEO: GSE56697⁵²), DNMTools discovered 76,227 distinct sperm HMRs, which we overlapped with DMRs for each Tet1 mutant genotype (Figure 5C, Table S4). A substantial portion of DMRs in Tet1^V/V (87%, 1235/1411), Tet1^HxD/HxD (58%; 1444/2488), and Tet1^−/− sperm (85%; 5113/6005) were indeed located within sperm HMRs. 521 probes overlapping sperm-specific HMRs were commonly hypermethylated in Tet1^V/V, Tet1^HxD/HxD, and Tet1^−/− sperm, likely representing sperm HMRs that require full catalytic TET1 capacity for reprogramming. We also assessed whether DMRs fell within sperm HMRs that are typically methylated in the oocyte (Figure 5B, such as ICRs), and confirmed that, in addition to ICRs, 25-40% of DMRs, depending on the genotype, were in regions that are differentially methylated between sperm and oocyte genomes (Supplemental Figure 6A).

DNMT3A is inhibited by H3K4me3-enriched regions during de novo methylation in prospermatogonia⁴⁶. We performed CUT&RUN for H3K4me3 on WT prospermatogonia to identify regions that are enriched for this chromatin mark during the de novo methylation period. Genome-wide, H3K4me3 signals of E17.5 prospermatogonia showed strong overlap with sperm HMRs, confirming that methylation was indeed excluded from regions that are enriched for H3K4me3 (Supplemental Figure 6C). We next overlapped prospermatogonia H3K4me3 signals with Tet1 mutant DMRs to determine: 1) if TET1-dependent DMRs were enriched in regions typically excluded from de novo methylation (as the sperm genome is largely hypermethylated elsewhere), and 2) whether dysregulated DNA methylation occurred in regulatory regions. We observed a significant presence of H3K4me3 signals corresponding to Tet1^−/−, Tet1^V/V, and Tet1^HxD/HxD DMRs (Figure 5E, Table S4). Overall, 3841/6005 Tet1^−/−, 942/1411 Tet1^V/V, and 939/2488 Tet1^HxD/HxD DMRs overlapped with H3K4me3 peaks in prospermatogonia. Because H3K4me3 is more commonly used as a marker of active or bivalent promoters in most cell types (in addition to its role in excluding DNMT3A/L), we assessed the genomic locations of H3K4me3-overlapping TET1-DMRs. Counter to our expectations, most H3K4me3-overlapping DMRs were located within gene bodies (exons and introns) or intergenic regions instead of the expected annotated promoters or TSSs (Figure 5G). Indeed, only ~20% of DMRs that overlapped with H3K4me3 peaks were mapped to annotated promoters, regardless of the Tet1 genotype (Figure 5G). This was also true at later stages of spermatogenesis (spermatogonia⁵³, pachytene spermatocyte⁵⁴, round spermatid⁵⁴, and sperm⁵⁵; Supplemental Figure 6E–G), where majority of TET1-DMRs that overlap with H4K3me3 peaks did not map to annotated promoters. Representative examples of DMRs in non-promoter H3K4me3-marked regions include exon 3 of Dyrk2 and exon5-6 of Fat1 (Figure 5D, F). Hypermethylation of Fat1 and Dyrk2 in Tet1^V/V, Tet1^HxD/HxD, and Tet1^−/− sperm supports the requirement of TET1 catalytic activity to curtail ectopic 5mC deposition during germline reprogramming. Although these H3K4me3 peaks did not fall within gene promoters, we noted that similar to ICRs (see Peg1 as an example, Supplemental Figure 6D), H3K4me3 enrichment remained throughout spermatogenesis at these TET-DMRs (e.g. Fat1, Supplemental Figure 6B). Finally, we assessed methylation levels of select non-ICR, TET1-dependent sperm HMRs in E14.5 WT and Tet1 mutant PGCs. While Fat1 DMRs indeed failed to reprogram in mutant PGCs, Dyrk2 and Shroom3 DMRs reached hypomethylated states to varying degrees, suggesting that Dyrk2 and Shroom3 require ox-mC generation by TET1 to maintain hypomethylated states after the PGC reprogramming period (Figure 5H). Taken together, our data demonstrated a dependency for TET1’s oxidative activity in many regions that will eventually be excluded from methylation in the male germline.

Genes associated with Tet1 DMRs are expressed throughout spermatogenesis

We referred to publicly available single-cell RNA sequencing (scRNAseq) data of adult mouse testes to determine the spermatogenesis stages in which genes associated with TET1-dependent DMRs are expresed⁵⁶. Previous analysis of scRNAseq data identified 12 germ cell (GC) clusters that correspond to spermatogonial stem cells (SSC, Figure 6A, GC1), two transitional preleptotene stages (GC2-3), 5 stages of spermatocytes undergoing meiosis (GC4-8), 3 stages of post meiotic spermatids (GC9-11), and elongating spermatids (GC12)⁵⁶. We mapped all DMRs to the nearest annotated genes, which resulted in 4358 DMR-associated genes. 3207 of these genes have detectable expression in at least one GC stage (Figure 6A). DMR-associated genes are expressed in all germ cell clusters, with median expression level comparable to that of all expressed genes at each stage of spermatogenesis (Figure 6A).

Figure 6.

DMR-associated genes are expressed throughout spermatogenesis.

A) Comparison of DMR-associated gene expression and all genes expressed throughout spermatogenesis based on normalized gene expression from publicly available scRNAseq (GEO: GSE112393⁵⁶). Fat1 (B,C) and Dyrk2 (D,E) 5’ RACE analyses of Tet1^+/+ and Tet1^−/− testes cDNA and the corresponding BLAT mapping of alternative (alt) transcripts to TET1 DMRs. “Main” indicates predicted main product and “alt” indicates identified alternative product.

A subset of DMRs overlapped with H3K4me3 peaks outside of known promoters throughout all analyzed germ cell stages (Supplemental Figure 6B, E–G). We next investigated whether these regions serve as alternative TSSs. We conducted rapid amplification of cDNA ends (RACE) to identify alternative 5’ ends on the Fat1 and Dyrk2 loci (Figure 6B–E) using adult testes cDNA to obtain pools from all spermatogenic stages. From the scRNAseq data, Fat1 is expressed in GC1 cluster (SSC), while Dyrk2 is expressed in GC8-GC11 clusters (late spermatocyte, spermatids). Using 3’ gene specific primers downstream of the DMRs, we detected alternative amplicons (Figure 6B, D) that are shorter than the expected transcripts from the RefSeq annotated promoters at each locus (Fat1 main: ~5.5 kb, Fat1 alt: 1586 kb; Dyrk2 main: ~1.5 kb, Dyrk2 alt: 745 bp). We subcloned and sequenced the alternative amplicons, which mapped their 5’ end to the DMRs at exon 5-6 of Fat1 and exon 3 of Dyrk2 (Figure 6C, E). These results demonstrated that at select loci, TET1-dependent sperm hypomethylated regions can be used as alternative transcriptional start sites during spermatogenesis.

Discussion

The discovery of TET proteins as DNA dioxygenases led to a paradigm shift epigenetics, where it was previously thought that DNA methylation was largely erased only through the lack of maintenance during cell division. There is now ample evidence supporting the importance of active DNA demethylation pathways during reprogramming to pluripotency and during PGC epigenetic reprogramming^{16–18,25,27,29,39,57,58}. What remains to be answered are questions of the regulatory potential of higher order oxidized cytosine bases (i.e. 5fC and 5caC) and whether TET1 acts to promote demethylation or as a safe-guard against re-methylation during the germline reprogramming period. Expanding upon the biochemical discovery of functional mutations within the catalytic domain of TET1 and TET2 that alter catalytic activities, we generated 5hmC dominant Tet1^V and catalytically inactive Tet1^HxD mouse lines to 1) determine the requirement of TET1 oxidative activity, and 2) study the importance of 5fC/5caC generation in the context of PGC epigenetic reprogramming^27,32.

The mutations evaluated here are confined to the catalytic domain and the TET1 active site, which is a buried surface that captures 5mC everted from the DNA duplex. Extensive prior biochemical analysis has shown that mutations validated with the catalytic domain recapitulate the behavior of the full-length TET enzymes³², with molecular dynamics modeling also indicating that the mutants are entirely non-disruptive to structure and dynamics of TET protein as a whole⁵⁹. Therefore, established biochemical and computational biophysics approaches with these mutants have established their value in studying our biological questions.

Our results show that select ICRs, the most well characterized TET1-dependent loci during germline reprogramming, require iterative oxidation of 5mC by TET1, including the capacity to generate 5fC/5caC to achieve methylation erasure in the mouse germline, as Tet1^V/V-PGCs failed to complete reprogramming similar to Tet1^HxD/HxD and Tet1^−/− PGCs¹¹. This finding contradicts previous views that the role of TET in germline reprogramming is restricted to the generation of 5hmC to repel DNMT1 binding and promote passive dilution^12,18,28,30. Our time course demethylation analysis of representative ICRs strongly suggests that PGCs lacking either TET1 or TET1 catalytic activity failed to complete reprogramming by E12.5. Our data suggests that each ICR may have subtle differences in its requirement for TET1’s iterative catalytic activity to achieve demethylation. Most importantly, there was no evidence of ectopic gain of methylation in Tet1 mutant PGCs during mitotic or meiotic arrest periods, at least among representative ICRs and meiotic promoters that were analyzed in our developmental time course. We acknowledge that Hill and colleagues proposed a non-traditional role of TET1 based upon genome-wide analysis of methylation in PGCs, and thus we cannot exclude the possibility that at these developmental time points other ICRs may behave distinctly from the loci we analyzed¹⁸. Similarly, we do not discount the possibility that protection against ectopic remethylation by TET1 occurs in time points where Dnmt3a and Dnmt3l are expressed at high levels during the de novo methylation period (in prospermatogonia or secondary growing oocytes). Similar to ICRs, meiotic gene promoters showed dependency on TET1 proficient in oxidizing beyond 5hmC.

We also conducted genome-wide methylation analysis using the Infinium Methylation BeadChip to assess methylation patterns in WT and mutant sperm, which reveals more extensive methylome defects than previously reported in Tet1^−/− sperm^15,18. We elected to conduct genome-wide methylation analysis using the array to achieve the high statistical power required to distinguish the differing requirement of 5hmC vs 5fC/5caC generation during epigenetic reprogramming and elucidate the resulting consequence on the germline genome^43,44. Comparison to the Tet1^−/− sperm methylome is essential in our studies because it reveals the distinct methylation defects that result from the presence of 5hmC-dominant TET1^V or the catalytically inactive TET1^HxD proteins (Figure 4). Of note, we showed that TET1^V and TET1^HxD can partially rescue the hypermethylation phenotype of Tet1^−/− sperm. Interestingly, TET1^V and TET1^HxD rescue Tet1^−/− hypermethylated loci to a similar degree (Figure 4E, F), suggesting that a large proportion of TET1-dependent loci in the germline do not actually require its catalytic activity or processivity to achieve or maintain normal methylation. TET1 contains a CXXC zinc finger domain within its N-terminus that specifically recognizes unmethylated CpGs in clusters ^60,61. It is plausible that DMRs that are rescued by the presence of full-length, catalytically mutant forms of TET1 are those that do not require TET1’s oxidative capability, but instead require TET1 binding and protection from the DNMT3 complex during de novo methylation in prospermatogonia. This may proceed through TET1 physical localization or TET1-mediated recruitment of histone modifiers to generate a DNMT3-repelling chromatin environment. In contrast, DMRs not rescued by TET1^V are those that require 5fC/5caC generation for normo-methylation. It is important to note that these DMRs are enriched for ICRs and meiosis-associated gene promoters, supporting the critical involvement of TET1’s iterative catalytic activity during germline reprogramming.

The role of TET1 in patterning the sperm methylome is supported by the finding that the majority of DMRs in all three mutant genotypes overlap sperm HMRs. The sperm genome is distinctively hypermethylated; this methylation pattern is acquired during the prospermatogonia stage when DNMT3A/3L complex targets the full genome, excluding protected regions^46,53,62,63. To date, H3K4me3 is thought to be the most dominant DNMT3-repelling epigenetic mark in the germline genome⁴⁶. We determined that Tet1-mutant DMRs overlapped with regions enriched for H3K4me3 in prospermatogonia, suggesting these DMRs are indeed normally excluded from DNA methylation. This result is similar to previous reports in mouse embryonic fibroblasts, where combined loss of TET1 and TET2 promoted methylation invasion into discrete hypomethylated canyons within the genome⁶⁴. It is not entirely understood why the hypermethylated sperm genome is interspersed with discrete HMRs. Included within these regions are maternally methylated ICRs, as well as evolutionarily conserved TSS and subfamilies of species-specific retrotransposons^4,48,49. Clinically, hypermethylation of sperm HMRs has been associated with idiopathic infertility and poor outcomes in couples undergoing fertility treatments^65–68.

Interestingly, the majority of TET1-dependent, H3K4me3-enriched sperm HMRs are located within gene bodies rather than annotated promoters. Using 5’ RACE, we identified Tet1-DMRs at gene bodies of Fat1 and Dyrk2 as bona fide alternative promoters. FAT1 is a protocadherin that controls cell proliferation in kidney development⁶⁹ and DYRK2 is a dual specificity tyrosine kinase that regulates ciliogenesis and Hedgehog signaling during embryogenesis⁷⁰. While these genes have not been previously studied in spermatogenesis, scRNAseq data confirmed their expression in germ cell development. Surprisingly, while abundant expression of lncRNAs and alternatively spliced RNAs have been described for the testes, the use of alternative promoters during spermatogenesis has not been studied in mammals^71,72. In Drosophila, the transition from proliferating spermatogonia to differentiating spermatocytes is accompanied by dramatic change in transcription, with about a third of new transcripts originating from alternative promoters⁷³. Some uses of alternative promoters include rendering tissue specificity, regulating timing of expression during development, and controlling translation efficiency of transcripts⁷⁴. While the function of mammalian sperm HMRs remains enigmatic, our study reveals at least one compelling use of these regions as alternative transcriptional start sites during spermatogenesis.

One of the unexpected findings in catalytically inactive Tet1^HxD/HxD sperm is the unique hypomethylation signature that is not observed in other mutants. There is currently scant data on the relative binding kinetics of TET’s catalytic domain for ox-mCs compared to 5mC or unmethylated cytosine. If the TET catalytic domain prefers to engage with 5mC, it is possible that loss of 5hmC catalysis in Tet1^HxD/HxD sperm may promote prolonged occupancy by TET1^HxD, which then mediates the recruitment of chromatin modifiers to form a DNMT-inaccessible environment. Figure 7 summarizes our finding for the requirement of iterative oxidative capability of TET1 to generate a reprogrammed genome.

Figure 7.

Summary of findings of TET1 iterative oxidative roles during germline reprogramming. TET1 generation of 5hmC and higher ordered oxidized bases 5fC/5caC are required for demethylation of select ICRs, meiosis-associated gene promoters, and some sperm HMRs, which are generated via exclusion of DNMT3 during de novo methylation in prospermatogonia. 5hmC generation by TET1^V is insufficient for these loci to achieve a hypomethylated state prior to mitotic arrest at E14.5. Due to incomplete methylation erasure during reprogramming, a subset of sperm HMRs in Tet1 mutant sperm become hypermethylated, similar to the rest of the sperm genome.

In summary, we conclusively demonstrate the requirement for full catalytic proficiency in TET1 for methylation erasure of select ICRs and meiosis-associated promoters during PGC reprogramming. In the male germline, we observe distinct methylation defects following expression of both 5hmC-stalling TET1^V or the catalytically inactive TET1^HxD mutants compared to Tet1^−/−, allowing us to determine the dependency of select loci on ox-mC generation to achieve complete reprogramming. While normo-methylation can be restored in many loci with catalytically-inactive TET1, loci reliant on efficient oxidation to 5hmC and beyond also encompass a larger portion of the male germline genome than the previously thought. Overall, our study supports the roles of TET1 not only as a component of germline reprogramming, but also as a contributor to patterning the eventual sperm epigenome.

Limitations of study

We employed Illumina Mouse Infinium BeadChip array for our whole genome analysis of the sperm methylome. While the array was designed to represent the mouse genome, only a minority of genomic CpGs (~285,000 CpGs) are examined. There are many more relevant CpGs, including those at ICRs, where probe design is not feasible. Moreover, the array is overrepresented for cancer- and aging-related promoters and CpGs at sparse regions. WGBS will likely capture other TET1-dependent regions and regions that are uniquely altered in the various mutants. We used locus-specific analyses to study demethylation dynamics of select ICRs and meiotic promoters. While we did not detect evidence of re-methylation of these representative TET1’s targets in mutant PGCs, our results cannot discount potentially distinct behavior of other TET1’s targets genome wide. Additionally, while we have included analyses of XX PGCs in the Tet1 catalytic mutants, we have yet to fully characterized the fertility phenotype of Tet1^V/V and Tet1^HxD/HxD females. Finally, we have not measured oxidized 5mC bases (5fC and 5caC). They are highly transient and present at very low abundance during germline reprogramming or in other relevant tissues. In fact, even global 5hmC levels immediately diminish following PGC specification. Thus, the quick turnover and low abundance of ox-mCs in PGCs are challenging to measure using current technologies, as error rates for these available methods exceed the anticipated levels of ox-mC bases.

STAR Methods

RESOURCE AVAILABILITY

Lead Contact

Lead contact of this study is Marisa S. Bartolomei (bartolom@pennmedicine.upenn.edu). Further information and requests for resources and reagents should be directed to and will be fulfilled by co-coresponding authors: Marisa S. Bartolomei (bartolom@pennmedicine.upenn.edu) and Rahul M. Kohli (rkohli@pennmedicine.upenn.edu).

Materials Availability

All unique reagents and mouse lines generated in this study are available from the Lead Contacts with a completed Materials Transfer Agreement.

Data and Code Availability

The accession number for raw and processed Illumina Mouse Infinium Methylation BeadChip CUT&RUN, and RNA-seq data generated in this paper is submitted to GEO: GSE224459. Accession numbers for existing, publicly available data are referenced as appropriate. Any additional information required to reanalyze the data reported in this paper is available from lead contacts upon request.

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Mouse husbandry and maintenance

All experiments were approved by the Institutional Animal Care and Use Committee of the University of Pennsylvania (protocol number: 804211). Mice are housed in polysulfone cages within a pathogen-free facility with 12-12 h light-dark cycle and ad libitum access to water and standard chow (Laboratory Autoclavable Rodent Diet 5010, LabDiet, St. Louis, MO, USA). Tet1 knockout⁵⁷ (017358; B6; 129S4-Tet1^tm1.1Jae/J) and Oct4-GFP⁷⁵ (008214; B6; 129S4-Pou5f1^tm2.Jae/J) were purchased from The Jackson Laboratory and were backcrossed for at least 10 generations to C57BL/6J (B6; The Jackson Laboratory, 000664) background. Oct4-GFP allele was maintained as homozygous in Tet1 heterozygote breeders (Tet1^+/−; Oct4^GFP/GFP). Mouse genomic DNA for genotyping using polymerase chain reaction (PCR) was isolated from ear punches as previously described¹⁶. Primers used for genotyping of Tet1⁻, Tet1^V, Tet1^HxD, Oct4-GFP alleles as well as sex genotyping are listed in Table S5. Timed mating was determined by visual detection of a vaginal sperm plug where E0.5 was taken to be 12.00h (noon) on the day the plug was observed. Visual staging of embryonic age was done at the time of dissection.

Generation and validation of Tet1^V and Tet1^HxD mouse lines

Mutational insertions at exon 10 of endogenous Tet1 allele to generate T1642V or H1654Y;D1656A amino acid substitutions within catalytic domain of TET1 were done using easi-CRISPR-Cas9 editing in the C57BL/6J (B6) and B6D2 (hybrid of B6 and DBA/2J) as previously described^88–90. Briefly, single-strand homology-directed repair (HDR) donor templates carrying the appropriate nucleotide substitutions in Tet1 exon 10 were synthesized as 4 nM Ultramers from Integrated DNA Technologies (see Table S4). The Tet1 exon 10 guide RNAs (gRNAs) were amplified using synthesized oligos and the pX335 gRNA scaffolding vector, in vitro transcribed using MEGAshortscript T7 transcription kit (Ambion), and purified using MEGAclear Transcription Clean-up kit (Ambion). The purified gRNA (50 ng/uL), Cas9 mRNA (100 ng/uL), and HDR donor templates (100 ng/uL) were injected into the pronucleus of single-cell B6 x B6D2 embryos and transferred to pseudopregnant dams. The mosaic founders from the CRISPR injection were ~75% B6 genetic background. Founders were screened by exon 10 restriction fragment length polymorphism (RFLP) using HaeIII enzyme to distinguish Tet1^HxD allele and HphI enzyme to distinguish Tet1^V allele from WT, of PCR amplified product flanking exon 10 (see Table S4). Genotypes of founders were further verified by Sanger sequencing. The targeted Tet1 allele was validated in heterozygote and homozygote animals after backcrossing to B6 strain for at least 3 generations using Southern blot as previously described⁹¹, with restriction enzymes and probes indicated in Supplemental Figure 1 and Table S4. All mice included in this study had been backcrossed to B6 strain for at least 4 generations unless noted otherwise.

METHOD DETAILS

Tissue collection

Sperm

Adult male mice (>10 weeks of age) were housed with a sexually mature female for at least 5 days, and then isolated for at least 3 days. After euthanasia, the caudal epididymis was dissected and the epididymal sperm was collected on a needle and capacitated in EmbryoMax Human Tubal Fluid media (HTF, EMD Millipore) for 30 minutes at 37°C. Motile sperm were collected by removing the supernatant, spun down for 5 minutes at 650 xg and incubated for 15 minutes on ice with somatic cell lysis buffer (0.1% SDS, 0.5% Triton-X-100) to remove any nonsperm contaminants. Following treatment with somatic lysis buffer, the sperm were counted, spun down for 5 minutes at 10,000 xg and snap frozen for storage at −80°C until further processing.

Embryonic germ cells

Embryonic Oct4-GFP⁺ gonads were harvested from embryos at E11.5 to E14.5 as well as E17.5 gonads.. The gonads were dissected in calcium- and magnesium-free phosphate buffered saline (PBS, Gibco) and transferred into 500 μL of 0.25% Trypsin-EDTA (Gibco). Gonads were incubated in Trypsin-EDTA for 10 minutes at 37°C and quenched with equal volume of Hank’s Balanced Salt solution (HBSS, Gibco) containing 5% fetal bovine serum (FBS). To achieve single cell suspension, gonads were triturated using p1000 tips (Denville), followed by p200 tips (Denville) and 22G needle (BD Biosciences). The single cell suspension was centrifuge for 5 minutes at 650 xg and resuspended in 5% FBS in HBSS prior to sorting. GFP+ PGCs were sorted using FACSAria Fusion or FACS Jazz cell sorter (Becton Dickinson). For bisulfite mutagenesis, PGCs were snap frozen for storage at −80°C until further processing. For CUT&RUN, PGCs were immediately processed for permabilization and binding to concanavalin A beads.

Somatic tissues

E10.5 whole-embryo was collected from timed-mating and immediately snap frozen for storage at −80°C until further processing. PND0 brain, tongue, and liver samples were collected following decapitation of neonates. Whole testis, liver, and cortex were dissected from male mice following euthanasia by CO₂ asphyxiation. Tissues were snap frozen for storage at −80°C until further processing.

Tissue homogenization and DNA extraction

Embryonic, neonatal, and adult tissues were digested in lysis buffer (50 mM Tris, pH8.0, 100 mM EDTA, 0.5% SDS) with proteinase K (180 U/mL; Sigma-Aldrich) overnight at 55°C. Sperm pellets were resuspended in sperm lysis buffer (20 mM Tris-HCl, pH8.0, 200 mM NaCl, 20 mM EDTA, 4% SDS) with the addition of 5 μL of β-mercaptoethanol and proteinase K (180 U/mL) at 55°C overnight. Genomic DNA was isolated using Phenol:Chroloform:Isoamyl Alcohol (25:24:1; Sigma-Aldrich) and ethanol precipitation and resuspended in TE buffer (10 mM Tris-HCl pH8.0, 0.5 mM EDTA).

RNA-sequencing

Total RNA was extracted from 5000 E14.5 PGCs (from a single embryo) using Quick-RNA Microprep kit (Zymo Research). After confirming RNA integrity using TapeStation, mRNA library was generated using Oligo d(T)₂₅ mRNA Magnetic Isolation Module and Ultra II RNA Library Prep Kit (New England Biolabs). Library quality was assessed using TapeStation and sequencing was performed on NextSeq 1000 (Illumina, San Diego, CA, USA).

Bisulfite mutagenesis

2 μg of genomic DNA was bisulfite treated using the EpiTect Bisulfite Kit (Qiagen) and eluted in 20 μL of 1:10 of the supplied EB buffer. Snap frozen PGC pellet was directly lysed using the LyseAll Lysis Kit (part of the EpiTect FAST Bisulfite Conversion Kit, Qiagen) and was bisulfite treated using the standard Epitect Bisulfite reagent mix following the low-input protocol. PGC bisulfite-treated DNA was resuspended in 20 μL of 1:10 of the supplied EB buffer.

Library preparation for bisulfite sparse sequencing

Whole genome BS libraries preparation was adapted from Luo et al.⁹² using xGEN Adaptase module (Integrated DNA Technology) following the workflow for single-cell Methyl-Seq (snmC-Seq), which includes random priming step, per manufacturer’s protocol. Additional components for random priming step are as followed: Klenow Exo-DNA Polymerase at 50 U/μL supplied with Blue Buffer (Enzymatics), Exonuclease I at 20 U/μL (Enzymatics), Shrimp Alkaline Phosphatase (NEB), 10 mM dNTP (Promega). Following random priming step, samples were eluted in 10 μL Low EDTA TE (included in the xGEN Adaptase module) and proceeded with Adaptase reaction per manufacturer’s protocol. To determine cycle numbers for enrichment PCR, 1 μL of library from Adaptase reaction was used to run qRT-PCR with the following condition, 0.5 uL of 10uM custom Illumina I7 and I5 primers to accommodate stubby adapter tails on the random primers (final concentration of 0.5 μM and all unique dual index primer sequences are listed in Luo et al.⁹²), 7 μL of 2x NEBNext Library Quant Master Mix with 1:100 low ROX (NEB), and 1.5 μL ddH2O. Indexing PCRs were done with 3 cycles less than the determined qRT-PCR cycle threshold (Ct) using KAPA HiFi HotStart ReadyMix (KAPA Biosystems) with final custom Illumina I7 and I5 concentrations at 1 μM. Amplified libraries were cleaned using two rounds of 0.8X SPRISelect beads and eluted in 13 μL of EB Buffer. Libraries were quantified using NEBNext Library Quant Kit and libraries sizes were determined using Bioanalyzer High Sensitivity DNA Kit (Agilent). Indexed libraries were pooled and sequenced on an Illumina MiSeq using a MiSeq Reagent Kit v2 (150x150; Illumina) with 10% PhiX spike-in to achieve ~65,000 aligned reads per library.

Locus specific DNA methylation analysis using pyrosequencing or targeted next-generation bisulfite-sequencing

Pyrosequencing PCRs and sequencing reactions for ICR methylation (H19/Igf2, Peg1, Peg3, KvDMR, and Snrpn) were described by SanMiguel et al¹⁶. Primers are listed in Table S5. Targeted DNA methylation analyses using next-generation sequencing were modified from IMPLICON protocol⁹³. For assay design, genomic DNA sequences of the regions of interest were obtained from UCSC Genome Browser and imported into MethPrimer⁹⁴ or BiSearch⁹⁵ to identify primer pairs with optimal amplicon size of 300 bp with a minimum of 5 CpGs. Primers are listed in Table S5. Stubby Illumina adapter sequences were added to the forward (ACACTCTTTCCCTACACGACGCTCTTCCGATCT) and reverse (GTGACTGGAGTTCAGACGTGTGCTCTTCCGATCT) primers. 1 μL of bisulfite treated DNA was amplified for the regions of interest using PyroMark PCR kit (Qiagen) with final primer concentration of 0.4 μM. Amplicons of similar sizes (+/− 50 bp) were pooled for column purification (Thomas Scientific) and eluted in TE buffer. 25 ng of pooled amplicons were loaded into indexing PCR using Multiplex PCR kit (Qiagen). Indexed reactions were purified using SPRIselect beads (0.9x; Beckman Coulter), eluted in TE buffer and quantified using Qubit Flourometer dsDNA BR Assay Kit (Invitrogen). Indexed libraries were pooled and sequenced on an Illumina MiSeq using a MiSeq Reagent Nano Kit v2 (250x250; Illumina) with 10% PhiX spike-in.

Western blot

Testes were homogenized by mortar and pestle in 1x RIPA buffer (Cell Signaling Technology) supplemented with Complete Protease Inhibitor (Roche). Tissue lysates were incubated on ice for 20 minutes, followed by centrifugation at 13,000 xg for 5 minutes at 4°C. Supernatant was collected and the protein concentration was quantified using bicinchoninic acid protein assay (BCA assay; Pierce, Thermo Scientific). 50 μg of protein lysate was denatured and were run on a 4-12% SDS-PAGE gel. The gel was transferred onto a PVDF membrane at 250 mA for 120 minutes, and then blocked for 1 hour at RT with shaking in 5% non-fat dry milk in TBS-Tween (TBS-T). For full-length TET1 detection, membranes were probed with 1:1000 rabbit anti TET1 (N3C1, Genetex) overnight at 4°C. For loading control, membranes were probed with 1:10,000 rabbit anti-GAPDH (2118, Cell Signaling Technology) overnight at 4°C. Membranes were washed 3x in TBS-T following primary antibody incubation, and probed with 1:20,000 goat anti-rabbit IgG HRP secondary antibody (Invitrogen) for 1 hour at RT. Membranes were developed using Immobilon Western Chemiluminescent HRP Substrate and imaged on an Amersham Imager 600.

Genome-wide DNA methylation profiling using Infinium Mouse Methylation BeadChip

1000 ng of bisulfite-treated sperm DNA was loaded onto Illumina Infinium Mouse Methylation-12v1-0 BeadChip (llumina) and was ran on an Illumina iScan System (Illumina) per manufacturer’s standard protocol. The samples were processed at the Center for Applied Genomics Genotyping Core at the Children’s Hospital of Philadelphia. Biological replicates for each genotype are as followed, Tet1^+/+ n = 8, Tet1^V/V/ n =10, Tet1^HxD/HxD n= 8, Tet1^−/− n = 10.

E17.5 prospermatogonia CUT&RUN

E17.5 Oct4-GFP⁺ prospermatogonia were collected using FACS as detailed above. Cells were hold on ice until CUT&RUN processing. Gonads from several embryos were pooled and CUT&RUN was done on 130,000 freshly sorted cells per motif (H3K4me3 and IgG) as previously described⁹⁶. Sorted cells were spun down at 600 xg for 3 minutes at room temperature, and washed three times with 1.5 mL of Wash buffer (20 mM HEPES pH 7.5, 150 mM NaCl, 0.5 mM spermidine, with 1x Roche complete protease inhibitor EDTA Free). Cells were bound to concanavalin A-coated magnetic beads (Bangs Laboratories) pre-washed in binding buffer (20 mM HEPES (pH 7.5), 10 mM KCl, 1 mM CaCl₂, 1 mM MnCl₂) for 10 minutes at room temperature. Cells were resuspended with H3K4me3 (Cell Signaling Technology 9751) or rabbit IgG antibodies (Cell Signaling Technology 2729) at a final dilution of 1:100 in 200 μL of antibody binding buffer (same as Wash buffer with 0.05% digitonin and 2 mM EDTA) and incubated overnight at 4°C. Cells were washed twice in Wash buffer with 0.05% digitonin, and incubated with pAG-MNase fusion protein (final concentration 700 ng/mL) for 1hr at 4°C. Following two washes in Wash buffer with 0.05% digitonin, samples were resuspended in Incubation buffer (3.5 mM HEPES (pH 7.5), 10 mM CaCl₂, 0.05% digitonin) to activate cleavage on ice for 5 minutes. 2xSTOP solution (170 mM NaCl, 20 mM EGTA, 0.05% digitonin, 50 μg/mL RNase A, 25 μg/mL Glycogen, 2 pg/mL yeast chromatin spike-in) was added to quench the reaction. To release antibody bound fragments, samples were incubated at 37°C for 30 minutes and supernatant was isolated from the magnet-bound beads. 2 μL of 10% SDS and 2.5 μL of 20 mg/mL Proteinase K was added to the supernatant and samples were incubated at 50°C for one hour to digest any protein, and DNA was isolated using Phenol:Chloroform extraction twice. DNA pellets were resuspended in 30 μL of Tris EDTA buffer (1mM Tris HCl, pH 8; 0.1 mM EDTA). CUT&RUN library was prepared using KAPA HyperPrep Kit (KAPA Biosystems) following the manufacturer’s protocol. Libraries were amplified for 18 cycles using KAPA HiFi Hot Start Ready Mix (Roche). Amplified libraries were cleaned using 1x KAPA Pure Beads. Libraries were quantified using NEBNext Library Quant Kit and libraries sizes were determined using Bioanalyzer High Sensitivity DNA Kit (Agilent).

RNA extraction, reverse transcription, qRT-PCR, and pyrosequencing for allelic expression analysis

For adult testes or PND0 livers, tissue lysates were divided in half by volume and added to TRIzol reagent (Thermo Fisher Scientific). Chloroform was added to achieve phase separation, followed by recovery of the aqueous phase. Equal volume of ethanol was added to the aqueous phase and RNA was bind to Zymo Research RNA miniprep column. RNA purification was done following Quick-RNA Miniprep Plus Kit as specified by manufacturer’s protocol including in column DNAseI treatment (Zymo Research). RNA quantity was determined by NanoDrop ND-1000 spectrophotometer (Thermo Fisher Scientific). RNA samples were reverse transcribed with Superscript III Reverse Transcriptase (Invitrogen) according to the manufacturer’s protocol. Quantitative real-time PCR (qRT-PCR) was performed using Power SYBR Green Master Mix (Applied Biosystems) on a QuantStudio 7 Flex Real-Time PCR system. Relative expression levels were determined using the Pffafl method normalized to the housekeeping gene Nono. To assess expression of Tet1^HxD or Tet1^V allele in heterozygote or homozygote PND0 brain, modified protocol for pyrosequencing for imprinting expression (PIE) was used as previously described in⁹⁰. 10 ng of cDNA was amplified using Pyromark PCR kit (Qiagen) with final primer concentration of 0.4 uM, in which the reverse primer was biotinylated. 4 uL of PCR product was sequenced on the Pyromark Q96 MD Pyrosequencer (Biotage, AB), using PyroMark Gold Q96 CDT Reagents (Qiagen), and Streptavidin Sepharose beads (GE Healthcare). Quantification of allele-specific expression was performed using Pyromark Q96 MD software based on the presence of a SNP (introduced in CRISPR mutagenesis) in the cDNA amplicon. Primers are listed in Table S5.

5’ rapid amplification of cDNA ends (RACE)

SMARTer RACE 5’/3’ Kit from Takara Bio was used following the manufacturer’s instructions. 3’ gene specific primers (Table S5) were designed downstream of the hypothesized DMRs using Primer BLAST with parameters specified by SMARTer RACE kit (23-28 nucleotides long, 50-70% GC content, with Tm > 70°C) with GATTACGCCAAGCTT- overhang on the 5’ ends to allow for cloning into the provided pRACE vector. Reverse transcription was done on 1 μg of total RNA from adult testes. PCR condition modifications to amplify RACE products are as followed using SeqAmp DNA Polymerase (Takara Bio): Fat1 (T_A: 68°C and extension of 6 minutes) and Dyrk2 (T_A: 65°C and extension of 3 minutes). RACE products were gel isolated using Qiaquick Gel Extraction Kit (QIAGEN) and cloned into pRACE vector. Sanger sequenced products were subjected to BLAT alignment to mm10 genome.

Quantification and statistical analysis

Statistics were performed using GraphPad Prism or R. Comparison of > 2 independent groups were performed using one-way ANOVA followed by Tukey’s post-hoc test. Fisher’s exact test was performed to determine significance of the frequency of hypermethylated F1 offspring of Tet1 catalytic mutant males. Bernoulli distribution test was conducted to determine distribution of DMRs genomic compartment as compared to genomic distribution of all probes in the array. Statistical significances are denoted by different letters or asterisks in the graph. Information on statistical tests performed, exact values of n, and degrees of significance are provided in the figure legends.

Analyses of sparse-seq and targeted next-generation bisulfite sequencing

Sequenced reads were trimmed using Trim Galore (version 0.6.7 https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/) in paired-end mode. For sparse-seq, in addition to low quality bases and adaptors, 15 bps were removed from 5’ end of read 1 and 30 bps were removed from 5’ end of read 2 to remove synthetic sequences introduced by Adaptase Module random priming step. Trimmed sequenced reads were aligned to mouse mm10 genome with Bismark (v0. 23.0 https://www.bioinformatics.babraham.ac.uk/projects/bismark/) in paired-end mode. The following Bismark parameters were used to align sparse-seq reads: --score_min L, 0, -0.6 -non_directional. Reads were deduplicated using Picard Toolkit (v2.25.7-0 https://broadinstitute.github.io/picard/) MarkDuplicates. Non-deaminated reads were filtered out based on the presence of ≥ 3 consecutive instances of non-CG methylation (Bismark function: filter_non_conversion; parameters: --paired --threshold 3 --consecutive). Bedgraph files were prepared using Bismark Methylation Extractor function to calculate percent methylation at each CpG and global modified Cytosine levels (in CpG, CHG, CHH, and unknown context). For locus specific analysis, percent methylation at each CpG was calculated from at least 30x coverage.

Genome-wide DNA methylation profiling analyses

Processing of raw IDAT files was done as previously described by Vrooman et al.⁹⁷ using SeSAMe R Package⁴⁵ (v1.10.4) and the MM285 array manifest file (vM25) to obtain methylation β-values (getBetas function). Probes that did not pass SeSAMe’s quality control pOOBAH approach for signal-to-background thresholding are masked with NA⁴⁵. To determined differentially methylated probes, we included only CG probes with no NA values in all biological replicates, totaling in 218,483 probes out of 287,172 that are printed on the array. We first used SeSAMe DML (Differential Methylation Locus) function which models the DNA methylation levels using mixed linear model, a supervised learning framework that identifies CpG loci whose differential methylation is associated with known co-variates (i.e. sample genotype in our experiment)⁴⁵. Only CG probes, as defined in “Infinium Mouse Methylation v1.0 A1 GS Manifest File.bpm”⁴², were included in F-test that is conducted as part of SeSAMe DML function and multiple-testing adjustment. To finalize the lists differentially methylated probes for each genotype compared to Tet1^+/+ was done using SeSAMe DMR function with false discovery rate cut off of 5% (Seg_Pval_adj < 0.05) and minimum difference of 10% between the mean of WTs and mutants. We used SeSAMe built in knowYourCG tool to test enrichments (testEnrichment) for probe design groups, chromHMM chromatin states, and transcription factor binding sites on differentially methylated probes for each mutant genotype compared to WT.

All downstream analysis was conducted using the mm10/GRCm38 mouse genome assembly. Differentially methylated regions (DMRs, each DMR corresponds to one array probe) were assigned to nearby genes and prioritized to genomic features based on proximity using HOMER annotatePeaks.pl function⁷⁸. We calculated Bernoulli distribution of DMRs’ genomic features distributions as compared to the genomic features distribution of all probes in the array. CpG densities of DMRs were assigned using the annotatr R package⁷⁷. Venn diagrams were generated using the R package BioVenn⁸⁷. Partially supervised clustering was conducted on DMRs of all genotypes to determine hyper- or hypomethylated signatures using the R package pheatmap with average clustering method. To assess rescue or partial rescue by TET1^V or TET1^HxD, DMRs that were found in Tet1^−/− were assessed in Tet1^V/V or Tet1^HxD/HxD samples. DMRs were classified as partially rescued if the FDRs were less than 0.05 in Tet1^V/V or Tet1^HxD/HxD but mean differential methylation levels between Tet1^V/V or Tet1^HxD/HxD samples and Tet1^+/+ were less than 10%. DMRs were classified as rescued if the FDRs were greater than 0.05 in Tet1^V/V or Tet1^HxD/HxD samples compared to Tet1^+/+. Volcano plots were made using the R package ggplot2⁸⁶. DMRs were examined for overlap with sperm HMRs, H3K4me3 ChIP-seq or CUT&RUN peaks for each stages of spermatogenesis using BEDtools⁷⁹ (version 2.27.1) intersect. To generate heatmaps for H3K4me3 signals at DMRs, DMRs were first binned into 250 bp non-overlapping windows using BEDtools merge function, heatmaps and metaplots were generated using deepTools⁸⁰(version 3.4.0).

Analyses of ChIP-seq and CUT&RUN

Publicly deposited ChIP-seq (DRA006633⁵³, SRA097278⁵⁴, GSE135678⁵⁵) and CUT&RUN fastq files were trimmed using Trim Galore (version 0.6.7, default parameters). ChIP-seq and CUT&RUN reads were aligned to mouse mm10 reference genome using Bowtie2 (version 2.5.0, parameters: --N 1 -sensitive -local). Following alignment, low quality reads (QMAP ≤ 10) and non-primary alignments were removed using SAMtools⁸¹ (version 1.16.1, view -q 10 -F 256). Duplicated reads were removed using SAMtools rmdup with default parameter and mitochondrial reads were removed using grep function. Alignment BAM files were converted to BED files and blacklisted regions⁹⁸ were remove using BEDtools (version 2.27.1). bigwig files normalized to count per million (CPM) were prepared using deepTools (version 3.5.1, function: bamCoverage, parameters: -bs 1 --normalizeUsing CPM). Peak calling was performed using Model-based Analysis of ChIP-Seq⁸²(macs2, version 2.1.0; parameters: --qvalue 0.01).

Identification of sperm hypomethylated regions

Sperm hypomethylated regions (HMRs) were determined using hmr function of DNMTools⁵¹, which uses hidden Markov model approach to identify methylation canyon in a supplied whole genome bisulfite sequencing methylation call data set (GEO: GSE56697⁵²). In total, 76227 HMRs were identified in sperm WGBS data set. To identify hypomethylated regions in the sperm genome as compared to the oocyte genome, we used the R package methylKit⁸³ calculateDiffMeth function.

Processing of RNAseq

Raw reads were quality-trimed using Trim Galore (version 0.6.7 https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/). Quality of raw reads were assessed using FastQC. Reads were aligned to the GRCm39/mm10 reference genome using STAR version 2.7.10b with default parameters and a maximum fragment size of 2000bp⁹⁹. Properly paired primary alignments were retained for downstream analysis using Samtools version 1.9. Count matrices were generated using featureCounts 2.0.3⁸⁴ against RefSeq gene annotation. Sample batch effect was corrected suing R-package RUVseq v1.34.0 using the RUVr method, using all genes for unwanted variation estimation and the parameter “k=1”¹⁰⁰. Batch-corrected counts were fed into DESeq2 v1.40.2 ¹⁰¹ to perform normalization and differential expression analysis with FDR cut off < 0.05 and minimum fold change of 1.5.

Processing of testis single cell RNAseq (scRNAseq)

Normalized gene expression matrices for adult mouse scRNA-seq spermatogenic germ cell clusters are publicly available (GSE112393⁵⁶). DMRs were matched to their nearest annotated gene using the annotatePeaks function of the Rpackage ChIPSeeker⁸⁵ (version 1.22.1; parameter: TxDb=TxDb.Mmusculus.UCSC.mm10.knownGene). Boxplots depicting normalized expression for DMR-linked genes vs. background gene expression for germ cell clusters were prepared using ggplot2⁸⁶.

Supplementary Material

2

Table S1. E14.5 primordial germ cells RNAseq analyses, related to Figure 1.

3

Table S2. Differential methylation analyses from Illumina Mouse Infinium Methylation BeadChip, related to Figure 4.

4

Table S3. knowYourCG analyses of Tet1 mutant differentially methylated regions, related to Figure 4.

5

Table S4. Overlap of Tet1 mutant differentially methylated regions and H3K4me3 enrichment at different stages of spermatogenesis, related to Figure 5.

6

Table S5. List of oligonucleotides used in this study, related to Figure 1, Figure 2, Figure 3, Figure 5, and Figure 6.

Key Resources Table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit polyclonal anti-TET1	Genetex	Cat#: N3C1; RRID: AB_11176491
Rabbit monoclonal anti-GAPDH	Cell Signaling Technology	Cat#: 2118; RRID: AB_561053
Goat anti-Rabbit IgG-HRP	Invitrogen	Cat#: 31460; RRID: AB_228341
Rabbit monoclonal Anti-H3K4me3	Cell Signaling Technology	Cat#: 9751; RRID: AB_2616028
Normal Rabbit IgG	Cell Signaling Technology	Cat#: 2729; RRID: AB_1031062
Chemicals, peptides, and recombinant proteins
Proteinase K	Sigma-Aid rich	P2308
Phenol:Chloroform:Isoamyl Alcohol	Sigma-Aid rich	P3803
Albumax	Thermo Fisher	11020021
EmbryoMax Human Tubal Fluid Media	EMD Millipore	MR-070-D
Triton X-100	Supelco	12659
Phosphate Buffer Saline (calcium, magnesium free)	Gibco	14190144
0.25% Trypsin-EDTA	Gibco	25200056
Hank’s Balanced Salt Solution	Gibco	14024076
Fetal Bovine Serum (FBS), heat inactivated	Gibco	10082147
Klenow Exo-DNA Polymerase	Enzymatics	P701-LC-L
Shrimp Alkaline Phosphatase	NEB	M0371S
dNTP	Promega	U1511
SPRISelect beads	Beckman Coulter	B233317
Concanavalin A-coated beads	Bang Laboratories	BP531
Streptavidin Sepharose beads	GE Healthcare	GE17-5113-01
pAG-MNase	Cell Signaling Technology	40366
PhiX Control v3	Illumina	FC-110-3001
TRIzol	Thermo Fisher	15596026
RIPA Buffer	Cell Signaling Technology	9806
Complete Protease Inhibitor EDTA Free	Roche	11873580001
Cytobuster Protein Extraction Reagent	Millipore	71009
Immobilon HRP Chemiluminescent Substrate	Millipore	WBKLS0100
Critical commercial assays
Bioanalyzer DNA High Sensitivity Chip	Agilent	5067-4626
Qubit Fluorometer dsDNA BR Assay Kit	Invitrogen	Q32850
GoTaq Green Master Mix	Promega	M7121
Epitect Bisulfite Kit	Qiagen	59104
Epitect Fast Bisulfite Conversion Kit	Qiagen	59824
PyroMark PCR Kit	Qiagen	978703
Quick-RNA Miniprep Plus	Zymo Research	R1057
QiaQuick Gel Extraction Kit	Qiagen	28704
Multiplex PCR Kit	Qiagen	246145
MiSeq Reagent Nanot Kit v2 (500 cycles)	Illumina	MS-102-2003
MiSeq Reagent Micro Kit v2 (300 cycles)	Illumina	MS-102-2002
PyroMark Gold Q96 CDT Reagents	Qiagen	972824
NEBNext Library Quant Kit	NEB	E7630
xGEN Adaptase Module	IDT	10009826
KAPA HiFi HotStart Readymix Kit	KAPA Biosystems	KR0370
Illumina Infinium Mouse Methylation BeadChip	Illumina	20041558
KAPA HyperPrep Kit	KAPA Biosystems	KK8504
Superscript III Reverse Transcriptase	Invitrogen	18080093
Power SYBR Green Master Mix	Applied Biosystems	A46111
SMARTer RACE 5’/3’ Kit	Takara	634859
Deposited data
Raw and processed Illumina Infinium Mouse Methylation BeadChip. See also Table S2 and S3.	This paper	GSE224459
Raw and processed CUT&RUN. See also Table S4.	This paper	GSE224459
Raw and processed RNAseq of primordial germ cell. See also Table S1.	This paper	GSE224459
Whole genome bisulfite sequencing for sperm and oocyte	Wang et al. 201452	GSE56697
Single cell RNA-seq of adult testes	Green et al. 201856	GSE112393
E17.5 prospermatogonia CUT&RUN	This paper	GSE224459
PND0 spermatogonia H3K4me3 ChIP-seq	Kawabata et al 201953	DRA006633
Pachytene and round spermatid H3K4me3 ChIP-seq	Lesch et al. 201354	SRA097278
Sperm H3K4me3 ChIP-seq	Lismer et al. 202155	GSE135678
Experimental models: Organisms/strains
Mouse: Tet1tm1.1Jae/J (Tet1−/−)	Dawlaty et al. 201157	Jackson Laboratory Strain #: 017358; RRID: IMSR_JAX:017358
Mouse: Tet1V/V	This paper	N/A
Mouse: Tet1HxD/HxD	This paper	N/A
Mouse: C57BL/6J	Jackson Laboratory	Jackson Laboratory Strain #: 000664; RRID: IMSR_JAX:000664
Mouse:Pou5f1tm2Jae/J (Oct4-GFP reporter)	Lengner et al. 200775	Jackson Laboratory Strain #: 008214; RRID: IMSR_JAX:008214
Oligonucleotides
Mouse Tet1 CRISPR/Cas9 mutagenesis primers and homology directed repair templates (See Table S5 for sequences)	This paper	N/A
Genotyping primers for Tet1v or Tet1HxD alleles RFLP (See Table S5 for sequences)	This paper	N/A
qRT-PCR primers for Tet isoforms and Tet1 allelic pyrosequencing (See Table S5 for sequences)	This paper	N/A
Primers for generation of Southern blot probes (See Table S5 for sequences)	This paper	N/A
Primers for bisulfite sequencing of candidate DMRs using MiSeq (See Table S5 for sequences)	This paper	N/A
Primers for bisulfite sequencing of ICRs using pyrosequencing (See Table S5 for sequences)	This paper; de Waal et al., 201476	N/A
Primers for 5’ RACE (See Table S5 for sequences)	This paper	N/A
Software and algorithms
Trim Galore (version 0.6.7)	Felix Kruger	https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/
Bismark (v0.23.0)	Felix Kruger	https://www.bioinformatics.babraham.ac.uk/projects/bismark/
Picard Toolkit (v2.25.7-0)	Broad Institute	https://broadinstitute.github.io/picard/
SeSAMe (R package, v1.10.4)	Zhou et al. 201845	https://www.bioconductor.org/packages/release/bioc/html/sesame.html
annotatr (R package)	Cavalcante et al. 201777	https://bioconductor.org/packages/release/bioc/html/annotatr.html
HOMER (v4.11)	Heinz et al. 201078	http://homer.ucsd.edu/homer/index.html
BEDtools (version 2.27.1)	Quinlan et al. 201079	https://github.com/arq5x/bedtools2/releases
deepTools (version 3.4.0)	Ramirez et al. 201680	https://deeptools.readthedocs.io/en/develop/
SAMtools (version 1.16.1)	Danecek et al. 202181	https://samtools.sourceforge.net/
macs2 (version 2.1.0)	Zhang et al. 200882	https://github.com/macs3-project/MACS/wiki/Install-macs2
DNMTools	Song et al. 201351	https://dnmtools.readthedocs.io/en/latest/
methylKit (R package)	Akalin et al. 201283	https://www.bioconductor.org/packages/release/bioc/html/methylKit.html
featureCounts	Liao et al. 201484	https://subread.sourceforge.net/featureCounts.html
ChIPseekers (R package, version 1.22.1)	Yu et al. 201585	https://bioconductor.org/packages/release/bioc/html/ChIPseeker.html
ggplot2 (R package)	Ginestet, C. 201186	https://cran.r-project.org/web/packages/ggplot2/index.html
BioVenn (R package)	Hulsen et al. 200887	https://cran.r-project.org/web/packages/BioVenn/index.html
Pheatmap (R package)	Raivo Kolde	https://cran.r-project.org/web/packages/pheatmap/index.html
RUVseq (R package)	Davide Risso	https://bioconductor.org/packages/release/bioc/html/RUVSeq.html
DESeq2 (R package)	Michael Love	https://bioconductor.org/packages/release/bioc/html/DESeq2.html
GraphPad Prism 9	GraphPad Software	www.graphpad.com

Prasasya et al. investigate the role of TET1 catalytic activity in germline reprogramming. Using catalytically inactive and 5hmC-stalling TET1 mice, they demonstrate the requirement of 5mC oxidation into 5fC/5caC for demethylation of select imprinting control regions, meiotic gene promoters, and an additional subset of loci classified as sperm hypomethylated regions.

Highlights.

• Non-catalytic and 5hmC-stalling TET1 mice decouple active demethylation pathways

• TET1 proficiency to generate 5fC/5caC is required for germline reprogramming

• Full length TET1 catalytic mutants rescue most methylation defects of Tet1-KO sperm

• TET1 catalytic activity is important for integrity of sperm hypomethylated regions