MORC1 represses transposable elements in the mouse male germline

The Microrchidia (Morc) family of GHKL ATPases are present in a wide variety of prokaryotic and eukaryotic organisms but are of largely unknown function. Genetic screens in Arabidopsis thaliana have identified Morc genes as important repressors of transposons and other DNA-methylated and silent genes. MORC1-deficient mice were previously found to display male-specific germ cell loss and infertility. Here we show that MORC1 is responsible for transposon repression in the male germline in a pattern that is similar to that observed for germ cells deficient for the DNA methyltransferase homologue DNMT3L. Morc1 mutants show highly localized defects in the establishment of DNA methylation at specific classes of transposons, and this is associated with failed transposon silencing at these sites. Our results identify MORC1 as an important new regulator of the epigenetic landscape of male germ cells during the period of global de novo methylation.

T wo Morc genes in A. thaliana, AtMorc1 and AtMorc6, were identified in forward genetic screens for novel transcriptional repressors 1,2 . AtMORC1 and AtMORC6 are required for silencing of a variety of transposons and are essential for higher-order chromatin compaction. The single Morc gene in Caenorhabditis elegans was also shown to be required for silencing of a repetitive transgene locus 1 . The founding member of the Morc gene family is mammalian Morc1. MORC1 is highly expressed in the blastocyst and male germline but is not expressed in most differentiated cells 3 . Mice deficient for MORC1 are normal, except that homozygous mutant males are infertile with small testicles (hence the name microrchidia) 4,5 . Male germ cells in the Morc1 mutant do not undergo successful chromosomal pairing during the zygotene stage of meiosis and instead undergo apoptosis, with no germ cells surviving to complete prophase I.
During germ cell development, most DNA methylation is lost between E8.5 and E13.5. Then, between E13.5 and birth (BE19), the genome undergoes global de novo methylation [6][7][8] . Failure to establish DNA methylation at this time causes transposon upregulation and meiotic failure. Indeed, the meiotic block in the Morc1 mutant is similar to that observed for mice that have defects in DNA methylation and transposon repression, including mice deficient for DNA methyltransferases [9][10][11] or the pre-meiotic Piwi-interacting RNA (piRNA) pathway 12,13 . Therefore, we hypothesized that MORC1 might be a critical factor for transposon silencing and DNA methylation in the mouse germline. Here we demonstrate that MORC1-deficient male germ cells undergo transposon derepression starting in late embryogenesis and continuing through the onset of meiosis. We also demonstrate that this phenotype is associated with failed locus-specific de novo methylation targeted specifically towards late-methylating transposon sequences.

MORC1 represses transposons in the male germline.
To further characterize MORC1 we used a previously described FVB/N Morc1 mutant (Morc1 tg ) mouse strain in which a tyrosinase gene was integrated into the Morc1 locus 5 . Transgene insertion resulted in loss of exons 2-4, eliminating a large region of the GHKL ATPase domain including residues predicted to be critical for catalysis and ATP binding 14 (Fig. 1a). Consistent with previous reports, we found that Morc1 tg/tg mice have a spermatogenesis defect with a complete absence of post-meiotic spermatids and spermatozoa ( Supplementary Fig. 1a-c).
Quantative reverse transcription-PCR (qRT-PCR) from wild type (WT), embryonic whole testis indicates that Morc1 messenger RNA becomes detectable at E14.5 and peaks at E16.5 ( Supplementary Fig. 2a), which intriguingly is a period of rapid transposon methylation in the male germline. We generated an antibody against the coiled-coil domain of mouse MORC1 ( Supplementary Fig. 2b) and found that MORC1 was localized to the nucleus of male germ cells at E16.5 (Fig. 1b). Conversely, MORC1 protein was undetectable in Morc1 tg/tg mutant germ cells (Fig. 1b). To test for transposon derepression in Morc1 tg/tg germ cells, we performed immunostaining for LINE1 ORF1p and intracisternal particle A (IAP) at postnatal day 14.5 (P14.5). We found that both of these transposon classes were ectopically expressed during early postnatal development, with particular enrichment of LINE1 ORF1p in the meiotic cells towards the centre of the tubule (Fig. 1c,d). LINE1 ORF1p was also derepressed at E16.5 and E18. 5, showing that transposon derepression in Morc1 tg/tg arises well before the apparent meiotic defect ( Supplementary Fig. 2c).
To identify which genes and specific transposons are targets of MORC1-mediated repression, we performed mRNA-Sequencing (RNA-Seq) of whole testes from Morc1 tg/tg and Morc1 tg/ þ heterozygous controls at P10.5 and P14.5. Given that germ cells make up only a small percentage of the testis during the embryonic early and postnatal period, we also performed RNA-Seq on ribosomal RNA-depleted total RNA from sorted germ cells at E16.5, E18.5, P2.5 and P10.5. To purify germ cells, we crossed the Morc1 tg allele into the Oct4-IRES-eGfp reporter strain, which exhibits a distinct eGFP þ population from e9.5 to P2.5 ( Supplementary Fig. 3a) 15,16 . To isolate germ cells at P10.5, we gated the side scatter (SSC) low (SSC Lo ), epithelial cell adhesion molecule (EPCAM) high (EPCAM Hi ) and major histocompatibility complex negative (MHCI À ) population of testicular cells (Supplementary Fig. 3b). A list of sequencing samples and the experiments to which they correspond is included in Supplementary Data 1. Consistent with RT-PCR and published data 5,17 , Morc1 mRNA was confirmed to show high expression at E16.5 and lower expression at later time points in the sorted germ cells (Fig. 2a).
The RNA-Seq data showed a broad transposon derepression defect in Morc1 tg/tg starting at E16.5 (Fig. 2c-e and Supplementary Data 2). Quantitative RT-PCR analysis of various transposable element classes gave similar results as RNA-Seq data ( Supplementary Fig. 4a). In addition, RNA-Seq on sorted Morc1 tg/tg and control germ cells at P10.5 resembled the pattern of transposon derepression observed in whole testes ( Supplementary Fig. 4a). Heterozygous Morc1 tg/ þ mice showed no marked increase in transposon expression relative to WT Morc1 þ / þ mice, confirming their validity as littermate controls ( Supplementary Fig. 4b).
Different transposon classes showed different patterns of derepression in Morc1 tg/tg . Some classes (RLTR4, RLTR6, MuRRS and Etn) were upregulated during embryogenesis but silenced even in the knockout at later time points (Fig. 2b). Other transposons (MMERVK10C, GLN and some IAP species) were most highly upregulated at postnatal time points (Fig. 2c,d). LINEs were upregulated both in late embryogenesis and again at P14.5 after the onset of meiosis (Fig. 2e), which was confirmed by immunofluorescence ( Fig. 1d and Supplementary Fig. 2c). These fluctuations in transposon upregulation may reflect differences in the inherent transcriptional programmes of certain transposon classes, as well as varied effectiveness of other, partially redundant transposon repression pathways at different times. In the aggregate, however, these results indicate that MORC1 constitutes a new participant in transposon repression in the mammalian germline, acting on many different elements. Notably, MORC1 silences many transposon classes well after it is downregulated, consistent with it acting through an epigenetic mark such as DNA methylation.
piRNA biogenesis occurs normally in the Morc1 mutant. During this period in germline development (E14.5 to birth), germ cells undergo mitotic arrest and global nuclear reprogramming that most notably involves genome-wide de novo DNA methylation mediated by the DNMT3A/DNMT3L complex 11,18 . The premeiotic piRNA pathway, involving the nuclear PIWI protein MIWI2, is also active during this period in promoting transposon silencing. To evaluate whether MORC1 acts on the same transposon classes as DNMT3L or MIWI2, we performed RNA-Seq on whole testes from Dnmt3l À / À (ref. 10) and Miwi2 À / À (ref. 12) mice and their respective controls at P10.5, and compared this with the Morc1 tg/tg P10.5 whole testis data set. Morc1 tg/tg and Dnmt3l À / À exhibited derepression of an overlapping set of transposons, primarily long terminal repeat (LTR) retrotransposons, while the Miwi2 À / À mutant testes had a milder phenotype and showed derepression of specific LINE elements and the IAP-Ey class of retrotransposons, which was not affected in Morc1 tg/tg or Dnmt3l À / À mutant testes (Fig. 2f-h).
The lack of overlap between Morc1 and Miwi2 predicts that piRNAs would be unperturbed in Morc1 tg/tg mice. To test this, we performed small RNA sequencing of the testis at E16.5 to examine piRNA production. Our data revealed that the ratio of piRNA/microRNA and the generation of antisense piRNAs were unaltered in Morc1 tg/tg (Table 1), indicating that the piRNA pathway remains largely intact in Morc1 tg/tg testis at E16.5, and that transposon derepression in Morc1 tg/tg is most likely to be independent of the piRNA pathway. In fact, at P10.5 we observed an increase in the fraction of piRNAs derived from LTR retrotransposons, especially of the IAP family, in the MORC1-deficient testis ( Supplementary Fig. 5a-c), similar to that observed in Dnmt3l À / À 19 . These LTR transposon-derived piRNAs corresponded to primary sense piRNAs ( Supplementary  Fig. 5d,e), suggesting that they are probably more abundant simply because the underlying mRNA species are derepressed in Morc1 tg/tg , and some fraction are converted to piRNAs ( Supplementary Fig. 5f). Hence, our results indicate that MORC1 acts in a transposon-silencing pathway independent of piRNA production.   self-renewing spermatogonial stem cell population 20,21 . In contrast to Dnmt3l À / À mutant germ cells that show a dramatic global reduction in DNA methylation 22 , we found no change in global levels of methylation at any time point in Morc1 tg/tg -sorted germ cells (Fig. 3a). Thus, despite the similar morphological phenotypes and transposon expression defects of Morc1 tg/tg and Dnmt3l À / À mice, MORC1 does not act by controlling de novo or maintenance methylation at a genomewide level.
In mammals, DNA methylation is very dynamic and promoter DNA methylation frequently correlates with gene repression. To determine whether there may be localized defects in DNA methylation in Morc1 tg/tg , and whether these are associated with derepressed transposons identified by RNA-Seq, we calculated statistically significant differentially methylated regions (DMRs) in the Morc1 tg/tg germ cells relative to the Morc1 tg/ þ control. At E16.5, we found very few DMRs (Fig. 3b). However, at P2.5 we identified 6,309 hypomethylated regions (Supplementary Data 3) but only 145 hypermethylated regions (Supplementary Data 4), indicating that Morc1 tg/tg germ cells have locus and stage-specific DNA methylation defects (Fig. 3b). In addition, the overwhelming majority of regions identified as hypomethylated at P2.5 remain hypomethylated at P10.5 ( Fig. 3c) and only a few regions lost methylation between P2.5 and P10.5 (Fig. 3d).
The hypomethylated DMRs in Morc1 tg/tg germ cells were highly enriched for LINE and LTR transposons rather than protein-coding genes compared with control regions (see Supplementary Methods), consistent with the transposon expression defects observed in Morc1 tg/tg germ cells (Fig. 4a). Indeed, 93.9% of hypomethylated DMRs contained an LTR or LINE, compared with 40.6% of control DMRs. The hypomethylated DMRs were strongly concentrated in the categories of transposons that showed evidence of derepression ( Fig. 4b-d) during some stage of development before meiosis.
A partial exception to this trend were IAP elements. Hypomethylated DMRs were strongly enriched for IAP elements and corresponding LTRs ( Supplementary Fig. 6a), but there was a poor correspondence between the extent to which a subcategory of IAPs was upregulated and the frequency of overlap with DMRs ( Fig. 2d and Supplementary Fig. 6b). This is probably because certain highly similar repetitive elements such as LTR1 give very few uniquely mapping reads and are therefore missing from the data set. To overcome this, we also mapped the BS-seq data to RepBase consensus sequences for relevant transposons. We confirmed hypomethylation of the upregulated IAPLTR1 class ( Supplementary Fig. 6c). Mapping to repeat consensus sequences also confirmed hypomethylation of LINE and LTR classes, which frequently overlap with DMRs ( Fig. 4c-e).
Only 20 protein-coding genes contained an annotated transcription start site (TSS) within 1 kb of a hypomethylated DMR (Supplementary Data 5) and only 3 contained a TSS within a DMR. Interestingly, all three of these genes (Nebulin, Tmc2 and Cdkl4) contain an RLTR10A transposable element immediately upstream of the TSS and all three genes showed a statistically significant increase in expression (Supplementary Fig. 7 and Supplementary Data 5). Thus, at a very few loci, MORC1 regulates genic expression, probably as a byproduct of its transposon repression activity in the local neighbourhood.   Considering MORC1's role as local modulator of DNA methylation, we examined changes in methylation in the three well-characterized paternally methylated imprinted loci 23 . Methylation occurred normally at two of the three loci (H19 and Dlk1-Gtl2), but the imprinting control region of Rasgrf1 showed increased transcription and hypomethylation in the Morc1 mutant ( Supplementary Fig. 8). Interestingly, this is a transposon-rich area, which has previously been demonstrated as a target of the piRNA pathway 24 (see Discussion below).   BS-seq data were mapped to annotated RepBase consensus sequences for each transposon class and methylation is plotted at each CG in the annotated RepBase consensus sequence relative to the consensus sequence. The first 50 bases from each element, which often have low read coverage, are omitted. As it was usually not possible to determine the orientation of LTR-derived reads relative to LTR transposons, the average methylation level for the relevant LTR species is shown. (e) Global hypomethylation of LTR species corresponding to upregulated transposon classes. Again, BS-seq data were mapped to RepBase consensus sequence for these LTRs.
Of the very few hypermethylated loci observed in the Morc1 mutant, most were not conserved across time points and are probably a consequence of biological or statistical noise. However, 15 hypermethylated DMRs were reproducible between P2.5 and P10.5. Nine of these 15 were embedded in 2 transcripts upregulated in Morc1 tg/tg : 6 DMRs contained within the body of the Cdkl4 gene described above and 3 DMRs in an unannotated transcript probably originating from a hypomethylated IAPLTR1 element (Supplementary Fig. 9). These are probably examples of transcriptional run-through from a nearby promoter causing methylation of a locus, a phenomenon that has been described for some imprinted loci 25 .
DMRs are sites of transposon transcriptional initiation. The highly localized affect of MORC1 on the germline epigenome suggests that MORC1 may function at the transcriptional start sites of transposons to facilitate their silencing and methylation. In support of this, we discovered that hypomethylation in Morc1 tg/tg mutant germline cells was concentrated at the 5 0 -ends of LINE elements coincident with the location of transcriptional initiation (Fig. 4c) 26 . Furthermore, LTR transposons, which are typically flanked by LTRs that serve promoter and enhancer functions 27 , showed hypomethylation on both ends in Morc1 tg/tg (Fig. 4d), and the LTRs themselves are heavily hypomethylated (Fig. 4d,e).
We also noted that hypomethylated DMRs in Morc1 tg/tg germ cells were late targets for de novo methylation during the course of epigenetic reprogramming, since in control Morc1 tg/ þ cells these genomic regions were also hypomethylated relative to the genome average at E16.5 (Fig. 3c,d and Supplementary Fig. 10). This suggests that these loci are somewhat resistant to de novo methylation. Consistent with this possibility, we also discovered that these Morc1 affected genomic regions have increased H3K4me3 relative to control regions of the genome in WT cells E13.5 ( Supplementary Fig. 11a) 28 . To determine whether these loci are enriched in H3K4me3 during the dynamic de novo methylation of the germline genome, we performed chromatin immunoprecipitation (ChIP) analysis for H3K4me3 at E16.5 and confirmed that these regions still exhibited higher H3K4me3 (Fig. 5a). In contrast, other chromatin marks such as H2B10ac, K3K27ac and H3K27me3 showed no correlation with Morc1 tg/tg DMRs ( Supplementary  Fig. 11b). It is well established that H3K4 methylation antagonizes de novo DNA methylation by blocking DNMT3A/ 3L binding to histone H3 (ref. 29). Thus, the presence of H3K4me3 could potentially explain why these loci methylate with slow kinetics and require an additional factor (MORC1) for eventual silencing and methylation. It is also well established that H3K4me3 is a mark of transcriptional start sites, consistent with the idea that many DMRs are TSSs for transposons that are active in the embryonic germline.
At E16.5, we found that RNA transcripts were significantly elevated at Morc1 tg/tg -hypomethylated DMRs relative to the surrounding areas, even in control Morc1 tg/ þ germ cells (Fig. 5b). However, by P10.5, this RNA expression was severely repressed in control Morc1 tg/ þ germ cells with modest but increased expression in Morc1 tg/tg (Fig. 5b). This data is consistent with a model in which the hypomethylated DMRs correspond to TSSs that are normally methylated and suppressed during development.
To further confirm that these DMRs correspond to transcriptional start sites, we employed ATAC-seq, which can be used to identify areas of open chromatin that are a signature of promoter and enhancer sites 30 . We confirmed that ATAC-seq can be accurately adopted for small sample sizes, that reads cluster near transcriptional start sites in E16.5 germ cells, and that ATAC-seq read density at the TSS correlates with gene expression (Supplementary Fig. 12). We found that ATAC-seq peaks overlapped tightly with DMRs (Fig. 5c,d) and most DMRs showed ATAC-seq reads substantially elevated over background (Fig. 5e). In contrast, reads from the naked DNA control were not enriched over DMRs (Fig. 5d). At P10.5, a more limited subset of DMRs exhibited elevated ATAC-seq reads (Fig. 5e), consistent with the observation that transcription is retained only at some DMRs in postnatal germ cells (Supplementary Fig. 13a,b). Other sites lose ATAC peaks and transcription ( Supplementary  Fig. 13c), either because relevant transcription factors are absent or because other mechanisms of transposon silencing are effective. Importantly, at E16.5, where we observe expression from DMR regions in both Morc1 tg/ þ and Morc1 tg/tg cells, we also observed a high ATAC-seq signal in both Morc1 tg/ þ and Morc1 tg/tg cells (Fig. 5d). In contrast, at P10.5, where DMRs are silenced in heterozygotes but remain expressed in Morc1 tg/tg , we only observed high ATAC-seq signal in the Morc1 tg/tg cells (Fig. 5d). These results support the view that DMRs in Morc1 tg/tg cells correspond to promoters of transposons that fail to silence properly, leading to an inappropriately open chromatin state, and ectopic transposon expression, which is retained at P10.5, even after MORC1 expression has ceased.

Discussion
The results of this study identify MORC1 as a critical regulator of transposon repression in the male germline. MORC1 does not act as a global regulator of DNA methylation. Instead, MORC1 functions to facilitate DNA methylation of a variety of transposons in the germline with very little effect on the expression or methylation of protein-coding genes. The observation that Morc homologues are required for gene silencing in Arabidopsis, C. elegans and now mammals suggests that the Morc family of proteins constitute conserved epigenetic regulators that probably function in a wide variety of eukaryotic organisms and developmental contexts.
The Morc1 tg/tg phenotype of transposon derepression and a block in meiosis prophase I superficially resembles the phenotype observed in mice deficient for proteins involved in the prepachytene piRNA pathway, including Mili 31,32 , Miwi2 (ref. 12 39,40 . What distinguishes Morc1 tg/tg from these characterized pre-pachytene piRNA mutants, however, is the apparently normal piRNA biogenesis in Morc1 tg/tg (Table 1). We do note similarities in the pattern of hypomethylation in Morc1 tg/tg and Mili À / À mutant germ cells, including the Rasgrf1 imprinting control region 24 , as well as many of the same transposon families 41 . The dissimilarity in transposon repression observed in Miwi2 À / À and Morc1 tg/tg germ cells (Fig. 2f,h) suggests that MORC1's role in the nucleus is independent from the nuclear piRNA pathway mediated by MIWI2. It is possible that MORC1 participates downstream of the nuclear piRNA pathway during embryogenesis and has a separate, piRNAindependent silencing role during the postnatal stages. This could cause Morc1 tg/tg to have a broader transposon derepression phenotype than Miwi2 À / À . Alternatively, there may exist a MILI-dependent, MIWI2-independent mechanism for promoting methylation of target loci.
TEX19.1 has also been implicated in transposon repression in the male germline and has no known link to the piRNA pathway 42 . However, TEX19.1 is cytoplasmic 42,43 , shows dysregulation only of MMERVK10C elements 42,44 and Tex19.1 À / À has an incomplete infertility defect 42,43 . Thus, the Morc1 tg/tg and Tex19.1 À / À defects are fairly dissimilar and there is no evidence that they participate in the same pathway.   Although we have revealed a critical role for MORC1 in transposon silencing, the actual mechanism by which MORC1 promotes DNA methylation in the male germline is unknown.
Our study suggests at least three potential routes by which MORC1 represses transposons and facilitates DNA methylation. One possibility is that MORC1 directly silences transcription, perhaps using its ATPase activity to compact chromatin, thereby reducing H3K4 methylation levels at target sites. This silencing would allow for normal de novo methylation by DNMT3L. A second possibility is that MORC1 could recruit an H3K4 demethylase, which would similarly promote DNA methylation. Either mechanism agrees with our observation that MORC1hypomethylated DMRs originate from loci with increased H3K4me3 at E13.5. A third non-mutually exclusive possibility is that MORC1 directly recruits the DNA methylation machinery to target loci, mediating methylation and silencing.
In conclusion, a robust genome defense system in the male germline is critical to safeguard genome integrity. We have identified a new participant that acts by facilitating DNA methylation of specific repetitive elements classes.

Methods
Mice. FVB/N-Morc tg/ þ (Tyr)1Az /J mice (Morc1 tg ) were recovered from cryopreservation at the Jackson Laboratory and maintained by intercrossing brothers and sisters in the FVB/N background. Male Morc1 tg/tg mice were viable healthy but infertile, whereas female Morc1 tg/tg mice were viable healthy and fertile. For PCR genotyping, the WT allele was detected as a 347-bp band with the following primers; forward: 5 0 -ATGCAACTTGAGGGGAAACA-3 0 and reverse: 5 0 -GCAGGAG TTATGCGATGTCA-3 0 , and the mutant allele was detected as a 244-bp band with the following primers; forward: 5 0 -AGTTAGCCGTTATTAGTGGAGAGG-3 0 and reverse: 5 0 -AGAAAGCCTGCCTCAAAACA-3 0 . PCR conditions involved ten cycles of 94°C, 65°C and 68°C, followed by 28 cycles of 94°C, 50°C and 72°C. For sorting germ cells from E16-5-P2.5, Morc1 tg /tg females were crossed into the Oct4-IRES-Gfp mixed background. For embryonic staging, timed pregnancies were established and the day a vaginal plug was identified was called embryonic day 0.5 (E0.5). For postnatal time points, the day a litter was first observed was referred to a postnatal day 0.5 (P0.5).
All animal experiments were approved by The UCLA Institutional Animal Care and Use Committee, also known as the Chancellor's Animal Research Committee.
Antibodies. Murine Morc1 coiled-coil domain (amino acids 788-950), expressed in and purified from bacteria, was provided by Jiamu Du and Dinshaw Patel (Sloan Kettering). Anti-Morc1 antibody was raised in rabbit in collaboration with Rockland Immunochemicals.
Anti-LINE Orf1p antibody was provided by Alex Bortvin (Carnegie Institution for Science) and anti-IAP Gaga antibody was provided by Bryan Cullen (Duke).
Embryonic germ cell purification. Collection of embryonic testes were performed following institutional approval for appropriate care and use of laboratory animals. Pregnant females were euthanized using CO 2 and the embryos removed from the womb and stored on a 10-cm dish filled with chilled 1 Â PBS. Testicles were removed from the embryos, placed in an individual 15-ml falcon tube with 3 ml of 0.25% Trypsin, with 3 ml of DNAse I 1 Unit per 1 ml (Life Technologies). Testes were incubated for 15 min at 37°C. After incubation, the cells were agitated into suspension gently by pipetting. The trypsin was then quenched using 5 ml DMEM/ 10% fetal bovine serum (Life Technologies). The cells were centrifuged at 278g for 5 min and resuspended in 500 ml FACS buffer (1 Â PBS 1% BSA). 7-Aminoactinomycin D was added at a 1:50 dilution (BD Biosciences) and the cells strained through BD FACS tubes (Corning) before analysis. Green fluorescent proteinpositive cells were sorted into Buffer RLT (Qiagen) or ATL (Qiagen) for RNA or DNA extraction, respectively.
Postnatal germ cell purification. Pups were euthanized using isoflurane. The testes were removed using tweezers, placed in a 1.5-ml centrifuge tube and chilled on ice. When all testes had been removed, each pair was placed in 1 ml of type IV collagenase (Invitrogen) in an ultra-low-attachment six-well plate (Corning).
All extraneous tissue and the tunica were removed and the seminiferous tubules were teased apart. The samples were then incubated at 37°C for 15 min and centrifuged for 5 min at 278g. Testes were then resuspended in 500 ml of 0.25% Trypsin (Life Technologies) and incubated for 5 min at 37°C. After the incubation period, the testes were agitated gently into suspension by pipetting. Five hundred microlitres of DMEM/10% fetal bovine serum was added and the samples were centrifuged for 5 min at 200g.
For the P2.5 timepoints, green fluorescent protein-positive cells were sorted as with embryonic time points. To sort germ cells at P10.5, the cells were washed with 1 ml FACS buffer and then resuspended in 500 ml FACS buffer. Cells were then incubated with 1:160 EPCAM PE (Biolegend 118205) and 1:250 ml H2-K q 647 (Biolegend 115106) on ice for 20 min in the dark, then centrifuged 5 min at 200g and resuspended in 500 ml FACS buffer. DAPI was added (1:1,000, Life Technologies) and the cells were strained through BD FACS tubes (Corning) before analysis. SSC lo EpCAM hi H2-K q À cells were sorted into Buffer RLT or ATL for RNA or DNA extraction, respectively.
qRT-PCR of Morc1. For embryonic samples, gonads were pooled from approximately five to seven mice per time point. RNA was extracted by the TRIzol method and DNase-treated (Qiagen) before complementary DNA conversion (Superscript III, Life Technologies). Quantitative amplification of cDNA was performed in triplicate using SYBR Green quantitation (PCR primers listed below) on a 7900 HT Fast Real Time PCR System (Applied Biosystems).
RNA preparation. RNA for was extracted from whole testes or cells using the RNeasy Micro Kit (Qiagen 74004). The material was quantified using a Nanodrop ND-1000 (Nanodrop) for RNA from whole testis or the Qubit RNA High Sensitivity Assay (Life Technologies) for RNA from sorted germ cells. RNA quality for material from whole testis was assessed by gel electrophoresis and visualization of the 28S and 18S rRNA bands.
DNA preparation. DNA for bisulfite sequencing was extracted using the QiaAMP DNA Micro kit (Qiagen) and quantified using the Qubit dsDNA High Sensitivity Kit (Life Technologies).
qRT-PCR of retrotransposons. qRT-PCR for retrotransposons was conducted using published primer sets 45 . One microgram total RNA was treated with DNAse I Amplification Grade (Life Technologies) and converted to cDNA using SuperScript II Reverse transcriptase and random hexamers as primer (Life Technologies). The samples were digested with RNAse H in accordance with manufacturer's protocol. RT-PCR was then performed using iQ SYBR Green Mastermix (BioRad) with 750 nM concentration of each primer. The samples was amplified (PCR programme: 95°C 10:00, 50x (95°C 30s, 55°C 30s, 72°C 30s)) with detection of PCR product after each elongation step and determination of melting temperature after the completion of PCR. The reaction was performed using an Agilent Technologies Mx3005p qPCR System (Stratagene). Upregulation of transposon transcript in the mutant is estimated using difference of squares with glyceraldehyde 3-phosphate dehydrogenase as a control.
RNA-seq library preparation. RNA from whole testes was processed for sequencing using a TruSeq RNA Sample Preparation Kit v2 (Illumina) with 250 ng-2 mg total RNA as starting material. Mutant and controls were always matched for starting RNA content. RNA from sorted germ cell was processed using the Ovation Human FFPE RNA-Seq Multiple kit (Nugen) using custom primers for depletion of murine rRNA provided by the manufacturer, using 10 ng of total RNA. Each library was prepared using RNA from one individual mouse.
Bisulfite library preparation. Libraries were prepared using the Ovation Ultralow Methyl-Seq Library System (Nugen). Five to 25 ng DNA was used as starting material. Matched mutant and control samples always contained identical quantities of DNA. Unmethylated Lambda phage DNA (NEB) was spiked in at 0.5% input DNA quantity to determine conversion efficiency, which was consistently 498%. Each library was prepared using DNA from one individual mouse.
ChIP sequencing. The ChIP sequencing (ChIP-seq) protocol was adapted from published sources 28 . FACS-sorted cells from an individual mouse were diluted to 292 ml with 1 Â PBS at room temperature. Formaldehyde (Sigma) was added to a final concentration of 1% and the sample was incubated for 10 min at room temperature with rocking. One molar glycine was then added to yield a final concentration of 0.14 M and the samples were quenched 30 min with rocking. Cells were then spun at 425g for 10 min at room temperature. The cell pellet was flash frozen.
After thawing, the cells were resuspended in 200 ml lysis buffer (50 mM Tris-Cl pH 8.0, 20 mM EDTA pH 8.0, 1% SDS, 1 Â Complete Protease Inhibitor (Roche)) and incubated on ice for 10 min. Samples were then subjected to a 9-min disruption using a Bioruptor on 'High' setting, with 30 s/30 s off disruption (hence, 4.5 min of disruption in total). Samples were spun at 14,000g for 10 min, to remove insoluble material. The soluble sample was diluted to 500 ml with dilution buffer (16.7 mM Tris pH 8, 0.01% SDS, 1.1% TritonX-100, 1.2 mM EDTA, 167 mM NaCl) and 10% of material was saved as input. Sample was precleared with 30 ml Protein A Dynabeads (Life Technologies) and preincubated for 1 h. The cleared material was incubated with 1 ml anti-H3K4me3 antibody (Millipore 04-745) overnight.
The input samples were thawed and diluted with 200 ml buffer. Cross-linking of ChIP and input samples was reversed by incubating 16 h at 65°C. Samples were cooled and treated with 1.5 ml of 10 mg ml À 1 RNaseA (PureLink RNAse A, Invitrogen 12091-021) for 30 min at 37°C. One hundred micrograms of Proteinase K was then added and the samples treated for 2 h at 56°C. The samples were then purified using a Qiagen Minelute kit.
Samples were amplified by a SeqPlex DNA Amplification kit (Sigma) and then converted to libraries using an Ovation Rapid Library kit.
ATAC-seq library construction. Libraries were generated using a method adapted from published protocol 30 . Briefly, FACS-collected cells from individual mice were spun at 500g for 5 min at 4°C. Cells were resuspended in 50 ml lysis buffer (10 mM Tris pH 7.4, 10 mM NaCl, 3 mM MgCl 2 , 0.1% NP40, 1 Â Complete Protease Inhibitor(Roche)) and spun at 500g for 10 min. at 4°C to collect nuclei. The nuclei were resuspended in 50 ml Transposase reaction (25 ml 2 Â Tagmentation buffer, 22.5 ml water, 2.5 ml Tn5 Transposase enzyme) and reacted for 30 min at 37°C on a PCR machine. The material was purified using a Qiagen MinElute protocol, eluting with 14 ml EB (Qiagen).
To amplify ATAC-seq libraries from the treated material, we amplified using the Ad1 primer below and a different Ad2 primer for each sample, which functions as a barcode The eluted material was amplified in 50 ml volume using 1.25 mM primer concentration and a 1 Â concentration NEBNext High-Fidelity Master-Mix (NEB) (programme: 72°C 5:00, 98°C 30 s, 5 Â (98°C 10 s, 63°C 30 s, 72°C 1 min), 4°C hold). After these five cycles of amplification, the tube was kept on ice.
A 5-ml aliquot was then removed and used to perform a 15-ml side reaction with identical concentrations of primer and enzyme as above, except that 0.6 Â SYBR Green (Invitrogen S-7563) is included to monitor amplification. This side reaction was amplified on a Stratagene Mx3005p qPCR (Agilent) system with the following amplification conditions (98°C 30 s, 20 Â (98°C 10 s, 63°C 30 s, 72°C 1:100)). The number of additional cycles 'N' required to reach one-fourth maximum fluorescence was observed. The purpose of this side reaction was to minimize the number of PCR cycles required used to generate the libraries, as length and GC bias increases with more amplification The remaining 45 ml of the reaction was then further amplified (98°C 30 s, N Â (98°C 10 s, 63°C 30 s, 72°C 1:00), 4°C) and the libraries were puried by a Qiagen MinElute kit eluting with 20 ml volume. Libraries were visualized by running on a 5% TBE gel and imaged by incubating for 20 min in 1 Â SYBR Green/1 Â TBE. Libraries quantified using the KAPA Library Quantification Kit (Kapa Biosystems).
Analysis on repeat families. Reads were then mapped to the mm9 genome allowing no mismatches and keeping reads that map up to 10,000 sites in the genome using Bowtie 46 . Each mapping read was assigned a score of 1/n, where n is the number of sites in the genome the read mapped to. Repeats were obtained from RepeatMasker. Expression values for each repeat family was calculated by adding the scores contained within the repeat body, divided by the total million reads mapped and average length (kb) of repeats within the family.
Analysis on individual genes and repeats. Reads were then mapped to known mm9 gene and repeat annotations by allowing up to two mismatches and only retaining reads that mapped to one location. When reads did not map to the annotated genes and repeats, the reads were mapped to the mm9 genome. Number of reads mapping to genes and repeats were determined by using HTSeq (doi: 10.1101/002824) using default parameters. Expression values were calculated as reads per kilobase of exons per million mapping reads. Differential gene and repeat expression was determined by using DESeq 47 , by using default parameters.
Whole-genome bisulfite sequencing. Reads were split into 50 bp reads before mapping. Reads were mapped to the mm9 genome as well as the lambda genome using BS seeker2 (ref. 48) using default parameters. Methylation levels were determined by #C/(#C þ #T). For identifying DMRs, the genome was tiled into 500 bp bins and CG methylation levels in knockout and control were compared within each bin. Bins that had a methylation level difference of 50% as well as a false discovery rateo0.05 calculated by Fisher's exact test corrected by the Benjamini-Hochberg procedure were selected. Finally, DMRs containing at least four cytosines in CG contexts, each covered by at least four reads were retained. Control regions were defined completely randomly, except that control regions have (1) exact same coverage of cytosines in CG contexts as the Morc1 þ /tg data within DMRs; (2) WT CG methylation levels are similar as the Morc1 þ /tg data within DMRs (o5%). (3) same number of regions per chromosome as DMRs. We defined genes as associated with DMRs when the TSS of an Ensembl transcript model was within 1 kb of a DMR.
Small RNA sequencing. Sequence adaptors were removed using a customdesigned dynamic programming algorithm that recognizes both exact and inexact matches, and the trimmed reads were aligned to the mm9 genome following a custom suffix array-based procedure 50 . Reads with lengths 424 nt were considered for piRNA analysis. Based on alignment coordinates, the reads were annotated as derived from exons, introns, transposons and other repeats according to the genome annotation obtained via UCSC Genome Bioinformatics 51 . Reads that had multiple valid alignments were annotated based on ten alignments selected at random, and the majority annotation was assigned as the final annotation. In case of ties, annotation was picked based on a fixed hierarchy principle 50 . Sense or antisense annotation was assigned to piRNA reads with respect to the strandedness of an underlying genomic feature. If a piRNA read contained U in position 1, such piRNA was considered as primary, while the presence of A in position 10 defined secondary piRNAs.
ChIP-seq analysis. Previously published ChIP-seq data 28 was obtained from GSE38165 in Gene Expression Omnibus. Reads were mapped to the mm9 genome using Bowtie by allowing up to two mismatches and only retaining reads that mapped to one location in the genome. Reads mapping to the same location were collapsed into one read. For all analyses, the data were normalized to total number of mapping reads in the library ATAC-seq analysis. Data were collected using 50 bp paired end sequencing on a HiSeq. In keeping with established methodologies 30 , reads were aligned to mm9 using Bowtie 52 with the parameters -X2000 and -m1. The -X2000 parameter allows the fragments o2 kb to align and only unique aligning reads were collected ( À m1). Duplicated reads were removed with samtools (rmdup function) 53 . Previous results show that for Tn5 transposase, the transposon binds as a dimer and insert two adaptors separated by 9 bp 54 . Thus, all reads aligned to the positive strands were offset by þ 4 bp and all reads aligned to the negative strands were offset by À 5 bp.