SPOCD1 is an essential executor of piRNA-directed de novo DNA methylation

Abstract

In mammals, the acquisition of the germline from the soma provides the germline with an essential challenge: the need to erase and reset genomic methylation¹. In the male germline, RNA-directed DNA methylation silences young, active transposable elements^2-4. The PIWI protein MIWI2 (PIWIL4) and its associated PIWI-interacting RNAs (piRNAs) instruct DNA methylation of transposable elements^3,5. piRNAs are proposed to tether MIWI2 to nascent transposable element transcripts; however, the mechanism by which MIWI2 directs the de novo methylation of transposable elements is poorly understood, although central to the immortality of the germline. Here we define the interactome of MIWI2 in mouse fetal gonocytes undergoing de novo genome methylation and identify a previously unknown MIWI2-associated factor, SPOCD1, that is essential for the methylation and silencing of young transposable elements. The loss of Spocd1 in mice results in male-specific infertility but does not affect either piRNA biogenesis or the localization of MIWI2 to the nucleus. SPOCD1 is a nuclear protein whose expression is restricted to the period of de novo genome methylation. It co-purifies in vivo with DNMT3L and DNMT3A, components of the de novo methylation machinery, as well as with constituents of the NURD and BAF chromatin remodelling complexes. We propose a model whereby tethering of MIWI2 to a nascent transposable element transcript recruits repressive chromatin remodelling activities and the de novo methylation apparatus through SPOCD1. In summary, we have identified a previously unrecognized and essential executor of mammalian piRNA-directed DNA methylation.

Extended Data Figure 1. Expression pattern and presence of nuclear localization signals for novel MIWI2 interactors.

a, b, Relative expression of indicated transcripts as measured by Affymetrix microarray in E16.5 gonocytes (n=2), adult spermatogonia (n=3), spermatocytes (n=3), MEFs (n=3) and bone marrow (n=2). Data are mean and s.e.m. NLS indicates presence of a nuclear localization signal as predicted by cNLS mapper.

Extended Data Figure 2. Homology alignment of SPOCD1 SPOC and TFIIS-M domains.

a, Multiple sequence alignment of the SPOC domain from SPOCD1 with representative vertebrate sequences from PHF3, DIDO1 and SPEN orthologues. The numbering for mouse SPOCD1 is shown above. Secondary structure elements for the human SHARP SPOC domain (PDBid 1OW1, SHARP is the human SPEN orthologue) are shown below the sequence, with dark rectangles for alpha helices and lighter arrows for beta strands. b, Multiple sequence alignment of TFIIS-M domain of SPOCD1 with equivalent sequences from TFIIS, PHF3 and DIDO1. Secondary structure elements from human PHF3 (PDBid 2DME) and human TFIIS (PDBid 3NDQ) are shown below, using the same annotation as in (a). Sequences are coloured according to sequence identity.

Extended Data Figure 3. Phylogeny of SPOCD1.

Bayesian phylogeny of Spocd1 (blue) and its vertebrate paralogs Phf3 (red) and Dido1 (green) inferred from cDNA sequences. Posterior probabilities of splits are shown as node labels. Branch lengths measure the expected substitutions per site as indicated in the scale bar.

Extended Data Figure 4. Generation and characterisation of the Spocd1^null mouse allele.

a, Schematic representation of the Spocd1 locus and the encoded 1015 amino acids (aa) protein (transcript XM_017320505.1) as well as design of the sgRNA targeting Spocd1 exon 7, which harbours part of the TFIIS-M domain. b, Schematic representation and sequencing trace (lower) of the part of Spocd1^null exon 7 harbouring the mutation site. The mutated site, highlighted in red, contains 2 premature stop codons and causes a frame-shift. Sequencing was repeated with identical results on n=3 animals. c, Representative image of genotyping results for Spocd1^+/+ , Spocd1^+/- and Spocd1^-/- animals. Similar results were obtained for all animals of the Spocd1^- line. d, Number of E16.5 embryos per plug from matings of mice with the indicated Spocd1 genotypes are presented. Mean and s.e.m. from n=7 Spocd1^+/+ dams mated to n=3 Spocd1^+/+ studs and n=12 Spocd1^-/- dams mated to n=5 Spocd1^+/- studs is plotted. NS, non-significant difference (P~0.98), two-tailed Student’s t-test. e, Representative PAS and haematoxylin stained histological testis sections of different stages of the seminiferous cycle are shown of (n=3) Spocd1^+/+ and Spocd1^-/- animals, indicating a germ cell differentiation arrest at the early pachytene stage. Scale bar, 5 μm. eP, early pachytene; RS, round spermatids, eS(13) elongating spermatids (step 13); PL, pre-leptotene; P, pachytene; L, leptotene; Z, zygotene; m2, secondary meiocytes. f, Representative images of zygotene spermatocytes in wildtype and Spocd1^-/- adult testis sections stained for the synaptonemal complex proteins SCP1 (red) and SCP3 (green). DNA stained with DAPI (blue). Scale bar, 1 μm. The representative images presented in panels e and f are from n=3 mice per genotype. g, h Analysis of TE expression in P20 testes from n=3 wildtype, Spocd1^-/- and Miwi2^-/- mice by RNA-seq. g, Comparison of TE expression in Miwi2^-/- and wildtype testes is shown. TEs with a significantly different (P<0.01, Benjamini-Hochberg adjusted two-sided Wald’s test) change in expression (>2-fold) are highlighted in red and the top 12 most up-regulated TEs in Miwi2^-/- testes are labelled. h, Comparison of TE expression in Spocd1^-/- and wildtype testes is shown. TEs with a significantly different (P<0.01, Benjamini-Hochberg adjusted two-sided Wald’s test) change in expression (>2-fold) are highlighted in red and same TEs as in (a) are labelled. i, Comparison of TE expression in Spocd1^-/- and Miwi2^-/- testes is shown. TEs with a significantly different (P<0.01, Benjamini-Hochberg adjusted two-sided Wald’s test) change in expression (>2-fold) are highlighted in red. TEs which are significantly up-regulated in Miwi2^-/- relative to wildtype are highlighted in black.

Extended Data Figure 5. CpG Methylation analysis of different genomic features and TE families.

Analysis of genomic CpG methylation of undifferentiated P14 spermatogonia from (n=3) wildtype, Spocd1 ^-/- and Miwi2 ^-/- mice is presented. a, b, Scatter plots comparing CpG methylation levels for the respective genomic features between wildtype and Spocd1^-/- or Miwi2^-/- (a) and between Spocd1^-/- or Miwi2^-/- spermatogonia (b) are shown. c, d, Scatter plots comparing CpG methylation levels for the respective TE families between wildtype and Spocd1^-/- or Miwi2^-/- (c) and between Spocd1^-/- or Miwi2^-/- spermatogonia (d) are shown. Data is mean from n=3 biological replicates per genotype and shown as individual data points (grey) overlayed by a density map.

Extended Data Figure 6. Methylation analysis of TE families.

Analysis of genomic CpG methylation of undifferentiated P14 spermatogonia from (n=3) wildtype, Spocd1 ^-/- and Miwi2 ^-/- mice is presented. a, Metaplots of CpG methylation over L1Md_A, IAPEy and MMERVK10C elements and adjacent 2 kb are shown. Schematic representation of the element is shown below. b, Metaplots of mean CpG methylation over LINE1 elements and adjacent 1 kb are shown. The Methyl-seq datasets of P14 wildtype, Miwi2^-/- and Spocd1^-/- spermatogonia are compared to WGBS datasets of adult Mili^-/- spermatocytes (Molaro et al. 2014) and P10 Dnmt3c^+/-, Dnmt3c^-/-, Dnmt3l^+/-, Dnmt3l^-/- germ cells (Barau et al. 2016). Schematic representation of LINE1 is shown below. c, Correlation analysis of mean CpG methylation loss relative to wildtype over individual elements of the indicated TE family in relation to their divergence from the consensus sequence is shown for Miwi2^-/- and Spocd1^-/- spermatogonia.

Extended Data Figure 7. piRNA analysis.

piRNA analyses of small RNAs sequenced from E16.5 testis from (n=3) Spocd1^+/- and Spocd1^-/- mice are presented. a, Relative frequency of piRNAs mapping to LINE1 and IAP families from Spocd1^+/- and Spocd1^-/- E16.5 testes. Plots are shown for all piRNA or anti-sense piRNAs. Data are mean and s.e.m. Adjusted P-values are listed, P=1.0 values are denoted as NS (Bonferroni adjusted two-sided Student’s t-test). b, Scatter plots showing mean expression of all (n=124411) piRNAs. The identity line is shown in red. r, Pearson’s correlation coefficient. c, Nucleotide features of piRNA from Spocd1^+/- and Spocd1^-/- E16.5 testes. Frequency of mapped piRNAs with a U at position 1 (1U) and with an A at position 10 (10A) are shown for L1Md_T elements. Data represent the mean and s.e.m. Adjusted P-values are shown (Bonferroni adjusted two-sided Student’s t-test) d, Ping-pong analysis of piRNAs from Spocd1^+/- and Spocd1^-/- E16.5 testis. Relative frequencies of the distances between 5’ ends of complementary piRNAs are shown for the indicated LINE1 and IAP families. e, Nucleotide features of piRNA from Spocd1^+/- and Spocd1^-/- E16.5 testis. Relative frequencies of piRNAs with a U at position 1 (1U) and with an A at position 10 (10A) are shown for respective elements shown in (d). Data are mean and s.e.m. Adjusted P-values are listed, P=1.0 values denoted as NS (Bonferroni adjusted two-sided Student’s t-test) f, Positions of piRNAs mapped to the consensus sequence of L1Md_T. Positive and negative values indicate sense and antisense piRNAs, respectively. Schematic representation of L1Md_T is shown above.

Extended Data Figure 8. TE and gene expression in Spocd1^-/- gonocytes.

Analysis of TE and gene expression in E16.5 Spocd1^+/- and Spocd1^-/- gonocytes by RNA-seq from n=3 mice per genotype. a, Comparison of TE expression in Spocd1^+/- and Spocd1^-/- gonocytes is shown. TEs up-regulated in Miwi2^-/- testes at P20 are highlighted in black. b, Comparison of gene expression in Spocd1^+/- and Spocd1^-/- gonocytes is shown. Significantly expressed genes (P<0.01, Benjamini-Hochberg adjusted two-sided Wald’s test, >2-fold change) are highlighted in red.

Extended Data Figure 9. Generation of the Spocd1^HA mouse allele.

a, Schematic representation of the SPOCD1 protein and Spocd1 locus as well as design of the sgRNA targeting the 3’ UTR near the translation termination site on Spocd1 exon 15. The Spocd1^HA allele encodes for a carboxy-terminal GGGGS linker followed by the HA epitope tag. The protospacer adjacent motif (PAM) site was mutated to inhibit re-targeting of the Spocd1^HA allele by the sgRNA-CAS9 complex. All inserted nucleotides and corresponding encoded amino acids are highlighted in red. The SPOCD1-HA protein is shown as a schematic representation. b, Schematic representation of the targeting strategy to generate the Spocd1^HA allele with a short single stranded DNA oligo donor (ssODN) of 200 nucleotides containing 5’ and 3’ homology arms (5’HA and 3’HA) of 72 nucleotides. c, Representative image of genotyping results for Spocd1^+/+ , Spocd1^HA/+ and Spocd1^HA/HA animals. Similar results were obtained for all animals of the Spocd1^HA line. d, Sequencing trace of part of a PCR amplicon of the HA epitope tag insertion site from a Spocd1^HA/HA animal. The experiment was repeated with identical results on n=2 animals. e, f, g, Representative images of wildtype (e), Spocd1^HA/+ (f) and Miwi2^HA/+ (g) testis sections at the indicated developmental time point probed with anti-HA antibody in green. DNA stained with DAPI in blue. Scale bars, 10 μm. The representative images presented in panels e to g are from experiments done n=3 mice as biological replicates with similar results.

Extended Data Figure 10. Co-immunoprecipitation experiments of SPOCD1 and DNMT3A/L/C in HEK cells.

Western blot analysis of co-immunoprecipitation of SPOCD1-HA with DNMT3L-FLAG, DNMT3A-FLAG, DNMT3C-FLAG or GFP in HEK cells. Shown are lysate sample (L), control IP (protein G beads) (B) and anti-HA IP (IP) for 4 experiments. For uncropped source data, see Supplementary Figure 1.

Figure 1. Definition of the MIWI2 interactome and identification of SPOCD1 from gonocytes undergoing de novo genome methylation.

a, Volcano plot showing enrichment (log2(mean LFQ ratio of HA-MIWI2 IP/control IP from Miwi2^+/+ foetal testis) and confidence (-log10(P-value of two-sided Student’s t-test)) of proteins co-purifying with HA-MIWI2 from E16.5 testis lysates (n=3). Dotted line indicates factors with enrichment >4-fold and significance P<0.05. Red: Known piRNA pathway members, blue: SPOCD1. b, List of known piRNA biogenesis factors and non-piRNA pathway-associated proteins co-purifying with HA-MIWI2. Novel identified MIWI2 interactors are underlined. c, Relative expression of indicated transcripts as measured by Affymetrix microarray in E16.5 gonocytes (n=2), adult spermatogonia (n=3) and spermatocytes (n=3). Data are mean and s.e.m., normalized to peak expression of each transcript. d, Schematic representation of SPOCD1 domain structure.

Figure 2. SPOCD1 is required for spermatogenesis and LINE1/IAP silencing.

a, Number of E16.5 embryos per plug of studs with the indicated Spocd1 genotype mated to wildtype females are presented. Data are mean and s.e.m. from n=3 Spocd1^+/+ studs (9 plugs total) and n=4 Spocd1^-/- studs (10 plugs total). **P~0.001, two-sided Student’s t-test. b, Representative images of PAS & Haematoxylin stained epididymis sections from (n=3) adult mice with the indicated genotype are shown. Scale bars, 20 μm. c, Average testicular weight in mg from adult mice with the indicated Spocd1 genotype is plotted. Insert shows a representative image of wildtype (left) and Spocd1^-/- testes. Data are mean and s.e.m. from n=3 wildtype and n=5 Spocd1^-/- mice. **P~0.01, two-sided Student’s t-test. d, Representative PAS & Haematoxylin stained testis sections of (n=3) adult mice of the indicated Spocd1 genotype is shown. Scale bars, 50 μm. e, f, Representative images of testis sections from (n=3) adult wildtype and Spocd1^-/- mice stained for LINE1 ORF1p (e) or IAP-GAG protein (f) (red) are shown. DNA was stained with DAPI (blue). Scale bars, 50 μm. g, RNA-seq derived heat maps depicting fold-change of expression relative to wildtype for the 10 most up-regulated LINE and ERVK TEs in (n=3) Miwi2^-/- and Spocd1^-/- P20 testis. h, i, Representative images of testis sections of (n=3) adult wildtype and Spocd1^-/- mice stained for the DNA damage response marker γH2AX (h) and TUNEL staining revealing apoptotic cells (i) (red). DNA was stained with DAPI (blue). Scale bars, 50 μm.

Figure 3. SPOCD1 is required for de novo TE DNA methylation loci but not piRNA expression.

a-e, Analyses of genomic CpG methylation of undifferentiated P14 spermatogonia from (n=3) wildtype, Spocd1 ^-/- and Miwi2 ^-/- mice are presented. a, b, Percentages of CpG methylation levels of the indicated genomic features (with genic, promoter and CpG islands non-overlapping TEs and intergenic non-overlapping TEs or genes) or TEs (non-overlapping genes) for (n=3) biological replicates per genotype is shown as box plots. Boxes represent interquantile range from 25^th to 75^th percentile, the horizontal line the median, whiskers denote the data range of median ± 2x interquantile range and dots datapoints outside of this data range. c, Metaplots of mean CpG methylation over LINE1 elements and adjacent 2 kb are shown. d, Correlation analysis of mean CpG methylation loss relative to wildtype for individual TEs of the indicated LINE1 family in relation to their divergence from the consensus sequence is shown for Spocd1^-/- spermatogonia. e, Heatmap of mean CpG methylation level of indicated maternal and paternal imprinted regions is shown. Rasgrf1 imprinted control region is shown in detail. f-h, piRNA analysis of small RNAs sequenced from E16.5 testes from (n=3) Spocd1^+/- and Spocd1^-/- mice is presented. f, Nucleotide (nt) length distribution of small RNAs is shown. Data represent the mean and s.e.m. No significant differences were observed (P=1.0, Bonferroni adjusted two-tailed Student’s t-tests). g, Annotation of piRNAs from merged replicates. h, Ping-pong analysis of piRNAs: Relative frequency of the distance between 5’ ends of complementary piRNAs mapping to the LINE1 L1Md_T family is shown. i, Representative images (of n=3 wildtype and Spocd1^-/- mice) of MIWI2 localization in E16.5 Spocd1^+/- and Spocd1^-/- gonocytes. Scale bars, 15 μm. Insert shows a zoom in of the indicated cell. Scale bars, 2 μm.

Figure 4. SPOCD1 is a nuclear protein that associates with the de novo DNA methylation machinery and repressive chromatin remodelling complexes.

a, Schematic representation of the Spocd1^HA allele and the C-terminal HA-tagged SPOCD1 protein. b, Representative image of SPOCD1-HA localization in gonocytes at E16.5 from (n=3) Spocd1^HA/+ mice. Scale bar, 20 μm. Insert shows a zoom in of the indicated foetal gonocyte. Scale bar, 2 μm. c, d, Representative images of expression of SPOCD1-HA (c) and HA-MIWI2 (d) in gonocytes at the indicated time points from (n=3) Spocd1^HA/+ and Miwi2 ^HA/+ mice, respectively, are shown. Scale bars, 2 μm. e, Volcano plot showing enrichment (log2(mean LFQ ratio of SPOCD1-HA IP/control IP from Spocd1^+/+ foetal testis) and statistical confidence (-log10(P-value of two-sided Student’s t-test)) of proteins co-purifying with SPOCD1-HA from E16.5 testis lysates (n=4). Dotted line indicates enrichment >4-fold and significance P<0.05. DNMT3L and DNMT3A (green), members of the NURD (violet) and BAF (blue) complexes are highlighted. f, Schematic representation of selected proteins co-purifying with SPOCD1. g, Volcano plot as presented in panel e of proteins co-purifying with MIWI2-HA from E16.5 testis Benzonase-solubilized extracts (n=4). h, Schematic representation of overlap of proteins co-purifying with both SPOCD1 and MIWI2.