Morc1 reestablishes H3K9me3 heterochromatin on piRNA-targeted transposons in gonocytes

Significance

Transposable elements (TEs) constitute a major threat to the genome integrity of the germline; therefore, the host has several mechanisms to silence them. In mouse embryonic germ cells, called gonocytes, TEs are transiently activated and resilenced by as-yet-unknown mechanisms. Here, we demonstrate that a key epigenetic mark, H3K9me3, and its regulator, Morc1, preserve genome integrity by repressing TEs in gonocytes. Morc1 specifically targets the activated TEs with the aid of nuclear PIWI protein containing PIWI-interacting RNAs (piRNAs) that has a complementary sequence against TEs, suggesting a highly ordered regulatory system for germline TEs, in which transcription occurs first, followed by the induction of small RNA-mediated heterochromatin formation. Thus, our study provides insights into how TEs are regulated in mammalian germ cells.

Abstract

To maintain fertility, male mice re-repress transposable elements (TEs) that were de-silenced in the early gonocytes before their differentiation into spermatogonia. However, the mechanism of TE silencing re-establishment remains unknown. Here, we found that the DNA-binding protein Morc1, in cooperation with the methyltransferase SetDB1, deposits the repressive histone mark H3K9me3 on a large fraction of activated TEs, leading to heterochromatin. Morc1 also triggers DNA methylation, but TEs targeted by Morc1-driven DNA methylation only slightly overlapped with those repressed by Morc1/SetDB1-dependent heterochromatin formation, suggesting that Morc1 silences TEs in two different manners. In contrast, TEs regulated by Morc1 and Miwi2, the nuclear PIWI-family protein, almost overlapped. Miwi2 binds to PIWI-interacting RNAs (piRNAs) that base-pair with TE mRNAs via sequence complementarity, while Morc1 DNA binding is not sequence specific, suggesting that Miwi2 selects its targets, and then, Morc1 acts to repress them with cofactors. A high-ordered mechanism of TE repression in gonocytes has been identified.

Heterochromatin has a crucial role in repressing transposable elements (TEs) and other repeats that cause genome instability in eukaryotes (1). In mammalian germ cells, deregulation of TEs leads to meiotic arrest, and ultimately an infertile phenotype, forcing the host side to equip a machinery to deposit DNA methylation (DNAme) for silencing (2, 3). In contrast to somatic cells, genome-wide DNAme levels vary widely along with the germline development (4). This indicates that the expression of TEs would also fluctuate in germ cells; however, much of such dynamics remain unclear.

Mouse gonocytes/prospermatogonia are male germ cells from embryonic day (E) 13.5 to postnatal day (P) 3, representing an essential intermediate between primordial germ cells (PGCs) and spermatogonia in germline development. In developing PGCs, the level of DNAme on the whole chromosome gradually decreases, a phenomenon called epigenetic reprogramming (5, 6). This makes the level of DNAme in gonocytes at E13.5 the lowest in the life cycle of mice (7–10). Despite the absence of DNAme, most TEs are not derepressed in the germ cells at that time (11, 12). Interestingly, at E16.5, a few days later, over 10,000 copies of TEs are transcriptionally activated (12). This temporal discrepancy between DNAme erasure and TE activation indicates that the activity of TEs cannot be explained by DNAme alone in gonocytes. These active TEs are enriched in specific genomic regions, called differentially accessible domains (DADs) (12). The activation of TEs at E16.5 is considered to trigger the de novo DNAme during gonocytes (12). On the other hand, however, the sustained active state of such TEs beyond the gonocyte stage leads to defects in meiotic prophase, particularly double-strand break formation (13). All these situations suggest that some machineries reestablish the transcriptional repression of activated TEs, of which mechanisms may include deposition of repressive histone modification in addition to DNAme.

Two types of mechanisms are known to specifically suppress TEs: one that recognizes their DNA sequences and the other that recognizes RNA sequences. The KRAB-ZFPs and PIWI-piRNA constitute the former and the latter, respectively. In mouse embryonic stem cells, KRAB-ZFPs recognize specific DNA sequences of TEs and then recruit SetDB1, a methyltransferase for H3K9, with the assistance of Trim28 (14–18). PIWI proteins and their associated small noncoding PIWI-interacting RNAs (piRNAs) form the piRNA-induced silencing complexes (piRISCs) (19–21). Mice have three PIWIs (Miwi, Mili, and Miwi2), of which Mili and Miwi2 are expressed in gonocytes. Mili degrades cytoplasmic TE transcripts through piRNA-guided endonucleolytic cleavage, producing piRNAs with a complementary sequence to TEs (22, 23). The resulting piRNAs guide nuclear PIWI protein Miwi2 to active TE loci by tethering the piRISC to the nascent transcript and trigger DNAme (24–26), suggesting that piRNAs act as a determinant of target specificity (27). Thus, the host triggers enrichment of multiple epigenetic marks, such as DNAme and H3K9me3, depending on the biological context to ensure complete suppression of TEs in a sequence-specific manner. In gonocytes, however, it remains elusive which epigenetic mark is responsible for both the activation and resilencing of TEs. In addition, it is also unclear how the multiple silencing mechanisms relate to each other and what types of TEs are preferentially targeted by each mechanism.

In this paper, we found that around 14,000 TEs are activated in the mid-gonocyte stage together with the loss of H3K9me3. After a few days, about half of them returned to transcriptionally silent state with the accumulation of H3K9me3. This suggests that H3K9me3-marked heterochromatin governs the activity of TEs in gonocytes. Further bioinformatic analysis suggested that the enrichment of H3K9me3 on once-activated TEs was dependent on Morc1, which is a GHKL ATPase that is widely conserved in prokaryotes and eukaryotes (28, 29). Similar to Miwi2, Morc1 is known to be involved in TE silencing via DNAme (30), although how Morc1 recognizes TEs and triggers their repression remains largely unclear. The types of TEs that accumulate DNAme in a Morc1-dependent manner only slightly overlap with those with H3K9me3, implying that Morc1 uses two ways of suppression for TEs. Comparison of these Morc1-regulated TEs with those under the control of Miwi2 showed that both TEs considerably overlap, suggesting that PIWI-piRNA machinery guides Morc1 to specifically target TEs. In sum, our data revealed a highly ordered regulatory system for TEs, in which transcription occurs first, followed by the induction of heterochromatin formation. Through this sequence of events, the TE-derived mutagenic transcripts are repurposed to trigger Miwi2/Morc1-dependent heterochromatin formation.

Results

Group I TEs Activated during the Gonocyte Stage Return to a Transcriptionally Silent State by Accumulation of H3K9me3.

We previously reported that >10,000 copies of TEs, such as long interspersed nuclear element 1 (LINE-1) and long terminal repeat (LTR) retrotransposons, were derepressed for a few days in gonocytes (12), while other TEs, including their remnants, remain inactive (Fig. 1A). The activated TEs return to inactive state in the following developmental stage, spermatogonia (31), indicating the existence of specialized machinery that resilences them. With ATAC-seq analysis, we found 14,218 TEs assumed more accessible chromatin at E16.5 than E13.5. We termed these activated TEs as DAD TEs since more than 60% of them tended to reside in DADs (12). The DAD TEs produced more transcripts at E16.5 than E13.5 (Fig. 1B); then, such transcripts were reduced at P0, suggesting that the DAD TEs are transcribed in the middle of the gonocyte stage (Fig. 1B).

Fig. 1.

Resilencing of TEs in gonocytes coincides with enrichment of H3K9me3. (A) Schematic of TE dynamics during the gonocyte stage. TEs are indicated by squares with rounded corners. The color inside the squares denotes the degree of TE silencing. A dark color indicates a more repressive chromatin state on each TE. (B) Dynamics of the transcript abundance of TEs with higher chromatin accessibility at E16.5 than E13.5 during the gonocyte stage. (C) K-mer clustering of DAD TEs based on their dynamics of chromatin accessibility over the TSS region. The number of clusters supplied as input was 4. (D) Genome browser views of representative ATAC-seq peaks categorized to each group. An asterisk marks peaks identified by Homer using ATAC-seq data at E16.5. (E) Differential transcript levels of TEs for each group in gonocytes between E16.5 and P0. The Mann–Whitney U test was used for statistical hypothesis testing. *P < 0.001. (F) Line plots of the average level of either H3K4me3 or H3K9me3 around the TSS of TEs for each group. Regions surrounding 5 kb upstream and downstream from peaks are depicted.

Clustering analysis based on the temporal change of chromatin accessibility grouped the DAD TEs into four clusters (Fig. 1C). DAD TEs in clusters 1 and 2 returned to closed chromatin at P0 in a similar fashion (Fig. 1 C and D), although the extent to which the accessibility reverted to the closed state at P0 was more prominent for TEs in cluster 1. Cluster 3 TEs showed more open chromatin at P0 than earlier stages (Fig. 1 C and D). Accessibility over cluster 4 TEs at P0 was overall the same level with that at E16.5 (Fig. 1 C and D). Because the kinetics of chromatin accessibility was similar in clusters 1 and 2, we categorized them together as group I TEs (6,852; 48.2% of DAD TEs). Hereafter, we termed TEs in clusters 3 and 4 as group II and III TEs, respectively. In line with the dynamics of chromatin accessibility, the levels of transcripts from group I TEs were more reduced compared with those from group II and III TEs from E16.5 to P0 (Fig. 1E). These data suggest that repressive chromatin was reestablished on group I TEs at P0. With public datasets, we found that H3K4me3 and H3K9me3 levels were lower and higher on group I TEs, respectively, than the other two groups at P0 (Fig. 1F). For group I TEs, the special machinery may specifically cause H3K9me3-dependent heterochromatin formation.

Morc1 Triggers Heterochromatin Formation by Reestablishing H3K9me3 around Group I TE Genomic Regions.

To elucidate the silencing pathway for group I TEs, we focused on Morc1. Morc1 was previously shown to mediate DNAme for TEs (30). Since there has been little attempt to identify physical interaction partners of mouse Morc1, it is unclear whether it directly triggers de novo DNAme on TEs. To reveal how Morc1 leads to transcriptional silencing of TEs, we performed ATAC-seq analysis at three time points during the gonocyte stage, E13.5, E16.5, and P0, using Morc1 homozygous (Morc1 KO) and Morc1 heterozygous (Morc1 het) knockout mice. While there was no clear difference between Morc1 KO and Morc1 het gonocytes until E16.5, numerous peaks in Morc1 KO gonocytes showed higher chromatin accessibility at P0 (Fig. 2A). We defined 3,537 peaks that assumed open chromatin in Morc1 KO gonocytes at P0 (Fig. 2B), 96.9% of which overlapped with TEs such as L1 and LTR retrotransposons (Fig. 2C). These results support the idea that Morc1 triggers chromatin compaction of TEs at the later stage of gonocytes. Because H3K9me3 tended to show a negative correlation with TE activity in gonocytes (Fig. 1F), we performed H3K9me3 ChIP-seq at each time point in Morc1 KO and Morc1 het gonocytes. ChIP-seq reads over 8,491 peaks were significantly reduced in Morc1 KO gonocytes at P0 (Fig. 2 D and E), 95.1% of which were on L1 or LTR retrotransposons (Fig. 2F). ATAC-seq peaks with higher chromatin accessibility in Morc1 KO gonocytes showed a lower signal for H3K9me3 than Morc1 het gonocytes (Fig. 2G). This supports the notion that Morc1 plays a role in chromatin compaction via deposition of repressive histone mark on TEs.

Fig. 2.

Morc1 triggers the formation of closed chromatin together with enrichment of H3K9me3 over TSSs of group I TEs. (A) Scatter plots of the normalized number of reads mapped to ATAC-seq peaks for Morc1 het and Morc1 KO gonocytes. The color of each dot indicates the range of the ratio between the two strains. Red, light red, gray, light blue, and blue indicate greater than 4, greater than 2 but no more than 4, greater than −2 but no more than 2, greater than −4 but no more than −2, and no more than −4, respectively. (B) Bar plot of the numbers of peaks that showed differential chromatin accessibility in Morc1 KO gonocytes identified by MAnorm. Red and blue bars indicate the number of peaks up-regulated and down-regulated in Morc1 KO gonocytes, respectively. (C) Pie chart of the annotation of ATAC-seq peaks with open chromatin in Morc1 KO gonocytes at P0 showing that 96.9% of them overlapped with LINE-1 or LTR-type retrotransposons. (D) Scatter plots of the normalized number of reads mapped to peaks of H3K9me3 for Morc1 het and Morc1 KO gonocytes. The value was normalized to the abundance of H3. The color of each dot indicates the range of the ratio between the two strains. Red, light red, gray, light blue, and blue indicate greater than 4, greater than 2 but no more than 4, greater than −2 but no more than 2, greater than −4 but no more than −2, and no more than −4, respectively. (E) Bar plot of the numbers of peaks that showed a differential level of H3K9me3 in Morc1 KO gonocytes identified by MAnorm. Red and blue bars indicate the number of peaks up-regulated and down-regulated in Morc1 KO gonocytes, respectively. (F) Pie chart of the annotation of ChIP-seq peaks with less H3K9me3 signals in Morc1 KO gonocytes at P0 showing 95.1% of them overlapped with LINE-1 or LTR-type retrotransposons. (G) Metaplot and heatmap plot analysis of the abundance of H3K9me3 surrounding ATAC-seq peaks that showed significantly higher accessibility in Morc1 KO gonocytes compared with Morc1 het gonocytes at P0. Regions spanning 5 kb upstream and downstream from peak regions are depicted. (H) Temporal transition of H3K9me3 abundance around TSSs of TEs in each group. The H3K9me3 level on group I TEs was more recovered at P0 than that of the other two groups. (I) Expression profile of Morc1 during the gonocyte stage. (J) ATAC-seq and H3K9me3 profiles of Morc1 het and Morc1 KO gonocytes at E16.5 and P0. The positions of LINE-1 and LTR-type retrotransposons from the RepeatMasker database are shown at the bottom. Gray boxes indicate regions showing a difference between Morc1 het and Morc1 KO gonocytes. (K) Pie chart showing the percentages of TEs repressed by Morc1 among TEs in each TE group. (L) As described in (G), but the depicted genomic coordinates were TSSs of group I TEs. (M) MAplot of RNA-seq reads mapped to TEs in Morc1 KO and Morc1 het gonocytes. Blue dots indicate individual TEs that were differentially expressed between Morc1 KO and Morc1 het gonocytes (FDR < 0.05). (N) Number of TE copies that were significantly up-regulated or down-regulated in Morc1 KO gonocytes compared with Morc1 het gonocytes (log2 FC > 1 or log2 FC < 1, FDR < 0.05). (O) Boxplots showing the abundance of transcripts from group I TEs in Morc1 KO and Morc1 het gonocytes. A t test was applied for statistical hypothesis testing. *P < 0.001.

All three groups of DAD TEs defined above (Fig. 1 C and D) were enriched with H3K9me3 at E13.5 and then lost H3K9me3 from their genomic region at E16.5 upon activation (Fig. 2H). Because H3K9me3-marked heterochromatin was reestablished more selectively on group I TEs after Morc1 began to express (Fig. 2 H and I), we suspected that Morc1 was involved in heterochromatin formation on group I TEs. Visual inspection of some representative genomic insertion sites of group I TEs revealed that their promoter regions lost H3K9me3 and adopted open chromatin in Morc1 KO gonocytes (Fig. 2J). Group I TEs included the largest number of Morc1 targets among the three TE groups (Fig. 2K). Importantly, reestablishment of H3K9me3 on group I TEs was severely affected in Morc1 KO gonocytes (Fig. 2L). RNA-seq analysis revealed that 9,152 TEs were derepressed in Morc1 KO gonocytes at P0 (Fig. 2 M and N), in which the transcripts from group I TEs were more abundant in the mutant (Fig. 2O). These data support the idea that Morc1 suppresses group I TE expression preferentially over other TE groups by restabilizing heterochromatin in gonocytes from E16.5 to P0.

SetDB1 Deposits H3K9me3 on Morc1-Dependent Group I TEs in Gonocytes.

Mice express four methyltransferases that target H3K9 residues, SetDB1, Suv39h1, G9a, and GLP (32–37). Suv39h1 is responsible for maintenance of H3K9me3 on centromeric repeats (38, 39). G9a and GLP are involved in converting unmethylated H3K9 to H3K9me2, but not to H3K9me3 (40). In contrast to these proteins, SetDB1 mediates silencing of TEs through H3K9me3-marked heterochromatin formation in some biological contexts, including mouse embryonic stem cells (mESCs) (16, 18). Therefore, we hypothesized that this enzyme deposited H3K9me3 on Morc1-dependent TEs in gonocytes. To test this hypothesis, we applied an organ culture method to embryonic testes (41), as this method enabled evaluation of the effect of a SetDB1 inhibitor (Fig. 3A). We chose three genomic sites around TSSs of specific TEs as representative loci targeted by Morc1. Site-specific ChIP-qPCR confirmed that H3K9me3 had accumulated on all three TEs from E16.5 to P0 in vivo (Fig. 3B). We then examined whether heterochromatin formation on Morc1 targets was recapitulated by the organ culture. Testes extracted from E16.5 embryos were incubated in a culture dish for 3 d, and then, germ cells were applied to ChIP analysis (Fig. 3A). The data show that the H3K9me3 level was increased over Morc1-target TEs during the 3 d of in vitro culture (Fig. 3B). Moreover, adding SETDB1-TTD-IN-1, the SetDB1 inhibitor, led to poor heterochromatin formation on such TEs (Fig. 3B). Although our results may be indirect, since this compound affects both somatic and germline cells in the tissue, these findings support the possibility that SetDB1 is involved in H3K9me3 deposition on Morc1 targets.

Fig. 3.

SetDB1 deposits H3K9me3 onto Morc1-dependent TEs in gonocytes. (A) Scheme of gonocytes in vitro culture. Testes were excised from E16.5 embryos. The embryo harbored a heterothallic Ddx4-Venus transgene, so that germ cells could be isolated by fluorescence. (B) ChIP-qPCR analysis of H3K9me3. The names of four samples used were indicated in the box: E16.5, germ cells isolated from E16.5 testes; P0, germ cells isolated from P0 testis; with DMSO, germ cells isolated from testis cultured in vitro for 2 d with DMSO; with inhibitor, germ cells isolated from testes cultured in vitro for 4 d with a SetDB1 inhibitor. The promoter region of β-actin (actb) and the pericentromeric region (cen.) were used as negative and positive controls, respectively. Three TE regions tested were among Morc1-dependent TEs. A t test was applied for statistical hypothesis testing. *P < 0.05.

Morc1 Deposits Two Distinct Epigenetic Marks to Suppress Different Types of TEs.

Morc1 triggers accumulation of DNAme on TEs (30). In this study, we decided to compare the families/subfamilies of TEs that gained DNAme [Morc1-dependent TEs (DNAme) or MdTE (DNAme)] (30) with those that gained H3K9me3 [Morc1-dependent TEs (K9me3) or MdTE (K9me3)]. This revealed that 1,026 TEs out of 8,491 MdTE (K9me3) and 6,302 MdTE (DNAme) were shared between the two groups (Fig. 4A). The remaining TEs belonged to one of the groups.

Fig. 4.

Certain subclasses of TEs acquire H3K9me3 in a Morc1-dependent manner accompanied by little or opposite effect on DNA methylation. (A) Venn diagram showing overlaps between peaks that lost H3K9me3 in Morc1 KO gonocytes and differentially methylated regions (DMRs) defined in a previous study (30). (B) Families/subfamilies of TEs were sorted by the extent to which certain TEs were enriched in MdTE (K9me3) (Methods). Enrichment values are also shown for MdTE (DNAme) on the right and sorted by the order of MdTE (K9me3) on the left. (C) Top 20 TE families/subfamilies of MdTE (K9me) and MdTE (DNAme). TEs in red are not shared between the two lists. Values shown were calculated as described in (B). (D) Subtracted difference of the DNAme level between Morc1 het and Morc1 KO gonocytes at MdTE (K9me3). The number of TEs with a value under 50% and 25% are shown. (E) Line plots showing the average level of H3K9me3 (Top) and DNAme (Bottom) over regions around each TE family/subfamily in Morc1 KO and Morc1 het gonocytes. (F) Scatter plots showing normalized reads of TEs in three TE families/subfamilies in Morc1 KO and Morc1 het gonocytes. (G) Box plots of the ratio of chromatin accessibility between Morc1 KO and Morc1 het gonocytes for three TE categories: TEs that lose H3K9me3 and DNAme in Morc1 KO gonocytes, TEs that lose only H3K9me3 in Morc1 KO gonocytes, and TEs with no significant change in H3K9me3 or DNAme in Morc1 KO gonocytes.

A more detailed classification analysis of MdTE (K9me3) and MdTE (DNAme) was performed to determine which families/subfamilies of TEs were enriched as preferential targets. While the overall pattern of families/subfamilies between MdTE (K9me3) and MdTE (DNAme) appeared similar (Fig. 4B), there were some distinct differences between them. For example, L1MdA_I was a preferential target among MdTE (K9me3), while it was avoided as a target of MdTE (DNAme) (Fig. 4B). Such a difference was obvious when the top 20 families/subfamilies of each group were compared (Fig. 4C). Some subfamilies of L1MdA (L1MdA_I, II, and III) were included in the top 20 for MdTE (K9me3), but they were not in the counterpart of MdTE (DNAme) (Fig. 4C), suggesting that Morc1 deposits two different epigenetic marks on different genomic regions.

In support of this, 20,794 and 23,618 MdTE (K9me3) out of 26,005 were not included in previously annotated differentially methylated regions (DMRs) (30), when the threshold of the differential DNAme level between Morc1 KO and Morc1 het gonocytes was set at 25% and 50%, respectively (Fig. 4D). An average line plot of the epigenome marks over each family/subfamily also showed such a trend (Fig. 4E). H3K9me3 over promoter regions of L1MdA, L1MdTf, and L1MdGf were reduced in Morc1 KO gonocytes (Fig. 4E). In contrast, the degree of DNAme loss over L1MdA in Morc1 KO is little or very limited. For LTR retrotransposons, while both RLTR10 and MMERVK10Cint lost DNAme in Morc1 KO gonocytes, H3K9me3 was reduced only for RLTR10 (Fig. 4 C and E). These results support the notion that Morc1 suppressed L1MdA and MMERVK10Cint via either H3K9me3 or DNAme, respectively, while L1MdTf was repressed by Morc1 via both epigenomic marks. In terms of transcripts, numerous copies of the three families/subfamilies (L1MdA, L1MdTf, and MMERVK10Cint) were derepressed in Morc1 KO gonocytes, suggesting that both H3K9me3 and DNAme suppressed the transcription of these TEs (Fig. 4F). It is noted that unlike in the case of ChIP-seq and ATAC-seq, multiple mapping reads were used for RNA-seq analysis (Methods), so there may be uncertainty at each TE insertions site.

Chromatin accessibility was increased over specific MdTE (K9me3) that were not categorized as MdTE (DNAme) (Fig. 4G), supporting the notion that H3K9me3 alone exerted a repressive effect on TEs. The idea that Morc1 induces two different modes of suppression for different types of TEs was further supported.

Since some reports have shown a functional relationship between H3K9me3 and DNAme in mammalian cell lines (42, 43), we examined whether such interplay also occurred in Morc1-mediated TE silencing. Interestingly, among the 7,130 TE copies in which H3K9me3 on their TSS was not reduced but DNAme was affected in Morc1 KO gonocytes, there are a certain number of TEs that acquire DNAme in the process up to spermatogonial stem cells, and vice versa (SI Appendix, Fig. S1 A and B). These data suggest that MdTE (K9me) acquire DNAme and MdTE (DNAme) acquire H3K9me3 at the later stage of germline development possibly for their strict silencing.

Miwi2-piRISC Is Involved in Target TE Selection for Heterochromatin Formation by Morc1.

More than 90% of Morc1 targets were TEs (Fig. 2 C and F). Similar to Morc1 in worms (44), mouse Morc1 binds directly to DNA, although such binding did not show obvious sequence specificity (SI Appendix, Fig. S2 A and B). These data imply that other factor(s) conferred target specificity to Morc1. Miwi2 was previously shown to be involved in transcriptional silencing of TEs in a sequence-specific manner in gonocytes (24–26, 45), prompting us to test whether Miwi2 would help Morc1 to selectively suppress TEs. Among the three PIWI family proteins in mice, Miwi2 is the only nuclear protein expressed in gonocytes, whose expression pattern resembles with Morc1 (Fig. 5A). To compare TEs regulated by Morc1 with those regulated by Miwi2, we reanalyzed the H3K9me3 ChIP-seq data from Miwi2 KO germ cells (45), and found substantial overlap of ChIP-seq peaks targeted by either Morc1 or Miwi2 (Fig. 5B). The H3K9me3 level over Miwi2-dependent peaks was decreased in Morc1 KO gonocytes and vice versa (Fig. 5C). The effect of Miwi2 KO around genomic insertion sites of each family/subfamily was similar to that of Morc1 KO (Fig. 5D), although the distribution patterns of H3K9me3 over TEs were different, possibly because of the difference in ChIP-seq library preparation. piRNAs loaded on Miwi2 were rich in antisense sequences for their target TEs. Therefore, the abundance of mapped piRNAs over a specific TE family/subfamily are useful to measure the degree of dependency on the PIWI-piRNA pathway for its repression. This revealed that suppression of TEs targeted by Morc1 was as dependent on PIWI-piRNA pathway as that by Miwi2 (Fig. 5E). Such a trend was not observed for randomly selected TEs (Fig. 5E). Importantly, TEs targeted by Trim28, which mediates TE silencing via KRAB-ZFPs, showed significantly less dependency on the PIWI-piRNA pathway than the others (Fig. 5E). Additionally, more than half of DMRs in Miwi2 KO gonocytes showed severe loss of DNAme in Morc1 KO gonocytes (Fig. 5F). These data suggested that Morc1 and Miwi2 shared their target TEs and supported a model that Morc1 recognized its targets with the aid of the PIWI-piRNA pathway.

Fig. 5.

A substantial number of TEs targeted by Morc1 are preferentially repressed by Miwi2. (A) Temporal change of the abundance of Miwi2 transcripts during the gonocyte stage. Shaded gray indicates the pattern of Morc1 expression shown in Fig. 2H. (B) Venn diagram showing the overlaps between peaks that lost H3K9me3 in Morc1 KO and Miwi2 KO gonocytes (45). (C) Average plots showing the H3K9me3 level in Morc1 het and Morc1 KO gonocytes over H3K9me3 peaks dependent on Miwi2 (Left). Average plots showing the H3K9me3 level in Miwi2 het and Miwi2 KO gonocytes over H3K9me3 peaks dependent on Morc1 (Right). (D) Heat map showing the average level of H3K9me3 over regions containing the indicated TE families/subfamilies in Morc1 KO and Morc1 het gonocytes (Top), and Miwi2 KO and Miwi2 het gonocytes (Bottom). (E) Cumulative plots showing the distributions of piRNA reads mapped to each copy of the shown TE families. In addition to the target TEs of Morc1 (orange), Miwi2 (blue), and Trim28 (black), the plots of randomly selected TEs (gray) are also shown. The x-axis shows the number of reads mapped to each TE copy. The y-axis shows the cumulative fraction. The two-sided Kolmogorov–Smirnov test was applied for statistical hypothesis testing. *P < 0.001. (F) As described in Fig. 4D, the subtracted difference in the DNAme level between Morc1 het and Morc1 KO gonocytes at DMR in Miwi2 KO gonocytes.

Morc1 Represses the Expression of Host Genes without Inducing Heterochromatin Formation on Their Promoters.

To investigate the effect of Morc1 on genomic elements other than TEs in terms of target selectivity, we profiled changes in host gene expression of Morc1 KO gonocytes. More up-regulated than down-regulated genes were found in Morc1 KO gonocytes at P0 and P3 (Fig. 6A). There were 203 and 323 up-regulated genes at P0 and P3, respectively, among which 137 genes were shared between the two stages (Fig. 6 B and C). These results suggest that Morc1 regulate host gene expression as well as TEs. However, in contrast to the situation for TEs, we could barely see the higher accessibility in the mutant at TSS regions of such affected genes (Fig. 6D). These up-regulated genes were significantly closer to the Morc1-dependent TEs than randomly picked TEs (Fig. 6E). Genome browser view around the up-regulated genes revealed that the reads of RNA-seq were mapped to a part of their exons (faded red boxes in Fig. 6F) and not to their TSS regions. This indicates that instead of using the intrinsic promoter of the host gene, the neighboring TE (faded blue boxes in Fig. 6F) served as an alternative promoter. A time course RNA-seq dataset from wild-type gonocytes revealed that the Morc1-dependent 137 genes were up-regulated at E16.5 and then reduced at P0 in wild-type gonocytes (Fig. 6G), which was reminiscent of the dynamics of group I TEs (Figs. 1 and 2). These 137 genes encode proteins that may not have clear function in testes, such as Ranbp3l, which is highly expressed in kidneys, and Sult2a5, which is exclusively expressed in the liver (46). This indicates that Morc1 corrected the host gene transcriptome through suppression of nearby TEs that would otherwise disrupt the host gene expression network. Importantly, H3K9me3 was not established on the TSSs of such genes (Fig. 6H). Thus, the transient expression, per se, does not induce Morc1-dependent heterochromatin formation. In addition to such nascent transcript, TE-derived piRNA sequence would also constitute additional target determinant for the heterochromatin formation. In sum, our analysis revealed that heterochromatin on group I TEs is formed when some key events occur in an orderly fashion: first, the transcription at the TEs is activated; then, such transcripts work as a hub for downstream events, such as recruitment of the PIWI-piRNA complex; and finally, this recognition triggers Morc1-dependent heterochromatin formation in the later stage of gonocytes (Fig. 7).

Fig. 6.

Morc1 reshapes the host gene transcriptome through suppression of neighboring TEs. (A) MAplots of RNA-seq reads mapped to host genes in Morc1 KO and Morc1 het gonocytes at P0 and P3. Blue dots denote differentially expressed genes (DEG) (q < 0.001). (B) Bar plots of the number of DEGs at the indicated time points. (C) Venn diagram showing overlaps of up-regulated DEGs in Morc1 KO gonocytes between P0 and P3. (D) Line plots showing the average level of chromatin accessibility around TSS regions of up-regulated DEGs in Morc1 KO gonocytes at the indicated time points in Morc1 KO and Morc1 het gonocytes. (E) Box plots showing the distance on chromosome between MdTE (ATAC) and up-regulated genes (UP), and randomly selected genes (random). A t test was applied for statistical hypothesis testing. *P < 0.01. (F) Genome browser views of exons downstream from Morc1-dependent TEs inserted in the corresponding intron producing aberrant transcripts in Morc1 KO gonocytes. Asterisks indicate exons upstream from the same TEs. Note that such exons did not produce more transcripts in Morc1 KO gonocytes than in Morc1 het gonocytes. (G) Box plots showing the abundance of transcripts of up-regulated DEGs in Morc1 KO gonocytes using wild-type RNA-seq datasets at each developmental stage. A t test was applied for statistical hypothesis testing. *P < 0.001. (H) As described in (D), but the H3K9me3 level is plotted.

Fig. 7.

Model showing how TEs are regulated by H3K9me3, Morc1, Miwi2, and SetDB1 during the gonocyte stage. From E13.5 to E16.5, TEs lose H3K9me3 from their TSS regions, leading to their transcriptional activation together with accumulation of their transcripts. These TE-derived transcripts serve as a source of piRNAs that are loaded onto Miwi2. As gonocytes enter the P0 stage, activated TEs gain H3K9me3 and return to the transcriptionally repressed state. Morc1 and SetDB1 family proteins are involved in reestablishing heterochromatin on the TEs with the guidance of Miwi2.

Discussion

Morc1-Dependent De Novo Formation of Heterochromatin on TEs.

This study focused on the mechanism of how to “establish” TE silencing during the mouse life cycle. In gonocytes, >10,000 copies of TEs are up-regulated to reorganize the overall chromatin structure, so that de novo DNA methyltransferases can gain access to the genome and catalyze DNAme along whole chromosomes (12). While this ectopic activation of numerous TEs would trigger chromatin reorganization, deregulation of TEs causes fatal genomic instability in general. Therefore, there should be certain machineries that resilence such active TEs. Here, we revealed that Morc1, which was previously shown to participate in DNAme on TEs, triggered enrichment of SetDB1-dependent H3K9me3 and chromatin compaction over TEs in gonocytes. Target recognition of Morc1 was most likely mediated by the PIWI-piRNA pathway that recognizes nascent transcript from transiently activated TEs, suggesting a critical role of Morc1 in reestablishing heterochromatin.

Small RNA Pathway in Morc1-Dependent Chromatin Modification.

A previous report suggested that Morc1 and Miwi2 act separately in silencing TEs (30). This result was based on RNA-seq data from whole testes at P10, suggesting that the levels of transcripts from isolated germ cells at embryonic or newborn stages are different from the mixture of somatic and germ cells in testes. In our study using H3K9me3 ChIP-seq datasets, we observed a substantial overlap between Miwi2 and Morc1 targets. In worms, Morc-1 acts in the RNAi pathway and plays a crucial role in germline immortality (47). Moreover, the H3K9me3 level on targets of HRDE-1, an Argonaute protein in worms, is reduced in Morc-1-mutant worms (48). These data support our model that mouse Morc1 exerts its repressive effect on TEs via a specific small RNA pathway.

Complex Interplay between DNA Methylation and H3K9me3 on TEs in Gonocytes.

With the aid of Trim28, specific KRAB-ZFPs recruit SetDB1 to genomic regions of TEs, triggering their repression. Using a specific reporter containing the TE sequence, deletion of Trim28 or SetDB1 abolishes both H3K9me3 and DNAme (42). Similarly, G9a is required for DNAme on TEs in mESCs, indicating that H3K9me3 precedes DNAme in such processes (43). Conversely, we observed some differences between MdTE (K9me3) and MdTE (DNAme) at the family/subfamily level. This is consistent with results from mESCs, in which the overall level of DNAme at TEs was unchanged or only modestly reduced in SetDB1 KO cells (49). Moreover, DNAme around TSSs of host genes is only weakly correlated to H3K9me3 in somatic cells and mESCs (50, 51). Thus, Morc1 induces the accumulation of two types of repressive epigenomic marks, probably because it forms different types of protein complexes depending on the chromatin environment around individual TE insertion site. It is known that, for example, Trim28 is also involved in DNA methylation at specific genomic loci (52). Specific KRAB zinc-finger protein recruits Trim28 to specifically induce DNA methylation around the loci it binds (53, 54). Therefore, it is possible that Morc1, like Trim28, assembles multiple protein complexes to distinctively accumulate DNAme and H3K9me3 at each TE locus.

Transcription of TEs Occurs Prior to Morc1-Dependent Heterochromatin Formation.

In plants and yeasts, there has been a well-established concept that the targets of nuclear RNAi machinery are transcribed to initiate transcriptional silencing (55). In Arabidopsis thaliana, an atypical RNA polymerase, Pol V, produces long noncoding RNA transcripts which serve as a scaffold for recruiting the nuclear RNAi machinery (56, 57). Then, this RNA-RNA hybridization triggers DNAme at cognate locus for silencing. In Schizosaccharomyces pombe, the centromeric repeats, where constitutive heterochromatin is formed, are transcribed by RNA polymerase II during S phase (58, 59). The resultant transcripts initiate the H3K9me3-marked heterochromatin formation in the following cell cycle. Much like these pathways, we found that mammalian nuclear RNAi-machinery triggered heterochromatin formation at TEs where transcripts were produced beforehand. Thus, this is a recurring model in small RNA-mediated transcriptional silencing observed across eukaryotes (60, 61).

In summary, we found that TEs in gonocytes, whose expression was tolerated by the host system, were resilenced by Morc1 at the chromatin level with accumulation of H3K9me3. Because Morc1 KO male mice show severe hypogonadism, this reestablishment of heterochromatin over activated TEs would ensure proper male fertility. Importantly, for group II and III TEs defined in Fig. 1, both their chromatin accessibility and their expression level decreased at P3 eventually (SI Appendix, Fig. S3 A and B), emphasizing the importance to suppress TE in the germline lineage. Our analysis also revealed that host genes near to Morc1-dependent TEs were activated at E16.5 in wild-type gonocytes (Fig. 6G). Therefore, during this time window, genes that are not transcribed from a normal TSS are expressed. The appearance of genes with a new domain conformation allows such genes to be domesticated by the host system (62, 63). In this regard, the gonocyte period could be a specific time window in which new genes emerge and evolve. Regardless, Morc1 plays a crucial role in coordinating interactions between host genes and TEs via multiple epigenetic modification.

Methods

Animal Care and Use.

All animal procedures were approved by the Institutional Safety Committee on Recombinant DNA Experiments and the Animal Research Committee of The University of Tokyo. Animal experiments were performed in accordance with the guidelines for animal experiments at The University of Tokyo.

Generation of Morc1 Knockout Mice.

The Morc1 KO mouse was generated by introduction of Cas9 protein (317–08441; NIPPON GENE), tracrRNA (GE-002; FASMAC) and synthetic crRNA (FASMAC), and ssODN into C57BL/6J (CLEA Japan) fertilized eggs by electroporation. The synthetic crRNAs were designed to GCACTGGTTAAAAGGCCGTG (TGG) of the first intron of Morc1 and ATAAGGGACCAGATGAACAG (TGG) in the 20th intron. ssODN: 5′- G​TGT​TAT​TAC​TGG​ACA​CCA​AGC​AGA​TTC​CAC​TGt​tta​ttt​att​GTG​TGG​AGG​GCG​GGTCACAAGAGAGATCGTGTG -3′ was used as a homologous recombination template. The electroporation solution contained 10 μM of tracrRNA, 10 μM of synthetic crRNA, 0.1 μg/μL of Cas9 protein, and 1 μg/μL of ssODN in Opti-MEM I Reduced Serum Medium (31985062; Thermo Fisher Scientific). Electroporation was carried out by following previous reports (64, 65). After electroporation, the embryos were cultured in KSOM (66) medium for O/N and then transferred into the oviducts of ICR (CLEA Japan) foster mothers at two-cell stage.

Isolation of Germ Cells from Testes.

Testes were obtained from E13.5 and E16.5 Mvh-Venus TG embryos and newborn male Mvh-Venus TG pups (67). After removing the tunica, dissociation buffer [500 µL Dulbecco’s modified eagle medium (DMEM), 10 µL fetal bovine serum (FBS), 7.5 µL of 100 mg/mL hyaluronidase (Tokyo Kasei, Japan, H0164), 2.5 µL of 10 mg/mL DNAse (Sigma, D5025-150kU), 10 µL of 100 mg/mL collagenase (Worthington, CLS1), and 25 µL of 14,000 U/mL recombinant collagenase (Wako, 036-23141)] was applied at 37 °C for 20 min. Subsequently, rigorous pipetting was performed until testicular cells were completely dissociated. After resuspending the cells in 2% FBS/PBS, Venus-positive cells were isolated by fluorescence-activated cell sorting using a FACS Aria III (BD). Before sorting cells, propidium iodine was added to select viable cells.

ATAC-Seq Library Construction.

Using the original protocol (68), an ATAC-seq library was constructed from gonocytes as described previously (12).

ChIP-Seq Library Construction.

Venus-positive testicular germ cells (2 × 10⁴) were fixed with 1% formaldehyde for 10 min. Cells were resuspended in Swelling buffer [20 mM Hepes (pH 7.9), 1.5 mM MgCl₂, 10 mM KCl, 0.1% NP-40, and 1 mM DTT]. After incubation on ice for 20 min, pelleted nuclei were resuspended in 1× shearing buffer (Covaris, 520154) and fragmented to ~500 bp with a sonicator (BRANSON, SFX150). Fragmented products diluted with RIPA buffer [50 mM Tris-HCl (pH 8.0), 150 mM NaCl, 2 mM EDTA (pH 8.0), 1% NP-40, 0.5% sodium deoxycholate, and 0.1% SDS] were mixed by rotation with Dynabeads-ProteinG (Thermo Fisher, 10009D) or Dynabeads M-280 Sheep anti-mouse IgG (Thermo Fisher, 11201D) for 1 h at 4 °C. Immunoprecipitation was performed with 2 µL anti-H3K9me3 antibody (Active Motif, 39161) on Dynabeads-ProteinG (Thermo Fisher, 10009D) or 1 µL anti-H3 antibody (MBL, 16004) on Dynabeads M-280 Sheep anti-mouse IgG (Thermo Fisher, 11201D) overnight at 4 °C. Beads were washed with low buffer [0.1% SDS, 1% Triton X-100, 2 mM EDTA (pH 8.0), 150 mM NaCl, and 20 mM Tris-HCl (pH 8.0)] three times and then with high buffer [0.1% SDS, 1% Triton X-100, 2 mM EDTA (pH 8.0), 500 mM NaCl, snf 20 mM Tris-HCl (pH 8.0)] once. IPed products were eluted from beads by suspension in direct elution buffer [10 mM Tris-HCl (pH 8.0), 5 mM EDTA (pH 8.0), 300 mM NaCl, and 0.5% SDS] while shaking at 65 °C for 15 min. The sample was treated with proteinase K for 6 h at 37 °C, followed by reversal of cross-linking overnight at 65 °C. DNA was extracted by EtOH precipitation. Library preparation was carried out with a QIAseq Ultralow Input Library Kit (Qiagen, 180492) following the manufacturer’s instructions.

ATAC-Seq Analysis.

Reads trimmed to 25 bp were aligned to the mouse (mm10) genome using Bowtie2 with -N 1 and -X 2000 parameters. If TEs with very similar sequences are located at different locations on the genome, reads derived from such TEs may be mapped to multiple locations on the genome, creating noise in the analysis for each transposon insertion site. Therefore, to minimize such effect, only reads that map to only one locus on the genome were used in the ATAC-seq analysis over the TSS region of TEs and host genes by calling grep with the -v option. To examine the dynamics of chromatin accessibility over internal regions of TEs, we also counted multiple mapped reads because most of the reads derived from such internal region are discarded in the analysis using only unique reads. Although this may cause some uncertainty in calling a given read from young TEs to certain genomic location, ATAC-seq data from Morc1 het mutant work as suitable negative control in examining the role of Morc1 in regulating accessibility of TE chromatin. After removing reads aligned to regions in the blacklist (69) and PCR duplicates with Picard, we calculated Pearson correlation coefficients between 10 kb bins of biological replicates on chr1 using Deeptools. ATAC-seq reads were normalized to CPM. Peak calling was performed using Homer with the following parameters: -style dnase, -size 500, and -minDist 1000. To define significantly accessible regions between Morc1 het and Morc1 KO gonocytes, peak call outputs and treated bam files were inputted to MAnorm. Significantly accessible regions were determined by the following criteria: ≥1 log2 fold change in normalized read counts and P-value < 0.0001.

ChIP-Seq Analysis.

Reads trimmed to 50 bp were aligned to the mouse (mm10) genome using Bowtie2 with the -N 0 option, and only unique reads were used for downstream analysis. As in the ATAC-seq analysis, to examine the dynamics of the H3K9me3 level over internal regions of TEs, we also counted multiple mapped reads, because most of the reads derived from such internal region are discarded in the analysis using only unique reads. After removing reads aligned to regions in the blacklist (69) and PCR duplicates with Picard, we calculated Pearson correlation coefficients between 10 kb bins of biological replicates on Chr1 using Deeptools. ChIP-seq reads were normalized to CPM, and relative H3K9me3 enrichment was calculated by dividing ChIP enrichment of H3K9me3 by that of H3. Peak calling was performed using Homer with the following parameters: -style histone, -size 1000, and -minDist 2500. To define peaks that indicated significantly more or less H3K9me3 in mutant mice compared with wild-type mice, MAnorm and MAnorm2 were used. Significantly increased or decreased H3K9me3 region candidates were determined by the following criteria: ≥1 log2 fold change in normalized read counts and P-value < 0.0001 (calculated with MAnorm) or ≥0.8 log2 fold change in normalized read counts and P-value < 0.005 (calculated with MAnorm2). Among candidates, regions where fold changes of H3K9me3 read counts/that of H3 read counts were more than two were defined as significantly increased or decreased H3K9me3 regions. To construct line plots of the H3K9me3 level over the family/subfamily of TEs, we used mapping files that included multiple mapped reads.

Genome Enrichment Calculation.

RepeatMasker (open-0.4.5, mm10) was used to annotate TEs in the mouse genome (mm10). The proportion of each TE family was calculated for specific regions and the entire genome. The proportion in specific regions divided by that in the entire genome was defined as the enrichment value of Morc1 and Miwi2 targets. TEs occupying >0.1% of the TE population were included in the analysis. Values in figures were converted to logarithmic values.

piRNA Enrichment Analysis.

To clarify the interaction between Morc1 and piRNA pathway, Miwi2-interacting piRNA reads were mapped to the mm10 genome with the condition that includes multiple mapped reads. piRNA which overlapped with Morc1 or Miwi2 target TEs were counted. In this analysis, target TEs are TEs such that the level of H3K9me3 over their TSSs is reduced in Morc1 or Miwi2 mutants.

RNA-Seq Analysis.

Reads trimmed to 90 bp were aligned to the mouse (mm10) genome using Hisat2 with the --dta option. As closely related TEs are quite similar in their genomic sequences, we included the multiple-mapped reads for RNA-seq analysis. This is because most RNA-seq reads are derived from TE bodies that are very similar in sequence within the same TE family, making it technically difficult to unambiguously identify the TE insertion site from which a given read was generated. In contrast, only uniquely mapped reads were used for downstream analysis in the ATAC-seq experiment. After removing reads aligned to regions in the blacklist (69), we used FeatureCount to calculate read counts on TEs and genes with the following parameters: -p, -M (when calculating read counts on genes, -M was removed). Read counts were normalized to RLE using DESeq2. For gene annotation, we used mus_musculus_GRCm38_102.gtf (https://asia.ensembl.org/Mus_musculus/Info/Index).

DNA Methylome Analysis.

Datasets analyzed were deposited in Gene Expression Omnibus (GEO) under accession numbers GSE12757 and GSE63048, and in DDBJ under accession numbers DRP000638 and DRP002386. Raw reads trimmed to 50 bp were aligned to the mouse (mm10) genome using BSseeker2 (70) with default parameters. CGmaptools was used to calculate and visualize CG methylation levels on TEs.

Organ Culture.

Organ culture was performed as described previously with some minor modifications (41) E16.5 testes without the epididymis were extracted from Mvh-Venus TG embryos and cut in half. DMEM (Nacalai) supplemented with 10% FBS (Gibco) and penicillin–streptomycin (Gibco) was used for the culture medium. Each testicular explant was placed in a 50 µL drop of the medium hanging on the lid of a culture dish and incubated at 37 °C for 3 d. The medium was changed every day. The SetDB1 Inhibitor was SETDB1-TTD-inhibitor (Target Mol, T9742) dissolved in DMSO and used at 10 µM as the final concentration.

ChIP-qPCR.

The starting material used for ChIP-qPCR analysis was 8 × 10³ cells. ChIP was performed as described above. Real-time quantitative PCR was performed using PowerUp SYBR Green Master Mix (Thermo Fisher) and the StepOnePlus real-time PCR system (Applied Biosystems). Primer sequences used for qPCR are listed in SI Appendix, Table S1.

DNA Gel Electrophoretic Mobility Shift Assay.

The electrophoretic mobility shift assay was performed following a published method (71) with some minor modifications. A pCAGGS plasmid expressing Flag-tagged GFP or Flag-tagged Morc1 was transfected into HEK293T cells. The cells were resuspended in buffer C (20 mM Hepes-KOH, 25% glycerol, 0.42 M NaCl, 1.5 mM MgCl₂, 0.2 mM EDTA, 0.5 mM DTT, and 0.1% NP-40) and subjected to ultrasonication (SFX250, Branson). Samples were centrifuged to remove cell debris. The supernatant was incubated with ANTI-FLAG M2 Affinity Gel (Merck) for 4 h at 4 °C. The resin was washed three times with buffer C and then incubated with 3× FLAG peptide solution. The eluted fraction was applied to an Amicon column (3 KDa cutoff, Merck) to remove the 3× FLAG peptide. Purified proteins were incubated with 234, 40, or 20 bp DNA substrates in 20 mM Tris-HCl (pH 7.5), 50 mM NaCl, 1 mM MgCl₂, 1 mM DTT, and 0.1 mg/mL BSA at 30 °C for 30 min. The reactions were resolved by 1.5% (for 234 bp DNA) or 2% (for 20 and 40 bp DNA oligos) agarose gel electrophoresis in TAE buffer (pH 7.5) at 4 °C for 2 h. DNA (234 bp) was detected by SYBR Gold staining (Thermo Fisher Scientific). X-Rhodamine-labeled DNA oligos were visualized using a Typhoon FLA 9500 scanner (GE Healthcare). The sequences of DNA substrates used in the electrophoretic mobility shift assay were as follows: 234 bp, 5′- CCA AGT CGA CAA ACA GCT ATT GTT AAC CCC CCT CCA CCA GAG TAC ATA AAC ACT AAG AAG AGT GGG CGG TTG ACG AAT CAG CTG CAG TTC CTA CAG AGG GTT GTG CTG AAG GCC CTG TGG AAG CAC GGC TTC TCT TGG CCT TTC CAA CAG CCG GTG GAC GCC GTG AAA CTA AAG CTG CCT GAC TAT TAC ACC ATC ATA AAA ACC CCA ATG GAT TTA AAT ACA ATT AAG AAG CGG-3′; 40 bp, 5′-TTT TGT ATT ATC CTT ATA CTT ATT TAC TTT ATG TTC ATT T /36-TAMTSp/-3′; 20 bp, 5′-TAC ATT GCT AGG ACA TCT TT /36-TAMTSp/-3′.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All data in this study have been deposited in the Gene Expression Omnibus (GEO) under accession number GSE235429 (72).