Destabilization of mRNAs enhances competence to initiate meiosis in mouse spermatogenic cells

Sexual reproduction depends on meiosis: a specialized cell cycle that produces haploid gametes via one round of DNA replication followed by two rounds of chromosome segregation. The major chromosomal events of meiosis, including pairing, synapsis and crossing over of homologous chromosomes, are generally conserved across eukaryotes. By contrast, the mechanisms that govern the transition from mitosis to meiosis are less well conserved (Kimble, 2011). For example, in budding yeast, meiotic initiation occurs when multiple inputs converge to activate a master transcription factor (IME1) that upregulates meiotic gene expression (Kassir et al., 1988; Smith et al., 1990; reviewed by van Werven and Amon, 2011). In Drosophila melanogaster, the transition from mitosis to meiosis occurs via translational repression mediated by the RNA helicase Bgcn and its binding partners Bam and Tut (Chen et al., 2014; Gonczy et al., 1997; Insco et al., 2009, 2012; McKearin and Spradling, 1990).

In mammals, transcriptional activation induced by extrinsic signaling plays a central role in meiotic initiation (i.e. the meiotic G1/S transition). During the transition from mitosis to meiosis (i.e. the ‘premeiotic’ stage during oogenesis and the ‘preleptotene’ stage during spermatogenesis), retinoic acid induces meiotic entry by transcriptionally activating Stra8 and Meiosin (Anderson et al., 2008; Baltus et al., 2006; Dokshin et al., 2013; Ishiguro et al., 2020). In vitro, nutrient restriction synergizes with retinoic acid to induce meiotic entry in spermatogenic cells (Zhang et al., 2021). The STRA8-MEIOSIN heterodimer transcriptionally activates expression of G1/S cyclins and meiotic factors that orchestrate the chromosomal events of meiotic prophase I (Anderson et al., 2008; Baltus et al., 2006; Ishiguro et al., 2020; Kojima et al., 2019; Mark et al., 2008; Soh et al., 2015; Zhang et al., 2023). In mice lacking Stra8 or Meiosin on a C57BL/6 genetic background, premeiotic oogonia and preleptotene spermatocytes halt their development and fail to progress to meiotic prophase I (Anderson et al., 2008; Baltus et al., 2006; Dokshin et al., 2013; Ishiguro et al., 2020). Indeed, Stra8-null premeiotic oogonia and preleptotene spermatocytes, as well as Meiosin-null premeiotic oogonia, fail to undergo meiotic DNA replication (Anderson et al., 2008; Baltus et al., 2006; Dokshin et al., 2013; Ishiguro et al., 2020). [Meiosin-null preleptotene spermatocytes initiate an S phase, but it remains unclear whether it is mitotic or meiotic in nature (Ishiguro et al., 2020).]

In the mammalian testis, competence to initiate meiosis in response to retinoic acid is acquired during spermatogenesis. In mitotic spermatogonia, DMRT1 postpones the acquisition of competence, thereby preventing precocious meiotic entry, by repressing retinoic acid-dependent transcription as well as Stra8, and likely Meiosin, gene expression (Matson et al., 2010; Ishiguro et al., 2020). As a result, undifferentiated spermatogonia exposed to endogenous retinoic acid express Stra8 at low levels, but do not express Meiosin, and consequently become mitotic differentiating spermatogonia (Ishiguro et al., 2020; Matson et al., 2010). Differentiating spermatogonia prematurely exposed to exogenous retinoic acid may express STRA8 protein, but they do not initiate meiosis (Endo et al., 2015; Johnson et al., 2023). At the mitosis-to-meiosis transition, the SCF E3 ubiquitin ligase complex degrades DMRT1 (Matson et al., 2010; Nakagawa et al., 2017), thereby conferring upon spermatogenic cells the competence to express Stra8 and Meiosin, and to initiate meiosis in response to retinoic acid. To date, the SCF complex is the only known positive regulator of spermatogenic cell competence to initiate meiosis.

Post-transcriptional regulation of mRNA also plays a role in governing the mitosis-to-meiosis transition in mammals. After meiotic initiation, MEIOC acts together with the RNA helicase YTHDC2 and the RNA-binding protein RBM46 to regulate progression through meiotic S phase into meiotic prophase I (Abby et al., 2016; Bailey et al., 2017; Hsu et al., 2017; Jain et al., 2018; Qian et al., 2022; Soh et al., 2017; Wojtas et al., 2017). The three proteins are required to increase meiotic gene expression and also to repress the mitotic cell cycle program, thereby inhibiting a premature and aberrant metaphase several days before wild-type meiotic metaphase I. MEIOC, YTHDC2 and RBM46 are homologs of Drosophila proteins Bam, Bgcn and Tut, respectively (Bailey et al., 2017; Jain et al., 2018; Peart et al., 2022; Qian et al., 2022; Soh et al., 2017). Like their homologs, MEIOC, YTHDC2 and RBM46 interact with one another, as well as with mRNA (Abby et al., 2016; Bailey et al., 2017; Hsu et al., 2017; Jain et al., 2018; Li et al., 2022; Peart et al., 2022; Qian et al., 2022; Saito et al., 2022; Soh et al., 2017). YTHDC2 also interacts with exoribonuclease XRN1, while RBM46 recruits nonsense-mediated decay protein UPF1 and subunits of the cytoplasmic deadenylase CCR4-NOT complex, which suggests that YTHDC2 and RBM46 mRNA targets are degraded (Kretschmer et al., 2018; Li et al., 2022; Qian et al., 2022; Wojtas et al., 2017). The mechanism by which the MEIOC-YTHDC2-RBM46 complex post-transcriptionally regulates its mRNA targets after meiotic initiation remains poorly defined.

Here, we molecularly dissect the activity of MEIOC in mouse spermatogenic cells during the mitosis-to-meiosis transition using two parallel approaches: (1) single-cell RNA sequencing (scRNA-seq) analysis of postnatal testes and (2) bulk RNA-seq analysis of developmentally synchronized testes with histologically verified staging. Both approaches reveal that Meioc-null germ cells developmentally diverge from wild-type spermatogenic cells at the meiotic G1/S transition, earlier than previously appreciated. Furthermore, we find that MEIOC activity leads to an increase in meiosis-associated gene expression in late mitotic spermatogonia, before meiotic initiation. We discover that MEIOC destabilizes mRNAs encoding transcriptional repressors of meiotic gene expression. This inhibition of repressors facilitates activation of the master meiotic transcriptional regulator STRA8-MEIOSIN in response to retinoic acid. Therefore, a post-transcriptional repressor of mRNA, acting in parallel to the degradation of DMRT1 by the SCF complex, enhances the competence of spermatogenic cells to activate the meiotic transcriptional regulator and initiate meiosis.

To characterize the molecular consequences of MEIOC activity during spermatogenesis, we performed 10x Genomics Chromium-based scRNA-seq on Meioc-null and wild-type P15 testes, and identified germ cell clusters by cell type-enriched marker expression and transcriptome-based cell cycle analysis (Figs S1-S3, Table S2). These germ cell clusters, identified in both wild-type and Meioc-null testes, consisted of four clusters in mitosis [undifferentiated spermatogonia (Undiff), differentiating type A₁-A₄ spermatogonia (A1-4), differentiating intermediate and type B spermatogonia (In/B) and differentiating type B spermatogonia in G2/M phase (B G2M)]; three clusters spanning the mitosis-to-meiosis transition [preleptotene spermatocytes in G1, early S and late S phase (pL G1, pL eS and pL lS, respectively)]; and three clusters in meiotic prophase I [leptotene spermatocytes (L), zygotene spermatocytes (Z) and pachytene spermatocytes (P)]. Although previous histological characterization found that Meioc-null spermatogenic cells did not progress beyond early zygotene (Abby et al., 2016; Soh et al., 2017), some Meioc-null cells were designated as pachytene spermatocytes in the scRNA-seq clustering. This indicates that transcriptomic progression in Meioc-null spermatocytes is not reflected at the histological level. We also identified a cluster composed of Meioc-null spermatocytes, which we designated as ‘Mutant-only’ (Mut). As this cluster primarily comprised cells in G2/M phase (Fig. S1C), it likely includes the aberrant metaphase spermatogenic cells found in P15 Meioc-null testes, days before meiotic metaphase I occurs in the wild-type testis (Abby et al., 2016; Soh et al., 2017).

We then applied pseudotime analysis to computationally reconstruct the developmental trajectory of the spermatogenic cells from undifferentiated spermatogonia to pachytene spermatocytes (Fig. 1A). Meioc-null germ cells followed the same trajectory as wild-type cells during the mitotic stages of spermatogenesis but diverged in the G1 phase of the preleptotene stage, immediately before meiotic S phase. As spermatogenic cells are classically staged via histology, we used histologically staged samples to independently verify these pseudotime results. We developmentally synchronized spermatogenesis to obtain Meioc-null and wild-type testes enriched for preleptotene spermatocytes (Fig. S4A). After histologically verifying staging (Fig. S4B), preleptotene-enriched testes were analyzed via bulk RNA-seq for differential expression, and we identified over 2000 transcripts as differentially abundant (Fig. S4C, Table S9). We then asked whether MEIOC impacts the G1/S transition in the preleptotene-enriched testes. We compared percentile ranks for fold changes (wild type/Meioc knockout) for genes enriched at specific cell cycle phases (see Materials and Methods). As a positive control for the disrupted G1/S transition, we re-analyzed a bulk RNA-seq dataset of preleptotene spermatocytes sorted from wild-type and Stra8-null testes on a C57BL/6 background (Kojima et al., 2019). This analysis revealed that, in preleptotene spermatocytes, MEIOC increased transcript abundance for genes associated with G1/S, S and G2, similar to STRA8 (Fig. S4D, Table S10). In total, both scRNA-seq and bulk RNA-seq datasets demonstrate that MEIOC drives a major transcriptomic shift in spermatogenic cells at the meiotic G1/S transition. This is earlier than previously appreciated, as previous histological studies reported that Meioc-null spermatogenic cells diverge from wild-type controls after entering meiotic S phase (Abby et al., 2016; Soh et al., 2017).

MEIOC protein and its binding partner YTHDC2 are first detected immunohistochemically at the preleptotene stage, but YTHDC2 protein is also detected via western blotting in mitotic spermatogonia (Abby et al., 2016; Bailey et al., 2017; Hsu et al., 2017; Jain et al., 2018; Soh et al., 2017; Wojtas et al., 2017). The binding partner RBM46 is detected via immunostaining in mitotic spermatogonia through meiotic spermatocytes (Peart et al., 2022; Qian et al., 2022). We found that Meioc and Ythdc2 transcripts became abundant in late mitotic spermatogonia (In/B and B G2M clusters), whereas Rbm46 was highly abundant in all germ cell clusters examined (Fig. 1B, Table S2). These observations raised the possibility that MEIOC, potentially in collaboration with YTHDC2 and RBM46, shapes the germline transcriptome in late mitotic spermatogonia, before the protein can be reliably detected via immunostaining.

To assess how MEIOC impacts the transcriptome during spermatogenesis, we carried out scRNA-seq-based differential expression analysis in germ cell clusters associated with mitotic spermatogonia (Undiff, A1-4, In/B and B G2M clusters), the mitosis-to-meiosis transition (pL G1, eS and lS clusters) and meiotic prophase I (L and Z clusters), and identified genes whose transcript abundance is increased (log2 fold change wild type/Meioc knockout>0.1 and adjusted P<0.05) or decreased (log2 fold change wild type/Meioc knockout<−0.1 and adjusted P<0.05) by MEIOC (Fig. 1C, Table S3). The number of differentially abundant transcripts increased in the In/B and B G2/M clusters, mirroring Meioc and Ythdc2 expression patterns, and continued to increase thereafter. We conclude that MEIOC begins to affect the germline transcriptome before the meiotic G1/S transition.

Based on functional analysis, transcripts upregulated by MEIOC in the In/B through pL lS clusters were enriched for Gene Ontology (GO) annotations related to meiosis (Fig. 1D, Table S4). Several transcripts contributing to this enrichment became more abundant in mitotic spermatogonia and remained more abundant through early meiotic spermatocytes (Fig. 1E). However, in the L and Z clusters, these changes in meiosis-associated transcripts no longer represented an enrichment in meiosis-associated GO annotations (Fig. 1D). Bulk RNA-seq analysis of preleptotene-enriched testes also revealed enrichment of meiosis-associated factors among MEIOC-upregulated transcripts (Fig. S4C,E, Table S11). Therefore, MEIOC activity leads to increased meiotic transcript abundance in late mitotic spermatogonia, earlier than previously appreciated. Misregulation of these transcripts provides a molecular rationale for the delayed progression of Meioc-null spermatogenic cells from preleptotene to leptotene to zygotene stages (Abby et al., 2016; Soh et al., 2017).

Given that MEIOC lowers mitotic transcript abundance in fetal oogonia (Soh et al., 2017), we asked whether similar transcriptomic changes were induced by MEIOC in mitotic spermatogonia. MEIOC-downregulated transcripts in the pL lS to Z clusters, but not in developmentally earlier clusters, were enriched for the GO annotation ‘mitotic cell cycle’ (Fig. 1D,E, Table S5). This included mitotic G1/S and G2/M cyclin Ccna2 (Fig. 1E), the downregulation of which was first evident in B G2/M and pL G1 clusters, before the broader enrichment of mitotic cell cycle-associated transcripts was detected. Consistent with downregulation of Ccna2 transcript abundance by MEIOC, Meioc-null spermatocytes exhibit prolonged CCNA2 protein expression at leptotene and zygotene stages, when wild-type spermatocytes no longer express this protein (Soh et al., 2017). Based on a curated list of genes whose expression is linked to specific cell cycle phases (Hsiao et al., 2020), MEIOC-downregulated genes in pL lS to Z clusters were primarily associated with G2 or M phases (Fig. 1F). We conclude that the downregulation of mitotic cell cycle transcripts by MEIOC occurs primarily after the meiotic G1/S transition. The loss of this regulation likely contributes to the delayed progression of Meioc-null spermatogenic cells from preleptotene to leptotene to zygotene stages, and their premature progression into an aberrant metaphase state during meiotic prophase I (Abby et al., 2016; Soh et al., 2017).

In summary, MEIOC elevates the abundance of meiosis-associated transcripts beginning in late transit-amplifying spermatogonia through the mitosis-to-meiosis transition and early meiotic prophase I. Then, late in the mitosis-to-meiosis transition into meiotic prophase I, MEIOC broadly lowers mitotic transcript abundance.

In early spermatocytes, MEIOC localizes to the cytoplasm and interacts with the RNA-binding proteins YTHDC2 and RBM46, which recruit other proteins that degrade mRNA (Kretschmer et al., 2018; Li et al., 2022; Qian et al., 2022; Wojtas et al., 2017). Therefore, we hypothesized that MEIOC degrades its mRNA targets. We first defined transcripts that associate with MEIOC by reanalyzing previously published MEIOC RIP-seq data from P15 testes (Soh et al., 2017), identifying 1991 MEIOC-bound mRNA (Table S6). We assessed the molecular impact of the interaction of MEIOC with these transcripts via three approaches. First, we examined the representation of these targets among transcripts exhibiting statistically significant abundance changes that were MEIOC dependent (i.e. MEIOC upregulated or downregulated). We found that MEIOC targets were enriched among transcripts whose abundance decreased, but not those that increased, in response to MEIOC, in the B G2/M through Z clusters (Fig. 2A). Second, we examined all MEIOC targets, irrespective of whether they met a statistical cut-off in the differential expression analysis. MEIOC targets exhibited slightly lower fold changes (wild type/Meioc knockout) than nontargets in the A1-4 through Z clusters (Fig. 2B, Fig. S5A). Third, an analysis of estimated changes in transcript stability from bulk RNA-seq data showed that MEIOC targets had reduced transcript stability relative to nontargets (Fig. S6A, Table S7). Therefore, MEIOC lowers the abundance of transcripts that it targets, presumably by promoting their degradation, beginning in mitotic spermatogonia.

Given that MEIOC forms a complex with YTHDC2 and RBM46 (Abby et al., 2016; Li et al., 2022; Qian et al., 2022; Soh et al., 2017), we hypothesized that many MEIOC targets are also bound by YTHDC2 and RBM46. Based on a published YTHDC2 CLIP analysis from testes (Saito et al., 2022) and our re-analysis of RBM46 CLIP data from testes (Qian et al., 2022), MEIOC shared 202 mRNA targets with YTHDC2 and 1477 targets with RBM46, both of which are statistically significant overlaps (Fig. 2C). The three proteins had 187 mRNA targets in common. Only 15 mRNAs were bound by MEIOC and YTHDC2 but not RBM46, and these likely represent technical differences between datasets rather than a biologically meaningful group of transcripts. MEIOC-YTHDC2-RBM46 targets were enriched for GO annotations associated with RNA stability, whereas MEIOC-RBM46-only targets did not exhibit statistically significant enrichment (Table S6). We conclude that MEIOC shares many, but not all, of its mRNA targets with YTHDC2 and RBM46.

As YTHDC2 and RBM46 interact with proteins that degrade mRNA, we hypothesized that MEIOC has a greater impact on transcript stability when acting in partnership with YTHDC2 and RBM46 than when acting alone. To test this, we binned the targets of MEIOC as follows: MEIOC-YTHDC2-RBM46 targets, MEIOC-RBM46-only targets, MEIOC-YTHDC2-only targets and MEIOC-only targets. We compared how these groups of target genes respond to MEIOC in our scRNA-seq data. First, we examined whether each group was over-represented among transcripts whose abundance decreased in response to MEIOC. MEIOC-YTHDC2-RBM46 targets exhibited such enrichment beginning in mitotic spermatogonia (Fig. 2D). By contrast, MEIOC-YTHDC2-only targets were enriched in one meiotic cluster and MEIOC-only targets showed no such enrichment (Fig. 2D). MEIOC-RBM46-only targets were enriched beginning at the mitosis-to-meiosis transition (Fig. 2D), suggesting that MEIOC-RBM46 may impact the transcriptome independently of YTHDC2, but the biological significance of this regulation remains unclear. Second, we examined the effect of YTHDC2 and RBM46 on the stability of MEIOC targets, irrespective of statistical cutoff. Relative to MEIOC-only targets, MEIOC-YTHDC2-RBM46 targets exhibited the largest decrease in transcript abundance in all scRNA-seq germ cell clusters (Fig. 2E, Fig. S5B), as well as in transcript stability in the bulk RNA-seq analysis of preleptotene-enriched testes (Fig. S6B). Taken together, these data demonstrate that the destabilization by MEIOC of its target mRNAs occurs via its interaction with both YTHDC2 and RBM46.

Given that MEIOC-downregulated transcripts are enriched for mitotic cell cycle factors, we hypothesized that MEIOC-YTHDC2-RBM46 directly targets and destabilizes these factors. By comparing MEIOC-downregulated transcripts bearing the GO annotation ‘mitotic cell cycle’ with MEIOC-YTHDC2-RBM46 targets, we discovered a small but statistically significant overlap that included factors that impact cell cycle progression, such as Ccna2 and E2f6 (Fig. 2F). Some targets were downregulated by MEIOC in mitotic spermatogonia and early in the mitosis-to-meiosis transition, before mitotic cell cycle enrichment was evident among the MEIOC-downregulated genes (Fig. 1D,E). Therefore, the destabilization of mRNAs by MEIOC before and during the mitosis-to-meiosis transition may impact the ability of spermatogenic cells to establish a meiosis-specific cell cycle program in meiotic prophase I.

We also considered the possibility that some MEIOC-YTHDC2-RBM46 targets are stabilized, rather than destabilized, by the complex. In particular, given that MEIOC-upregulated transcripts are enriched for the ‘meiotic cell cycle’ annotation (Fig. 1D,E), we asked whether these transcripts were MEIOC-YTHDC2-RBM46 targets. MEIOC-upregulated ‘meiotic cell cycle’ transcripts and MEIOC-YTHDC2-RBM46 targets shared one transcript, Meioc (Fig. 2F), with this overlap representing neither statistical depletion nor enrichment. We conclude that MEIOC-YTHDC2-RBM46 does not directly regulate the stability of meiosis-associated transcripts. Overall, MEIOC collaborates with YTHDC2 and RBM46 to promote the decay of the transcripts that it targets, beginning in mitotic spermatogonia.

Most MEIOC-regulated transcripts, as defined by RNA-seq analysis of wild-type versus Meioc-null samples, do not directly interact with MEIOC-YTHDC2-RBM46 (Fig. 3A, Fig. S6C). We examined whether these changes in transcript abundance reflected altered transcriptional rates or altered transcript stabilities. Applying REMBRANDTS to our bulk RNA-seq data, we estimated mRNA abundance from exonic reads, pre-mRNA abundance (i.e. transcriptional rate) from intronic reads, and mRNA stability from the difference between exonic and intronic reads (Alkallas et al., 2017; Gaidatzis et al., 2015). MEIOC-regulated transcripts exhibited large changes in transcriptional rates and smaller changes in transcript stabilities (Fig. 3B). We conclude that MEIOC indirectly impacts transcription in ways that alter the abundance of transcripts that it does not bind.

We set out to identify MEIOC targets that might drive such transcriptional changes. The vast majority of MEIOC-upregulated transcripts are not directly bound by MEIOC (Figs 2A and 3A). We hypothesized that MEIOC-YTHDC2-RBM46 destabilizes a mRNA that encodes a transcriptional repressor; destabilization of this mRNA then derepresses (i.e. upregulates) gene expression. For this analysis, we focused on the In/B to pL G1 clusters, before spermatogenic cells have undergone the MEIOC-dependent transcriptomic shift observed in the scRNA-seq pseudotime analysis (Fig. 1A). Among 48 MEIOC-YTHDC2-RBM46 targets downregulated by MEIOC in the In/B to pL G1 clusters, we identified eight mRNAs that encode transcriptional repressors: Cbx1, E2f6, Hnrnpab, Mga, Pbrm1, Rad21, Sp3 and Suz12 (Fig. 3C,D). Strikingly, Cbx1, E2f6 and Mga all encode subunits of the noncanonical Polycomb Repressive Complex (PRC) 1.6 (Figs S7A and S9A), representing a statistically significant enrichment for the subunits of the complex (one-tailed hypergeometric test, P=3.55E-05). E2f6 and Mga are particularly attractive candidates because they encode sequence-specific DNA-binding subunits of PRC1.6 that are required to repress meiosis-specific genes in somatic and embryonic stem cells (Pohlers et al., 2005; Dahlet et al., 2021; Kehoe et al., 2008; Maeda et al., 2013; Suzuki et al., 2016; Kitamura et al., 2021; Uranishi et al., 2021). By contrast, CBX1 is not strictly required to repress gene expression, as other CBX proteins can compensate for its absence (Ostapcuk et al., 2018).

We hypothesized that the inhibition of E2f6 and Mga by MEIOC-YTHDC2-RBM46 impacts gene expression during spermatogenesis. To test this, we re-analyzed ChIP-seq and RNA-seq datasets from mouse embryonic stem cells (Dahlet et al., 2021; Qin et al., 2021; Stielow et al., 2018) to identify genes directly repressed by E2F6 or MGA (see Materials and Methods; Fig. S8A,B, Table S12). Given that genetic ablation of Max (which encodes the binding partner of MGA and a PRC1.6 subunit) induces embryonic stem cells to enter meiosis (Suzuki et al., 2016), these data are relevant to meiotic entry in spermatogenic cells. As E2F6 and MGA cooperate to repress expression of an overlapping set of genes (Dahlet et al., 2021) (Fig. S8C), we classified genes as repressed by both E2F6 and MGA, repressed by E2F6 alone, or repressed by MGA alone. Genes repressed by both E2F6 and MGA, but not those repressed by either factor alone, were over-represented among MEIOC-upregulated transcripts, beginning in the mitotic In/B cluster (Fig. 3E), coinciding with the destabilization of Mga mRNA by MEIOC in that cluster (Fig. 3D). This over-representation continued through the mitosis-to-meiosis transition (Fig. 3E), including the pL G1 cluster, when MEIOC destabilizes E2f6 and Mga mRNA (Fig. 3D). Genes repressed by both E2F6 and MGA were not enriched among MEIOC-downregulated transcripts (Fig. 3E). Bulk RNA-seq analysis of preleptotene-enriched testes largely confirmed these results, while also revealing that MEIOC-upregulated genes were enriched for genes repressed by E2F6 or MGA alone (Fig. S9B), likely because this analysis included lowly expressed genes that were not detected in the sparser scRNA-seq dataset. We conclude that the repression of E2F6 and MGA by MEIOC results in enhanced expression of specific genes in mitotic spermatogonia.

We also examined whether the repression of E2F6 and MGA by MEIOC results in diminished expression of other genes in spermatogenic cells. For this analysis, we used the mouse embryonic stem cell datasets to classify genes as activated by both E2F6 and MGA, by E2F6 alone, or by MGA alone. Genes activated by MGA alone were modestly enriched among MEIOC-downregulated genes within a limited set of clusters (pL lS and L; Fig. S8D). By contrast, genes activated by both E2F6 and MGA, or by E2F6 alone were not enriched among MEIOC-upregulated (or downregulated) genes (Fig. S8D). Bulk RNA-seq analysis of preleptotene-enriched testes confirmed these results (Fig. S9B). We conclude that the repression of E2F6 and MGA by MEIOC leads to upregulation of gene expression.

As many MEIOC-upregulated transcripts are meiotic cell cycle factors not bound by MEIOC-YTHDC2-RBM46 (Figs 1D and 2F), we hypothesized that repression of E2F6 and MGA enhances expression of these meiotic genes. To test this, we asked whether E2F6- and MGA-repressed genes are enriched among meiotic cell cycle transcripts upregulated by MEIOC. We again focused this analysis on the In/B to pL G1 clusters, before spermatogenic cells undergo MEIOC-dependent transcriptomic changes. We found that 13 out of 17 MEIOC-upregulated meiotic cell cycle transcripts are targeted for repression by E2F6 and/or MGA (Fig. 3F), and we confirmed these observations in bulk RNA-seq analysis of preleptotene-enriched testes (Fig. S9C). We conclude that the destabilization of E2f6 and Mga mRNAs by MEIOC de-represses meiosis-associated gene expression beginning in mitotic spermatogonia.

In mouse embryonic stem cells, E2F6 and MGA directly repress Meiosin (previously known as Gm4969 and Bhmg1) (Uranishi et al., 2021) (Fig. S8A,B). Consistent with the destabilization of E2f6 and Mga mRNAs by MEIOC, we found that MEIOC enhances Meiosin expression at both the transcript and protein levels during the mitosis-to-meiosis transition (Fig. 4A, Figs S4C, S10A, Table S13). By meiotic prophase I, Meioc-null germ cells exhibited delayed upregulation of Meiosin expression (Fig. 4A). We confirmed that Meiosin expression is also increased by YTHDC2 and RBM46, based on our examination of published RNA-seq data from postnatal testes (Jain et al., 2018; Peart et al., 2022). MEIOC protein does not bind Meiosin mRNA (Table S6). This suggests that the regulation of E2f6 and Mga by MEIOC-YTHDC2-RBM46 may be impacting Meiosin gene expression.

We examined whether MEIOC affects other known regulators of Meiosin gene expression. As retinoic acid transcriptionally activates Meiosin (Ishiguro et al., 2020), we considered whether MEIOC enhances retinoic acid-mediated transcription, but we found no evidence for this model. MEIOC did not increase the transcript abundance of any retinoic acid receptors (RARs) or retinoid X receptors (RXRs), which mediate transcriptional activation by retinoic acid (reviewed by Endo et al., 2019) (Fig. S10B,C). In addition, MEIOC did not upregulate additional retinoic acid-activated genes, such as Stra8, which encodes the binding partner of MEIOSIN, and Rec8, a meiotic cohesin (Koubova et al., 2014; Soh et al., 2015; Zhang et al., 2021) (Fig. 4A, Fig. S10C,D). We observed that MEIOC downregulated Stra8 abundance overall in the scRNA-seq data (Fig. 4A), but as Stra8 mRNA is not directly bound by MEIOC (Table S6), the molecular basis for this regulation remains uncharacterized. These observations indicate that MEIOC does not activate Meiosin gene expression by increasing retinoic acid-mediated transcription.

We also considered whether Dmrt1, encoding a protein that presumably represses Meiosin gene expression (Ishiguro et al., 2020), was regulated by MEIOC-YTHDC2-RBM46; again, we found no evidence for this alternative model. Dmrt1 transcript was not a target of MEIOC, YTHDC2 or RBM46 (Table S6), nor was it differentially expressed in response to MEIOC in the pL G1 or pL eS clusters of the scRNA-seq analysis or in the preleptotene-enriched testes of the bulk RNA-seq analysis (Fig. S11A,B). DMRT1 protein expression was similar in both wild-type and Meioc-null testes, with DMRT1 expressed in mitotic spermatogonia but absent from preleptotene spermatocytes (Fig. S11C). In addition, MEIOC did not affect the expression of DMRT1-regulated genes Tbx1 and Crabp2 (Matson et al., 2010) (Fig. S11D,E). DMRT1 also directly inhibits Stra8 expression (Matson et al., 2010), but as noted above, Stra8 was not activated by MEIOC (Fig. 4A). Therefore, changes in DMRT1 do not account for the upregulation of Meiosin gene expression by MEIOC. Taking these observations together, we conclude that the inhibition of E2f6 and Mga by MEIOC derepresses Meiosin gene expression during the mitosis-to-meiosis transition.

STRA8-MEIOSIN functions as an obligate heterodimer that transcriptionally activates gene expression during the mitosis-to-meiosis transition (Ishiguro et al., 2020; Kojima et al., 2019). Stra8 is highly expressed at both the transcript and protein levels in Meioc-null spermatogenic cells at the mitosis-to-meiosis transition (Fig. 4A, Figs S4C and S10A), as previously reported (Abby et al., 2016; Soh et al., 2017). We hypothesized that the upregulation of Meiosin gene expression by MEIOC leads to STRA8-MEIOSIN-mediated transcriptional changes. We tested three predictions of this hypothesis.

First, we tested whether, during the mitosis-to-meiosis transition, genes dependent on MEIOC also depend on STRA8. We defined STRA8-dependent (i.e. STRA8-upregulated or -downregulated) genes using bulk RNA-seq data from wild-type and Stra8-null preleptotene spermatocytes isolated via synchronization and sorting (Kojima et al., 2019). STRA8-upregulated genes were enriched among MEIOC-upregulated genes during the mitosis-to-meiosis transition (pL eS and lS clusters) and in meiotic prophase I (L cluster; Fig. 4B). Similarly, STRA8-downregulated transcripts were enriched among MEIOC-downregulated transcripts at these same stages (Fig. 4B). Bulk RNA-seq analysis of preleptotene-enriched testes produced similar results (Fig. S12A).

Second, focusing exclusively on STRA8-dependent genes, we looked for correlated changes in transcript abundance in the MEIOC scRNA-seq and STRA8 bulk RNA-seq datasets. STRA8-dependent genes exhibited significantly correlated effects in the MEIOC and STRA8 datasets that peaked at the mitosis-to-meiosis transition (pL eS and lS clusters; Fig. 4C, Figs S13A, S14A). STRA8-dependent genes similarly exhibited robustly correlated effects in the MEIOC bulk RNA-seq dataset from preleptotene-enriched testes and the STRA8 dataset (Fig. S12B). We conclude that MEIOC activity leads to STRA8-MEIOSIN-mediated changes in transcript abundance.

Third, we tested whether MEIOC-dependent genes are directly activated by STRA8-MEIOSIN. We identified STRA8-activated genes as those genes (1) whose expression increased in response to STRA8 and (2) whose promoters were bound by STRA8 (as defined by ChIP-seq from testes enriched for preleptotene spermatocytes; Kojima et al., 2019). We found that MEIOC-upregulated genes were enriched for STRA8-activated genes during the mitosis-to-meiosis transition (pL eS and lS clusters; Fig. 4D). However, de novo motif analysis failed to identify enrichment of any motif among these MEIOC-upregulated genes, likely due to the sparsity of scRNA-seq data. The enrichment of STRA8-activated genes among MEIOC-upregulated genes was also visible in the bulk RNA-seq analysis of preleptotene-enriched testes (Fig. S12C). Furthermore, the top motif identified by de novo motif analysis of MEIOC-upregulated genes from the bulk RNA-seq analysis matched the STRA8-MEIOSIN binding motif (Fig. S12D). The second top motif matched the binding site shared by E2F6 and other E2F proteins (Fig. S12D), as previously reported among STRA8-activated genes (Kojima et al., 2019). In total, ∼60-70% of MEIOC-upregulated genes are directly activated by STRA8-MEIOSIN.

Together, these analyses demonstrate that MEIOC indirectly activates the STRA8-MEIOSIN transcriptional regulator, thereby facilitating a massive shift in the transcriptome at the meiotic G1/S transition (Fig. 1A). With this new molecular insight, we revisited our interpretation of the Meioc-null phenotype. Based on previous histological analyses and DNA labeling experiments, Meioc-null spermatogenic cells successfully enter the preleptotene stage, express STRA8 protein and exhibit DNA synthesis. However, they then accumulate at the preleptotene stage and are delayed in entering meiotic prophase I (Abby et al., 2016; Soh et al., 2017). Previously, we interpreted these data as Meioc-null spermatogenic cells successfully undergoing the meiotic G1/S phase transition but thereafter exhibiting disrupted progression into meiotic prophase I. Given our present findings that Meioc-null germ cells are delayed in upregulating Meiosin expression and exhibit a disrupted meiotic G1/S phase transition, we now interpret the Meioc-null phenotype as defective meiotic initiation. We conclude that MEIOC enhances the competence of spermatogenic cells to initiate meiosis by activating the STRA8-MEIOSIN transcriptional regulator.

We observed that many STRA8-activated genes are repressed by E2F6 and/or MGA (Fig. S12E). Consistent with this observation, STRA8-activated genes are enriched for the binding motif of E2F6 and other E2F proteins (Fig. 1,2D; Kojima et al., 2019). Accordingly, we divided STRA8-activated genes into two groups, based on whether or not they were directly repressed by E2F6 and/or MGA in embryonic stem cells. As MEIOC represses E2F6 and MGA in mitotic spermatogonia and activates STRA8-MEIOSIN during the mitosis-to-meiosis transition, we predicted that genes experimentally defined as E2F6- and/or MGA-repressed and STRA8-activated would be upregulated by MEIOC, starting in mitotic spermatogonia, coincident with the destabilization of E2f6 and Mga mRNAs by MEIOC. Conversely, STRA8-activated genes not repressed by E2F6 or MGA should be upregulated by MEIOC later in the mitosis-to-meiosis transition, coincident with MEIOC-mediated upregulation of Meiosin gene expression. Consistent with these predictions, STRA8-activated, E2F6- and/or MGA-repressed genes were enriched among MEIOC-upregulated genes in the mitotic B G2/M cluster and during the mitosis-to-meiosis transition (pL G1, eS and lS clusters; Fig. 4E). Also as predicted, STRA8-activated genes not repressed by E2F6 or MGA exhibited enrichment in the pL eS and lS clusters (Fig. 4E). Both genesets were also enriched among MEIOC-upregulated genes in the bulk RNA-seq analysis (Fig. S12F). We conclude that, during late transit amplification, MEIOC derepresses a set of meiotic genes targeted by E2F6 and MGA; then, during the mitosis-to-meiosis transition, MEIOC elevates expression of these and other meiotic genes by activating STRA8-MEIOSIN.

In total, the destabilization of E2f6 and Mga mRNAs by MEIOC-YTHDC2-RBM46 in late mitotic spermatogenic cells derepresses E2F6-MGA-targeted genes, including Meiosin and other meiosis-associated genes. In turn, these steps drive activation of the STRA8-MEIOSIN transcriptional regulator by retinoic acid during the mitosis-to-meiosis transition (Fig. 5). As STRA8-MEIOSIN is the molecular vehicle through which retinoic acid induces the meiotic G1/S transition, the regulation of the STRA8-MEIOSIN complex by MEIOC enhances competence of spermatogenic cells to initiate meiosis.

Here, we demonstrate that Meioc-null spermatogenic cells developmentally diverge from their wild-type counterparts during the meiotic G1/S transition, earlier than previously appreciated. MEIOC is required to derepress expression of meiotic genes, including the transcriptional regulator Meiosin. In turn, MEIOC elevates expression of genes targeted by STRA8-MEIOSIN, which drives meiotic initiation. Therefore, MEIOC enhances the competence of spermatogenic cells to activate the meiotic transcriptional regulator and initiate meiosis in response to retinoic acid.

We find that MEIOC-YTHDC2-RBM46 destabilizes its mRNA targets, as suggested previously based on analyses of transcript abundance (Hsu et al., 2017; Qian et al., 2022; Saito et al., 2022; Soh et al., 2017). Here we distinguished changes in transcription versus transcript stability using two approaches. First, scRNA-seq allowed us to identify and analyze spermatogenic cells impacted by MEIOC before the onset of major transcriptional changes. Second, we employed a specialized pipeline to distinguish between transcriptional rate and RNA stability in bulk RNA-seq data of preleptotene-enriched testes (Alkallas et al., 2017). These complementary methods confirmed that MEIOC-YTHDC2-RBM46 reduces the stability of its target transcripts. Transcript stability may be the primary mechanism by which MEIOC-YTHDC2-RBM46 regulates its targets, as a recent study using ribosome profiling found that YTHDC2 does not affect translation in postnatal testes (Saito et al., 2022).

MEIOC-YTHDC2-RBM46 elevates the abundance of meiosis-associated transcripts without binding them. Here, we have demonstrated that these transcript abundance changes are driven by changes in transcription (Fig. 3B). Further, we provide a two-step mechanism for these indirect effects on transcript abundance. First, MEIOC-YTHDC2-RBM46 binds to and destabilizes E2f6 and Mga mRNAs, which encode transcriptional repressors whose genomic targets are enriched for meiosis-associated genes (Dahlet et al., 2021; Kehoe et al., 2008; Kitamura et al., 2021; Maeda et al., 2013; Pohlers et al., 2005; Suzuki et al., 2016; Uranishi et al., 2021). This regulation impacts gene expression and potentially differentiation, beginning in late mitotic spermatogonia, when MEIOC begins to destabilize Mga (Fig. 3D). Second, this MEIOC-YTHDC2-RBM46-mediated inhibition of E2f6 and Mga derepresses Meiosin gene expression and activates STRA8-MEIOSIN. This transcriptional regulator then activates meiotic gene expression and the meiotic G1/S transition. Thus, the inhibition of a transcriptional repressor by MEIOC-YTHDC2-RBM46 confers developmental competence to initiate meiosis.

MEIOC is also required in meiotic oocytes for progression through early meiotic prophase I (Abby et al., 2016; Soh et al., 2017), and it remains an unanswered question whether MEIOC supports activation of STRA8-MEIOSIN during meiotic initiation in premeiotic oogonia. There are some molecular differences in meiotic initiation between oogonia and spermatogenic cells. At stages when wild-type germ cells are undergoing meiotic DNA replication, Stra8-null oogonia fail to initiate any DNA replication, whereas Stra8-null spermatogenic cells initiate a DNA replication that is non-meiotic in nature, as meiotic cohesin REC8 is not loaded onto the chromosomes (Anderson et al., 2008; Baltus et al., 2006; Dokshin et al., 2013). STRA8 has the additional function of sequestering RB1, a key regulator of the G1/S transition, to promote a timely meiotic G1/S transition in oogonia, but this activity does not appear to impact spermatogenesis (Shimada et al., 2023).

In mitotic oogonia, loss of PRC1 activity via genetic ablation of Ring1 and Rnf2 causes premature meiosis due to precocious expression of Stra8 and other meiosis-associated genes (Yokobayashi et al., 2013). In addition, genetic ablation of Max, which encodes a subunit of PRC1.6, induces a similar phenotype with precocious expression of Stra8, Meiosin and meiosis-associated genes (Suzuki et al., 2024). During spermatogenesis, PRC1.6 may also regulate gene expression and meiotic initiation, but the germline roles for PRC1.6, E2F6 or MGA remain uncharacterized. Although E2f6-null mice are fertile, spermatogenesis was reportedly disrupted, without detailed characterization (Storre et al., 2002). Genetic ablation of Mga causes embryonic lethality, precluding analysis of spermatogenesis (Burn et al., 2018; Washkowitz et al., 2015). Further studies will be required to characterize the roles of E2F6 and MGA in spermatogenesis, including whether loss of E2F6 and MGA rescues the developmental competence of Meioc-null spermatogenic cells. Other transcriptional repressors targeted by MEIOC-YTHDC2-RBM46 may also control Meiosin gene expression.

As some Meioc-null spermatogenic cells upregulate Meiosin gene expression and enter meiotic prophase I on a delayed timeline, MEIOC is not strictly required for competence to initiate meiosis. Perhaps residual activity of YTHDC2 and RBM46 in the absence of MEIOC can support meiotic initiation in some cells. Consistent with this possibility, YTHDC2 protein is still expressed in Meioc-null spermatogenic cells (Abby et al., 2016; Soh et al., 2017), but whether loss of MEIOC affects RBM46 protein expression remains unclear. Alternatively, acquisition of competence mediated by the SCF complex may enable meiotic initiation on this delayed timeline. Regardless, we have found that the Meioc-null, Ythdc2-null or Rbm46-null phenotype of delayed entry into meiotic prophase I is the result of reduced STRA8-MEIOSIN transcriptional activation and is less severe than the arrest at the G1/S transition, before meiotic prophase I, exhibited by Stra8-null or Meiosin-null spermatogenic cells on an inbred C57BL/6 background. Intriguingly, Stra8-null spermatogenic cells on a mixed genetic background exhibit the less severe phenotype seen with Meioc-null, Ythdc2-null or Rbm46-null spermatogenic cells. Our analyses of the Meioc-null phenotype suggest that, on a mixed genetic background, Stra8-null spermatogenic cells transcriptionally activate some meiotic gene expression, perhaps due to MEIOSIN acting as a homodimer, but further studies will be required to test this possibility.

Our model of mammalian meiotic initiation exhibits parallels to the molecular network that governs meiotic initiation in budding yeast. In yeast, inhibition of a transcriptional repressor (Rme1p) via mRNA destabilization activates expression of the key transcription factor (Ime1p) that governs meiotic entry (Blumental-Perry et al., 2002; Bushkin et al., 2019; Covitz et al., 1991; Kahana et al., 2010; Kassir et al., 1988; Mitchell and Herskowitz, 1986; Smith et al., 1990). We conclude that in both unicellular eukaryotes and multicellular organisms, destabilization of a transcriptional repressor at the transcript level controls activation of the meiotic transcriptional program.

We have placed transcriptional activation by STRA8-MEIOSIN and post-transcriptional repression of mRNA by MEIOC-YTHDC2-RBM46 in a positive-feedback loop that facilitates the mitosis-to-meiosis transition (Fig. 5). This model generates a new question: how is MEIOC-YTHDC2-RBM46 activated before retinoic acid activates STRA8-MEIOSIN, particularly when E2F6 and MGA repress Meioc gene expression (Fig. 5)? One possibility is illustrated by the Drosophila MEIOC homolog Bam, which (translationally) represses its mRNA targets in late mitotic spermatogonia to facilitate the transition from mitosis to meiosis (Insco et al., 2009, 2012). This transition requires that Bam protein accumulates to a critical threshold level (Insco et al., 2009). Whether mammalian MEIOC protein levels must also clear a critical threshold to activate the MEIOC-YTHDC2-RBM46 complex requires further investigation.

In conclusion, by destabilizing its mRNA targets, MEIOC-YTHDC2-RBM46 inhibits transcriptional repressors E2F6 and MGA, thereby allowing spermatogenic cells to activate Meiosin expression in response to retinoic acid. In turn, this activates the key meiotic transcriptional regulator STRA8-MEIOSIN, which amplifies expression of meiosis- and cell cycle-associated genes, and drives the meiotic G1/S transition. In total, the post-transcriptional activity of MEIOC-YTHDC2-RBM46 enhances the activity of the meiotic transcriptional regulator. This regulatory pathway, acting in parallel with SCF complex-mediated degradation of DMRT1, enhances the competence of spermatogenic cells to initiate meiosis in response to retinoic acid.

All experiments involving mice were performed in accordance with the guidelines of the Massachusetts Institute of Technology (MIT) Division of Comparative Medicine and Cincinnati Children's Hospital Medical Center (CCHMC) Division of Veterinary Services, which are overseen by their respective Institutional Animal Care and Use Committees (IACUC). The animal care programs at MIT/Whitehead Institute and CCHMC are accredited by the Association for Assessment and Accreditation of Laboratory Animal Care, International (AAALAC) and meet or exceed the standards of AAALAC as detailed in the Guide for the Care and Use of Laboratory Animals. This research was approved by the MIT IACUC (0617-059-20) and CCHMC IACUC (2022-0061).

Mice carrying the Meioc-null allele Meioc^tm1.1Dpc (Soh et al., 2017) were backcrossed to C57BL/6N (B6N) from Taconic Biosciences for at least 10 generations. Mice used for scRNA-seq experiments were also heterozygous for Hspa2^tm1Dix (RRID: IMSR_JAX:013584; Dix et al., 1996) and homozygous for Gt(ROSA)26Sor^{tm9(CAG-tdTomato)Hze} (ROSA26^tdTomato; RRID: IMSR_JAX:007909; Madisen et al., 2010) with the floxed stop codon intact; both of these genotypes exhibit normal spermatogenesis.

Single-cell sequencing libraries were prepared and sequenced in two batches, with each batch containing one wild-type and one Meioc-null pup at P15. One P15 testis per pup was enzymatically dissociated into single cells (see supplementary Materials and Methods) and resuspended in 0.05% bovine serum albumin (BSA) in phosphate-buffered saline (PBS) for a target concentration of 1000 cells per microliter. Cell suspensions were loaded onto the Chromium Controller, aiming for recovery of 10,000 cells per sample. Libraries were generated using the Chromium Next GEM Single Cell 3′ v3.1 (10x Genomics), according to manufacturer's instructions, and sequenced as 150 bp paired-end reads on an Illumina NovaSeq 6000 system with an S4 flow cell.

Alignment, filtering, barcode counting and UMI counting were carried out using the count function in Cell Ranger v.4.0.0 with default settings using Cell Ranger's mm10-2020-A reference package (i.e. the GRCm38/mm10 mouse genome assembly with GENCODE vM23/Ensembl 98 annotation). Using Seurat v.3.2.3 (Stuart et al., 2019), cells were filtered for less than 10% mitochondrial reads, more than 1000 detected features and a doublet score (generated by the bcds in scds v.1.2.0) of less than 0.4. Using protein-coding genes, UMI counts from both wild-type and Meioc-null samples were integrated and clusters identified using the first 30 dimensions.

The wild-type samples were used to assign cell types to clusters. Five somatic cell types were identified based on cell type-enriched gene expression: fetal Leydig cells (Dlk1 and Cyp11a1) (Kaftanovskaya et al., 2015; Ye et al., 2017), peritubular myoid cells (Acta2 and Myh11) (Chen et al., 2016; Cool et al., 2008), vascular endothelium (Tm4sf1) (Shih et al., 2009), testicular macrophages (Cd14, Adgre1 and Itgam) (Bhushan et al., 2020; Landmann et al., 2000) and Sertoli cells (Sox9 and Cldn11) (Mazaud-Guittot et al., 2010; Sekido et al., 2004). Germ cells were identified via Ddx4 and Dazl. Then, using additional cell type-enriched gene expression along with UMAP-based cluster relationships, subclusters of spermatogenic cells were assigned to the following cell types: spermatogonial stem cells (Id4, Gfra1 and Etv5) (He et al., 2007; Helsel et al., 2017; Sun et al., 2015; Wu et al., 2011), undifferentiated spermatogonia (Zbtb16) (Hobbs et al., 2012), type A spermatogonia (Kit, Stra8 and Ccnd2) (Beumer et al., 2000; Endo et al., 2015; Schrans-Stassen et al., 1999; Yoshinaga et al., 1991), intermediate/type B spermatogonia (Kit) (Schrans-Stassen et al., 1999; Yoshinaga et al., 1991), preleptotene spermatocytes (Stra8), leptotene and zygotene spermatocytes (Meiob) (Chen et al., 2018; Souquet et al., 2013) and pachytene spermatocytes (Piwil1) (Chen et al., 2018; Deng and Lin, 2002).

Spermatogenic clusters were further refined based on cell cycle phase using the CellCycleScoring function in Seurat to identify type B spermatogonia in G2/M phase, as well as preleptotene spermatocytes in G1, early S and late S phases (Fig. S1C,D). Clusters were merged as needed. Cell cycle designations of the preleptotene clusters were confirmed via gene expression patterns in wild-type cells independently of the gene:cell cycle phase pairings used by Seurat's CellCycleScoring function (Fig. S3). These final cell type designations were then applied to the Meioc-null samples. The mutant-only (Mut) cluster was primarily composed of Meioc-null spermatocyte cells. Four wild-type germ cells assigned to the Mut cluster (Table S8) were considered to be an scRNA-seq artifact and were thus excluded from subsequent analysis.

In addition to the somatic and germ cell clusters shown in Fig. S1A,B, one additional somatic cell cluster and one additional germ cell cluster were identified but could not be assigned a cell type using marker expression. This unassigned germ cell cluster is likely a technical artifact as it exhibits a median number of features lower than that of all other assigned germ cell subpopulations (Fig. S2A). The unassigned somatic and germ cell clusters, along with all other somatic clusters, were excluded from subsequent analyses. See supplementary Materials and Methods for additional details on cluster-specific enrichment or depletion in wild-type cells and Gene Set Enrichment Analysis (GSEA).

Pseudotime trajectories were built in Monocle 3 v1.0.0 (Qiu et al., 2017a,b; Trapnell et al., 2014) using the following procedure: the Seurat object containing germ cell clusters only was imported as a Monocle object; data were normalized and pre-processed [function preprocess_cds(); options: num_dim=100]; batch effects were removed [function align_cds(); options: alignment_group=’batch’]; dimensionality reduction was carried out via UMAP [function reduce_dimension()]; cells were clustered [function cluster_cells()]; a principal graph was learned from the reduced dimension space [function learn_graph()]; and cells were ordered by selecting the Undiff cluster as the starting branch of the trajectory [function order_cells()]. Cluster cell-type assignments made in Seurat were maintained in the pseudotime trajectories.

For differential expression analysis between wild-type and Meioc-null germ cells, log₂ fold change was defined as wild-type over Meioc-null germ cells, such that the value reflects the activity of MEIOC in the unperturbed wild-type state. Differential expression analysis of scRNA-seq data was carried out on each germ cell cluster on genes with a minimum absolute log₂ fold change of 0.1 and that were detected in at least 25% of either population using Seurat's FindMarkers function (options: logfc.threshold=0.1, min.pct=0.25); P values were adjusted for multiple hypothesis testing of tested genes across all nine germ cell clusters via the Bonferroni correction using the P.adjust function in R. Additional expressed genes (i.e. genes detected in at least 25% of cells in wild-type or Meioc-null cells per cluster) that failed to meet the log₂ fold change threshold for differential expression testing were identified via the FindMarkers function (options: logfc.threshold=0, min.pct=0.25) and their P values were marked as ‘nd’ for differential expression testing ‘not done’ (see supplementary Materials and Methods for additional details on Gene Ontology analysis, cell cycle analysis of differentially expressed genes and de novo motif analysis). Dot plots were generated using Seurat's DotPlot function with parameter scale=FALSE to maintain average expression from wild-type and Meioc-null samples on the same scale. Code for scRNA-seq analysis is available on GitHub (https://github.com/Mikedis-Lab/2024_Meioc_spermatogenesis/).

Spermatogenesis was synchronized using a protocol originally developed by Hogarth et al. (2013) and modified by Romer et al. (2018) (see supplementary Materials and Methods for details).

Total RNA was extracted from four wild-type samples and three Meioc-null samples, using 1.5 synchronized testes from a single pup per sample. TRIzol Reagent (Thermo Fisher Scientific) was added to freshly thawed whole testes, and ERCC RNA ExFold RNA Spike-In Mix 1 and 2 (Thermo Fisher Scientific) were added to wild-type and Meioc-null samples, respectively, at a concentration of 1 µl of a 1:100 dilution of spike-in mix per 1 mg of testis tissue. (Spike-in mixes were ultimately not used for data analysis.) Total RNA was then isolated with chloroform following the manufacturer's protocol, precipitated using isopropanol and resuspended in RNase-free water. RNA-seq libraries were prepared with the TruSeq Stranded Total RNA kit with the Ribo-Zero Gold rRNA Removal kit. The barcoded libraries were pooled and sequenced with 50 bp single-end reads on an Illumina HiSeq 2500 machine.

Reads were quality trimmed using cutadapt v1.8 (options: -q 30 --minimum-length 20 -b AGATCGGAAGAGC). Expression levels of all transcripts in the mouse Gencode Basic vM15 gene annotation were estimated using kallisto v0.44.0 (Bray et al., 2016) with sequence-bias correction (--bias) and strand-specific pseudoalignment (--rf-stranded). Quantified transcripts were filtered for protein-coding genes, transcript-level estimated counts and transcripts per million (TPM) values were summed to the gene level, and TPMs were renormalized to transcript-per-million units.

To identify the MEIOC-dependent differential expression program, read counts from kallisto were rounded to the nearest integer and then supplied to DESeq2 v1.26.0 (Love et al., 2014). Genes were filtered for a minimum TPM of 1 in at least three out of seven samples. Differential expression was defined using a cutoff of adjusted P value <0.05. See supplementary Materials and Methods for additional details on cell cycle analysis, Gene Ontology analysis and de novo motif analysis.

For analysis of transcript stability and transcriptional rates, reads were mapped to the mouse genome (mm10) with the GENCODE Basic vM15 gene annotation via STAR v2.7.1a (Dobin et al., 2013) (options: --outFilterMultimapNmax 1 --alignEndsType Extend5pOfRead1 --outFilterMismatchNmax 2 --outSAMattributes None). All other parameters were set to default. Counts were quantified by htseq v0.11.0 (options: -m union --stranded=reverse) at the gene level (-type gene) and exon level (-type exon), and intron levels were calculated as gene-level counts minus exon-level counts. Changes in transcript stability, transcriptional rate and abundance were calculated for each gene for each sample using REMBRANDTS (Alkallas et al., 2017) with a stringency of 0.80 and linear bias mode.

The cumulative distribution of the log₂ fold change (wild type/Meioc knockout) in transcript stability or transcriptional rate in MEIOC-upregulated or -downregulated transcripts was compared with expressed but unchanged transcripts via a two-tailed Wilcoxon rank sum test with Bonferroni correction for multiple hypothesis testing using the wilcox.test and P.adjust functions in R. MEIOC-upregulated genes were defined as log₂ fold change wild type/Meioc knockout>0 and P<0.05 and MEIOC-downregulated genes were defined as log₂ fold change wild type/Meioc knockout<0 and P<0.05 based on the DESeq2 analysis of the bulk RNA-seq data.

RNA-seq data from Stra8-null and Stra8-heterozygote (i.e. phenotypically wild-type) preleptotene spermatocytes (NCBI GEO GSE115928; Kojima et al., 2019) were re-analyzed. Genes with STRA8-bound promoters identified by ChIP-seq in testes enriched for preleptotene spermatocytes were extracted from Kojima et al. (2019) (supplementary file 2: https://doi.org/10.7554/eLife.43738.028) (see supplementary Materials and Methods for details).

MEIOC RIP-seq from P15 testes (NCBI GEO GSE96920; Soh et al., 2017) were re-analyzed, with additional details provided in the supplementary Materials and Methods. YTHDC2-bound mRNAs identified via CLIP-seq in P8 and P10 testes were extracted from Saito et al. (2022) (Tables S1 and S2: http://genesdev.cshlp.org/content/suppl/2022/01/19/gad.349190.121.DC1/Supplemental_Tables.xlsx). RBM46 eCLIP data from P12-P14 testes (NCBI GEO GSE197282) (Qian et al., 2022) were re-analyzed, with additional details provided in the supplementary Materials and Methods. See supplementary Materials and Methods for details on target comparison to MEIOC scRNA-seq and bulk RNA-seq datasets.

To identify transcriptional repressors that are inhibited by MEIOC-YTHDC2-RBM46, we first identified transcripts that were (1) downregulated by MEIOC in the In/B, B G2M and/or pL G1 clusters; and (2) targeted by MEIOC-YTHDC2-RBM46. We then manually examined the ‘Function’ description of the UniProtKB database (release 2020_01; www.uniprot.org) for these mouse proteins. Any proteins that were annotated as repressing transcription or gene expression were considered to be transcriptional repressors.

The following datasets were re-analyzed here: E2F6 ChIP-seq and input data from wild-type and E2f6 knockout mouse ESCs (NCBI GEO GSE149025) (Dahlet et al., 2021); MGA and IgG ChIP-seq from wild-type mouse ESCs (ArrayExpress E-MTAB-6007) (Stielow et al., 2018); mouse ESC RNA-seq data from wild-type and E2f6 knockout samples (NCBI GEO GSE149025) (Dahlet et al., 2021); and wild-type and Mga knockout samples (NCBI GEO GSE144141) (Qin et al., 2021). See supplementary Materials and Methods for further details.

Immunostaining was carried out in tissue sections using the following primary antibodies: anti- DMRT1 (Santa Cruz Biotechnology sc-377167, 1:200 dilution for fluorescent staining); anti-MEIOSIN (guinea pig polyclonal from Ishiguro et al., 2020, 1:100 dilution for fluorescent staining); anti-STRA8 (Abcam ab49405, 1:500 dilution for chromogenic staining or 1:200 for fluorescent staining); and anti-STRA8 (Abcam ab49602, 1:200 dilution for fluorescent staining). See supplementary Materials and Methods for further details.

We thank Kei-Ichiro Ishiguro for the MEIOSIN antibody; Mary Goodheart, Mina Kojima, Holly Christensen and Peter Nicholls for technical help; and Abigail Groff and Linyong Mao for advice on bioinformatics analysis. We are grateful to the members of the Page Lab for valuable feedback, and to Winston Bellott, Jordana Bloom, Z. Y. Chen, Jennifer Hughes, Mina Kojima and Adrianna San Roman for comments on the manuscript. We thank the Whitehead Genome Technology Core for library preparation and Illumina sequencing, and the Whitehead Institute FACS Facility for cell sorting. Fluorescent microscopy was carried out at the Whitehead Institute's WM Keck Foundation Biological Imaging Facility and Cincinnati Children's Bio-Imaging and Analysis Facility (RRID: SCR_022628).

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.202740.reviewer-comments.pdf