The microRNA miR-202 prevents precocious spermatogonial differentiation and meiotic initiation during mouse spermatogenesis

Spermatogenesis begins with the proliferation and differentiation of spermatogonia and ends with a large population of spermatozoa. Mouse spermatogonia undergo eight to nine consecutive mitotic divisions to generate many subtypes, including type A (SG-A), intermediate (SG-I) and type B (SG-B) spermatogonia (de Rooij and Russell, 2000; Griswold, 2016). Consequently, it is estimated that one spermatogonial stem cell (SSC) produces 2048 or 4096 spermatozoa by the end of spermatogenesis (de Rooij and Russell, 2000). These subtypes can also be roughly divided into undifferentiated (SG_undiff) and differentiating (SG_diff) spermatogonia. Whereas SG_undiff are marked by the expression of PLZF (ZBTB16) (Buaas et al., 2004; Costoya et al., 2004), SG_diff are positive for KIT, a well-known marker for transit-amplifying cells in some other tissues (Besmer et al., 1993). SSCs are a functionally defined subpopulation of SG_undiff without specific markers (La et al., 2018; Yoshida, 2019). The transition from SG_undiff to SG_diff occurs cyclically, resulting in the concurrent presence of several types of cells originating from overlapping cycles to form a fixed number of unique associations (12 in mice) known as seminiferous stages (Russell et al., 1990). Spermatogonial differentiation ends with the mitosis-to-meiosis transition, also known as meiotic initiation (de Rooij and Russell, 2000; Griswold, 2016). Disruption of spermatogonial differentiation reduces the sperm population (Hobbs et al., 2012; Maezawa et al., 2017). By contrast, precocious differentiation/meiosis leads to exhaustion of the spermatogonial pool (Matson et al., 2010).

Spermatogonial development is under the control of both extracellular and intracellular factors. For example, the low level of glial cell line-derived neurotrophic factor (GDNF) in Gdnf^+/− mice was found to result in seminiferous tubules containing only Sertoli cells, whereas overexpression of GDNF in transgenic mice resulted in overproliferation of spermatogonia and testicular tumors (Meng et al., 2000). Retinoic acid (RA) is the key molecule that induces spermatogonial differentiation and meiotic initiation by activating the expression of a large number of genes (Bowles et al., 2006; Koubova et al., 2006; Anderson et al., 2008; Endo et al., 2015; Wang et al., 2016; Griswold and Hogarth, 2018). The stimulated by retinoic acid gene 8 (Stra8) is a direct target and mediator of RA signaling, and plays an essential role in meiotic initiation in both males and females (Oulad-Abdelghani et al., 1996; Anderson et al., 2008; Griswold and Hogarth, 2018; Ishiguro et al., 2020). STRA8 has also been shown to promote spermatogonial differentiation (Endo et al., 2015).

A number of regulators for spermatogonial differentiation and/or meiotic initiation have also been identified, including transcriptional/epigenetic regulating factors such as SOX3 (Raverot et al., 2005), SOHLH1 (Ballow et al., 2006), SOHLH2 (Hao et al., 2008), DMRT1 (Matson et al., 2010), SALL4 (Hobbs et al., 2012), DMRT6 (DMRTB1) (Zhang et al., 2014), MAX (Suzuki et al., 2016), PRC1 (Maezawa et al., 2017) and MEIOSIN (Ishiguro et al., 2020), RNA-binding proteins such as NANOS2 (Suzuki and Saga, 2008), DAZL (Lin et al., 2008), AGO4 (Modzelewski et al., 2012) and BCAS2 (Liu et al., 2017), and the ubiquitylation-related protein β-TrCP (Nakagawa et al., 2017). DMRT proteins are transcription factors, and some members have been reported to participate in multiple steps of mammalian spermatogenesis (Zhang and Zarkower, 2017). DMRT1 is required for the establishment and maintenance of SSCs (Zhang et al., 2016), and prevents the precocious activation of the mitosis-meiosis switch (Matson et al., 2010). DMRT6 acts in SG_diff to coordinate an orderly transition from the mitotic program to the meiotic program (Zhang et al., 2014).

MicroRNAs (miRNAs) are believed to be essential for mammalian spermatogenesis based on the infertile phenotypes of gene knockout (KO) mice, including KOs of Dicer (Dicer1), Drosha and Dgcr8, which encode key regulators of miRNA biogenesis (Hayashi et al., 2008; Maatouk et al., 2008; Korhonen et al., 2011; Romero et al., 2011; Greenlee et al., 2012; Wu et al., 2012; Zimmermann et al., 2014; Modzelewski et al., 2015; Hilz et al., 2016). In somatic stem/progenitor cells, miRNA function is intricately regulated to promote and stabilize cell fate determination (Shenoy and Blelloch, 2014). Whereas global miRNA loss resulting from knock out or mutations of key regulators in miRNA biogenesis induces dramatic phenotypic changes in almost all examined tissues, mice with individual miRNA KOs often lack dramatic phenotypic consequences, implying that miRNAs act in a redundant manner (Park et al., 2012; Shenoy and Blelloch, 2014). However, there have been no reports on the in vivo functions of single miRNAs in spermatogenesis based on mouse genetic studies.

The intergenic miRNA gene miR-202 (Mir202) belongs to the highly conserved let-7 family (Roush and Slack, 2008) and generates miR-202-3p and miR-202-5p, which are highly expressed in mouse testes (Chen et al., 2017). We previously reported that miR-202 plays an important role in maintaining the stem cell state of cultured mouse spermatogonia (Chen et al., 2017). In the present study, we investigated the in vivo function of miR-202 in spermatogonial differentiation and meiotic initiation using KO mice. We found that both spermatogonial differentiation and meiotic initiation occurred precociously upon miR-202 knock out. We also identified that SYCP3, STRA8 and DMRT6 were aberrantly and prematurely expressed in the absence of miR-202, and found that DMRT6, the mRNA of which is a direct target of miR-202-5p, mediated the function of miR-202. Our results contribute to our understanding of how a single miRNA safeguards spermatogonial differentiation and meiotic initiation.

The miR-202 KO mice were produced by CRISPR-Cas9 technology, as described in detail in another study (Chen et al., 2021 preprint). No apparent defects were observed in the seminiferous tubules of KO mice at 2 months of age, and the germ cell loss and fertility reduction became progressively more severe in an age-dependent manner (Fig. 1A,B, Fig. S1A,B). At 12 months of age, about 2% of the KO seminiferous tubules were agametic (Fig. 1A). This phenotype was reminiscent of that of Plzf KO mice. PLZF is required for maintenance of the SSC pool, as Plzf KO mice are depleted of germ cells in an age-dependent manner (Buaas et al., 2004; Costoya et al., 2004). Therefore, we examined changes in the PLZF⁺ SG_undiff in postnatal day (P) 9 and adult KO mice and found that the numbers of PLZF⁺ SG_undiff per tubule were comparable between KO and wild-type (WT) mice at P9 but reduced from 7.4 to 4.2 in KO littermates compared with adult WT mice (Fig. 1C, Fig. S1C).

As miR-202 is expressed in Sertoli cells (Chen et al., 2017), we also examined its function in Sertoli cells by immunostaining for the Sertoli cell marker WT1 and by transplanting WT SSCs into WT or KO recipient mice. We found no change in the number of WT1⁺ Sertoli cells or in the number of colonies formed by transplanted SSCs between WT and KO recipients, indicating that miR-202 knock out does not affect the function of Sertoli cells (Fig. 1D,E). Interestingly, the numbers of SG_diff, represented by KIT⁺ cells inside the tubules, were similar in KO and WT mice (Fig. S2A-C), indicating that the population of SG_diff was not changed in the KO mice. These results indicated that miR-202 knock out resulted in reduction of the SG_undiff pool in adults, independently of Sertoli cells, but had no effect on initial establishment of the SG_undiff population.

We next carried out RNA-sequencing (RNA-seq) analysis on SG-A, which were isolated using the STAPUT method (Gan et al., 2013), to investigate gene expression changes resulting from miR-202 knock out. The proportion of PLZF⁺ cells in the isolated SG-A was 92-94%, indicating that the majority of SG-A are SG_undiff (Fig. S3A). The differentially expressed genes (DEGs) between KO and WT mice included 146 upregulated and 721 downregulated genes (Fig. 1F, Table S1). The upregulated genes were enriched with gene ontology (GO) terms such as ‘spermatogenesis’, ‘sperm motility’, ‘cell differentiation’, ‘spermatid development’ and ‘binding of sperm to zona pellucida’ (Fig. 1G, Table S1). Gene-set enrichment analysis (GSEA) also revealed that genes enriched in the GO term ‘sperm motility’ were significantly upregulated in KO SG-A (Fig. S3B). We compiled a list of genes expressed in spermatogenic cells (16,816; RPKM >0.1) based on one of our previous studies (Lin et al., 2016) and a list of genes expressed in THY1⁺ SG_undiff (15,800; RPKM >0.1) based on a study by others (Maezawa et al., 2020). The intersection between these two lists of genes contained 13,690 genes. We found that 88% of our upregulated genes and 94% of the downregulated genes in the miR-202 KO mice are in this intersection gene set. This result supports the conclusion that the majority of the DEGs are expressed in the isolated SG-A.

We conducted qRT-PCR to validate the downregulation of genes involved in SG_undiff maintenance [including Gfra1 and Oct4 (Pou5f1)], and the upregulation of genes promoting spermatogonial differentiation and meiotic initiation/progression (such as Sohlh1, Sohlh2, Kit, Stra8, Dazl, Spo11, Sycp3, Sycp1, Dmc1, Rad51 and Mlh1) in KO SG-A (Fig. 1H) (La et al., 2018). As miR-202 lies within the long non-coding (lnc)RNA Gm2044-201, we also examined whether the deletion of miR-202 might perturb lncRNA Gm2044-201 expression. As shown in Fig. S1D, lncRNA Gm2044-201 is not misregulated in KO mice. These results indicate that miR-202 knock out disrupts expression of differentiation- and meiosis-related genes in spermatogonia.

We then examined the timing of meiotic initiation using pre-pubertal mice, in which the first wave of spermatogenesis occurs in a synchronized manner (de Rooij and Russell, 2000). In WT mice at P9, only ∼20% of the tubules contained cells expressing SYCP3, which marks meiotic spermatocytes (Mahadevaiah et al., 2001), whereas in the KO mice, about 72% of tubules contained SYCP3⁺ cells (Fig. 2A, Fig. S4A). SYCP3 and γH2AX co-staining of chromosomal spreads (Mahadevaiah et al., 2001) from mice at P9 showed that the most advanced spermatogenic cells in the WT mice were leptotene spermatocytes, whereas 23% of spermatocytes were zygotene spermatocytes in the KO mice at this time (Fig. 2B). These results showed that meiosis was initiated precociously in KO mice.

We also examined the expression pattern of STRA8 as it is a key factor promoting spermatogonial differentiation and meiotic initiation (Anderson et al., 2008; Mark et al., 2008; Endo et al., 2015; Ishiguro et al., 2020). Whereas 31% tubules in WT mice were STRA8⁺, this percentage increased to 68% in KO mice at P9 (Fig. 2C, Fig. S4B). The number of tubules with robust STRA8 expression in adult KO mice was also significantly increased compared with the WT littermates (25% versus 17%; Fig. 2D). The upregulation of STRA8 was confirmed by western blot analyses (Fig. 2E,F).

We next examined whether STRA8 was aberrantly expressed in SG_undiff by co-immunostaining of STRA8 and PLZF. STRA8 is expressed at low levels in spermatogonia and at greatly elevated levels in preleptotene spermatocytes (plpSCs), in which the mitosis-meiosis transition occurs (Zhou et al., 2008). STRA8 expression overlaps with PLZF in stages VII-VIII, and is limited to PLZF-low and -negative spermatogonia thereafter in stages IX-X (Endo et al., 2015). We found that more PLZF⁺ cells became STRA8⁺ cells in KO than in WT at P9 and 4 months of age (Fig. 3A-D). Moreover, in adult mice, STRA8 expression also overlapped with PLZF in stages I-VI and IX-XII in KOs, showing the widespread and early presence of STRA8 independent of seminiferous stages (Fig. 3C,E). Taken together, the loss of miR-202 in SG_undiff caused precocious spermatogonial differentiation and meiotic initiation.

As miR-202 KO mice are similar to Dmrt1 KO mice in terms of precocious meiosis initiation (Matson et al., 2010), we examined whether DMRT1 expression was changed in our KO mice. This turned out not to be true (Fig. S5A,B), suggesting that the function of miR-202 may not be mediated by DMRT1. We next focused on Dmrt6 because: (1) Dmrt6 plays a crucial role in the mitosis/meiosis switch (Zhang et al., 2014); (2) its 3′ UTR is predicted to contain an miR-202-5p-binding site (Paraskevopoulou et al., 2013); and (3) its expression was upregulated in miR-202 KO SG-A (Fig. 1F,H).

We examined whether miR-202-5p directly targeted the Dmrt6 mRNA as predicted. First, we tested this using a dual luciferase assay, in which a putative binding sequence was fused to the luciferase gene and the plasmid construct was transfected into 293FT cells, and the luciferase activities expressed by this construct were compared with that by a control construct with a mutant binding site. Our results showed that Dmrt6 mRNA was indeed a direct target of miR-202-5p (Fig. 4A). Second, the targeting of Dmrt6 mRNA by miR-202-5p was also validated in cultured SSCs using a biotin-labeled miRNA pull-down assay, in which biotinylated miRNA mimics were used to pull down targets that were subsequently detected by qRT-PCR (Fig. 4B,C) (Lal et al., 2011).

We found that the KO testes contained more DMRT6⁺ tubules (74%) than did WT (28%) at P9 (Fig. 4D, Fig. S4C). The adult KO mice also contained more DMRT6⁺ tubules than the WT littermates (75% versus 68%; Fig. 4E), and the upregulation of DMRT6 was also confirmed by western blot analyses in P9 and adult testes (Fig. 2E,F). To examine whether DMRT6 was prematurely expressed in SG_undiff like STRA8, we co-immunostained DMRT6 and PLZF in P9 and adult mice (Fig. 5). Consistent with a previous study (Zhang et al., 2014), we found that DMRT6 and PLZF did not colocalize in germ cells of WT mice (Fig. 5A,B). In contrast, they colocalized in 24% and 37% of SG_undiff in P9 and adult mice, respectively (Fig. 5A,B), indicating the premature expression of DMRT6 in these cells. We also found that DMRT6 expression overlapped with SYCP3 in P9 KO testes (Fig. S6), suggesting that the expression of SYCP3 might also be premature.

To test further our hypothesis that the miR-202/DMRT6 axis regulates the initiation of meiosis, we took advantage of SSC lines established from the miR-202 KO mice and their WT littermates, which have been described in detail in a separate study (Chen et al., 2021 preprint). The establishment rate of KO mice was significantly lower than that of the WT littermates (3/7 versus 11/11; Chi-squared test, P<0.005). The cells of these SSCs were more appropriately named SG_undiff because only a small fraction were bona fide SSCs based on transplantation assays (Kanatsu-Shinohara et al., 2003; Kubota et al., 2004). Using these cell lines, we first assessed proliferation, and found that KO SG_undiff were more mitotically active than WT SG_undiff based on cell counting and bromodeoxyuridine (BrdU) incorporation assays (Fig. 6A, Fig. S7A,B). However, KO SG_undiff also exhibited a higher apoptotic rate, as revealed by terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assays (Fig. 6B, Fig. S7C). Interestingly, we did not detect differences in the amounts of mitotic and apoptotic germ cells between WT and KO P9 testes in vivo (Fig. S8A,B), probably because the proportion of SG_undiff among all germ cells in the seminiferous tubules is too low. Alternatively, the non-natural in vitro growth condition tends to amplify the differences between KO and WT spermatogonia. When we compared the stem cell activity of WT and KO SG_undiff using transplantation assays (Fig. 6C), we found that KO SG_undiff exhibited a 25% lower stem cell activity than WT (Fig. 6D), consistent with our previous report (Chen et al., 2017). These results indicated that miR-202 facilitates the maintenance of a more stem cell-like state of SG_undiff that possesses lower mitotic and apoptotic rates and higher stem cell activity.

We next investigated the mechanisms of miR-202 function using an in vitro meiosis-initiation model, whereby SG_undiff cultures can be induced to initiate meiosis when they are plated on a feeder layer of neonatal Sertoli cells (Wang et al., 2016). We monitored differentiation and meiotic initiation by examining the expression of KIT and SYCP3 (Fig. 6E, Fig. S7D). One day after induction, we found that 18% of the KO cells started to express KIT, whereas only 7% WT cells did so. On day 2 of induction, more KO cells than WT cells began to express SYCP3 (28% versus 12%; Fig. 6E, Fig. S7D). We further conducted co-staining of SYCP3 and γH2AX on chromosome spreads to examine the types of WT and KO SYCP3⁺ cells that were generated 6 days after induction, and found that the most advanced spermatogenic cells in KO were zygotene spermatocytes, in contrast to leptotene spermatocytes in WT (Fig. 6F). These in vitro results again indicated that KO SG_undiff can be induced to initiate differentiation and meiosis more readily, and that miR-202 deletion predisposes SG_undiff to a differentiation- and meiosis-primed state.

When we examined whether STRA8 and DMRT6 were prematurely expressed in KO SG_undiff that were PLZF⁺ (Fig. S7E), we did indeed detect both STRA8 and DMRT6 in KO but not in WT cells, consistent with the in vivo results (Fig. 6G). We further tested whether the precocious meiotic initiation of miR-202 KO SG_undiffin vitro could be rescued by a miR-202-5p mimic or by the knockdown of Dmrt6 mRNA. We selected Dnmt3b (which is involved in DNA methylation) as a potential negative control based on evidence that: (1) Dnmt3b mRNA was upregulated in KO SG-A (Fig. 1H, Table S1); (2) Dnmt3b was a direct target of miR-202-3p in dual luciferase assays (Fig. S9A) (Agarwal et al., 2015); and (3) Dnmt3b conditional knock out in germ cells manifested no phenotypic abnormality in spermatogenesis (Kaneda et al., 2004). SG_undiff cultures were pretreated with miR-202-5p mimic or with siRNAs targeting Dmrt6 and Dnmt3b (Fig. S9B) before induction was conducted. On day 2 of induction, about 27% of the KO cells pretreated with control miRNA mimic or control siRNAs were SYCP3⁺, which was similar to the percentage without pretreatment (Fig. 6E,H,I); this percentage was reduced to ∼15% by pretreatment with the miR-202-5p mimic, which approximated the value observed in the WT cells pretreated with a control miRNA mimic (Fig. 6H). Moreover, knockdown of Dmrt6 with two different siRNAs also reduced the percentages of SYCP3⁺ cells significantly, whereas knockdown of Dnmt3b with either of two different siRNAs did not (Fig. 6I). These in vitro results thereby indicate that the function of miR-202 in inhibiting precocious meiotic initiation is mediated by its downstream target Dmrt6.

Spermatogonial differentiation and meiotic initiation must be precisely regulated to ensure both productive generation of spermatozoa and maintenance of the stem cell pool to support life-long spermatogenesis (Griswold, 2016; Griswold and Hogarth, 2018). However, only a small number of regulators for these two important processes have been identified. In the present study, we found that knock out of miR-202 resulted in precocious spermatogonial differentiation, meiotic initiation and a reduced SG_undiff pool. We also demonstrated that loss of miR-202 results in premature expression of STRA8 and DMRT6, two key factors promoting differentiation and meiosis (Oulad-Abdelghani et al., 1996; Anderson et al., 2008; Zhang et al., 2014; Endo et al., 2015; Griswold and Hogarth, 2018; Ishiguro et al., 2020) in SG_undiff, and found that Dmrt6 mRNA was directly targeted by miR-202-5p. Therefore, a module composed of miR-202, DMRT6 and STRA8 has been identified in the context of the spermatogonial fate-decision regulatory network (Fig. 7A).

As far as we know, our study is the first to show that knock out of a single miRNA gene in mice results in abnormal spermatogonial proliferation and differentiation. Original studies using Dicer conditional KO mice mostly failed to examine spermatogonial defects over extended time periods and by co-immunostaining of marker proteins such as PLZF and STRA8/DMRT6 (Hayashi et al., 2008; Maatouk et al., 2008; Korhonen et al., 2011; Romero et al., 2011). Besides, these studies came to different conclusions regarding the function of Dicer, varying from a role early in the differentiation process (in spermatogonia), slightly later (in meiosis) or later still (in spermatids), almost certainly resulting from varying Cre efficiencies and incomplete Dicer deletions. Non-deletant germ cells can often compensate for mutant, non-viable ones, masking potential spermatogonial defects. Hayashi et al. have previously reported that Dicer KO spermatogonia can be cultured in vitro for 2 weeks with reduced capacity for proliferation, suggesting that Dicer also plays a role in spermatogonial proliferation (Hayashi et al., 2008). Moreover, Tong et al. reported that two miRNA clusters, Mirc1 and Mirc3, are involved in the regulation of spermatogonial differentiation in mice, and that male germ cell-specific Mirc1 KO mice exhibit some tubules containing only Sertoli cells (Tong et al., 2012). Our results, together with these previous studies, suggest that miRNAs – either as a group or individually – are involved in spermatogonial fate decision. It is important to point out that the function of miRNAs in gametogenesis is not evolutionarily conserved as sperm and oocyte formation and reproduction are normal in dicer KO zebrafish (Giraldez et al., 2005; Houwing et al., 2007).

The precocious initiation of spermatogonial differentiation and meiotic initiation in miR-202 KO mice are supported by multiple lines of evidence, including the premature expression of SYCP3, STRA8 and DMRT6, and the upregulation of a number of meiotic and post-meiotic genes. In WT mice, the expression of STRA8 in SG_undiff and plpSCs is restricted in stages VII-VIII of the seminiferous cycle (Endo et al., 2015), whereas in KO mice STRA8 expression was observed in all stages. Moreover, more SG_undiff become STRA8⁺ in the absence of miR-202 (Fig. 3A-D). DMRT6 is normally expressed in SG_diff that are negative for PLZF expression (Zhang et al., 2014). However, in the absence of miR-202, the expression of DMRT6 was not only augmented but was also detected as early as in PLZF⁺ SG_undiff (Figs 4D,E and 5A,B)_. These results suggest that the precise control of spermatogonial differentiation and meiotic initiation was disrupted upon miR-202 knock out (Fig. 7A).

The regulators that govern spermatogonial differentiation and/or meiosis identified so far include transcription factors, epigenetic regulators, RNA-binding proteins, and proteins involved in RA pathways. Polycomb repressive complex 1 (PRC1) promotes spermatogonial differentiation by the timely activation of the expression of germline genes as knock out in mice of Rnf2, which encodes the catalytic subunit of PRC1, causes severe defects in spermatogonial differentiation and in gene expression (Maezawa et al., 2017). The effect of PRC1 is mediated by SALL4, a transcription factor that promotes spermatogonial differentiation by sequestering PLZF, thus preventing DNA binding (Hobbs et al., 2012; Maezawa et al., 2017). In contrast, MAX, one component of an atypical PRC1 complex (PRC1.6), prevents spermatogonial differentiation and meiotic onset as indicated by the meiosis-like cytological changes induced by its knockdown in cultured germline stem cells (Maeda et al., 2013; Suzuki et al., 2016). Additional meiotic initiation activators include DAZL and MEIOSIN, and suppressors include NANOS2, AGO4 and DMRT1. DAZL is a key regulator that facilitates meiotic initiation (Lin et al., 2008) and activates the expression of Stra8 by targeting the 3′ UTRs of Stra8 mRNA (Li et al., 2019). Interestingly, the expression of Dazl is repressed by NANOS2, which suppresses meiotic initiation, and is activated by BCAS2, which promotes the process. Both NANOS2 and BCAS2 are RNA-binding proteins (Suzuki and Saga, 2008; Kato et al., 2016; Liu et al., 2017). AGO4, a small RNA-binding protein, prevents precocious meiotic initiation, and its deletion results in loss of many miRNAs, including miR-202 (Modzelewski et al., 2012). It is likely that the function of AGO4 in meiotic initiation is at least partially mediated by miR-202. MEIOSIN interacts with STRA8 to activate meiotic initiation (Ishiguro et al., 2020). DMRT1 inhibits meiosis in SG_undiff by limiting RA-dependent transcription, and by specifically blocking Stra8 transcription at plpSCs (Matson et al., 2010). β-TrCP, a component of an E3 ubiquitin ligase complex, targets DMRT1 for degradation and thereby activates meiotic initiation (Nakagawa et al., 2017). The relationships of these regulators are illustrated schematically in Fig. 7B.

miR-202 KO mice phenocopy the mutants of DMRT1, NANOS2, MAX and AGO4 in that premature meiotic initiation is observed. However, there are some differences in the phenotypes. For example, spermatogonial differentiation in Dmrt1 KO mice is truncated resulting in SG_diff depletion (Matson et al., 2010), whereas this does not occur in miR-202 mutants as shown by the unaffected number of KIT⁺ SG_diff (Fig. S2A-C). As both spermatogonial differentiation and meiosis occur precociously and the former increases but the latter decreases the number of KIT⁺ cells, the net effect could be the unchanged population of KIT⁺ SG_diff in adult mice. This is also not at odds with the increased number of STRA8⁺ spermatogonia at P9, at which time the rate of spermatogonial differentiation is greater than that of meiosis initiation for the first wave spermatogenesis. NANOS2 suppresses meiosis and activates a male-specific genetic program in gonocytes, indicating that NANOS2 plays a key role earlier than miR-202. MAX prevents differentiation/meiosis via interaction with distinct co-factors in cultured cells, and its in vivo function warrants further investigation (Maeda et al., 2013; Suzuki et al., 2016).

Based on our findings and those of published studies, it can be seen that many regulators cooperate to regulate the fate decision of spermatogonial differentiation and meiotic initiation, and miR-202 and its direct target Dmrt6 as well as Stra8, the expression level and timing of which are also regulated by miR-202, act together as a module in this expanding regulatory network (Fig. 7B). One interesting research direction in the future is to elucidate how the expression and action of miR-202 are regulated by known or novel players in the network.

All of the animals used in this study were approved by the Animal Ethics Committee of the Institute of Zoology at the Chinese Academy of Science. All of the procedures were conducted in accordance with institutional guidelines. Animals were specific-pathogen free. All mice had access to food and water ad libitum, were maintained on a 12:12 h light-dark artificial lighting cycle, with lights off at 19:00, and were housed in cages at a temperature of 22-24°C.

The generation of the miR-202 KO mice using the CRISPR/Cas9 gene-editing approach has been described in detail in a separate study (Chen et al., 2021 preprint). Briefly, the miR-202 locus was targeted by two gRNAs (gRNA-1s: TTGCAGGGGATCAATCCTTCTGG; gRNA-1a: GCTAATGACTTTGTTTGGGTGGG; PAM sites of sgRNAs are underlined), which were designed and used in a previous study (Chen et al., 2017). Cas9 mRNA and these two sgRNAs were injected into zygotes of the C57BL/6J background. A mutant allele from a male founder mouse lacking a 155-bp fragment containing the whole transcribed region of miR-202 was passed at least to F3 mice for phenotypic evaluation. All of the mice were maintained on a C57BL/6J;ICR mixed background. The genotyping of the miR-202 KO mice was conducted using genomic PCR, and the primers used are listed in Table S2.

Testes dissected from WT and KO mice immediately after euthanasia were fixed in 4% paraformaldehyde (PFA) or Bouin's solution for up to 24 h, dehydrated using an increasing ethanol gradient, treated with xylene, and then embedded in paraffin. Five-micrometer-thick sections were cut and mounted on glass slides. After deparaffinization in xylene and re-hydration in a decreasing ethanol gradient, Bouin-fixed testis cross-sections were used for Hematoxylin and Eosin (HE) staining, and PFA-fixed sections were used for immunohistochemistry, immunofluorescence analyses and TUNEL assays.

The reproductive performance of KO mice was examined in another study (Chen et al., 2021 preprint). We re-analyzed the data by classifying the males into three types based on their ages of mating. Briefly, each male mouse was housed with 8-week-old C57BL/6J WT female mice. Female mice that were vaginal-plug positive were separated and observed for 3 weeks for pregnancies. The number of litters per female were recorded and calculated.

All antibodies used in this study are listed in Table S3. We obtained the anti-DMRT6 and anti-DMRT1 antibodies from Prof. David Zarkower in University of Minnesota (Matson et al., 2010; Zhang et al., 2014, 2016), as generous gifts.

Deparaffinized sections were boiled for 15 min in a sodium citrate buffer for antigen retrieval. Then, the slides were incubated with primary antibodies overnight at 4°C and then incubated with secondary antibodies for 2 h at room temperature. For immunofluorescence, signals were visualized using fluorophore-conjugated secondary antibodies. For immunohistochemistry, the sections were stained with a horseradish peroxidase-conjugated secondary antibody.

Mouse testes were dissected to remove the tunica albuginea, and seminiferous tubules were untangled. Tubules were fixed in 4% PFA overnight at 4°C, permeabilized with 0.1% Triton X-100 in PBS for 4 h at room temperature, and blocked with 5% bovine serum albumin (BSA) in PBS for 2 h at room temperature. The tubules were incubated with primary antibodies overnight at 4°C and subsequently with fluorophore-conjugated secondary antibodies for 4 h at room temperature. The nuclei were counterstained with DAPI.

Chromosome spreads of the testicular samples were performed using the drying-down technique previously described by Peters et al. (1997). Briefly, the testes were dissected, and the seminiferous tubules were washed in PBS. Then, tubules were placed in a hypotonic extraction buffer for 30-60 min. Subsequently, the tubules were torn to pieces in 0.1 M sucrose (pH 8.2) on a clean glass slide and were pipetted repeatedly to make a suspension. The cell suspensions were then dropped onto slides containing 1% PFA and 0.15% Triton X-100 (pH 9.2). The slides were dried for at least 2 h in a closed box with high humidity. Finally, the slides were washed twice with 0.4% Photo-Flo 200 (Kodak) and dried at room temperature. The dried slides were stored at −20°C for immunofluorescence staining. Immunolabeled chromosome spread nuclei were imaged on confocal laser-scanning microscopes (Leica TCS SP8 or Carl Zeiss LSM780) using a 63× oil-immersion objective.

We used TargetScan or microT-CDS to predict mRNA target sites of miR-202-3p and -5p and found that the 3′UTR of Dmrt6 mRNA contains a binding site for miR-202-5p (Paraskevopoulou et al., 2013; Agarwal et al., 2015). The predicted 3′ UTRs for candidate genes were amplified from mouse testis cDNA (primers are listed in Table S2) and inserted into the pMIR-REPORT luciferase vector. Binding sequence with four mutations in the seed region was used as negative control.

Dual luciferase assays were performed following our previous report (Chen et al., 2017). Briefly, 293FT cells were plated in a 96-well plate. About 24 h later, 100 nM miR-202 or scrambled negative control (NC) mimics were first transfected using the Lipofectamine RNAiMAX Reagent (Invitrogen). Ten hours later, 50 ng of the firefly luciferase report plasmids (pMIR-REPORT) and 5 ng of the Renilla luciferase internal control plasmid (pRT-TK) were co-transfected using Lipofectamine 2000 Reagent (Invitrogen). Forty-eight hours after plasmid transfection, luciferase activity was examined using the Dual-Luciferase Report Assay System (Promega) on the Synergy4 (BioTek) platform. Data were first normalized to the empty vector and then to the NC mimic.

We performed biotin-labeled miRNA pull-down as previously described (Lal et al., 2011). Briefly, WT SSCs were transfected with biotin-labeled miR-202-5p or NC mimic (100 nM, GenePharma) with 100 nM RA. After 1 day, the cells were lysed in lysis buffer [20 mM Tris (pH 7.4), 100 mM KCl, 0.3% NP-40, 5 mM MgCl₂, protease inhibitor cocktail and RNase inhibitor]. Streptavidin magnetic beads (Promega) were then added to the cell lysate and incubated at 4°C for 4 h with rotation. The beads were washed three times with lysis buffer. RNA bound to the beads (pull-down RNA) or from 10% of the lysate (input RNA) was isolated using Trizol reagent (Invitrogen) and quantified by qRT-PCR.

Mice at 6-7 days post-partum were used for SG-A isolation by the STAPUT method following our previous report (Gan et al., 2013). Mice were sacrificed, and then the testes were removed and decapsulated. The seminiferous tubules were cut into small pieces. The pieces were incubated in DPBS containing Collagenase I (Gibco) and then in Trypsin (Gibco) containing DNase I. The dispersed cells were filtered through a 40-μm cell strainer (BD Falcon). After filtering, the cells were resuspended in 0.5 ml of DMEM containing 0.5% BSA. Cells were bottom-loaded into a 10-ml syringe, and this was followed by 10 ml of a 2-4% BSA gradient in DMEM. After 2 hours of velocity sedimentation at unit gravity, the cell fractions (0.5 ml/fraction) were collected from the bottom of the syringe at a rate of about 1 ml/min. The cell types and purity in each fraction were initially assessed using a light microscope based on their diameters and morphological characteristics. Fractions highly enriched in SG-A were pooled and the purity was assessed to be 92-94% by immunostaining of PLZF, which is a marker of SG_undiff (Fig. S3A).

Primers used in qRT-PCR are listed in Table S2. Total RNAs were isolated using Trizol regent (Invitrogen). Total RNAs were reverse-transcribed using a High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems) and the reactions were performed using Hieff qPCR SYBR Green Master Mix (Yeasen) in a Roche LightCycler 480 Real-Time PCR system. The data were then analyzed using the comparative Ct method (ΔCt), with β-actin RNA used as the internal control (Gan et al., 2013).

All cells were maintained at 37°C under 5% CO₂ and tested negative for mycoplasma contamination. Establishment and maintenance of SSCs (SG_undiff in this paper) and Sertoli cells was described in detail in another study (Chen et al., 2021 preprint). Sertoli cells were treated with mitomycin C and used for induction of meiotic initiation. The 293FT cell lines were maintained in DMEM medium supplemented with 10% FBS.

Cell lysates from mouse testes were prepared by homogenizing small pieces of organs with glass homogenizers in an RIPA buffer (Beyotime) supplemented with Protease Inhibitor Cocktail (Sigma-Aldrich). Cultured SG_undiff were directly lysed in RIPA buffer supplemented with Protease Inhibitor Cocktail. Lysates were then centrifuged at 20,000 g for 10 min at 4°C, and supernatants were used for western blot analyses. Briefly, lysates were run in SDS-PAGE gel and transferred to PVDF membranes. The blots were blocked with 5% BSA for 2 h, incubated with primary antibodies overnight at 4°C and then incubated with HRP-conjugated secondary antibodies at room temperature for 2 h. The proteins were detected using SuperSignal West Pico PLUS Chemiluminescent Substrate (Thermo Fisher) on the ChemiDoc XRS⁺ system (Bio-Rad) (Fig. S10A,B). After detected for one antibody, blots were stripped and then reprobed for another antibody. The band densities were analyzed by ImageJ.

The SG_undiff were seeded on Laminin-coated glass in a plate and were treated with 10 μg/ml of BrdU for 12 h. Then, the cells were fixed with 70% ice-cold ethanol, denatured with 2 M of HCl and neutralized with 0.1 M of sodium borate (pH 8.5). Afterward, the cells were analyzed using a standard procedure similar to that used for immunofluorescence.

Transplantation of the SSCs/SG_undiff into recipient testes was performed as described previously (Kubota et al., 2004; Wang et al., 2015). For the examination of miR-202 function in Sertoli cells, 10⁵ WT SSCs labeled by EGFP were transplanted into each testis of busulfan-pretreated WT or KO recipient mice. For the examination of miR-202 function in SSCs, 5×10⁴ internal control SSCs labeled by mRuby2 together with 10⁵ WT or KO SSCs labeled by EGFP were mixed and transplanted into each testis of busulfan-pretreated WT recipient mice. One month later, the colony number was counted and analyzed.

SG_undiff were digested with Accutase (Gibco), resuspended in DMEM medium supplemented with 10% fetal bovine serum and plated on a dish. After 30 min, MEF feeder cells, but not SG_undiff, attached to the dish bottom firmly. Floating SG_undiff were collected and plated onto a plate containing mitomycin C-treated Sertoli cells. The induced germ cells were harvested for characterization. For rescue experiments, 100 nM of miRNA mimics or mRNA siRNAs were transfected into SG_undiff using Lipofectamine RNAiMAX reagent 10 h prior to induction (Table S4).

Cultured SG_undiff were treated with 0.25% trypsin to dissociate cells. The cells were filtered through a 40-μm Nylon Cell Strainer (BD Falcon), fixed in 4% PFA for 15 min, and permeabilized with 0.1% Triton X-100 for 30 min (dispensable for the anti-KIT antibody). After blocking with 5% BSA in PBS for 30 min, the cells were incubated with primary antibodies at 37°C for 1 h and then incubated with secondary antibodies at 37°C for 1 h. Analyses were performed using a CytoFLEX research cytometer.

Total RNAs were isolated by Trizol. Two biological replicates for SG-A were used. The RNA libraries were constructed using the NEBNext Ultra Directional RNA Library Prep Kit for Illumina following the manufacturer's recommendations and Oligo (dT) beads (NEB) were used to isolate poly(A) mRNAs. High-throughput sequencing was performed on a Hiseq-PE150 platform.

The correlation coefficients of gene expression levels from biological duplicates were all >0.95. The sequencing reads were mapped to the mouse genome (UCSC, mm9) using TopHat. The mRNA expression level was represented by fragments per kilobase of transcript sequence per millions (FPKM) calculated by Cufflinks. In addition, DEGs were identified by Cuffdiff based on the threshold of P<0.01 together with fold change >3/2 or <2/3. RefSeq mRNAs downloaded from UCSC (mm9) were used as the reference mRNAs.

GO analysis was performed using DAVID bioinformatics tools (Huang et al., 2009) and GO terms with P<0.05 were considered to be significant. GSEA analysis was performed using GSEA v4.0.1 software (Subramanian et al., 2005) with FPKM data.

All experiments reported here were repeated at least three independent times except for RNA-seq for SG-A, which was performed with two biological samples. All of the values in the figures are shown as mean±s.e.m. unless otherwise stated. Excel 2016 or GraphPad Prism 7 were used to perform statistical analyses. For statistical analysis of differences between two groups, two-tailed, unpaired Student's t-tests were used. For the statistical analysis in Fig. 2B, χ² test was used. No samples or animals were excluded from analyses. Sample size estimates were not used. Mice analyzed were littermates and sex-matched whenever possible. Investigators were not blinded to mouse genotypes or cell genotypes during experiments.

In all figures, statistical significance is represented as *P<0.05, **P<0.01 and ***P<0.001. The comparison was considered to be not statistically significant (NS) if P>0.05.

We thank David Zarkower of the University of Minnesota for antibodies. We thank Shiwen Li, Shuguang Duo, Xili Zhu, Xia Yang, Hua Qin and Qing Meng of the Institute of Zoology, Chinese Academy of Sciences for their technical assistance.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.199799.