MFN2 interacts with nuage-associated proteins and is essential for male germ cell development by controlling mRNA fate during spermatogenesis

Mitochondria are dynamic organelles that play crucial roles in multiple biological processes, including energy generation, calcium homeostasis, signal transduction, redox homeostasis and apoptosis (Chen and Chan, 2004; Westermann, 2010). Emerging evidence indicates that mitochondrial function is a crucial determinant of spermatogenesis and male fertility (Ren et al., 2019; Zhang et al., 2016). Moreover, the mitochondria and endoplasmic reticulum (ER) form contacts called mitochondria-associated ER membranes (MAMs), which are reported to be essential for supporting cell structural and functional inter-organelle communications in multiple organs, including the brain, liver and skeletal muscle (Paillusson et al., 2016; Poston et al., 2013; Rieusset, 2018; Theurey et al., 2016). Alteration of MAMs was reported in several human diseases, highlighted by neurodegenerative diseases (Paillusson et al., 2016). Interestingly, our previous work has demonstrated an abundance of MAMs in mouse and human testes and identified a large portion of MAM proteins in testicular cells (Wang et al., 2018), suggesting that MAM proteins may have essential functions in spermatogenesis and male fertility.

Mitofusins (MFN1/2), the ortholog of Fzo in yeast and Drosophila, are the crucial regulators of mitochondrial fusion in mammalian cells and enriched in MAMs (Hales and Fuller, 1997; Santel and Fuller, 2001; Wang et al., 2018). Loss of either Mfn1 or Mfn2 in mice leads to embryonic lethality, and cultured cells obtained from the mutant mouse embryos display overtly fragmented mitochondria (Chen et al., 2003). Specific ablation of Mfn2 in anorexigenic pro-opiomelanocortin (POMC) neurons in the hypothalamus results in aberrant MAMs, defective POMC processing, ER stress-induced leptin resistance, hyperphagia and obesity (Schneeberger et al., 2013). Interestingly, the mutations in human MFN2, but not MFN1, leads to Charcot–Marie–Tooth disease type 2A, a neurodegenerative disorder characterized by progressive sensory and motor loss in the limbs (Bouhy and Timmerman, 2013; Chen et al., 2007; Misko et al., 2012; Rouzier et al., 2012; Züchner et al., 2004). In addition, conditional knockout of Mfn1, but not Mfn2, in growing oocytes results in female infertility (Hou et al., 2019). Although MFN1/2 deficiency in male germ cells leads to mitochondrial defects and male infertility (Chen et al., 2020; Varuzhanyan et al., 2019; Zhang et al., 2016), the underlying mechanism of how mitofusins regulate spermatogenesis and post-meiotic male germ cell development remains largely unknown.

During spermatogenesis, multiple specific germinal structures, termed nuage, are present in the cytoplasm of male germ cells; these are electron-dense, non-membranous structures, close to mitochondria and/or nuclei. They are of varying size, yielding to different compartments that harbor different components. These primarily include the inter-mitochondrial cement (IMC), piP-body, and chromatoid body (CB) (Aravin et al., 2009; Eddy, 1974, 1975). The IMC and CB are tightly associated with mitochondria. IMC clusters formed between mitochondria are present in gonocytes, prospermatogonia, spermatogonia and meiotic spermatocytes until late pachytene spermatocytes (PS) (Eddy, 1974, 1975). The cytoplasmic IMC is considered the primary site for Piwi-interacting RNA (piRNA) pathway proteins, also known as pi-body (Aravin et al., 2009; van der Heijden et al., 2010). The chromatoid body first appears as thick cytoplasmic fibers and granulated material in the interstices of mitochondrial clusters and the perinuclear area of late PS, condenses into a single lobulated, perinuclear granule in round spermatids (RS) and disassembles later during spermatid elongation (Eddy, 1970, 1974; Fawcett et al., 1970). Several proteins have been identified and localized to the nuage, such as MIWI (also known as Piwil1), DDX4, MAEL, TDRKH and GASZ (Asz1) (Wang et al., 2020). Interestingly, MIWI, DDX4 and MAEL are expressed in both IMC and CB, whereas TDRKH and GASZ localize only to the IMC (Ma et al., 2009; Saxe et al., 2013; Takebe et al., 2013; Toyooka et al., 2000). Importantly, GASZ interacts with MFN1 to mediate spermatogenesis (Zhang et al., 2016). The nuage is the proposed site of germ cell functions with multiple RNA processing events, including translational regulation, RNA-mediated gene silencing, mRNA degradation, piRNA biogenesis and nonsense-mediated mRNA decay (Kotaja and Sassone-Corsi, 2007; Wang et al., 2020). Translational regulatory proteins such as MIWI, TDRD6 and MAEL have been found to be localized to the chromatoid body and associated with mRNA translational machinery and piRNA biogenesis (Castañeda et al., 2014; Fanourgakis et al., 2016; Grivna et al., 2006; Takebe et al., 2013; Vasileva et al., 2009). Thus, many components necessary for transcript processing are physically connected to the nuage during spermatogenesis.

In mice, spermatogenesis undergoes two rounds of transcriptional cessations: one is during meiosis, which includes the synapsis, desynapsis and meiotic sex chromosome inactivation (MSCI), and the other is the process in which histones are replaced by transition proteins, protamines, during late spermiogenesis (Sassone-Corsi, 2002; Wykes and Krawetz, 2003). During transcriptional cessations, a large number of germ cell-specific mRNAs are synthesized and stored in germ cells (meiotic spermatocytes or elongating spermatids) before their translation (Braun, 1998). Post-transcriptional and translational regulation of stored mRNAs ensures timely synthesis of proteins essential for male germ cell development (Braun, 1998; Kleene, 2013). For example, Y-box proteins FRGY2 (Xenopus)/MSY2 (mouse) (also known as YBX2), have been identified to play an essential role in the regulation of transcription and mRNA storage/translational delay in germ cells by a Y-box promoter-bound mechanism preferentially (Matsumoto and Wolffe, 1998; Yang et al., 2005a). However, limited knowledge exists about the specific regulators and how this regulation is achieved in male germ cells.

In this study, we report that MFN2, a mitochondrial fusion protein, interacts with nuage-associated proteins, including MIWI, DDX4, GASZ and TDRKH in germ cells and is essential for male germ cell development. The conditional knockout of Mfn2 in mice led to male sterility characterized by mitochondria defects, structurally abnormal MAMs and precocious transcriptional/translational processes in germ cells. We further revealed that MFN2 interacts with MSY2 to control the fate of MSY2-bound mRNAs in testes, such as the early-translated SPATA19. Collectively, our findings uncovered a novel role for MFN2 in mitochondria-associated germinal structure formation and regulation of the spermatogenic mRNA fates during male germ cell development and spermatogenesis.

Although MFN2 is a ubiquitously expressed protein located on both the outer mitochondrial membrane (OMM) and the ER surface in mammalian cells (de Brito and Scorrano, 2008), to date no information is available on the pattern of its expression in male germ cells during spermatogenesis. By performing RT-qPCR and western blot assays, both Mfn2 mRNA and protein were detected in mouse testes (Fig. S1A,B) and expressed from postnatal day (P) 0 to P56 (Fig. S1C,D). Immunofluorescence staining of MFN2 in the developing testes further revealed that MFN2 is primarily expressed in prospermatogonia at P0 and spermatogonia at P7, and finally highly expressed in PS and RS at P14 to P28 (Fig. S1E).

Subcellular localization of MFN2 during spermatogenesis was determined by immunostaining in adult testes. We observed the presence of MFN2 throughout most of the germ cell development, including the mitotic spermatogonia, meiotic spermatocytes (pre-leptotene to diplotene), and RS (Fig. 1A). Interestingly, MFN2 expression was relatively high in the cytoplasm of PS and RS (Fig. 1A,B). To further determine the expression pattern of MFN2 in male germ cells, we co-stained MFN2 and GCNA1 (GCNA; a germ cell marker) in adult WT testes, and found that MFN2 distributes in almost all types of germ cells, but is nearly imperceptible in Sertoli cells (Fig. S1F). This observation was further confirmed in different isolated testicular cells by RT-qPCR and western blot assays (Fig. S1G-I). As Mfn2 encodes an OMM protein, we confirmed its co-localization with ATP5A (an OMM marker) in adult testes by co-immunofluorescence staining (Fig. 1C). Together, these data indicate that MFN2 is highly expressed in the cytoplasm of meiotic spermatocytes and post-meiotic RS and may play an essential role in spermatogenesis.

To elucidate the physiological role of MFN2 in spermatogenesis, we generated germline-specific knockout mice. We used Stra8-Cre transgenic mice, in which Cre is expressed in undifferentiated spermatogonia at P3 (Sadate-Ngatchou et al., 2008), to delete exon 6 of the Mfn2 gene (Fig. S2A). The efficiency of this Mfn2 conditional knockout mouse (Stra8-Cre; Mfn2^loxp/Del, herein called Mfn2-cKO) was verified by determining levels of Mfn2 mRNA and protein in testes. Both Mfn2 mRNA and protein were dramatically reduced in Mfn2-cKO testes compared with that of control testes (Fig. 2A-D). Thus, we successfully generated the male germ cell-specific Mfn2 mutants.

Although Mfn2-cKO males were viable and appeared to be grossly normal, they displayed sterility after mating with fertility-proven adult wild-type (WT) females for 5 months (zero out of six male mice were fertile). Consistent with this infertile phenotype, the testes from Mfn2-cKO mice were significantly smaller than their control littermates (Fig. 2E). The ratio of testis weight/body weight of Mfn2-cKO mice was decreased substantially at various ages, from P35 to P120, compared with controls (Fig. 2F). Histological analyses showed that, as the adult Mfn2-cKO mice aged, the number of severely atrophic abnormal seminiferous tubules increased (Fig. 2G), suggesting that spermatogenic defects are age-dependent in Mfn2-cKO mice. Interestingly, when the first wave of spermatogenesis culminated at P42, the late-stage spermatocytes and RS continually decreased in number and appeared to gradually vacuolize (Fig. 2G). In comparison, no discernible abnormality was found in the Mfn2-cKO testes from P7 to P28 (Fig. S2B). Consistent with these histological results, the TUNEL assay revealed that the number of apoptotic cells in Mfn2-cKO testes at P14 to P28 was comparable with control testes but increased significantly at P35 and P56 (Fig. S2C,D). Moreover, the number of spermatozoa retrieved from adult Mfn2-cKO cauda epididymis was dramatically reduced compared with controls (Fig. 2H,I). Only ∼2% of Mfn2-cKO epididymal sperm showed normal morphology, compared with ∼80% of the sperm in controls (Fig. 2J,K). These data indicate that, upon Mfn2 deletion in postnatal testes, most germ cells are gradually lost from P35 by apoptosis, and the remainder is morphologically defective.

MFN1 and MFN2 are the two Fzo homologs in humans and mice. They interact with each other in mammalian cells to coordinately regulate mitochondrial fusion (Chen et al., 2003), raising the question of whether MFN1 and MFN2 interact in testicular cells. We first examined their interaction in adult mouse testes by MFN1 and MFN2 co-immunoprecipitation (Co-IP). In both the MFN2 and MFN1 antibody immunoprecipitants, each protein was detected in testes (Fig. S3A). These results confirmed that MFN1 and MFN2 are indeed bona fide interacting partners in the testes. In contrast, both MAEL (a piP-body component; Soper et al., 2008; Takebe et al., 2013) and GAPDH were not detected in the MFN2 antibody immunoprecipitants (Fig. S3B), confirming the high specificity of MFN2 antibody in Co-IP assays. To determine the role of MFN1 during spermatogenesis and male germ cell development, we then generated Stra8-Cre-mediated Mfn1 conditional knockout mice through the deletion of exon 4 (Stra8-Cre; Mfn1^loxp/Del, herein called Mfn1-cKO) of the Mfn1 gene (Fig. S3C). As expected, Mfn1-cKO mice were also infertile. Spermatogenesis was disrupted as evidenced by smaller testes, decreased testis weight and disrupted seminiferous tubule structure (Fig. S3D-G). Importantly, we observed that Mfn1-cKO testes exhibited severe vacuolization defects as early as P28. In comparison, the first wave of spermatogenesis in Mfn2-cKO testes proceeded normally until P42. This disparity between Mfn1-cKO and Mfn2-cKO mouse phenotypes shows that MFN2 and MFN1 are essential for spermatogenesis but each has a distinct role.

To better understand the physiological roles of mitofusins in spermatogenesis, we next simultaneously deleted Mfn1 and Mfn2 in mouse testis using Stra8-Cre-mediated Cre-Loxp strategy (Stra8-Cre; Mfn1^loxp/Del-Mfn2^loxp/Del, herein called Mfn1/2-cDKO). Notably, the testicular disruption in Mfn1/2-cDKO mice was much more severe than in either single Mfn1 or Mfn2-cKO mice (Fig. 2L). Compared with the control mice, the diameters of seminiferous tubules from Mfn1/2-cDKO mice appeared to be much smaller, as early as P14 (Fig. S4A,B). Almost no RS were observed in Mfn1/2-cDKO testes at P25 (Fig. 2L). These data indicate that Mfn1/2-cDKO mice failed to complete meiosis, which is consistent with a previous report (Varuzhanyan et al., 2019). Consistent with the expected phenotype, the adult Mfn1/2-cDKO testis was much smaller than that of adult Mfn1 or Mfn2 single cKO mice (Fig. S4C,D). Histological analyses of adult testicular sections further revealed increased atrophic and vacuolated seminiferous tubules in Mfn1/2-cDKO testes compared with single Mfn1 or Mfn2-cKO testes (Fig. S4E). Strikingly, unlike the uniformly distributed donut of mitochondria in the control cytoplasm, mitochondria were aggregated to one side of the cytoplasm in the Mfn1/2-cDKO spermatocytes at P25 (Fig. 2M; Fig. S4F). However, this abnormal mitochondria distribution was not observed in Mfn1-cKO or Mfn2-cKO mice (Fig. S4F). Together, these results suggest that MFN2 could interact with MFN1 to regulate the distribution of mitochondria in the testes, thereby contributing to spermatogenesis and male germ cell development.

Given that the distribution of the mitochondria was disrupted in Mfn1/2-cDKO testes, we examined the ultrastructure of mitochondria and its associated organelles (MAMs and ER) in mitofusin-deficient germ cells. Using transmission electron microscope (TEM) analysis, we observed that mitochondria exhibit swelling and fragmentation in germ cells from adult Mfn2-cKO or Mfn1-cKO mouse testes (Fig. 3A,B; Fig. S5A). The thickness of IMC increased significantly in Mfn2-cKO PS (Fig. 3A; P=0.024, two-tailed Student's t-test). We then focused on Mfn2-cKO RS to further detail the mitochondrial defects. The mitochondrial aspect ratios (AR; major axis/minor axis of mitochondria) were used to calculate the length of mitochondria in RS in control and Mfn2-cKO mice. Although the AR of all mitochondria was not significantly decreased in Mfn2-cKO RS (Fig. 3C), the AR distribution displayed an increased percentage of short mitochondria with AR≤1.5 compared with that of controls (Fig. 3D). These data suggest that developmental defects of male germ cells in both Mfn1- and Mfn2-cKO mice could be due to the fragmentation and mitochondrial swelling in germ cells.

Given that MFN2 supports structural and functional communication between the mitochondria and ER, we then investigated whether the deletion of Mfn2 in postnatal male germ cells disrupts MAM and ER homeostasis. We observed that the distance between mitochondria and ER was increased by ∼16% in Mfn2-cKO spermatids compared with control spermatids. Similarly, the percentage of mitochondria-ER contacts was significantly reduced by nearly 50% in Mfn2-cKO testes compared with that of controls (Fig. 3E,F). In addition, the ER-mitochondria contact coefficient (ERMICC) (Naon et al., 2016) was also reduced by more than 30% in Mfn2-cKO testes compared with that of controls (Fig. 3G). Notably, the ER displayed tube-like cisternae structures in control spermatids, whereas it was fragmented in Mfn2-cKO spermatids (Fig. 3B), implying that the ER structure was disrupted upon MFN2 deletion. To further confirm these observations, calreticulin (a MAM and ER marker protein) immunostaining was employed to examine the MAM structure in P18 and P60 testes. Calreticulin displayed a diffused granular pattern in the cytoplasm of Mfn2-cKO spermatocytes at P18 instead of the continuous perinuclear localization exhibited in the controls (Fig. 3H). Moreover, in P60 testes, the calreticulin signals in Mfn2-cKO germ cells appeared to be reduced and displayed diffused granular distribution compared with that of controls (Fig. 3I). Of note, unlike the observation in Mfn2-cKO spermatocytes, calreticulin does not exhibit any diffused signal pattern in Mfn1-cKO spermatocytes (Fig. S5B). These data suggest that MFN2 but not MFN1 may function in the regulation of the formation of MAM and ER structures.

Dynamic mitochondrial fusion and fission are known processes that regulate mitochondrial DNA (mtDNA) stability and energy production (Palmer et al., 2011; Westermann, 2010). To examine each, we first determined mtDNA copy number in Mfn1-cKO and/or Mfn2-cKO adult testes by quantitative real-time PCR (qPCR). Unexpectedly, we found that the mtDNA copy number was increased significantly in both Mfn1-cKO and Mfn2-cKO adult testes (Fig. 3J), which is in contrast to previous reports showing that conditional single knockout of either Mfn1 or Mfn2 in skeletal muscle did not alter mtDNA copy number (Chen et al., 2010). This difference implies that an additional MFN1/2 mechanism regulates mtDNA stability. We directly stained cytochrome c oxidase (COX, complex IV) and succinate dehydrogenase (SDH, complex II) activity to assess whether the function of respiratory complexes was affected by MFN1- or MFN2 deficiency. In line with the increased mtDNA copy number, COX activity increased in both Mfn1-cKO and Mfn2-cKO testis sections, whereas the SDH activity was unaltered (Fig. 3K). These results suggest that MFN1/2 deficiency in male germ cells leads to increased mitochondrial fission and mitochondrial metabolism disorder.

As MFN2 displays a granular cytoplasmic localization in male germ cells (Fig. 1), we assumed that MFN2 is involved in mitochondria-associated germinal structure formation. To test this hypothesis, we selected four mitochondrial-associated germinal granule proteins (also known as nuage-associated proteins), MIWI, DDX4, TDRKH and GASZ, to perform Co-IP in the testis and an HEK293T cell line. We found that, in MFN2 antibody immunoprecipitants, MIWI, DDX4, GASZ and TDRKH were strongly detected in WT testes (Fig. 4A). Likewise, in MIWI, TDRKH, DDX4 and GASZ antibody immunoprecipitants, MFN2 was detected in testes (Fig. 4B-E). These results suggest that MFN2 interacts with MIWI, TDRKH, DDX4 and GASZ in testes. We also examined their interaction in the cell line. When the full length of MFN2 was ectopically co-expressed with MIWI, TDRKH and DDX4 in HEK293T cells, MFN2 was pulled down by MIWI, TDRKH and DDX4, indicating the ability of MFN2 to interact with MIWI, TDRKH and DDX4 proteins (Fig. 4F,G). To validate the interaction specificity of MFN2 with these proteins, we chose hnRNPA2, a nuclear protein, which does not interact with MFN2 as a negative control. The result showed that Myc-hnRNPA2 could not interact with Flag-MFN2 in HEK293T cells (Fig. S5C).

Next, we examined whether the expression of MIWI, DDX4, GASZ and TDRKH was affected by the deletion of MFN2 in testes. Co-immunostaining revealed reduced levels of MIWI, DDX4 and GASZ in IMC and CB structures of Mfn2-cKO testes, whereas TDRKH displayed no apparent change (Fig. 4H; Fig. S5D-F). In Mfn2-cKO testes, both MIWI and DDX4 showed weaker staining in the IMC of PS and was barely detected in the CB of RS (Fig. 4H; Fig. S5D,E). GASZ also showed a decreased signal in both PS and RS within Mfn2-cKO testicular sections (Fig. 4H; Fig. S5F). As MFN1 was reported to interact with GASZ and is essential for spermatogenesis (Zhang et al., 2016), we also compared the expression and localization of DDX4, GASZ and TDRKH in Mfn1-cKO adult testes. However, unlike Mfn2-cKO testes, we found no obvious changes in expression or localization in Mfn1-cKO testes (Fig. S5G-I).

Nuage-associated proteins are documented in regulating piRNA biogenesis. This prompted us to determine whether MFN2 participates in the piRNA pathway. We chose P25 testes to examine the abundance and size distribution of piRNAs through small RNA-seq, as the morphology and the cell populations were comparable at P25 between WT and Mfn2-cKO testes. The results showed ∼50% decrease in the total piRNAs normalized to miRNA counts (Fig. 4I; Table S1), suggesting that MFN2 participates in piRNA biogenesis through its interaction with the nuage proteins. To investigate whether the abundance of piRNA precursors is affected by MFN2 deletion, we analyzed piRNA precursor levels in the Mfn2-cKO testis at P25. We found that the levels of piRNA precursors in Mfn2-cKO were comparable with those of WT (Fig. 4J). Next, we asked whether the retrotransposons were affected in Mfn2-cKO testes, because the nuage-associated proteins have been reported to repress retrotransposons to maintain genome integrity. We analyzed the levels of LINE1 and IAP mRNAs, and found that there were no significant changes between WT and Mfn2-cKO testes (Fig. 4K). Consistent with the level of LINE1 mRNA, the LINE1 ORF1 protein showed no activation in P60 Mfn2-cKO testes (Fig. 4L). Taken together, these results indicate that MFN2 interacts with nuage-associated proteins and participates in piRNA pathways during spermatogenesis, but does not seem to directly participate in retrotransposon silencing.

To investigate the molecular consequences of the loss of MFN2 in spermatogenic cells, we compared the transcriptomes of purified PS and RS from WT and Mfn2-cKO adult testes using RNA-seq. Hierarchical clustering of differentially expressed genes (DEG) showed that, in PS, a total of 4046 genes were upregulated and 5324 genes were downregulated in Mfn2-cKO mice compared with WT mice. In RS, a total of 3756 genes were upregulated and 3186 genes were downregulated in Mfn2-cKO mice compared with WT mice (Fig. 5A,B; Fig. S6A,B; Tables S2,S3). To validate the RNA-seq data, we randomly selected the genes associated with ribosome transport (Uba52, Rps20, Rpl21, Fau and Rpl26), spliceosome and mRNA transport (Pabpc1, Hspa2, Sf3b4, Acin1, Ncbp1 and Nup188) pathways for RT-qPCR analyses. The expression of genes related to ribosome transport pathways was increased more than twofold (except Rps20) in both Mfn2-cKO PS and RS compared with controls, in accord with the fold-changes determined by RNA-seq. Moreover, the expression of genes related to spliceosome and mRNA transport pathways were reduced by more than 50% in both Mfn2-cKO PS and RS (Fig. S6C,D). Taken together, these RNA-seq analyses suggest that MFN2 deficiency in testes influences the expression of genes that control multiple biological processes, including mRNA processing.

As shown above (Fig. 4I), we observed a ∼50% reduction of piRNA populations in Mfn2-cKO testes. Next, we asked whether the decreased level of piRNAs dysregulate transcripts in Mfn2-cKO PS and RS. As the function of piRNAs in regulating mRNA stability and directed meiotic transcript cleavage require the slicer activity of MIWI, a nuage-associated protein (Goh et al., 2015; Zhang et al., 2015), we compared the RNA-seq dataset of Mfn2-cKO with the published RNA-seq dataset of Miwi-KO (GEO accession number GSE64138) (Goh et al., 2015), including PS and RS. First, we compared the upregulated genes identified from Mfn2-cKO PS and RS with the 330 putative piRNA-targeting mRNAs reported by Goh et al. (2015). We found that ∼10% (34/330 in PS and 32/330 in RS) of piRNA-targeting mRNAs were upregulated in Mfn2-cKO spermatogenic cells (Fig. 5C,D; Tables S4,S5). With fewer piRNA-targeting mRNAs overlapped than expected (representation factor=0.9 and 0.8, respectively), the probability of this overlap is non-significant, with P-values of 0.077 (PS) and 0.230 (RS). When compared with the results from Goh et al. (2015), ∼20% (72/330) piRNA-targeting mRNAs were reported to be upregulated in Miwi-KO RS (Goh et al., 2015). Interestingly, only three transcripts overlapped between 32 upregulated piRNA-targeting mRNAs in Mfn2-cKO RS and 72 upregulated piRNA-targeting mRNAs in Miwi-KO RS (Table S5), likely reflecting the complexity of MFN2 in regulating spermatogenesis, which is distinct from MIWI. These analyses suggest that the 50% decrease of piRNA observed in Mfn2-cKO could influence the expression of ∼10% piRNA-targeting mRNAs. Second, we compared the overlapped dysregulated genes between Mfn2-cKO and Miwi-KO spermatogenic cells, including PS and RS. We took the Mfn2-cKO versus WT to scatter plots and mapped the genes upregulated in Miwi-KO PS (Fig. 5E) and RS (Fig. 5F), downregulated in Miwi-KO PS (Fig. 5G) and RS (Fig. 5H), and unchanged in Miwi-KO (Fig. 5I,J). The results showed that 74 genes are co-upregulated in both Mfn2-cKO and Miwi-KO PS (Fig. 5E; Fig. S7A), and 152 genes are co-upregulated in both Mfn2-cKO and Miwi-KO RS (Fig. 5F; Fig. S7B). This overlap of ∼10% (74/680, representation factor=0.8, P=0.049) and ∼7% (152/1971, representation factor=0.7, P=1.137e-7) upregulated genes, respectively, is significantly less than was expected. As such, these overlapping upregulated genes in PS and RS of Mfn2-cKO and Miwi-KO were only a small portion of the two gene sets, which indicates that they are unlikely to be co-regulated by MIWI and MFN2 in testes.

Furthermore, 304 and 108 genes were found to be commonly downregulated in Mfn2-cKO and Miwi-KO PS (Fig. 5G; Fig. S7C) and RS (Fig. 5H; Fig. S7D), respectively, which accounts for ∼25% (304/1541, representation factor=1.1, P=0.004) and ∼6% (108/1757, representation factor=0.7, P=1.846e-7) of downregulated genes in Miwi-KO PS and RS, respectively. Interestingly, the 304 overlapping genes identified in PS was significantly more than expected, whereas the 108 overlapping genes in RS was significantly less than expected, indicating that the 304 overlapping downregulated genes in PS are related, whereas those overlapping downregulated genes in RS are not. Moreover, the 304 downregulated genes in PS overlaps are mostly related to germ cell development, spermatogenesis and meiotic division (Fig. S7E,F). Together, MFN2 and MIWI (a nuage protein) only regulate a smaller overlapping subset of RNAs than expected, suggesting that although MFN2 and MIWI are bound in testes, they regulate mostly non-overlapping sets of mRNAs.

Because some of the MFN2 interaction proteins, like MIWI and DDX4, are associated with RNA processing and translational machinery in spermatogenesis (Grivna et al., 2006; Nagamori et al., 2011), we wanted to determine whether MFN2 plays a role in mRNA translation. To test this relationship, we first subjected adult testicular extracts to sucrose density polysome gradients. MFN2 co-sedimented with both the monosome (80S) and polysome fractions (Fig. 6A). Similar to RPS6, the addition of EDTA to dissociate the large and small ribosomal subunits shifted MFN2 to the ribonucleoprotein (RNP) fractions of the gradients. These data suggest that MFN2 associates with cytoplasmic monosomes and polysomes.

Given that MSY2 was exclusively enriched in both the meiotic and post-meiotic germ cells (Gu et al., 1998), and has been reported to participate in both transcription and mRNA translation (Yang et al., 2005b), we next asked whether MFN2 complexes with MSY2 to regulate mRNA translation in testes. We found that MFN2 and MSY2 could reciprocally pull down each other independent of RNA, indicating a direct interaction between these two proteins in vivo (Fig. 6B,C). Because our unbiased RNA-seq data showed an increased level of Msy2 mRNA in both Mfn2-cKO PS and RS, we then examined the protein level of MSY2 in adult control and Mfn2-cKO testes by immunofluorescence. In control testes, MSY2 was broadly expressed in nearly all types of spermatogenic cells and showed a disperse distribution in the cytoplasm and nucleus in PS and RS (Fig. 6D). However, in Mfn2-cKO testes, the cytoplasmic MSY2 increased in PS and RS, and the MSY2 foci within the RS nucleus had almost disappeared (Fig. 6D). Considering that MSY2 marks specific mRNAs in the nucleus for cytoplasmic storage and links transcription and mRNA storage/translational delay in meiotic and post-meiotic male germ cells (Yang et al., 2005a), we analyzed the transcript levels of MSY2-bound and -unbound mRNAs in our RNA-seq data. Intriguingly, for MSY2-bound gamete-specific mRNAs, of which ∼57% (26/46) is upregulated in Mfn2-cKO RS, including Msy2 itself, ∼9% (4/46) downregulated and the remaining 34% (16/46) mRNAs are unaffected (Fig. 6E; Fig. S8A; Table S6). In contrast, MSY2-unbound mRNAs in Mfn2-cKO RS were either downregulated (34%, 16/48) or unaffected (54%, 26/48), and only 12% of MSY2-unbound mRNA was upregulated (6/48) (Fig. 6E; Fig. S8B; Table S7). The situation in Mfn2-cKO PS is almost the same as in RS (Fig. 6F; Tables S8,S9). These results indicate that MFN2 interacts with cytoplasmic MSY2 in spermatogenic cells to regulate the fates of MSY2-bound gamete-specific mRNAs.

To further explore the role of MFN2 during translation, we assessed whether any proteins were mistranslated from the MSY2-bound gamete-specific mRNAs. As SPATA19 is a mitochondrial protein previously reported to be expressed in haploid spermatids in P28 testes (Doiguchi et al., 2002; Matsuoka et al., 2004) and germ cell-specific Spata19 knockout mice displayed male sterility (Mi et al., 2015), we chose to examine SPATA19 at the protein level in P25 testes from control and Mfn2-cKO mice. Notably, we found that the SPATA19 protein could be detected in P25 Mfn2-cKO testes, whereas it was not detected in control testes (Fig. 6G,H). To determine whether the early-translated SPATA19 protein observed in Mfn2-cKO testes is caused by translational activity, we measured cytoplasmic translation by quantifying the polysome distribution of Spata19 mRNAs, with β-actin (Actb) as an internal control in WT and Mfn2-cKO testes at P25. The distribution of Spata19 mRNA shifted toward the heavier polysome fractions, whereas the distribution of Actb mRNA showed no apparent change (Fig. 6I,J), indicating increased SPATA19 translational activity in the cytoplasm of Mfn2-cKO testicular cells. Taken together, these results indicate that MFN2 is associated with polysomes and interacts with MSY2 (non-nuage protein) to control the fates of MSY2-bound gamete-specific mRNAs (e.g. Spata19) during spermatogenesis.

Although two recent studies have reported that both MFN1 and MFN2 play essential functions in spermatogonial differentiation and meiosis through a dynamic mitofusin-mediated mitochondrial and ER stress mechanism in spermatogenesis (Chen et al., 2020; Varuzhanyan et al., 2019), its roles in post-meiotic spermatogenic cell development have remained elusive. In this study, we showed that loss-of-function of either MFN1 or MFN2 in postnatal male germ cells results in male sterility in mice and, more importantly, we revealed a previously unknown role of MFN2 upon interaction with nuage-proteins and MSY2 that controls the fate of many spermatogenic cell mRNAs. These findings uncovered a novel role of MFN2 in late spermatogenesis and added another layer of MFN1/2 molecular functions onto post-meiotic male germ cell development. Interestingly, we found that the Mfn2-cKO mouse phenotype is age-dependent, with germ cells beginning their decline after P35. In comparison, Mfn1-cKO mice showed a severe phenotype with an observation of seminiferous tubule vacuolization as early as P28. Similarly, Vasa-Cre-mediated Mfn1 or Mfn2 single knockout mice (Vasa-Cre; Mfn1^flox/flox or Vasa-Cre; Mfn2^flox/flox) were also infertile and displayed a distinct meiotic arrest phenotype, with a relative earlier blockage of meiosis in Mfn1-cKO than in Mfn2-cKO mice (Chen et al., 2020). The discrepancy of phenotypes between this study and the previous study possibly reflects the differences in Cre recombinase expression, as Vasa-Cre appears much earlier than Stra8-Cre in the male germline. Thus, the phenotypic differences between Mfn1- and Mfn2-cKO mice mediated by either Stra8-Cre or Vasa-Cre suggest that MFN1 plays a distinct role from MFN2, not only in regulating spermatogonial differentiation but in promoting meiotic and post-meiotic spermatogenic cell development. Notably, consistent with previous reports (Varuzhanyan et al., 2019), the Stra8-Cre induced Mfn1/2 double knockout mice presented a defective meiosis phenotype of the near loss of PS, suggesting that MFN1 cooperates with MFN2 to regulate spermatogonial differentiation and meiosis. However, based on transgene rescue experiments (Chen et al., 2020), it does not appear to be likely that MFN1 mutually compensates with MFN2 in male germ cell development. This is consistent with the view that the regulation of spermatogenesis by these two mitochondrial fusion proteins at different stages is independent of other currently unknown functions beyond the mitochondrial fusion mechanism. The identity of the distinct underlying molecular mechanism between MFN1 and MFN2 in regulating male germ cell development needs further investigation.

Several MFN2 interactive proteins identified in this work, including MIWI, DDX4, GASZ and TDRKH, were also present in the MAM fractions in our previous mass spectroscopy data (Wang et al., 2018). Remarkably, the proteins identified located in MAMs are possibly responsible for regulating piRNA biogenesis and mRNA translation as they are a feature of nuage-associated proteins in germ cells. The fact that we observed reduced pachytene piRNA production in Mfn2-cKO testes suggests that mitofusins might have a function in piRNA biogenesis. We observed an ∼50% reduction in Mfn2-cKO pachytene piRNA, which may explain, to a lesser extent, the elevated mRNA levels in Mfn2-cKO PS and RS determined by RNA-seq, because pachytene piRNA could target meiotic mRNA cleavage (Goh et al., 2015). Surprisingly, the transposable elements (TEs) were not activated despite the decrease in pachytene piRNA in Mfn2-cKO testes. These data are in agreement with a previous report, which showed that the activity of TEs was not perturbed in Vasa-Cre-mediated Mfn2 conditional knockout mice (Chen et al., 2020). The possible reason for this is that the reduced piRNAs in Mfn2-cKO are mostly not associated with TEs or repeats (NRapiRNAs) (Larriba and Del Mazo, 2018). However, we could not exclude the possibility of reduced piRNA biosynthesis in this study as an indirect effect caused by a decrease of piRNA biogenesis proteins, such as MIWI, DDX4, GASZ, etc. in Mfn2-cKO testes.

Interestingly, our data reveal that both MIWI and DDX4 are absent in the CB of Mfn2-cKO RS, resembling MIWI absence in the CB of Tdrkh-cKO spermatids (Ding et al., 2019). Unlike the reduced MIWI and GASZ expression, TDRKH expression was not affected in Mfn2-cKO spermatogenic cells, even though TDRKH was identified as interacting with MFN2. TDRKH expression was also unchanged in the Miwi-KO mice (Ding et al., 2019). The observations can be reconciled given that TDRKH is instrumental in binding to mitochondria to recruit MIWI to IMC and CB. These are the two typical nuage structures localized at spermatocytes and spermatids, respectively. Given the feature that TDRKH is a mitochondrial membrane-anchored protein, and MFN2 is a mitochondrial fusion protein (Hales and Fuller, 1997; Saxe et al., 2013), it has raised the possibility that the recruitment function of TDRKH requires the coordinated regulation of MFN2.

MFN2 is enriched in monosome and polysome fractions in spermatogenic cells and interacts with MSY2 to mediate MSY2-bound gamete-specific mRNA fates. MSY2 showed increased cytosolic expression in both PS and RS upon MFN2 depletion. This discovery raises the possibility that MFN2 partners with MSY2-bound mRNA for its function, because MSY2 was enriched in the RNP fraction and regulates mRNA stability, storage and translation delay in spermatogenesis (Yang et al., 2005a). Interestingly, MFN2 is not a known transcription factor, chromatin regulator or predicted RNA-binding protein. The global changes in the transcriptome of Mfn2-cKO spermatogenic cells might be indirectly caused by decreased piRNA abundance and misregulation of MSY2. The ∼50% decrease of piRNA might dysregulate a portion of meiotic transcripts, for example the piRNA-targeting transcripts. Besides, most of the interaction proteins of MFN2 are associated with RNA processing and the perturbation of the interaction protein could alter RNA transcript levels. Many MSY2-bound mRNAs are gamete-specific transcripts that are usually expressed in meiotic and post-meiotic cells. They should be subject to translational delay when they are stored in the germ cell cytoplasm (Yang et al., 2007). MSY2 is known to repress translation of and prevent degradation of target RNAs (Xu and Hecht, 2008; Yang et al., 2005a). In the present study, if MFN2, acting as a partner protein, is required for both of those functions, we would expect that MSY2 target RNA levels would decrease in the Mfn2-cKO, even in the presence of increased MSY2. In contrast, though, we observed that 57% of the MSY2-bound RNAs showed higher expression in the Mfn2-cKO (versus 12% of MSY2-unbound RNAs). At the same time, the SPATA19 protein (encoded by MSY2 target RNA) was expressed prematurely at P25 in the Mfn2-cKO. One possible explanation is that absence of MFN2 compromises the translational repression by MSY2 but does not disturb its ability to protect target RNAs from degradation. Thus, increased MSY2 in the Mfn2-cKO may be more efficient at keeping the target RNAs safe, but without MFN2 it would not prevent their ectopic translation. However, further experiments are needed to elucidate the effects of MFN2 on the expression of MSY2 binding/nonbinding mRNAs during germ cell development. Specifically, it will be important to determine whether other MSY2 targets besides SPATA19 are expressed early in the Mfn2-cKO testes.

In conclusion, we report a novel role of MFN2 in controlling mRNA fates during spermatogenesis through interactions with both nuage-associated proteins and translation regulator MSY2. We propose that MFN2 associates with several mRNA translation-associated proteins such as MSY2, MIWI and DDX4 at the cytoplasmic nuage and/or MAM, and it likely serves as a scaffold to recruit mRNA, polysomes and other RNA binding proteins in both PS and RS to control gamete-specific transcript delay during spermatogenesis (Fig. S8C,D). This finding extends beyond the known functions of MFN2 in regulating mitochondrial fusion processes and spermatogenesis.

All animal procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of Tongji Medical College, Huazhong University of Science and Technology, China, and the mice housed in the specific pathogen-free facility of Huazhong University of Science and Technology. All experiments with mice were conducted ethically according to the Guide for the Care and Use of Laboratory Animal guidelines. Floxed Mfn1 or Mfn2 mice were previously created (Chen et al., 2007). The mice harboring the floxed Mfn1 or Mfn2 allele purchased from the Jackson Laboratory (stock no. 026401 and 026525). Mfn1^flox/flox and/or Mfn2^flox/flox female mice were crossed with Stra8-Cre male mice (Jackson Laboratory, stock no. 008208) to obtain Stra8-Cre; Mfn1^+/flox and/or Mfn2^+/flox males, then the Stra8-Cre; Mfn1^+/flox and/or Mfn2^+/flox male mice were further bred with Mfn1^flox/flox and/or Mfn2^flox/flox to obtain Stra8-Cre; Mfn1^flox/del and/or Mfn2^flox/del (designated as Mfn1-cKO and Mfn2-cKO) males.

The details of all commercial antibodies used in this study are presented in Table S10.

Testes and epididymides from control and cKO mice were collected and fixed in Bouin's solution (Sigma-Aldrich, HT10132) at 4°C overnight and then washed with 75% alcohol five times for 30 min each time. Samples were then embedded in paraffin and 5 μm sections were cut using a ThermoFisher Scientific CryoStar NX50 cryostat and stained with periodic acid-Schiff (PAS) after being dewaxed and rehydrated. Slides were then mounted with Permount (Fisher Chemical™ SP15-100) and imaged using a ZEISS microscope. For immunofluorescence staining, testes were fixed in 4% paraformaldehyde (PFA) in PBS overnight at 4°C and embedded in Tissue-Tek optimal cutting temperature (OCT) compound (Sakura Finetek, 4583) on dry ice. A 5 μm thick section was cut and microwaved in 0.01 M sodium citrate buffer (pH 6.0) for 5 min to retrieve antigen. After rinsing with PBS three times, the sections were blocked in blocking solution (containing 3% normal goat serum and 3% fetal bovine serum in 1% bovine serum albumin) for 1 h at room temperature (RT). Testis sections were then incubated with primary antibodies (Table S10) diluted in blocking solution overnight at 4°C. After washing with PBS, sections were incubated with Alexa Fluor 488 goat anti-rabbit IgG (1:500; A32731, Invitrogen) and/or Alexa Fluor 594 goat anti-mouse IgG (1:500, A11032, Invitrogen) for 1 h at RT and then stained with DAPI for 5 min, washed in PBS and mounted using 80% glycerol. Confocal fluorescence microscopy was conducted using a confocal A1 laser microscope (Nikon, A1 HD25).

Testes were fixed in Bouin's solution, embedded in paraffin and sectioned (5 μm). TUNEL staining was performed using One Step TUNEL Apoptosis Assay Kit (Meilunbio, MA0223) according to the manufacturer's instructions. Images were obtained using a FluoView 1000 microscope (Olympus).

Transmission electron microscopy was performed as previously described with some modifications (Wang et al., 2018). In brief, testes from control and cKO mice were fixed in 0.1 M cacodylate buffer (pH 7.4) containing 3% PFA and 3% glutaraldehyde plus 0.2% picric acid for 2 h at 4°C, then for 1 h at RT. After three washes with 0.1 M cacodylate buffer, the samples were post-fixed with 1% OsO4 for 1 h at RT. Then the samples were dehydrated in an ethanol series and embedded in an Eponate mixture (Electron Microscopy Sciences) for polymerization for 24 h at 60°C. Ultrathin sections (∼70 nm) were cut using a diamond knife. The sections were re-stained with uranyl acetate and lead citrate and then photographed using a TEM (FEI Tecnai G2 12). For the calculation of the ERMICC contact index, the formula is: LIN/(PerM×DistER-M). LIN represents the interface length between mitochondria and ER; PerM indicates mitochondria perimeter; DistER-M is the distance between mitochondria and ER. Approximately 100 mitochondria were used to calculate for Mito-ER distance, Mito-ER contacts, and ERMICC contact index.

HEK293T (293T) cells were cultured in DMEM medium with 10% fetal bovine serum (Gibco, 10270106). Transfection was performed using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. For the transfection of the plasmid, 4 μg DNA was used in a 45 mm diameter plate.

For Sertoli cell isolation, testes from adult mice were dissected in DHANKS medium and incubated in DMEM/F12 medium containing 1 mg/ml collagenase IV, 0.5 mg/ml deoxyribonuclease I and 0.5 mg/ml hyaluronidase for 10 min at 37°C. After centrifugation (100 g), the precipitates were collected and incubated in the DMEM/F12 medium with 2.5 mg/ml trypsin and 0.5 mg/ml deoxyribonuclease I for 10 min at 37°C. Then seminiferous tubules were incubated with DMEM/F12 containing 10% fetal bovine serum for 5 min to inhibit the trypsin digestion. Subsequently, Sertoli cells were collected after filtering through 40 μm pore-size nylon mesh and stored at −80°C for the experiments.

Protein extracts were prepared from mouse tissues using RIPA lysis buffer [150 mM sodium chloride, 1.0% NP-40, 0.5% sodium deoxycholate, 0.1% SDS and 50 mM Tris (pH 8.0)]. Protein extracts were denatured with 2× loading buffer [4% SDS, 10% 2-mercaptoethanol, 20% glycerol, 0.004% bromophenol blue, 0.125 M Tris-HCl (pH 6.8)] at 95°C for 10 min and run on a 10% SDS-PAGE, followed by transfer to PVDF membrane. The membrane was probed with primary antibodies (Table S10) followed by secondary antibody treatment at 1:10,000 dilutions (anti-mouse HRP and anti-rabbit HRP, Abbkine, A25012 and A21020, respectively). Immun-Star™ HRP (1705040, Bio-Rad) was used for chemiluminescence detection, and was photographed using the ChemiDoc XRS+ system (Bio-Rad).

For in vivo Co-IP, mouse testes were homogenized in lysis buffer [20 mM HEPES (pH 7.3), 150 mM NaCl, 2.5 mM MgCl₂, 0.2% NP-40 and 1 mM DTT] with protease inhibitor cocktail (P1010, Beyotime). The relevant antibody and pre-cleaned magnetic protein A/G beads (blocking beads with bovine serum albumin 10% for 1 h to reduce the background before use) were added to the tissue lysate. RNaseA digestions were performed by adding RNase A (TransGen Biotech, GE101-01) at a final concentration of 500 μg/ml, followed by incubation for 3 h at 4 C followed by 30 min at RT. The lysate was incubated overnight at 4°C with gentle agitation to form the immunocomplex, and the magnetic beads were washed using the lysis buffer five times. The pellet was re-suspended with 40 μl elution buffer [1% SDS, 10 mM EDTA, 50 mM Tris (pH 8.0), 1× protease inhibitor] for 30 min at 70°C (with vortexing every 5-10 min). The eluate was transferred to a fresh tube, then 10 μl 5× SDS loading buffer was added, followed by boiling for 10 min. For in vitro Co-IP, full-length Mfn2 complementary DNA (cDNA) was cloned into a modified pcDNA3 vector encoding a 2× FLAG-tag, and full-length Miwi, Ddx4, Msy2 and Tdrkh cDNAs were cloned into a pCMV vector containing the N-terminal c-Myc epitope tag (Clontech Laboratories, K6003-1). HEK293T cells were transfected with indicated plasmids using Lipofectamine 2000 (Life Technologies). After 48 h, immunoprecipitation was performed using anti-MYC rabbit polyclonal antibody (4 μg antibody in 500 μl cell lysate, 10828-1-AP, Proteintech), followed by a western blot to identify protein interactions.

To quantify the amount of mtDNA present per nuclear genome, total DNA was extracted using the DNeasy Blood and Tissue Kit (Qiagen) and used for the detection of mtDNA copy number by quantitative real-time PCR (qPCR) using Nd1, 16S and Hk2 primers (Table S11) and quantified using the 2^−ΔΔCt method.

COX and SDH activities were investigated separately using procedures modified from Ross (2011) and Jiang et al. (2017). Briefly, fresh testes tissues were quickly collected, rapidly frozen in liquid nitrogen, embedded in OCT compound, cryosectioned into 10-μm-thick sections, mounted on slides and left to air dry briefly. For COX staining, freshly prepared incubation buffer A (0.8 ml 5 mM 3,3′-diaminobenzidine tetrahydrochloride, 0.2 ml 500 mM cytochrome c, 10 ml catalase) was added to the slides. After incubation for 40 min at 37°C, slides were washed three times with 0.1 M PBS (pH 7.0) and dehydrated in an ethanol series (70%, 70%, 95%, 95%, 99.5%). Slides were then cleared with at least two changes of xylene and mounted with Permount Mounting Medium (Sinopharm, 10004160). For SDH staining, freshly prepared incubation buffer B [0.8 ml 1.875 mM nitroblue tetrazolium (NBT), 0.1 ml 1.3 M sodium succinate, 0.1 ml 2.0 mM phenazine methosulphate, 10 ml 100 mM sodium azide] was applied and incubated for 40 min at 37°C. Slides were washed in three changes of 0.1 M PBS, incubated with acetone solutions (30%, 60% and 90% in increasing then decreasing concentration) to remove unbound NBT, and mounted with an aqueous mounting medium for bright-field microscopy.

Total RNA was extracted from mouse tissues using Trizol reagent (Invitrogen, 15596-025) following the manufacturer's instructions. For cDNA synthesis, 1 μg of RNA was treated with DNase I (Promega, M6101) to remove the residual genomic DNA and reverse transcribed with cDNA Synthesis Kit (Thermo Fisher Scientific, 4368814). RT-qPCR was performed using an SYBR Premix on a StepOne Plus machine (Applied Biosystems, 4309155). Relative gene expression was analyzed using the 2^−ΔΔCt method with Arbp as an internal control. All primers are shown in Table S11.

The STA-PUT method based on sedimentation velocity at unit gravity (Bellve, 1993; Bellve et al., 1977) was used to purify the PS and RS from adult WT and Mfn2-cKO mouse testes. We assessed cellular morphology using phase contrast microscopy to determine the purity of the cell population. PS and RS with ≥90% purity were used for RNA-seq analyses.

After purification of PS and RS from WT and Mfn2-cKO testes, total RNA was isolated using the Trizol reagent following the manufacturer's protocol and treated with DNase I to digest residual genomic RNA. The purity, concentration and integrity were assessed using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific), a Qubit RNA Assay Kit in Qubit 2.0 Fluorometer (Life Technologies) and a Nano 6000 Assay Kit of the Bioanalyzer 2100 system (Agilent Technologies), respectively. Then, a total amount of 2 µg of RNA per sample was used to prepare poly(A+)-enriched cDNA libraries using the NEBNext Ultra RNA library Prep Kit for Illumina (New England Biolabs) according to the manufacturer's instructions, and base pairs (raw data) were generated by the Illumina Hi-Seq 2500 platform. Raw reads were processed with cutadapt v1.9.1 to remove adaptors and perform quality trimming. Trimmed reads were mapped to the UCSC mm10 assembly using HiSAT2 (V2.0.1) with default parameters. Differentially expressed genes for all pairwise comparisons were assessed by DESeq2 (v1.10.1) with internal normalization of reads to account for library size and RNA composition bias. Differentially regulated genes in the DESeq2 analysis were defined as twofold changes with an adjusted Ρ-value of <0.05. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses were conducted using the database for annotation, visualization and integrated discovery (DAVID) (Dennis et al., 2003). The rich factor=number of DEGs in GO term/total number of genes in GO term. The larger the rich factor is, the higher enrichment is.

For preparing small RNA-seq libraries, two biological testis samples at P25 from WT and Mfn2-cKO mice were collected. Total RNAs were extracted using TRIzol reagent (Invitrogen) and subsequently processed to prepare small RNA-seq library and sequenced as previously described (Dong et al., 2019). For the length distribution of piRNA analysis, data were normalized by miRNA reads (21-23 nt) from small RNA-seq.

Testicular seminiferous tubules were dissected and collected according to previous literature with minor modifications (Aboulhouda et al., 2017; Karamysheva et al., 2018). Briefly, after adding polysome extraction buffer [20 mM Tris-HCl (pH 7.4), 100 mM KCl, 5 mM MgCl₂, 1 mM DTT, 0.5% NP-40, 1× protease inhibitor cocktail (EDTA-free), 200 μg/ml CHX, 200 units/ml RNase inhibitor], the tubules were transferred to a small (0.5-1.0 ml) Dounce homogenizer to disrupt the tissues with seven to eight strokes of the glass pestle. The samples were centrifuged at 12,000 g at 4°C for 10 min to collect the supernatant into a new tube. Then, DNase I (final 5 U/ml) was added into the tube, and the tube was placed on ice for 30 min to allow the DNase I to degrade any DNA contamination. Equal amounts of cytoplasmic lysates were carefully loaded on the top of a 10-50% sucrose gradient. All tubes were equally balanced, and then the gradients were centrifuged for 3 h in SW41Ti swinging bucket rotor at 160,000 g at 4°C using a Beckman ultracentrifuge. Fractions (300 μl/tube) were collected manually and immediately transferred to an ice bucket. The absorbance was detected at 260 nm to display the polysome profile of the gradients. The proteins were precipitated with nine volumes of 96% ethanol.

All data are presented as mean±s.e.m. unless otherwise noted in the figure legends. Statistical differences between datasets were assessed using one-way ANOVA or two-tailed Student's t-test using the SPSS16.0 software. The Mann–Whitney U-test (a non-parametric statistical method) was used to analyze Fig. 3C-G. P-values are denoted in figures by *P<0.05; **P<0.01; ***P<0.001.

The representation factor calculation was performed to identify whether the expected number of overlaps was found between Mnf2-cKO and Miwi-KO PS and RS up- or downregulated genes using the online resource from Lund lab available at http://nemates.org/MA/progs/overlap_stats.html. A representation factor <1.0 has fewer than expected overlap, whereas a representation factor >1.0 has more overlaps than expected. The significance of this overlap was calculated using the exact hypergeometric distribution or normal approximation of the hypergeometric distribution calculations where appropriate (Cahill et al., 2018; Fury et al., 2006; Lahiri et al., 2007). The hypergeometric distribution is similar to the one-tailed Fisher's exact test, and is commonly used to assess statistical significance of gene overlaps and/or enrichment (Fury et al., 2006; Plaisier et al., 2010). This calculation describes the expected number of successes (overlapping genes) in a series of draws from a finite population (experiment 1 genes and experiment 2 genes) without replacement (Plaisier et al., 2010). The normal approximation of the hypergeometric distribution is a method used under circumstances that the number of genes in experiment 1 (here taken to be the up-/downregulated Mnf2-cKO genes)×10<the total genes (total sequenced genes in PS/RS in the Mnf2-cKO) (Kim et al., 2001; Lahiri et al., 2007). Significance was considered to have a P-value<0.05.

We are grateful for engaging discussions with colleagues from Huazhong University of Science and Technology, China, in the very initial phase of the project.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/148/7/dev196295/