Nuclear translocation of MTL5 from cytoplasm requires its direct interaction with LIN9 and is essential for male meiosis and fertility

Meiosis is essential for the generation of gametes and sexual reproduction, yet the factors and underlying mechanisms regulating meiotic progression remain largely unknown. Here, we showed that MTL5 translocates into nuclei of spermatocytes during zygotene-pachytene transition and ensures meiosis advances beyond pachytene stage. MTL5 shows strong interactions with MuvB core complex components, a well-known transcriptional complex regulating mitotic progression, and the zygotene-pachytene transition of MTL5 is mediated by its direct interaction with the component LIN9, through MTL5 C-terminal 443–475 residues. Male Mtl5c-mu/c-mu mice expressing the truncated MTL5 (p.Ser445Arg fs*3) that lacks the interaction with LIN9 and is detained in cytoplasm showed male infertility and spermatogenic arrest at pachytene stage, same as that of Mtl5 knockout mice, indicating that the interaction with LIN9 is essential for the nuclear translocation and function of MTL5 during meiosis. Our data demonstrated MTL5 translocates into nuclei during the zygotene-pachytene transition to initiate its function along with the MuvB core complex in pachytene spermatocytes, highlighting a new mechanism regulating the progression of male meiosis.

Introduction Infertility, which affects roughly 15% of reproductive-aged couples worldwide, is emerging as both a serious medical issue as well as a major social problem [1][2][3][4]. Although artificial assisted reproductive technology can enable an increasing number of infertile couples to give birth to their own children, men with infertility caused by meiotic dysfunction do not benefit from these techniques due to the lack of gametes [5,6]. Thus, studies regarding genes that are essential for meiotic progression help increase our fundamental understanding of the mechanisms controlling meiosis [7,8].
Meiotic prophase I, which contains a series of meiosis-specific events, including meiotic recombination and synapsis, is equivalent to the G2 phase of the cell cycle for somatic cells [9][10][11][12][13][14]. It takes longer (lasting around 13 days in mice) than mitotic cells to guarantee the orderly progression of these meiotic events. The meiotic G2 phase (meiotic prophase I) is divided into a series of sub-stages, including leptotene, zygotene, pachytene, diplotene, and diakinesis, according to the chromosome configurations and structure [15]. Its progression is strictly regulated by different checkpoints that monitor the proper occurrence of meiotic events, which lead to meiotic arrest (usually at pachytene stage) when meiotic errors take place. This suggests the existence of a cell cycle regulator that promotes or blocks meiotic progression under the regulation of meiotic checkpoints.
The MuvB core complex, which consists of LIN9, LIN54, LIN37, LIN52, and RBBP4, is a well-known cell cycle regulator that serves essential roles during various phases of mitosis cell cycle by binding and directing different key transcription factors to the promoters of cell cycle genes [16,17]. During the G0 and early G1 phase, the MuvB core interacts with E2F4/5, DP1/2, and p130/p107 to form the DREAM complex (dimerization partner (DP), RB-like, E2F and MuvB) that represses the expression of genes. These co-factors are then released from the MuvB core complex to allow the recruitment of B-MYB to MuvB core complex (MMB complex) in the S and early G2 phase, promoting G2/M transition. This complex further recruits FOXM1 in the late G2 phase to activate G2/M genes [16,17]. However, it is unknown whether and how this complex functions during meiotic prophase I. MTL5 (Metallothionein-like 5, also known as Tesmin) is a testis-specific metallothioneinlike protein which contains tandem cysteine-rich regions (CRC domain, the DNA-binding domain), and it shares sequence similarity with LIN54 [18]. Mouse Mtl5 gene locates on chromosome 19 and it has two transcript isoforms, with the longer and more common transcript, encoding a 475 amino acids (aa) protein, specifically expressed in mouse testes [19]. As a testis-specific metallothionein, Mtl5 mRNA expression is first detected at 8 days post-partum (dpp), which is consistent with the initiation of meiosis [20]. Despite of the early expression of mRNA, homozygous Mtl5 deletion in mice does not cause any obvious defects in the leptotene and zygotene stages, but results in male infertility due to spermatogenic arrest at the pachytene/diplotene stage [21].
Intriguingly, immunocytochemical analysis of MTL5 localization in mouse testes by Matsuura et al. showed that it shuttles between the cytoplasm and nuclei of spermatocytes during spermatogenesis [22]. Sutou et al. found that MTL5 mainly locates in the cytoplasm of spermatocytes in stages I-IX tubules, which corresponds to early and mid-pachytene spermatocytes, but then translocates into the nuclei of spermatocytes at the late pachytene or diplotene stages, which are present in stage X-XII tubules [19]. Taken together, these reports suggested a possible relationship between the nuclear translocation of MTL5 and its function.
To uncover the mechanism and biological significance of the translocation of MTL5 from the cytoplasm to nucleus, we first studied its localization in mouse testes and found that MTL5 translocates to the nucleus around zygotene-pachytene transition. We then showed that MTL5 interacts with all the MuvB core complex components except LIN54 in spermatocytes and transportes into the nuclei directly by LIN9 in cultured cells. We also showed that the nuclear translocation of MTL5 is carried out by LIN9, the subunit of the MuvB core complex which is well-known in the DREAM complex [23,24], through its direct interaction with the C-terminal 443-475 aa region of MTL5. A truncation mutation of MTL5 (p.Ser445Arg fs � 3), which generated the Mtl5 c-mu/c-mu mouse line, eliminated its interaction with LIN9, and consequently led to cytoplasmic retention of MTL5 in pachytene spermatocytes. Male mice homozygous for the mutation showed spermatogenesis arrest at pachytene and azoospermia, exactly the same phenotype as in Mtl5 knockout (Mtl5 -/-) mice, indicating that MTL5 is transported into the nuclei by LIN9 during zygotene-pachytene transition to promote meiotic progression beyond pachytene stage.

MTL5 shifts from the cytoplasm to nuclei of spermatocytes during zygotene-pachytene transition
Previous studies in mice reported that MTL5 localized to the cytoplasm or nucleus in a spermatogenic cell type-specific manner [19], but the stage at which MTL5 translocates from cytoplasm into nuclei of spermatocytes during spermatogenesis was not fully resolved. To determine the exact stage in which MTL5 translocates from cytoplasm to nucleus, we analyzed the expression and localization patterns of MTL5 during spermatogenesis of mice in detail. MTL5 protein is expressed specifically in mouse testes (S1A Fig) and was first detected at 10 dpp in mouse testes (S1B Fig), which corresponds to the leptotene and zygotene stages of the first wave of spermatogenesis. Consistent with this finding, RT-PCR on isolated spermatogenic cells showed that Mtl5 expression was initially detected in leptotene and zygotene spermatocytes, then increased sharply in pachytene and diplotene spermatocytes, and finally disappeared in spermatids (S1C Fig).
The localization of MTL5 was determined by the co-staining of MTL5 and γH2AX on either testicular sections or smeared testicular cells (Fig 1A and 1B). The MTL5 signal was first observed in cytoplasm of leptotene and zygotene spermatocytes, which showed dispersed γH2AX signals, and became stronger in the nuclei of pachytene and diplotene spermatocytes, while no MTL5 signals were detected in pre-meiotic and Sertoli cells (Fig 1A and 1B). To detect the dynamic nuclear localization of MTL5 in pachytene/diplotene spermatocytes, immunostaining of MTL5 and PNA were performed on testicular sections from 10-week-old wild-type mouse and seminiferous tubule stages were distinguished according to the morphology of PNA staining (Fig 1C). The localization of MTL5 was observed in both cytoplasm and nuclei of the early pachytene spermatocytes in stage II, and then gradually accumulated in the nuclei of pachytene/diplotene spermatocytes from stage III to VIII (Fig 1C). Almost no cytoplasm localization of MTL5 was observed in pachytene/diplotene spermatocytes from tubules after stage VIII ( Fig 1C). Then, we stained H1t, which is expressed in germ cells after the midpachytene stage, to further detect the dynamics of MTL5 nuclear localization on spermatocyte spreads ( Fig 1D). MTL5 signals were first observed in the nuclei from early-pachytene Immunostaining for MTL5 (red) and γH2AX (green) in smeared testicular cells of 10-week-old wild-type mice for MTL5 (red) and γH2AX (green). Nuclei were spermatocytes, and became stronger in the nuclei of the mid-late pachytene and diplotene spermatocytes (Fig 1D).
To further confirm the translocation of MTL5, we separated the cytoplasm and nuclei of spermatocytes from 8 dpp to 25 dpp mice and determined the level of MTL5 by Western blot (Fig 1E). Consistent with the immunofluorescence localization, no MTL5 was detected in either cytoplasmic or nuclear lysates of 8 dpp testes, but was subsequently observed in cytoplasm lysate of 10 dpp testes and then for the first time in nuclei of spermatocytes in 12 dpp testes ( Fig 1E). With the progression of the spermatogenic wave, MTL5 protein levels gradually increased in both cytoplasmic and nuclear lysates of spermatocytes from 15 dpp onward ( Fig  1E).
These results indicate that MTL5 expression is restricted to the cytoplasm of spermatocytes during the leptotene and zygotene stages, and then starts to translocate into the nuclei of germ cells after the zygotene-pachytene transition.

MTL5 is transported into the nuclei by the MuvB core complex component LIN9
Based on protein structure analysis, MTL5 does not have a clear nuclear localization sequence (NLS), strongly suggesting that other proteins mediate or contribute to its transport into nuclei. To identify the proteins that help the nuclear translocation of MTL5, we thus performed immunoprecipitation with an antibody against MTL5 in the pachytene and diplotene spermatocytes enriched from 10-week-old wild-type mice testes using the STA-PUT method [25], followed by mass spectrometry analysis (IP-MS) (Fig 2A and 2B). In total, 27 putative candidate MTL5 interacting proteins which contain at least 3 unique peptides were identified (S1 Table). Interestingly, besides MTL5 and MYBL1, 4 out of 5 components of MuvB core complex, including RBBP4, LIN9, LIN52, and LIN37, appeared at the top of candidate list ranked by coverage (%) (Fig 2C). Notably, MYBL1 is the homologue of MYBL2, which has been experimentally proven to interact with MuvB core complex in somatic cells [26][27][28]16]. To confirm the interaction of MTL5 with the MuvB core complex, we performed co-immunoprecipitation (co-IP) in lysates of HEK293T cells after co-expressing MTL5 with all the components of MuvB core complex (including LIN9, LIN37, LIN52, RBBP4, and LIN54) as well as MYBL1. As expected, the co-IP results showed that MTL5 interacted with MYBL1 and all MuvB core complex components ( Fig 2D) except LIN54 which is particular lowly expressed in testis (S2 Fig).
To determine which component(s) of MuvB core complex interact(s) directly with MTL5, we examined the interactions between MTL5 (prey, transformed into the Y187 yeast haploid strain) and LIN9, LIN37, LIN52, RBBP4, or MYBL1 (baits, transformed into the AH109 yeast haploid strain) by using yeast two-hybrid (Y2H) screening. Only the transformed AH109 and Y187 yeast haploid cells with MTL5 and LIN9 or MTL5 and RBBP4 could grow on all selective medium plates (S3 Fig), indicating that MTL5 interacts with LIN9 and RBBP4 directly.
Based on the above findings, we propose that the nuclear translocation of MTL5 from cytoplasm is mediated by binding to LIN9 or/and RBBP4. We thus examined the localization of MTL5 in HEK293T cells co-transfected with plasmids encoding MYC-Mtl5 and plasmids counterstained with Hoechst 33342 (blue). White arrowheads indicate pachytene or diplotene spermatocytes. Green dotted ovals indicate zygotene spermatocytes. Scale bars, 10 μm. (C) Immunostaining of MTL5 (green) and PNA (red) in testicular sections of 10-week-old wild-type mice. Nuclei were counterstained with Hoechst 33342 (blue). Enlarged views are indicated by white boxes. Scale bars, 50 μm. (D) Immunostaining for SYCP3 (red), H1t (gray), and MTL5 (green) in spread spermatocytes of 10-week-old wild-type mice. Scale bars, 10 μm. (E) Western blot analyses were performed using cytoplasmic and nuclear protein extracts from 8, 10, 12, 15, 20, and 25 days post-partum (dpp) mouse testes. GAPDH and SYCP3 served as cytoplasmic and nuclear controls, respectively.
https://doi.org/10.1371/journal.pgen.1009753.g001 encoding GFP-Lin9 or GFP-Rbbp4, respectively (for the controls, GFP-Lin37, GFP-Lin52, or GFP-Mybl1 encoded plasmids were also co-transfected with MYC-Mtl5 constructs respectively). MTL5 was diffusely localized in the cytoplasm of HEK293T cells when transfected with MYC-Mtl5 alone or with the pEGFP-N1 vector (Fig 3A and 3B). As expected, the cytoplasmic localization of MTL5 was also observed in cells with MTL5 co-expressed with LIN37, LIN52, or MYBL1, because no direct interaction of MTL5 with them was detected in Y2H screening (Fig 3C-3E). What is more, RBBP4, a previously identified direct interacting partner of MTL5, only showed a cytoplasmic localization and failed to transport MTL5 into the nucleus ( Fig 3F). Instead, co-expression with LIN9, another MTL5-interacting protein, caused a dramatic redistribution of MTL5 from cytoplasm to nuclei in HEK293T cells (Fig 3G), indicating that LIN9 may serve as the key factor for MTL5 nuclear translocation.

LIN9 transports MTL5 into nuclei by the direct interaction with the Cterminal 443-475 amino acids of MTL5 in cultured somatic cells
To determine which regions of MTL5 are responsible for direct interaction with LIN9, and subsequent importation into nuclei, we performed Y2H screening assays using truncated variants of MTL5 to identify which exhibited direct interactions with LIN9. MTL5 consists of 475 aa and contains a CRC domain in the region from 263 aa to 370 aa, as well as two predicted helix domains (from 420 aa to 439 aa and from 444 aa to 468 aa) in the C terminus. Thus  Y187 yeast haploid strain and then hybridized with BD-LIN9, which was transformed into the AH109 yeast haploid strain. Only the transformed Y187 and AH109 yeast haploid cells transformed with AD-MTL5-C and BD-LIN9 could grow on both double and quadruple dropout medium plates (Fig 4A), suggesting that the region mediating direct interaction between MTL5 and LIN9 is located in the MTL5 C-terminal 371-475 aa. We also cloned the fragments of LIN9 (S4B Fig) into the pGBKT7 expression vector, performed Y2H screening assays with AD-MTL5-C, and found that the C-terminus (341-559 aa) region of LIN9 exhibited direct interaction with the C terminus of MTL5-C (371-475 aa) ( Fig 4B).
To map more precisely the region(s) of MTL5 that interact with LIN9, we further cloned three fragments of MTL5 C-terminal 371-475 aa into pGADT7 expression vectors to generate AD-MTL5-C1 (371-418 aa), AD-MTL5-C2 (419-442 aa), and MTL5-C3 (443-475 aa) (S4A Fig). Only the transformed Y187 and AH109 yeast haploid cells with AD-MTL5-C3 and BD-LIN9-C grew on all selective medium plates (Fig 4C), suggesting that the C-terminal 443-475 aa of MTL5, which contains the second predicted helix domain from 444 aa to 468 aa, is essential for its direct interaction with LIN9. Given that the C-terminal 443-475 aa of MTL5 is necessary for its direct interaction with LIN9, we hypothesized that this direct interaction might be essential for the nuclear translocation of MTL5 from the cytoplasm. To test this, we generated a MYC-Mtl5-ΔC3 vector that encodes a truncated MTL5 lacking the C-terminal 443-475 aa (S4C Fig), and then transfected it into HEK293T cells with the GFP-tagged mouse LIN9 (GFP-Lin9) vectors. It is shown that LIN9 failed to transport MTL5-ΔC3 into the nuclei of transfected cells (Fig 4D), but was able to transport the truncated MTL5 with deletion in other regions into nuclei (S5 Fig). These results further indicate that the C-terminal 443-475 aa of MTL5 is essential for its translocation from the cytoplasm to the nucleus, mediated by the direct interaction with LIN9.

Translocation of MTL5 into the nuclei of spermatocytes requires the MTL5 C-terminus and is essential for meiotic progression and sperm production
Given the function of LIN9 to translocate MTL5 into the nuclei of somatic cells, we hypothesized that LIN9 could be essential for the nuclear translocation of MTL5 in spermatocytes. To test this, we first examined whether LIN9 showed nuclear localization in pachytene spermatocytes. The localization of LIN9 was detected in 10-week-old mouse testes by immunofluorescence staining, and we found that LIN9 signals were observed in the nuclei of Sertoli cell, spermatogonia and pachytene/diplotene spermatocytes, while no obvious LIN9 signals were observed in leptotene/zygotene spermatocytes (S6A Fig). Furthermore, LIN9 protein was first detected in spermatocytes around 12 dpp, which is consistent with its appearance in pachytene spermatocytes when nuclear translocation of MTL5 occurs (S6B Fig).
To further test whether the interaction between MTL5 and LIN9 is also essential for the nuclear translocation of MTL5 in spermatocytes, we generated MTL5 C-terminal truncated mice (Mtl5 c-mu/c-mu ), which produced a truncated protein p.Ser445Arg fs � 3. This mutation is predicted to ablate the second potential helix domain within C-terminal 443-475 aa of MTL5, which abolished LIN9 interaction (Fig 5A and 5B). At the same time, we obtained the MTL5 knockout mice (Mtl5 -/-) (S7 Fig). Indeed, we detected the presence of truncated MTL5 at the expected size in Mtl5 c-mu/c-mu testes by Western blot using the anti-MTL5 N-terminal antibody (epitope 1-220 aa) (Fig 5C). The truncated MTL5 (p.Ser445Arg fs � 3) abolished its interaction with LIN9 in HEK293T cells (Fig 5D). The disturbed interaction between MTL5 and LIN9 was also confirmed in 14 dpp Mtl5 c-mu/c-mu mice testes ( Fig 5E). As expected, we did not see any MTL5 signals present in the nuclei of pachytene spermatocytes in Mtl5 c-mu/c-mu testes examined using immunofluorescence staining (Fig 5F). Consistent with the immunofluorescence results, no MTL5 protein was detected in the nuclear extract of Mtl5 c-mu/c-mu mice ( Fig 5G). Taken together, these results further indicate that the C-terminal 445-475 aa of MTL5 is responsible for MTL5 nuclear translocation in spermatocytes and highlights an essential role of the LIN9-MTL5 interaction in MTL5 nuclear transportation in spermatocytes during spermatogenesis.
To determine whether the nuclear translocation of MTL5 is required for its function in vivo, we examined the spermatogenesis and meiosis in Mtl5 c-mu/c-mu mice, and compared with those in Mtl5 +/+ and Mtl5 -/mice (Fig 6). The Mtl5 c-mu/c-mu mice showed smaller testes than their Mtl5 +/+ littermates, with a significant decrease in the testis/body weight ratio to 2.16, which is comparable to that of Mtl5 -/mice (Fig 6A and 6B). H&E staining showed spermatogenic arrest at the primary spermatocyte stage in Mtl5 c-mu/c-mu mice, which was indistinguishable from that of Mtl5 -/mice ( Fig 6D). In agreement with these findings, no sperm were observed in the cauda epididymis from Mtl5 c-mu/c-mu mice, same as that seen in the Mtl5 -/mice cauda epididymis (Fig 6C and 6D). This observation is also consistent with those in the reported Mtl5 knockout mice [21]. Additionally, all of the germ cells were SYCP3 positive, To further determine the exact stage at which meiosis is arrested in Mtl5 c-mu/c-mu mice, we performed immunostaining for SYCP3 and SYCP1, the synapsis makers between homologous chromosomes, on spermatocyte spreads to examine meiotic prophase progression in Mtl5 +/+ , Mtl5 c-mu/c-mu , and Mtl5 -/mice. It is shown that 207 diplotene spermatocytes, out of 921 meiotic prophase I cells examined, were observed in Mtl5 +/+ testes, while no diplotene cells were seen in either Mtl5 c-mu/c-mu or Mtl5 -/testes after scoring 767 and 976 prophase I spermatocytes, respectively (Fig 6E and 6F). Compared with those in controls, the ratios of leptotene and zygotene spermatocytes increased significantly in Mtl5 c-mu/c-mu and Mtl5 -/testes due to the lack of diplotene spermatocytes, while the ratio of pachytene spermatocytes was not significantly changed, indicating loss of pachytene spermatocytes in Mtl5 c-mu/c-mu and Mtl5 -/mice (Fig 6E and 6F). Consistent with this, TUNEL assay showed a significant increase in the number of apoptotic spermatocytes in testes of Mtl5 c-mu/c-mu and Mtl5 -/mice (S9 Fig).
Given the similar ratios of spermatocytes at different sub-stages in Mtl5 c-mu/c-mu and Mtl5 -/mice, we further determined whether the C-terminal mutation of MTL5 cause spermatogenesis arrest at the same stage with Mtl5 -/mice. First, we compared the testicular histomorphological sections at 10, 12, 15, 20, and 25 days post-partum (dpp) from control, Mtl5 c-mu/c-mu , and Mtl5 -/mice, and observed nuclear morphology abnormality in germ cells from Mtl5 c-mu/c-mu and Mtl5 -/mice since 15 dpp, which were distinguishable from control mice, indicating spermatogenesis arrests between 12 to 15 dpp (before late pachytene) in both Mtl5 c-mu/c-mu and Mtl5 -/mice (S10 Fig). To further confirm the spermatogenesis arrest before late pachytene stage, we stained H1t on Mtl5 +/+ , Mtl5 c-mu/c-mu , and Mtl5 -/spermatocyte spreads (S11 Fig). We found no pachytene spermatocytes with thicker ends of lateral elements or strong H1t signal (characteristics of late pachytene) in Mtl5 c-mu/c-mu and Mtl5 -/mice, indicating similar spermatogenesis arrest before late pachytene in both mutant mice (S11 Fig). To examine whether homologous recombination was affected in Mtl5 c-mu/c-mu and Mtl5 -/mice, we counted foci of meiotic DSB markers RPA2 (S12A and S12B Fig) and DMC1 (S12C and S12D Fig) in zygotene and early-pachytene spermatocytes. No obvious difference in the numbers of RPA2 or DMC1 foci were observed in Mtl5 c-mu/c-mu zygotene and early-pachytene spermatocytes, when compared with the wild-type controls (S12 Fig). We also stained SYCP1 and SYCP3 to detect synapsis of pachytene spermatocytes in Mtl5 +/+ , Mtl5 c-mu/c-mu , and Mtl5 -/mice. We found that no synapsis defects were observed in about 60% pachytene spermatocytes, even in the remained pachytene spermatocytes, the synapsis defects were only involved in one or two autosomal pairs in the mutants (S13 Fig). Thus, it indicates that there is only a mild synapsis defect in our mutant mice and this synapsis defect is not likely the main cause of pachytene arrest.
Taken together, these results indicate that the nuclear translocation of MTL5 in spermatocytes, mediated via direct interaction with LIN9, is essential for meiotic progression during spermatogenesis, sperm production, and thus for male fertility (Fig 7).

Discussion
In the current study, we characterized the dynamic localization of MTL5 in spermatocytes of different meiotic sub-stages and highlighted the significance of MTL5 nuclear translocation in the zygotene-pachytene transition during spermatogenesis. We showed that the MTL5 protein accumulates in the cytoplasm of leptotene and zygotene spermatocytes and then begins to translocate into the nuclei of spermatocytes around the zygotene-pachytene transition. Further experiments indicated that MTL5, mediated by its C-terminus, interacts directly with LIN9, a component of MuvB core complex, by which it is transported into the nuclei of spermatocytes from the cytoplasm. The reproductive phenotype of MTL5 mutant (deleting C-terminal 445-475 aa of MTL5) and knockout mice demonstrated that the nuclear translocation of MTL5 in the spermatocytes is required for MTL5 to function, as the deletion of MTL5 C-terminal 445- 475 aa prevented its nuclear translocation and led to meiosis arrest at the pachytene stage during spermatogenesis, exactly the same phenotype as observed in Mtl5 -/mice.
Spermatogenesis is a complex developmental progress and requires a series of proteins to shuttle between the cytoplasm and nuclei of germ cells [29]. For example, cyclin D2 was observed in the nuclei of pachytene spermatocytes (G2 stage), and cyclin D3 was detected in the cytoplasm of elongating spermatids [30,31], while in somatic cells, the D-type cyclins are generally accumulated in the nucleus at the G1 phase and then are rapidly exported to cytoplasm for degradation at the onset of the S phase [32,33], suggesting a specific mechanism that regulates the nuclear and cytoplasmic transition of functional proteins. However, it remained unknown how this mechanism is achieved to facilitate spermatogenesis. Here, we identified the localization of MTL5 in spermatocytes and revealed that MTL5 begins to translocate into the nuclei around the zygotene-pachytene transition. Thus, the specific cytoplasm and nucleus distribution of MTL5 in spermatocytes of different stages provides us a novel model to understand the mechanism and function of the cytoplasmic-nucleus translocation of proteins during spermatogenesis.
The translocation of proteins into the nucleus is a fundamental process and is critical to the basic function of cells. Although many proteins move into the nucleus by the conserved nuclear localization sequence (NLS) [34,35], no canonical NLS is observed in the MTL5 sequence. We demonstrated that the MTL5 cytoplasmic-nucleus translocation is achieved by LIN9, which is a nuclear and chromatin-associated protein [36], through their direct interaction. Prevention of the interaction between MTL5 and LIN9 results in cytoplasmic retention of MTL5, which consequently leads to spermatogenesis arrest at the pachytene stage, same as observed in Mtl5 -/mice. Intriguingly, no LIN9 protein was observed in leptotene and zygotene spermatocytes in wild-type testicular sections (S6A Fig) and LIN9 protein was first detected in the spermatocytes around 12 dpp (S6B Fig), which is consistent with the nucleus translocation of MTL5, further confirming the essential role of LIN9 in transporting MTL5 into the nuclei of spermatocytes around the zygotene-pachytene transition. Thus, the transportation of MTL5 into the nuclei of spermatocytes by LIN9 is necessary for meiosis progress beyond the pachytene stage during spermatogenesis.
Mtl5 knockout mice was reported to experience arrest of spermatogenesis at the pachytene stages, with approximately 18.8% spermatocytes at the late pachytene and diplotene stages [21]. However, we did not find late pachytene or diplotene spermatocytes in either Mtl5 c-mu/cmu or Mtl5 -/mice. This inconsistency may be due to the criteria used to distinguish different meiotic sub-stages [9,37,38]. Oji et al. defined the late-pachytene and diplotene spermatocytes based on the formation of an intact sex body, which could also be observed in early or midpachytene spermatocytes [39][40][41]. However, based on the spermatocyte spreads stained for synaptonemal complex (SC) lateral element protein (SYCP3) and central element protein (SYCP1) as well as H1t staining signal, we did not find any cells with strong H1t signal and thicker ends of lateral elements or desynapsed SCs, characteristics of late pachytene or diplotene spermatocytes [37], in Mtl5 c-mu/c-mu or Mtl5 -/mice, indicating the arrest of spermatogenesis occurred before the late pachytene stage in these mice.
Homologous recombination and synapsis defects are two major causes of meiotic arrest at pachytene stage. However, we detected no obvious difference in the numbers of RPA2 and DMC1 foci among Mtl5 +/+ , Mtl5 c-mu/c-mu and Mtl5 -/spermatocytes at zygotene and earlypachytene stage, indicating the meiotic arrest in Mtl5 mutants was not a result of homologous recombination defect. Though we did observe mild synapsis deficiency in Mtl5 c-mu/c-mu and Mtl5 -/spermatocytes, the severity of the asynapsis is not expected to result in meiotic pachytene arrest [42,43].
We have demonstrated that MTL5 interacts with MYBL1 and all MuvB core complex components except LIN54 in mouse testes and cultured human somatic cells. This interaction with the MuvB core complex has been demonstrated in Arabidopsis thaliana [44]. It has been shown recently that TCX5, a tesmin/TSO1-like CXC domain-containing protein, is a component of the TCX5/6-containing multi-subunit complexes, which are also known as DREAM complex (a complex containing E2F/DP-p130/p107 and MuvB core complex) [44,16,17].The MuvB core complex is known to play essential roles during various phases of cell cycle, by binding and directing different key transcription factors to the promoters of cell cycle genes [26,45,28,16,17]. During mitosis, the MuvB core complex binds E2F/DP-p130/p107 to form the DREAM complex to repress cell cycle dependent gene expression in the G0 and early G1 phases, while it recruits B-MYB (MYBL2) proteins to form the Myb-MuvB (MMB) complex during the S and early G2 phases and recruits FOXM1 to the promoters of the G2/M genes to activate their expression [46,26,28,17]. In meiosis, deletion of MYBL1 (A-MYB, the paralog of B-MYB), which is highly expressed after meiotic entry, also leads to spermatogenic arrest before the late pachytene stage [47], the similar meiotic arrest stage in MTL5 deficiency mice, indicating an essential role for the Myb-MuvB complex in promoting meiotic progression through the mid-pachytene stage. Given that MYBL1 is pulled down by MTL5 together with the MuvB core complex components in the testes, our findings indicates that MTL5 is essential for meiosis progression beyond the mid-pachytene stage during spermatogenesis and that its deficiency leads arrest of spermatogenesis at the pachytene and consequently to azoospermia. Thus, it is likely that MTL5, after its nuclear translocation mediated by binding to LIN9 of the MuvB core complex, functions together with the Myb-MuvB complex to promote meiosis progression beyond the mid-pachytene stage during spermatogenesis (Fig 7).
It is well-known that cyclins and cyclin-dependent kinases (CDKs) are well-defined regulators of the cell cycle. In meiosis, cyclin/CDK complex is known to have unique characteristics and requirements [48,49,11,[50][51][52][53][54]13]. Some of them (such as Ccnb1, Ccnd2 and Ccnb2) only affect the meiotic cell cycle in female mice. Furthermore, deletion of Cyclin A1, which is highly expressed in testis, leads to male meiotic arrest at mid-diplotene [48,55];Cyclin E1 deficient testes are morphologically normal as wild-type testes, Cyclin E2 deletion leads to spermatid maturation arrest [50], while double mutant of Cyclin E1 and E2 causes meiotic arrest at a pachytene-like stage with severe synapsis defects [50];CDK2 is essential for the attachment of telomere to the nuclear envelope, and its deficiency leads to defects in homolog pairing due to the failure of bouquet formation [51]. Given the unique phenotype caused by Mtl5 mutation, we think that MTL5 regulates meiotic progression by affecting factors which are previously unknown.
MTL5 can pull down all of the components of the MuvB core complex except LIN54, a paralog of MTL5 [18,56] and is expressed at low levels in mouse testes (S3 Fig). In somatic cells, LIN54 was reported to serve as a key factor for directing the MuvB core complex to bind the cell cycle genes homology region (CHR), an regulatory element in the promoter regions of cell cycle-regulated genes [18], through its DNA binding domain (CRC domain) which shares sequence similarity with MTL5 [18,57,56]. Thus, we believe that MTL5 substitutes the function of LIN54 in the MuvB core complex and then directs the MuvB core complex to the promoters of essential genes required in pachytene spermatocytes. Taken together, our results showed that MTL5 is translocated into the nucleus by LIN9 to promote meiotic progression beyond the pachytene stage. The function of MTL5 is likely achieved by substituting LIN54 to form a meiosis-specific MuvB core complex to promote the meiotic cell cycle progression. Thus, our study not only identified MTL5 as a novel and germ-cell specific regulator of cell cycle progression to function at a specific stage in meiosis to ensure male fertility, but also sheds light on further research to find the master regulatory proteins that directly promote meiotic progression (Fig 7).

Ethics statement
All the experiments on mice followed the guidelines of the Institutional Animal Care Committee of the University of Science and Technology of China with the approval number USTCACUC1301021.

Mice
CRISPR/Cas9 genome editing was used to generate Mtl5 knockout and mutant mice as described previously [58,59]. The sgRNAs and Cas9 mRNA were microinjected into the zygotes of C57BL/6J mice, and the zygotes were then transferred into pseudo-pregnant ICR females. Genomic DNA was extracted from toe biopsies of founder mice for Sanger sequencing, and heterozygous founder mice were bred to produce homozygous mice. The founder female mice with a 19 bp deletion in exon 2 or a 76 bp deletion in exon 9 of Mtl5 (ENSMUST00000025840.15) was selected to produce homozygous Mtl5 knockout (Mtl5 -/-) and mutant (Mtl5 c-mu/c-mu ) mice, respectively. All the mice were nourished with proper food, ddH 2 O and maintained in a 12-hour photoperiod (lights on 08:00-20:00). The primers for genotyping and sgRNAs used in this study are listed in S2 Table.

Hematoxylin and eosin (H&E) staining and immunohistochemistry
The mice were euthanized by cervical dislocation, testes and epididymides were detached and immediately fixed overnight in Bouin's solution for H&E staining or in 4% paraformaldehyde (PFA) for immunohistochemistry at 4˚C. For immunohistochemistry, testicular and epidydimal sections were incubated with the primary antibodies overnight at 4˚C, and then probed with secondary antibodies for 1 hour at 37˚C mounting with Vectashield (Vector Laboratories, H-1000) containing Hoechst 33342 (Invitrogen, H21492). To reduce inter-experiment variation, testes from wild-type, knockout, and mutant mice were processed simultaneously. Slides were visualized and all images were captured using a Nikon Eclipse 80i microscope equipped with digital cameras (Nikon DS-Ri1 for H&E staining or Hamamatsu C4742-80 for immunohistochemistry). Some images were also captured using a Nikon C2 Plus Confocal Laser Scanning Microscope system.

RNA isolation and RT-PCR
Total RNA extraction and cDNA synthesis were performed using Trizol reagents (TaKaRa, 9109) and the PrimeScript RT kit (TaKaRa, RR047A) according to the respective manufacturer's protocols. RT-PCRs were performed using EasyTaq DNA Polymerase (TransGen Biotech, AP111) under the following conditions: 94˚C for 3 min, followed by 25~30 cycles of 30 s at 94˚C, 60˚C for 30 s, and at 72˚C for 30 s. The cDNA templates (S1B Fig) from previous studies in our laboratory were used to RT-PCR, as previously described [25]. Primers used for RT-PCR are listed in S2 Table.

Spermatocyte spreads and immunofluorescence staining
Spermatocyte spreads were prepared and immunofluorescence staining was performed as we previously described [25,60,61]. Vectashield (Vector Laboratories, H-1000) was used to seal the slides and images of meiotic cells were captured using a BX61 microscope (Olympus) connected to a CCD camera, and analyzed using Image-Pro Plus software (Media Cybernetic). The antibodies used for meiotic analysis are listed in S3 Table.

Co-immunoprecipitation (Co-IP) and mass spectrometry
The pachytene and diplotene spermatocytes were enriched using STA-PUT method [25] and then were lysed in cold RIPA buffer (Beyotime, P0013C) with phenylmethylsulfonyl fluoride (Thermo Fisher, 36978) and protease inhibitor cocktail (Roche, 04693116001). The lysates were sonicated for 8-10 cycles (3 s on/off) with 8% pulses and centrifuged at 15,000 xg in 4˚C for 15 minutes. The supernatant was divided into two aliquots, and each aliquot was incubated with pre-cleared Protein A/G agarose beads (Santa Cruz, sc-2003) and 2 μg anti-MTL5 C-terminal antibody (epitope from residues 221-475) or rabbit IgG non-specific antibody (ABclonal, AC005). After overnight incubation at 4˚C, the agarose beads were washed in RIPA buffer for 3 times and twice in 10 mM Tris buffer, and then the immunocomplexes were dissociated from the beads with elution buffer (0.2 M Glysine, 0.15% NP-40, pH 2.3) at room temperature. After validation of the eluted samples by silver staining, samples were analyzed by mass spectrometry in National Center for Protein Science Shanghai. The candidate interacting proteins of MTL5 are listed in S1 Table. TUNEL assay Cell apoptosis was detected in testicular sections by terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay according to the manufacturer's specifications (Roche, 11684795910, Basel, Switzerland). The images were captured using a Nikon ECLPSE 80i microscope (Nikon) which is equipped with a CCD camera (Hamamatsu) and analyzed with the NIS-Element Microscope imaging software (Nikon).

Cytoplasmic and nuclear protein extraction
The extraction of cytoplasmic and nuclear proteins was performed according to the manufacturer's specifications (Beyotime, P0027). GAPDH and SYCP3 were served as the loading controls of cytoplasm and nucleus, respectively.

Western blot
Western blot was carried out as we previously described [25] and the bands were visualized and examined with chemiluminescence (GE Healthcare, ImageQuant LAS 4000). The antibodies used in this study are listed in S3 Table.

Plasmids
The typical mouse Mtl5 (ENSMUST00000025840.15) full-length coding sequence was reverse transcribed from mouse testis cDNAs and cloned into a MYC-tagged expression vector under control of CMV promotor to form MYC-Mtl5. The mouse Lin9, Lin37, Lin52, Rbpp4, Mybl1, and Lin54 full-length coding sequences were also reverse transcribed from mouse testis cDNAs, and cloned into the pEGFP-N1 expression vector to form GFP-Lin9, GFP-Lin37, GFP-Lin52, GFP-Rbbp4, GFP-Mybl1, and GFP-Lin54 through the ClonExpress II one Step
Twenty-four hours of after transfection, cells were fixed in 4% paraformaldehyde followed by immunofluorescence. The primary antibodies were incubated at 4˚C overnight and followed with secondary antibodies for 1 hr at 37˚C. Finally, the cells were mounted with Vectashield. Pictures were captured using a Nikon C2 Plus Confocal Laser Scanning Microscope. The antibodies used for cell immunofluorescence are listed in S3 Table. For Co-IP, the cells were harvested 36 hours after transfection and lysed in pre-cold NP-40 buffer (Beyotime, P0013F) with phenylmethylsulfonyl fluoride (Thermo Fisher, 36978) and protease inhibitor cocktail (Roche, 04693116001). Lysate was then centrifuged and 1/10 volumes of supernatant were used as input samples. The remaining supernatant was incubated with Myc-Tag (9B11) mouse mAb (Magnetic bead conjugate) (Cell Signaling Technology, #5698). After incubation at 4˚C overnight with rotation, the magnetic beads were washed in NP-40 buffer and collected by magnetic stand. The proteins were eluted from the beads with sample buffer on a 100˚C hot block for 10 min.

Statistical analysis
The testis/body weight ratio, the spermatocyte proportions, TUNEL positive tubules and cells analysis, RPA2 and DMC1 foci analysis, H1t-staining analysis were compared between the wild-type, knockout, or mutant mice using the One-way ANOVA. The percentages of cotransfected cells with different MTL5 localization were analyzed by Student's t-test. The results were presented as means ± SEM. Statistical significance was set at P < 0.05. and LIN9 proteins were detected by immunostaining with antibodies against MTL5 (red) and GFP (green) and the protein localization was analyzed by a confocal laser scanning microscope. Nuclei (blue) were counterstained with Hoechst 33342. The "n" in brackets indicates the number of co-transfected cells analyzed. Data are presented as mean ± SEM for three independent experiments. �� p<0.01; ��� p<0.001; ���� p<0.0001. P values were determined by Student's t-test. Scale bars, 10 μm.