Dual functions for the ssDNA-binding protein RPA in meiotic recombination

Meiotic recombination permits exchange of genetic material between homologous chromosomes. The replication protein A (RPA) complex, the predominant ssDNA-binding complex, is required for nearly all aspects of DNA metabolism, but its role in mammalian meiotic recombination remains unknown due to the embryonic lethality of RPA mutant mice. RPA is a heterotrimer of RPA1, RPA2, and RPA3. We find that loss of RPA1, the largest subunit, leads to disappearance of RPA2 and RPA3, resulting in the absence of the RPA complex. Using an inducible germline-specific inactivation strategy, we find that loss of RPA completely abrogates loading of RAD51/DMC1 recombinases to programmed meiotic DNA double strand breaks, thus blocking strand invasion required for chromosome pairing and synapsis. Surprisingly, loading of MEIOB, SPATA22, and ATR to DNA double strand breaks is RPA-independent and does not promote RAD51/DMC1 recruitment in the absence of RPA. Finally, inactivation of RPA reduces crossover formation. Our results demonstrate that RPA plays two distinct roles in meiotic recombination: an essential role in recombinase recruitment at early stages and an important role in promoting crossover formation at later stages.


Introduction
During sexual reproduction, meiotic recombination permits reciprocal exchange of genetic materials between homologous chromosomes and ensures faithful chromosome segregation [1,2]. Abnormalities in meiotic recombination are a leading cause of aneuploidy, infertility, and pregnancy loss in humans [3]. Meiotic recombination is initiated by the formation of programmed DNA double strand breaks (DSBs) in germ cells and involves a large number of single-stranded DNA (ssDNA)-binding proteins [2]. These DSBs undergo end resection to generate 3' ssDNA overhangs; subsequent loading of RAD51 and DMC1 recombinases and other proteins enables strand invasion into duplex DNA for homologue pairing and recombination intermediate formation [4][5][6][7]. All meiotic DSBs are repaired but only a subset lead to crossovers, which are critical for proper segregation of homologous chromosomes during the first meiotic cell division.
The replication protein A (RPA) complex, comprised of RPA1, RPA2, and RPA3, is the predominant ssDNA-binding heterotrimeric complex in DNA metabolism [8]. RPA1, the largest subunit, is responsible for the majority of ssDNA-binding activity of the RPA complex. RPA protects ssDNA from degradation and prevents secondary structure formation. RPA interacts with both RAD51 and DMC1 [9], suggesting that RPA may direct RAD51 and DMC1 to these ssDNA overhangs. RAD51 and DMC1 form nuclear complexes on meiotic chromosomes [10,11]. The Hop2-Mnd1 heterodimer interacts with the RAD51/DMC1 recombinases and stimulates their enzymatic activity [12]. MEIOB is a meiosis-specific ssDNA-binding RPA1 homologue [13,14]. The MEIOB/SPATA22 heterodimer interacts with RPA and could also position critical proteins for meiotic recombination [15]. These interactions suggest that RPA may play multiple roles in meiotic recombination [16]. A long-standing question remains as to whether RPA is required for loading of these ssDNA-binding proteins to DSBs in vivo.
RPA is ubiquitously expressed and essential for DNA replication, repair, and recombination [8]. However, the physiological role of RPA in mammalian meiosis remains elusive, due to embryonic lethality of the Rpa1 mutation in mice [17]. As a result, the interplay of RPA with other ssDNA-binding proteins during meiotic recombination in vivo remains enigmatic. Here we employed a germ cell-specific inducible deletion approach and uncovered the functions of RPA in meiosis.
always colocalized with MEIOB and that RPA2 also always colocalized with MEIOB [13]. By SIM, we found that RPA1 foci were present in leptotene, zygotene, early pachytene, and mid pachytene spermatocytes, but absent in late pachytene and diplotene spermatocytes (Fig 1A-1F). We next examined colocalization of RPA2 with other DNA-binding proteins that are known to be involved in meiosis. While most RPA2 foci did not overlap with DMC1 foci ( Fig  1G), RPA2 foci always colocalized with MEIOB foci (Fig 1H). These results were consistent with previous conventional immunolocalization findings, suggesting that RPA may function in meiotic recombination in mouse [13,18].
To overcome the expected embryonic lethality of Rpa1 inactivation, we generated an Rpa1 floxed (Rpa1 fl ) allele that allows for germ cell-specific inactivation of Rpa1. We first produced Rpa1 fl/-Ddx4-Cre males. Ddx4-Cre begins to express only in germ cells at embryonic day 15 [19]. Rpa1 fl/-Ddx4-Cre males were viable but sterile. Testes from 10-week-old Rpa1 fl/-Ddx4-Cre males lacked all germ cells (S1A Fig). We also generated Rpa1 fl/-Stra8-Cre males. Stra8-Cre begins to express in spermatogonia postnatally, prior to meiosis [20]. Testes from adult Rpa1 fl/-Stra8-Cre males displayed a heterogeneous phenotype: some tubules were nearly depleted of all germ cells, while other tubules had apparently normal spermatogenesis (S1B Fig), presumably due to inefficient Stra8-Cre-mediated deletion as observed previously [21]. Although these genetic studies revealed an essential role for RPA1 in spermatogenesis, the loss of all germ cells in testes from Rpa1 fl/-Ddx4-Cre males or the presence of all germ cells in testes from Rpa1 fl/-Stra8-Cre males precluded investigating its function in meiosis.
To circumvent these hurdles, we performed tamoxifen-inducible inactivation of Rpa1 following crossing with Ddx4-Cre ERT2 mice (Fig 2A) [22]. Intraperitoneal injection of 8-weekold Rpa1 fl/fl Ddx4-Cre ERT2 (referred to as Rpa1 cKO ) male mice with tamoxifen resulted in the deletion of exon 8, which led to a frame shift in the resulting mutant transcript (Fig 2A and  2B). The abundance of the Rpa1 transcript (both pre and post exon 8 regions) decreased at days 2, 4, and 6 post tamoxifen treatment, suggesting that the Rpa1 transcript is depleted due to nonsense mediated decay and/or depletion of spermatocytes (S2 Fig). Immunofluorescence analysis of spread nuclei revealed the absence of RPA1 foci in mutant spermatocytes at 4 days Functions of RPA in meiotic recombination post-tamoxifen treatment (dpt), showing efficient depletion of Rpa1 in zygotene-like spermatocytes under this regimen ( Fig 2C). Interestingly, although RPA2 and RPA3 formed foci in wild type spermatocytes, they failed to form foci on meiotic chromosomes in zygotene-like Rpa1 mutant spermatocytes ( Fig 2C). These results were confirmed by the dramatic reduction in the abundance of RPA1, RPA2, and RPA3 proteins in tamoxifen-treated testes ( Fig 2D). These data demonstrate that the stability of both RPA2 and RPA3 depends on RPA1 and that loss of RPA1 causes the depletion of the RPA complex.

RPA1 is essential for meiotic progression
Inactivation of Rpa1 results in severe defects in meiotic progression in males (Fig 3A). At 4 dpt, Rpa1 cKO seminiferous tubules (Stage IX) were partially devoid of leptotene spermatocytes (Fig 3A). At 6 dpt, both leptotene (stage IX) and zygotene (stage XII) spermatocytes were lost in Rpa1 cKO testes. At 8 dpt, in addition to the loss of leptotene and zygotene spermatocytes, there was a partial loss of pachytene spermatocytes (Fig 3A and 3C). At 12 dpt, pachytene spermatocytes exhibited a significant loss (Fig 3A). Progressive loss of spermatocytes was expected to contribute partially to the reduction in protein abundance of RPA and other proteins in the mutant testes ( Fig 2D).
We monitored meiotic progression and determined the spermatocyte composition in Rpa1 cKO males by nuclear spread analysis (Fig 3B and 3C). At 4 dpt, Rpa1 cKO testes lacked normal zygotene spermatocytes, characterized by partial chromosomal synapsis, but contained defective zygotene-like spermatocytes, which were characterized by lack of synapsis and presence of unusually strong γH2AX (Fig 3B). The γH2AX intensity in both Rpa1-deficient leptotene and zygotene-like spermatocytes was significantly higher than wild type, suggesting that DSBs were generated but not repaired in the absence of RPA1 (Fig 3B). Zygotene-like spermatocytes were only present in Rpa1 cKO testes at 2 to 6 days post tamoxifen treatment with a peak at 2 dpt and 4 dpt (Fig 3C and 3D). Consistent with histological analysis (Fig 3A), only pachytene and diplotene spermatocytes were present in Rpa1 mutant testes at 8 dpt (Fig 3C  and 3D). Apoptosis in spermatocytes increased dramatically in stage IV tubules from Rpa1 cKO testes (S3 Fig), suggesting that Rpa1-deficient zygotene-like spermatocytes were eliminated by the pachytene checkpoint in response to early meiotic defects ( Fig 3D).

RPA1 is required for pre-meiotic S phase DNA replication
The synchronized depletion of meiotic germ cells in mutant testes suggests that RPA1 is required for pre-meiotic S phase DNA replication in preleptotene spermatocytes. To test this, we performed a BrdU incorporation (pulse-labelling) assay after tamoxifen treatment (0 dpt) ( Fig 4A). Preleptotene spermatocytes were only present in stage VII-VIII tubules and were nearly all BrdU-positive in~80% of such tubules in wild type testes, whereas preleptotene spermatocytes in the remaining 20% of wild type stage VII-VIII tubules were all BrdU-negative ( Fig 4B). Partial positive tubules with a mixture of BrdU-positive and BrdU-negative preleptotene spermatocytes were not observed in wild type testes but accounted for 50% of tubules in Rpa1 cKO testes ( Fig 4A and 4C), supporting the requirement for RPA1 in pre-meiotic S phase DNA replication in BrdU-negative preleptotene spermatocytes. While not observed in wild type testes, nearly 10% of stage VII-VIII Rpa1 cKO tubules lacked preleptotene spermatocytes, suggesting that RPA1 is also essential for mitotic DNA replication in spermatogonia, direct precursors of preleptotene spermatocytes.

Loading of recombinases RAD51 and DMC1 to DSBs is dependent on RPA
During meiotic recombination, DNA recombinases RAD51 and DMC1 form nuclear complexes on meiotic chromatin [10,11] and generate presynaptic filaments on ssDNA that drive strand invasion into homologous DNA duplex [23,24]. While RAD51 and DMC1 formed foci on meiotic chromosomes in wild type mice (Fig 5A), such foci were surprisingly absent in Rpa1-deficient zygotene-like spermatocytes (Fig 5A and 5B), although both proteins were  present at reduced abundance in the mutant testes as shown by Western blotting (Fig 2D). This result raises the possibility that RPA binding to ssDNA precedes and is a pre-requisite for loading of RAD51 and DMC1 recombinases (Fig 5C). In combination with our colocalization data (Fig 1), our results suggest that the interaction between RPA and RAD51/DMC1 is transient and that RPA is replaced by RAD51/DMC1 in the foci that form on DSBs. As such, loss of RPA blocks the critical strand invasion step in meiotic recombination and leads to a complete failure in chromosomal synapsis as observed in zygotene-like spermatocytes (Fig 5C).
ATR promotes loading of RAD51 and DMC1 to DSBs during meiotic recombination [25,26]. In wild type spermatocytes, ATR formed foci on the synaptonemal complex at the zygotene stage and localized to the XY body at the pachytene stage (S4 Fig), consistent with the previous reports [26][27][28]. In pachytene spermatocytes from Rpa1 cKO testes (4 dpt), ATR still localized to the XY body (S4B Fig). In addition, ATR still formed foci in zygotene-like Rpa1 mutant spermatocytes (S4A Fig). These results show that ATR is not sufficient for loading of RAD51 and DMC1 to DSBs in the absence of RPA.

Localization of MEIOB/SPATA22 to DSBs is independent of RPA
We next assessed if MEIOB contributes to RPA-dependent regulation of meiotic recombination. In wild type mice, the MEIOB/SPATA22 dimer always colocalized with RPA in foci on meiotic chromosomes (Fig 1H) [13]. In sharp contrast with RAD51/DMC1, MEIOB and SPATA22 still co-localized in Rpa1-deficient zygotene-like spermatocytes, suggesting that the formation and loading of MEIOB/SPATA22 dimers on DSBs are independent of RPA ( Fig 5). In addition, this result demonstrates that MEIOB/SPATA22 dimers are insufficient to recruit RAD51/DMC1 to DSBs (Fig 5C). Taken together with the previous finding on the presence of RPA foci in Meiob-deficient spermatocytes [13], the RPA trimer and the MEIOB/SPATA22 dimer localize to DSBs independently and function non-redundantly in meiotic recombination.

RPA promotes crossover formation
Following strand invasion, a subset of DSBs are repaired into crossovers at the pachytene stage. At the subsequent diplotene and metaphase stages, crossover sites become chiasmata to physically link bivalent homologs till chromosome segregation. Strand invasion into the homologue duplex creates the displacement loop (D-loop), which is single stranded. RPA formed foci in early-to-mid pachynema, presumably binding to the D loop and the single stranded second end (Fig 1). We first examined the RPA depletion efficiency in pachytene spermatocytes at 2 dpt and 6 dpt by nuclear spread analysis. At 2 dpt, the number of RPA2 foci in early-mid pachytene spermatocytes was comparable between control and Rpa1 cKO testes (S5 Fig), even though the RPA2 protein abundance was reduced in Rpa1 cKO testes (Fig 2D). However, at 6 dpt, in contrast with control early-mid pachytene spermatocytes, foci of RPA1, RPA2, or RPA3 were dramatically reduced in number and intensity in the early-mid pachytene mutant spermatocytes (S6 Fig). Therefore, to investigate whether RPA plays a role in crossover formation, we analysed pachytene spermatocytes from Rpa1 cKO testes at 6 or 8 days post-tamoxifen treatment.
We examined proteins involved in meiotic recombination that form foci at the pachytene stage. MEIOB and SPATA22 foci were still present, but RAD51 and DMC1 foci were absent in Rpa1-deficient early-mid pachytene spermatocytes (S7 Fig). This localization result was similar to that in zygotene-like Rpa1-deficient spermatocytes (Fig 5), confirming the unique role of RPA in RAD51/DMC1 localization throughout meiosis.
We next analysed two meiosis-specific factors, TEX11 and MSH4 [29][30][31], which modulate crossover formation. These two factors formed foci in control early-to-mid pachytene spermatocytes (Fig 6). However, the number of TEX11 foci was significantly reduced in earlymid pachytene spermatocytes from Rpa1 cKO testes at 6 dpt ( Fig 6A). Likewise, the number of MSH4 foci decreased significantly (Fig 6B). The reduction in the number of TEX11 and MSH4 foci in Rpa1-deficient pachynema (Fig 6) suggests a role of RPA in crossover formation.
To directly assess the effect of RPA on crossovers, we immunostained for MLH1, which localizes specifically to future crossover sites. MLH1 foci were present but significantly reduced in number in Rpa1-deficient pachynema compared with control at both 6 dpt and 8 dpt ( Fig  7A and 7B). Analysis of metaphase nuclei revealed that 20% of metaphase I spermatocytes had univalent chromosomes in Rpa1 cKO testes (Fig 7C and 7D). Univalents were observed for both autosomes and sex chromosomes. The presence of univalent chromosomes leads to chromosome mis-segregation and triggers the spindle checkpoint ( Fig 3D). Indeed, we detected Spermatocytes from untreated adult males were used as controls. The number of TEX11 foci (A) or MSH4 foci (B) in mid-pachytene spermatocytes is lower than in early pachytene spermatocytes. Early and mid-pachytene spermatocytes were categorized by the morphology of XY chromosomal axis and the intensity of the synaptonemal complex staining as previously described [41]. Early pachytene stage is characterized by an extended XY configuration and low intensity of the synaptonemal complex staining, whereas the mid-pachytene spermatocytes exhibit a U-shaped XY axis and higher synaptonemal complex staining. n, number of spermatocytes counted from four experiments (A) or three experiments (B); �� , p< 0.01; ��� , p<0.001. https://doi.org/10.1371/journal.pgen.1007952.g006 Functions of RPA in meiotic recombination massive apoptosis of spermatocytes in stage XII seminiferous tubules from Rpa1 cKO testes, which contained metaphase and anaphase spermatocytes (Figs 7E, 7F and 3A-12 dpt). These results demonstrate that RPA promotes crossover formation and thus is required for proper chromosome segregation during male meiosis (Fig 7G).

Discussion
RPA is often included in in vitro recombination reactions but its specific requirement is unknown. Recombinases RAD51 and DMC1 form helical nucleoprotein filaments on ssDNA [7,32] and RPA is necessary for efficient RAD51 filament formation [32]. The DNA strand exchange activities catalysed by RAD51 or DMC1 nucleoprotein filaments are rather inefficient but strongly stimulated by RPA [6,7,32]. These biochemical studies support RPA as an important accessary factor in recombination. While studies of yeast hypomorphic Rpa1 mutants also support a role for RPA in meiotic recombination [33], these studies did not address whether RPA is required for pre-synaptic filament formation. Using loss of function Rpa1 mouse mutants, we demonstrate that RPA loading to meiotic DSBs is a pre-requisite for RAD51/DMC1 loading in vivo (Fig 5C). One possible explanation is that RPA loading prevents formation of ssDNA secondary structure and recruits RAD51/DMC1 to these sites of recombination. An alternative but less likely explanation is that RAD51/DMC1 nucleation on ssDNA is independent of RPA, but that their localization to DSBs is stabilized by RPA. Finally, the fact that MEIOB/SPATA22 still localizes as foci in Rpa1-null spermatocytes suggests that the requirement of RPA in RAD51/DMC1 loading is unique and cannot be compensated for by other ssDNA-binding complexes such as MEIOB/SPATA22. MEIOB, SPATA22, and RPA colocalize in foci on meiotic chromosomes [13,34]. In Meiob or Spata22-deficient germ cells, RPA foci are still present [13,14,34,35]. Here we find that MEIOB and SPATA22 still form foci in RPA1-deficient spermatocytes. Collectively, although the MEIOB/ SPATA22 dimer and the RPA trimer colocalize in foci and interact with each other [15], their localizations to DSBs are independent. A previous study suggests that SPATA22 is required for the maintenance but not formation of RAD51 foci in rat spermatocytes [35]. RAD51 and DMC1 foci are present in Meiob-deficient spermatocytes [13]. Therefore, the presence of MEIOB/ SPATA22 foci and the absence of RAD51/DMC1 foci in RPA-deficient spermatocytes support that RPA but not MEIOB/SPATA22 is required for formation of RAD51/DMC1 foci.
Our genetic study identifies an essential role for RPA in meiotic recombination-strand invasion at the zygotene stage (Fig 7G). Failure in strand invasion causes persistence of DSBs and DNA damage in early Rpa1-deficient zygotene-like spermatocytes, which are eliminated by the first round of apoptosis in stage IV seminiferous tubules due to the pachytene checkpoint (Fig 3D, Fig 7G and S3 Fig). Therefore, Rpa1-deficient zygotene-like spermatocytes would not be expected to progress through the pachytene stage.
Our study uncovers a second role for RPA in meiotic recombination-crossover formation at the pachytene stage ( Fig 7G). We postulated that the pachytene spermatocytes from Rpa1 cKO testes were already at the pachytene stage and lost RPA1 upon tamoxifen injection, and thus were not likely to be derived from the defective zygotene-like spermatocytes. Inactivation of RPA1 in pachytene spermatocytes reduces crossover formation but does not completely abrogate it. This partial effect was likely due to the conditional deletion approach. RPA was dramatically depleted in Rpa1 cKO pachytene spermatocytes at 6 dpt, but was still detectable as extremely weak foci at a greatly reduced number (S6 Fig). A second possibility is that the existing RPA1 protein may have played a role in crossover formation before being degraded. Lastly, MEIOB/SPATA22 may exert a more important role in crossover formation than does RPA, since MEIOB/SPATA22 still forms foci in Rpa1-deficient pachytene spermatocytes. We hypothesize that, in the absence of RPA, MEIOB/SPATA22 binds to the recombination intermediates to promote the formation of most but not all crossovers (Fig 7G). Regardless of these scenarios, reduction in crossover formation in Rpa1-deficient pachytene spermatocytes leads to chromosome mis-segregation at the anaphase I stage spermatocytes, which are depleted by the second round of apoptosis due to the spindle checkpoint activation (Figs 3D and 7G).
Previously, a knockdown approach was employed to investigate the function of an essential gene Rad51 in mouse spermatogenesis [36]. In this study, we have used a powerful inducible germline-specific inactivation strategy to elucidate the function of RPA, an essential gene involved in DNA metabolism. This approach can be used to study the functions of any somatically essential genes in meiosis. Our results demonstrate that RPA plays a dual function in meiotic recombination: an essential role in presynaptic filament formation and an important role in crossover formation.

Ethics statement
Mice were maintained and used for experimentation according to the protocol approved by the Institutional Care and Use Committee of the University of Pennsylvania.

Antibody production
The mouse RPA2 full-length and RPA3 full-length were expressed as 6xHis-RPA2 and 6xHis-RPA3 fusion proteins in E. coli using the pQE-30 vector and affinity purified with Ni-NTA agarose. Two rabbits were immunized with each fusion protein (Cocalico Biologicals Inc.). Two guinea pigs were also immunized with the RPA2 recombinant protein. The resulting antisera UP2436 (rabbit anti-RPA2), GP111 (guinea pig anti-RPA2), and UP2439 (anti-RPA3) were used for immunofluorescence and western blotting analyses.

Histological analysis and TUNEL assays
For histological analysis, testes were fixed in Bouin's solution, embedded with paraffin, and sectioned. Sections were stained with hematoxylin and eosin. For TUNEL analysis, testes were fixed in 4% paraformaldehyde overnight at 4˚C, dehydrated in 30% sucrose and sectioned. Sections were performed with the TUNEL Enzyme and Label Kit (Roche Boehringer Mannheim).

Surface nuclear spread analyses
Nuclear spread analysis of spermatocytes was performed as previously described [40]. Primary antibodies that were used for immunofluorescence were listed in S1 Table. For quantification of foci, images of spermatocytes from two to four animals were captured and analysed. Axial element markers SYCP2 or SYCP3 were used to classify the stage of meiotic prophase on nuclear spreads [43,44]. Early and mid-pachytene spermatocytes were distinguished by the morphology of XY chromosomal axis and the intensity of synaptonemal complex [41]. Characteristics of early pachynema: relatively low intensity of synaptonemal complex staining, short synapsed pseudoautosomal regions, and/or unsynapsed ends of a few autosomes. Characteristics of mid pachynema: strong intensity of synaptonemal complex staining, full synapsis of all autosomal pairs, and often U-shaped XY axis. Characteristics of late pachynema: accumulation of SYCP2/3 proteins at the synaptonemal complex ends and figure -8 shaped XY chromosomes. The abnormal zygotene-like spermatocytes from Rpa1 mutant testes were characterized by extremely strong γH2AX signals and clusters of ends of axial elements. Metaphase spread cells were stained with 4% Gurr Giemsa.

Imaging
Color histological images were captured on a Leica DM5500B microscope with a DFC450 digital color camera (Leica Microsystems). Immunolabeled chromosome spreads and testis TUNEL assay images were captured with an ORCA Flash4.0 digital monochrome camera (Hamamatsu Photonics) on a Leica DM5500B microscope (Leica Microsystems) and processed using Photoshop (Adobe) software packages. Super-resolution imaging microscopy analysis was performed using a Nikon NSIM super-resolution microscope system and NIS-Elements 2 image processing software.

BrdU incorporation assay
Adult control (wild type) and Rpa1 cKO male mice were intraperitoneally injected with 50 mg/ kg body weight of BrdU (Sigma, B5002). Adult Rpa1 fl/fl Ddx4-Cre ERT2/+ male mice were treated with daily injection of tamoxifen for five consecutive days as described above. 12 hours after the last tamoxifen injection, a single dose of BrdU was injected. After 2 h, the mice were sacrificed. Testes were collected and fixed in the fixative solution (30% formaldehyde, 15% ethyl alcohol, 5% glacial acetic acid) overnight, embedded in paraffin, and cut in 5 μm sections. After deparaffinization and rehydration, slides were immersed in 20 mM Tris-HCl, pH9.0, at 95˚C for 15 min. The slides were blocked with 1% BSA for 1h at room temperature followed by incubating with anti-BrdU and anti-SP10 antibodies at 37˚C overnight. The morphology of acrosome revealed by anti-SP10 (ACRV1) antibody was used to precisely identify stage VII-VIII tubules, in which preleptotene spermatocytes are present [42]. Slides were rinsed with PBS and then incubated with rabbit anti-rat (Vector, FI-4001) and goat anti-guinea pig secondary antibodies (Novus, NB-1206906) at 37˚C for 1h. Mounting medium with DAPI was added to the slides for imaging. Three pairs of control and Rpa1 cKO males (0 day post tamoxifen treatment) were analyzed. At least thirty five stage VII-VIII tubules were counted for each mouse (Fig 4B and 4C).

Statistics
Statistical analysis was performed with Student's t-test.