Low retinoic acid levels mediate regionalization of the Sertoli valve in the terminal segment of mouse seminiferous tubules

In mammalian testes, undifferentiated spermatogonia (Aundiff) undergo differentiation in response to retinoic acid (RA), while their progenitor states are partially maintained by fibroblast growth factors (FGFs). Sertoli valve (SV) is a region located at the terminal end of seminiferous tubule (ST) adjacent to the rete testis (RT), where the high density of Aundiff is constitutively maintained with the absence of active spermatogenesis. However, the molecular and cellular characteristics of SV epithelia still remain unclear. In this study, we first identified the region-specific AKT phosphorylation in the SV Sertoli cells and demonstrated non-cell autonomous specialization of Sertoli cells in the SV region by performing a Sertoli cell ablation/replacement experiment. The expression of Fgf9 was detected in the RT epithelia, while the exogenous administration of FGF9 caused ectopic AKT phosphorylation in the Sertoli cells of convoluted ST. Furthermore, we revealed the SV region-specific expression of Cyp26a1, which encodes an RA-degrading enzyme, and demonstrated that the increased RA levels in the SV region disrupt its pool of Aundiff by inducing their differentiation. Taken together, RT-derived FGFs and low levels of RA signaling contribute to the non-cell-autonomous regionalization of the SV epithelia and its local maintenance of Aundiff in the SV region.

www.nature.com/scientificreports/ terminal segment of ST may regulate the directional flow of the luminal fluid and sperm 24,25 , yet the exact function of the SV in spermatogenesis remains unclear. The SV epithelia lack most spermatogenic activity, mainly composed of GFRα1-positive stem/progenitor spermatogonia and proliferative Sertoli cells, which are biased to the proximal part of the SV region (SV niche) [26][27][28] . It is likely that certain unique signaling states within the SV epithelia (e.g., high GDNF expression) contribute to maintaining the GFRα1-positive A undiff in the SV region 26,29 . However, the regionalization of the SV epithelia and their molecular/cellular characteristics are unclear.
In this study, we examined the postnatal development of the SV region and demonstrated the non-cell autonomous regionalization of the SV is due to active degradation of RA and, potentially, is modulated by FGF signaling from the adjacent RT region. We also investigated the molecular characteristics of SV epithelia by global mRNA analyses, comparing gene expression profiles of isolated SV fragments with those of ST and RT fragments.

Results
Constitutively activated AKT signals in the SV epithelia. The SV epithelia in mouse testis comprise unique Sertoli cells with distinct characteristics from those in the convoluted seminiferous epithelia, represented by high-level expression of acetylated tubulin (ace-TUB) [26][27][28][29] . However, the mechanism of its regional specificity and homeostasis still remain to be discovered. In order to understand the specification of Sertoli cells within the SV region, we first screened the potential signaling pathways activated specifically in the SV epithelia and revealed constitutively high expression of phosphorylated AKT (p-AKT) in the Sertoli cells located within the SV region ( Fig. 1a-f). In the convoluted seminiferous tubule (ST), p-AKT immunoreactivity was observed in the cytoplasm of Sertoli cells in a seminiferous epithelial cycle-dependent manner. p-AKT expression was high at seminiferous epithelial stages II-VI (Fig. 1b), whereas low at the seminiferous epithelial stages IX-XII (Fig. 1b), where the expression levels of phosphorylated ERK (p-ERK) are known to be high 30 . These data suggest the distinct regulation of AKT or ERK signaling pathway in the convoluted seminiferous epithelia with seminiferous epithelial stage dependency. Interestingly, Sertoli cells within the ace-TUB-positive SV region showed constitutively high expression levels of p-AKT expression, in comparison to the fluctuating expression pattern of p-AKT in the convolute seminiferous epithelia (Fig. 1d,e). In the developing postnatal mouse testes, the terminal segment of seminiferous tubules was partially positive for anti-p-AKT and ace-TUB staining at postnatal day (P) 7 (i.e., prior to the formation of the SV structure and the initiation of luminal fluid flow) (Fig. 1f). These signals became clearer in the Sertoli cells at P14, leading to the establishment of the SV structures with high expression of p-AKT by P28 (Fig. 1f). In the germ cell-depleted W/W v mutant mice, p-AKT expression remained high in the SV region compared to the convoluted ST ("W/W v " in Fig. 1d,e), suggesting no appreciable contribution of spermatogenic cells to the high p-AKT expression in the SV epithelia (Fig. 1e). In contrast, p-ERK immunoreactivity was not restricted to the SV nor proximal ST regions, neither in wild-type and W/W v testes (Fig. S1a). Signals of anti-phosphorylated p38 nor anti-phosphorylated JNK staining were not restricted to the SV region either (Fig. S1b). We further showed the ubiquitous expression of AKT1/2 in Sertoli cells of both ST and SV regions (Fig. S1c). The p-AKT signals were lost by the pretreatment with alkaline phosphatase prior to immunostaining, confirming the specificity of the anti-p-AKT antibody to the phosphorylated forms of AKT (Fig. S1d). Based on these findings, we conclude that p-AKT signals can be used as a marker of Sertoli cells in the SV region.

Regeneration of the SV epithelia with AKT phosphorylation by the transplanted Sertoli cells.
To examine whether the immature Sertoli cells adjacent to RT are pre-determined to form the SV epithelia or not, we conducted a Sertoli cell transplantation assay using the diphtheria toxin (DT)-treated AMH-Treck Tg male mice with Sertoli cell-ablated STs 31 . In brief, we collected Sertoli cell from the distal part of the convoluted STs (i.e., the region without RT and proximal ST) of CAG-EGFP (cytoplasmic/nuclear EGFP) or R26-H2B-mCherry (nuclear mCherry) mouse testes, and then injected the Sertoli cell suspension intratubularly into the DT-pretreated AMH-Treck host males as reported previously (Fig. 2a) 30 . At day 10 post-Sertoli cell transplantation, the donor-derived (GFP-positive) SOX9-positive Sertoli cells had settled in the presumptive SV regions adjacent to the host-derived RT in the recipient AMH-Treck Tg males (Fig. 2b), indicating the replacement of SV Sertoli cells with the transplanted donor-derived Sertoli cells. At day 45 post-transplantation, spermatogenesis was recovered in some of the proximal STs ( Fig. 2c-f), in which the donor-derived (mCherry/SOX9-double positive) Sertoli cells had settled, including the presumptive SV region connecting to RT (Fig. 2d,e). Interestingly, the presumptive SV regions colonized by mCherry/SOX9-double positive transplanted Sertoli cells were also positive for anti-ace-TUB/p-AKT staining, reconstituting the valve-like structures enriched with GDNF expression (Fig. 2c,d,f). These findings imply that the SV structure was non-cell autonomously constructed by the transplanted Sertoli cells located adjacent to the RT, suggesting the possibility that the factors derived from the RT and its surrounding tissues contribute to the regionalization of mouse SV epithelia.
Expression profile of FGF-related genes and potential contribution of FGF signaling to the AKT phosphorylation in Sertoli cells. Having identified the SV-specific AKT phosphorylation, we then asked which signaling molecule contributes to this local AKT activation. Based on our previous study showing the SV-specific HSPG (Heparan Sulfate Proteoglycan) enrichment in hamsters 26 , we hypothesized that HSPG, which functions as a reservoir and co-receptor of FGF ligands 32 , modulates the SV-specific AKT activation through its local storage and reception of ligands produced by the RT. To test this hypothesis, we focused on FGFs as a potential upstream of AKT activation at the SV. We first confirmed the enrichment of HSPG in the basement membrane of the SV region in adult wild-type mouse testis (Fig. 3a), which is consistent with our previous work in hamster. HSPG was observed predominantly in the SV region, while the RT region marked by ECAD (E-cadherin, also known as Cadherin-1) immunoreactivity only showed little expression of HSPG (Fig. 3a'). www.nature.com/scientificreports/   (Fig. 3b). The expression levels of Fgf9 were significantly higher in the RT also in the W/W v mutant mice, suggesting that its region-specific difference is not due to the existence of germ cells (Fig. 3b). Next, we examined the mRNA expression levels of two major FGF receptor genes in Sertoli cells, Fgfr1 and Fgfr2 33 . As a result, Fgfr2 was expressed significantly higher in the SV region compared to the ST region both in W/W v and wild-type mice (Fig. 3b). Fgfr1 expression in the SV was significantly higher in the W/W v mice, but not in wild-type mice potentially due to the contamination of germ cells (Fig. 3b). We further performed in situ hybridization in the wild-type mice to confirm the localization of mRNA in each gene. As a result, the expression of Fgf9 was detected in the RT region, while hardly any Fgf9 signals were observed in SV and ST regions (Fig. 3c). On the other hand, both Fgfr1 and Fgfr2 mRNA were detected both in the SV and RT www.nature.com/scientificreports/ www.nature.com/scientificreports/ regions (Fig. 3c). Such co-expression of HSPG and FGFRs in the SV region suggests the potential binding and activation of FGFs within the SV region.
To investigate the correlation between FGF signaling and AKT phosphorylation at the SV region, we next examined the effect of exogenous FGF9 on the p-AKT signals in convoluted seminiferous tubules by using an in vivo bead transplantation assay (n = 5) 13,34 . In this experiment, W/W v mutant mice were used to exclude the effect of FGF9 on germ cells. Even without the germ cells, p-AKT signals were observed sporadically in the convoluted seminiferous tubules of the W/W v mutant mice (Fig. S2a). However, the signal intensity of such p-AKT was relatively weak compared to that in the SV region (Fig. S2b), and not all the Sertoli cells within the cross-sectioned seminiferous tubules expressed p-AKT in the W/W v mutant mice. In contrast, p-AKT signals were greatly increased in the convoluted ST adjacent to the FGF9-soaked beads, but not in the ST far away from the FGF9-soaked beads nor the ST adjacent to the BSA-soaked beads (Fig. 3d). Therefore, these findings suggest the possible contribution of FGF-FGFR signaling to the SV-specific expression of p-AKT.
Identification of Cyp26a1 as an SV-specifically upregulated gene. To further understand the regulatory mechanism of the SV region on a molecular basis, we next performed the global mRNA analyses, aiming to identify the genes specifically expressed in the SV region. The RT, SV, and convoluted ST fragments were manually separated from adult W/W v mice upon intratubular trypan blue injection (Fig. 4a), and then subjected to the microarray analyses (20 testes for each set of microarrays; 3 sets for RT, 4 sets for SV and ST).
The transcriptomic analysis resulted in the identification of 623 and 2,304 upregulated probes in the SV and RT respectively, compared to the convoluted ST (> twofold change with p-value < 0.05). Since 512 probes were shared by the 2 groups, 111 probes were identified as SV-specific probes, which are exclusively expressed in the SV region (Fig. 4b). Gene ontology (GO) analysis of the 111 SV-specific probes suggested the terms "retinoic acid metabolic process" (e.g., Cyp26a1, Rbp1; p = 7.4E -02 ), "positive regulation of protein phosphorylation" (e.g., Il34, Camp, Gpnmb, Ptpn5; p = 6.9 -02 ), "detection of mechanical stimulus involved in sensory perception" (e.g., Asic2, Serpine2; p = 2.5E -02 ) and "positive regulation of cell proliferation" (e.g., Il34, Sfrp1, Ctsh; p = 2.5E -02 ). On the other hand, GO analysis of the 1792 RT-specific genes identified by microarray analyses yielded the terms "positive regulation of epithelial cell proliferation" (e.g., Fgf1, Fgf9, Bmp4, Notch1, Igf1, Vegfa; p = 7.8E -06 ), "Wnt signaling pathway" (e.g., Wnt7b, Wnt9a; p = 1.5E -02 ) and "Wound healing" (e.g., Tgfa, Cx3cl1; p = 4.4E -02 ). These results suggest the secretion of various signaling factors in the RT region, which may affect the terminal end of the ST to obtain its unique phenotypes. Differentially expressed gene (DEG) analysis resulted in the identification of 862 DEGs (p < 0.01, Fig. 4c). 107 DEGs were upregulated both in the SV and RT regions (orange dots in Fig. 4d), while 24 DEGs were exclusively upregulated in the SV region (red dots in Fig. 4d). Among such SV-specific DEGs, we identified Cyp26a1 as the most significantly upregulated gene with top fold change (Fig. 4d). Since Cyp26a1 encodes a cytochrome P450 enzyme which controls the metabolic inactivation of RA 35 , SV-specific expression of Cyp26a1 implies unique RA metabolization within the SV region. Being consistent with the microarray data, in situ hybridization confirmed the region-specific expression of Cyp26a1 in the Sertoli cells located within the SV region of W/W v mutant mice (Fig. 4e). The expression of Cyp26a1 was also observed in the SV region of wild-type mice (Fig. 4f). These data indicate constitutively high Cyp26a1 expression in the Sertoli cells within the SV region irrespective of the presence of spermatogenic cells, and in turn suggest the constitutively active degradation of RA in the SV region.
Exogenous RA treatment induces ectopic c-KIT-positive spermatogonial patches and subsequent disruption of the SV structure. Finally, we examined the effect of exogenous RA on the structure and function of the SV region in wild-type mouse testes ( Fig. 5a-f). In this experiment, we transplanted RAsoaked microbeads locally around the SV region to temporarily increase the local RA levels (Fig. 5a). Being consistent with our previous report 26 , control (DMSO-treated) mouse testes had hardly any c-KIT-positive A diff within the proximal part of the SV region (0 ~ 50 μm from the edge of rete testis), where GFRα1-positive/c-KITnegative A undiff were constitutively maintained (Fig. 5b). On the other hand, RA-treated mouse testes showed a reduced number of GFRα1-positive A undiff within the proximal SV region (Fig. 5b,c; "GFRα1 + Total" in Fig. 5e), while c-KIT-positive A diff were enriched within the SV region at day 1 after the RA treatment (Fig. 5b,d; "c-KIT + " in Fig. 5e). Since RARγ-positive subpopulation was previously shown to maintain the differentiation competence of GFRα1-positive spermatogonia 15,36 , we then quantified the RARγ-negative/positive subpopulation of GFRα1-positive cells in the proximal SV region treated with RA and DMSO respectively. As a result, we found a drastic reduction of RARγ-positive subpopulation (13.0% of the control value) together with a mild reduction of a RARγ-negative subpopulation (51.4% of the control value) in the RA-treated group (n = 5, two right graphs in Fig. 5e). In contrast, RA treatment did not seem to affect the expression of GDNF and CyclinD1 (CCND1) in the Sertoli cells located within the SV region (Fig. S3), which are other characteristics of the SV region 26 . These findings suggest that the c-KIT-positive spermatogonial patches observed at the proximal SV region in the RA-treated mice may have resulted directly from the A undiff -A diff transition of pre-existing GFRα1/ RARγ-double positive spermatogonia by the exogenous exposure to excess RA. Interestingly, in some severely affected SVs at day 3 after RA treatment, the ace-TUB-positive valve-like structures were disrupted (Fig. 5f). These data suggest that low RA levels in the SV region support the maintenance of the local A undiff , as well as contributing to the physical valve-like structure.

Discussion
In this study, we demonstrated the non-cell-autonomous regeneration of the SV epithelia by taking advantage of the AMH-Treck Tg system. Although the transplanted Sertoli cells were derived from the convoluted ST, donorderived Sertoli cells colonized in the presumptive SV region (i.e., within 200-300 µm from the RT) regenerated www.nature.com/scientificreports/ www.nature.com/scientificreports/ not only the valve-like structure, but also expressed several SV markers such as p-AKT and ace-TUB, similar to the SV epithelia in the normal intact testes (Fig. 2f). Consistent with the previous observations in wild-type mouse testis 26 , spermatogenesis was repressed in such regenerated SV epithelia (Fig. 2c-f), while patches of active spermatogenesis were frequently observed in the convoluted STs ("ST" in Fig. 2c-f). Although the function of the regenerated SV region as a stem/progenitor cell niche remains unclear, our data suggest that the SV region is non-cell autonomously specified by the Sertoli cells adjacent to the RT. In conjunction with the data showing the upregulation of p-AKT and ace-TUB in the Sertoli cells adjacent to the RT in P7 testes, prior to the formation www.nature.com/scientificreports/ of the valve-like structure (Fig. 1f), we speculate that the regionalization of the SV epithelia is induced and/or maintained, together with the activation of the AKT signaling pathway and tubulin acetylation, presumably by local soluble factors derived from the RT and its surrounding tissues at the terminal segment of the ST. In this study, we revealed the constitutively high expression of p-AKT in the SV epithelia, together with seminiferous epithelial cycle stage-dependent expression of p-AKT in the convoluted ST. This is reminiscent of the expression pattern of phosphorylated signal transducer and activator of transcription-3 (STAT-3), which shows constitutively high expression in the SV epithelia with some local expression in the convoluted ST at seminiferous epithelial cycle stages I-VI 29 . The onset of p-STAT3 activation in the SV Sertoli cells takes place in P7 testis 29 , coinciding with the onset of p-AKT expression (Fig. 1f), suggesting that the regionalization of the SV region takes place around P7 in the postnatal testicular development.
In the fetal mouse testes, FGF9 was previously shown to induce the proliferation of Sertoli cells through FGFR2 [37][38][39] . FGF9, as well as FGF10, is upregulated in a testis-specific manner by 11.5 dpc and contributes to synchronous testiculogenesis of the pre-Sertoli cells in the fetal testes 40 . Concomitantly with the onset of Fgf9 expression, testis-specific AKT activation also takes place in presumptive pre-Sertoli cells at 11.5 dpc 41 , suggesting a potential association between FGF9 signals and p-AKT activation in the developing fetal testes. The present study demonstrated the increased p-AKT signals in the Sertoli cells of convoluted ST caused by exogenous FGF9 (Fig. 3d). Since FGF9 is known to activate the PI3K/AKT pathway as well as MEK/ERK pathway 42 , RT-derived FGF9 can be associated with the activation of AKT signals in the Sertoli cells within the SV region. Interestingly, the basement membrane at the SV region has the region-specific enrichment of HSPGs, which works as a reservoir for several FGF ligands 31 . Therefore, it might be possible that the RT-derived FGFs bind to the HSPGs at the SV region and subsequently cause signal transduction via local FGFRs. Since FGF9 is known to work over a short-range 32 , and the SV region is adjacent to the RT, we assume that Sertoli cells in the SV can perceive the RT-derived FGF9. Taken together, RT-derived FGF ligands and their local perception through the HSPG can be one of the causes for the activation of AKT within the SV region. Microarray analyses revealed the high expressions of various cytokine and growth factor genes in RT, including those encoding candidate ligands upstream of AKT/STAT3 signaling, such as FGF9, suggesting their potential contribution to non-cell autonomous specification and regionalization of the SV epithelia in vivo. Meanwhile, our microarray analysis indicated the expression of various cytokines and growth factor genes in the RT region aside from FGF9, which can also be candidate ligands upstream of AKT/STAT3 signaling. Therefore, a further study on such ligands is required to comprehensively understand the RT-derived stimuli, which play a role in the non-cell autonomous regionalization and the constitutive p-AKT/p-STAT3 activation of the SV epithelia in vivo.
It is traditionally known that FGFs and RA signaling acts in an opposing manner in the regionalization of various organs such as limb, neural pattern, segmentation, and somite formation [43][44][45] . In fetal and postnatal testes, FGF9 antagonizes RA-dependent meiotic differentiation of germ cells in part by maintaining Nanos2 expression 7,8,46 . In adult testes, FGFs including FGF2 and FGF5 maintain the self-renewal of stem/progenitor spermatogonia independent of GDNF signals 14,47 , while RA deficiency supports the A undiff to retain their stem/ progenitor state in vivo 1,4,17 . FGF9, as well as FGF2, maintains these stem/progenitor spermatogonia in vitro 45 . FGFs including FGF5 were shown to be produced by the lymphatic endothelial cells around the seminiferous tubule to support A undiff 14 . Considering that lymphatic vessels from convoluted STs join together around RT and SV regions 48 , the source of the FGF ligands in the RT can be its epithelial cells as well as its surrounding lymphatic tissue. Similarly, excess expression of a niche factor, GDNF, represses the differentiation of A undiff , leading to inactive spermatogenesis in vivo 11,34,49,50 . Since GDNF is also highly expressed in the SV region 26 , GDNF and FGFs may contribute to the structure of SV epithelia by suppressing the differentiation of local A undiff in rodent testes.
In mammalian spermatogenesis, RA is involved in (1) the A undiff -A diff transition of spermatogonia, (2) spermiation, and (3) regulation of the seminiferous epithelial cycle 18 . RA has been reported to be dispensable for the meiotic initiation of the germ cells [51][52][53] , whereas injection of exogenous RA promotes the differentiation of A undiff in vivo 54 . Similarly, in our study, local administration of exogenous RA promoted the ectopic appearance of c-KIT-positive A undiff in the SV region. Together with the SV-specific expression of Cyp26a1, it is reasonable to speculate that the levels of RA in the SV region are maintained low, preventing the A undiff -A diff transition of the local spermatogonia within the SV region. Considering that another RA metabolizing enzyme, Cyp26b1, expresses in the peritubular myoid cells to block the entry of interstitial RA into the seminiferous tubules 55,56 , Cyp26a1 in the SV region may also function as a local catabolic barrier insulating the A undiff from RT-derived RA. Since Cyp26a1-null mutant mouse testes were reported to have no appreciable defects in the convoluted ST 35 , the compensational roles of Cyp26a1 and Cyp26b1 in the SV region cannot be neglected, yet their detailed phenotypes of the SV remain unclear. Meiotic germ cells express Aldh1a1-3, which make the germ cell one of the major sources of RA in the testis 54 , and the lack of meiotic germ cells in the SV region can also be a cause for the low level RA within the SV region. Although RA is synthesized by both Sertoli cells and germ cells, germ cell-derived RA has been reported to play important roles in the A undiff -A diff transition of spermatogonia at the initial onset of spermatogenesis 54 . Taken together, the lack of meiotic germ cells within the SV region contributes to its local low RA signaling state, in combination with the local degradation of RA by Cyp26a1 expression. In our proposed model, the specification and regionalization of the SV epithelia are non-cell autonomously mediated by the factors derived from the RT, resulting in the AKT phosphorylation in the SV Sertoli cells (Fig. 6). Particularly, FGFs (e.g., FGF9) may contribute to the maintenance of A undiff in the SV region, through its regionspecific enrichment of HSPG in the basal lamina. Moreover, the high expression of Cyp26a1 in the SV region leads to the region-specific low-RA signaling states, which may promote the maintenance of GFRα1-positive A undiff and repress their differentiation into c-KIT-positive A diff . A lack of differentiated germ cells (i.e., c-KITpositive A diff and meiotic/post-meitoic germ cells) may also physically allow the formation of unique Sertoli cell morphology within the SV region, extending the cytoplasmic processes into the adluminal compartment to form the valve-like structure within the SV region. In this study, exogenous RA treatment induced the ectopic Considering the potential contribution of meiotic germ cell-derived RA for the initial A undiff -A diff progression of the stem/progenitor cell population 50 , the lack of advanced spermatogenic cells within the SV region may be also crucial for the physical and endocrinological maintenance of the SV epithelia.
Although the existence of the SV region has been recognized for half a century, the majority of its functionality and molecular mechanisms remain unknown. This study provided the first direct evidence of non-cell autonomous regionalization of mouse SV epithelia, as well as its unique gene expression profile. A further study on SV-specific genes may resolve the molecular mechanisms underlying the functional and biological significances of the SV region, as well as the mammalian stem/progenitor cell niche.

Methods
Animal care and use. C57BL/6-Tg AMH-Treck mice (AMH-Treck mice) 31 , C57BL/6-R26-H2B-mCherry knock-in mice (H2B-mCherry mice) 57 and C57BL/6-Tg (CAG-EGFP) mice (GFP mice; SLC Japan) were used for Sertoli cell transplantation experiments. Wild-type mice (C57BL/6 and ICR mice) and W/W v mice (SLC Japan) were used for in situ hybridization, immunohistochemistry, microarray, and quantitative reverse transcription-polymerase chain reaction (qRT-PCR) analyses. All animal experiments were performed in accordance with Guidelines for Animal Use and Experimentation at the University of Tokyo, and all the experimental procedures herein performed were approved by the Institutional Animal Care and Use Committee, Graduate School of Agricultural and Life Sciences, The University of Tokyo (approval IDs, P13-762 and P13-764).

Sertoli cell transplantation.
For Sertoli cell transplantation, AMH-Treck (3 ~ 4 weeks old) and GFP/ mCherry (1 week old) male mice were used as recipients and donors, respectively 31 . First, AMH-Treck male mice were treated with diphtheria toxin (DT; Sigma-Aldrich, 4 μg/kg i.p.) 4 days before transplantation. To obtain donor Sertoli cells, the convoluted STs were isolated from the distal two-thirds of a whole testis (i.e., the testes without RT, SV, and proximal ST) of GFP/mCherry mice. A single-cell suspension (1.0 × 10 7 cells/mL) including immature Sertoli cells was prepared by two-step enzymatic digestion as previously reported 31 and transplanted through the efferent ducts into the testes of DT-pretreated AMH-Treck recipients (Fig. 2a). On days 10 and 45 after transplantation, all recipient testes were analyzed for histology and immunohistochemistry. The proportion of donor-derived cells (relative number of mCherry-positive cells per total cells) were 96.9 ± 0.5% in Sertoli cells and 75.2 ± 15.5% in germ cells on 45 days after transplantation (n = 4). www.nature.com/scientificreports/ In vivo treatment of exogenous FGF9 and retinoic acid. For FGF9 treatment, FGF9 (R&D systems, 0.1 mg/ml)-soaked beads (Affi-gel blue, Bio-Rad, ~ 50 μm) were injected into the interstitial regions of W/W v mouse testes as previously reported 13,34 . The beads were labeled with DiI prior to the transplantation, which makes it distinguishable in the section. At 2 ~ 24 h after the bead treatment, testes were isolated for the subsequent histology and immunohistochemical analyses. For RA treatment, wild-type male mice (ICR, 7-8 weeks old) were injected with beads (BioMag Amine) soaked in 40.0 mg/ml all-trans RA (Sigma-Aldrich) in 16% (vol/vol) dimethyl sulfoxide (DMSO) or DMSO (Fig. 5a). In this experiment, the beads were injected locally around the SV region (Fig. 5a) to examine the effects of locally elevated RA levels on the SV function. At 1 and 3 days after the treatment, testes were isolated for subsequent histology and immunohistochemical analyses.
Immunohistochemistry and morphometry. The testes were fixed in 4% paraformaldehyde (PFA) and routinely embedded in paraffin or OCT compound. Cryosections (8 µm thickness) or deparaffinized sections (5 µm thickness) were subjected to immunohistochemical staining as described previously 58  In some stained sections, the seminiferous cycle stages were identified by acrosome staining using rhodaminelabeled soybean agglutinin (SBA; Vector Stain). To verify antibody specificity for phosphorylated AKT, two serial sections were pretreated with or without alkaline phosphatase from calf intestine (ALP; #47785055; Oriental Yeast) at 34 °C for 1 h before incubation with the primary antibody.
For whole-mount immunohistochemistry, SV and ST fragments were carefully isolated and fixed in 4% PFA. After methanol permeabilization, the fragments were incubated for 12 h at 4 °C with anti-c-KIT (1:100 dilution; R&D Systems), anti-GFRα1 (1:100 dilution; R&D Systems) or anti-RARγ (1:100 dilution; Cell Signaling) antibody. After washing with PBS, the samples were incubated for 2 h at room temperature with Alexa-488/594 conjugated secondary antibodies, including DAPI. Stained whole-mount samples were photographed, and the numbers of GFRα1-positive/RARγ-negative, GFRα1-positive/RARγ-positive and c-KIT-positive cell were quantified respectively in the proximal part of SVs (i.e., 0 ~ 50 μm apart from the edge of the rete testis) as previously shown 26 . In addition, in intact wild-type mice, a proportion of RARγ-positive cells out of total GFRα1-positive spermatogonia in the SV region was at 20.1 ± 7.2%, similar to 21.3 ± 7.8% in the convoluted ST region. All of the stained samples were analyzed using an Olympus fluorescence microscope (BX51N-34-FL-2) and a Leica TCS SP8 (Leica Microsystems GmbH) confocal laser microscope. In addition, any non-specific signals inside the STs were detected in sections incubated with normal IgG instead of the primary antibody (data not shown).

Isolation of the tubular fragments from the RT, SV and ST.
To visualize the RT, SV and proximal ST, trypan blue solution (0.4%; Sigma-Aldrich) was micro-injected through the efferent ducts into the testes of W/W v mutant mice or wild-type ICR mice (7 weeks old) (Fig. 4a). The tunica albuginea was then removed from the testes, and the RT, SV and distal convoluted ST fragments were isolated manually by using sharp forceps and 31G needles under a dissecting microscope. RT samples were collected from the main part just beneath the tunica albuginea, and SV samples were collected as tiny, tubular fragments (< 1 mm) adjacent to the RT region. ST samples were collected from the distal part of the testis far apart from the RT and SV regions. The border of the RT and SV regions were distinguishable by their distinct tubular diameters with different intensities of the trypan blue color. Of note, as a limitation of this study, it is highly likely that SV fragments contained a certain amount of the RT-derived cells due to the continuous tubular structure of the RT and SV regions. The samples were then subjected to the RNA analyses.
RNA extraction, microarray, and qRT-PCR analyses. Total  www.nature.com/scientificreports/ RT, SV and ST from wild-type mice (n = 4 animals each) or W/W v mice (n = 5 animals each) as described above, and then reverse-transcribed with random hexamers by using the Superscript-III cDNA synthesis kit (Invitrogen). Specific primers and fluorogenic probes for Fgf9 (Mm00442795_m1), Fgf10 (Mm00433275_m1), Fgfr1 (Mm00438930_m1), Fgfr2 (Mm01269930_m1) and Actb (Endogenous Control; 4352341E) were purchased from Applied Biosystems. PCR was performed using the Applied Biosystems Step One Real-Time PCR System. The relative levels of the transcripts were normalized to that of Actb as an endogenous reference.
In situ hybridization. In situ hybridization was conducted as described previously 59  Statistical analyses. The qRT-PCR data were analyzed by Paired t-tests using R software 60 . Morphometric data were analyzed by Student's t-test. A p-value < 0.05 was considered indicative of statistical significance. Quantitative data are presented as mean ± standard error of the mean (SEM). Throughout this study, "n" refers to the number of animals, except for the microarray analysis, in which 20 testes (10 animals) were used for each set of microarray analysis.