Fetal Leydig cells dedifferentiate and serve as adult Leydig stem cells

Testicular Leydig cells are functionally highly differentiated cells that produce androgens to induce masculinization. It is well accepted that two types of Leydig cells, fetal Leydig cells (FLCs) and adult Leydig cells (ALCs), develop in the eutherian testis at prenatal and postnatal stages, respectively (Roosen-Runge and Anderson, 1959). FLCs are essential for masculinization of fetal tissues such as the brain and external genitalia, whereas ALCs play important roles in developing male-specific secondary sex characteristics. Although both FLCs and ALCs produce androgens, they show different gene expression patterns. Specifically, FLCs do not express 17β-HSD, an enzyme that catalyzes final reaction of testosterone biosynthesis, and therefore they predominantly produce androstenedione, whereas ALCs express all the enzymes required for testosterone production (O'Shaughnessy et al., 2000; Shima et al., 2013). There has been considerable debate about the fate of FLCs in the postnatal testis. In rodents, it was generally believed that FLCs completely disappear and they are replaced by newly developed ALCs after birth (Griswold and Behringer, 2009). However, several researchers claimed that FLCs persist in the postnatal testis (Ariyaratne and Chamindrani Mendis-Handagama, 2000; Kerr and Knell, 1988; Mendis-Handagama et al., 1987), and recent lineage-tracing analyses of FLCs clearly showed that FLCs persist as a small subpopulation of Leydig cells in the adult testis (Kaftanovskaya et al., 2015; Shima et al., 2015). The origins of FLCs and ALCs have been also intensively investigated. FLC progenitor cells are thought to exist in the interstitial space of the fetal testis (Inoue et al., 2016; Stevant et al., 2018). Meanwhile, ALC progenitor cells were identified at the peritubular and perivascular regions of the postnatal testis (Davidoff et al., 2004; Ge et al., 2006). Based on these studies, it was assumed that both FLC progenitors and ALC progenitors reside within the fetal testis, and several researchers claimed that FLCs and ALCs share a common progenitor pool in the fetal testis (Barsoum et al., 2013). However, the lineage relationship between FLCs and ALCs has not been fully clarified, and how the two distinct Leydig cell populations sequentially develop remains an open question (Shima and Morohashi, 2017; Svingen and Koopman, 2013).

The orphan nuclear receptor NR5A1 (previously known as Ad4BP or SF-1) was identified as a factor essential for development of reproductive tissues, such as the ventromedial hypothalamus, pituitary gonadotrope, and gonad (Luo et al., 1994; Morohashi, 1997). We have clarified that Nr5a1 gene expression in each tissue is regulated by a tissue-specific enhancer, and most recently we identified that an upstream enhancer of Nr5a1 gene has the potential to induce gene expression in FLCs but not in ALCs. Therefore, we designated this region as the fetal Leydig enhancer (FLE) of the Nr5a1 gene (Shima et al., 2012). In this study, we performed lineage-tracing analyses of FLCs using the FLE, and revealed that FLCs dedifferentiate at fetal stages, and thereafter redifferentiate into ALCs at the pubertal stage. In addition, dedifferentiated cells also contribute to the peritubular myoid cells (PTMCs) and vascular pericytes (VPs). Although steroidogenic gene expression was completely absent, these cells retained the potential to redifferentiate into ALCs. FLC-lineage-specific Nr5a1 gene disruption resulted in almost complete loss of FLCs as well as a significant decline of ALCs, suggesting that most, if not all, ALCs arise from dedifferentiated FLCs. We also generated a mouse line in which the FLE was deleted. In these mice, both FLCs and ALCs were completely absent, suggesting that the FLE is essential for not only initial FLC differentiation but also ALC redifferentiation at the pubertal stage.

For FLC lineage tracing, previously generated mFLE-CreERT mice (Fig. 1A) (Shima et al., 2015) were crossed with CAG-CAT-EGFP mice (Kawamoto et al., 2000), and pregnant mothers were treated with tamoxifen at embryonic day (E) 12.5 when FLCs start to develop (Miyabayashi et al., 2013) (Fig. 1B). Two days after tamoxifen treatment (E14.5), EGFP expression was detected in the round-shaped interstitial cells that express NR5A1 (Fig. 1C). Furthermore, all the EGFP⁺ cells were positive for HSD3B1 in the E14.5 testis (Fig. S1). These results indicate that FLCs were specifically labeled by mFLE-CreERT. Further investigation of E18.5 and postnatal day (P) 10 testis revealed that NR5A1⁻ peritubular cells were labeled with EGFP in addition to the round-shaped FLCs (Fig. 1D-F). Careful observation of the P10 testis revealed that EGFP was expressed in the peritubular cells (Fig. 1G,G′). Notably, a subset of peritubular cells expressed laminin (Fig. 1G″) as well as collagen type IV (collagen IV) (Fig. 1H,H′) and α-smooth muscle actin (α-SMA) (Fig. 1I,I′), and these α-SMA⁺ cells did not express the Leydig cell marker protein HSD3B1 (Fig. S2). These results indicate that FLCs contribute to the PTMC population. In addition, EGFP expression was also observed in a subset of perivascular α-SMA⁺ cells, indicating that FLCs contribute to the VP population (Fig. 1J,J′). Cell-counting analyses revealed that EGFP⁺/NR5A1⁻ cells (undifferentiated interstitial cells) increased from E14.5 to E18.5, and EGFP⁺/NR5A1⁻/laminin⁺ cells (PTMCs and VPs) appeared and increased from E18.5 to P10 (Fig. 1K). These results suggest that FLCs dedifferentiate at fetal stages, and a subset of dedifferentiated cells transdifferentiate into PTMCs and VPs at the neonatal stages.

The fate of EGFP-labeled FLCs was further traced at adult stage (P56), and we found that FLCs contributed to HSD3B1⁺ Leydig cells (Fig. 2A-A″), α-SMA⁺ PTMCs (Fig. 2B-B″) and VPs. It was considered that HSD3B1⁺ cells contain both FLCs and ALCs. Therefore, to distinguish ALCs from FLCs, we used an antibody against HSD3B6 (Yamamura et al., 2014), which is expressed in ALCs but not in FLCs (Miyabayashi et al., 2017; O'Shaughnessy et al., 2002a; Shima et al., 2015). Immunostaining for EGFP and HSD3B6 revealed that FLCs contributed to both round-shaped HSD3B6⁺ cells (ALCs) (Fig. 2C-C″) and round-shaped HSD3B6⁻ cells (FLCs) (Fig. 2D-D″), and cell-counting analyses revealed that 17% of total EGFP⁺ cells (which correspond to 20% of round-shaped EGFP⁺ cells) were ALCs (Fig. 2E). We previously performed lineage-tracing analyses by labeling FLCs at E14.5 and found that about 5% of round-shaped EGFP⁺ cells in the adult testis were ALCs (Shima et al., 2015). As the current result seemed to contradict our previous result, we next compared the fate of FLCs labeled at E12.5 or E14.5 in the fetal testis and found that only a small proportion of FLCs labeled at E14.5 (less than 5%) contributed to undifferentiated cells and PTMCs/VPs at P10 (Fig. S3). These results suggest that the FLCs at earlier stages have greater potential for dedifferentiation than the FLCs at later stages. Moreover, these findings suggest that dedifferentiated FLCs thereafter contribute to ALCs in the adult testis.

Because the mFLE-CreERT mice could label only a limited population of FLCs (approximately 20% of total FLCs at E14.5), we also used previously generated mFLE-Cre mice (Fig. 3A,B) (Shima et al., 2015) to label the FLC lineage more effectively. Most of the EGFP signals were restricted to FLCs at E15.5 (Fig. 3C-E), whereas EGFP⁺ and NR5A1⁻ peritubular cells and perivascular cells emerged at E18.5 (Fig. 3F-H). These peritubular and perivascular cells expressed Ki67 (Mki67) at a higher frequency than round-shaped FLCs, suggesting that dedifferentiated FLCs proliferate in the fetal testis (Fig. S4). FLC-derived peritubular cells and perivascular cells were also observed in P10 testis (Fig. 3I,J,K,K′), and some of them did not express laminin (Fig. 3J′), but laminin⁺ cells were also observed (Fig. 3J″,J‴). These peritubular and perivascular cells were positive for α-SMA and negative for HSD3B1 (Fig. 3L-L‴), indicating that FLCs contribute to PTMCs and VPs. Cell-counting analyses revealed that peritubular and perivascular cells increased from E15.5 to E18.5, and PTMCs and VPs appeared and increased from E18.5 to P10 (Fig. 3M). Immunostaining analyses revealed that FLCs contributed to 30% of total Leydig cells in the adult testis (Fig. S5). Because our previous study revealed that postnatal FLCs make up 5-10% of the total Leydig cells at P56 (Shima et al., 2015), this suggests that FLCs contribute to a considerable portion of ALCs. Moreover, the percentage of EGFP⁺ cells in total Leydig cells increased to about 80% in the Nr5a1^+/− background (Fig. S5). These results support the results of the lineage-tracing analyses and suggest that FLCs dedifferentiate at fetal stages and thereby contribute to a considerable proportion of ALCs in the adult testis.

To fractionate the FLCs and their descendants, testicular cells were dispersed from mFLE-Cre;CAG-CAT-EGFP double-transgenic mice at P10, and immunostained with a combination of rabbit anti-laminin antibody and Alexa Fluor 594-conjugated anti-rabbit IgG, and subjected to fluorescence-activated cell sorting (FACS). Flow cytometry analyses identified three distinct cell populations – (I) EGFP^Bright, (II) EGFP^Dim/laminin^–, and (III) EGFP^Dim/laminin⁺ – and these populations were considered to correspond with FLCs, dedifferentiated peritubular and perivascular cells, and PTMCs and VPs, respectively (Fig. 4A). Consistently, population I showed high side-scatter intensity, whereas populations II and III showed low side-scatter intensities (Fig. 4B-D). Quantitative RT-PCR (qRT-PCR) analyses revealed that Cre, EGFP, Nr5a1, Hsd3b1 and Insl3 were highly expressed in population I, whereas Acta2 (α-SMA) was highly expressed in population III (Fig. 4E). Population II expressed Arx and Pdgfra, which are marker genes for undifferentiated interstitial cells (Ge et al., 2006; Miyabayashi et al., 2013) (Fig. 4E), as well as previously reported marker genes for ALC stem cells, such as Nr2f2 (COUP-TFII) (Kilcoyne et al., 2014), Nes (Davidoff et al., 2004), Itgav (CD51) (Jiang et al., 2014), Ngfr (Zhang et al., 2013) and Thy1 (CD90) (Li et al., 2016) (Fig. S6).

In addition, the whole transcriptomes of population II and population III were determined by mRNA-sequencing analyses (Table S4) and compared with the previously presented data of single cell RNA expression profiles in the fetal testis (Stevant et al., 2018). As shown in Fig. S7, cluster analyses revealed that both population II and population III were classified in cluster 3, which corresponds to the interstitial progenitor population in the fetal testis. These results strongly suggest that ALC progenitor cells are enriched in population II and, moreover, suggest the possibility that FLC-derived PTMCs and VPs serve as ALC stem cells.

To elucidate the fate of each cell population in the adult testis, sorted cells were transplanted into the 4-week-old nude mouse testis, and cell fate was analyzed at 8 weeks after transplantation (Fig. 5A). The recipient testis sections were immunostained for EGFP, NR5A1 and laminin, and NR5A1⁺ cells (Leydig cells), NR5A1⁻/laminin⁻ cells (undifferentiated interstitial cells) and NR5A1^–/laminin⁺ cells (PTMCs and VPs) were counted (Fig. 5B-C″). In a separate experiment, the recipient testis sections were immunostained for EGFP and HSD3B6, and the percentages of HSD3B6⁺ cells (ALCs) and HSD3B6⁻ cells (FLCs) were calculated (Fig. 5D-E″). As shown in Fig. 5F, these analyses revealed that the cells in population I persisted as HSD3B6^– cells (FLCs) in the recipient testis even at 8 weeks after transplantation, suggesting that FLCs rarely dedifferentiate after the neonatal period. More than 60% of the cells in population II gave rise to ALCs, supporting our hypothesis that population II contains ALC progenitor cells. Intriguingly, about 40% of the cells in population III also gave rise to ALCs as well as the other interstitial cell types, suggesting that the dedifferentiated cells reversibly transform to PTMCs and VPs, and these cells at the neonatal stage still have the potential to differentiate to ALCs.

We next attempted to disrupt the Nr5a1 gene in FLCs. mFLE-Cre mice, Nr5a1 heterozygous mice (Shinoda et al., 1995) and Nr5a1 flox (f/f) mice (Zhao et al., 2001) were crossed and FLC-specific Nr5a1 knockout mice (Nr5a1^ΔFLC/−, hereafter abbreviated as ΔFLC/−) were generated (Fig. 6A). Considering that Nr5a1 knockout led to gonadal agenesis (Luo et al., 1994), it was expected that FLCs would disappear as a result of cell-specific Nr5a1 gene disruption. As expected, HSD3B1⁺ FLCs mostly disappeared from the ΔFLC/− testis at E18.5 (Fig. 6B,C). Likewise, histological analyses confirmed the disappearance of lipid droplet-containing FLCs from the testicular interstitium. Instead, many fibroblastic cells occupied the interstitial space (Fig. 6D,E). qRT-PCR analyses of control and ΔFLC/− testes revealed that the expression of Leydig cell marker genes, such as Star, Hsd3b1 and Insl3, was broadly decreased, whereas the expression of Sertoli and germ cell marker genes was unaffected (Fig. 6F). Correspondingly, testosterone production was significantly decreased in the ΔFLC/− testes (Fig. 6G), and severe defects were observed in androgen-dependent tissues. Namely, the vas deferens was underdeveloped (Fig. 6H-I′, arrows), and the external genitalia were feminized (Fig. 6J-O). Also, owing to decreased Insl3 expression and low testosterone production, the testes were undescended (Fig. 6H,I). Taken together, this phenotypic picture of ΔFLC/− mice at E18.5 indicates that NR5A1 is essential for FLC differentiation, and that FLCs play pivotal roles in the masculinization of male fetuses.

In the adult ΔFLC/− mice, testes were localized in the abdominal cavity (Fig. 7A), and weighed approximately 5% of control testes (Fig. 7B,C). Histological analyses revealed that the interstitial space of the ΔFLC/− testes was occupied by numerous fibroblast-like cells, and that lipid droplet-containing cells – observed abundantly in the control testis – were absent (Fig. 7D,E). Moreover, Masson's trichrome staining identified massive interstitial fibrosis in the ΔFLC/− testis (Fig. 7F,G). Immunostaining for NR5A1 and HSD3B1 revealed only a small number of Leydig cells in the ΔFLC/− testis (Fig. 7H,I). The expression of Leydig cell marker genes (O'Shaughnessy et al., 2002b) as well as ALC-specific genes (O'Shaughnessy et al., 2002a; Shima et al., 2013) was decreased in the ΔFLC/− testis compared with the control (Fig. 7J). Correspondingly, intratesticular testosterone content was significantly decreased (Fig. 7K). This decreased testosterone led to morphological changes in testosterone-dependent tissues, including the disappearance of the seminal vesicles and vas deferens, the appearance of nipples, and feminized external genitalia (Fig. S8). These lines of evidences suggest that a large part of ALCs are derived from dedifferentiated FLCs.

We traced the fate of FLCs in the ΔFLC/− testes, and found that a multilayer of EGFP⁺ and laminin⁺ cells surrounded the seminiferous tubules (Fig. S9). These results support a previous finding that NR5A1 is required for maintenance of Leydig cell identity, but not for survival of Leydig cells (Buaas et al., 2012). Moreover, it was presumed that massive interstitial fibrosis in the ΔFLC/− testes was induced by a large amount of extracellular matrix synthesized by the Nr5a1 gene-disrupted cells.

As shown in Fig. S10, SOX9⁺ Sertoli cells were observed in the seminiferous tubules of the ΔFLC/− testis, but TUBB3 expression pattern was severely disorganized. Cytoskeleton formation by microtubules in Sertoli cells is a crucial event for sperm maturation (De Gendt et al., 2011), and indeed spermatogenesis was arrested in ΔFLC/− testes. A small number of spermatocytes in the ΔFLC/− testes expressed SCP3 (Sycp3) and γH2AX, indicating that they had undergone meiosis. However, elongated spermatids were not observed, and Prm2 (protamine 2) expression was not detected (Fig. S10).

As the labeling efficiencies of mFLE-Cre in the adult testis were different between the Nr5a1^+/+ and Nr5a1^+/− backgrounds (Fig. S5), we compared the phenotypes of Nr5a1^ΔFLC/ΔFLC (ΔFLC/ΔFLC) mice and ΔFLC/− mice. These analyses revealed that a small number of FLCs persisted in the ΔFLC/ΔFLC testis at fetal stages, whereas FLCs almost completely disappeared from the ΔFLC/− fetal testis. Moreover, the interstitial space of the testis was occupied by ALCs in the ΔFLC/ΔFLC testis, whereas only a small number of ALCs developed in the ΔFLC/− testis (Fig. S11). These phenotypic differences between ΔFLC/ΔFLC mice and ΔFLC/− mice seemed to be consistent with our hypothesis that FLCs dedifferentiate and proliferate in the fetal testis, and that these dedifferentiated FLCs thereafter serve as ALC progenitor cells.

To examine the functional importance of the FLE of the Nr5a1 gene (Shima et al., 2012) in vivo, we generated mice lacking FLE (ΔFLE mice) using CRISPR/Cas9 technology (Jinek et al., 2012) (Fig. 8A,B, Fig. S12). Homozygous ΔFLE mice (FLE^−/− mice) demonstrated phenotypes closely similar to those of ΔFLC/− mice. Namely, the testes were not descended (Fig. 8C,D), and the vas deferens was severely underdeveloped at E18.5 (Fig. 8D′). Immunostaining analyses revealed that FLCs disappeared and ARX⁺ undifferentiated interstitial cells increased in the FLE^−/− fetal testis (Fig. 8E-H), whereas Sertoli cells developed normally and the basement membrane was not affected (Fig. S13). At adult stage (P56), the FLE^−/− testis was undescended and much smaller than that of the control (FLE+/−) (Fig. 8I). Histological analyses identified no mature sperms (Fig. 8J,K), and massive interstitial fibrosis was induced in the FLE^−/− testis (Fig. 8L,M). Immunostaining analyses revealed no ALCs in the FLE^−/− testis at P56 (Fig. 8N,O), and, moreover, as in the ΔFLC/− testis, multilayered laminin⁺ peritubular cells surrounded the seminiferous tubules in the FLE^−/− testis (Fig. 8P,Q). These results indicated that FLE is essential not only for initial FLC differentiation, but also for ALC redifferentiation.

Based on the results of this study, we propose a model of Leydig cell development (Fig. 9). FLCs differentiate from FLC progenitor cells in the fetal testis. Thereafter, a subset of FLCs undergo dedifferentiation at the fetal stages, whereas the rest of the FLCs persist in the postnatal testis (postnatal FLCs, pFLCs). The dedifferentiated cells redifferentiate to ALCs at the pubertal stage. In contrast, a subset of dedifferentiated FLCs transdifferentiate to PTMCs and VPs, and these non-steroidogenic cells serve as potential ALC stem cells. The phenotypes of ΔFLC/− mice strongly suggested that a large part of ALCs originate from dedifferentiated FLCs, but we could not completely exclude the possibility that some ALCs arise independently of FLCs. Both FLCs and ALCs completely disappeared from FLE^−/− testis, indicating that the FLE is essential for both initial FLC differentiation and ALC redifferentiation. A previous study showed that the FLE induces EGFP expression in FLCs but not in ALCs (Shima et al., 2015), suggesting that additional regulatory element(s), in addition to the FLE, is required for Nr5a1 gene expression in ALCs.

In this study, we clarified that FLCs dedifferentiate at fetal stages and thereby contribute to ALCs. It may sound unusual that functionally highly differentiated Leydig cells spontaneously return to the undifferentiated state. However, several lines of evidences seem to support our conclusion. First, histological analyses of human testis at prenatal and neonatal stages suggested that FLCs dedifferentiate during the second trimester of pregnancy, and redifferentiate into another postnatal Leydig cell population, neonatal Leydig cells (Teerds and Huhtaniemi, 2015). Second, it is assumed that adult adrenal cortex is formed by dedifferentiation and redifferentiation of fetal adrenal cells. A previous study suggested that fetal adrenal cells give rise to adult adrenal cells (Zubair et al., 2008), whereas another study suggested that the undifferentiated cells at the peripheral zone of the adrenal gland give rise to adult adrenal cells (King et al., 2009). Unifying these conflicting models together, it was proposed that fetal adrenal cell dedifferentiate and localize at the peripheral zone, and thereby redifferentiate into adult adrenal cells (Wood and Hammer, 2011).

The results of our lineage-tracing analyses suggest that FLCs at early fetal stages have the potential of dedifferentiation, but FLCs at later stages rarely dedifferentiate and persist as postnatal FLCs. This phenomenon seems similar to the case of fetal adrenal cells. A previous study reported that fetal adrenal cells labeled at E11.5 contribute to a large portion of adult adrenal cells, whereas fetal adrenal cells labeled at E14.5 rarely contribute to adult adrenal cells (Zubair et al., 2008). These lines of evidence suggest that common mechanisms underlie the developmental transition from FLCs to ALCs as well as fetal adrenal cells to adult adrenal cells. Future studies are expected to give mechanistic insight into this unique developmental process of Leydig cells and adrenocortical cells.

We previously revealed that the FLE induces gene expression in FLCs but not in ALCs (Shima et al., 2015; Shima et al., 2012). However, contrary to our expectations, deletion of the FLE resulted in complete loss of FLCs as well as ALCs. These results suggest that the FLE is essential for NR5A1 expression in both FLCs and ALCs. Although the mechanism that regulates Nr5a1 gene expression in ALCs is still unknown, we speculate that additional unknown regulatory elements, in addition to the FLE, are involved in Nr5a1 gene expression in ALCs.

It is well known that fetal cells and postnatal cells show distinct gene expression profiles in various tissues. Moreover, a recent study revealed that the enhancers activated in fetal cells are demethylated not only in fetal cells but also in adult cells in which the enhancers are inactivated. These results suggested that DNA methylation serves as epigenetic memory of fetal development (Hon et al., 2013). In this study, we showed that FLE deletion resulted in complete loss of FLCs as well as ALCs. These results lead us to speculate that epigenetic modifications of the FLE serve as a memory of FLC development, and this memory is required for reactivation of the Nr5a1 gene and ALC redifferentiation. Future studies are expected to clarify DNA methylation and other epigenetic modifications of the FLE in FLCs and ALCs, and these analyses will give insight into the role of the FLE in the transition from FLCs to ALCs.

Previous studies have shown that VPs serve as stem cells in various tissues (Dore-Duffy and Cleary, 2011; Kramann and Humphreys, 2014). In the case of the testicular interstitium, peritubular cells in the neonatal testis (Ge et al., 2006) as well as VPs in the adult testis (Davidoff et al., 2004) have been reported to be ALC stem cells. Consistent with these observations, we showed that PTMCs as well as VPs in the neonatal testis have the potential to redifferentiate into ALCs, i.e. these non-steroidogenic cells serve as potential ALC stem cells.

As shown in this study, PTMCs and VPs contact the basement membrane (BM). Laminin, one of the components of BM, is known to regulate the maintenance, expansion and differentiation of tissue stem cells in various organs. For example, hepatic progenitor cells need contact with laminin to maintain their naïve characteristics, and loss of such contact induces the transition from progenitor cell to hepatocyte (Williams et al., 2014). As shown in this study, PTMCs and VPs are enveloped by laminin. Moreover, dedifferentiated peritubular and perivascular cells are in contact with laminin, whereas mature ALCs lose the contact, suggesting that ALC differentiation correlates with laminin association. The molecular function of BM components, especially laminin, in ALC differentiation should be evaluated in future studies.

In ΔFLC/− mice, FLCs were almost completely absent and ALCs were substantially decreased. Although we cannot completely exclude the possibility that a small proportion of ALCs arise independently of FLCs, our data strongly suggest that FLCs contribute to the majority of ALCs. Moreover, our results also suggest that FLCs contribute to the PTMC and VP populations in the postnatal testis. PTMCs play essential roles in both maintaining the structure of seminiferous tubules and regulating spermatogenesis (Maekawa et al., 1996; Potter and DeFalco, 2017), whereas VPs play important roles in angiogenesis (Bergers and Song, 2005). Severe structural defects in the testis of both ΔFLC/− mice and ΔFLE mice might be attributable to not only decreased testosterone production but also abnormal development of PTMCs and VPs in the postnatal testis.

Our two mouse models, ΔFLC/− and ΔFLE, showed similar phenotypes, such as cryptorchidism, decreased testosterone production and impaired spermatogenesis. It is important to note that these phenotypes are similar to the symptoms observed in human testicular dysgenesis syndrome (TDS). TDS is one of the major causes of male infertility, and recent epidemiologic analyses of TDS patients identified exposure to environmental factors (e.g. maternal smoking) in fetal life as a potential risk factor for TDS (Juul et al., 2014). Our study showed that dysfunction of FLCs perturbed testis development in the postnatal stages and phenocopied TDS symptoms, suggesting that TDS pathogenesis at postnatal stages might be attributable to damage to the FLCs sustained in utero. We expect this study will give clues to the mechanisms underlying male infertility and TDS in human patients.

The mFLE-CreERT mice and mFLE-Cre mice were previously described (Shima et al., 2015). CAG-CAT-EGFP mice (Kawamoto et al., 2000), Nr5a1 flox mice (Zhao et al., 2001) and Nr5a1 knockout mice (Shinoda et al., 1995) were also previously reported. Nr5a1^+/− male mice were crossed with mFLE-Cre;Nr5a1 flox (f/f) female mice to generate mFLE-Cre(+);Nr5a1 (f/−) mice (FLC lineage-specific Nr5a1 knockout mice, referred to as Nr5a1^ΔFLC/− mice or ΔFLC/− mice). All animal protocols were approved by the Animal Care and Use Committee of Kyushu University and the Animal Care and Use Committee of Kawasaki Medical School.

The mFLE-CreERT mice (Shima et al., 2015) were crossed with CAG-CAT-EGFP mice (Kawamoto et al., 2000), and 100 mg/kg body weight of tamoxifen (Sigma) dissolved in corn oil containing 10% ethanol was administered intraperitoneally to pregnant females at E12.5. The fetuses or postnatal animals were harvested at E14.5, E18.5, P10 and P56, and the testes were collected and subjected to immunofluorescence analyses. In a separate experiment, mFLE-Cre mice were crossed with CAG-CAT-EGFP mice, and the double-transgenic mouse testes were analyzed by immunofluorescence analyses at E15.5, E18.5, P10 and P56. Fetal testis sections were stained with antibodies for EGFP, NR5A1 and laminin, and the percentages of EGFP⁺/NR5A1⁺ cells and EGFP⁺/NR5A1⁻ cells were calculated. P10 testis sections were stained with antibodies for EGFP, NR5A1 and laminin, and the percentages of EGFP⁺/NR5A1⁺ cells, EGFP⁺/NR5A1⁻ cells and EGFP⁺/NR5A1⁻/laminin⁺ cells were calculated. P56 adult testis sections were stained with antibodies for EGFP, HSD3B1 and α-SMA, and the percentages of EGFP⁺/HSD3B1⁺ cells (Leydig cells), EGFP⁺/HSD3B1⁻/α-SMA⁻ cells (undifferentiated cells) and EGFP⁺/HSD3B1⁻/α-SMA⁺ cells (PTMCs and VPs) were calculated. As the EGFP⁺/HSD3B1⁺ population was considered to contain both FLCs and ALCs, we performed immunostaining of P56 testis sections for EGFP and HSD3B6 in a separate experiment, and the percentages of round-shaped EGFP⁺/HSD3B6⁻ cells (FLCs) and round-shaped EGFP⁺/HSD3B6⁺ cells (ALCs) were calculated. To calculate the precise numerical ratio of each cell type, a design-based cell counting method, stereology, was used. Randomly selected 50-µm sections were immunostained with fluorescence-tagged antibodies and z-stacked images were obtained using a Zeiss LSM700 confocal microscope. The stacked images were used for stereology analyses with Stereo Investigator (MBF Bioscience). Cell numbers per unit volume were measured and the ratio of each cell type was calculated. We used three to six animals at each stage, and three sections from each sample were used for the analyses.

Testes were fixed by perfusion fixation and/or immersion fixation with 4% paraformaldehyde (Wako Pure Chemical Industries) and 0.5% glutaraldehyde (Nissin EM Corporation) in PBS. The samples were then embedded in OCT compound (SAKURA Finetek) and 50- or 10-µm-thick sections were cut using a Leica CM3050 S cryostat. The 50-µm sections were stained by the free-floating staining method (Jinno et al., 1998), whereas the 10-µm sections were stained as previously described (Shima et al., 2015). The primary and secondary antibodies used in this study are listed in Table S1. DAPI (4′6′-diamidino-2-phenylindole) (Sigma) was used for nuclear staining. Immunofluorescence was examined using a Zeiss LSM700 confocal microscope (Zeiss).

Testes were fixed by perfusion fixation followed by immersion fixation for 36 h with 2% paraformaldehyde (Wako Pure Chemical Industries) and 2% glutaraldehyde (Nissin EM Corporation) in PBS. The tissues were embedded in paraffin wax or resin and then sectioned at 5 µm or 0.4 µm thickness, respectively. The 5-µm sections were subjected to Hematoxylin and Eosin (H&E) staining and Masson's trichrome staining (Merck Millipore), whereas the 0.4-µm sections were subjected to Toluidine Blue staining.

Testes were collected from mFLE-Cre;CAG-CAT-EGFP double-transgenic mice at P10 and the cells were dispersed as previously described (Shima et al., 2013). Dispersed cells were incubated in 3% bovine serum albumin (BSA)/PBS containing 1 mg/ml anti-laminin antibody (L9393, Sigma-Aldrich) at 4°C for 30 min. The cells were washed with 3% BSA/PBS three times and then incubated with 3% BSA/PBS containing Alexa Fluor 594-conjugated anti-rabbit IgG (Thermo Fisher Scientific) at 4°C for 30 min. The cells were washed with 3% BSA/PBS three times, filtered with a 70-µm cell strainer (Falcon), and subjected to FACS using JSAN (Bay Bioscience) to fractionate the cells into three populations based on the intensities of EGFP fluorescence and Alexa Fluor fluorescence (I: EGFP^Bright; II: EGFP^Dim/laminin⁻; III: EGFP^Dim/laminin⁺). Sorted cells were then used for qRT-PCR analyses, mRNA sequence analyses, and cell transplantation experiments.

qRT-PCR was performed as previously described (Shima et al., 2013). In brief, total RNA was prepared from the sorted cells or tissues using the RNeasy micro kit (Qiagen) and 50 ng of total RNAs were subjected to reverse transcription using SuperScript II Reverse Transcriptase or M-MLV reverse transcriptase (Thermo Fisher Scientific). Quantitative PCR was performed with an ABI 7500 real-time PCR system (Applied Biosystems) or a CFX96 Real-Time System (Bio-Rad) using THUNDERBIRD SYBR qPCR mix (TOYOBO) or SYBR Select Master Mix (Applied Biosystems). Gene expression levels were standardized to those of Actb (β-actin) and reported as mean±s.e.m. Primers used for PCR are listed in Table S2.

mRNA sequence analyses were performed as previously described (Miyabayashi et al., 2017). Briefly, poly(A) RNAs were isolated from total RNAs (50 ng from each sample) prepared from sorted cells (cell populations II and III, n=2 for each population). Sequence libraries were constructed using the NEBNext Ultra Directional RNA Library Prep Kit for Illumina (NEB). cDNAs were sequenced using a Hiseq 1500 (51-bp single-end; Illumina), and the fastq data were deposited in DDBJ (accession number DRA007347). Sequence data were mapped to the reference genome (UCSC mm10) using the TopHat (version 2.1.1) with the option ‘--library-type fr-secondstrand’. Because the mice have a transgene (mFLE-Cre), we detected a fusion transcript consisting of the 5′ untranslated region of the Nr5a1 gene and Cre cDNA in the sequence library. This artificial fusion transcript was distinguished from the endogenous transcripts as previously described (Miyabayashi et al., 2017). Cufflinks (version 2.2.1) was used with the option ‘-u --library-type fr-secondstrand’ for quantification of the transcripts, and gene expression data were presented as FPKM (Table S4). Gene expression profiles in the sorted cell populations II and III were compared with those in the previously described fetal testicular cells (Stevant et al., 2018). We identified 21,984 genes in our data and 16,459 genes in the previous data. Of these genes, 15,459 genes that appeared in both data sets were extracted and expression levels were normalized by quantile normalization (preprocessCore, bioconductor package in R). Of the 811 genes used for the cluster analyses in the previous study, 776 genes were found in our data set. These genes were extracted and t-distributed stochastic neighbor embedding (t-SNE) analysis was performed according to the method described in the previous study (Stevant et al., 2018).

P28 nude mice (BALB/cSlc-nu/nu, Japan SLC) were used as recipients. Sorted cells were suspended in 3% BSA/PBS at a concentration of 1×10⁵ cells in 100 µl and then injected into the interstitial space of the testes of recipient mice using a glass pipette. The testes were harvested 8 weeks after transplantation and subjected to immunofluorescence analyses. To trace the fate of transplanted cells, randomly selected 50-µm sections were immunostained with antibodies for EGFP, NR5A1 and laminin, and the percentages of EGFP⁺/NR5A1⁺ cells (Leydig cells), EGFP⁺/NR5A1⁻/laminin⁻ cells (undifferentiated cells) and EGFP⁺/NR5A1⁻/laminin⁺ cells (PTMCs and VPs) were calculated. As the EGFP⁺/NR5A1⁺ population was considered to contain both FLCs and ALCs, we performed immunostaining of recipient testis sections for EGFP and HSD3B6 in a separate experiment, and the percentages of round-shaped EGFP⁺/HSD3B6⁻ cells (FLCs) and round-shaped EGFP⁺/HSD3B6⁺ cells (ALCs) were calculated.

Testes were collected from control mice and Nr5a1^ΔFLC/− mice at E18.5 and P56. Intratesticular testosterone concentration was measured by liquid chromatography with tandem mass spectrometry analyses (ASKA Pharma Medical Co.).

To delete the FLE of the Nr5a1 gene, genome editing experiments were performed using CRISPR/Cas9 technology (Jinek et al., 2012). Guide RNAs were designed to target the upstream and downstream regions of the FLE using CRISPR direct (http://crispr.dbcls.jp/), and oligonucleotides complementary to the guide RNAs were inserted into the PX458 plasmid [pSpCas9(BB)-2A-GFP; Addgene plasmid #48138, deposited by Feng Zhang]. The plasmids targeting the upstream and downstream sequences of the FLE were mixed and injected into the pronuclei of the fertilized eggs. The eggs were then transferred into the oviducts of recipient mothers, and genotypes of the pups were analyzed by PCR. The sequences of guide RNAs and genotyping primers are shown in Table S3. Four distinct F0 mice with different deletion patterns were obtained. The FLE was homozygously deleted in a female F0 founder and this mouse was sterile (A1 in Fig. S12). The other three males (A4, A5 and B2 in Fig. S12) showed heterozygous deletion of the FLE, and the pups of each mouse were further crossed with C57BL/6J mice more than five times. After the backcross, homozygous FLE-deletion mice were generated. The three distinct mouse lines showed exactly the same phenotypes, and line B2 was chosen for data presentation.

At least three animals were used for each experimental group unless otherwise noted. qRT-PCR analyses were performed in triplicate. Data are presented as mean±s.e.m., and statistical differences between experimental groups were examined by the two-tailed Student's t-test.

We thank Professor Jun-ichi Miyazaki (Osaka University, Japan) for kindly providing the CAG-CAT-EGFP mice. We also thank Ms Kaoru Akiyama (Hanaichi Ultrastructure Research Institute, Okazaki, Japan) for her technical support with the histological analyses. We deeply thank Professor Ken-ichirou Morohashi (Kyushu University, Japan) for his critical advice and support on this project. We appreciate the technical support provided by the Research Support Center, Graduate School of Medical Sciences, Kyushu University, and Central Research Center, Kawasaki Medical School. We thank Edanz Group (www.edanzediting.com/ac) for editing a draft of this manuscript.