Glutamine protects mouse spermatogonial stem cells against NOX1-derived ROS for sustaining self-renewal division in vitro

Spermatogonial stem cells (SSCs) are the only stem cells in the body that transmit genetic information to the next generation (de Rooij and Russell, 2000; Meistrich and van Beek, 1993). Throughout the lifespan of male animals, SSCs constantly undergo self-renewal to sustain sperm production. Although SSCs make up only a small population in the testis, they undergo self-renewal in a unique microenvironment (i.e. a niche) that provides self-renewing factors for spermatogenic cells (Griswold, 2018). Several cytokines, such as GDNF and FGF2, are considered to be essential for self-renewal division (Kanatsu-Shinohara and Shinohara, 2013); however, much remains unknown with respect to other requirements for SSC maintenance. Previous studies have shown that spermatogenic cells on the basement membrane strongly express glycolysis-related genes (Nakamura et al., 1984). Consistent with this observation, we and others have shown that glycolysis is crucial for self-renewal division (Kanatsu-Shinohara et al., 2016; Helsel et al., 2017). The stimulation of glycolysis by chemical induction or lipid component removal enhances SSC proliferation, while the inhibition of glycolysis impairs proliferation.

Although glucose (Gluc) is a major energy source for SSCs, amino acids also influence stem cell metabolism. For example, embryonic stem (ES) cells depend on Pro metabolism and downstream metabolites of that pathway may serve as signaling molecules (Casalino et al., 2011). Similarly, Thr (mouse) or Met (human) catabolism is required for ES cell growth and differentiation because these amino acids serve as precursors for donor molecules that are used in histone methylation and acetylation (Wang et al., 2009; Shiraki et al., 2014). Hematopoietic stem cells (HSCs) are sensitive to the loss of branched-chain amino acids, such as Ile or Val (Wilkinson et al., 2018). However, the amino acid requirements of SSCs have not been investigated. This is probably because SSCs make up a small population of all germ cells; moreover, their close association with Sertoli cells in the seminiferous tubules prevents direct analysis of their metabolic requirements in germ cells.

A valuable approach to this problem involves the recapitulation of self-renewal division in vitro. We have previously described a long-term culture system for mouse SSCs (Kanatsu-Shinohara et al., 2003b). Addition of GDNF and FGF2 to freshly isolated testis cells induces spermatogonia colony formation. These cells, designated as germline stem (GS) cells, can proliferate for more than 2 years and produce offspring upon microinjection into the seminiferous tubules of infertile mice (Kanatsu-Shinohara and Shinohara, 2013). GS cells are the primary population of spermatogonia, which also includes a small number of SSCs. We have previously used this culture system to demonstrate that reactive oxygen species (ROS) are required for SSC self-renewal (Morimoto et al., 2013). Supplementation of GDNF and FGF2 increased ROS generation via NOX1; Nox1-deficient SSCs underwent limited self-renewal division both in vitro and in vivo (Morimoto et al., 2013, 2019). These findings differ from the results in most other stem cell types, which are sensitive to increased levels of ROS (Bigarella et al., 2014). For example, ES cells that are exposed to hydrogen peroxide quickly undergo apoptosis, while the same treatment enhances proliferation in GS cells (Mori et al., 2021). Similarly, increased levels of ROS cause HSCs to undergo senescence via p38 mitogen-activated protein kinase activation (Ito et al., 2006). Because large numbers of GS cells can be collected for molecular and biochemical characterization, functional analysis can be performed to identify factors involved in ROS regulation and self-renewal.

In this study, we used GS cells to analyze the amino acid requirements of SSCs. Feeder-free cultures allowed us to define specific requirements for SSCs in vitro, without the effects of exogenous molecules secreted from feeder cells (Kanatsu-Shinohara et al., 2005). Our analyses revealed that Gln is essential for SSC propagation in vitro. Gln has major roles in several metabolic pathways, such as nitrogen metabolism, ammonia detoxification, acid-base homeostasis, osmotic regulation, and cell signaling and proliferation (Watford, 2015). Gln is converted to Glu by mitochondrial glutaminase activity in nitrogen-donating reactions of nucleotide synthesis, as well during the production of reduced glutathione. Two forms of glutaminases are known: Gls encodes the kidney-type isoenzyme, whereas Gls2 encodes the liver-type isoenzyme. A spliced variant, known as glutaminase C, is closely related to cancer initiation and progression. Most Gln is degraded by glutaminase to Glu; it is then used as a substrate for urea and Gluc synthesis, or as a fuel (ATP production). However, a small fraction of Gln is also converted to Glu by Asn synthetase (ASNS), thereby producing structurally related Asn (Watford, 2015). Gln is one of the most abundant amino acids in the seminiferous tubules (Hinton, 1990). However, Gln is not necessarily essential for cell culture because some cells can be cultured using high concentrations of Glu (Levintow and Eagle, 1961). Nevertheless, our analysis suggests that Gln is essential for protection against NOX1-derived ROS and ensures ROS-dependent self-renewal division via Myc induction.

To examine the amino acid requirements of GS cells, we first evaluated GS cell proliferation in the absence of specific amino acids. We established feeder-free cultures of GS cells to avoid the influence of nutrients supplied by feeder cells (Kanatsu-Shinohara et al., 2005). GS cells were plated on a laminin-coated dish and recovered by trypsin digestion after 4 days of culture. Compared with complete culture medium that contained all amino acids, GS cell proliferation was significantly reduced when Ala, Arg, Asn, Cys, Gln, Gly, Leu, Lys, Ser, Tyr or Val was removed from the culture (Fig. 1A). Among these nutrient-deficient conditions, GS cells grown without Ala, Arg, Gln, Leu or Lys failed to proliferate during the 4-day culture period; Lys deficiency had the strongest effect, with only ∼13% recovery compared with the complete medium. To determine whether feeder cells could rescue the phenotypes, we cultured GS cells on mouse embryonic fibroblasts (MEFs), which are routinely used for GS cell maintenance. Although we were able to rescue proliferative defects caused by Ala deficiency, we were unable to rescue defects caused by other amino acids (Fig. 1B,C). This result suggested that MEFs produced Ala and contributed to GS cell proliferation.

We then measured the amounts of amino acids in the culture media to evaluate which amino acids are most extensively used in complete medium. GS cells were plated on a laminin-coated dish and the amounts of amino acids were measured using high-performance liquid chromatography (HPLC). Except for Ala and Glu, the levels of all tested amino acids were significantly reduced after 4 days of culture (Fig. 1D). Among the tested amino acids, Gln exhibited the greatest reduction in the spent culture medium. The total amount of Gln in complete medium (6 mM) was significantly greater than the amounts of other amino acids, which suggested that GS cells depend heavily on Gln metabolism. We also tested the impact of Gln concentration on GS cell proliferation; we found that GS cells proliferated most efficiently in the presence of ∼6 mM of Gln (Fig. 1E), but their proliferation decreased at higher doses, suggesting that the Gln concentration in the original medium was optimal for GS cell cultures.

To examine the biological impact of Gln deficiency on SSCs, we carried out phenotypic analysis of GS cells. Real-time polymerase chain reaction (PCR) analysis of spermatogonial markers revealed dynamic changes in spermatogonia marker expression. In particular, we noted increased expression of SSC-associated markers, including Gfra1, Id4 and Nanos2 (Fig. 1F). These molecules are expressed in A_single or A_paired undifferentiated spermatogonia, which are considered to contain SSCs. Conversely, we also found that genes associated with differentiation, such as Neurog3, Sohlh1 and Kit, were significantly downregulated after Gln deprivation. Although the frequency of SSCs in GS cell culture is very low (1-2%) (Kanatsu-Shinohara and Shinohara, 2013), the upregulation of SSC-related markers and downregulation of differentiation markers suggested that Gln deprivation increases the frequency of SSCs in GS cell cultures.

To evaluate the frequency of SSCs in GS cell culture after Gln deprivation, we performed spermatogonial transplantation (Brinster and Zimmermann, 1994). After 3 days of Gln deprivation, GS cells were dissociated into single cells and transplanted into the seminiferous tubules of infertile mouse testes. Recipient testes were analyzed 2 months after transplantation. Unexpectedly, the number of colonies generated from Gln-deprived cultures was significantly reduced (Fig. 1G,H). These results suggest that Gln deprivation reduces the numbers of SSCs in GS cell cultures.

To examine the mechanism of Gln uptake in GS cells, we performed reverse transcription-PCR (RT-PCR) analysis of Gln transporters. In normal GS cell culture conditions with Gln supplementation, GS cells expressed Slc38a1, Slc38a2, Slc38a5, Slc38a6, Slc38a7, Slc38a9, Slc38a10 and Slc1a5 (Fig. 2A). Real-time PCR analysis showed that Slc38a9 and Slc1a5 were relatively strongly expressed in GS cells (Fig. 2B). Gln deprivation did not significantly change the expression of Slc38a9, while Slc1a5 was significantly downregulated (Fig. 2C); this finding suggested that Slc1a5 is regulated by Gln. We performed gene knockdown (KD) by lentivirus-mediated short hairpin RNA (shRNA) to study the functional significance of these genes. Cell recovery was significantly decreased when Slc38a9 or Slc1a5 was depleted (Fig. 2D). Real-time PCR analysis confirmed the downregulation of target genes (Fig. S1A).

To determine how Gln was metabolized in GS cells, we examined the expression patterns of glutaminases that convert Gln into Glu. Real-time PCR analysis showed higher expression levels of Gls than Gls2 (Fig. 2E). Gln deprivation slightly increased the expression of Gls (Fig. 2F). To examine the functions of glutaminases, we first used a Gln analog, L-DON, that competes with Gln for glutaminolysis and inhibits both glutaminase and other enzymes that use Gln. As expected, L-DON was able to block GS cell proliferation (Fig. 2G). We then examined the function of glutaminases using specific chemical inhibitors. Two glutaminase-specific inhibitors, BPTES and 968, inhibited proliferation (Fig. 2G). We confirmed these results of chemical inhibitors by KD experiments, which showed significantly decreased proliferation upon depletion of Gls (Fig. 2H). Real-time PCR confirmed the downregulation of the target genes (Fig. S1B). These results suggested that Gls was essential for GS cell proliferation.

We then compared the effects of Gln deprivation and Gluc deprivation on GS cell proliferation. We removed Gln or Gluc from the medium and compared the numbers of cells. After 1 week of culture under Gln deprivation conditions, cell proliferation was significantly reduced (Fig. 3A,B). This reduction of cell recovery was comparable with the reduction after Gluc deprivation conditions. Gln-driven oxidative phosphorylation is a major ATP source; thus, we evaluated the impact of Gln deprivation on ATP production. The removal of Gln or Gluc significantly reduced ATP levels. In contrast, ADP levels were significantly reduced by Gluc deprivation but not Gln deprivation (Fig. 3C). We then calculated the ADP/ATP ratio, which has been used to differentiate apoptotic cell death from necrotic cell death. The ADP/ATP ratio is expected to be 0.11 or less in healthy viable cells, between 0.11 and 1.0 in cells undergoing apoptosis, and more than 1.0 in necrotic cells because of the intracellular energy dissipation associated with necrosis (Blenn et al., 2006). However, our analyses revealed that GS cells exhibited relatively higher values, regardless of whether they were cultured with Gln or Gluc. Nevertheless, we found a greater impact of Gluc on ADP/ATP levels, which suggested that Gluc deprivation induces necrosis in more cells.

The low cell recovery observed after Gln deprivation suggested that Gln influences either the cell cycle or apoptosis. To distinguish between these possibilities, we first carried out MKI67 immunostaining (Fig. 3D). Consistent with the poor ATP production mentioned above, Gln deprivation reduced the proportion of cells expressing MKI67. Terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) staining also showed that the number of apoptotic cells had increased (Fig. 3E). This suggested that the Gln deprivation-induced low cell recovery was the result of reduced proliferation and increased apoptosis.

Apoptosis may be induced by the presence of increased ROS levels. Although we previously showed that ROS are required for SSC self-renewal, excessive levels of ROS are harmful for SSCs (Morimoto et al., 2013). Because Gln is potentially involved in ROS regulation (Watford, 2015), we examined ROS levels after Gln deprivation. ROS are produced by NOX enzymes and mitochondria (Morimoto et al., 2021). To study the impact of NOX enzymes, we focused on Nox1. This was because SSCs in Nox1 knockout (KO) testes undergo limited self-renewal division (Morimoto et al., 2013). To examine the impact of mitochondria-derived ROS, we used GS cells from Top1mt KO mice (Khiati et al., 2015). Top1mt is a mitochondria-specific topoisomerase. Top1mt KO mice initially demonstrate normal spermatogenesis, which gradually diminishes; they exhibit significantly smaller testes and empty seminiferous tubules at more than 1 year of age, along with infertility (Morimoto et al., 2021).

We carried out flow cytometric analysis and examined ROS levels. Analyses of wild-type cells showed a significant increase in total ROS levels upon Gln deprivation. We also observed an increased production of mitochondria-derived ROS, which could be detected by MitoSOX. Contrary to our expectations, however, both types of mutant GS cells showed reduction in total ROS levels and in mitochondria-derived ROS levels, even in Gln-supplemented cultures (Fig. 4A,B); these findings suggested interactions between ROS generated from different compartments. Gln depletion further increased ROS levels in both cell types. Therefore, these mutant cells exhibited similar staining patterns in terms of the ROS response.

To examine the impact of Gln deprivation in a functional manner, we cultured both types of mutant cells in the absence of Gln. When total cell recovery was compared with wild-type cells, Nox1 KO GS cells showed significantly increased cell recovery (Fig. 4C). However, no significant changes were found in Top1mt KO GS cells (Fig. 4D). Although MKI67 staining did not show any apparent changes (Fig. 4E), the number of apoptotic cells was significantly reduced in Nox1 KO GS cells (Fig. 4F). In contrast, Top1mt KO GS cells and wild-type GS cells showed similar staining patterns (Fig. 4G,H). These results suggested that reduced ROS production from NOX1 contributed to enhanced GS cell survival upon Gln deprivation.

The above results suggested that Gln was essential for protection against NOX1-generated ROS. We then examined how Gln deprivation caused poor proliferation through increased levels of ROS. We previously showed that MYC is downregulated in Nox1 KO GS cells (Morimoto et al., 2021). Moreover, Myc family genes are generally involved in Gln metabolism through the induction of glutaminases and Gln transporters (Wise et al., 2008). We first examined the impact of Gln on Myc expression. Real-time PCR analysis showed that Mycn expression depended on Gln, while Myc expression did not (Fig. 5A). This relationship was also confirmed by western blotting (Fig. 5B). Therefore, Gln was essential for inducing MYC, which enhances SSC activity of fresh testis cells after overexpression (Kanatsu-Shinohara et al., 2014).

We then studied the impact of Myc on Gln metabolism using Myc DKO GS cells because Myc KO and Mycn KO cells do not show an apparent phenotype (Kanatsu-Shinohara et al., 2016). Myc has been shown to promote Gln uptake through direct transactivation of SLC1A5 expression (Nicklin et al., 2009); the depletion of Slc1a5 led to impaired GS cell proliferation (Fig. 2D). Therefore, we examined the expression of Slc1a5 using real-time PCR. As expected, Slc1a5 was significantly reduced in Myc DKO GS cells (Fig. 5C). We also examined the expression patterns of glutaminases; however, we did not find any significant changes using real-time PCR, regardless of Gln deprivation (Fig. 5D). A recent study suggested that MYC was also involved in Gln synthesis (Bott et al., 2015); therefore, we analyzed Glul expression, which produces Gln from Glu and ammonia. Although Gln deprivation led to downregulation of Glul in wild-type cells, the levels of Glul were comparable in Myc DKO GS cells (Fig. 5E), which suggested that Myc stimulates Glul expression. To confirm this observation, we carried out western blot analysis. However, western blot analysis revealed GLUL upregulation in both wild-type and Myc DKO GS cells in the absence of Gln (Fig. 5F). These results suggested that Myc is not involved in GLUL expression.

Finally, we examined the impact of Gln deprivation on Myc DKO GS cells. Cell recovery after Gln deprivation showed that Myc DKO GS cells continue to require Gln for proliferation (Fig. 5G). Consistent with this observation, immunostaining confirmed a significant decrease in MKI67⁺ cells after Gln deprivation (Fig. 5H). Because reduction of proliferation occurred in the absence of Myc genes, these results suggested that Gln stimulates proliferation in Myc-independent manner. Moreover, TUNEL staining also showed a decrease in the number of apoptotic cells, indicating that MYC is involved in apoptosis caused by Gln deprivation (Fig. 5I). Taken together, these results suggested that Gln attenuates MYC-induced apoptosis but has additional targets to promote GS cell proliferation even in the absence of Myc genes.

To understand the mechanism underlying apoptosis during Gln deprivation, we examined Trp53; we have previously found that apoptosis is suppressed in Trp53 KO GS cells after exposure to hydrogen peroxide (Mori et al., 2021). Because apoptosis was attenuated in Nox1 KO GS cells, Gln deprivation was presumed to enhance ROS production and induce TRP53 expression, thereby leading to apoptosis. Western blot analysis revealed that Gln deprivation led to increased TRP53 expression (Fig. 6A). To test the direct involvement of TRP53, we used GS cells that had been derived from Trp53 KO mice. Trp53 KO GS cells proliferated efficiently even in the absence of Gln (Fig. 6B,C). As expected, although these cells did not show any apparent changes in terms of MKI67 staining (Fig. 6D), they exhibited a significant decrease in TUNEL staining (Fig. 6E). These results suggest that Gln deprivation causes GS cell apoptosis via TRP53 activation.

The reduction of apoptosis in Nox1 KO GS cells suggested that excessive ROS production was responsible for Gln deprivation-induced cell death; thus, we measured glutathione (GSH) levels. Glu derived from Gln is used in GSH synthesis, and the reduced form of GSH is a scavenger for ROS. ROS detoxification results in oxidation of the reduced GSH into glutathione-disulfide (GSSG). Both GSSG and GSH levels were significantly reduced in GS cells (Fig. 7A). This suggested that Gln-derived GSH protects GS cells from excessive ROS production. To determine whether GSH improved survival, we added GSH to GS cell cultures after Gln deprivation. As expected, adding 1 mM of GSH significantly increased GS cell recovery (Fig. 7B). Although this suggested that GSH production could rescue GS cell apoptosis, these cells disappeared after the second and third passages; long-term GS cell rescue was not possible. Therefore, ROS suppression alone did not guarantee long-term GS cell proliferation, suggesting that other Gln-derived metabolites contribute to the maintenance of SSC self-renewal.

Because Gln is deaminated to produce Glu, we presumed that the addition of excessive amounts of Glu might rescue the GS cell phenotype. The ability to use Glu instead of Gln appeared to be a fairly regular characteristic of primary cultures (Levintow and Eagle, 1961). However, the addition of Glu (up to 15 mM) did not have any impact on GS cells that were cultured without Gln (Fig. 7C). To overcome this problem, we then attempted rescue using Asn, which is produced from Gln by ASNS (which converts Asp and Gln to Asn and Glu, respectively). We suspected that the addition of Asn might promote Gln production and rescue the GS cells. However, low concentrations (0.2 mM) of Asn did not show any effects. Nevertheless, the addition of 15 mM of Asn led to cell recovery comparable with Gln-supplemented cultures (Fig. 7D,E), although long-term culture proliferation was partially reduced (Fig. 7F).

To confirm the SSC activity of the cultured cells, we carried out spermatogonial transplantation. After 4 weeks of culture, these cells were able to restore spermatogenesis in recipient testes (Fig. 7G). Although we were able to confirm normal-appearing spermatogenesis (Fig. 7H), Asn-supplemented cultures generated fewer colonies than did Gln-supplemented cultures; however, this difference was not statistically significant (Fig. 7I). Because this finding suggested poor germ cell quality after Asn supplementation, we tested whether sperm generated in the seminiferous tubules were functionally normal. The cells were cultured for 83 days with Asn supplementation. Three months after transplantation, donor spermatogenic cells in the recipient testes were recovered by repeated pipetting of the green seminiferous tubules to release the sperm. These sperm were microinjected into oocytes using the intracytoplasmic sperm injection technique. In total, 43 embryos were constructed. After 24 h of in vitro culture, 40 two-cell embryos were transferred into the oviducts of pseudo-pregnant mothers. Fifteen offspring were born from Gln-deprived GS cells (Fig. 7J). Green fluorescence in the offspring indicated the donor cell origin. These results confirmed that the genome in the sperm was capable of undergoing syngamy with the oocyte genome and producing offspring.

This study analyzed the amino acid requirements of GS cells; it shows that deficiencies of several amino acids, including Gln, prevent GS cell proliferation. To our knowledge, this is the first report of amino acid requirement of GS cells and the data may be useful for future studies. Additional analysis of spent culture media revealed that Gln was the most extensively used amino acid in GS cell culture medium. The concentration of Gln in the blood is 0.5-0.8 mM, while most cell culture media contain 2-4 mM Gln (Krall et al., 2016). The importance of Gln in spermatogenesis was first demonstrated by testis organ culture studies. These studies indicated that Gln was essential for spermatogenesis progression. Although small numbers of spermatocytes were found in conventional cultures containing 2 mM of Gln, greater amounts of Gln (4 mM) could effectively induce spermatocyte generation (Steinberger and Steinberger, 1966). Therefore, Gln was considered essential for spermatocyte differentiation. However, there have been no reports of the role of Gln in premeiotic cells, including SSCs.

Real-time PCR analysis of GS cells showed increased expression of SSC-related markers after Gln-deprivation. Therefore, it was possible that Gln deprivation eliminated progenitor cells and enriched SSCs in vitro. However, transplantation assays showed that Gln deprivation decreases SSC frequency in GS cell cultures. One possibility for this discrepancy is the impact of Gln on translation efficiency. It is known that Gln influences translation efficiency, often via the mechanistic targeting of rapamycin or integrated stress response pathways (Gameiro and Struhl, 2018); therefore, Gln deficiency likely changed the global protein translation patterns in GS cells and resulted in the discrepancy between the stem cell marker genes and functional identity. Alternatively, Gln deprivation might have induced ectopic expression of SSC markers in progenitor cells. Because these markers were identified in SSCs during steady-state spermatogenesis, little is known about the changes in stem cell phenotype during regeneration. It is possible that excessive stimulation of self-renewal division by unphysiological levels of cytokines induced these stem cell markers to be ectopically expressed in progenitor cells.

We examined the effects of Gln deprivation on Nox1 and Top1mt KO GS cells because ROS are produced by NOX enzymes and mitochondria (Bigarella et al., 2014). In another study, we demonstrated that, compared with mitochondria-derived ROS, NOX1-derived ROS are more important for self-renewal than mitochondria-derived ROS (Morimoto et al., 2021). Consistent with this notion, Top1mt KO GS cells, which have accumulated mutations in mitochondrial DNA and have reduced mitochondrial ROS levels, did not show any apparent changes in cell recovery after Gln deprivation. In contrast, significantly attenuated apoptosis of Nox1 KO GS cells provided evidence that ROS generated by NOX1 were responsible for Gln deprivation-induced cell death. Because Nox1 KO GS cells proliferated poorly only when they were cultured under hypoxia (Morimoto et al., 2021), we initially thought that Nox1 KO GS cells behave similarly to wild-type GS cells. However, reduced apoptosis of Nox1 KO GS cells suggested that NOX1-derived ROS have a role in inducing apoptosis even under normoxia. The cell death occurred despite reductions in total and mitochondria-derived ROS in both Nox1 KO and Top1mt KO GS cells. Although we currently do not know how Gln specifically confers protection against NOX1-derived ROS, a previous report showed NOX1 is involved in Gln metabolism. NOX1 levels were inversely correlated with levels of mitochondrial glutamate dehydrogenase GDH1 in two cell lines (Bertram et al., 2015). In these cells, Nox1 depletion upregulated GDH1, which catalyzes reversible oxidative deamination of Glu to α-ketoglutarate. This might have allowed oxidation of Glu in the medium to α-ketoglutarate and provided the TCA cycle with a substrate, thereby increasing the survival of cells. Although this possibility needs to be tested in future study, the present results confirm our previous findings that ROS from two different origins have distinct roles in SSCs and progenitor cells in vivo (Morimoto et al., 2021).

Subsequently, we focused on Myc to determine the mechanism by which Gln withdrawal causes poor proliferation; we previously found that MYC is downregulated in Nox1 KO GS cells (Morimoto et al., 2021). Moreover, MYC is closely associated with Gln metabolism. MYC promotes Gln uptake by activating Gln transporters, including SLC1A5; the depletion of Slc1a5 has been shown to compromise GS cell proliferation. This depletion also enhances the translation of Gls and causes enhanced glutaminolysis (Dong et al., 2020). Consistent with these previous observations, we observed the downregulation of Slc1a5 in Myc DKO GS cells. However, we failed to find significant changes in Gls and GLUL levels. On the other hand, Gln promoted MYCN expression. Because MYCN enhances self-renewal division by stimulating glycolysis (Kanatsu-Shinohara et al., 2016), Gln indirectly promotes glycolysis, thereby enhancing GS cell proliferation (Kanatsu-Shinohara et al., 2016). However, MYCN induction alone is not sufficient to explain the impact of Gln because Myc DKO GS cells also proliferated poorly in the absence of Gln. Although these results suggest additional target genes, it is possible that the remaining members of Myc family genes (Bmyc, Mycl or Mycs), also contribute to Gln-induced proliferation. For example, because Bmyc KO mice also show abnormal spermatogenesis (Turunen et al., 2012), it is likely that Bmyc compensated for the simultaneous loss of Myc and Mycn genes upon Gln deprivation.

We also observed the involvement of Trp53 in apoptosis after Gln deprivation. Trp53 has been implicated in the quality control of germ cells (Yin et al., 1998). Mice lacking Trp53 exhibit reduced spontaneous apoptosis but sire fewer offspring. Thus far, there is minimal information concerning the regulation of Trp53 in vivo. Based on our results, Gln deficiency may be a potent inducer of Trp53. Trp53 deficiency per se does not directly influence SSC activity (Ishii et al., 2014), although it may maintain the GS cell genome by promoting apoptosis in ROS-exposed cells (Mori et al., 2021). High base excision repair activity in GS cells protects DNA from ROS by upregulating Ogg1, a protein essential for oxidative stress-induced DNA demethylation. Most of the previous studies used irradiation to study the role of Trp53 in spermatogonia. However, irradiation is apparently not a physiological stimulus for the analysis of spermatogenesis; much remains unknown how Trp53 is regulated during spermatogenesis under physiological conditions. We speculate that competition for nutrition, including Gln, may promote the survival of cells with ROS levels that are appropriate for spermatogenesis maintenance. It will be intriguing to investigate whether deficiencies of other amino acids similarly induce Trp53 activity and trigger apoptosis.

Because apoptosis was attenuated in Nox1 KO GS cells, we examined the impact of GSH supplementation. Glu is used in GSH synthesis through the condensation with Cys and Gly by Glu-Cys ligase and GSH synthetase (Watford, 2015). Notably, testes have higher levels of GSH than do other organs (Calvin and Turner, 1982). However, cellular distribution of GSH is heterogenous, and it was proposed that depletion of the mitochondrial GSH pool frequently correlated better with toxic cell death than with overall loss of intracellular GSH (Marí et al., 2020). This may also contribute to the phenotypic difference between Top1mt KO and Nox1 KO GS cells. The importance of GSH has been demonstrated by defective spermatogenesis in several KO mice, which lack genes that are involved in redox regulation. For example, the lack of genes such as Sod1 or Atm causes spermatogenic abnormalities (Takubo et al., 2008; Ishii et al., 2005; Nguyen-Powanda and Robaire, 2021). However, it remains unknown how much GSH, which is produced in germ cells, contributes to spermatogenesis because it is difficult to distinguish the effect of germ cell-derived GSH from the effect of Sertoli cell-derived Gln. Thus, our findings concerning Gln-mediated ROS suppression reveal an important role for GSH production in germ cells for the first time.

Although GSH supplementation could improve cell survival, it was not possible to ensure long-term rescue of GS cells, which suggested that other Gln metabolites also contribute to SSC maintenance. We found that Asn supplementation was sufficient to rescue proliferation in Gln-deprived GS cells. The blood concentration of Asn is much lower than the concentration of Gln (0.05-0.1 mM versus 0.5-0.8 mM); some media, such as Dulbecco's modified Eagle's medium, lack Asn. The Asn concentration required to rescue cell proliferation was substantially higher than the concentration present in the GS cell culture medium (15 mM versus 0.2 mM). In the seminiferous tubules, the concentration of Asn is relatively lower than the concentrations of other amino acids (Hinton, 1990). Although there is no reported spermatogenesis phenotype caused by Asn deficiency, the lack of Asn in vitro compromised GS cell proliferation (Fig. 1A). Asn is often considered a metabolic dead-end. However, Asn impacts global intracellular amino acid levels and may be used in the uptake of extracellular amino acids (Krall et al., 2016). It also stimulates GLUL and promotes MYC accumulation in the absence of Gln (Luo et al., 2018). Asn is synthesized by ASNS; although this enzyme typically exhibits low expression, it can be rapidly induced in response to reduced levels of Gluc, Asn, Leu and Ile, or in response to an imbalanced dietary amino acid composition (Huang et al., 2017). In addition, a recent report showed that exogenous Asn is required for cells to maintain translation when extracellular Gln is limited (Pavlova et al., 2018). Therefore, it is possible that Asn allows the cells to process translation and maintain protein synthesis despite Gln deprivation.

There are several caveats that need to be considered in interpretating our results. First, GS cells used in this study were derived from immature pups. Because spermatogonia proliferate actively in this stage (Shinohara et al., 2001), it is easier to derive GS cells from immature pups than from adult testes. In addition to proliferating activity, selection of SSCs from adult testes requires enrichment of SSCs by magnetic cell sorting or fluorescence-activated cell sorting, because the frequency of SSCs is very low. Therefore, we cannot ignore the potential differences that may exist between SSCs in pup and adult testes. Second, GS cells proliferate more actively in vitro than in vivo because they were exposed to high levels of self-renewal factors. It has been shown that GS and SSCs in vivo have different gene expression patterns. For example, GS cells express KIT, while SSCs in vivo do not express KIT (Morimoto et al., 2009). We currently do not know whether this difference is caused by the enhanced proliferation or by the stress of in vitro culture, such as loss of contact with Sertoli cells. Third, the frequency of SSCs in GS cells is very low (1-2%) (Kanatsu-Shinohara and Shinohara, 2013). Therefore, most of the cultured cells may reflect the phenotype of committed spermatogonia without self-renewal potential. Finally, our study was carried out under normoxia. Because SSCs in vivo are maintained under hypoxia, such a difference in oxygen tension can result in significant changes in the response of cells. For example, we recently found that Cdkn1a- or Nox1-deficient GS cells do not show an apparent phenotype under normoxia (Morimoto et al., 2021). Because oxygen tension has a large impact on cell metabolism, it is likely that GS cells do not fully recapitulate the same phenotype in vitro. Therefore, caution is necessary to extend our results to SSCs in vivo.

The close interaction between germ cells and Sertoli cells has impeded the analysis of SSC nutritional requirements. Although our experiments have several caveats, we exploited the feeder-free GS cell culture system as a model for studying the role of Gln. Our analysis shows a crucial role for Gln in protection against excessive ROS production in cultured SSCs; Gln also promoted and stimulated proliferation by increasing MYC expression. We identified several other amino acids that are required for SSCs in vitro, which may be useful for further analysis and optimization of the metabolic requirements of GS cells. Such studies may promote the development of new technologies for animal transgenesis, while increasing our understanding of the mechanisms of male infertility; such investigations may lead to newer therapeutic strategies.

GS cells used in the current study have been described previously (Kanatsu-Shinohara et al., 2003b, 2004, 2016; Morimoto et al., 2021), using 2- to 10-day-old pup testes. All lines were tested for contamination. Two independent lines of GS cells were analyzed at different passages. Cells were maintained on laminin (20 µg/ml; Invitrogen)-coated plates using rat GDNF and human FGF2 (Peprotech). MEFs were used in some experiments. To determine the amino acid requirements, the cells were plated on laminin-coated plates with Isocove's modified Dulbecco's medium (IMDM) containing all amino acids (Sigma-Aldrich). On the next day, amino acid-deficient IMDM was added to the cell culture after cells had been washed with phosphate-buffered saline (PBS). Amino acid-deficient IMDM was purchased from the Cell Science and Technology Institute (Miyagi, Japan); the medium was supplemented with all other GS cell culture components and individual amino acids. For Gluc or Gln starvation cultures, DMEM/Hams's F12 medium lacking Gluc (Nacalai Tesque) or Gln (Invitrogen) was used. L-DON (50 µM; 17586; Cayman), BPTES (2.5 µM; SMLO601; Sigma-Aldrich) and 968 (20 µM, 17199; Cayman) were added at indicated concentrations. All amino acids were purchased from Nacalai Tesque.

To analyze the spent culture medium, GS cells were cultured for 4 days with all amino acids. The spent medium was analyzed for amino acid content using the Nexera HPLC instrument equipped with a Nexera SPD-M30A photodiode array detector (Shimazu) and a 3.0×100 mm Inertsil ODS-4 column (GL Sciences).

For lentivirus transfection into GS cells, KD vectors for shRNA experiments were purchased from Open Biosystems. All shRNA vectors used in the experiment are listed in Table S1. Scramble shRNA was used as a control (Addgene 1864). The multiplicity of infection (moi) was adjusted to 20.0. Lentivirus infection was carried out, as described previously (Morimoto et al., 2021). AxCANCre (RIKEN BRC) was used to produce Myc DKO GS cells by infecting GS cells with Myc^f/f/Mycn^f/f alleles (moi=2.0). AxCANLacZ was used as a control (RIKEN BRC) (Takehashi et al., 2007).

Total RNA was isolated using TRIzol (Invitrogen). First-strand cDNA was produced using a Verso cDNA synthesis kit for RT-PCR (Thermo Fisher Scientific). For real-time PCR, StepOnePlus real-time PCR system (Applied Biosystems,) and Fast SYBR Green PCR Master Mix (Applied Biosystems) were used in accordance with the manufacturers' protocols. Transcript levels were normalized to the levels of Hprt. PCR conditions were as follows: 95°C for 10 min, followed by 40 cycles of 95°C for 15 s and 60°C for 1 min. Each PCR was performed at least in triplicate. All PCR primers used are listed in Table S2.

To evaluate the colony numbers, we used 4- to 5-week-old C57BL/6 (B6)×DBA/2 F1 (BDF1) mice (Japan SLC). These mice were treated with busulfan (44 mg/kg) to abolish endogenous spermatogenesis, as described previously (Ogawa et al., 1997). For offspring production, we used 4- to 6-week-old WBB6F1-W/W^v mice (Japan SLC), which congenitally lack endogenous spermatogenesis. These mice were treated with anti-CD4 antibody to allow allogeneic spermatogenesis (Kanatsu-Shinohara et al., 2003a). Donor cells were microinjected into the seminiferous tubules through the efferent duct (Ogawa et al., 1997). Each injection filled ∼75-85% of the seminiferous tubules. The Institutional Animal Care and Use Committee of Kyoto University approved all of the animal experimentation protocols.

The levels of ATP, ADP, GSSG and GSH were measured by a competition-based ELISA using an ADP/ATP ratio assay kit (ab65313; Abcam) and a GSSG/GSH quantification kit (G257; Dojindo).

For GS cell staining, the cells were dissociated using trypsin, and single cell suspensions were concentrated on glass slides by centrifugation using a Cytospin 4 unit (Themo Electron). Cells were then fixed in 4% paraformaldehyde for 1 h at room temperature. The slides were incubated in 0.01% Triton-X in PBS for 2 min to achieve permeabilization. To block non-specific antibody binding, the slides were treated with 3% bovine serum albumin (BSA) in PBS for 1 h at room temperature. The slides were then incubated with primary and secondary antibodies diluted in 3% BSA in PBS, overnight and for 1 h, respectively. Slides were then washed with PBS. All antibodies used are listed in Table S3. Hoechst 33342 (Sigma-Aldrich) was used for counterstaining.

GS cells were stained with CellROX Deep Red (5 mM; Thermo Fisher Scientific) for 30 min in GS cell culture medium at 37°C. Mitochondrial-derived ROS were detected with MitoSOX Red (5 mM; Thermo Fisher Scientific). Cells were incubated for 10 min at 37°C. After cells had been washed, they were analyzed using a FACSAria III (BD Biosciences) with FlowJo software (Tree Star).

Samples were separated by SDS-PAGE, transferred to Immobilon-P membranes (Merck Millipore) and incubated with primary antibodies. All antibodies used in the experiments are shown in Table S3. All images were analyzed using ImageJ software.

Apoptotic cells were detected using an In Situ Cell Death Detection Kit: TMR red (Roche Applied Science) in accordance with the manufacturer's protocol. Cells were counterstained with Hoechst 33342 (Sigma-Aldrich).

Testes were collected and refrigerated overnight before microinsemination (Ogonuki et al., 2006). Germ cells were collected by mechanically dissociating the seminiferous tubule segments that exhibited green fluorescence under UV light. These cells were microinjected into BDF1 oocytes using a piezo-micropipette-driving unit (PrimeTech). Embryos at the two-cell stage after 24 h in culture were transferred to the uteri of ICR recipient females.

Significant differences between means for single comparisons were determined by paired two-tailed Student's t-tests. Multiple comparison analyses were carried out using ANOVA followed by Tukey's honestly significant difference test. Graphs were produced using GraphPad Prism software.

We thank J. Yang and S. Watanabe for technical assistance.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.201157