According to our knowledge, final testis size, number of germ cells and sperm output are closely connected to the number of SCs. This emphasizes the significance of SC proliferation. When porcine testis was treated with AR antagonists during postnatal proliferation, an increased number of SCs per testis was detected (52). However, SC numbers were increased in young ram lambs following prenatal exposure to testosterone (53). This is not contradictory, because administration of exogenous testosterone increases negative hypothalamic feedback and may reduce intratesticular testosterone levels, leading to decreased androgen signaling. However, when comparing the number of SCs per testis in Sertoli cell androgen receptor knockout (SCARKO) mice and androgen receptor knockout (ARKO) mice with the control group, researchers found that the SCARKO mice did not show significant differences compared to the controls, whereas the ARKO mice showed a significant decrease in SC number, indicating that in mice AR in SCs has little impact on SCs proliferation (54, 55). A similar phenomenon was found when researchers used a testicular feminization mouse (Tfm) model that lacks functional AR in all cells (56). It is difficult to reconcile these results. We think that the expression of AR in SCs is undetectable or very weak during most of the periods in which these cells proliferate. While AR on other cell types such as peritubular myoid cells, which express AR at relatively high level, may contribute to SC proliferation. Hu et al. recently found that AR could interact directly with PI3K regulatory subunit p85α to activate kinase Akt and that this represses the expression of Cytokeratin 18 (CK18) (a marker of immature and dysfunctional SCs) and induces the expression of Ki67 (a marker of SC proliferation), thereby preserving SC proliferation in mice (57). Since the experiments were conducted on mature mice, additional experiments on prepubertal mice are needed to test this result, although it is true that some immature SCs are present in mature mice. Considering these results as a whole, it seems that role of AR signaling in SC proliferation is species-specific and phase-specific. We suggest that future studies test different species with the aims of understanding and combining their AR expression patterns in SCs, their intratesticular androgen levels and their SC proliferation patterns to provide a better understanding of the role of AR in SC proliferation.

Followed by cessation of SC proliferation comes to SC maturation. Around SC maturation, AR expression level is progressively upregulated until all SCs express AR (58). AR in SCs is key mediator for SC molecular and cytoskeleton maturation (59–61).

In molecular level, several molecules important for maturation are under AR signaling regulation (51) and the level of heterochromatin (62). Among them, Anti Mullerian hormone (AMH) has been widely researched. AMH is crucial for fetal sex differentiation by regressing the Müllerian ducts in male (13, 63). Secreted by immature SCs, AMH has become a biomarker for testicular function in males during the prepubertal period which is tightly regulated by AR (64). During the postnatal period, AMH level can be maintained by androgen epi-testosterone through non-classical AR signaling pathway when intracellular AR level is very low (65). Other transcription factors, such as sex determining region Y-box 9 (SOX9), steroidogenic factor 1 (SF1) can bind to their own response elements to promote Amh transcription (66). Around the onset of puberty, upregulation of AR expression downregulates AMH expression by blocking SF1 binding to its response elements or interacting with SF1 response elements to prevent SF1from exerting a stimulatory effect on Amh transcription (67). An in vitro study using TM4 Sertoli cells and C3H10T1/2 cell line demonstrated that AR could downregulate AMH expression by repressing SOX9, consistent with the findings in the azoospermia patients and mice in vivo (68).

Besides molecular level, AR signaling is necessary for SC cytoskeleton maturation. In SCARKO mice, relatively normal SCs nuclei were displaced from the basal basement and dispersed throughout the seminiferous tubules, probably resulting from the disrupted position of vimentin and the presence of thicker basal lamina, both of which participate in the linkage between nuclei and the cell membrane (69). Another group observed that SC nuclei formed one or two layered rings located in the center of tubule in mice in which AR function had been ablated (70). Suppressed expression of connexin 43(Cx43) in mice resulted in an intermediate state of SCs between proliferation and maturation, along with weaker AR immunostaining (71). The impaired SCs maturation observed in this model probably resulted from disrupted AR function. AR can also regulate microtubular composition and structure in SC so as to yield proper SC shape via Class III β tubulin (TUBB3) in mice and rats through classical signaling pathway (72).

Premature expression of ARs in postnatal mouse SCs resulted in fewer SCs, although GCs development accelerated in a gain-of-function mouse model (73). When the AR transgene was injected into a prenatal mouse model, precocious maturation of SCs occurred and limited the window for SC proliferation. The production of fewer SCs can lead to lower-than-normal sperm output (74).

Precise regulation of Smad2/Smad3 is important both for SC proliferation and SC maturation (72). Type II A activin promotes immature testis growth along with timely appropriate AR expression via Smad3. The expression of Smad3 is then downregulated, and this is associated with cessation of SC proliferation and progression of the cells to terminal differentiation. AR activation can induce Smad2 transcription, which will further promote SC maturation through shifting from using Smad3 only to using both Smad2 and Smad3 under activin A activation. In Smad3^-/- mice, delayed maturation of SCs was accompanied by reduced levels of AR and Smad2 (75).

In recent years, microRNAs (miRNAs) have been found to regulate SC proliferation and maturation. miR-124a binds to the 3’-UTR of Ar mRNA and downregulate AR expression, thereby inhibiting the proliferation of immature porcine SCs, while the coactivator really interesting new gene (RING) finger protein 4 (RNF4) binds to AR and promotes porcine SCs proliferation by elevating Proliferation cell nuclear antigen (a marker of cell proliferation) mRNA transcript levels (76). miR-762 promotes porcine SC maturation by inhibiting binding of RNF4 to AR (77). Moreover, miR-130a has been shown to inhibit AR expression, although its impact on SC proliferation and maturation remains to be elucidated (78). Additionally, the relationship of miR-133b (79) and miR-638 (80), which regulate SC proliferation and maturation, to AR remains to be investigated ( Figure 2 ).

Importantly, the relationship between SC maturation and spermatogenesis is not well understood. We hypothesize that AR-regulated lipid metabolism in mature SCs can provide energy for spermatogenesis (81); for example, AR signaling in SCs induces the expression of ubiquitin-conjugating enzyme E2B (Ube2b) which regulates the expression of genes related to glycosylphosphatidylinositol (GPI)-anchor biosynthesis and oxidative phosphorylation (82). So does AR-regulated glucose uptake. The non-classical pathway participates in the regulation of glucose uptake via the cytochrome C oxidase (83). Also, BTB formed between SCs can facilitate normal spermatogenesis, as we will discuss later.

Wnt5a, which refers to Wingless-type MMTV Integration Site Family Member 5A, is secreted by SCs in mice. Several studies found that Wnt5a expression is regulated by androgen and that it promotes SSCs self-renewal. Both in vitro and in vivo studies confirmed that Wnt5a can increase SSC self-renewal division after androgen blockage in mice SCs (84). In addition, Crespo et al. found that in zebrafish one androgen, 11-ketotestosterone (11-KT), indirectly inhibits SSC self-renewal by inhibiting prostaglandin E2 (PGE2) in SCs. Otherwise, PGE2 promotes SSC self-renewal by upregulating Wnt5a expression (85). However, more investigations should focus on the existence of AREs within the promoter of Wnt5a to confirm the exact signaling pathway in SCs through which androgen exerts its effects.

Promyelocytic leukemia zinc finger (Plzf), a key transcription suppressor gene, has been characterized as a marker for undifferentiated SSCs in rodents (86) and primates (87). Plzf is important for SSC maintenance (45). Recently, in vivo and in vitro studies found that after activation of AR signaling in SCs, AR homodimer binds to the ARE region of GATA binding protein 2 (GATA-2) to inhibit its expression; this can further decrease WT1 transcription factor (Wt1) expression since Wt1 expression requires binding of GATA-2to its promoter. Wt1 binds to β1-integrin and increases its expression in SCs; β1-integrin will interact with unknown molecules on the surface of SSCs, to induce Plzf expression. As a result, activation of classical AR signaling pathway can promote SSC differentiation by downregulating Plzf expression. We propose that the exact molecules that interact with β1-integrin on SSC surface should be identified to determine a more concise relationship between SCs and SSCs (88).

In zebrafish, Insulin like growth factor 3 (Igf3) mRNA transcript level was found to be upregulated by androgen; this can promote SSC differentiation through non-classical signaling pathway (89). Reproductive homeobox 5 gene (Rhox5), which is directly regulated by classical AR pathway, also contributes to spermatogonia differentiation (90). Besides, in vitro study found that testosterone can increase Glial-derived neurotrophic factor (GDNF) secretion by Sertoli cell (91). Glial-derived neurotrophic factor secreted will promote Gdnf family receptor α1 and proto-oncogene Rearranged expression during Transfection expression for SSC self-renewal and maintenance (92).

The fact that AR expression peaks at stages VI-VIII of meiosis, coinciding with the preleptotene and leptotene stages, suggests that the AR might mediate the preparation of SCs for chromosomal synapsis and recombination. Male germ cells can enter prophase I of meiosis normally and undergo normal DSBs formation in the absence of AR signaling (5). However, the problems happened on DSBs repair and chromosome synapsis. In SCARKO mice, chromosome synapsis was incomplete, as shown by the fact that Chen et al. detected a significantly larger number of univalent chromosomes in spermatocytes than in the control group. Additionally, RAD51 and DMC1 which are required for resolution of DSBs and homologous recombination, along with TEX15, BRCA1, BRCA2 and PALB2 which load RAD51 and DMC1 onto DSBs sites, showed decreased protein expression in SCARKO mice. Furthermore, activation of AR can negatively regulate Epidermal growth factors (EGFs), including Egf, betacellulin (Btc) and neuregulin 1 (Nrg1). These EGFs bind to their corresponding receptors (EGFR, ERBB4) on the surface of spermatocytes and stimulate the accumulation of above homologous recombination factors for appropriate prophase preparation (94). At the synaptonemal complex, nuclear autoantigenic sperm protein (NASP) inhibits formation of CDC2/cyclinB1 complex via HSP70-2 which are important for G2/M transition. NASP showed elevated expression after androgen manipulation. Other genes that show changes in transcript after androgen manipulation, such as Mapre1, Ruvbl1, Tuba1c, Tuba3a and Tubb2c, are related to spindle dynamics and the formation of dynactin complexes (99). Although fewer spermatocytes may gain competence to enter metaphase I, they cannot complete meiosis due to the lack of normal spindle dynamics, and this may account for the disruption in post-meiotic differentiation. More proteins account for chromosome synapsis and dynamics that is under AR regulation deserve to be confirmed to establish the AR downstream signaling network by taking advantage of the development of proteomics.

GC apoptosis, especially apoptosis of spermatocytes, is another cause of meiotic arrest. This is not difficult to understand, since spermatocytes are cells that undergo meiosis. Through single cell transcript analysis technology, genes related to apoptosis have been identified. Aldh1a1, Igfbps, and mitochondrial membrane and oxidative phosphorylation transcripts were upregulated in SCARKO mice (98). Another group found that androgen deprivation led to a decreased expression of Aldh2, Prdx6 and Gstm5, which encode protectors of oxidative stress (99). Other genes related to androgen-dependent apoptosis, such as Diablo, Cbl, Hsf1 showed increased expression in mice after androgen deprivation or knocking out AR in SCs (97, 100). Hu et al. showed that Rhox5 can repress expression of Unc-5 Netrin Receptor C(UNC5C) which can promote spermatocyte apoptosis in vitro (101). Since Rhox5 is regulated by classical signaling pathway (90, 102, 103), it needs to be determined whether AR signaling represses UNC5C expression and thereby protects spermatocytes from apoptosis. Recently, Marina et al. found that ubiquitin carboxyl-terminal hydrolase L1(UCHL1) is negatively regulated by androgen. Decreased level of androgen leads to increased level of UCHL1 in spermatocytes, which will lead to deubiquitination of the pro-apoptotic factor p53. Deubiquitination of p53 triggers germ cell death (104). In addition, phagocytic clearance of apoptotic germ cells by SCs is vital for meiosis because phagocytosis of dying GCs provides lipid sources for SCs which will provide ATP for GCs development. Lower androgen increases the expression of miR-471-5p, resulting in inhibition of Dock1, Tecpr1, Atg12 and Becn1. This leads to impaired LC3-associated phagocytosis. Besides, inhibition of Dock1, Rac1-GTPase in this model leads to disrupted engulfment of SCs (11). Both of these inhibitions impair clearance of apoptotic GCs by SCs. Another protein Elmo1, which is upregulated by classical AR signaling in vivo, participates in apoptotic germ cells clearance by SCs and functions downstream of Dock1 and Rac1-GTPase (73, 105).

Many of the transcription factors and regulators related to AR signaling that have been found are not mentioned here. These factors can participate in the regulation of meiosis. For additional information on these factors, please see Table 1 . For additional genes that show changes in expression after androgen manipulation, please see the supplied references (98, 99, 123).

Based on the evidences presented above, we propose the reasons for meiotic arrest when AR signaling in SC is impaired. Chromosomal dysfunction or breakage can induce oxidative stress that cannot be repressed because of the decreased level of oxidative stress protectors. This leads to apoptosis of some spermatocytes. The presence of increased number of apoptotic cells, the failure to clear apoptotic cells combined with DNA damage trigger the inner meiotic checkpoint and cause meiosis to halt in prophase I. Future studies might focus on the roles of genes related to AR signaling that have been mined through mRNA transcript analysis in meiosis and attempt to determine whether they play permissive roles or instructive roles for meiosis.

Classical signaling pathway regulates proper tight junction dynamic. In mouse models with mutations in exon1 of AR, increased permeability of BTB was observed (124, 127). Although the BTB appears to be present in this model under microscopic examination, it is incomplete as we found biotin enter into adluminal area with the help of biotin tracer (127). Reversibly, BTB were strengthened after testosterone treatment of an SC culture system with disrupted AR function (128). To further investigate the molecular basis of these effects, microarray analysis was used to identify the genes that may be involved. The mRNA transcript level of tight junction protein including Claudin 3, Claudin 11, Claudin 13 were all downregulated while Occludin and Claudin 26 showed elevated transcript levels after ablation of exon1 of AR in mice (129). Among them, Claudin 3 which is not essential for fertility, is expressed in the newly formed BTB after preleptotene spermatocytes migrate past the old BTB (127). Claudin 13 shows a similar expression pattern to Claudin 3. CHIP-seq assays found putative AREs in the promoter regions of Claudin 13 and tight junction protein 2 isoform 3, indicating a potential classical signaling regulation pattern. In contrast to Claudin 3, Claudin 11 preserves its expression throughout the seminiferous cycle. Besides, membrane-associated GUK family of proteins including tight junction protein 1, tight junction protein 2 isoform 1, tight junction protein 2 isoform 2, and tight junction protein 2 isoform 3, can function as a bridge that links tight junction proteins to the cytoskeleton like actin filaments. Their mRNA transcripts were found to be downregulated in mouse SCs after ablation of AR (129–131). What’s more, it was found that tight junction proteins can enter the nuclei of SCs and regulate their own transcription (132), indicating possible tight junction proteins compensatory effects in Cldn3^-/- mutant mice.

Apart from its directly regulation of TJ proteins, non-classical signaling can regulate TJ proteins via tissue type plasminogen activator (tPA) during stages VII-VIII of spermatogenesis when BTB undergoes restructuring (133). We also know that tPA is involved in BTB degradation (134), indicating a potential role of tPA in BTB dynamics. Moreover, Dietze et al. and Bulldan et al. found that Claudin 1, Claudin 5 and Claudin 11 protein levels were upregulated by CREB protein (49, 135, 136). These results further support the idea that non-classical androgen signaling can maintain BTB integrity by regulating tight junction dynamics.

Besides, in heat-treated monkey testes, AR regulates reversible changes in BTB integrity via Par complex (Par6-Par3, Par6-aPKC, and Par6-Cdc42) targeting TJ proteins (ZO-1 and occludin) and basal ES proteins (N-cadherin, E-cadherin, ɑ-catenin, β-catenin, and γ-catenin) (137).

Based on the results described above, we hypothesize that although the transition of preleptotene spermatocytes into adluminal compartment is not blocked when AR signaling in SC is prevented, the amount of claudin available for remodeling BTB (Claudin 3, Claudin 11, Claudin 13) is insufficient under these conditions. Besides, the downregulated tight junction proteins are not sufficient to enter the nucleus and increase their own transcription, let alone to recruit more claudins to the BTB. While Occludin and Claudin26 show relatively upregulated mRNA transcript levels, this is not enough to compensate for the loss of other tight junction proteins. Thus, humoral immunity against self-antigens is mounted, resulting in disruption of normal spermatogenesis.

Connexin 43 (Cx43), one gap junction protein, also under AR regulation as shown by its reduced mRNA transcript levels and protein expression levels in adult pig testes treated with flutamide, an anti-androgen (138). Interestingly, we also found weak AR immunostaining and partial disruption of AR signaling in SCs in Sertoli cell specific knock-out of connexin 43 mice (71). Xia et al. found that androgen can regulate the expression of Wt1 indirectly by inhibiting GATA-2expression. Androgen inhibits GATA-2 expression through the classical pathway. Wt1 can bind to the promoter of Cx43 and inhibit its expression, indicating that androgen can promote Cx43 expression by downregulating GATA-2 and Wt1 (139). What’s more, ouabain binding to its receptor ATP1A1can upregulate Cx43 expression via non-classical signaling pathway in vivo (140). The relationship between Cx43 and AR deserves further investigation. It remains to be determined whether reversing one of them (Cx43 or AR) can restore the other to normal level in human and correct impaired spermatogenesis, though it is possible that treating hpg mice with DHT can restore the expression and localization of Cx43 to the BTB (141).

The endocytosis of old BTB proteins for renewal is also important for BTB integrity (142). Testosterone can promote the endocytosed occludin to the BTB location through recycling pathway (124). Other cells such as Madin-Darby canine kidney cells and epithelial cells can internalize occluding and claudin via clathrin- and caveolin-mediated pathways (143, 144). Coincidentally, Su et al. found that testosterone increases levels of clathrin and caveolin-1 in SCs (145), this suggests that AR signaling may participate in tight junction protein endocytosis for renewal via the clathrin and caveolin-1 pathways. This possibility needs further investigation.

Since AR signaling is important for BTB integrity, what will happen if AR is overexpressed under normal conditions? The influence of a tighter BTB on spermatogenesis remains unknown. We suggest an in vitro AR overexpression experiment to answer this question. Recently, in muscle cells, it was found that AR can participate in IGF-1/IGF-1R-PI3K/Akt-mTOR pathway (146). We also know that mTOR signaling can regulate BTB dynamic by balancing the levels of F-actin binding proteins epidermal growth factor receptor pathway substrate 8 and actin related protein 3 (147). Therefore, we suggest investigation into crosstalk between AR and mTOR pathway and its downstream molecules related to actin-binding proteins in germ cell lines.

The molecular mechanism through which Sertoli cell - Spermatid adhesion occurs has been elucidated. Based on experiments in which adenovirus constructs were used to express different AR mutants that impair the classical or nonclassical pathways and on other experiments in which inhibitors of ERK kinase and Src kinase were added to an SC-GC coculture system, it is proposed by study that non-classical AR signaling pathway contributes dominantly to Sertoli cell-Spermatid adhesion, while activation of the classical AR signaling pathway is not sufficient to permit Spermatids binding to SCs (151). Testosterone can promote the attachment of spermatids to SCs in Sertoli cell -Spermatid co-culture system by activation ERK and Src, which belong to the non-classical signaling pathway (152). Terada et al. found that testin, which can be induced by non-classical signaling pathway, is important for Sertoli cell - Spermatid adhesion. Overexpression of testin can block spermatid differentiation, probably due to excessive adhesion to SCs (113). However, an in vivo study adopting TE implants brought about not only the loss of spermatids but also the activation of ERK before sperm release, along with breakage association of N-cadherin and β-catenin indicating disruption of ES (153). This result seems contradictory to in vitro study. One possible reason maybe that the formation of ES requires ERK activation only at the beginning of formation rather than during the entire process.

Round spermatids develop into elongated sperm before final releasing.

Gonadotropin Regulated Testicular Helicase (GRTH/DDX25) is significant in this process, mainly by participating on the nuclear export and transport of specific mRNAs as well as the structural integrity of Chromatoid Bodies of round spermatids (154). As an RNA helicase and only DEAD-box family member regulated by androgen in spermatids, GRTH/DDX25 transcription in spermatid is regulated by AR signaling pathway in Sertoli cells through a paracrine fashion. Germ cell nuclear factor (GCNF) in spermatids can promote GRTH/DDX25 transcription and expression in response to androgen regulation (155–158). However, the signal relayed to spermatids remains to be further elucidated.

Before releasing sperm, SCs endocytose the residual body from maturing sperm. Kumar et al. found that AR dimerization can recruit corepressor nuclear receptor corepressor 1 (NCoR1) to the ARE of phosphatidylinositol binding clathrin assembly protein (Picalm) and early endosome antigen 1 (Eea1) and thereby inhibit their expression. Coactivator steroid receptor coactivator-1 (Src1) can be recruited to ARE of syntaxin 5 (Stx5a) and induce its expression (112, 159, 160). Since the products of these 3 genes are located around tubule bulbar complexes and participate in endocytosis and intracellular transport, it is possible that they are involved in residual body absorption.

Activation of Src is important for sperm release (161, 162). The injection of Src inhibitor into rat testis can block the release of elongating spermatids, while injection of an ERK inhibitor has little effect on sperm release (152, 163). Additionally, O’Donnell et al. pointed out that the disengagement complex, including α6β1 integrin which associates with Src, and tyrosine-phosphorylated focal adhesion kinase (FAK), remains associated with sperm that fail to be released. Both Src and FAK can be phosphorylated on activation of non-classical androgen signaling (149). Further study confirmed that activated non-classical signaling target Src is near the head of sperm to be released (164).

Also, sperm motility is significant during sperm release. Located in elongated spermatids, ATPase Ca⁺⁺ transporting plasma membrane 4 (PMCA4) is critical for sperm motility (165). Recently, it was found that PMCA4 is positively regulated by AR signaling in SCs both in vitro and in vivo (121).

In a summary, non-classical AR signaling promotes Sertoli cell – Spermatid adhesion by activating both ERK and Src while promoting sperm release by activating Src. Considering the above study, future studies might focus on achieving a deeper understanding of the different effects of Src to immature spermatid and mature spermatid as well as the impact of both ERK and Src to other kinds of proteins in apical ES.