Retinoblastoma protein represses E2F3 to maintain Sertoli cell quiescence in mouse testis

ABSTRACT Maintenance of the differentiated state and cell cycle exit in adult Sertoli cells depends on tumor suppressor retinoblastoma protein (RB, also known as RB1). We have previously shown that RB interacts with transcription factor E2F3 in the mouse testis. Here, we investigated how E2f3 contributes to adult Sertoli cell proliferation in a mouse model of Sertoli cell-specific knockout of Rb by crossing these mice with an E2f3 knockout mouse line. In the presence of intact RB, E2f3 was redundant in Sertoli cells. However, in the absence of RB, E2f3 is a key driver for cell cycle re-entry and loss of function in adult Sertoli cells. Knockout of E2f3 in Sertoli cells rescued the breakdown of Sertoli cell function associated with Rb loss, prevented proliferation of adult Sertoli cells and restored fertility of the mice. In summary, our results show that RB-mediated repression of E2F3 is critical for the maintenance of cell cycle exit and terminal differentiation in adult mouse Sertoli cells. Summary: The quiescent and differentiated state of Sertoli cells during spermatogenesis depends on negative regulation of E2F3 by retinoblastoma protein, which maintains cell cycle exit and terminal differentiation.


INTRODUCTION
Sertoli cells are highly differentiated somatic cells that control and support sperm development in the adult testis. As the fetus develops, Sertoli cells are the first cell type to undergo male sex differentiation as a result of SRY gene activation (Capel, 2017;Kashimada and Koopman, 2010;Mirza et al., 2006) and Sertoli cell proliferation drives testicular organogenesis and expansion of the testis cords (Ungewitter and Yao, 2013). Sertoli cells gradually cease proliferation and terminally differentiate during puberty (Vergouwen et al., 1991). Adequate Sertoli cell proliferation is important for fertility; Sertoli cell number determines sperm output in the adult, as a single Sertoli cell can support only a finite number of germ cells (Ranganathan et al., 1994;Rebourcet et al., 2017).
Sertoli cells were long thought to permanently exit the cell cycle as they terminally differentiate in the mouse (Tarulli et al., 2012). However, based on data from different species, it has been suggested that Sertoli cells are in a state of continuous cell cycle repression rather than terminally differentiated (Tarulli et al., 2012). First, Sertoli cell proliferation is reactivated during the reproductive season in seasonal breeders such as the Djungarian hamster, where Sertoli cell numbers fluctuate depending on the photoperiod and associated gonadotropin levels (Tarulli et al., 2006). Second, when adult mouse or human Sertoli cells are removed from their physiological environment to in vitro culture conditions, some of them can re-enter the cell cycle (Ahmed et al., 2009). Sertoli cells are maintained in a non-proliferative state in adulthood by the retinoblastoma protein (RB, also known as RB1) tumor suppressor protein (Nalam et al., 2009;Rotgers et al., 2014). Loss of Rb in Sertoli cells (SC-RbKO mouse line) results in reactivation of Sertoli cell proliferation in adult mice, which ultimately leads to spermatogenic failure as a result of Sertoli cell dysfunction (Nalam et al., 2009;Rotgers et al., 2014). We showed that shRNA-mediated in vivo knock-down of transcription factor E2F3 could ameliorate the testicular phenotype of SC-RbKO mice (Rotgers et al., 2014).
A key function of RB is to regulate the E2F transcription factors, which are crucial players in the control of cell cycle, apoptosis and differentiation (reviewed in Chen et al., 2009). The classical view of RB-E2F-mediated cell cycle control is that upon mitogenic signaling, cyclin-dependent kinases (CDKs) inactivate RB by phosphorylation, which results in the release of E2F and activation of genes required for cell cycle progression. The E2F protein family has eight members, traditionally classified as transcriptional activators (E2F1-E2F3a) and repressors (E2F3b-E2F8) . These classifications are not as clear-cut in vivo; the activator E2Fs can switch to repressors during differentiation through forming complexes with RB family proteins (Chong et al., 2009). All the E2Fs share the same consensus DNA-binding sequence (Zheng et al., 1999) and there is considerable functional redundancy between family members (Tsai et al., 2008).
Adult mouse Sertoli cells express two E2F family members, E2F3 and E2F4 (El-Darwish et al., 2006); however, two E2f3 isoforms are transcribed from the same gene locus through the use of an alternative promoter He et al., 2000;Leone et al., 2000). E2F3a is classified as a transcriptional activator, while the shorter E2F3b isoform is a transcriptional repressor . E2F3a and E2F3b can have opposing roles in regulating cell fate in vivo, but their function is dependent on the cellular context. In neural progenitor cells E2F3a promotes differentiation, while E2F3b promotes progenitor fate and supports the expansion of the neural precursor colony (Julian et al., 2013). However, in myotubes E2F3b activates the expression of genes required for differentiation (Asp et al., 2009). Furthermore, increased E2f3 activity in the absence of Rb leads to defects in both proliferation and apoptosis in lens and neuronal cell lineages (Chong et al., 2009;Ziebold et al., 2001). Thus, the function of E2F3 is highly dependent on the cell lineage in question and their RB expression status.
We investigated the contribution of E2F3 RB-mediated cell cycle control in Sertoli cells by interbreeding the SC-RbKO (Amh-cre;Rb flox/flox ) strain with the E2f3 flox/flox strain to achieve a simultaneous conditional knockout of Rb and E2f3 in Sertoli cells (Lécureuil et al., 2002;Marino et al., 2000;Wu et al., 2001). In adult males, a complete loss of E2f3 rescued the testicular dysfunction caused by Rb deficiency. However, in the absence of Rb, even one E2F3 has been shown to be expressed in adult mouse Sertoli cells and spermatogonia (El-Darwish et al., 2006), but the expression pattern of the two isoforms, E2F3a and E2F3b, in the postnatal testis was unknown.
We analyzed both E2F3 protein and mRNA expression using immunohistochemistry and RNA in situ hybridization on postnatal day (P)6, P10, P20 and P40 testes to assess the expression pattern during the major events in postnatal Sertoli cell development (Fig. 1). On P6, Sertoli cells are actively proliferating, on P10 they gradually begin to exit the cell cycle, by P20 Sertoli cell proliferation has ceased and in P40 mouse testis full spermatogenesis is reached. Surprisingly, E2F3a protein was expressed in only a subpopulation of Sertoli cells on P6 (Fig. 1A, black arrowheads), but the proportion of E2F3a-positive Sertoli cells increased gradually as the animals matured. By P20 all the Sertoli cells were positive for E2F3a (Fig. 1A). To analyze whether E2F3a expression in immature Sertoli cells was associated with a non-proliferative state as a sign of commitment to terminal differentiation, consecutive sections of P6 and P10 testes were stained for E2F3a and KI67 (Fig. 1A, black double arrows and KI67 in inset). There was no correlation between KI67 and E2F3a staining, as the E2F3a-positive Sertoli cells were seen to be both KI67-positive and -negative (Fig. 1A, black and red double arrows, respectively). As expected, the signal using an antibody against both E2F3 isoforms (E2F3ab) followed a similar pattern in Sertoli cells as the E2F3a-specific antibody (Fig. 1B, red arrowheads show E2F3ab-negative cells, black arrowheads show E2F3ab-positive cells). However, in spermatogonia the E2F3a isoform was not detected at either P6 or P10 (Fig. 1A, yellow arrowheads), and only the antibody against both isoforms (E2F3ab) gave a positive signal (Fig. 1B, orange arrowheads). The expression pattern of E2f3a mRNA, and of both E2f3a or E2f3b mRNA detected by a non-isoform-specific probe (E2f3ab mRNA hereafter) followed that of protein expression (Fig. 1C). In the testis, both E2f3a and E2f3ab mRNA could be detected in Sertoli cells (Fig. 1C). In general, the E2f3 mRNA signal was relatively weak throughout postnatal development, but the positive control probes yielded a strong signal especially in the adult testis (Fig. 1C, second and third rows). These data suggest that E2F3a expression is associated with Sertoli cell differentiation irrespective of the proliferation status of the Sertoli cells, while E2F3b is the primary isoform expressed in spermatogonia.

E2F3 is redundant in mouse Sertoli cells
To analyze the effect of E2f3 loss on Sertoli cells, Amh-cre and E2f3 flox/flox strains were crossbred (SC-E2f3 −/− ). Surprisingly, E2f3 appeared to be redundant in mouse Sertoli cells. The SC-E2f3 −/− mice were viable and in good health at the ages of two and five months. The relative testis weights of the 2-month-old SC-E2f3 −/− males were comparable to the controls, but at the age of five months there was a slight decrease in relative testis weight (Fig. S1A). The SC-E2f3 −/− males sired healthy pups and there were no significant differences in the number of pups per litter as compared with the control males (Fig. S1B). Testicular histology appeared normal at both ages (Fig. S1C) and no changes were observed in the testicular cell populations on flow cytometric analysis (Fig. S1D). TUNEL-positive germ cells were quantified from seminiferous tubule crosssections in stages VII-VIII to analyze germ cell apoptosis. The number of apoptotic cells was similar in control and SC-E2f3 −/− seminiferous tubules (Fig. S1E). Sertoli cells express not only E2F3, but also E2F4 (El-Darwish et al., 2006) and there could be functional redundancy between these factors leaving E2f3 dispensable for Sertoli cell development and function.
Knockout of E2f3 in Sertoli cells rescues the spermatogenic failure associated with Rb loss Loss of Rb in mouse Sertoli cells (SC-RbKO) results in infertility due to Sertoli cell failure in adult life (Nalam et al., 2009;Rotgers et al., 2014). We have previously shown that shRNA-mediated knockdown of E2f3 in vivo leads to a partial rescue of the disruptive testicular phenotype of SC-RbKO mice (Rotgers et al., 2014). Here, we cross-bred the SC-RbKO (Amh-cre+; Rb flox/flox ) animals with E2f3 flox/flox mice to study whether a complete loss of E2f3 could fully rescue the SC-RbKO phenotype. In E2f3 flox/flox mice, exon 3 of the E2f3 gene is flanked by loxP sites, and the sequence coding for the DNA-binding domain is excised by Cre recombinase and direct transcriptional control of gene expression by E2f3 is abolished (Wu et al., 2001). In the Amh-cre mouse strain, constitutive expression of Cre begins on embryonic day (E)14.5 in Sertoli cells (Lécureuil et al., 2002).
The SC-Rb −/− E2f3 +/+ (Amh-cre+; Rb flox/flox ; E2f3 wt/wt ) animals had a similar testicular phenotype to the previously described SC-RbKO mice (Fig. 2, Table 1) (Rotgers et al., 2014), despite a different hybrid genetic background. The primary time-point chosen for this study was the age of 2.5 months, because at this age the SC-RbKO males showed a clear testicular phenotype, which allowed us to evaluate whether the loss of E2f3 could rescue testicular defects (Rotgers et al., 2014).
Interestingly, loss of only one E2f3 allele was not sufficient to fully overcome the effect of Rb loss. The SC-Rb −/− E2f3 +/− testes had some seminiferous tubules with a severe disruption of spermatogenesis ( Fig. 2B, asterisks). The disruption became widespread and the testes resembled those of SC-Rb −/− E2f3 +/+ males as the SC-Rb −/− E2f3 +/− males approached the age of 4 months (data not shown). The morphological changes associated with the SC-Rb −/− E2f3 +/+ phenotype were similar to those previously shown and consisted of complete disruption of the seminiferous tubule architecture with massive germ cell loss and Leydig cell hyperplasia ( Fig. 2B) (Nalam et al., 2009;Rotgers et al., 2014).
To assess whether the testicular phenotype had induced compensatory changes in the hypothalamic-pituitary axis, serum levels of luteinizing hormone (LH) and follicle-stimulating hormone (FSH) were measured in serum from 2.5-month-old animals. Despite the severe testicular phenotype in SC-Rb −/− E2f3 +/+ males, no significant alterations were observed in either FSH or LH serum levels in the experimental groups as compared with controls. Androgen levels were analyzed from testicular homogenates using radio-immunoassay and testosterone content was normalized to the weight of the tissue (Table 1). Intra-testicular testosterone content was significantly higher in SC-Rb −/− E2f3 +/+ testes as compared with SC-Rb −/− E2f3 +/− and a similar trend was observed when comparing SC-Rb −/− E2f3 +/+ to the control and SC-Rb −/− E2f3 −/− testes, although this was not significant because of high variation between individuals. Since there is excessive loss of germ cells in the SC-Rb −/− E2f3 +/+ testis, the relative abundance of Leydig cells per mg of tissue is higher, which likely explains the increased testicular testosterone level when compared with control testes that have full spermatogenesis. However, we cannot exclude the possibility of Leydig cell hyperplasia resulting from impaired paracrine signaling in the testis. Insets show proliferation marker KI67 on a consecutive section to identify the same Sertoli cells as in the E2F3a staining. Black double arrows, Sertoli cells stained for both E2F3a and KI67; red double arrows, E2F3a+ KI67− Sertoli cells. Both proliferative KI67+ and non-proliferative KI67− Sertoli cells expressed E2F3a. Scale bars: 25 µm (B) Antibody detecting both E2F3a and E2F3b isoforms (E2F3ab) showed positive staining in germ cells (orange arrowheads) and some Sertoli cells (black arrowheads, E2F3ab+; red arrowheads, E2F3ab−) on P6 and P10. Scale bars: 25 µm. (C) RNA in situ hybridization using probes specifically against mRNA encoding the E2f3a isoform, or a probe detecting both E2f3a and E2f3b isoforms (E2f3ab), showed mRNA expression in the same cell types where protein expression was detected. Immunohistochemical detection of vimentin was coupled to the RNA in situ hybridization to aid Sertoli cell identification in the P20 and P40 testes. Black, pan-E2F3; red, E2F3a; yellow, vimentin (VIM); blue, Mayer's hematoxylin. Arrowheads, E2F3a/E2F3ab-positive Sertoli cell. Lower panels show negative and positive controls for RNA in situ hybridization. PC, positive control probes (targeting common housekeeping gene); NC, negative control probes. Scale bars: 10 µm.

E2F3 induced cell cycle re-entry in adult Sertoli cells in the absence of RB
Loss of Rb impairs cell cycle control in adult Sertoli cells and causes an increased Sertoli cell proliferation rate in adult SC-RbKO testes (Nalam et al., 2009;Rotgers et al., 2014). Sertoli cell proliferation rate in adult testes was analyzed by quantifying KI67-and SOX9-positive Sertoli cells. As expected, loss of Rb resulted in a cell cycle re-entry of adult Sertoli cells, shown by an increased frequency of KI67-positive Sertoli cells in SC-Rb −/− E2f3 +/+ testes (Fig. 3A,B). KI67-positive Sertoli cells were also observed in SC-Rb −/− E2f3 +/− testes, suggesting that even one copy of E2f3a was able to drive Sertoli cell cycle re-entry in the absence of Rb. In control testes, no proliferating Sertoli cells were detected (Fig. 3A,B). Lastly, no KI67-positive Sertoli cells were observed in the testes of 2.5-monthold SC-Rb −/− E2f3 −/− mice, showing that complete knockout of E2f3 blocked ectopic Sertoli cell proliferation caused by Rb loss (Fig. 3A, B). In contrast to adult Sertoli cells, adult female supporting cells, granulosa cells, undergo active proliferation during folliculogenesis.
The cell cycle is tightly regulated, and different layers of cell cycle control can act redundantly when balance of the system is disrupted. In the Rb −/− fetal testis, loss of Rb is compensated by an induction of cyclin-dependent kinase inhibitor 1B (CDKN1B or p27 Kip ) to achieve mitotic arrest of germ cells (Spiller et al., 2010). p27 can be inhibited by proteasomal degradation, and is targeted for proteasomal degradation by S-phase kinase-associated protein 2 (SKP2) (Carrano et al., 1999;Polyak et al., 1994). p27 did not show significant change in its expression, although it did tend to be higher in SC-Rb −/− E2f3 +/+ testes (Fig. 3C). Interestingly, expression of p27-inhibiting Skp2 was significantly downregulated in the SC-Rb −/− E2f3 +/+ testes and normalized after E2f3 deletion (Fig. 3D), suggesting that p27-dependent compensatory mechanisms operates not only in fetal germ cells, but also in adult The reduced relative testis weight (testis weight/body weight) (mg/g) observed in SC-Rb −/− E2f3 +/+ mice is restored to the same level as control in SC-Rb −/− E2f3 +/− and SC-Rb −/− E2f3 −/− mice. ***P<0.001. Statistical significance was calculated by one-way ANOVA followed by Tukey's test, n=8/group. Bars represent mean±s.e.m. (B) Representative PAS-stained testis histological sections. SC-Rb −/− E2f3 +/+ testes are severely atrophied at the age of 2.5 months. The histology of the SC-Rb −/− E2f3 −/− testes is comparable to control, while SC-Rb −/− E2f3 +/− testes show focal seminiferous tubule atrophy. Squares in upper panels outline magnified area in lower panels. Asterisks, atrophied tubules; black arrowheads, Sertoli cells; red arrowheads, mitotic cell. Scale bar: 250 µm in upper panels, 100 µm in the lower panels.
Sertoli cells to ensure cell cycle exit in the absence of Rb. Despite reactivated Sertoli cell proliferation (Fig. 3A,B), the expression level of Sertoli cell-associated transcripts (Wt1, Sox9, Gata1, Gdnf ) remained unaffected (Fig. 3E), which is in agreement with the absence of expansion of the Sertoli cell population and tumor formation. The expression of markers of immature Sertoli cells (Amh and Thra) was not significantly increased in SC-Rb −/− E2f3 +/+ testes, suggesting that, despite resuming proliferation, Rb-deficient Sertoli cells are capable of maintaining a transcriptional program typical of mature Sertoli cells (Fig. S3).

Loss of E2F3 improved the supportive capacity of RBdeficient Sertoli cells
Rb-deficient Sertoli cells failed to support spermatogenesis, which resulted in increased germ cell apoptosis shown in a TUNEL-assay (Fig. 4A,B) (Rotgers et al., 2014). The number of apoptotic germ cells (TUNEL-positive, SOX9-negative) was quantified from seminiferous epithelial stages VII-VIII to increase the sensitivity of detecting non-physiological germ cell apoptosis, because physiological germ cell apoptosis is absent in these stages. However, the epithelial stages could not be identified from SC-Rb −/− E2f3 +/+ testes owing to germ cell loss, and thus random seminiferous tubule cross-sections were included in the analysis. When compared with the control, the SC-Rb −/− E2f3 +/− testes had more apoptotic germ cells in stages VII-VIII than the control (Fig. 4B), showing that the Sertoli cells were unable to support germ cell survival. However, the loss of both E2f3 alleles in the SC-Rb −/− E2f3 −/− Sertoli cells resulted in a rescue of the supportive function of Sertoli cells, and germ cell apoptosis was comparable to the control testes (Fig. 4B).
To gain a quantitative insight into the changes in testicular cell population sizes, flow cytometry analysis was performed on enzymatically dissociated testicular tissue (Rotgers et al., 2015).
As expected, based on the breakdown of spermatogenesis and infertility, the haploid cell population consisting of round and elongating spermatids was significantly smaller in the SC-Rb −/− E2f3 +/+ testes than in controls. Subsequently, this led to a relative overrepresentation of the diploid cell population, which included spermatogonia and somatic cells (Fig. 4C). SC-Rb −/− E2f3 +/− testes showed a slight decrease in the haploid cell population as compared with controls, while SC-Rb −/− E2f3 −/− testes did not differ from controls (Fig. 4C).
The expression of several pro-and anti-apoptotic genes was analyzed using targeted RNA sequencing. The relative expression of anti-apoptotic Bcl-2 was significantly increased in SC-Rb −/− E2f3 −/− testes in comparison with controls ( Fig. 4D). By contrast, the level of apoptotic Bax was significantly increased in SC-Rb −/− E2f3 +/+ , but not changed in SC-Rb −/− E2f3 −/− testes (Fig. 4D). Consequently, the Bcl-2:Bax ratio was increased in SC-Rb −/− E2f3 −/− testes (Fig. 4D), indicating activation of anti-apoptotic signaling. Proapoptotic Trp53 expression was also decreased in SC-Rb −/− E2f3 −/− testes as compared with controls ( Fig. 4D). Since the assay was performed on bulk RNA, it was not possible to distinguish gene expression changes between germ cells and Sertoli cells. However, altered Bcl-2 levels can affect Sertoli cell biology. Transgenic overexpression of Bcl-2 in Sertoli cells results in a failure of spermatogenesis due to Sertoli cell dysfunction (Yamamoto et al., 2001). Increased Bcl-2 expression in SC-Rb −/− E2f3 −/− testes may reflect the activation of crucial anti-apoptotic machinery that contributes to the survival of Sertoli cells despite lacking Rb.
In addition, Sertoli cell dysfunction in SC-Rb −/− E2f3 +/+ animals resulted in failure of the spermatogonial stem cell niche function, since relative expression levels of several spermatogonial stem cell markers [Oct4 (also known as Pou5f1), Lin28 (also known as Lin28a) and Gfra1] were significantly decreased in the SC-Rb −/− E2f3 +/+ testes as a sign of loss of this cell population, whereas in Table 1. Characterization of the SC-Rb −/− E2f3 −/− and SC-E2f3 −/− mouse line phenotypes One-way ANOVA followed by Tukey's post-test was used as a statistical test for the organ weight, fertility test, FSH and LH in the SC-Rb −/− E2f3 −/− study. Non-parametric Kruskal-Wallis test followed by Dunn's multiple comparison test was used to analyze the intratesticular testosterone assay in the SC-Rb −/− E2f3 −/− study. Unpaired t-test was used to analyze the SC-E2f3 −/− data at the two-and five-month time points. Statistical significance was analyzed compared to control, with the exception of values marked * ,a , which were compared with SC-Rb −/− E2f3 +/− . ***P<0.001, **P<0.01, *P<0.05. n.a., not analyzed.
SC-Rb −/− E2f3 −/− testes the levels were normal (Fig. 4E). Surprisingly, expression of markers for more differentiated spermatogonia [Plzf (also known as Zbtb16), Stra8 and Kit] was not altered (Fig. 4E). The Leydig cell hyperplasia likely contributed to the high Kit mRNA level in SC-Rb −/− E2f3 +/+ testes, since Kit is also expressed in Leydig cells (Manova et al., 1990). SC-Rb −/− E2f3 +/+ testes also showed a significant decrease in meiotic cell markers (Sycp1 and Spo11) and spermatid marker (Piwil1) mRNA, as expected based on the histological findings, and with E2f3 deletion, these were normalized (Fig. 4E). These data indicate that RB is required in Sertoli cells to maintain their ability to support the spermatogonial stem cell niche, meiosis and haploid germ cell differentiation. The loss of Rb in Sertoli cells led to a breakdown of Sertoli-germ cell junctions and sloughing of immature germ cells to the cauda epididymis (Fig. 5A, black arrows). This sloughing was rescued by additional loss of E2f3, since only spermatozoa were detected in the cauda epididymis of SC-Rb −/− E2f3 −/− epididymides (Fig. 5A).
Integrity of the seminiferous epithelium is maintained by the blood-testis barrier that forms between Sertoli cells and Sertoligerm cell junctions such as the ectoplasmic specializations. Claudin 11 (CLDN11, blood-testis barrier) and espin (Espn, ectoplasmic specializations) showed a disrupted expression pattern in all the seminiferous tubules of SC-Rb −/− E2f3 +/+ testes (Fig. 5B, asterisk), while the majority of the seminiferous tubules in SC-Rb −/− E2f3 +/− testes were intact, and in SC-Rb −/− E2f3 −/− testes no disruption was observed (Fig. 5B). Relative expression of Cldn11 was not significantly altered by loss of Rb in Sertoli cells (Fig. 5C). Claudin 3 (CLDN3) localizes to newly formed tight junctions between Sertoli cells in seminiferous epithelial stages VIII-X (Meng et al., 2005;Smith and Braun, 2012). Moreover, Cldn3 expression has been shown to be co-activated by RB and androgen receptor (AR) in Sertoli cells in vitro (Wu et al., 2013) and loss of Ar in vivo results in a loss of Cldn3 expression (Meng et al., 2005). Correspondingly, SC-Rb −/− E2f3 +/+ testes showed significantly decreased expression of Cldn3 mRNA, while the additional loss of E2F3 in Sertoli cells restored Cldn3 expression to close to normal (Fig. 5E). This suggests that Rb is involved in controlling Cldn3 expression in Sertoli cells in vivo and that this control is mediated by transcription factor E2F3.
We hypothesized that the focal disruption of the seminiferous epithelium in SC-Rb −/− E2f3 +/− testes (Fig. 2B) was associated with a defect of Sertoli cell cycling during the seminiferous epithelial cycle. The stage-specific cycling behavior of Sertoli cells was analyzed by using AR expression pattern as a proxy, since AR staining intensity fluctuates depending on the seminiferous epithelial stage (Bremner et al., 1994). The stage-specificity of AR expression was retained in Sertoli cells in all experimental groups, with a strong AR signal in Sertoli cell nuclei in seminiferous epithelial stages V-VI and a low signal in stages X-XI (Fig. 5F). Even in SC-Rb −/− E2f3 +/+ testes, where discerning the seminiferous epithelial stages based on germ cell morphology was not possible, a clear fluctuation of AR intensity between tubule cross sections could be observed (Fig. 5F, tubules marked with H for high expression and L for low expression). Furthermore, relative Ar expression showed no changes when compared to control (Fig. 5D). These results indicate that the RB-E2F3 pathway does not control AR expression in Sertoli cells.
Since E2F3 is a transcription factor, we were interested in analyzing the effect of E2f3 loss on the expression of putative E2F target genes in the testis to find possible downstream effectors of E2f3. The putative E2F target genes were chosen both directly from literature (Bracken et al., 2004) and through cross-comparisons of experimental data. The lists of differentially expressed genes in earlier SC-RbKO studies (Nalam et al., 2009;Rotgers et al., 2014) were compared to ChIP-on-CHIP datasets, where binding of E2F1 and E2F3 to promoters was studied in mouse cells (Asp et al., 2009;Cao et al., 2011). As expected, based on the experimental design, control and SC-Rb −/− E2f3 +/+ testes showed differential expression of many of these genes as shown in the heatmap (Fig. 6F). SC-Rb −/− E2f3 +/− and SC-Rb −/− E2f3 −/− more closely resembled expression patterns in control testes than SC-Rb −/− E2f3 +/+ testes (Fig. 6F).

DISCUSSION
Rb controls the maintenance of Sertoli cell quiescence and terminal differentiation in adulthood (Nalam et al., 2009;Rotgers et al., 2014). We have previously shown that RB interacts with E2F3 during testicular development, and that shRNA-mediated knockdown of E2f3 alleviated the degenerative phenotype associated with loss of Rb in Sertoli cells (Rotgers et al., 2014). In this study, we used E2f3 flox/flox mice to achieve a complete knockout of E2f3 in Sertoli cells and to obtain a more robust model for E2f3 knockdown than in vivo shRNA for studying the role of the RB-E2F3 pathway in mouse Sertoli cells.
Sertoli cell dysfunction in SC-Rb −/− E2f3 +/+ testes was identical to that previously published (Rotgers et al., 2014) and these animals showed a disruption of spermatogenesis due to a combined failure of the spermatogonial stem cell niche, increased germ cell apoptosis and breakdown of Sertoli-germ cell junctions. Sertoli cell dysfunction was rescued by loss of E2f3. SC-Rb −/− E2f3 +/− testes showed a complete rescue of spermatogenesis in the majority of seminiferous tubules on a morphological level when compared with SC-Rb −/− E2f3 +/+ testes in 2.5-month-old animals. However, by the age of four months, SC-Rb −/− E2f3 +/− animals, with one remaining functional E2f3 allele, showed a similar level of testis disruption as SC-Rb −/− E2f3 +/+ animals (data not shown). In addition, SC-Rb −/− E2f3 +/− testes showed Sertoli cell proliferation at a similar rate as in SC-Rb −/− E2f3 +/+ testes in the 2.5-month-old mice, suggesting that a single copy of E2f3 is sufficient to drive adult Sertoli cell proliferation in the absence of Rb, and this ultimately leads to a complete disruption of spermatogenesis. However, a complete loss of E2f3 in Sertoli cells of SC-Rb −/− E2f3 −/− mice led to a sustained rescue of spermatogenesis and Sertoli cell quiescence status. Moreover, when Rbf (Drosophila homolog of RB) is knocked down in hub cells, which are the somatic component of the stem cell niche in the Drosophila testis, the hub cells exit the quiescent state and form ectopic stem cells niches. Similarly to our model, ectopic hub cell proliferation is rescued by a simultaneous knockdown of E2F, indicating a conserved role of an RB-E2F switch in the control of testicular nurse cell quiescence (Greenspan and Matunis, 2018). E2Fs also drive cell proliferation in the absence of RB in several other cell types; loss of E2f3 can rescue defects in proliferation and apoptosis in Rb-deficient lens and CNS cells (Ziebold et al., 2001) and pituitary tumors in Rb −/− mice can be rescued by E2f3 deletion (Ziebold et al., 2003). In conclusion, RB-mediated inhibition of E2F3 is essential for the maintenance of quiescence in adult Sertoli cells, which is in agreement with the role of E2F3 as a driver of cell cycle progression.
The E2f3 locus encodes two isoforms, E2F3a and E2F3b, through the use of alternative promoters, and these can have opposing roles in regulating cell fate (Danielian et al., 2008). E2F3a expression was gradually induced in mouse Sertoli cells upon postnatal maturation. RB and E2F3 interact in mouse Sertoli cells on P10, which coincides with the induction of the cell cycle exit period as Sertoli cells mature in puberty (Rotgers et al., 2014;Zimmermann et al., 2015). It is possible that transcription factor E2F3a recruits RB to the promoters of genes associated with Sertoli cell proliferation to induce gene silencing, and/or to genes associated with maturation, to activate transcription. We hypothesized that acquiring E2F3a expression in Sertoli cells was associated with cell cycle exit during maturation. However, expression of proliferation marker KI67 and E2F3a did not correlate in the juvenile testis as some E2F3a-positive Sertoli cells were positive for KI67 while others were negative. Furthermore, E2f3 is dispensable for cell cycle progression of mouse Sertoli cells, since loss of E2f3 did not affect testis size in SC-E2f3 −/− mice, indicating that Sertoli cell proliferation was not compromised during development. This could result from compensation for E2f3 loss by the other E2Fs, since the activator E2Fs show extensive redundancy during development (Kong et al., 2007). Only E2F3 and the repressive E2F4 are expressed in wild-type mouse Sertoli cells (El-Darwish et al., 2006) and SC-Rb −/− E2f3 +/+ testes showed a decreased expression of E2f4, rather than an increase. However, we cannot exclude the possibility that E2F4 compensates for E2F3 in transcriptional control of Sertoli cells since RB can interact with both E2F3 and E2F4, and the E2Fs share a similar consensus DNAbinding site (Moberg et al., 1996;Zheng et al., 1999).
One of the E2F target genes that showed a high level of induction in SC-Rb −/− E2f3 +/+ testes was Fst (Müller et al., 2001). Fst is highly expressed in immature mouse testes and its expression gradually decreases as the animals mature (Barakat et al., 2008). Interestingly, transgenic expression of Fst in the testis results in a phenotype very similar to that of SC-Rb −/− E2f3 +/+ and SC-Rb −/− E2f3 +/− testes (Guo et al., 1998). In particular, Fst transgenic testes exhibit a similar focal seminiferous tubule disruption as observed in SC-Rb −/− E2f3 +/− testes (Guo et al., 1998). Fst expression was dependent on the E2F3 status of Sertoli cells, with expression significantly increased in SC-Rb −/− E2f3 +/+ testes and decreased in response to the loss of E2f3, in a dose-dependent manner. We hypothesize that RB and E2F3 act as a complex to repress Fst expression in adult wild-type Sertoli cells. By contrast, E2F3 induces Fst expression in the absence of RB in SC-Rb −/− E2f3 +/+ testes, and increased Fst expression could contribute to the loss of Sertoli cell function. Alternatively, rather than being a driver of the phenotype, increased Fst expression could merely be a consequence of reactivated Sertoli cell proliferation in adulthood, as FST inhibits activin A, which promotes Sertoli cell proliferation (Archambeault and Yao, 2014).
Cldn11, an integral part of tight junctions on the blood-testis barrier, results in both failure of supporting spermatogenesis and reactivation of Sertoli cell proliferation in the adult, but no apparent dedifferentiation (Gow et al., 1999;Mazaud-Guittot et al., 2010). Knockout of gap junction protein connexin-43 (CX43, also known as GJA1) leads to a similar phenotype with continued Sertoli cell proliferation in adulthood, sloughing of Sertoli cells to seminiferous tubule lumen and a failure of spermatogenesis (Sridharan et al., 2007). In our study, Rb appeared to have a direct role in controlling the expression of Cldn3, a tight junction protein associated with the newly formed blood-testis barrier (Smith and Braun, 2012), since Cldn3 mRNA levels were dramatically decreased in SC-Rb −/− E2f3 +/+ testes. It has been shown in vitro that RB and AR co-activate Cldn3 expression in Sertoli cells (Wu et al., 2013) and that a loss of Ar in Sertoli cells decreases Cldn3 expression and increases the permeability of the blood-testis barrier (De Gendt et al., 2004;Meng et al., 2005). The loss of E2f3 in SC-Rb −/− E2f3 −/− testes restored expression of Cldn3 to a normal level. Our results suggest that Rb controls Cldn3 expression in Sertoli cells via regulation of E2f3, also in vivo.
In summary, Rb is crucial for the maintenance of cell cycle quiescence in adult mouse Sertoli cells and the main function of RB is to inhibit transcription factor E2F3. Without functional RB, E2F3 can drive adult Sertoli cells to resume proliferation. As a result, Sertoli cells are unable to support the spermatogonial stem cell niche, spermatogenesis and blood-testis barrier integrity, which ultimately leads to infertility. This study elucidates the significance of cell cycle status for the maintenance of Sertoli cell function and uncovers the critical contribution of the RB-E2F3 pathway in this process. Mature Sertoli cells need to remain mitotically quiescent to be able to carry out their function as supporters of spermatogenesis.

Animals
Animals (strains and sources listed below) were housed under environmentally controlled conditions (12 h light:12 h darkness; temperature 21±1°C) in the animal facility of the University of Turku. They were fed soy-free RM-3 (E) mouse chow (Special Diet Services, Essex, UK) and tap water ad libitum. All procedures were carried out according to the institutional and ethical policies of the University of Turku and approved by the local ethics committee on animal experimentation.
Animals were euthanized by carbon dioxide inhalation and cervical dislocation. For mRNA and protein expression analysis of the target of study, control male mice were euthanized at P6, P10, P20 and P40. For the SC-E2f3 −/− study, male mice were euthanized at a young adult age (2 months old) and after aging (5 months old). For the SC-Rb −/− E2f3 −/− study, 2.5 months was chosen as the primary experimental timepoint, because at this age a clear disruptive phenotype could be seen in the SC-Rb −/− E2f3 +/+ testes, which enabled assessing the rescue effect of E2f3 loss.

Fertility
Fertility testing was conducted by breeding four males per experimental group with four C57Bl6/NCrl females for 24 h (or 48 h if a copulatory plug was not observed after 24 h). For the SC-E2f3 −/− strain the test was performed at the ages of 2 and 5 months. For the SC-Rb −/− E2f3 −/− strain a single time point of 2.5 months was used in order to target the same age as that primarily analyzed for adults in the previous publication on SC-Rb −/− E2f3 +/+ animals in the hybrid genetic background (Rotgers et al., 2014). The number and sex ratio of the pups was recorded and displayed as mean±s.e.m. pups per litter. Statistical testing was performed using paired t-test for the SC-E2f3 −/− strain and one-way ANOVA followed by Tukey's test for the SC-Rb −/− E2f3 −/− strain.

Serum gonadotrophin and intratesticular testosterone assays
Serum was collected from 2.5-month-old control, SC-Rb −/− E2f3 +/+ , SC-Rb −/− E2f3 +/− and SC-Rb −/− E2f3 −/− males and 2-and 5-month-old controls and SC-E2f3 −/− males (n=8/group). Serum FSH and LH were analyzed using immunofluorometric assay (IFMA) (Delfia, Wallac Oy, Turku, Finland). For the FSH assay 65 µl of serum was diluted to 1:5. The sensitivity of the assay was 0.1 µg/l (van Casteren et al., 2000). For the LH assay, 25 µl of serum was used. The sensitivity of the assay was 0.03 µg/l in 25 µl (Haavisto et al., 1993). For the intratesticular testosterone assay, pieces of testicular tissue were weighed and homogenized in PBS. 200 µl of the homogenate was extracted twice with diethyl ether and dried through evaporation. The residues were reconstituted in PBS with 0.1% BSA and measured using a radio-immunoassay kit (TESTO-CT2; Cisbio Bioassays, Codolet, France). The analytical detection limit was better than 0.1 nmol/l and functional detection limit is approximately 0.3 nmol/l. The intra-assay and inter-assay coefficients of variation for testosterone were 10.5% and 17.3% in the SC-E2f3 −/− analysis and 10.8% and 19.3% in the SC-Rb −/− E2f3 −/− . All serum samples within an analysis were assayed at the same time using specific antiserum and radiolabeled hormones diluted the same day.

Histology
For the histological analysis the testes were fixed in Bouin's fixative overnight at room temperature. After serial washes in ethanol followed by dehydration, the testes were embedded in paraffin. Sections at 4 µm thick were cut from the testes and every tenth section was collected for the analysis. The sections were stained with periodic acid-Schiff (PAS) and imaged using Pannoramic 250 slide scanner with the 20× objective (3DHISTECH Ltd).

Immunohistochemical sample processing
Testes were fixed overnight in 4% paraformaldehyde (Electron Microscopy Sciences) and embedded in paraffin. Sections at 4 µm thick were cut. When the cell counts were quantified, every tenth section was chosen for the analysis. The analyzed sections were ten sections apart to avoid reanalyzing the same cells on each section. The TUNEL and SOX9 assay was conducted as described previously (Rotgers et al., 2014). The antibodies used and their dilutions are summarized in Table S1, except where indicated otherwise.
Germ cell apoptosis was only quantified from stage VII-VIII seminiferous tubule cross-sections to increase the sensitivity in detecting excess apoptosis, since there is a very low level of apoptosis in these stages in normal testes (Yan et al., 2000). Intratubular TUNEL-positive SOX9negative cells were classified as apoptotic germ cells. For SC-Rb −/− E2f3 +/+ testes, random tubule sections were chosen for the analysis because it was not possible to identify seminiferous tubule stages VII-VIII in the absence of spermatids.
Sertoli cell proliferation was quantified from SOX9-and KI67-stained testicular sections. The number of double-positive Sertoli cells for two hundred Sertoli cells on a section was quantified. Three sections, each 50 µm apart, were quantified for each animal.
p130 and p107 were detected on testicular sections with a Novolink chromogenic immunohistochemistry kit (Leica) according to the manufacturer's instructions. After detecting the signal with DAB, samples were counterstained using Mayer's hematoxylin, dehydrated and embedded with Pertex. The slides were imaged using a Pannoramic MIDI FL slide scanner with a 20× Plan Apochromat objective (Zeiss).

RNA in situ hybridization
An Advanced Cell Diagnostics (ACD) RNAscope 2-plex chromogenic assay kit (ACD-320700,) was used to detect E2f3a and E2f3ab mRNA on mouse PFA-fixed paraffin-embedded mouse testicular samples. Custom probes were designed by ACD for the detection of the E2f3 isoforms. In addition, the positive (ACD-310751) and negative (ACD-320741) control probes accompanying the kit were used. E2f3a mRNA signal was detected using the red Vulcan Fast Red chromogen provided in the kit. The E2f3ab signal was detected using an alternative black chromogen ( provided as a custom reagent by ACD). The assay was run according to the manufacturer's instructions with some modifications. The slides were incubated in the 'pretreat 2' reagent for 15 min and for 20 min in 'pretreat 3'. Mayer's hematoxylin was used as a counterstain and water at pH 11 was used to induce blue color in the hematoxylin stain. The samples were air-dried at room temperature and mounted using Ecomount. For the RNA in situ hybridization coupled to immunohistochemistry (IHC) of vimentin, the slides were processed as described above until the chromogenic reaction for both of the probes was ready. Then they were incubated with 5% horse serum in TBS followed by an overnight incubation with the primary antibody [anti-vimentin (VIM) diluted 1:3000 in 5% horse serum in TBS] at +4°C. Residual HRP activity was blocked by an incubation in 3% hydrogen peroxide in water. The primary antibody was detected using a biotinylated horse anti-rabbit antibody (1:500 in 5% horse serum in TBS) followed by Vector ABC-HRP kit. HIGHDEF yellow IHC chromogen (HRP) (Enzo Life Sciences) was used to visualize the vimentin staining. The slides were airdried at room temperature and mounted using Ecomount. The slides were imaged using a Pannoramic MIDI FL slide scanner with a 40×/Korr 0.95 Plan Apochromat objective (Zeiss) using an extended focus to better visualize the RNAScope signal.

Targeted RNA sequencing
A custom assay panel was designed for the TruSeq targeted RNA sequencing (Illumina) ( Table S3). The panel consisted of a selection of testicular cell type marker genes and known cell cycle-related genes. In addition, putative E2F target genes were chosen both from the literature (Bracken et al., 2004) and by comparing the differentially expressed genes in juvenile SC-Rb −/− E2f3 +/+ testis (Rotgers et al., 2014) and adult Sertoli cells (Nalam et al., 2009) to ChIP-on-chip data on E2f3 and E2f1 target genes from murine and human cells (Asp et al., 2009;Cao et al., 2011) and overlapping genes were selected to be added to the panel. For the assay, whole testis mRNA was extracted from animals of various genotypes and ages as described above. For a detailed experimental setup see Table S2.
The library preparation and sequencing were performed at the Finnish Functional Genomics Centre (FFGC, Turku, Finland). Prior to the library preparation, the quality of the total RNA samples was ensured using an Agilent Bioanalyzer 2100 and an Advanced Analytical Fragment Analyzer. Total RNA samples were pure, intact and all samples had similar quality. Bioanalyzer RNA integrity number (RIN) values were >9.0. 50 ng of RNA was taken for the library preparation. Library preparation was done according to the Illumina TruSeq Targeted RNA Expression Guide (15034665). The high quality of the libraries was confirmed using an Advanced Analytical Fragment Analyzer. The libraries were normalized and pooled for the automated cluster preparation, which was carried out using an Illumina MiSeq instrument. The 44 libraries were pooled into a single pool. The samples were sequenced in a single run on the Illumina MiSeq instrument using v2 sequencing chemistry. Single-read sequencing with 1×50 bp read length was used, followed by 6 bp+8 bp dual index run. Technical quality of the MiSeq run was good and the cluster amount was as expected. Greater than 90% of all bases above Q30 was requested.
The results were submitted to the Illumina BaseSpace data management and analysis interphase. The TruSeq Targeted RNA sequencing application v1.0 was used to analyze differential expression of the transcripts in the experimental groups. Briefly, the depth of sequencing between samples was estimated and the expression levels for each replicate were normalized based on the total number of aligned reads. The normalized transcript abundance was modeled by a negative binomial distribution model and this model was used to derive a P-value for the differential expression of each transcript. Finally, the application corrected for multiple hypothesis testing by adjusting the P-value for false discovery rate (FDR) using the Benjamini-Hochberg method. Hprt, Rpl19 and Ppia were used as reference genes for normalization during the analysis. The gene expression results are visualized as fold changes of the pairwise comparisons between control and experimental groups.