Oocyte-dependent activation of MTOR in cumulus cells controls the development and survival of cumulus–oocyte complexes

The ultimate goal of ovarian folliculogenesis is to produce a mature oocyte (egg) for reproduction, and to generate hormones and growth factors, particularly steroids, for endocrine functions (Eppig, 2001). To achieve these goals, the germinal and somatic compartments of the follicle must coordinate precisely in responding to the myriad endocrine, paracrine, autocrine and juxtacrine signals prevalent in the follicular microenvironment to ensure the proper development and normal functions of the follicle (Su et al., 2009). Bi-directional communication between the oocyte and its companion granulosa cells is a crucial factor in this coordination. Granulosa cells promote oocyte growth and global silencing of transcription (De La Fuente and Eppig, 2001; Hutt et al., 2006), maintain oocyte meiotic arrest (Zhang et al., 2010) and mediate the resumption of meiosis induced by the preovulatory luteinizing hormone surge (Park et al., 2004), whereas signals from oocytes promote granulosa cell proliferation and differentiation by coordinating with signals induced by follicle-stimulating hormone (FSH) (Diaz et al., 2007a; El-Hefnawy and Zeleznik, 2001; Gilchrist et al., 2001; Latham et al., 2004; Li et al., 1995; Vanderhyden et al., 1990, 1992), suppress granulosa cell apoptosis and premature luteinization (Eppig et al., 1997; Hussein et al., 2005; Orisaka et al., 2006), and enable expansion of the cumulus cell population and ovulation (Buccione et al., 1990; Joyce et al., 2001).

An ‘oocyte–granulosa-cell regulatory loop’ involving complementary metabolic and signaling pathways between oocytes and granulosa cells controls the development and functions of both compartments, and highlights a crucial role of oocytes in this dialog (Eppig, 2001; Su et al., 2004). Importantly, oocyte control of the metabolic cooperativity between the oocyte and its companion granulosa cells is a cornerstone of this regulatory loop (Su et al., 2009). For example, oocytes are unable to carry out some metabolic processes – including glycolysis and cholesterol biosynthesis – and to take up certain amino acids, such as L-alanine, owing to their deficient expression of genes encoding related enzymes and amino acid transporters, and therefore rely on cumulus cells to provide them with the corresponding metabolites that support their development (Biggers et al., 1967; Eppig et al., 2005; Su et al., 2008; Sugiura et al., 2005). Remarkably, oocytes enable this metabolic cooperativity by promoting the expression of these related key metabolic enzymes and amino acid transporters in cumulus cells (Eppig et al., 2005; Su et al., 2008; Sugiura et al., 2005, 2007). In addition to the role of compensating for its own metabolic deficiencies, the oocyte probably controls the growth rate and survival of follicles by promoting cumulus cell metabolism (Eppig et al., 2002; Su et al., 2009).

How cues from oocytes result in metabolic action in cumulus cells remains uncertain, but the mechanistic target of rapamycin (MTOR; also known as mammalian target of rapamycin, mTOR) pathway is likely to be involved. MTOR is an evolutionarily conserved kinase and an essential component of the classic energy-sensing pathway within cells, and plays essential roles in controlling cell growth and development by integrating and transmitting multiple extracellular cues, including nutrient supplies and signals affecting growth and stress (Albert and Hall, 2015; Aramburu et al., 2014; Huang and Fingar, 2014). Several studies have shown that this pathway might participate in the control of granulosa cell proliferation (Kayampilly and Menon, 2007; Yu et al., 2012), differentiation (Alam et al., 2004) and follicle growth (Yu et al., 2011). However, whether or not this pathway is regulated by oocyte-derived signals, and whether it is involved in the control of oocyte developmental competence and cumulus cell survival is unresolved and is therefore the focus of this study.

Many of the regulatory effects of oocytes on their companion granulosa cells are mediated by the oocyte-derived paracrine factors (ODPFs) – growth-differentiation factor 9 (GDF9) and/or bone morphogenetic protein 15 (BMP15), two members of the transforming growth factor superfamily (Pangas and Matzuk, 2004). Both GDF9 and BMP15 are predominantly expressed by oocytes, and are essential for female fertility in a variety of species studied (Pangas and Matzuk, 2004; Su et al., 2009). Gdf9-knockout (Gdf9^−/−) female mice are sterile owing to the arrest of primary follicle development (Dong et al., 1996), whereas Bmp15-null (Bmp15^−/−) females are subfertile owing to defective ovulation (Yan et al., 2001). Gdf9^+/− Bmp15^−/− (hereafter referred to as double mutants) females display a more severe subfertility than Bmp15^−/− mice, which is largely caused by abnormal development of cumulus cells (Yan et al., 2001). Therefore, GDF9 and BMP15 are essential for cumulus cell development both before and after the luteinizing hormone surge (Su et al., 2004). GDF9 and BMP15 can form heterodimers that are more active than each of the homodimers (Mottershead et al., 2015; Peng et al., 2013; Wigglesworth et al., 2013).

Taking advantage of the defective development of the double-mutant cumulus cells, we have described previously a transcriptomic analysis of double-mutant cumulus cells using microarrays in order to gain a global perspective of the role of ODPFs in cumulus cell development (Su et al., 2008). Surprisingly, we have found that the major theme associated with GDF9- and BMP15-regulated transcripts in cumulus cells is metabolism, which includes cholesterol biosynthesis, glycolysis and inositol metabolism (Su et al., 2008). Moreover, levels of DNA-damage-inducible transcript 4-like (Ddit4l) mRNA is strongly upregulated in Bmp15^−/− and double-mutant cumulus cells (Su et al., 2008). Because DDIT4L (also known as development and DNA-damage response 2, REDD2, or RTP801-like, RTP801L) is an upstream negative regulator of MTOR (Corradetti et al., 2005), we hypothesized that oocytes enable the activation of the MTOR pathway in cumulus cells, at least in part, by suppressing the expression of DDIT4L. Here, we tested this idea and investigated the potential biological functions relevant to oocyte-controlled activation of MTOR in cumulus cells.

By mining our previously published dataset (Su et al., 2008), we found that the mRNA levels of Ddit4l, but not of its closely related paralog Ddit4, were significantly upregulated in Bmp15^−/− and double-mutant cumulus cells (Fig. 1A; Fig. S1A). This upregulation was validated by quantitative real-time RT-PCR (qRT-PCR) analysis (Fig. 1A). Immunohistochemistry revealed that in wild-type large antral follicles, DDIT4L was predominantly expressed by mural granulosa cells adjacent to the follicular basal lamina, and there were very few cumulus cells that stained positively for DDIT4L (Fig. 1B,C; Fig. S1B). In contrast to the wild-type follicles, the difference in DDIT4L expression level between mural granulosa cells and cumulus cells was diminished in double-mutant antral follicles, and there was a large proportion (∼60%) of cumulus cells that stained positively with the antibody against DDIT4L (Fig. 1B,C; Fig. S1B).

Because both Bmp15 and Gdf9 are exclusively expressed by oocytes, the upregulation of Ddit4l mRNA and protein in double-mutant cumulus cells implies that mouse oocytes suppress the expression of Ddit4l. We tested this possibility by using microsurgery to remove the oocyte from normal wild-type COCs and culturing the resultant oocytectomized (OOX) cumulus cells in vitro. As shown in Fig. 2A,B, Ddit4l mRNA was upregulated in oocytectomized cumulus cells after 20 h of culture, this upregulation was completely prevented by co-culture of oocytectomized cumulus cells with wild-type fully grown oocytes. However, neither the Bmp15^−/− nor the double-mutant oocytes were able to prevent the increase of Ddit4l mRNA in oocytectomized cumulus cells as effectively as the wild-type oocytes; they only partially suppressed the upregulation caused by oocytectomization (Fig. 2B). Interestingly, Ddit4 mRNA was unchanged in oocytectomized cumulus cells (Fig. S1A). Treating oocytectomized cumulus cells with recombinant mouse GDF9 (500 ng/ml) also effectively prevented the upregulation of Ddit4l mRNA. Recombinant mouse GDF9–BMP15 heterodimer elicited a stronger inhibitory effect on the expression of Ddit4l mRNA in oocytectomized cumulus cells; it completely prevented the upregulation of Ddit4l mRNA even at the concentration of 1 ng/ml, which was 500 times as efficient as the GDF9 monomer (Fig. 2D).

The SMAD2-dependent pathway mediates regulatory signals from oocytes to companion granulosa cells (Diaz et al., 2007b; Mottershead et al., 2012; Su et al., 2010). We therefore tested whether this pathway also participates in oocyte-mediated suppression of Ddit4l mRNA expression in cumulus cells. As shown in Fig. 3A, when COCs were treated with 10 μM SB431542, a SMAD2–SMAD3 inhibitor (Inman et al., 2002), Ddit4l mRNA expression in cumulus cells was upregulated. However, the same effect did not occur when COCs were treated with 20 μM SIS3, which inhibits SMAD3 only (Jinnin et al., 2006), rather, there was a slight decrease in Ddit4l mRNA in cumulus cells. SB431542, but not SIS3, also effectively abolished the suppressive effect of GDF9 on Ddit4l mRNA expression in oocytectomized cumulus cells; SIS3 partially enhanced the suppressive effect of GDF9 on Ddit4l mRNA expression in oocytectomized cumulus cells (Fig. 3B).

In situ hybridization revealed that Ddit4l mRNA was robustly expressed by mural granulosa cells but was barely detectable in cumulus cells within large antral follicles of normal wild-type mouse ovaries (Fig. 4A,B). This differential pattern of Ddit4l mRNA expression was further confirmed by performing qRT-PCR analysis using cumulus and mural granulosa cells that had been isolated from large antral follicles (Fig. 4C). Similarly, immunohistochemical analysis revealed that DDIT4L protein was also differentially expressed within large antral follicles; the positive signal, as indicated by red staining, was enriched in granulosa cells lining the follicle wall, and a very low signal, if any, was found in the cumulus cells and peri-antral granulosa cells (Fig. 4D). However, unlike DDIT4L, DDIT4 was expressed uniformly in all types of granulosa cells within large antral follicles (Fig. S1C).

Because DDIT4L is an upstream negative regulator of the MTOR signaling pathway (Corradetti et al., 2005), the observation that oocytes suppress the expression of Ddit4l mRNA in cumulus cells and that Ddit4l is highly expressed by mural granulosa cells indicates that oocytes might promote differential activation of the MTOR pathway in cumulus cells by suppressing Ddit4l expression. To test this possibility, we assessed whether the oocyte and its secreted paracrine factors, GDF9 and GDF9–BMP15 heterodimer, control the activation of MTOR in cumulus cells. As shown in Fig. 5A,B, when oocytes were removed from COCs by performing micro-surgery and cultured for 20 h, the levels of the active forms of MTOR (phosphorylated MTOR, p-MTOR) and its downstream effectors – RPS6KB1 (phosphorylated RPS6KB1, p-RPS6KB1) and EIF4EBP1 (phosphorylated EIF4EBP1, p-EIF4EBP1) (also known as S6K1 and 4E-BP1, respectively) – in the resultant oocytectomized cumulus cells were reduced. Co-culture of oocytectomized cumulus cells with normal wild-type fully grown oocytes effectively prevented this reduction in the levels of p-MTOR, p-RPS6KB1 and p-EIF4EBP1. Treating oocytectomized cumulus cells with 500 ng/ml of recombinant GDF9 or 3.5 ng/ml of GDF9–BMP15 heterodimer was also effective in preventing the reduction of p-MTOR, p-RPS6KB1 and p-EIF4EBP1. Is the MTOR pathway differentially activated in cumulus cells? Immunofluorescent analysis revealed that the active form of ribosomal protein S6, phosphorylated RPS6 (p-RPS6), another major downstream effector of MTOR activation, was strongly expressed in cumulus cells and peri-antral granulosa cells but barely detectable in mural granulosa cells (Fig. 5C).

Given that DDIT4L is reported to be an upstream negative regulator of MTOR in other types of cells (Corradetti et al., 2005), the mutually exclusive patterns of expression of DDIT4L and p-RPS6 in mural and cumulus granulosa cells suggest that oocyte suppression of Ddit4l expression in cumulus cells might directly contribute to the activation of MTOR signaling. This possibility was tested here in cultured monolayer cumulus cells by knocking down Ddit4l expression and then assessing the activation of the MTOR pathway. Because the efficiency of gene knockdown in the multi-layered oocytectomized complexes cultured in the three-dimensional form is low (Sugiura et al., 2009), cumulus cells were cultured in monolayer. As shown in Fig. 6A, Ddit4l mRNA was upregulated about fourfold after being seeded in monolayer and cultured for 24 h. Transfection of these initially cultured monolayered cumulus cells with specific small interfering RNAs (siRNAs) against Ddit4l, but not with the control siRNAs, reduced the expression of Ddit4l mRNA by 97% (Fig. 6B). Importantly, levels of p-MTOR, p-RPS6KB1 and p-EIF4EBP1 were significantly elevated in cumulus cells with knocked down Ddit4l, which were 2.4-, 2.3- and 1.8-fold the levels in control groups, respectively (Fig. 6C,D).

As a key intracellular energy-sensing pathway, MTOR signaling is crucial for sustaining the development and survival of a variety of cell types (Albert and Hall, 2015; Aramburu et al., 2014; Huang and Fingar, 2014). Because cumulus cells in COCs play an indispensable nurturing role in supporting oocyte development and maturation, we hypothesized that the oocyte-controlled activation of MTOR in cumulus cells promotes oocyte developmental competence and cumulus cell survival. We tested this possibility by inhibiting the MTOR pathway in intact COCs in culture and subsequently by assessing the post-fertilization developmental competence of oocytes and the levels of apoptosis in cumulus cells. As shown in Fig. 7A, Torin1, a specific inhibitor of MTOR (Liu et al., 2010), inhibited the phosphorylation of RPS6KB1 and EIF4EBP1 in isolated COCs at all the dosages tested. Both the rate of fertilization and the development of two-cell-stage embryos into blastocysts were significantly reduced in all the groups of COCs that had been initially treated with various doses of Torin1 for 24 h (Fig. 7B). Moreover, Torin1 dose dependently increased the occurrence of apoptotic cumulus cells, as revealed by whole-mount terminal deoxynucleotidyl transferase dUTP nick-end labeling (TUNEL) staining of COCs (Fig. 7C,D).

In order to gain further insights into the roles that MTOR plays in cumulus cells, particularly those specific to cumulus cells, and to reveal the potential mechanisms responsible for Torin1-mediated reduction in oocyte developmental competence and cumulus survival, transcriptomic analysis of COCs that had been treated with 5 μM Torin1 or DMSO vehicle for 14 h in culture was performed using the Affymetrix^® Mouse Gene 2.1 ST Arrays. A total of 1023 transcripts (indicated by the probe IDs) were changed significantly in COCs that had been treated with Torin1, of which 461 were upregulated and 549 were downregulated. Unsupervised hierarchical clustering revealed that these transcripts were changed to similar extents across all the samples tested (Fig. S2). Molecule Annotation System (MAS) software was used to identify the enriched pathways associated with the changed transcripts. Among the top 30 pathways identified, MAPK signaling and extracellular matrix (ECM)-related processes were associated most significantly with upregulated transcripts (Fig. S3), whereas metabolic pathways, particularly cholesterol and sterol biosynthesis, were predominantly associated with downregulated transcripts (Fig. S4).

Because the MAPK pathway is essential for COC maturation and other ovulation-related processes after the luteinizing hormone surge (Fan et al., 2009; Su et al., 2002, 2003, 2006), and ECM formation is key for cumulus expansion (Richards, 2007), we analyzed the changed transcripts in more detail. To our surprise, many transcripts that have been reported previously to be up- or downregulated after the luteinizing hormone surge in COCs were also changed in the same direction in Torin1-treated COCs (Hernandez-Gonzalez et al., 2006). By comparing with another set of microarray data that we generated independently to examine the luteinizing-hormone-induced transcriptomic changes in cumulus cells in vivo, we found that 74.8% and 60.7% of the up- and downregulated transcripts, respectively, in Torin1-treated COCs were also changed in a similar manner in cumulus cells 6 h after injection of human chorionic gonadotropin (hCG) (Fig. 8A). In particular, transcripts essential for cumulus expansion – i.e. Has2, Ptgs2, Tnfaip6 and Ptx3, and those encoding EGF-like growth factors, Areg, Ereg and Btc – were strongly upregulated by treatment with Torin1. These transcripts (Fig. 8D,E) and some of the other transcripts that were most dramatically upregulated (Fig. 8B) were validated with qRT-PCR analysis. For the downregulated transcripts, we validated those encoding the amino acid transporter SLC38A3, the glycolytic enzyme ALDOC and the essential enzymes for cholesterol biosynthesis, as well as those encoding important receptors (NPR2, EGFR, FSHR, AR and TGFBR3) on cumulus cells, by performing qRT-PCR (Fig. 8C, F).

Oogenesis and folliculogenesis are highly coordinated processes requiring complex cell-to-cell communication and a myriad of endocrine and paracrine factors that interact within the follicular microenvironment. Pituitary-derived FSH and factors produced by oocytes are of particular importance in the control of ovarian follicular development. FSH is classically known to be crucial for granulosa cell differentiation and follicle maturation, such that genetic deletion of Fshr or Fshb prevents the development of follicles significantly beyond the early antral stage (Dierich et al., 1998; Kumar et al., 1997). Whereas ODPFs have a more profound impact on follicular development starting from follicle formation up to the process of ovulation (Dong et al., 1996; McNatty et al., 2004; Soyal et al., 2000). Oocytes orchestrate the rate of follicle growth and appear to be a dominant factor in ovarian follicular development (Eppig et al., 2002). Interestingly, ODPFs and FSH exert opposing actions in directing patterns of gene expression in granulosa cells during the transition from preantral to antral follicles (Diaz et al., 2007b). By over-riding the effects of FSH in controlling the expression of Lhcgr mRNA and other genes that are characteristic of mural granulosa cells (Diaz et al., 2007b), oocytes promote the development and function of cumulus cells (Eppig, 2001). Accumulating evidence indicates that reciprocal signals and complementary metabolites generated by the oocyte and its companion granulosa cells promote the development and function of both compartments in a loop of interconnected regulatory processes (Eppig, 2001; Su et al., 2009). Here, we demonstrated that oocytes promote cumulus cell development and survival by suppressing Ddit4l expression, thereby enabling MTOR activation in cumulus cells, which in turn ensures the production of functionally normal eggs. These observations reveal an underlying mechanism governing the complex interplay between the oocyte and its companion granulosa cells in orchestrating the growth and function of ovarian follicles.

Mammalian DDIT4L is encoded by the mammalian ortholog of the Drosophila gene charybde that was originally identified to be a hypoxia-inducible negative regulator of the Drosophila TOR pathway (Reiling and Hafen, 2004). It is stimulated under certain stress conditions and functions as a negative regulator of MTOR by modulating the availability of free TSC2, the immediate upstream inhibitor for MTOR (Miyazaki and Esser, 2009). Despite the well-studied roles of DDIT4L in the pathogenesis of certain diseases (Cuaz-Perolin et al., 2004; Imen et al., 2009), virtually nothing was known about its ovarian expression and function. Nevertheless, we demonstrated here that Ddit4l, but not its closely related paralog Ddit4 (Corradetti et al., 2005), is differentially expressed in mouse ovarian large antral follicles before the luteinizing hormone surge, with its mRNA and protein expressed strongly in mural granulosa cells lining the follicle wall but not in the peri-antral granulosa cells or cumulus cells that are juxtaposed to the oocyte and follicular antrum. Notably, some DDIT4L protein was also detected in oocytes despite the presence of relatively lower levels of Ddit4l mRNA. This suggests that the expression of DDIT4L protein in oocytes might be subjected to different control mechanisms – it might be either actively translated or be stabilized immediately after translation. In striking contrast to the pattern of Ddit4l expression, p-RPS6 was predominantly found in cumulus cells and peri-antral granulosa cells but not in the mural granulosa cells. Given that DDIT4L is reported to be a negative regulator of the MTOR pathway in other cell types, and that Ddit4l expression and MTOR activation were found here to be mutually exclusive in granulosa cells, it is conceivable that DDIT4L expression contributes to the regulation of MTOR signaling in follicles. This possibility was supported by our observation that Ddit4l knockdown in cumulus cells significantly enhanced the activation of MTOR pathway. Therefore, before the luteinizing hormone surge, lower levels of Ddit4l in cumulus cells and peri-antral granulosa cells result in high levels of p-RPS6, whereas higher levels of Ddit4l in mural granulosa cells attenuate the levels of p-RPS6.

DDIT4L is upregulated under hypoxia (Shoshani et al., 2002). Although there is some controversy whether the microenvironment within antral follicles is hypoxic, it is generally accepted that the absence of a vasculature within the follicular basal lamina results in a poorly oxygenated environment for oocytes and granulosa cells before the luteinizing hormone surge (Neeman et al., 1997; Thompson et al., 2015). Therefore, low oxygen could be an initiating factor that promotes Ddit4l expression in mural granulosa cells, although other factors, particularly FSH, might be also involved. Interestingly, oxygen is also predicted to be lower in the vicinity of COCs (Clark and Stokes, 2011); and cumulus and peri-antral granulosa cells express higher levels of vascular endothelial growth factor (VEGF), a known effector of hypoxic stress (Shweiki et al., 1993; Wigglesworth et al., 2015). Cumulus and peri-antral granulosa cells should, in theory, express more Ddit4l; however, they were observed here unexpectedly to express much lower levels of Ddit4l and to possess higher levels of p-RPS6. These data suggest that there must be additional key factors restricting Ddit4l expression in cumulus cells to facilitate the activation of the MTOR pathway. Indeed, our data indicate that ODPFs, specifically GDF9 and GDF9–BMP15 heterodimer, suppress the expression of Ddit4l and enable MTOR activation in cumulus cells. Therefore, oocytes control the differential expression of Ddit4l and activation of the MTOR pathway in granulosa cells before the luteinizing hormone surge.

Cumulus cells surround the oocytes intimately, and along with peri-antral granulosa cells, also directly contact the follicular fluid, they thus have early access to GDF9 and GDF9–BMP15 heterodimer diffusing from the oocyte, as well as to other ODPFs moving into the follicular fluid. In contrast, mural granulosa cells are sequestered from follicular fluid by the adjacent peri-antral granulosa layers and therefore have limited exposure to ODPFs. This spatial difference between mural granulosa cells and cumulus cells in their relative proximity to the oocyte and follicular fluid might be the key for oocytes to drive the differential expression of Ddit4l and activation of MTOR pathway within antral follicles. A similar mechanism has been proposed for oocyte control of the differential expression of many other functionally important genes within antral follicles (Eppig et al., 1997; Eppig et al., 2005; Su et al., 2008; Sugiura et al., 2005) and to direct the global patterns of gene expression in granulosa cells (Diaz et al., 2007b; Emori et al., 2013; Wigglesworth et al., 2015). Although the actual concentration of GDF9 in follicular fluid is unlikely to be as high as the dose (500 ng/ml) that we have observed here to be effective in regulating Ddit4l expression and MTOR activation, genetic studies clearly show that both GDF9 and BMP15 are present in follicular fluid and function in synergy (Su et al., 2004). Interestingly, GDF9 and BMP15 have recently been found to form a heterodimer that strongly stimulates cumulus cell functions in vitro (Mottershead et al., 2015; Peng et al., 2013; Wigglesworth et al., 2013). We also observed here that mouse recombinant GDF9–BMP15 heterodimer suppressed Ddit4l expression in oocytectomized cumulus cells even at 1 ng/ml, which was 500 times as potent as GDF9 homodimer. Given that the same heterodimer might also form in vivo, GDF9–BMP15 heterodimer that is present in follicular fluid, even at a very low concentration, could preferentially reach peri-antral and cumulus cells. In addition, as proposed in a recent study (Watson et al., 2012), the differential expression of certain types of heparan sulfate proteoglycan co-receptors for GDF9 and/or GDF9–BMP15 heterodimer in mural, peri-antral and cumulus granulosa cells could facilitate the cell-type specific regulatory functions of GDF9 and GDF9–BMP15 heterodimer. The same scenario might also account for the differential expression of Ddit4l and activation of MTOR in these cells.

We observed that even though oocytes could still mature to the metaphase stage of meiosis II (MII) after MTOR activity in COCs was inhibited with Torin1 for 24 h, their post-fertilization developmental competence was dramatically compromised. These data suggest that the oocyte-dependent activation of MTOR in cumulus cells is essential for maintaining oocyte developmental competence. Although completely disrupting the activation of MTOR complex 1 (MTORC1) in mouse oocytes through conditional knockout of Rptor (also known as Raptor), a specific and essential component of MTORC1, does not affect folliculogenesis and female fertility (Gorre et al., 2014), we still can not rule out the possibility that some of the effects of Torin1 on oocyte developmental competence in COCs observed here could be attributable to the direct inhibition of MTOR signaling in oocytes. The role of MTOR in oocytes awaits further investigation. Interestingly, exogenous ODPFs augment the developmental competence of oocytes in cultured COCs (Mottershead et al., 2015; Yeo et al., 2008), and the developmental competence of double-mutant mouse oocytes and human oocytes expressing low levels of GDF9 and/or BMP15 is low (Gode et al., 2011; Su et al., 2004; Wu et al., 2007). Therefore, oocytes influence their own development through an integrated ‘oocyte–granulosa cell regulatory loop’, and activation of MTOR in cumulus cells is an essential step in getting this loop fully connected and functioning.

We also observed that Torin1 dose dependently increased cumulus cell apoptosis in COCs. Hence, MTOR activation in cumulus cells is essential for COC survival. This anti-apoptosis role of the oocyte-dependent activation of MTOR in cumulus cells is consistent with a previous finding made by Hussein et al. (2005), which showed that oocytes suppress apoptosis in cumulus cells through specific ODPFs. Therefore, oocyte suppression of Ddit4l expression, thereby enabling MTOR activation in cumulus cells, is essential for maintaining the healthy vigor of COCs. Given the well-conserved role of MTOR in stimulating cell growth and proliferation, and the fact that cumulus cells are more proliferative than mural granulosa cells (Hirshfield, 1986), oocyte-dependent activation of MTOR in cumulus cells might also contribute to the promotion of cumulus cell proliferation. Indeed, oocytes are known to be able to promote cumulus cell proliferation (Vanderhyden et al., 1992).

Lastly, we revealed that the oocyte-dependent activation of MTOR in cumulus cells is crucial for maintaining the integrity of the transcriptome of COCs. Torin1 treatment caused significant alteration of the global profile of COC gene expression, and most strikingly, the expression of genes that was changed by Torin1 treatment in COCs was also found to be changed in a similar way in cumulus cells isolated from mice that had been primed with hCG for 6 h. In particular, the most dramatically upregulated genes included those that are essential for cumulus expansion – i.e. Has2, Ptgs2, Ptx3 and Tnfaip6 – and those encoding EGF-like growth factors – i.e. Areg, Ereg, and Btc. The expression of these genes is required for re-initiation of oocyte meiotic maturation and cumulus expansion, two normally integrated processes that are commonly referred to as COC maturation (Park et al., 2004; Shimada et al., 2006). Therefore, it is plausible that Torin1 treatment could induce phenotypic changes similar to those of COC maturation. However, because the culture system applied in this experiment (i.e. serum-free medium containing 5 μM milrinone that maintains oocyte meiotic arrest and does not support cumulus expansion) was not optimal for assessing COC maturation, we did not observe evident changes in these parameters. Further studies are needed to resolve this issue. Because COC maturation only takes place in follicles after the luteinizing hormone surge, upregulation of the genes essential for COC maturation through MTOR inhibition suggests that the oocyte-dependent activation of MTOR in cumulus cells prevents the precocious onset of COC maturation before the luteinizing hormone surge. Therefore, fine-tuned MTOR signaling in cumulus cells is crucial for synchronizing COC maturation and ovulation. Downregulation of a specific set of metabolism-associated transcripts, especially those encoding the essential enzymes for cholesterol biosynthesis, might account for the greatly reduced oocyte developmental competence and cumulus survival. This is because oocyte and preimplantation embryos up to the blastocyst stage are deficient in producing cholesterol and/or other sterols, and require the surrounding cumulus cells to synthesize and provide cholesterol to them in order to support preimplantation development (Comiskey and Warner, 2007; Pratt 1982; Su et al., 2008). Apart from these transcripts that are involved in metabolism, the downregulation of some other transcripts, such as those encoding important receptors – e.g. Fshr, Ar, Npr2 and Tgbr3 – might also contribute to the defects observed in Torin1-treated COCs. Of note, conditional deletion of Ar in granulosa cells results in decreased follicular health and fewer competent oocytes (Walters et al., 2012).

We propose that Ddit4l, a heretofore functionally unexplored gene in mammalian ovaries, is differentially expressed in mural granulosa cells and that the MTOR pathway is selectively activated in cumulus and peri-antral granulosa cells, within the large antral follicles before the luteinizing hormone surge. ODPFs, specifically GDF9 and/or GDF9–BMP15 heterodimer, regulate the differential expression of Ddit4l and activation of MTOR in granulosa cells. The oocyte-dependent activation of MTOR signaling in cumulus cells encompasses a myriad of aspects of the metabolic and nutritional processes – key elements of the oocyte-cumulus cell regulatory loop – that promote the health and function of the cumulus cells and, in turn, the oocytes themselves.

Unless otherwise specified, all chemicals and reagents were purchased from Sigma-Aldrich.

Female 22- to 24-day-old B6SJLF1 mice, as well as Bmp15^−/−, double-mutant and the wild-type littermate control mice on a B6129 mixed genetic background, were used in experiments. Mice were raised under the standard conditions at the investigators' colonies at the Jackson Laboratory and/or the Research Animal Center at Nanjing Medical University. All mouse procedures and protocols were approved by the Animal Care and Use Committee of each institution, and conducted in accordance with the institutional guides for the care and use of laboratory animals.

Female mice were primed with 5 IU equine chorionic gonadotropin (eCG; EMD Biosciences) for 46 h to stimulate follicular development, and COCs, cumulus cells and oocytes were then isolated as described previously (Su et al., 2008). Oocytectomized cumulus cells were prepared through microsurgical removal of oocytes from COCs, as described previously (Buccione et al., 1990). COCs, oocytectomized cumulus cells and oocytectomized cumulus cells plus oocytes (two oocytes/μl of medium) were cultured in a drop of medium covered with washed mineral oil at a density of one COC or oocytectomized cumulus cells/μl of medium. The culture medium used in all experiments was minimum essential medium α (MEMα) that had been prepared exactly as described previously (Su et al., 2008). Cells were cultured for various times according to each experimental design and then collected for mRNA or protein expression analyses.

To test the effects of GDF9 and GDF9–BMP15 heterodimer on the expression of Ddit4l mRNA and phosphorylation of MTOR, RPS6KB1 and EIF4EBP1 protein in cumulus cells, oocytectomized cumulus cells were treated with 100- or 500-ng/ml recombinant mouse GDF9 (catalog no. 739-G9-010, R&D Systems) and/or various doses (0, 0.35, 1, 3.5 ng/ml) of recombinant mouse GDF9–BMP15 heterodimer (a gift from Prof. Martin M. Matzuk, Baylor College of Medicine, TX) and cultured for 20 h. To test the effect of inhibitors for SMAD 2 and/or SMAD3 on the expression of Ddit4l mRNA in cumulus cells of COCs, COCs were cultured in medium supplemented with or without 10 μM SB431542 or 20 μM SIS3 (EMD Biosciences) for 20 h. To test the effect of the same SMAD inhibitors on GDF9-controlled expression of Ddit4l mRNA, oocytectomized COCs cells were first treated with 10 μM SB431542 or 20 μM SIS3 for 1 h, and then treated together with 500 ng/ml GDF9 for another 19 h. The doses of SB431542 and SIS3 were chosen based on our previous studies (Diaz et al., 2007b; Su et al., 2010). SB431542 and SIS3 were dissolved in DMSO at stock concentrations of 10 and 20 mM, respectively. To study the effect of MTOR inhibition on COCs, COCs were treated with various concentrations (0, 1, 5, 10 μM) of Torin1 and cultured for either 14 or 24 h depending on the specific experimental designs. Torin1 was dissolved in DMSO at a stock concentration of 10 mM. Groups treated with only 10 μM DMSO served as controls.

Cumulus cells were stripped from intact COCs by repeated mouth-controlled pipetting, and seeded into the wells of a 96-well plate at a density of 50 COCs/150 μl/well. Cumulus cells were firstly cultured in 10%-FBS-containing medium for 24 h to form monolayers. Then the medium was changed into Opti-MEM containing no antibiotics, and the cells were transfected with Ddit4l-specific siRNAs (a mixture of three siRNAs targeting different regions of Ddit4l mRNA) or control siRNA at final concentrations of 40 nM for 24 h. Then, Opti-MEM was replaced by MEMα medium containing 10% FBS and antibiotics, and cells were further cultured for another 24 h. At the end of culture, cells were collected for both western blot and real-time RT-PCR analyses. Ddit4l siRNA and control siRNA were purchased from GenePharma company (Beijing, China). Their sequences were: Ddit4l-mus-392: 5′-CCUCAACGAGGUGGUAUUUTT-3′ and 5′-AAAUACCACCUCGUUGAGGTT-3′; Ddit4l-mus-460: 5′-CCAGAUCAAAGCAAACCAATT-3′ and 5′-UUGGUUUGCUUUGAUCUGGTT-3′; Ddit4l-mus-626: 5′-GGAUAGGAUCGUGUGUGAUTT-3′ and 5′-AUCACACACGAUCCUAUCCTT-3′. Control siRNA: 5′-UUCUCCGAACGUGUCACGUTT-3′ and 5′-ACGUGACACGUUCGGAGAATT-3′.

After being treated with various doses (0, 1, 5, 10 μM) of Torin1 in medium containing 5 μM milrinone for 24 h, COCs were washed with medium containing 10% FBS but without milrinone three times and were then cultured in the same medium and treated with 10 ng/ml of EGF for 14 h to induce maturation. After maturation, cumulus cells were removed by treating COCs with 100 ng/ml of hyaluronidase, and oocytes with extruded first polar bodies were counted and inseminated with normal sperm, as described previously (Su et al., 2004). The formation of two-cell-stage embryos was scored 24 h after in vitro fertilization. Thereafter, two-cell-stage embryos were cultured in KSOM medium (EMD Millipore Corporation, Billerica, MA), and the number of blastocysts was counted 3–5 days later.

To examine the effect of MTOR inhibition on transcriptomic changes in cumulus cells, COCs were cultured in medium containing milrinone (5 μM) and treated with 5 μM Torin1 or an equal concentration of DMSO for 14 h, and then were collected in RLT lysis buffer for RNA extraction. To compare global changes of gene expression in cumulus cells after hCG administration in vivo, cumulus cells were isolated from large antral follicles of mice at the time of hCG injection (designated as hCG 0 h) and at 6 h after hCG injection (designated as hCG 6 h). Three sets of samples were collected for each experiment, with each sample containing 100 COCs or cumulus cells from 100 COCs. Total RNA extraction, quantification and quality evaluation, sample amplification, labeling and hybridization, as well as data acquisition and analysis were performed as described previously (Su et al., 2008). Pairwise comparison was used to resolve the statistical differences of gene expression between groups with different treatments. Differentially expressed genes are defined at a false discovery rate (FDR) q-value threshold of 0.05. Molecule Annotation System (MAS) software (http://bioinfo.capitalbio.com/mas3/) was used to identify the enriched pathways and gene ontology (GO) terms associated with the changed transcripts.

Total RNA extraction, in vitro transcription and real-time PCR analyses were performed as described previously (Su et al., 2008). The primer sequences for the tested genes are listed in Table S1. The relative fold change in mRNA levels was calculated using the method of 2^−ΔΔCt, as described previously (Su et al., 2007), and is presented as relative changes to a specific group (control) in which the expression level of the mRNA of interest was set at 1. Rpl19 served as an internal control.

Protein sample preparation, SDS-PAGE gel electrophoresis and transfer, and the subsequent immunodetection were carried out following the exact procedures as described previously (Su et al., 2001). The primary antibodies used were against p-MTOR (phosphorylated at Ser2448; clone D9C2; XP^® rabbit monoclonal; catalog no. 5536; 1:1000), p-RPS6 (phosphorylated at Ser235 and Ser236; catalog no. 2211; 1:1000), p-4E-BP1 (phosphorylated at Thr37 and Thr46; clone 236B4; rabbit monoclonal; catalog no. 2855; 1:1000) and phosphorylated p70 S6 Kinase (phosphorylated at Thr389; catalog no. 9205; 1:1000). All these antibodies were purchased from Cell Signaling Technology. The expression of β-actin (ACTB) detected by anti-β-actin (ACTB) antibody (clone AC-15; mouse monoclonal; catalog no. A1978; 1:5000) was used as internal control of each sample.

Female mice were primed with eCG for 46 h to stimulate antral follicle development. Ovaries were then isolated and fixed in 4% paraformaldehyde solution for 4 h. The fixed ovaries were then embedded in paraffin and sectioned at 5-μm thickness for in situ hybridization, immunohistochemistry and immunofluorescence analyses. In situ hybridization was performed as described previously (Eppig et al., 2002) using the ³³P-labeled anti-sense riboprobe spanning nucleotides 571–1480 of the ovarian Ddit4l mRNA sequence. Immunohistochemistry and immunofluorescence detection of DDIT4L and/or p-RPS6 proteins was performed following the exact procedure as described previously by Diaz et al. (2007b). The VECTASTAIN ABC-AP KIT (Rabbit IgG, catalog no. AK-5001) and VECTOR Red Alkaline Phosphatase (AP) Substrate Kit (catalog no. AK-5100) purchased from Vector Laboratories were used for immunohistochemistry staining. Primary antibodies used were: anti-DDIT4L polyclonal antibody (catalog number 12094, Proteintech Group; 1:50 dilution); anti-p-RPS6 (phosphorylated at Ser235 and Ser236) antibody (catalog no. 2211, Cell Signaling Technology; 1:200 dilution for immunohistochemistry, 1:100 dilution for immunofluorescence). For immunofluorescence detection of p-RPS6, the Alexa-Fluor-594 donkey anti-rabbit IgG was used as the secondary antibody.

Apoptosis was detected using the TMR red in situ Cell Death Detection Kit (catalog no. 12156792910, Roche Diagnostics GmbH) according to the manufacturer's instructions. Images were taken under the LSM 710 confocal laser scanning microscope (Zeiss). Z-stack images were recorded at 0.2-μm intervals over the entire depth of the nuclei using the same image acquisition settings for controls and inhibitor-treated groups, and the maximum projection images were computed. The fluorescence intensity of the projected images was calculated using the ImageJ software (National Institutes of Health) and was presented as levels relative to those of the control group.

All experiments were repeated at least three times, and data are presented as mean±s.e.m. In experiments with more than two treatments, differences between groups were calculated by one-way ANOVA followed by Tukey's honest significant difference test (HSD) test using JMP software (SAS Institute, Cary, NC), whereas for experiments with only two treatments, differences were analyzed with a t-test using Microsoft Excel. P<0.05 was defined as a significant difference.

We thank Prof. Martin M. Matzuk at Baylor College of Medicine for providing the Bmp15^−/− and Gdf9^+/−Bmp15^−/− mice, and the recombinant mouse GDF9–BMP15 heterodimer.