FOXO1 mediates hypoxia-induced G0/G1 arrest in ovarian somatic granulosa cells by activating the TP53INP1-p53-CDKN1A pathway

Ovarian functions in mammals are mediated through a specialized compartment called the ovarian follicle, comprising a single oocyte surrounded by multiple layers of somatic granulosa cells (GCs). In developing follicles, GCs proliferate rapidly to support follicular growth, maturation and ovulation (Gougeon, 1996). The mitogenic activity of GCs is regulated, at least in part, by hormones secreted from the pituitary and neighbouring cells (Robker and Richards, 1998). In addition, several intraovarian growth factors, such as insulin-like growth factor-I (IGF-I), epidermal growth factor (EGF), transforming growth factor-beta (TGFβ) and fibroblast growth factor (FGF), can affect GC proliferation (Vanderhyden et al., 1992). Although significant progress has been made in recent years, the mechanisms involved in controlling the GC cell cycle are not fully understood.

The ovarian follicular vascular system directly supplies the nutrition, oxygen and hormones that are required for follicle development (Brown and Russell, 2014). However, angiogenesis is confined to the theca cells outside the basement membrane, a physical barrier that blocks the capillary network from penetrating into the interior of the follicle, thus inhibiting the transport of O₂ to the GC layer. During follicular development, the proliferation of GCs and thickening of the GC compartment further increases the distance from GCs to the follicular vasculature, leading to a progressively deficient supply of O₂ (Bianco et al., 2005; Redmer and Reynolds, 1996; Suzuki et al., 1998). Therefore, it is commonly accepted that GCs exist in a hypoxic environment within ovarian follicles (Nishimura and Okuda, 2015).

In recent years, evidence has accumulated that cells can respond to varying O₂ levels by altering their rates of proliferation (Ortmann et al., 2014). In fact, many different cell types display cell cycle arrest and proliferative retardation during hypoxia (Ortmann et al., 2014). Conversely, rapid proliferation of cells accelerates O₂ consumption, leading to a reduction in cellular O₂ tension (Ortmann et al., 2014). Indeed, regular follicular development is accompanied by a high proliferation rate of GCs. However, Zhang et al. (2018), reported that the expression of some mitosis-related genes in GCs tends to decrease with follicular growth, implicating a possible reduction of GC proliferation activity during folliculogenesis. Interestingly, transcriptome sequencing in porcine large antral follicles revealed gene pathways associated with hypoxia (Munakata et al., 2016). Moreover, our earlier study showed that cobalt chloride (CoCl₂), a hypoxia-mimicking agent, could inhibit cell cycle progression in cultured GCs (Li et al., 2020). This prompted us to investigate whether hypoxia is actually involved in the regulation of the cell cycle in ovarian GCs during folliculogenesis.

Cellular proliferation is controlled by molecules that positively or negatively affect transitions through the checkpoints of the cell cycle. Briefly, entry into the cell cycle is positively regulated by a family of protein kinases referred to as cyclin-dependent kinases (CDKs) (Ortmann et al., 2014). By contrast, CDK inhibitors (CKIs) are negative regulators that hinder the cell cycle progression. CKIs, such as CDKN1A (cyclin-dependent kinase inhibitor 1A), function by binding to CDKs and inhibiting their interaction with cyclins, resulting in cellular growth arrest (Ortmann et al., 2014). The expression of CDKN1A is mainly controlled at the transcriptional level by the transcription factor p53 (Li et al., 1994). On activation by p53 during a variety of stresses, CDKN1A directly inhibits the formation of cyclin E-CDK2 complexes, leading to p53-mediated G1-stage cell cycle arrest (Harper et al., 1995). As one type of stress, hypoxia has been suggested to induce cell cycle arrest at G0/G1 phase and to block G1-to-S transition (Gardner et al., 2001). According to our previous studies, chemical-simulated hypoxia in primary cultured GCs could arrest cells at the G1/S boundary (Li et al., 2020). However, current evidence does not indicate whether and how real hypoxia acts on regulating GC cell cycle in vivo.

Here, using high-throughput sequencing technology, we demonstrated hypoxia as a primary component within developing porcine follicles that prevents cell cycle progression of GCs. Our findings suggest an inhibitory role for hypoxia in regulating the G1-S transition by activating the JNK-14-3-3-FOXO1-TP53INP1-p53-CDKN1A signalling axis in ovarian GCs.

Using flow cytometry, we determined the cell cycle distribution of GCs collected from porcine ovarian antral follicles. As shown in Fig. 1A-C, the proportion of GCs at G0/G1 phase tended to increase, whereas the S-and-G2/M phase population decreased with the growing sizes of the follicles, indicating an impaired GC proliferation activity during follicular development. Because follicular growth is supposed to cause hypoxia in GCs (Bianco et al., 2005; Redmer and Reynolds, 1996; Suzuki et al., 1998), it raises the possibility that hypoxia might influence GC proliferation. Interestingly, individual follicles of the same diameter also showed large variations in G0/G1 distribution of GCs (Fig. 1A-D). To investigate whether hypoxia or other factors might contribute to GC cell cycle asynchrony among similar-sized follicles, GCs retrieved from developing follicles around 4 mm in diameter were classified into groups with high G0/G1 distribution (high groups) and low G0/G1 distribution (low groups), which were then subjected to transcriptomic profiling by RNA sequencing (RNA-seq). We found that the transcriptome of the high groups was significantly different from that of the low groups by showing that 638 transcripts were downregulated and 686 transcripts were upregulated in the former (Fig. 1E,F). The expression of ten randomly selected transcripts were verified using quantitative RT-PCR (qRT-PCR), indicating the reliability of the RNA-seq data (Fig. S1). Gene ontology (GO) annotation was performed to assign a functional classification to these differentially expressed genes (DEGs). The top 15 significantly enriched GO terms of the DEGs for each category are shown in Fig. 1G,H. For downregulated biological processes, the DEGs were mainly assigned to cell cycle-related functions (Fig. 1G), implicating that our sampling procedures were specific. Notably, response to hypoxia was enriched at the top list for the upregulated biological process (Fig. 1H), consistent with our above-mentioned speculation regarding the effects of hypoxia on GCs cell cycle. To explore the biochemical pathways further, all DEGs were mapped to terms in the Kyoto Encyclopedia of Genes and Genomes (KEGG) database (https://www.kegg.jp; Fig. 1I). Among the top 20 enriched pathways, we paid greater attention to FOXO1 signalling, because our previous studies suggested FOXO1 to be a crucial mediator of the stress response in ovarian GCs (Li et al., 2020; Shen et al., 2018, 2017, 2012, 2014), and FOXO1 is the only FOXO family member present in our transcriptomic data.

To validate the possible involvement of hypoxia and FOXO1 in modulating GC cell cycle, experiments were performed in porcine ovarian GCs harvested from growing follicles at various stages, including small antral follicles (1-3 mm), medium antral follicles (3-5 mm) and large antral follicles (≥5 mm). As a transcription factor, FOXO1 functions by entering the cell nucleus (Shen et al., 2012). For medium and large antral follicles, increased nuclear distribution of FOXO1 was observed in those with a higher prevalence of GCs in G0/G1 phase (Fig. S2F,K). Linear regression analysis showed that the fraction of the population arrested at G0/G1 was positively correlated with FOXO1 expression in the nuclei (Fig. S2G,L). To test whether hypoxia might be relevant to regulation of the GC cell cycle in vivo, the protein level of hypoxia inducible factor-1alpha (HIF-1α), a key endogenous marker of hypoxia, was determined in ovarian GCs. As shown in Fig. S2H,I,M,N, elevated proportions of GCs at G0/G1 correlated strongly with a progressively more hypoxic status as reflected by the accumulation of cellular HIF-1α. Moreover, a significant positive correlation was determined between nuclear distribution of FOXO1 and HIF-1α accumulation in ovarian GCs as mentioned above (Fig. S2J,O). However, no significant correlation was detected among intranuclear FOXO1 protein levels, HIF-1α abundance, and cell cycle distribution in GCs collected from small antral follicles (Fig. S2A-E). Therefore, our in vivo findings indicate a potential interaction between hypoxia and FOXO1 signalling in controlling the cell cycle transition of GCs in medium and large antral follicles.

To clarify further the relationship between hypoxia and GC cell cycle without interference from other follicular factors, we applied hypoxia treatment to primary cultured GCs harvested from porcine antral follicles. As shown in Fig. 2A,B, hypoxia significantly increased the percentage of cells in G0/G1 phases, but reduced the fraction in S phase, implicating an inhibitory effect of hypoxia on G1-to-S transition in GCs. To assess whether FOXO1 might be involved in this process, we measured the subcellular localization of FOXO1 in hypoxia-treated GCs. As shown in Fig. 2C,D, hypoxia facilitated nuclear translocation of FOXO1, as suggested by the increased percentage of cells with FOXO1 in the nucleus. To examine whether FOXO1 might be required for hypoxia-mediated G0/G1 cell cycle arrest, we blocked FOXO1 expression using RNA interference. Six siRNAs against FOXO1 were first screened for their potency and specificity. Two out of these candidates were identified as the most potent and specific siRNAs in silencing FOXO1 and were used for subsequent experiments (Fig. S3A). As shown in Fig. 2E,F, knocking down FOXO1 expression markedly restored the proportion of GCs in the S-phase population by driving cells out of the G0/G1 phase. Collectively, these data suggest that FOXO1 plays a pivotal role in the hypoxia-induced G0/G1 arrest of GCs.

Among the stress-responsive kinases, SAPK/JNK, p44/42 MAPK and p38 MAPK are known to target FOXOs and regulate their activity through a cytoplasmic-nuclear shuttling mode (Klotz et al., 2015). To explore further the mechanism of FOXO1 regulation under hypoxia, we examined whether these pathways would affect FOXO1 localization and cell cycle progression in primary cultured GCs. As shown in Fig. 3A-C and Fig. S4A-C, unregulated JNK activity, as reflected by elevated phosphorylation status at Thr183/Tyr185, correlated strongly with the increased nuclear distribution of FOXO1 in GCs with prolonged hypoxic treatment (Pearson correlation test; P<0.01). However, p44/42 MAPK and p38 MAPK presented no evident response to hypoxia. We also determined the possible involvement of AKT, a major upstream suppressor of FOXO1 activity (Klotz et al., 2015). Paradoxically, the active form of AKT, namely p-AKT (phosphorylated at Ser473), showed elevated expression during hypoxic exposure (Fig. 3A), suggesting that AKT might be irrelevant to hypoxia-mediated FOXO1 activation in GCs. To verify the impact of JNK on FOXO1 subcellular localization, we blocked JNK activity using the inhibitor SP600125. Immunofluorescence staining indicated that the nuclear translocation of FOXO1 upon hypoxia stimulation was inhibited after SP600125 treatment (Fig. 3D,E). Consistent with this, in GCs subjected to hypoxia the JNK inhibitor significantly increased the proportion of cells in S phase, but reduced the population in G0/G1 phases, implicating a recovery of the G1-to-S transition (Fig. 3F,G).

JNK has been documented to induce the phosphorylation of 14-3-3β and ζ at Ser186 and 184, respectively, thereby preventing the association of 14-3-3 with its target proteins (Sunayama et al., 2005). As the formation of the FOXO1-14-3-3 complex is required for the cytoplasmic transport of FOXO1 in mammalian cells (Huang and Tindall, 2007), we wondered whether hypoxia-induced activation of JNK would affect the binding between 14-3-3 and FOXO1. To test this assumption, GCs exposed to hypoxia were harvested for immunoprecipitation with an antibody against FOXO1. As shown in Fig. 3H and Fig. S4D, 14-3-3 was co-precipitated by the FOXO1 antibody. The binding affinity of 14-3-3 to FOXO1 was remarkably decreased after incubation with 1% O₂, along with an elevated level of phosphorylated 14-3-3 under hypoxic conditions. In addition, blocking of JNK activity augmented hypoxia-induced phosphorylation of 14-3-3, followed by a restored association between 14-3-3 and FOXO1 in hypoxic GCs (Fig. 3I, Fig. S4E). Thus, we conclude that dissociation of the FOXO1-14-3-3 complex caused by 14-3-3 phosphorylation through the JNK pathway might be an essential step in hypoxia-mediated suppression of the G1-S transition in ovarian GCs.

Growing evidence indicates that FOXOs can regulate transcription directly through consensus-binding motifs or indirectly via partner proteins (Huang and Tindall, 2007; Ramaswamy et al., 2002). To investigate further how FOXO1 functions in the nucleus, GCs cultured under hypoxia were transfected with vectors encoding FOXO1, including wild-type (WT) FOXO1 (FOXO1-WT) and FOXO1-DBD/FOXO1^N208A,H212R (containing two point mutations in the DNA-binding domain), after knocking down endogenous FOXO1 (Fig. S5A). Western blot analysis demonstrated a marked increase in FOXO1 levels in GCs with enforced expression of FOXO1 (Fig. S5A). RNA interference showed no apparent influence on the expression of Flag-tagged FOXO1 detected in both the FOXO1-WT and FOXO1-DBD groups, enabling us to specify the effects of exogenous FOXO1 (Fig. S5A). By performing flow cytometric assay, we found that overexpression of FOXO1-WT restored hypoxia-induced G0/G1 arrest in GCs pretreated with FOXO1 siRNA (Fig. S5B,C). By contrast, the FOXO1-DBD mutant showed no evident effects on this process (Fig. S5B,C). These results suggest that binding to the recognition motifs in target genes might be required for FOXO1-induced G0/G1 arrest in GCs exposed to hypoxia.

To identify the downstream effectors of FOXO1 signalling, we blocked FOXO1 activity using RNAi, and then determined the expression of checkpoint factors required for G1-S progression in hypoxia-treated GCs. qRT-PCR assay revealed that, among the tested transcripts, the expression patterns of TP53INP1 and CDKN1A were significantly correlated with that of FOXO1 (P<0.01) in response to hypoxia and FOXO1 siRNA treatment (Fig. S6A). Consistent results were obtained by western blot analysis (Fig. S6B). As shown in Fig. S6B, knockdown of FOXO1 abrogated hypoxia-upregulated expression of TP53INP1 and CDKN1A. In contrast, the protein levels of TP53INP1 and CDKN1A inhibited by FOXO1 silencing were restored after overexpression of FOXO1-WT, whereas the DBD mutant had no effect (Fig. S6C). These data thus suggest TP53INP1 and CDKN1A as potential downstream target genes of FOXO1 in regulating GC cell cycle during hypoxia exposure.

In fact, the above-mentioned results of CDKN1A expression in hypoxia-treated GCs are expected, as CDKN1A has been well described as a transcriptional target of FOXOs (Huang and Tindall, 2007). Compared with CDKN1A, there is little evidence regarding whether and how FOXO1 could directly regulate TP53INP1 expression. However, our RNA-seq data indicated TP53INP1 as a significantly upregulated DEG in ovarian GCs with high G0/G1 distribution (Table 1, Table S5), consistent with the qRT-PCR results as shown in Fig. S6. To test whether FOXO1 would directly activate the expression of TP53INP1 at the promoter level, a 2072-bp fragment of the TP53INP1 5′ untranslated region (UTR) containing the putative FOXO-recognized element (FRE) site or mutated FRE sites was cloned upstream of a luciferase reporter gene as pGL3-TP53INP1 (WT), pGL3-TP53INP1 (M1), pGL3-TP53INP1 (M2) and pGL3-TP53INP1 (M3), respectively (Fig. 4A). These constructs were then co-transfected with pRL-TK and pCMV5-FLAG-FOXO1 (FOXO1-WT) or the control pCMV5 into GCs. As shown in Fig. 4B, enforced expression of the FOXO1-WT vector significantly increased TP53INP1 promoter activity. In contrast, FRE site mutations markedly reduced the responsiveness of the luciferase reporter to FOXO1 overexpression (Fig. 4B). To confirm whether FOXO1 binds to TP53INP1 promoter, we next performed chromatin immunoprecipitation (ChIP) assays. GC chromatin was first digested into fragments of 250-700 bp with micrococcal nuclease (Fig. S7) and then used for ChIP experiments. As shown in Fig. 4C-E, the FOXO1 antibody precipitated the TP53INP1 promoter region around the FRE site. Notably, the signals were significantly enhanced in GCs with hypoxia exposure. Therefore, these data suggest that FOXO1 was recruited directly to the TP53INP1 promoter for its transcriptional activation in GCs upon exposure to hypoxia.

To test whether TP53INP1 is required for FOXO1-mediated G0/G1 arrest, we transfected GCs with siRNAs against FOXO1 and/or TP53INP1 (Fig. S3B). As shown in Fig. 4F,G, blocking the expression of FOXO1 and TP53INP1 both facilitated the passage from G1 to S phase in hypoxic GCs, but FOXO1 siRNA failed to further promote G1-S transition when TP53INP1 was transcriptionally silenced. In addition, hypoxia-induced CDKN1A expression was remarkably reduced after knocking down the expressions of FOXO1 and/or TP53INP1 (Fig. 4H, Fig. S8). Thus, these data confirm that FOXO1 induces TP53INP1-dependent G0/G1 arrest in hypoxic GCs.

We next sought to identify the downstream effectors of TP53INP1-triggered G0/G1 arrest. The results presented in Fig. 4 indicated that TP53INP1 might be required for hypoxia-induced CDKN1A expression in GCs. Indeed, it has been well documented that CDKN1A is also transcriptionally regulated by p53 (Li et al., 1994). Of note, the KEGG analysis in Fig. 1H revealed p53 signalling to be one of the top 20 enriched pathways in ovarian GCs with high G0/G1 rates. This prompted us to test whether p53 might exert any effects on hypoxia-induced G0/G1 arrest of GCs. As shown in Fig. 5A and Fig. S9A, the increased level of CDKN1A during hypoxia was accompanied by an evident upregulation of p53 expression. To test whether p53 might be required for hypoxia-induced CDKN1A expression, GCs were treated with pifithrin-α, an inhibitor that blocks the transcription activity of p53. The results in Fig. 5B and Fig. S9B showed that the hypoxia-triggered expression of CDKN1A was significantly decreased in GCs treated with the p53 inhibitor. Interestingly, pifithrin-α treatment also inhibited TP53INP1 expression upon hypoxia stimulation. To clarify the relationship among p53, TP53INP1 and CDKN1A, we silenced TP53INP1 activity using RNAi, and then determined the expression of p53 and CDKN1A. As shown in Fig. 5C, Fig. S9C and Fig. 4F, blocking TP53INP1 markedly reduced the protein level of CDKN1A, but had little effect on p53 expression during hypoxia exposure. Therefore, these data indicate the presence of a p53-TP53INP1-CDKN1A axis in hypoxic GCs.

We then examined whether TP53INP1 might contribute to p53-mediated transcriptional activation of CDKN1A. By performing ChIP assay, we observed that the anti-p53 antibody bound to the consensus motif of the p53 binding site located within the CDKN1A promoter (Fig. 5D). Notably, knocking down the expression of TP53INP1 abolished the interaction between p53 and the CDKN1A promoter in hypoxia-exposed GCs (Fig. 5D-F). In accordance with this, inhibitors or siRNAs against p53 and TP53INP1 revealed the approximate level of suppression of CDKN1A expression (Fig. S10), thus suggesting TP53INP1 to be a crucial activator of p53 transcriptional function in hypoxic GCs.

We also determined whether the p53-CDKN1A axis might affect cell cycle progression of GCs exposed to hypoxia. Flow cytometric analysis showed that inhibition of p53 and CDKN1A both markedly restored the proportion of GCs in S phase by driving cells out of G0/G1 phase (Fig. 5G,H, Fig. S3C). Taken together, these data suggest that activation of the p53-CDKN1A axis by TP53INP1 might constitute a primary suppressive mechanism responsible for hypoxia-mediated G0/G1 arrest in GCs.

To test whether our model of cell cycle regulation would be applicable in vivo, GCs collected from developing follicles around 4 mm in diameter were classified into groups with a high G0/G1 distribution (high groups) or a low G0/G1 distribution (low groups). As expected, elevated JNK activity was associated with a more hypoxic status as reflected by the accumulation of HIF-1α in GCs with higher G0/G1 distribution (Fig. 6A, Fig. S11A). In addition, JNK activity was proportional to the intranuclear level of FOXO1, both of which were increased in the high groups (Fig. 6B, Fig. S11B). Moreover, GCs with higher G0/G1 distribution showed upregulated level of phosphorylated 14-3-3, along with an impaired binding affinity of 14-3-3 to FOXO1 (Fig. 6C). Of note, compared with the low groups, the JNK activity, 14-3-3 phosphorylation, and the binding status between 14-3-3 and FOXO1 showed similar trends of change in the high groups (Fig. S11C). These findings suggest the possible involvement of JNK-14-3-3-FOXO1 signalling in controlling GC proliferation under hypoxia in vivo.

We next determined the downstream effectors of FOXO1. As shown in Fig. 6D and Fig. S11D, the high groups of GCs showed increased TP53INP1 expression, accompanied by an elevated level of intranuclear FOXO1. Using ChIP-qPCR assays, we further observed increased binding of FOXO1 to the promoter of TP53INP1 in the high groups (Fig. 6E, Fig. S11E). Concomitantly, the expression of TP53INP1 was proportional to that of CDKN1A in the above-mentioned GC groups (Fig. 6F, Fig. S11F). Further investigation revealed that the upregulated expression of CDKN1A was accompanied by an enrichment of p53 at the promoter region of CDKN1A (Fig. 6G, Fig. S11G). Furthermore, TP53INP1 expression was proportional to the p53 binding level at the CDKN1A promoter (Fig. S11H). Using histological sections of porcine ovaries, we performed immunofluorescence staining to determine PCNA (reflecting cellular proliferation status) colocalization with TP53INP1 or CDKN1A in GCs within the same follicles (Fig. 6H). The results showed increased expression of both TP53INP1 and CDKN1A in inner layers of GCs, where the levels of PCNA tended to decrease in cells with higher abundance of TP53INP1 or CDKN1A. These data further confirm the relevance of the TP53INP1-CDKN1A axis to GC proliferation retardation at the single-cell level in vivo. Taken together, these results suggest that the FOXO1-TP53INP1-p53-CDKN1A axis might mediate hypoxia-induced G0/G1 arrest of ovarian GCs in vivo.

Various environmental stresses, including oxidative stress, heat stress, cold stress and restraint stress, have been suggested to negatively influence the development of ovarian follicles (Shen et al., 2012). Upon stress stimulation, an arrest or delay of cell cycle progression was observed in immortalized cells of human tumours and other mammalian cell lines (Douglas and Haddad, 2003). Growing evidence implicates hypoxia as one such stress that appears to affect cell cycle activity directly (Douglas and Haddad, 2003). In the ovary, GCs are believed to live in a hypoxic microenvironment inside the follicles (Nishimura and Okuda, 2015). During follicular development, the progressive deficiency in O₂ supply is believed to exacerbate the hypoxic conditions in GCs (Bianco et al., 2005; Redmer and Reynolds, 1996; Suzuki et al., 1998). Our earlier study revealed that chemically mimicked hypoxia caused cell cycle arrest in cultured GCs (Li et al., 2020). However, no definitive evidence exists regarding whether the hypoxia in ovarian follicles exerts any effects on cell cycle regulation of GCs in vivo. Here, we show for the first time that hypoxia might represent a principal inhibitory factor in the follicle that blocks the cell cycle transition in ovarian GCs.

The follicular antrum is filled with fluid that absorbs nutrients, hormones and O₂ from the capillary network outside the basement membrane (Brown and Russell, 2014). In particular, the O₂ supplied by blood vessels is transported into GCs by diffusion (Hirshfield, 1991). Predictably, the O₂ content in GCs should be proportional to the diameter of follicles. However, we found that GCs from follicles with similar sizes also showed large variations in hypoxic status as reflected by variations in the accumulation of HIF-1α, a key indicator of cellular hypoxia (Fig. S2). This intriguing phenomenon suggests a possible asynchrony between follicular growth and local angiogenesis. However, the current study discovered that hypoxia is responsible for G0/G1 arrest in ovarian GCs. As the progression from G1 to S phase is a rate-limiting step of cell proliferation (Bertoli et al., 2013), our data indicate that GC proliferation is tightly controlled by the O₂ tension within ovarian follicles. Previous studies have demonstrated a close relationship between GC proliferation and follicular development/atresia (Hirshfield, 1991). Under normal physiological conditions, most follicles become atretic during the 8th or 9th generations of GCs, a time point corresponding to the transition of the growing follicles into mature follicles. At this stage, the GCs have exhausted their proliferative potential, and the follicles do not continue to enlarge. These developmentally arrested follicles will finally be degraded by atresia (Hirshfield, 1991). Therefore, it is conceivable that the asynchronous development of follicles and follicular vasculature interrupts the timely transportation of blood O₂ into the follicular cavity, leading to impaired GC proliferation via cell cycle arrest, which might also elicit selective atresia of certain ovarian follicles.

FOXO1 is a multifunctional transcription factor that plays essential roles in diverse cellular and physiological processes, including cell cycle arrest, proliferation, differentiation, apoptosis, metabolism and aging (Huang and Tindall, 2007). In GCs in particular, FOXO1 has been demonstrated as a key regulator of the stress response (Li et al., 2020; Shen et al., 2018, 2017, 2012, 2014). We previously discovered that a ‘pseudo-hypoxia’ status mimicked by CoCl₂ treatment could induce FOXO1-dependent cell cycle arrest in cultured GCs (Li et al., 2020). Here, using high-throughput sequencing technology, we further demonstrated that FOXO1 signalling acts as a crucial mediator of cell cycle inhibition induced by follicular hypoxia in ovarian GCs. Of note, FOXO1 is activated via different molecular mechanisms under ‘real hypoxia’ versus ‘false hypoxia’. Our previous work indicated that PI3K/AKT, a canonical inhibitory pathway of FOXO1, was suppressed during CoCl₂ treatment, resulting in the dephosphorylation and nuclear localization of FOXO1 in cultured GCs (Li et al., 2020). By striking contrast, our current data showed that AKT activity was actually increased under ‘real hypoxia’ conditions, but FOXO1 remained in the nuclei of hypoxic GCs. As reported, AKT-mediated phosphorylation of FOXO1 creates binding sites for 14-3-3 chaperone proteins, which promote the nuclear export of FOXO1 (Huang and Tindall, 2007). It thus raised the possibility that hypoxia might block the association between 14-3-3 and FOXO1. Besides AKT, MAPK-related proteins, including SAPK/JNK, p44/42 MAPK and p38 MAPK, are known to regulate FOXO activity upon stress stimuli (Klotz et al., 2015). In this study, we identified JNK as a major upstream activator of FOXO1 in hypoxic GCs. Importantly, JNK enhanced the nuclear translocation of FOXO1 by phosphorylating 14-3-3β/ζ at Ser184/186, which blocked its interaction with FOXO1. Thus, this result might provide an explanation of why AKT failed to influence the cytoplasmic-nuclear shuttling of FOXO1 during ‘real hypoxia’.

TP53INP1 (tumor protein 53-induced nuclear protein 1) is a stress-induced p53 target gene that plays essential roles in cell cycle arrest and p53-mediated apoptosis (Tomasini et al., 2003). Recent studies suggested that FOXO1 activation might increase the expression of TP53INP1 (Guan et al., 2015; Zhang et al., 2017), but there is little evidence regarding whether this gene is a direct transcription target of FOXO1. Our results showed that FOXO1 stimulated TP53INP1 expression and activity of its promoter, as well as causing G0/G1 arrest in GCs (Figs S6 and S7). In fact, FOXO1 was found to bind directly to the FRE site within the TP53INP1 promoter (Fig. 4). Under hypoxic conditions, nuclear transportation of FOXO1 increased its binding with the FRE sequence, which was correlated with upregulated expression of TP53INP1 and increased proportion of cells at G0/G1 (Fig. 3, Fig. S5, Fig. 4). Moreover, we found that CDKN1A, a well-described transcriptional target gene of FOXO1 (Huang and Tindall, 2007), was also transcriptionally regulated by TP53INP1. This thus provides evidence for a potential regulation mode of FOXO1-mediated CDKN1A transcription in a TP53INP1-dependent manner. Collectively, these data suggest that when GCs are exposed to hypoxia, FOXO1 activates the transcription of TP53INP1 and its downstream cell cycle inhibitors, such as CDKN1A, thereby triggering G0/G1 arrest in GCs.

Because TP53INP1 expression is regulated by both FOXO1 and p53, we speculate that knockdown of FOXO1 alone cannot completely inhibit the function of TP53INP1. However, we observed that blocking the expression of FOXO1 and TP53INP1 showed similar effects on restoring the passage from G1 to S phase in hypoxic GCs (Fig. 4), implicating FOXO1 as the only upstream activator of TP53INP1-mediated G0/G1 arrest. Paradoxically, p53 seems to be irrelevant to this process. In fact, we found that there is a bidirectional regulation between TP53INP1 and p53, i.e. p53 promotes the expression of TP53INP1, and TP53INP1 in turn activates p53 transcriptional activity, causing a further increase in TP53INP1 expression levels and p53 transcriptional activity, leading to upregulation of CDKN1A expression and the resultant suppression of G1-S transition (Fig. 5). Further investigations revealed TP53INP1 to be a major activator of p53 transcriptional function in hypoxic GCs (Fig. S10), indicating that p53 is actually a downstream effector of the FOXO1-TP53INP1 pathway. Therefore, we conclude that activation of the TP53INP1-p53-CDKN1A axis is required for FOXO1-induced G0/G1 arrest in hypoxic GCs.

In summary, this study provides the first evidence suggesting hypoxia as a key stimulus of GC proliferation retardation during follicular development. In addition, we revealed that the activation of FOXO1 by JNK-triggered 14-3-3 phosphorylation is responsible for hypoxia-induced G0/G1 arrest in ovarian GCs. Further investigations identified a novel transcriptional target gene of FOXO1, namely TP53INP1, which acts through TP53INP1-p53 positive feedback and p53-CDKN1A signalling to mediate FOXO1-dependent G0/G1 arrest upon hypoxia exposure (Fig. 6I). These findings reveal further insights into the functions of hypoxia in the regulation of folliculogenesis, and provide potential targets for the clinical treatment of reproductive disorders caused by impaired follicular development.

SP600125 (S1460) and pifithrin-α (S2929) were purchased from Selleck Chemicals. Antibodies against FOXO1 (2880), p-FOXO1 (9461), TUBA1A (2125), p44/42 MAPK (4695T), p-p44/42 MAPK (4370), Histone H3 (4499), 14-3-3 (8312), p53 (2527), HIF-1α (14179), phospho-SAPK/JNK (Thr183/Tyr185; 4668), SAPK/JNK (9252), AKT (9272), p-AKT (4060) and FLAG (2368) were obtained from Cell Signaling Technology. Antibodies against CDKN1A (10355-1-AP) and p38 MAPK (66234-1-Ig) were obtained from Proteintech. 14-3-3β/ζ (phospho Ser184/186; ABP56527) antibody was purchased from Abbkine. Antibodies against Phospho-p38 MAPK (Thr180/Tyr182; AF4001) and TP53INP1 (DF8731) were obtained from Affinity Biosciences.

Ovaries from mature Duroc-Yorkshire-Landrace sows were collected at a local slaughterhouse and subsequently transferred to the laboratory within 2 h. Ovaries without corpus luteum were selected to isolate morphologically healthy follicles with small forceps. Generally, healthy follicles are rounded with evenly distributed blood vessels; they display pink or yellow in appearance and have a visible cumulus-oocyte complex immersed in clear follicular fluid. After isolation, the follicles were subjected to size measurement, and then individually torn apart to obtain the mural GCs by scraping the follicular wall. GCs were immediately processed for the determination of cell cycle distribution, or restored for RNA-seq, qRT-PCR or immunoblotting analysis. The detailed procedure for the collection of porcine ovarian GCs is depicted as a schematic in Fig. S12. For cell culture, GCs isolated from antral follicles were centrifuged at 1000 g for 5 min, washed with PBS (Gibco), and then cultured in DMEM/F-12 (1:1) medium (Life Technologies) supplemented with 10% fetal bovine serum (FBS; Sigma-Aldrich) and 100 units/ml penicillin plus 100 μg/ml streptomycin (Gibco) for 2 days. In the control group, cells were maintained at 37°C in a 5% CO₂, 95% air incubator (20% O₂). In the hypoxic group, cells were maintained in a modular incubator chamber flushed with a gas mixture containing 1% O₂, 5% CO₂ and 94% N₂ at 37°C. For drug administration, GCs were treated with the JNK inhibitor SP600125 (10 μM) or the p53 inhibitor pifithrin-α (30 μM) for 2 h, and then incubated 1% O₂ for 12 h or 24 h. For RNA interference, GCs were transfected with FOXO1 siRNA, CDKN1A siRNA, TP53INP1 siRNA or scrambled control siRNA for 12 h, and then exposed to 20% O₂ or 1% O₂ for the indicated time. In some experiments, cells treated with FOXO1 siRNA for 12 h were transfected with the Flag-tagged FOXO1 plasmids or an empty control plasmid, and then cultured with 20% O₂ or 1% O₂ for 12 h or 24 h before they were used for the next assay.

GCs collected from antral follicles around 4 mm (4±0.05 mm) in diameter were classified into four groups with two high G0/G1 distribution groups (high groups) and two low G0/G1 distribution groups (low groups), which were then lysed with Trizol reagent (Invitrogen). The cDNA libraries were constructed for each RNA sample using the TruSeq Stranded mRNA Library Prep Kit (Illumina) following the manufacturer's protocol. The libraries were sequenced on an Illumina HiSeq X platform (Illumina) to obtain raw reads. After filtering out low-quality reads and removing the adaptor sequences, we aligned RNA-seq reads to pig genome (Sscrofa11.1, NCBI) using Hisat2 software, and calculated the fragments per kilobase of transcript per million (FPKM) of each gene using Cufflinks. DEGs and corresponding adjusted P-values were determined using DEseq2. The biological significance of DEGs was explored by GO term enrichment analysis based on Fisher's exact test and then KEGG pathway enrichment analysis of DEGs was also performed with Fisher's exact test.

Cytosolic and nuclear extracts were prepared according to the manufacturer's instructions of NE-PER Nuclear and Cytoplasmic Extraction Reagents (Thermo Fisher Scientific, 78833). Briefly, GCs were washed twice with ice-cold PBS buffer and centrifuged at 500 g for 3 min. After the addition of Cytoplasmic Extraction Reagent I, the lysate was incubated on ice for 10 min, and Cytoplasmic Extraction Reagent II was added to the suspension, which was then centrifuged at 16,000 g at 4°C for 5 min. The supernatant was collected as the cytoplasmic extracts. The insoluble (pellet) fraction was resuspended in Nuclear Extraction Reagent, incubated on ice for 40 min, and then centrifuged at 16,000 g at 4°C for 10 min. The supernatant was collected as the nuclear extracts. Both nuclear and cytosolic extracts were used for subsequent western blotting experiments.

Total cell lysate was prepared using ice-cold RIPA Lysis Buffer (Beyotime, P0013B) containing a complete protease inhibitor cocktail (Roche, 04693132001), and protein concentrations were determined with a bicinchoninic acid (BCA) protein assay kit (Beyotime, P0012). Equal amounts of proteins were separated by electrophoresis on a 4-20% Sure PAGE gel (Genscript) and transferred to PVDF membranes (Millipore) by electroblotting. Nonspecific binding sites were blocked for 1 h with 5% bovine serum albumin (BSA) diluted in Tris-buffered saline with Tween 20 (TBST; Solarbio, Shanghai, China), followed by incubation with the primary antibodies (1:1000) in blocking solution overnight at 4°C. After washing in TBST three times, the membranes were incubated at room temperature for 1 h with an appropriate secondary antibody (1:2000). The immunoblotting signals were then detected under a LAS-4000 luminescent image analyser (Fuji Film Co.) using WesternBright ECL HRP substrate kit (Advansta). The relative expression of each target protein was normalized to that of TUBA1A.

Total RNA extracted with TRIZOL reagent (Invitrogen) was reverse transcribed into cDNA using the PrimeScript RT Master Mix (Takara Bio). Real-time PCR was carried out using AceQ qPCR SYBR Green Master Mix (Vazyme) and gene-specific primers (see Table S1 for primer sequences) on the ABI QuantStudio5 system (Applied Biosystems). Specificity of each PCR amplification was verified by melting curve analysis. Data were normalized to the expression of the housekeeping gene Tuba1a.

GCs grown on coverslips were subjected to the desired treatments, and fixed with 4% paraformaldehyde (Sigma-Aldrich, P-6148) for 1 h. Cells were washed in PBS (Gibco; 1 mM KH₂PO₄, 155 mM NaCl, 3 mM Na₂HPO₄-7H₂O) and then permeabilized with 0.5% Triton X-100 (Sigma-Aldrich, T8787) for 10 min at 4°C. After blocking with 1% BSA (Sigma-Aldrich, A3059) for 1 h at room temperature, the cell climbing sheets were stained with antibody against FOXO1 (2880, Cell Signaling Technology; 1:100) for 1 h at 37°C. Immunoreactivity was detected by incubation with a Rhodamine (TRITC)-conjugated goat anti-rabbit IgG (33109ES60, YEASEN; 1:100) for 1 h at 37°C, and the nuclei were counterstained with DAPI (Sigma-Aldrich, D8417) for 20 min. The slides were then visualized under a Zeiss LSM 710 META confocal microscope (Carl Zeiss). For immunohistological staining, porcine ovaries were fixed in buffered paraformaldehyde (4%), embedded in paraffin, sectioned to approximately 5 μm, and mounted on glass slides. The ovarian sections were deparaffinized in xylene, rehydrated, and retrieved by microwave heating with citrate buffer in 0.05% Tween 20, pH 6.0, for 0.5 h. Endogenous peroxidase activity was eliminated by exposure to 3% H₂O₂ (Sigma-Aldrich, 216763-100 ml) for 10 min. After 1 h of blocking with 1% BSA, sections were incubated with antibodies against TP53INP1 (DF8731, Affinity Biosciences; 1:200), CDKN1A (10355-1-AP, Proteintech; 1:200) or PCNA (Proteintech, 60097-1-Ig; 1:200), and corresponding secondary antibodies [Cy3-AffiniPure goat anti-rabbit IgG (111-165-003, Jackson ImmunoResearch; 1:200) and FITC-AffiniPure goat anti-mouse IgG (115-095-003, Jackson ImmunoResearch; 1:200)], and then mounted with VECTASHIELD Mounting Medium plus DAPI and examined under an LSM 710 META laser-scanning confocal microscope (Carl Zeiss).

GCs with desired treatments were digested in 0.25% EDTA-free trypsin (Gibco), rinsed with cold PBS, and fixed with 70% ethanol for 2 h. After twice washing with PBS, cells were centrifuged at 1000 g for 3 min, and then incubated with propidium iodide (PI)/RNase A staining solution (RNase A:PI=9:1; KeyGEN). The cell cycle distribution was determined using a BD Accuri C6 flow cytometer (Becton Dickinson).

siRNAs specific for FOXO1, TP53INP1, CDKN1A and the scrambled control siRNA were obtained from GenePharma (Shanghai, China). Transfection of siRNAs (see Table S2 for siRNA sequences) was performed with Lipofectamine 3000 reagent (Invitrogen, L3000015) following the manufacturer's instructions.

GCs washed with PBS (Gibco) were lysed on ice with IP lysis buffer (Pierce, 26149) containing protease inhibitor cocktail (Roche, 04693132001). Whole-cell lysates (WCL) were then subjected to immunoprecipitation with anti-FOXO1 (Cell Signaling Technology, 2880). For each immunoprecipitation reaction, we added 5 μl of antibody to 500 μl of cell lysate to reach a final concentration of 20 ng/μl, and the mixture was incubated at 4°C overnight. After addition of 25 μl Protein A/G Magnetic Beads (88802, Thermo Fisher Scientific), the incubation was continued for 1 h at 4°C. The beads were pelleted with magnetic force and the supernatant was discarded. The immunoprecipitates were washed using 1×cell lysis buffer, magnetized again to remove the supernatant, eluted with SDS loading buffer (SunShineBio, Nanjing, China), and then processed for immunoblotting with the indicated antibodies.

The ChIP assay was performed using a Pierce Agarose ChIP Kit (Thermo Fisher Scientific, 26156) according to the manufacturer's instructions. Briefly, cells were cross-linked with 1% formaldehyde at room temperature for 10 min, and the reaction was stopped by 1× glycine. After washing with cold PBS containing protease inhibitors, cells were lysed in 1 ml SDS lysis buffer (1% SDS, 10 mM EDTA, 50 mM Tris, pH 8.1). Chromatin was then digested with micrococcal nuclease to generate 250-700 bp fragments. For the immunoprecipitation assay, the fragmented chromatin was incubated with anti-FOXO1 (Cell Signaling Technology, 2880), anti-p53 (Cell Signaling Technology, 2527) or rabbit IgG (as control) at 4°C overnight. After degrading proteins in the precipitated complexes with proteinase K, levels of ChIP-purified DNA were quantified with qRT-PCR (see Table S3 for primer sequences). The amount of immunoprecipitated DNA for each ChIP reaction was calculated using the Percent Input Method and was presented as percentage input.

The TP53INP1 promoter was amplified with PCR from pig genomic DNA (2072 bp, 5′-UTR of TP53INP1), cloned into the pGL3-Basic plasmid (Promega) at the KpnI and XhoI (Takara Bio) sites and named pGL3-TP53INP1(WT) (see Table S4 for primer sequences). By searching the JASPAR database (http://jaspar.genereg.net/), we identified three FRE (TGTTTC) motifs within the TP53INP1 promoter region. Using a Mut Express II Fast Mutagenesis Kit (Vazyme, China), pGL3-TP53INP1 (M1), pGL3-TP53INP1 (M2) and pGL3-TP53INP1 (M3) were individually constructed by introducing mutations into each of these FRE sites. Transfection was performed with Lipofectamine 3000 (Invitrogen). Three plasmids were co-transfected in each treatment: (1) 1.6 μg expression vector pCMV5-FLAG-FKHR (WT); (2) 1.6 μg reporter construct pGL3-TP53INP1(WT), pGL3-TP53INP1(M1), pGL3-TP53INP1(M2) or pGL3-TP53INP1(M3); and (3) 32 ng control reporter pRL-TK (Promega). Luciferase activities were measured 36 h post-transfection using the Dual-luciferase Reporter Assay System (Promega) and Modulus Microplate Luminometer (Turner BioSystems) according to the manufacturers’ protocols. Data are presented as firefly luciferase reporter values normalized to Renilla values.

Statistical analysis was performed using the SPSS software version 20.0 (SPSS). All experiments were repeated at least three times, and the data are presented as mean±s.e.m. Pairwise comparisons were assessed by the Student's t-test, and P-values <0.05 were considered to be statistically significant.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.199453