The involvement of hypoxia-inducible factor 1α (HIF1α)-stabilising factors in steroidogenic acute regulatory (STAR) protein-dependent steroidogenesis in murine KK1 granulosa cells in vitro

Keywords: factor inhibiting HIF (FIH), granulosa cells, HIF1α, hypoxia-inducible factor 1 (HIF), prolyl hydroxylases (PHD1, 2, 3), steroidogenesis, steroidogenic acute regulatory (STAR) protein, von Hippel Lindau (VHL) suppressor.

Enclosed in ovarian follicles and surrounded by a semi-permeable basal lamina, granulosa cells exert their roles in an environment characterised by a restricted vascular supply (Suzuki et al. 1998; Martelli et al. 2009; Rodgers and Irving-Rodgers 2010). Consequently, the provision of nutrients and oxygen (O₂) for granulosa cells and the oocyte depends on diffusion and varies between different stages of folliculogenesis (Redding et al. 2008). A study of pigs has shown that the O₂ tension of follicular fluid correlates negatively with follicular size (Basini et al. 2004). Following ovulation, the follicular basal lamina is breached and vascular supply spreads into the follicular cavity during the formation of the corpus luteum (CL). However, during early luteal development, angiogenesis falls behind the intense luteal cell proliferation and the O₂ supply is further limited (Amselgruber et al. 1999). Similarly, during luteal regression, reduced vascular density leads again to hypoxia, thereby contributing to functional and structural luteolysis (Meidan et al. 2013; Nishimura and Okuda 2020). Cumulatively, hypoxia plays important roles in regulating reproductive events, including follicular and luteal function.

Hypoxia-inducible factor (HIF)-complexes are major transcriptional responders, activating the cell specific response to hypoxia (Semenza 1998; Semenza 2000). They are heterodimers consisting of α- and β-subunits connected at the helix–loop–helix (bHLH)-Per-Arnt-Sim domain (Wang and Semenza 1995; Jiang et al. 1996a). Only HIFα-expression is O₂ dependent, whereas HIFβ is constitutively expressed (Huang et al. 1996; Semenza et al. 1996). Of the three known isoforms of HIFα, HIF1α is the best characterised and is expressed ubiquitously in mammalian tissues (Wenger et al. 1996, 1998). Its O₂-dependent cellular availability is regulated enzymatically; it undergoes hydroxylation and degradation initiated by factor inhibiting HIF (FIH) and prolyl hydroxylases (PHDs), utilising O₂ as a cosubstrate (Elkins et al. 2003). Accordingly, their hydroxylation abilities are limited in hypoxia. FIH hydroxylates the asparaginyl residue on HIF1α and inhibits the binding of CBP/P300, a transcriptional cofactor and stabiliser of dimerisation of HIF1-complexes (Mahon et al. 2001; Dames et al. 2002; Freedman et al. 2002). PHDs hydroxylate prolyl residues serving as binding sites for the von Hippel-Lindau (VHL) tumor suppressor protein (Bruick and McKnight 2001; Epstein et al. 2001). Once bound, VHL, which is a ubiquitin E3 ligase, marks HIF1α for proteasomal degradation (Ivan et al. 2001; Jaakkola et al. 2001). There are three known isoforms of PHDs in mammalian cells: PHD1 (EGLN2), PHD2 (EGLN1), and PHD3 (EGLN3) (Bruick and McKnight 2001; Epstein et al. 2001). Their expression varies among different tissues and cell types, with PHD2 having the highest and most widely spread expression, while PHD3 protein levels are below detectable limits in several cell types (Lieb et al. 2002; Oehme et al. 2002; Cioffi et al. 2003; Appelhoff et al. 2004). PHD2 seems to be the major isoform involved in regulating HIF1α (Berra et al. 2003; Appelhoff et al. 2004; Minamishima et al. 2008), whereas suppression of PHD1 and PHD3 expression had no effect on HIF1α in normoxia, silencing of PHD2 upregulated HIF1α levels in several cell types (Berra et al. 2003) and, in contrast with the other isoforms, its prenatal knockout in mice is lethal (Minamishima et al. 2008). Therefore, the PHD2 is considered a critical O₂ sensor (Berra et al. 2003).

HIF1α also plays indispensable regulatory roles in the ovary. Studies of mice have shown that suppressing the DNA binding capacity of HIF1-complexes with echinomycin prevents ovulation (Kong et al. 2005; Kim et al. 2009). Further, the link between hypoxia and steroidogenesis has been previously established and HIF1α is involved in steroidogenic acute regulatory (STAR) protein-dependent steroidogenesis in granulosa cells (Jiang et al. 2011; Kowalewski et al. 2015; Fadhillah et al. 2017). STAR is a key factor regulating the provision of steroidogenic substrate to mitochondria and its function is rate-limiting for steroidogenesis (Clark et al. 1997). In vitro studies with granulosa cells demonstrated that blocking HIF1α activity with echinomycin, or decreasing its expression by siRNA, suppresses basal, and dbcAMP-stimulated STAR expression, both under moderately lowered (10%) and, importantly, ambient (20%) O₂ tension (Kong et al. 2005; Kowalewski et al. 2015; Fadhillah et al. 2017). These findings show the need for transcriptionally active HIF1-complexes under 20% O₂, strongly emphasising the importance of HIF1α for STAR-mediated steroidogenesis. They also suggest the possibility of O₂-independent regulation of HIF1α in granulosa cells. Such mechanisms, mediated through FSH or inflammatory stimuli, have been indicated previously (Jiang et al. 1996b; Kim et al. 2009). Conversely, both severely lowered O₂ and exaggerated stabilisation of HIF1α (caused by exposure to CoCl₂) proved to be detrimental for STAR expression and steroidogenesis in the ovary and the testis (Koos and Feiertag 1989; Basini et al. 2004; Jiang et al. 2011; Kumar et al. 2014; Kowalewski et al. 2015; Fadhillah et al. 2017).

Cumulatively, these observations underline the need for precise regulation of FIH-, PHDs- and VHL-dependent regulation of HIF1α availability and activity for the proper steroidogenic function of granulosa cells, under both ambient and reduced O₂ conditions. While this is becoming our working hypothesis, surprisingly, none of this has been shown previously for steroidogenic cells. Consequently, in the present study, we have used a well-characterised model, immortalised KK1 granulosa cells (Kananen et al. 1995; Kowalewski et al. 2009, 2015; Manna et al. 2009), to provide the first insights into the regulation of HIF1α-stabilising factors in steroidogenic cells. The presence of HIF1α stabilising machinery was also confirmed in murine ovaries and the respective mRNA localised. Using the steroidogenic model, the response of FIH, VHL and PHDs to different O₂ conditions was assessed in vitro, and the possible effects on STAR protein expression were addressed.

Ovaries were collected from wild type not mutated wt/wt C57BL/6 mice (n = 3). This tissue was derived from other experiments performed at the University of Zurich with the permit number 042/2018, and was shared with us using the 3R (reduce, replace, refine) approach. The animals did not receive any treatment and were killed by overdose with pentobarbital. The tissue material was collected and processed immediately after the animals were killed. Isolated ovaries were trimmed of the surrounding tissues and either preserved for 24 h in RNAlater (Ambion Biotechnology GmbH, Wiesbaden, Germany) and subsequently frozen at −80°C until isolation of RNA, or fixed for 24 h in phosphate-buffered 4% formaldehyde and then paraffin-embedded, for in situ hybridisation studies.

For all in vitro experiments, immortalised murine KK1 granulosa cells (kindly provided by Dr. Ilpo Huhtaniemi, Hammersmith Campus, Imperial College, London, UK) were used. All experiments were performed at least three times with cells from different passages. The KK1 cells are well-characterised and respond in a dose-dependent manner to stimulation with N6,2-dibutyryladenosine-3,5-cyclic monophosphate (dbcAMP/cAMP; Sigma-Aldrich Chemie GmbH, Buchs, Switzerland) (Kananen et al. 1995; Kowalewski et al. 2009, 2010, 2015; Manna et al. 2009). The cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM)/F-12 (1:1) containing 10% heat inactivated fetal bovine serum (FBS), 1% gentamycin (10 mg/mL) and sodium bicarbonate (NaHCO₃, 1.125 g/L) in a humified incubator at 37°C, 20% O₂, 5% CO₂ (DMEM/F-12 and FBS were purchased from Thermo Fisher Scientific AG, Reinach, Switzerland; NaHCO₃ from Sigma-Aldrich Chemie GmbH; and gentamycin from Chemie Brunschwig AG, Basel, Switzerland). Twenty-four hours prior to experiments, cells were trypsinised (Chemie Brunschwig AG), seeded on six-well plates to grow overnight to 75–85% confluence and preincubated under 20%, 10%, or 1% O₂. On the day of the experiments, serum-containing medium was removed, cells were washed with sterile 1× phosphate buffered saline (PBS; Chemie Brunschwig AG), and serum-free medium (preconditioned 24 h prior to stimulation, according to the respective experimental conditions) was added for stimulation.

The cells were stimulated with dbcAMP in a hypoxic incubator for 6 h, the time point associated with the highest STAR, phospho(P)-STAR and progesterone (P4) output (Kowalewski et al. 2010, 2015). In blocking experiments, cells were preincubated with 5 nm echinomycin (Cayman Chemical Inc., Ann Arbor, MI, USA), or 25 μM, 50 μM, or 100 μM roxadustat (FG-4592; MedChemExpress, Monmouth Junction, NJ, USA) for 40 min prior to combined stimulation with dbcAMP. Echinomycin blocks the DNA-binding activity of HIF1α, but not its expression (Kong et al. 2005; Kim et al. 2009; Kowalewski et al. 2015; Turhan et al. 2021). Roxadustat is a prolyl hydroxylase inhibitor (Chen et al. 2019; Dhillon 2019), and was used within the concentrations recommended by the manufacturer.

African green monkey kidney cells (COS-1), referred to here as COS (Horowitz et al. 1983), served as controls, as recommended by the manufacturers of the antibodies used for the western blots. Unstimulated COS were incubated in DMEM with 10% FBS and 1% penicillin/streptomycin at 20% O₂ and 1% O₂ following the same experimental procedure as the KK1.

TRIzol® reagent (Thermo Fisher Scientific) was used for total RNA isolation from mouse ovarian tissue and KK1 granulosa cells, following the manufacturer’s directions. The quantity and quality of isolated RNA was determined using a NanoDrop 200C® spectrophotometer (Thermo Fisher Scientific). Qualitative PCR was performed to confirm the presence of transcripts encoding for FIH, VHL and PHD1-3 in ovaries and KK1 cells. Therefore, reverse transcription (RT) using random hexamers as primers, and conventional hot start PCR reactions, were run in a Mastercycler® (Eppendorf AG, Hamburg, Germany) as previously described (Kowalewski et al. 2006b). After RT, complementary DNA (cDNA) was amplified using the AmpliTaq Gold™ DNA Polymerase with Gold Buffer and MgCl₂ (Applied Biosystems by Thermo Fisher, Foster City, CA, USA). Self-designed primers (Table 1) were ordered from Microsynth AG (Balgach, Switzerland). The conditions for amplification were set as follows: 10 min at 95°C, 39 cycles of 1 min at 94°C, 1 min at 60°C and 2.5 min at 72°C, 10 min at 72°C. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH), proved to be stably expressed under varying O₂ concentration in KK1 (Kowalewski et al. 2015), served as control. Amplicons were visualised on ethidium bromide stained 2% agarose gel. The bands obtained were isolated using a QIAquick® Gel Extraction Kit (50) (Qiagen, Hilden, Germany) and commercially sequenced (Microsynth AG) to confirm the primers’ specificity.

Table 1.  List of primers used for RT-PCR (a) and in situ hybridisation (ISH) (b).

Because the antibodies used for western blot analysis lacked the specificity required for immunohistochemistry, the localisation of factors involved in HIF1α-stabilisation (FIH, VHL, PHD2) in the ovary was assessed at the mRNA level by non-radioactive ISH, utilising the protocol described previously (Kowalewski et al. 2006a, 2008). Therefore, complementary RNA (cRNA) probes were synthesised by using homogenised murine ovarian tissue in PCR with primers specific for FIH, VHL and PHD2 (Table 1b). The PCR products were separated on a 2% agarose gel, stained with ethidium bromide and isolated using the QIAquick® Gel Extraction Kit (50) (Qiagen) followed by ligation into the pGEM-T plasmid (Promega, Dübendorf, Switzerland) and transformation of bacteria with the plasmid. Colonies containing the insert were selected and plasmids were purified using the PureYield™ Plasmid Miniprep System (Promega). Final products were submitted for sequencing to Microsynth. Restriction enzymes Nco1 and Not1 were used for linearising plasmids in the course of synthesis of sense- and antisense cRNA. The cRNAs synthesis and labelling were done by using the DIG-RNA labelling kit (Roche Diagnostics, Basel, Switzerland) according to the manufacturer’s manual. Dot-blot analyses were performed for semi-quantification of DIG-labelled ribo-probes in serial dilutions on positively charged nylon membranes (Roche Diagnostics). To perform non-radioactive ISH, paraffin-embedded mouse ovaries were cut into 2 μm-thick slices using a microtome and then placed on SuperFrost Plus microscope slides (Menzel-Gläser, Braunschweig, Germany). The sections were dewaxed, digested with proteinase K (70 μg/mL; Sigma-Aldrich) for 20 min to increase tissue permeability, and post-fixed with 4% paraformaldehyde. Hybridisation was performed overnight at 37°C in a formamide chamber. To detect the DIG-labelled cRNA, the sections were incubated with alkaline phosphatase-conjugated sheep anti-DIG Fab Fragments (diluted by 1:5000; Roche Diagnostics) in 1% ovine serum overnight at 4°C. The signals were visualised using 5-bromo-4-chloro-3-indolyl phosphate and nitroblue tetrazolium (BCIP/NBT; Roche Diagnostics).

Protein preparation and immunoblotting were performed as described previously (Gram et al. 2013). Briefly, after stimulation, cells were collected using Net-2-lysis buffer (50 mM Tris–HCl, pH 7.4, 300 mM NaCl, 0.05% Nonidet P-40) containing 10 μL/mL protease inhibitor cocktail (Sigma-Aldrich Chemie). Protein homogenates (20–30 μg total protein) were separated using 8–15% polyacrylamide gels and transferred onto methanol-activated polyvinylidene difluoride (PVDF) membranes at 100 volts (V) for 1 h. The membranes were blocked with 5% skimmed milk in PBS/0.25% Tween 20 (PBST) and incubated with specific antibodies, diluted in 2% skimmed milk/PBST (Table 2) at 4°C overnight. Beta-actin (ACTB) (Table 2) served as an internal loading control. Therefore, membranes were first stained for the target proteins and then reblotted and stained against the ACTB to semi-quantify the expression of the targets. Secondary antibodies (Table 2) were used for 1 h. The signals were visualised using SuperSignal™ West Pico or SuperSignal West Femto Chemiluminescent Substrate (Pierce Biotechnology by Thermo Fisher Scientific) to initiate signals in a ChemiDoc™ XRS+ with Image Lab™ Sofware (Bio-Rad Laboratories AG, Cressier, Siwtzerland) in the presence of Precision Plus Protein marker (Bio-Rad). The optical density of proteins was measured with ImageJ software (Wayne Rasband, National Institutes of Health, USA) and ACTB was used as the reference to calculate the standardised optical density (SOD) of protein bands.

Table 2.  Antibodies used for western blot analysis.

All experiments were repeated at least three times. Cells from different passages were used for each experiment and representative blots are shown. GraphPad 3.06 (GraphPad Software, San Diego, CA, USA) software was used for statistical analysis with one-way analysis of variance (ANOVA) followed by Tukey–Kramer multiple comparison post-test. The WB optical density results are presented as mean ± s.d. P < 0.05 was considered significant.

In order to verify the hypoxia-related stabilisation of HIF1α and its effects on STAR in KK1 granulosa cells (Fig. 1) (Kowalewski et al. 2015), control experiments were set up with cells exposed to 20% O₂ and severely lowered (1%) O₂, stimulated with a low background (0.1 mM) of dbcAMP in the presence/absence of echinomycin. Additionally, as for subsequent analyses, control experiments were performed with COS (Fig. 1) to verify the specificity of the anti-HIF1α antibody. Although dbcAMP and echinomycin did not affect the levels of HIF1α protein, strongly lowered O₂ (1%) increased its expression significantly (P < 0.01 and P < 0.001 compared with control and treated cells respectively; details in Fig. 1a). As expected, the expression of STAR was induced by dbcAMP, and this positive effect was inhibited by echinomycin (Fig. 1b). Concomitant with the strongly induced HIF1α levels, 1% O₂ had detrimental effects on STAR levels in the experimental groups, i.e. in control and treated cells (P < 0.05, for statistical details see Fig. 1b).

Fig. 1.  Effects of hypoxia and functional suppression of HIF1-complexes on HIF1α and STAR expression in KK1 granulosa cells. Cells were stimulated with 0.1 mM dbcAMP and treated with 5 nM echinomycin (Echino/E, functional blocker of HIF1-complexes) under 20% and 1% O₂. HIF1α (a) (115 kDa) and STAR (b) (30 kDa) protein expression as determined by western blot analysis. The results were normalised against ACTB and calculated as the average SOD. Representative western blots are shown. For (a) and (b) one-way ANOVA was applied (P < 0.0001 each), followed by a Tukey–Kramer multiple comparisons post-test. Bars with asterisks differ at: *P < 0.05, **P < 0.01, ***P < 0.001 as indicated on the figure. P < 0.05 was considered significant. Results are presented as mean ± s.d.

Because FIH, VHL, PHD1, PHD2 and PHD3 are involved in the stabilisation of HIF1α, enabling cells to adapt to hypoxia by balancing the availability of HIF1α (Semenza 1998), we verified their mRNA and protein expression in murine ovaries (Fig. 2) and in the KK1 granulosa cells used as a steroidogenic model (Fig. 2). The basic expression of all factors was confirmed at the transcript level in both ovaries and KK1 cells (Fig. 2a). Similarly, all of the proteins were detected in ovaries by western blot, indicating that PHD2 had the highest abundance compared with lower levels of PHD1 and PHD3 (Fig. 2b). However, in KK1 cells, both PHD1 and PHD2 were abundantly expressed, while PHD3 expression seemed to be below the detection level (Fig. 2b). FIH and VHL were clearly detectable and abundantly expressed in both KK1 cells and ovaries. COS were used as positive controls for all antibodies, as recommended by the manufacturers and supported by the available literature (Duan et al. 1995; Huang et al. 1996; Bracken et al. 2006). They displayed the clearly detectable presence of FIH and PHD1, and apparently lower VHL, PHD2 and PHD3 content, pointing towards tissue/cell dependent expression of HIF1α stabilising factors (Fig. 2b).

Fig. 2.  Basal mRNA and protein expression of HIF1α-stabilising factors: FIH (42 kDa), VHL (17 kDa), PHD1 (44 kDa), PHD2 (50 kDa) and PHD3 (27 kDa) in murine ovaries, KK1 granulosa cells and COS cells serving as positive controls. Representative western blots are shown. (a) Results from qualitative PCR. (b) Protein expression determined by western blot. ACTB was used as loading control.

The antibodies used for the western blots were found to be unsuitable for immunohistochemical detection in the localisation studies. Therefore, ISH was applied, allowing the localisation of the respective transcripts for FIH, VHL and PHD2. PHD2 was the most abundantly expressed PHD in the ovary (Fig. 2b), and has been shown to be the predominant isoform involved in the stabilisation of HIF1α in several systems (Berra et al. 2003; Minamishima et al. 2008; Katschinski 2009). Due to our use of the 3R approach to source ovarian tissue, an accurate staging of the ovaries was not possible prior to tissue collection. Nevertheless, all the functional structures characteristic of an adult ovary were found in histological sections. This included secondary and tertiary follicles of different sizes and luteal structures, mostly of mature and regressing CL (Fig. 3). The signals of all three factors were localised in the granulosa cells of growing follicles (Fig. 3). No quantification of signals intensity was performed. However, signals appeared weaker in the CL. Small, primordial and primary follicles as well as interstitium stained negatively and revealed a slight background staining only. The sense probes employed as negative controls (NC) remained unstained.

Fig. 3.  Localisation of transcripts encoding for FIH, VHL and PHD2 murine ovaries as determined by non-radioactive ISH. FIH (a, b), VHL (c, d, e, f) and PHD2 (g, h) signals are localised predominantly in granulosa cells of growing follicles (black arrows). The signal intensity appears weaker in mature and regressing corpora lutea (white arrows), no or weak background staining is observed in interstitial compartments (grey arrows). There is no staining in negative control (NC).

The effects of functional blocking of HIF1α-complexes on the expression of HIF1α stabilising factors were examined to evaluate the possible self-regulatory loops; control and dbcAMP-stimulated KK1 granulosa cells were used, in the presence of echinomycin under decreasing O₂ content (20%, 10%, 1%) (Figs 4 and 5). Unstimulated COS cells were applied as positive controls under 20% and 1% O₂, and did not show any changes in protein expression for VHL, FIH, PHD1, or PHD2. PHD3 responded to decreased O₂ and seemed to be more highly represented under 1% O₂ (Figs 4 and 5). Similarly, for KK1 cells, VHL, FIH and PHD1 did not change significantly in response to either dbcAMP or echinomycin, nor to changing O₂ levels, and did not respond to changing O₂ content (ANOVA P = 0.92, P = 0.96 and P = 0.8, for VH, FIH and PHD1, respectively) (Figs 4 and 5). They seemed, therefore, to be constitutively expressed. In contrast, the expression of PHD2 (Fig. 5b) increased gradually in response to decreasing O₂ content; it was significantly upregulated at 10% O₂ compared with 20% (P < 0.001), followed by a further increase towards 1% O₂ (P < 0.001) (Fig. 5b). None of the treatments affected the responsiveness to decreasing O₂, thus PHD2 did not respond to dbcAMP or echinomycin. As for PHD3, its expression was below detection limits in KK1 cells (Fig. 1), and remained unaffected in response to lowered O₂ (Fig. 5c).

Fig. 4.  Effects of changing oxygen content (20%, 10%, 1% O₂) and functional blocking of HIF1α on FIH and VHL in KK1 granulosa cells. Cells were stimulated with 0.3 mM dbcAMP and treated with echinomycin (Echino/E). Western blot analysis was performed. COS cells served as positive controls. Representative immunoblots are shown. Expression of FIH (a) and VHL (b) were examined, normalised against ACTB and calculated as the average SOD. One-way ANOVA was applied: FIH (P = 0.92), VHL (P = 0.96). P < 0.05 was considered significant. Results are presented as mean ± s.d.

Fig. 5.  Effects of changing oxygen content (20%, 10%, 1% O₂) and functional blocking of HIF1α on PHD1, PHD2 and PHD3 in KK1 granulosa cells. Cells were stimulated with 0.3 mM dbcAMP and treated with echinomycin (Echino/E). Western blot analysis was performed. COS cells served as positive controls. Representative immunoblots are shown. Expression of PHD1 (a), PHD2 (b) and PHD3 (c) was examined, normalised against ACTB and calculated as the average SOD. (a, b) One-way ANOVA was applied revealing P = 0.8 for PHD1 (a), and P < 0.0001 for PHD2 (b), in case of P < 0.05, followed by a Tukey–Kramer multiple comparisons post-test: *P < 0.001. Results are presented as mean ± s.d.

The expression of HIF1α and STAR protein were assessed in KK1 cells treated with increasing dosages (25 μM, 50 μM and 100 μM) of roxadustat, a functional PHD blocker (Dhillon 2019) (Fig. 6). Echinomycin was included, additionally confirming its inhibitory effects and, thereby, the involvement of HIF1α in STAR expression under ambient O₂ (Fig. 6b). Cells were exposed to 20% O₂, allowing the maintenance of a low basal expression of HIF1α, which was clearly detectable in all experimental groups. Blocking of PHDs with roxadustat gradually stabilised HIF1α levels. Thus, its abundance was increased in cells treated with 25 μM roxadustat (P < 0.001), was further potentiated towards 50 μM roxadustat (P < 0.05, compared with 25 μM), and remained highly stabilised under 100 μM roxadustat treatment (Fig. 6a). As expected, STAR protein responded positively to dbcAMP treatment (P < 0.05) and displayed the highest levels in response to 25 μM roxadustat (P < 0.001 compared with untreated control, and P < 0.01 compared with 0.1 mM dbcAMP). However, the levels of STAR were apparently lower in cells treated with dbcAMP in the presence of 50 μM roxadustat, followed by strong suppression of its expression in cells treated with 100 μM roxadustat (P < 0.001 compared with 25 μM). Interestingly, there was also a significantly lower expression of STAR following functional suppression of HIF1α with echinomycin, when compared with cells in which HIF1α was stabilised with 25 μM roxadustat, showing the strongest STAR expression (Fig. 6b).

Fig. 6.  The effect of functional blocking of PHDs on HIF1α and STAR protein expression in KK1 granulosa cells. Cells were stimulated with 0.3 mM dbcAMP and PHD function was blocked using increasing dosages of roxadustat (25 μM, 50 μM and 100 μM); HIF1α activity was blocked with echinomycin (Echino). HIF1α (a) and STAR (b) protein levels were examined by western blot. Representative immunoblots are shown. The protein expression was normalised against ACTB and calculated as the average SOD. One-way ANOVA was applied (P < 0.0001 each), followed by a Tukey–Kramer multiple comparisons post-test. P < 0.05 was considered significant. Results are presented as mean ± s.d. Bars with asterisks differ at: *P < 0.05, **P < 0.01, ***P < 0.001 as indicated on the figure.

The transcriptional activity of HIF1-complexes is required for adequate functionality of ovarian follicles, including the steroidogenic properties of granulosa cells (Kowalewski et al. 2015; Nishimura and Okuda 2015, 2020; Baddela et al. 2020). Their interaction with STAR plays fundamental roles in this process (Kowalewski et al. 2015; Nishimura and Okuda 2015). However, extremely high levels of HIF1α are detrimental, highlighting the need for regulation of its abundance under altered oxygenation conditions, also clearly indicated by in vitro studies utilising abundant (20%) O₂ levels (Kowalewski et al. 2015; Fadhillah et al. 2017). In the present study, the expression and potential role of HIF1α regulating factors were examined in the murine model. Their expression was proved in the ovary, and in vitro experiments were applied to investigate their responsiveness to changing O₂ saturation in KK1 granulosa cells.

The responsiveness of KK1 cells to changing O₂ levels was examined by exposing them to higher and lower O₂ (20% vs 1%) and verifying the HIF1α and STAR expression. In accordance with our previous findings (Kowalewski et al. 2015), strongly reduced O₂ diminished the basal and dbcAMP-induced expression of STAR. At the same time, the effects of echinomycin confirmed that HIF1α is needed for adequate STAR activation. In addition, protein isolates from COS cells were included to test the specificity of the anti-HIF1α antibody, and showed their high responsiveness to lower O₂, mirrored in a strong activation of HIF1α under severe hypoxia. Interestingly, although not the main focus of the study, the hypoxia-induced HIF1α expression in COS appeared to exceed those levels observed in KK1 cells under the same experimental conditions, pointing towards a cell/tissue specific response to hypoxia.

However, the present study focused on the expression and regulation of HIF stabilising factors (FIH, VHL, PHD1, PHD2, PHD3). The presence of their respective transcripts and proteins was investigated in ovaries and KK1 cells. While the expression of all targets was confirmed at the mRNA level in all tissue/cell types, the expression patterns and the apparent abundance of the respective proteins appeared to provide the first hints for possible regulatory mechanisms. In particular, the expression of PHDs differed strongly. While FIH and VHL were clearly detectable, only PHD2 seemed to be abundantly present in the ovary, apparently exceeding the levels of PHD1 and PHD3. In turn, PHD3 seemed to be under detection limits in KK1 cells, which presented abundant expression of PHD1 and PHD2. Thus, the isoform commonly strongly represented in ovary and KK1 cells was the PHD2, pointing towards its possible predominant role in these tissue/cells. This situation differed, however, from what has been observed in COS, in which FIH and PHD1 seemed to be the predominant HIF1α regulators. The individual, tissue specific distribution of PHDs, and the prominence of PHD2 compared with PHD3, have also been observed in other biological targets, for example, in Pc-3 cells (human prostate adenocarcinoma) or human kidney (Hirsilä et al. 2003; Appelhoff et al. 2004). Moreover, HIFα specificity has been postulated for PHD2 and PHD3, implying their selective affinity to HIF1α or HIF2α respectively (Appelhoff et al. 2004). Whether this applies to the ovary is not known, as the expression and potential roles of HIF2α in the ovarian cellular compartments have not yet been investigated. However, compared to HIF1α, the expression of HIF2α appears to be targeted to certain cell types (Wiesener et al. 2003; Loboda et al. 2010). Regarding reproductive organs, HIF1α has been described as being constitutively expressed in testicular Leydig cells (Palladino et al. 2011). The lack of detectable levels of PHD3 in KK1 cells compared with the ovary could relate to the cellular source of the protein within the organ, with whole ovarian isolates presenting relatively low, yet detectable amounts of the protein. Conversely in COS, PHD3 seemed to be activated by lower O₂ content.

Next, we felt prompted to investigate the cellular distribution of VHL, FIH and PHD2 in the ovary. PHD2 was selected based on our observations, supported by the previously available information emphasising its role as the predominant regulator of HIF1α (Berra et al. 2003; Wiesener et al. 2003; Katschinski 2009; Mollenhauer et al. 2012). ISH indicated that granulosa cells of growing follicles were the main cellular target of all factors. This is an important finding implying an active local regulation of HIF1α availability within the ovary, possibly preventing its exaggeratedly high levels and thereby supporting the underlying hypothesis of the present study. Interestingly, the regressing CL in particular seemed to stain more weakly, supporting the higher availability of functionally active HIF1α during luteal regression observed in other species, for example, in the cow or dog (Nishimura et al. 2006; Sousa et al. 2016; Nishimura and Okuda 2020).

By balancing HIF1α expression in an O₂-dependent manner, factors stabilising HIF1α act in a tissue and cell type specific manner (Los et al. 1996; Berra et al. 2003; Appelhoff et al. 2004; Soilleux et al. 2005). It was, thus, interesting to observe the stably expressed levels of VHL, FIH and PHD1 in KK1 granulosa cells under changing O₂ saturation. Neither dbcAMP, nor echinomycin-mediated modulation of HIF1-complexes activity, affected their levels, which seemed to be constitutively and abundantly present. Based on this observation, it could be concluded that the FIH- and PHD1-mediated hydroxylation of HIF1α in granulosa cells depends predominantly on O₂ availability; the less oxygen is available, the more HIF1α is stabilised due to inactivation of FIH and PHD1 (Semenza 1998). In this context, the ubiquitination of HIF1α by VHL depends on PHD activity, whereas decreased functionality of FIH allows the cofactor CBP to attract HIF1-complexes to their transcriptional targets.

It was interesting and of functional importance to see the gradual increase of PHD2 in granulosa cells following decreasing O₂ saturation. This expression pattern resembled the previously published HIF1α levels observed in KK1 cells under progressively decreasing O₂ concentrations (Kowalewski et al. 2015) and led us to hypothesise that PHD2 could be the major O₂ responsive regulator of HIF1α in granulosa cells. Being induced by decreasing O₂ saturation, PHD2 might act as a specific regulator preventing exaggeratedly high HIF1α levels and thereby protecting the cells from its possible inhibitory effect. Such backup mechanisms, highlighting the importance of intracellular protection mechanism against toxic levels of HIF1α, have been previously indicated in other systems. Thus, blocking of PHD2 expression in HeLa cells at ambient (20%) O₂ resulted in initially increased HIF1α expression that diminished over time with progressively decreasing HIF1α levels (Berra et al. 2003). Although the exact molecular mechanisms are not fully understood, the existence of such backup mechanisms underlines the complexity and importance of balancing HIF1α in steroidogenic cells. As for other factors, the expression of PHD2 remained stable under exposure to dbcAMP and echinomycin. This differs from what has been observed in other cell types, for example, HeLa cells, in which the hypoxia-dependent increased expression of PHD2 was inhibited by blocking HIF1α expression with siRNA (Berra et al. 2003), or in rat C6 glioma cells (D’Angelo et al. 2003), and points towards cell type- or species-specific regulation of PHD2 expression.

Nevertheless, taking into account its responsiveness to changing O₂ content, and its abundance in ovarian isolates compared with other isoforms, PHD2 appears to be a major player regulating the stabilisation of HIF1α in granulosa cells. These findings corroborate observations made in other cell and tissue types (Berra et al. 2003; Appelhoff et al. 2004; Minamishima et al. 2008).

Finally, in order to gain a deeper functional insight into the role of PHDs in the HIF1α-dependent expression of STAR, experiments utilising roxadustat, a specific functional blocker of PHDs, were performed. Roxadustat was initially developed to treat anemia caused by chronic kidney disease in humans (Dhillon 2019). It increases HIF1α levels, followed by transcription of HIF1 target genes, including erythropoietin (Besarab et al. 2015). Roxadustat is not selective to different PHD isoforms. KK1 cells responded in a dose-dependent manner to the treatment by presenting increasing HIF1α levels under ambient O₂, clearly indicating the involvement of PHDs in the maintenance of HIF1α in steroidogenic granulosa cells. Most importantly, constituting one of the most important observations from the present study, these effects were accompanied by a biphasic response in STAR protein expression. It was significantly upregulated in response to lower concentrations of roxadustat, to above the levels detected in cells treated with dbcAMP only, apparently mimicking the positive effects of moderate hypoxia on STAR expression described previously (Kowalewski et al. 2015). Conversely, enhanced stabilisation of HIF1α with higher dosages of the compound, significantly reduced STAR to basal levels, i.e. significantly lower than those observed in dbcAMP-treated cells, resembling the effects of strongly lowered (1%) O₂, confirmed in the present study. Consequently, it became obvious that the more HIF1α is stabilised, crossing the not yet defined threshold, the less STAR is present in steroidogenic cells (Fig. 7). The underlying molecular mechanisms need further clarification. Similar effects on steroid synthesis, resulting from overexpression of HIF1α, were shown in luteinised bovine granulosa cells (Jiang et al. 1996b) or canine lutein cells (Sousa et al. 2016) treated with CoCl₂. The latter is a non-specific blocker of HIF1α degradation, affecting the functionality of PHDs, FIH and VHL, leading to extremely high intracellular HIF1α levels (Muñoz-Sánchez and Chánez-Cárdenas 2019).

Fig. 7.  Schematic representation of proposed regulation of HIF1α availability and its importance in STAR-dependent steroidogenesis in granulosa cells. HIF1α expression and the activity of HIF1-complexes, are requested for STAR expression and steroidogenesis. VHL, FIH and PHDs, modulate (inhibit) the presence of HIF1α depending on the availability of O₂. However, FIH and VHL are expressed in an O₂-independent manner in granulosa cells, while PHD2 (but not PHD1 or PHD3) responds to changing O₂ levels by increasing its expression. Enhanced PHD2 activity under ambient and/or moderately lowered O₂ (20% and 10% O₂) appears to have a protective role in regulating HIF1α availability by preventing an exaggeratedly high expression. When overexpressed, e.g. during severely lowered (1%) O₂, HIF1α exerts negative effects over STAR expression and steroidogenesis. With this, the effects of HIF1α appear to be biphasic, i.e. while it is required for STAR expression, when overexpressed it exerts negative effects. The underlying molecular mechanisms remain to be elucidated.

Cumulatively, the present study shows the ovarian expression of factors regulating cellular availability of HIF1α and indicates the importance of regulating its expression in steroidogenic cells. Although studies utilising primary cell cultures are needed to support the present findings, by actively responding to changing oxygenation conditions, PHD2 might be the major player, among the PHDs, involved in stabilising HIF1α in granulosa cells. It could be involved in a self-regulatory loop protecting these cells from detrimentally high levels of HIF1α (Fig. 7). PHD-dependent maintenance of HIF1α regulates STAR availability, and when overexpressed HIF1α downregulates it, impeding steroidogenic cell function. Nevertheless, HIF1-complexes are required for efficient STAR synthesis (Fig. 7). Thus, our conclusion that balancing HIF1α levels is crucial for the steroidogenic function of granulosa and, presumably, other steroidogenic cells is of translational value.

The data that support this study will be shared upon reasonable request to the corresponding author.

The authors declare that they have no conflicts of interest.

This project was supported by the Vetsuisse Faculty of Zurich via the Institute of Veterinary Anatomy.

TG was involved in developing the concept of the study, experimental design, generating data, analysis and interpretation of data, and drafting the original manuscript. MPK designed and supervised the project, was involved in analysis and interpretation of the data, and editing and revising the manuscript. Both authors read and approved the final manuscript.