All biochemicals and enzymes were of analytical grade and were obtained from commercial suppliers. Insulin (dissolved in 0.9% NaCl), Lithium chloride (dissolved in H₂O), Wortmannin and Rapamycin were purchased from Sigma Aldrich. BML-257 and MG132 were purchased from Cayman chemicals. Unless otherwise stated, all substances were dissolved in DMSO.

HEK 293 and HepG2 cells were cultured under normoxia (16% O₂, 79% N₂, and 5% CO₂ [v/v]) in MEM supplemented with 10% FBS. MCF-7 cells were cultured under normoxia (16% O₂, 79% N₂, and 5% CO₂ [v/v]) in DMEM containing 10% FBS.

All cell lines were tested mycoplasma negative by using the MycoAlert Detection Kit (Lonza). In all experiments the number of cell passages used was below 10. For protein extraction, cells were seeded into 60 mm dishes and, when necessary, transfections were performed the next day for 12 h. Prior to treatment, cells were starved in serum-free media for 16 h, then treated either alone or in combination with 100 nM insulin, 20 nM wortmannin, 12.5 µM BML-257, 10 nM rapamycin or 500 nM LiCl and further cultured for 4 h under normoxia or mild hypoxia (5% O₂, 90% N₂, and 5% CO₂ [v/v]). In all experiments control cells were treated with DMSO.

The construct for full-length HIF-3α1 was amplified by polymerase chain reaction (PCR) from HepG2 cDNA using the following primer set: human HIF-3α1-F 5′- cac​cat​gga​ctg​gca​aga​c-3′ and human HIF-3α1-R 5′- gtc​agc​ctg​ggc​tga​gcc​tg-3′. The resulting PCR product was directly cloned into the pcDNA3.1D/V5-His-TOPO vector according to manufacturer’s protocol. The pcDNA6/Myc-His-Gal4-HIF-3α-TAD was generated by inserting a HindIII-BamHI Gal4DBD fragment (1–147) of the pSG424 (Sadowski and Ptashne, 1989) into the pcDNA6/Myc-His vector to create a pcDNA6/Myc-His-Gal4 construct. The HIF-3α-TAD (451–528) was amplified by PCR from HepG2 cDNA using the following primer set: TAD-HIF-3α-F (5′-CTG GGG ATC CAC AGA CTC TCC ACT GCC CGG-3′) and TAD-HIF-3α-R (5′-CCT GCT CGA GTC GAC TTG TCG TCG TCG TCC TTG TAG TCC GGC AAC AGG CCAT GGA AGC T-3′) and ligated into the BamHI and SalI sites of pcDNA6/Myc-His-Gal4 allowing generation of a Gal4 fusion protein. The construct for pG5E1B-Luc was described previously (Krüger et al., 1997). The vector expressing HA-tagged full-length PKB has been already described (Franke et al., 1995). The constructs for full-length HIF-3α1 and Gal4-HIF-3α-TAD with mutation in serine 524 were generated using the QuickChange mutagenesis kit (Promega). All constructs were verified by sequencing in both directions.

4 × 10⁵ cells per 60 mm dish were transfected as previously described (Roth et al., 2004). To investigate HIF-3α transactivity, cells were cotransfected with 2 µg of reporter construct pG5-E1B-Luc and 500 ng of the Gal4-HIF-3α-TAD or respective mutant construct. After 24 h, the medium was changed, cells were treated with insulin (100 nM) and further cultured in serum-free medium for 8 h under normoxic conditions. The detection of luciferase activity was performed with the Dual- Luciferase™ Reporter Gene Assay Kit (Berthold, Pforzheim, Germany).

Western blot analysis was carried out as described previously (Kietzmann et al., 1999). In brief, lysates from MCF-7, HEK 293 and HepG2 cells were collected, and 50–100 µg of protein was loaded onto a 7.5% or 10% sodium dodecyl sulfate (SDS)- polyacrylamide gel. After electrophoresis and electroblotting onto a nitrocellulose membrane, proteins were detected with monoclonal antibodies against V5 tag (#R960-25; 1:5,000; Invitrogen), hemagglutinin (HA-Tag #sc-7392; 1:500; Santa Cruz), phospho-GSK3β (Ser9) (#9323; 1:1,000; Cell Signaling), E-cadherin (#610181; 1:1,000; BD Biosciences), α-smooth muscle (#A2547; 1:1,000; Sigma Aldrich), ubiquitin (#646304; BioLegend; 1:1,000) or α-Tubulin (#T5168; 1:10,000; Sigma-Aldrich). Polyclonal antibodies were against phospho-AKT (Ser473) (#9271; 1:1,000; Cell Signaling), phospho-AKT (Thr308) (#9275; 1:1,000; Cell Signaling), total AKT (#9272; 1:1,000; Cell Signaling) or phospho-p70S6 Kinase (Thr421/Ser424) (#9204; 1:1,000; Cell Signaling). The secondary antibody was either an anti-mouse (#1706516) or an anti-rabbit immunoglobulin G conjugated (#1706515) to horseradish peroxidase (1:5,000; Bio-Rad). The ECL system (Amersham) was used for detection.

For half-life studies, HEK 293 cells were transfected together with expression vectors encoding full-length V5-tagged HIF-3α1 WT or mutant HIF-3α1 S524A. 48 h post transfection, cells were treated with protein synthesis inhibitor cycloheximide (CHX, 20 μg/mL, (Sigma-Aldrich)). Afterwards, cells were harvested at the indicated time points and protein levels were measured by immunoblot analysis. For immunoprecipitation and ubiquitylation assays, HEK 293 cells were treated with proteasome inhibitor MG 132 (50 μM; Calbiochem). Four hours later, cells were scraped in lysis buffer (50 mM Tris/HCL, pH 7.5, 150 mM NaCl, 1% Triton X-100, 2 mM EDTA, 2 mM EGTA, 1 mM PMSF and complete protease inhibitor cocktail tablet (Roche)). After scraping, lysates were incubated with continuous shaking at 4°C for 20 min and centrifuged at 12.000 g at 4°C for 15 min. To recover anti-V5 and anti-HA immunoprecipitates, 150 µg of protein was incubated with 2 µg antibody for 1 h at 4°C before protein G Sepharose beads (30 µL per reaction mixture; #GE17-0,618-01; GE Healthcare) were added for 12 h. Thereafter, the beads were washed five times with lysis buffer and recovered, pellets were dissolved in 2 X Laemmli buffer, loaded onto a 7.5 % SDS gel, blotted and detected with antibodies against ubiquitin and V5- or HA-epitope.

Cell viability was assessed by a MTT [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide] reduction assay. MCF-7 or HEK 293 cells were transfected with expression vectors for full-length wild-type HIF-3α1 or HIF-3α1 S524A mutant, seeded in 24-well plates (5 × 10⁴ cells per well) and cultured under normoxia (16% O₂). After 24 h cells were treated with insulin (100 nM) and further cultured in serum-free medium for 8 h under normoxic conditions (16% O₂). Then 500 µL of MTT (4 mg/mL) was added to the wells and incubated at 37°C for 2 h. Formation of MTT–formazane was measured with a Tecan plate reader at a wavelength of 570 nm.

For colony formation, MCF-7 or HEK 293 cells were transiently transfected with plasmids for full-length V5-tagged HIF-3α1 WT or mutant HIF-3α1 S524A. After transfection, cells were further cultured for 20 h under normoxia (16% O₂) and then seeded into 6-well plates at a density of 2,000 cells/well. After 7 days, cells were fixed with 4% PFA and stained with crystal violet (Applichem). Plates were either scanned or images of the colonies were taken using the Leica MZFLIII microscope at original magnification ×40 and analyzed using the ImageJ software.

For adhesion assay, MCF-7 cells were transiently transfected with plasmids for full-length V5-tagged HIF-3α1 WT or mutant HIF-3α1 S524A. 24 h post transfection, cells were seeded into 6-well plates and grown under normoxia (16% O₂) until they reached confluency (2–3 days). The 6-well plates were then transferred to an orbital shaker and rotated at 250 rpm for 8 h at 37°C. Control plates remained in the incubator. After rotation, the medium was aspirated from the cells followed by washing with 1x PBS and fixation with 4% PFA for 20 min at room temperature. The cells were briefly washed with 1x PBS and stained with crystal violet solution. Subsequently, the staining solution was aspirated, and the wells were rinsed with water to remove excess stain and air dried. Plates were scanned and analyzed using the ImageJ software for quantification.

For scratch wound assay, MCF-7 cells were transiently transfected with plasmids for full-length V5-tagged HIF-3α1 WT or mutant HIF-3α1 S524A. 24 h post transfection, cells were seeded onto 96-well Essen ImageLock Plates (Essen Bioscience) (3 × 10³ cells per well) in DMEM with 10% FBS. The following day the confluent cell monolayer was wounded with the 96 PTFE pin Wound Maker (Essen Bioscience), cells were washed with 1x PBS and further cultured in DMEM supplemented with 0.1% FBS and insulin (100 nM). The live wound closure was phase contrast-imaged for 72 h in 4 h intervals and the confluence analysis was performed using the IncuCyte software.

Escherichia coli BL21 (DE3) cells transformed with pGEX-GST-HIF-3α-TAD or pGEX-5X1 were grown at 37°C in Luria-Bertani media containing 100 µg of ampicillin/mL and induced with 0.1 mM isopropyl-β-D-thiogalactoside (IPTG) for 4 h. Cells were then harvested, and glutathione S-transferase (GST)-HIF-3α-TAD and GST proteins were prepared essentially as described previously (Liu et al., 2004). The integrity and yield of purified GST proteins were assessed by SDS-polyacrylamide gel electrophoresis followed by Coomassie blue staining.

The purified wild-type or mutant GST-HIF-3α-TAD-N fusion proteins (20 µg) were incubated in kinase buffer [0.2 M MOPS (morpholinepropanesulfonic acid; pH 7.4), 0.5 M EDTA, 0.1 M Mg(CH₃COO)₂] in the presence of active AKT1 (50 mU) (#SRP0353; Sigma Aldrich) and 1 μCi/μL [γ-³³P]ATP (#NEG302H; PerkinElmer). After 30 min of incubation at 30°C, samples were loaded onto a 10% SDS gel, and after electrophoresis and blotting onto a polyvinylidene difluoride membrane, phosphorylated proteins were visualized by phosphorimaging.

Isolation of total RNA was performed using RNeasy Mini Kit (Qiagen, Hilden, Germany) following the manufacturer’s instructions. Total RNA concentration and purity were measured on a NanoDrop spectrophotometer. One μg of total RNA was used for cDNA synthesis using iScript cDNA Synthesis Kit (Bio-Rad, München, Germany).

Quantitative real-time PCR was performed in an Applied Biosystems 7,500 Real-Time PCR System (Life Technologies, Helsinki, Finland) by using an Applied Biosystem Power SYBR^® green PCR master mix (Life Technologies). The following primer sets were used: human HIF-3α-F (5′-GTC​GGA​GAG​TAT​CGT​CTG​TGT​C-3′), human HIF-3α-R (5′-TCT​GCG​AGA​GTG​TTG​CTC​CGT​T-3′), human β-actin-F (5′-GTT​GTC​GAC​GAC​GAG​CG-3′) and human-β-actin-R (5′-GCA​CAG​AGC​CTC​GCC​TT-3′). The experiments for each data point were carried out in triplicate. The relative quantification of gene expression was determined using the ΔΔCt method (Livak and Schmittgen, 2001).

Co-immunoprecipitation and Mass spectrometry were conducted as follows. HEK 293 cells were transfected with an expression vector encoding full-length V5-tagged HIF-3α1 WT and full-length HA-tagged PKB and incubated for 24 h. Cells were collected by scraping and lysed in lysis buffer containing 50 mM Tris/HCl (pH 7.5), 150 mM NaCl, 10 mM Na₄P₂O₇, 1% Triton X-100, 2 mM EDTA, 2 mM EGTA, and complete protease inhibitor cocktail tablet (Roche). The lysate was incubated with continuous shaking at 4°C for 20 min and then centrifuged at 12,000 g for 15 min at 4°C. To recover anti-V5, 150 µg of protein was incubated with 2 µg of antibody for 1 h at 4°C, followed by the addition of protein G Sepharose beads (30 µL per reaction mixture; #GE17-0,618-01; GE Healthcare) for 12 h. The beads were washed five times with lysis buffer, and the protein was recovered. The pellets were dissolved in 2X Laemmli buffer and loaded onto a 10% SDS gel. Subsequently, the gel was washed briefly with ddH₂O and stained with Coomassie Blue for 1 h at room temperature. After overnight destaining, the gel piece containing a ∼72 kDa polypeptide was excised. The gel piece was completely destained with 40 mM ammonium bicarbonate in 40% acetonitrile, treated with 20 mM DTT for 30 min at room temperature, and alkylated by adding 45 mM iodoacetamide (IAA) for 30 min at room temperature. The gel piece was washed once with the same solvent and twice with ddH₂O. Trypsin (Sigma proteomics grade, 5 ng/μL in 40 mM ammonium bicarbonate/9% acetonitrile) was added (20 µL) and incubated overnight at room temperature.

For data acquisition, a Nano LC 1000 system (Thermo Scientific, United States) coupled with a Lumos Fusion Orbitrap mass spectrometer (Thermo Scientific, United States) was used. Samples were trapped onto a Symmetry C18 100Å 5 µm 0.180 × 20 mm precolumn and eluted over nanoEaseBEH C18 130Å 1.7 µm 0.075 × 150 mm analytical column. The LC conditions included a flow rate of 0.3 μL/min and a gradient from 97% solvent A (0.1% formic acid in water) to 35% solvent B (0.1% formic acid in acetonitrile) over 90 min. Mass spectrometry data were acquired in data-dependent acquisition (DDA) mode with survey scans at a resolution setting of 120.000 over a mass range of 375–1,500. The automatic gain control (AGC) target was set to 4e5, and the maximum fill time was under 100 ms. MS/MS interrogation was triggered for ions with multiple charges and intensity above 2.5e4 in the Orbitrap, while ions between 1e3 and 2.5e4 were analyzed in the ion trap. AGC for Orbitrap analysis was set to 5e4 with a maximum fill time of 200 ms and a resolution of 15,000. The ion trap was filled for a maximum of 300 ms with AGC set to 10.000. HCD fragmentation with 30% collision energy was used for MS/MS.

Densitometry data were plotted as fold induction of relative density units, with the zero-value absorbance in each figure set arbitrarily to 100%. Statistical comparisons of absorbance differences were performed by the Mann-Whitney test (Statview 4.5, Abacus Concepts, Berkeley, CA), and p values p ≤ 0.05 were considered significant. Luc values presented are means +SEM. Results were compared by ANOVA for repeated Luc measurements followed by the Newman-Keuls test. A probability level p ≤ 0.05 was accepted as significant.

To investigate the potential role of the PI3K/PKB signaling pathway in insulin-mediated modulation of HIF-3α1 protein expression, we conducted experiments using MCF-7 cells because they express insulin receptors and respond to insulin stimulation (Kietzmann et al., 1999). First, we treated MCF-7 cells with insulin and demonstrated by qRT-PCR analysis that insulin upregulated HIF3A mRNA levels under both normoxia and mild hypoxia, which is in line with previous studies (Kietzmann et al., 2003; Heidbreder et al., 2007) (Supplementary Figure S1). Next, we transfected MCF-7 with wild-type V5-tagged HIF-3α1 and subsequently treated for a duration of 4 h with insulin, wortmannin (a potent PI3K inhibitor), and BML-257 (a specific PKB inhibitor) alone or in combination under normoxia or mild hypoxia. Our results demonstrated that treatment with insulin alone enhanced HIF-3α1 protein levels by about 2-fold under normoxia and by about 2.5-fold under mild hypoxia (Figures 1A, B). In contrast, when cells were treated with wortmannin alone or in combination with insulin no changes were observed in HIF-3α1 protein levels compared to the control cells, although PKB activation was inhibited (Figures 1A, B). This suggests that the insulin-induced HIF-3α1 expression involves the PI3K pathway.

However, treatment of cells with BML-257, a specific PKB inhibitor, alone resulted in a significant reduction of HIF-3α1 protein expression by about 50% under both normoxic and hypoxic conditions. Furthermore, the combined treatment of BML-257 and insulin did not reverse this decrease in HIF-3α1 protein levels (Figures 1A, B). Thus, these findings strongly suggest that the activation of HIF-3α1 protein expression by insulin is mediated through the PKB signaling pathway.

Next, we investigated whether PKB downstream targets, such as glycogen synthase kinase-3 (GSK-3) or mammalian target of rapamycin (mTOR), are involved in the regulation of HIF-3α1 protein expression. To achieve this, we treated MCF-7 cells with insulin and LiCl both well-known inhibitors of GSK-3 and rapamycin, a mTOR inhibitor alone or in combination under normoxia and mild hypoxia. Consistent with our previous findings, treatment with insulin alone increased HIF-3α1 protein expression under both normoxic and mild hypoxic conditions. However, treatment with LiCl and rapamycin alone slightly reduced HIF-3α1 protein levels compared to the control cells (Figures 1C, D). Furthermore, exposure of cells to these inhibitors in combination with insulin reduced insulin-mediated induction of HIF-3α1 protein levels but to a lesser extent than Wortmannin and BML-257 (Figures 1C, D). Collectively, these results suggest that PKB itself may regulate HIF-3α1 expression in response to insulin more potent and directly than the PKB downstream targets mTOR and GSK3.

To substantiate the finding that insulin is able to modulate HIF-3α1 protein expression via PKB, we first used computational approaches to search for the presence of putative PKB phosphorylation sites within HIF-3α1. We found that serine 524 matched the consensus motif, which is also highly conserved in mouse and rat HIF-3α (Figure 2A). This result was confirmed by mass spectrometry (MS) analysis of HIF-3 immunoprecipitates, which showed that serine 524 is indeed phosphorylated (Supplementary Figure S2). Notably, serine 524 is located within the transactivation domain (TAD), suggesting that phosphorylation of this residue may affect HIF-3α transactivity.

To investigate this, we transfected HepG2 cells with a Gal4 construct expressing a fusion protein with the wild-type TAD of HIF-3α (Gal4-HIF-3α-TAD) or a fusion protein in which the respective PKB site has been mutated (Gal4-HIF-3α-TAD-S/A). The cells were then treated with insulin for 8 h under normoxia and mild hypoxia. We found that mild hypoxia induced Luc activity by about 200% in both the wild-type and mutant construct (Supplementary Figure S3). Treatment with insulin enhanced Luc activity by about 600% under normoxia and by about 450% under mild hypoxia using the wild-type Gal4-HIF-3α-TAD construct, whereas insulin had no effect on the Luc activity on the mutant Gal4-HIF-3α-TAD-S/A construct (Supplementary Figure S3). Thus, insulin appears to regulate HIF-3α transactivity via serine 524.

To further support the idea that HIF-3α1 might be a direct target of PKB, we performed a phosphorylation assay. When using the GST-HIF-3α-TAD protein as a substrate we found that PKB phosphorylated the HIF-3α-TAD, whereas the GST protein alone was not phosphorylated (Figures 2B, C). Next, we used the GST-HIF-3α-TAD-S/A protein in which the putative PKB site was mutated and found that PKB no longer phosphorylated this protein (Figures 2B, C).

To validate the notion that HIF-3α1 is a direct target of PKB, we performed coimmunoprecipitation analysis to examine whether PKB forms a complex with HIF-3α1. Indeed, the results demonstrated an interaction between PKB and wild-type HIF-3α1 as they could be found in the same complex. Conversely, when lysates from cells expressing the HIF-3α1 S524A mutant were used, no interaction between PKB and HIF-3α1 could be detected (Figure 2D).

To determine whether the insulin-mediated HIF-3α1 regulation involves ubiquitination, we first performed coimmunoprecipitation analysis using proteins from HEK 293 cells transfected with vectors for V5-tagged wild-type HIF-3α1 or the HIF-3α1 S524A mutant. Subsequently, the cells were treated with insulin and the proteasome inhibitor MG132 for 4 h before harvesting. Although both wild-type and mutant HIF-3α1 proteins were found to be ubiquitinylated following immunoprecipitation with the V5-tag antibody, the extent of the ubiquitination was different. Treatment of cells with insulin inhibited the ubiquitination of wild-type HIF-3α1 whereas no differences in ubiquitination could be detected with the HIF-3α1 mutant upon insulin treatment (Figure 2E). In addition, we determined the half-life of HIF-3α1 and the respective PKB site mutant and could show that the HIF-3α1 S524A mutant´s half-life is prolonged compared to that of wild-type HIF-3α1 (Supplementary Figure S4). Together, these data suggest that insulin can modulate the protein expression of HIF-3α1 via PKB and that serine 524 is crucial for the binding of PKB within HIF-3α1 and may alter its protein stability.

HIF-3α has been proposed to play important roles in regulation of cell proliferation and growth (Tanaka et al., 2009; Ando et al., 2013; Xue et al., 2016; Zhou et al., 2018). The proper function of a protein in cells and tissues is tightly regulated by several mechanisms; of these, phosphorylation is one of the most important. Therefore, we next examined the effects of substituting serine 524 in HIF-3α1 on the ability of cells to form colonies. To this end, cells expressing either wild-type HIF-3α1 or the HIF-3α1 S524A mutant were seeded at low density and grown for 1 week before being subjected to crystal violet staining. Counting of cells showed that overexpression of HIF-3α1 variants (WT and S/A) reduced the number of colonies compared to the control (Ctl) (Figure 3A). Interestingly, overexpression of HIF-3α1 WT produced about 2.5-fold larger colonies compared to the control (Ctl). Notably, transfection of cells with the HIF-3α1 S524A induced even more widespread colony growth, where the colony size was about 8-fold larger in comparison to that of the control cells (Figures 3B–D). In line with this, the reduction in the number of colonies as well as the production of larger colonies could be completely abolished by using two different siRNAs against HIF-3α1 (Supplementary Figure S5).

Next, we investigated whether insulin might have an influence on the HIF-3α1 modulated cell viability as well as colony formation. Again, HIF-3α1 WT and the corresponding HIF-3α1 S524A mutant expressing cells were used. We found that application of insulin improved the cell viability of HIF-3α1 WT transfected HEK 293 cells by approximately 180% and increased it by approximately 240% in MCF-7 transfected cells. (Figures 3E, F). In addition, the number of colonies in HIF-3α1 WT transfected cells increased upon insulin treatment (Figures 3G, H). In contrast, treatment of cells with insulin did not affect viability or colony number of cells transfected with the HIF-3α1 S524A mutant (Figures 3E–H). Thus, these findings further support the conclusion that serine 524 is crucial for the insulin-mediated regulation of HIF-3α1.

Since we observed that overexpression of HIF-3α1 WT or the HIF-3α1 S524A mutant reduced colony number but increased colony size compared to control cells, we next investigated whether HIF-3α1 also affects cell adhesion. To do this, we transfected cells with HIF-3α1 WT or the HIF-3α1 S524A mutant and subjected them to rotation on an orbital shaker for 8 h. After rotation, ∼70% of the control cells were detached. Similarly, cells overexpressing HIF-3α1 WT showed a comparable level of adhesion (Figures 4A, B). By contrast, only about 40% of cells were detached when the HIF-3α1 S524A mutant construct was used (Figures 4A, B). Additionally, we investigated the expression of E-cadherin which plays an important role in cell-cell adhesion and tissue formation as well as of α-smooth muscle actin (α-SMA), a protein constituent of the contractile apparatus. Overexpression of full-length HIF-3α1 WT appeared to induce both E-cadherin and α-SMA expression (Figures 4C, D). Consistent with the reduced detachment in adhesion assay, the highest expression levels of these two proteins were seen in cells expressing the HIF-3α1 S524A mutant (Figures 4C, D).

The above findings indicated that HIF-3α1 and its insulin-dependent regulation may play a role in cell adhesion as well as colony-forming ability of cells. Next, it was investigated whether HIF-3α1 mediates the mobility of cells in a wound healing assay. The results of the wound healing assay showed that cells overexpressing wild-type HIF-3α1, but not the HIF-3α1 S524A mutant, migrated 2-fold faster than non-transfected cells (Figures 5A, B). As expected, exposure of wild-type HIF-3α1 transfected cells with insulin promoted cell migration and increased the rate of wound closure by about 40% and closed the wound in 72–74 h when compared with the control (Figures 5C, D), whereas insulin was able to cause a decrease in the rate of wound closure in HIF-3α1 S524A transfected cells in comparison to the control (Figures 5C, D). Overall, these findings offer the potential to affect HIF-3α-dependent processes by targeting PKB.