Hypoxia-inducible Factor-2α-dependent Hypoxic Induction of Wnt10b Expression in Adipogenic Cells*

Background: In obesity, enlarged adipocytes become hypoxic, which inhibits adipocyte differentiation. Results: Hypoxia induces the expression of Wnt1 and Wnt10b in both human and mouse adipogenic cells in a hypoxia-inducible factor (HIF)-2α-dependent manner. Conclusion: Hypoxia enhances the secretion of Wnt ligands, which trigger Wnt signaling in the neighboring cells. Significance: Wnt10b is a novel HIF-2α-specific target gene and a paracrine factor in hypoxia. Adipocyte hyperplasia and hypertrophy in obesity can lead to many changes in adipose tissue, such as hypoxia, metabolic dysregulation, and enhanced secretion of cytokines. In this study, hypoxia increased the expression of Wnt10b in both human and mouse adipogenic cells, but not in hypoxia-inducible factor (HIF)-2α-deficient adipogenic cells. Chromatin immunoprecipitation analysis revealed that HIF-2α, but not HIF-1α, bound to the Wnt10b enhancer region as well as upstream of the Wnt1 gene, which is encoded by an antisense strand of the Wnt10b gene. Hypoxia-conditioned medium (H-CM) induced phosphorylation of lipoprotein-receptor-related protein 6 as well as β-catenin-dependent gene expression in normoxic cells, which suggests that H-CM contains canonical Wnt signals. Furthermore, adipogenesis of both human mesenchymal stem cells and mouse preadipocytes was inhibited by H-CM even under normoxic conditions. These results suggest that O2 concentration gradients influence the formation of Wnt ligand gradients, which are involved in the regulation of pluripotency, cell proliferation, and cell differentiation.

Direct measurements of O 2 concentrations in adult tissues have revealed hypoxic conditions characterized by O 2 levels ranging from 2 to 9% . Tissue hypoxia is caused by an imbalance between oxygen supply and metabolic demand. Within the microenvironment of adult stem cells, which are often found distal to blood vessels, O 2 concentrations are estimated to be as low as 1% (7.2 mmHg) (1)(2)(3). In obesity, enlarged adipocytes become hypoxic because they expand up to 150 -200 m in diameter, which exceeds the normal O 2 diffusion distance of ϳ100 m from the vessels (4,5). Hypoxiainducible factor (HIF) 2 plays a major role in adaptive responses to hypoxia, including angiogenesis, vasodilation, and glycolysis, by inducing target genes such as those encoding vascular endothelial growth factor (VEGF), inducible nitric oxide synthase, and many glycolytic enzymes. HIF is a heterodimeric transcription factor consisting of two subunits, ␣ and ␤. HIF binds to E-box-like DNA sequences within its target genes, named hypoxia-responsive elements (HREs). The first HIF-␣ isoform, HIF-1␣, was identified by HRE affinity purification, whereas HIF-2␣/EPAS-1 was discovered in a homology search (1,6).
In contrast to the constitutive HIF-␤ subunit, the two HIF-␣ isoforms are ubiquitinated and rapidly degraded under normoxia and are stabilized under hypoxia. Although both HIF-1␣ and HIF-2␣ are co-expressed in many cell types and share many target genes, HIF-1␣ and HIF-2␣ knock-out mice both display embryonic lethality (7,8). Following exposure to hypoxia, HIF-1␣ protein levels increase maximally after 4 h and decrease after 24 h, whereas HIF-2␣ protein levels steadily increase after 24 h. Thus, acute hypoxia preferentially increases HIF-1␣ protein levels, whereas chronic hypoxia enhances HIF-2␣ protein levels (9,10). Chromatin binding analysis in MCF7 breast cancer cells revealed higher levels of HIF-1␣ occupancy in most of the hypoxia-induced genes and especially in glycolytic pathway genes, whereas HIF-2␣ showed preferential binding for only a few genes such as OCT4, Arginase1, and cyclin D1 (11)(12)(13).
Wingless/integration (Wnt) ligands are secreted glycoproteins that control tissue remodeling by regulating the proliferation and differentiation of specific cell types (14,15). Canonical Wnts bind to transmembrane receptors, frizzled (Fzl) and lipoprotein receptor-related protein 5 or 6 (LRP5/6). These interactions between canonical Wnts and their receptors cause dissociation of the Axin-APC-GSK3b complex, resulting in stabilization of ␤-catenin. Stabilized ␤-catenin then enters the nucleus and promotes the transcription of Wnt target genes by interacting with its transcriptional partner, lymphoid enhancer-binding factor (LEF)/T cell factor (TCF) (16). A canonical Wnt isoform, Wnt10b, maintains preadipocytes in an undifferentiated state through the Wnt/␤-catenin signaling pathway and by repressing adipogenesis (17)(18)(19). In this study, we found that the Wnt10b/Wnt1 locus is occupied by HIF-2␣ but not by HIF-1␣. In addition, we demonstrate for the first time that hypoxia induces Wnt10b expression in a HIF-2␣-dependent manner.
Cell Culture and Adipocyte Differentiation-3T3-L1 (ATCC, catalog number CL-173) preadipocytes and NIH3T3 (ATCC, catalog number CRL-1658) cells were maintained in DMEM containing 10% (v/v) bovine calf serum. For differentiation of preadipocytes into adipocytes, postconfluent 3T3-L1 cells were exposed to a standard mixture (MDI) composed of 0.5 mM IBMX, 1 M dexamethasone, and 5 g/ml insulin in DMEM containing 10% FBS for the first 2 days. Cells were then cultured in DMEM supplemented with 10% FBS and containing 5 g/ml insulin for the following 2 days, after which they were maintained in DMEM supplemented with 10% FBS in a humidified atmosphere of 95% air and 5% CO 2 at 37°C. The medium was changed every 2 days. hADSCs were obtained from two different donors (catalog number 510070, lot numbers 1199 and 2152, Invitrogen) and expanded in basal medium. For adipogenesis, hADSCs were cultured in adipogenic medium (catalog number A1007001, Invitrogen) according to the manufacturer's instructions. HIF-2␣ knock-out mouse embryonic fibroblasts (MEFs) were isolated from HIF-2␣ Ϫ/Ϫ embryos at embryonic day 12.5 and cultured in DMEM containing 10% (v/v) FBS as described previously (20). Hypoxic treatment of cells was achieved by incubating the cells in an anaerobic incubator (Ͻ0.5% O 2 , Model 1029, Forma Scientific, Inc.) or an InVivo 2 200 hypoxia work station (5% or 3% O 2 , Ruskin). Accumulated lipids in adipocytes were visualized and measured by staining with Oil Red-O, as described previously (21).
Quantitative Reverse Transcription-PCR (qRT-PCR)-Steadystate mRNA expression was measured by quantitative real-time PCR using Power SYBR Green PCR master mix (Applied Biosystems) on an ABI 7000 real-time PCR system. The C t value of a target mRNA was normalized against the C t value of endogenous 18 S rRNA (⌬C t ϭ C ttarget Ϫ C t18 S ). C t is the threshold cycle of quantitative PCR (qPCR) defined by an ABI 7000 realtime PCR system. The relative mRNA level of a target gene is obtained by 2 Ϫ⌬⌬ Ct ; ⌬⌬C t ϭ ⌬C ttreated value Ϫ C tuntreated value . Copy numbers of Wnt10b mRNA were determined by using a plasmid encoding mouse Wnt10b cDNA (accession number: NM_011718, 334 -1503 bp) as a standard PCR template for the standard curve method as described previously (22). Copy number was calculated using the equation: 1 g of 1000 bp of DNA ϭ 9. Reporter Analysis-3T3-L1 preadipocytes were seeded at a density of 2 ϫ 10 4 cells/well in 24-well plates, and after 24 h, the cells were transfected with Wnt10b reporter constructs (250 ng) or with HIF-1␣ (200 ng) or HIF-2␣ (200 ng) expression vectors, together with the pRL-CMV plasmid (50 ng) using the Lipofectamine reagent (Invitrogen). After 24 h, transfected cells were cultured in normoxic or hypoxic conditions with or without MDI or Bt2-cAMP (300 M). NIH3T3 or 293 cells were seeded at a density of 2 ϫ 10 4 cells/well in 24-well plates, and after 24 h, the cells were transfected with 250 ng of Wnt/␤catenin reporter plasmid (Super 8ϫTOPFlash, Addgene Inc.). After 24 h, the transfected cells were treated with 1 ml of the indicated CM, which was used to culture 3T3-L1 or hADSCs in normoxic or hypoxic conditions. The control medium was DMEM containing 10% FBS. After exposure to the CM for 24 h, the transfected cells were harvested, and the luciferase activity was measured using the luciferase assay system (Promega). Luciferase activity was normalized against Renilla luciferase activity.
Statistical Analysis-All quantitative measurements were performed in at least two independent experiments. Data were presented as the mean Ϯ S.D. or mean Ϯ S.E. For comparisons between two groups, p values were calculated using paired twotailed Student's t tests. A p value of less than 0.05 was considered statistically significant.

Hypoxia Increases Expression of the Wnt10b Gene in Mouse
3T3-L1 Cells-Previously, we showed that both physiological (3% O 2 ) and severe hypoxia (Ͻ0.5% O 2 ) inhibit adipogenesis of mouse 3T3-L1 preadipocytes by blocking the induction of C/EBP␣ and PPAR␥ adipogenic transcription factors, which are the target genes of C/EBP␤ (21). During adipogenesis of 3T3-L1 cells, the level of ␤-catenin protein gradually decreased, whereas under hypoxia, it remained stable. The level of HIF-1␣ protein was also stabilized under hypoxic conditions (Fig. 1A). Following treatment with adipogenesis-inducing hormones (IBMX, Dexamethasone, Insulin, MDI) for 6 h, Wnt10b expression initially decreased to the level usually found in mature adipocytes regardless of the oxygen concentration. After treatment for 24 h and under physiological hypoxia (3-5% O 2 ), Wnt10b expression was comparable with the level found in preadipocytes, whereas after treatment for 48 h and under severe hypoxia (Ͻ0.5% O 2 ), the level of Wnt10b expression was greater than that in preadipocytes (Fig. 1, B and C). Similar to Wnt10b expression, the mRNA levels of Axin2, a target gene of Wnt/␤-catenin signaling, initially decreased and then gradually recovered after 24 h of hypoxic exposure (Fig. 1D). Exposure to severe hypoxia (Ͻ0.5% O 2 , for 48 h) also caused an increase in mRNA levels of Wnt10b as well as in BCL2/adenovirus E1B 19-kDa protein-interacting protein 3 (Bnip3), a hypoxia-inducible gene in 3T3-L1 preadipocytes and MEFs (Fig. 1E).
IBMX, a nonselective phosphodiesterase inhibitor, is responsible for suppressing Wnt10b expression by activating cAMP signaling (23,24). Bt2-cAMP, a cAMP mimetic, was sufficient to repress Wnt10b within 6 h, even under hypoxia. However, after 24 h, hypoxia facilitated the recovery of Wnt10b expression even in the presence of MDI or Bt2-cAMP. Hypoxia did not alter the phosphorylation of CREB (Fig. 2, A and B). Instead, hypoxic exposure for 24 h enhanced the binding of CREB to the Wnt10b promoter, which was inhibited by MDI or Bt2-cAMP (Fig. 2, C  and D). The reporter assay revealed that both MDI and Bt2-cAMP repressed Wnt10b promoter activity (from Ϫ715 to ϩ286 bp), which contains a cAMP-response element. In contrast to endoge-nous Wnt10b expression, hypoxia did not lead to recovery of the Wnt10b promoter activity (Fig. 2E), implying that hypoxia indirectly increased CREB binding to the Wnt10b promoter through other regulatory regions.
HIF-2␣ Occupies the Wnt10b/Wnt1 Locus in Mouse Chromosome 15-An extended search up to Ϫ15 kb upstream of the mouse Wnt10b gene in chromosome 15 revealed additional HREs near the Wnt1 gene, which is encoded by an antisense strand of the Wnt10b gene (Fig. 4A). The Wnt1 coding region was flanked by several HREs. ChIP analysis using primer sets covering the Wnt10b/Wnt1 locus showed that HIF-2␣ bound to HREs located Ϫ0.2 kb from the Wnt1 transcription start site (Ϫ11 kb from the Wnt10b gene), whereas HIF-2␣ did not bind to the region located Ϫ7.0 kb from this site, where no HREs were found. By contrast, HIF-1␣ binding was not observed in the Wnt10b/Wnt1 locus (Fig. 4B). ChIP analysis using anti-H3K9K14Ac or anti-H3K4me3 antibodies revealed that, on the transcription start site of either the Wnt10b or the Wnt1 gene, hypoxic exposure increased the acetylation of Lys-9/Lys-14 residues and trimethylation of the Lys-4 residue of histone3, which are found in active chromatin domains (Fig. 4C).

Hypoxia Induces Expression of Wnt10b and Wnt1 Genes in
Mature Adipocytes-To test whether hypoxia induces Wnt10b expression in 3T3-L1 mature adipocytes, we differentiated 3T3-L1 cells into mature adipocytes and then exposed them to hypoxia as indicated in Fig. 5A. Hypoxic exposure inhibited lipid accumulation in mature adipocytes (Fig. 5A) and promoted expression of Wnt10b, Wnt1, and Axin2, whereas PPAR␥2 was repressed (Fig. 5B). In addition, hypoxia in mature adipocytes led to stabilization of HIF-1␣ and HIF-2␣ proteins and expression of Wnt10b and Wnt1 proteins (Fig. 5C). Hypoxic exposure also increased the protein stability of ␤-catenin, but decreased the protein levels of PPAR␥ and C/EBP␣ in mature adipocytes (Fig. 5D). On the other hand, hypoxia failed to increase the expression of Wnt10b in mature adipocytes derived from shHIF-2␣ 3T3-L1 cells (Fig. 5E). ChIP analysis confirmed that hypoxic exposure increased HIF-2␣ binding to HREs within the Wnt10b gene in mature mouse adipocytes (Fig. 5F).
Similarly to the mouse Wnt10b/Wnt1 locus, the human Wnt1 gene is encoded by an antisense strand of the Wnt10b gene in chromosome 12q13. Upon exposure to hypoxia, expression of both Wnt10b and Wnt1 was induced in hADSCs (Fig. 6,  A-C). We induced the differentiation of hADSCs into mature adipocytes for 12 days and exposed them to hypoxia for an additional 12 days as indicated in Fig. 6, D and E. Hypoxia promoted the expression of both Wnt10b and Wnt1 in mature adipocytes derived from hADSCs (Fig. 6F).
H-CM Contains Wnt Signals-H-CM, N-CM, and Np-CM were spent media harvested from adipocytes or preadipocytes cultured under hypoxic or normoxic conditions (Fig. 7A). To test whether hypoxic cells secrete Wnt ligands into the culture media, we transfected NIH3T3 cells with a reporter plasmid encoding luciferase driven by ␤-catenin/TCF binding sites and then treated cells with the various CM for 24 h under normoxic conditions. Based on luciferase activity, H-CM and Ha-CM dis-played Wnt activity of ϳ1-1.25 ng/ml recombinant mouse Wnt3a protein (rmW3a) (Fig. 7B). We next induced adipogenesis of 3T3-L1 preadipocytes cultured under normoxic conditions for 6 days in various culture media. Oil Red-O staining showed that MDI in N-CM induced lipid accumulation in 3T3-L1 cells comparable with that in fresh medium, but MDI in H-CM did not (Fig. 7C). Treatment with H-CM, Ha-CM, and Wnt3a-CM prevented MDI-mediated induction of PPAR␥ and C/EBP␣ expression (Fig. 7D). In addition, treatment with H-CM or Wnt3a-CM increased the phosphorylation of the serine residue at position 1490 in LRP6, reduced the Axin1 protein  level, and elevated the Axin2 mRNA level (Fig. 7, E and F). These findings suggest that H-CM contains canonical Wnt signals, which can induce the phosphorylation of LRP6, destabilization of Axin1 protein, and expression of Axin2. In addition, H-CM and Ha-CM from cultured hADSCs contained Wnt signals and inhibited adipogenesis of hADSCs (Fig. 7, G-I).
Hu-CM, which is CM collected from cultured undifferentiated hADSCs under hypoxia, elevated LRP6 phosphorylation in HEK293 cells. The addition of DKK1, a Wnt antagonist, prevented this Hu-CM-mediated increase of LRP6 phosphorylation (Fig. 7J). We also collected Hu-CM from hADSCs exposed to hypoxia in the presence of IWP2 (Hu-CMϩIWP2), a Porcupine inhibitor. Porcupine is a membrane-bound acyltransferase that is essential for the secretion of Wnt ligands (27). Treatment with Hu-CMϩIWP2 did not elevate LRP6 phosphorylation in HEK293 cells (Fig. 7J). These results indicate that, in both mouse and human adipogenic cells, hypoxia induces the expression of Wnts including Wnt10b and Wnt1, which can be secreted as paracrine factors to trigger Wnt signaling in adjacent cells.

DISCUSSION
This is the first study to show that hypoxia increases the expression of Wnt10b and Wnt1 in both human and mouse adipogenic cells and that the hypoxic induction of both Wnt10b and Wnt1 depends on HIF-2␣ but not on HIF-1␣. Most of the hypoxic target genes are induced by both HIF-1␣ and HIF-2␣ or exclusively by HIF-1␣ (1, 28). Wnt10b and Wnt1 belong to a small subset of hypoxic target genes that are exclusively induced by HIF-2␣. Both Wnt10b and Wnt1 are canonical Wnt ligands, which trigger the stabilization of ␤-catenin. The stabilized ␤-catenin interacts with the transcriptional partners LEF/ TCF to activate Wnt target gene expression. Taken together with previous findings showing that HIF-1␣/␤ increases the expression of LEF-1 and TCF-1 (29), this study shows that hypoxia increases levels of not only Wnt downstream signaling molecules but also Wnt ligands. Our findings that H-CM elevated ␤-catenin/TCF-driven luciferase activity and LRP6 phosphorylation in normoxic cells and that DKK1, a Wnt antagonist, prevented H-CM from initiating Wnt signaling suggest that adipogenic cells exposed to hypoxia can secrete Wnt ligands. Because Wnt ligands can diffuse in interstitial fluid, O 2 concentration gradients can rebuild gradients of Wnt ligands in these microenvironments. Having measured the diffuse average concentration of Wnts in H-CM, we postulate that those cells in close proximity to hypoxic areas are exposed to a higher local concentration of Wnt ligands than the concentration found in H-CM (Fig. 7K).
Wnt10b expression was first found to be elevated in breast cancer cell lines and was therefore classified as a proto-oncogene. Wnt10b and other canonical Wnts including Wnt1, Wnt6, and Wnt10a are regulators of mesenchymal stem cell fate and inhibit adipogenesis and stimulate osteoblastogenesis. Ectopic expression of Wnt1 and Wnt10b suppresses the expression of both C/EBP␣ and PPAR␥2 in MDI-treated 3T3-L1 cells (30 -32). However, the molecular mechanisms by which Wnt/␤-catenin signaling suppresses the expression of C/EBP␣ and PPAR␥2 genes are poorly understood. Nonca-nonical Wnts such as Wnt5a also inhibit adipogenesis not through the ␤-catenin/TCF pathway, but through repression of PPAR␥ target genes (33,34). Wnt5a activates an NF-B essential modulator (Nemo)-like kinase, which phosphorylates a H3K9 methyltransferase, SET domain bifurcated 1 (SETDB1). Phosphorylated SETDB1 in turn forms a co-repressor complex with DNA-bound PPAR␥ to repress PPAR␥ target genes through H3K9 methylation (34).
Simon and colleagues (29) demonstrated that under conditions of 1.5% O 2 , which can maintain cell proliferation, the HIF-1␣/Arnt heterodimer increases the expression of LEF-1 and TCF-1, which leads to increased sensitivity of mouse embryonic stem cells to the proliferation-inducing effects of Wnt/␤catenin signaling. Wnt/␤-catenin signaling promotes G 1 progression through the induction of c-Myc, which plays dual roles, such as up-regulating cyclin D and repressing the cell cycle inhibitors p27 and p21 (35). By contrast, under severe hypoxia (1% O 2 ), cell growth is arrested at G 1 phase by the induction of p27 and p21 (36,37). Paraskeva and colleagues (38) examined whether severe hypoxia inhibits cell growth through repression of ␤-catenin/TCF activity. Using SW480 and HCT116 colon cancer cells with constitutively high ␤-catenin/ TCF activity, they showed that, under conditions of 1% O 2 , HIF-1␣ competes with TCF4 for binding to ␤-catenin, leading to dissociation of ␤-catenin from TCF4. On the other hand, the ␤-catenin/HIF-1␣ interaction enhances HIF-1␣-mediated transcription (38). Although hypoxia (Ͻ1% O 2 ) induces expression of both Wnt10b and Wnt1, which promote cell cycle reentry, mesenchymal stem cells can remain in a quiescent and undifferentiated state under severe hypoxia (3,39). By contrast, the cell proliferation rate is enhanced more by physiological hypoxia (1.5-3% O 2 ) than by normoxia (20% O 2 ) (20,29). Nevertheless, the molecular mechanisms responsible for such dramatic changes in the rate of cell proliferation within such a narrow concentration range of O 2 remain unknown. As paracrine factors, Wnt ligands can diffuse from a severe hypoxic area to areas in which O 2 concentration is above the threshold that would permit their mitogenic effects.
In obesity, adipocyte hyperplasia and hypertrophy can lead to many responses commonly associated with the growth of solid tumors, such as hypoxia, enhanced cytokine secretion, macrophage recruitment, and metabolic dysregulation (40,41). Many studies have revealed that hypoxia increases the expression of inflammatory adipokines, including IL-6, macrophage migration inhibitory factor, plasminogen activator inhibitor type1, and adrenomedullin (4,42,43). The production of inflammatory adipokines has been implicated in the development of insulin resistance and metabolic syndrome. Canonical Wnts lead to insulin resistance and dedifferentiation of adipocytes (44). Together with these findings, our results suggest that hypoxic areas in adipose tissue can cause dedifferentiation and insulin resistance by inducing Wnts as well as inflammatory adipokines.
Although HIF-1␣ and HIF-2␣ share a dimerization partner as well as many target genes, they play different roles in cell cycle regulation, metabolism, cell differentiation, and inflammation (1,13,45). This study proposes that the Wnt10b/Wnt1 locus is a novel HIF-2␣-specific target. However, the molecular mechanisms of how this locus distinguishes HIF-2␣ from HIF-1␣ remain to be investigated.