Induction of WNT11 by hypoxia and hypoxia-inducible factor-1α regulates cell proliferation, migration and invasion

Changes in cellular oxygen tension play important roles in physiological processes including development and pathological processes such as tumor promotion. The cellular adaptations to sustained hypoxia are mediated by hypoxia-inducible factors (HIFs) to regulate downstream target gene expression. With hypoxia, the stabilized HIF-α and aryl hydrocarbon receptor nuclear translocator (ARNT, also known as HIF-β) heterodimer bind to hypoxia response elements (HREs) and regulate expression of target genes. Here, we report that WNT11 is induced by hypoxia in many cell types, and that transcription of WNT11 is regulated primarily by HIF-1α. We observed induced WNT11 expression in the hypoxic area of allograft tumors. In addition, in mice bearing orthotopic malignant gliomas, inhibition with bevacizumab of vascular endothelial growth factor, which is an important stimulus for angiogenesis, increased nuclear HIF-1α and HIF-2α, and expression of WNT11. Gain- and loss-of-function approaches revealed that WNT11 stimulates proliferation, migration and invasion of cancer-derived cells, and increases activity of matrix metalloproteinase (MMP)-2 and 9. Since tumor hypoxia has been proposed to increase tumor aggressiveness, these data suggest WNT11 as a possible target for cancer therapies, especially for tumors treated with antiangiogenic therapy.

In this manuscript, we further investigated the mechanistic link between WNT11 and cancer. We report that WNT11 is induced by hypoxia in many cell types, and that transcription of WNT11 is regulated primarily by HIF-1α . Elevated endogenous WNT11 increases activity of MMP2 and MMP9, and promotes proliferation, migration and invasion of cancer-derived cells. Finally, elevated WNT11 expression was observed in the hypoxic area of allograft tumors and in human malignant glioma xenografts after treatment with antiangiogenic therapy. Although antiangiogenic therapy is thought to hold significant potential for the treatment of cancer, limited effectiveness, and increased tumor invasiveness and metastasis have been reported [23][24][25] . Taken together, our results provide a possible mechanism by which WNT11 induced by hypoxia and the HIF pathway regulates cell migration and invasion through activation of MMPs, and this could be a potential mechanism that drives the deleterious action of antiangiogenic therapy.

Results
Hypoxia or hypoxic mimetics induce expression of WNT11 in a wide variety of cell types. During a screen of Wnt genes regulated by hypoxia, we observed that mRNA of Wnt11 was robustly and specifically induced by hypoxia mimetics, such as cobalt chloride CoCl 2, deferoxamine (DFO), and dimethyloxalylglycine (DMOG) in fully differentiated adipocytes (Fig. 1A). In contrast, expression of Wnt5a and secreted frizzled related protein-5 (Supplementary Figure 1A) were both decreased under these conditions. Moreover, culture of ear mesenchymal stem cells (EMSCs) or C2C12 cells at various stages of differentiation in 1% O 2 increased Wnt11 (Fig. 1B,C and Supplementary Figure 1B,C). Interestingly, we also observed increased WNT11 protein in response to hypoxia or hypoxia-mimetic reagents in a macrophage cell line (Supplementary Figure 1D), and human cancer-derived cell lines (HeLa; Fig. 1D, MDA-MB-231; Fig. 1E). Expression of WNT11 protein was increased by DMOG in a concentration-dependent manner (Fig. 1E,F and Supplementary Figure 1E). Thus, induction of WNT11 appears to be a general response to hypoxia found in a wide variety of cell types.
To determine how rapidly hypoxia influences WNT11 expression, a time course after exposure of EMSCs to DMOG was assessed. Hypoxia increased Wnt11 mRNA as early as 2 hr, and WNT11 protein at 4 hr ( Fig. 1G-I; Supplementary Figure 1F). Two hours was also the first time point where increased Vegf mRNA was also detected, although the magnitude of induction was much larger in this case (Fig. 1G). VEGF is a well-characterized HIF target gene 7 , suggesting that elevated WNT11 may also be a direct target of HIF-1α and/or − 2α . Following DMOG treatment of MDA-MB-231 cells, HIF-1α is rapidly and transiently induced whereas the elevation of HIF-2α is slightly delayed but more sustained (Fig. 1H,I) To test whether effects of hypoxia to increase Wnt11 mRNA are transcriptional, we next explored the possibility that hypoxia affects Wnt11 promoter activity. Bioinformatic analysis suggested the presence of two potential HIF-binding DNA elements in the Wnt11 promoter (Supplementary Figure 1G). To this end, 1064 bp upstream of the mouse Wnt11 transcription start site was cloned into a luciferase reporter vector. After transient transfection into C2C12 cells, we observed that Wnt11 promoter activity was induced by both DMOG and hypoxia (Fig. 1J), suggesting that regulation of Wnt11 by hypoxia is mediated at the transcriptional level.
Von Hippel-Lindau (VHL) represses expression of WNT11. Hypoxia is well known to activate gene transcription through the HIF pathways, but signaling through mechanisms independent of HIFs have also been described 26 . To determine if WNT11 induction by hypoxia is HIF-dependent, we first studied cells and tissues lacking the VHL protein, which forms part of an E3 ubiquitin ligase that targets HIF-1α /2α for degradation. To generate Vhl −/− cells, EMSCs isolated from the Vhl fl/fl mouse were infected with lentivirus carrying Cre recombinase (for Vhl −/− ) or GFP (for control). Vhl −/− EMSCs had increased amounts of HIF-1α , HIF-2α and WNT11 proteins compared to non-infected or GFP-infected control EMSCs under standard culture conditions (21% O 2 ). The induction was similar to that observed with DMOG or 1% O 2 ( Fig. 2A,B). To study regulation of WNT11 by VHL, we analyzed mice lacking Vhl in specific tissues. Consistent with our observations on induction of WNT11 by hypoxia (Fig. 1), knockout of the tumor suppressor Vhl in liver or duodenum resulted in increased Wnt11 mRNA (Fig. 2C,D), suggesting that the classical hypoxia signaling pathway regulates Wnt11.
HIF-1α is an important transcriptional regulator of WNT11 expression. Activation of WNT11 by VHL disruption raised the possibility that HIF-1α and/or HIF-2α may regulate Wnt11 expression. Through gain-and loss-of-function experiments the specific effects of HIF-1α and/or HIF-2α on Wnt11 expression under hypoxic conditions were assessed. EMSCs from Hif-1α fl/fl or Hif-2α fl/fl mice were infected with lentivirus carrying either the GFP gene (for control) or Cre recombinase (for knockout), and EMSCs from Hif-1α-Tg fl-Stop or Hif-1α-Tg fl-Stop mice were infected with lentivirus-GFP (for control) or Cre recombinase (for overexpression). As shown in Fig. 3A, WNT11 was induced by DMOG treatment in control cells (non-infected cells or cells infected with lenti-GFP), but both basal and DMOG-induced WNT11 expression were markedly attenuated in Hif-1α  27 , we observed a reciprocal relationship between HIF-1α and HIF-2α expression: HIF-2α expression was elevated in HIF-1α knockout cells in response to hypoxia (Fig. 3A) and HIF-2α levels were decreased with HIF-1α overexpression (Fig. 3C). These data suggest that HIF-1α is both necessary and sufficient to regulate WNT11 expression.
We next evaluated the potential regulation of WNT11 by HIF-2α , and found that knockout of HIF-2α was associated with elevated WNT11 under basal and DMOG-induced conditions; however, both effects are likely secondary to compensatory increases in HIF-1α protein levels (Fig. 3D). Consistent with regulation of WNT11 residing predominantly with HIF-1α , overexpression of HIF-2α had little effect on WNT11 protein level ( Fig. 3B; right panel and 3E). HIF-1α binds to hypoxia response elements (HREs) of target genes as a heterodimer with ARNT 1-4,6 ; thus, we next considered whether ARNT also regulates Wnt11 transcription. Co-expression of HIF-1α and ARNT markedly induced Wnt11 promoter reporter activity, whereas expression of either protein separately, or co-expression of HIF-2α and ARNT had only minor effects (Supplementary Figure 2A,B). To explore promoter occupancy of HIF-1α in the regulation of Wnt11 expression, HIF-1α chromatin immunoprecipitation analysis was performed and we found that Flag-HIF-1α binds to the Wnt11 promoter in MDA-MB-231 cells (Fig. 3F). As a positive control, the VEGFα promoter was also specifically enriched by Flag-HIF-1α ChIP (Fig. 3F). These physical interactions provide a mechanistic basis for transactivation of Wnt11 promoter by HIF-1α . Taken together, these data provide evidence that Wnt11 expression is primarily regulated by VHL, HIF-1α and ARNT.  WNT/β -catenin signals regulate WNT11 expression in a variety of cell types and species 12 . To evaluate whether β -catenin is required for hypoxia-induced WNT11 expression, we generated EMSCs with stable expression of shRNAs against β -catenin. Whereas DMOG treatment caused a transient decline in β -catenin, basal WNT11 and expression of WNT11 after DMOG treatment was strongly suppressed in cells with a β -catenin deficiency (Fig. 3G). Consistent with data presented above, increased HIF-2α with β -catenin knockdown is not sufficient to elevate WNT11 (Fig. 3G). Although hypoxia did not induce accumulation of nuclear β -catenin (Supplementary Figure 2C), β -catenin is essential for WNT11 expression, in accordance with the previous reports that β -catenin regulates WNT11 expression in other contexts 12 . Furthermore, we observed additive induction effects on Wnt11 promoter activity when β -catenin was co-transfected with HIF-1α and ARNT expression vectors (Fig. 3H). Taken together, these data indicate that HIF-1α and β -catenin are required for WNT11 expression.

WNT11 is important for cell migration and invasion.
HIF-1α has extensive direct and indirect effects on gene expression 2,3,5 , and a subset of HIF-1α targets play important roles in diverse aspects of cancer biology, including migration and invasion [28][29][30] , and metastasis [31][32][33] . The induction of WNT11 by hypoxia, and the reported ability of WNT11 to influence cell mobility 15,[20][21][22][34][35][36] led us to hypothesize that WNT11 plays an intermediary role in the effects of HIF-1α on cell migration and invasion. To understand effects of WNT11 on cancer cells, we generated stable overexpression of WNT11 in BT-474 cells (Supplementary Figure 3A), which have amongst the lowest expression of WNT11 in cell lines tested (Supplementary Figure 3B). We observed that ectopic expression of WNT11 facilitates both migration and invasion of these cells in vitro (Fig. 4A,B). We then examined the functional consequences of endogenous WNT11 induced by hypoxia.  Figure 4E,F). Although DMOG had no effect on cell growth during this time frame, incubation in 1% O 2 decreased growth rate of both cell types (Supplementary Figure 4E,F). To minimize potential effects of differential growth rates on migration and invasion, we optimized plating numbers and used short incubation times during which effects of WNT11 on number of viable cells by MTT assay were not observed (Supplementary Figure 4G,H). These results confirm previous reports that WNT11 functions to increase mobility of cancer cells 15,22 , and demonstrate the intermediary role of WNT11 in regulation of migration and invasion by hypoxia. WNT11 regulates MMP activity. To determine the mechanism by which WNT11 regulates migration and invasion, we evaluated MMP activities using zymography. HIF-1α and β -catenin, two factors we describe as upstream of WNT11, both regulate mRNA and/or enzyme activity of MMP-2 and MMP-9 28,38-40 . Overexpression of WNT11 in EMSCs and BT-474 breast cancer cell lines resulted in higher MMP-2 and/or MMP-9 zymographic activities (Fig. 5A). To explore potential roles of endogenous WNT11 in modulation of MMP activities, we used shRNA to block induction of WNT11 by hypoxia. In MDA-MB-231 cells, DMOG increased MMP-9 and MMP-2 activity between 6 and 24 hrs after treatment, and knockdown of WNT11 decreased basal and DMOG-induced MMP activities (Fig. 5B; left panel). Knockdown of Wnt11 also decreased baseline and DMOG-induced WNT11 in EMSCs without influencing induction of HIF-2α ( Fig. 5B; right panel). Whereas hypoxia did not increase MMP-9 activity in MDA-MB-231 ( Fig. 5C; left panel) or EMSCs ( Fig. 5C; right panel), increases in MMP-2 activity were observed in both cell models. Knockdown of WNT11 decreased WNT11 and both MMP activities under standard and hypoxic cell culture conditions (Fig. 5C). We further determined whether altered MMP activity was explained by amount of MMP protein. Interestingly, we found decreased MMP-2 protein level in media corresponding to decreased WNT11 expression, and a reciprocal change arising from WNT11 overexpression (Fig. 5D). No change in MMP-9 protein levels was observed even when MMP-9 activity was regulated by WNT11 expression (Fig. 5A-C), suggesting regulation through other factors such as tissue inhibitor metalloproteinases (TIMP) and/or cell surface activator of proMMPs (MT-MMP). Furthermore, recombinant human WNT11 induced both MMP-2 protein and activity in the media of WNT11 knockdown cells (Fig. 5E) with decreased MMP-2 in cell lysates (Fig. 5E), suggesting that hypoxia-WNT11 pathway regulates MMP2 expression and secretion. In addition, MMP-2 inhibitor (ARP-100) suppressed migration of MDA-MB-231 cells expressing both GFP and exogenous WNT11 (Fig. 5F). Taken together, our results suggest that in response to hypoxia, elevated WNT11 promotes migration and invasion by increasing the activity of MMP-2 and MMP-9. expression in EMSC (lenti-Cre infectied Hif-2α-Tg fl-Stop ). (F) HIF-1α binds to WNT11 and VEGF promoters as assessed by ChIP analyses (n = 3 each condition). (G) WNT11 expression was suppressed in cells with a β -catenin deficiency. EMSCs stably expressing shRNAs against β -catenin or scrambled control were treated with 0.1 mM DMOG for 24 hrs and analyzed by immunoblotting. (H) Co-transfection with expression vectors for β -catenin, Hif-1α and ARNT stimulates further induction of Wnt11 promoter activity. Constructs encoding Hif-1α , Hif-2α , ARNT, β -catenin and Wnt11 promoter-luciferase reporter, were transiently transfected into HEK293T cells. Cells were harvested and luciferase activities were measured 48 hrs after transfection (n = 3). For panels (A-E,G), laminin, α -tubulin, ERK1/2 and lamin A/C were used as loading controls, WNT11 normalized to α -Tubulin was shown under the bots. For panels (F,H), values are mean ± s.e.m. *p < 0.05, **p < 0.01 Scientific RepoRts | 6:21520 | DOI: 10.1038/srep21520 WNT11 is induced by hypoxia and regulates tumor cell growth in vivo. To extend our results with cultured cells, we investigated whether WNT11 is expressed in hypoxic regions of allografts. After validation  Figure 5), we observed that WNT11 is elevated in the area of tumor hypoxia where HIF-1α and pimonidazole staining are colocalized (Fig. 6A). In addition, we observed decreased tumor growth in mice injected with 4T1 cells expressing shRNAs against Wnt11 (Fig. 6B,C), and reduced proliferation in WNT11 knockout cells in vitro (Fig. 4E), suggesting that impaired tumor progression in Wnt11 shRNA allograft is secondary to reduced proliferation. These observations are consistent with previous reports Scientific RepoRts | 6:21520 | DOI: 10.1038/srep21520 that WNT11 plays a key role in tumor progression 12,15 . Although these data are interesting, the impaired growth of Wnt11 knockdown cells precluded evaluation of metastasis or effects of antiangiogenic treatments.
As another in vivo model, we chose to evaluate malignant gliomas because these tumors secrete high levels of VEGF and are refractory to bevacizumab, a monoclonal antibody to inhibit VEGF 23,25 . Using an orthotopic malignant glioma model in which U87Δ EGFR cells are injected into the frontal lobe of athymic mice, we observed increased expression of WNT11 mRNA in animals treated with bevacizumab (Fig. 7A). However, we saw no effects on WNT1 and WNT10b (Fig. 7A), which were recently reported to be induced by hypoxia-HIF-2α in adipogenic cells 41 . Furthermore, bevacizumab treatment stimulated expression of WNT11 protein, as well as nuclear HIF-1α and HIF-2α (Fig. 7B). These data provide further evidence that WNT11 is induced by hypoxia in vivo, and raise the possibility that induction of WNT11 is involved in hypoxia-induced metastatic and invasive potential of tumor cells.

Discussion
This study provides the first demonstration that WNT11 expression is regulated by hypoxia and the HIF-1α pathway in normal and cancer-derived cells (Supplementary Figure 6). Previous studies have identified factors that directly or indirectly regulate WNT11 expression, including glial cell-line derived neurotrophic factor signaling and estrogen/ERR-α 12 (Supplementary Figure 6). Signaling by the Wnt/β -catenin pathway also enhances expression of WNT11, consistent with two conserved Tcf/LEF binding sites in the Wnt11 promoter 12 , and our results support a β -catenin requirement for the induction of WNT11 (Fig. 3G,H). Relationships between hypoxia and Wnt signaling have been explored in neural stem cells where HIF-1α enhances β -catenin activation and expression of downstream effectors LEF-1 and TCF-1 9 . Hypoxia causes nuclear accumulation of both HIF-1α and β -catenin in induced pluripotent stem cells 42 . We did not observe accumulation of nuclear β -catenin in cells under hypoxia or DMOG treatment (Fig. 3G, Supplementary Figure 2C), and other investigators have observed that hypoxia suppresses β -catenin nuclear localization and/or TOP-Flash activity 8,43 . Thus, effects of hypoxia on β -catenin appear to be cell type dependent.
Hypoxia is an important regulator of cell migration and invasion under physiological and pathological conditions [28][29][30][31][32][33] . In the hypoxia associated with cancer, HIF-1α and a subset of its targets play important roles in migration, invasion [28][29][30] , and metastasis [31][32][33] . Our finding that expression of WNT11 is induced by hypoxia/HIF-1α is of particular interest because WNT11 increases cell migration/invasion during both development 15,20,21,[34][35][36] and carcinogenesis 22 . In line with these findings, loss-and gain-of-function experiments reveal that WNT11 plays an essential role in triggering cell migration and invasion (Fig. 4, Supplementary Figures 4 and 6). Recent studies using time-lapse microscopy revealed that at least 14 hours were required for breast cancer cells to acquire enhanced motility under hypoxic conditions 44 . Our experiments are consistent with this timeframe, and cells required pretreatment with DMOG before the migration and invasion assay in order to induce WNT11 expression (Figs 1G-I and 4C,D). Interestingly, stimulation of RhoA and Rock by HIFs is important for regulation of cell motility under hypoxia 44 ; thus, the observation that WNT11 also induces activation of RhoA is consistent with an intermediary role 22 . It should be noted however that another report suggests a tumor suppressor role for WNT11 in hepatocellular carcinoma cells 37 .
Inhibiting key players in tumor angiogenesis such as VEGF and VEGF receptor 7 holds significant potential for treatment of cancers; however, a limited number of patients respond to antiangiogenic agents. Furthermore, this therapy may in some cases increase tumor invasiveness and metastasis [23][24][25] . For example, anti-VEGF treatment of malignant glioma appears to have therapeutic benefit in patients, but long-term outcomes have not been improved 45 , perhaps because treatment induces a phenotypic shift towards more aggressive forms 46 . Whereas tumor hypoxia following antiangiogenic treatment is proposed as the cause of poor outcome, little is known about the underlying mechanisms. Although elevated WNT11 was not observed in tumors derived from glioblastoma spheroids from two patients implanted in rat 23 (data not shown), we tested effects of bevacizumab in an orthotopic malignant glioma mouse model, and found that anti-VEGF treatment increased HIF-1α and HIF-2α , and also induced expression of WNT11 mRNA and protein (Fig. 7). The effects of antiangiogenic therapy in our study appear to be specific to WNT11 in that we did not see effects on WNT1 and WNT10b, which are induced by hypoxia and HIF-2α in adipogenic cells 41 . Further work will be required to evaluate whether elevated WNT11 is necessary or sufficient to cause the mesenchymal transition associated with long-term antiangiogenic therapy of glioma tumor tissue in vivo 46 . We also observed that growth of Wnt11 deficient allografts was significantly slower (Fig. 6B,C). In addition, we performed metabolomic experiments using conditioned media from Wnt11 knockdown and control 4T1 cells. As often occurs in proliferating cancer cells in the presence of oxygen 47 , we observed that control 4T1 cells consumed more fuel and amino acids and produced more lactic acid than Wnt11 knockdown cells (data not shown). It may be interesting if induction of WNT11 by tumor hypoxia protects against cell death, as was observed in cardiomyocytes 48,49 . Taken together, these data suggest WNT11 as a possible target for cancer therapies, especially with tumor hypoxia and/or tumors treated with antiangiogenic therapy. Blocking WNT11, either by inhibiting upstream regulators (HIFs and/or β -catenin) or with WNT11 neutralizing antibody 48,49 can be a novel therapeutic strategy. Further studies will be needed to examine effectiveness of this treatment.
Wnt11 luciferase reporter assay. A 1064bp DNA fragment in mouse Wnt11 promoter proximal region was cloned into XhoI/HindIII sites on pGL3-basic plasmid. HEK293T or C2C12 cells were transfected with Wnt11 luciferase plasmid and pRL Renilla Luciferase Vector, used as an internal control. Relative cellular luciferase activities were examined using the Dual-Luciferase Reporter Assay System (Promega, Madison, WI), and the luciferase signals were recorded with a Wallac Victor 1420 Plate Reader (Perkin-Elmer). The reporter activity was expressed as arbitrary luciferase units (firefly/renilla). Three independent experiments were carried out.
Plasmids. Lenti-Cre (LV-Cre pLKO.1) and LV-GFP plasmids were purchased from Addgene (Cambridge, MA). pcDNA3-HIF1α , HIF2α , ARNT, pcDNA3-β -catenin S33Y have been described previously 55 Hospital Research Institute) was used as template to clone mouse Wnt11 coding sequence by PCR. To prepare the vector, Cre coding sequence was cut out from LV-Cre pLKO.1 plasmid and mouse Wnt11 coding sequence was ligated in by using XbaI/KpnI sites. The plasmid was sequenced to ensure the validity of the DNA sequence. Both human and mouse WNT11 were stably knocked down by expression of an shRNA from the pLKO.1 puro vector. The shRNA was designed according to the manufacturer's instructions to target the following sequence of mouse WNT11: 5′-GGATGTGGGAGTTACAGAAAT-3′, human WNT11: 5′-TGTGGAAGCTACAGAAATA-3′, and β -catenin: GCACACGAATGGATCACAA.
Animal study. All animal procedures were approved by the University of Michigan Committee on the Use and Care of Animals, or were in accordance with the approved guidelines of the Animal Research Committee of Okayama University. Brain xenograft was essentially as described previously 25 . Briefly, 2 μ l of the U87Δ EGFR cell suspension (1.0 × 10 5 /μ l) was injected into the right frontal lobe of athymic mice (BALB/c-nu/nu; CLEA Japan). Mice were randomly assigned to two groups, and PBS or bevacizumab (6 mg/kg) were administered three times per week, intraperitoneally, starting on day 5 after tumor cell implantation. Animals were sacrificed 18 days after implantation and tumors dissected for analyses.
For the syngeneic breast cancer model, 4T1 cells derived from BALB/c mammary tumors were grown to 80% confluence, were trypsinized, mixed with Matrigel 1:1 by volume (BD Bioscience), and injected into the fourth inguinal mammary pad of 10-week old female mice (5.0 × 10 5 cells). Tumor size was measured by digital calipers, and tumor volume was calculated as (4/3) × π × (L/2) × (L/2) × (H/2). Five weeks after tumor cell injection, mice were euthanized, and tumor tissue harvested immediately and snap-frozen in liquid nitrogen. Hypoxia was detected using Hypoxyprobe following the manufacturer's protocol (Natural Pharmacia International). We monitored invasion and metastasis by bioluminescent imaging using the IVIS Lumina Imaging System (Xenogen) after luciferin (Promega) injection into mice.
Statistical analyses. All data are presented as mean ± s.e.m. and were analyzed by 2-tailed Student's t-test or analyses of variance (ANOVA). The differences were considered to be significant if p < 0.05. Whereas a power equation was not used to predetermine sample sizes, our sample sizes were chosen to be similar to those reported in previous publications for in vitro 12,13,15,[20][21][22] and animal studies 23-25 of this nature. No animals were removed from analyses.