Hypoxic ER stress suppresses β-catenin expression and promotes cooperation between the transcription factors XBP1 and HIF1α for cell survival

Hypoxia occurs in many human solid tumors and activates multiple cellular adaptive-response pathways, including the unfolded protein response (UPR) in the endoplasmic reticulum (ER). Wnt/β-catenin signaling plays a critical role in tumorigenesis, and β-catenin has been shown to enhance hypoxia-inducible factor 1α (HIF1α)-activated gene expression, thereby supporting cell survival during hypoxia. However, the molecular interplay between hypoxic ER stress, Wnt/β-catenin signaling, and HIF1α-mediated gene regulation during hypoxia remains incompletely understood. Here, we report that hypoxic ER stress reduces β-catenin stability, which, in turn, enhances the activity of spliced X-box–binding protein 1 (XBP1s), a transcription factor and signal transducer of the UPR, in HIF1α-mediated hypoxic responses. We observed that in the RKO colon cancer cell line, which possesses a Wnt-stimulated β-catenin signaling cascade, increased ER stress during hypoxia is accompanied by a reduction in low-density lipoprotein receptor-related protein 6 (LRP6), and this reduction in LRP6 decreased β-catenin accumulation and impaired Wnt/β-catenin signaling. Of note, β-catenin interacted with both XBP1s and HIF1α, suppressing XBP1s-mediated augmentation of HIF1α target gene expression. Furthermore, Wnt stimulation or β-catenin overexpression blunted XBP1s-mediated cell survival under hypoxia. Together, these results reveal an unanticipated role for the Wnt/β-catenin pathway in hindering hypoxic UPR-mediated responses that increase cell survival. Our findings suggest that the molecular cross-talks between hypoxic ER stress, LRP6/β-catenin signaling, and the HIF1α pathway may represent an unappreciated mechanism that enables some tumor subtypes to survive and grow in hypoxic conditions.

Hypoxia, a common stress state in most human solid tumors due to the highly chaotic architecture of blood vessels, exerts critical influences upon tumorigenesis and tumor growth (1). Tumor cells must survive in a hostile microenvironment characterized by low levels of oxygen, glucose, and pH value. In response to hypoxia, tumor cells undergo a variety of adaptive alterations for survival and growth, including O 2 -sensitive pathways that are orchestrated by the hypoxia-inducible factor family of transcription factors (HIFs) 3 (2). HIF1 is composed of two subunits, the constitutive nuclear HIF1␤ subunit and the O 2 -regulated cytosolic HIF1␣ subunit. Under normoxic conditions, HIF1␣ is subjected to hydroxylation catalyzed by prolyl hydroxylase enzymes, leading to its proteasomal degradation. Under hypoxia (typically less than 3-5% O 2 ) states, however, HIF1␣ is stabilized and translocates to the nucleus where it can dimerize with HIF1␤. The HIF1 heterodimer in turn recognizes and binds to hypoxia-response elements (HREs) in the genome to activate the transcription of a series of hypoxia-responsive genes (3). Clinical studies utilizing oxygen electrodes and molecular markers of hypoxia have shown that tumors are poorly oxygenated, whereas such oxygenation is extremely heterogeneous (4). Depending on the intensity and duration of the hypoxic stimuli, HIF1 can trigger activation of the transcription of genes involved in either adaptive responses or detrimental processes. When exposed to hypoxia, cells undergo temporary cell cycle arrest, reduced energy consumption, and secrete various survival and pro-angiogenic factors.
Hypoxia can also cause perturbations in the loading and processing of newly synthesized proteins, thus triggering endoplasmic reticulum (ER) stress in cancer cells (5)(6)(7). In order to cope with hypoxic tumor microenvironment-induced ER stress, tumor cells initiate an adaptive response that aims to restore cellular homeostasis, referred to as the unfolded protein responses (UPR) (8). The ER-localized transmembrane proteins, inositol-requiring enzyme 1 (IRE1), protein kinase R-like endoplasmic reticulum kinase (PERK), and activating transcription factor 6 (ATF6), act to mediate three canonical UPR signaling branches. Activation of the UPR allows cells to alleviate ER stress under hypoxic conditions within tumors. However, chronic activation of the UPR can also lead to cell death under severe and/or persistent ER stress (9). Emerging evidence indicates that activation of the UPR is linked to hypoxia tolerance and tumor growth (5). For instance, activation of PERK and the subsequent phosphorylation of eukaryotic initiation factor 2␣ (eIF2␣) act through two major mechanisms in protein translational control in response to hypoxic ER stress (10). Although globally repressing the initiation of mRNA translation and rapidly down-regulating protein synthesis, activation of the PERK-eIF2␣ branch can also selectively up-regulate the expression of a subset of genes by the transcription factor ATF4, thereby promoting cell survival (11). The transcription factor X-box-binding protein 1 (XBP1) has also been documented as another important player in handling hypoxic ER stress (12). XBP1 is regulated by the protein Ser/Thr kinase/ endoribonuclease IRE1␣, which is activated through autophosphorylation and dimerization/oligomerization upon ER stress (13). IRE1␣ catalyzes the nonconventional splicing of Xbp1 mRNA, thereby generating the spliced form XBP1s that activates a key transcriptional program of the UPR (8). Under hypoxic conditions, XBP1s has been shown to be a critical cell survival factor in vitro and required for optimal tumor growth in vivo (12). Recently, genome-wide mapping of the XBP1 transcriptional regulatory network revealed that XBP1s drives tumorigenesis of triple-negative breast cancer by assembling a transcriptional complex with HIF1␣ to cooperatively activate the expression of the HIF1␣ gene-expression program (14).
Many signaling pathways in cell development and growth control are engaged in hypoxia responses, including the evolutionarily conserved Wnt/␤-catenin pathway that has essential roles in embryonic development, tissue homeostasis, and tumorigenesis (15)(16)(17). In the absence of Wnt stimulation, cytoplasmic ␤-catenin protein forms a destruction complex with the scaffolding protein Axin, the tumor suppressor adenomatous polyposis coli gene product (APC), casein kinase 1 (CK1), and glycogen synthase kinase (GSK) 3␤. Upon phosphorylation by CK1 and GSK3␤, ␤-catenin is targeted by the E3 ubiquitin ligase ␤-Trcp for proteosomal degradation. Canonical Wnt/␤-catenin signaling is initiated by binding of Wnt proteins to the Frizzled family member receptors and subsequent complex formation with the low-density lipoprotein receptorrelated protein 5/6 (LRP5/6) co-receptors (18 -20). Stimulation by Wnt signals leads to disassembly of the destruction complex and thus inhibition of the ␤-catenin breakdown, allowing for its accumulation and localization in the nucleus. As a transcriptional co-activator, ␤-catenin interacts with the T-cell transcription factor (TCF)/lymphoid enhancer-binding factor family of DNA-binding proteins to activate the expression of Wnt target genes such as CyclinD1 and c-MYC proto-oncogenes. It has been well-documented that deregulated Wnt/␤catenin signaling is associated with cancer (21)(22)(23). Genetic alterations in the APC and CTNNB1 (␤-catenin) genes leading to abnormal accumulation of intracellular ␤-catenin occur very commonly in human colon cancer as well as other malignancies (24,25). Moreover, LRP6 expression has also been found to be frequently up-regulated in several types of cancer (26,27). Notably, canonical Wnt/␤-catenin signaling was reported to cross-talk with hypoxia-response pathways in tumor progression and metastasis, and direct interaction has been found between ␤-catenin and HIF1␣, implying potential competition for ␤-catenin between HIF1␣ and TCF-4 (28).
Apparently, complex interplays exist between the cell survival signaling network and multiple adaptive-response pathways in the face of hypoxia. It remains incompletely understood, however, how the Wnt/␤-catenin signaling and the UPR branches are integrated with the HIF1␣ pathway in context-dependent and/or cell type-selective manners to manage hypoxic stress and promote cell survival. In this study, we investigated whether Wnt/␤-catenin signaling interconnects with the UPR branch and HIF1␣-regulated hypoxia-response program in RKO colon cancer cells possessing normal Wnt/␤-catenin signaling. We found that hypoxic ER stress resulted in destabilization of ␤-catenin, largely because of reduced LRP6 production. Interestingly, we also found that ␤-catenin could negatively regulate XBP1s-mediated promotion of the HIF1␣-activated transcriptional program, suggesting that hypoxic suppression of ␤-catenin may facilitate a more efficient XBP1s-HIF1␣ cooperation for cell survival.

Hypoxia leads to activation of the UPR with simultaneously decreased ␤-catenin signaling
To examine whether hypoxia stress influences both the UPR and Wnt/␤-catenin-signaling pathways, we utilized the RKO colon cancer cell line without aberrant ␤-catenin activation. Consistent with the reported findings that UPR activation constitutes an important component of the hypoxia response (10,29,30), we observed that in parallel with a higher accumulation of HIF1␣ protein (Fig. 1A), phosphorylation of IRE1␣ (at Ser 724 within its kinase domain) as well as eIF2␣ significantly increased in RKO cells in response to hypoxia (1% O 2 ). Suggesting a hypoxic ER stress state, we also detected increased Xbp1 mRNA splicing along with elevated mRNA abundance of the XBP1 target genes, the ER chaperone BiP, and the ER-localized DnaJ 4 (ERdj4) (Fig. 1B). In contrast to the accumulation of HIF1␣ protein, we observed significantly decreased ␤-catenin protein levels (Fig. 1A), but no significant changes in its mRNA levels (Fig. 1C), following exposure to hypoxia. Consistently, gene expression profiling analyses revealed significantly downregulated mRNA levels of ␤-catenin target genes, including DKK1, c-MYC, AXIN2, and CyclinD1 (Fig. 1D), along with markedly up-regulated mRNA levels of typical HIF1␣ targets such as VEGFa, GLUT1, PDK1, DDIT4, ID1, JMJD1a, and MCT4 (Fig. 1E). Moreover, similar hypoxic UPR activation accompanied by decreased ␤-catenin signaling was also observed in HEK293 cells (Fig. S1). These results showed that hypoxic ER stress accompanied the down-regulation of ␤-catenin signaling, suggesting that hypoxia may link the Wnt/␤-catenin signaling pathway to the UPR-mediated adaptive response.

␤-Catenin hinders hypoxic UPR-mediated adaptive response ER stress results in decreased stability of ␤-catenin
To affirm that ER stress can affect ␤-catenin signaling, we treated RKO cells with two chemical ER stressors, tunicamycin (TM, which blocks the first step of N-linked glycosylation) and thapsigargin (TG, which inhibits the ER Ca 2ϩ -dependent ATPase to cause depletion of calcium from the ER lumen). Indeed, both TM-and TG-induced ER stress, as indicated by elevated eIF2␣ phosphorylation, resulted in gradual decreases in the Wnt3a-stimulated accumulation of ␤-catenin in RKO cells ( Fig. 2A). Likewise, similar decreases in Wnt3a-stimulated accumulation of ␤-catenin were also observed in TM-and TGtreated HEK293 cells (Fig. S2A). Consistently, such ER stressinduced reduction in ␤-catenin accumulation was associated with decreased luciferase activity from the TOPFlash reporter (31) that can be activated by the ␤-catenin/TCF complex (Fig.  S2B). Next, we tested whether the chemical chaperones, tauroursodeoxycholic acid (TUDCA) and 4-phenylbutyric acid (4-PBA) that are known for their ability to alleviate ER stress (32), could have an effect upon hypoxia-induced decrease of ␤-catenin protein in RKO cells. Both TUDCA and 4-PBA reduced hypoxia-induced eIF2␣ phosphorylation and exhibited a partial reversal effect upon the decrease of ␤-catenin protein following exposure to hypoxia (Fig. S3).
Given that ER stress had no effect upon the levels of ␤-catenin mRNA in RKO or HEK293 cells (Fig. S4A), we wondered whether the ER stress-induced decrease in ␤-catenin

␤-Catenin hinders hypoxic UPR-mediated adaptive response
protein was related to the ubiquitin proteasome system that is considered to be the major mechanism for controlling the stability of ␤-catenin via its phosphorylation by GSK3␤ and CK1␣. In the absence or presence of Wnt3a stimulation, treatment of RKO cells with the proteasome inhibitor MG132, but not the inhibitor of lysosomal endopeptidases chloroquine (CQ), resulted in dramatic increases in ␤-catenin protein levels and prominently diminished the suppression by TG-induced ER stress of ␤-catenin protein accumulation (Fig. 2B). Furthermore, treatment with LiCl, which inhibits GSK-3␤ activity, led to increased accumulation of ␤-catenin in a dose-dependent fashion even in the presence of TG-induced ER stress (Fig. 2C), despite that such effects were less pronounced in the presence of Wnt3a stimulation. Together, these data suggest that ER stress could promote the degradation of ␤-catenin, likely through impairing the Wnt-signaling pathway upstream of GSK-3␤.

ER stress and hypoxia cause destabilization of ␤-catenin via down-regulation of LRP6
To determine how experimental ER stress and hypoxia could promote ␤-catenin's destabilization via affecting the Wnt sig-

␤-Catenin hinders hypoxic UPR-mediated adaptive response
naling cascade, we examined the level of LRP6 protein, the single-span transmembrane co-receptor that is essential for canonical Wnt signaling (19,20). Notably, chemically-induced ER stress resulted in significant reduction of matured LRP6 protein in RKO or HEK293 cells treated with TM and TG (Fig.  3A) without alterations in its mRNA abundance (Fig. S4B), and an LRP6 protein band of lower molecular weight was detected in TM-treated cells, most likely representing its nonglycosylated form due to TM inhibition of N-glycosylation. Because LRP6 protein contains an extracellular domain with autoinhibitory activity, we utilized the N-terminally truncated LRP6⌬N, a constitutively activated receptor that lacks most of the extracellular domain and can bind to Axin in the absence of Wnt ligand stimulation (31), in order to test the effect of restored LRP6 signaling upon ER stress-promoted ␤-catenin degradation. Interestingly, overexpression of LRP6⌬N in HEK293 cells markedly increased ␤-catenin protein accumulation and prevented its degradation upon TM or TG treatment (Fig. 3B), whereas Wnt3a stimulation that did not affect LRP6 protein levels had no such effect. In addition, gradual decreases in LRP6 protein levels, but not its mRNA abundance (Fig. S4C), were also detected in RKO and HEK293 cells following exposure to hypoxia (Fig. 3C), and restored overexpression of LRP6⌬N or full-length LRP6 was able to blunt hypoxia-induced reduction in ␤-catenin accumulation in HEK293 cells (Fig. 3D).
These results indicate that hypoxia and ER stress resulted in impaired cellular production of LRP6, which in turn led to promotion of ␤-catenin destabilization and down-regulation of Wnt/␤-catenin signaling.

Activated Wnt/␤-catenin pathway blunts the HIF1␣ transcriptional program in RKO cells
To test whether hypoxic ER stress-induced down-regulation of ␤-catenin is implicated in the regulation of the hypoxia response, we examined the effects of Wnt/␤-catenin signaling activation upon the HIF1␣-responsive gene expression program. Wnt3a stimulation of RKO cells with WCM under normoxia dramatically increased ␤-catenin accumulation, whereas exposure to hypoxia significantly reduced it (Fig. 4A) without affecting its mRNA abundance (Fig. 4B). However, Wnt3a treatment had no effect upon HIF1␣ protein induction following exposure to hypoxia (Fig. 4A) but resulted in significant suppression of hypoxia-evoked up-regulation of HIF1␣ target genes (Fig. 4B). Next, we utilized cobalt chloride (CoCl 2 ) that can enhance the stability of HIF1␣ to mimic the inducing effect of hypoxia upon HIF1␣ without affecting the stability of ␤-catenin in RKO cells under normoxia (Fig. 4C). Exposure to CoCl 2 significantly enhanced the expression of HIF1␣ targets genes, whereas Wnt3a treatment blunted such induction by CoCl 2 (Fig. 4D). Furthermore, overexpression of ␤-catenin could also markedly diminish CoCl 2 -induced up-regulation of HIF1␣ target genes (Fig. 4, E and F). Therefore, activation of the Wnt/␤-catenin pathway could exert a negative regulatory action upon the HIF1␣-directed gene expression program, and ER stress-associated suppression of Wnt/␤-catenin signaling might facilitate the fine-tuning of the adaptive response to hypoxia.

Interactions between ␤-catenin, HIF1␣, and XBP1s
Given that hypoxia could induce the activation of the IRE1␣-XBP1 pathway (Fig. 1, A and B) and XBP1s has been documented to enhance the HIF1␣-activated gene expression program in hypoxia (14), we wondered if there exist mutual interactions between ␤-catenin, HIF1␣, and XBP1s that may constitute a dynamic regulatory network to govern the transcriptional program in response to hypoxia. Co-immunoprecipitation analyses showed that in HEK293 cells under hypoxia, HIF1␣ could associate with both ␤-catenin and XBP1s, and decreased HIF1␣-XBP1s interaction was detectable in the presence of overexpressed ␤-catenin (Fig. 5A). Moreover, XBP1s

␤-Catenin hinders hypoxic UPR-mediated adaptive response
was able to interact with both ␤-catenin and HIF1␣, and the presence of overexpressed HIF1␣ apparently reduced ␤-catenin-XBP1s association (Fig. 5B), whereas overexpression of XBP1s resulted in decreased ␤-catenin-HIF1␣ association under hypoxia (Fig. 5C). Curiously, neither ␤-catenin nor XBP1s overexpression influenced HIF1␣-HIF1␤ interaction in HEK293 cells under hypoxia (Fig. 5, A and C). Although the precise interactions and their regulatory modes, or possible competitions between these transcriptional regulators on the target chromatins are currently unclear, these data suggest that there might be a favorable HIF1␣-XBP1s interaction relative to their association with ␤-catenin under hypoxic stress.

Wnt/␤-catenin blunts XBP1s augmentation of HIF1␣activated transcription
We then asked whether Wnt/␤-catenin signaling was able to affect the transcriptional cooperation between XBP1s and HIF1␣. Remarkably, using HIF1␣-responsive 5xHRE luciferase reporter, we observed that overexpression of XBP1s not only markedly activated its transcription in HEK293 cells under normoxia, but also enhanced its induction in response to hypoxia (Fig. 6A). Interestingly, Wnt3a treatment had no effect upon XBP1s-activated transcription, but it prominently abolished XBP1s' augmentative effect in response to hypoxia (Fig. 6A). Overexpressed ␤-catenin exhibited similar blunting effects upon XBP1s-promoted HIF1␣ activation of the 5xHRE reporter (Fig. 6B). In addition, Wnt3a also suppressed XBP1s-mediated up-regulation of HIF1␣ target genes in CoCl 2 -treated RKO cells (Fig. 6C). These results indicate that Wnt/␤-catenin signaling could act to disrupt the cooperative activation by XBP1s-HIF1␣ of the hypoxic gene expression program.

␤-Catenin hinders hypoxic UPR-mediated adaptive response Wnt/␤-catenin hinders XBP1s-promoted cell survival under hypoxia
To determine the functional impact of Wnt/␤-cateninmediated suppression of the XBP1s-HIF1␣ transcriptional cooperation upon cell survival in the face of hypoxia stress, we first examined XBP1s' ability to affect cell survival under hypoxia. Overexpression of XBP1s significantly improved the survival of RKO cells following exposure to hypoxia (Fig. 7C), which might largely result from its alleviating effects upon apoptosis induced by such prolonged exposure to hypoxia (Fig. 7, A  and B). Notably, Wnt3a treatment significantly blunted the promoting effect of XBP1s upon the survival of RKO cells (Fig.  7D), and overexpression of ␤-catenin also diminished XBP1smediated improvement of survival in HEK293 cells (Fig. 7E). Thus, these data suggest that hypoxic ER stress suppression of Wnt/␤-catenin signaling may facilitate the efficiency of XBP1s-HIF1␣ cooperation in promoting the ability of cells to survive the hostile hypoxia environment.

Discussion
Hypoxia is a hallmark of many solid tumors, eliciting multiple cellular adaptive-response pathways to promote angiogenesis, metastasis, and therapy resistance (5). The hypoxic tumor microenvironment disturbs cellular proteostasis to trigger ER stress (33), which is thought to affect tumor progression. It remains to be better understood how tumor cells employ their hypoxic stress-response pathways in connection with developmental control signaling, such as the Wnt/␤-catenin cascade to survive the hostile microenvironment during tumor progression. As depicted by the schematic model (Fig. 7F), this study shows that hypoxic ER stress may exert dual actions in promoting the survival of RKO cells during deprivation of oxygen: activation of the IRE1␣-XBP1 branch of the UPR to elicit XBP1s-mediated enhancement of the HIF1␣-regulated gene expression program, and suppression of Wnt/␤-catenin signaling to alleviate its negative effect upon XBP1s-HIF1␣ cooperation in regulating the hypoxia stress response. Our results indicate that to gain adaptive advantage for cell survival, dynamic interplays are at work between these signaling regulators of the hypoxic stress response and developmental control mechanism.
Wnt/␤-catenin signaling is a key cascade in developmental regulation, which is tightly associated with many types of cancer (34,35). Aberrant Wnt/␤-catenin cascade in tumorigenesis has been most commonly documented in colorectal cancer. A previously reported study (28) showed that in SW480 and HCT116 colon cancer cells with a constitutively high level of ␤-catenin, hypoxia could inhibit ␤-catenin/TCF-4 complex formation and transcriptional activity, and ␤-catenin was found to bind directly to HIF1␣ and enhance its transcriptional activity, thereby promoting cell survival under hypoxia. In contrast, here we found that in RKO colon cancer cells with normal Wnt/ ␤-catenin signaling capacity, hypoxia could inhibit ␤-catenin accumulation by reducing the LRP6 protein level, leading to impairment of Wnt signaling transduction. Furthermore, Wnt stimulation or ␤-catenin overexpression exhibited an attenuating effect upon HIF1␣-regulated gene expression program in hypoxia. More notably, ␤-catenin was able to diminish the cooperative cross-talk between the HIF1␣ and XBP1s pathways that could be provoked by hypoxic ER stress. Thus, our results from RKO cells suggest that hypoxia may influence the Wnt/␤catenin cascade through distinct mechanisms with different functional consequences in different colon cancer subtypes. The net effect of hypoxia upon the interplays between ␤-catenin, HIF1␣, and XBP1s is most likely context-dependent and can be influenced by tumor cell origins or its subtypes. It is also noteworthy that hypoxia has been shown to block the processing and secretion of Wnt proteins through ER stress induction, leading to suppression of the Wnt/␤-catenin pathway and growth of RKO cells (36). Our results revealed hypoxic ER stress-associated ␤-catenin destabilization resulted from decreases in matured protein levels of the LRP6 co-receptor, mimicking its alterations under TG-induced experimental ER stress. It remains to be defined, however, whether this hypoxiainduced reduction of LRP6 involves suppression of global cellular protein translation, membrane protein processing and maturation within the ER, or ER-associated degradation processes. Another open question that has yet to be deciphered is the exact mechanism(s) implicated in regulating the modes of the mutual interactions between ␤-catenin, HIF1␣, and XBP1s, as well as their functional output(s) as the key nodes in reprogramming metabolism, protein synthesis, and cell cycle progression to fine-tune the adaptive hypoxia responses. A, HEK293 cells were transiently co-transfected with 5xHRE-luciferase reporter plasmid together with vector or XBP1s-expressing plasmid for 24 h. Cells were cultured in LCM or WCM and exposed to normoxia (N) or 1% O 2 hypoxia (H) for 12 h. Cell lysates were used for luciferase activity measurement. B, HEK293 cells were co-transfected with 5xHRE-luciferase reporter together with the indicated plasmids for 24 h, followed by measurement of luciferase activity. C, RKO cells were transfected with vector or XBP1s plasmid for 24 h and cultured in LCM or WCM in the absence or presence of 100 M CoCl 2 for 12 h. Quantitative RT-PCR analysis of the mRNA abundance for the indicated HIF1␣ target genes is shown. All data are shown as the mean Ϯ S.D. from three independent experiments after normalization to the values of vector control or LCM-cultured cells. *, p Ͻ 0.05; **, p Ͻ 0.01; and ***, p Ͻ 0.001 by Student's t test.

␤-Catenin hinders hypoxic UPR-mediated adaptive response
Sustained ER stress is thought to endow malignant cells with a higher capacity for tumorigenesis, metastasis, and drug resistance (6,37). Hypoxic activation of the UPR-signaling pathways may be exploited by tumor cells to cope with the hostile microenvironment, and such adaptation to hypoxia is a critical determinant of tumor progression. Increasing lines of evidence have indicated that the IRE1␣-XBP1 branch of the UPR is implicated in promoting the progression of various types of tumor, including triple-negative breast cancer (14,38), hepatocellular carcinoma (39), glioblastoma (40), ovarian cancer (41,42), prostate cancer (43), as well as colon cancer (44). Preclinical studies have also shown promising anti-tumor potential of targeted inhibition of the IRE1␣-XBP1 pathway (45), particularly when employed in combination with standard chemotherapy drugs (46). Although therapeutic agents targeting the IRE1␣-XBP1 and HIF1␣ pathways have been actively pursued (47,48), chemical modulators of the Wnt/␤-catenin cascade are also being developed as anti-tumor entities (34,35,49). Therefore, it is of great significance for precision targeting to dissect the molecular hallmarks of hypoxia and ER stress responses together with the activation state of Wnt/␤-catenin signaling in different tumors or subtypes. Given the presumably distinct inter-relationships between XBP1s, HIF1␣, and ␤-catenin in different tumors, it requires more in-depth characterization of the molecular signatures of the cross-talk between these signaling cascades in order to develop more effective combination anti-tumor therapies.
In summary, our findings revealed the occurrence of tumor cell subtype-selective or context-dependent cross-talks between the Wnt/␤-catenin signaling cascade and the UPR and HIF1␣ stress-response pathways, which is likely to allow tumor cells for more efficient adaptation to the unfavorable microenvironment of hypoxia. Further elucidation of the molecular dialogues in these signaling networks will advance our understanding of the molecular basis of malignant progression.

␤-Catenin hinders hypoxic UPR-mediated adaptive response
LRP6⌬N expression plasmids were kindly provided by Dr. Xi He from Harvard University. The ␤-catenin-FLAG expression plasmid was purchased from Addgene. The XBP1s and XBP1s-FLAG expression plasmids and adenoviruses were used as described previously (50). HIF1␣-HA and 5xHRE reporter plasmids were gifts from Dr. Mingliang Zhang at Shanghai Jiao-Tong University.

Cell culture and transfection
HEK293T cells, mouse L cells, Wnt3a-producing mouse L cells (stably transfected with the Wnt3a expression plasmid and maintained in the presence of G418, kindly provided by Dr. Xi He), and RKO cells were maintained in DMEM supplemented with 10% fetal bovine serum and 100 g/ml penicillin and 100 g/ml streptomycin. HEK293T cells were transfected in 12-well or 24-well plates by Lip2000 (Invitrogen) at 0.5 g/well (12-well plates) or 0.2 g/well (24-well plates). Cells were used for subsequent analysis after incubation for 24 h at 37°C with 5% CO 2 .

Hypoxia
For exposure to hypoxia, cells in fresh medium were maintained for varying periods of time in a hypoxia incubator chamber (STEMCELL Technologies Inc.) containing 1% O 2 , 5% CO 2 -balanced N 2 . For normoxia, cells were cultured in an incubator containing 21% O 2 , 5% CO 2 . For CoCl 2 treatment, cells were grown to 50 -80% confluency and cultured with 100 M CoCl 2 for 4 h.

Wnt stimulation
To analyze Wnt/␤-catenin signaling activation, cells were treated with Wnt3a-conditioned medium (Wnt3a CM) or control medium (LCM) that was prepared as described previously (31). Briefly, Wnt3a-producing mouse L cells or control L cells were cultured to full confluency in DMEM and then replaced with serum-free DMEM. The culture medium was collected after 3 and 7 days, and the two collections were mixed and filtered using 0.22-m filters. After a 2-fold dilution, medium was used for cell treatment after the concentration of Wnt3a was evaluated by its ability to stabilize ␤-catenin as compared with recombinant Wnt3a protein (R&D Systems, Minneapolis, MN), usually in the range of 150 -200 ng/ml.

Cell survival assay
RKO or HEK293 cells at 2.5 ϫ 10 5 cells per well (24-well plate) or 5 ϫ 10 5 cells per well (12-well plate) were exposed to hypoxia and counted after 1, 2, or 3 days. Cells were stained with crystal violet dye after fixation with methanol, followed by microscopic analysis.

TUNEL assay
RKO cells were transfected with the vector or XBP1s expression plasmid, or they were infected with Ad-EGFP or Ad-XBP1s adenovirus (50) for 24 h. Cells were then incubated under hypoxia conditions for 48 h. After fixation with 4% paraformaldehyde for 10 min, cells were subjected to TUNEL analysis according to the manufacturer's instructions (one-step TUNEL apoptosis assay kit, Beyotime). Apoptotic cells labeled with Cy3.0-dUTP were quantified from micrograph images or by flow cytometry.

Immunoblot analysis and co-immunoprecipitation
Cell lysates were prepared using RIPA buffer (150 mM NaCl, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS, 50 mM Tris-HCl, pH 7.4). Immunoprecipitation was conducted using the desired antibody by incubation overnight on the shaker to allow formation of protein/antibody complexes. Thirty l of protein G-agarose beads (GE Healthcare) were added for another 2 h, and the beads were collected by centrifugation at 2000 ϫ g for 1 min and washed four times with lysis buffer. For immunoblotting, SDS-PAGE was performed, and separated proteins from cell lysates or the beads were transferred onto a polyvinylidene difluoride membrane filter and incubated with the desired antibodies. The blots were then developed with Thermo Fisher Scientific SuperSignal West Pico Chemiluminescent substrate or Millipore Immobilon Western Chemiluminescent horseradish peroxidase substrate.

Luciferase reporter assay
HEK293T cells were transfected with TOPFlash or 5xHRE reporter plasmid and Renilla (10:1) for 24 h (51). Cell lysates were used for luciferase activity measurement with the Dual-Luciferase reporter system (Promega).

Statistical analysis
All data represent results from at least three independent experiments. Statistical significance was determined by Student's t test, and p Ͻ 0.05 was considered significant.