PanIN-Specific Regulation of Wnt Signaling by HIF2α during Early Pancreatic Tumorigenesis

Like most adult epithelial malignancies, pancreatic ductal adenocarcinoma (PDAC) arises from noninvasive precursor lesions following a prolonged period of local inflammation and stress (1). Among the precursor lesions, the most common and extensively studied is pancreatic intraepithelial neoplasia (PanIN). The exact sequence of molecular events leading to full blown invasive cancer is still not clear, although genetic analyses have disclosed that PanINs already exhibit abnormalities in many of the same genes and pathways altered in PDAC, including Kras, CyclinD1, Smad4, Notch, Hedgehog, and β-catenin/Wnt (2, 3). Although tight regulation of β-catenin/Wnt-signaling is required for the initiation of Kras-induced ductal reprogramming (4), our knowledge about the exact role β-catenin/Wnt-signaling may play during PanIN progression remains limited. Cells overcome hypoxic stress through multiple mechanisms, including the stabilization of the family of hypoxia inducible factor (HIF) transcriptional regulators. Two distinct transcription factors, HIF1α and HIF2α, regulate overlapping sets of target genes in hypoxic cells, although each protein also has unique functions (5–7). In the setting of normoxia, HIFs are ubiquinated through an oxygen-dependent interaction with von Hipple–Lindau (VHL) protein (8, 9). In humans, VHL is expressed in normal pancreatic ducts, but its expression is lost as early as in PanIN1As (10). HIF1α is absent in PanINs but is highly expressed in pancreatic cancer and has been recognized as an important resistance factor against chemotherapy and radiotherapy (11, 12). However, so far there have been no reports on the role of HIF2α in pancreatic cancer. Given the absence of VHL and HIF1α in PanINs (10, 11), we set out to study the potential role of HIF2α in pancreatic premalignant lesions. Specifically, we hoped to determine whether loss of Hif2α would influence the progression of Kras-induced pancreatic neoplasia. We generated the Ptf1aCre;Kras^G12D;Hif2α^f/f mice, and studied Kras-mediated mPanIN progression in the absence of HIF2α.

Together, our findings show that β-catenin/Wnt-signaling in the mPanINs is regulated by Smad4 expression and HIF2α plays a key role in orchestrating this process.

Human pancreatic adenocarcinoma specimens were collected from patients who underwent surgical procedures. All studies with human pancreatic tissues were approved by the University of Pittsburgh Institutional Review Board.

Mice used in these studies were maintained according to protocols approved by the University of Pittsburgh Institutional Animal Care and Use Committee. The Hif2α-floxed strain was generated by Guo-Hua Fong (Supplementary Results and Discussion); the Ptf1Cre (13), Kras^G12D (14), and Rosa26-HA-HiIF1dPA (15) strains were obtained from the Mutant Mouse Regional Resource Centers, National Cancer Institute Mouse Repository, and The Jackson laboratories, respectively.

Tissue processing and immunostaining were conducted as previously described (16). For SMAD4 staining, antigen retrieval was done in Tris-EDTA Buffer (10 mmol/L Tris Base, 1 mmol/L EDTA Solution, 0.05% Tween 20, pH 9.0) for 20 minutes at 95°C in steamer. For HIF2α, antigen retrieval was carried out in IHC-Tek Epitope Retrieval Solution for 20 minutes at 95°C in steamer. Images were acquired on a Zeiss Imager Z1 microscope with a Zeiss AxioCam driven by Zeiss AxioVision Rel.4.7 software.

Bromodeoxyuridine BrdUrd labeling was conducted by injecting BrdUrd (0.2 mg/g body weight; Sigma) intraperitoneally 2 hours before sacrifice.

Low-passage cancer cells were isolated from PdxCre;Kras^G12D;p53^+/− tumors in Dr. M. T. Lotze laboratory (University of Pittsburgh Cancer Institute, Pittsburgh, PA) as previously described (17). In brief, tumor fragments were digested with 2 mg/mL collagenase V at 37°C while agitating, then strained and washed to retrieve single cells. Cells were subsequently plated in T25 flasks and expanded. After 3 passages, they were tested for mesothelin expression. Tumor cells were frozen down and used for further expansion and testing as needed. Cells were cultured in RPMI-1640 medium (Gibco) supplemented with 10% FBS, 5% penicillin–streptomycin, and 5% glutamax. Cultures were maintained in 5% CO₂ at 37°C in a humidified incubator.

Endogenous Hif2α, β-catenin, and Smad4 in CPK cells were silenced using ON-TARGETplus SMARTpool siRNA (Dharmacon) according to the manufacturer's instructions.

Harvesting, tissues processing, and reactions were conducted as previously described (16). PCR primers were purchased from Qiagen (QuantiTect Primer Assays, Qiagen) and are listed in Supplementary Table S2.

Western blotting was done as described previously (18).

To study the role of HIF2α in pancreatic development and malignant transformation, we generated a Hif2α-floxed strain (Supplementary Fig. S1), and crossed the homozygous mice (Hif2α^f/f) with Ptf1aCre transgenic mice, which targets Cre recombinase expression to the epithelial lineages of the embryonic pancreas (13). Mice with homozygous deletion of Hif2α in the pancreas (Ptf1aCre;Hif2α^f/f) were born at the expected frequency, and showed normal pancreatic cytoarchitecture and differentiation throughout postnatal life (n = 15 mice, ages 1–12 months) as evident by exocrine as well as endocrine immunostaining analyses (Supplementary Fig. S2). More importantly, Ptf1aCre;Hif2α^f/f mice did not develop pancreatic neoplasms.

We first examined HIF2α expression in well-preserved surgical specimens from 11 patients with PDAC. In the region of histologically normal tissue, HIF2α could not be detected in the acinar or ductal compartments, but it was highly expressed in the endocrine islets (Fig. 1A). Immuohistochemical analyses of human pancreatic adenocarcinoma tissues showed nuclear and/or cytoplasmic HIF2α localization within the metaplastic ducts (inset in Fig. 1A) and in the early-stage mPanINs (Fig. 1B). The levels of HIF2α protein gradually declined as the lesions progressed to more advanced stages (Fig. 1C). Similarly, immunostaining analyses of Ptf1aCre;Kras^G12D pancreata (PK mice) showed expression of HIF2α in the islets (Fig. 1E) and a more prominent presence of HIF2α in less advanced lesions (Fig. 1F) compared with PDAC (Fig. 1G). As shown in Fig. 1D and H, adjacent cells within the same lesion in both human and mouse samples could display varying levels of HIF2α protein. Interestingly, although cells with low or no HIF2α expression could be occasionally found in PanIN1s, the presence of HIF2-negative or HIF2^Low cells was more noticeable in the more advanced PanINs. These findings show that the dynamic expression of HIF2α observed in human PanIN and PDAC is associated with PanIN progression and is recapitulated in a mouse model of PDAC.

To study the loss of Hif2α in the setting of oncogenic Kras expression, Hif2α^f/f mice were crossed with PK mice to generate the Ptf1aCre;Kras^G12D;Hif2α^f/f (PKH2 mice) strain. The absence of HIF2α in the early PanIN lesions in PKH2 mice confirmed the pancreatic inactivation of Hif2α (Supplementary Fig. S1E). In the PK pancreas, only a few PanIN1As could be detected at 1 month of age (Fig. 2A and B), whereas at 3 months early morphologic changes including ADM and PanIN1A lesions could be found (Fig. 2C). By contrast, at 1 month, PKH2 mice showed acceleration of PanIN progression, with higher incidence of both metaplastic ducts and mPanIN1A lesions, although still in the context of intact lobule architecture (Fig. 2A and B).

Between 1 and 3 months, there was evidence of further progression in PKH2 mice, with severe periductal stromal response, secondary loss of the lobular acinar parenchyma, as well as increased duct ectasia and PanIN1 lesions. At 3 months of age, PKH2 pancreata already displayed PanIN1B lesions, with a few of them transitioning to PanIN2 (Fig. 2C). Compared with their age-matched PK cohort, 3-month-old PKH2 mice also showed robust desmoplastic response, diffuse replacement of the acini by lobules with mucinous epithelium (PanIN1), and luminal dilation (with significant increase in the acinar-to-ductal ratio; Fig. 2C). From 3 months onward, the PK mice displayed classic stepwise mPanIN formation (Fig. 2D–F), and transition to PDAC (Fig. 2G). Interestingly, the 9-month-old PKH2 pancreas contained a similar percentage of PanIN1s as the PK mice (Fig. 2A and D). The PK mice, however, contained more PanIN1A lesions, whereas the PKH2 mice contained more PanIN1B lesions (Fig. 2A). Nevertheless, despite the higher PanIN incidence in younger mice, there was a decrease in the number of PanIN2s in older PKH2 mice, and more advanced PanIN3s were rarely detected (Fig. 2A and D–F). The changes within the lobules showed increased cytoplasm with progressive increase in dilation and papillary projections into the dilated lumina (Fig. 2G).

These findings suggest that HIF2α is dispensable for initiation of mPanINs, but it is required for mPanIN progression.

Notch and Wnt-signaling pathways are reactivated during PanIN formation. Expression of the Notch target gene Hes1 was significantly higher in the PKH2 pancreas compared with the age-matched PK cohort (Fig. 3A). In addition, some targets of Wnt/β-catenin pathway, such as Lef1 (Fig. 3B), CyclinD1 (Fig. 3C), and Axin-2 (Fig. 3D) were found to be expressed at lower levels in the PKH2 pancreas. In contrast, the expression of another Wnt target, cMyc, was substantially higher in the PKH2 pancreas (Fig. 3E).

PKH2 pancreata have a higher incidence of ADM and PanIN lesions at earlier time points (Fig. 2A). We thus quantified the proliferation rate in the lesions of both PK and PKH2 mice. Interestingly, despite an overall 4-fold increased BrdUrd⁺ cell number in the PKH2 pancreatic epithelium (Fig. 3F), the increased proliferative rate was more prominent among the metaplastic ducts, whereas cells within the PanINs were mostly BrdUrd⁻ (Fig. 3G and H). In addition, compared with age-matched PK, the PKH2 PanINs contained fewer CyclinD2-expressing cells (Fig. 3I and J).

Several studies have highlighted an important role for Wnt/β-catenin signaling in PDAC (4, 19–21). Thus, the observed downregulation of Wnt pathway in PKH2 mice, encouraged us to study β-catenin protein expression in PK and PKH2 pancreas (Fig. 4). As expected, immunostaining in PK mice revealed that β-catenin was present in all epithelial cells, including metaplastic ducts and PanINs (Fig. 4A and D). In contrast, in PKH2 pancreata β-catenin was absent in the vast majority of PanINs, although it could be detected in the endocrine and acinar cells, as well as in the normal and metaplastic ducts (Fig. 4B and E). Expression of E-cadherin confirmed the epithelial characteristics of the β-catenin-deficient PKH2 PanIN cells (Fig. 4E). The deficiency of β-catenin was related neither to the age of the mice, nor to the PanIN stage (Supplementary Fig. S3). The absence of CyclinD1 in PKH2 PanINs, further confirmed the impairment of β-catenin/Wnt-signaling in these animals (Fig. 4G and H).

Deletion of Hif1α often leads to enhanced Hif2α expression, and similarly Hif2α deletion results in higher levels of Hif1α expression (22). We thus speculated that the phenotype observed in PKH2 mice could be due to compensatory expression of Hif1α rather than deletion of Hif2α. To investigate this hypothesis, pancreatic Hif1α gene expression was analyzed in samples obtained from 3- to 12-month-old PK and PKH2 mice (Fig. 4I). We found that Hif1α transcript levels were undetectable in 3-month-old PK pancreas, but gradually increased in the older animals, confirming what previous studies had reported in human PanINs (11). In contrast, the deletion of Hif2α in the PKH2 pancreas leads to an accelerated and higher Hif1α expression (Fig. 4I). To further determine whether the absence of β-catenin in Hif2α-deficient PanINs was primarily the result of HIF1α function, we bred the PK mice into the Rosa26-HA-HiF1-dPA strain (15). Our analyses of 3-month-old Ptf1aCre;Kras^G12D;R26^Hif1α pancreas showed normal expression of β-catenin in PanINs (Fig. 4C and F). These data indicate that the absence of β-catenin in the PKH2 PanINs is likely to be independent of the ectopic presence of HIF1α in the Hif2α-deficient pancreas.

Our transgenic studies strongly suggest that the presence of β-catenin in PanINs is correlated with HIF2α activity. To determine whether loss of Hif2α affects β-catenin at a transcriptional and/or translational level, CKP cells were treated with Hif2α siRNA (siHif2α) and harvested after 48 or 72 hours (Fig. 5). The CPK cell line is a mouse pancreatic cancer cell line isolated from PdxCre;Kras^G12D;p53^+/− tumors. As shown in Fig. 4A, HIF2α and β-catenin are both expressed in this cell line at basal level. After siHif2α treatment, we detected a 60% reduction in Hif2α mRNA at 48 hours, which was further reduced to more than 95% at 72 hours. Consistently, the expression of Hin1, which is a direct HIF2α target gene (23), was also reduced 50% at 48 hours and 75% at 72 hours (Fig. 5A). Interestingly, β-catenin levels and downstream Wnt-target genes CyclinD1 and cMyc were not appreciably affected after 48 hours, but showed significant reduction only after 72 hours (Fig. 5A). Western blot analysis also confirmed reduction of β-catenin protein levels 72 hours after siHIF2α treatment (Fig. 5B, and Supplementary Fig. S4A and S4B). Overall, our results show that HIF2α regulates the transcription of β-catenin in pancreatic lesions.

Given the oncogenic nature of Wnt signaling, we looked for other mechanisms that could account for the accelerated and higher PanIN incidence observed in PKH2 mice. Active Notch pathway promotes epithelial transformation and PanIN progression in human and mouse (24, 25). As shown in Fig. 3, Hes1 expression was significantly higher in the PKH2 pancreas. However, we could not detect any changes on mRNA level for Notch targets such as Hes1, Hey1, and Hey2 following HIf2α silencing in CKP cells (Fig. 5C).

Loss of Smad4 is a relatively late event during PanIN development (26), and inactivation of Smad4 in the setting of Kras-driven neoplasia is associated with acceleration in pancreatic tumorigenesis (27, 28). Our immunohistochemical analyses showed that SMAD4 was present in the normal- as well as in the metaplastic ducts in both the PK (Fig. 6A and Supplementary Fig. S5A) and the PKH2 pancreas (Fig. 6B and C and Supplementary Fig. S5B–S5D). As expected, early PanINs in PK mice expressed Smad4 (Fig. 5A and Supplementary Fig. S5A), however, the PanINs in the hif2α-mutant pancreas lacked Smad4 expression (Fig. 6B and C and Supplementary Fig. S5B–S5D). The observed loss of SMAD4 in PKH2 pancreas occurred concomitantly with transformation of metaplastic ducts to PanIN1A (Fig. 6B and C). Similarly to β-catenin, the loss of Smad4 in the Hif2α-deficient lesions was not age-dependent (Supplementary Fig. S5B–S5D).

To verify whether the expression of Smad4 is dependent on HIF2α, we analyzed Smad4 expression in siHif2α-treated and control CPK cells (Fig. 6D). No significant changes could be detected in Smad4 expression at 48 hours. However, at 72 hours both Smad4 mRNA and protein levels had reduced significantly compared with controls, confirming that HIF2α also regulates Smad4 transcription (Fig. 6D and E and Supplementary Fig. S4C). Finally, the presence of SMAD4 in Ptf1aCre;Kras^G12D;R26^Hif1α PanINs confirmed that the loss of SMAD4 in PKH2 PanINs is HIF2α-dependent (Fig. 6F).

We have shown that inactivation of Hif2α (95% reduction in siHif2α-treated cells or complete inactivation in the PKH2 pancreas) leads to significant reduction of SMAD4 and β-catenin levels. These data suggest that the observed decrease in HIF2α levels in more advanced lesions (Fig. 1C and F) would result in reduced Smad4 and β-catenin expression. While this finding is consistent with Smad4 expression during PanIN progression (26), it is in contrast to what has been proposed for the β-catenin activity (4). However, this discrepancy could reflect a dose-dependent transcriptional activity of HIF2α. To determine whether HIF2α regulates Smad4 and β-catenin expression in a dose-dependent manner, we silenced Hif2α under suboptimal conditions. A 60% reduction in Hif2α mRNA levels 72 hours after siHif2α treatment resulted in decreased SMAD4, but had no effect on β-catenin expression (Fig. 7A, and Supplementary Fig. S4D).

Next, we studied whether the reductions in Smad4 and β-catenin mRNA are 2 independent events, or if the 2 genes are synchronous hierarchically downstream of HIF2α. To further investigate the relation between SMAD4 and β-catenin, we silenced the expression of Smad4 (siSmad4) and β-catenin (siβ-cat) separately in CKP cells, and determined SMAD4 and β-catenin mRNA and protein expression in comparison with nontarget siRNA–treated cells (Fig. 7B and C, and Supplementary Fig. S4E). Interestingly, we found that expression of β-catenin was significantly increased in Smad4-silenced CKP cells both at the mRNA and protein levels (Fig. 7B and C). Consistently, the expression of Wnt targets such as Axin2 and CyclinD1, were also higher in siSmad4-treated CKP cells (Fig. 7B). On the other hand, silencing of β-catenin did not have any effect on Smad4 expression at mRNA (data not shown) and protein level (Fig. 7C).

Overall, our findings indicate that in pancreatic cancer cells HIF2α promotes β-catenin and SMAD4 expression independently. In addition, HIF2α regulates Smad4 expression in a dose-dependent manner, and the expression of β-catenin is negatively regulated by SMAD4.

The loss of VHL and the subsequent activation of the hypoxic pathway is one of the earliest events in Kras-induced pancreatic neoplasia, and it is evident as early as in PanIN1As (10). Here, we evaluated the role of HIF2α in pancreatic carcinogenesis in vivo, and propose a molecular mechanism, which could explain the phenotype observed. Histologic analysis of the PKH2 pancreata showed that loss of Hif2α significantly accelerated the onset of acinar-to-ductal metaplasia and the appearance of PanIN lesions. However, despite the increased acinar-to-ductal metaplasia and PanIN incidence early on, the subsequent PanIN progression was halted, with a predominance of low-grade PanINs in the older PKH2 mice. The phenotype in PKH2 pancreas was associated with a higher Notch activity, as evident by increased Hes1 expression. Nevertheless, Hif2α silencing in CKP cells did not have any effect of Notch targets, thus it is likely that the in vivo difference could reflect higher number of PanINs in the PKH2 pancreas. The PKH2 pancreas also displayed overall lower transcript levels of Wnt-target genes, such as Axin2, lef1, and CyclinD1, but higher levels of cMyc. However, increased expression levels of cMyc is a common emergent phenotype in most neoplastic cells and in addition, cMyc is expressed by the acinar cells (29).

Recent studies imply that Wnt-signaling must be tuned to appropriate levels at key time points during transformation to specify the PanIN-PDAC lineage (3, 4). Here, we report loss of β-catenin expression and impaired Wnt signaling in PKH2 PanINs. Of note, β-catenin was not uniformly absent in all Hif2α mutant PanIN cells. Indeed, in some lesions only a subpopulation of cells was β-catenin negative, whereas in few other lesions all cells were β-catenin positive. This finding is consistent with the absence of HIF2α in a subpopulation of PanIN cells. Similar to β-catenin, in the PKH2 mice, SMAD4 could be detected in the metaplastic ducts; however, its expression was lost in early PanINs. Together, these findings suggest that the accelerated and higher PanIN incidence seen in the PKH2 pancreas may be due to the absence of SMAD4. At the same time, the impaired PanIN progression could be the result of β-catenin loss.

Our in vitro studies revealed that, like β-catenin, Smad4 was also regulated by HIF2α at the transcriptional level. However, the lack of HIF-binding motifs in the regulatory elements of β-catenin or Smad4 would suggest that HIF2α likely does not directly regulate β-catenin or Smad4 gene expression. Furthermore, while silencing of Hif2α in CKP cells resulted in decreased transcript levels of Hin1 (a direct HIF2α target) within 48 hours, any significant decrease in β-catenin or Smad4 gene expression was detected only after 72 hours.

In addition to directly binding to DNA and promote transcription of target genes, HIFs may also regulate transcription through heterodimer-mediated protein–protein interaction with other transcription factors (30). For example, HIF2α accentuate MAX/MYC or β-catenin/TCF-driven transcription (31, 32). Thus, we speculate that HIF2α may be part of DNA-binding protein complexes that regulate β-catenin and Smad4 expression. More detailed analyses are needed to better understand the mechanisms by which β-catenin and Smad4 expressions are regulated by HIF2α.

To determine whether loss of β-catenin was the result of HIF1α expression, we generated a mouse in which HIF1α was expressed in conjunction with oncogenic Kras in the pancreas. The presence of β-catenin and SMAD4 in the PanIN lesions in the Ptf1aCre;Kras^G12D;R26^Hif1α pancreas confirmed that the expression profile observed in PKH2 PanINs is the result of HIF2α loss. Smad4 deficiency in the setting of oncogenic Kras expression is associated with increased proliferation of the neoplastic epithelium (27, 28). Consistently, PKH2 mice displayed a higher proliferative rate among metaplastic ducts compared with their PK littermates. However, the PanINs in PKH2 mice incorporated significantly less BrdUrd, likely due to the impaired Wnt signaling and the absence of CyclinD1.

On the basis of the collective results of our in vivo and in vitro experiments, we propose that HIF2α modulates Wnt signaling during mPanIN progression, by maintaining appropriate levels of Smad4 and β-catenin (Fig. 7D). Transformation of metaplastic ducts into PanIN1A lesions is associated with downregulation of VHL, which leads to accumulation of high levels of HIF2α in the early PanINs. The abundance of HIF2α at this stage promotes expression of both Smad4 and β-catenin in PanINs. SMAD4 prevents overexpression of β-catenin, thus maintaining β-catenin levels at the necessary threshold required for PanIN progression (4). This finding is consistent with recent reports showing that β-catenin expression is negatively regulated by SMAD4 (33, 34). As PanINs progress, the levels of HIF2α decline. Lower levels of HIF2α maintain β-catenin expression, but may not be sufficient to promote Smad4 expression. In the absence of SMAD4, β-catenin/Wnt-signaling levels reach the necessary threshold, required for PanIN-PDA transformation. It stands to reason that HIF2α may interact with 2 separate transcriptional factors, both competing for binding HIF2α to drive β-cateinin or Smad4 transcription, respectively. It is tempting to speculate that the complex promoting β-catenin may have higher affinity to bind HIF2α, and as the result being less sensitive to HIF2α levels present in the PanIN cell. This model, while intriguing, requires validation from more precise biophysical studies. Future studies should elucidate the mechanisms that distinguish the tumor promoting, or tumor suppressive activities of HIF2α protein during pancreatic tumorigenesis.

No potential conflicts of interest were disclosed.

Conception and design: A. Criscimanna, M. Lotze, G.K. Gittes, G.H. Fong, F. Esni

Development of methodology: A. Criscimanna, D. Duan, M. Lotze, G.H. Fong, F. Esni

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Criscimanna, D. Duan, V. Fendrich, E.D. Wickline, D.J. Hartman, S.P.S. Monga, G.H. Fong

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Criscimanna, D. Duan, V. Fendrich, D.J. Hartman, S.P.S. Monga, G.H. Fong

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J.A. Rhodes, G.H. Fong

Writing, review, and/or revision of the manuscript: A. Criscimanna, D.J. Hartman, S.P.S. Monga, M. Lotze, G.H. Fong, F. Esni

Study supervision: G.H. Fong, F. Esni

The authors thank Drs. S.D. Leach (Johns Hopkins University, Baltimore, MD) and S.F. Konieczny (Purdue University, W. Lafayette, IN) for their insightful comments.

This work was supported by the Concern Foundation (F. Esni), The Cochrane–Weber endowed Fund in Diabetes Research (F. Esni and A. Criscimanna), The Children's Hospital of Pittsburgh of UPMC (F. Esni), and NIH 5R01EY019721 (G.H. Fong).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.