Hypoxia-Inducible Factor-1-Dependent Repression of E-cadherin in von Hippel-Lindau Tumor Suppressor–Null Renal Cell Carcinoma Mediated by TCF3, ZFHX1A, and ZFHX1B

Epithelial cell-cell adhesion in humans and other mammals is mediated by intercellular junctional complexes consisting of tight junctions, adherens junctions, and desmosomes. E-cadherin, which is the principal component of adherens junctions and desmosomes in epithelial cells, mediates adhesion by homophilic interactions between cells (1). A defining step in the pathogenesis of carcinomas is the epithelial-mesenchymal transition, during which E-cadherin–mediated cell-cell adhesion is lost, and cells acquire invasive and metastatic properties (2, 3). In addition to its direct effects on cell-cell adhesion, E-cadherin loss of function in cancer cells also activates signal transduction pathways that promote proliferation, invasion, and metastasis (1).

E-cadherin loss of function occurs via a variety of molecular mechanisms in different human cancers (4), including mutations in the CDH1 gene that encodes E-cadherin (5, 6), increased E-cadherin degradation (7), and decreased CDH1 mRNA expression associated with CDH1 hypermethylation indicating transcriptional silencing (8, 9). The repressors SLUG, SNAIL, TCF3 (also known as E12/E47), ZFHX1A (also known as δEF1 or ZEB1), and ZFHX1B (also known as SIP1 or ZEB2) have been shown to bind to the proximal promoter of CDH1, repress its transcription, and induce the epithelial-mesenchymal transition (10–16). Increased expression of one or more of these repressors has been correlated with decreased CDH1 mRNA expression in human cancers, but the molecular basis for these observations has not been determined.

The conventional clear cell type of renal cell carcinoma (RCC) is associated with loss of function of the von Hippel-Lindau tumor suppressor (VHL) protein as a result of VHL gene mutation or hypermethylation (17). A principal function of VHL is to negatively regulate hypoxia-inducible factor-1 (HIF-1), a transcriptional activator of genes that play key roles in angiogenesis, cell proliferation/survival, energy metabolism, invasion, and resistance to radiation and chemotherapy (18–23). The constitutively expressed HIF-1β subunit forms functional heterodimers with either HIF-1α or HIF-2α (24, 25). VHL binds to HIF-1α and HIF-2α and targets them for ubiquitination and proteasomal degradation (26). VHL binding is dependent upon the O₂-dependent hydroxylation of specific proline residues in HIF-1α and HIF-2α (27, 28).

VHL loss of function leading to HIF-1 gain of function is the earliest detectable molecular event in the pathogenesis of RCC (29). However, the molecular basis for the epithelial-mesenchymal transition in RCC has not been determined, and the identification of proteins regulated by VHL in addition to HIF-1 has raised the possibility that key steps in RCC pathogenesis are independent of HIF-1 (17). In this study, we investigated whether dysregulated HIF-1 activity contributes to down-regulation of E-cadherin expression, loss of cell-cell adhesion, and the epithelial-mesenchymal transition in RCC.

Construction of enhanced green fluorescent protein and HIF-1αDN retroviral shuttle vectors. Enhanced green fluorescent protein (EGFP) sequences were amplified using pEGFP-N (Clontech, Palo Alto, CA) as template and inserted into retroviral vector pQCXIH (Clontech). The open reading frame from pCEP-4-HIF-1αDN (30) was amplified by PCR and inserted into retroviral vector pQCXIN (Clontech).

Retrovirus production and generation of stable cell lines. 293-T cells were seeded at 7 × 10⁶ per 10-cm dish. Retroviral vector DNA (12 μg) encoding EGFP or HIF-1αDN was cotransfected with plasmids encoding gag/pol (6 μg) and VSVG (1.5 μg) using LipofectAMINE 2000. Viral supernatant was collected 48 hours after transfection and added to RCC4 cells in the presence of polybrene (8 μg/mL; Sigma, St. Louis, MO). The procedure was repeated at 60 and 72 hours after transfection. Hygromycin (400 μg/mL) or G418 (800 μg/mL) was added to the culture media, starting at 96 hours after transfection of pQCXIH-EGFP or pQCXIN-HIF-1αDN, respectively. Pooled clones of resistant cells were cultured in DMEM (Mediatech, Herndon, VA) with 10% fetal bovine serum (FBS; Gemini, Calabasas, CA) and 1% penicillin/streptomycin starting 24 hours before all studies.

RNA isolation and quantitative real-time reverse-transcription PCR. Total RNA was isolated from cells using QIAshredder and RNeasy Mini Kits (Qiagen, Chatsworth, CA). cDNA was prepared using the iScript cDNA Synthesis kit (Bio-Rad, Richmond, CA). cDNA samples were diluted 1:15, and real-time PCR was performed using IQ SYBR Green Supermix and specific primers in the iCycler Real-time PCR detection system (Bio-Rad). Primers were designed using Beacon Designer 2.1 software (sequences available on request). Annealing temperature was optimized for each primer pair. The expression of each target mRNA relative to 18S rRNA (R) was calculated based on the threshold cycle (C_t) as R = 2^−Δ(ΔCt), where ΔC_t = C_{t, target} − C_{t, 18s} and Δ(ΔC_t) = ΔC_{t, RCC4} − ΔC_{t, RCC4-VHL} or ΔC_{t, RCC4-GFP} − ΔC_{t, RCC4-DN} or ΔC_{t, GFP-EV} − ΔC_{t, GFP-shr2265}.

Immunoblot assays. Cells were lysed in radioimmunoprecipitation assay buffer supplemented with protease inhibitors; 60-μg aliquots of protein were resolved by 7% PAGE and transferred to a nitrocellulose membrane. A monoclonal antibody (mAb) directed against the extracellular domain (clone HECD1, Zymed Laboratories, South San Francisco, CA; 1:3,000 dilution) or intracellular domain (clone 36, BD Biosciences, San Jose, CA; 1:3,000 dilution) of E-cadherin was used. Immunoblot assays of HIF-1α (mAb H1α67) and HIF-2α (rabbit polyclonal antibody, NB-100-122, Novus Biologicals, Littleton, CO) were performed using 100-μg aliquots of lysate. Horseradish peroxidase (HRP)–conjugated secondary antibody against mouse (goat anti-mouse, Santa Cruz Biotechnology, Santa Cruz, CA) or rabbit (donkey anti-rabbit, Amersham, Arlington Heights, IL) immunoglobulin was used for detection. β-Actin was detected using a goat polyclonal antibody against β-actin and HRP-conjugated donkey anti-goat secondary antibody (Santa Cruz Biotechnology). Immunoblots were developed using enhanced chemiluminescence (Amersham Biosciences, Piscataway, NJ).

Transepithelial resistance assays. In a modified Boyden chamber (BD Biosciences), 1 × 10⁴ cells resuspended in 2.5 mL of media were added to the upper chamber, and 3.5 mL of media were added to the lower chamber, in triplicate. Resistance was measured using an epithelial voltohmmeter (EVOM, World Precision Instruments, Sarasota, FL). A blank with no cells was maintained, and the average resistance from the blank was subtracted from the sample. Resistance (R) per unit area was calculated as (R_sample − R_blank) × πd²/4, where d = diameter (in cm) of the chamber.

Immunocytochemistry. Cells (5 × 10³ per well) were seeded into an eight-well chamber slide precoated with poly-d-lysine. Cells were fixed with 4% paraformaldehyde for 10 minutes and washed with 0.5% bovine serum albumin (BSA) in PBS. Blocking was carried out at room temperature for 30 minutes using normal goat serum. Upon permeabilization with 0.1% Triton X-100 in 0.5% BSA/PBS, cells were incubated with primary antibody against E-cadherin (clone 36, BD Biosciences) overnight. Cells were incubated with phycoerythrin-conjugated secondary antibody (BD Biosciences) at 1:500 dilution for 1 hour at room temperature. Before mounting the slide, 4′,6-diamidino-2-phenylindole (Molecular Probes, Eugene, OR) was added to 300 nmol/L. Phycoerythrin-conjugated IgG2A (R&D Systems, Inc., Minneapolis, MN) was used as an isotype-matched control. The sections were photographed under fluorescent microscopy (Olympus BX51 mounted with Olympus DP70 camera).

Retroviral expression of short hairpin RNA in RCC4 cells. Short hairpin RNA (shRNA) specific for HIF-1α was identified using the manufacturer's design tool (OligoEngine, Seattle, WA). Oligonucleotides meeting the standards was annealed and cloned into BglII- and HindIII-digested pSUPER-Retro-neo-GFP (OligoEngine). In this vector, shRNA and GFP mRNA expression are driven by H1 and phosphoglycerate kinase promoter sequences, respectively. Retrovirus was generated by cotransfecting 293T cells with the shRNA vector and a second vector encoding gag/pol and VSVG using LipofectAMINE 2000 (Invitrogen, San Diego, CA). The supernatant was collected 48 hours after transfection, centrifuged, and directly used for infection of 1 × 10⁵ RCC4 cells per well in a six-well plate, using 2 mL of viral supernatant and polybrene at a final concentration of 8 μg/mL. Three rounds of retroviral infection were performed at 24-hour intervals. A pool of cells was selected by treatment with G418 at a concentration of 800 μg/mL. Cells were maintained in DMEM containing 10% FBS, 1% penicillin/streptomycin, and 600 μg/mL G418.

Immunohistochemistry. Formalin-fixed, 6-μm-thick sections were stained for HIF-1α using Catalyzed Signal Amplification (DAKO, Carpinteria, CA) and mAb H1α67 (1 mg/mL, Novus Biologicals; 1:10,000 dilution) as previously described (31). Nuclei were counterstained with hematoxylin. As a negative control, nonimmune serum was used in lieu of the primary antibody. Immunohistochemistry for E-cadherin was performed using mouse anti-human E-cadherin clone 4A2C7 (Invitrogen). Epitope retrieval was performed in 10 mmol/L citrate buffer (pH 6) for 20 minutes in a 1,200-W microwave oven at 90% power. Antibody incubation and detection were carried out on a NEXes instrument (Ventana Medical Systems, Tucson, AZ). Endogenous peroxidase activity was blocked with hydrogen peroxide. E-cadherin antibody was diluted 1:100 in PBS, applied, and incubated overnight at room temperature. Primary antibody was detected using biotinylated goat anti-mouse antibody followed by streptavidin-HRP conjugate and visualized with 3,3-diaminobenzidene enhanced with copper sulfate. Slides were washed in distilled water, counterstained with hematoxylin, dehydrated, and mounted.

Statistical analysis. Student's t test and ANOVA were performed using Excel (Microsoft, Redmond, WA).

Expression of E-cadherin in RCC4 cells is positively regulated by VHL. To investigate the effect of VHL loss of function on E-cadherin expression in clear cell renal carcinoma, we performed immunoblot assays on VHL-null RCC4 cells and a subclone transfected with an expression vector encoding VHL (RCC4-VHL). RCC4 cells expressed high levels of HIF-1α and HIF-2α protein under nonhypoxic conditions, as previously described (32), showing loss of physiologic regulation in the absence of VHL, which was corrected by restoration of VHL expression (Fig. 1A). In contrast, levels of the HIF-1β subunit and β-actin were unaffected by the presence or absence of VHL. E-cadherin protein was not detected in lysates of VHL-null RCC4 cells using antibodies directed against either the extracellular domain, which is the site of homophilic interactions, or the intracellular domain, which interacts with catenins that link E-cadherin to the cytoskeleton. Upon restoration of VHL expression, increased levels of E-cadherin protein were detected with both antibodies (Fig. 1A). Immunocytochemistry confirmed the low levels of E-cadherin protein in RCC4 cells and the markedly increased E-cadherin levels in RCC4-VHL cells (Fig. 2).

To determine whether the dysregulation of E-cadherin protein reflected changes in steady-state mRNA levels, quantitative real-time reverse transcription-PCR was performed. A 27-fold decrease in E-cadherin mRNA expression in RCC4 cells relative to RCC4-VHL cells was shown (Fig. 1B). In contrast, vascular endothelial growth factor (VEGF) mRNA expression was increased 5.5-fold in RCC4 cells compared with RCC4-VHL cells, which is consistent with previous reports (32). Thus, VEGF expression is induced, whereas E-cadherin expression is repressed in RCC4 cells with VHL loss of function.

Expression of E-cadherin in RCC4 cells is negatively regulated by HIF-1. To determine whether the dysregulated expression of HIF-1α and HIF-2α in RCC4 cells contributed to the loss of E-cadherin expression, the RCC4-DN subclone was established by stable transfection of RCC4 cells with a retroviral expression vector encoding a dominant-negative form of HIF-1α that competes with endogenous HIF-1α and HIF-2α for binding to HIF-1β, resulting in the formation of HIF-1αDN:HIF-1β heterodimers that cannot bind to DNA or activate transcription (30). In addition, the dominant-negative form of HIF-1α unexpectedly inhibited HIF-2α protein expression (Fig. 1C). RCC4-DN cells showed increased expression of E-cadherin protein relative to RCC4-GFP cells, which were stably transfected with a retroviral expression vector encoding GFP.

Loss of HIF-1 activity was associated with 4.2-fold decreased VEGF mRNA expression in RCC4-DN cells compared with RCC-GFP cells (Fig. 1D). RCC4-DN cells showed 9.8-fold increased expression of E-cadherin mRNA relative to RCC4-GFP cells (Fig. 1D). Increased E-cadherin protein expression in RCC4-DN cells relative to RCC-GFP cells was also shown by immunocytochemistry (Fig. 2). Thus, in RCC4 cells, the loss of E-cadherin expression that is associated with VHL loss of function is mediated by HIF-1.

Evidence of epithelial-mesenchymal transformation of VHL-null RCC4 cells. To investigate the functional consequences associated with the presence or absence of E-cadherin expression in RCC4 cells, we analyzed transepithelial resistance as a measure of cell-cell adhesion in RCC4 and RCC4-VHL cells over 6 days in culture. VHL loss of function was associated with a highly significant reduction in transepithelial resistance at all time points (Fig. 3A). Comparison of RCC4-GFP and RCC4-DN cells revealed that expression of the dominant-negative form of HIF-1α, which was associated with a derepression of E-cadherin expression, also resulted in an increase in transepithelial resistance at all time points relative to RCC4-GFP cells (Fig. 3C). The rate of RCC4 and RCC4-GFP cell proliferation was increased compared with RCC4-VHL and RCC4-DN, respectively (Fig. 3B and D). Thus, the transepithelial resistance measurements, which were not corrected for cell number, understate the differences between subclones. From these results, we conclude that VHL loss of function in RCC4 cells is associated with a HIF-1–dependent loss of transepithelial resistance, which is consistent with a loss of adherens junctions due to reduced E-cadherin expression.

HIF-1α is specifically required for E-cadherin repression in RCC4 cells. HIF-1αDN competes with both HIF-1α and HIF-2α for dimerization with HIF-1β. HIF-1 heterodimers containing HIF-1α or HIF-2α have an overlapping but distinct set of transcriptional targets (32, 33). To determine whether HIF-1α was specifically required for the loss of E-cadherin expression in VHL-null RCC4 cells, we established by retroviral infection stable RCC4 subclones expressing GFP alone (RCC4-GFP-EV) or GFP and a shRNA (RCC4-GFP-shr2265), which modestly reduced the levels of HIF-1α protein (Fig. 4A) and mRNA (Fig. 4B) and the levels of VEGF mRNA. Despite the modest level of inhibition, E-cadherin protein and mRNA levels were increased significantly in RCC4-GFP-shr2265 compared with RCC4-GFP-EV cells. Immunocytochemistry revealed that compared with RCC4-GFP-EV cells, E-cadherin expression in RCC4-GFP-shr2265 cells was markedly increased (Fig. 4C). The increase in E-cadherin expression was associated with a significant increase in the transepithelial resistance of RCC4-GFP-shr2265 cells compared with RCC4-GFP-EV cells (Fig. 4D). Thus, interference with HIF-1α expression in RCC4 cells is sufficient for derepression of E-cadherin and increased cell-cell adhesion.

Mutually exclusive expression of HIF-1α and E-cadherin in human RCC biopsies. To investigate whether the relationship between VHL, HIF-1α, and E-cadherin that we have shown in RCC4 cells is clinically relevant, we analyzed E-cadherin expression by immunohistochemistry. In sections of normal kidney, E-cadherin was specifically detected within the distal convoluted tubules but not in proximal tubules (Fig. 5, top). Next, 13 RCC biopsies were analyzed. Eleven of these biopsies were classic clear cell renal carcinomas, in which HIF-1α was strongly expressed in the nuclei of all tumor cells, and E-cadherin expression was not detected (Fig. 5, middle). These data are difficult to interpret because it is not possible to determine whether the tumors arose from proximal or distal tubules. The most informative cases were two RCCs with mixed clear cell and granular morphology, both of which manifested the following pattern of HIF-1α and E-cadherin expression. In the areas of clear cell morphology, nuclear HIF-1α overexpression was detected, and E-cadherin expression was absent (Fig. 5, bottom). In contrast, the contiguous areas of granular morphology lacked detectable HIF-1α expression and manifested strong E-cadherin expression. Thus, within individual human RCC biopsies, expression of HIF-1α and E-cadherin was mutually exclusive.

HIF-1–mediated induction of repressors of CDH1 gene transcription in RCC4 cells. To investigate the mechanism by which increased HIF-1 activity in VHL-null RCC4 cells led to reduced E-cadherin mRNA and protein expression, we determined the levels of mRNAs encoding known transcriptional repressors of the CDH1 gene, which encodes E-cadherin. Analysis of mRNA expression by quantitative RT-PCR revealed increased levels of TCF3 (E12/E47), ZFHX1A (δEF1/ZEB1), and ZFHX1B (SIP1/ZEB2) mRNAs in RCC4 compared with RCC4-VHL cells (Fig. 6A,, top) and in RCC4-GFP compared with RCC4-DN cells (Fig. 6A , bottom). Among these three repressors, ZFHX1B showed the greatest loss of expression in response to VHL or HIF-1αDN and was the only repressor whose expression was decreased in response to shr2265 (data not shown), which may reflect the more modest reduction in HIF-1 transcriptional activity achieved in RCC4-GFP-shr2265 cells compared with RCC4-VHL or RCC4-DN cells.

Taken together, our data support a model in which dysregulated HIF-1 activity resulting from VHL loss of function in RCC4 cells induces the expression of multiple known repressors of CDH1 gene transcription, leading to a dramatic loss of E-cadherin mRNA and protein expression and decreased cell-cell adhesion (Fig. 6B). Whereas E-cadherin mRNA is repressed in RCC4 cells in a HIF-1–dependent manner, exposure of RCC4-VHL cells to hypoxia paradoxically represses TCF3, ZFX1A, and ZFX1B mRNA and induces E-cadherin mRNA expression in a HIF-1–independent manner (data not shown), thus indicating that there is considerable complexity underlying the regulated expression of E-cadherin and its repressors.

The demonstration of a reciprocal relationship between HIF-1α and E-cadherin expression in human RCC biopsies suggests that the mechanism deduced from molecular analysis of RCC4 subclones may have clinical relevance. VHL loss of function leading to HIF-1 gain of function is the earliest detectable molecular event in the pathogenesis of RCC (29), but the key HIF-1–mediated changes in gene expression that induce RCC transformation were previously undetermined. Our data implicate HIF-1–mediated loss of E-cadherin expression as a critical event underlying the epithelial-to-mesenchymal transition in RCC. However, as described in the Introduction, multiple mechanisms resulting in E-cadherin loss of function have been reported in other cancers, and there is no reason to suppose that these may not also occur in some cases of RCC. In support of this hypothesis, the VHL-null RCC line 786-0 does not express E-cadherin after VHL rescue (data not shown). Further studies are required to determine the extent to which this mechanism is responsible for loss of E-cadherin expression in other human cancers.

Grant support: NIH/USPHS grant P50-CA103175.

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.

We thank Linzhao Cheng (Johns Hopkins University, Baltimore, MD) for providing retroviral packaging vectors and advice on viral production: Luis Chiriboga (New York University, New York, NY) for technical assistance, Garry Cutting (Johns Hopkins University) for use of his voltohmmeter, Karen Padgett (Novus Biologicals, Littleton, CO) for providing anti-HIF-2α antibodies, and Celeste Simon (University of Pennsylvania, Philadelphia, PA) for providing RCC4 and RCC4-VHL cells.