Identification of Aryl Hydrocarbon Receptor as a Putative Wnt/β-Catenin Pathway Target Gene in Prostate Cancer Cells Free

The etiology of prostate cancer (CaP) across the affected population likely comprises a diverse array of selected growth and survival traits; therefore, multiple levels of complexity probably underlie the overall CaP phenotype. The wnt/β-catenin signaling pathway has been described recently to be a potentially important contributor to this disease (see Ref. 1 and citations therein). Several normal functions have been ascribed to canonical wnt signaling and its pathway components, namely, the effector molecule β-catenin, in orchestrating morphogenetic processes across a wide spatial (tissue/cell-specificity) and temporal (embryogenesis versus adult tissue homeostasis) spectrum (2, 3, 4, 5, 6, 7). These observations tie wnt/β-catenin signaling to modes of tissue-specific stem cell maintenance, proliferation, and differentiation; quite significantly, analogous cellular properties are triggered in human cancer (1, 8, 9, 10, 11). Possibly, the most overt, and likely the most influential, outcome of canonical wnt signaling is the activation of β-catenin, which entails freeing β-catenin from post-translational degradation by a complex containing the adenomatous polyposis coli (APC) gene product and Axin tumor suppressors, thereby permitting its nuclear accrual (11, 12). Once in the nucleus, β-catenin is proposed to effect changes in gene expression by trans-activating promoter-bound transcription factors such as those of the T-cell factor/lymphoid enhancer factor (TCF/LEF) family (13, 14).

The pervasive role of wnt/β-catenin-mediated gene expression in animal development and neoplasia, together with the realization that one hallmark of cancer is tissue specificity, now forms the backdrop for current research of this pathway in CaP. Links between this signal transduction pathway and prostate (normal and CaP) have only recently come to light (1). Genetic analyses of prostate tumors have uncovered potentially oncogenic (i.e., selected) mutations in β-catenin at a rate of approximately 5–10% (Refs. 11, 15, 16, 17). These alterations likely produce transcriptionally active β-catenin, as mutation-positive lesions manifest up-regulated nuclear and cytoplasmic β-catenin localization (15, 16). Despite absence of direct β-catenin mutation in certain, very different prostate tissues (normal rat prostate and advanced metastatic CaP), nuclear β-catenin has been observed, perhaps arguing that various mechanisms can operate in prostate cells to stimulate elevated β-catenin stability (18, 19). One might infer from these observations of nuclear β-catenin across this diverse group of prostate tissues that its ultimate downstream effects vary, depending on the context. Cell biological evidence supports the hypothesis that mutant β-catenin, or up-regulation of the β-catenin protein by other means, enhances gene expression in CaP cell lines (e.g., TCF/LEF-dependent transcription; Refs. 15, 18, 19, 20, 21). The potential relevance of wnt signaling to prostate physiology, possibly representing a prostate-specific function, is additionally exemplified by novel reports of an interaction between the wnt/β-catenin axis and androgen receptor; these pathways intersect leading to enhancement of ligand-dependent androgen receptor function and repression of β-catenin/TCF-dependent transcription (1, 18, 20, 21, 22, 23, 24, 25, 26, 27). It is necessary to dissect these genetic and cellular phenomena in a more physiologically relevant setting (i.e., in vivo); indeed, initial foundations have been laid recently using transgenic models that conditionally target activated β-catenin expression to the prostate, among other tissues, and directly associate β-catenin up-regulation to abnormal prostate growth (28, 29).

What are the wnt/β-catenin target genes in normal and transformed prostate tissue? The answer to this question could provide rationale behind the selection of ectopic β-catenin activity in the pathogenesis of certain CaPs; such data may additionally bridge the gap between classifying tumor genotype (e.g., β-catenin mutation-positive) and applying this toward predicting tumor phenotype. In this study, we have approached this problem by applying cDNA-microarray technology (30) along with standard cell culture techniques. Our study complements, but also differs from, recent work by Bierie et al. (29) investigating the effects of β-catenin up-regulation on expression of putative target genes in prostate. Of the several candidate genes we have scanned, certain focus will be drawn to the induction of one gene in particular, the aryl hydrocarbon receptor.

Cells were purchased from the American Type Culture Collection (DU145, PC3, LNCaP, HEK-293, control-, or wnt3a-transduced mouse L cells) or were furnished to us by Dr. John Isaacs [Johns Hopkins University, Baltimore, MD; CWR22-Rv1 (CW) and LAPC-4] and were incubated at 37°C/5% CO₂/90% humidity in medium containing 10% FCS. Specific aspects for each cell line are listed in previous reports (18, 22). All media were purchased from Invitrogen (Carlsbad, CA). Lithium chloride (LiCl) and 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) were purchased from Sigma (St. Louis, MO).

With minor modification, control- and wnt3a-conditioned media were prepared as described in a protocol provided by the American Type Culture Collection with the purchase of the appropriate mouse L cells (31). Briefly, 1 million mouse L fibroblasts (control or wnt3a cDNA-transduced) were plated to 100-mm plates in 10 ml of DMEM/10% FCS and incubated for 4 days. Medium was removed, clarified with a 0.45 μm Millex-HA syringe-driven filter unit (Millipore, Bedford, MA), and placed to 4°C. Fresh medium (10 ml) was additionally incubated on the L cells for 3 days, at which time the medium was removed and processed as before; the first and second batches of conditioned media were mixed together and stored at 4°C.

pcDNA 3.1 (referred to as pcDNA) was obtained from Invitrogen. The background and design for pcDNA-Del-β-catenin (referred to as Del-β-catenin), pOT, pOF, and ΔN-TCF4 are described in detail in references (15, 18, 22). pOT, pOF, and ΔN-TCF4 were furnished to us by Dr. Kenneth Kinzler (Johns Hopkins University). Plasmids for constructing recombinant adenoviruses, pAdTrack and pAdEasy-1 were originally engineered by the laboratory of Bert Vogelstein (Johns Hopkins University) and Kenneth Kinzler and furnished to us by Dr. Ron Rodriguez (Johns Hopkins University). The plasmid for expression of enhanced green fluorescent protein (GFP) was purchased from Clontech (Palo Alto, CA). For expression of hemagglutinin (HA)-tagged Del-β-catenin, we subcloned the Del-β-catenin open reading frame (starting at the second residue) downstream and in frame of the sequence encoding the HA epitope (YPYDVPDYA), and a start codon, all within the pcDNA backbone. pRL-CMV was purchased from Promega (Madison, WI); this vector encodes Renilla luciferase under the control of the cytomegalovirus promoter and was used to monitor cell transfection efficiency in luciferase assays.

Antibodies against the listed proteins were purchased from the following companies, BD Biosciences PharMingen (San Diego, CA; β-catenin mouse monoclonal antibody), Oncogene Research Products (San Diego, CA; α-tubulin), Affinity BioReagents (Golden, CO; aryl hydrocarbon receptor, CYP1A1), Pierce (Rockford, IL; rhodamine-coupled antimouse IgG, horseradish peroxidase-coupled antimouse and antirabbit IgG), Molecular Probes (Eugene, OR; Alexa 488-coupled antirabbit IgG), NeoMarkers (Lab Vision Corporation, Fremont, CA; β-catenin rabbit polyclonal antibody), Sigma (HA rabbit polyclonal antibody), and Torrey Pines BioLabs (Houston, TX; green fluorescent protein).

The following protocol briefly lists the major steps involved in the AdEasy recombinant adenovirus production system; for complete directions, we referred to a protocol provided by Q-Biogene (Carlsbad, CA). Adenoviral plasmid constructs containing either an empty cassette or Del-β-catenin were prepared using standard recombinant cloning techniques (32). These constructs, pAdTrack-empty and pAdTrack-Del-β-catenin, were recombined with pAdEasy-1 in bacteria. Using Fugene-6 reagent (Roche, Indianapolis, IN), the newly recombined AdTrack/AdEasy constructs were then transfected into the HEK-293 helper cell line, which permits productive adenoviral infection. Adenoviral particles were acquired by subjecting infected cells to multiple free-thaw cycles using methanol/dry ice and 37°C water baths. Particles were amplified by sequential infection of HEK-293 cells. Quantification of infectious particles per ml was carried out by entering infected cell populations (24 h) into flow cytometric analyses (Becton Dickinson FACScan) to detect the degree of green fluorescent protein activity emitted from individually infected cells. Calculations were verified by cell infection (HEK-293 and other cell lines including CW) followed by green fluorescent protein expression analyses using immunofluorescence and Western blot procedures; similarly, Del-β-catenin expression was examined. Adenoviral preparations were snap frozen and stored at −70°C. For microarray experiments, 1 × 10⁷ CW cells were plated to T75 flasks and infected the following day at a multiplicity of infection of 10 with either control or Del-β-catenin adenovirus preparations. Cells were entered into an RNA preparation procedure 24 h after infection. Infection of CW cells at a multiplicity of infection of 10 permitted near complete transduction of the cell population with limited cell rounding.

Total RNA was prepared from CW cells infected with either control or Del-β-catenin-expressing adenoviruses using TRIzol reagent (Invitrogen). Cells were lysed in 8 ml of TRIzol, followed by 5 min of vortexing. Chloroform (0.2 volumes) was mixed with each preparation by vigorous shaking. Mixes were incubated for 3 min and then spun at 10,000 × g for 15 min at 4°C in a Beckman centrifuge. The aqueous portion of the phase separation was removed and gently mixed with an equal volume of 70% ethanol. The mixtures were then entered into RNeasy Midi columns (Qiagen, Valencia, CA) and centrifuged twice. The columns were then processed according to the manufacturer’s directions. RNA was eluted in 500 μl of diethyl pyrocarbonate-treated water and then concentrated using Microcon YM-100 filters (Millipore). The integrity of isolated RNA was assessed using the RNA 6000 Nano assay on an Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA).

Using 65 μg of total RNA as a template, cDNA was generated using Superscript II reverse transcriptase (Invitrogen) and oligodeoxythymidylic acid [oligo(dT)] primer (Sigma Genosys, The Woodlands, TX). Briefly, 2 μg of oligo(dT) was annealed to total RNA in a 20 μl reaction by incubation at 68°C for ten min. The reactions were then placed on ice for 2 min, followed by the addition of first-strand reaction mix composed of first-strand buffer (Invitrogen), low T deoxynucleoside triphosphate mix (Amersham Biosciences, Piscataway, NJ), 1 mm Cy5 or Cy3 dUTP (Amersham Biosciences), DTT (Invitrogen), and 400 units of Superscript II in a 40-μl reaction. Reactions were then incubated at 42°C for 30 min after which time another 400 units of Superscript II was added for continued incubation (90 min). Reactions were terminated by addition of 0.5 m EDTA (pH 8.0)/1 N NaOH and then incubation at 65°C for 15 min. They were then cooled to room temperature and neutralized by the addition of 1 M Tris-HCl. The entire reaction volume was combined with 10 μg CoT-1 DNA (Invitrogen) and 10 mm Tris (pH 7.5)-1 mm EDTA, placed in a Microcon YM-100 filtration unit (Millipore) and concentrated.

The fluor-labeled cDNA probes were combined; cDNA from cells infected with control and Del-β-catenin-expressing adenoviruses were labeled with Cy5 and Cy3, respectively. The pooled, labeled cDNAs were then combined with hybridization reagents comprised of 8 μg poly dA (Amersham Biosciences), 4 μg yeast tRNA (Sigma), 20× SSC, 50× Denhardt’s solution (Sigma), and 10% SDS. The hybridization solution was incubated at 98°C for 2 min, placed on ice for 25 s, applied onto a 24 × 50 mm glass coverslip (Corning, Big Flats, NY), and touched with an inverted microarray slide. The microarray slide was placed in a hybridization chamber and incubated in a 65°C water bath for 16 to 20 h. After this incubation the microarray slides were washed with 0.5× SSC/0.01% SDS until the glass coverslip separated from the microarray. The slides were then transferred to new wash solution, gently shaken for 1 min and let stand for 1 min; this procedure was likewise performed in a wash solution of 0.06× SSC. The slides were centrifuged at 900 rpm to dry the microarrays.

Slides were scanned using an Axon Scanner 4000A setup. Image analysis was performed using IPLab (Scanalytics, Fairfax, VA) with custom extensions for array analysis. Data from all hybridizations were imported into a FileMaker Pro database for additional analysis.

The microarray slides, which consisted of either 6,500 or 12,000 elements, were obtained from the National Cancer Institute and the Johns Hopkins Oncology Center Microarray Core Facility, respectively. Both sets of microarrays measure relative gene expression based on competitive hybridization between reverse-transcribed, fluor-labeled experimental cDNA populations to spotted cDNA clones. RNA prepared from four separate adenoviral-mediated expression experiments, all comprised of identical parameters (see above), was entered into two separate analyses using either type of microarray slide.

RNA for validation studies was prepared using the GenElute Mammalian Total RNA Miniprep kit (Sigma) following the manufacturer’s protocol. All of the RNA preparations were stored at −70°C and analyzed by agarose gel electrophoresis to ensure product integrity. One μg of total RNA was used as template for synthesis of cDNA using Superscript II reverse transcriptase (Invitrogen) and oligo(dT) primer (see protocol above) in a 20-μl reaction. One μl of the reverse transcription products was entered into PCR reactions (50 μl) for amplification of the cDNA of interest. All of the reactions used Taq DNA polymerase from Roche and 10% DMSO (Sigma). The primer sequences and PCR parameters (times and temperatures) used for each particular gene amplification reaction are available on request. Five or 10 μl of PCR products were analyzed by 1.5% agarose gel electrophoresis, ethidium bromide staining, and visualization under a UV lamp.

The luciferase assays in this study were performed exactly as those described previously (22), except as noted here. One day after plating, cells were transfected in the context of DMEM/10% FCS using 1 μl Fugene-6 transfection reagent per 0.2-μg transfection DNA. Twelve or 24 h before analysis of luciferase activity (firefly and Renilla), medium was replaced with 50 μl of either control or wnt3a-conditioned medium (CM).

Western blot analysis of protein expression was performed as described in a previous report (22) using standard protocols (32). Briefly, cells were lysed in radioimmunoprecipitation assay (RIPA) buffer [150 mm NaCl, 1.0% NP40, 0.5% deoxycholate, 0.1% SDS, and 50 mm Tris (pH 8.0)] containing a protease inhibitor mixture (Roche) and assayed for total protein concentration by BCA protein assay (Pierce). Released DNA was sheared by brief sonication, after which appropriate amounts of protein lysate were loaded onto large Hoeffer gels and electrophoresed alongside prestained molecular weight protein markers (Bio-Rad, Hercules, CA). Gels were transferred to nitrocellulose (Amersham), blocked in 5% nonfat dried milk/0.1% Tween 20/1× PBS and probed sequentially with appropriate primary and secondary antibodies. Probed membranes were entered into an enhanced chemiluminescence (ECL) analysis (Amersham). For certain experiments, developed films were analyzed with a scanning densitometer (Perkin-Elmer, Torrance, CA).

Immunofluorescence analyses were performed as described in previous work (22). For antibody combinations and controls used in the present studies, refer to the Fig. 5 legend. 4′,6-Diamidino-2-phenylindole staining was applied for all of the immunofluorescence work to monitor nuclei/cell number, and Prolong AntiFade (Molecular Probes) was used for final slide preparation. Staining was analyzed by immunofluorescence microscopy (Zeiss AxioScope), and images were acquired and normalized using IPLab.

In our earlier attempts to uncover potential wnt/β-catenin target genes in CaP cells by cDNA microarray analysis, we applied an inducible system (Invitrogen). A mutant, hyperactivated form of β-catenin (Del-β-catenin, Δ24–47; Ref. 15) was placed under the regulatory control of ponasterone, permitting strong overexpression (data not shown). These assays produced a high number of false-positive readings that were probably a result of multiple factors: (a) the “prostate” cancer cells entered into this screen (TSU-Pr1) were categorized recently as having originated from bladder carcinoma (33), interestingly, a malignancy not shown to select for β-catenin up-regulation in Western populations (34); (b) stable cell clones were used, increasing the risk of uncovering a clonal gene response to β-catenin signaling that is not broadly applicable to other types of CaP cells; and (c) despite efforts to control for the strong side-effects exerted by ponasterone on gene expression, an underlying propensity existed for uncovering artifacts.

To circumvent problems associated with studying stable clones, we used the AdEasy/AdTrack adenovirus expression system (35) to express Del-β-catenin in CaP cell lines (CW, LAPC-4, LNCaP, and DU145). Western blotting, immunofluorescence, RT-PCR, and flow cytometric analyses were used to determine the parameters (infection duration and infectious particle dose) to achieve the highest Del-β-catenin expression accompanied by the least cell rounding and toxicity.³ Because CW cells demonstrated robust adenoviral-mediated overexpression of the transgene, and represent a suitable in vitro CaP model (36), we entered these cells into our microarray analyses. Total RNAs from CW cells infected at an multiplicity of infection of 10 for 24 h with either Del-β-catenin-expressing adenovirus or control (empty) adenovirus were compared by microarray analysis (30). In general, alterations in gene expression were subtle; indeed, the overall fold increase in β-catenin transcription was quite modest (Table 1; data not shown) despite ample mRNA levels as measured by RT-PCR (Fig. 1,A). Table 1 lists the fold differences in expression of β-catenin and of those putative target genes of which the expression changes (i.e., induction) have subsequently been verified by various RT-PCR analyses (see below). These genes [AHR, transmembrane protein 2 (TMEM2), aldehyde dehydrogenase 1 family member A1 (ALDH1A1)] have not been implicated previously as downstream targets of the wnt pathway. For comparison purposes, included in Table 1 is the α-catenin gene, of which the transcription is not known to be responsive to nuclear β-catenin, and Axin2/conductin, a bona fide β-catenin/TCF target gene (37, 38, 39). c-Myc and cyclin D1, two β-catenin target genes discovered in colorectal cancer cell lines (40, 41, 42), were not induced under these experimental conditions; their calibrated fluorescence ratios were 1.027 ± 0.087 and 0.900 ± 0.109, respectively. It is important to note here that several candidate target genes (∼12) were initially assayed for differential expression; however, only a few genes (above) consistently exhibited unambiguous up-regulation in β-catenin target induction experiments independent from those based on adenoviral transduction. This general outcome may result from the above-mentioned observation that putative target genes showed only modest up-regulation by microarray analysis, thereby not standing apart from genes elevated in a manner unrelated to β-catenin overexpression. The application of four individual microarray experiments partly alleviated this problem.

We applied RT-PCR to follow up and confirm certain of the changes in gene expression noted in Table 1. To test the redundancy of these proposed β-catenin pathway endpoints in CaP cells, we examined these changes in gene expression in both CW and LAPC-4 cells, the latter of which originate from an advanced, hormone-refractory tumor (43). As expected, infection of both cell lines with the Del-β-catenin-expressing adenovirus led to an ample increase of total β-catenin transcripts (Fig. 1, row 3). Using primers designed to detect both endogenous and exogenous β-catenin mRNAs, expression of the Del-β-catenin species was detected as a band of slightly faster gel migration (Fig. 1, row 4). As a known β-catenin/TCF target, the induction of Axin2 by mutant β-catenin (Fig. 1, row 5) additionally validates the utility of our microarray analyses and this RT-PCR experiment; of note here, we have found by transient transfection that Axin2 is responsive to β-catenin across most CaP cell lines and follows an induction profile consistent with published luciferase reporter assays (Ref. 18; data not shown). AhR and TMEM2 gene expression was induced in both cell lines after infection with the Del-β-catenin-expressing adenovirus (Fig. 1, rows 6 and 7), whereas c-myc and cyclin D1 showed no apparent changes under the same conditions (data not shown). Not all of the genes showing up-regulation by microarray analysis displayed true differential expression in this RT-PCR assay and assays described below (data not shown); the identification of false-positives by these secondary screens additionally provided to us the sense that genes manifesting genuine up-regulation do not simply reflect a global, nonspecific pattern of elevated expression. In total, these mRNA expression data conform well with our microarray results and support the hypothesis that AhR and TMEM2 positively respond to β-catenin signals.

The putative target genes listed in Table 1 have been shown to be sensitive to elevated levels of Del-β-catenin introduced by ectopic expression techniques (viral transduction and plasmid transfection, Fig. 1; data not shown). Therefore, we were interested in determining whether or not putative target genes are likewise up-regulated by different modes of canonical wnt pathway induction. Such an inquiry might provide resolution to an important question: is endogenous, wild-type β-catenin in CaP cells competent to perform target gene transactivation, or are the aforementioned results dependent on gross overexpression of mutant β-catenin? Posing this question should also shed light on the degree to which prostate cells can sustain alternative modes of β-catenin activation (i.e., are wnt pathway components intact in these cells?). We first tested the effects of LiCl on the transcript levels of putative target genes; lithium inhibits glycogen synthase kinase-3β, a component of the β-catenin degradation complex, and renders higher levels of stable/activated β-catenin (44). As determined by a 24 h dose-response assay, 40 mm LiCl was found to amply induce β-catenin protein levels in human embryonic kidney (HEK-293) cells (Fig. 2,A); these cells contain very low levels of β-catenin stabilized to cell-cell adhesions, and as such presented a convenient means by which to examine the effects of LiCl on soluble β-catenin. Various CaP cell lines were then treated with 40 mm LiCl for certain time periods and analyzed for β-catenin protein expression (Fig. 2,B). Total β-catenin protein displayed increased levels after 12 h of LiCl treatment, but dropped lower after 24 h with the exception of CW cells, which maintained increased β-catenin protein through 48 h of treatment. Because we were able to detect noticeable changes in β-catenin metabolism by LiCl application in all of the cell lines tested (Fig. 2,B; data not shown), we next analyzed the extent to which such β-catenin stabilization/induction could impact target gene expression. Excluding that gene expression observed for LNCaP and PC3 cells, the known wnt pathway target gene Axin2 exhibited increased transcript messages after 12- and/or 24-h treatment with 40 mm LiCl (Fig. 2,C). This result is consistent with earlier work (18, 22) demonstrating that CW, LAPC-4, DU145, and HEK-293 cells, but not PC3 and LNCaP cells, are capable of driving target gene transcription with β-catenin overexpression. The putative target genes, TMEM2 and AhR, followed similar response profiles with Axin2 induction in CW, LAPC-4, and HEK-293 cells, but not in DU145 cells. Interestingly, PC3 cells did demonstrate up-regulation of TMEM2 but not of AhR; LNCaP cells were tested for AhR gene induction by lithium and showed no response (Fig. 2 C). Altogether, these data support the proposal that the target genes in question (AhR and TMEM2) are sensitive to both ectopically and endogenously derived β-catenin. The differences between induction of an established wnt target gene (Axin2) and of novel target genes (AhR and TMEM2), as well as certain disparities that exist between these data and published luciferase reporter results (18), could be indicative of fundamental differences in the respective modes of regulatory control placed on these genes (i.e., cellular specificity).

We wished to test a third methodology for inducing wnt/β-catenin signaling to additionally investigate the fidelity of these differential expression results and to possibly illuminate upon the cell line-specific responses to LiCl-stimulated changes in gene expression. Simple overexpression of ectopic wnt factors to potentiate canonical pathway activity in cultured cells usually leads to poor β-catenin induction; however, others have demonstrated that, in particular instances, wnt factors secreted into media retain induction potential (31, 45, 46, 47). We chose a commercially available cell culture system for preparation of human wnt3a: control (no secreted product) and wnt3a-CM were prepared using parental or retrovirally transduced, stable wnt3a-expressing mouse L cells, respectively (31, 45). Various batches of these CM were prepared and tested on CaP cell lines to measure β-catenin/TCF-related transcriptional activity (CRT) by luciferase assay (Fig. 3,A); the reporter plasmids used contain either three wild-type (OT) or mutant (OF) consensus TCF/LEF binding elements and positively respond (OT only) to overexpressed β-catenin in CaP cell lines (18, 48). In cell lines tested (CW, DU145, and HEK-293), wnt3a-CM elicited strong luciferase activity using the experimental OT reporter, but not the mutant OF reporter (Fig. 3,A); therefore, certain CaP cell lines are able to respond appropriately to wnt factors in vitro (i.e., drive specific β-catenin/TCF-dependent transcription, or CRT). Finally, these data correlate closely with similar luciferase assays for CRT reported previously (mutant β-catenin-driven CRT; Ref. 18). Applying these same CM preparations, we analyzed β-catenin protein expression and localization to additionally understand the nature of wnt-activated β-catenin in CaP cell lines. As shown in Fig. 3,B, wnt3a-CM does act to increase total β-catenin protein in cells compared with control CM. CW cells up-regulated β-catenin after 24-h incubation with wnt3a-CM, whereas DU145 cells exhibited elevated β-catenin levels only at the earlier 12-h time point. HEK kidney cells displayed robust protein up-regulation by wnt3a at all of the time points collected. Regarding CW CaP cells, and to a lesser extent DU145 CaP cells, we were intrigued by the seeming disparity between their respective Western blot and luciferase assay data (Fig. 3, A and B), having expected to detect a stronger increase in β-catenin protein expression to parallel the relatively pronounced wnt3a-induced CRT. Hence, we assessed by immunofluorescence analysis the localization of β-catenin in CW and DU145 cells treated with either CM and discovered that wnt3a treatment augments cell-cell border staining, but not nuclear staining (data not shown). This result agrees with previous work showing that one of the more noticeable effects of wnt signaling is greater cell border staining (46, 49); newly stabilized β-catenin can interact with adhesion complexes. Only in HEK cells, which lack prominent cell-cell adhesions, did we note faint positive nuclear staining. These data would give the impression that β-catenin nuclear signaling need not primarily be associated with detectable nuclear staining; thus, it will be important to determine the extent to which β-catenin nuclear signaling may occur across the gamut of human prostate tumors in a manner independent of nuclear localization. On the other hand, recent evidence could argue that this increase in β-catenin membrane staining is the molecular event responsible for promoting elevated nuclear signaling (50).

We examined the effect of wnt3a-induced β-catenin protein accumulation on putative target gene expression. Serving as a positive control based on previous experiments, Axin2 appropriately exhibited strong induction in CW, LAPC-4, DU145, and HEK-293 cells, whereas its induction in LNCaP and PC3 cells was lacking (Fig. 3,C; data not shown). After 12- and 24-h incubation periods, the putative wnt/β-catenin targets AhR, TMEM2, and ALDH1A1 were all observed to show induction by wnt3a-CM in CW cells; LAPC-4 cells likewise showed notable induction of AhR and TMEM2 with the same CM (Fig. 3,C). Although DU145 had elevated Axin2 expression in response to wnt3a-CM, induction of AhR and TMEM2 was minimal. LNCaP and PC3 CaP cells, together with HEK-293 cells, contained no elevated putative target gene expression in response to wnt3a (Fig. 3,C). Overall, these results and those in Figs. 1 and 2,C strengthen the assertion that certain putative target genes, as originally judged by microarray analysis of cells overexpressing mutant β-catenin (Table 1), show comparable up-regulation in response to two different means of stimulating wnt/β-catenin activity. This conclusion is especially the case for CW and LAPC-4 cells; however, discrepancies do exist. PC3 CaP cells up-regulated TMEM2 in response to lithium treatment (Fig. 2,C), but did not similarly respond to wnt3a-CM (Fig. 3,A). Discrepancies likewise exist between the outcome of either induction mode, lithium versus wnt3a, and certain target gene responses in LNCaP and HEK-293 cells. We find it interesting to point out that, although certain cell lines demonstrated Axin2 expression on wnt3a treatment (Fig. 3 C), they did not in turn up-regulate c-myc or cyclin D1 to any real extent. This general result conforms with data provided by the microarray work.

Does the basis for increased expression of the putative target genes, AhR and TMEM2, involve direct up-regulation of gene transcription? Knowing the general answer to this question should lend insight into the mechanism(s) responsible for the above observations. Because utilization of wnt3a-CM offered us a relatively easy and expeditious way for which to efficiently modulate β-catenin signaling, we used CM together with the protein translation inhibitor cycloheximide to test the extent to which de novo protein synthesis might be required for induction of putative target RNA expression by β-catenin activation. In Fig. 3 E, we observed that, despite the application of 20 μg/ml cycloheximide to LAPC-4 cells incubating with wnt3a-CM, AhR and TMEM2 remained elevated compared with that expression occurring in control CM. The gene expression profiles of these genes matched that profile for Axin2, which served as a control in its capacity as a known transcriptional target of β-catenin/TCF. Therefore, it is plausible that the various modes of putative target RNA up-regulation described in this report derive from stabilized β-catenin, not newly synthesized β-catenin or other proteins, interfacing directly with certain nuclear regulatory components to alter gene expression. These data, however, do not exclude the possibility that the observed escalation in target gene RNA levels results from other mechanisms, such as increased RNA half-life.

An important question to consider was whether or not certain of the novel wnt/β-catenin pathway target gene candidates, of which the RNA levels are augmented by multiple modes of β-catenin activation, similarly respond at the translational level. We chose to study AhR protein expression, a decision influenced by both the fact that AhR function has been fairly well studied (51, 52), and the commercial availability of anti-AhR antibodies. Understanding wnt/β-catenin signaling on AhR protein expression would potentially offer more credibility to our differential RNA expression results, as well as build on hypotheses explaining the role of wnt signaling in CaP. Simple overexpression of the constitutively active mutant Del-β-catenin in CaP cell lines CW and LAPC-4 led to higher levels of AhR protein as determined by Western blot analysis (Fig. 4,A). We also asked whether or not a dominant-negative form of TCF4 (Refs. 8, 10; this transcription factor has been widely implicated in integrating β-catenin signals) might impinge upon this up-regulatory activity. Coexpression of ΔN-TCF4 with Del-β-catenin resulted in dampening of this response in AhR protein expression in CW cells, but not LAPC-4 cells (Fig. 4,A); with regards to this inhibitory effect in CW cells, this result may indicate that AhR induction by β-catenin relies on TCF/LEF-mediated modulation of AhR regulatory DNA sequences. Such an association between β-catenin-mediated expression of AhR and TCF function would be consistent with the data shown in Fig. 3 E, insofar as it relates to the direct action of a putative β-catenin/TCF transcriptional complex near or at the 5′ upstream sequence of the AhR gene.

We next tested the responsiveness of AhR protein expression to wnt3a-CM. Accounting for our RT-PCR analyses in Fig. 3,C, we were not overly surprised to observe AhR protein induction upon cell stimulation with wnt3a (Fig. 4,B). Next, we tested the extent to which certain cells that did not exhibit strong wnt3a-mediated induction of AhR message (see Fig. 3,C) might alter AhR protein expression on wnt3a stimulation. AhR protein expression paralleled RNA expression in DU145 and LNCaP CaP cells with β-catenin activation, whereas HEK-293 kidney cells did show a modest increase in AhR protein (Fig. 4,C; data not shown). These data in Fig. 4, B and C, in composite, demonstrate a direct correlation between AhR RNA and protein levels, at least as it concerns the metabolic changes accompanying wnt/β-catenin signals.

As discussed above, β-catenin in wnt3a-stimulated CW cells showed no detectable nuclear accumulation; therefore, we wished to determine by immunofluorescence analysis whether or not a constitutively active form of β-catenin, which does permit high-level detection of nuclear staining (15), may impact AhR expression. If AhR induction is apparent by this assay, we should also be able to assess on a cell-to-cell basis the degree to which AhR protein expression occurs concomitantly with nuclear β-catenin expression. Interestingly, there was no obvious correlation between the expression of AhR and ectopic β-catenin; cells exhibiting elevated AhR expression did not manifest intense nuclear β-catenin staining, and vice versa (Fig. 5,A). This result may suggest that, although activation of β-catenin does lead to elevated AhR RNA and protein expression, it does so in a manner not easily distinguished by the occurrence of gross nuclear β-catenin accumulation alone. For example, the response of AhR protein expression to the β-catenin signal may both initiate, and terminate, before β-catenin accrues to appreciable levels. Elevated AhR expression in this experiment was due to transient expression of Del-β-catenin, as cells transfected with an empty vector showed only uniform staining. It is important to note that the localization of β-catenin-induced AhR expression in comparison to basal AhR expression did not noticeably differ (Fig. 5 A); AhR showed predominant nuclear staining. Additional work has shown that CaP cell lines (CW and LAPC-4) exhibit both cytoplasmic and nuclear AhR staining under basal conditions (i.e., no exogenous AhR ligand present; data not shown), an observation having precedent (53).

To additionally address the issue of correlating mutant β-catenin expression with augmented AhR staining, we transfected CW cells with an HA-tagged Del-β-catenin construct and double-stained cells for AhR and HA (Fig. 5,B). Our unpublished studies have shown that, in CW cells bearing transient HA-Del-β-catenin expression, a majority of transfected cells exhibit HA positivity restricted to cell/cell borders. This localization event, which is not discernable when using anti-β-catenin antibodies (endogenous, membranous β-catenin expression is high), is associated with the β-catenin/E-cadherin interaction (54), because an HA-tagged species of β-catenin incompetent in E-cadherin binding localizes entirely to the nucleus. We were interested in deciphering the extent to which AhR induction may be associated with membranous β-catenin overexpression. Perhaps such an analysis would allow detection of that β-catenin localization event (nuclear versus membranous), which is more likely indicative of concurrent target gene up-regulation. In Fig. 5,B, we observed that, for the most part, strong expression of AhR was not associated with strong nuclear β-catenin expression. Instead, strong AhR staining was often associated with HA positivity (β-catenin) at the membrane (consider examples indicated by arrows). This result may imply that AhR induction by β-catenin, and by extension other target genes, occurs in a fashion that does not necessitate a strong nuclear β-catenin presence. Conditional β-catenin up-regulation in the prostate elicits elevated β-catenin expression at cell-cell borders that is tied to an increased, but diffuse, cell body expression pattern (i.e., nuclear β-catenin is not predominant; see immunofluorescence images in Refs. 28, 29); although difficult to make a direct comparison, this previous observation may be consistent with our result in Fig. 5,B. On the other hand, however less frequently, we did detect cells with strong nuclear β-catenin staining that showed a concomitant increase in AhR staining (Fig. 5,B, arrowhead), albeit modest. Altogether, these results (Figs. 4 and 5) lead us to conclude that AhR protein, and not only its RNA, is up-regulated upon activation of the wnt/β-catenin pathway.

AhR is a transcription factor capable of trans-activating target gene expression upon binding ligand. Although no in vivo ligand has been defined for AhR, several environmental carcinogens are known to interact with AhR and, in turn, up-regulate its transcriptional activity (51, 52); one such ligand is TCDD. We were interested in investigating whether or not wnt/β-catenin-mediated up-regulation of AhR expression can couple with TCDD to enhance the overall output of AhR activity. Although both of these pathways likely exert multiple, complex effects on CaP cells, approaching this question experimentally may yield an early view into how these pathways interact. We found that, as described above, wnt3a-CM induced AhR protein expression to a similar degree as described above (12- and 24-h time points); cotreatment with 10 nm TCDD led to variable responses in AhR protein levels by CW and LAPC-4 cells (Fig. 6). LAPC-4 cells generally exhibited a pronounced decrease in AhR protein with exposure to TCDD, and this response occurred regardless of wnt3a-CM treatment. This effect corresponds well with reports demonstrating that TCDD leads to decreased AhR protein half-life; ubiquitin-mediated proteasomal degradation has been implicated in this post-translational down-regulation (55, 56). Conversely, CW cells demonstrated an increase in AhR protein expression after 12-h TCDD treatment, but similar to the response in LAPC-4 cells, wnt3a stimulation did not impact this response. The cytochrome P450 enzyme, CYP1A1, is a target gene for AhR transcriptional activity (52); gene-knockout experiments have demonstrated that CYP1A1 expression may be totally dependent on AhR (57). As such, we evaluated the degree to which CYP1A1 protein expression responded to TCDD treatment. As shown in Fig. 6, both cell lines maintain moderate basal CYP1A1 protein expression; surprisingly, TCDD strongly suppressed CYP1A1 levels in CW and LAPC-4 cells at both time points tested. These early results suggest that, based on the parameters used for this study, wnt/β-catenin signaling does not converge with and synergistically potentiate AhR signaling in CaP cells in vitro.

In this report, we have begun to dissect the nature of target gene expression promoted by the wnt/β-catenin pathway in CaP cells, in vitro. This general question is of relevance to the CaP field, as this pathway has been linked previously to certain aspects of this disease (1). Focus was drawn toward the expression, although still putative, of real β-catenin target genes in CaP cells; we have demonstrated previously that β-catenin can potentiate TCF-mediated transcription from artificial reporter plasmids in these cells (18, 22), thereby providing both a premise and basic framework for broaching the present topic. cDNA microarrays, which have become a main tool in our laboratory for analyzing differential gene expression in prostate tissue (30, 58), were used to probe this question, and in doing so, uncovered certain genes of which the mRNA is elevated in a manner associated with increased β-catenin activation. The putative targets highlighted in our report, AhR and TMEM2, were up-regulated at the RNA level in response to multiple modes of β-catenin up-regulation, these being overexpression of a mutant, hyperactive form of β-catenin (Del, Δ24–47), which has been implicated previously in advanced disease (15), and treatment of cells with LiCl or wnt3a, both of which up-regulate endogenous β-catenin activity (31, 44). Current work is aimed at using these strategies to verify the expression of other microarray candidate genes not presented in this report. Perhaps of greater significance, we additionally demonstrated that RNA up-regulation can be correlated with protein expression; AhR protein levels are elevated upon mutant β-catenin overexpression and wnt3a treatment. The implications for this observation are discussed below.

We were led to the above changes in target gene expression by microarray analysis of CW CaP cells transduced to overexpress Del-β-catenin. Upon validating the altered expression patterns of certain putative target genes, we subsequently showed that they were similarly regulated in LAPC-4 cells; but, to a large extent, these changes did not universally carry over to other CaP cell lines (LNCaP, DU145, and PC3). For these latter cell lines, we uncovered no conspicuous changes in expression of AhR and TMEM2 in response to wnt3a-CM, whereas in certain cases, such changes were apparent after LiCl exposure. Such discrepancies may derive from the pleiotropic effects known to be elicited by LiCl; besides its role in GSK-3β inhibition, lithium is known to shape a number of other signal transduction events (59). If Axin2 RNA induction represents any true measure of wnt/β-catenin signaling (37, 38, 39) in CaP cells, then any substantial deviation in induction of a putative target gene compared with that of Axin2 may flag such a change as being unrelated to bona fide β-catenin activity. For example, despite induction of TMEM2 in LNCaP cells treated with LiCl, it may be of no surprise that TMEM2 does not likewise manifest up-regulation by wnt3a-CM, given that Axin2 expression does not change with either wnt3a-CM or LiCl (see Fig. 2,C and Fig. 3 C). This latter result is credible, given that past data have shown that LNCaP cells may be recalcitrant to β-catenin/TCF signaling (18). LNCaP cells not withstanding, we find it of interest to note that those CaP cell lines (in this study, CW and LAPC-4) better resembling prostate tumor cells in vivo (e.g., express androgen receptor and prostate-specific antigen; Refs. 36, 43) are more capable of potentiating putative wnt/β-catenin target gene induction. This result is in comparison to DU145 and PC3 cell lines, both of which are believed to represent cells relatively atypical to most prostate tumors; they have fewer of those luminal/secretory differentiation qualities (60, 61) usually ascribed to the bulk of the cells composing human prostate tumors. Both of these cell lines have been shown to display very modest or total lack of β-catenin/TCF signaling potential (18). Altogether, these data may point to that cellular compartment within prostate tumors that may be of relevance when considering wnt/β-catenin signaling in CaP; that being said, more CaP cell lines must be examined to additionally refine this general hypothesis.

We examined the inducibility of certain genes, namely AhR and TMEM2, to wnt/β-catenin signaling in CaP cell lines, which provided very much an in vitro setting. It will important to know the status of such sensitivity in CaP cells in vivo; but in addition, a question worth raising is whether or not these genes respond in kind to β-catenin signaling in normal prostate epithelial cells. The former question can be addressed histologically: are putative target gene products such as AhR up-regulated in CaP specimens that show etiological evidence for aberrant β-catenin activity (e.g., expression in tumors bearing β-catenin mutation and/or abnormal localization)? One study has focused on AhR protein in prostate and found it to be expressed throughout many human prostate tissues (normal, benign prostatic hypertrophy, carcinoma; 62); therefore, AhR protein expression may be relevant to understanding different facets of prostate biology and could be amenable for future studies attempting to link it with in vivo β-catenin activation. The latter question on the nature of target gene expression in normal prostate has been raised by Bierie et al. (29), who have recently used microarrays to investigate gene expression in mouse prostate bearing mutant β-catenin. In their model, mutant β-catenin is expressed in the context of otherwise normal prostate cells, leading to hyperplasia and transdifferentiation (these observations are fairly similar to those described in a similar study; Ref. 28). Several differentially expressed genes were culled out when comparing mutant β-catenin-expressing prostate tissue with that from wild-type; however, as duly noted by the authors, the extent to which these genes represent targets closely associated with β-catenin activity in prostate cells, or are only manifested from the heavy transdifferentiation ascribed to the experimental tissue, remains at issue. Aside from the fact that we used a cultured, homogenous population of CaP cells (CW cell line), our microarray work has potentially focused attention on genes targeted by β-catenin in the context of transformed/neoplastic prostate cells, potentially explaining the lack of similarity between our results and those of Bierie et al. (29). Perhaps an interesting inquiry to broach using this mouse model will be to assess the global expression pattern of mutant β-catenin-expressing prostate tissue upon its progression to full-blown carcinoma; such a setting could possibly be achieved by either chemical mutagenesis or introduction of alleles that potentiate tumorigenesis. Knowing the answer to this question may help elucidate the nature of aberrant β-catenin activation during cell transformation and selection (i.e., in the prostate, when is increased β-catenin activity a selected trait and how does it mediate a growth advantage?).

With the exception of HEK-293 kidney cells, we did not examine the responsiveness of AhR and TMEM2 gene expression to wnt/β-catenin in other tissue-specific cell lines aside from those derived from CaP. That being said, we would be interested to know if these genes are subject to the same responsiveness in other cell/tissue types. Understanding this question (learning through analogy) may reflect upon the role these particular target genes play in wnt signaling in CaP. Posing a problem to answering this question, however, is the relative scarcity of strong evidence demonstrating that β-catenin target genes reported for one particular tissue type are likewise regulated in other tissues. Perhaps this general observation is unsurprising, as tissues assume specific form and function, and, therefore, demand specific types of output (e.g., gene expression) from “common” signal transduction events throughout tissue development and homeostasis. As shown in this report and in unpublished studies, we have been unable to detect strong up-regulation of genes identified previously as β-catenin targets (c-myc and cyclin D1). However, we have been able to detect ample Axin2/conductin gene sensitivity to β-catenin. These data altogether suggest that target genes broadly fall into two different categories, those that respond to wnt/β-catenin signaling in a tissue-restricted manner and those that show conservation between various tissue types. Tissues that are believed to share certain characteristics might be expected to exhibit similar signaling endpoints. For example, Bierie et al. (29) make mention that a portion of those genes up-regulated in prostate tissue bearing mutant β-catenin expression overlap with those similarly up-regulated in mammary gland (63). Although mutant β-catenin expression imposed a squamous transdifferentiation program on prostate and mammary glands—indeed, this phenotype itself may partly account for the shared differential gene expression—one could argue that the overlap in gene induction is related to the general degree of similarity held between the function and hormonal regulation of these tissues.

We extended our analyses to examine not only the expression of AhR, but also its function as a transcription factor. Although AhR protein is expressed in both normal and neoplastic prostate tissue (62), its function remains poorly understood. Cytoplasmic AhR protein is thought to bind ligand, upon which time it dissociates from chaperonin proteins (e.g., heat shock protein-90) and translocates into the nucleus. Nuclear AhR heterodimerizes with AhR nuclear translocator; this complex binds to “xenobiotic response” elements in promoter sequences, a process leading to transcriptional up-regulation of AhR-targeted genes. Interestingly, with TCDD treatment of CW and LAPC-4 CaP cells, we observed the exact opposite response; expression of the P450 enzyme CYP1A1, which is known to be targeted by liganded AhR (52), was strongly abrogated compared with basal CYP1A1 expression. Our working hypothesis had predicted that CYP1A1 expression would increase with TCDD treatment. Because both cell lines exhibited clear nuclear localization of AhR and demonstrated moderately high CYP1A1 expression under normal cell culture conditions, we suggest that these cells bear elevated AhR-mediated transcriptional activity at a basal level. It is possible that this heightened activity in the face of little or no AhR agonist results from expression of certain AhR interaction partners; for example, Antenos et al. (64) have shown that the ubiquitin-like protein Nedd8 augments AhR activity under culture conditions lacking exogenous AhR ligand. The nature of TCDD-associated down-regulation of CYP1A1 protein in our experiments is unknown. It could conceivably be related to putative modulation by TCDD/AhR on other modes of signaling that are known to be able to suppress CYP1A1 expression (e.g., see Refs. 65, 66, 67). Because CYP1A1 is known to be up-regulated in rat prostate by TCDD exposure (68, 69), the apparent inconsistency that our data present may be due to inherent differences in context afforded by CaP cell lines versus normal prostate epithelia in vivo. Regardless of these data, we did not see any overt interaction between the wnt/β-catenin and AhR pathways as it concerns the degree of change in CYP1A1 expression; wnt3a did not rescue CYP1A1 expression, although wnt3a did up-regulate AhR protein (±TCDD). Investigating the effects of wnt/β-catenin signaling on AhR expression and activity represents an important inquiry, because both pathways are implicated in tissue development and oncogenesis (1, 2, 11, 57); furthermore, AhR itself and its target gene, CYP1A1, have been associated with these processes in the prostate (70, 71, 72). Additional studies using more parameters (treatment times, AhR ligand type, AhR target genes analyzed, and additional cell lines) should be of benefit for answering this question as to whether or not wnt/β-catenin activity, by virtue of its stimulatory effects toward AhR protein levels, leads to synergistic augmentation of AhR-mediated transcription.

The mechanism(s) at play in induction of the putative β-catenin-targeted genes (AhR and TMEM2) were not heavily investigated. Our work with the protein synthesis inhibitor cycloheximide may, however, indicate that the nature of the RNA up-regulation of these genes is due, in part, to the effects of β-catenin either directly on gene transcription or on RNA stability. A huge body of evidence already exists linking β-catenin activation to the former mechanism; therefore, we speculate that β-catenin may potentiate the expression of these genes via a DNA-binding factor such as TCF/LEF. Indeed, for AhR induction by mutant β-catenin overexpression, we did observe a dampening of this induction upon transient expression of a dominant-negative form of TCF4 in CW cells, but not in LAPC-4 cells. This experiment implies that the AhR gene may be regulated, in part, by TCF proteins. A scan of the putative regulatory region upstream of the AhR transcription start site (−5000 bp) finds three consensus TCF/LEF binding sites (C/T-C-T-T-T-G-A/T-A/T; Ref. 41) present at positions −1221→ −1214, −1280→ −1273 and −4177→ −4170. Scanning the analogous region of the TMEM2 gene finds two consensus TCF/LEF binding sites (−3204→ −3197 and −3510→ −3503). Future reporter and DNA binding studies should take these sequences into account when honing into that region potentially necessary for β-catenin-mediated AhR up-regulation. It will also be important to consider the impact of other transcription factors beside those of the TCF/LEF family that have been found more recently to collaborate with β-catenin (1, 73, 74).

In conclusion, we have used microarray analysis of a CaP cell line to uncover genes that may be sensitive to wnt/β-catenin signaling in CaP. Multiple modes of wnt pathway stimulation were used to validate the sensitivity of certain genes to such signaling. It is important to note that our use of established CaP cell lines, as opposed to a more in vivo setting, may partially impede interpretation of the present study. Therefore, future work will necessitate focus on β-catenin-associated gene expression in the context of a prostate tumor model, not limited to CaP cell lines or the context of otherwise normal prostate. Additional attention is merited to further understand the mechanisms responsible for target gene up-regulation and to determine the extent to which these target genes are involved in aspects of human CaP for which abnormal β-catenin activity is predicted to occur.

We thank John Isaacs, Ron Rodriguez, Kenneth Kinzler and Bert Vogelstein (all Johns Hopkins University, Baltimore, MD) for providing to us certain reagents used in this study. We also thank Dr. Tony Brown (Cornell Medical School, New York, NY) for advice pertaining to the preparation of conditioned media from mouse L cells.