Overexpression of the Retinoic Acid-responsive Gene Stra6 in Human Cancers and Its Synergistic Induction by Wnt-1 and Retinoic Acid

The relationship of Wnt-1 signaling to cancer was initially realized by the observation of tumor formation in mice aberrantly expressing the Wnt-1 gene in mammary tissue (1). Although Wnt-1 itself remains unproven as a human oncogene, at least three components in the signaling pathway are now known to contribute directly to human cancer (2, 3, 4, 5, 6). The inactivation of one of these genes, the tumor suppressor APC,² is the most common genetic defect in colorectal cancer (7, 8). Inactivation of APC results in the stabilization of β-catenin, an outcome also achieved by stimulation of cells with the secreted ligand Wnt-1 (9, 10). Stabilization of β-catenin also occurs in response to missense mutations and interstitial deletions that affect amino acid sequence in the NH₂-terminal region of the protein (4, 5, 10). These mutations prevent the recognition of β-catenin by an E3 ubiquitin ligase that is required for the targeted degradation of the protein (11, 12, 13). Oncogenic mutations in the β-catenin gene have now been identified in a wide variety of human cancers originating from hepatic, colorectal, endometrial, dermal, renal, and other tissues (14, 15, 16, 17, 18, 19, 20, 21). More recently, mutations in axin, an additional negative regulator of Wnt-1 signaling, were detected in human hepatocellular and in colorectal cancers (6, 22). In both cases, the tumors containing axin mutations were exclusive to those with β-catenin mutations, implying that either defect produces a similar outcome in tumor progression.

A common outcome resulting from mutation of APC, β-catenin, or axin in human cancers is the constitutive activation of gene transcription (23). This is accomplished by the direct binding of β-catenin to members of the HMG family of transcription factors TCF4 and LEF1 (24, 25). The COOH-terminal sequence in β-catenin has been proposed to both directly participate in gene activation and to recruit transcriptional coactivators, such as p300/CBP, to the sites of TCF/LEF DNA binding (26, 27, 28). An underlying assumption is that the genes activated by the interaction of β-catenin with TCF/LEF transcription factors contribute to tumor progression. The nature of some of the target genes identified to date, which include c-myc and cyclin D1, support this (29, 30). The promoter elements of the cyclin D1 and c-myc genes contain canonical binding sites for TCF/LEF transcription factors, which when mutated abrogate their activation by the Wnt-1 signaling pathway. Activation of other Wnt-1 target genes, however, might require signals in addition to those supplied by the β-catenin/TCF complex. For example, the activation of the WISP-1 gene by Wnt-1 is dramatically reduced by mutation of a CREB binding site residing near a TCF/LEF site in the WISP-1 promoter (31, 32). This suggests that a higher level of order exists for the activation of some Wnt-1 target genes and that additional signals cooperate with Wnt-1 for their activation. For example, overexpression or inhibition of the CREB-binding protein p300/CBP affects induction of the Xenopus gene Siamois by β-catenin, but does not influence the activation of the cyclin D1 gene by β-catenin (27).

The potential for Wnt-1 signaling to cooperate with other signaling pathways was also revealed in a recent study demonstrating cross-talk between RAR signaling and the Wnt-1 pathway (33). Treatment of cells with retinoids inhibited the activation of a synthetic reporter gene responsive to Wnt-1 signaling; conversely, Wnt-1 signaling potentiated the activation of an RAR-responsive test gene by retinoids. Thus, signaling by RAR might impact negatively on Wnt-1 signaling or Wnt-1 might cooperate with retinoids to synergistically activate RAR-responsive genes. In the latter case, genes could conceivably undergo activation by Wnt-1 signaling yet not contain binding sites for TCF/LEF in their promoter elements. This could result from the direct interaction of Wnt-1 and RAR signaling molecules. Indeed, the β-catenin and RAR-α proteins were reported to physically bind to each other in vitro (33). Alternatively, Wnt-1 signaling could activate the expression of RARs, which would in turn heighten the responsiveness of those cells to retinoids. RAR-γ has been proposed as a target of Wnt signaling in Xenopus embryos (34).

We performed a screen to identify genes induced by Wnt-1 stimulation of C57MG mouse mammary epithelial cells, a cell line known to undergo morphological transformation in response to Wnt-1. Unexpectedly, one of the mRNA transcripts highly induced by Wnt-1 was identified as Stra6, a gene activated by the stimulation of mammalian cells with retinoids (35). We found that Wnt-1 and retinoids acted synergistically to induce Stra6 expression and that Stra6 mRNA was highly overexpressed in the majority of human colorectal cancers that were examined.

GeneCalling was performed essentially as described (36). Briefly, each sample set consisted of three individual samples. Double-stranded cDNA was synthesized from the mRNA in each sample. These cDNA pools were digested by pairs of restriction enzymes with 6-bp recognition sites. The reactions were performed in triplicate for each sample in the sample sets. Ninety-six restriction enzyme pairs were chosen such that representative coverage of most of the possible sequences in a given cDNA sample was achieved. The digestions and PCR amplifications using specific linkers were carried out as described previously (36). The PCR products were denatured, and the single-stranded, 6-carboxyfluorescein-labeled fragments were size resolved by capillary electrophoresis (MegaBACE; Molecular Dynamics) rather than by the thin-slab gel Niagara instrument described previously. The data were queried against public and proprietary databases to generate GeneCalls. The identities of the GeneCalled fragments were confirmed via platform-consistent competitive PCR reactions (36). A partial sequence for mouse Stra6 was detected in a differentially expressed transcript, and clones containing the full-length sequence for human Stra6 were isolated by inverse PCR. Briefly, human Stra6 was cloned using the following primers for inverse PCR: 100038.sn.r1 (GTCCTGTGTGTCACCAGTGAGCGC) and 100038.sn.f1 (CAACCTTCGAGCTCTGCACCGAG). The inverse PCR amplicons were ligated using T4 DNA ligase. An oligonucleotide (100038.p1; CACCTTTCTTCTCTTCCCCCTCAATGTGCTGGT-jGGGTGCCATGGT) was radiolabeled and used as a probe to identify the human Stra6 clones by standard colony lift procedures.

C57MG and C57MG/Wnt-1 cells were grown in DMEM supplemented with 10% FBS, 2 mm l-glutamine, and 2.5 μg/ml puromycin (Edge Biosystems). C57MG cells with tetracycline-repressible Wnt-1 expression were grown in complete medium without puromycin, supplemented with 400 μg/ml G418 (Life Technologies, Inc.), 100 μg/ml hygromycin B (Life Technologies), and 50 ng/ml tetracycline (37). For Wnt-1 induction studies, cells were washed with PBS, cultured in tetracycline-free medium for 10, 24, 48, 72, and 96 h, and then harvested. A 0-h control dish was maintained entirely in medium containing tetracycline. All dishes were harvested simultaneously, and total RNA was extracted for each time point. RT-PCR was carried out with gene-specific primers and probes on 100 ng of total RNA from each sample.

The human colon adenocarcinoma cell lines HCT116 and WiDr were obtained from the American Type Culture Collection (Rockville, MD. HCT116 cells were maintained in McCoy’s 5A medium supplemented with 10% FBS. WiDr cells were maintained in DMEM supplemented with 10% FBS. For RA induction studies, cells were plated at 10⁵ cells/60-mm dish containing 2.5 ml of medium and allowed to grow for 24 h. Cells were treated with vitamin D₃, ATRA (Spectrum Laboratory Products), or 9-cis-RA (Toronto Research Chemicals Inc.; 1 μm final concentration in DMSO unless otherwise specified) for the indicated times. Control cells were treated with an equal volume of DMSO.

For the treatment of C57MG parent cells with Wnt-3A-conditioned medium, cells were incubated in regular medium or conditioned medium from L-cells or L/Wnt-3A cells in the presence or absence of 1 μm 9-cis-RA or ATRA for the indicated times. Conditioned medium was prepared as described previously (38).

Total RNA from cells in culture was isolated using Qiagen RNeasy minicolumns. Total RNA from tissue samples was isolated using RNA Stat-60 (Tel-Test). All samples were processed according to the manufacturers’ instructions.

Quantitative RT-PCR was performed using TaqMan assay reagents (Perkin-Elmer, Applied Biosystems). RT-PCR reactions (total volume, 50 μl) contained 5 μl of 10× TaqMan Buffer A, 300 μm each dNTP, 5 mm MgCl₂, 10 units of RNase inhibitor, 12.5 units of MuLV Reverse Transcriptase, 1.25 units of AmpliTaq Gold DNA Polymerase, 200 nm probe, 500 nm primers, and 100 ng of RNA. Reaction conditions consisted of reverse transcription at 48°C for 30 min, denaturation at 95°C for 10 min, and 40 thermal cycles of 95°C for 25 s and 65°C for 1 min. Reaction products were analyzed on 4–20% polyacrylamide gels (Novex).

Standard curves were used to determine relative levels of expression for each gene of interest as well as the GAPDH housekeeping gene for each sample analyzed. Relative normalized units were obtained by dividing the mRNA level for the gene of interest by the GAPDH mRNA level. Relative normalized units were compared between experimental samples and controls to determine fold of induction. Mouse Wnt-1, mouse and human Stra6, mouse and human GAPDH, mouse TCF-1, and mouse COX-2 primer sequences used for quantitative RT-PCR are as follows:

Mouse Wnt-1: forward primer, 5′-TCGAGAAATCGCCCAACTTC-3′; reverse primer, 5′-TGCAAGCTCGTCCAGCTGT-3′; probe, 5′-CCTCGGCCACAGCACAGCAG-3′

Mouse Stra6: forward primer, 5′-AGCCAAGTCAGACTCCAAGAG-3′; reverse primer, 5′-CAGAGAGCACACTAACTTCTTTCA-3′; probe, 5′-CCCCACTGAGCTGCCCTCTCC-3′

Human Stra6: forward primer, 5′-AGACCAGGTCCCACACTGA-3′; reverse primer, 5′-TTCATAATAGCCAAAGGCATAAAA-3′; probe, 5′-CTGCCCACACTCGAGAGCCAGAT-3′

Mouse GAPDH: forward primer, 5′-TGCACCACCAACTGCTTAG-3′; reverse primer, 5′-GGATGCAGGGATGATGTTC-3′; probe, 5′-CAGAAGACTGTGGATGGCCCCTC-3′

Human GAPDH: forward primer, 5′-GAAGATGGTGATGGGATTTC-3′; reverse primer, 5′-GAAGGTGAAGGTCGGAGTC-3′; probe, 5′-CAAGCTTCCCGTTCTCAGCC-3′

Mouse TCF-1: forward primer, 5′-GGTCTTCTATACAGCCCTGGTT-3′; reverse primer, 5′-CCAGCACTTGCGAGATAGAG-3′; probe, 5′-TCTTGGAATTCACAGAGGTCCCCC-3′

Mouse COX-2: forward primer, 5′-AAGTTAGAACTTAGGCTGTTGGAAT-3′; reverse primer, 5′-ACATGCTTGGGTCAGTCAATA-3′; probe, 5′-TGGATCCTATGCAGTCTGCTTTATGCG-3′

Northern blotting was performed using the Northern Max kit (Ambion). For each sample, 10 μg of total RNA were subjected to formaldehyde-agarose gel electrophoresis and transferred to BrightStar-Plus nylon membranes (Ambion). RNA from C57MG cells was hybridized with ³²P-labeled PCR product corresponding to nucleotides 2863–2973 of the mouse Stra6 cDNA. RNA from HCT116 and WiDr cells was hybridized with ³²P-labeled PCR product corresponding to nucleotides 2767–2855 of the human Stra6 cDNA. Membranes were stripped by boiling in 0.1% SDS and rehybridized with a ³²P-labeled β-actin probe.

³³P-labeled sense and antisense riboprobes were transcribed from an 874-bp PCR product corresponding to nucleotides 432-1247 of the human Stra6 coding sequence. Formalin-fixed, paraffin-embedded tissue sections were processed as described previously (31). ATRA- and vehicle-treated HCT116 cell pellets were processed concurrently as controls.

Cells were lysed in Triton X-100 lysis buffer [20 mm Tris-HCl (pH 8.0), 137 mm NaCl, 1% Triton X-100, 1 mm EGTA, 10% glycerol, 1.5 mm MgCl₂, 1 mm DTT, 1 mm sodium vanadate, 50 mm sodium fluoride, and complete protease inhibitor cocktail (Boehringer Mannheim)], and protein-equivalents were subjected to SDS-PAGE and immunoblotting. Blots were incubated with 0.2 μg/ml affinity-purified rabbit polyclonal antibody against β-catenin (39), 0.1 μg/ml anti-ERK2 mAb (Transduction Laboratories), 1 μg/ml anti-smooth muscle actin mAb (NeoMarkers), or rabbit polyclonal antisera against RARγ-1 (1:2000 dilution; Affinity Bioreagents). For the WiDr cell line, cells were treated with 1 μm ATRA for 48 h and then lysed in Triton X-100 lysis buffer and processed as indicated above. Blots were incubated with anti-Stra6 peptide B mAb (mAb 2880; 1:50 dilution). Blots were developed using the ECL system (Amersham).

Two extracellular domains of the human Stra6 protein—peptide A (amino acids 229–295) and peptide B (amino acids 532–667)—were expressed separately as peptides in the E. coli cytoplasm with an NH₂-terminal polyhistidine leader having the amino acid sequence MKHQHQHQHQHQHQMHQ. This leader provides for optimal translation initiation, purification on a nickel chelation column, and efficient removal if desired with the TAGZyme system (Unizyme Laboratories). Transcription is controlled by the E. coli alkaline phosphatase promoter (40), and the trp operon ribosome binding site (41) provides for translation. Downstream of the translation termination codon, is the λt(o) transcriptional terminator (42) followed by the rare codon tRNA genes: pro2, argU, and glyT (43, 44).

The two Stra6 extracellular domain-coding sequence DNA fragments were prepared by PCR from a full-length cDNA clone and inserted into the expression vector pST239. After DNA sequence verification, the new Stra6 expression plasmids pE148380A and pE148380B were transformed into the E. coli strain 58F3 [fhuAΔ(tonAΔ) lonΔ galE rpoHts (htpRts) ΔclpP lacIq ΔompTΔ (nmpc-fepE) ΔslyD]. Luria broth cultures of these transformants were first grown overnight at 30°C, and then diluted 100-fold in a phosphate-limiting medium to induce the alkaline phosphatase promoter. After 24 h at 30°C with shaking, the cultures were centrifuged, and the cell pastes were frozen until the start of peptide purification.

E. coli pastes (6–10 g pellets) were resuspended in 10 volumes (w/v) of 20 mm Tris (pH 8) containing 7 m guanidine HCl. Solid sodium sulfite and sodium tetrathionate were added to make final concentrations of 0.1 and 0.02 m, respectively, and the solution was stirred overnight at 4°C. The solution was clarified by centrifugation and loaded onto a Qiagen Ni-NTA metal chelate column equilibrated in 20 mm Tris (pH 7.4) containing 6 m guanidine HCl. The column was washed with additional buffer containing 50 mm imidazole (Ultrol grade; Calbiochem). The protein was eluted with buffer containing 250 mm imidazole. Fractions containing the desired protein were pooled, dialyzed against 1 mm HCl, and stored at 4°C.

Five BALB/c mice (Charles River Laboratories, Wilmington, DE) were hyperimmunized with purified Unizyme-conjugated amino acid peptide, corresponding to amino acids 532–667 of human Stra6, in Ribi adjuvant (Ribi Immunochem Research, Inc., Hamilton, MO). B-cells from popliteal lymph nodes were fused with mouse myeloma cells (X63.Ag8.653; American Type Culture Collection) as described previously (45). After 10–14 days, supernatants were harvested and screened for antibody production by direct ELISA. Eight positive clones, showing the highest immunobinding by direct ELISA and immunohistochemistry after two rounds of subcloning by limiting dilution, were injected into pristane-primed mice for in vivo production of mAb (46). The ascites fluids were pooled and purified by Protein A affinity chromatography (Pharmacia fast protein liquid chromatography; Pharmacia, Uppsala, Sweden) as described previously (45). The purified antibody preparations were sterile filtered (0.2 μm pore size; Nalgene, Rochester, NY) and stored at 4°C in PBS.

WiDr cells were treated with 9-cis-RA or DMSO, and then detached and pelleted by low-speed centrifugation. Cell pellets were fixed overnight in 10% neutral-buffered formalin, dehydrated, and embedded in paraffin. Immunohistochemistry was performed using anti-Stra6 peptide B mAb (mAb 2880) or nonspecific mouse isotype IgG2A as primary antibodies, followed by detection using avidin-biotin complex method with diaminobenzidine as chromogen (Vectastain Elite Kit; Vector Laboratories) as described previously (47). Sections were counterstained with hematoxylin.

Ro 61-8431 (kindly provided by Dr. L. Foley, Hoffmann-La-Roche, Nutley, NJ) was dissolved in absolute ethanol at a concentration of 10 mm, and then diluted in culture medium to the required final concentration. Retinoid stock solutions were kept at −80°C.

Overexpression of Wnt-1 in the C57MG mouse mammary epithelial cell line is sufficient to induce a partially transformed phenotype, characterized by cells that exhibit an elongated, refractile morphology and loss of contact inhibition (48, 49). The morphological differences that are observed between the parent and the C57MG/Wnt-1 cells suggested that these cell lines would be useful for the analysis of mRNA transcript changes that contribute to the transformed phenotype. A novel transcript-profiling method, GeneCalling, was used to identify genes that were up-regulated in the C57MG/Wnt-1 cell line compared with the parent C57MG cells (36). This procedure involves the fragmentation of cDNA generated from cellular mRNA followed by fluorescent labeling and high-resolution separation of the fragments by capillary electrophoresis. The relative intensities of specific cDNA fragments derived from treated and untreated cells were compared with those in the GeneCalling database to analyze differentially expressed mRNA transcripts. The identities of fragments that showed differential modulation were confirmed by a competitive PCR reaction with gene-specific primers or by cloning and sequencing the novel fragments. Quantitative real-time RT-PCR and Western blot analysis also confirmed the up-regulation of mRNA and protein expression, respectively.

A portion of a set of traces showing a number of fluorescently labeled cDNA fragments resolved according to size is presented in Fig. 1,A. The traces from the two cell lines illustrate that most of the transcripts represented by these fragments are expressed at approximately equal levels. However, a cDNA fragment that migrates at position 241.8 was present at ∼11-fold higher intensity when derived from the C57MG/Wnt-1 cell line compared with the C57MG parent cells. Competitive PCR revealed that the differentially expressed transcript was identical to that transcribed by the mouse Stra6 gene. To confirm the change in expression levels, RNA was prepared from the two cell lines, and quantitative RT-PCR using Stra6-specific primers was performed. Gel analysis of the PCR products and Northern blot analysis confirmed that the Stra6 mRNA was up-regulated in the Wnt-1-expressing cell line relative to the parental cells (Fig. 1, B and C).

A C57MG cell line that conditionally expresses Wnt-1 was used to determine the kinetics of Stra6 induction by Wnt-1. The C57MG cells express Wnt-1 under control of a tetracycline-repressible promoter and produce low amounts of Wnt-1 in the repressed state, but show a strong induction of Wnt-1 after removal of tetracycline (37). The levels of Wnt-1, Stra6, and GAPDH mRNA isolated from these cells at various times after tetracycline removal were assessed by quantitative RT-PCR. The level of β-catenin protein was also measured by Western blot analysis. Strong induction of both Wnt-1 and Stra6 mRNA was seen as early as 10 h after tetracycline removal, with a maximum level of expression for both observed at 72 h (Fig. 2,A). In contrast, expression of the housekeeping gene GAPDH did not change during the course of the experiment. The levels of β-catenin protein were elevated after the induction of Wnt-1, which is consistent with the stabilization of this protein by Wnt-1 signaling (Fig. 2 B). These data support our initial observation that the expression of Stra6 correlates with that of Wnt-1. Moreover, the acute induction of Stra6 mRNA after Wnt-1 expression suggests that it is an early response to Wnt-1 signaling.

Transgenic mice that express the Wnt-1 proto-oncogene in the mammary gland typically exhibit hyperplastic lesions and develop neoplasms in this tissue (50). To determine whether Stra6 was overexpressed under these conditions, quantitative RT-PCR was performed on RNA extracted from hyperplastic mammary glands and mammary gland tumors from Wnt-1 transgenic mice. Strong expression of both Wnt-1 and Stra6 was seen in all hyperplastic mammary glands (Fig. 3,A) and mammary gland tumors (Fig. 3 B) compared with its expression in normal mammary glands from age-matched nontransgenic mice.

cDNA clones of human Stra6 were isolated, and the encoded amino acid sequence was 74% identical to mouse Stra6 (data not shown). Two variants of a human Stra6 cDNA were isolated. One clone was similar to the reported mouse sequence, whereas a second clone lacked nine amino acids encompassed by residues 89–97 relative to the first sequence. Additional polymorphisms affecting amino acid positions 518 and 527 were also noted when the sequences were compared. The Stra6 protein contains 9 potential transmembrane domains and 14 cysteine residues. The human Stra6 gene was localized to chromosome 15q23 as determined by UNIGENE. Preliminary fine mapping indicated that Stra6 is located in the sequence-tagged site interval D15S124–D15S160, and the GeneMap’98 position corresponded to 244.52 on the G3 map.

The majority of colorectal tumors contain mutations in either APC or β-catenin (51). These mutations result in the stabilization of β-catenin, a hallmark of Wnt-1 signaling, which leads to activation of specific genes. Therefore, it was of interest to determine whether Stra6 was overexpressed in human colorectal tumors relative to normal colorectal mucosa. The level of Stra6 transcript present in equivalent amounts of RNA isolated from 14 adenocarcinomas and their matched normal mucosa was assessed by quantitative RT-PCR. Stra6 mRNA was significantly increased (2- to >100-fold) in all of the human colon tumors examined compared with their corresponding normal tissues (Fig. 4,A). The cycle threshold values obtained by RT-PCR indicated that Stra6 mRNA levels were extremely low or possibly absent in many of the normal mucosa samples. As a second method of detection, the products obtained after completion of the RT-PCR reactions (40 cycles each) were subjected to electrophoresis in polyacrylamide gels and visualized by ethidium bromide staining (Fig. 4 B). In accordance with the quantitative data, substantially greater amounts of PCR product were generated from Stra6 mRNA in tumor samples compared with their normal counterparts. In contrast, a comparable level of product from the internal control GAPDH mRNA was generated from all samples. Stra6 mRNA expression in colon adenocarcinomas was localized to the epithelial tumor cells by ISH (data not shown).

Tissue-specific expression of human Stra6 was characterized by RT-PCR analysis of adult multiple tissue total RNA panels. High Stra6 expression was seen in the adult kidney, trachea, breast, prostate, testis, and uterus, and moderate expression was detected in the lung (Fig. 5). Little or no expression was detected in the brain, heart, liver, bone marrow, colon, spleen, stomach, thymus, small intestine, and skeletal muscle. ISH confirmed Stra6 expression in kidney tubular epithelial cells, myometrium, and stromal cells surrounding breast ducts and lobules, whereas little or no expression was detected in sections of brain, liver, spleen, pancreas, heart, lung, stomach, small intestine, colon, prostate, and adrenal cortex (data not shown). Of the normal tissues examined by ISH, the highest expression levels were seen in the placenta and adrenal medulla, which were not included in the RT-PCR analyses. Several tumor types other than colon adenocarcinomas, which frequently harbor alterations leading to abnormal activation of β-catenin signaling, also showed high levels of Stra6 expression. These included three of three melanomas (Fig. 6, A and B), three of four endometrial adenocarcinomas (Fig. 6, C and D), two of three ovarian adenocarcinomas, and a Wilm’s tumor of the kidney (Fig. 6, E and F). The Stra6 ISH signal in these various tumors was considerably greater than in colon adenocarcinomas, which is consistent with our data showing relatively high expression levels in normal kidney and uterus and low levels in normal colon. Because Stra6 was detected in normal adrenal medulla, we therefore examined two pheochromocytomas, which are tumors derived from this tissue. In these tumors, Stra6 expression exceeded that of any other tumor or tissue examined in this study (Fig. 6, G and H). Although Stra6 was detected in normal kidney and was strongly expressed in Wilm’s tumor, it was not detected in renal cell carcinomas. In kidney transitional cell carcinomas, tumor-associated stromal cells rather than tumor epithelial cells expressed Stra6 (data not shown).

Easwaran et al. (33) previously reported enhanced activation of a synthetic RA-responsive reporter gene when MCF-7 cells were cotransfected with mutant β-catenin and treated with retinoids. Considering this, together with the original identification of Stra6 as a RA-inducible gene (35), we asked whether RA could synergize with Wnt-1 to increase the level of Stra6 in the C57MG cell line. Treatment of parental C57MG cells with either ATRA or 9-cis-RA for 48 h significantly increased the level of Stra6 mRNA, whereas DMSO and vitamin D₃ had no effect (Fig. 7 A). As expected, the C57MG/Wnt-1 cells treated with either DMSO or vitamin D₃ exhibited enhanced levels of Stra6 mRNA relative to the parent cell line. The level of Stra6 induction by Wnt-1 was comparable to that observed on stimulation of the parental C57MG cells with RA. However, treatment of the C57MG/Wnt-1 cell line with either ATRA or 9-cis-RA (1 μm) induced a further 10-fold increase in Stra6 mRNA relative to either untreated C57MG/Wnt-1 or RA-treated C57MG parent cells. C57MG/Wnt-1 cells treated with doses as low as 10 nm also showed significant synergistic induction of Stra6 (data not shown).

It was possible that potential clonal variations in the C57MG/Wnt-1 cell line that were unrelated to Wnt-1 expression accounted for their differential response to RA relative to parental control cells. To address this, we tested the response of the parental C57MG cells to stimulation by soluble Wnt-3A in the presence or absence of 9-cis-RA. Wnt-3A is a Wnt-1 homologue that exhibits transforming properties similar to those of Wnt-1 and can be produced as a soluble ligand in mouse l cells (52). Conditioned medium from cultured l cells expressing Wnt-3A, but not from control l cells, slightly induced the expression of Stra6 in the C57MG cells (Fig. 7,B). A higher level of induction was observed on treatment of C57MG cells with 9-cis-RA. However, the combination of 9-cis-RA and Wnt-3A resulted in levels of Stra6 transcript greatly exceeding that seen with either agent alone. This effect was also confirmed by Northern blot analysis of RNA extracted from C57MG cells in which Wnt-1 was conditionally expressed for 48 h by removal of tetracycline. The intensity of the band corresponding to the Stra6 transcript was dramatically increased by treatment of Wnt-1-expressing cells with ATRA (Fig. 7 C).

If the induction of Stra6 in the RA-treated C57MG/Wnt-1 cells was potentiated by increased β-catenin levels, then one might expect a similar induction of Stra6 in response to RA in human colon carcinoma cells containing mutations in either β-catenin or APC. To determine whether this occurs, Stra6 mRNA levels were measured before and after RA treatment in HCT116 cells, which carry an activating mutation in β-catenin, and in WiDr cells, which have lost both copies of wild-type APC. In both cell lines, a significant increase in Stra6 mRNA levels was seen after treatment with either ATRA or 9-cis-RA compared with DMSO or vitamin D₃ (Fig. 8,A). Activation of the Stra6 gene by ATRA in these cell lines was confirmed by Northern blot analysis (Fig. 8,B). Induction of Stra6 expression in the HCT116 cell line was also demonstrated by ISH of cells pelleted by centrifugation (data not shown). Induction of the predicted 73-kDa Stra6 protein band in WiDr cells treated with ATRA was also detected by Western blot analysis with a Stra6-specific mAb (Fig. 8,C). Immunohistochemical analysis of the WiDr cells revealed that the increase in Stra6 protein in response to ATRA was localized to the plasma membrane (Fig. 8 D).

Our data show that Wnt-1 and Wnt-3A are capable of activating Stra6 in the absence of added retinoids. This implies that the canonical Wnt pathway involving the interaction of β-catenin with TCF/LEF transcription factors might directly regulate activity of the Stra6 gene. However, it is also possible that basal RAR signaling, which occurs in the absence of added retinoids, is still required to support this activation. To examine this, we treated C57MG cells with a potent RAR pan-antagonist, Ro 61-8431. In parental C57MG cells treated with ATRA, Ro 61-8431 was effective in abrogating the retinoid induction of Stra6 (Fig. 9,A). After treatment of C57MG/Wnt-1 cells with Ro 61-8431 in the absence of ATRA, Wnt-1 induction of Stra6 was also inhibited, suggesting an absolute requirement for RAR activity in the induction of Stra6 by Wnt. In contrast, the RAR antagonist had no effect on the activity of the canonical Wnt pathway, as demonstrated by the induction of TCF-1 and COX-2, two genes shown previously to be induced by Wnt-1 (Fig. 9 A; Refs. 53 and 54).

The RARγ gene has been proposed as a target for Wnt signaling in Xenopus embryos, and the induction of Stra6 by retinoids was shown to be dependent on the presence of this RAR subtype (34, 55). Together, these observations suggest that the synergistic activation of Stra6 by Wnt and retinoids might be caused by the activation of RARγ expression by Wnt signaling. To determine whether Wnt signaling had any influence over the levels of RARγ in mammalian cells, we performed Western blots for the receptor on lysates prepared from C57MG cells treated with conditioned medium containing Wnt-3A. RARγ was detected in cells treated with control medium, and its expression gradually increased over time in the presence of Wnt-3A media (Fig. 9,B). The presence of RA alone did not appreciably affect this result. Although these data demonstrate that Wnt signaling promoted the up-regulation of RARγ, the kinetics of Stra6 induction by ATRA plus Wnt-3A suggest that additional mechanisms might be responsible for the synergy (Fig. 9 C). Stra6 mRNA was detectable as early as 6 h after the addition of soluble Wnt-3A plus ATRA, whereas RARγ in these same samples was not significantly increased until 24 h posttreatment. The data indicate that Wnt and RA signaling cooperate in the acute induction of Stra6 mRNA expression, which might be further supported by a subsequent increase in the levels of RARγ.

Gene expression profiling approaches are based on unbiased detection of mRNA transcripts and can therefore lead to unexpected insights into the mechanisms by which gene activation occurs. Here we have shown that Wnt-1 promoted the induction of the RA-responsive gene Stra6, suggesting a connection between signaling pathways elicited by Wnt and RA. This connection was further supported by demonstration of the synergistic induction of Stra6 by a combination of Wnt and RA. The cross-regulation of these signaling pathways was observed when endogenous levels of the receptors were stimulated by natural ligands and did not require the ectopic overexpression of any intracellular signaling components. For this reason we believe that the interaction of the Wnt and RA signaling pathways is likely to occur in genuine biological settings such as development and disease.

There are several possible mechanisms by which Wnt signaling could influence the activation of RA-responsive genes. Easwaran et al. (33) demonstrated enhanced activation of an RAR response element reporter gene by expression of β-catenin in cells treated with RA. The RAR response element used in their study did not contain a consensus sequence for TCF binding, which suggests that TCF/LEF transcription factors might not be involved in this mode of Wnt signaling. We found that the activation of Stra6 by Wnt-1 was dependent on RAR activity, whereas activation of TCF-1 and COX-2 by Wnt-1 was independent of RAR activity. Thus, the activation of Stra6 by Wnt signaling appears to use a mechanism distinct from that which drives the activation of genes by the canonical TCF-dependent signaling. The strict dependence of Stra6 induction by Wnt on RAR signaling might also explain the lack of a good correlation between Wnt-1 levels and Stra6 levels in the mouse mammary tumors (Fig. 3). The level of RARs and the metabolism of retinoids themselves likely influence the magnitude with which Wnt-1 would activate Stra6 in these tissues.

We have also shown that Wnt promotes the up-regulation of RARγ, which could potentiate the response of the cell to retinoids. That RARγ is the appropriate receptor for the activation of the Stra6 gene is consistent with a previous study in which the specific disruption of the RARγ gene greatly reduced activation of the Stra6 gene by retinoids (55). The defect was then rescued on reexpression of RARγ (55). Although our data show that Wnt signaling resulted in increased levels of RARγ, the kinetics of the synergistic induction of Stra6 by Wnt and RA suggest that these signaling pathways might also converge at a step that precedes the increase in RARγ. The demonstration that β-catenin binds to RARs offers an attractive explanation for the acute effects of Wnt and retinoids on the induction of Stra6 mRNA (33). The synergistic induction of a RA-responsive gene by Wnt-1 signaling implies that human cancers that harbor genetic defects in the Wnt-1 pathway would exhibit overexpression of Stra6. The vast majority of colorectal tumors contain mutations in the genes coding for either the APC tumor suppressor or β-catenin (51), and accordingly, we have detected overexpression of the Stra6 transcript in 14 of 14 colorectal tumors relative to matched normal tissue. Activating mutations in β-catenin have also been identified in cancers of the ovary and endometrium, Wilm’s kidney tumors, and melanomas, demonstrating that defects in Wnt-1 signaling contribute to the progression of these cancers (5, 18, 19, 21, 56, 57). All four of these human cancers displayed overexpression of Stra6 mRNA as determined by ISH. Pheochromocytoma, a tumor derived from the adrenal medulla, exhibited extremely high levels of Stra6 mRNA; however, the status of these tumors with respect to mutations in the Wnt-1 signaling pathway has not been reported.

The function of Stra6 is not known, making it difficult to speculate on a potential role for its overexpression in cancer. Administration of retinoids in various cancer models has been inconsistent, yielding both suppression and enhancement of tumor progression depending on tumor type, genetic background, and administration protocol (reviewed in Ref. 58). Therefore, expression of Stra6 in cancer could play a positive or negative role in the progression of disease. Irrespective of the function of Stra6, it is clear from our data that the activation of the Wnt signaling pathway can have a profound effect on the expression of genes regulated by RA. Accordingly, the activation of certain genes responsive to retinoids might be greatly increased in cancers that are driven by defects in the Wnt-1 pathway relative to their responses in wild-type tissues. RA-responsive genes that are preferentially activated in such an oncogenic background could, therefore, be exploited to improve the therapeutic index of drugs in cancer therapy.

We thank A. Levine for the C57MG and C57MG/Wnt-1 cell lines; K, Willert and R. Nusse for the tet-repressible C57MG/Wnt-1 cells and the l-W3A cells, and the CuraGen Genomics Facility for generation of the GeneCalling data. We also thank D. Lewin and F. Mehraban from CuraGen’s Department of Gene Discovery for development and implementation of analysis strategy, P. Dowd for radiation hybrid mapping, J. Brush and S. Johnson for human Stra6 cDNA sequencing, D. Vandlen for Stra6 peptide purification, B. Wright for performing immunohistochemistry, L Foley for providing the retinoid antagonist Ro 61-8431, and V. Dixit for discussions.