Restoration of Wnt-7a expression reverses non-small cell lung cancer cellular transformation through frizzled-9-mediated growth inhibition and promotion of cell differentiation.

The Wnt signaling pathway is critical in normal development, and mutation of specific components is frequently observed in carcinomas of diverse origins. However, the potential involvement of this pathway in lung tumorigenesis has not been established. In this study, analysis of multiple Wnt mRNAs in non-small cell lung cancer (NSCLC) cell lines and primary lung tumors revealed markedly decreased Wnt-7a expression compared with normal short-term bronchial epithelial cell lines and normal uninvolved lung tissue. Wnt-7a transfection in NSCLC cell lines reversed cellular transformation, decreased anchorage-independent growth, and induced epithelial differentiation as demonstrated by soft agar and three-dimensional cell culture assays in a subset of the NSCLC cell lines. The action of Wnt-7a correlated with expression of the specific Wnt receptor Frizzled-9 (Fzd-9), and transfection of Fzd-9 into a Wnt-7a-insensitive NSCLC cell line established Wnt-7a sensitivity. Moreover, Wnt-7a was present in Fzd-9 immunoprecipitates, indicating a direct interaction of Wnt-7a and Fzd-9. In NSCLC cells, Wnt-7a and Fzd-9 induced both cadherin and Sprouty-4 expression and stimulated the JNK pathway, but not beta-catenin/T cell factor activity. In addition, transfection of gain-of-function JNK strongly inhibited anchorage-independent growth. Thus, this study demonstrates that Wnt-7a and Fzd-9 signaling through activation of the JNK pathway induces cadherin proteins and the receptor tyrosine kinase inhibitor Sprouty-4 and represents a novel tumor suppressor pathway in lung cancer that is required for maintenance of epithelial differentiation and inhibition of transformed cell growth in a subset of human NSCLCs.

Deregulation of developmental signaling pathways is a common theme in human cancers. In this regard, the Wnt family encodes 19 distinct proteins that serve as extracellular signaling molecules controlling diverse morphogenetic and develop-mental programs (1). Signaling proceeds in an autocrine and paracrine fashion and is mediated by a family of 10 distinct seven-membrane receptors known as Frizzled (Fzd) 1 (2). A well defined cell signaling pathway has been established whereby Wnt binding to Fzd ultimately results in the inhibition of glycogen synthase kinase-3␤ present in a complex with adenomatous polyposis coli (APC) and Axin. In the absence of Wnt signaling, ␤-catenin is phosphorylated by glycogen synthase kinase-3␤, targeting it for ubiquitylation and 26 S proteasomemediated degradation. Suppression of glycogen synthase kinase-3␤ following Wnt/Fzd binding allows ␤-catenin to accumulate and function in the nucleus as a transcriptional coactivator with the T cell factor (TCF)/lymphoid enhancer factor transcription factors (3).
The oncogenic potential of the wnt genes was first appreciated upon the identification of the murine wnt-1 gene as the site of mouse mammary tumor virus integration, leading to deregulated Wnt-1 expression and mammary tumorigenesis (4). In human colon cancers, Wnt pathway components, including APC and ␤-catenin, are frequently mutated, leading to tumorigenesis. At least 85% of human familial colorectal cancers have loss-of-function mutations in APC or gain-of-function mutations in ␤-catenin (5). By contrast, mutations in APC or ␤-catenin are rare in human breast carcinomas (6). However, if stabilization and nuclear accumulation of ␤-catenin are used as indicators, then the Wnt pathway is likely to be activated in a majority of breast cancers as well (6). The Wnt pathway has also been implicated in many other tumor types, including hepatocellular carcinoma, uterine carcinoma, and melanoma (7)(8)(9). Thus, increased activity of the Wnt signaling pathway appears to be a major oncogenic input in diverse human cancers.
Non-small cell lung cancer (NSCLC) cells exhibit frequent loss of function in several defined tumor suppressor genes, including p53 (ϳ50%), Rb (retinoblastoma protein; 15-30%), and p16 INK4 (30 -70%) (10). In addition, variable deletion of putative and ill defined tumor suppressor genes residing on chromosome 3p occurs frequently in NSCLC cells (10). Although evidence indicates that the Wnt pathway is central to development of the lung (11)(12)(13)(14), the role of the Wnt pathway in lung carcinogenesis is ill defined. Similar to breast cancer, mutations in APC and ␤-catenin are rare in lung cancers (15,16), yet the possible dysregulation of specific Wnt proteins in lung cancer cells leading to oncogenic signaling has not been examined. In this study, we demonstrate that re-expression of Wnt-7a and signaling through Fzd-9 are associated with increased differentiation and additionally represent a novel tumor suppressor pathway in NSCLC cells.

EXPERIMENTAL PROCEDURES
Cell Culture and Retrovirus-mediated Gene Transfer-NSCLC cells of the adenocarcinoma (A549, H2122), large cell (H1334, H460, H661), and squamous (H226, H157) phenotypes as well as a mesothelioma (H28) cell line were cultured in RPMI 1640 medium supplemented with 10% fetal bovine serum at 37°C in a humidified 5% CO 2 incubator. Normal human short-term bronchial epithelial (STBE) cultures were obtained from Clonetics Corp. (San Diego, CA) and cultured according to the manufacturer's protocol.
wnt-3 and wnt-7a cDNAs inserted into pcDNA3 encoding a C-terminal hemagglutinin (HA) epitope were kindly provided by Dr. Jan Kitajewski (Columbia University). The wnt cDNAs were excised from pcDNA3 with HindIII and NotI and ligated between the HindIII and NotI sites of the retroviral expression vector pLNCX2 (Clontech). A murine fzd-9 cDNA was excised from pCS2 ϩ with HindIII and NotI and ligated between the HindIII and NotI sites of the retroviral expression vector pLPCX (Clontech). The resulting vectors, LNCX2-HA-Wnt and LPCX-mFzd-9, were packaged into replication-defective retrovirus using 293T cells and the retrovirus component expression plasmids SV-⌿ Ϫ -A-MLV and SV-⌿ Ϫ -env Ϫ -MLV as described (17)(18)(19). The secreted retroviruses were collected and incubated with NSCLC cell lines H157, A549, and H2122, and the transduced cells were selected in growth medium containing G418 (250 g/ml) for LNCX2 vectors or puromycin (1 g/ml) for LPCX vectors. NSCLC cells expressing both HA-Wnt and Fzd-9 were first transduced with LNCX2-HA-Wnt, and subclones expressing the HA-tagged Wnt proteins were subsequently transduced with LPCX-Fzd-9. The cells were maintained in medium containing G418 and puromycin.
For the measurement of transfected cell growth on plastic dishes, 50,000 cells of the different transfectants were seeded per well of a 24-well culture plate in complete growth medium. On subsequent days as indicated, the cells were trypsinized from the wells with 100 l of trypsin, diluted with 400 l of growth medium, and counted using a hemocytometer. For measurement of anchorage-independent cell growth, 2500 cells were plated in triplicate in 35-mm wells of a 6-well plate in a volume of 1.5 ml of growth medium containing 0.3% Nobel agar onto a base of 1.5 ml of growth medium containing 0.5% agar. The plates were incubated in a 37°C CO 2 incubator for 21 days, and the colonies were counted using a microscope. The data are presented as cloning efficiency calculated by dividing the mean number of colonies per well by the number of cells (2500) plated.
Semiquantitative and Quantitative PCRs-RNA was extracted from cultured cells with the RNeasy minikit (Qiagen Inc.). Total RNA that had been extracted from primary lung tumors as well as uninvolved lung was provided by the Specialized Program of Research Excellence Lung Cancer Tissue Procurement Core of the University of Colorado Health Sciences Center. Aliquots of the RNA (10 g) were converted to cDNA with Superscript II (Invitrogen) and random hexamers according to the manufacturer's specifications. Primer sets for semiquantitative and quantitative PCRs are shown in Table I. Semiquantitative PCR was carried out using a PerkinElmer Life Sciences GeneAmp PCR System 9600 with an initial denaturation step of 94°C for 10 min, followed by 20 cycles at 94°C for 1 min, 55°C for 1 min, and 72°C for 2 min and a final extension cycle at 72°C for 10 min. Each reaction contained 1 l of cDNA template, 5 pmol each of forward and reverse primers, 0.75 units of AmpliTaq Gold, 200 M each dNTP, and 1.5 mM MgCl 2 . PCR products were visualized on a 1.5% agarose gel with ethidium bromide. Each sample was amplified twice using the same set of cDNAs and once using cDNAs derived from independently isolated RNAs.
Aliquots (1 l) of reverse transcription reactions were subjected to PCR conditions of 95°C for 10 min, followed by 95°C for 15 s and 60°C for 1 min for 40 cycles in 50-ml reactions with SYBR® green Jumpstart Taq Readymix (Sigma) using the primer sets shown in Table I. Initial real-time PCR amplification products were resolved by electrophoresis on 5% polyacrylamide gels to verify that the primer pairs amplified a single product of the predicted size. ␤-Actin mRNA levels were measured by quantitative PCR in the samples as a control gene. The realtime PCR data were analyzed with the Smart Cycler® software (Version 1.2d) to calculate the threshold cycle values for the different samples and are presented as mRNA levels in arbitrary units.
Transient Transfections and Luciferase Assays-Aliquots of lung cancer cells (2 million cells in 100 l) were electroporated at 220 V and 250 microfarads with a GeneZAPPER (IBI, New Haven, CT) in Gene Pulser 0.4-cm electrode gap cuvettes (Bio-Rad). Cells were transfected with 2 g of TOPflash and 1 g of pCMV-␤-gal for determination of transfection efficiency and with other expression plasmids as indicated in the figure legends. For analysis of c-Jun activity in cells, cells were similarly transfected with 100 ng of plasmid c-Jun-Gal4, 2 g of 5xUAS-TK-Luc, and 1 g of pCMV-␤-gal. Following electroporation, cells were plated in 10-cm dishes in complete medium. After 3 days of incubation, the cells were collected, washed once with ice-cold phosphate-buffered saline, and resuspended in 250 l of luciferase reporter lysis buffer (Promega, Madison, WI). The cell lysates were centrifuged in a microcentrifuge, and aliquots (80 l) of the supernatants were assayed for luciferase activity using a Monolight 2010 luminometer (Analytical Luminescence Laboratory, Ann Arbor, MI) and luciferase assay substrate (Promega). Aliquots (15 l) of the extracts were also assayed for ␤-galactosidase to correct for transfection efficiency. The data are presented as relative light units/milliunit of ␤-galactosidase.
Immunoblotting and Immunoprecipitation-For immunoblotting of HA-Wnt, cell extracts were prepared in MAPK lysis buffer (0.5% Triton X-100, 50 mM ␤-glycerophosphate (pH 7.2), 0.1 mM sodium vanadate, 2 mM MgCl 2 , 1 mM EGTA, 1 mM dithiothreitol, 2 g/ml leupeptin, and 4 g/ml aprotinin) as described previously (20). For immunoblot analysis of Fzd-9, the cells were suspended in hypotonic lysis buffer (10 mM Tris-Cl (pH 8.0), 1 mM EDTA, 4 g/ml phenylmethylsulfonyl fluoride, 2 g/ml leupeptin, 4 g/ml aprotinin, and 1 mM dithiothreitol) and forced three times through a 26-gauge syringe needle to lyse the cells, and the homogenate was submitted to centrifugation at 1000 ϫ g for 5 min to pellet nuclei and unbroken cells. The supernatant was subsequently centrifuged at 10,000 ϫ g for 10 min to collect the membrane fragments, which were then resuspended in 100 l of MAPK lysis buffer. Aliquots of the different extracts were resolved by 10% SDS-PAGE and transferred to nitrocellulose. The filters were blocked in Tris-buffered saline (10 mM Tris-Cl (pH 7.4) and 140 mM NaCl) containing 0.1% Tween 20 and 3% nonfat dry milk and then incubated with the same blocking solution containing the indicated antibodies at 1 g/ml for 12-16 h. The HA epitope was detected with monoclonal antibody 12CA5 (Roche Applied Science), and Fzd-9 was detected with a previously described rabbit polyclonal antibody (60). The filters were extensively washed with Tris-buffered saline containing 0.1% Tween 20, and bound antibodies were visualized with alkaline phosphatase-coupled secondary antibodies and LumiPhos reagent (Pierce) according to the manufacturer's directions.
Three-dimensional Cell Culture, Immunofluorescence Analysis, and Image Acquisition-Cells were grown in three-dimensional basement membrane cultures according to Debnath et al. (21) with the following modifications. Growth factor-reduced Matrigel (BD Biosciences) was combined in a 1:1 ratio with complete serum medium (RPMI 1640 medium ϩ 10% fetal bovine serum). 90 l was added to each well of an 8-well glass slide chamber and allowed to solidify for 2 h in a 37°C incubator. Cells were trypsinized, counted, and diluted to 25,000 cells/ ml. A 10% Matrigel solution was prepared in complete serum medium. The cell suspension was combined in a 1:1 ratio with the 10% Matrigel solution, and 200 l of this mixture was added to each well for a final concentration of 5000 cells/well in 5% Matrigel. Cells were consequently fed complete serum medium containing 5% Matrigel every 3 days for a defined term of culture. After 8 days of incubation, the cultures were fixed with 2% paraformaldehyde for 20 min, and permeabilization was performed with phosphate-buffered saline containing 0.5% Triton X-100, followed by rinsing several times with phosphate-buffered saline/glycine. The primary and secondary blocks were performed according to Debnath et al. (21). Primary antibodies against rabbit laminin V (Chemicon International, Inc., Temecula, CA) and mouse phospho-ERM (ezrin/radixin/moesin; Cell Signaling Technology, Inc., Beverly, MA) were both used at 1:100 dilution and were incubated overnight (15)(16)(17)(18) h) at room temperature. Alexa-conjugated rabbit and mouse secondary antibodies were both added at 1:500 dilution and incubated for 50 min at room temperature. The cells were rinsed several times and mounted with ProLong antifade reagent (Molecular Probes, Inc., Eugene, OR) for 15 min. Confocal analyses were performed using LSM410 confocal microscopy systems (Carl Zeiss MicroImaging, Inc.). Composite images were generated using MetaMorph software (Version 4.6r6).
JNK Assay-Cells were lysed in MAPK lysis buffer. Following microcentrifugation at 10,000 ϫ g for 5 min, aliquots of the extracts containing 200 g of protein were incubated for 2 h at 4°C with GST-c-Jun-(1-79) immobilized on glutathione-agarose (10 l of packed beads/sample containing ϳ5-10 g of protein). The GST-c-Jun-(1-79)-agarose complexes were washed three times by repetitive centrifugation in MAPK lysis buffer and then incubated for 20 min at 30°C in 40 l of 50 mM ␤-glycerophosphate (pH 7.6), 0.1 mM sodium vanadate, 10 mM MgCl 2 , and 20 M [␥-32 P]ATP (25,000 cpm/pmol). The reactions were terminated with 10 l of SDS-PAGE sample buffer and submitted to 10% SDS-PAGE. The GST-c-Jun-(1-79) polypeptides were identified on Coomassie Bluestained gels, excised, and counted in a scintillation counter.

Wnt-7a mRNA Expression Is Reduced in NSCLC Cells and Primary Tumors and Re-expression Inhibits Proliferation in a
Subset of Cell Lines-The Wnt proteins leading to ␤-catenin accumulation through stimulation of the Fzd receptors are widely invoked as an oncogenic signaling pathway in diverse tumors (1,5). However, recent evidence indicates that specific Wnt proteins may function as tumor suppressors in certain instances where loss of Wnt expression is observed in cancer cells (22)(23)(24). In this regard, a recent study demonstrated a frequent loss of Wnt-7a mRNA expression in lung cancer cell lines and primary lung tumors (25), indicating that Wnt-7a may represent a novel tumor suppressor in lung cancer. Our quantitative reverse transcription (RT)-PCR experiments revealed that Wnt-7a mRNA levels were undetectable in six of the eight NSCLC cell lines compared with the four normal human STBE cells. Two of the eight cell lines (H1334 and H2122) had detectable, but markedly lower expression (Fig.  1A, left panel). Importantly, Wnt-7a mRNA expression was significantly lower in the 13 primary NSCLC tumors relative to the 13 matched uninvolved lung tissues as assessed by quantitative RT-PCR analysis (Fig. 1A, right panel). By contrast, Wnt-3, Wnt-4, Wnt-5a, and Wnt-10b mRNAs were present at approximately equal levels in STBE cells relative to the panel of NSCLC cell lines (Fig. 1B) and Wnt-1, Wnt-2, Wnt-5b, Wnt-6, and Wnt-8b mRNAs were not reproducibly detected in either the STBE or NSCLC cells (data not shown). Thus, our results demonstrate decreased expression of Wnt-7a mRNA in lung cancer cell lines and primary tumor samples relative to STBE cells and uninvolved lung tissue, consistent with a previous study (25), and indicate that loss of Wnt-7a expression is a frequent molecular event that accompanies oncogenesis in the lung.
If loss of Wnt-7a expression in NSCLC cells contributes to lung cancer progression, then re-expression of Wnt-7a is predicted to negatively influence the growth of NSCLC cells lacking Wnt-7a. To this end, a retroviral vector encoding HA-tagged Wnt-7a was constructed and packaged into retrovirus (see "Experimental Procedures"). As a control, retroviruses encoding the empty LNCX vector or HA-Wnt-3, a Wnt protein ubiquitously expressed in NSCLC cell lines and STBE cells (Fig. 1), were generated. NSCLC cell lines H157, A549, and H661 were transduced with the packaged retroviruses, and stable transfectants were selected for resistance to G418. Fig. 2 shows an anti-HA immunoblot of cell extracts prepared from A549 and H157 transfectants transduced with the indicated retroviruses. Whereas transfected HA-Wnt-3 was readily detected in both A549 and H157 cells, transfected HA-Wnt-7a was observed only in transduced H157 cells, with only trace levels of HA-Wnt-7a detected in multiple independent A549 transfectants. Moreover, retroviral infection of H661 cells with the HA-Wnt-7a retrovirus failed to generate any G418-resistant H661 cells in multiple retroviral infections. This result suggests that Wnt-7a expression exerts a strong negative influence on growth of A549 and H661 cells, but not H157 cells.
The influence of Wnt-7a expression on cell proliferation measured on standard tissue culture plastic in complete medium was tested. The findings revealed that expression of Wnt-7a significantly reduced the growth rate of A549 cells compared with cells transduced with the empty vector (Fig. 3A) despite the fact that the HA-tagged Wnt-7a protein was only weakly expressed ( Fig.  2A). By contrast, the growth rate of H157 cells expressing substantially higher levels of HA-Wnt-7a protein was not different from that of the LNCX controls (Fig. 3B). In preliminary studies, expression of Wnt-7a also failed to influence the growth rate of H2122 cells (data not shown). Importantly, expression of Wnt-3 did not significantly affect the growth of any of the cell lines tested (Fig. 3, A and B), indicating that the growth inhibition observed in A549 cells with Wnt-7a was not simply a general effect of Wnt overexpression.  (Table I) as described under "Experimental Procedures." The relative mRNA abundance for Wnt-7a in the different samples was normalized to ␤-actin mRNA measured by RT-PCR in the same samples. B, total RNAs from the indicated cell lines were submitted to semiquantitative RT-PCR analyses as described under "Experimental Procedures" using the Wnt primer pairs listed in Table I. was affected by Wnt-7a expression, but that the growth of A549 and presumably H661 cells was reduced or strongly inhibited by Wnt-7a expression suggests that A549 and H661 cells express a specific Fzd protein that is not present in H157 or H2122 cells and that transmits signals leading to cell growth inhibition. To investigate the expression status of different Fzd mRNAs in NSCLC cell lines, we performed semiquantitative RT-PCR with the primer pairs listed in Table I to screen for six Fzd mRNAs (Fzd-2, Fzd-3, Fzd-5, Fzd-6, Fzd-7, and Fzd-9). Among the Fzd mRNAs tested, Fzd-3, Fzd-6, and Fzd-7 were detected in all of the lung cancer cell lines as well as in STBE cells (Fig. 4). Fzd-2 and Fzd-5 mRNAs were not reproducibly detected in any of the lung cancer cell lines or the STBE cells (data not shown). By contrast, Fzd-9 mRNA was not detected in the STBE cells, but was detected in four of the eight NSCLC cell lines (Fig. 4A). This finding was confirmed by quantitative RT-PCR, revealing that Fzd-9 mRNA was expressed at markedly elevated levels relative to STBE cells in the four NSCLC cell lines in which A549 and H661 cells showed the highest expression level (Fig. 4B). Consistent with the quantitative RT-PCR data, an immunoblot of membrane fractions prepared from the panel of NSCLC cell lines demonstrated high levels of Fzd-9 protein expression in A549 and H661 cells, but weak expression in the other NSCLC cell lines (Fig. 4C). In addition, Fzd-9 mRNA levels in total RNA from 13 primary lung tumor samples and 13 matched uninvolved normal lung samples were measured by quantitative PCR and revealed that Fzd-9 mRNA expression levels were significantly higher in primary lung tumors relative to uninvolved lung tissues (Fig. 4B), consistent with a general increase in expression of Fzd-9 in the NSCLC cell lines.
Inspection of the data in Figs. 1 and 4 revealed that the ability to successfully express Wnt-7a in the NSCLC cell lines correlated with a lack of expression of Fzd-9 mRNA. The H157 cell line, in which Wnt-7a could be efficiently expressed, showed little or no Fzd-9 mRNA, whereas A549 and H661 cells, in which transfection of Wnt-7a proved extremely difficult, expressed readily detectable Fzd-9 mRNA. In fact, H661 cells, which failed to yield stable Wnt-7a transfectants, expressed the highest level of Fzd-9 mRNA among the NSCLC lines. Thus, we hypothesized that coexpression of Wnt-7a and Fzd-9 establishes growth inhibitory signaling and that loss of Wnt-7a expression provides a means for escape from this negative growth control. To test this hypothesis, the fzd-9 cDNA was inserted into LPCX, and retrovirusmediated gene transfer was used to stably express Fzd-9 in H157 cells previously transduced with the empty LNCX vector or LNCX2-Wnt-7a-HA. In addition, H2122 cells, which express detectable endogenous Wnt-7a mRNA (Fig. 1), were similarly transduced with LPCX or LPCX-Fzd-9. An anti-Fzd-9 immunoblot of membrane fractions prepared from H157 and H2122 cells in which the fzd-9 cDNAs were expressed by the indicated retroviruses is shown in Fig. 2B and   FIG. 2. Immunoblot analysis of transfected HA-Wnt and Fzd-9 in NSCLC cell lines. A, extracts were prepared from the indicated HA-Wnt-transfected H157 and A549 clones with MAPK lysis buffer. Aliquots containing 100 g of protein were resolved on SDS-10% polyacrylamide gels, transferred to nitrocellulose, and probed with anti-HA monoclonal antibody 12CA5. A nonspecific (ns) endogenous cellular protein recognized by the antibody is indicated. B, membrane fractions were isolated from the indicated transfectants as described under "Experimental Procedures," and aliquots containing 75 g of protein were resolved on SDS-10% polyacrylamide gels, transferred to nitrocellulose, and probed with rabbit anti-Fzd-9 polyclonal antibody. C, extracts were prepared from the indicated H157 transfectants with radioimmune precipitation assay buffer as described under "Experimental Procedures." Aliquots containing 100 g of protein were resolved by SDS-PAGE and immunoblotted (IB) with anti-Fzd-9 antibody to verify expression of Fzd-9 protein (left panel). In addition, aliquots of the extracts containing 200 g of protein were immunoprecipitated (IP) with protein G-Sepharose and mouse anti-HA monoclonal antibody 12CA5 to collect HA-Wnt-7a (middle panel) or rabbit anti-Fzd-9 antibody or normal rabbit serum (NRS) (left panel) as indicated. Following SDS-PAGE, the filters were immunoblotted with the indicated antibodies. The mouse anti-HA immunoprecipitate was blotted with rabbit anti-HA antibody (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) to avoid detection of the mouse IgG heavy chains.

FIG. 3. Transfected Wnt-7a and Fzd-9 inhibit proliferation of NSCLC cells.
The indicated NSCLC transfectants were seeded in growth medium containing 10% fetal bovine serum into 24-well plates, and cell number was measured daily as described under "Experimental Procedures." demonstrates the successful expression of Fzd-9 in these cells. The H2122 cells expressing Fzd-9 exhibited a markedly reduced growth rate relative to cells transduced with the empty LPCX vector (Fig. 3C). Likewise, H157 cells transduced with both Wnt-7a and Fzd-9 exhibited a decreased growth rate relative to empty vector-transfected cells (Fig.  3D). As with expression of Wnt-7a alone (Fig. 3B), H157 cells transfected with Fzd-9 alone proliferated at a rate equal to empty vector-transfected cells. Importantly, coexpression of Wnt-3 alone or combined with Fzd-9 had no effect on the rate of H157 cell proliferation (Fig. 3, B and D), indicating the specificity of Wnt-7a for growth inhibition. In addition to the rate of proliferation on tissue culture plastic, coexpression of Wnt-7a and Fzd-9 in H157 and H2122 cells inhibited anchorage-independent growth in soft agar-containing medium. The soft agar cloning efficiency of H2122 cells expressing Fzd-9 was Ͻ1% of that of cells transfected with the empty LPCX vector (Fig. 5). The cloning efficiency of H157 cells transfected with Fzd-9 alone or with Wnt-7a alone was not different from that of cells transfected with the empty LNCX and LPCX vectors, but the cloning efficiency of cells expressing both Fzd-9 and Wnt-7a was inhibited by ϳ65% (Fig. 5). Thus, coexpression of Wnt-7a and Fzd-9 in either H2122 or H157 cells inhibited cell proliferation as well as anchorage-independent growth, the latter being an excellent in vitro measure of cellular transformation. These data combined suggest that re-establishment of a Wnt-7a and Fzd-9 signaling system exerts a tumor suppressor phenotype.
The simplest interpretation of growth inhibition dependent upon coexpression of Fzd-9 and Wnt-7a is that Wnt-7a binds to Fzd-9 and stimulates growth inhibitory signaling pathways. To test whether Wnt-7a can bind to Fzd-9, cell extracts from H157 cells expressing the empty LNCX and LPCX vectors or coexpressing HA-Wnt-7a and Fzd-9 were submitted to an anti-Fzd-9 immunoprecipitation reaction, and the immunoprecipitated proteins were immunoblotted for the presence of HAtagged Wnt-7a. As shown in Fig. 2C, HA-tagged Wnt-7a was detected in anti-Fzd-9 immunoprecipitates, but not in complexes precipitated with normal rabbit serum, providing evidence for the direct physical interaction of Wnt-7a and Fzd-9.

Coexpression of Wnt-7a and Fzd-9 Stimulates Epithelial Differentiation Accompanied by Induction of Sprouty-4 and Cad-
herin Proteins-Normal epithelia maintain cell-to-cell contacts that contribute to the proper development and maintenance of epithelial polarity and architecture (26 -29). The loss of these inputs has been proposed as a mechanism whereby epithelial cells lose their characteristic phenotype and acquire motile and invasive properties (29). There is ample evidence to support that an epithelial-to-mesenchymal transition permits dissemination of single carcinoma cells from the sites of primary tumors and is involved in the dedifferentiation program that leads to malignant carcinoma (30). To examine the effect of Wnt-7a and Fzd-9 coexpression on cell polarity and the epithelial phenotype, a three-dimensional Matrigel culture assay was employed (see "Experimental Procedures"). Fig. 6 demonstrates that transfection of Fzd-9 had a dramatic effect on the morphologic architecture of H2122 cells cultured in a threedimensional Matrigel matrix. Compared with H2122 cells transfected with the empty LPCX vector (H2122-LPCX cells), H2122 cells transfected with Fzd-9 (H2122-Fzd-9 cells) showed a polarized deposition of laminin V, a basement membrane  4. RT-PCR analysis of Fzd mRNAs in NSCLC cell lines and primary lung tumors. A, total RNA was isolated from the indicated NSCLC cell lines and submitted to semiquantitative RT-PCR with the primers listed in Table I as described under "Experimental Procedures." The PCRs were resolved by agarose gel electrophoresis, and the DNA was stained with ethidium bromide and photographed. The results shown are representative of at least three independent RT-PCRs for each sample with two independent RNA samples. B, total RNAs from the cell lines in A or from primary NSCLC tumors and uninvolved lung tissues were submitted to quantitative RT-PCR with the Fzd-9 primer pair as described under "Experimental Procedures." C, membrane fractions purified from the indicated NSCLC cell lines were immunoblotted with anti-Fzd-9 antibody.
component. In addition, H2122-Fzd-9 cells stained with antibodies against phospho-ERM revealed a more organized spheroid structure, with localization of the anti-ERM antibody to the apical and subapical regions relative to cells transfected with LPCX, which grew as unorganized aggregates (Fig. 6). Although not pronounced, many of the epithelial structures generated by H2122-Fzd-9 cells showed some evidence of cavitation or cyst formation (Fig. 6). A more modest epithelial polarization response was seen in H157 cells expressing Wnt-7a and Fzd-9 compared with empty vector-transfected H157 control cells, in which formation of cysts was not noted (data not shown). These data indicate that coexpression of Wnt-7a and Fzd-9 in an NSCLC cell line can induce the acquisition of a more differentiated epithelial phenotype.
The dominant signaling pathways implicated in transformation of NSCLC cells are mutations in K-Ras and overexpression of epidermal growth factor (EGF) receptor family members (31). EGF-stimulated protein tyrosine phosphorylation of specific cellular proteins was markedly inhibited in H2122 cells expressing Fzd-9 relative to empty vector-transfected control cells (Fig. 7). Neither the phosphotyrosine content of the EGF receptor nor the expression level of the EGF receptor was affected by Wnt-7a and Fzd-9 transfection (Fig. 7). This finding suggests that the ability of the EGF receptor to phosphorylate cellular targets is reduced upon expression of Wnt-7a and Fzd-9. Sprouty proteins are known intracellular antagonists of receptor tyrosine kinases (32, 33), and immunoblot analysis of the expression levels of Sprouty proteins in the transfected NSCLC cell lines revealed that Sprouty-4 was induced by 2-3fold in H157 cells expressing Wnt-7a and Fzd-9 and by 5-10fold in H2122 cells transfected with Fzd-9 (Fig. 8). Sprouty-1 and Sprouty-2 were not induced in either cell line (data not shown). These data suggest coexpression of Wnt-7a and Fzd-9 induces Sprouty-4, an antagonist of receptor tyrosine kinase activity, and may contribute to the reduced in vitro growth of transfected NSCLC cells.
The cadherin family of proteins is critical for maintenance of normal epithelial function and architecture (28,34). Immunoblot analysis of E-cadherin revealed a marked induction in H2122 cells transfected with Fzd-9 relative to empty vectortransfected control cells (Fig. 8). Interestingly, expression of E-cadherin in H157 cells cotransfected with Wnt-7a and Fzd-9 was not different from that in empty vector-transfected control cells. Rather, expression of N-cadherin was markedly induced (Fig. 8). Thus, increased epithelial differentiation by Wnt-7a and Fzd-9 coexpression is associated with increased expression of specific cadherin proteins, which are likely to contribute to the epithelial differentiation response observed in Fig. 6.
Growth Inhibition by Transfected Wnt-7a and Fzd-9 Is Associated with Activation of the JNK Pathway, but Not the ␤-Catenin/TCF Pathway-Stabilization of ␤-catenin leading to transcriptional activation of gene expression through the TCF/ lymphoid enhancer factor is a major signaling pathway regulated in response to Wnt proteins (1,3). H157 cells stably transfected with empty vector, Wnt-3, or Wnt-7a with or without Fzd-9 ( Figs. 1 and 4) were transiently cotransfected with the TOPflash reporter containing tandomized TCF-binding sites linked to firefly luciferase. Wnt-7a and Wnt-3 coexpressed with or without Fzd-9 were expressed at comparable levels (data not shown). As shown in Fig. 9A, coexpression of Wnt-3 and Fzd-9, which did not inhibit cell proliferation (Fig. 3A) or anchorage-independent growth, increased TOPflash activity by were serum-starved for 24 h and then incubated with or without EGF for 15 min as indicated. Cell extracts were prepared with MAPK lysis buffer, resolved by SDS-PAGE, and probed with anti-phosphotyrosine (PY) antibody. The filter was stripped and reprobed for the EGF receptor (EGFR) as indicated. Anti-phosphotyrosine antibody-stained polypeptides of ϳ50 kDa were reduced in the extracts from H2122-Fzd-9 cells. ϳ20-fold relative to the empty vector or Wnt-3 alone. By contrast, coexpression of Wnt-7a and Fzd-9 failed to significantly increase TOPflash activity (Fig. 9A). Thus, this experiment indicates that Wnt-7a exerts its growth inhibitory action in NSCLC cells through a pathway independent of the ␤-catenin/ TCF pathway. In addition, the results demonstrate that a distinct Fzd protein can engage different signaling pathways depending on the specific Wnt protein it binds.
The JNK pathway has been identified as an alternative signaling pathway regulated by Wnt proteins in both invertebrate and mammalian cells (35,36). As measured by an in vitro kinase assay, JNK activity was increased in both H157 and H2122 cells following coexpression of Wnt-7a and Fzd-9 (Fig. 9B). H157 cells coexpressing Wnt-7a and Fzd-9 (H157-Wnt-7a/Fzd-9 cells) exhibited a statistically significant 1.8 Ϯ 0.2-fold increase in JNK activity compared with empty vector-transfected control cells, whereas H2122 cells transfected with Fzd-9 exhibited a 2.6 Ϯ 0.8-fold increase relative to empty vector-transfected control cells. In addition, an anti-phospho-c-Jun immunoblot was performed on replicate aliquots of the extracts used for measurement of JNK activity and demonstrated a marked increase in endogenous c-Jun phosphorylation in H157 cells expressing both Wnt-7a and Fzd-9 relative to empty vector-transfected control cells. All together, these data indicate that Wnt-7a and Fzd-9 signal through the JNK pathway. In keeping with a role for the Rho family of monomeric G proteins in the actions of Wnt proteins in invertebrate systems (37,38), treatment of H2122 cells with the Rho-associated kinase inhibitor Y-27632 partially inhibited the phosphorylation of c-Jun observed in cells transduced with Fzd-9 (Fig. 9D). This finding is consistent with a recent report demonstrating that Rho kinase mediates JNK activation by Rho proteins (39). Measurement of c-Jun transcriptional activity by transient transfection of a c-Jun-Gal4 and 5xUAS-TK-Luc reporter system (see "Experimental Procedures") revealed a 2.0 Ϯ 0.1-and 2.3 Ϯ 0.3-fold stimulation of luciferase activity in H157 and H2122 cells, respectively, coexpressing Wnt-7a and Fzd-9 relative to the empty vector. Cotransfection of dominantnegative Rac (RacN19) decreased c-Jun-Gal4 activity to 0.3 Ϯ 0.2-and 0.3 Ϯ 0.1-fold in H157-Wnt-7a/Fzd-9 and H2122-Fzd-9 cells, respectively. Thus, JNK activation and c-Jun phosphorylation stimulated by coexpressed Wnt-7a and Fzd-9 appear to proceed through Rho family member signaling and Rho-associated kinases.
To test the role of the JNK pathway in the transformed growth inhibition induced by Wnt-7a and Fzd-9, H2122 cells were transduced with a retroviral vector encoding a previously described gain-of-function JNK polypeptide (40). This construct encodes a fusion protein of JNK1␣ and MKK7, its dual-specificity kinase activator. The forced proximity of MKK7 and JNK1 induces phosphorylation of the JNK1 moiety, leading to constitutive activation (41). Fig. 10A demonstrates the constitutive phosphorylation of the transfected MKK7-JNK1␣ fusion protein detected with anti-phospho-JNK antibody. Also, increased c-Jun Ser 73 phosphorylation was detected in MKK7-JNK1␣-transfected cells, indicating a functional activation of the JNK pathway (Fig. 10A). Two independent H2122 cell lines transfected with MKK7-JNK1␣ exhibited markedly reduced anchorage-independent growth in soft agar-containing medium (Fig. 10B). Thus, consistent with the increased JNK activity resulting from coexpression of Wnt-7a and Fzd-9, constitutive activation of the JNK pathway in H2122 cells reduces transformed cell growth, indicating that the JNK pathway is a likely mediator of the reduced transformed growth stimulated by Wnt-7a and Fzd-9.

DISCUSSION
Based on ample precedent in the literature for a transforming role of Wnt proteins in human cancers, we anticipated that one or more Wnt mRNAs would be induced in NSCLC cell lines. By contrast, our study instead unveiled the loss of a specific Wnt (Wnt-7a) in NSCLC cell lines and primary tumors. Importantly, our findings that re-expression of Wnt-7a in NSCLC cell lines reverses multiple indicators of cellular transformation provide evidence to support the function of Wnt-7a as a tumor suppressor in NSCLC. Although the overwhelming majority of published reports invoke a transforming role for Wnt proteins, a limited number of studies have also invoked a tumor suppressor role for specific Wnt proteins (23,42). Wnt-7a is upregulated in endometrial tissue by progestogens, a finding that may account for the antineoplastic effect of these hormones on the endometrium. Also, analysis of uterine leiomyomas demonstrated a frequent reduction in Wnt-7a expression relative to the adjacent myometria (22). The latter two findings may be related to the established importance of Wnt-7a in development and differentiation of the female reproductive tract (43). Loss of Wnt-5a expression has been observed in several tumor cell types, including renal cell carcinoma and breast cancer (24,42). Transfection of Wnt-5a into RCC23 renal cell carcinoma cells (42) or transformed uroepithelial cells (44) reverses the transformed phenotype, similar to our findings with transfection of Wnt-7a. In addition, an antisense strategy to decrease Wnt-5a expression induces transformation of C57MG mammary epithelial cells (42). Finally, Wnt-5a has been recently shown to inhibit proliferation of B cells and to function as a tumor suppressor in hematopoietic tissue (23). It is likely that expression of specific Wnt proteins may have dominant roles in tissue homeostasis and maintenance of epithelial cell differentiation and that loss of expression of these Wnt proteins will contribute to tumorigenesis in specific tissues.
A well recognized program of molecular events occurs upon cellular transformation and mediates an epithelial-to-mesenchymal transition (30). Normal epithelia possess cadherin-dependent cell-cell contacts that promote proper development of these tissues during embryogenesis and maintenance and homeostasis of adult epithelial structures. It has been proposed that as epithelial tumor cells progressively lose their epithelial phenotype, they obtain a more mesenchymal phenotype, which is associated with increased cell motility and migration (29). The reverse of epithelial-to-mesenchymal transition is mesenchymal-to-epithelial transition, and evidence supports the involvement of the Wnt proteins in mesenchymal-to-epithelial transition during morphogenesis (30). Our findings provide support for a role of Wnt-7a in maintenance of lung epithelial differentiation. Expression of Wnt-7a in the context of Fzd-9 inhibited NSCLC cell proliferation, transformed growth in soft agar, and induced a more differentiated epithelial phenotype as assessed in three-dimensional culture. In addition, E-cadherin and N-cadherin, markers of differentiated epithelial cells, were induced by coexpression of Wnt-7a and Fzd-9. Interestingly, Fzd-9 expression in H2122 adenocarcinoma cells induced Ecadherin, whereas coexpression of Wnt-7a and Fzd-9 in H157 squamous carcinoma cells resulted in N-cadherin (but not Ecadherin) induction. The induction of E-cadherin or N-cadherin by the Wnt-7a/Fzd-9 interaction may in part explain the mesenchymal-to-epithelial transition observed in this study.
We suggest that Wnt-7a and Fzd-9 signaling inhibits growth of NSCLC cells lines in part through activation of the JNK pathway, but not the ␤-catenin/TCF signaling pathway. Our data showing that Wnt-7a and Fzd-9 failed to engage the ␤-catenin pathway are consistent with a previous study dem-onstrating that Wnt-7a failed to engage the ␤-catenin signaling pathway when coexpressed with Fzd-9 in 293 cells (45). Importantly, genetic analyses in Drosophila revealed Wnt-dependent regulation of two epithelial programs (planar polarity and dorsal closure) that proceed through JNK activation, but independently of ␤-catenin (36). Recent studies have established that the mammalian counterpart of planar polarity, termed convergent extension, is also a Wnt-5a-dependent, ␤-cateninindependent program (35,46). Moreover, studies in Xenopus revealed that Wnt-5a-regulated convergent extension requires the JNK pathway (35). Although the JNKs are widely invoked as components of pro-apoptotic signaling cascades (47), substantial literature has emerged to support the JNK pathway as a required element in development, morphogenesis, and cell differentiation (48). We have previously demonstrated a role for the JNK pathway in neural differentiation modeled in PC12 cells (49) and increased muscle gene expression in vascular FIG. 9. Transfected Wnt-7a stimulates JNK activity and c-Jun phosphorylation, but not the ␤-catenin/TCF pathway. A, H157 cells transfected with Wnt-3 or Wnt-7a with or without Fzd-9 were transiently transfected with TOPflash and pCMV-␤-gal as described under "Experimental Procedures." The cells were incubated for 3 days, and luciferase and ␤-galactosidase activities were measured. The data are presented as relative light units/milliunit of ␤-galactosidase activity. B, the indicated stable H157 and H2122 transfectants were assayed for basal JNK activity with the GST-c-Jun-(1-79) adsorption assay as described under "Experimental Procedures." The data were normalized to the JNK activity measured in empty vector-transfected cells and are presented as the means Ϯ S.E. of four and three independent experiments for H157 and H2122 transfectants, respectively. *, statistically significant difference (p Ͻ 0.05) compared with the activity in control cells by Student's t test. C, aliquots of extracts from H157 cells transfected with or without Wnt-7a and Fzd-9 were immunoblotted for c-Jun phosphorylated at Ser 73 or total c-Jun protein as indicated. D, H2122 cells expressing the empty LPCX vector or Fzd-9 were treated for 30 min with or without 10 M Y-27632, a Rho-associated protein kinase inhibitor (ROCK Inhib.), and extracts were prepared and immunoblotted for c-Jun phosphorylated at Ser 73 or total c-Jun as indicated.

FIG. 10. Expression of gain-of-function JNK in NSCLC cells inhibits transformed growth.
A, H2122 cells expressing the MKK7-JNK1␣ fusion protein or the empty TRE vector were immunoblotted with anti-phospho-JNK, antiphospho-Ser 73 c-Jun, or anti-total c-Jun antibody to demonstrate expression of the activated MKK7-JNK1␣ polypeptide. B, the indicated H2122 transfectants were assayed for anchorage-independent growth in soft agar-containing medium. The colonies were counted after 3 weeks of incubation. smooth muscle cells (50). In both instances, a modest JNK activation similar to the activation observed in response to Wnt-7a and Fzd-9 coexpression (Fig. 9) was observed. With regard to epithelial cell differentiation and related to our own findings, a recent study demonstrated a JNK requirement for the in vitro formation of polarized acinar structures in a breast epithelial cell line (51).
Our data indicate that the growth inhibitory action of Wnt-7a is mediated by Fzd-9 and imply the existence of signaling specificity by distinct Wnt and Fzd combinations. The fact that coexpression of Wnt-3 and Fzd-9 did not have an effect on in vitro cell growth or anchorage-independent growth provides additional support for the specificity of Wnt/Fzd interactions yielding distinct cellular responses. This has in fact long been appreciated with other members of the seven-membrane receptor superfamily such as the adrenergic and cholinergic receptors. In Drosophila, DFz1 signals through the JNK pathway rather than ␤-catenin (36). In addition, Sheldahl et al. (52) have shown that signaling through Fzd-1, Fzd-7, and Fzd-8 results in ␤-catenin activation, whereas signaling through Fzd-2, Fzd-3, Fzd-4, and Fzd-6 stimulates protein kinase C activity through a G protein-dependent mechanism. Beyond signaling specificity inherent to the different Fzd proteins, our findings shown in Fig. 9A indicate that Fzd-9 engages the ␤-catenin/ TCF pathway when stimulated by Wnt-3, but the JNK pathway when stimulated by Wnt-7a. Thus, distinct Fzd molecules appear to engage different signaling pathways depending on the specific Wnt protein they bind. Our results shown in Fig.  2C indicate that Wnt-7a directly interacts with Fzd-9 to induce growth inhibition, increased epithelial differentiation, and JNK activation.
Overexpression of EGF receptor family members is frequently observed in NSCLC (53)(54)(55). More recently, gain-of-function EGF receptor mutations have been identified in some NSCLC cells (56,57). In this regard, our observation that Wnt-7a and Fzd-9 coexpression resulted in the induction of the receptor tyrosine kinase antagonist Sprouty-4 is novel and potentially important for the reduced anchorage-dependent and -independent growth of NSCLC cells. Severe defects in lobulation and lung hypoplasia during lung development are observed in the mammalian fetus overexpressing Sprouty-4, suggesting a role for Sprouty-4 as a growth-inhibiting protein (58).
Our findings showing little or no Wnt-7a mRNA expression in NSCLC cell lines and primary NSCLC tumors are consistent with a previous report by Calvo et al. (25). The mechanism for reduced expression of Wnt-7a was not addressed in the present study, but it is interesting to note that the wnt-7a gene maps to chromosome 3p.25.1, a recognized site of frequent genomic deletion in lung cancer (10,25,42). Although genomic deletion is a possible mechanism for loss of Wnt-7a expression, decreased Wnt-7a mRNA expression has been observed in pancreatic carcinoma cell lines, and treatment with inhibitors of DNA methylation or histone deacetylases restores Wnt-7a expression (59). Furthermore, direct analysis revealed that the wnt-7a gene is frequently (59%) methylated in pancreatic carcinoma cell lines (59). We are presently pursuing the mechanism for loss of Wnt-7a expression in NSCLC cell lines and primary tumors. The results of our studies predict that reinduction of Wnt-7a expression in human lung cancers would significantly decrease the transformed phenotype of the cancer cells in situ.