α-Catenin Inhibits β-Catenin Signaling by Preventing Formation of a β-Catenin·T-cell Factor·DNA Complex*

α-Catenin and β-catenin link cadherins to the cytoskeleton at adherens junctions. β-Catenin also associates with members of the T-cell factor (Tcf) family of transcription factors, and mutations in β-catenin lead to activation of Tcf-dependent transcription and increased cell growth. Although the loss of α-catenin expression can also promote cell growth, the role of endogenous α-catenin in β-catenin signaling is unclear. Here we show that loss of α-catenin expression in a colon cancer cell line correlates with increased Tcf-dependent transcription. The presence of α-catenin in colon cancer cell nuclei suggests that it inhibits transcription directly, and, in agreement with this, ectopic expression of α-catenin in the nucleus represses Tcf-dependent transcription. Furthermore, recombinant α-catenin disrupts the interaction between the β-catenin·Tcf complex and DNA. We conclude that α-catenin inhibits β-catenin signaling in the nucleus by interfering with the formation of a β-catenin·Tcf·DNA complex.

␤-Catenin has two separable functions; it is a component of the Wnt signal transduction pathway that results in axis formation and cell fate choice during embryonic development (1,2), and it is part of the cadherin cell adhesion complex (3). The central armadillo repeat domain of ␤-catenin is necessary for both of these functions, because it directly binds Tcf/LEF-1 1 family transcription factors to transduce Wnt signals and cadherins to promote cell-cell adhesion.
In the current model of the Wnt signaling pathway, the cytosolic level of ␤-catenin is controlled by phosphorylation and ubiquitin-dependent degradation. Wnt signals stabilize ␤-catenin, allowing it to associate with Tcf/LEF-1 family members and enter the nucleus to regulate gene expression. ␤-Catenin can also enter the nucleus independently, in a manner similar to that of the nuclear transport protein importin-␤ (4,5), suggesting that ␤-catenin may be a specific importin for other proteins. Support for this comes from studies of Armadillo (the Drosophila homologue of ␤-catenin), which co-imports the transcription factor Teashirt into the nucleus (6). Several tumor types harbor mutations in the genes encoding the tumor suppressor APC or ␤-catenin. These mutations result in permanent activation of Wnt target genes because the ␤-catenin⅐Tcf complex is no longer regulated by degradation of ␤-catenin (7)(8)(9). Importantly, proteins that are not involved in ␤-catenin stability can also regulate ␤-catenin signaling. NEMO-like kinase, for example, inhibits the interaction between the ␤-catenin⅐Tcf complex and DNA by phosphorylating Tcf/LEF proteins (10).
␣-Catenin is another protein that can inhibit ␤-catenin signaling independently of stabilizing mutations in ␤-catenin or APC (11,12). ␣-Catenin associates directly with ␤-catenin (13,14), linking cadherins to the actin cytoskeleton, an interaction that is essential for strong cell-cell adhesion (15,16). The expression of ␣-catenin is often reduced during tumor progression (17), and several cancer-derived cell lines have mutations in the ␣-catenin gene (18,19). Reintroduction of ␣-catenin into such lines reduces cell growth (18) and attenuates tumor formation (19). Although the tumor suppressor function of ␣-catenin is generally believed to result from its promotion of cell-cell adhesion, the possibility that ␣-catenin influences tumor progression by regulating ␤-catenin signaling has not been investigated.
We examined ␤-catenin signaling in colon cancer cell lines and found that loss of ␣-catenin expression correlated with increased ␤-catenin⅐Tcf-dependent transcription. Furthermore, ␣-catenin was found in cell nuclei, suggesting a role in processes other than cell adhesion. In support of this possibility, ectopic expression of ␣-catenin in the nucleus inhibited ␤-catenin⅐Tcf signaling, and ␣-catenin disrupted the interaction between the ␤-catenin⅐Tcf complex and DNA in vitro.

EXPERIMENTAL PROCEDURES
DNA Plasmid Constructs-Tcf cDNAs, TopFlash, and FopFlash reporter DNAs (20) were gifts from Marc van de Wetering and Hans Clevers (Utrecht University, Utrecht, The Netherlands). Human ␣Ncatenin cDNA (21) was provided by Christine Petit (Institute Pasteur, Paris, France). Plasmids encoding GST-␣-catenin constructs (22) and purified GST-␣-catenin were generously provided by Vania Braga (Medical Research Laboratory for Molecular Cell Biology), with permission from David Rimm (Yale University). GFP-␣N-catenin cDNA (23) was provided by Ravinder Sehgal and Louis Reichardt (University of California, San Francisco). To make GFP-␣-NLS, a double-stranded oligonucleotide (MWG Biotech, Germany) encoding the nuclear localization signal from SV40 large T antigen (24) and appropriate restriction sites was ligated into GFP-␣N-catenin cut with AflII and HindIII. This resulted in the insertion of the sequence AARDPKKKRKV after residue 902 of avian ␣N-catenin, followed by a stop codon. GFP-␣⌬-NLS was made by deleting internal sequences in GFP-␣-NLS using XhoI, resulting in a fusion protein containing residues 1-214 of ␣N-catenin followed by the nuclear localization signal.
Cell Culture and Transfections-HCT116 cells, SW480 cells, RKO cells, and DLD-1b cells were obtained from the Institute of Cancer Research (London, UK). DLD ␣ϩ cells were isolated by collecting and subcloning floating cells from the culture medium of DLD-1b cells. DLD-1 ␣ϩ cells will be characterized fully at a later date. Parental DLD-1 and DLD ␣Ϫ cells (25) were kindly provided by Dr. S. T. Suzuki (Institute for Developmental Research, Aichi Human Service Center, Aichi, Japan). HCT116 cells were grown in McCoy's medium with 10%  fetal calf serum, all other cells, including COS 7 and Neuro-2A cells,  were grown in Dulbecco's modified Eagle's medium supplemented with  10% fetal calf serum. Transient transfections were done according to manufacturers protocols using LipofectAMINE Plus reagent (Life Technologies, Inc.) in 6-well Falcon tissue-culture plates (Becton Dickinson), or, for immunocytochemistry, in 2-well Lab-Tek chamber glass slides (Nalge Nunc) precoated with 30 g/ml collagen for 5 min (Vitrogen 100, Collagen GmbH, Ismaning, Germany). 293 cells were transfected with 1 g of pMT23 vector or pMT23 ␤Amyc, a phosphorylation site mutant of ␤-catenin (12), and 0.5 g of GFP or GFP-␣⌬-NLS. Neuro-2A cells were transfected in 10-cm plates using either 1.5 g of p33 Tcf-1 or p45 Tcf-1, 1 g of GFP-␤-catenin, and 1.5 g of GFP-␣-catenin for immune precipitation assays or with 4 g of p45 Tcf-1 for gel shift assays. For transcription assays each well of a 6-well plate contained 25 ng of RSV ␤-galactosidase, 175 ng of pTopFlash (or pFopFlash), 100 ng of p45 Tcf-1 (COS 7 cells only), 100 ng of ␤-catenin construct (COS 7 cells only), and 500 ng of each GFP fusion construct. After 24 h cells were either harvested for preparation of extracts (except for 293 cells, which were harvested after 48 h) or fixed for analysis by fluorescence microscopy.
Immunofluorescence-Cells expressing GFP fusion proteins were washed in PBS and fixed for 15 min using 3% paraformaldehyde. For immunofluorescence, fixed cells were permeabilized with 0.2% Triton X-100, 150 mM NaCl, 50 mM Tris-Cl, pH 7.6, for 1 min or using cold methanol for 1 min to optimize detection of nuclear antigens, rinsed in PBS, and blocked for 30 min in PBS containing 1% bovine serum albumin (Fraction V, Sigma). Polyclonal anti-␣-catenin (1:500) and monoclonal anti-␤-catenin (1:250; Transduction Laboratories) were diluted in blocking buffer and added for 2 h. Cells were then washed four times for 10 min in blocking buffer with 0.05% Triton X-100, incubated for 1 h with Texas Red-conjugated (1:400) or fluorescein isothiocyanateconjugated (1:200) secondary antibodies (Jackson Inc.), and then washed again as above. After staining, slides were mounted in gelvatol. Confocal images were obtained using a laser scanner (MRC 1024, Bio-Rad) attached to a Nikon microscope (Optiphot 2) and processed using Adobe Photoshop.

␣-Catenin Is Present in the Nuclei of SW480 and DLD-1 Colon Cancer Cells-
The majority of colon cancer cell lines harbor mutations in APC or in ␤-catenin that result in the stabilization of ␤-catenin and formation of transcriptionally active ␤-catenin⅐Tcf complexes in the nucleus. Although ␣-catenin in normal cells is found at cell junctions and in the cytoplasm, its localization in colon cancer cells has not been examined in detail. We compared the subcellular localization of the catenins in colon cancer cells by immunocytochemistry (Fig. 1). As expected, ␤-catenin was present in the nuclei of SW480 cells (Fig. 1A), HCT116 cells (Fig. 1C), and DLD-1 cells (Fig. 1E), all of which are colon cancer cell lines with mutations in APC or in ␤-catenin. ␤-Catenin was difficult to detect in RKO colon cancer cells (Fig. 1G), probably because these cells express no cadherin and do not have mutations in APC or ␤-catenin (30). Interestingly, in addition to its localization in the plasma membrane and cytoplasm, ␣-catenin was present in the nuclei of SW480 cells (Fig. 1B) and DLD-1 cells (Fig. 1F) but not in the nuclei of HCT116 cells (Fig. 1D) or RKO cells (Fig. 1H). We examined two additional cell lines, DLD ␣Ϫ and DLD ␣ϩ. These are clonal variants of DLD-1 cells that form a disordered monolayer in culture; DLD ␣ϩ cells express ␣-catenin (Fig. 2), whereas DLD ␣Ϫ cells do not (25). The subcellular distribution of ␤-catenin in DLD ␣ϩ (Fig. 1I) and DLD ␣Ϫ (Fig. 1K) cells was comparable with that in DLD-1 cells (Fig. 1E), which form an ordered monolayer. Similarly, the distribution of ␣-catenin in DLD ␣ϩ cells (Fig. 1J) resembled that in DLD-1 cells (Fig.  1F). As expected, ␣-catenin was not detected in DLD ␣Ϫ cells ( Fig. 1L). These results, together with results obtained using other cell lines, 2 suggest that the nuclear localization of ␣-catenin is associated with that of ␤-catenin. However, other factors must also regulate the nuclear localization of ␣-catenin because it is found predominantly outside the nucleus in HCT116 cells, which contain nuclear ␤-catenin. In addition, formation of an ordered epithelial monolayer does not appear to influence the nuclear localization of the catenins, because they were distributed similarly in the membrane, cytoplasm, and nucleus of parental DLD-1 cells and DLD ␣ϩ cells.
Loss of Endogenous ␣-Catenin in DLD-1 Colon Cancer Cells Results in Increased ␤-Catenin⅐Tcf Transcriptional Activity-Although ␣-catenin overexpression can inhibit ␤-catenin signaling (11,12,23), it is unclear whether endogenous ␣-catenin plays a similar role. To determine whether endogenous ␣-catenin influences ␤-catenin signaling, we used the DLD cell lines described above. A parental line from a different source was included in this analysis (DLD-1b) to control for clonal differences among the cell lines. Except for DLD ␣Ϫ cells, which lack ␣-catenin ( Fig. 2A, lane 4), all the DLD cell lines expressed comparable levels of cadherin, ␤-catenin, ␣-catenin, and vinculin ( Fig. 2A). Compared with the DLD cell lines, SW480 cells expressed similar levels of ␣-catenin and vinculin but less cadherin and more ␤-catenin (lane 5). To compare the cadherincatenin complexes in the DLD cell lines, ␣-catenin immune precipitates from cell extracts were probed for the presence of ␤-catenin and cadherin by Western blotting (Fig. 2B). The cell lines that expressed ␣-catenin (lanes 1-3) contained an intact cadherin⅐␤-catenin⅐␣-catenin complex. As expected, anti-␣catenin immune precipitates from DLD ␣Ϫ cells did not contain cadherin or ␤-catenin (lane 4). The amount of cadherin associated with ␣-catenin in SW480 cells (lane 5) reflected the low level of cadherin in these cells.
To compare ␤-catenin signaling in the DLD cell lines, we used a transcriptional reporter assay (20). In this assay, cells are transiently transfected with plasmids encoding luciferase driven by a promoter containing either Tcf-binding sites (Top-Flash) or mutated Tcf-binding sites (FopFlash), together with a plasmid encoding ␤-galactosidase driven by the RSV promoter (to measure transfection efficiency). The ratio of the TopFlash and FopFlash values provides a measure of ␤-catenin signaling activity in the cells. Interestingly, when the DLD cell lines were compared using this assay (Fig. 2C), transcriptional activity was significantly elevated in DLD ␣Ϫ cells (p Ͻ 0.01). Importantly, the transcriptional activity in DLD ␣Ϫ cells was 30% higher than that in DLD ␣ϩ cells, suggesting that the increase in transcriptional activity results from the loss of ␣-catenin expression rather than from a change in cell morphology.
Targeted Expression of ␣-Catenin in the Nucleus Inhibits ␤-Catenin⅐Tcf Signaling-The inhibitory effects of ␣-catenin overexpression on ␤-catenin⅐Tcf transcription have been proposed to result from the sequestration of ␤-catenin in the cytoplasm (11). However, the presence of endogenous ␣-catenin in the nuclei of some colon cancer cells (Fig. 1) suggests that inhibition may occur in the nucleus. To test this possibility, we targeted ␣-catenin expression to the nucleus by adding a nuclear localization signal to a GFP-␣-catenin fusion protein. This form of ␣-catenin (GFP-␣-NLS) was expressed to similar levels as GFP-␣-catenin and associated with ␤-catenin as expected (data not shown). In contrast to GFP-␣-catenin, which was found both in the cytoplasm and in cell junctions in COS 7 cells (12), GFP-␣-NLS was found exclusively in the nucleus, where it had a diffuse staining pattern (Fig. 3A). Coexpression of GFP-␣-NLS with ␤-catenin in COS 7 cells resulted in the formation of nuclear rod-like structures (Fig. 3B) that colocalized with transfected ␤-catenin (data not shown). This pattern of localization is similar to that for ectopically expressed ␤-catenin (11,12) and contrasts with the diffuse nuclear localization pattern of GFP-␣-NLS in SW480 cells (Fig. 3C), which contain nuclear ␤-catenin⅐Tcf complexes. GFP-␣-catenin without a nuclear localization signal also had a diffuse nuclear localization pattern  in SW480 cells (Fig. 3D), consistent with a model in which ␤-catenin⅐Tcf complexes that are not associated with endogenous ␣-catenin transport transfected GFP-␣-catenin into the nucleus.
We reasoned that if ␣-catenin inhibits transcription by sequestering ␤-catenin in the cytoplasm, GFP-␣-NLS would be a less effective inhibitor than GFP-␣-catenin. However, GFP-␣catenin and GFP-␣-NLS inhibited transcription to a similar extent both in COS 7 cells transfected with ␤-catenin and Tcf and in SW480 cells (Fig. 4A), which contain endogenous ␤-catenin⅐Tcf complexes. These results are consistent with inhibition of ␤-catenin function occurring in the nucleus rather than by sequestration of ␤-catenin in the cytoplasm.
The central and C-terminal domains of ␣-catenin associate with a number of cytoskeletal proteins including vinculin (31, 32), ␣-actinin (33), ZO-1 (34), and actin (22). To determine the importance of these interactions for the inhibitory effects of ␣-catenin on transcription, we made a truncation mutant of ␣-catenin (GFP-␣⌬-NLS). GFP-␣⌬-NLS contains residues 1-214 of ␣-catenin, which includes the ␤-catenin-binding site (33,35,36) and a homo-dimerization site (37) but not in vivo binding sites for other known proteins. GFP-␣⌬-NLS was expressed to a similar level as GFP-␣-NLS and localized exclusively to the nucleus in transiently transfected COS 7 cells (data not shown). Interestingly, GFP-␣⌬-NLS reduced ␤-catenindependent transcription to a similar extent as GFP-␣-NLS in DLD ␣Ϫ cells and DLD ␣ϩ cells (Fig. 4B), as well as in SW480 cells and COS 7 cells (data not shown). The extent of inhibition by the GFP-␣-catenin constructs was greater in DLD ␣Ϫ cells than in DLD ␣ϩ cells, most likely because the ␤-catenin⅐Tcf complex in DLD ␣ϩ cells is partially inhibited by endogenous ␣-catenin. The ability of GFP-␣⌬-NLS to inhibit transcription in DLD ␣Ϫ cells indicates that the mechanism of inhibition does not involve binding to endogenous ␣-catenin. Taken together, these results suggest that the inhibition of ␤-catenin⅐Tcf signaling by nuclear ␣-catenin does not require interactions with other known cytoskeletal proteins.
To determine whether ␣-catenin also regulates transcription of genes relevant to growth control, we examined cyclin D1, which is a direct gene target of the ␤-catenin⅐Tcf complex (38,39). Western blots of extracts from 293 cells transfected with a stabilized form of ␤-catenin and either GFP or GFP-␣⌬-NLS were probed for cyclin D1 (Fig. 4C). As reported by others (39) ␣-Catenin Disrupts the Interaction of the ␤-Catenin⅐Tcf Complex with DNA-Our results suggest that ␣-catenin inhibits the ␤-catenin⅐Tcf complex directly in the nucleus. This might occur by disruption of the ␤-catenin⅐Tcf complex itself, disruption of its interaction with DNA, or disruption of interactions among ␤-catenin⅐Tcf and other proteins, such as components of the basal transcription machinery (40). To distinguish these possibilities, we first determined whether ␣-catenin could stably associate with the ␤-catenin⅐Tcf complex (Fig. 5A). ␣-Catenin, ␤-catenin, and Tcf-1 were coexpressed in Neuro-2A cells (these cells express low levels of endogenous catenins). Two isoforms of Tcf-1 were used, p45 (Fig. 5A, lanes 1 and 2), which associates with ␤-catenin, and p33 (lane 3), which does not. ␣-Catenin co-immune precipitated with p45 Tcf-1 (Fig. 5A, lane 4) but not with p33 Tcf-1 (lane 6). In addition, ␣-catenin did not co-immune precipitate with p45 Tcf-1 in the absence of ␤-catenin (lane 5), indicating that ␣-catenin interacts indirectly with Tcf-1 via ␤-catenin. Immune precipitates were also prepared from nuclear extracts of SW480 cells and probed for endogenous Tcf-4 (Fig. 5A, lanes 7-10). Two isoforms of Tcf-4 were detected in SW480 cell extracts (lane 7), which likely correspond to Tcf-4B and Tcf-4E (20). As expected, Tcf-4 associated with ␤-catenin (lane 8) but not with cadherin (lane 9). Tcf-4 also associated with endogenous ␣-catenin (lane 10). Thus, ␣-catenin can stably associate with the ␤-catenin⅐Tcf complex both in transfected cells and in colon cancer cells.
To determine whether ␣-catenin affected the interaction between the ␤-catenin⅐Tcf complex and its DNA target, we conducted gel retardation assays (Fig. 5B). Cell extracts were prepared from Neuro-2A cells that had been transfected with p45 Tcf-1. When the extracts were incubated with a Tcf oligonucleotide probe, a single protein-DNA complex was detected (lane 3) that could be supershifted with an anti-Tcf-1 antibody  (lane 10). Interestingly, addition of GST-␣-catenin abrogated binding of the ␤-catenin⅐Tcf-1 complex to DNA, although it did not affect binding of Tcf-1 to DNA (lane 7). Furthermore, a GST fusion to the N-terminal domain of ␣-catenin (N228, containing the ␤-catenin-binding site) also reduced the interaction of the ␤-catenin⅐Tcf-1 complex with DNA (lane 9), whereas a GST fusion to a C-terminal domain of ␣-catenin (C447, which does not bind to ␤-catenin) had no effect (lane 8). These results were quantitated by densitometric scanning of this and other representative gels and then adjusting the amount of each ␤-catenin⅐Tcf-1⅐DNA complex relative to the amount of Tcf-1⅐DNA complex. The amount of ␤-catenin⅐Tcf-1⅐DNA complex formed in the presence of GST-␣-catenin (lane 7) or N228 (lane 9) was between 32 and 50% of the amount of complex formed in the absence of ␣-catenin (lane 5) or in the presence of C447 (lane 8). These results suggest that the association of ␣-catenin with the ␤-catenin⅐Tcf-1 complex reduces the interaction between this complex and DNA. This may account for the inhibitory effects of ␣-catenin on ␤-catenin signaling in vivo. DISCUSSION The tumor suppressor function of ␣-catenin is generally believed to result from its role in cell-cell adhesion. However, ␣-catenin may also regulate adhesion-independent functions of ␤-catenin. In this report we provide evidence supporting a role for ␣-catenin in the nucleus.
The mechanism by which ␣-catenin overexpression inhibits ␤-catenin⅐Tcf-dependent transcription has been proposed to be by sequestration of ␤-catenin in the cytoplasm (11). This seems unlikely, first, because endogenous ␣-catenin and exogenously expressed GFP-␣-catenin can be found in cell nuclei, and second, because when ␣-catenin is targeted to the nucleus, it inhibits transcriptional activity to the same extent as nontargeted ␣-catenin. Because ␣-catenin associates stably with the ␤-catenin⅐Tcf complex in cells expressing all three proteins, it may inhibit ␤-catenin⅐Tcf transcription by preventing DNA binding. The results of gel retardation experiments (Fig. 5B) support this model, although they do not rule out that ␣-catenin also alters ␤-catenin⅐Tcf interactions with other proteins that regulate transcription in vivo.
␣-Catenin was present in the nuclei of SW480 and DLD-1 cells but not in the nuclei of HCT116 cells. The reason for this is presently unknown, but it suggests that nuclear localization of ␣-catenin is not determined solely by ␤-catenin and that it may be regulated by other mechanisms. Nuclear entry of ␣-catenin may also be regulated in nontransformed cells, for example during Wnt signaling. Sensitive assays may be required to test this, because it may involve transient changes in the localization of a fraction of the total cellular ␣-catenin.
Factors that regulate nuclear entry of ␣-catenin would link signaling events in the cytoplasm to ␤-catenin-dependent transcription in the nucleus. One possibility is that ␣-catenin responds to changes in the actin cytoskeleton. ␣-Catenin associates with a number of cytoskeletal proteins including ␣-actinin, vinculin, ZO proteins, and actin (3). The binding of these proteins (except possibly actin; see below) to ␣-catenin does not appear to be necessary for inhibition of transcription, because GFP-␣⌬-NLS inhibits transcription to a similar extent as GFP-␣-NLS. However, such interactions may regulate the ability of endogenous ␣-catenin to enter the nucleus and repress transcription.
An interesting possibility is that the effects of ␣-catenin on gene expression involve its ability to bind actin. Actin is normally rapidly exported from the nucleus (41), but it can accumulate there under certain conditions (42), and actin mutants that accumulate in the nucleus reduce cell proliferation (41). Moreover, changes in the level of actin can regulate transcriptional activity (43). ␣-Catenin has two domains that bind actin in vitro, a major site in the C-terminal domain and a second site close to the ␤-catenin-binding site (22). Although GFP-␣⌬-NLS does not contain the major site, it retains the second actinbinding site. Thus, it remains possible that binding to actin plays a part in the effects of ␣-catenin on ␤-catenin-dependent transcription.
Our results suggest that the relative levels of ␤-catenin⅐Tcf and ␣-catenin⅐␤-catenin⅐Tcf complexes in the nucleus will determine the overall transcriptional activity. Thus, in SW480 cells, where the amount of ␤-catenin far exceeds that of ␣-catenin, one would predict that endogenous nuclear ␣-catenin will not significantly affect transcription. In contrast, in DLD-1 cells, which contain far lower levels of ␤-catenin than SW480 cells (Fig. 1A), one would predict that loss of ␣-catenin would result in increased ␤-catenin⅐Tcf transcriptional activity. This appears to be the case because Tcf-dependent activity in DLD ␣Ϫ cells is 30% higher than in any of the other DLD cell lines (Fig. 2C). These results were obtained using TopFlash and FopFlash reporter plasmids. Because the basal (Tcf-independent) activity measured by FopFlash can be high in some cells (30), the apparent Tcf-dependent transcriptional activities calculated (determined as the ratio of TopFlash and FopFlash measurements) may obscure differences in Tcf-specific activity among the cell lines. We have also compared Tcf-dependent transcriptional activity using the optimized OT and OF reporter plasmids (30). The low basal (Tcf-independent) activity measured using OF results in a more accurate calculation of Tcf-dependent activity (as determined by the ratio of OT and OF measurements). The activity in DLD ␣Ϫ cells is 50% higher than in the other DLD cell lines when measured using the OT and OF reporters, confirming the results obtained using Top-Flash/FopFlash. 2 To date we have been unable to examine the effect of loss of endogenous ␣-catenin on ␤-catenin signaling in other colon cancer cell lines because there are no other cells known to have mutations both in ␣-catenin and in either ␤-catenin or APC.
The expression of ␣-catenin is often reduced during tumor progression (17), and several cancer-derived cell lines have mutations in the ␣-catenin gene. Reintroduction of ␣-catenin into such lines reduces cell growth (18) and attenuates tumor formation (19). Furthermore, DLD-1 cells that have lost ␣-catenin expression are more invasive than the parental cell line (44). An important question that will be addressed in the future is whether the increased invasiveness of DLD ␣Ϫ cells is due to a reduction in cell adhesion or to an increase in ␤-catenin⅐Tcf transcription activity. Interestingly, in a recent study of samples from colorectal carcinoma patients (45), many of the tumors examined had reduced expression of either E-cadherin (29%) or ␣-catenin (56%), but increased tumor cell invasion and metastasis correlated with reduced expression of ␣-catenin. This suggests that ␣-catenin may function in an adhesionindependent manner to regulate invasion and metastasis. All the results presented in this report were obtained using cultured cells, and therefore, further analysis will be required to determine whether similar mechanisms operate in vivo. If this proves to be the case and if the effects of ␣-catenin on cell adhesion and transcriptional activity are separable, the targeted expression of ␣-catenin to the nucleus may be a useful approach for treating cancer cells that have activating mutations in the Wnt signaling pathway.