APC/β-catenin-rich complexes at membrane protrusions regulate mammary tumor cell migration and mesenchymal morphology

The adenomatous polyposis coli (APC) tumor suppressor gene was initially identified as the gene mutated in familial adenomatous polyposis (FAP), a rare inherited colorectal cancer syndrome that predisposes carriers to hundreds to thousands of adenomatous polyps and aggressive, early-onset colon cancer [1‐3]. Identification of APC mutations and loss of heterozygosity (LOH) in approximately half of sporadic colorectal adenomas and majority of adenocarcinomas indicates that APC inactivation is an early event in tumor progression (i.e. initiation). It is now appreciated that APC mutation is relatively rare in other solid tumors but it is silenced by gene methylation in multiple tumor types, including breast cancer (reviewed in [4]). Recent studies from our laboratory demonstrated that APC is required for maintaining epithelial homeostasis in the mouse mammary gland and its mutation cooperates with other oncogenic alterations, such as overexpression of the Polyoma Middle T antigen (PyMT) oncogene [5, 6], to enhance primary mammary tumorigenesis. These studies have indicated that the effects of Apc mutation in both normal mammary homeostasis and mammary tumor development are not solely dependent on Wnt pathway regulation, which is APC’s best-characterized molecular activity. Through de-repression of the canonical Wnt pathway, APC loss results in β-catenin stabilization, accumulation and nuclear translocation where it associates with TCF/LEF transcription factors to regulate Wnt target genes. In addition to inducing tumor cell proliferation, Wnt target genes also promote cell survival, stem cell self-renewal, and matrix remodeling [7]. Yet, emerging data like those in the mammary models described above indicate that the consequences of APC inactivation are not solely restricted to Wnt pathway activation.

Among the functions that have been attributed to APC independent of regulating Wnt signaling are mediating genomic stability, apoptosis, DNA repair, proliferation, and apical-basolateral and front-rear polarity (reviewed in [8, 9]). Robust microtubule-dependent localization of endogenous and overexpressed APC to cell protrusion ends in migrating cells [10‐13] suggests that this major pool of APC is likely to be involved in guiding front-rear cell polarity and motility. In fact, APC is required for directed motility of astrocytes [14, 15]. Despite these findings, it is not clear how this pool of APC, and its role at the leading edge and at membrane protrusions in migration, contributes to tumor suppression mediated by APC and drives tumor initiation or progression when APC is mutated. In the current study, we have characterized APC/β-catenin complexes at membrane protrusions in epithelial cells undergoing an epithelial-to-mesenchymal transition (EMT) and in human and mouse breast cancer cells. Using an in vitro model of APC depletion in the 4T07 mouse mammary tumor cells, we demonstrate that disruption of these complexes inhibits tumor cell migration and disrupt mesenchymal cell morphology. These results indicate that invasive tumor cell behavior is dependent on APC/β-catenin complexes at membrane extensions and suggests that perturbation of these complexes in motile tumor cells via APC inactivation may significantly impact tumor cell dissemination and colonization.

To examine whether endogenous β-catenin localized to membrane protrusions like APC [10] and exogenous mutant β-catenin [20], β-catenin was stabilized with the proteasome inhibitor ALLN in MDCK kidney epithelial cells. ALLN treatment resulted in altered cell morphology compared to vehicle controls; specifically, treated cells appeared flatter and more spread and had abundant long, thin membrane processes (Figure 1A). Treatment with the proteasome inhibitor also significantly increased expression of total β-catenin protein by western blotting over a range of concentrations (Figure 1B). To further implicate β-catenin in these phenotypes, β-catenin was knocked down in MDCK cells using siRNAs. Staining of the cells with phalloidin and anti-β-catenin antibody by IF confirmed that the compact epithelial morphology and tight cell-cell interactions of MDCK cells were perturbed by β-catenin silencing in the absence of ALLN (Figure 1C). β-catenin down-regulation abrogated the effect of ALLN on protrusion formation compared to a non-targeting control siRNA (Figure 1C).

To characterize the cellular processes that formed following stabilization of endogenous β-catenin, IF for β-catenin was performed on ALLN-treated MDCK cells. In control cells, the localization of β-catenin was restricted to the membrane at sites of cell-cell contact as expected, but stabilized endogenous β-catenin was at the membrane, heterogeneous in nuclei, and, unexpectedly, prominently localized at the tips of long membrane protrusions in ALLN-treated cells (Figure 1D). Consistent with previous observations [10, 21], APC was localized to the leading edge and in clusters at the ends of membrane protrusions of subconfluent MDCK cells (Figure 1D). In cells treated with ALLN, β-catenin co-localized with APC at the tips of membrane protrusions (Figure 1D). To explore whether there were other circumstances in which β-catenin co-localized with APC at protrusion ends in epithelial cells, we analyzed the well-characterized model of TGFβ-induced EMT in MDCK cells. Unlike untreated MDCK cells, upon TGFβ exposure, β-catenin localization was evident at the tips of membrane extensions, where it co-localized with APC (Figure 1E). These data indicate that in epithelial cells in which endogenous β-catenin is stabilized or that are undergoing EMT, β-catenin shifts from an exclusively membrane-associated localization to an additional complex at the ends of membrane protrusions where it co-localizes with the APC tumor suppressor.

β-catenin stabilization and subsequent Wnt pathway activation, as well as EMT, are associated with the progression of many tumor types, including breast cancer. Therefore, we next explored whether APC/β-catenin complexes at protrusion ends were found in breast cancer cells. For this, we turned to a progression series set of mouse mammary cell lines, all derived from the same genetic background, that range from non-transformed (EpH4) to tumorigenic but non-metastatic (67NR) to highly invasive (4T07) [18]. Like MDCK cells, nontransformed EpH4 mouse mammary epithelial cells have β-catenin restricted to the lateral membrane at cell-cell contacts and co-localized with E-cadherin, whereas, APC is in a punctate localization pattern at the few membrane protrusions these cells have (Figure 1E). The membrane-associated pool of β-catenin was maintained in 67NR cells, and compared to EpH4, these cells had a more elongated appearance with decreased cell-cell interactions and obvious membrane protrusions; yet, neither β-catenin nor APC was associated with these protrusions (Figure 1E). 4T07 cells exhibited a spindle-like morphology and lacked obvious cell-cell contacts with E-cadherin or β-catenin membrane localization. However, β-catenin and APC were prominently localized at the ends of the long membrane protrusions in 4T07 cells (Figure 1E). Human breast cancer cells, including the ER-negative Hs578T line, showed the identical pattern of APC/β-catenin co-localization at protrusion ends (Figure 1F).

To determine the importance of APC/β-catenin-containing complexes at the protrusions in mammary tumor cell behavior, we stably suppressed APC expression using lentiviral-mediated expression of APC shRNAs. We first introduced GFP into the 4T07 cells for visualization and then generated stable 4T07 pools that expressed one of two different APC-specific shRNAs, a scrambled shRNA (SHC-002) or empty vector (pLKO.1). The expression of APC mRNA in the control and APC shRNA-infected cell lines was analyzed by qRT-PCR, and we found that APC expression was significantly reduced by 70-90% in the APC shRNA-infected cells relative to the pLKO.1-infected cells (Figure 2A; p < 0.01). Total cell lysates from each cell line were analyzed by western blotting with an APC antibody and show that there is a significant reduction in full-length APC protein expression upon knockdown (64-87% depending on the construct) (Figure 2B), consistent with the mRNA levels. The cells were then analyzed for expression and localization of APC protein by IF (Figure 2C). Both APC shRNA cell lines had a lower intensity of staining compared with the control lines, and the specific localization of APC at cell protrusion ends was absent. Additionally, the spindle-shaped morphology of the cells was attenuated, and the cells displayed a slightly more rounded shape. Suppression of APC expression and the morphological change was concomitant with a loss of β-catenin also at protrusion ends (Figure 2C). Despite the fact that APC loss is often associated with β-catenin stabilization and nuclear accumulation, this was not observed in the APC-knockdown 4T07 cells, either by IF or western blotting (Figure 2C and data not shown). Consistently, there was no induction of β-catenin/TCF-mediated transcriptional activity as determined by a TOPflash reporter assay (Figure 2D). These data are also consistent with our previous observations that heterozygous Apc mutation (in vivo) or APC knockdown (in vitro) was insufficient for hyperactivation of canonical Wnt signaling in mammary epithelial cells [5]. Moreover, there was not a statistically significant difference in BrdU incorporation, a measure of S-phase entry, between control and APC-knockdown cell lines, although there was a trend to increased proliferation in the APC shRNA1 cells (Figure 2E). Since APC has been shown to control the G₁/S transition by regulating the β-catenin/TCF transcriptional targets c-myc and cyclin D1 [22], these results suggest that APC suppression in these invasive breast cancer cells impacts neither β-catenin-mediated transcription nor proliferation.

The localization of APC/β-catenin complexes to protrusion ends and the subtle disruption of the spindle-shaped morphology of 4T07 cells with APC knockdown suggests that these complexes may be important in tumor cell motility and invasion, particularly in response to a stimulus such as wounding or growth factors/cytokine signaling. Wound-filling assays were performed in the presence of the cell cycle inhibitor mitomycin C in control and APC-knockdown cells. Images were acquired over the time-course (up to 36 h), and representative phase-contrast images are shown, demonstrating that the cells with APC knockdown had an impaired ability to fill in the wound compared to control cells (Figure 3A). Control cells at the wound edge had a spindle-shaped morphology with long membrane protrusions, but very few of the APC-knockdown cells had this morphology. Quantification of the unfilled area of the wound at each time point (0, 12, 24, and 36 h post-wounding) indicated that there was a statistically significant larger wound area in the APC-knockdown cell lines at all time points compared to the vector-infected control cells (Figure 3B). Because of this effect and the morphological changes in the APC-knockdown cells noted earlier, we examined tumor cell morphology more thoroughly. Control and APC-knockdown 4T07 cells were plated at subconfluent density and stained with phalloidin so that cell protrusions could be readily visualized. The morphology of the cells was characterized as having two protrusions (bipolar), typical of parental 4T07 cells (Figure 1E), or rounded. With knockdown observed a significant decrease in the bipolar morphology and concomitant increase in the rounded phenotype (Figure 3C-D). Additionally, more frequent cell-cell contacts and generally blunted cell protrusions were observed in the cells with APC knockdown (Figure 3D).

Given that APC associates with the plus ends of microtubules, binds to other plus end-binding proteins, and its localization to cortical membrane clusters is microtubule-dependent [10, 21, 23‐27], we next determined whether protrusion formation and APC/β-catenin complexes required intact microtubules. Incubation of MDCK cells in nocodazole attenuated the ALLN-induced phenotype regardless of the order of ALLN and nocodazole treatment (Figure 4A), and quantification of protrusions confirmed these observations (Figure 4B). Importantly, treatment of 4T07 cells with nocodazole over a 4–12 h time frame inhibited the localization of both APC and β-catenin at protrusion ends by 4 h and significantly abrogated protrusion formation by 9–12 h (Figure 4C). Additionally, disruption of microtubules in these cells promoted cell-cell interactions and β-catenin localization to the cell-cell contacts (Figure 4C, arrows). In fact, this phenotype observed with nocodazole treatment was reminiscent of the dramatic change in morphology and β-catenin distribution that we observed with acute knockdown of APC expression using transient expression of APC-specific siRNAs in 4T07 cells (Figure 4D). It should be noted that background staining is frequently observed in APC-knockdown cells upon staining with an anti-APC antibody which could reflect non-specific immunoreactivity of the antibody or residual endogenous APC expression, although the APC localization at protrusion-associated puncta is consistently absent upon APC silencing. Collectively, these data suggest that both APC and intact microtubules are required for β-catenin localization at protrusion ends, the spindle-shaped, mesenchymal morphology and membrane protrusions that are important for tumor cell migration.

While loss of the APC tumor suppressor has been demonstrated in numerous studies to facilitate tumor initiation by activating Wnt signaling and driving proliferation, a role for APC in later stages of tumor progression, including morphological changes and tumor cell migration, has not been clearly demonstrated. Moreover, the importance of the distinct subcellular pools of β-catenin at the lateral membrane and nucleus in tumor development have been well documented; yet, the presence of exogenous mutant β-catenin at protrusion ends suggests that this may represent a fraction of β-catenin with a distinct activity. In this study, we have described APC/β-catenin-rich clusters at protrusion ends in multiple models, including 1) epithelial cells in which endogenous β-catenin has been stabilized, 2) epithelial cells undergoing EMT and 3) mouse and human breast tumor cell lines, all of which have a mesenchymal morphology with robust membrane protrusions. Using RNA interference via lentiviral-mediated delivery of shRNAs, we demonstrated that APC is required for maintaining β-catenin localization to the protrusion tips and the mesenchymal phenotype of the 4T07 mouse mammary tumor cell line but does not significantly affect proliferation or β-catenin/TCF-dependent transcriptional activation. Importantly, the migratory behavior of these tumor cells was also dependent on APC.

Several previous studies have demonstrated localization of APC to the leading edge or membrane protrusions of migratory non-transformed cells, including subconfluent epithelial cells, fibroblasts, astrocytes, neurons and endothelial cells [10, 14, 23, 27‐29]. As far as we are aware, this work is the first to address the role of these protrusion-associated APC/β-catenin complexes in tumor cells. In these other contexts, APC localizes to the plus-ends of microtubules and the localization of APC to membrane protrusions is microtubule- but not actin-dependent [14, 15, 25]. This association of APC with microtubule plus ends involves a direct interaction of APC’s carboxy-terminal basic domain with tubulin and indirect association through the plus-end microtubule binding protein, EB1, and the KAP3/Kif3 plus-end-directed motor proteins [30‐32]. Recent studies suggest that microtubule-associated APC transports RNA molecules to protrusion ends in fibroblasts [33]. Our observation that APC suppression reduces tumor cell migration is consistent with functional studies indicating that APC is required for directed cell motility in astrocytes [14, 15], as well as protrusion formation and wound filling in fibroblasts and osteosarcoma cells [11]. In the latter model systems, this activity for APC was linked to changes in microtubule stability [11]. These data are consistent with our observations in mammary tumor cells that microtubule integrity is required for membrane protrusions driven by APC/β-catenin-complexes and that APC suppression mimics the morphological defects induced by the microtubule de-stabilizing drug, nocodazole. Because of limitations of our work that include primarily providing functional data with one tumor cell line and characterizing in vitro tumor cell behavior, it will be important to extend these observations to other tumor models and in vivo studies to more fully dissect APC molecular action in cell motility.

In vivo studies have also demonstrated that APC loss or mutation impedes enterocyte migration along the crypt-villus axis [34, 35] and forced overexpression of APC results in disordered, non-adhesive intestinal cell migration, presumably by changing the relative activity of cadherin-catenin complexes [36]. Additional support for our observation that APC controls cell-cell interactions is provided by studies in APC-mutant colorectal cancer cells in which APC re-introduction induced β-catenin/E-cadherin localization at the adherens junction and promoted a more epithelial, less motile phenotype [37]. It is likely that inconsistencies in the net impact on migration with APC loss or overexpression between these models and in vitro studies are due to cell-type specific activities, analyzing specific APC mutations vs. loss vs. ectopic expression, differences between studying normal and tumor cells, and the contribution of tissue architecture and substratum in guiding epithelial cell migration. It is also interesting that our 4T07 cells with the greatest extent of APC knockdown (APC shRNA1) had a more modest effect on migration but slightly more robust impact on morphology. Because these are antibiotic-resistant pools upon lentiviral infection, the populations are heterogeneous and may have various levels of APC from cell to cell at a wound edge for example, a concern could be mitigated by subcloning. It is also possible that these differences reflect a biologically significant threshold or local concentration of APC necessary for protrusion formation, cell motility and cell-cell interactions. However, we are confident in the specificity of the knockdown with both shRNAs because the same phenotypes are observed with multiple, independent shRNAs, and APC expression and defects in other cell lines associated with these constructs have been rescued with ectopic expression of a silencing-resistant human full-length APC cDNA (JRP and KHG, manuscript in preparation).

The data presented here indicate that the observed effects of APC depletion are independent of Wnt signaling activation, since β-catenin/TCF transcriptional activity is not detected in APC-knockdown 4T07 cells. Despite the fact that Wnt signaling has been implicated in promoting EMT and tumor cell motility and invasion [38‐41], our data support a separation of these activities in this model system. Additionally, these Wnt-independent effects of APC in 4T07 mammary tumor cells are consistent with our previous characterization of APC-mediated regulation of epithelial polarization and tissue architecture and tumorigenesis in the mammary gland [5, 6] that do not involve Wnt pathway hyperactivation, suggesting that Wnt-independent APC activities may be context- or tissue-specific. We cannot rule out that there is redundant action of APC2, for example, in the APC-knockdown cells sufficient for restricting Wnt pathway activation. In any case, a direct effect of APC on migration or invasion may reflect that its regulation of membrane protrusions is required for both migration and invasion or may involve the association of APC with the actin cytoskeleton, for example. Unfortunately, we were unable to separate a contribution of APC specifically to tumor cell invasion from its impact on motility (M.O and K.H.G., unpublished observations). Although APC localization to protrusion ends is not actin-dependent, APC binds directly to actin and promotes its nucleation [42, 43] and alters actin dynamics through its interactions with IQGAP, an effector of Rac and Cdc42 [13], Asef1 and Asef2, Rac-specific guanine-exchange factors [44‐46] and mDia, a Rho GTPase effector [27]. Moreover, recent data from our lab and others demonstrate that APC mutation result in dramatic changes in cell-matrix signaling pathways, including focal adhesion kinase (FAK) and Src [6, 47]. Collectively, these findings suggest that APC inactivation in tumor cells may elicit widespread changes in cytoskeletal dynamics and integrin signaling to modify tumor cell behavior.

Our results demonstrating that stabilized endogenous β-catenin localizes to APC clusters at membrane extensions are generally consistent with studies in which an exogenous stabilized mutant β-catenin localized to MDCK protrusions [20] but also extends these findings to tumor models. Recent studies indicated that, in MDCK cells, phosphorylation of β-catenin at its amino-terminus leads to its localization to cell protrusion ends as well as adherens junctions and that the pool at protrusion ends associates with APC and is phosphorylated by GSK3β but is not degraded [48]. Phosphorylated β-catenin is also induced by proteasome inhibition in 293T cells [49]. Also consistent with our results, Faux et al. demonstrated that protrusion-associated β-catenin complexes required APC [48]. In endothelial cells, phosphorylation of APC by GSK3β and CK1 is required for APC and phosphorylated β-catenin to be associated with microtubules in clusters at protrusion ends and for cell migration [28]. Moreover, these clusters are thought to have a high-rate of turnover via the proteasome; whereas a more stable pool of phosphorylated APC and phosphorylated β-catenin is found at the lateral membrane [28], suggesting that complex turnover and post-translation modifications may control the dynamics of distinct APC and β-catenin subcellular pools. Other components of these APC/β-catenin clusters may include the APC binding partners Dlg and Scribble, which have been associated with APC at protrusion ends in astrocytes [15, 50] and epithelial cells ([12]; KHG, unpublished results). A detailed analysis to identify all of the complex constituents and the mechanisms that control complex dynamics has not yet been performed, but will be necessary to uncover a clear molecular picture of the membrane-protrusion-associated APC/β-catenin complex in these in vitro model systems. The next challenge will be applying this knowledge to elucidate the physiological relevance of these protrusion-associated complexes to normal tissue homeostasis and tumor initiation and progression in vivo.

These results also suggest that APC loss late in tumor progression may impact the invasive or metastatic tumor phenotype. Spindle-shaped 4T07 mammary tumor cells are capable of forming primary tumors, intravasating, surviving the blood stream, and extravasating into lungs, but they fail to efficiently colonize to form distant metastases (reviewed in [51]). Depletion of APC in these cells suppressed tumor cell migration but did not significantly impact proliferation. One prediction from these data is that, in contrast to exclusively driving tumor initiation as known to do in colorectal cancer for example, APC loss would actually prevent primary tumor invasion or metastasis to a distant site. However, the consequence of APC inactivation might be dependent on the timing of loss. For example, if APC is mutated or down-regulated after tumor cell extravasation, it may actually promote tumor cell colonization by promoting cell-cell or cell-matrix interactions. Although neither possibility has yet been tested experimentally, recent studies indicate that APC gene silencing by methylation is associated with reduced disease-free overall survival in breast cancer patients [52], serum APC levels are associated with colorectal cancer metastases [53] and APC somatic mutations and LOH is observed frequently in brain metastases from lung and colorectal primary tumors [54, 55]. It will also be important to determine if there are distinct consequences of APC mutation and down-regulation on cell motility and tumor progression, which differ in frequency among tumor types. This is likely to be the case because mutant APC appears to have a gain-of-function binding to the Asef proteins that control tumor cell migration [44‐46]. Careful comparison of APC mutation and silencing in well-defined model systems is required to address this issue. However, our findings support the concept that the timing of APC loss during tumor progression may be important in dictating the biological consequences and suggest that APC, and its downstream effectors, might even represent novel therapeutic targets to modulate tumor metastasis.