Cdc42-interacting protein 4 is a Src substrate that regulates invadopodia and invasiveness of breast tumors by promoting MT1-MMP endocytosis

Tumor metastasis requires a subset of tumor cells to acquire the ability to break through the basement membrane and invade through dense networks of interstitial extracellular matrix (ECM) proteins (Yamaguchi et al., 2005a). The ability of invasive tumor cells to degrade ECM and migrate through tissues is promoted by actin-rich membrane protrusions called invadopodia (Gimona et al., 2008; Buccione et al., 2009). Invadopodia are finger-like projections of the ventral membrane that are supported by a core of filamentous actin (F-actin), much like podosomes in macrophages (Linder, 2007). Podosomes and invadopodia promote ECM degradation by the recruitment of transmembrane type I matrix metalloprotease (MT1-MMP) from secretory vesicles, and by localized activation of secreted MMP-2 and MMP-9 (Seiki, 2003; Steffen et al., 2008). In addition to promoting invasion via ECM degradation, invadopodia also function in directional cell migration of breast tumor cells towards epidermal growth factor (EGF) (Desmarais et al., 2009). The role of invadopodia in tumor metastasis in vivo is less clear, but increased expression of EGF receptor (EGFR) has been shown to correlate with increased spontaneous lung metastases in breast tumor cells (Xue et al., 2006; Le Devedec et al., 2009). Furthermore, upregulation of EGFR and numerous regulators of actin assembly, invadopodia and cell motility comprise a gene expression signature of invasive breast tumor cells (Wang et al., 2004). However, in experimental metastasis assays, invadopodia were not required for breast tumor metastases to the lung, but did promote tumor progression via enhancing growth of tumor vasculature (Blouw et al., 2008). Thus, regulators of invadopodia are potential therapeutic targets for reducing tumor metastasis and cancer morbidity in humans.

Formation of invadopodia is regulated by the impact of signaling pathways on lipid metabolism, vesicle trafficking and actin regulatory proteins (Gimona et al., 2008; Poincloux et al., 2009). In breast tumor cells, EGF treatment leads to rapid formation of invadopodia following a burst of actin polymerization from free barbed ends induced by the actin-severing protein cofilin (Oser et al., 2009). A regulatory cycle of EGF-induced cortactin tyrosine phosphorylation was shown to control its interaction with cofilin and its F-actin-severing activity (Oser et al., 2009). Invadopodia are also induced by expression of activated Src protein-tyrosine kinase (Src-Y527F) in breast tumor cells (Artym et al., 2006). Compared with EGF, Src is a more potent inducer of invadopodia and ECM digestion (Oser et al., 2009), and is a key regulator of breast tumor growth and metastases in mouse models (Myoui et al., 2003; Rucci et al., 2006).

Src is also a potent inducer of podosomes in normal cells, including fibroblasts (Chen, 1989), osteoclasts (Destaing et al., 2008) and vascular smooth muscle cells (Zhou et al., 2006; Furmaniak-Kazmierczak et al., 2007). Src directly phosphorylates many key regulators of invadopodia precursor formation, including cortactin (Bowden et al., 2006), and the adaptor protein Tks5 (Seals et al., 2005; Blouw et al., 2008). Src-induced podosome precursors form in areas of membrane with elevated levels of phosphatidylinositol-(3,4)-bisphospate [PtdIns(3,4)P₂] that recruits the Tks5 adaptor along with N-WASP, which binds the SH3 domains of Tks5 (Oikawa et al., 2008). Tks5 also recruits Nck adaptor, which binds N-WASP to trigger actin assembly (Stylli et al., 2009). The finger-like membrane protrusions within podosomes and invadopodia contain inverse-BAR domain proteins IRSp53 and IRTKS (Li et al., 2010), which also drive formation of membrane protrusions in filopodia (Mattila and Lappalainen, 2008). These membrane protrusions are stabilized by rapid polymerization of branched and bundled F-actin from barbed ends generated by cofilin (Oser et al., 2009). The Cdc42–N-WASP–Arp2/3 pathway promotes branched actin assembly within invadopodia (Yamaguchi et al., 2005b). Invadopodia formation also depends on polymerization of unbranched F-actin by Diaphanous-related formins (Lizarraga et al., 2009), and the actin bundling protein fascin, which stabilizes invadopodia and promotes invasion of tumor cells (Li et al., 2010). Maturation of invadopodia into ECM-degrading structures requires the recruitment of MT1-MMP from secretory vesicles, and activation of the secreted proteases MMP-2 and MMP-9 (Seiki, 2003; Artym et al., 2006; Steffen et al., 2008; Poincloux et al., 2009). Surface expression of MT1-MMP within invadopodia is further regulated by endocytosis, which is blocked by Src-induced tyrosine phosphorylation of endophilin A2, which disrupts dynamin binding and membrane recruitment (Wu et al., 2005).

CIP4 (also known as TRIP10) is an adaptor protein that was identified as a potential Src SH3-domain binding partner and substrate (Lock et al., 1998; Dombrosky-Ferlan et al., 2003). CIP4 also interacts with known regulators of invadopodia, including Cdc42 (Aspenstrom, 1997), N-WASP (Tian et al., 2000), dynamin (Itoh et al., 2005), and Diaphanous-related formins (Aspenstrom et al., 2006a). CIP4 is the founding member of the CIP4 subfamily of Fer–CIP4 homology Bin–Amphiphysin–Rvs (F-BAR) proteins, which are defined by their conserved N-terminal F-BAR domains (Aspenstrom et al., 2006b; Chitu and Stanley, 2007; Heath and Insall, 2008). Two CIP4-related adaptor proteins Toca-1 (Ho et al., 2004) and FBP17 (Kamioka et al., 2004) share a similar domain organization consisting of F-BAR, protein-kinase-C-related kinase homology region 1 (HR1), and SH3 domains. F-BAR domains form crescent-haped dimers that bind membrane phospholipids through clusters of basic residues on the concave face (Henne et al., 2007; Shimada et al., 2007; Frost et al., 2008). F-BAR proteins participate in clathrin-mediated endocytosis by promoting membrane invagination, actin assembly and vesicle scission (Suetsugu et al., 2010). In addition to plasma membranes, the F-BAR domain of CIP4 also confers localization to Rab5 endosomes (Hu et al., 2009). The HR1 domain of CIP4 binds GTP-loaded Cdc42 (Aspenstrom, 1997; Kobashigawa et al., 2009), whereas the SH3 domain interacts with numerous ligands involved in actin assembly and endocytosis (Chitu and Stanley, 2007).

The physiological functions of CIP4 are beginning to emerge, based on RNA interference and gene-knockout studies. CIP4-knockout mice are viable, but have altered glucose metabolism (Feng et al., 2010). This is consistent with cell-based studies implicating CIP4 in regulating both the trafficking of GLUT4 storage vesicles to the plasma membrane (Lodhi et al., 2007), as well as GLUT4 endocytosis via CIP4 interactions with N-WASP and dynamin-2 (Hartig et al., 2009). CIP4 has also been implicated in positioning of the microtubule-organizing center for directed exocytosis of cytolytic granules in natural killer (NK) cells (Banerjee et al., 2007), although a recent study of CIP4-knockout mice showed normal NK cell function (Koduru et al., 2010). However, this study also shows that CIP4 promotes integrin-mediated recruitment of effector T cells to sites of cutaneous inflammation (Koduru et al., 2010). CIP4 also functions in endosome to lysosome trafficking of EGFR (Hu et al., 2009). The Drosophila CIP4 orthologue functions in the Cdc42–WASP pathway, which controls E-cadherin endocytosis and actin-based vesicle motility downstream of both WASP and WAVE (Leibfried et al., 2008; Fricke et al., 2009). Using microinjection of CIP4 fragments into primary macrophages, an early study showed that CIP4 can regulate formation of podosomes (Linder et al., 2000). This study showed that injection of CIP4 fragments lacking WASP or microtubule-binding sites could block podosome formation, possibly by disrupting localization of WASP to podosomes. A recent study showed that FBP17 localizes to podosomes in monocytes and promotes actin assembly and podosome formation (Tsuboi et al., 2009). Taken together, these results point to a potential role for CIP4 and related F-BAR proteins as potential regulators of invadopodia, owing to their similarity with podosomes.

In this study, we show that CIP4 is a Src substrate in MDA-MB-231 breast tumor cells. Furthermore, we demonstrate that Src induces localization of CIP4 to invadopodia. Stable shRNA-driven CIP4 knockdown in MDA-MB-231 cells and a derivative expressing activated Src, results in increased numbers of invadopodia and matrix degradation. This is due to increased surface levels of MT1-MMP, as a result of reduced MT1-MMP endocytosis in CIP4-knockdown cells. Consistently, CIP4-knockdown cells are significantly more invasive through matrigel, compared with control cells. These results implicate CIP4 as a suppressor of Src-induced invadopodia formation and invasiveness of human breast tumor cells.

Src protein-tyrosine kinase is a key regulator of breast tumor progression, invasion and metastasis (Myoui et al., 2003; Rucci et al., 2006). Although CIP4 was identified as a potential Src substrate (Lock et al., 1998; Dombrosky-Ferlan et al., 2003), the potential involvement of CIP4 in Src signaling in breast tumor cells has not been reported. To test this, MDA-MB-231 cells were transduced with either a control retrovirus (MDA-vec) or a retrovirus expressing activated Src-Y527F (MDA-SrcYF). Consistent with previous studies (Artym et al., 2006; Oser et al., 2009), MDA-SrcYF cells formed more invadopodia than MDA-vec cells, as marked by F-actin dots over areas of ECM digestion (supplementary material Fig. S1). As expected, MDA-SrcYF cells displayed elevated levels of Src phosphorylation on the activating Y416 residue, compared with levels in MDA-vec cells (Fig. 1A, top panel). Although the total Src levels appeared similar (Fig. 1A, second panel; 1.5-fold increase in MDA-SrcYF), the avian SrcYF protein was poorly recognized by this pan-Src antibody (data not shown; but did crossreact with pY416-Src antisera). Next, we tested the effects of Src-Y527F expression on tyrosine phosphorylation of CIP4. Although endogenous CIP4 tyrosine phosphorylation was detected in MDA-vec cell lysates, this was increased dramatically in MDA-SrcYF cell lysates (Fig. 1A, third panel). This was not due to altered levels of CIP4 expression or recovery between cell lines (Fig. 1A, bottom panel). To test further whether CIP4 is a Src substrate in MDA-MB-231 cells, we treated MDA-SrcYF cells with Src inhibitors PP2 or SU6656, or with vehicle control (DMSO). We found that pretreatment of MDA-SrcYF cells with either Src inhibitor, caused a dramatic reduction in CIP4 tyrosine phosphorylation (Fig. 1B). This correlated with the reduced levels of activated Src detected in cell lysates pretreated with Src inhibitor (Fig. 1B, bottom panel). Quantification of results from several experiments revealed a ~threefold increase in CIP4 tyrosine phosphorylation levels in MDA-SrcYF cells compared with MDA-vec cells, which was dependent on Src activity (Fig. 1C). Taken together, our results demonstrate that CIP4 is part of the Src signaling axis in MDA-MB-231 breast tumor cells.

Src substrates such as Tks5 undergo a dramatic change in subcellular localization in response to expression of activated Src (Abram et al., 2003). To test whether Src also affects CIP4 localization in breast tumor cells, we examined the subcellular localization of endogenous CIP4 by immunofluorescence (IF) and epifluorescence microscopy. In MDA-vec cells, CIP4 showed a punctate localization throughout the cytosol, with highest levels in the perinuclear region (Fig. 2A). Invadopodia could be identified by F-actin dots that form in the perinuclear region of MDA-MB-231 cells (Artym et al., 2006). We observed some F-actin dots in the perinuclear region of the same cells that partially colocalized with CIP4 (Fig. 2A). This is consistent with localization of CIP4 to invadopodia in MDA-MB-231 cells. However, owing to the diffuse nature of CIP4 localization within the cytosol, we tested whether CIP4 localized to SrcYF-induced invadopodia, which form in clusters at the cell periphery (Artym et al., 2006). Indeed, CIP4 localized more at the cell periphery in MDA-SrcYF cells, within clusters of invadopodia marked by F-actin and tyrosine-phosphorylated cortactin (pY-Cort; Fig. 2B).

To gain more insight into CIP4 localization to invadopodia that are actively engaged in ECM digestion, we performed 3D confocal imaging of endogenous CIP4 and pY-Cort in MDA-SrcYF cells plated on fluorescent ECM. We observed similar colocalization of CIP4 and pY-Cort at invadopodia formed at the cell periphery (Fig. 3A, see arrows). The orthogonal view of this region of the cell shows CIP4 localization to several invadopodia at various stages of protruding into the ECM. CIP4 overlapped with the upper portion of the F-actin core marked by pY-Cort (and F-actin) that is within the dorsal cell protrusion within the ECM (Fig. 3A, see x-z panel below). In some invadopodia, CIP4 localized to the tip of the protrusion (Fig. 3A, see y-z panel on the right). Similar results were found in MDA-SrcYF cells transfected with GFP–CIP4: GFP–CIP4 exhibited extensive colocalization with F-actin dots (Fig. 3B, see x-y axes) and F-actin columns at the cell periphery (Fig. 3B, lower panels). Taken together, these results provide evidence that CIP4 localizes to Src-induced invadopodia in breast tumor cells.

To test the function of CIP4 in breast tumor cells, and the potential for CIP4 to regulate Src-induced invadopodia formation, we generated MDA-SrcYF cells with stable shRNA-driven CIP4 knockdown. To test for effects of CIP4 KD on Src-induced invadopodia, MDA-SrcYF-vector control cells and CIP4-knockdown (KD) cells were plated on TRITC-labelled fibronectin (ECM) for 18 hours, before staining for F-actin. Compared with control MDA-SrcYF cells, both CIP4 KD cell pools showed more extensive degradation of ECM (Fig. 4A). Quantification of ECM degradation area per cell revealed a ~2.5-fold increase in CIP4 KD cells, compared with the control (Fig. 4B, asterisks indicate significant differences from vector control cells, P<0.01). The extent of CIP4 KD was tested by immunoblot with CIP4 antisera which revealed a 70–80% reduction in CIP4 levels in CIP4 KD cells, compared with vector control cells (Fig. 4C; ERK served as a loading control). Also, CIP4 KD did not alter levels of the other CIP4 subfamily proteins, Toca-1 and FBP17 (Fig. 4C). The consistent effects observed with two shRNAs against CIP4 in terms of knockdown and ECM digestion, suggested that the phenotype was not due to off-target effects. However, we further tested this by rescue experiments in which human shRNA-resistant or mouse GFP–CIP4 constructs were transiently transfected in CIP4 KD MDA-SrcYF cells. We examined fields of view showing GFP-positive and -negative cells (marked by F-actin) with visible ECM digestion (Fig. 4D). Quantification of the ECM degradation area per cell revealed that compared with expression of GFP, which had no effect on ECM digestion, CIP4 KD cells expressing GFP–CIP4 (mouse or human) showed >60% decrease in ECM digestion (Fig. 4E). These results suggest that CIP4 is a negative regulator of Src-induced ECM digestion by invadopodia.

To test whether knockdown of CIP4 altered the numbers of invadopodia, we compared the numbers of Src-induced invadopodia marked by pY-Cort and F-actin dots over areas of ECM degradation (Fig. 5A). CIP4 KD cells formed extensive clusters of invadopodia at the cell periphery that were actively degrading ECM, again to a greater extent than the vector control cells (Fig. 5A). Scoring of invadopodia for multiple fields revealed a ~twofold increase in the average number of invadopodia in both CIP4 KD MDA-SrcYF cell lines, compared with the control (Fig. 5B, asterisks indicate significant differences from vector control cells, P<0.01). These results suggest that CIP4 is a negative regulator of Src-induced invadopodia formation and ECM degradation in breast tumor cells.

Recent studies have shown that EGF signaling leads to formation of invadopodia in MTLn3 rat mammary tumor cells (Oser et al., 2009). To investigate whether EGF could induce invadopodia in MDA-MB-231 cells and determine the involvement of CIP4, we generated a stable CIP4 KD cell pool from MDA-MB-231 cells. Immunoblotting with CIP4 antisera revealed a >80% reduction in CIP4 levels, compared with the vector control (Fig. 5C). Immunoblotting with EGFR antisera revealed no significant difference in EGFR levels upon knockdown of CIP4 (data not shown). Next, we plated MDA-MB-231 vector and CIP4 KD cells on fluorescent ECM in the presence of EGF (Fig. 5D). Although the extent of ECM digestion was modest in comparison to MDA-SrcYF cells, we observed invadopodia marked by F-actin and pY-Cort in both vector and CIP4 KD cells. Quantification revealed an increase the average number of invadopodia per cell in EGF-treated CIP4 KD cells compared with the vector control (Fig. 5E). Likewise, ECM degradation in CIP4 KD cells was elevated (~2.5-fold) compared with vector control cells (Fig. 5F). Taken together, these results suggest that CIP4 is a negative regulator of Src- and EGF-induced invadopodia formation and ECM degradation in breast tumor cells.

MT1-MMP is required for invadopodia formation, and is a marker of mature invadopodia in breast tumor cells (Artym et al., 2006). To further define how CIP4 regulates invadopodia formation, we tested MT1-MMP localization relative to the invadopodia markers F-actin and pY-Cort in MDA-SrcYF control cells and CIP4 KD cells. In CIP4 KD cells (shRNA2), MT1-MMP colocalized extensively with F-actin and pY-Cort at the cell periphery (Fig. 6A). Quantification of these results for several fields revealed a significant increase in MT1-MMP puncta colocalized with F-actin, pY-Cort, or both markers, in CIP4 KD MDA-SrcYF cells, compared with control cells (Fig. 6B; similar results observed with shRNA1 cells, data not shown). These results suggest that CIP4 negatively regulates the number of mature invadopodia.

To test the potential role of CIP4 in MT1-MMP endocytosis, we performed surface biotinylation assays on MDA-SrcYF vector and CIP4 KD cells. The total surface levels of MT1-MMP were compared under conditions that prevent endocytosis (4°C) by anti-MT1-MMP immunoblotting of proteins bound to streptavidin-conjugated beads (Fig. 6C). Interestingly, the relative levels of surface MT1-MMP was ~twofold higher in CIP4 KD cells, compared with control cells (Fig. 6C; tubulin immunoblot was used as a control for input protein concentration). In addition, biotinylated cells were shifted to 37°C for 15 minutes to allow endocytosis to proceed, before stripping of surface biotin and capture of the internalized pool of biotinylated proteins. Interestingly, we observed a 40–50% reduction in MT1-MMP endocytosis in CIP4 KD cells, compared with the vector control (Fig. 6C, lower panels). SrcYF expression in MDA-MB-231 cells also leads to increased secretion of MMP-2 and MMP-9 (Artym et al., 2006). We tested conditioned medium from MDA-vec, MDA-SrcYF vector and CIP4 KD cells using gelatin zymograms and found no major defects in levels of activated MMP-2 and MMP-9 released by MDA-SrcYF CIP4 KD cells, compared with the control (supplementary material Fig. S2). Taken together, these results identify CIP4 as a negative regulator of MT1-MMP accumulation within invadopodia, which acts by promoting MT1-MMP endocytosis. Furthermore, the differences in surface MT1-MMP in CIP4 KD cells probably accounts for their elevated levels of ECM degradation compared with that observed in control cells.

Because invadopodia promote directional cell motility and invasion of breast tumor cells (Stylli et al., 2008; Desmarais et al., 2009), we compared these properties in both MDA-SrcYF control and CIP4 KD cells. Confluent cells were subjected to wound-healing assays, and we observed a similar ability of all three cell lines to close most of the wound area within 24 hours (Fig. 7A). Quantification of the cell-free areas from several wounds revealed no requirement for CIP4 in migration of MDA-SrcYF cells (Fig. 7B).

Because CIP4 KD MDA-SrcYF cells degraded more ECM than control cells, we tested whether these differences result in altered ability to invade through a layer of matrigel. Indeed, we observed a significant increase in the invasiveness of CIP4 KD MDA-SrcYF cells, compared to control cells (Fig. 7C, asterisks indicate significant differences from vector control cells, P<0.01). These results identify CIP4 as a negative regulator of breast tumor invasion. To test whether CIP4 regulates migration or invasion in parental MDA-MB-231 cells, we performed wound healing migration, which revealed no major defects in CIP4 KD cells (supplementary material Fig. S3A). Consistent with our results showing increased invadopodia in CIP4 KD cells, we observed increased invasion through matrigel, compared with invasion of control MDA-MB-231 cells (supplementary material Fig. S3B). Taken together, our results suggest that CIP4 is a negative regulator of invadopodia in both weakly invasive MDA-MB-231 parental cells, and in the highly invasive MDA-SrcYF cell line.

Because Y471 in CIP4 was previously reported to be a potential Src phosphorylation site (Dombrosky-Ferlan et al., 2003), we prepared a Y471F mutation in a GFP–CIP4 expression plasmid, and performed co-transfections of GFP–CIP4 with or without SrcYF plasmid in HEK293 cells. As expected, SrcYF expression induced robust tyrosine phosphorylation of CIP4 (Fig. 8A, compare lanes 3 and 4), which was largely blocked by the Y471F mutation (compare lanes 3 and 5). These differences were not due to reduced expression or recovery of the GFP–CIP4^Y471F protein (Fig. 8A, second panel). However, it is worth noting that with longer exposures, some CIP4 tyrosine phosphorylation was still observed with the Y471F mutant (data not shown), suggesting the presence of other secondary tyrosine phosphorylation sites in CIP4. CIP4 is probably a direct Src substrate, based on the ability of purified, active Src to phosphorylate recombinant, purified CIP4 in vitro (data not shown). To test the potential effect of CIP4 tyrosine phosphorylation on its function in invadopodia, we performed transient transfections of MDA-SrcYF cells with GFP and wild-type and Y471F GFP–CIP4, and measured ECM degradation area per transfected cell (Fig. 8B). Although GFP–CIP4 expression caused a decrease in ECM digestion relative to GFP control, we observed a more pronounced inhibitory effect with GFP–CIP4^Y471F (Fig. 8C). These results suggest that tyrosine phosphorylation of CIP4 alters its invadopodia-suppressive activity.

In this study, we add CIP4 to the expanding list of Src substrates that both positively and negatively regulate invadopodia formation and ECM degradation in tumor cells (Stylli et al., 2008; Buccione et al., 2009). We demonstrate that activated Src expression induces both phosphorylation of CIP4 and localization of CIP4 to invadopodia. Using stable silencing of CIP4 in MDA-MB-231 and the more invasive derivative MDA-SrcYF, we show that CIP4 regulates surface levels of MT1-MMP, a marker of mature invadopodia. We uncover a potential mechanism for CIP4 regulation of invadopodia, by showing that CIP4 promotes MT1-MMP internalization (Fig. 9). This probably relates to the roles of BAR and F-BAR domain proteins in inducing or stabilizing membrane invaginations through their BAR and F-BAR domains (Frost et al., 2008; Takano et al., 2008; Wang et al., 2008), while recruiting key regulators of actin assembly (e.g. N-WASP) or endocytosis (e.g. dynamin) to facilitate vesicle scission (Suetsugu et al., 2010). In our model (Fig. 9), we propose that F-BAR adaptors such as CIP4 function in the formation of membrane invaginations supported by localized actin assembly, whereas BAR domain adaptors such as endophilin further constrict the membrane and recruit dynamin to promote vesicle scission. Src phosphorylation of the endophilin SH3 domain (Y315) was shown to block dynamin binding and reduce MT1-MMP endocytosis (Wu et al., 2005). We hypothesize that Src-induced CIP4 tyrosine phosphorylation will also alter its ability to recruit and activate N-WASP (or other SH3 ligands) at sites of MT1-MMP endocytosis. Consistent with a potential negative effect of CIP4 phosphorylation on its scaffolding function, ectopic expression of a CIP4 mutant lacking the major Src phosphorylation site (Y471F), led to more dramatic inhibition of ECM digestion in MDA-SrcYF cells, compared with that observed in the control. Thus, CIP4 functions in a negative regulatory pathway that promotes endocytosis of proinvasive proteins such as MT1-MMP within invadopodia, and this pathway might be inhibited by Src (Fig. 9).

Our results extend on the emerging evidence that CIP4 and related F-BAR proteins regulate the early steps of receptor endocytosis (Itoh et al., 2005; Tsujita et al., 2006). We also found that CIP4 localizes to Rab5 endosomes, and regulates lysosomal targeting of activated EGFR in A431 cells (Hu et al., 2009). Interestingly, internalized MT1-MMP is also found in Rab5 endosomes before either degradation in lysosomes or recycling to the cell surface (Wu et al., 2005). Because the major fraction of internalized MT1-MMP recycles from endosomes to the cell surface (Remacle et al., 2003; Itoh and Seiki, 2006; Li et al., 2008), our observations of more surface MT1-MMP expression and ECM degradation in CIP4 KD cells, compared with control MDA-SrcYF cells, is consistent with CIP4 promoting internalization and/or inhibiting recycling of MT1-MMP (Fig. 9). MT1-MMP is also focally delivered to invadopodia by directed exocytosis, which involves cortactin, IQGAP1 and exocyst vesicle-docking complex (Poincloux et al., 2009). Delivery of key cargo and membranes to podosomes and invadopodia relies on positioning of the microtubule organizing center (MTOC) and microtubules (Gimona et al., 2008). Because CIP4 regulates MTOC positioning and exocytosis of cytolytic granules in Natural Killer cells (Banerjee et al., 2007), CIP4 might also regulate invadopodia formation by affecting the delivery of vesicles containing invadopodia components.

The CIP4 subfamily of F-BAR proteins promote actin assembly via the Cdc42/N-WASP pathway in a number of model systems, including Drosophila, Xenopus and mammalian cells (Ho et al., 2004; Takano et al., 2008; Fricke et al., 2009). Dimerization of N-WASP by interaction with F-BAR protein dimers, was recently shown to enhance Arp2/3 binding and F-actin branching (Padrick et al., 2008; Soderling, 2009). However, it is unlikely that CIP4 family adaptors are solely responsible for N-WASP activation and formation of the F-actin core of invadopodia, because adaptors such as Tks5 and Nck can also recruit and activate N-WASP in invadopodia precursors (Yamaguchi et al., 2005b; Stylli et al., 2009). Although N-WASP localized throughout the F-actin core of invadopodia, active N-WASP was restricted to the base of the invadopodia (Lorenz et al., 2004). We observed localization of CIP4 at the base of invadopodia, with long protrusions into the ECM, and in some cases, CIP4 localized throughout the invadopodia with shorter protrusions (this paper). Further experiments will be needed to determine at which stage of invadopodia formation or disassembly CIP4 is recruited, and which domains are involved. Early events in Src-induced podosome (and presumably invadopodia) formation are driven by localization of PH- and PX-domain-containing adaptors (e.g. Tks5, AFAP110) (Gatesman et al., 2004; Stylli et al., 2009), to ventral membranes enriched in PtdIns(3,4)P₂ (Oikawa et al., 2008). Although primarily an outward area of membrane curvature, the neck of the invadopodia might provide appropriate curvature to recruit F-BAR proteins such as CIP4. In the case of the F-BAR adaptor Pacsin-2, it was shown to localize to the base of microspikes via its F-BAR domain (Shimada et al., 2010). Similar localization of Toca-1 to the base of filopodia has also been reported recently (Bu et al., 2009; Hu et al., 2010). Future studies will be required to fully understand the potential roles of F-BAR proteins in invadopodia formation or dynamics.

It is worth noting that invadopodia are also actin comet-based structures (Baldassarre et al., 2006). N-WASP localizes to the head of the actin comets, whereas the Arp2/3 complex, cortactin and fascin localize to the tails that move in a corkscrew fashion within areas of ECM degradation (Li et al., 2010). Phosphoinositides might direct N-WASP localization to the head of actin comets, either by direct binding with N-WASP (Caldieri et al., 2009) or possibly indirectly through CIP4 subfamily F-BAR proteins. Considering the emerging role for Toca-1 in N-WASP pathways and actin comets (Ho et al., 2004; Leung et al., 2008), it will be important to test the involvement of Toca-1 as a potential positive effector of actin assembly within invadopodia. One recent study showed that FBP17 KD in monocytic THP-1 cells blocked podosome formation (Tsuboi et al., 2009). It would be interesting to test the effects that knockdown of other CIP4 subfamily proteins have on invadopodia formation and invasiveness of breast tumor cells. Despite their similarities in domain organization and potential binding partners, it is likely that the expansion of the CIP4 subfamily in mammals allows for distinct functions for CIP4, Toca-1 and FBP17 in cells that express all three members.

During revision of our paper, another group reported that CIP4 localizes to invadopodia in MDA-MB-231 breast tumor cells, and positively regulates invadopodia formation, ECM degradation, cell migration and invasion (Pichot et al., 2010). The authors show that CIP4 forms a transient, EGF-induced association with N-WASP, and that EGF-induced N-WASP pY256 was reduced in CIP4 KD cells. In their study, CIP4 or N-WASP KD resulted in similar defects in invadopodia, migration and invasion of MDA-MB-231 cells. Although our results for CIP4 localization to invadopodia are similar to theirs, the effects of CIP4 KD on invadopodia parameters are not. This could reflect the fact that the transient siRNA-induced CIP4 KD used by Pichot and colleagues was more acute than our stable shRNA-driven CIP4 KD. Perhaps 10–20% of CIP4 expression is sufficient to support this positive role in a weakly invasive cancer cell. It is worth noting that in a study of the role of Toca-1 in bacterial actin comet formation in epithelial cells, the positive role of Toca-1 in this N-WASP pathway required >95% depletion of Toca-1 (Leung et al., 2008). It is also possible that during the selection of our stable CIP4 KD cell pools, compensatory changes in expression of other N-WASP regulatory gene products occurred. However, we found no effect of knockdown of CIP4 on the other CIP4 subfamily proteins (this paper). Another important difference between these studies is that the invadopodia parameters were studied in different conditions, using different fluorescent ECM substrates. We used EGF-treated MDA-MB-231 cells with CIP4 knockdown to score invadopodia and ECM digestion (this paper), whereas Pichot and co-workers used serum conditions in which cells formed very few invadopodia (<3 per cell) (Pichot et al., 2010). Similarly to SrcYF expression, treatment with EGF leads to cortactin phosphorylation, and the release of cofilin (Oser et al., 2009). This leads to increased actin severing to generate new barbed ends for further polymerization. Under conditions favoring cofilin activity, F-actin branching pathways might become less dependent on N-WASP pathways, and more dependent on actin nucleators such as formins. This could explain why SrcYF-induced invadopodia also had shorter lifetimes than those induced by EGF (Oser et al., 2009). It is possible that the negative regulatory effect of CIP4 we report here, involves a role for CIP4 in promoting invadopodia disassembly. Future experiments will be required to fully define the molecular mechanisms by which CIP4 regulates the invadopodia.

The ability of a cancer cell to form invadopodia and digest ECM is directly associated with its ability to invade through matrigel basement membranes (Monsky et al., 1993). In this paper, we show that stable silencing of CIP4 promotes invadopodia formation and invasion of MDA-MB-231 parental and MDA-SrcYF breast tumor cells, without effecting cell migration. By contrast, Pichot and colleagues find that transient CIP4 knockdown results in a significant reduction in both migration and invasion of MDA-MB-231 cells (Pichot et al., 2010). In their experimental system, CIP4 knockdown resulted in a migration defect, which would also contribute to the defect in cell invasion independently of potential effects on invadopodia. We performed similar wound-healing migration assays in MDA-MB-231 cells (and MDA-SrcYF), and observed no defect in migration of CIP4 KD cells (this paper). Probably the best way to address the role of CIP4 in this process will be to extend these studies to mammary tumor progression and metastasis models in vivo. The stable CIP4 knockdown approach used in our study should allow for stable silencing of CIP4 expression during mouse xenograft experiments with MDA-SrcYF cells. Given the ability of Src inhibitors to limit breast tumor metastases (Rucci et al., 2006), it will be possible then to test whether CIP4 actually regulates tumor metastasis in vivo. One might also predict that loss of CIP4 expression is a mechanism for establishing a more invasive breast tumor phenotype. In fact, a previous study of estrogen receptor (ER) target genes identified CIP4 (also called TRIP10), and found that in ER-negative breast tumors, the CIP4 promoter becomes hypermethylated and CIP4 mRNA is downregulated (Leu et al., 2004). This mode of regulation of CIP4 expression was also recently shown for mesenchymal stem cells, wherein CIP4 downregulation by promoter methylation leads to increased differentiation (Hsiao et al., 2010). This implies that the levels of the CIP4 adaptor protein within cells are tightly controlled, and that CIP4 downregulation via epigenetic or post-translational modifications can have dramatic effects on cell phenotype and function. Future studies addressing levels of CIP4 protein and CIP4 tyrosine phosphorylation in breast tumor microarrays might provide further insights into the potential involvement of CIP4 in human breast cancer.

Rabbit polyclonal CIP4 antiserum was raised against purified 6-His-tagged, full-length human CIP4 within the Queen's University Animal Care Services facility. Mouse anti-Toca-1 hybridoma supernatant was kindly provided by Giorgio Scita (IFOM, Milan, Italy). Mouse anti-phosphotyrosine (PY99) and anti-Src were purchased from Santa Cruz Biotech. Rabbit anti-pY416-Src was purchased from Cell Signaling Technology. Rabbit antisera specific for human pY421-cortactin (pY-Cort) was purchased from Medicorp. Goat anti-FBP17 antiserum was from IMGENEX. CIP4 monoclonal antibody was purchased from BD Bioscience. Polyclonal GFP antibody was obtained from Chemicon. Monoclonal MT1-MMP antibody was obtained from Abcam. PP2 and SU6656 were purchased from Calbiochem. TRITC-conjugated phalloidin, fibronectin, Coomassie Brilliant Blue, gelatin, and bovine serum albumin (BSA) were all purchased from Sigma. GammaBind Sepharose and Glutathione–Sepharose were purchased from GE Healthcare. Transwell chambers (8 μm pore) and matrigel were purchased from BD Bioscience. Phalloidin conjugated to Alexa Fluor 350 and Alexa Fluor 488 was purchased from Invitrogen. Streptavidin-conjugated Sepharose and biotin (EZ LINK NHS-SS) were from Pierce.

MDA-MB-231 (MDA) cells and HEK293T cells were grown in DMEM supplemented with 10% FBS (Sigma), 1% glutamine (Invitrogen), antimycotics and antibiotics (Invitrogen) at 37°C in a humidified atmosphere with 5% CO₂. MDA-MB-231 cells were transduced with pWZL-hygro (empty vector) and pWZL-hygro-SrcY527F retroviruses and selected with hygromycin as previously described (Mukhopadhyay et al., 2009). The resultant cell pools were designated MDA-vec and MDA-SrcYF and maintained in medium containing hygromycin (100 μg/ml). For Src inhibitor pretreatment, cells were treated with PP2 to a final concentration of 10 μM, SU6656 to a final concentration of 20 μM, or equivalent volume of DMSO (vehicle control) for 1 hour in DMEM complete medium. Soluble cell lysates (SCLs) were prepared as previously described (Hu et al., 2009), and subjected to immunoprecipitation with rabbit anti-CIP4 and GammaBind-Sepharose beads. Following extensive washing, bound proteins and SCLs were subjected to immunoblotting with pY416-Src, PY99 and CIP4 antisera. Transient transfections of MDA-SrcYF cells plated in 24-well plates with gelatin-coated glass coverslips were performed using GFP-CIP4^WT expression plasmid (Hu et al., 2009) and FuGENE HD reagent (Roche) according to the manufacturer's instructions. Cells were harvested at 48 hours post transfection, fixed in 2% paraformaldehyde and stained with rabbit anti-GFP followed by Alexa Fluor 488 goat anti-rabbit IgG and TRITC-phalloidin. For rescue and overexpression experiments, MDA-SrcYF and/or MDA-SrcYF-shRNA2 cells were transiently transfected as above with plasmids encoding GFP, GFP fused to mouse CIP4 (Chang et al., 2002), GFP fused to an shRNA2-resistant GFP–CIP4 (human; silent mutations engineered into shRNA target sequence by QuikChange mutagenesis, Stratagene). After 48 hours, cells were trypsinized, counted and plated on fluorescent ECM as described below. HEK293T cells were transiently transfected with GFP–CIP4^WT or GFP–CIP4^Y471F plasmids (Y471F mutation made using QuikChange mutagenesis) with or without SrcYF plasmid by the calcium phosphate method. Lysates were prepared and subjected to immunoprecipitation and immunoblotting as described above.

Lentiviruses encoding empty pLKO.1 vector and five different shRNAs specific for human CIP4 were obtained from Open Biosystems (RHS4533). Lentiviruses were produced by transfection of HEK293 cells (grown on 100 mm plates) with pLKO.1-based plasmid (15 μg), pCMVΔR8.91 packaging plasmid (15 μg), and pMD.2G envelop plasmid (6 μg) via the calcium phosphate method. Two batches of conditioned medium were collected at 48 and 72 hours, filtered through 0.45 μm sterile filters, and stored at −80°C in aliquots. Viruses were titered according to the manufacturer's instructions (Open Biosystem) using A431 cells. MDA-SrcYF cells were transduced with viral supernatants of similar titre (≈1×10⁵ TU/ml), and cell pools were selected using puromycin (1.5 μg/ml). After several passages, a vector cell pool, and two separate CIP4-specific shRNA-expressing cell pools were selected based on the efficiency of CIP4 KD (shRNA1 target sequence: 5′-GCCCATAATAGCCAAGTGCTT-3′, TRCN0000063187, clone A9; shRNA2 target sequence: 5′-GCAACAGTCCTTCGTACAGAT-3′, TRCN0000063187, clone A10).

MDA-SrcYF cells were plated on gelatin-coated glass coverslips for the time indicated in figure legends before fixation in 2% paraformaldehyde and permeabilization with 0.2% Tween-20. Following blocking with 3% BSA (Sigma) at room temperature, primary antibodies were added for overnight incubation at 4°C as indicated: mouse anti-CIP4 (BD Bioscience, 1:25), rabbit anti-pY421-cortactin (pY-Cort, 1:100), rabbit anti-GFP (1:200), mouse anti-MT1-MMP (1:50). After extensive washing in PBS/0.2% Tween-20, coverslips were incubated with either Alexa-Fluor-488-conjugated or Alexa-Fluor-633-conjugated goat anti-mouse or rabbit IgG (1:400) at room temperature for 1 hour. In experiments showing F-actin staining, coverslips were incubated with either TRITC-conjugated or Alexa-Fluor-488-conjugated phalloidin (1:200) at room temperature for 1 hour. After washing, coverslips were mounted on slides and analyzed by either confocal microscopy (HCX PL APO DIC 63×/1.32 Oil CS objective; Leica TCS SP2 Multi Photon; Queen's Protein Function Discovery facility) or epifluorescence microscopy [Olympus BX51 microscope equipped with a Q Color5 digital camera (60× objective; images acquired using QCapturePro software)]. For 3D imaging experiments, images were acquired using a Quorum WaveFX spinning disk confocal system (Quorum Technologies, Guelph ON, Canada). Briefly, the system is based on the Olympus BX-1 inverted stand with five laser emission lines (405 nm, 440 nm, 491 nm, 568 nm, 633 nm) provided from a Synapse merge module (Spectral Applied Research, Richmond Hill, ON). Images were collected with a Hamamatsu EM-CCD camera (model 09100-13). Data acquisition and analysis were performed using Metamorph imaging software.

Cells were lysed using kinase lysis buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM EDTA, 1% Nonidet P-40, 0.5% sodium deoxycholate, 10 μg/ml aprotinin, 10 μg/ml leupeptin, 1 mM Na₃VO₄, 100 μM phenylmethylsulfonyl fluoride). Cell debris was removed by centrifugation at 12,000 r.p.m., for 10 minutes at 4°C. SCLs were subjected to immunoprecipitation with rabbit anti-CIP4. Antibody–protein complexes were recovered with 20 μl Gamma Bind Sepharose beads (a 50% [vol./vol.] slurry in phosphate-buffered saline (PBS); Amersham Pharmacia Biotech) and washed three times with PBS. Proteins were recovered by adding 25 μl of 2× SDS sample buffer to the beads. Immunoblotting (IB) was carried out following the transfer of proteins to Immobilon P membranes (Millipore) by using a semidry transfer apparatus (Bio-Rad). Blots were probed with antibodies directed against: rabbit CIP4 (1:1000), ERK1/ERK2 (ERK1/2; 1:2000), Src (1:1000), pY416-Src (1:1000), phosphotyrosine (PY99; 1:1000) Antibody complexes were detected with either horseradish-peroxidase-conjugated sheep anti-mouse immunoglobulin (1:5000; Amersham Pharmacia Biotech) or horseradish-peroxidase-conjugated goat anti-rabbit immunoglobulin (1:10,000; Amersham Pharmacia Biotech) and revealed by enhanced chemiluminescence (ECL; Applied Biological Materials).

MDA-SrcYF vector control and CIP4 knockdown cell lines were trypsinized, counted, and plated on gelatin-coated glass coverslips (5000 cells/well) containing a layer of TRITC-conjugated fibronectin, prepared as previously described (Webb et al., 2007). After incubation for 18 hours, cells were fixed and stained for either F-actin and/or pY-Cort as described above. Epifluorescence microscopy was performed and the areas of ECM degradation underneath of cells stained with F-actin were quantified as digestion area per cell using Image Pro Plus software (Media Cybernetics). At least 30 cells were quantified for each condition and digestion/cell was averaged from three independent experiments. Gelatin zymography was performed as described previously (Hauck et al., 2002). Briefly, conditioned medium from MDA-vec, MDA-SrcYF vector and CIP4 knockdown cells was collected, clarified by centrifugation, resolved in nonreducing gels containing 0.05% (w/v) gelatin, and processed for zones of gelatin degradation by Coomassie Blue staining. The number of invadopodia in MDA-SrcYF cell lines were quantified from confocal micrographs as described previously (Cortesio et al., 2008). Briefly, dot-like structures containing both pY-Cortactin and F-actin were scored as invadopodia, and the numbers of invadopodia per cell were quantified.

Cell-surface protein biotinylation and endocytosis assays were performed as previously described (Wu et al., 2005). Briefly, subconfluent MDA-SrcYF vector and CIP4 KD cells growing on six-well plates were chilled on ice, washed twice with ice-cold PBS before incubation in cold biotinylation buffer [154 mM NaCl, 10 mM HEPES pH 7.6, 3 mM KCl, 1 mM MgCl₂, 0.1 mM CaCl₂, 10 mM glucose, 0.6 mg/ml biotin (EZ LINK NHS-SS; Pierce)] for 40 minutes at 4°C. For surface MT1-MMP level determination, labeled cells were lysed, precipitated with streptavidin-conjugated Sepharose beads (Pierce), and subjected to immunoblotting with anti-MT1-MMP. For MT1-MMP endocytosis assays, cells were warmed to 37°C for 15 minutes to allow internalization of biotinylated receptors before addition of cold MESNA buffer [100 mM NaCl, 50 mM Tris-HCl pH 8.6, 1 mM MgCl₂, 0.1 mM CaCl₂, 50 mM MESNA (Sigma)] to remove cell-surface biotin. As a negative control, cells were treated with MESNA buffer directly (without incubation at 37°C). Lysates were prepared and subjected to streptavidin–Sepharose pull-downs, and immunoblotting with anti-MT1-MMP. To normalize for input protein amounts between cell lines, lysates were subjected to immunoblotting with anti-tubulin. Densitometry was performed to determine the relative MT1-MMP levels (ratio of MT1-MMP/tubulin).

Wound-healing assays were performed as previously described (Malliri et al., 1998). MDA-MB-231 and MDA-SrcYF vector and CIP4 KD cells were plated on a gelatin-coated glass coverslips in a six-well plate at (10⁶ cells/well) and cultured for 1 day until a uniform monolayer was formed. Wound areas were made using 200 μl sterile pipette tips, and washed with serum-free medium to remove suspended cells. Cells were fixed immediately (0 h), or following 24 hour incubation in DMEM + EGF (10 ng/ml) before staining of F-actin, as described above. Wound areas were analyzed by epifluorescence microscopy and the average cell-free areas (three fields of views for triplicate samples) were calculated for each cell line based on pixel areas scored using Corel PhotoPaint histogram function.

Invasion assays were performed as described previously (Rucci et al., 2006). Briefly, Transwell inserts (8 μm pore) were coated with 100 μl of ice-cold Matrigel (1:5 dilution in DMEM with 0.5% FBS) and incubated at 37°C for 30 minutes to solidify. MDA-MB-231 and MDA-SrcYF vector or CIP4 KD cells were trypsinized, counted and seeded in DMEM serum-free medium in the top chamber (50,000 cells) which was placed in 24-well plates containing 0.5 ml NIH-3T3 conditioned medium as chemoattractant. After incubation for 24 hours in the 37°C incubator, the filters were fixed in methanol, cells and matrigel on top of the filter were removed, and cells attached to the underside of the filter were stained with DAPI. Epifluorescence microscopy was performed to detect DAPI-stained nuclei on the filters (numbers of invaded cells), and quantification was performed using Image Pro Plus software (Media Cybernetics).

Differences between experimental groups were examined for statistical significance (P<0.05 or P<0.01) using paired Student's t-test in Excel (Microsoft, Redmond, WA).

We thank Alan Saltiel (University of Michigan, Ann Arbor, MI) for the mouse CIP4 construct, Chris Nicol for use of his epifluorescence microscope, Rob Eves for technical assistance, and the expert staff of the cell imaging facility at Queen's University for help with confocal microscopy. This work is supported by a grant to A.W.B.C. from Canadian Breast Cancer Foundation (CBCF) Ontario chapter, and grants to A.S.M. from Canadian Institutes for Health Research (CIHR; MP78468) and Ontario Heart & Stroke Foundation (T5829). Salary support was provided by CBCF predoctoral fellowship to J.H. and a CIHR New Investigator award to A.W.B.C.