Oncogene (1999) 18, 4440 ± 4449 ã 1999 Stockton Press All rights reserved 0950 ± 9232/99 $12.00 http://www.stockton-press.co.uk/onc An invasion-related complex of , and PKCm associates with invadopodia at sites of degradation

Emma T Bowden1, Mara Barth1, Dianne Thomas1, Robert I Glazer2 and Susette C Mueller*,1

1Department of Cell Biology, Lombardi Cancer Center, Georgetown University Medical School, Washington DC 20007, USA; 2Department of Pharmacology, Lombardi Cancer Center, Georgetown University Medical School, Washington DC 20007, USA

Invasive breast cancer cells have the ability to extend molecules such as a-actinin, , tensin and paxillin membrane protrusions, invadopodia, into the extracel- concentrate in or around invadopodia (Mueller et al., lular matrix (ECM). These structures are associated 1992; Chen, 1989). with sites of active matrix degradation. The amount of A number of lines of evidence suggest that cortactin matrix degradation associated with the activity of these may have an important role in modulating invasive protrusions has been shown to directly behavior. Cortactin binds to F- in vitro, co- correlate with invasive potential. We demonstrate here localizes with cortical actin at ru‚ing membranes, that microinjection of polyclonal anti-cortactin antibo- and possesses actin-bundling activity that is modulated dies blocks matrix degradation at invadopodia support- by c-Src, suggesting a role in membrane motility (Wu ing the hypothesis that cortactin has a direct role in and Parsons, 1993; Huang et al., 1997). In v-Src- invasive behavior. MDA-MB-231, invasive breast cancer activated ®broblasts transformed by Rous sarcoma cells were sheared from the surface of a gelatin matrix to virus, cell motility and invasiveness are increased, isolate invadopodia. Cortactin, paxillin and cortactin becomes phosphorylated on residues kinase C (PKC) m, a kinase, were co-immunopre- enabling an association with v-Src in cipitated as a complex from invadopodia-enriched (Okamura and Resh, 1995). Overexpression of membranes. We con®rmed the subcellular distribution cortactin in NIH3T3 ®broblasts or endothelial cells of these by immunolocalization and Western increases cell motility in vitro (Patel et al., 1998; Huang blotting. We also determined that, in contrast to its et al., 1998). Cortactin overexpression has also been presence in invasive cells, this complex of proteins was implicated in breast cancer progression. The human not detected in lysates from non-invasive cells that do not homologue of the cortactin , EMS1 (Schuuring et form invadopodia. Taken together, these data suggest al., 1992) is overexpressed in a number of breast cancer that the formation of this cortactin-containing complex cell lines (Schuuring et al., 1993; Campbell et al., 1996). correlates with cellular invasiveness. We hypothesize that In human breast carcinoma, the majority of tumors this complex of molecules has a role in the formation and showing an 11q13 overampli®cation, also show an function of invadopodia during cellular invasion. increase in expression of cortactin, and this has been correlated with poor patient prognosis (Schuuring et Keywords: invasion; cortactin; paxillin; PKCm; breast al., 1992). In cells with this ampli®cation, cortactin has cancer been shown to redistribute to contact sites at the margins of cells, concurrent with an increase in its levels of (van Damme et al., 1997). In untransformed cells, cortactin is a downstream target Introduction for tyrosine phosphorylation following cell adhesion to ®bronectin (Vuori and Ruoslahti, 1995). As a When invasive cells such as Rous sarcoma virus- consequence of this, it has been suggested that transformed chicken embryo ®broblasts or metastatic cortactin also may have a role in the regulation and cancer cell lines are cultured on ECM, actin-containing remodeling of the in response to speci®c protrusions, called invadopodia, extend into the matrix extracellular stimuli (Vuori and Ruoslahti, 1995). and participate in matrix degradation (Chen, 1989; Paxillin, a protein (Turner et al., Coopman et al., 1996; Monsky et al., 1994; Mueller 1990), is phosphorylated in response to - and Chen, 1991). There is a correlation between the mediated adhesion (Burridge et al., 1992). V-Src ability of cells to locally degrade the matrix at transformation causes a partial redistribution of invadopodia, and their invasive potential as measured paxillin from focal adhesions to invadopodia in Rous in other in vitro and in vivo assays for motility and sarcoma virus-transformed chicken embryo ®broblasts invasion (Coopman et al., 1998). Membrane-associated grown on a gelatin matrix (Mueller et al., 1992). Viral proteases including seprase, membrane type-1 matrix transformation also causes a signi®cant increase of metalloproteinase, and matrix metalloproteinase-2 are both paxillin expression and tyrosine phosphorylation thought to mediate ECM degradation at sites of in invasive and metastatic cervical tumors (McCormick invadopodial extension (Nakahara et al., 1996, 1997; et al., 1997). Since tyrosine phosphorylation of paxillin Monsky et al., 1993, 1994). Actin and actin-associated increases in response to migratory stimuli, a direct role in motility has been suggested (Abedi et al., 1995). However, the functional role of paxillin both in *Correspondence: SC Mueller Received 26 October 1998; revised 11 March 1999; accepted 11 transformed and normal cell types remains largely March 1999 unidenti®ed. Complex of cortactin, paxillin and PKCm in invadopodia ET Bowden et al 4441 Another downstream response to integrin-mediated man et al., 1996). We have demonstrated previously adhesion includes activation of PKC. PKC activity is that the in vitro invasive behavior of cells grown on involved in integrin-mediated responses including gelatin and measured by phagocytosis is similar to their adhesion, motility and membrane ru‚ing (Lewis et behavior in this assay when grown on collagen I and al., 1996; Klemke et al., 1994; Yebra et al., 1995; Matrigel (Coopman et al., 1998). We have also Miyata et al., 1989). It can also a€ect invasive behavior demonstrated that transformed cells will make (Ways et al., 1995). PKC activation has been invadopodia on a range of ECM substrates including demonstrated as a controlling factor in the reabsorb- ®bronectin, collagen type I, collagen type IV and tion activity exhibited by osteoclasts, a non-trans- laminin (Kelly et al., 1994). We use crosslinked gelatin formed cell type, responsible for bone remodeling in our in vitro assays because it is easily handled and (Moonga et al., 1996). PKC activity is also implicated because in these assays, it allows a rapid and visual in certain breast cancers where expression is higher in di€erentiation between non-invasive and invasive cells. tumors compared with normal tissue (O'Brian and Initially, we cultured cells in the absence of serum in Ward, 1989). order to block matrix degradation prior to microinjec- In this report, we describe that in invasive human tion. Following a period of recovery after microinjec- breast cancer cells, cortactin, paxillin and the PKC tion, serum free culture media was replaced with serum isoform PKCm, are associated as an invasion-related containing media. This released the serum dependant complex in invadopodia. We describe a correlation block in matrix degradation and the cells were allowed between formation of this complex of molecules and to invade into the ®lms for 5 h. Microinjected antibody the invasive phenotype in breast cancer cells. These was visualized using Texas red conjugated anti-rabbit data identify a novel invadopodia associated distribu- secondary antibody and nuclei were visualized by co- tion for both cortactin and PKCm. We con®rm the injected 4,6-diamidino-2-phenylindole (DAPI). Micro- colocalization of cortactin, paxillin and PKCm at injection with the preimmune IgG did not signi®cantly invadopodia by immuno¯uorescence. We report that inhibit matrix degradation (Figure 2, preimmune). microinjection of the anti-cortactin poyclonal antibody Microinjection of the anti-cortactin pAb (#170-10) (pAb) disrupts invadopodia function, blocking the signi®cantly inhibited hole formation (Figure 2, matrix degradation associated with these structures. Immune versus preimmune cells, P=0.006). An This suggests that cortactin has a direct role in the examination by both phase contrast and immunofluor- coordination of local invasion mediated by the escence microscopy revealed that microinjected cells formation of invadopodia. The interaction of cortactin remained well spread and that there were no signi®cant with paxillin and PKCm at invadopodia suggests that morphological changes. Therefore, we conclude that these molecules have a direct role in the regulation of cortactin function is required for either the formation invadopodia activity.

Results

Invasive breast cancer cells extend invadopodia into the matrix Ultrathin vertical sections of MDA-MB-231 cells grown on crosslinked gelatin beads were examined by transmission electron microscopy. Invasive breast cancer cells extend long protrusions, or invadopodia, into the surface of the bead from their ventral membrane surface (Figure 1a and b, arrows). The region of interest examined in the electron micrograph (Figure 1b) is highlighted in the schematic diagram (Figure 1a, shaded box). Frequently, there are clear areas around the extending protrusions where invado- podia-associated proteolytic activity presumably has perturbed the crosslinked gelatin bead. Invasion is often concentrated in an area at the ventral membrane under or near the nucleus.

Microinjection of anti-cortactin pAbs blocks local invasion To test the hypothesis that cortactin is required for the Figure 1 Invasive breast cancer cells produce invadopodia that invade and extend into the matrix substratum. Bar=1 mm. (a)A invasive function of invadopodia, MDA-MB-231 cells cartoon representation shows the localization of invadopodia on were cultured on ¯uorescein isothiocyanate (FITC)- the ventral surface of an invasive cell in vertical cross section. The crosslinked gelatin matrices. Cells were microinjected shaded area corresponds to an area similar to that shown in the with either preimmune IgG or the anti-cortactin pAb electron micrograph in (b). (b) Electron micrograph of an (#170-10). As previously described, these cells locally ultrathin vertical section through an MDA-MB-231 cell cultured on gelatin crosslinked beads shows the ultrastructure of degrade an FITC matrix to leave dark areas, easily invadopodia. Invadopodia form long thin protrusions associated detectable by immuno¯uorescence microscopy (Coop- with the disruption of the gelatin bead surface (arrows) Complex of cortactin, paxillin and PKCm in invadopodia ET Bowden et al 4442 antibody that recognizes focal adhesions demonstrated that there was no vinculin staining material associated with the gelatin matrix suggesting that focal adhesions have been removed from the surface of the gelatin (data not shown). This is consistent with immunoblotting experiments in which focal adhesion kinase (FAK) was detected in the CBM but not the INV fraction and suggests that focal adhesions have been scraped away from the surface of the gelatin (data not shown). We performed immunoprecipitation experiments from each of the fractions in order to identify molecules that interact with cortactin. Cortactin immunoprecipitations and supernatants were exam- ined by Western blot analysis. Cortactin was immuno- precipitated eciently from each of the fractions, CBM, INV, and CYT, leaving only trace amounts detected in the supernatant from the immunoprecipita- tion (Figure 3c, Cortactin Blot, compare IP versus IP SUP). The blot was then stripped and reprobed for the presence of potential interacting molecules including vinculin, FAK, PKCa, PKCd and PKCm. We found that paxillin and PKCm co-immunoprecipitated with cortactin from the INV and CYT fractions (Figure 3c, IP Cortactin, Reblot Paxillin and PKCm: INV and Figure 2 Microinjection of anti-cortactin pAb (#170-10) blocks CYT). The supernatants clearly showed that although matrix degradation at invadopodia. MDA-MB-231 cells grown there were signi®cant amounts of both paxillin and overnight in serum free media were microinjected with either PKCm in the CBM there was little, if any, co- preimmune IgG or anti-cortactin pAb (#170-10). Injected cells immunoprecipitation of either with cortactin (Figure were allowed to recover before being transferred to serum containing media. Cells were ®xed and processed as described 3c, Paxillin Reblot and PKCm Reblot, compare CBM for immuno¯uorescence microscopy. An equal number of injected from IP versus IP SUP). There was also a fraction of (preimmune or immune) and non-injected cells were examined. the total pool of paxillin and PKCm that was Cells were scored as invasive if a cluster of at least four holes in associated with the CYT that did not co-immunopre- the FITC-gelatin matrix was present underneath the cell body. Cortactin pAb (#170-10) injected cells showed approximately a cipitate with cortactin (Figure 3c, compare CYT from sevenfold decrease in the percentage of invasive cells compared to IP and IP SUP). Thus, a complex of cortactin, paxillin, cells injected with the preimmune sera (P=0.006) and PKCm co-immunoprecipitated predominantly from the INV and CYT fractions of MDA-MB-231 cultured on crosslinked gelatin. and/or function of invadopodia as measured by matrix We also examined a time course of cells fractionated degradation. at 6, 24 and 48 h on gelatin (Figure 3d). We wanted to determine if there was any shift in the localization of the complex over time and with increasing cell density. An immunoprecipitable complex of cortactin, paxillin Again, we performed cortactin immunoprecipitations and PKCm from each fraction at each time point and analysed the To identify the components of MDA-MB-231 invado- results by SDS ± PAGE and Western blotting. Initially, podia that interact with cortactin, we isolated we examined the immunoprecipitation of cortactin to invadopodia-enriched subcellular fractions. Cell bodies determine the level of immunoprecipitation from each were sheared away from the surface of a crosslinked fraction (Figure 3d, Cortactin Blot, 6, 24 and 48 h: gelatin matrix to obtain a fraction enriched in CBM, INV and CBM). The Western blots were then invadopodia (INV) (Mueller et al., 1992). The cell stripped and reprobed for paxillin and PKCm. Co- bodies were further fractionated by centrifugation to immunoprecipitation was evident between cortactin, yield two fractions, one containing the cell body paxillin and PKCm in the INV fraction (Figure 3d, membranes (CBM) and the other the cytosol (CYT). Paxillin and PKCm Reblots, 6, 24 and 48 h: INV). b1 integrin localized predominantly to the CBM and However, although there was co-immunoprecipitation INV membane containing fractions, demonstrating an at 6 and 24 h in the CYT fraction, the interaction was ecient separation of membranes and cytosol (Figure no longer detected at 48 h although cortactin was 3a, b1 integrin: CBM and INV). The presence of b1 present in this fraction (Figure 3d, Paxillin and PKCm integrin in the INV fraction is consistent with a Reblots, 6, 24 and 48 h: CYT, Cortactin Blot, 48 h: previously described localization to invadopodia CYT). (Mueller et al., 1992; Coopman et al., 1996). We also wanted to explore the possibility that there We used immuno¯uorescence to visualize structures was a correlation between an invasive phenotype and associated with the gelatin after shearing (Figure 3b). formation of the complex. We examined both invasive Phalloidin staining of F-actin revealed clusters of and non-invasive breast cancer cell lines for the punctate structures in association with the gelatin presence of the complex. Non-invasive breast cancer (Figure 3b, solid arrows). There were, however, few if cell lines form few or no invadopodia compared with any, streak like structures. Staining with an anti- invasive breast cancer cell lines (Table 1) (Coopman et Complex of cortactin, paxillin and PKCm in invadopodia ET Bowden et al 4443 Table 1 Characterization of invasiveness of breast cancer cell lines Cell line MCF-7 MDA-MB-468 MDA-MB-453 SKBR3 MCF-7 ADR MDA-MB-231 HS578T Cell outgrowth in Matrigela Sphc Sph Sph Sph Inv Inv Inv Behavior in nude micea P P N N P Li HM Chemoinvasion assaya ++ + + + ++ +++++ +++++ Chemotaxisa ++ +++ + + +++ +++++ +++++ Film degradationb 7 + 7 7 ++ +++++ +++++ aThe invasion assay data of `Cell Outgrowth in Matrigel', `Behaviour in Nude Mice', `Chemoinvasion', and `' of breast cancer cells are from a previous study (Thompson et al., 1992). bThe invasion assay data of `Film Degradation' is from Coopman et al. (1998). cTable abbreviations include: Sph, spherical colony/non-invasive cluster formation in Matrigel; Inv, invasive colony formation; P, primary tumor only; N, non-tumorigenic; Li, local invasion; HM, hematogenous metastases to lungs. The chemoinvasive and chemotaxis assays are all graded in terms of a percentage of the measured activity of MDA-MB-231 cells, i.e., +, 0 ± 20%; ++, 20 ± 40%; +++, 40 ± 60%; ++++, 60 ± 80% and +++++, 80 ± 100%. The ®lm degradation assay is graded in terms of a percentage of the highest measured activity; +, 0 ± 20%; ++, 20 ± 40%; +++, 40 ± 60%; ++++, 60 ± 80%; +++++, 80 ± 100%

Figure 3 Subcellular fractionation of MDA-MB-231 cells and coimmunoprecipitation of cortactin, paxillin and PKCm.(a) After 24 h, cells were fractionated by shearing away the cell bodies to leave invadopodia-enriched membranes (INV) embedded in the matrix. The cell bodies were further fractionated into cell body membranes (CMB) and cytosol (CYT) fractions. Ten mg of lysate, INV, CBM, and CYT fractions were analysed by 10% SDS ± PAGE. b1 integrin was detected by Western blotting. (b) After shearing away cell bodies, as described above, a gelatin plate was washed once in YPP bu€er and, after ®xation, the clusters of invadopodia that remained embedded in the gelatin matrix were visualized by staining with Texas red phalloidin (closed arrows). Bar=10 mm. (c) Protein from each fraction (300 mg) was used to immunoprecipitate cortactin. Twenty mg of the supernatant from each immunoprecipitation was concentrated and Western blotted (IP SUP) together with the immunoprecipitated samples (IP). Immunoprecipitates were analysed by 8% SDS ± PAGE and Western blotting. Paxillin (Paxillin Reblot) and PKCm (PKCm Reblot) co-immunoprecipitate with cortactin (Cortactin Blot) from the INV and CYT fractions (IP, INV and CYT). (d) 300 mg of protein from each fraction at 6, 24 and 48 h was used to immunoprecipitate cortactin. Immunoprecipitations were analysed by 8% SDS ± PAGE and Western blotting. Paxillin and PKCm co-immunoprecipitate with cortactin from the INV fraction at 6, 24 and 48 h, but only from the CYT fraction at 6 and 24 h Complex of cortactin, paxillin and PKCm in invadopodia ET Bowden et al 4444 al., 1998). Paxillin immunoprecipitates from total cell crosslinked gelatin. Cortactin localized with F-actin, lysate of cell lines grown on tissue culture treated identifying invadopodia as punctate sites of labeling plastic were analysed by SDS ± PAGE and Western (Figure 5a and b, solid arrows). This puntcate staining blotting (Figure 4a). The Western blot was ®rst probed represents short protrusions that extend into the gelatin for paxillin to establish the level of paxillin that was matrix beneath the cells perpendicular to the plane of immunoprecipitated from lysates of the cell lines view (Figure 5a, cortactin (red) and Figure 5b, examined (Figure 4a, Paxillin Blot). Cortactin and cortactin colocalizing with actin (yellow)). We con- PKCm co-immunoprecipitated with paxillin from the ®rmed this using confocal microscopy (data not whole cell lysates of only those cell lines characterized shown). Cortactin also colocalized with actin at as having a more motile and invasive nature, MCF-7- ru‚ing membranes (Figure 5a and b, arrowhead), ADR, MDA-MB-231, and HS578T (Figure 4a, whereas it did not colocalize with actin stress ®bers. Cortactin and PKCm Reblots and Table 1) (Thomp- To verify that the membrane protrusions identi®ed son et al., 1992). Vinculin, FAK, PKCa and PKCd did by labeling with cortactin were associated with ECM not co-immunoprecipitate with paxillin under these degrading activity, we utilized an immuno¯uorescent conditions (data not shown). We examined the assay that demonstrates membrane-associated degrada- expression levels of each of the proteins screened for tion. When MDA-MB-231 cells are cultured on co-immunoprecipitation with paxillin. In every case, crosslinked FITC-gelatin, the cells produce dark holes there was easily detectable expression of each protein of degradation against a background of bright (Figure 4b). Thus, we conclude that co-immunopreci- ¯uorescent matrix. We used two-color ¯uorescence pitation of cortactin and PKCm with paxillin does not microscopy to visualize either the FITC-gelatin matrix correlate with the expression levels of these proteins in (Figure 5c, e and f) or cortactin localization (Figure 5d, breast carcinoma cell lines, but rather with their e and f). Anti-cortactin monoclonal antibodies (mAbs) invasive and/or motile behavior. Since the complex localized cortactin to ru‚ing membranes (Figure 5d, was isolated from cells grown on plastic, we also arrowhead) and areas of punctate staining (Figure 5d, conclude that the cells do not require exogenous ECM solid arrow). The sites of punctate cortactin staining to generate the complex. co-localized with sites of invasion identifying these structures as invadopodia (Figure 5e and f, solid arrows). Image analysis was used to detect cortactin Cortactin, paxillin and PKCm co-localize at invadopodia and this overlay image, outlining cortactin staining, by immuno¯uorescence was merged with the two-color image to visualize the To con®rm the co-distribution of cortactin with areas of overlap (Figure 5f). We also performed double paxillin and PKCm at invadopodia, we immunoloca- labeling experiments with antibodies against cortactin lized these molecules in intact cells cultured on and paxillin on cells plated onto crosslinked gelatin. Paxillin localized to focal adhesions (Figure 5g, open arrow) as well as punctate sites where it colocalized with cortactin (Figure 5g and g', closed arrows). PKCm has previously been described as a Golgi localized protein (Prestle et al., 1996). The association of PKCm with cortactin and paxillin in the INV fraction indicates a novel plasma membrane localiza- tion. We con®rmed this previously unreported localiza- tion for PKCm by immuno¯uorescence. First, we examined the immuno¯uorescence localization of PKCm using a number of commercially available mAbs and pAbs. Anti-PKCm mAb (clone P26720) produced a perinuclear, Golgi-like staining (Figure 5h, solid arrow) consistent with the Golgi associated localization previously reported in the literature (Prestle et al., 1996). However, using two commer- cially available pAbs, we observed staining of the cell membrane, particularly at ru‚es (Figure 5i, arrow- head), as well as a perinuclear staining (Figure 5i, closed arrow). Staining, however, was particularly sensitive to detergent extraction. We con®rmed the Figure 4 (a) Immunoprecipitation of a complex of cortactin, paxillin and PKCm from invasive breast cancer cell lines. Breast speci®city of one of the pAbs using a synthetic peptide cancer cell lines were seeded into tissue culture ¯asks and lysates antigen to block speci®c binding. The peptide prepared after 24 h of culture. The invasion-related complex was eliminated the staining of both the Golgi and plasma isolated by immunoprecipitation of paxillin followed by SDS ± membrane components, con®rming the speci®city of PAGE and Western blotting as described in Materials and the antibody for PKCm (data not shown). methods. The complex immunoprecipitated with anti-paxillin antibodies contained cortactin, paxillin and PKCm and was The extensive membrane staining and Golgi isolated from the invasive cell lines MCF-7 ADR, MDA-MB- localization of PKCm rendered it extremely dicult to 231 and HS578T. (b) Breast cancer cell lines all express paxillin, observe immuno¯uorescent staining at invadopodia. cortactin and PKCm. Western blots of equal protein loading of However, we were able to detect the presence of PKCm total cell lysates from each of the cell lines were probed with anti- paxillin, cortactin, or PKCm mAbs. The characteristics of the in invadopodia after shearing away cell bodies. We breast cancer cell lines have been reported previously (Thompson ®xed cells in formaldehyde and then scraped the cell et al., 1992; Coopman et al., 1998) bodies from the surface of the FITC-gelatin matrix. Complex of cortactin, paxillin and PKCm in invadopodia ET Bowden et al 4445 After a brief Triton X-100 extraction, the remaining cellular material was immunostained with anti-PKCm antibodies. This process is in contrast to the cell fractionation protocol where there was no ®xation before shearing the cells. Fixation leaves ru‚es (Figure 5j, arrowhead) and invadopodia (Figure 5j, solid arrows) in contact with the matrix. Using this technique we con®rmed that PKCm staining co- localized with sites of matrix degradation (Figure 5j, solid arrows).

Discussion

In the present study, we demonstrate by immunofluor- escence, cell fractionation and Western blotting that cortactin, paxillin, and PKCm form a novel complex associated with invadopodial membranes that extend into a crosslinked matrix. Its presence in both invasive cell lines and its association with subcellular protru- sions at sites of matrix degradation suggest that the formation and membrane association of this complex contribute to the invasive phenotype of these cells.The inhibition of localized matrix degradation following microinjection of anti-cortactin antibodies supports the hypothesis that cortactin, in particular, plays a major role in mediating cellular invasion. Cortactin is an actin-bundling protein (Huang et al., 1997). Therefore, one role of cortactin might be to participate in ®lopodia formation by bundling actin. Alternatively, since cell surface associated proteases mediate matrix degradation, an association of cortactin with the membrane, as a component of the complex, may participate in the localization of matrix degrading proteases to invadopodia. Our working model illus- trates these two potential mechanisms leading to the formation of an active invadopodium (Figure 6). Our data currently cannot distinguish between these possibilities. Figure 5 Immuno¯uorescent localization of actin, cortactin, The speci®c function of cortactin in invadopodia paxillin and PKCm at invadopodia. MDA-MB-231 cells were may be modulated by its interactions with paxillin and cultured on FITC labeled or unlabeled gelatin ®lms coated onto glass coverslips. Primary antibodies were detected using either the protein serine kinase, PKCm. The invasion-related Texas red (red) or FITC (green) secondary antibodies. Actin was complex of cortactin/paxillin/PKCm was isolated from detected using FITC-phalloidin (green) and nuclei were detected invadopodia-enriched fractions and was not present at using DAPI (blue). Images were collected by conventional wide signi®cant levels in the CBM fraction although the ®eld ¯uorescence microscopy (a±f, h±j) or confocal microscopy (g). (a±f) Actin and cortactin co-localized in invadopodia. (a) component molecules were present. Since all three Cortactin localized to invadopodia in punctate spots (closed proteins, cortactin, paxillin, and PKCm also localize to arrows) and at the cell margin in ru‚ing membranes (arrowhead). invadopodia using immuno¯uorescence microscopy, (b) A three color overlay of DAPI-stained nuclei, cortactin and we conclude that the complex is localized to the actin illustrates co-localization of cortactin and actin at both membranes of invadopodia. Using a variety of ru‚ing membranes (arrowhead) and invadopodia (solid arrows). Invadopodia are visualized as a yellow color produced by the colocalization of red and green. (c±f) The anti-cortactin mAb labeled invadopodia at sites of invasion in an FITC-gelatin matrix. (c) Invasive MDA-MB-231 cells degrade areas of the FITC-gelatin (green) to leave dark spots in the matrix (solid magni®cation in g'). Again, invadopodia are identi®ed by the arrow). (d) Immuno¯uorescence localization of cortactin in the yellow produced by the colocalization of red and green. Note that same cells highlighted invadopodia (solid arrow), and ru‚ing cortactin did not stain focal adhesions. (g) Scale bar=10 mm. (g') membranes (arrowhead). (e) A two-color overlay of (c) and (d) Scale bar=2.5 mm. (h) MDA-MB-231 cells grown on a glass showed co-localization of cortactin staining at sites of invasion coverslip were probed with anti-PKCm mAb P26720. Golgi-like into the FITC-gelatin matrix (closed arrow). (f) Image analysis staining was observed in the perinuclear area (solid arrow). (i) can be used to highlight areas of cortactin staining (white outline) Cells grown on a glass coverslip were probed with anti-PKCm superimposed onto the two-color image. The formation of polyclonal D20. This staining was representative of that seen invadopodia changes over time however, and holes in the matrix using either the D20 or N20 anti-PKCm pAbs and showed ru‚ing leave a record of their activity. Therefore, some areas of membranes (arrowhead) and bright perinuclear staining (closed degradation are not associated with cortactin labeled invadopo- arrow). (h and i) Scale bar=10 mm. (j) Fixed cells were sheared dia. (a±f) Scale bars=10 mm. (g) Paxillin (Texas-Red) and from the surface of the FITC gelatin matrix as described in cortactin (FITC) colocalization at invadopodia. Paxillin predo- Materials and methods. Punctate sites of invadopodia labeled minantly stained adhesions (open arrow). Cortactin and paxillin with PKCm pAbs (red) and co-localized with areas of degradation colocalized at invadopodia (g, closed arrow, and at higher (closed arrows). Scale bar=5 mm Complex of cortactin, paxillin and PKCm in invadopodia ET Bowden et al 4446 size that the association of paxillin with cortactin in invadopodia in¯uences a downstream signaling path- way that is related to invasion and matrix degrada- tion. All three proteins from the complex localized to the CYT fraction, and the complex could also be immunoprecipitated from this fraction. Isolation of the complex from the cytosol might simply re¯ect a release from the membrane during the fractionation procedure. Alternatively, it is possible that the complex is in equilibrium between the cytosol and membrane. It is unlikely that the complex detected in the INV fraction represents complex formed in the cytosol that non-speci®cally sticks to the INV membranes. In support of this interpretation, the time course of distribution of the complex between the fractions demonstrated that there is no longer any complex associated with the CYT fraction at 48 h although it is present in the INV fraction. Previously, we and others, have demonstrated that with increasing time and cell density, there is a decrease in invasive behavior in vitro (Coopman et al., 1998; Tamaki et al., 1996; Xie et al., 1994). The absence of the complex from the CYT at 48 h suggests that the association of the complex with the invadopodia fraction is speci®c and supports our theory that the association of the complex with the CYT fraction is more transitory in nature. The loss of the complex in the CYT fraction therefore correlates with a decrease in invasive capacity suggesting that the cytosolic pool of the complex may be depleted. The role of the protein serine kinase PKCm in metastatic breast disease is unknown. However, several PKC isoforms have been directly implicated in the Figure 6 A model for the role of the cortactin-containing aggressive behavior of a number of tumor types, complex in the formation of invadopodia. The formation of an including PKCa in glioblastoma cells, where antisense active invadopodium involves a coordination of cytoskeletal rearrangements to form a membrane protrusion and the oligonucleotide therapy inhibited tumor growth localization and activation of ECM-degrading proteases at this (Yazaki et al., 1996). The presence of PKCm in these protruding membrane. Two possible sequences of events are complexes was initially puzzling given its previously illustrated. The cortactin-containing complex may ®rst bind to the reported localization to the Golgi (Prestle et al., 1996). membrane, coupling the actin cytoskeleton to the membrane to generate a membrane protrusion. The protease becomes linked to However, the association of PKCm with invadopodia- the assembling membrane protrusion by interaction with a localized complexes was demonstrated by co-immuno- transmembrane protein, e.g. an integrin or activating transmem- precipitation with paxillin and cortactin from lysates of brane protease. Alternatively, the complex may associate with the invasive cell lines and from invadopodia-enriched membrane protein ®rst and this complex might recruit ECM- membrane fractions. This association was speci®c, degrading proteases. Then, the coupling of the actin cytoskeleton to this complex would activate clustering of the complexes at sites other PKC isoforms including PKCa and PKCd were of membrane protrusion not co-immunoprecipitated. We found that the immuno¯uorescence localization of PKCm at invado- podia was dicult to detect because of sensitivity to detergent extraction. However, our results suggest that techniques we established that focal adhesions and polyclonal antibodies directed against PKCm, more focal adhesion associated proteins including FAK and sensitively detect other sites of subcellular localization vinculin, are not present in our invadopodia enriched including a punctate staining at the plasma membrane. fractions. We were somewhat surprised that FAK Even under optimized ®xation conditions, the mono- failed to associate with the complex of proteins at clonal anti-PKCm antibody failed to recognize mem- invadopodia as FAK binds to paxillin (Hildebrand et brane and cytosolic PKCm. We speculate that the role al., 1995; Tachibana et al., 1995). However, although of PKCm in invadopodia may be as a regulator of we did detect the co-association of paxillin and FAK invasive function via speci®c serine phosphorylation following immunoprecipitation of paxillin from the events. Thus, one role of PKCm may be to modulate CBM fraction, we did not detect this association in either complex formation or activity at invadopodia. the INV fraction (data not shown). We conclude that Finally, to reiterate, our data support the conclusion FAK and FAK-paxillin complexes are not prominent that the formation of a complex comprising cortactin, components of invadopodia. In summary, our data paxillin and PKCm is associated with the invasive suggest that FAK associates with paxillin in focal phenotype of breast cancer cells and that cortactin is adhesions but not in invadopodia. Paxillin may required for invadopodia mediated matrix degradation. therefore engage in di€erential signaling pathways We suggest that the activity of this complex may be to based upon its subcellular localization. We hypothe- coordinate the aggregation and/or activation of matrix Complex of cortactin, paxillin and PKCm in invadopodia ET Bowden et al 4447 degrading proteases together with cytoskeletal rearran- PKCm (data not shown). Fluorescently labeled and unlabeled, gements at forming invadopodia. Protein kinase anity puri®ed anti-mouse and anti-rat IgG+IgM secondary activity may regulate multiple steps during this process antibodies were from Jackson Laboratories (West Grove, PA, that results in the formation of matrix degrading USA) and HRP-conjugated secondary antibodies were from competent invadopodia. These points of regulation Rockland (Gilbertsville, PA, USA). remain to be elucidated and will be the subject of further studies. Transmission electron microscopy Cells (56105/ml) in complete culture medium were incubated 6 Materials and methods overnight with approximately 2610 large (4100 mm) gelatin beads at 378C. The mixture was rotated continuously, end- over-end, to promote the even attachment and spreading of Cell culture cells on the surface of the beads. The pellet of gelatin beads All cell lines were obtained from the Lombardi Cancer coated with cells was ®xed with 2% glutaraldehyde/3% Center Tissue Culture Shared Resource and originated from paraformaldehyde in PBS (150 mM NaCl, 1.9 mM the American Type Culture Collection (ATCC, Rockville, NaH2HPO4, pH 7.2) for 2 h. Post-®xation was in 1% MD, USA). Cells were maintained in Falcon ¯asks (Becton osmium tetroxide in distilled water for 1 h before processing and Dickinson Labware, Plymouth, UK) in a 1 : 1 mixture of for conventional ultrathin section electron microscopy as RPMI 1640 : DMEM (Mediatech, Washington, DC, USA) previously described (Coopman et al., 1996). Sections were supplemented with 10% bovine calf serum (Hyclone post-stained with uranyl acetate and lead citrate and Laboratories Inc., Logan, Utah, USA) and 5% NU-serum photographed with a JEOL 1200EX-transmission electron (Becton Dickinson Labware, Franklin Lakes, NJ, USA) and microscope operated at 60 kV. 2mML-glutamine (Gibco BRL, Gaithersburg, MD, USA). Cells were grown in the presence of 7.5% CO and passaged 2 Microinjection regularly to avoid post-con¯uent growth. Cells were examined regularly for the presence of Mycoplasma contam- MDA-MB-231 cells were seeded onto Eppendorf grided ination. coverslips in serum-free media (Fisher, Rockville, MD, USA) previously coated with FITC-crosslinked gelatin ®lms. After culture overnight at 378CinaCO2 incubator, Antibodies cells were microinjected using the Eppendorf microinjection Mouse anti-cortactin mAbs (clone 4F11) were generously system (Fisher) with either puri®ed preimmune IgG or the provided by Dr T Parsons (Department Of Microbiology, anti-cortactin IgG (#170-10) at 1 mg/ml, mixed 1 : 1 with University of Virginia, Charlottesville, VA, USA) or were 40 mg/ml DAPI in sterile PBS. Coverslips were returned to obtained from Upstate Biotechnology (Lake Placid, NY, the CO2 incubator for a 1 h recovery period. After an USA). Polyclonal antibodies (pAb) to chicken cortactin were exchange from serum-free, to serum-containing media, the prepared by injecting a glutathione S-transferase (GST)- cells were cultured for an additional 5 h. Finally, cells were cortactin fusion protein into rabbits. The constructs used to ®xed as described previously (Mueller et al., 1992). Injected generate GST-cortactin were also the kind gift of Dr Tom IgG was visualized by staining the coverslips with secondary Parsons. Fusion proteins were bacterially expressed and IgG Texas Red goat anti-rabbit IgG (Jackson Laboratories). was puri®ed from both immune and preimmune rabbit sera Microinjected cells were identi®ed using a Zeiss Photo- using Protein A chromatography (Pierce, Rockford, IL, scope II equipped with a 636 Neo¯uor 1.4 Zeiss lens. USA). Anti-GST antibodies were removed from both Injected cells were identi®ed as positive by the presence of preparations by anity chromatography using GST-Sephar- DAPI stained nuclei and cytosolic IgG immunolabeling. ose as described by the manufacturer (Pierce). We veri®ed Coverslips were analysed blind, and individual cells were that the anti-cortactin pAb (#170-10) recognized the same classi®ed as invasive if a cluster of at least four holes in the protein as the anti-cortactin mAb (4F11) by immunoprecipi- FITC-gelatin matrix was present underneath the cell body. tation and immuno¯uorescence experiments. Cortactin, 20 ± 50 non-microinjected cells and 20 ± 50 microinjected cells immunoprecipitated by the mAb (4F11) was Western blotted were counted per coverslip. The data are presented as a using either the pAb (#170-10) or the mAb (4F11). Both of percentage of invasive cells relative to total cells and the these antibodies recognized protein bands at p80/p85, mean and standard error were calculated from three corresponding to the molecular weight of cortactin (data experiments carried out on separate days. not shown). The anti-cortactin pAb (#170-10) also immu- nostained invadopodia, co-localizing with immunostaining of Cell fractionation and Western blotting the anti-cortactin mAb (4F11) in double labeling experiments (data not shown). MDA-MB-231 cells were cultured at 56106 cells/30 cm The mouse anti-paxillin (clone 349), anti-FAK (clone 2A7) diameter plate on crosslinked gelatin ®lms for 6, 24 or and anti-PKCm (clone P26720) mAbs were from Transduc- 48 h, according to a previously described protocol (Mueller et tion Laboratories (Lexington, KY, USA). The rat anti-b1 al., 1992). At 6 and 24 h cells were 60 ± 70% con¯uent. At integrin antibody (mAb 13) was a kind gift from Dr Steven 48 h post plating, cells were con¯uent. At each time point cell Akiyama (National Institute of Environmental Health bodies were sheared from the surface of the plates to leave an Science). Mouse anti-phosphotyrosine (clone 4G10) and invadopodia-enriched fraction embedded in the gelatin. The anti-vinculin mAbs were from Upstate Biotechnology. cell body fraction was further separated into the cell body Rabbit D20, C20 and N20 anti-PKCm polyclonal antibodies membranes and cytosol fractions (Mueller et al., 1992). (pAb) were from Santa Cruz Biotechnology, Inc., (Santa Brie¯y, each plate was washed in TBS and rinsed in tyrosine Cruz, CA, USA). D20 epitope binding-site blocking peptide phosphorylated protein (YPP) bu€er (10 mM MOPS, pH 6.8, was also from Santa Cruz Biotechnology, Inc. We determined 100 mM KCl, 2.5 mM MgCl2,1mM CaCl2, 0.3 M sucrose, that the anti-PKCm mAb recognized the same antigen as the 1mM vanadate, 2 mM PMSF, 2 mg/ml leupeptin, 2 mg/ml pAbs D20, C20 and N20 by immunoprecipitation and aprotinin and 0.02% NaN3). Cell bodies were sheared with a Western blotting. PKCm, immunoprecipitated by the mAb glass rod into a small volume of YPP bu€er. CBM were (P26720) was Western blotted using either the pAbs or the separated from the CYT by centrifugation at 9000 g for mAb (P26720). All of these antibodies recognized a protein 20 min at 48C. The invadopodial membranes, embedded in band at p115, corresponding to the molecular weight of the gelatin matrix, were rinsed in YPP bu€er before being Complex of cortactin, paxillin and PKCm in invadopodia ET Bowden et al 4448 scraped up with the crosslinked gelatin into immunoprecipi- 18 mm gelatin coated coverslip) were incubated overnight at tation bu€er (50 mM Tris, pH 8.0, 50 mM NaCl, 0.005% 378C, then ®xed and labeled by indirect immuno¯uorescence deoxycholate, 1% NP40, 1 mM vanadate, 50 mM NaF, 2 mM as previously described (Mueller et al., 1992). To assist in the PMSF, 2 mg/ml leupeptin, 2 mg/ml aprotinin). Each fraction visualization of invadopodia, cell bodies were mechanically was solubilized in immunoprecipitation bu€er at 48C for sheared from the surface of the gelatin, after the cells had 20 min and samples were taken for protein determination been ®xed, using a glass rod. Indirect immuno¯uorescence (bicinchoninic acid assay, Pierce). For immunoprecipitation was performed using a range of mAbs and immunostaining experiments 500 mg of each fraction was incubated with 2 mg was visualized using anity-puri®ed Texas red goat anti- of either anti-cortactin or anti-paxillin mAbs for 2 h at 48C. mouse antibodies. All primary antibodies were used at 1 : 50 ± A15ml suspension of Protein A-Sepharose beads (Boehringer 1 : 300 dilution in PBS and all secondary antibody incuba- Mannheim, Indianapolis, IN, USA) was incubated with 2 mg tions were at a 1 : 200 dilution in PBS. Actin was localized unlabeled anity-puri®ed rabbit anti-mouse IgG+IgM for using FITC labeled phalloidin (Molecular Probes, Inc., 2 h at 48C. After the preincubation period, beads were added Eugene, OR, USA) at a dilution of 1 : 400. Nuclei were to the lysate and primary antibody mixture, and incubated stained with 4,6-diamidino-2-phenylindole (DAPI) at a ®nal overnight at 48C. Beads were washed six times in concentration of 80 ng/ml. Coverslips were mounted using immunoprecipitation bu€er and proteins eluted from the the ProLong antifade kit (Molecular Probes, Inc.). Conven- beads by boiling in SDS ± PAGE sample bu€er under tional immuno¯uorescence images were digitized using Image reducing conditions. Protein immuno-complexes were sepa- Pro Plus from Media Cybernetics (Silver Spring, MD, USA) rated by SDS ± PAGE in 6% gels and analysed by Western and a Toshiba integrating camera from I-Cube (Crofton, blot analysis. MD, USA). Confocal images were collected on a Zeiss 410 To characterize the invadopodia fraction morphologically, confocal microscope at a resolution of 0.1323 pixels per representative plates were also processed for immunofluores- micron using a 636/1.4 Zeiss immersion lens. cence. After shearing, and a brief wash in YPP bu€er, gelatin plates and their embedded invadopodia were ®xed according to the described protocol for immuno¯uorescence. The plates were stained with either Texas-Red phalloidin to visualize F- actin, or anti-vinculin antibodies to visualize focal adhesions. Finally, the plates were ¯ooded with PBS and examined using a Zeiss 406 water immersion lens. Acknowledgements For total protein lysates, cells were grown in tissue culture We gratefully acknowledge Tom Parsons for mouse mAb ¯asks to 70 ± 80% con¯uency, washed in ice-cold TBS anti-cortactin clone 4F11 and the GST-cortactin constructs (150 mM NaCl, 50 mM Tris/HCl, 50 mM Tris pH 7.5) and used to generate fusion protein for pAb production. We suspended in 4 ml of immunoprecipitation bu€er. Cells were would like to thank Carolyn L Smith and the Light solubilized for 15 min at 48C and samples taken for protein Imaging Facility (National Institute for Neurological determination. Five hundred mg of total protein lysate was Disorders and Stroke) for assistance with confocal used in immunoprecipitation experiments as described above. microscopy. We also thank Sandra McLeskey and Steven Also Western blots of lysate and cell fractions were run using Byers for critical reading of the manuscript and Donna 10 mg of protein per lane. Almario and Maozheng Dai for technical assistance. This work was supported in part by the National Institutes of Health grants R01DK48910, R21CA62232 and Immuno¯uorescence microscopy R01CA61273 to SCM, R01CA57244 and R01NS34431 to MDA-MB-231 cells were cultured on crosslinked gelatin or RIG and by the Lombardi Cancer Center Microscopy/ FITC-labeled, crosslinked gelatin ®lms overnight as pre- Imaging and Tissue Culture shared resources supported by viously described (Mueller et al., 1992). Cells (76104 per US Public Health Service Grant 1P30CA51008.

References

Abedi H, Dawes KE and Zachary I. (1995). J. Biol. Chem., Lewis JM, Cheresh DA and Schwartz MA. (1996). J. Cell 270, 11367 ± 11376. Biol., 134, 1323 ± 1332. Bellis SL, Perrotta JA, Curtis MS and Turner CE. (1997). McCormick SJ, Brazinski SE, Moore JL, Werness BA and Biochem. J., 325, 375 ± 381. Goldstein DJ. (1997). Oncogene, 15, 265 ± 274. Burridge K, Turner CE and Romer LH. (1992). J. Cell Biol., Miyata Y, Nishida E, Koyasu S, Yahara I and Sakai H. 119, 893 ± 903. (1989). J. Biol. Chem., 264, 15565 ± 15568. Campbell DH, de Fazio A, Sutherland RL and Daly RJ. Monsky WL, Kelly T, Lin C-Y, Yeh Y, Stetler-Stevenson (1996). Int. J. Cancer, 68, 485 ± 492. WG, Mueller SC and Chen W-T. (1993). Cancer Res., 53, Chen W-T. (1989). J. Exp. Zool., 251, 167 ± 185. 3159 ± 3164. Coopman PJ, Do MTH, Thompson EW and Mueller SC. Monsky WL, Lin C-Y, Aoyama A, Kelly T, Mueller SC, (1998). Clin. Cancer Res., 4, 507 ± 515. Akiyama SK and Chen W-T. (1994). Cancer Res., 54, Coopman PJ, Thomas DM, Gehlsen KR and Mueller SC. 5702 ± 5710. (1996). Mol. Biol. Cell, 7, 1789 ± 1804. Moonga BS, Stein LS, Kilb JM and Dempster DW. (1996). Hildebrand JD, Schaller MD and Parsons JT. (1995). Mol. Calcif. Tiss. Int., 59, 105 ± 108. Biol. Cell, 6, 637 ± 647. Mueller SC and Chen W-T. (1991). J. Cell Sci., 99, 213 ± 225. Huang C, Lui JL, Haudenschild CC and Zhan X. (1998). J. Mueller SC, Yeh Y and Chen W-T. (1992). J. Cell Biol., 119, Biol. Chem., 273, 25770 ± 25776. 1309 ± 1325. HuangC,NiYS,WangT,GaoYM,HaudenschildCCand Nakahara H, Howard L, Thompson EW, Sato H, Seiki M, Zhan X. (1997). J. Biol. Chem., 272, 13911 ± 13915. Yeh YY and Chen W-T. (1997). Proc. Natl. Acad. Sci. Kelly T, Mueller SC, Yeh Y and Chen W-T. (1994). J. Cell USA, 94, 7959 ± 7964. Physiol., 158, 299 ± 308. Nakahara H, Nomizu M, Akiyama SK, Yamada Y, Yeh Y Klemke RL, Yebra M, Bayna EM and Cheresh DA. (1994). and Chen W-T. (1996). J. Biol. Chem., 271, 27221 ± 27224. J. Cell Biol., 127, 859 ± 866. Complex of cortactin, paxillin and PKCm in invadopodia ET Bowden et al 4449 O'Brian CA and Ward NE. (1989). Cancer Metast. Rev., 8, Turner CE, Glenney Jr JR and Burridge K. (1990). J. Cell 199 ± 214. Biol., 111, 1059 ± 1068. Okamura H and Resh MD. (1995). J. Biol. Chem., 270. van Damme H, Brok H, Schuuring-Scholtes E and 26613 ± 26618. Schuuring E. (1997). J. Biol. Chem., 272, 7374 ± 7380. Patel AS, Schecter GL, Wasilenko WJ and Somers KD. Vuori K and Ruoslahti E. (1995). J. Biol. Chem., 270, (1998). Oncogene, 16, 3227 ± 3232. 22259 ± 22262. Prestle J, P®zenmaier K, Brenner J and Johannes FJ. (1996). Ways DK, Kukoly CA, de Vente J, Hooker JL, Bryant WO, J. Cell Biol., 134, 1401 ± 1410. Posekany KJ, Fletcher DJ, Cook PP and Parker PJ. Schuuring E, Verhoeven E, Litvinov S and Michalides RJ. (1995). J. Clin. Invest., 95, 1906 ± 1915. (1993). Mol. Cell Biol., 13, 2891 ± 2898. Wu H and Parsons JT. (1993). J. Cell Biol., 120, 1417 ± 1426. Schuuring E, Verhoeven E, Mooi WJ and Michalides RJ. Xie B, Bucana CD and Fidler JJ. (1994). Am.J.Pathol.,144, (1992). Oncogene, 7, 355 ± 361. 1058 ± 1067. Tachibana K, Sato T, D'Avirro N and Morimoto C. (1995). Yazaki T, Ahmad S, Chahlavi A, Zylber-Katz E, Dean NM, J. Exp. Med., 182, 1089 ± 1099. Rabkin SD, Martuza RL and Glazer RI. (1996). Mol. Tamaki M, McDonald W and Del MR. (1996). J. Pharmacol., 50, 236 ± 242. Neurosurg., 84, 1013 ± 1019. Yebra M, Filardo EJ, Bayna EM, Kawahara E, Becker JC Thompson EW, Paik S, Brunner N, Sommers CL, Zugmaier and Cheresh DA. (1995). Mol. Biol. Cell, 6, 841 ± 850. G, Clarke R, Shima TB, Torri J, Donahue S, Lippman ME, Martin GR and Dickson RB. (1992). J. Cell Physiol., 150, 534 ± 544.