Functional role of α-actinin, PI 3-kinase and MEK1/2 in insulin-like growth factor I receptor kinase regulated motility of human breast carcinoma cells

Differentiated epithelial cells are organized in contiguous layers that protect the underlying connective tissue from the external environment and comprise the parenchyma of the inner organs such as the glands. The lateral membranes of these adherent cells are connected via highly specialized adhesive junctions whose formation and maintenance is critically dependent on linkage to the cytoskeleton: tight and adherens junctions are linked to the cortical actin, whereas desmosomes are connected to the network of intermediate filaments (Cowin and Burke,1996; Davies and Garrod,1997). This unique tissue architecture allows epithelial cells to remain relatively static, except for during cell separation and migration that occur in development, tissue regeneration and metastasis(Gumbiner, 1996).

Some mechanistic aspects and common sequential changes that precede cell separation have been determined using animal and cellular models (reviewed inHay, 1995;Thiery and Chopin, 1999;Savagner, 2001). For example,epithelial colonies can be induced to separate by treatment with physiological peptides, including hepatocyte growth factor/scatter factor (HGF/SF),neuregulin, members of the fibroblast growth factor (FGF), epidermal growth factor (EGF) and transforming growth factor (TGF) families(Stoker and Perryman, 1985;Savagner et al., 1997;Chausovsky et al., 1998;Müller et al., 1999). Peptides with cell scattering function induce a disruption of intercellular junctions, a reorganization of the actin cytoskeleton and an increase in local cell motility. These effects are mediated by receptor tyrosine kinases(Sachs et al., 1996), of which the most extensively studied is Met, the receptor for HGF/SF (reviewed inFurge et al., 2000). Activation of Met has a strong effect on scattering and invasiveness of the rodent mammary epithelial cells and various human carcinoma cell lines, and similar disorganization is believed to lead to metastasis (Wiedner et al.,1990; Liang et al., 1996;Firon et al., 2000). It is less clear, however, whether HGF/SF and Met control motile behavior of human breast tumor cells. No scattering response to HGF/SF has been seen in human breast carcinoma cell lines in vitro(Stoker et al., 1987); in breast tissue biopsies, the intensity of Met staining was lower in tumor tissue than in normal cells adjacent to mammary ducts(Tsarfaty et al., 1992).

Another member of the receptor tyrosine kinase family is the receptor for IGF-I, a protein of particular interest for the biology of the human mammary gland. There is a general requirement for IGF-I in terminal end bud formation and ductal morphogenesis in the developing gland(Kleinberg et al., 2000). In humans, the IGF-IR is expressed by both normal and tumor mammary cells; yet the level of the receptor expression and its tyrosine kinase activity is higher in malignant compared with benign and normal tissues(Happerfield et al., 1997;Resnik et al., 1998). The signaling mechanisms evoked by IGF-IR kinase are well documented (reviewed byLeRoith, 2000). Ligand-induced activation of the IGF-IR causes the sequential phosphorylation of three conservative tyrosines within the catalytic domain, followed by phosphorylation of the cytoplasmic domain of the receptor, recruitment of adapter molecules and initiation of signaling cascades. The two best-characterized pathways triggered by the activation of this receptor are the PI3-kinase signaling via the insulin receptor substrates (IRSs) and the MAP kinase pathway via adaptor molecules Shc. Although IGF-I may have a pleiotropic effect on cell proliferation, differentiation and migration(Leventhal and Feldman, 1997;Lee et al., 1998;Chernicky et al., 2000), the range of IGF-IR activities is critically dependent on the cell context(Baserga, 1999). There is good evidence that in cultured mammary epithelial cells the IGF-IR stimulates not only proliferative but also motile cell responses(Kohn et al., 1990;Doerr and Jones, 1996;Mira et al., 1999). We have recently reported that IGF-I treatment of breast cancer cells overexpressing IGF-IR results in a disruption of the epithelial sheet, implying a role for this receptor in concomitant reorganization of the actin cytoskeleton and cellular junctions (Guvakova and Surmacz,1999).

The most prominent epithelial structures relying on the actin cytoskeleton are adherence junctions. The core of these junctions is composed of transmembrane E-cadherins and cytoplasmic α,β- and γ-catenins(reviewed by Gumbiner, 2000).α-catenin directly or via actin-binding proteins α-actinin and vinculin couples cadherin-catenin complexes to microfilaments(Knudsen et al., 1995;Nieset et al., 1997;Vasioukhin et al., 2000). Such a linkage is essential for the stabilization of intercellular adhesion and the maintenance of mammary epithelial tissue(Tsukatani et al., 1997). On the basis of the correlation between upregulation of the IGF-IR in breast tumors and the effects of the overexpressed IGF-IR on the actin cytoskeleton,we hypothesize that high levels of IGF-IR activity can induce destabilization of cell-cell contacts. We report a novel signaling effect of the IGF-IR kinase in coupling α-actinin and actin-containing microspikes by a PI 3-kinase-dependent mechanism at an early stage of epithelial cell separation.

MCF-7 human breast cancer cell line (ATCC, HTB-22) and its derivatives were all grown in Dulbecco's Modified Eagle Medium: Nutrient Mixture F-12 (Ham) 1:1(D-MEM/F-12) containing 5% calf serum (CS), penicillin and streptomycin at 37°C in 5% CO₂. For overnight serum starvation, MCF-7 cells were incubated in phenolred- and serum-free D-MEM/F-12 containing 0.5 mg/ml bovine serum albumin (BSA). The derivation of a clonal line of MCF-7 that overexpressed the wild-type human IGF-IR 18-fold has been previously described(Guvakova and Surmacz, 1997). This line is referred as MCF-7/IGF-IR/WT in this study. D-MEM/F-12,phenol-red-free D-MEM/F-12, CS and Geneticin (G418) were purchased from Gibco BRL, Life Technologies. LY 294002, PD 98059 and the Akt inhibitor were from Calbiochem, UO126 from Promega and human recombinant IGF-I from Bachem.

MCF-7/IGF-IR/DK kinase-deficient cells were generated using a IGF-IR βsubunit with mutations in the activation loop of the catalytic domain (three tyrosine residues, Y1131, Y1135, and Y1136 were substituted by phenylalanines). This DNA was a gift from D. Leroith, NIH(Kato et al., 1994). The mutant IGF-IR cDNA was subcloned from the pBluscript II KS+ vector into the EcoRI and XbaI sites of the pcDNA3 expression vector(Invitrogen Corp., San Diego, CA) under the control of the CMV promoter. MCF-7 cells expressing the mutant receptor were derived by calcium-phosphate-precipitation-mediated transfection with a plasmid encoding the mutant hIGF-IR followed by selection in 2.5 mg/ml G418. Surface expression of the IGF-IR was confirmed by fluorescence-activated cell sorting (FACS)using a monoclonal antibody against the α subunit of the IGF-IR and a secondary FITC-conjugated goat anti-mouse antibody (Vector). Stable clones expressing approximately 1.0×10⁶ IGF-IRs per cell were isolated.

For functional analysis of α-actinin, new DNA constructs encoding the truncated α-actinins with enhanced green fluorescent protein (EGFP)tagged to the C-terminus were generated. A full-length chicken sarcomericα-actinin (GenBank™ accession number X59247) was obtained from J. Sanger (University of Pennsylvania)(Dabiri et al., 1997). A 230 amino-acid N-terminal truncation, including residues 1-24 in the potential site for regulation of high-affinity binding of α-actinin to actin(Xu et al., 1998) and residues 108-189 in the actin-binding site determined by mutagenesis(Hemmings at al., 1992), was produced by excision of a unique 5′ Eco47III-EcoRV restriction enzyme fragment, followed by blunt-end religation of the truncated plasmid DNA. In this mutant α-actinin RNA, the authentic methionine start codon is removed, and the translation is initiated at the AUG codon for methionine 231. The deletion of the C-terminal tail of α-actinin was generated by excision of a SacI fragment from the plasmid encoding full-length α-actinin, followed by religation of the fragment into the SacI site of pEGFPN1 vector. The authentic initiating codon is retained in this mutant. The removed region encodes 168 amino acids, including a large part of a vinculin-binding site [residues 713-749(McGregor et al., 1994)] and both EF-hand Ca²⁺-binding motifs(Baron et al., 1987). The proper deletions were confirmed by restriction mapping of the N- and C-terminal regions of α-actinin cDNA. The position of the truncatedα-actinins in frame with EGFP was verified using the Vector NTI version 3.0 program (InforMax, Inc.).

In transient transfection experiments, 5×10⁴MCF-7/IGF-IR/WT or MCF-7/IGF-IR/DK cells were plated into a 35 mm glass-bottomed culture dish (MatTek Corporation, Ashland, MA). The next day,the cells were transfected with plasmids encoding EGFP-α-actinin constructs or the pEGFP-N1 control vector (Clontech) using Lipofectamine Plus Reagent (Gibco, BRL). After 3 hours, the cells were washed in D-MEM/F-12 and incubated overnight in the conditioned medium of MCF-7/IGF-IR/WT cells to improve the survival of the transfectants. The living cells were used in experiments 48 hours after transfection, when approximately 40% of cells produced fluorescently labeled protein.

Cells were lysed in buffer containing 50 mM Hepes pH 7.5, 150 mM NaCl, 1.5 mM MgCl, 1 mM EGTA, 10% glycerol, 20 μg/ml aprotinin, 1 mM phenylmethylsulfonyl fluoride, 2 mM Na orthovanadate and either 1% Triton X-100 or 1% SDS. For western blots, 20 μg of total protein was resolved by SDS-polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to Hybond ECL nitrocellulose (Amersham Pharmacia Biotech). In a series of experiments,the blots were probed with antibodies against IGF-IR (rabbit polyclonal anti-IGF-IR β subunit, Santa Cruz, SC); IRS-1 (rabbit polyclonal anti-IRS-1, Upstate Biotechnology Inc., UBI, Lake Placid, NY); α-actinin(mouse IgG clone AT6/172, UBI and goat polyclonal anti-α-actinin antobodies, SC); α-catenin (rabbit polyclonal antibodies, Sigma); GFP(mouse monoclonal antibody, Clontech Laboratories, Inc.); phospho-Akt (rabbit anti-phospho-Akt (Ser 473), rabbit anti-phospho-Akt (Thr 308), New England Biolab, NEB); Akt (rabbit anti-Akt, NEB); phospho MAPK (mouse monoclonal E10,anti-phospho-p44-42 MAPK (Thr 202/Tyr 204), NEB); ERK1/ERK2 (rabbit polyclonal k23 anti-ERK1, SC), and phospho-tyrosine-containing proteins (mouse monoclonal PY-20 anti-phosphotyrosine, SC). For immunoprecipitation of α-actinin,500 μg of total protein from 1% SDS cellular extracts was incubated with 4μg of anti-α-actinin monoclonal IgG₁, clone AT6/172 (UBI)and 50 μl anti-mouse IgG-agarose beads (Sigma) in HNTG buffer (20 mM Hepes pH 7.5, 150 mM NaCl, 0.1% Triton X-100, 10% glycerol) overnight at 4°C. The immunoprecipitated proteins were resolved by SDS-PAGE and blotted with anti-α-actinin and PY 20 antibodies. To visualize primary antibody-bound protein, the secondary antibodies conjugated to horseradish peroxidase(1:10,000 dilution; Calbiochem) and the ECL detection solutions (Amersham Pharmacia, UK) were applied. The chemiluminescent intensity of the bands was digitized using the Image Analyser LAS-1000 plus system and the Image Reader LAS-1000 Lite version 1.0 software (Fuji). When several proteins were to be detected on the same membrane, nitrocellulose was incubated in stripping buffer (100 mM 2-mercaptoethanol, 2% SDS, 62.5 mM Tris-HCl pH 6.7) for 30 minutes at 50°C, then washed three times in washing buffer (100 mM Tris pH 7.5, 1M NaCl, 1% Tween-20), blocked in 5% BSA in PBS and re-probed. The relative activity of the protein kinase B (PKB/Akt) and mitogen-activated protein (MAP) kinases was quantified as the ratio of phosphorylated protein to the corresponding total protein detected on the same blot.

To visualize filamentous actin (F-actin), cells grown on 22 mm (no. 1)glass coverslips were fixed with 3.7% formaldehyde in PBS for 15 minutes,permeabilized with 0.05% Triton X-100 in PBS for 5 minutes and stained with tetramethylrhodamine B isothiocyanate (TRITC)-phalloidin (1 μg/ml) (Sigma)for 30 minutes. Changes in the intracellular distribution of α-actinin were assessed in formaldehyde-fixed cells using an anti-α-actinin antibody (clone BM-75.2; Sigma) (1:200) followed by fluorescein isothiocyanate(FITC)-conjugated goat anti-mouse antibody (Vector). Distribution ofα-catenin was assessed in formaldehyde-fixed cells using an anti-α-catenin antibody (Sigma) (1:200) followed by a Rhodamine-Red-conjugated goat anti-rabbit antibody (Jackson ImmunoReseasrch Laboratories, Inc). Fascin was visualized in methanol-fixed cells using antibody 55k-2 as described previously(Guvakova et al., 2002). The samples were examined under a Nikon Eclipse E600 MRC 1024 confocal laser-scanning (Bio-Rad) microscope with a Plan Apo 60×/1.4 oil objective lens (Nikon). Images were acquired using the Bio-Rad LaserSharp 2000 software run under the OS/2 Warp ConnectTM operating system. Images were collected as stacks of xy optical sections (z-series) taken at different focal planes within the sample.

Cells were grown to confluence in regular culture medium on 60 mm tissue culture dishes (Falcon). Following overnight serum starvation, cells were either stimulated with 50 ng/ml IGF-I for 15 minutes or kept in serum-free medium at 37°C until fixing. Cells were fixed in fresh fixative (1%glutaraldehyde, 1% OsO₄ in 0.05M phosphate buffer, pH 6.2) for 45 minutes at 4°C (Tilney and Tilney,1994), washed three times in water at 4°C to remove phosphate,stained with 1% uranyl acetate overnight in the dark at 4°C, dehydrated in ethanol and 2-hydroxypropyl methacrylate and then embedded in an epon substitute LX-112 (Ladd Research Industries Inc, VT). Sections were cut with a diamond knife and examined with a Philips 200 electron microscope (Philips Scientific, Mahwah, NJ). Photographs were taken of sections mounted on uncoated grids. Images of the scanned photographs were processed as Adobe Photoshop files for the figures presented.

Cells grown in glass-bottomed dishes in D-MEM/F-12 containing 15 mM HEPES buffer were placed within a chamber heated at 37°C mounted on the stage of a Nikon Eclipse TE 300 confocal laser-scanning (Bio-Rad) microscope. Multiple replicate fields of cells were observed over 10 minutes in serum-starved cultures or were followed for 10 minutes in the presence of 50 ng/ml IGF-I. For the IGF-I time course study, four-dimensional (4D) data sets (sequences of z-series collected over time) were acquired using the acquisition function of the LaserSharp 2000 (Bio-Rad) software. Digital confocal images were captured in real time as a z series (z-step=0.5 μm) throughout the thickness of each cell. The 4D series was reconstructed into a 3D projection sequence using the LaserSharp processing function. Single sections were selected for export into Photoshop files for the figures presented.

MCF-7, MCF-7/IGF-IR/WT or MCF-7/IGF-IR/DK cells were plated at low density in DMEM/F-12 in Slideflasks (Nunc, Denmark) and allowed to adhere, spread over and establish cell-cell contacts. Time-lapse video microscopy was carried out in a 37°C environmental chamber using a Zeiss Axiovert 100 microscope linked by a Sony SS-M37OCE CCD camera to a video recorder driven by an EOS BAC 900 animation controller. Recordings were made at a rate of 10 frames per minute over a 60-80 minute period for each sample. Each sample contained between 46 to 102 cells per field, and all conditions were analyzed in three independent experiments. Alterations in cell motility and shape were measured from traces of the video images documenting (a) paths of the cell centroid,which were used to calculate cell velocities as movement over time; (b) the changes in cell outline over time, from which the percentage of cells with edge ruffles was calculated; and (c) the appearance of apical ruffles, which appeared as dynamic phase dark ridges at the nuclear and perinuclear regions of the apical cell membrane. In experiments involving IGF-I treatment, video recording was started within 1 minute of adding the growth factor. A QuickTime movie was generated from video records using the Improvision Openlab software package.

In the present study, we examined the effects of signaling from the IGF-IR kinase on motile behavior of human mammary epithelial cells. For this purpose,we have used parental MCF-7 cells, an MCF-7 cell line overexpressing the wild-type IGF-IR [MCF-7/IGF-IR/WT(Guvakova and Surmacz, 1997)]and novel clonal lines expressing a kinase-dead form of the IGF-IR (designated MCF-7/IGF-IR/DK). Quantitative western blotting was used to compare the level of the IGF-IR in parental MCF-7 cells with clonal lines, which overexpressed either the active or kinase-inactivated IGF-IR. In both transfected cell lines, the receptor was about 20-fold higher than the endogenous level in parental MCF-7 cells (Fig. 1a).

To compare receptor signaling over time in transfected cells,phosphorylation of the IGF-IR and IRS-1 was examined in the wild-type and kinase-dead transfectants treated with IGF-I. Cells expressing the wild-type IGF-IR had higher tyrosine phosphorylation of both the IGF-IRβ subunit and IRS-1 between 5 and 60 minutes following IGF-I treatment. By contrast,MCF-7/IGF-IR/DK cells showed a complete lack of tyrosine phosphorylation of either the IGF-IRβ subunit or IRS-1(Fig. 1b). These data establish that IGF-IR signaling is blocked in MCF-7/IGF-IR/DK cells at the first step of receptor kinase activation and IRS-1 phosphorylation.

We next compared migratory behavior of the parental MCF-7 cell line and its derivative clones in response to IGF-I. Parental cells treated with IGF-I became migratory and pulled apart from one another at cell-cell contacts within colonies. Cell surfaces changed markedly, exhibiting intensive membrane ruffling over the apex and at the cell margins. When clones were treated with IGF-I, MCF-7/IGF-IR/WT but not MCF-7/IGF-IR/DK cells developed protrusions,separated from one another and became motile in the direction of intense peripheral ruffling. The MCF-7/IGF-IR/WT cells initially located at the edges of colonies became most motile; these cells detached completely from neighboring cells and migrated freely as single cells until they came into contact with other cells (Movie 1; available at jcs.biologists.org/supplemental).

To estimate the differences in cell migratory behavior in response to IGF-I, we compared ruffling activity and cell movement rates in MCF-7/IGF-IR/WT and MCF-7/IGF-IR/DK cells. Before stimulation, small zones of minor membrane ruffles at the free edges of cells were observed occasionally in both cell types, but apical membrane activity was undetectable. Stimulation with IGF-I resulted in a dramatic increase in the membrane ruffling of MCF-7/IGF-IR/WT over the period of 1 hour and that correlated temporally with cell-cell separation (Fig. 2A). 100% of IGF-I-stimulated MCF-7/IGF-IR/WT cells developed prominent apical membrane activity, and the percentage of cells with peripheral, phase dark ruffling activity doubled (Fig. 2Ba-c). The rates of cell motility were related to the initial position of cells within colonies and varied between 20.7±2.4μm/hour (mean±s.e.m.) for cells located inside the colonies and 59.7±7.6 μm/hour (mean±s.e.m.) for cells at the periphery of colonies.

In sharp contrast, MCF-7/IGF-IR/DK cells showed no increase in cortical ruffling upon IGF-I stimulation, and less than 20% of MCF-7/IGF-IR/DK cells developed apical membrane protrusions (Fig. 2Bd-f). Furthermore, only 17% of the cells showed any motile activity with mean velocities 3.8±1.6 μm/hour inside the colonies and 13.9±1.9 μm/hour at the colony margins. These results established that in MCF-7 cells IGF-IR tyrosine kinase activity was essential for the formation of apical membrane projections, the increase of lateral ruffling and separation of cells followed by migration.

In MCF-7 cells, the activation of the IGF-IR causes reorganization of the actin cytoskeleton (Guvakova and Surmacz,1999). Here, we used indirect immunofluoresence to assess whether in separating cells reorganization of F-actin in cell-cell contacts is coupled to the distribution of α-catenin and α-actinin, two principle actin-binding proteins that connect microfilaments to adherens junction receptor complexes. In serum-starved MCF-7/IGF-IR/WT cells, α-catenin and α-actinin formed continuous lines along the borders of mature cell-cell contacts, where F-actin was also localized extensively(Fig. 3Aa,e,i). Within 5 minutes of IGF-I stimulation, about 30-35% of cells had punctuated instead of continuous distribution of α-catenin, which indicated the onset of the disruption of adherens junctions. By contrast, α-actinin appeared highly concentrated at cell margins, and the short actin-enriched projections appeared at cell-cell contacts (Fig. 3Ab,f,j). The formation of the cortical α-actinin and F-actin structures correlated temporally with the loosening of cell-cell contacts and membrane ruffling documented by time-lapse video microscopy. By 15 minutes, the majority of the cells (up to 85%) had discontinuousα-catenin staining at cell-cell borders. In sharp contrast,α-actinin and F-actin continue to highly concentrate within marginal membrane protrusions (Fig. 3Ac,g,k). At 60 minutes, α-catenin localized diffusely in the cytoplasm and in the rare remaining cell-cell contacts, whereasα-actinin and F-actin concentrated at the edges of multiple membrane protrusions (Fig. 3Ad,h,l). These findings indicated that IGF-I-induced separation of MCF-7 cells was associated with the progressive reduction of α-catenin and the increase of α-actinin and F-actin at cell-cell contacts.

The reduction of α-catenin in cell-cell contacts can be achieved experimentally by lowering the concentration of extracellular calcium ions. This process is thought to be associated with unclustering and/or internalization of Ecadherin-catenin complexes(Kusumi et al., 1999). When confluent MCF-7/IGF-IR/WT cells were shifted to PBS containing 0.5 mM EDTA or PBS free of calcium, α-catenin quickly translocated from cell-cell contacts to the cytoplasm before dissociation of the cells. This procedure,however, did not stimulate the assembly of cortical structures containingα-actinin and F-actin or cell motility (data not shown). We propose that IGF-IR-mediated reorganization of the actin cytoskeleton at cell-cell contacts is specific for the process of `active cell separation' followed by migration as opposed to experimentally induced `passive cell dissociation' unrelated to cell movement.

To determine the role of the IGF-IR kinase activity in active cell separation, we analyzed the effect of IGF-I on localization of α-actinin and F-actin in MCF-7/IGF-IR/DK cells, in which the dominant-negative receptor no longer activated IGF-IR signaling. In these cells, IGF-I treatment did not alter localization of either α-actinin or F-actin over a 60 minute period (Fig. 3B). The cells continued to have apico-basal polarity and to be organized in tight patches. Thus, in MCF-7 cells, both disorganization of cell-cell contacts and the concomitant concentration of α-actinin and F-actin at cell peripheries are dependent on the catalytic activity of the IGF-IR kinase.

To examine in detail the sites of α-actinin and F-actin colocalization before and after IGF-I stimulation, we used double immunofluorescence staining (Fig. 4A). In serum-starved MCF-7/IGF-IR/WT cells, α-actinin co-distributed with F-actin in a circumferential band, small cytoplasmic aggregates and the tips of the stress fiber-like filaments(Fig. 4Ac). Within 5 minutes of IGF-I stimulation, F-actin appeared to be depleted from the cytoplasm and was instead concentrated in short (∼2 μm) bundles that form highly ordered arrays of spikes at cell peripheries (Fig. 4Ad). At the same time, α-actinin showed a striking enhancement in its colocalization with F-actin at the basal portions of actin bundles within microspikes. Notably, there was no α-actinin in the extremities of microspikes where F-actin persisted in the tight bundles(Fig. 4Af). These experiments revealed that IGF-I activates localization of α-actinin at the base of actin microspikes organized in well ordered arrays in lateral zones of separating cells.

To examine more deeply the structure of microspikes in the separating cells, we repeated the IGF-I stimulation experiments, this time processing cells for transmission electron microscropy (TEM). Confluent serum-starved cells had intercellular contacts typical of polarized epithelium(Fig. 4Ba). Within 15 minutes of IGF-I stimulation, finger-like projections formed in the lateral zones of the opposing cells (Fig. 4Bb). That correlated temporally with appearance of α-actinin-actin microspikes identified by immunofluorescence. The average diameter of the membrane-flanked individual projection was 0.24±0.03 μm(mean±s.e.m.), the lengths of projections measured from tip to base varied from 0.27 μm to 1.21 μm. The shorter projections were in close contact with each other, whereas the elongated ones were separated by gaps. Clearly, the core of each projection consisted of a strap of bundled actin oriented perpendicularly to the cell-cell contact zones. Actin bundles traversed projections, extended through the cell body and embedded in the actin meshwork, which coincided well with positioning of the cortical cytoskeleton relatively to the apical membrane and cell nucleus. The average length of the individual actin strap was 1.94±0.15 μm(mean±s.e.m.). That agreed well with the estimated lengths of actin bundles visualized by immunofluorescence. Thus, the ultra-structural analysis confirmed that lateral projections induced by IGF-I in separating cells are actin-enriched microspikes.

To study the dynamics of α-actinin in living cells, we took advantage of confocal laser-scanning microscopy and examined the spatial and temporal effects of IGF-I on full-length α-actinin tagged with EGFP. The interior details from multiple focal planes were recorded as they changed over time (4D imaging). For thorough analysis of α-actinin localization, the representative sections are shown in Fig. 5.

Typically, without IGF-I treatment, α-actinin lined the borders of the adjacent MCF-7/IGF-IR/WT cells. The thin band of α-actinin appeared close to the lateral cell membranes in the most apical optical section and through the cell thickness of 10-12 μm. α-actinin was also localized within small aggregates scattered throughout the cytoplasm and in the fine dotted or streak-like focal structures about 2.5 μm above the substrate surface (Fig. 5Aa,d,g). Once cells were stimulated with IGF-I, the aggregates of α-actinin cleared off the cytoplasm and an increased amount of α-actinin became associated with cell boundaries where the nascent projections were formed. Within a few minutes, the propelling microvilli-like projections containing α-actinin extended and retracted laterally and upward; the most elongated projections(of 6-10 μm in length) bent across the cell apex(Fig. 5Ab,c). The short (of 2-4μm in length) α-actinin projections were also seen through a series of the middle optical sections spanning 10 μm in height(Fig. 5Ae,f). Within 5 minutes,some cells vanished from view in the most apical sections and then from the series of lateral sections (compare circled areas inFig. 5Aa with b,d and e). In the same cells, α-actinin motile ruffles formed along the boundaries of the basal membrane. These basal ruffles resembled growth-factor-induced ruffles described in KB cells (Kadowaki et al., 1986). In parallel, the initially minuteα-actinin-containing focal structures enlarged and oriented in the direction of cell spreading (Fig. 5Ag-i, circled). By 9 minutes of IGF-I stimulation, these focal structures appeared closer to the substratum than they were in the untreated cells (1.5 μm compared with 2.5 μm from the substrate level). Thus, we found that in separating living cells, α-actinin is localized in the dynamic apico-lateral microspikes besides focal adhesions and the edges of the basal ruffles.

In MCF-7/IGF-IR/DK cells, despite the continuous treatment with IGF-I,α-actinin remained predominantly localized to lateral membranes of cells tightly adhered to each other. Only minor time-dependent redistribution ofα-actinin occurred in the cytoplasm and at the cell borders(Fig. 5B); these changes were typical of all live MCF-7-derived cells regardless of IGF-I stimulation. Taken together these observations clearly show that activation of the IGF-IR causes a time-dependent accumulation of α-actinin-EGFP in the apico-lateral microspikes of separating cells.

In the α-actinin molecule, the N-terminal head contains highly conservative actin- and phospholipid-binding sites, the central rod domain is composed of four spectrin-like repeats implicated in dimerization ofα-actinin monomers and interaction with other proteins and the C-terminal end of the rod provides binding sites for ions of calcium as well as proteins (Hemmings et al.,1992; Matsudaira,1994; McGregor et al.,1994; Fukami et al.,1996). The characterization of α-actinin mutants has been particularly difficult in non-muscle cells owing to the rapid lethality of cells expressing the mutant proteins (Scultheiss et al., 1992;Hijikata et al., 1997). To define the contribution of individual domains to the function ofα-actinin in microspikes, we have constructed new pEGFP N1-vector-based plasmids encoding the mutant α-actinins as C-terminal chimeras with EGFP(Fig. 6A), and that allowed us to monitor fusion proteins soon after transfection. 230 amino-acid residues in the N-terminus were deleted to yield a mutant α-actinin lacking the entire actin-binding region (ΔN). To inactivate α-actinin in the C-terminus (ΔC), the last 168 amino acids were removed (described in the Materials and Methods).

MCF-7/IGF-IR/WT cells were transfected with plasmid DNA encoding either control EGFP or EGFP chimeras with a full-length wild-type (WT), mutantΔN and ΔC α-actinin. Cell extracts were resolved electrophoretically, and proteins transferred on the membrane were probed sequentially with antibody to GFP and α-actinin. Western immunoblotting has confirmed protein expression and the expected molecular weights of exogenous EGFP (27 kDa), WT (127 kDa), ΔN (104 kDa) and ΔC (109 kDa) EGFP-α-actinins (Fig. 6B).

To visualize the dynamics of α-actinin-EGFP fusion proteins specific to the regions of cell-cell contacts, we used confocal laser-scanning microscopy and collected stacks of images from the middle optical sections crossing nuclei. In serum-starved cells with mature cell-cell contacts, only the wild-type and ΔC α-actinins incorporated into pre-existing cell-cell junctions. Unlike the wild-type and ΔC α-actinin,ΔN α-actinin did not associate with the circumferential actin ring and was instead localized diffusely throughout the cell body(Fig. 6Ca-c).

In WT α-actinin transfectants, IGF-I triggered the expected reorganization of cell-cell contacts and the assembly of α-actinin-actin microspikes in the lateral zones (Fig. 6Cd). ΔN α-actinin did not colocalize with actin microspikes and remained diffused throughout the cell body(Fig. 6Ce). The number ofΔC α-actinin-transfected cells with α-actinin-actin spikes was half that of wild-type transfectants. Moreover, the spatial relationship between ΔC α-actinin and F-actin was disturbed, as microspikes did not organize into well ordered arrays (Fig. 6Cf). In two independent experiments, the appearance of microspikes was assessed quantitatively(Table 1). Collectively, these results demonstrate that in the α-actinin molecule both the head and tail regions are required for the function of microspikes. The N-terminal part of α-actinin is necessary for association of α-actinin with actin bundles, whereas the C-terminal tail of the rod is needed for correct assembly and/or positioning of microspikes at cell peripheries.

To address the role of IGF-IR kinase signaling in regulating microspike assembly in relation to cell separation, we used four specific pharmacological inhibitors of PI 3-kinase and MAP kinase signaling pathways(Fig. 7). They are LY 294002, a specific inhibitor of the catalytic subunit of the PI 3-kinase; a D3-modified phosphatidylinositol (PI) ether lipid analog (also called Akt inhibitor); PD 98059, a synthetic inhibitor of the MAP kinase-activating enzyme, MAPK/ERK kinase (MEK1); and UO 126, an inhibitor of MEK1 and MEK2. In MCF-7/IGF-IR/WT cells treated with LY 294002, the normal induction of phosphorylation of the PKB/Akt by putative PDK2 on serine 473 was downregulated by 90% and 84% and by 3-phosphoinositide-dependent kinase 1 (PDK1) on threonine 308 by 28% and 20%,and at 5 and 15 minutes, respectively. The 3D-modified PI-analog also downregulated signaling associated with the activities of PI 3-kinase-generated phospholipids, although its effect on phosphorylation of PKB/Akt was much less. The relative phosphorylation of PKB/Akt on serine 437 was reduced by 8% and 19%, and on threonine 308 by 9% and 11% at 5 and 15 minutes of IGF-I stimulation (average of five experiments)(Fig. 7a). PD 98059 reduced the relative phosphorylation of ERK1/2 MAP kinases on threonine 202 and tyrosine 204 by 73% and 67% at 5 and 15 minutes of IGF-I treatment (average of four experiments). UO 126 was a much more potent inhibitor of ERK1/2 MAPK kinases;it blocked completely IGF-I-stimulated and basal phosphorylation of these kinases (Fig. 7b and data not shown).

Pretreatment of MCF-7/IGF-IR/WT cells with 10 μM LY294002 and 40 μM PI analog before the addition of 50 ng/ml IGF-I blocked the formation of cortical microspikes and cell separation in response to IGF-I [compareFig. 8A,b with c,d also with Guvakova et al. (Guvakova et al.,2002)]. LY 294002- and PI-analog-pretreated cells remained organized in tight colonies during at least 60 and 30 minutes of the IGF-I treatment, respectively. By contrast, there was no apparent inhibitory effect of 50 μM PD 98059 on cell separation; cells, exposed to this drug or not,appeared spindle-shaped, separated from each other and migrated within 60 minutes of IGF-I stimulation. The formation of well orderedα-actinin-actin microspikes was slightly disturbed by PD 98059 (compareFig. 8A,b with e). The more potent inhibitor of MEK1 and MEK2, UO 126, did not prevent IGF-stimulated concentration of α-actinin and actin in the peripheral microspikes(Fig. 8Af); however, it inhibited cell migration radically within first 60 minutes of IGF-I stimulation (data not shown).

The above findings prompted our suggestion that MEK1/2 signaling is unnecessary for initial steps of cell-cell separation but may be essential for movement of the cells separated in response to PI 3-kinase signaling. To test this idea, we compared the effect of the PI analog and UO 126 on IGF-stimulated reorganization of cell-cell contacts, this time monitoring distribution of α-catenin, a molecular marker of adherens junctions, and fascin, a marker of the peripheral actin microspikes. In contrast toα-actinin, fascin localizes along the length of actin bundles induced by IGF-I and that allows visualizing microspikes without co-staining for F-actin(Guvakova et al., 2002). Pretreatment of cells with the PI analog blocked IGF-I-stimulated redistribution of α-catenin and thereby dissolution of adherens junctions. This inhibitor also blocked the formation of microspikes containing fascin (Fig. 8Ba,c; inset in c,fascin microspikes induced by IGF-I). In sharp contrast, total inhibition of MEK1/2 activity by UO 126 did not prevent a loss of α-catenin from adherens junctions and did not block the formation of fascin microspikes at cell peripheries (Fig. 8Bb,d). Strikingly, inhibition of MEK1/2 by UO 126 had a marked effect on stress fibers. Normally, activation of the IGF-IR in MCF-7 cells results in a rapid disassembly of stress fibers followed by re-assembly of microfilaments after 15 minutes of IGF-I stimulation. The latter coincides with advancing long protrusions and moving the cell body [seen inFig. 3Ai-l and described in detail in (Guvakova and Surmacz,1999)]. Pretreatment with UO 126 had no effect on disassembly of stress fibers; however, it prevented their re-assembly as well as the development of the long membrane protrusions after 15 minutes of IGF-I stimulation (compare Fig. 8Ca with b).

We conclude that separation of MCF-7 breast cancer cells in response to IGF-I is dependent on signaling from the activated IGF-IR through the PI 3-kinase and its phospholipid products, and that signaling through MEK1/2 may be necessary for subsequent cell locomotion.

The IGF system is essential for normal development of the mammary gland,and its disruption in human breast cancer leads to an excess of IGF-IR signaling. The biological relevance of the enhanced catalytic activity of the IGF-IR kinase to breast tumor progression has not been established. In this study, using a combination of molecular genetic, biochemical and pharmacological approaches with a complementary monitoring of living cell behavior, we have identified a novel mechanism by which the IGF-IR kinase induces separation and migration of human breast carcinoma cells.

It has previously been established that polypeptide growth factors acting through receptor tyrosine kinases can induce transition of cells from an epithelial to a motile fibroblastic phenotype, a phenomenon that resembles epithelial-mesenchymal transition, which occurs during development(Boyer and Thiery, 1993). IGF-I is mainly known as a mitogenic growth factor, and signaling through the IGF-IR in mammary epithelial cells has been chiefly studied in respect to effects on cell proliferation and apoptosis (reviewed byLee et al., 1998;Chernicky et al., 2000). Nevertheless, a role for the IGF-IR in motility of breast carcinoma cells had been proposed more than a decade ago on the basis of the results of the Boyden chamber assay (Kohn et al.,1990). It should be noted, however, that in this assay cells are artificially separated from each other before examination, and it is therefore not possible to analyze mechanisms of epithelial sheet disintegration.

In this study we examined in detail the effects of IGF-I on motile behavior of cells within an epithelial sheet. We have now demonstrated that, like other growth factors with scattering function, IGF-I induced cell motile responses in a biphasic manner, first causing separation and subsequently causing migration of the cells. The phase of active separation lasted only for 15-30 minutes and was unusually short compared with the 5-16 hours reported for HGF/SF, FGF-1 and EGF (Stoker and Perryman, 1985; Savagner et al., 1997; Müller et al.,1999). During the separation phase, IGF-I stimulated highly dynamic activity of the entire cellular membrane: formation of apical ruffles and motile apico-lateral microspikes, dissolution of adherence junctions and an increase in basal membrane ruffling. Although the precise mechanism of such rapid morphological alterations is unknown, the appearance of dynamic microspikes at the cell-cell contacts and temporal correlation of this event with disintegration the adherens junction complex led us to suggest that microspikes facilitate cell-cell detachment. In addition to acute cell separation, video tracking has revealed uncommonly fast rates of migration of single MCF-7/IGF-IR/WT cells. Collectively these data establish for the first time that activation of the IGF-IR can in the short-term cause destabilization of cell-cell interactions, leading to quick separation and migration of breast cancer epithelial cells.

EGF modulates the interaction between the adherens junction receptor E-cadherin and the actin cytoskeleton, although associated cytoskeletal reorganizations have not been characterized(Hazan and Norton, 1998). Our previous work has shown that in MCF-7 cells, IGF-I induces rapid disassembly of stress fibers and extensive reorganization of the cortical actin(Guvakova and Surmacz, 1999). We have now demonstrated that disassembly of the cortical actin belt coincides with mobilization of α-actinin and actin into apicolateral microspikes and displacement of α-catenin from the adherens junctions. This agreed with the recent report that activation of the IGF-IR induces redistribution of E-cadherin and β-catenin from the adherens junctions into the cytoplasm(Morali et al., 2001).

What drives α-actinin into cell-cell contacts? As an actinbinding protein, intact α-actinin shows a prominent colocalization with filamentous actin. Yet α-actinin can bind to a number of molecules, in addition to actin (Crawford et al.,1992; Otey et al.,1993; Kroemker et al.,1994; McGregor et al.,1994; Christerson et al.,1999) that perhaps determine the localization of α-actinin. For example, in Cos and PtK2 cells, mutated α-actinins with a deleted actin-binding site were not incorporated into stress fibers, although they localized to ruffled membranes and adherens junctions(Hemmings et al., 1992;Hijikata et al., 1997). The prominent localization of α-actinin to adherens junction has been related to the direct association between the central rod domain ofα-actinin and the C-terminus of α-catenin(Knudsen et al., 1995;Nieset et al., 1997). In epithelial culture, α-actinin colocalizes with vinculin and the cadherin-catenin complex. In this study, we used GFP-labeled α-actinin to follow distribution of α-actinin in living cells before and after IGF-I stimulation. Our results with a GFP-labeled mutant of α-actinin,in which the α-catenin-binding site was preserved while the actin-binding domain was removed, support the idea that both α-catenin and actin-binding sites of α-actinin are essential for localization of this protein into the mature cell-cell junctions. The interaction of the C-terminal end of α-actinin with vinculin does not seem to be required because a GFP-tagged ΔC α-actinin mutant in which the vinculin-binding site was disrupted did not interfere with the ability ofα-actinin to localize to intercellular junctions. The development of microspikes in response to IGF-I required at least two fully functional regions of the α-actinin molecule. As expected, the head ofα-actinin was obligatory for binding to actin spikes. The C-terminal tail of α-actinin was required for correct assembly and/or stabilization of microspikes because the truncation of this region reduced the number ofα-actinin/actin microspikes and caused their misalignment at the cell periphery. According to the previous suggestion, it is possible that the actin-binding activity of α-actinin requires a ternary interaction with the C-tail as modifications in the EF-hand domain of α-actinin have deleterious effects on the activity of its actin-binding domain(Witke et al., 1993;Dubreuil and Wang, 2000). Additionally, direct interaction of α-actinin with vinculin, which is probably disturbed in the ΔC α-actinin mutant, might be required for correct organization of microspikes. Thus our findings establish that functional α-actinin is necessary for the development of microspikes and is not passively reorganized following actin.

Our results, while supporting previous findings on IGF-I-stimulated actin polymerization at the cell periphery(Kadowaki et al., 1986;Isumi et al., 1988), further demonstrate that activation of the IGF-IR coordinates reorganization of actin into peripheral microspikes. Localization of α-actinin at the base rather than at the tips of microspikes implies that other actin-crosslinking protein(s) compose the core of a spike. A good candidate for this role is the 55 kDa actin-bundling protein fascin, which acts as a crosslinker, causing aggregation of actin into tight bundles(Edwards and Bryan, 1995;Kureishy et al., 2002). In vitro, fascin and α-actinin synergistically affect mechanical properties of actin filaments (Tseng et al.,2001). In vivo, fascin plays a role in extending the membrane during cell spreading and migration, and overexpression of fascin in pig epithelial cells caused the disorganization of cell-cell contacts(Yamashiro et al., 1998). Fascin also acts as a negative regulator of cell-cell interactions in rat mammary epithelial cells (Wong et al.,1999). Activation of the IGF-IR in MCF-7 cells induces rapid relocalization of fascin from the cytoplasm to the peripheral actin microspikes (Guvakova et al.,2002). We reason now that the activated IGF-IR induces cortical actin dynamics beneath the plasma membrane by increasing actin polymerization and promoting crosslinking of actin by fascin to stabilize the cores of microspikes and by α-actinin to properly position these nascent projections at the cell periphery. Although the significance of microspikes in cell-cell separation remains hypothetical, one possibility is that microspikes play an active cytomechanical role in the destabilization of interactions between opposing cells. Alternatively, microspikes may be the form of cell-cell contacts that are transient and more flexible than adherens junctions and therefore provide the intermediate strength of adhesion between separating cells.

Cell-cell detachment and microspike formation was completely blocked in MCF-7/IGF-IR/DK cells, clearly indicating the direct causative relationship between IGF-IR activation and α-actinin redistribution. Although it is conceivable that IGF-IR signaling affects the phosphorylation status ofα-actinin, at least two lines of evidence imply that tyrosine phosphorylation of α-actinin is not relevant to the development of microspikes in MCF-7 cells. First, there was no detectable difference in the tyrosine phosphorylation of α-actinin immunoprecipitated from cells stimulated with IGF-I (see Materials and Methods). Second, the WTα-actinin-EGFP chimera, which lacked the only identified phopshotyrosine residue 12 within the amino-acid motif QTNDY(Izaguirre et al., 2001), was nevertheless functionally active and localized to microspikes.

As relocalization of α-actinin into cortical structures was detected after 1-2 minutes of IGF-I stimulation, we predicted that the early signaling triggered by the ligand-receptor interaction promotes the development of microspikes. In MCF-7 cells, activation of the IGF-IR kinase quickly drives the formation of a complex between the receptor and substrates IRS and Shc that brings about activation of the downstream cascades of serine/threonine kinases and the PI 3-kinase. Interestingly, in different epithelial cells,different kinases play roles in cell motility. In MDCK canine epithelial cells, activation of both the PI 3-kinase and MAP kinase was found to be essential for the motile response to HGF/SF(Royal and Park, 1995;Potempa and Ridley, 1998). The function of MAP kinases was delineated in scattering triggered by HGF/SF in HT29 colon carcinoma cells and haptotaxis in FG carcinoma cells and T47D breast carcinoma cells (Klemke et al.,1997; Herrera,1998; Spencer et al.,2000), whereas pretreatment with the MEK1 inhibitor, PD 98059, was dispensable for the chemotactic response of MCF-7 cells to IGF-I(Manes et al., 1999). Others have reported that IGF-I stimulates colonic epithelial cell migration in a wound healing assay through multiple signaling pathways including PI 3-kinase,MAP kinases and PKC-γ and -δ(Andre et al., 1999). In MCF-7 cells expressing the dominant-negative receptors, activity of the PI-3 and the ERK1/2 MAP kinases is downregulated via markedly reduced signaling through IRS-1 and Shc pathways (in this paper and M.A.G., unpublished). To dissect signaling pathways essential for cell motility, we applied selective pharmacological inhibitors of PI 3-kinase and MEK. The MEK inhibitor also inhibits ERK1/2 MAP kinases (summarized inFig. 9a). Using immunofluorescence staining coupled with biochemical analysis, we found that the activity of PI 3-kinase rather than that of the MAP kinases was required for disassembly of adherens junctions and redistribution of actin,α-actinin and fascin into microspikes during cell separation. Subsequently, the activity of MEK1/2 appeared to be necessary for re-assembly of stress fibers, which may be involved in the extension of cell protrusions and the translocation of the cell body (model inFig. 9b). As activities of the ERK1/2 MAP kinases may enhance myosin light chain kinase (MLCK) activity,leading to the phosphorylation of myosin light chains (MLC)(Klemke et al., 1997), it seems possible that IGF-IR signaling sequentially coordinates cell-cell separation and phospho-MLC-dependent cytoskeletal contraction culminating in cell movement.

The function of the PI 3-kinase targets implicated in actin reorganization is dependent on PI 3-kinase-generated phospholipids(Nobes et al., 1995). Moreover, α-actinin may bind to the PI kinase and its product,phosphatidylinositol (3,4,5)-trisphosphate[PtdIns(3,4,5)P₃](Shibasaki et al., 1994;Greenwood et al., 2000). To investigate the role of PI-3-kinase-generated phosphatidylinositides in cell separation, we used a D3-modified PI analog, a competitive inhibitor of the PI 3-kinase with respect to PI, synthesized to selectively block the effect of myo-inositol-derived second messengers of the PI 3-kinase(Hu et al., 2000). This compound prevented a loss of α-catenin from adherens junctions as prevented redistribution of α-actinin and fascin into actin microspikes,suggesting a role for D3 phospholipids in IGF-I-stimulated cell separation. Although PKB/Akt possesses a pleckstrin-homology (PH) domain and may bind to PI-3-kinase-generated phospholipids, its phosphorylation in MCF-7 cells was only slightly reduced by the PI analog. Thus, the precise role of this kinase as well as other proteins bearing PH domains, including several adaptor proteins and GTP/GDP exchange and GTP-activating proteins for the ARF/Rho/Rac family of GTP-binding proteins, has yet to be investigated.

In summary, we have identified the first biochemically defined pathway in which the IGF-IR tyrosine kinase, through activities of the PI 3-kinase andα-actinin, promotes active separation in human breast cancer cells. We have also established the requirement for MEK1/2 activity in IGF-IR-mediated movement of these cells. Further development of α-actinin mutants and the measurements of the direct association between α-actinin and other molecules in living cells should make it possible to understand the dynamic regulation of microspikes and their role in cell separation. Further studies are needed to reveal the precise mechanisms by which MAP kinases control motility in the separated breast cancer cells.

We are very grateful to Derek LeRoith (NIH) for providing the mutant IGF-IR cDNA, Joseph Sanger (University of Pennsylvania) for the generous gift of EGFP-α-actinin encoding plasmid, and Lewis Tilney (University of Pennsylvania) for expertise for the actin study. We thank Mark Shipman for assistance and advice with use of Improvision software, image processing and movie production, Irina Chernysh for help with use of the LaserSharp 2000 software for 4D imaging in live cells and Patricia Connelly for excellent technical assistance with TEM. We would especially like to thank Francois Huber for discussing strategies on the DNA construct design and Laura Lynch for critical comments on the manuscript. This work was supported by the Breast Cancer Research Program of the Department of Defense DAMD 17-97-1-7211 award(to M.A.G.), National Institute of Health grant CA 16502 (to D.B.), and Wellcome Trust grant 038284 (to J.C.A.).