αPIX Associates with Calpain 4, the Small Subunit of Calpain, and Has a Dual Role in Integrin-mediated Cell Spreading*

Binding of integrins to the extracellular matrix results in actin cytoskeletal rearrangements, e.g. during cell spreading, by regulating the activity of Rho GTP-ases. We have shown previously that αPIX (Cool-2 or ARHGEF6), a Rac1/Cdc42-specific guanine nucleotide exchange factor (GEF), binds to β-parvin/affixin and colocalizes with integrin-linked kinase in actively spreading cells, suggesting that αPIX is involved in integrin-induced signaling leading to activation of Rac1/Cdc42. Here we report calpain 4, the small subunit of the proteases μ-calpain and m-calpain, as a novel binding partner of αPIX. This association was identified by the CytoTrap system and confirmed by coimmunoprecipitation and glutathione S-transferase pull-down assays. The αPIX triple domain SH3-DH-PH was found to be required for calpain 4 binding. During integrin-dependent spreading of CHO-K1 cells, αPIX colocalized with μ- and m-calpain, integrin-linked kinase, and β1 integrin in early integrin-containing clusters. Overexpression of αPIX wild type but not the GEF-deficient mutant (L386R/L387S) resulted in enhanced formation of characteristic cellular protrusions during cell spreading, suggesting that αPIX GEF activity is necessary for this specific actin cytoskeletal reorganization. The calpain inhibitors calpeptin and calpain inhibitor IV significantly inhibited integrin-dependent cell spreading. However, concomitant overexpression of αPIX wild type or the L386R/L387S mutant restored cell spreading. Together, these data suggest that αPIX is a component of early integrin clusters and plays a dual role in integrin-dependent cell spreading. Whereas αPIX GEF activity contributes to enhanced formation of cellular protrusions, the GEF-independent association with calpain 4 leads to induction of a yet unknown signaling cascade resulting in cell spreading.

nisms involve changes in organization of the actin cytoskeleton. Interaction of cells with the ECM is mediated by transmembrane receptors termed integrins. Binding of integrins to components of the ECM induces clustering of these receptors that subsequently results in the intracellular recruitment of structural and signaling molecules at the sites of matrix attachment, thereby providing links to the actin cytoskeleton (1,2).
Upon activation of integrins via ECM engagement, regulation of the actin cytoskeletal dynamics occurs primarily via the Rho family of small GTPases, i.e. Cdc42, Rac, and Rho (3). Rho GTPases function as binary switches that cycle between an active GTP-bound form and an inactive GDP-bound form (4,5). In particular, activation of Rho increases cell contractility and leads to the formation of focal adhesions and actin stress fibers (3,6). Cdc42 and Rac activation propagates the formation of filopodia, lamellipodia, and peripheral membrane ruffles as well as focal contacts/complexes (7,8). The regulated activation of Rho GTPases by growth factor receptors and G proteincoupled receptors has been studied extensively (9). Considerably less is known about how the ECM acts as an insoluble stimulus of the reorganization of the cytoskeleton. Recent studies provide evidence that members of the Rho family of GTPases are activated after signaling through integrins and are involved in integrin-induced cell spreading (10 -13). However, the upstream and downstream signaling pathways of Rho GTPases need to be defined in detail.
Activation of Rho GTPases depends on the regulated action of guanine nucleotide exchange factors (GEFs). Vav2, an exchange factor for Rac1, Cdc42, and RhoA (14), has been shown to be necessary for integrin-dependent activation of Rac leading to lamellipodia formation in fibroblasts (15), suggesting that GEFs are also involved in integrin-induced activation of Rho GTPases. ␣PIX/Cool-2/ARHGEF6, an exchange factor for Rac1 and Cdc42 (16,17), mediates PAK activation upon cell adhesion to fibronectin (18). Moreover, ␣PIX stimulates platelet-derived growth factor-induced peripheral spreading of Xenopus mesoderm aggregates on fibronectin, suggesting a role of ␣PIX in integrin-mediated cell adhesion and spreading (18). ␣PIX and its close homologue, ␤PIX, are part of a large protein complex including PAK, GIT1 (G protein-coupled receptor kinase interactor 1)/p95PKL, and associated proteins implicated in actin cytoskeletal regulation (19 -22). Moreover, GIT1/ p95PKL is critically involved in the regulation of actin cytoskeletal changes that accompany integrin engagement with the ECM as well as subsequent cell spreading and motility (23).
Other molecules that play a role in integrin-induced signaling are the intracellular, Ca 2ϩ -dependent calpain proteases with the two major forms -calpain and m-calpain, both consisting of a large catalytic and a small regulatory subunit (24,25). The 80-kDa catalytic subunits of theand m-forms are encoded by different genes, whereas the 28-kDa regulatory subunit, calpain 4, is common to both forms. Recent data sug-gest that calpain activity is regulated during signaling pathways leading to cell adhesion, spreading, and motility (26,27). Moreover, calpain is required for the formation of a new type of integrin complex that induces, by yet unknown means, the generation of focal complexes and adhesions during integrininduced cell spreading (28,29).
Recently, it has been shown that ␣PIX interacts with the focal adhesion protein ␤-parvin/affixin (30 -32). ␤-Parvin, integrin-linked kinase (ILK), and ␤3 integrin assemble in a protein complex that transduces signals from the ECM to intracellular effector proteins (33,34). The interaction of ␤-parvin with ␣PIX and colocalization of ␣PIX and ILK during cell spreading suggested an involvement of ␣PIX in integrin-induced signaling, which in turn results in cytoskeletal rearrangements via the GTPases Rac1 and Cdc42 (32,35). In the present study, we identified calpain 4, the regulatory subunit of bothand m-calpain, as an ␣PIX-interacting protein. We show that ␣PIX regulates integrin-mediated cell spreading in a GEF-dependent and -independent manner.

Yeast Two-hybrid Screening
We used the yeast two-hybrid system CytoTrap (Stratagene), also designated Sos Recruitment System (36), to identify ␣PIX-interacting proteins. Therefore, ␣PIX was fused to the N-terminal 1070 residues of human SOS and used as bait. We screened a human fetal brain plasmid cDNA library (Stratagene) with 5.3 ϫ 10 6 primary colonies and an average insert size of 1.3 kb according to the manufacturer's instructions. The prey cDNAs were fused to the myristoylation signal of v-Src that anchors the fusion proteins to the plasma membrane. All yeast transformations were done by the standard lithium acetate method. First, yeast strain cdc25H was cotransformed with pSos-␣PIX together with pYES-mGAP to reduce isolation of Ras GTPase false positive clones (37). For the CytoTrap screening, the pretransformed cdc25H yeast strain (see above) was transformed with 3 g of pMyr-cDNA library plasmids. Resulting transformants were grown for 5 days at 22°C on selective minimal glucose plates (complete supplement mixture-LEU-URA-TRP). After replica plating onto selective minimal galactose plates, a total of 54 colonies of ϳ2 million transformants showed galactose-dependent growth under restrictive conditions. Plasmid DNA was isolated, transformed into Escherichia coli DH10B, and selected on 50 ng/l chloramphenicol for the presence of the cDNA insert-containing pMyr plasmid. Protein interactions of putatively positive colonies were confirmed by retransformation of the cdc25H yeast strain with both the cDNA-containing pMyr plasmid and pSos-␣PIX or empty pSos (negative control). Only those clones growing on galactose media at the restrictive temperature of 37°C after 6 days were defined as "true" positives. Cotransformation of cdc25H cells with pSos-MafB and pMyrtarget cDNA, pSos and pMyr-target cDNA, pSos-MafB and pMyr-Lamin C, pSos and pMyr, pSos-␣PIX and pMyr-Lamin C, and pSos-␣PIX and pMyr served as negative controls, whereas cotransformation of cdc25H cells with pSos-MafB and pMyr-MafB served as a positive control.
In subsequent yeast two-hybrid experiments, different domains of ␣PIX were tested for interaction with calpain 4, using various bait constructs (see "Plasmid Constructs") in combination with plasmid pMyr-calpain 4 as prey. Expression of Sos-␣PIX fusion proteins was confirmed by immunoblotting of yeast lysates with mouse monoclonal anti-Sos1 antibody (1:400; BD Biosciences) and HRP-conjugated antimouse antibody (1:5000; Amersham Biosciences).
For data base searches, we used the NCBI BLAST Network Service. Deduced protein sequences were searched for functional domains using PROSITE and SMART. GenBank accession numbers are AF207831 and D25304 for the ␣PIX sequence and NM_001749 for calpain 4.

Plasmid Constructs
␣PIX Constructs Used as Bait in CytoTrap-Full-length and various ␣PIX constructs were established by using specific PCR primers and the KIAA0006 clone as template, and PCR products were ligated unidirectionally into pSos (Stratagene) via SalI and NotI restriction sites.
Generation of N-terminal HA-tagged ␣PIX Constructs-Various HAtagged ␣PIX constructs were generated by using specific PCR primers and the KIAA0006 clone as template, and PCR products were purified and cloned as NotI-EcoRI fragments in eukaryotic expression vector pMT2SM-HA. ␣PIX ⌬DH was generated by PCR-mediated cloning. Two overlapping PCR products lacking the DH domain were amplified and applied to megaprime PCR. The ␣PIX L386R/L387S construct was generated by PCR-mediated mutagenesis (38). Mutations of these two residues abolish GEF activity in ␣PIX (39) and ␤PIX (17). Two overlapping cDNA fragments were amplified with the desired mutations and applied to megaprime PCR.
Generation of N-terminal FLAG-tagged Calpain 4 Construct-The calpain 4 coding region was amplified from human testis cDNA (Invitrogen), and the purified PCR product was cloned into "donor vector" pENTR/D-TOPO (Invitrogen) according to the protocol provided. The DNA insert was sequenced for integrity and pENTR/D-TOPO-calpain 4 was used for cloning the calpain 4 coding region into pFLAG-CMV-4-cassetteA (32) via LR reaction according to the manufacturer's protocol.
Generation of N-terminal GST-tagged Calpain 4 Construct-pENTR/ D-TOPO-calpain 4 was used for cloning the coding region of calpain 4 into Gateway pDEST27 vector (Invitrogen) via LR reaction.
All described constructs were sequenced for integrity, and large and pure amounts of plasmid DNA were prepared by using a plasmid midi or maxi kit (Qiagen). Primer sequences are available on request.

GST Fusion Affinity Precipitations
CHO-K1 cells cotransfected with pDEST27-calpain 4 (5 g of DNA) and various pMT2SM-HA-␣PIX constructs (5 g of DNA) were lysed with ice-cold lysis buffer, and lysates were clarified by centrifugation for 10 min at 20,000 ϫ g at 4°C. The supernatants were incubated with 75 l of GST-Bind Resin (CN Biosciences Novagen) overnight at 4°C and subsequently washed three times with ice-cold Tris-buffered saline (50 mM Tris-HCl, pH 7.4, 150 mM NaCl). Precipitates were eluted from the beads and subjected to SDS-PAGE together with total lysates followed by immunoblotting. Proteins were detected with HRP-conjugated goat polyclonal anti-GST antibody (1:4000; Amersham Biosciences) and HRP-conjugated rat monoclonal anti-HA antibody clone 3F10 (1:4000; Roche Applied Science).
For affinity precipitations in the presence of calpain inhibitors, cell culture medium was supplemented with Me 2 SO (control), calpeptin (75 g/ml in Me 2 SO), or calpain inhibitor IV (100 g/ml in Me 2 SO). Cells were cotransfected with pDEST27-calpain 4 (5 g of DNA) and various pMT2SM-HA-␣PIX constructs (5 g of DNA) and incubated overnight in supplemented medium. Then cells were lysed with ice-cold lysis buffer containing Me 2 SO (control), calpeptin (75 g/ml in Me 2 SO), or calpain inhibitor IV (100 g/ml in Me 2 SO).

Cell Spreading
Coating of Coverslips-Sterile coverslips were acid-washed, treated with 2% 3-aminopropyltriethoxy-silane (Sigma), and overlaid with 0.25% glutaraldehyde. Subsequently, coverslips were coated with 10 g/ml fibronectin (Sigma) in PBS, blocked with 2% bovine serum albumin (Sigma), and finally washed with PBS before cell plating. This coating procedure prevented unspecific cell spreading on the coverslips and enabled the investigation of fibronectin-induced (and therefore integrin-mediated) cell spreading.
Cell Plating-Serum-starved CHO-K1 cells were washed with PBS and detached with 0.5% trypsin-EDTA (Invitrogen). Trypsin reaction was stopped using 0.25 mg/ml soybean trypsin inhibitor. To eliminate growth factors, cells were washed once with PBS, once with Puck's saline, and twice with 1% bovine serum albumin in nutrient mixture F-12 (HAM). Finally, cells were resuspended in 0.2% bovine serum albumin in nutrient mixture F-12 (HAM), and 2-3 ϫ 10 4 cells were plated onto fibronectin-coated coverslips. To ensure that cell spreading was indeed fibronectin-induced, we seeded serum-starved CHO-K1 cells onto coverslips coated with fibronectin versus coverslips without fibronectin. Quantification of cell spreading showed that in the case of fibronectin-coated coverslips, 83% of cells could spread, whereas only 22% of cells plated in the absence of fibronectin were able to spread (data not shown). For ectopic expression of various HA-tagged ␣PIX proteins, CHO-K1 cells were seeded at 1.2 ϫ 10 6 cells/100-mm cell culture dish in complete culture medium. After 15 h, serum was reduced to 0.5%, and 24 h later, cells were transfected with pMT2SM-HA-␣PIX wild type, pMT2SM-HA-␣PIX (L386R/L387S), or pMT2SM-HA-␣PIX ⌬DH (6 g of DNA each) in the absence of serum according to the manufacturer's protocol. Cells were incubated overnight at 37°C in 5% CO 2 and prepared for cell spreading experiments on the next day.
Inhibition of Cell Spreading Using Calpain Inhibitors-Before plating onto fibronectin-coated coverslips, detached cells were incubated in nutrient mixture F-12 (HAM) supplemented with calpain inhibitor for 30 min. The following inhibitors were tested: the reversible, cell membrane-permeable inhibitor calpeptin (75 g/ml; Calbiochem), which reacts with the active site of both -calpain and m-calpain; the irreversible, cell membrane-permeable calpain inhibitor IV (100 g/ml; Calbiochem), which nucleophilically attacks the active site of m-calpain; and the reversible, cell membrane-permeable inhibitor PD 151746 (100 g/ml; Calbiochem), which binds predominantly to the Ca 2ϩ -binding site of -calpain.

Immunofluorescence
Transfected and untransfected cells, respectively, were plated onto fibronectin-coated coverslips, incubated for 30 min at 37°C, rinsed with PBS, and fixed in 4% paraformaldehyde containing PBS for 10 min. After residual formaldehyde had been quenched with PBS for 10 min, cells were incubated with 2% bovine serum albumin, 3% goat serum, and 0.5% Nonidet P-40 for 60 min to permeabilize the membrane and block nonspecific antibody binding. For detection of cells expressing HA-tagged ␣PIX proteins, cells were incubated either with rabbit polyclonal antibody anti-HA (2.5 g/ml; Sigma) or with mouse monoclonal antibody HA.11 (1.0 g/ml; Eurogentec) followed by incubation with Alexa Fluor 488 goat anti-rabbit IgG (4 g/ml; Invitrogen), Alexa Fluor 546 goat anti-rabbit IgG (4 g/ml; Invitrogen), or Alexa Fluor 546 goat anti-mouse IgG (4 g/ml; Invitrogen). Calpain 4 was detected by mouse anti-calpain small subunit monoclonal antibody (1:100; Chemicon), calpain 1 was detected by mouse anti-calpain I large subunit monoclonal antibody (1:100; Chemicon), and calpain 2 was detected by rabbit anticalpain II large subunit polyclonal antibody (10 g/ml; Chemicon). For ILK detection, cells were stained with mouse anti-ILK vinculin monoclonal antibody (clone 65.1.9; 10 g/ml; Upstate Biotechnology), and for the detection of ␤1 integrin, we used mouse anti-integrin ␤1 monoclonal antibody (clone DE9; 10 g/ml; Upstate Biotechnology). Paxillin was detected using mouse anti-paxillin monoclonal antibody (10 g/ml; BD Biosciences), and for vinculin staining, cells were incubated with mouse anti-human vinculin hVIN-1 monoclonal antibody (1:200; Sigma). As secondary antibody, Alexa Fluor 488 goat anti-mouse IgG (4 g/ml; Invitrogen) or Alexa Fluor 488 goat anti-rabbit IgG (4 g/ml; Invitrogen) was used. After washing twice with high salt PBS (650 mM NaCl) and three times with PBS, cells on coverslips were mounted in glycerol gelatin (Sigma) on microscopic slides. Cells were examined using a Leica DMRA immunofluorescence microscope, and images were acquired on a Leica TCS-NT confocal microscope equipped with an Apo 40-by-1.0 oil immersion objective lens.

RESULTS
Identification of Calpain 4, the Regulatory Subunit of Calpain, as a Novel ␣PIX Binding Partner-To identify novel protein interaction partners, the coding region of ␣PIX was used in a CytoTrap screen of a human fetal brain cDNA library. Of a total of 54 clones that restored growth of cdc25H cells in a galactose-dependent manner, only those clones that also showed ␣PIX-dependent growth were defined true positives (Fig. 1A). Two independent clones encode the C-terminal region of calpain 4, the small regulatory subunit of the protease calpain, and contain ϳ400 bp of the 3Ј-untranslated region (Fig.  1B). Calpain 4 consists of two domains, the N-terminal glycin-rich domain V, which is often referred to as a hydrophobic domain, and C-terminal domain VI, which contains five EFhand Ca 2ϩ -binding sequences (24). The two cDNAs obtained encode only four EF-hand calcium-binding motifs (Fig. 1B).
To verify the interaction between ␣PIX and calpain 4 in mammalian cells, we performed coimmunoprecipitation experiments in CHO-K1 cells. ␣PIX was detected by polyclonal anti-PIX antibodies (17) that recognize, in addition to ␣PIX, two isoforms of ␤PIX (␤1PIX and ␤2PIX) (40), a close ␣PIX homologue. Endogenous ␣PIX was coimmunoprecipitated with FLAG-calpain 4 (Fig. 1C, top panel, lane 2), but not with empty FLAG-vector control (Fig. 1C, top panel, lane 1). In addition, an in vivo GST pull-down assay was performed using CHO-K1 cells cotransfected with both GST-calpain 4 and wild-type HAtagged ␣PIX. As shown in Fig. 2 (middle panel, lanes 1 and 2), GST-calpain 4 fusion protein bound full-length HA-␣PIX, whereas GST alone failed to bind ␣PIX. Together, these data suggest that calpain 4 is a binding partner of ␣PIX in vivo.
The  12, 14, 16, and 18). Remarkably, an ␣PIX protein lacking only the DH domain (␣PIX ⌬DH) showed no affinity to calpain 4 ( Fig. 2, middle panel, lane 20). These data suggest that the integrity of the three ␣PIX domains SH3-DH-PH is necessary for efficient binding of the small subunit of calpain (calpain 4). These data are in line with our findings obtained by the CytoTrap system using various ␣PIX bait constructs in combination with pMyr-calpain 4 as prey. Again, the three domains SH3-DH-PH of ␣PIX are required for binding to calpain 4 (data not shown).

␣PIX Localizes in Small Dotted Structures at the Cell Periphery and Enhances the Formation of Cellular Protrusions upon
Integrin-induced Cell Spreading-By immunofluorescence analysis, we examined the effect of wild-type and various mutant HA-tagged ␣PIX proteins upon integrin-induced cell spreading after 30 min. Overexpression of ␣PIX wild-type protein caused extensive morphological changes including the formation of various cellular protrusions that represent lamellaesheets (Fig. 3, A1 and A4) with small finger-like protrusions at their end (Fig. 3, A3 and A6). ␣PIX was located in a punctual pattern at the cell periphery. These aggregates most likely correspond to the early integrin clusters described previously (29,41). In contrast, cells expressing a GEF-deficient ␣PIX mutant (L386R/L387S) or ␣PIX ⌬DH did not show this characteristic and enhanced formation of protrusions (Fig. 3, B1 and C1). ␣PIX mutant L386R/L387S was distributed in the cytoplasm as well as at the cell periphery, whereas ␣PIX ⌬DH localized diffusely in the cytoplasm and was not enriched at the plasma membrane. These data suggest that the GEF activity of ␣PIX is necessary for the formation of distinct cellular protrusions during integrin-dependent cell spreading. Moreover, the DH domain seems to be required for proper targeting of ␣PIX to the cell periphery.
␣PIX Colocalizes with the Small and Large Subunits of mand -calpain, ILK, and ␤1 Integrin in Early Integrin Clusters during Cell Spreading-To determine the type of ␣PIX-containing protein aggregates during integrin-dependent cell spreading, we analyzed the colocalization status of ectopically expressed HA-␣PIX with endogenous calpain subunits in serum-starved CHO-K1 cells (Fig. 4, AϪC). 30 min after seeding on fibronectin, ␣PIX wild type-expressing cells formed enhanced lamellipodia-like sheets during cell spreading, and ␣PIX localized in punctual clusters at the cell periphery (Fig. 4,  A1, B1, and C1). Staining of endogenous calpain 4 and the large subunits of -calpain (calpain 1) and m-calpain (calpain 2) revealed that calpain 4 and calpain 1 were localized around the nucleus (Fig. 4, A2 and B2), whereas calpain 2 was distributed in the cytoplasm in small dots (Fig. 4, C2). Nonetheless, all three proteins colocalized with ␣PIX in small dots in lamellae sheets or in small finger-like protrusions close to the membrane (Fig. 4, A4, B4, and C4). The presence of bothand m-calpain (large and small subunits) in a punctual pattern at the cell periphery suggests that these protein complexes represent initial integrin clusters that form immediately upon integrin-induced signaling (29). ␣PIX is a novel component of these early clusters and is possibly involved in integrin-mediated activation of Rac1 and/or Cdc42.
To elucidate the composition of these initial integrin clusters, we stained ␣PIX-overexpressing CHO-K1 cells with various antibodies during integrin-dependent cell spreading. We ob-served colocalization of ␣PIX wild type with ILK and ␤1 integrin in dotted structures at the cell surface (Fig. 4, D4 and E4). In contrast, no colocalization was found for ␣PIX and paxillin or vinculin in actively spreading cells (Fig. 4, F4 and G4).
Both Wild-type ␣PIX and the GEF-deficient Mutant (L386R/ L387S) Restore Cell Spreading and Associate with Calpain 4 in the Presence of Calpain Inhibitors-To examine whether cell spreading of CHO-K1 cells can be suppressed by inhibition of calpain, we serum-starved CHO-K1 cells and pretreated the cells with various calpain inhibitors and Me 2 SO, respectively. 77% and 82% of cells incubated with calpeptin or calpain inhibitor IV were unable to form early lamellae protrusions and appeared round (Fig. 5, A and C, untransfected controls, middle and bottom panels). In contrast, 75% of control cells were still able to spread and formed lamellipodia after 30 min ( Fig. 5A  and C, untransfected controls, top panel). Cells transiently overexpressing ␣PIX wild type or the L386R/L387S mutant extended membrane protrusions similar to lamellipodia in the FIG. 2. The ␣PIX SH3-DH-PH triple domain binds calpain 4. The domain structure of various ␣PIX proteins is shown. CHO-K1 cells were cotransfected with the indicated HA-tagged ␣PIX constructs and GST-calpain 4. As a control, empty GST-vector was used. The GST-tagged protein complexes were isolated using glutathione-Sepharose beads (GST pull-down) and subjected to immunoblot analysis with anti-HA antibody. GST-calpain 4 was able to bring down all HA-tagged ␣PIX proteins containing the triple domain SH3-DH-PH (middle panel, lanes 2, 4, 6, 8, and 10). In contrast, no HA-␣PIX was precipitated from cells transfected with ␣PIX constructs lacking at least one of these domains (middle panel, lanes  12, 14, 16, 18, and 20). None of the expressed ␣PIX proteins was coprecipitated with GST alone (middle panel, lanes 1, 3, 5, 7, 9, 11, 13, 15, 17, and 19). Expression of HA-tagged ␣PIX proteins in total lysates was confirmed by immunoblot (top panel, lanes 1-20). In precipitates, expression of GST-calpain 4 or GST was detected by reprobing membranes with anti-GST antibody (bottom panel, lanes 1-20).
presence of calpain inhibitor IV or calpeptin, 30 min after seeding on fibronectin (Fig. 5, B1-B4; data not shown). In contrast, cells expressing ␣PIX ⌬DH appeared round and did not form membrane extensions in the presence of calpain inhibitor IV (Fig. 5, B5 and B6) or calpeptin (data not shown). Quantitative analysis showed that 52-62% of cells expressing ␣PIX wild type or L386R/L387S, respectively, could spread in the presence of calpain inhibitors (Fig. 5C, middle and bottom  panels). Conversely, only 24% and 15% of cells expressing the ␣PIX ⌬DH mutant were able to spread under the respective condition that is comparable to untransfected control cells (Fig.  5C, middle and bottom panels). When cells were treated only with Me 2 SO, 76% of ␣PIX wild type-overexpressing cells, 73% of ␣PIX L386R/L387S-overexpressing cells, and 69% of ␣PIX ⌬DH-overexpressing cells exhibited a spreading phenotype (Fig. 5C, top panel). The calpain inhibitor calpeptin was known to inhibit bothand m-calpain, whereas the specificity of calpain inhibitor IV is restricted to m-calpain. The -calpainspecific inhibitor PD 151746 was found not to significantly inhibit integrin-mediated cell spreading because only 31% of the treated cells were unable to spread, which is comparable with 25% of Me 2 SO-treated control cells (data not shown).
In GST pull-down experiments, GST-calpain 4 was able to bring down HA-␣PIX wild type and HA-␣PIX L386R/L387S in both the absence and presence of calpain inhibitors (Fig. 5D,  middle panel, lanes 1, 2, 4, 5, 7, and 8), whereas ␣PIX ⌬DH failed to interact with GST-calpain 4 under either of the conditions analyzed (Fig. 5D, middle panel, lanes 3, 6, and 9). This finding is in agreement with the results obtained by GST pull-down experiments shown in Fig. 2 and indicates that calpain inhibitors do not impair the association between ␣PIX and calpain 4, an interaction that is likely crucial for spreading of CHO-K1 cells in the presence of calpain inhibitors. DISCUSSION In this study we describe a novel interaction between the Rac1/Cdc42-specific guanine nucleotide exchange factor ␣PIX and calpain 4, the small subunit of /m-calpain. Calpains are Ca 2ϩ -dependent cysteine proteases with a large number of substrates including several cytoskeletal proteins and signaling molecules (24).
We showed that ␣PIX colocalizes with calpain 4, calpain 1 and 2, ILK, and ␤1 integrin in small protein aggregates at the cell periphery during cell spreading, whereas no colocalization was observed with paxillin and vinculin. Our data suggest that these protein aggregates correspond to the early integrin clusters described by Bialkowska et al. (29), and ␣PIX provides a direct link between integrin-induced signaling and cytoskeletal reorganization during cell spreading in CHO-K1 cells. However, it has been reported that vinculin is a protein component of the integrin clusters in bovine aortic endothelial cells (29). Possibly, vinculin is recruited to these clusters at a later stage of their development.

FIG. 5.
Inhibition of m-calpain results in significant repression of integrin-mediated cell spreading that can be restored by overexpression of ␣PIX wild type and L386R/L387S, but not ␣PIX ⌬DH. A, CHO-K1 cells were serum-starved for 24 h, trypsinized, and pretreated with culture medium containing either Me 2 SO, calpeptin, or calpain-inhibitor IV for 20 min. Subsequently, cells were replated onto fibronectin-coated coverslips and incubated for 30 min. Control cells were still able to spread, whereas cell spreading was significantly inhibited in cells treated with calpeptin or calpain inhibitor IV. Pictures were taken from representative areas out of three independent experiments. B, immunofluorescence analysis of CHO-K1 cells expressing various ␣PIX proteins in the presence of calpain inhibitor IV. Serum-starved CHO-K1 cells expressing HA-tagged ␣PIX wild type (B1), ␣PIX L386R/L387S (B3), or ␣PIX ⌬DH (B5) were treated as described in A using calpain inhibitor IV. After a 30-min incubation, cells were fixed and permeabilized, and HA-␣PIX was stained with anti-HA antibody. Cells expressing ␣PIX wild type or L386R/L387S extended lamellae protrusions in the presence of calpain inhibitor IV (arrowheads in B2 and B4), whereas those expressing ␣PIX ⌬DH did not and appeared round (B6). Pictures were taken from representative areas out of three separate experiments. C, quantitative analysis of cell spreading in untransfected, ␣PIX wild type-, ␣PIX L386R/L387S-, or ␣PIX ⌬DH-expressing CHO-K1 cells replated onto fibronectin in the presence of Me 2 SO, the calpain inhibitor calpeptin, or calpain inhibitor IV. The results shown are the means Ϯ S.D. of three independent experiments. D, both ␣PIX wild type and ␣PIX L386R/L387S are able to bind calpain 4 in the presence of calpeptin or calpain inhibitor IV. The indicated HA-tagged ␣PIX constructs were cotransfected with GST-calpain 4 into CHO-K1 cells. Subsequently, cells were incubated in medium containing Me 2 SO, calpeptin, or calpain inhibitor IV for 14 h and lysed with buffer containing Me 2 SO, calpeptin, or calpain inhibitor IV. The GST-tagged protein complexes were isolated with glutathione-Sepharose beads and subjected to immunoblot analysis with anti-HA antibody. ␣PIX of the actin cytoskeleton during the formation of cellular processes. Active calpain is necessary for lamellipodia formation, activation of Rho GTPases, and actin filament organization (25,28). Calpain-cleaved ␤3 integrin was found in transient integrin clusters that form at early stages after integrin-induced adhesion of bovine aortic endothelial cells (29). Moreover, calpain cleaves RhoA during integrin-induced spreading, thereby generating a dominant negative form that significantly decreases cell spreading (42). Additional proteins have been detected in the early integrin clusters such as ␣-actinin and skelemin (41). Nonetheless, it is not known whether these molecules are cleaved by active calpain. Similarly, we cannot yet exclude the possibility that ␣PIX is cleaved by m-calpain in the early integrin clusters in CHO-K1 cells.
Inhibition of m-calpain in CHO-K1 cells led to significant impairment of cell spreading that could be restored by overexpression of ␣PIX wild type or the GEF activity-deficient L386R/ L387S mutant, but not by the ␣PIX ⌬DH mutant, suggesting that the interaction between ␣PIX and calpain 4/m-calpain is necessary for cell spreading rather than the GEF activity of ␣PIX. Indeed, association of calpain 4 and ␣PIX wild type or L386R/L387S mutant was found in the presence of calpain inhibitors. Together, these data suggest that ␣PIX acts downstream of calpain. Although the GEF activity of ␣PIX is not primarily necessary for restoring cell spreading in the presence of calpain inhibitors, the integrity of the ␣PIX DH domain seems to be required for this process. Besides cleaving integrin/ cytoskeletal proteins, m-and -calpain are also involved in signal transduction pathways, e.g. they play a critical role during integrin-induced actin remodeling and cell spreading (27,28). Inhibition of calpain with membrane-permeable inhibitors or by expression of a catalytically inactive form resulted in an inability of cells to spread. Remarkably, overexpression of constitutively active forms of Rac and RhoA reversed this effect, suggesting that calpain acts upstream of both Rac and RhoA in integrin-induced cell spreading in bovine aortic endothelial cells (28). It has been shown that -calpain but not m-calpain is involved in these morphological changes (28,29). In contrast, m-calpain but not -calpain was detected at sites of focal contact formation during T-cell adhesion and fibronectindependent spreading (43). We found m-calpain to be implicated in integrin-mediated cell spreading in CHO-K1 cells because inhibition of CHO-K1 cells with the m-calpain-specific calpain inhibitor IV abrogated integrin-dependent cell spreading. Conversely, inhibition of -calpain did not significantly disturb spreading. Because none of the synthetic calpain inhibitors presently available is completely specific for one calpain form, we cannot exclude an implication of -calpain in this process, for example at a different time point. In NIH 3T3 cells, stable overexpression of the natural calpain inhibitor calpastatin resulted in a decreased level of m-calpain mRNA and impaired the ability of cells to extend actin-rich processes such as lamellipodia and filopodia and to spread (27). In a recent study that aimed to address the isoform-specific functions of calpain 1 and calpain 2 in regulating membrane protrusion, it has been shown that calpain 2 (m-calpain) is necessary for the protrusion and lamellipodia formation at the leading edge (44). Together, these data implicate m-calpain in integrin-induced signaling events. It seems that both calpain isoforms are probably involved in integrin-mediated cell spreading and that the cell type and/or the stage of cell spreading determines which calpain form is required.
The association of ␣PIX with calpain 4 occurs via the SH3-DH-PH triple domain. In the majority of Rho GEF proteins, the DH-PH module is responsible for the exchange activity in vivo (45), whereas the SH3 domain is involved in interaction with FIG. 6. Proposed model for the molecular signaling cascade during integrin-mediated reorganization of the actin cytoskeleton and cell spreading in CHO-K1 cells. Upon attachment of cells to the extracellular matrix, early integrin clusters are formed that contain ILK, calpain protease, ␤-parvin, ␣-actinin, and ␣PIX. The proteases m-and -calpain are necessary for the formation of these clusters and act upstream of ␣PIX, which promotes the formation of characteristic cellular protrusions during cell spreading. Activation of Rho GTPases is mediated by the guanine nucleotide exchange activity of ␣PIX and leads to reorganization of the actin cytoskeleton. In addition, a second ␣PIXmediated signaling cascade exists, resulting in cell spreading that is independent of the ␣PIX exchange activity. Although association between ␣PIX and calpain 4 seems to be required for the GEF-independent pathway, additional components implicated in this signaling cascade are yet unknown.
wild type and ␣PIX L386R/L387S coprecipitated with GST-calpain 4 (middle panel, lanes 1 and 2) , lanes 6 and 9). Expression of various HA-tagged ␣PIX proteins in total lysates was confirmed by immunoblot (top panel, lanes 1-9). In precipitates, abundance of GST-calpain 4 was shown by reprobing the membrane with anti-GST antibody (bottom panel, lanes 1-9). other proteins, in case of ␣PIX with PAK (16,17). The activity of GEF proteins is regulated by direct protein-protein interaction of the DH and/or PH domains with other molecules (45). For example, the catalytic DH-PH module of ephexin was shown to associate with the EphA receptor, suggesting that EphA modulation of ephexin activity might occur through hindrance of the GEF activity (46). A similar mechanism for modulation of ␣PIX GEF activity by calpain 4 might be possible.
Overexpression of wild-type ␣PIX dramatically enhanced the formation of cellular protrusions during integrin-dependent cell spreading, whereas no morphological changes were observed when the ␣PIX GEF-deficient L386R/L387S mutant or the ⌬DH mutant was expressed. The protrusions formed upon overexpression of wild-type ␣PIX resemble lamellae sheets with long and thin filopodia-like structures at their end. ␣PIX is located in a punctual pattern at the end of these hand-like structures, which most likely represent early integrin clusters. Together, these data suggest that Rac and Cdc42 activity might be elevated in these cells. In the majority of reports, cell morphology changes rather than direct determination of Rac activity have been documented during integrin-mediated cell spreading at very early time points (11,12,28,29,41). Vav1 overexpression in Jurkat cells that were plated on fibronectin for 10 min did not induce an increase in Rac1 and Cdc42 activities, although the cells showed enhanced lamellipodia formation. Based on these findings, it has been suggested that integrin engagement induces activation of a very restricted Rac1/Cdc42 pool within the cell, resulting in a very modest increase in total Rac1/Cdc42 activity (47). Similarly, we were not able to detect increased Rac1 or Cdc42 activity by PBD (p21-binding domain of PAK) pull-down (data not shown). In line with this finding, the GEF activity of ␣PIX has been described to be very low (17,40), and additional cofactors, e.g. PAK1, Cdc42, Rac1, platelet-derived growth factor receptor, G␤␥, and RP1 (calponin homology 1 domain of affixin), are required for the detection of ␣PIX exchange activity in vivo (18,35,39). However, ␣PIX exchange activity for Rac1 and Cdc42 has already been shown in vitro (17,48) and seems to be differentially regulated depending on the monomer-dimer equilibrium of ␣PIX (49). Together, these observations suggest that although ␣PIX expression alone is not sufficient for detecting GEF activity in vivo, it is apparently adequate to cause morphological changes during cell spreading and in fully spread cells (32,40,47). Furthermore, our data suggest that the GEF activity of ␣PIX is required for enhanced formation of cellular protrusions, whereas a GEF activity-independent pathway may also exist. Exchange factor-independent functions of ␣PIX are known, e.g. PAK1 activity was enhanced solely by association with ␣PIX (39). Moreover, an exchange factor-independent role of the Rho, Rac, and Cdc42-specific GEF Vav1 (50) in integrin-mediated T-cell spreading has also been identified (47). Thus, we propose that ␣PIX may exert a dual effect: one depends on its GEF activity and leads to enhanced formation of characteristic cellular protrusions, and the other one regulates cell spreading via yet unidentified signaling components.
Recently, we identified an association of ␣PIX with ␤-parvin (affixin) (32), a novel focal adhesion protein (30,31). ␤-Parvin binds to ILK (31), a serine/threonine protein kinase, which is able to associate with the cytoplasmic domains of ␤1 and ␤3 integrin (51) and binds to ␣-actinin in an ILK kinase activity-dependent manner (52). Both expression of the ␣-actinin binding domain of ␤-parvin in CHO-K1 cells and ␤-parvin knockdown by short interfering RNAs resulted in blockage of cell spreading and lamellipodia formation (52). These data are in contrast to those reported very recently by Zhang et al. (53) showing that depletion of ␤-parvin in HeLa cells replated on fibronectin for 25 min had no effect on cell spreading. The direct association of ␣PIX with ␤-parvin and calpain 4 as well as colocalization of ␣PIX with ␤1 integrin and ILK during cell spreading provides evidence that integrin-induced signaling leads to the formation of large protein complexes, the early integrin cluster (29). Thus, we suggest that upon attachment of cells to the extracellular matrix, early integrin clusters are formed containing ␤1 integrin, ILK, calpain proteases, ␤-parvin, ␣-actinin, and ␣PIX. These clusters promote cell spreading via at least two distinct mechanisms. Activation of Rho GTPases is mediated by the GEF exchange activity of ␣PIX and leads to reorganization of the actin cytoskeleton. However, an additional, ␣PIX GEFindependent signaling cascade exists that most likely depends on ␣PIX-calpain 4 association and results in cell spreading. In this pathway, calpain acts upstream of ␣PIX (Fig. 6).
Because mutations of ␣PIX (ARHGEF6) are implicated in X-linked nonspecific mental retardation in humans (54), the question arises whether integrin-dependent signaling and cell spreading might also be involved in the regulation of neurite outgrowth and the morphology of dendrites and/or dendritic spines. In this context, it is of interest to note that ILK was found to be important for integrin-dependent neurite outgrowth in N1E-115 cells (55) as well as nerve growth factormediated neurite outgrowth in rat adrenal pheochromocytoma PC12 cells (56). However, additional studies are required to elucidate the biological function of ␣PIX in integrin-mediated signaling and cell spreading during neuronal processes.