IQGAP1 Integrates Ca2+/Calmodulin and Cdc42 Signaling*

Calmodulin regulates diverse Ca2+-dependent cellular processes, including cell cycle progression and cytoskeletal rearrangement. A recently identified calmodulin-binding protein, IQGAP1, interacts with both actin and Cdc42. In this study, evidence is presented that, in the absence of Ca2+, IQGAP1 bound to Cdc42, which maintained Cdc42 in the active GTP-bound state. Addition of Ca2+ both directly abrogated the effect of IQGAP1 on the intrinsic GTPase activity of Cdc42 and, in the presence of calmodulin, dissociated Cdc42 from IQGAP1. In addition, in vitro binding assays revealed that calmodulin associated with both the calponin homology domain and the IQ motifs of IQGAP1. Moreover, F-actin competed with Ca2+/calmodulin for binding to the calponin homology domain, but not the IQ motifs, of IQGAP1. Analysis of cell lysates revealed that calmodulin bound to IQGAP1 in a ternary complex with Cdc42. Increasing the Ca2+ concentration enhanced the interaction between calmodulin and IQGAP1, with a concomitant decrease in the association of IQGAP1 with Cdc42. Our data suggest that IQGAP1 functions as a scaffolding protein, providing a molecular link between Ca2+/calmodulin and Cdc42 signaling.

The Ras superfamily comprises a group of small GTPases that function in intracellular signaling cascades, affecting cell growth and differentiation as well as cytoskeletal organization, vesicle trafficking, and nuclear transport (1). These proteins act as molecular switches, alternating between an active GTP bound form and an inactive GDP bound form. Regulators include GTPase-activating proteins (GAPs) 1 that facilitate conversion from active to inactive states, guanine nucleotide exchange factors that catalyze release of GDP from the GTPase, and GDP-dissociation inhibitory factors that inhibit both GAPmediated and intrinsic GTP hydrolysis. Deregulation of both GTPases and their regulators may be involved in cell cycle derangement and oncogenesis (2).
The Rho family GTPase Cdc42 is involved in cytoskeletal rearrangement and cell cycle progression (3,4). Specifically, active Cdc42 stimulates filopodium and microspike formation in fibroblasts (5,6) and polarization of actin and microtubules in T cells (7). In addition, microinjection of Cdc42 into Swiss 3T3 fibroblasts stimulates DNA synthesis (4). It is therefore not surprising that Cdc42 has been implicated in carcinogenesis. Transforming potential has been confirmed by overexpression studies in which dominant negative mutants of Cdc42 inhibit Ras-mediated transformation, whereas constitutively active mutants promote anchorage-independent growth in Rat1 fibroblasts (8).
Downstream effectors of Cdc42 are being elucidated and may include phosphatidylinositol 3-kinase (9) and pp70 S6K (10), which are involved in cell cycle progression, and the Wiskott-Aldrich syndrome protein (11) and n-chimaerin (12), which may regulate actin polymerization. Cdc42 also stimulates p21activated Ser/Thr kinases, which in turn regulate the activation of the nuclear mitogenic protein kinases, c-Jun kinase and p38 (13,14). Interestingly, another potential downstream effector of Cdc42, IQGAP1, displays significant sequence similarity to Sar1 and the tumor suppressor neurofibromin (15), and is the major calmodulin-binding protein in Ca 2ϩ -free breast cell lysates (16).
Calmodulin is a highly conserved, ubiquitous protein involved in diverse Ca 2ϩ -dependent cellular processes, including cell cycle progression and proliferation, cyclic nucleotide metabolism, glycogen metabolism, cytoskeletal arrangement, and smooth muscle contraction (17). It possesses four Ca 2ϩ -binding sites, occupation of which effects a conformational change that facilitates association with multiple target proteins. Binding to calmodulin occurs via either basic amphiphilic ␣-helices or IQ motifs, 23 amino acid sequences with the consensus IQXXXRGXXXR (18). A "complete" IQ motif contains a C-terminal arginine in its consensus sequence, dictating no Ca 2ϩ requirement for calmodulin binding. Alternatively, an "incomplete" IQ motif without arginine requires Ca 2ϩ for binding (18).
A substantial body of evidence implicates calmodulin in carcinogenesis. For example, the level of calmodulin is significantly increased in malignant tissue (19), including breast carcinoma (20), and overexpression of calmodulin alters cell morphology and shortens the cell cycle (21). Although a causal relationship between calmodulin concentration and malignancy has not been demonstrated, it is hypothesized that increased concentrations of calmodulin may contribute to neoplastic transformation.
The downstream effectors of such transformation are unknown. One candidate is the recently isolated 189-kDa protein, IQGAP1. IQGAP1 contains three complete IQ motifs, one incomplete IQ, and a N-terminal region homologous to the actin and calmodulin-binding domain of calponin (22,23) (Fig. 1). In addition, IQGAP1 contains a region with significant sequence similarity to the catalytic domain of Ras-GAPs (15). IQGAP1 binds Rac and Cdc42 (22) and also cross-links microfilaments (24).
Recently, it was demonstrated that calmodulin binds the N-terminal region of IQGAP1 (22), which contains both the calponin homology domain (CHD) and the IQ motifs. Ca 2ϩ / calmodulin attenuates the association of IQGAP1 with Cdc42 (16) and F-actin (24). IQGAP1 thus appears to be an actinassociated protein that can transduce Ca 2ϩ /calmodulin signals to Cdc42 at the cytoskeleton. To further explore the functional sequelae of the interaction between Ca 2ϩ /calmodulin and IQ-GAP1, we isolated both full-length endogenous human IQ-GAP1 and glutathione S-transferase (GST) fusion constructs containing selected regions of IQGAP1. We show here that Ca 2ϩ binds directly to IQGAP1 and modulates the IQGAP1mediated inhibition of Cdc42-catalyzed GTP hydrolysis. We also present evidence that Ca 2ϩ /calmodulin competes with Factin for binding to the CHD of IQGAP1. Finally, we demonstrate that Ca 2ϩ enhances the binding of calmodulin to IQ-GAP1 thereby inducing the release of Cdc42 from IQGAP1. We conclude that IQGAP1 may provide a molecular link between Ca 2ϩ /calmodulin signaling pathways and Cdc42-mediated processes.

EXPERIMENTAL PROCEDURES
Materials-Tissue culture reagents were obtained from Life Technologies Inc. Fetal bovine serum was from Biowhittaker. Production of the GST-Cdc42 fusion protein has been described (22). Restriction enzymes and DNA Polymerase I (Klenow fragment) were purchased from New England Biolabs, Inc. Nucleotide primers for polymerase chain reaction were obtained from Genemed Biotechnologies. Radionucleotides were from DuPont. Calmodulin-Sepharose was purchased from Pharmacia Biotech Inc. G-actin was obtained from Sigma. All other reagents were of standard analytical grade.
Antibodies-Anti-GST and anti-Cdc42 polyclonal antibodies were purchased from Upstate Biotechnology Inc. and Santa Cruz Biotechnology, respectively. Anti-myoglobin monoclonal antibody was kindly provided by Dr. Jack Ladenson (Washington University School of Medicine, St. Louis, MO). Production of anti-calmodulin monoclonal antibody has been described previously (25). To produce anti-IQGAP1 antibody, GST-IQGAP1 N (N-terminal region, amino acids 1-863) was affinity purified, digested with thrombin, and GST was removed by glutathione affinity chromatography. Anti-IQGAP1 antiserum was raised by injecting rabbits with the purified N-terminal fragment of IQGAP1. For immunoprecipitation, anti-IQGAP1 antiserum was purified by protein A-Sepharose chromatography.
GST-IQGAP1 Fusion Constructs-The production of GST-IQGAP1 N and GST-IQGAP1 C (C-terminal region, amino acids 864 -1657) ( Fig. 1) has been described (22). The cDNA for GST-IQGAP1 CHD (calponin homology domain, amino acids 1-232) was generated by digestion of pGEX-2T-IQGAP1 N with EcoRI, gel purification of vector, and religation. To obtain the cDNA for GST-IQGAP1 IQ (IQ motifs, amino acids 740 -869), polymerase chain reaction on pcDNA3 vector containing full-length cDNA of IQGAP1 was performed using primers flanking nucleotides 2220 and 2608 of IQGAP1, with the 5Ј primer containing a BglII site. The resulting 388-base pair product was gel purified and cut with EcoRI. Blunt ends were generated with DNA Polymerase I (Klenow fragment). The fragment was then cut with BglII and subcloned into the BamHI site of pGEX-2T. GST-IQGAP1 fusion proteins were expressed in Escherichia coli. Bacteria were lysed by sonication in 2 mM EDTA, 0.2 mM phenylmethylsulfonyl fluoride, 5 mM benzamidine, and 10 mM dithiothreitol. Triton X-100 was added to 1% (v/v) and debris removed by centrifugation. Samples were loaded on glutathione-Sepharose columns and washed with PBS containing 10 mM dithiothreitol. GST fusion proteins were eluted with reduced glutathione and dialyzed against PBS. All GST fusion proteins migrated to the expected position on SDS-PAGE (see Fig. 4A).
Cdc42-catalyzed GTPase Activity Assay-2 g of purified GST-Cdc42 was preincubated with 1 mM EDTA in a buffer containing 25 mM Tris (pH 7.5), 1 mM dithiothreitol, and 125 mM NaCl for 10 min. Cdc42 was then loaded with radiolabeled GTP by incubation with 0.1 mM [␥-32 P]GTP in 30 M MgCl 2 for 15 min. Various combinations of purified full-length IQGAP1, calmodulin, and 1 mM CaCl 2 were added to equal aliquots of GTP-loaded GST-Cdc42. GTP hydrolysis was initiated by adding 7.5 mM MgCl 2 at 22°C. Control samples without added Mg 2ϩ were processed in parallel. The assay was terminated after 15 min by adding ice-cold PBS containing 5 mM MgCl 2 . Samples were spotted on Millipore HA 0.45-m filter membranes and filtered under vacuum with a Millipore Sampling Manifold Apparatus. Membranes were washed twice with ice-cold PBS containing 5 mM MgCl 2 , and radiolabeled GTP retained by Cdc42 was quantified by liquid scintillation spectrometry. GTP hydrolysis was the difference between [␥-32 P]GTP retained in the presence and absence of Mg 2ϩ .
In Vitro Binding Assays-Approximately 20 pmol of GST, GST-IQGAP1 N , GST-IQGAP1 C , GST-IQGAP1 CHD , and GST-IQGAP1 IQ were incubated with calmodulin-Sepharose in buffer A containing 1 mM CaCl 2 or 1 mM EGTA for 1 h at 22°C on a rotator. Where indicated, 200 pmol of F-actin or bovine serum albumin were added. Samples were washed four times in buffer A. Complexes were resolved by SDS-PAGE and transferred to PVDF membrane. Blots were probed with anti-GST antibody. Antigen-antibody complexes were visualized with horseradish peroxidase-conjugated secondary antibody and developed by ECL.
Actin competition analysis was also performed by incubating 0.16 M GST fusion protein and different concentrations of F-actin in 50 mM Tris (pH 7.4), 75 mM NaCl, 75 mM KCl, 1 mM CaCl 2 , and 0.1% Triton X-100 for 15 min at 22°C. After adding 1.2 M calmodulin, samples were incubated for 60 min at 22°C. Fusion proteins were isolated with glutathione-Sepharose and washed 4 times with PBS. Samples were resolved by SDS-PAGE, transferred to PVDF membrane, and blots were probed with anti-calmodulin antibody. Antigen-antibody complexes were visualized with horseradish peroxidase-conjugated secondary antibody and developed by ECL. Full-length IQGAP1 was isolated from MCF-7 cell lysates with calmodulin-Sepharose. After washing the beads four times with buffer A containing 1 mM CaCl 2 , 0.8 -6.4 M F-actin was added and samples were incubated in buffer A with 1 mM CaCl 2 for 1 h at 22°C. Following four washes with buffer A, samples were resolved by SDS-PAGE, transferred to PVDF membrane, and blots were probed with anti-IQGAP1 antibody.
In Vivo Binding Assays-MCF-7 cells were lysed in buffer A containing 1 mM CaCl 2 or 1 mM EGTA and subjected to centrifugation at 15,000 ϫ g for 5 min at 4°C to remove debris. The resultant supernatants were equalized for protein concentration. GST or GST-Cdc42 was incubated with 1 mM EDTA for 10 min at 22°C to remove bound guanine nucleotides, and then incubated with 75 M GTP␥S in 5 mM MgCl 2 for 30 min to reload. MCF-7 lysates were incubated with 1 g of GST-GTP␥S or GST-Cdc42-GTP␥S for 30 min at 22°C, and complexes were isolated with glutathione-Sepharose for 30 min at 22°C on a rotator. In separate experiments, equal amounts of protein lysate from MCF-7 cells were incubated with calmodulin-Sepharose in buffer A containing either 1 mM CaCl 2 or 1 mM EGTA for 1 h at 22°C on a rotator. Complexes were washed 4 times with buffer A, resolved by SDS-PAGE, and transferred to PVDF membrane. Blots were probed with anti-IQGAP1, anti-calmodulin, or anti-Cdc42 antibodies. Antigenantibody complexes were visualized with horseradish peroxidase-conjugated secondary antibody and developed by ECL.
Immunoprecipitation-MCF-7 cells were lysed in buffer A containing 1 mM EGTA and subjected to centrifugation at 15,000 ϫ g for 5 min at 4°C. Supernatants were divided into two equal aliquots, one of which was brought to 5 mM CaC1 2⅐ Samples were precleared with protein A-Sepharose for 30 min at 4°C. Anti-IQGAP1 antiserum or non-immune rabbit serum was added to the supernatant for 2 h at 4°C. Immune complexes were collected by incubating samples for 2 h at 4°C with protein A-Sepharose. Complexes were sedimented by centrifugation, washed five times with buffer A, and heated for 2 min at 100°C in solubilization buffer. Samples were resolved by SDS-PAGE, transferred to PVDF, and blots were probed with anti-IQGAP1, anti-Cdc42, or anti-calmodulin antibodies. Antigen-antibody complexes were visualized with the appropriate (rabbit or mouse) horseradish peroxidaseconjugated secondary antibody and developed by ECL.
Immunodepletion of IQGAP1 with Anti-calmodulin Antibody-MCF-7 cells were lysed in buffer A containing 1 mM CaCl 2 or 1 mM EGTA and subjected to centrifugation at 15,000 ϫ g for 5 min at 4°C. The supernatants were equalized for protein concentration. Equal aliquots of lysate were precleared with blocked Affi-Gel in buffer A with 1 mM CaCl 2 or 1 mM EGTA for 1 h at 4°C on a rotator, followed by the addition of either anti-calmodulin or anti-myoglobin monoclonal antibody linked to Affi-Gel for an additional 3 h at 4°C. Complexes were sedimented by centrifugation and removed. Proteins remaining in the supernatants were precipitated by trichloroacetic acid, heated at 100°C for 2 min in solubilization buffer, resolved by SDS-PAGE, and transferred to PVDF. An equal aliquot of untreated MCF-7 lysate was processed in parallel. Blots were probed with anti-IQGAP1 antibody. Antigen-antibody complexes were visualized with horseradish peroxidaseconjugated secondary antibody and developed by ECL.
Miscellaneous-Protein assays were performed using the DC Protein Assay from Bio-Rad. Densitometry of ECL signals was performed in triplicate and analyzed with NIH Image. G-actin was incubated with 1 mM ATP, 0.2 M KCl, and 1 mM MgCl 2 for 2 h to assemble F-actin. Statistical significance was assessed by Student's t test using InStat software (GraphPad Software, Inc.).

Ca 2ϩ Inhibits the Effect of IQGAP1 on Cdc42-catalyzed
GTPase Activity-We previously demonstrated that Ca 2ϩ /calmodulin modulates the association of IQGAP1 with Cdc42 (16), but no analysis of function was performed. To examine the role of both Ca 2ϩ /calmodulin and IQGAP1 on Cdc42 function, GTPase activity was assessed by [␥-32 P]GTP-Cdc42 filter binding assays (Fig. 2). Ca 2ϩ alone did not significantly alter GTP hydrolysis in the control samples. Similarly, neither Ca 2ϩ / calmodulin nor apocalmodulin produced a significant alteration of the intrinsic GTPase activity of Cdc42 (Fig. 2). In contrast, IQGAP1 inhibited GTPase activity of Cdc42 by 38 Ϯ 13% (mean Ϯ S.E., n ϭ 4) (p Ͻ 0.05) in the absence of Ca 2ϩ . This inhibition was not affected by the addition of calmodulin (Fig. 2). However, addition of Ca 2ϩ abrogated the effect of IQGAP1 on Cdc42-catalyzed GTP hydrolysis. Incubation of Ca 2ϩ /calmodulin, which prevents the association of Cdc42 with IQGAP1 (16), concurrently with IQGAP1 similarly eliminated regulation of the GTPase activity of Cdc42 by IQGAP1 (Fig. 2).
IQGAP1 Binds Ca 2ϩ -The direct effect of Ca 2ϩ on the modulation of Cdc42 activity by IQGAP1 suggested that IQGAP1 may bind Ca 2ϩ . To test this hypothesis, GST, GST fusion proteins containing the N-or C-terminal regions of IQGAP1, and purified IQGAP1 were immobilized on PVDF (Fig. 3A). Amido Black staining confirmed that all proteins bound to the membrane with similar affinity (data not shown). 45 Ca 2ϩ bound directly to full-length IQGAP1 and to the N-, but not the C-terminal region of IQGAP1 (Fig. 3A). To further localize the Ca 2ϩ -binding site, GST fusion proteins of the IQ and CHD regions of IQGAP1 were evaluated. This strategy demonstrated Ca 2ϩ binding to the CHD of IQGAP1, but not to the IQ region (Fig. 3B). GST alone did not bind Ca 2ϩ .

FIG. 2. Effect of Ca 2؉ /calmodulin on Cdc42-catalyzed GTPase activity.
[␥-32 P]GTP-loaded GST-Cdc42 was incubated with vehicle (control), calmodulin (CaM), IQGAP1, or calmodulin and IQGAP1 together (CaM/IQGAP1) in the absence (clear bars) or presence (hatched bars) of Ca 2ϩ as described under "Experimental Procedures." GTP hydrolysis was initiated by the addition of Mg 2ϩ . Radiolabeled GTP that remained bound to Cdc42 was quantified by liquid scintillation spectrometry. The data are expressed as a percentage, relative to control (no proteins added) and represent the mean Ϯ S.E. from four independent experiments. There was no statistically significant difference in GTP hydrolysis between the controls with Ca 2ϩ and without Ca 2ϩ . (*, significantly different from control (p Ͻ 0.05)). (22), but the exact location of the calmodulin-binding site on IQGAP1 had not been determined. Candidate regions include the four IQ motifs and the CHD, which displays sequence similarity to the calmodulin-binding region of calponin (22). Therefore, we examined the binding of GST fusion proteins containing the N-terminal, C-terminal, CHD, or IQ motifs of IQGAP1 to calmodulin-Sepharose. As anticipated, calmodulin bound to both the N-terminal domain and IQ motifs of IQGAP1 (Fig. 4B). As was observed with full-length IQGAP1 (16), binding affinity of calmodulin to both of these fusion proteins was higher in the presence than in the absence of Ca 2ϩ . Ca 2ϩ /calmodulin also bound to the CHD of IQGAP1, but apocalmodulin did not bind (Fig. 4B). Based on densitometry, it is estimated that the binding affinity of Ca 2ϩ /calmodulin for the CHD was approximately 20-fold lower than for the IQ motifs. Consistent with previous data (22), neither GST-IQGAP1 C nor GST alone bound to calmodulin (Fig. 4B). The lower molecular weight bands present in the GST-IQGAP1 N and GST-IQGAP1 IQ lanes likely represent degradation fragments of the fusion proteins (see Fig. 4A).

Calmodulin Binds to Both the CHD and IQ Motifs of IQ-GAP1-Calmodulin binds to the 95-kDa N-terminal region of IQGAP1
Actin Inhibits the Binding of Calmodulin to the CHD of IQGAP1-The CHD displays sequence similarity to the actinbinding region of calponin (22,23). Since actin associates with IQGAP1 (24), the ability of actin to compete with calmodulin for binding to the CHD of IQGAP1 was examined. Addition of F-actin prevented the binding of the GST-IQGAP1 CHD to calmodulin (Fig. 5). An equimolar amount of bovine serum albumin had a negligible effect, impeding binding by only 15% (Fig.  5A).
F-actin competition analysis was also performed with calmodulin in solution and complexes were isolated with glutathione-Sepharose (Fig. 5B). Probing the resultant blots with anticalmodulin antibody confirmed that F-actin substantially decreased calmodulin binding to the CHD of IQGAP1. Bovine serum albumin did not decrease calmodulin binding to the CHD (data not shown). Separate analysis performed by actin immunoblotting revealed that F-actin bound to the CHD of IQGAP1 (data not shown). By contrast, calmodulin binding to the IQ region (Fig. 5B) and to full-length IQGAP1 (Fig. 5C) were not altered by F-actin; increasing the F-actin concentration by 8-fold (up to 6.4 M) had no effect (data not shown). There is substantially more calmodulin in the GST-IQGAP1 IQ sample than in the GST-IQGAP1 CHD sample in Fig. 5B because of the significantly higher affinity of calmodulin for the IQ region (see Fig. 4B).
Ca 2ϩ /Calmodulin Prevents the in Vivo Association of IQ-GAP1 with Cdc42-Ca 2ϩ /calmodulin inhibits the in vitro binding of IQGAP1 to Cdc42 (16). To characterize this interaction in a normal cellular milieu, we examined the effect of calmodulin on the binding of GST-Cdc42 to endogenous IQGAP1 in MCF-7 human breast epithelial cell lysates. Endogenous IQGAP1 bound to activated GST-Cdc42 (Fig. 6A). Binding affinity was greater in the absence than in the presence of Ca 2ϩ . In addition, calmodulin was isolated from cell lysates by GST-Cdc42 affinity chromatography (Fig. 6A). Because calmodulin does not associate directly with Cdc42 (16,22), the calmodulin was presumably bound to IQGAP1 in a ternary complex with Cdc42. The reduced amount of calmodulin detected upon addition of Ca 2ϩ (Fig. 6A) can be accounted for by two possible mechanisms. First, Ca 2ϩ /calmodulin may mediate a decrease in the affinity of IQGAP1 for Cdc42 (16). An alternative mechanism that has to be excluded is a decrease in the affinity of IQGAP1 for Ca 2ϩ /calmodulin. To distinguish between these two possibilities, calmodulin-Sepharose affinity chromatography was performed. Ca 2ϩ increased the binding of endogenous IQGAP1 to calmodulin by 2.39 Ϯ 0.25-fold (mean Ϯ S.E., n ϭ 4), with a concomitant decrease in the amount of endogenous Cdc42 that was retained by the calmodulin-Sepharose (Fig.   FIG. 3. 45 Ca 2؉ binding to IQGAP1. GST, purified IQGAP1 and GST fusion proteins (GST-IQGAP1 N (N), GST-IQGAP1 C (C), GST-IQGAP1 IQ (IQ), and GST-IQGAP1 CHD (CHD)) were applied to PVDF membrane and incubated with 45 CaCl 2 as described under "Experimental Procedures." After washing, Ca 2ϩ -binding proteins were visualized by autoradiography. The experiments depicted in panels A and B were performed on different days. The data are representative of three independent experiments. 6B). These data suggest that Ca 2ϩ enhances the binding of calmodulin to IQGAP1 and that this binding effects the release of Cdc42 from IQGAP1.
Ca 2ϩ Modulates the Binding of Cdc42 and Calmodulin to IQGAP1-The specificity of the anti-IQGAP1 antibody was evaluated before using it in immunoprecipitation analysis. IQ-GAP1 was the only protein recognized in MCF-7 cell lysates by the anti-IQGAP1 antibody (Fig. 7A). The antibody, which was raised against the N-terminal fragment of IQGAP1, bound to GST-IQGAP1 N but not GST-IQGAP1 C (data not shown). The effect of Ca 2ϩ on the interaction among calmodulin, Cdc42, and IQGAP1 was analyzed further by co-immunoprecipitation. Ca 2ϩ did not alter the binding of IQGAP1 to its antibody (Fig.  7B). By contrast, Ca 2ϩ decreased by 1.5-fold the amount of Cdc42 that co-immunoprecipitated with IQGAP1 (Fig. 7B). Consistent with the calmodulin-Sepharose data in Fig. 6, substantially more calmodulin co-immunoprecipitated with IQ-GAP1 in the presence of Ca 2ϩ than in the absence of Ca 2ϩ (Fig.  7B). Overexposure of the ECL image revealed that a small amount of calmodulin co-immunoprecipitated with IQGAP1 in the absence of Ca 2ϩ (data not shown). The specificity of the interactions is revealed by the absence of immunoreactive proteins in the samples immunoprecipitated with non-immune serum (Fig. 7B, lanes 3 and 4).
Anti-calmodulin Antibody Depletes MCF-7 Lysates of IQ-GAP1-Previous in vitro assays have shown that Ca 2ϩ does not impair the binding of IQGAP1 to Cdc42 (16). However, less IQGAP1 from MCF-7 lysates bound to GST-Cdc42 in the presence than in the absence of Ca 2ϩ (Fig. 6A). Moreover, Ca 2ϩ decreased the amount of Cdc42 that co-immunoprecipitated with IQGAP1 (Fig. 7). An explanation for these findings may be that in the normal cellular milieu, Ca 2ϩ /calmodulin binds a substantial amount of endogenous IQGAP1, thereby preventing its association with Cdc42. To evaluate this hypothesis, MCF-7 cell lysates were subjected to immunodepletion with anti-calmodulin antibody. Notably, in the presence of Ca 2ϩ , but not EGTA, incubation with anti-calmodulin antibody reduced by 53 Ϯ 1% (mean Ϯ S.E., n ϭ 3) the amount of IQGAP1 in MCF-7 lysates (Fig. 8). Incubation with an irrelevant isotypeidentical monoclonal antibody (anti-myoglobin IgG 1 ) did not alter the amount of IQGAP1 remaining in the supernatant. These data confirm that a significant proportion of endogenous FIG. 5. Effect of actin on the association of IQGAP1 with calmodulin. A, GST-IQGAP1 CHD (CHD) was preincubated with vehicle (control), F-actin, or bovine serum albumin (BSA). GST-IQGAP1 CHD and GST were then incubated with calmodulin-Sepharose in the presence of 1 mM Ca 2ϩ as described under "Experimental Procedures." Complexes were resolved by SDS-PAGE, transferred to PVDF membrane, and blots were probed with antibody to GST. Immune complexes were visualized by ECL. The position of migration of GST-IQGAP1 CHD is depicted. The data are representative of two independent experiments. B, GST-IQGAP1 CHD (CHD) or GST-IQGAP1 IQ (IQ) was preincubated in the presence of 1 mM Ca 2ϩ with (ϩ) or without (Ϫ) 0.8 M F-actin. Calmodulin was added and complexes were isolated with glutathione-Sepharose. Proteins were resolved by SDS-PAGE, transferred to PVDF membrane, and blots were probed with anti-calmodulin antibody. Immune complexes were visualized by ECL. The position of migration of calmodulin (CaM) is depicted. C, IQGAP1 was isolated from cell lysates with calmodulin-Sepharose as described under "Experimental Procedures." Samples were incubated with (ϩ) or without (Ϫ) 0.8 M F-actin and processed as described for panels A and B, except blots were probed with anti-IQGAP1 antibody. All data are representative of at least two independent experiments. IQGAP1 is bound to endogenous Ca 2ϩ /calmodulin. Taken together with the data in Figs. 6 and 7, these results suggest that Ca 2ϩ increases the binding of calmodulin to IQGAP1 and thus elicits the dissociation of Cdc42 from IQGAP1.

DISCUSSION
Cdc42 facilitates cell cycle progression and cytoskeletal rearrangement (3,4). However, the regulators that control Cdc42 function and the downstream effectors that link activated Cdc42 to actin reorganization have remained obscure. In this paper, we characterize the regulatory pathways connecting Ca 2ϩ , calmodulin, IQGAP1, and Cdc42.
Previous in vitro data support a model in which Ca 2ϩ /calmodulin regulates Cdc42-mediated GTPase activity through the intermediary protein IQGAP1 (16,22). The C-terminal region of IQGAP1 has significant sequence similarity to the catalytic domain of all Ras-GAPs and has been hypothesized to act as a GAP (15). However, data from Hart et al. (22) indicate that recombinant IQGAP1 actually stabilizes the GTP-bound state of Cdc42. To elucidate this interaction, the effect of purified, full-length, endogenous IQGAP1 on Cdc42 activity was examined. We demonstrate that Ca 2ϩ , independently of calmodulin, modulates the regulation of Cdc42 activity by IQ-GAP1. Specifically, while IQGAP1 alone maintains Cdc42 in its active GTP-bound state, Ca 2ϩ /IQGAP1 fails to inhibit the intrinsic GTPase activity of Cdc42. As Ca 2ϩ alone does not impair the binding of IQGAP1 to Cdc42 (16), this Ca 2ϩ effect is not secondary to dissociation of IQGAP1 from Cdc42.
We theorized that the direct effect of Ca 2ϩ on IQGAP1 may be due to direct association of Ca 2ϩ with IQGAP1. Notably, amino acids 48 -161 in the N-terminal region of IQGAP1 display significant sequence similarity to MP-20, a putative Ca 2ϩbinding Drosophila muscle protein (15,26). Indeed, the Nterminal region and CHD (which includes amino acids 48 -161) of IQGAP1 were shown to bind Ca 2ϩ directly by 45 Ca 2ϩ overlay. As the GAP homology region that associates with Cdc42 is in the C-terminal half of IQGAP1 (see Fig. 1), Ca 2ϩ binding presumably promotes a conformational change in IQGAP1.
Our previous work revealed that calmodulin in the presence of Ca 2ϩ effects the dissociation of IQGAP1 from Cdc42 (16). In vitro, Ca 2ϩ /calmodulin appears then to be a negative regulator of Cdc42 activity on two levels. In the absence of Ca 2ϩ , IQGAP1 binds Cdc42 and maintains it in its active GTP bound state. An increase in Ca 2ϩ concentration both abolishes the inhibition of Cdc42 GTPase activity by IQGAP1 and, in the presence of calmodulin, dissociates IQGAP1 from Cdc42. Although we have not demonstrated this thesis in vivo, the data in Figs. 6 and 7 provide support for this model in the normal cell milieu. The results presented in Fig. 8 further substantiate a physiological role for calmodulin in IQGAP1 function, as a substantial proportion of endogenous IQGAP1 is bound to Ca 2ϩ /calmodulin.
A second mechanism for Ca 2ϩ /calmodulin regulation of Cdc42 via IQGAP1 involves subcellular localization. Bashour et al. (24) have shown that IQGAP1 interacts with and cross-links microfilaments. Calmodulin inhibits this interaction. In addition, immunohistochemical studies colocalize IQGAP1 with cytochalasin D-sensitive microfilaments in lamellipodia and membrane ruffles (24). Here, we demonstrate that Ca 2ϩ /calmodulin binds to the CHD of IQGAP1 and F-actin diminishes this association. Since F-actin does not bind directly to calmodulin (17), our data indicate that F-actin binds to the CHD of IQGAP1. Because Ca 2ϩ /calmodulin and F-actin compete for binding to the CHD, IQGAP1 may couple Cdc42 to microfilaments in the absence of Ca 2ϩ . Indeed, GTP-bound Cdc42 coimmunoprecipitated with IQGAP1 and F-actin (27) and enhanced F-actin cross-linking by IQGAP1 (28). In the presence of Ca 2ϩ , calmodulin may dissociate IQGAP1 from not only Cdc42 but also from F-actin, insuring a separation of Cdc42 from microfilaments. However, if the results of our in vitro GTPase analysis are replicated in intact cells and IQGAP1 does indeed inhibit the intrinsic GTPase activity of Cdc42 in the absence of Ca 2ϩ , an alternative possibility exists. Active GTP-bound Cdc42 may act upstream of IQGAP1 and may inhibit a function of IQGAP1 at the cytoskeleton. In the absence of Ca 2ϩ , IQGAP1 may bind activated Cdc42, stabilize its active GTP-bound state, and localize it to cytoskeletal structures. There, active Cdc42 may inhibit the effect of IQGAP1 on cytoskeletal rearrangement. Increased intracellular Ca 2ϩ may abolish the inhibition of Cdc42 GTPase activity by IQGAP1 and may dissociate IQGAP1 from both Cdc42 and microfilaments. In this model, IQGAP1 is both a regulator of Cdc42 localization and activity, and a downstream target of Cdc42 function. The N-terminal of IQGAP1 binds Ca 2ϩ /calmodulin and actin, while the C-terminal interacts with Cdc42. As such, IQGAP1 may serve as a scaffold for a multimeric actin complex, providing a molecular link between the cytoskeleton and Cdc42 and mediating a regulatory role by Ca 2ϩ /calmodulin.
The functional sequelae of the interaction of Ca 2ϩ /calmodulin with IQGAP1 remain under investigation. The interaction between calmodulin and IQGAP1 is likely physiologic, as IQ-GAP1 is the predominant calmodulin-binding protein in Ca 2ϩfree breast carcinoma cell lysates (16) and a substantial fraction of endogenous IQGAP1 is bound to Ca 2ϩ /calmodulin in the normal cellular milieu. As Cdc42 participates in cell proliferation and regulation of the cytoskeleton, involvement of Ca 2ϩ and calmodulin in such processes may involve IQGAP1 as an intermediary.