IQGAP1 Binds Rap1 and Modulates Its Activity*

IQGAP1 is a scaffolding protein involved in multiple fundamental cellular activities, including transcription, cell-cell attachment, and regulation of the cytoskeleton. To function in these pathways, IQGAP1 associates with numerous proteins such as actin, calmodulin, E-cadherin, β-catenin, CLIP-170, and components of the mitogen-activated protein kinase pathway. Moreover, IQGAP1 binds to active Cdc42 and Rac1 but not RhoA or Ras. Here we show that IQGAP1 also binds to the small GTPase Rap1. In vitro analysis demonstrates a direct interaction between Rap1 and IQGAP1, which is augmented by activation (GTP loading) of Rap1. Cdc42 does not modulate the interaction between Rap1 and IQGAP1. In contrast, the association is eliminated by calmodulin both in the absence and presence of Ca2+. The binding of Rap1 to a point mutant IQGAP1 construct that is unable to interact with calmodulin is 2.5-fold more than to wild type IQGAP1. Consistent with these findings, Rap1 binds to the IQ region of IQGAP1. Confocal microscopy demonstrates that Rap1 and IQGAP1 co-localize at the periphery of human epithelial cells but not in the cytoplasm. The interaction has functional sequelae. Overexpression of IQGAP1 substantially reduces adhesion-mediated activation of Rap1. In addition, Rap1 activation by cAMP is attenuated in cells that overexpress IQGAP1 and enhanced in cells lacking IQGAP1. These findings reveal that the interaction of IQGAP1 with Rap1 differs in several respects from its interaction with other small GTPases. Furthermore, our data suggest that IQGAP1 may link the calmodulin and Rap1 signaling pathways.

The Ras superfamily of small GTPases comprises ϳ150 members in humans (1,2). Members of the Ras family, which contains 36 proteins, include the Ras proteins (H-Ras, K-Ras, and N-Ras) and several related proteins such as R-Ras proteins (R-Ras, R-Ras2, and M-Ras), Ral and Rap (2). Like Ras, Rap1 cycles between an active GTP-bound form and an inactive GDP-bound form. Guanine nucleotide exchange factors (GEFs), 2 such as Epac (exchange protein directly activated by cAMP), C3G, and CD-GEF, mediate Rap1 activation through exchange of GDP for GTP (3). Rap1 is inactivated by GTPaseactivating proteins (GAPs), which catalyze the hydrolysis of bound GTP to GDP. A wide variety of extracellular stimuli, for example thrombin, growth factors, and interferon, activate Rap1. These effects are mediated by second messengers, including Ca 2ϩ , diacylglycerol, and cAMP, which directly stimulate Rap1 GEFs. Ras and Rap1 are very similar at the amino acid level, with identical effector regions (4). Nevertheless, the proteins have distinct functions. Rap1 has been implicated in the modulation of a number of cellular responses ranging from Ca 2ϩ signaling and secretion to neurite outgrowth and cell proliferation (3). A well characterized function of Rap1 is in cell adhesion, both integrin-mediated adhesion and cadherin-mediated cell junction formation (5). Rap1 interacts with a large number of proteins that contribute to regulating these aspects of cell function (5).
In addition to the binding partners mentioned above, IQGAP1 also interacts with selected small GTPases. Direct binding of IQGAP1 to active (GTP-bound) Cdc42 (13,14), Rac1 (14), and TC10 (24) has been documented. However, IQGAP1 does not bind RhoA or H-Ras (14,25). In this study we show for the first time that IQGAP1 interacts directly with a GTPase that is not a member of the Rho subfamily. The association of IQGAP1 with Rap1, which differs in several respects from its interaction with other small GTPases, alters the activation of Rap1.
We generated maltose-binding protein (MBP) fusion constructs of wild type Rap1A, Rap1A-63E, and Rap1A-17N. The BamHI-HindIII fragment containing Rap1A was cut from pRSET-Rap1A and inserted into pMAL-c2X at BamHI and HindIII sites. The proteins migrated to the expected position on SDS-PAGE.
GST-Rap1A was generated by PCR on full-length Rap1A using the primers 5Ј-CGGGATCCCCGTGAGTTACAAGTC-TAGTGG-3Ј and 5Ј-CGGAATTCTCAGAGCAGCAGACAT-GATTTC-3Ј. The PCR product was cut with BamHI and EcoRI and inserted into pGEX4T-1 at the same restriction sites.
IQGAP1 was tagged with red fluorescent protein (RFP) using monomeric mRFP which was generously donated by Roger Tsien (University of California, San Diego, CA). A PCR product of mRFP was generated using primers 5Ј-GCTCTAGAATGG-CCTCCTCCGAGGACG-3Ј and 5Ј-GAAGATCTGGCGCCG-GTGGAGTGGCGTCT-3Ј with pRSETB-mRFP1 as template. The PCR product was cut with XbaI and BglII, and the insert was replaced with a NheI-BglII fragment bearing the DsRed gene on pDsRed2-C1 (Clontech). To generate RFP-wild type IQGAP1, pcDNA- Myc-IQGAP1 was cut with XbaI and made  blunt end with T4 polymerase then partially digested with  BamHI. The BamHI-XbaI fragment bearing the whole IQGAP1  gene was inserted into mRFP-C1 at BglII-SmaI site. The RFP  tag was attached to IQGAP1⌬IQ and IQGAP1 IQ3,4R using RFP-tagged wild type IQGAP1. The PacI-SpeI fragment was cut from mRFP-IQGAP1(wild type) and replaced with the same fragment from pcDNA-myc-IQGAP1⌬IQ or pcDNA-Myc-IQGAP1 IQ3,4R.
Gel Filtration Chromatography-MCF-7 and MCF-siIQ8 cells (8 dishes/10 cm each) were lysed in 2 ml of buffer composed of 50 mM Tris, pH 7.4, 150 mM NaCl, 1 mM EGTA, 1% Nonidet P-40, 20% glycerol, and protease inhibitors. The lysate was loaded onto a Superose 6 column (Amersham Biosciences) pre-equilibrated with the same buffer. The sample was fractionated at 1 ml/min (0.5-ml fractions were collected) by fast protein liquid chromatography separation performed on an AKTATMFPLC system (Amersham Biosciences) equipped with a UPC-900 monitor and a P-920 pump. The system was monitored and controlled by methods run by the UNICORN control system (Version 5.01). The column was calibrated using apoferritin (443 kDa), ␤-amylase (200 kDa), and bovine serum albumin (66 kDa) as standards. Forty-microliter aliquots of selected fractions were resolved by Western blotting, then probed for IQGAP1 and Rap1.
Assay for Activation of Rap1 and Cdc42-Active Cdc42 (31) and active Rap1 (32) were assessed essentially as described pre-viously. Briefly, cells were washed with ice-cold PBS and lysed with buffer A. Lysates were cleared by centrifugation, and active Rap1 and Cdc42 were precipitated with glutathione-Sepharose beads precoupled to a GST fusion protein of the Rap1 binding domain of RalGDS for Rap1 and the GTPase binding domain of WASP (Wiskott-Aldrich syndrome protein) for Cdc42. Samples were incubated for 40 min (for Rap1) or 30 min (for Cdc42). Beads were sedimented by centrifugation, washed three times with buffer A, and solubilized in SDS sample buffer. A portion of the cell lysate was reserved for analysis of total Rap1 or Cdc42 content. Cdc42 and Rap1 were detected by Western blotting.
8CPT-2Me-cAMP Treatment-Equal numbers of MCF-7 or MCF/I cells were serum-starved in Dulbecco's modified Eagle's medium containing 1% bovine serum albumin (BSA). After 24 h, cells were stimulated with 100 M 8CPT-2Me-cAMP for the times indicated in the legend to Fig. 9. GTP-bound Rap1 was detected by pulldown with GST-RalGDS as described above. Similar analysis was done with MEFs derived from IQGAP1-null mice and normal littermate controls.
In Vitro Binding Analyses-For binding assays using GSTtagged proteins, MBP-Rap1A was loaded with either GDP or GTP␥S essentially as previously described (13,33). Briefly, MBP-Rap1A was incubated in buffer B (50 mM Tris, pH 7.4, 150 mM NaCl, and 1% Triton X-100) with 1 mM EDTA. After 10 min at 22°C, 3 mM MgCl 2 and 140 M GTP␥S or GDP were added, and samples were incubated for 30 min at 22°C. Samples were precleared with 40 l of glutathione-Sepharose for 1 h, and equal amounts of GST-IQGAP1 in 500 l of buffer B containing 1 mM EGTA were added for 3 h at 4°C. Complexes were isolated with glutathione-Sepharose, washed three times in buffer B, and resolved by SDS-PAGE. Western blots were probed with anti-Rap1 antibodies.
For binding assays with MBP-tagged proteins, equal amounts of wild type MBP-Rap1A (loaded with GTP␥S or GDP), dominant-negative Rap1A-17N, or constitutively active Rap1A-63E were incubated with purified IQGAP1 in 500 l of buffer B for 3 h at 4°C. Complexes were isolated with amylose beads, washed, and resolved by SDS-PAGE, and blots were probed with anti-IQGAP1 antibodies (26).
To compare binding of Rap1A and Rap1B to IQGAP1, we used GST-tagged Rap1 proteins. IQGAP1 was cleaved from GST using tobacco etch virus protease as described (29). GST, GST-Rap1A, and GST-Rap1B were loaded with GTP␥S as described above. Equal amounts of purified IQGAP1 were incubated with equal amounts of GST␥S-loaded GST, GST-Rap1A, or GTP Rap1B in 20 mM Tris-HCl, pH 8.0, 137 mM NaCl, 0.2% Triton X-100, 10% glycerol, and 1 mM EGTA for 2 h at 4°C. Complexes were isolated with glutathione-Sepharose, washed, and resolved by SDS-PAGE. The gel was cut into two pieces; the top portion was transferred to polyvinylidene difluoride membranes, and blots were probed with anti-IQGAP1 antibody. The lower half was stained with Coomassie Blue to visualize GSTtagged proteins.
[ 35 S]Methionine-labeled transcription and translation (TNT) products were generated with the TNT Quick Coupled Transcription/Translation system (Promega) according to the manufacturer's instructions. Briefly, 0.5 g of the IQGAP1 plasmids was incubated with 40 l of TNT Quick Master mix and 20 Ci of [ 35 S]methionine (PerkinElmer Life Sciences) at 30°C for 1 h. Products were confirmed by SDS-PAGE and autoradiography. Equal amounts of radiolabeled peptide were incubated with His-Rap1A-63E. Complexes were isolated with Ni 2ϩ -NTA agarose, resolved by SDS-PAGE, and processed by autoradiography.
Competitive Inhibition Analysis-[ 35 S]Methionine-labeled IQGAP1 was preincubated with 5 g of calmodulin, 5 g of constitutively active Cdc42 (Cdc42-Q61L), or 5 g of BSA in 500 l of buffer B containing 1 mM Ca 2ϩ or 1 mM EGTA for 1 h at 4°C. Equal amounts of His-Rap1A-63E protein were added, and samples were incubated another 1 h. The complexes were washed and isolated with Ni 2ϩ -NTA agarose. Samples were resolved by SDS-PAGE, and bands were analyzed by autoradiography of the dried gel. For calmodulin-Sepharose chromatography, [ 35 S]methionine-labeled IQGAP1 was preincubated with 5 g of His-Rap1A-63E or 5 g of BSA in the presence of Ca 2ϩ or EGTA for 1 h at 4°C. Equal amounts of calmodulin-Sepharose or empty Sepharose were added. After 1 h beads were isolated and washed three times in buffer B containing Ca 2ϩ or EGTA. Proteins bound to calmodulin were resolved by SDS-PAGE and processed by autoradiography of the dried gel.
Immunofluorescence Staining and Confocal Microscopy-MCF-7 or MCF-siIQ8 cells were transiently transfected with GFP-Rap1A alone or in combination with RFP-tagged IQGAP1 constructs. Cells were plated on glass coverslips that had been coated overnight with 10 g/ml type I collagen or 10 g/ml poly-L-lysine (Sigma-Aldrich) and blocked with 0.5% BSA. Immunocytochemistry was performed as described (23,31) with a few modifications. Cells were washed twice with PBS and incubated in 4% paraformaldehyde-PBS for 20 min at 22°C. After washing twice in PBS, cells were permeabilized in 0.2% Triton X-100 with 3% BSA in PBS for 1 h at 22°C. Polyclonal anti-IQGAP1 antibody was added in 0.2% Triton, 1% BSA for 16 h at 4°C followed by Alexa-Fluor 543-labeled anti-rabbit IgG secondary antibody (Molecular Probes) for 1 h. Where appropriate, actin was visualized with Alexa-Fluor 543-conjugated phalloidin. Rabbit IgG was used as a control for the corresponding primary antibody. The stained cells were analyzed using a Zeiss LSM 510 confocal microscope and analyzed with Zeiss LSM software.
Miscellaneous-Densitometry of enhanced chemiluminescence (ECL) signals was analyzed with UN-SCAN-IT software (Silk Scientific Corp.). Statistical analysis was assessed by Student's t test with Instat software (GraphPad Software, Inc.). Protein concentrations were determined with the DC protein assay (Bio-Rad).

RESULTS
Direct Interaction between IQGAP1 and Rap1-In vitro analysis was conducted to examine a possible association between IQGAP1 and Rap1. Purified MBP-Rap1A was loaded with either non-hydrolyzable GTP␥S to form active Rap1A or with GDP to produce inactive Rap1A. Equal amounts of the Rap1A constructs were incubated with purified GST-IQGAP1. Pulldown with glutathione-Sepharose beads revealed that Rap1A binds to IQGAP1. IQGAP1 binds more efficiently to GTP-Rap1A than to the inactive form of Rap1A ( Fig. 1A). Binding was specific, as no Rap1A associated with GST alone. Coomassie staining reveals that equal amounts of GST-IQGAP1 were present in the assay (Fig. 1A).
To confirm these findings, the reverse experiment was done. MBP-Rap1A was incubated with purified IQGAP1. Pulldown with amylose beads reveals that more IQGAP1 binds to GTPloaded Rap1A than to GDP-Rap1A (Fig. 1B). Similar analysis was conducted with MBP-tagged forms of the constitutively active Rap1A-63E and dominant negative Rap1A-17N. Consistent with the data obtained with guanine nucleotide loading, IQGAP1 associates with Rap1A-63E but not with Rap1A-17N (Fig. 1B). The amounts of MBP-Rap1A present in the assay were equivalent (Fig. 1B). Collectively, these data reveal that Rap1A interacts directly with IQGAP1 in a GTP-regulated manner.
Rap1 has two isoforms, Rap1A and Rap1B, which differ in only a few amino acids, predominantly in the hypervariable region (3). To determine whether IQGAP1 interacts with both isoforms, we evaluated in vitro binding. Equal amounts of GST-Rap1A and GST-Rap1B (Fig. 1C) were loaded with GTP␥S and incubated with purified IQGAP1. The amounts of IQGAP1 bound to Rap1A and Rap1B were approximately equal ( Fig.  1C), implying that the binding domain on Rap1 lies outside the hypervariable region. All subsequent in vitro analyses were conducted with Rap1A.
Calmodulin Abrogates the Binding of Rap1A to IQGAP1-Ca 2ϩ / calmodulin is known to influence the binding of IQGAP1 to several target proteins, most likely by altering the conformation of IQGAP1 (15). Therefore, we investigated the effect of calmodulin on the interaction between IQGAP1 and Rap1A. [ 35 S]Methionine-labeled IQGAP1 was preincubated with calmodulin, and purified His-Rap1A-63E was subsequently added. Complexes were isolated with Ni 2ϩ -agarose, and the association was detected by autoradiography. Calmodulin abrogates binding of IQGAP1 to Rap1A ( Fig. 2A). The effects of calmodulin are independent of Ca 2ϩ , as disruption occurred both in the presence of Ca 2ϩ and when Ca 2ϩ was chelated with EGTA. Disruption was specific for calmodulin because an equivalent amount of BSA did not reduce binding ( Fig. 2A).
To extend these findings, we evaluated the effect of Rap1A on the binding of IQGAP1 to calmodulin. Radiolabeled IQGAP1 was preincubated with Rap1A-63E, and complexes were isolated with calmodulin-Sepharose. In contrast to the inhibition by calmodulin, active Rap1A did not modify the binding of IQGAP1 to calmodulin (Fig. 2B). Note that more IQGAP1 binds to calmodulin in the presence than in the absence of Ca 2ϩ , as observed previously by us (13) and others (11).
Cdc42, an IQGAP1 binding partner, regulates the association of selected targets with IQGAP1 (15). Therefore, we examined the effect of constitutively active Cdc42 (Cdc42-Q61L) on the interaction between IQGAP1 and Rap1A. Preincubation of Cdc42-Q61L with IQGAP1 did not significantly affect the binding of IQGAP1 to Rap1A-63E ( Fig. 2A). This finding was unexpected as we anticipated that Rap1A and Cdc42 would bind to the same region of IQGAP1 and compete with one another.
Identification of the Rap1A Binding Domain on IQGAP1-To ascertain the region of IQGAP1 to which Rap1 binds, pulldown assays with [ 35 S]methionine-labeled IQGAP1 were performed.  ]methionine-labeled IQGAP1 was preincubated with 5 g of calmodulin (CaM), 5 g of constitutively active Cdc42 (Cdc42Q61L) or 5 g of BSA in the presence of Ca 2ϩ or EGTA. After incubating for 1 h at 4°C, constitutively active His-Rap1A-63E was added for an additional 1 h. Radiolabeled IQGAP1 was also incubated with Ni 2ϩ -NTA resin (Ni 2ϩ ) alone as control. Complexes were isolated with Ni 2ϩ -NTA-agarose beads and resolved by SDS-PAGE. Gels were dried and processed by autoradiography. Data are representative of three independent experiments. B, [ 35 S]methionine-labeled IQGAP1 was preincubated with 5 g of His-Rap1A-63E (63E) or 5 g of BSA in the presence of Ca 2ϩ or EGTA for 1 h at 4°C. Calmodulin-Sepharose (CaM-Seph) or empty Sepharose (Seph) beads was added for an additional 1 h. Samples were processed by SDS-PAGE and autoradiography. Data are representative of two independent experiments.
Selected constructs of IQGAP1 (depicted in Fig. 3A) were labeled with [ 35 S]methionine in a reticulocyte lysate and incubated with His-tagged Rap1A-63E. Constructs that bound to Rap1A were identified by autoradiography. As was shown in Fig. 2, full-length IQGAP1 binds Rap1A-63E (Fig. 3B). The N-terminal half (comprising amino acids 2-863) of IQGAP1 also binds to Rap1A, but no binding of the C-terminal half of IQGAP1 was detected. To further narrow the binding site, selected IQGAP1 constructs lacking specific protein interaction motifs in the N-terminal half were used. Analysis showed that the WW domain residue (643-744) was not necessary for IQGAP1 to bind Rap1A (Fig. 3B). In contrast, absence of the region that includes the IQ domain (717-916) completely disrupted the association between IQGAP1 and Rap1A. The data in Fig. 3D reveal that the amounts of radiolabeled IQGAP1 constructs used were equivalent. Collectively, these data suggest that the IQ domain of IQGAP1 is necessary for Rap1A binding. To determine whether the IQ domain is sufficient for Rap1A binding, we incubated with Rap1A-63E a peptide comprising amino acids 717-916 of IQGAP1 which encompasses the IQ region. This reveals that the IQ domain of IQGAP1 interacts directly with Rap1A (Fig. 3C), implying that it is sufficient to mediate Rap1 binding.
Binding of Mutant IQGAP1 to Rap1-The binding of IQGAP1 to Rap1 in a normal cell milieu was examined. His-Rap1A-63E, incubated with lysates from 293H cells transfected with IQGAP1, was isolated with Ni 2ϩ -agarose. Probing the resultant Western blots revealed that IQGAP1 binds specifically to Rap1A (Fig. 4A). We were unable to reproducibly and specifically co-immunoprecipitate Rap1 and IQGAP1 from cell lysates. The reason for this is unknown. This may be due to masking of the antibody recognition epitope in the complex or conformation of the proteins when bound. Another possible explanation is that the high concentrations of endogenous calmodulin prevent detection. To test this hypothesis, we used an IQGAP1 mutant construct, termed IQGAP1 IQ3,4R. This construct has point mutations of selected Arg residues in the IQ domain that abrogate calmodulin binding (28). IQGAP1 IQ3,4R and wild type IQGAP1 were transiently transfected to equivalent levels in 293H cells (Fig. 4A, upper panel). Analysis reveals that the amount of IQGAP1 IQ3,4R pulled down by Rap1A is 2.5-fold greater than that of wild type IQGAP1 (Fig. 4). Equal amounts of His-Rap1A were present in each sample (data not shown). These data demonstrate that elimination of calmodulin binding to IQGAP1 facilitates the binding of Rap1.
Rap1 and IQGAP1 Co-elute on Gel Filtration Chromatography-Gel filtration was used to gain further insight into the interaction between the IQGAP1 and Rap1 in a normal cell milieu. Lysates from MCF-7 cells were resolved by gel filtration chromatography. Although it has a molecular mass of ϳ21 kDa, the majority of the endogenous Rap1 eluted in a single peak corresponding to ϳ150 -450 kDa (Fig. 5A). This pattern coincides with that of the majority of endogenous IQGAP1 (Fig.  5A). The elution of IQGAP1 is consistent with our earlier findings, obtained with purified protein, that IQGAP1 oligomerizes and elutes in peaks corresponding to multimers (29). A small fraction of Rap1 elutes after the 60-kDa marker. Some of this is likely to be unbound Rap1.
If binding to IQGAP1 accounts for the appearance of Rap1 in a higher molecular mass complex, one would anticipate that reduction of endogenous IQGAP1 would induce a shift in the distribution of Rap1 in the gradient. To test this hypothesis, we used MCF-7 cells that have stable integration of small interfer- Each IQGAP1 construct was also incubated with Ni 2ϩ -NTA resin alone (Ni 2ϩ ) as control. Complexes were isolated with Ni 2ϩ -NTA-agarose beads and resolved by SDS-PAGE. Gels were dried and processed by autoradiography (Pulldown). In addition, an aliquot of [ 35 S]methionine-labeled TNT product (equivalent to 10% of the amount that was subjected to pulldown) was resolved by SDS-PAGE, dried, and processed by autoradiography (Input). D, 10% of the total amount of [ 35 S]methionine-labeled IQGAP1 constructs used in the pulldown assay depicted in B were resolved by SDS-PAGE, and autoradiography was performed on the dried gel. Data are representative of three independent experiments. ing RNA specifically targeted to IQGAP1 (23). Termed MCF-siIQ8, these cells have 80% reduction in IQGAP1 levels (23). When lysates from MCF-siIQ8 cells were resolved by gel filtration chromatography, the elution of Rap1 was changed (Fig.  5A). Virtually no Rap1 eluted in fractions 24 -30, which corresponds to ϳ150 -300 kDa. Moreover, the amount of Rap1 that eluted in fractions 16 -22 was substantially reduced, with a concomitant increase in the amount of Rap1 detected at lower molecular masses. Interestingly, the elution of IQGAP1 in MCF-siIQ8 cells was different from that in MCF-7 cells (Fig. 5A). The IQGAP1 shifted to a lower molecular mass, perhaps due to attenuated oligomerization when the IQGAP1 concentration is reduced. Taken together, the gel filtration data suggest that Rap1 associates with IQGAP1 in MCF-7 cells.
Rap1 Co-localizes with IQGAP1 at the Cell Periphery-We next examined the localization of Rap1 and IQGAP1 in MCF-7 cells transfected with GFP-Rap1. In agreement with previous reports (17,30), endogenous IQGAP1 is diffusely distributed throughout the cytoplasm, with accumulation at cellcell contacts in MCF-7 cells (Fig.  5B). The distribution of transfected Rap1 is similar. GFP-tagged Rap1 is found throughout the cell with some accumulation at the plasma membrane (Fig. 5B). Analysis with specific co-localization software revealed that IQGAP1 and Rap1 co-localize exclusively at the cell periphery; essentially no co-localization is detected in the cytoplasm (Fig. 5B). We were unable to obtain specific immunostaining with commercially available antibodies to Rap1, precluding analysis of the endogenous protein by immunocytochemistry. Similar limitations have been reported by other investigators (34). The GFP tag does not affect Rap1 activity (35).
Based on our analyses in this study, one would anticipate that IQGAP1⌬IQ, which does not bind Rap1, will exhibit less co-localization with Rap1 than that seen with wild type IQGAP1. Moreover, Rap1 should have greater co-localization with IQGAP1 IQ3,4R than with wild type IQGAP1. To test these hypotheses, we transfected GFP-Rap1A and RFP-tagged versions of the pertinent IQGAP1 constructs. As seen with endogenous IQGAP1, transfected wild type IQGAP1 co-localizes with Rap1A predominantly at the cell periphery (Fig. 5C). As anticipated, Rap1A has weaker co-localization with IQGAP1⌬IQ than with wild type IQGAP1. Note that in cells transfected with Rap1A and IQGAP1⌬IQ, the two proteins co-localize in a perinuclear region (Fig. 5C). Also consistent with our prediction, Rap1A exhibits enhanced co-localization with  4R). After 48 h cells were fixed and processed for immunocytochemistry. Merge represents a composite of the two channels, with yellow indicating co-localization of IQGAP1 and Rap1. Colocalization of GFP-Rap1 (green) and each RFP-IQGAP1 construct (red) was analyzed with Zeiss LSM software on the confocal microscope. Representative images are depicted. Scale bar, 10 m. IQGAP1 IQ3,4R, which is mainly observed at the cell periphery, particularly in areas of ruffling (Fig. 5C).
Transfection of Rap1A Alters the Morphology of Cells with Reduced IQGAP1-The effects of transfecting GFP-Rap1A on the morphology of MCF-7 cells expressing different IQGAP1 levels was compared. When expressed in MCF-siIQ8 cells, GFP-Rap1A was diffusely distributed (Fig. 6). The cell morphology was altered with the appearance of multiple, elongated, actin-rich processes at the cell periphery (Fig. 6B). It is not possible to discern whether these result from membrane retraction or are extensions from the cell margin. Nevertheless, these structures were seen more frequently when Rap1A was transfected into MCF-siIQ8 cells than into MCF-7 cells. Analysis of Ͼ200 cells transfected with GFP-Rap1A revealed that elongated processes were present in 73 and 16% of MCF-siIQ8 and MCF-7 cells, respectively. In addition, there seemed to be less accumulation of GFP-Rap1A at MCF-siIQ8 cell-cell junctions than at junctions of MCF-7 cells (Fig. 6B).

Effects of IQGAP1 on Activation of Rap1 by Extracellular
Matrix-Rap1 is activated during cell adhesion (36,37). We used this knowledge to investigate whether interaction with IQGAP1 alters Rap1 function. MCF-7 cells that stably overexpress IQGAP1 were used. Termed MCF/I, these cells contain 3-fold more IQGAP1 than MCF-7 cells (Fig. 7, A and C) (18,23). Probing the blots for tubulin verifies equal loading. MCF/I and MCF-7 cells were seeded on plates coated with fibronectin or collagen I, and adherent cells were harvested at different time intervals. Adhesion to fibronectin increases the amount of the active, GTP-bound form of Rap1 in MCF-7 cells in a time-dependent manner, reaching 13.7-fold at 90 min (Fig. 7, A and B). Total Rap1 is not altered. Overexpression of IQGAP1 reduces the amount of GTP-Rap1 in non-adherent cells (Fig. 7, A and  B). The effect of increasing IQGAP1 on activation of Rap1 by adhesion is even more dramatic. Rap1 activation is markedly blunted in MCF/I cells in response to adhesion. Maximum levels of GTP-Rap1 in MCF/I cells are 14.3% of these in MCF-7 cells (Fig. 7, A and B). The data strongly suggest that IQGAP1 attenuates extracellular matrix-induced Rap1 activation.
To validate this hypothesis, we examined the effect of IQGAP1 on activation of Rap1 by cell adhesion to collagen I. Adhesion of MCF-7 cells to collagen I increases the amount of GTP-Rap1, although the magnitude of stimulation is substantially less than that seen with fibronectin (Fig. 7, C and D). Analogous to the findings with fibronectin, overexpression of IQGAP1 attenuates activation of Rap1 by adhesion to collagen I (Fig. 7, C and D). Although the variation among experiments with collagen I is large, the patterns are virtually identical to those seen with fibronectin. Collectively these data reveal that overexpression of IQGAP1 reduces activation of Rap1 by cell adhesion.
To confirm that the observed effect is mediated by Rap1 binding to IQGAP1, we repeated the analysis with cells transfected with IQGAP1⌬IQ, which does not interact with Rap1 (see Fig. 3). In contrast to the results obtained with wild type IQGAP1, IQGAP1⌬IQ neither reduces the amount of GTP-Rap1 in cells in suspension nor impairs the increase in GTP-Rap1 on cell adhesion (Fig. 8, A and B).
Cdc42 activation is different to that of Rap1. Adhesion of MCF-7 cells to fibronectin had little effect on Cdc42 activation (Fig. 7, A and B). A slight increase in active Cdc42 was evident only at 90 min. This is most likely due to the development of E-cadherin-mediated adherens junctions, which are known to augment GTP-Cdc42 (31). As we documented previously (23,30), overexpression of IQGAP1 significantly increases the amount of GTP-Cdc42 in cells (Fig. 7, A and B). Adhesion to fibronectin does not substantially alter active Cdc42 in MCF/I cells (Fig. 7, A and B).
Blockade of calmodulin has been reported to affect the GTP loading of Rap1 (38). It is conceptually possible that when overexpressed, IQGAP1 could perturb calmodulin signaling pathways by acting as a calmodulin "sink." To test this hypothesis, we used IQGAP1 IQ3,4R which does not bind calmodulin. MCF-7 cells were transiently transfected with IQGAP1 IQ3,4R, and the response of Rap1 to adhesion to fibronectin was examined. Analogous to wild type IQGAP1, IQGAP1 IQ3,4R substantially reduced Rap1 activation in response to adhesion (Fig.   FIGURE 6. Expression of GPR-Rap1A in MCF-7 and MCF-siIQ8 cells. MCF-7 or MCF-siIQ8 (siIQ8) cells were transiently transfected with GFP-Rap1A, and 48 h later were processed for immunocytochemistry. A, cells were incubated with anti-IQGAP1 polyclonal antibody followed by Alexa-Fluor 543-tagged anti-rabbit secondary antibody (red). Merge represents a composite of the two channels, with yellow indicating co-localization. B, cells were incubated with Alexa-Fluor 543-conjugated phalloidin to identify actin (red). Two different images are depicted for MCF-siIQ8 cells. Data are representative of images obtained from two experiments. Scale bar, 10 m. C and D). These data reveal that the inhibitory effect of IQGAP1 on activation of Rap1 is not mediated via calmodulin.

8,
IQGAP1 Regulates cAMP-induced Activation of Rap1-To determine whether IQGAP1 modulates activation of Rap1 by another pathway we examined Epac. Epac is a cAMP target and a Rap1-specific GEF (39). Therefore, we incubated cells with 8CPT-2Me-cAMP, which specifically activates Epac without activating protein kinase A (40). As anticipated, the cAMP compound induces an increase in active Rap1 in MCF-7 cells. GTP-Rap1 was ϳ2-fold higher at 30 min than in unstimulated MCF-7 cells (Fig. 9, A and B). By contrast, 8CPT-2Me-cAMP is unable to increase GTP-Rap1 in MCF/I cells. These data demonstrate that IQGAP1 overexpression modulates activation of Rap1 by cAMP.
The involvement of IQGAP1 in cAMP-mediated activation of Rap1 was also analyzed by a second strategy. If IQGAP1 is a negative regulator of Rap1, one would predict that reduction of IQGAP1 levels would enhance the amount of GTP-Rap1. We tested this hypothesis with MEF cells isolated from IQGAP1-null mice. These cells have no detectable IQGAP1 protein (Fig. 9C). The amount of GTP-bound Rap1 in IQGAP1 Ϫ/Ϫ MEFs is 2.3-fold greater than that in control MEFs (Fig. 9, C and D). Moreover, the ability of cAMP to activate Rap1 is significantly enhanced in the cells lacking IQGAP1 when compared with control MEFs.
IQGAP1 is known to interact with members of the Rho GTPase family, specifically Cdc42, Rac1, and TC10 (6,14,23,24). In contrast, IQGAP1 does not bind RhoA or H-Ras (14), nor have interactions with any other members of the Ras superfamily been previously documented (immunoaffinity purification of proteins cross-linked to constitutively active M-Ras identified 18 proteins including Rap1 and IQGAP1, but no direct binding of IQGAP1 to M-Ras or Rap1 was documented in that study (41)). The interactions of IQGAP1 with Cdc42 and Rac1 have been well characterized by several groups. Direct binding of Cdc42 and Rac1 to IQGAP1 is dependent on GTP (13,14,42). IQGAP1 does not function as a GAP; in fact it maintains Cdc42 and Rac1 in the active, GTP-bound form (10,14). This results in an increase in the GTP-bound forms of the small GTPases in cells that overexpress IQGAP1 (30). Importantly, the interaction has functional significance. Overexpression of IQGAP1 promotes cell motility and invasion at least in part via its association with Cdc42 and Rac1 (23). In this study we document for the first time a direct interaction between IQGAP1 and a GTPase that is not a member of the Rho subfamily.  JULY 13, 2007 • VOLUME 282 • NUMBER 28

JOURNAL OF BIOLOGICAL CHEMISTRY 20759
The interaction of Rap1 with IQGAP1 differs in several respects to the interaction of Cdc42 and Rac1 with IQGAP1. For example, although more GTP-Rap1 binds to IQGAP1, GDP-Rap1 also associates with IQGAP1. Moreover, binding to IQGAP1 does not appear to stabilize the GTP-bound form of Rap1, as overexpression reduces the amount of GTP-Rap1 in cell lysates.
A second novel aspect of the interaction of Rap1 with IQGAP1 is the effect of calmodulin. Calmodulin attenuates the interaction of IQGAP1 with all of its binding partners examined to date, namely Cdc42 (13), actin (11), ␤-catenin (18), and E-cadherin (17). The interaction of calmodulin with the four IQ motifs of IQGAP1 is complex (28) and augmented by Ca 2ϩ (11,13,28). In this context Ca 2ϩ is necessary for calmodulin to inhibit IQGAP1 binding to other proteins (11,13,28). Although the mechanism by which calmodulin produces this effect is not known, we have proposed that in the presence of Ca 2ϩ , calmodulin alters the tertiary conformation of IQGAP1, thereby reducing its ability to bind other proteins (6,13,15). It was, thus, surprising that calmodulin abrogates Rap1 binding to IQGAP1 both in the absence and presence of Ca 2ϩ . This observation implies that the molecular mechanism is different. Some insight was obtained when the site of Rap1 binding on IQGAP1 was identified. In contrast to Cdc42 and Rac1, which bind to a region that includes the GRD in the C-terminal half of IQGAP1 (10,14,43), Rap1 binds to the IQ domain in the N-terminal half of IQGAP1. Because calmodulin also binds to this region (28), it is likely that calmodulin directly competes with Rap1 by sterically hindering access of Rap1 to the IQ region. Thus, apocalmodulin, which binds only to the third and fourth IQ motifs (28), may restrict access of Rap1 without altering the conformation of IQGAP1. Additional support for this mechanism was obtained by examining the mutant construct IQGAP1 IQ3,4R. This construct contains replacement of selective hydrophobic residues in the third and fourth IQ motifs, which specifically abrogates binding to apocalmodulin (28). Rap1 binds more efficiently to IQGAP1 IQ3,4R in cell lysates than to wild type IQGAP1. It is noteworthy that Rap1 does not interfere with calmodulin binding to IQGAP1, possibly because the affinity of IQGAP1 for Rap1 is lower than its affinity for calmodulin. Finally, the different binding sites for the GTPases explain why Cdc42 did not attenuate the interaction between Rap1 and IQGAP1. Based on these in vitro data, it seems likely that IQGAP1 may bind simultaneously to Rap1 and Cdc42. Collectively, our findings suggest that the association of calmodulin with IQGAP1 is a critical regulatory factor in the interaction of Rap1 with IQGAP1.
Rap1 binds to the IQ region of IQGAP1. This is substantially different to other binding domains identified for Ras proteins. The IQ motif contains 20 -25 amino acids with a core consen- . Cells were plated on collagen I-coated culture dishes for the times indicated. Adherent cells were harvested with lysis buffer and analyzed by Western blotting. Equal amounts of protein lysate were also used in a GST-RalGDS (for active Rap1) pulldown followed by Western blotting with anti-Rap1 antibody. Protein lysates were also blotted for Myc (all IQGAP1 constructs are Myc-tagged), IQGAP1, and ␤-tubulin (as loading control). B, the relative amount of GTP-Rap1 in each sample was quantified by densitometry and corrected for total Rap1 in the corresponding cell lysate. Data, expressed relative to MCF-7 cells transfected with vector at time 0, are the means Ϯ S.D. of two independent experiments. C, MCF-7 cells were transiently transfected with empty vector (V), wild type IQGAP1 (WT), or IQGAP1 IQ3,4R (IQ3,4R). Cells were plated on culture dishes coated with fibronectin. After 90 min, cells were processed for Rap1 as described for A. D, the relative amount of GTP-Rap1 in each sample was quantified by densitometry and corrected for total Rap1 in the corresponding lysate. Data, expressed relative to MCF-7 cells transfected with vector at time 0, are representative of two independent experiments. FIGURE 9. IQGAP1 alters activation of Rap1 by cAMP. A, equal numbers of MCF-7 and MCF/I cells were stimulated with 100 M 8CPT-2Me-cAMP for the times indicated. GTP-bound Rap1 was detected by pulldown with GST-RalGDS as described under "Experimental Procedures." B, the amount of active Rap1 was quantified by densitometry and corrected for the total Rap1 in the corresponding lysate. Data are expressed relative to the amount of active Rap1 in MCF-7 at time 0. A representative experiment of three independent determinations is shown. C, equal numbers of MEF cells derived from normal mice (ϩ/ϩ) and from IQGAP1-null mice (Ϫ/Ϫ) were incubated with 8CTP-2Me-cAMP for the times indicated. GTP-bound Rap1 was detected by pulldown with GST-RalGDS as described under "Experimental Procedures." Equal amounts of protein lysate were also resolved by SDS-PAGE, and blots were probed for Rap1 (Total Rap1) and IQGAP1. D, the amount of active Rap1 was quantified by densitometry and corrected for the total Rap1 in the corresponding lysate. Data, expressed relative to the amount of active Rap1 in IQGAP1 ϩ/ϩ MEFs at time 0, are the means Ϯ S.E. of three independent experiments. The absence of error bars indicates that the S.E. is smaller than the size of the symbol. *, p Ͻ 0.05; **, p Ͻ 0.001.
sus IQXXXRGXXXR (where X is any amino acid) (44). A recent search identified IQ domains in 2479 proteins in the SMART non-redundant data base. Calmodulin, myosin light chains, and a few other Ca 2ϩ -binding proteins associate with IQ motifs. The crystal structures of IQ motifs of myosin V bound to calmodulin (45) and to myosin light chain 1 (46) have been solved recently. There are no published structures of the IQ domains of IQGAP1. It is not known how Rap1 associates with the IQ motifs of IQGAP1. Differences in the amino acid composition and size between the IQ domains of IQGAP1 and myosin V preclude extrapolation of the myosin V data to IQGAP1. Therefore, solving the structure of Rap1 bound to the IQ motifs of IQGAP1 is necessary to provide insight into the nature of this unusual interaction.
We (present study) and others (34,35) observed that Rap1 is found in the cytoplasm, particularly in the perinuclear area and at the plasma membrane where it is enhanced in ruffles. IQGAP1 has a similar distribution, with enhanced localization in the perinuclear area and at the plasma membrane (17,23,47). Notwithstanding the localization of both IQGAP1 and Rap1 in the cytoplasm, we observed co-localization exclusively at the cell periphery. This limited and specific co-localization suggests that functional interactions between IQGAP1 and Rap1 are restricted to the cell periphery. The reason for this is unknown. Nevertheless, this limited area of interaction may contribute to our inability to co-immunoprecipitate the endogenous proteins from cell lysates. Based on the region of colocalization, we speculate that IQGAP1 may serve to integrate Rap1 with the cytoskeleton at sites of cell contact both with other cells or potentially with the substratum. Support for the former hypothesis is that both IQGAP1 (17) and Rap1 (35) are known to participate in adherens junctions. Rap1 was shown to be recruited to matured epithelial cell-cell contact sites (35). In that study the authors raised the question as to which molecules are involved in anchoring Rap1 to cell-cell contacts (35). Based on our data, it is tempting to speculate that IQGAP1 binds GDP-Rap1 at adherens junctions, perhaps participating in anchoring the Rap1. A functional interaction between Rap1 and integrins is well recognized (5). However, an interaction of IQGAP1 with integrin is less well documented and is controversial (6). Further studies are necessary to determine whether the IQGAP1-Rap1 interaction contributes to the regulation of adherens junction formation and/or to the attachment of cells to the substratum.
Binding of Rap1 to IQGAP1 has functional sequelae as IQGAP1 modulates activation of Rap1. We observed that overexpression of IQGAP1 reduces the amount of active Rap1 in cells in suspension and substantially impairs the activation of Rap1 on cell adhesion. Consistent with these findings, a mutant IQGAP1 that lacks Rap1 binding did not attenuate activation of Rap1. Similarly, overexpression of IQGAP1 reduces the ability of cAMP to increase active Rap1. Moreover, cells lacking IQGAP1 have increased amounts of GTP-bound Rap1 and display an enhanced response to activation by cAMP. Collectively these data imply that IQGAP1 acts as a negative regulator of Rap1. The molecular mechanism underlying these observations is unknown. One possibility is that binding to IQGAP1 attenuates the interaction of Rap1 with GEFs, impairing the exchange of GDP for GTP. This could occur by altering Rap1 conformation or by changing the subcellular localization of Rap1 to microdomains lacking the GEFs. Another possibility is that IQGAP1 may facilitate the interaction of Rap1 with GAPs, resulting in enhanced GTP hydrolysis. These postulates are not mutually exclusive, and more than one may be operative. Regardless of the mechanism, the result is that IQGAP1 reduces GTP-Rap1, an effect opposite that produced on Rac1 and Cdc42, which are maintained in the GTP-bound form when bound to IQGAP1 (10,14). These differences most likely are the result of different modes of interaction of the GTPases with IQGAP1. Cdc42 and Rac1 bind to the GTPase regulatory domain (GRD) of IQGAP1, whereas Rap1 binds to the IQ domain.
In summary, our data document for the first time a direct interaction between a member of the Ras subfamily and IQGAP1. These findings extend the repertoire of binding partners for both IQGAP1 and Ras GTPases. We propose that IQGAP1 is a target for Rap1. Based on its well documented interaction with the cytoskeleton, we hypothesize that IQGAP1 integrates Rap1 with actin dynamics and cell-cell attachment. Finally, because calmodulin abrogates IQGAP1 binding to Rap1, it is tempting to speculate that IQGAP1 may serve as a link between the calmodulin and Rap1 signaling pathways.