Polarized secretion is a tightly regulated event generated by conserved, asymmetrically localized multiprotein complexes, and the mechanism(s) underlying its temporal and spatial regulation are only beginning to emerge. Although yeast Iqg1p has been identified as a positional marker linking polarity and exocytosis cues, studies on its mammalian counterpart, IQGAP1, have focused on its role in organizing cytoskeletal architecture, for which the underlying mechanism is unclear. Here, we report that IQGAP1 associates and co-localizes with the exocyst-septin complex, and influences the localization of the exocyst and the organization of septin. We further show that activation of CDC42 GTPase abolishes this association and inhibits secretion in pancreatic β-cells. Whereas the N-terminus of IQGAP1 binds the exocyst-septin complex, enhances secretion and abrogates the inhibition caused by CDC42 or the depletion of IQGAP1, the C-terminus, which binds CDC42, inhibits secretion. Pulse-chase experiments indicate that IQGAP1 influences protein-synthesis rates, thus regulating exocytosis. We propose and discuss a model in which IQGAP1 serves as a conformational switch to regulate exocytosis.
Polarized cell growth is essential for specialized cellular functions such as migration as well as for cell division and differentiation, and is generated by a core mechanism involving evolutionarily conserved multi-protein complexes (Johnson and Wodarz, 2003; Nelson, 2003). The generation of polarized cell growth is driven by directed exocytosis and requires the action of asymmetrically localized positional cues and localized activation of G-proteins followed by the recruitment of downstream effectors to a specific site, leading to the reorganization of the actin and microtubule cytoskeletons and of the secretory pathway (reviewed in Osman and Cerione, 2005). However, the molecular determinants of its spatial and temporal regulation are not well understood.
Yeast IQGAP1, Iqg1p, was identified as a target for Cdc42p GTPase (Osman and Cerione, 1998) and was found to bind Bud4p, the septin Cdc12p and the exocyst Sec3p, which serve as positional cues for cell polarity events such as axial bud-site selection and cytokinesis (Osman et al., 2002). Although there is no known mammalian homolog for Bud4p, a counterpart exists for each septin and exocyst subunit. The mammalian exocyst [Sec6-Sec8 (EXOC3-EXOC4) complex] has been co-purified and immunoprecipitated with septins (Hsu et al., 1996; Hsu et al., 1998; Hsu et al., 1999); like its yeast counterpart, it was found to influence polarized vesicle delivery (Hsu et al., 1999). Mammalian septins were individually identified and implicated in cytokinesis, exocytosis, vesicle-targeting and membrane dynamics (Spiliotis and Nelson, 2006; Beites et al., 1999; Trimble, 1999). Together, these findings argue that septins can cooperate with the exocyst in the regulation of polarized secretion, perhaps acting as positional markers, tethering or tracking filaments. One of the best-characterized mammalian septins is NEDD5, from here on referred to as SEPT2 (Spiliotis and Nelson, 2006; Kinoshita et al., 1997; Sakai et al., 2002; Vega and Hsu; 2003; Trimble, 1999). As is the case with all septins, its role in these processes remains unclear.
Whereas the yeast exocyst communicates with several members of the Rho subfamily of small GTPases, including Cdc42p (Lipschutz and Mostov, 2002; Novick and Guo, 2002), in mammals, only RAL (RALA) and TC10 (RHOQ) GTPases have been shown to bind the exocyst (Inoue et al., 2003), and no link has been reported with CDC42. Recently, genetic and biochemical evidence have shown that both yeast and mammalian exocysts communicate with the ER-translocon via an interaction with the Sec61β subunit (Lipschutz et al., 2003; Toikkanen et al., 2003), suggesting a role in protein synthesis/translocation events.
Mammalian IQGAP1 was also identified as a CDC42 target, acting through its C-terminal domain (Brill et al., 1996; Kinoshita et al., 1997; McCallum et al., 1996; Kuroda et al., 1996), and the domain structure of yeast and mammalian IQGAPs is conserved (Epp and Chant, 1997; Lippincott and Li, 1998; Osman and Cerione, 1998), implying a conserved function. IQGAP1 is a member of a three-isoform family of proteins that also includes IQGAP2 (Yamashiro et al., 2003) and IQGAP3 (Wang et al., 2007). IQGAP1, the most studied, binds calmodulin, cross-links actin filaments, integrates signaling networks (reviewed in Mateer et al., 2003; Brown and Sacks, 2006), and regulates cell-cell contacts (Fukata et al., 1999; Fukata et al., 2001) and the capture of microtubule plus-ends via association with CLIP-170 (Fukata et al., 2002). The current paradigm is that IQGAP1 regulates actin assembly in cooperation with the ARP2/3 complex and the Rho GTPases in different cell types to regulate cell outgrowth and migration (Watanabe et al., 2004; Bensenor et al., 2007; Le Clainche et al., 2007; Wang et al., 2007). This role, however, does not contradict with an IQGAP1 essential function in secretion, because the mammalian exocyst subunit EXO70 (EXOC7) has also been reported to interact with the ARP2/3 complex and modulate actin-based membrane dynamics (Zuo et al., 2006), similar to IQGAP1. Thus, in this work we present a novel role for IQGAP1 in the regulation of exocytosis.
Here, we investigated whether mammalian IQGAP1 associates with the exocyst-septin complex and influences secretion in a CDC42-regulated fashion. We used pancreatic β-cell lines for two reasons. First, we found that IQGAP1, the exocyst and septins are abundant proteins in these cells, offering insulin secretion as a physiologically relevant functional assay for exocytosis. Second, mastoparan, a tetradecapeptide from wasp venom, was reported to enhance insulin secretion in βHC-9 cell lines that overexpress wild-type CDC42 by stimulating its exchange activity (Daniel et al., 2002). Mastoparan is known to stimulate exocytosis independent of Ca2+ by activating G-proteins in a number of mammalian cell types, including β-cells (reviewed in Kowluru, 2003). Thus, it seemed plausible that CDC42 also would enhance an IQGAP1–exocyst-mediated secretion.
Surprisingly, we found that overexpression or activation of CDC42 by mastoparan, dominant-active mutations or IQGAP1 expression disrupted the endogenous association of IQGAP1 with the exocyst-septin complex and blocked secretion. This effect appeared to be mediated by the C-terminus of IQGAP1, which binds CDC42 and inhibits secretion. By contrast, the N-terminus of IQGAP1 binds to the exocyst-septin complex, enhances secretion and abrogates the inhibition caused by CDC42 or the depletion of IQGAP1, perhaps via the involvement of the N-terminus in protein synthesis, as indicated by pulse-chase experiments. These data raise a possibility that IQGAP1 serves as a regulator of secretion by acting as a conformational switch.
Association of IQGAP1 with the exocyst-septin complex is regulated by CDC42
Because the domain structure of yeast Iqg1p and mammalian IQGAP1 is conserved (Fig. 1A), the ability of IQGAP1 to interact with the exocyst-septin complex was examined. Because mammalian cells contain two ∼60% identical IQGAP isoforms in addition to IQGAP3, the interaction with both IQGAP1 and IQGAP2 was examined in pancreatic βTC-6 cells using antibodies specific for each protein. IQGAP1 but not IQGAP2 (or IQGAP3, not shown) co-precipitated with the exocyst-septin complex (Fig. 1B).
To examine the effects of CDC42 and its activation by mastoparan on the association of IQGAP1 with the exocyst-septin complex, βTC-6 cells stably expressing the vector control or wild-type CDC42 (CDC42WT) that were treated either with mastoparan (Mp) or vehicle alone were used for immunoprecipitation (IP). Unexpectedly, IQGAP1 formed complexes only in control cells lacking CDC42WT that were not treated with mastoparan (Fig. 1C, lane 3). CDC42 did not disrupt the exocyst-septin complex itself and only IQGAP1 was displaced (Fig. 1C, lanes 1, 2 and 4). In addition, antibodies against IQGAP1 or the V5 tag of its recombinant constructs were also able to co-immunoprecipitate endogenous EXO70, SEC8 and SEPT2 in untreated control β-cells and in six other mammalian cell lines (Fig. 3C; and supplementary material Fig. S1A,C). These data were identical in βHC-9 cells and indicate that IQGAP1 interacts with the exocyst-septin complex and that overexpression or activation of CDC42 by mastoparan disrupts this interaction.
Active alleles of CDC42 inhibit IQGAP1 association with the exocyst-septin complex
To ascertain that the disruption of the interaction between IQGAP1 and the exocyst-septin complex is specific to the activation of CDC42 and not due to other effects caused by mastoparan, we undertook a genetic approach using dominant CDC42 mutants in a transient-expression assay. We compared the effects of dominant alleles of CDC42 on IQGAP1-EXO70 co-immunoprecipitation in pancreatic β-cells that have not been treated with mastoparan (Fig. 2A,B). An HA-tagged pcDNA3 vector encoding the GTPase-defective, dominant-active mutant of CDC42 (Q61L; Fig. 2, lane 1), the dominant-negative (inactive) mutant (T17N; Fig. 2, lane 2), the wild-type (WT; Fig. 2, lane 3), the constitutively active mutant (F28L; Fig. 2, lane 4) or the empty vector for control (vector; Fig. 2, lane 5) were transiently expressed to similar levels (HA-CDC42, Fig. 2A, lower panel). When endogenous EXO70 was immunoprecipitated, more endogenous IQGAP1 co-precipitated in cells expressing the vector control or the dominant-negative CDC42 allele (T17N) but little in cells expressing WT or the dominant-active alleles (Q61L and F28L). This result is not due to differences in IQGAP1 or EXO70 expression, because equal amounts of IQGAP1 (Fig. 2A, bottom of the upper panel) and EXO70 (Fig. 2A, top panel) were found in the lysate used for IP in cells expressing each of the CDC42 alleles.
Three such blots were quantified for the intensities of the IQGAP1 bands co-precipitated with EXO70 in the presence of these alleles, and their averages are presented as relative amounts (Fig. 2B). These data were identical in both βHC-9 and βTC-6 cell lines, and indicate that dominant-active CDC42 alleles more effectively diminish the endogenous association between IQGAP1 and EXO70. Moreover, expression of the WT allele (CDC42-GDP or nucleotide-depleted, CDC42-ND) also blocked the interaction (see Discussion). Therefore, by two independent lines of evidence – pharmacologically using mastoparan and genetically using dominant mutant alleles – our data indicate that active CDC42 disrupts the interaction between IQGAP1 and the exocyst-septin complex; protein subunits implicated in polarized exocytosis.
A physical interaction between CDC42 and IQGAP1 is necessary for disrupting the IQGAP1-exocyst complex
To examine whether binding of CDC42 and IQGAP1 is necessary for the dissociation of IQGAP1 from the exocyst-septin complex, we constructed CDC42 and IQGAP1 mutants that are unable to bind to each other. The double mutant HA-CDC42-F28LC37A and the deletion mutant V5-IQGAP1-FΔMK, in which the 24 amino acids M1054-K1077 of IQGAP1, required for CDC42 binding, are deleted (Mataraza et al., 2003), were transiently expressed in βTC-6 cells, verified for binding (Fig. 2D) and co-precipitation of IQGAP1 with EXO70 was performed (Fig. 2C) as described above. The C37A mutation specifically reduced IQGAP1 binding to CDC42-F28L (Fig. 2D, upper panels) without altering the binding of other effectors (Lin et al., 1999) and the FΔMK deletion mutant significantly reduced binding of CDC42 to IQGAP1 (Fig. 2D, lower panel). Co-precipitation of IQGAP1 with EXO70 was enhanced significantly by these mutations compared with their cognate parental proteins (Fig. 2C), indicating that binding of CDC42 to IQGAP1 is necessary for its negative effects on IQGAP1 interaction with the exocyst-septin complex.
The region of IQGAP1 mediating the association with the exocyst and septin
Next, we investigated whether CDC42 displaces the exocyst-septin complex by competing for binding. Thus, to identify the domain(s) of IQGAP1 that mediate the interaction with EXO70 and SEPT2, antibodies against the V5 tag of the IQGAP1 constructs shown in Fig. 3A (supplementary material Fig. S1Ci) were used to co-precipitate two different subunits of the exocyst, SEC8 and EXO70, and SEPT2 (Fig. 3B, right panels). Exogenous full-length IQGAP1-F (F1; Fig. 3B, first lane), IQGAP1-N (N1; Fig. 3, second lane) and the IR-WW (Fig. 3B, third lane), but not IQGAP1-C (C2; Fig. 3B, fourth lane) or the empty vector (Fig. 3B, fifth lane), could associate with the exocyst in vivo. Conversely, antibodies for IQGAP1 co-precipitated the two-exocyst subunits and SEPT2 from different mammalian cell lines (Fig. 3B, left panels, supplementary material Fig. S1Ai), confirming the functionality of the recombinant IQGAP1 proteins.
To verify that IQGAP1-N mediates the interaction with EXO70-SEPT2, bacterially expressed GST-vector for control, GST-SEPT2 (Fig. 3C, upper panel) and GST-EXO70 (Fig. 3C, right, lower panel) were used as affinity reagents to pull down the recombinant V5-IQGAP1 domains expressed in COS7 cells (Fig. 3C, left). GST alone (Fig. 3C, represented in lane 1 of both panels with N1 cell lysate) did not pull down any domain, whereas GST-SEPT2 and GST-EXO70 pulled down V5-N1 (Fig. 3C, lane 2) but not V5-C2 (Fig. 3C, lane 4). Contrary to the in vivo data, an association between the V5-IR-WW domain and GST-EXO70 was not detected, and GST-SEPT2 pulled down a weak band (Fig. 3C, upper panel, lane 3), suggesting that the entire N-terminus is required for efficient in vitro interaction. The role of IQGAP1-N was confirmed further by conversely pulling down GST-SEPT2 or GST-EXO70 with 6×His-IQGAP1-N1 immobilized on cobalt resins (Fig. 3D). These results confirm that IQGAP1-N plays a role in mediating the association with the exocyst-septin subunits. Consistent with this finding, exogenous V5-N1 and V5-IR-WW, but not other IQGAP1 domains, co-immunoprecipitated with EXO70 in βTC-6-CDC42WT stable cell lines (supplementary material Fig. S1B), abrogating the inhibitory effects of CDC42 on the endogenous binding. Furthermore, these data suggest that the dissociation of the IQGAP1-exocyst-septin complex is unlikely because of a binding-site competition with CDC42, which binds to the C-terminus of IQGAP1. Rather, a different mechanism is involved (see Fig. 9).
Co-localization of endogenous IQGAP1 with EXO70 and the effect of IQGAP1 depletion
As an additional measure for IQGAP1 association with EXO70, their localization was compared in fixed double-stained HeLa cells by confocal microscopy (Fig. 4). As expected for interacting proteins, in >90% (n=100) of the examined cells, IQGAP1 and EXO70 concentrated together at the leading edge of the cells, consistent with previously reported localization of the individual proteins (Watanabe et al., 2004; Bensonor et al., 2007; Zuo et al., 2006). In addition, they co-localized in a perinuclear meshwork and cytoplasmic tubular structures, presumably the ER (Fig. 4, arrow). This overlapping localization is consistent with their biochemical association.
We further examined the effect of IQGAP1 on EXO70 localization by RNA interference (RNAi). Control RNAi did not affect IQGAP1 expression in HeLa cells (Fig. 5A, first lane), whereas IQGAP1-RNAi (Fig. 5A, second lane) diminished it by ∼90% without affecting the expression of IQGAP2 (nor IQGAP3, not shown), EXO70 or SEPT2, confirming the specificity of IQGAP1-RNAi. In >90% (for n=50) of IQGAP1-RNAi cells, EXO70 signal was faint, and not readily detectable at the membrane ruffles, requiring a four-times longer exposure to be visible (Fig. 5B). These data suggest that EXO70 might depend on IQGAP1 or factors organized by it such as actin or microtubules for proper localization. Alternatively, EXO70 polarized transport might be impaired, explaining the faint signal at the membrane ruffles.
Co-localization and effects of IQGAP1 expression on SEPT2 organization
Endogenous SEPT2 and IQGAP1 also co-localized to the plasma membrane both in HeLa and pancreatic β-cells (Fig. 6; supplementary material Fig. S2), in agreement with the reported SEPT2 localization to the plasma membrane in interphase and mitotic MDCK (Spiliotis and Nelson, 2006) and in PC12 cells (Beites et al., 1999).
Further, we measured the effects of IQGAP1 overexpression on EXO70 and SEPT2 localization. HeLa cells stably expressing equal but low levels (<10% of WT) of the V5-IQGAP1 domains did not show any measurable effect on EXO70 or SEPT2 localizations (not shown), and expression of IQGAP1-F or IQGAP1-C had no detectable effects on SEPT2 organization (not shown). However, cells expressing the IR-WW domain, which binds to the exocyst-septin complex, exhibited disorganized SEPT2 filaments, as characterized in 200 cells each from four independent experiments represented in Fig. 6A. In control cells, SEPT2 localized in a perinuclear position and prominently decorated the cell peripheries with a few visible SEPT2 filamentous structures in some cells (Fig. 6A, upper panels), whereas, in the IR-WW cells, SEPT2 filaments were elaborate, thick and randomly distributed across the cells or often clustered in the nuclear vicinity (Fig. 6A, lower panel), suggesting the involvement of IQGAP1 in septin-filament organization.
Next, we investigated the effect of IQGAP1 depletion on SEPT2 localization in 70 cells depleted of IQGAP1 by RNAi (Fig. 6B). Both at 100 nM (Fig. 6B, left) and 40 nM (Fig. 6B, right) concentrations of IQGAP1-RNAi, SEPT2 localization was not readily detectable at the plasma membrane, as was the case with EXO70. Therefore, IQGAP1 overexpression or depletion affects septin localization and filament organization.
In pancreatic β-cells, we observed a similar localization of IQGAP1 using ER and membrane markers (supplementary material Fig. S2). In single cells, IQGAP1 concentrated at a paranuclear position, but when cell-cell contacts were formed (islet-like cells), it concentrated in the cell peripheries. Although the reason behind this switch in localization is unclear, localization of the ER marker inositol 1,4,5-triphosphate receptor, type 3 (IP3R3, ITPR3) revealed an identical pattern to that of IQGAP1 in these cells (supplementary material Fig. S2A). IP3R3 is an ER resident protein shown to switch localization from the ER to the plasma membrane when MDCK cells polarize (Colosetti et al., 2003 ), presenting a possibility that pancreatic β-cells undergo polarization upon forming cell contacts and that IQGAP1 switches localization from the ER to the plasma membrane upon polarization of specific cell types. A similar pattern was also observed for EXO70 and SEPT2 in these cells (not shown).
To verify that these cell-cell contacts are the plasma membrane, and in particular that these proteins are cytosolic (supplementary material Fig. S2Aii), we used the t-SNARE syntaxin 1A as a membrane marker. IQGAP1 and syntaxin 1A overlapped in the cell-cell contacts in pancreatic β-cells (supplementary material Fig. S2B) and in HeLa cells (Fig. 8), indicating that IQGAP1 localizes at the plasma membranes of different cell types. Collectively, these data suggest that IQGAP1 co-localizes with and influences EXO70 localization and septin-filament organization, and that increasing or decreasing the dosage of IQGAP1 has cellular effects.
Differential effects of IQGAP1 domains on secretion
As a readout assay for functional association of IQGAP1-exocyst-septin, we measured insulin secretion in pancreatic βTC-6 cells, comparing basal and glucose-induced secretion in cells expressing equal levels of the V5-IQGAP1 domains (Fig. 7). The basal secretion levels were slightly enhanced by the expression of IQGAP1 constructs (Fig. 7A). However, expression of IQGAP1-C blocked the induced insulin exocytosis, whereas that of IQGAP1-F had no significant effect. By contrast, expression of IQGAP1-N or IR-WW enhanced exocytosis significantly (>50%). Similar results were obtained from βHC-9 cells using a different kit or using radiolabeled methods (not shown). Furthermore, IQGAP1-C appears to reduce the endogenous association between IQGAP1 and the EXO70-SEPT2 complex (supplementary material Fig. S1Cii,iii), mimicking the effect of activated CDC42. Thus, the domain of IQGAP1 that interacts with the exocyst-septin complex enhances exocytosis, whereas the domain that interacts with CDC42 acts as dominant-negative, inhibiting exocytosis.
To examine the role of endogenous IQGAP1, we measured the stimulated level of secreted insulin in βTC-6 cells treated with IQGAP1-RNAi (supplementary material Fig. S3). Depletion of IQGAP1 did not affect the expression levels of EXO70 or SEPT2 (Fig. 5A and supplementary material Fig. S3Ai), but decreased induced insulin secretion by ∼45%, which was rescued by co-transfection of an RNAi-refractory IR-WWR domain (supplementary material Fig. S3Aii,B). These data affirm the dominant-negative effects of IQGAP1-C and indicate further that IQGAP1-N serves as a dominant-positive (Fig. 7A). Although in the immune system delivery of secretory lysosomes requires the clearing of both actin and IQGAP1 from the target site of the plasma membrane (Stinchcombe et al., 2006), depletion of IQGAP1 in mast cells only mildly enhances agonist-stimulated histamine secretion (Psatha et al., 2007). Together, these data support the concept that the essential role of IQGAP1 in secretion is regulatory.
IQGAP1 influences protein synthesis
Our secretion results, however, do not differentiate between whether the enhancement of secretion is a result of increased release from cellular stores or due to an increase in protein synthesis. Therefore, we measured protein synthesis rates in the stable cell lines by pulse-chase experiments. The labeled basal levels of insulin in cells expressing IQGAP1-N or IQGAP1-C were higher compared with their vector-control cells (Fig. 7B, left panels), consistent with their basal secretion level (Fig. 7A), perhaps indicating a deregulation in the basal steady state. However, when these cells were chased for 20 minutes (not shown) or 1 hour and the labeled insulin levels measured by IP, IQGAP1-C cells retained higher levels of labeled proteins compared with their vector control and IQGAP1-N cells (Fig. 7B, right panels). The persistence of the labeled proteins in IQGAP1-C cells indicates a reduction in protein synthesis and exocytosis rates.
The disappearance of labeled proteins in IQGAP1-N cells indicates higher rates of both protein synthesis and exocytosis, thereby replacing the labeled with new proteins. To verify whether this was the case, we measured the total insulin levels in the labeled cells by immunoblotting, revealing that total insulin levels were similar in N1 and C2 cells, and higher than in their vector-control cells (Fig. 7B, lower panel). In addition, immunofluorescence of insulin in stable βTC-6 cells that were stimulated for 1 hour confirmed these immunoblot data, showing more insulin in IQGAP1-N and IQGAP1-C cells (Fig. 7C) caused by active protein synthesis in the former and defective synthesis/secretion rates in the latter. Taken together, these results indicate that both protein synthesis and exocytosis are impaired in IQGAP1-C cells because they accumulate labeled insulin and exhibit low secretion level. Thus, by three different measurement criteria – insulin-secretion assays, pulse-chase for protein synthesis rate, and immunofluorescence – our data support the conclusion that the C-terminus of IQGAP1 inhibits protein synthesis and exocytosis, whereas its N-terminus enhances both of them.
IQGAP1 localizes and associates with sites of protein synthesis and exocytosis
If IQGAP1 is involved in protein synthesis/translocation, then it would be expected to localize to the ER. The recent discovery that the vesicle-tethering exocyst has a conserved role in communicating with the ER-translocon complex (Lipschutz et al., 2003; Toikkanen et al., 2003) lends credence to such an idea. Furthermore, upon an ExPASy proteomic server (http://br.expasy.org/) search for motifs on IQGAP1, we identified an ER membrane-retention signal (KFYG) on the extreme C-terminus. Because IQGAP1 is a cytosolic protein and because an ER luminal signal (KDEL) is absent from IQGAP1, this membrane-retention signal might be involved in retrieval events at the ER membrane. Therefore, we examined whether the localization of IQGAP1 in the perinuclear meshwork observed in HeLa cells in Fig. 4 and Fig. 5B might be to the ER. This was tested by co-localization of IQGAP1 with two ER resident markers in a total of three cell types: with calnexin in HeLa cells (Fig. 8A), and with the IP3R3 in NIH3T3 cells (Fig. 8B) and β-cells (supplementary material Fig. S2A). In the three cell types, IQGAP1 overlapped with each marker in the perinuclear area, whereas it localized alone to the leading edges of NIH3T3 and HeLa cells (in ∼90% for n=200 of cells examined). However, we also observed non-overlapping spots of the proteins, perhaps reflecting specialized functions of each protein.
Thus, we investigated whether this ER localization indicates an association with the ER translocon, as is the case with its exocyst partner SEC8 subunit (Lipschutz et al., 2003). Co-immunoprecipitation of endogenous IQGAP1 with the ER translocon subunit Sec61β from HEK293NT cells (and MD-MB231 human breast cancer cell lines, not shown) showed that each antibody conversely precipitated the other protein (Fig. 8D, top panel). Thus, three lines of supporting evidence – an ER signal, localization with two ER markers and co-immunoprecipitation with the ER translocon – suggest that IQGAP1, like its exocyst partner, at least in part localizes to the ER and interacts with the translocation complex, providing additional support that it influences protein synthesis/translocation.
Furthermore, we examined whether IQGAP1 localization at the plasma membrane indicates a connection with membrane-fusion proteins in the secretory pathway. A likely candidate is the vesicle-fusion protein syntaxin 1A, which is a member of the t-SNARE complexes that are abundant in pancreatic β-cells. Syntaxin 1A has been shown to bind septins in order to regulate exocytosis (Beites et al., 1999), to bind the exocyst to complete cytokinesis (Gromley et al., 2005), and was recently shown to regulate docking and fusion of insulin granules in the first phase release (Ohara-Imaizumi et al., 2007). Our data show that IQGAP1 localized and co-precipitated with syntaxin 1A at polarized membrane regions in HeLa (Fig. 8C, Fig. 8D, bottom panel) and β-cells (supplementary material Fig. S2B). Future sub-cellular fractionation and/or immuno-electron-microscopy (immuno-EM) studies will be required to define exact points of overlap of these proteins. However, these findings provide additional support for an IQGAP1 role in exocytosis with the vesicle-tethering exocyst and vesicle-fusion SNAREs.
Interplay between CDC42 and IQGAP1 regulates secretion
Next, we examined whether CDC42 disruption of the association of IQGAP1 with the exocyst-septin complex consequently disrupts secretion. In such a case, we investigated whether expression of IQGAP-C, which binds CDC42, would abrogate the disruption by sequestering CDC42, thereby allowing endogenous binding of IQGAP1 and exocyst. Thus, we measured glucose-stimulated insulin secretion in pancreatic βTC-6 cells stably expressing CDC42WT, in which the association is disrupted, comparing it with those stably co-expressing the V5-IQGAP1 domains (Fig. 9A). First, our data affirmed a previous finding that stable expression of CDC42WT inhibits glucose-stimulated insulin secretion (Fig. 9A, last two columns) (Daniel et al., 2002) and further showed that expression of IQGAP1-F (F1) had no measurable effect on the inhibition (Fig. 9A, first columns). By contrast, expression of IQGAP1-N or IQGAP1-IR-WW abrogated the inhibition caused by CDC42WT, enhancing secretion significantly (Fig. 8A, second and fourth columns). Unexpectedly, expression of IQGAP1-C (C2), which binds CDC42, did not abrogate the secretion inhibition in CDC42WT stable cells (Fig. 8A, third columns). This finding is consistent with the result that exogenous N-terminus and IR-WW domain can bind EXO70 in CDC42WT cells (supplementary material Fig. S1B), whereas expression of IQGAP1-C depresses the level of endogenous bound IQGAP1-EXO70 (supplementary material Fig. S1C). Therefore, we investigated a possible mechanism that generates such an effect.
Several lines of evidence from yeast studies suggest that localized activation of Cdc42p occurs at membrane sites by positional cues, such as Iqg1p (reviewed in Osman and Cerione, 2005). Given our finding that activation of CDC42 disrupts the association of IQGAP1 with the exocyst-septin complex and inhibits secretion, we investigated whether expression of IQGAP1 activates endogenous CDC42, leading to such effects. Therefore, we measured the amounts of active CDC42 by using the CDC42/Rac interactive binding (CRIB) region of PAK, which specifically binds active CDC42 (GTP bound). Control β-cells that incorporated the vector alone (V) did not have measurable amounts of GTP-CDC42 (Fig. 9B, right panel), which also was the case with IQGAP1-N (N1) cells. However, expression of either IQGAP1-F (F1) or IQGAP1-C (C2) enhanced the level of GTP-CDC42. All cells had equal levels of endogenous total CDC42, as demonstrated in the lower panel in Fig. 9, and these results are not unique to β-cells, because they confirm similar findings from other cell types (Fukata et al., 2002; Mataraza et al., 2002; Grohmanova et al., 2004), affirming an apparently conserved function for IQGAP1 upstream of CDC42 (Osman et al., 2002). These data help explain why IQGAP1-C disrupts binding, secretion and protein synthesis, confirming that active CDC42 negatively regulates IQGAP1 function in secretion. They also indicate that IQGAP1 modulates CDC42 activity, in effect serving as a regulator of secretion. Thus, interplay between IQGAP1 and CDC42 plays a role in regulating exocytosis.
IQGAP1 links the exocyst-septin complex with CDC42 signaling to regulate protein synthesis and exocytosis
In this study, we presented evidence for a novel role of IQGAP1, a CDC42 target/effector, in regulating secretion by associating with and influencing the localization of EXO70 and the organization of SEPT2 (Figs 1, 2, 3, 4, 5, 6), two subunits of protein complexes implicated in cell polarization and directed exocytosis. The co-localization data agree with previously reported localization of the individual proteins in other cell types. IQGAP1 and the exocyst each localize to the basolateral membrane in gastric parietal cells and to cell-cell contacts in MDCK cells (Grindstaff et al., 1998; Katata et al., 2003; Zhou et al., 2003). By contrast, the isoform IQGAP2 localizes to the apical membranes (Yamashiro et al., 2003), explaining why it did not associate with the exocyst-septin complex. Similarly, septins localize to the plasma membrane, and with microtubule and actin cytoskeletons (Spiliotis and Nelson, 2006), and the effect of IQGAP1 on SEPT2 organization resembles that of the Borg3 protein (Joberty et al., 2001), another effector of CDC42. Borg proteins are absent from yeast and other metazoans but IQGAP1 is widely conserved; thus, these two CDC42 effectors might represent different `flavors' for fine-tuning septin functions/organization in different organisms and cell types, or for specificity in CDC42-divergent signaling pathways.
The data presented here suggest that IQGAP1 enhances protein synthesis and that it resides in the ER, associating with the translocation complex (Fig. 8), and at the plasma membrane, associating with vesicle-tethering exocyst and vesicle-fusion SNAREs (Figs 4, 8 and supplementary material Fig. S2A), similar to mammalian and yeast exocysts (Lipschutz et al., 2003; Toikkanen et al., 2003). These findings support the idea of involvement of these protein complexes in a positive-feedback loop for exocytosis. In mammals, the Sec61β subunit that associates with IQGAP1 and the exocyst binds non-translating ribosomes (Levy et al., 2001). By contrast, IQGAP1 and CDC42, but not other effectors such as WASP, were purified as binding-partners for the double-stranded RNA-binding polarity protein Staufen as part of RNA granules involved in the in vivo localization and translation of human mRNAs (Villace et al., 2004). Future investigation should reveal whether IQGAP1 tethers ribosomes to mRNAs, serving as a scaffolding positional marker.
Involvement of IQGAP1 in physiological secretion
Roles for the exocyst in insulin secretion and septins in glucose-stimulated growth-hormone release have been reported previously (Inoue et al., 2003; Beites et al., 1999), and the data presented here further indicate that IQGAP1 associates with these protein complexes to regulate their effects on secretion in cooperation with CDC42-GTPase (Figs 1 and 9). This agrees with the finding that stable expression of CDC42, while enhancing mastoparan-activated CDC42 insulin secretion, inhibited glucose-stimulated secretion (Fig. 9) (Daniel et al., 2002). Mastoparan is a toxin from wasp venom that stimulates insulin release independent of Ca2+. By contrast, glucose-stimulated insulin secretion occurs via a Ca2+-dependent pathway in response to nutrients (Kowluru, 2003), supporting the involvement of IQGAP1 in the Ca2+-dependent nutrient-stimulated pathway. Significantly, Ca2+ has been shown to dissociate IQGAP1 from CDC42 (Ho et al., 1999), lending support to our finding and to the model presented below that dissociation of CDC42 is necessary to allow the effects of IQGAP1 on secretion.
Interplay between CDC42 and IQGAP1 regulates the function(s) of IQGAP1
Our results indicate that IQGAP1 serves both as an upstream activator and downstream target for CDC42, in agreement with the previous finding that its yeast counterpart, Iqg1p, serves as an axial marker upstream of Cdc42p (Osman et al., 2002). Therefore, either a sequestration or an activation model could account for the negative effects of CDC42 and IQGAP1-C on secretion. In the first model, CDC42 binds and sequesters IQGAP1 from the exocyst-septin complex, thereby inhibiting secretion. In this scenario, ectopic expression of IQGAP1-C, which binds CDC42, would be expected to block the inhibition caused by CDC42. However, expression of IQGAP1-C slightly exacerbated that inhibition (Fig. 9) and depressed endogenous IQGAP1-EXO70 association (supplementary material Fig. S1C). Therefore, the sequestration model cannot alone explain these effects, but they are consistent with the activation model discussed below.
Because CDC42 disrupts the association of IQGAP1 with the exocyst-septin complex and expression of IQGAP1-F or IQGAP1-C increases the cellular level of active CDC42 (Fig. 9B) and decreases the endogenous IQGAP1-EXO70 association, these data provide a probable explanation as to why IQGAP1-F and IQGAP-C blocked secretion (Fig. 9A). IQGAP1 also binds nucleotide-depleted CDC42 (ND-CDC42) and GDP-CDC42 (Grohmanova et al., 2004), confirming its GEF-like activity on CDC42 and providing an explanation as to why wild-type CDC42 blocked both interaction and secretion (Figs 1 and 8) similar to its dominant active mutants, indicating that increasing the level of CDC42 enhances the pool of its active form, leading to the observed inhibition.
By contrast, IQGAP1-N, which binds the exocyst-septin complex (this study), was reported to self-associate and to inhibit CDC42 activation, perhaps by associating with C-terminus sequences of endogenous IQGAP1 (Mataraza et al., 2002; Le Clainche et al., 2007), which would present a synergistic mechanism for enhancing secretion. Together, these data affirm the idea that IQGAP1 modulates CDC42 activity, but, at present, it remains unclear how, because purified recombinant IQGAP1 failed to exhibit a direct GEF activity on CDC42. It is likely that IQGAP1 recruits a CDC42 GEF to facilitate an exchange activity. Nevertheless, CDC42 activation impairs IQGAP1 function in exocytosis by displacing IQGAP1 from its downstream effectors, such as the exocyst-septin complex (Fig. 1C, Fig. 2 and supplementary material Fig. S1), by facilitating the binding of an inhibitor, or by dissociating the exocyst and/or septins from IQGAP1, leading to regulated inhibition of secretion in response to intrinsic or extrinsic stimuli.
A mechanism for IQGAP1 in regulating protein synthesis and secretion
The finding that IQGAP1-C inhibits the function of IQGAP1 in exocytosis suggests that this domain serves as a dominant-negative (Figs 1, 6 and 8). Apparently, this represents a common mechanism for IQGAP1, because this domain also inhibits the ability of IQGAP1 to dissociate cell-cell contacts (Fukata et al., 2001). How this translates on the cellular level could be subject to many interpretations; however, a likely mechanism is that IQGAP1 acts as a conformational switch, operating in open or closed molecular states (Fig. 9C). With regard to secretion, IQGAP1-C mimics inactive (open) states and IQGAP1-N represents active (closed) states. This would explain why overexpression of full-length IQGAP1 had little effect on secretion and presents a mechanism whereby IQGAP1 could function positively or negatively with CDC42 in one state versus the other (Fig. 9C). In support of this model is the finding by independent groups that two different regions on the C-terminus form complexes in vitro (Fukata et al., 2001; Grohmanova et al., 2004) and that phosphorylation of serine 1443 in the second region prevents this interaction, increasing the binding of nucleotide-depleted CDC42 (Grohmanova et al., 2004). This model is generally consistent with that for DBL proteins (CDC42-GEFs) shown to exist in conformational states regulated by tyrosine phosphorylation, which causes an opening of their structure leading to the activation of Rho GTPases (Aghazadeh et al., 2000). In exocytosis, IQGAP1 operates in an inverse (closed) state in which an opening of its structure leads to binding and activation of CDC42, resulting in the inhibition of secretion (Fig. 9C). It is plausible, however, that other functions of IQGAP1, such as cytokinesis, are positively influenced in this open, CDC42-bound, state.
Materials and Methods
Construction and expression of IQGAP1 domains
Full-length IQGAP1, IQGAP-F1, IQGAP1-N (N1), IQGAP1-C (C2) and the IR-WW domain (Fig. 3A) were generated by high-fidelity PCR, cloned into the expression vector TOPO pCDNA3.1 (Invitrogen) and confirmed by DNA sequencing (Cornell BRC facility) to create V5-6×His double-tagged proteins. The QuickChange site-directed mutagenesis kit (Stratagene) was used to delete amino acids M1054-K1077, comprising the CDC42-binding region (Mataraza et al., 2003), from the V5-tagged pcDNA3.1 plasmid encoding IQGAP1-F1, thus generating the FDMK construct. Three silent mutations (underlined) were introduced into the V5-IR-WW plasmid at nucleotides 363-5′-TGCAATGGATGAAATTGGG-3′-381, creating an RNAi-refractory [adapted from Watanabe et al. (Watanabe et al., 2004)] IR-WWR domain. Plasmid HA-pCDNA3-CDC42F28L was similarly mutated to add C37A, which reduces the binding of IQGAP1 to CDC42. All constructs were confirmed by DNA sequencing and expression was further confirmed by recognition of the expected sizes with V5-tag antibodies (Invitrogen) as well as with IQGAP1 antibodies specific for the N- or the C-halves of IQGAP1 (Upstate and Santa Cruz Biotech).
Cell culture, transfection, RNAi and IP
βTC-6 cells were purchased from ATCC (CRL-11506) and mastoparan PTX was from Sigma-Aldrich. The transfection efficiency of βTC-6 cells was >70%, as determined by GFP, and, like βHC-9 cells, they lost glucose sensitivity with higher passage numbers. Therefore, early passages were used and the passage number was kept low. All cell lines were cultured in Dulbecco's modified Eagle medium (DMEM) supplemented with 10% calf serum (NIH3T3) or 10% fetal bovine serum (FBS; 15% for βTC-6 and 20% for βHC-9) and 100 units/ml penicillin, 100 μg/ml streptomycin (Invitrogen) in a humidified 5% CO2 incubator at 37°C. HeLa cells (ATCC) were cultured in MEM under the same conditions. Stable transfection of HeLa and β-TC6 cells was performed following the Invitrogen manual-selecting for clones with equal expression levels. For transient expression, cells from 100 mm plates were transfected with 9 μg DNA with Lipofectamine (Invitrogen) following the manufacturer's instructions. After 48 hours, the cells were washed with ice-cold phosphate-buffered saline (PBS) and lysed on ice for 20 minutes in buffer (25 mM HEPES, pH 7.4, 15 mM MgCl2, 150 mM NaCl, 1% NP40, 10 μg/ml each of leupeptin and aprotinin and 0.2 mg/ml phenylmethylsulfonic chloride) prior to centrifugation at 13,000 g for 10 minutes. Protein concentrations were determined with the Bio-Rad Dc kit and equal amounts precleared with beads (15 μL) for 1 hour at 4°C and used for IP. Briefly, the antibody was added to the precleared lysate, incubated on ice for 1 hour and 40 μl of PBS-equilibrated protein A or G beads were added and gently rocked overnight at 4°C. The beads were washed four times with 1 ml lysis buffer, boiled for 10 minutes in 40 μl 2×SDS sample buffer and loaded for SDS-PAGE.
Immunoblotting was performed with the antibodies indicated in the figures. Antibodies for IQGAP1 and IQGAP2 were from Upstate Biotech, for IQGAP3 were from Novus Biologicals, and those for EXO70 and SEPT2 were previously described (Vega and Hsu, 2001; Vega and Hsu, 2003). Mouse anti-rSec8 was from Stressgen Biotechnology and HA from Covance. Insulin antibodies were from Abcam and secondary antibodies were from Jackson's Laboratory or from Molecular Probe. The immunoblot images were acquired with a Bio-Rad ChemiDoc XRS imager.
The human IQGAP1 RNA 21-oligomers (5′-UGCCAUGGAUGAGAUUGGA-3′) were synthesized (Dharmacon) in both sense and antisense directions with dTdT overhang at the 3′ termini. The sequences were searched for in the GenBank database against the human genome, ensuring that only IQGAP1, and not its isoforms IQGAP2, IQGAP3 or other genes, was targeted. A scramble IQGAP1 sequence: 5′-CAGUCGCGUUUGCGACUGG-3′ and the siCONTROL non-targeting oligomer from Dharmacon were used as control. The oligos were transfected at 100 nM with Lipofectamine 2000 (Invitrogen) following the manufacturer's instructions. After 48 hours, the cells were fixed for immunofluorescence or lysed for western blotting. The transfection efficiencies in HeLa cells were 70-90%, as monitored by fluorescent-label RNAi from Dharmacon.
In vitro pull-down assays
The GST-EXO70 and GST-SEPT2 fusion proteins were previously described (Vega and Hsu, 2001; Vega and Hsu, 2003). These were expressed in Escherichia coli BL21 (DE3) and sonicated in buffer S (20 mM Tris, pH 8, 2 mM EDTA, 2 mM DTT, 150 mM NaCl, 0.1% Tween 20). The sonicates were centrifuged at 18,000 g for 15 minutes at 4°C, the supernatant was incubated with GST beads overnight with gentle rocking at 4°C, and the beads were washed three times in buffer S and incubated with equal amounts of proteins from COS7, precleared with GST beads, overnight at 4°C with gentle rocking. The beads were washed four times with excess buffer S, boiled in ∼40 μl 2×SDS sample buffer and used for immunoblotting. For the reciprocal experiment, His-IQGAP1 proteins were purified using the BD TALON cobalt-based resins (BD Biosciences), following the manufacturer's instructions, incubated with the bacterial extracts containing GST-SEPT2 or GST-EXO70 and processed as described above. The GST-PBD assay for CDC42 activity was performed using the CDC42 activation kit from Upstate following the manufacturer's instructions. The immunoblot images were acquired with a Bio-Rad ChemiDoc XRS imager and exported as TIFF files.
The cells were washed with PBS, fixed in –20°C methanol acetone for 10 minutes, blocked with 1 mg/ml BSA in PBS and incubated with primary antibodies or IgG as control followed by secondary antibodies (Texas red and Alexa-Fluor-488, Molecular Probes) for 1 hour each at room temperature. The nuclei were stained with DAPI (Sigma) and the images were captured with a Leica confocal or an Olympus fluorescence microscope fitted with a CCD camera and Slide Book software with no post-image-acquisition processing except for dark-field correction.
Insulin-secretion assays and immunofluorescence
Pancreatic βTC-6 cells were stably transfected with the pCDNA3.1 vector alone or encoding either the V5-IQGAP1 domains and/or HA-tagged CDC42WT. Early passages were used for insulin-secretion assays. Equal numbers of cells were grown and washed twice with Krebs-Ringer bicarbonate (KRB) buffer (129 mM NaCl, 5 mM NaHCO3, 4.8 mM KCL, 1.2 mM KH2PO2, 2.0 mM CaCl2, 1.2 mM MgSO4, 0.2% BSA, 10 mM HEPES, pH 7.4, and 0.1 mM glucose), incubated at 37°C in the same buffer for 30 minutes and treated with buffer alone (basal) or with 30 mM glucose (stimulated) as previously described (Daniel et al., 2002). Both sets were incubated at 37°C for another 30 minutes. Aliquots of the buffer containing the secreted insulin from basal or stimulated cells were collected and stored at –80°C, and the cells were washed twice with ice-cold PBS, lysed and the expression levels of the constructs were determined by immunoblotting. Cells with equal levels of domain expression were selected for comparison of insulin-secretion levels with the ultrasensitive mouse insulin enzyme-linked immunosorbent assay kit (Immunodiagnostic Systems) following the manufacturer's instructions. For immunofluorescence, the cells were seeded on chamber slides treated as above and stimulated with glucose, then washed and fixed.
35S-labeling and pulse-chase experiments
Pancreatic βTC-6 cells expressing equal amounts of the indicated IQGAP1 domains were seeded at 4×104 in 100-mm plates, rinsed once in methionine- and cysteine-free DMEM and incubated in the same medium containing 10% dialyzed FBS for 2-3 hours prior to labeling with 0.1 mCi/ml of 35S Met/Cys (Express Protein Labeling Mix, Perkin Elmer) for 30 minutes at 37°C in 3.5 ml of the same medium. The cells were washed once with 10 ml normal culture medium and incubated in fresh medium containing glucose for the indicated time points, washed in cold PBS and lysed on ice for 20 minutes in buffer A [40 mM HEPES, pH 7.5, 120 mM NaCl, 1 mM EDTA, 10 mM pyrophosphate, 10 mM glycerophosphate, 50 mM NaF, 0.5 orthovanadate, EDTA-free protease inhibitors (Roche) and 1% Triton X-100]. The lysate was cleared by centrifugation at 13,000 g for 10 minutes. The protein concentration was determined and 50 μg was loaded for the total, or equal amounts were used for IP, as described above, with insulin antibodies (Abcam) and resolved on 20% SDS-PAGE. The images were acquired with a phosphoimager.
βHC-9 cells were washed with cold PBS, homogenized in buffer H (150 mM NaCl, 100 mM Tris, pH 7.4, 10 mM MgCl2 and protease inhibitors) and centrifuged for 5 minutes at 3024 g. The supernatant was centrifuged at 27,216 g for 15 minutes prior to ultra-centrifugation at 100,000 X g for 90 minutes. The supernatant was saved as the cytosol fraction and the pellet suspended in the same buffer containing 1% NP40, incubated on ice for 30 minutes with regular mixing then centrifuged at 27,216 g for 30 minutes and the supernatant saved as the solubilized membrane fraction. Equal amounts of proteins in the cytosol and the membrane fractions were used for immunoblotting.
This work was supported by grants to M.A.O. from NIH-National Cancer Institute (#K22CA104285), the American Cancer Society and research funds from Cornell University. We thank John Hellmann and Ahmad Gaballa for facilitating the radiolabel experiments. We thank Richard Cerione for the CDC42 plasmids and comments on a draft; Geoffrey Sharp and Troitza (Trisha) Bratanova-Tochkova for βHC-9 cells; and Carmen Hassan and Joanne (Chung-Un) Kim for technical assistance. Finally, we appreciate the insights of the anonymous reviewers of this manuscript.