Regulation of Spine Density and Morphology by IQGAP1 Protein Domains

IQGAP1 is a scaffolding protein that regulates spine number. We now show a differential role for IQGAP1 domains in spine morphogenesis, in which a region of the N-terminus that promotes Arp2/3-mediated actin polymerization and branching stimulates spine head formation while a region that binds to Cdc42 and Rac is required for stalk extension. Conversely, IQGAP1 rescues spine deficiency induced by expression of dominant negative Cdc42 by stimulating formation of stubby spines. Together, our observations place IQGAP1 as a crucial regulator of spine number and shape acting through the N-Wasp Arp2/3 complex, as well as upstream and downstream of Cdc42.

However, and despite a well-recognized role for scaffolding and cytoskeletal crosslinker proteins in neuronal polarization and synaptic plasticity [1,2] the function of IQGAP1 in brain neurons remained largely unexplored until recently. Some of the initial evidences came from a study showing that impairment of Ncadherin-mediated ERK signaling is paralleled by redistribution of IQGAP1 from spines to dendritic shafts [17]. Later on, the same group identified IQGAP1 as a key regulator of dendritic spine number with a specific role in cognitive but not emotional or motivational processes [10]. Mice lacking IQGAP1 exhibited marked memory defects, including impaired long-term potentiation (LTP) in a weak cellular learning model [10]. Interestingly, hippocampal neurons from IQGAP1 2/2 mice displayed reduced spine number, lower levels of surface NR2A and impaired ERK activity. Other study demonstrated that IQGAP1 and the microtubule plus-end tracking protein, CLIP-170, cooperatively regulate dendritic arbor growth in both cortical and hippocampal pyramidal neurons [11]. Based on these observations and to gain insights into the mechanisms by which IQGAP1 regulates spine morphogenesis, in the present study we used several deletion mutants to identify domains of IQGAP1 that could be necessary for spine and synapse formation.

Results and Discussion
IQGAP1 possesses several sequentially arranged functional domains ( Figure 1) that enable direct binding to a wide spectrum of cytoskeletal, adhesion and signaling molecules [15,18,19]. In this study, we used deletion mutants (D) of the following IQGAP1 domains ( Figure 1): 1) The Calponin homology domain (D-CHD); 2) the RasGAP-related domain (D-RGD); and 3) the C-terminal region (D-CT), to test their involvement in spine morphogenesis.
First, a gain-of-function experiment was performed to evaluate the effect of overexpressing IQGAP1 wild-type (IQGAP1 WT) on spine number and shape. Cultured hippocampal pyramidal neurons were transfected with Green Fluorescent Protein (GFP) or Red Fluorescent Protein (RFP) plus myc-tagged IQGAP1 WT or mock vector 17 days after plating, fixed 18-20 hours later, double stained with MAP2 or synaptophysin and examined by confocal microscopy ( Figure 2); in some experiments, neurons were either co-transfected with GFP-PSD95 or stained with a mAb against PSD95. Serial confocal sections were obtained and 3-D reconstructions of dendritic shafts and spines performed using Imaris software. Spine number and type were evaluated by either manual counting or computer-assisted methods using published protocols (see Materials and Methods); both procedures gave similar results (Table S1 and Figure S1). The ectopic expression of myc-tagged IQGAP1 WT ( Figure 2) or GFP-tagged IQGAP1 WT (not shown) induced a significant increase in the number of dendritic spines. High-magnification views revealed that neurons overexpressing IQGAP1 WT display many dendritic spines with long necks and large bulbous heads, characteristic of mushroomshaped spines (Figure 2 H-J). As in the case of mock-transfected neurons these spines localize GFP-PSD95 ( Figure 2K-L) or PSD-95 immunofluorescence ( Figure S2) to their tips that in most cases was in contact with synaptophysin puncta (Figure 2M-N; Figure  S2). Quantitative analysis showed that the increase in spine density of myc-tagged IQGAP1 WT positive neurons was associated with a higher number of mushroom-and stubby-shaped dendritic spines and no significant changes in thin spines and filopodial-like protrusions ( Figure 2O, 2P); this analysis also revealed that ectopic expression of IQGAP1 WT induces a significant (p,0.0001) increase in spine head size ( Figure 2Q).
We then examined the consequences of expressing an IQGAP1 mutant lacking the CHD (Figure 1, D-CHD-IQGAP1). The CHD (amino acids 44-159) located at the N-terminus is responsible for actin and presumably N-WASP binding, as well as capable of promoting Arp2/3-mediated actin polymerization and branching [16]. Neurons expressing this mutant display an increase in the number of spine-like protrusions ( Figure 2O; Figure 3A). Highmagnification views ( Figure 3B) and 3-D reconstructions ( Figure 3C) revealed that many of these protrusions resembled filopodial-like extensions lacking a discernable head or displaying a small expansion at the tip. Therefore, we tested, whether or not these structures may represent thin spines or conventional filopodia [20][21][22][23][24] by co-expressing D-CHD-IQGAP1 with GFP-PSD95 or staining with anti-PSD95. Interestingly, co-expression of these constructs revealed that the majority (.95%) of the filopodial extensions display GFP-PSD95 fluorescence at their tips ( Figure 3D, 3E) and colocalize with synaptophysin puncta, typical of thin spines ( Figure 3F, 3G); quantitative analysis confirmed these observations ( Figure 3H-J). Collectively, these results suggest that the CHD of IQGAP1 is required for proper spine head formation, and that in its absence thin spines capable of contacting synaptic terminals are formed; they also raise the possibility that the increase in the number and size of mushroomshaped spines elicited by IQGAP1 WT could require N-WASP and/or Arp 2/3 expression. Previous work has suggested that the Arp 2/3 complex induces formation of the branched actin network in the spine head [21,23,25,26]. Therefore, we generated sh-RNAs specific for Arp3, Arp2 [27] and N-WASP ( Figure S3). As shown in Figure 3I, silencing of Arp3 significantly reduces spine number. The remaining dendritic protrusions resemble filopodial extensions; a similar phenotype was observed after Arp2 (not shown) or N-WASP suppression ( Figure 3I). Interestingly, co-expression of IQGAP1 WT in Arp 2, or 3 or N-WASP-suppressed neurons stimulated the formation of filopodial-like extensions, but not of mushroom-or stubby spines ( Figure 3K-M). This phenotype resembles the one observed after expression of D-CHD; however, the filopodial protrusions formed in the absence of Arp 2/3 or N-WASP failed to localize with synaptophysin puncta ( Figure 3J, 3N-Q) and do not contain PSD95 at their tips (not shown). This is consistent with the idea that modeling of the actin cytoskeleton has also a major role in organizing the post-synaptic density [22,26,28]. It has also been reported filopodial formation in the absence of Arp2/3 or WAVE [27,29]. Therefore, the protrusions observed in Arp2/3 or WASP-suppressed neurons co-expressing IQGAP1 WT may represent either immature thin spines or ''conventional filopodia''. Since our results also suggest that the CHD is dispensable for stalk/neck formation, we decided to examine the role of GRD.
This domain (Figure 1), which extends from amino acids 1025 to 1238 binds to the small GTPases Cdc42 and Rac1, both implicated in spine formation [3,4]. Expression of IQGAP1 D-GRD also increases spine number ( Figure 2O and Figure 4A-G, 4K); analysis of spine shape reveals a selective increase in the number of stubby spines ( Figure 4K) that contain GFP-PSD95 ( Figure 4E) or stain for PSD95 ( Figure S2) and are contacted by synaptophysin puncta (Figure 4F-G; Figure S2). These observations suggest that IQGAP1 interaction with Cdc42 and/or Rac could be important for stalk-neck formation; since a recent study has shown that stimulated spines have increased Cdc42 activation at the spine neck [30] we tested the effect of a Cdc42 dominant negative (DN) mutant (T17N) on the stimulatory effect of IQGAP1 WT on spine formation. In agreement with previous observations [26] expression of T17N reduced the number of mushroom-shaped spines, consistent with its proposed role in stalk-neck formation; intriguingly, T17N also induced a significant increase in the number of filopodial extensions ( Figure 4L). However, co-expression of IQGAP1 WT reverts the T17N phenotype, stimulating formation of stubby spines and reducing filopodial number ( Figure 4H-L), suggesting that IQGAP1mediated stalk-neck formation and/or extension requires Cdc42. Conversely, spine deficiency caused by IQGAP1 suppression ( Figures S4 and S5) can be rescued by co-expression of an active fast cycling mutant (F28L) of Cdc42 [31,32] ( Figure S5). Together, these observations favor the idea that IQGAP1 acts both upstream (e.g. a regulator) and downstream (e.g. an effector) of Cdc42 [18,33]; they also suggest that Cdc42 may not require IQGAP1 to promote spine formation.
Next, we evaluated the IQGAP1 C-terminus ( Figure 1, CT). This region, comprising amino acids 1563-1657, binds to several proteins including CLIP-170, adenomatous polyposis coli (APC), and E-cadherin [18]. Since an interaction between IQGAP1 and CLIP170 regulates dendritic growth [11] and dynamic microtubules containing +TIPs invade spines [34], it became of interest to explore if expression of a deletion mutant of IQGAP1 lacking the CT could affect spine number or shape. The results obtained show that deletion of this domain does not prevent the stimulatory effect of IQGAP1 on spine number or spine head size; however, this mutant fails to increase the total length of mushroom-shaped spines, as does full length IQGAP1 ( Figure 5).
In the final set of experiments we evaluated the possible participation of the CHD or GRD or CT domains of IQGAP1 in regulating the distribution of the NR2A, a subunit of the NMDA receptor. Previous studies have shown that IQGAP1 interacts with NR2A and that NR2A surface levels were significantly decreased in IQGAP12/2 mice [10]. To test if any of the IQGAP1 deletion mutants used in this study could affect the intracellular or surface distribution of NR2A, we cotransfected GFP-tagged NR2A at the N-terminus with each of the deletion mutants. By staining non-permeabilized transfected   neurons with a mAb against GFP, we were able to label surface NR2A [10]. The results obtained ( Figure 6) showed that none of the IQGAP1 deletion mutants alter the intracellular or surface distribution of NR2A, suggesting that other IQGAP1 domains or a combination of domains regulate the distribution of this NMDA subunit.

Conclusions
The present results confirm and extend recent studies [10,11,17] indicating that IQGAP1 is required for spine formation. The new information presented here suggests that IQGAP1 protein domains actively participate in spine morphogenesis and differentially affect spine number and shape acting through the N-WASP-Arp2/3 complex and Cdc-42 signaling.

Short-Hairpins RNA and Plasmid Constructs
The IQGAP1, Arp2, Arp3, N-WASP short hairpin (sh) RNAs and their corresponding scrambled control sequences were constructed using previously described procedures [35,36]. In brief, DNA fragments containing U6-sh-RNA and U6-scrambledsh were inserted into pCAG vector in which the GFP or HcRed cDNA is under the control of a chick actin-minimal (CAG) promoter [35,36].
The following targeting sequences were used: For IQGAP1-sh Forward: 59TGCCATGGATGAGATTGGAAAGCTTTC-CAATCTCATCCATGGCACTTTTTTG and.
The DNA was then cloned in pcDNA3.1 Myc. The IQGAP1-D-CHD mutant was generated from pcDNA3.1-IQGAP-WT by PCR and sub cloned in pcDNA3.1 by generating a SalI site using the following primers: Forward: 59-GTCGACGAGAAGTATGGCATCC and Reverse 59-TTACTTCCCGTAGAAC-39.
The transcript was then cloned into pEGFP C2 between EcoR1 and KpNI.
For IQGAP1-DCT a mutation was generated to create a stop codon at position 1377. The following primers were used: Forward: 59-GCCTGGAGATGAGAATGCATAAATG-GATGCTCGAACC-39.
Reverse Efforts were made to minimize animal suffering and to reduce the number of animals used.
Neuronal shape parameters were evaluated as described previously [38][39][40]. Briefly, maximal projection images showing the complete neuronal arbor of transfected neurons visualized by GFP or HcRed fluorescence or myc-tagged IQGAP1 immunofluorescence were created from confocal images acquired through a 606or 636, 1.4 NA oil objective. Spine number and shape were assessed manually by GFP or RFP fluorescence or myc-IQGAP1 immunofluorescence and the presence of synapses by co-expressing GFP-PSD95 and/or co-staining with synaptophysin as described by Tolias et al., [41]. We also used computer-assisted methods for evaluating spine number and shape using the procedures described by Rodriguez et al., [42,43]. No differences in spine number or shape were found between manual and computer-assisted methods; both procedures gave similar results ( Figure S1, Table S1). At least 10 dendritic segments (50 mm length/each) per cell (total 6 cells per culture) from at least 3 different cultures were analyzed for each experimental condition. Differences among experimental groups were analyzed by oneway ANOVA and Tukey's post hoc test.

Western Blot
Changes in the levels of IQGAP1, N-WASP and ARP2/3 after RNA interference treatments of CHO cells or cultured hippocampal pyramidal neurons were analyzed by Western blotting as described previously [35]. Densitometry of Western blots were performed using Scion Image software. Figure S1 Computer-assisted analysis of spine number and shape. A representative image of the computer display provided by the software used to automatically evaluate changes in spine number and shape. For further details see Ref. 42 and 43. (TIFF) Figure S2 PSD95 immunofluorescence in neurons expressing IQGAP1 and its mutants. (A-C) High magnification views of a dendritic segment from a 17 DIV hippocampal cell culture transfected with myc-tagged IQGAP1 WT (green) and double stained with anti-PSD95 (red); note that long spines stain for PSD95 (arrows). (D-F). A similar set of images but from a culture transfected with myc-tagged IQGAP1 WT (green) and double stained with anti-synaptophysin (red); note that spines colocalize with endogenous synaptophysin puncta (arrows). (G-I) High magnification views of a dendritic segment from a 17 DIV cultured hippocampal neuron transfected with myc-tagged D-GRD IQGAP1 (green) and double stained with anti-PSD95 (red); note that stubby spines stain for PSD95 (arrows). (J-L) A similar set of images but from a 17 DIV cultured hippocampal neuron transfected with myc-tagged D-GRD IQGAP1 (green) and double stained with anti-synaptophysin (red); note that stubby spines colocalize with endogenous synaptophysin puncta (arrows).  Table S1 A comparison of spine number and type as determined by manual counting vs. automated methods [42,43]. Each value is the mean 6 standard deviation. At least 10 dendritic segments (50 mm length/each) per cell (total 6 cells per culture) from at least 3 different cultures were evaluated. Note that both methods gave similar results. (DOC)

Supporting Information
Author Contributions