Interaction of Cupidin/Homer2 with two actin cytoskeletal regulators, Cdc42 small GTPase and Drebrin, in dendritic spines.

BACKGROUND
Homer is a postsynaptic scaffold protein that links various synaptic signaling proteins, including the type I metabotropic glutamate receptor subunits 1alpha and 5, the inositol 1,4,5-trisphosphate receptor, Shank and Cdc42 small GTPase. Overexpression of Homer induces changes in dendritic spine morphology in cultured hippocampal neurons. However, the molecular basis underpinning Homer-mediated spine morphogenesis remains unclear. In this study, we aimed to elucidate the structural and functional properties of the interaction between Cupidin/Homer2 and two actin-cytoskeletal regulators, Cdc42 small GTPase and Drebrin.


RESULTS
Cupidin/Homer2 interacted with activated Cdc42 small GTPase via the Cdc42-binding domain that resides around amino acid residues 191-283, within the C-terminal coiled-coil domain. We generated a Cupidin deletion mutant lacking amino acids 191-230 (CPDDelta191-230), which showed decrease Cdc42-binding ability but maintained self-multimerization ability. Cupidin suppressed Cdc42-induced filopodia-like protrusion formation in HeLa cells, whereas CPDDelta191-230 failed to do so. In cultured hippocampal neurons, Cupidin was targeted to dendritic spines, whereas CPDDelta191-230 was distributed in dendritic shafts as well as spines. Overexpression of CPDDelta191-230 decreased the number of synapses and reduced the amplitudes of miniature excitatory postsynaptic currents in hippocampal neurons. Cupidin interacted with a dendritic spine F-actin-binding protein, Drebrin, which possesses two Homer ligand motifs, via the N-terminal EVH-1 domain. CPDDelta191-230 overexpression decreased Drebrin clustering in the dendritic spines of hippocampal neurons.


CONCLUSION
These results indicate that Cupidin/Homer2 interacts with the dendritic spine actin regulators Cdc42 and Drebrin via its C-terminal and N-terminal domains, respectively, and that it may be involved in spine morphology and synaptic properties.

Dendritic spine morphology is dynamically changed in response to synaptic activity, which is associated with synaptic functions including the long-term maintenance of synaptic strengthening [15,16]. Impaired spine morphology is known to contribute to mental retardations [15,16]. We previously showed that Cupidin, identical to Homer2, is co-sedimented with filamentous actin (F-actin) via the EVH1 domain, and also interacts with the GTP-bound, activated form of Cdc42 small GTPase via the C-terminal region [6]. Interestingly, over-expression of Cupidin/ Homer2 suppressed Cdc42-induced formation of filopodia-like protrusions in HeLa cells [6]. Moreover, Cupidin/ Homer2 was partly colocalized with Drebrin, a dendritic F-actin-binding protein, in the dendrites of cultured hippocampal neurons [12] and cerebellar granule cells [17]. It is known that both Cdc42 [18][19][20] and Drebrin [21,22] are involved in dendritic spine morphogenesis by regulating actin-cytoskeletal organization. A previous study showed that over-expression of Homer1b together with Shank induced enlargement of the spine heads of hippocampal neurons [13]. Together, the results of these studies suggest that Homer family proteins are involved in the regulation and/or plasticity of spine morphology by interacting with two dendritic F-actin regulators, Cdc42 and Drebrin. However, little is known about the molecular basis underpinning the involvement of Homer proteins in actin cytoskeleton-based regulation of spine morphology.
In this study we analyzed the structural and functional properties of the Cupidin/Homer2 scaffolding that interacts with two dendritic spine F-actin organization modulators, Cdc42 and Drebrin. We defined the Cdc42-binding domain in the C-terminal region of Cupidin/Homer2 and revealed the functional significance of Cdc42-binding domain in spine and synapse formation by cultured hippocampal neurons, as well as in Cdc42-induced filopodia-like protrusion formation in HeLa cells. We also proved Drebrin to be a Homer EVH1-binding target and showed the effect of Cdc42-binding domain on the Drebrin accumulation in spines. These results strongly implicate the postsynaptic Homer scaffolding in the morphogenesis of dendritic spines.

Cupidin interacts with activated Cdc42 via the C-terminal coiled-coil region
Cupidin/Homer2 is comprised of an N-terminal EVH1 domain, a C-terminal coiled-coil (CC) motif and two Leu zipper motifs A and B (LZA and LZB, respectively) ( Fig. 1). Our previous study indicated that Cupidin interacts with Cdc42 small GTPase in a GTP-dependent manner via the C-terminal region [6]. To narrow down the region responsible for Cdc42 binding activity, we generated a series of glutathione S-transferase (GST) fusion constructs containing various regions of Cupidin (CPD), as shown in Fig. 1. These GST fusion proteins were blotted onto membrane filters and probed by a ligand overlay assay with GSTfused Cdc42 in the presence of [ 35 S]-GTPγS. As described previously [6], GST-CPD and GST-CPD C (C-terminal amino acids (aa) 111-343), but not GST-CPD N (N-terminal aa 1-110), were radioactively labeled (Fig. 1A). No specific radioactive labeling of GST-CPD C was observed in the presence of [ 35 S]-GDPβS and Cdc42, as previously reported [6].
To assess the multimerization ability of these Cupidin deletion mutants, the GST-proteins were treated with thrombin to remove the GST moiety, which are known to bind each other, and was subjected to a mobility shift assay using sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) after treatment with the cross-linker dimethyl pimalidate (DMP). Upon DMP treatment, immunoreactivity for CPDΔ191-230 containing both LZA and LZB was enhanced at a position indicative of multimers in comparison with that for CPDΔ191-283 and CPDΔ231-283, both of which contained LZB only ( Fig. 2A). To test a possible correlation between multimerization and Cdc-42 binding ability, we coexpressed both GFP-and Flag-tagged constructs of CPD or CPDΔ191-230 in COS cells together with either dominant-active Cdc42 (Cdc42 V12 ) or dominant-negative Cdc42 (Cdc42 N17 ), followed by evaluation of the interaction between GFP-and Flag-tagged constructs using immunoprecipitation with an anti-Flag antibody (Fig.  2B). With both CPD and CPDΔ191-230 constructs, GFPtagged constructs were coimmunoprecipitated with Flagtagged constructs in intact cells, regardless of whether they expressed active or inactive Cdc42. Although the coimmunoprecipitation efficiency of GFP-tagged CPDΔ191-230, which showed a loss of Cdc42-binding activity, was 30% lower than that of GFP-tagged CPD, these results indicated that Cdc42 binding to Cupidin is substantially independent of the self-multimerization of Cupidin.

The Cupidin-Cdc42 interaction influences actincytoskeletal organization and the morphology of HeLa cells
We previously showed that co-expressed Cupidin suppresses the dominant-active Cdc42-induced morphological and actin-cytoskeletal changes in HeLa cells [6]. To verify the cellular function of Cdc42 binding, we analyzed the influence of Cdc42-binding deficiency on these cellular phenotypes (Fig. 3A). In HeLa cells transfected with CPD full-length or CPDΔ191-230, stress fibers and bun-Multimerization assays with the Cupidin deletion mutants Figure 2 Multimerization assays with the Cupidin deletion mutants. Cross-linking assay using the chemical cross-linker DMP. Bacterially expressed GST-proteins were digested by thrombin, and the resulting untagged proteins were incubated in the presence (+) or absence (-) of DMP followed by Western analysis using anti-Cupidin (CPD C) antibody. One asterisk indicates the migration of multimers, and two asterisks indicate that of monomers. (B) Co-immunoprecipitation assay using extracts from cells in which the three distinct constructs shown in the upper panels were heterologously overexpressed in COS7 cells. I, Input; C, IP samples using non-immune serum; F, IP samples using anti-Flag antibody. Western blot analysis was performed using anti-Cupidin antibody, and the multimerization efficacies were calculated from the band signal intensities, and shown in the graph at the bottom.
dles of F-actin were formed in the cytoplasm and the cell periphery, respectively, as in cells transfected with vector alone (mock). Upon expression of Cdc42 V12 , a dominantactive form of Cdc42, many filopodia-like protrusions with actin bundles were drastically induced around the cell periphery. Co-expression of the CPD diminished the number of Cdc42 V12 -induced protrusions and actin bundles in the cell periphery. By contrast, co-expression of CPDΔ191-230 did not affect Cdc42 V12 -induced cellular phenotypes, as statistically proven by counting spike number around the cell periphery (Fig. 3B). These results indicated that the region containing amino acids 191-230 is required for the suppression of Cdc42 V12 -induced morphological and actin-cytoskeletal changes in HeLa cells.

The Cdc42-binding domain of Cupidin is involved in the formation of dendritic spines and synapses in hippocampal neurons
Cupidin is predominantly localized in the dendritic spines of cultured hippocampal neurons [12,23]. To investigate the role of Cdc42 binding in the postsynaptic targeting of Cupidin, we infected primary hippocampal cell cultures with recombinant adenovirus vectors containing enhanced green fluorescent protein (GFP)-fused CPD, GFP-fused CPDΔ191-230 and GFP alone. Punctate fluorescence of GFP-CPD and GFP-CPDΔ191-230 was observed throughout the dendrites of hippocampal neurons ( Fig. 4A). In neurons expressing GFP-CPD, approximately 60% of the total punctate fluorescent signals were located at spine heads, whereas less than 10% of these were detected in dendritic shafts (Fig. 4B). On the other hand, in neurons expressing GFP-CPDΔ191-230, about 40% and 30% of the total fluorescent puncta were localized in spine heads and dendritic shafts, respectively (Fig.  4B). These results suggested that the Cupidin C-terminal region possessing Cdc42 binding activity contributes to the targeting of Cupidin to dendritic spines.
Overexpression of Cupidin induced mushroom-type spines in hippocampal neurons as shown in Fig. 4C. On the other hand, overexpression of GFP-CPDΔ191-230 decreased the number of mushroom-type spines and increased filopodia-like or odd-shaped protrusions, although the total number of dendritic protrusions was only slightly reduced in neurons expressing GFP-CPDΔ191-230 compared with neurons expressing GFP-CPD and GFP alone (Fig. 4C). These results are consistent with the idea that Cdc42-binding domain of Cupidin is important for spine morphogenesis and/or maturation.
The effect of Cdc42 binding to Cupidin on synapse formation was analyzed by immunostaining for synaptophysin, a presynapse marker ( Fig 5A). The number of synaptophysin-positive puncta was drastically reduced in adjacent to dendrites of neurons overexpressing GFP-CPDΔ191-230. In addition, the number of synaptophysin puncta localized adjacent to GFP-CPDΔ191-230 puncta was also decreased. We further explored synaptic functions by measuring the peak amplitudes and frequencies of miniature excitatory postsynaptic currents (mEPSCs) in neurons overexpressing these constructs at 21 DIV ( pressing GFP-CPD (-9.26 ± 0.12 pA) did not significantly differ from those in neurons overexpressing GFP (-9.86 ± 0.15 pA, p > 0.04), the mEPSC amplitudes in neurons overexpressing GFP-CPDΔ191-230 were significantly reduced (-8.89 ± 0.11 pA, p < 0.001). The mEPSC frequency was enhanced by overexpression of either GFP-CPD (6.90 ± 0.81 Hz, p < 0.001) or GFP-CPDΔ191-230 (6.68 ± 0.94 Hz, p < 0.001) in comparison with overexpression of GFP alone (6.13 ± 0.77 Hz), although at present we could not indicate whether this increased frequency reflects an alteration in the presynaptic release probability or the number of functional synapses. Taken together, these results suggested that a deletion in the Cdc42-binding domain disturbs not only synapse formation, but also synapse electrophysiology.

The Cupidin-Cdc42 domain is involved in Drebrin targeting into dendritic spines
A homology search with the Homer ligand PPxxF motif identified two homologous sites (aa 592-596 and 674-678) in the C-terminal region of mouse Drebrin (Fig. 6A).
Coimmunoprecipitation with anti-Homer antibodies (either anti-Homer 1, 2 and 3 antibody mixture or antipan-Homer antibody) revealed an association between Homer proteins and Drebrin proteins (two splicing variants A and E) in P8 mouse brains (Fig. 6A). Since the anti-Drebrin monoclonal antibody used recognizes the C-terminal region near the PPxxF motifs, the anti-Drebrin antibody might interfere with the Cupidin-Drebrin interaction in a reverse coimmunoprecipitation of Cupidin with anti-Drebrin antibody. Thus, we carried out an overlay assay in which Cupidin was tested for binding to a polypeptide containing two Drebrin PPxxF motifs (aa 579-706 of mouse Drebrin A) or its double mutant (P593A and P675A) blotted onto a membrane filter (Fig.  6B). Cupidin bound to the PPxxF-containing polypeptide, but not to the mutant polypeptide, again indicating that Cupidin interacts with the C-terminal PPxxF motifs of Drebrin.
Endogenous Cupidin and Drebrin were both punctately distributed in dendritic spines of immunostained primary hippocampal neurons (Fig. 6C). Cupidin puncta (which are known to concentrate in the PSD [6,12]) largely overlapped with Drebrin puncta (which are known to concentrate throughout spine heads [21]) around the bottom half of spine heads. We next analyzed the effects of GFP, GFP-CPD, or GFP-CPDΔ191-230 overexpression on the dendritic distribution of Drebrin in primary hippocampal neurons ( Fig. 6D and 6E). The number of Drebrin puncta was slightly reduced by overexpressing Cupidin, but was significantly reduced by overexpressing GFP-CPDΔ191-230 ( Fig. 6D and 6E). Taken together, these results suggested that a deletion in the Cdc42-binding region of Cupidin disturbs dendritic Drebrin distribution in hippocampal neurons.

Discussion
Our study demonstrates the structural and functional interaction of Cupidin/Homer2 with two dendritic spine F-actin modulators, Cdc42 small GTPase, via the C-terminal region, and Drebrin, via the N-terminal EVH1 domain. Cdc42 regulates actin polymerization and is involved in filopodia formation [24] and dendritic spine morphogenesis [18-20]. Over-expression of Drebrin increases spine length [21] and promotes synaptic clustering of PSD-95 and F-actin [22]. Homer also interacts with Shank, another postsynaptic scaffold protein that binds to the GKAP/PSD-95/NMDA receptor complex [9]. Overexpression of Shank together with Homer induces enlargement of spine heads [13] and increases the level of the βPIX guanine nucleotide exchange factor (GEF) for Cdc42 in the PSD [13,25]. Moreover, oligophrenin-1 (Ophn-1), a Rho GTPase activating protein (GAP) that is involved in non-specific X-linked mental retardation and binds Homer1b/c, changes spine morphology in hippocampal neurons [26]. Taken together, these lines of evidence suggest that postsynaptic Homer-mediated scaffolding is involved in the regulation of dendritic spine morphology by interacting with the actin organization signaling molecules Cdc42 and Drebrin, as well as synaptic signaling molecules including the NMDA receptor complex, mGluR1α/5 subunits and InsP 3 R (Fig. 7). We

Conclusion
Cupidin/Homer2 interacts with activated Cdc42 via the Cdc42-binding domain within the C-terminal coiled-coil domain, which may play a role in the suppression of Cdc42-induced filopodia-like protrusion formation in HeLa cells and the formation of mushroom-type spines in hippocampal neurons. Cupidin/Homer2 interacts with a dendritic spine F-actin-binding protein Drebrin via the Nterminal EVH-1 domain. Drebrin possesses two Homer ligand motifs in the C-terminal region, and is mostly colocalized with Cupidin around the spine heads. Drebrin clustering in dendritic spines is disturbed by overexpression of Cupidin deficient in Cdc42 binding. These results suggest that Cupidin/Homer2 is involved in the modulation of spine morphology and function by scaffolding multiple target proteins, including the two dendritic spine actin regulators Cdc42 small GTPase and Drebrin.

Construction and expression of GST fusion proteins in E. coli
Glutathione S-transferase (GST) fusion constructs were generated by cloning various parts of Cupidinα/Homer2a cDNA [6] into the GST fusion vector pGEX-KG (see, Fig.  1). Escherichia coli JM109 expressing GST-fusion proteins were lysed in lysis buffer (50 mM Tris-HCl pH 7.4, 25% sucrose, 1% Triton X-100, 5 mM MgCl 2 ). GST fusion protein lysates (10 mg) were coupled to glutathione-Sepharose 4B (Amersham Pharmacia Biotech, Piscataway, NJ) Cupidin/Homer scaffolds for actin cytoskeleton modulators and postsynaptic proteins Figure 7 Cupidin/Homer scaffolds for actin cytoskeleton modulators and postsynaptic proteins. Cupidin/Homer2 acts as a scaffold tethering postsynaptic signalling proteins (such as mGluR1α/5, IP 3 R, and Shank) as well as the dendritic actin-binding protein Drebrin, a spine actin modulator, via the N-terminal EVH-1 domain, which recognizes the PPxxF consensus motif. It also links the GTP-bound, activated Cdc42 small GTPase, which is known to regulate filopodia formation, via the C-terminal CBD (Cdc42-binding domain) and forms a tetrameric scaffold complex via the C-terminal coiled-coil (CC) and Leu zipper motif (LZA and LZB) regions. The Homer scaffold complex is implicated in the facilitation of signalling crosstalk among tethered synapse proteins and regulation of the actin-based morphology of dendritic spines.

Regulation of dendritic spine actin polymerization
Homer multimers PIX Ophn-1 GDP GEF GAP by rotating for 1 h at 4°C. After washing three times with 1% Triton X-100/phosphate-buffered saline, GST fusion protein-coupled Sepharose was mixed with 1 mg of protein lysates prepared from mouse cerebella, which were pre-cleared with glutathione-Sepharose for 1 h at 4°C. After rotating for 1 h at 4°C, the GST fusion protein complex was washed five times with cell lysis buffer and subjected to immunoblotting.

Ligand overlay assay with Cdc42
Bacterially expressed GST-Cdc42 protein was purified using glutathione-Sepharose column chromatography according to a previously described procedure [6]. One μg samples of non-degraded GST-fusion proteins were separated by 10% SDS-PAGE and blotted onto nitrocellulose membranes (Hybond-ECL; Amersham Pharmacia Biotech, Piscataway, NJ). A ligand overlay assay was carried out as described previously [6]. Briefly, after the GSTfusion proteins on the blots were re-natured, the protein blots were probed by incubating with each GST-Rho family fusion protein loaded with [ 35 S]-GTPγS at an equal specific activity. After washing three times, the ligand-bound blots were air-dried and the radioactivities were analyzed using a BAS2000 Bioimaging analyzer (Fujix, Japan). The relative radioactivities were respectively measured from consistently sized areas using IPLab software (Scanalytics, Fairfax, VA), and normalized as described in the figure legends (Fig. 1).
Signal intensities in areas of consistent size were measured using IPLab software, and the efficiency of multimerization was calculated as described in Fig. 2.

Cell morphology of transfected HeLa cells
HeLa cells were transfected with CPD alone, CPDΔ191-230 alone, Cdc42 V12 alone, Cdc42 V12 and CPD, or Cdc42 V12 and CPDΔ191-230 using the calcium phosphate precipitation method described previously [6]. At 24 hours after transfection, cells were fixed with 4% formalin in PBS and stained with Alexa Fluor568-conjugated phalloidin (1:1000) (A12380, Invitrogen). Fluorescence was observed with a microscope (Eclipse E800; Nikon, Tokyo, Japan) equipped with a CCD camera (SPOT; Diagnostics Instruments Inc., Sterling Heights, MI). The phalloidin images were captured following confirmation of completing single/double transfection by detection of distinct fluoroprobes, as described by Shiraishi et., al [6]. The number of spikes protruding from the cellular edge were counted in 10 cells respectively, and represented as the means ± SE per 10 μm of cell edge; data were compared by a two-tailed unpaired Student t tests using Excel software (Microsoft Corporation, Tokyo, Japan).

Construction of and infection with recombinant adenovirus vectors
The EGFP-coding region (referred as to GFP in this study) derived from pEGFP-C1 (Clontech, Cambridge, UK) was fused in frame to the N-terminus of the full-length or mutated constructs of Cupidin to generate GFP-CPD. The GFP fragment was also fused to Cupidin with a deletion of amino acid residues 191-230 (CPDΔ191-230) to generate GFP-CPDΔ191-230. Replication-deficient adenovirus vectors carrying these GFP-fused constructs were generated by the COS-TPC method, as described previously [35]. Briefly, the DNA fragment of GFP-CPD or GFP-CPDΔ191-230 was inserted into the SwaI site of the pAxCAwt cosmid cassette (Takara, Tokyo, Japan

Analysis of spine morphology
Hippocampal cultures were fixed with 4% formaldehyde for 10 min and directly incubated with Oregon Green phalloidin (Molecular Probes, 1:200) overnight at 4°C. DiI (Molecular Probes) emulsion, mixed with cod liver oil at 1 μg/μl, was put onto the somata of neurons, which were identified by phalloidin staining, as a droplet of 20-30 μm in diameter, using a manually handled injector (Narishige, Tokyo, Japan). After incubation overnight at 4°C, the excess un-penetrated DiI emulsions were removed by suction, and DiI images were captured by confocal microscopy (MRC1024; BioRad, Hercules, CA) with 100×, 1.4 NA lens. Digital images were processed using Adobe Photoshop 6.0 software (Adobe Systems, San Jose, CA). Numbers of either GFP-CPD or GFP-CPD puncta were manually counted on the secondary dendrites of 10 neurons and the results presented as the means ± SE; data were compared by a two-tailed unpaired Student t test using Excel software. Spine morphology was categorized into five types as described in the legend for Fig. 4. Over 1,000 protrusions on the secondary dendrites of 20-30 neurons were analyzed for evaluation of spine morphology.

Immunocytochemistry
All immunocytochemical procedures were performed as described previously [12]. Briefly, primary-cultured neurons (21 DIV) overexpressing GFP-constructs by adenovirus-mediated infection were fixed with 4% paraformaldehyde for 30 min at 37°C, washed three times with PBS, and then permeabilized with 0.2% Triton X-100 in PBS for 10 min. After preincubation with 5% BSA in PBS for 1 h, cells were incubated with primary antibody (anti-synaptophysin or anti-Drebrin antibody) for 1 h at 37°C. After washing three times with PBS, the cells were incubated with Alexa Fluor 568-conjugated antimouse IgG (Invitrogen). Fluorescence and phase-contrast images of immunostained cells were captured by confocal microscopy (MRC1024; BioRad, Hercules, CA) to acquire a single focal plane with a 100×, 1.4 NA lens. Digital images were processed using Adobe Photoshop 6.0 software (Adobe Systems, San Jose, CA). The number of punctate immunopositive signals larger than 1 pixel (0.16 × 0.16 μm 2 /pixel with a 255-gradient signal intensity; signals lower than 165 on the scale were cut off to eliminate noise) was counted by measuring the area with a signal above 165 on the scale, using IPLab software. Scores from the secondary dendrites of 10 neurons were normalized to each control (= 1.0). Results presented as mean ± SE were compared by two-tailed unpaired Student t tests using Excel software.