A δ-Catenin Signaling Pathway Leading to Dendritic Protrusions*

δ-Catenin is a synaptic adherens junction protein pivotally positioned to serve as a signaling sensor and integrator. Expression of δ-catenin induces filopodia-like protrusions in neurons. Here we show that the small GTPases of the Rho family act coordinately as downstream effectors of δ-catenin. A dominant negative Rac preventedδ-catenin-induced protrusions, and Cdc42 activity was dramatically increased by δ-catenin expression. A kinase dead LIMK (LIM kinase) and a mutant Cofilin also prevented δ-catenin-induced protrusions. To link the effects of δ-catenin to a physiological pathway, we noted that (S)-3,5-dihydroxyphenylglycine (DHPG) activation of metabotropic glutamate receptors induced dendritic protrusions that are very similar to those induced by δ-catenin. Furthermore, δ-catenin RNA-mediated interference can block the induction of dendritic protrusions by DHPG. Interestingly, DHPG dissociated PSD-95 and N-cadherin from the δ-catenin complex, increased the association of δ-catenin with Cortactin, and induced the phosphorylation of δ-catenin within the sites that bind to these protein partners.

␦-Catenin is a component of the synaptic adherens junction that is necessary for normal learning and memory (1). In the absence of ␦-catenin, mice have severe deficits in several types of memory as well as synaptic plasticity. However, the functional basis for these deficits is not obvious, particularly because the morphological changes in ␦-catenin null mice are minimal. ␦-Catenin contains 10 Armadillo repeats (a 42-amino acid motif, originally described in the Drosophila segment polarity gene, armadillo) spaced in the characteristic arrangement of all members of this gene family which includes the prototypical member, p120 ctn , as well as p0071, ARVCF (Armadillo Repeat gene deleted in Velo-Cardio-Facial syndrome) (2), and the plako-philins, both components of the desmosome (3)(4)(5)(6). The core functions of this protein family are stabilization of cadherins by binding to a highly conserved sequence in the juxtamembrane region and regulatory coordination over Rho GTPases (7). ␦-Catenin is localized to the post-synaptic adherens junction, collaborates with Rho GTPases to set a balance between neurite elongation and branching, and robustly induces dendritic protrusions (8). Among the cadherin binding family members, ␦-catenin is the only one that is a neural-specific protein. However, ␦-catenin null mice develop normally, whereas p120 ctn can regulate synapse and spine development (9).
Given the minimal morphological changes in the brains of ␦-catenin gene-disrupted animals, the role of ␦-catenin will likely emerge with more dynamic experiments related to synaptic activation. Neuronal culture provides a means to assess an intrinsic role for the protein independent of the developmental confounders in the reported gene disruption study (1). The signaling pathway downstream of metabotropic stimulation is particularly germane because the selective group 1 mGluR agonist (S)-3,5-dihydroxyphenylglycine (DHPG) leads to calcium release from intracellular stores, local protein synthesis, and importantly, dendritic protrusive activity (18). Here we suggest that ␦-catenin lies at a hub in the metabotropic glutamatergic signaling network leading to actin cytoskeletal reorganization.

EXPERIMENTAL PROCEDURES
Cell Culture and DHPG Treatment-Primary cultures were prepared from hippocampi of embryonic day 18 Sprague-Dawley rats as described previously (19). They were maintained in Neurobasal medium (Invitrogen) containing B27 supplement (Invitrogen) and 0.5 mM L-glutamine. HEK293 cells were cultured in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum in the same atmospheric conditions. For DHPG treatment, 21 days in vitro (DIV) neurons were washed once with prewarmed Neurobasal medium. Immediately afterward, DHPG (Sigma) was added at a final concentration of 100 M, and cells were incubated for different time points at 37°C. For the phosphoprotein analysis, cortical neurons were treated with DHPG for 30 min.
DNA Preparation and cDNA Transfections-A full-length ␦-catenin cDNA was subcloned into pEGFP (Clontech, Palo Alto, CA) or into pCMVTag3B (Stratagene, La Jolla, CA) as described previously (20). pEGFP-␦-catenin* is a resistant fulllength ␦-catenin construct mutated at sites mapped to the region of Dplx1 siRNA, prepared using site-directed mutagenesis. Expression plasmids for HA-tagged LIMK1 and their mutants or for S3A Cofilin were constructed as described (21,22). The cDNA coding for Rac-dominant negative was a generous gift from Dr. Alan Hall (MRC, Laboratory for Molecular Cell Biology, London, UK). Three weeks after plating, hippocampal neurons were transfected with the Helios gene gun (Bio-Rad). To make the bullets, each cDNA was attached to gold particles. To achieve a uniform distribution of gold particles and reduce damage due to particle bombardment, mesh barriers (steel mesh plus 70-m nylon mesh) were used as a diffusion screen for dissociated culture neurons. The expression of transfected cDNAs was detected 24 -30 h later.
Immunoblot Analysis and Quantification-Proteins were transferred to polyvinylidene difluoride membrane (Bio-Rad), and ECL (Amersham Biosciences) was used for detection. Immunoblots of ␦-catenin knockdown and DHPG treatments were quantified using MetaMorph (Universal Imaging Corp., West Chester, PA) and ImageQuant software (Amersham Biosciences). ␤-Actin was used for loading normalization. Percent of inhibition was calculated relative to the mock-transfected samples. Statistical significance was calculated using Student's t test or analysis of variance.
Immunocytochemistry-Neurons were fixed with 4% paraformaldehyde and 4% sucrose in phosphate-buffered saline (PBS) for 15 min at room temperature, washed 3 times in PBS, and permeabilized for 5 min in PBS containing 0.25% Triton X-100. After washing and blocking, the cells were incubated with primary antibody for 90 min at room temperature or overnight at 4͉°C. Secondary antibodies were applied for 1 h at room temperature. Secondary antibodies conjugated to Alexa 488 and Alexa 568 were used for double labeling and Cy3-and Cy5-were used in the presence of GFP-expressing cells.
Immunoprecipitation-After each incubation, the medium was aspirated, cells were washed in ice-cold phosphate-buffered saline to stop the reaction and lysed with lysis buffer (15 mM Tris, pH 7.5, 5 mM EDTA, 2.5 mM EGTA, 1% Triton X-100, 0.1% SDS, 120 mM NaCl, 25 mM KCl, 1 mM Na3VO4, 20 mM NaF, and protease inhibitor mixture, Roche Applied Science), and protein G beads were added for preclearing for 30 min. After centrifugation, the supernatant was immunoprecipitated with the specific antibody overnight at 4͉°C, and the immunoprecipitates were captured by protein G beads or A beads (Pierce). Normal mouse or rabbit IgG were used as negative controls (Santa Cruz Biotechnology).
Phosphoprotein Sequence Analysis by Liquid Chromatography-Tandem Mass Spectrometry-Two-week-old cortical cultures were treated with vehicle control or 100-m DHPG or cine for 30 min. Cells were lysed, and ␦-catenin was immunoprecipitated using 50 -70 g of the ␦-catenin polyclonal antibody, separated on SDS-PAGE, and stained with Coomassie Blue. ␦-Catenin bands were cut from the gels and analyzed by the Taplin Biological Mass Spectrometry Facility (Harvard Medical School, Boston, MA) as follows. Gel pieces were subjected to a modified in-gel trypsin digestion procedure (24). On the day of analysis the samples were reconstituted in 5 l of HPLC solvent A (2.5% acetonitrile, 0.1% formic acid). A nanoscale reverse-phase HPLC capillary column was created by packing 5-m C18 spherical silica beads into a fused silica capillary (75-m inner diameter ϫ 12-cm length) with a flamedrawn tip (25). After equilibrating the column, each sample was pressure-loaded off-line onto the column. The column was then reattached to the HPLC system. A gradient was formed, and peptides were eluted with increasing concentrations of solvent B (97.5% acetonitrile, 0.1% formic acid).
As each peptide was eluted they were subjected to electrospray ionization, and then they entered into an LTQ linear ion-trap mass spectrometer (ThermoFinnigan, San Jose, CA). Eluting peptides were detected, isolated, and fragmented to produce a tandem mass spectrum of specific fragment ions for each peptide. Peptide sequences were determined by matching protein or translated nucleotide databases with the acquired fragmentation pattern by the software program, Sequest (ThermoFinnigan) (26). The modification of 80 mass units to serine, threonine, and tyrosine was included in the data base searches to determine phosphopeptides. Each phosphopeptide that was determined by the Sequest program was also manually inspected in ensure confidence.
Image Acquisition and Quantification-Confocal images were obtained with an Olympus Fluo-View1000 confocal laser scanning microscope. Olympus images were obtained using oil 60ϫ and 40ϫ objectives with sequential acquisition settings at a resolution of 1024 ϫ 1024 and 512 ϫ 512 pixels, respectively. Each image was a 0.5-m z-series of 7-13 images averaged 2-4 times. Confocal scanning settings of pinhole, brightness, and contrast were kept the same for all images when intensity was compared. Additional digital zoom factor of ϫ3-5 was used as well. MetaMorph software was used for morphometric analysis. Maximum projection of the z-stack images was flattened into a single image and was used for all quantitative analysis, in parallel with the z-stack image. The density of dendritic protrusions and their lengths were measured blinded in neurons transfected with GFP or GFP-␦-catenin constructs. Representative counts were confirmed by an independent observer.
DHPG/RNAi Quantification-After DHPG treatment the neurons were stained with ␦-catenin and PSD-95 antibodies. GFP was used to visualize the shape of the cell and the protrusions. 7-10 neurons of three experiments were analyzed for each condition. 2-3 segments of 50 -100-m secondary branched dendrites were counted from each neuron. All the

␦-Catenin Is a Signaling Integrator
measurements were done using the three-dimensional reconstructed image as well as the flatten image to make sure all the protrusions were counted.
The assignment of a protrusion as a spine or filopodium was based upon the criteria of Ziv and Smith (27) and Papa et al. (28) and discussion by others (29 -32). The category of 0 -3 m was predominantly represented by different types of spines (stubby, thin, and mushrooms). Mushroom-type spines were analyzed based on a thin neck and a relatively large tip (head) following these criteria; the distance of the spine head from the dendritic shaft is 0.5-2 m in length, and spine head diameter is Ͼ0.6 m in length. The category of 3-4-m protrusions occasionally contained a head-like structure of spines. Protrusions of 4 -10 m were defined as filopodialike and generally lacked heads at the tip and showed narrower and longer stalk properties than spines.
Rac1 and Cdc42 Activation Assay-Chinese hamster ovary cells were transfected with either GFP or GFP-␦-catenin, and cell homogenates were processed for Rac1 or Cdc42 activation assay according to the manufacturer's instructions (Cytoskeleton, Denver, CO). The Rac-GTP quantification was developed using an enzyme-linked immunosorbent assay according to the manufacturer's instructions (G-LISATM Rac activation assay biochem kit-BK125, Cytoskeleton, Denver, CO).

␦-Catenin Recruits Small Rho GTPases to Induce Protrusive
Activity-To understand the mechanistic basis for the observation that ␦-catenin can induce dendritic protrusions, we overexpressed ␦-catenin in mature primary cultured neurons and observed the expected increase in dendritic protrusions (Fig. 1). As shown previously (8,17,33), ␦-catenin dramatically induced protrusions. Neurons were transfected with either GFP alone or full-length GFP-␦-catenin and analyzed for the number of protrusions at 20 DIV based on GFP visualization, ␦-catenin, and PSD95 immunolocalization (Fig. 1). The elaboration of protrusions was restricted to the dendritic tree despite GFP- ␦-catenin signal throughout the neuron including the axon (Fig. 1A). We had previously linked ␦-catenin-induced protrusions and branching to RhoA inhibition (8), a finding supported by studies of ␦-catenin paralogs (34). Rac activation by ␦-catenin was demonstrated by increased Rac activity in Chinese hamster ovary cells which overexpressed ␦-catenin ( Fig. 2A,  a and b). A dominant negative Rac prevented ␦-catenin protrusive activity (Fig. 2Bi). Cdc42 activity also dramatically increased in the presence of ␦-catenin expression ( Fig. 2A, c  and d), and a dominant negative Cdc42 prevented ␦-catenin protrusive activity in 21 DIV neurons (Fig. 2C). Thus, activation of Rac and Cdc42 occurred simultaneously with Rho inhibition and collectively appeared to operate downstream of the ␦-catenin ability to induce protrusions. Our data are not consistent with Kim et al. (17), who detected no effect on Cdc42 and Rac1 activity in the presence of ␦-catenin overexpression. However, our data are consistent with many reports showing that the Rho family of small GTPases (particularly RhoA, Rac1, Cdc42) are well known regulators of the actin cytoskeleton that have profound influence on spine morphogenesis (35)(36)(37). An active form of Rac can increase the activity of LIMK1 (38). Activation of LIMK phenocopied the increased protrusive activity associated with ␦-catenin overexpression (Fig. 2B, g and h). Neurons co-transfected with ␦-catenin and an HA-tagged LIMK1-kinase dead mutant did not elaborate protrusions (Fig. 2B, c, d, and f) and Table  1). If LIMK acts downstream of ␦-catenin, one would expect to see a similar blockade of protrusive activity with mutant actin depolymerizing factor/Cofilin. Actin depolymerizing factor/Cofilin is the only substrate identified for LIMK (39) which phosphorylates Cofilin (at serine 3) when in its active phosphorylated state (phospho-LIMK1). The phosphorylation of Cofilin is an inactivation step that prevents actin binding and thereby promotes actin protrusive activity by reducing actin dynamics. S3A Cofilin is a mutant that cannot be phosphorylated by LIMK1 and, therefore, is constitutively active. S3A is, therefore, functionally equivalent to the LIMK kinase dead mutant, and it too blocked ␦-catenin-induced protrusions (Fig. 2Bj).   ␦-Catenin Is an Intermediary in Group I mGluR Activation-Interestingly, we found that Group I mGluR activation also operates through these small Rho GTPases. 5 min after DHPG stimulation, 21 DIV hippocampal neurons increased Rac and Cdc42 activity significantly (Fig. 3A). Furthermore, delivery of DHPG to neurons induced spine elongation and filopodial elaboration (18). We replicated these morphological effects of DHPG (Fig. 3, B and C). We next sought to determine whether reduced levels of ␦-catenin would interfere with the elaboration of DHPG-induced protrusions.
A variety of siRNAs were tested for their ability to knockdown ␦-catenin in heterologous cells by co-transfection of the siRNAs, GFP, and GFP-␦-catenin into HEK293 cells. Transfected proteins were immunoprecipitated with GFP antibody and analyzed by Western blot with GFP and ␦-catenin antibodies (Fig. 4A). Although all of the sequences showed some degree of knockdown of GFP-␦-catenin in HEK293 cells, sequences Dplx0 and Dplx1 were optimal (Fig. 4A). A developmental blot showed that ␦-catenin started expressing 1 day after plating (supplemental Fig. S1A), and therefore, siRNAs were introduced to neurons at three DIV in culture. The two siRNAs Dplx0 and Dplx1 suppressed ␦-catenin in neurons at 20 DIV by immunoblot (Fig. 4B) and by immunocytochemistry ( Fig. 4C and  supplemental Fig. S1B). On average, 80% of neurons showed inhibition of ␦-catenin expression. Neurons at 20 DIV, when spines are clearly present, were profoundly affected by ␦-catenin suppression (Fig. 4,  and C and D). Quantification of ␦-catenin signal indicated a 72% reduction with Dplx1 and a 35% reduction with Dplx0 (Fig. 4D). A dosage effect was observed in that the degree to which spines were reduced correlated with the degree of ␦-catenin suppression (Fig. 4D). Induction of the same phenotype with two different siRNA reduced the likelihood that the spine loss was due to an off-target effect.
Not surprisingly, ␦-catenin siRNA treatment of neurons also resulted in synapse loss and decreased complexity of the dendritic tree (see supplemental Figs. S2 and S3 and Table  SI). ␦-Catenin knockdown caused a concomitant loss of N-cadherin by immunoblot (Fig. 5A) as previously reported in the ␦-catenin gene-disrupted mouse (1). A significant loss of ␣N-catenin also occurred (Fig.  5A). Quantitatively there was significant loss of 34% of the protein level of N-cadherin and 32% of ␣N-catenin (Fig. 5B). ␦-Catenin RNAi induced a small reduction in MAP2 (data not shown) that just reached significance, perhaps due to the attenuation of the dendritic tree. The level of PSD-95 did not change by Western blot (data not shown).
To rescue the phenotypic effect of ␦-catenin inhibition on dendritic protrusions, we transfected a mutated form of GFP-␦-catenin* (resistant to Dplx1 siRNA) together with ␦-catenin siRNA Dplx1 (Fig. 5, C and D). The expression of the resistant GFP-␦-catenin* construct reversed the phenotype, and ϳ80% of the protrusions were restored using immunohistochemistry compared with the neurons co-transfected with GFP and nontargeting siRNA (Fig. 5D).
Neurons treated with ␦-catenin RNAi were unresponsive to DHPG (Fig. 6 and supplemental Fig. S4). Neurons transfected with non-targeting siRNA showed an increase in the total number and lengths of protrusions as expected after DHPG treatment (Fig. 3, B and C). In the control non-targeting siRNA sample, the mean total number of protrusions was 7.48 Ϯ 1/10 m, a value that decreased by 20% to 4.96 Ϯ   1.54/10 m (Fig. 6B, p Ͻ 0.001) after ␦-catenin RNAi. DHPG stimulation of the control non-targeting siRNA sample significantly increased total protrusions to 8.49 Ϯ 0.77/10 m (Fig. 6B, p Ͻ 0.05), but in the presence of ␦-catenin RNAi no DHPG effect occurred; the mean total number of protrusions remained similar to siRNA treatment alone (5.07 Ϯ 0.88/10 m).

siRNA N o n -T a r g e ti n g S M A R T P o o l D p lx 1 D p lx 2 D p lx 3 D p lx 4 D p lx
A more detailed analysis of the dendritic protrusion categories based on length and shape (see "Experimental Procedures") revealed more specific effects. Protrusions of 0 -3 m represent mostly mature dendritic spines, and they account for the majority (94%) of the total protrusions under control conditions (Fig.  6B). ␦-Catenin silencing affected this population dramatically; the number of protrusions in this category decreased by 46% (from 7.05 Ϯ 0.63 to 2.8 Ϯ 1.31/10 m, p Ͻ 0.001, Fig. 6B). Mushroom-type spines were measured as part of the 0 -3-m category, and they showed a significant loss in the ␦-catenin inhibited neurons (from 3.58 Ϯ 0.47 to 1.75 Ϯ 0.6/10 m, p Ͻ 0.001, Fig.  6B). Interestingly, the numbers of the protrusions in the 3-4-and 4 -10-m categories increased after ␦-catenin inhibition, probably due to loss of N-cadherin stability in the absence of ␦-catenin (40,41). Protrusions in these categories, especially the 4 -10-m category, resemble immature spines seen in developing neuronal cultures. Under control conditions, DHPG mainly increased the numbers of protrusions in the length categories of 3-4 m (from 0.31 Ϯ 0.28 to 1.01 Ϯ 0.46, p Ͻ 0.001) and the 4 -10 m (from 0.07 Ϯ 0.16 to 0.59 Ϯ 0.25, p Ͻ 0.001, Fig. 6B). This increase did not occur under ␦-catenin silencing conditions (Fig. 6B). Taken together, DHPG failed to induce its expected effects on dendritic protrusions in the presence of ␦-catenin knockdown (Fig. 6).
mGluR Activation Reconfigures ␦-Catenin Binding Partners and Induces ␦-Catenin Phosphorylation-To search for the effects of mGluR stimulation on ␦-catenin, 21 DIV hippocampal neurons were treated with DHPG, and changes in ␦-catenin binding partners were probed. After DHPG incubation, ␦-catenin was immunoprecipitated, and the precipitated pellets were blotted with antibodies against the ␦-catenin interactor proteins, Cortactin, PSD-95, and N-cadherin. DHPG induced a reconfiguration of the ␦-catenin complex over a similar time frame as the elaboration of increased dendritic protrusions. After treatment with DHPG for 20 min, co-immunoprecipitated Cortactin increased, whereas co-immunoprecipitated N-cadherin and PSD-95 decreased (Fig. 7).
One mechanism by which mGluR activation may act upon ␦-catenin is by inducing its phosphorylation. ␦-Catenin was immunoprecipitated from cultured rat cortical neurons either with or without DHPG stimulation, and mass spectroscopy identified phosphorylation sites with a DHPG-induced on-off state ( Fig. 8 and supplemental Table SII). Most of these sites lie in interaction domains with ␦-catenin binding partners. These  NOVEMBER 21, 2008 • VOLUME 283 • NUMBER 47

JOURNAL OF BIOLOGICAL CHEMISTRY 32787
phosphorylation sites include Ser-693 in the fourth Arm repeat, which interacts with classical cadherins (42), Ser-1242 at the penultimate serine related to PDZ interactions, and Ser-1063/ Thr-1064 and Ser-1098 two sites between the Arm repeats and the PDZ ligand sequence where Cortactin binds (8). These sites may regulate the composition of the ␦-catenin complex. No sites of DHPG-induced tyrosine phosphorylation were identified in any of these binding domains, consistent with previous observations that tyrosine phosphorylation induced a dissociation between Cortactin and ␦-catenin (8), and in these experiments with metabotropic stimulation, the binding to Cortactin increased (Fig. 7). Recently, phosphorylation at Thr-454 was shown to sequester p190RhoGEF (17). However, a tryptic peptide containing Thr-454 was recovered and was constitutively phosphorylated on Ser-457 or Ser-458, but not on Thr-454, indicating that DHPG did not induce a detectable phosphorylation at this site (supplemental Table SII). The variety of sequence contexts observed among these ␦-catenin phosphorylation sites suggests that several kinases collectively contribute to DHPG-induced ␦-catenin phosphorylation.

DISCUSSION
The results described here link ␦-catenin to group 1 mGluR activation upstream and to the Rho family GTPases downstream. The Rho family of small GTPases (particularly RhoA, Rac1, Cdc42) are well known regulators of the actin cytoskeleton that have profound influence on spine morphogenesis (35). RhoA inhibits, whereas Rac and Cdc42 promote, the growth and/or stability of dendritic spines. Consistent with the cellular phenotype induced by ␦-catenin, expression of ␦-catenin inhibits RhoA (8) by decreasing the binding between p190RhoGEF and RhoA, resulting in lower levels of GTP-RhoA (17). A similar mechanism of RhoA inhibition may operate via p120 ctn (43)(44)(45) in its role in organizing actin into filaments. Here we report that ␦-catenin also activates Rac and Cdc42, a finding that follows from the decreased Rho activity. A further parallel with p120 ctn is the observation that p120 ctn loss resulted in mis-regulation of Rho-family GTPases, with decreased Rac1 and increased RhoA activity (9). Farther downstream, ␦-catenin expression is associated with phosphorylation of LIMK and Cofilin. Consistent with these findings are other observations that LIMK knock-out mice have short and smaller dendritic spines and, like ␦-catenin knock-out mice, have enhanced hippocampal longterm potentiation (46). After exploratory activity in vehicle-treated rats compared with rats treated with the N-methyl-D-aspartate receptor antagonist (ϩ/Ϫ)-3-(2-carboxypiperazin-4-yl)propyl-1-phosphonic acid (CPP) dense phosphorylated Cofilin immunoreactivity appeared in enlarged spines of hippocampal field CA1 (47). The question remained as to what upstream stimuli trigger the pathway that links ␦-catenin to the small Rho GTPases. The type I mGluR activation was a logical receptor to study because stimulation of this receptor by DHPG induced dendritic protrusions that resembled those of ␦-catenin. Furthermore, DHPG stimulation, like ␦-catenin expression, also increased Rac and Cdc42 activity in 21 DIV hippocampal neurons (Fig. 3A). We found that activation of the type I mGluR receptors by DHPG dissociated ␦-catenin from PSD-95 and from N-cadherin while increasing its association with Cortactin (Fig. 7). mGluR activation also induced ␦-catenin phosphorylation at the proteinprotein interaction sites where PSD-95, N-cadherin, and Cortactin bind. Thus, dissociation of ␦-catenin from the post-synaptic scaffold releases a latent tendency to elaborate filopodia by multiple pathways related to both small Rho GTPases and Cortactin (Fig.  9). Previous studies have shown that Rac is locally activated in dendritic spines (48), and Cdc42 is a key module in dendritic spine formation (49). In older neurons with established synapses, Cortactin knockdown results in depletion of dendritic spines, and in response to synaptic stimulation and Nmethyl-D-aspartate receptor activation, Cortactin redistributes rapidly from spines to the dendritic shaft (50 FIGURE 9. The ␦-catenin pathway. The model indicates the series of ␦-catenin-related events that occur after group I mGluR activation. The reconfiguration of the ␦-catenin complex is indicated by green arrows, and a networked pathway of activation and inhibitory signals that lead to actin polymerization and increased dendritic protrusions are shown. DHPG-induced release of ␦-catenin from its submembranous partners N-cadherin and PSD-95 may set up a collaboration with Cortactin, Rho, Rac, Cdc42, LIMK, and Cofilin to reorganize actin. ROCK, Rho kinase; PAK, p21-activated kinase. NOVEMBER 21, 2008 • VOLUME 283 • NUMBER 47 JOURNAL OF BIOLOGICAL CHEMISTRY 32789 ␦-catenin binding partners, and a shift in its functional localization may induce filopodia (Fig. 9).

␦-Catenin Is a Signaling Integrator
Finally, RNAi knockdown of ␦-catenin prevented DHPG from inducing filopodia. Interestingly, reducing ␦-catenin had a unique phenotype that had some superficial parallels with ␦-catenin overexpression. In both cases, filopodial elaboration occurred. However, the magnitude of the response and, more importantly the qualitative nature, differed greatly. Reduced ␦-catenin resulted in fewer filopodia than overexpression, and the filopodia that did emerge were restricted to a size category consistent with immature protrusions (Fig. 6). We suggest that in contrast to ␦-catenin overexpression, which likely operates through small Rho GTPases and Cortactin, ␦-catenin suppression likely operates by destabilization of cadherin and the initiation of filopodia emergence reported under these conditions (40,41,51). In fact, we did observe reduced N-cadherin in the presence of ␦-catenin RNAi (Fig. 5).
Once dissociated from its submembranous partners, ␦-catenin collaborates with Rac, LIMK, Cofilin, and Cortactin in convergent pathways that assemble actin filaments (Fig. 9). Together these findings reveal a signaling pathway that captures a broad network of proteins to coordinate the precise effects on actin dynamics needed for plasticity. These coordinated effects may establish the antagonistic relationship between Rac activation and Rho inhibition that leads to protrusive activity (52). However, the picture is by no means complete. Additional proteins such as Cordon-Bleu (53) are likely to serve as actin nucleators to induce the unbranched filaments observed in filopodia.
The deletion of ␦-catenin in the genetic disorder cri du chat (54), the up-regulation of ␦-catenin in the lateral hypothalamus with cocaine addiction (55) and its up-regulation in prostatic carcinoma cells (56) as well as the interaction of ␦-catenin with the Alzheimer gene presenilin (57), whose ␥-secretase activity can disassemble the adherens junction (58), all support a role for this protein in several human pathologies. Several genes in the ␦-catenin pathway are associated with mental retardation. Williams syndrome is due to a deletion in chromosome 7q11.23, which includes LIMK1 in the critical region (59), and PAK3 mutations cause a non-syndromic mental retardation (60). Finally, ␦-catenin may be relevant to fragile X syndrome, which has been linked to deficits in group I mGluR signaling (61,62) based upon the presence of long thin spines (63,64) and the interaction of fragile X mental retardation protein (FMRP) with p21-activated kinase (PAK) (65).