Molecular determinants for PICK1 synaptic aggregation and mGluR7a receptor coclustering: role of the PDZ, coiled-coil, and acidic domains.

PSD-95/Disc-large/ZO-1 (PDZ) domain-containing proteins play a central role in synaptic organization by their involvement in neurotransmitter receptor clustering and signaling complex assembly. The protein interacting with protein kinase C (PICK1), a synaptic PDZ domain protein that also contains a coiled-coil and acidic domain, binds to several synaptic components including the metabotropic glutamate receptor mGluR7a. Coexpression of PICK1 and mGluR7a in heterologous cells induces coclustering of these two proteins. To examine the role of the different structural motifs of PICK1 in synaptic aggregation of PICK1 and mGluR7a coclustering, several PICK1 mutants were generated to analyze their distribution in transfected hippocampal cultured neurons and to test their ability to induce coclusters with mGluR7a when coexpressed in fibroblast cells. The PDZ and coiled-coil domains are both required, whereas the acidic region plays an inhibitory role in these processes. Our data suggest that synaptic aggregation and receptor coclustering depend on PICK1 binding to a target membrane receptor, e.g. mGluR7a, by a PDZ-mediated interaction and on PICK1 oligomerization through the coiled-coil domain. This study defined three structural signals within PICK1 regulating its synaptic localization and receptor coclustering activity, which could represent molecular substrates involved in synaptic development and plasticity.

A characteristic and important feature of central nervous system synapses is the clustering of receptors and signaling molecules at both postsynaptic and presynaptic sites, which underlies an efficient synaptic transmission in the brain. For instance, at glutamate excitatory synapses of the hippocampus, the ionotropic glutamate N-methyl-D-aspartate and ␣-amino-3hydroxy-5-methylsoxazole-4-propionate (AMPA) 1 receptors are both concentrated at postsynaptic sites (1), whereas the G protein-coupled metabotropic glutamate receptor mGluR7a is specifically targeted to the presynaptic active zone of nerve terminals (2). Recent studies showed the importance of PDZ domain-containing proteins in mediating the specific synaptic aggregation of neurotransmitter receptors and in regulating receptor signaling at synapses (3)(4)(5). Among them, protein interacting with protein kinase C (PICK1) is a PDZ domain protein originally cloned on the basis of its interaction with protein kinase C (PKC) by a yeast two-hybrid screen (6). PICK1 has subsequently been shown to interact with the Eph receptor tyrosine kinases and ephrin-B ligands (7), GluR2/3/4c subunits of the AMPA receptor (8,9), mGluR7a receptor (10,11), and dopamine and norepinephrine transporters (12).
Functionally, PICK1 can induce coclustering of the AMPA and mGluR7a receptors in heterologous cells, suggesting a role of PICK1 in the synaptic clustering of these receptors (8 -11). Indeed, disruption of the interaction between PICK1 and mGluR7a by deletion of the PICK1 binding domain in mGluR7a abolished the presynaptic clustering of recombinant mGluR7a in hippocampal neurons while retaining its axonal targeting (10). However, PICK1 might not be primarily involved in the synaptic clustering or stabilization of AMPA receptors (13) but rather was hypothesized to be involved in AMPA receptor internalization (14,15). Blocking interaction between PICK1 and GluR2 by the infusion of competitive peptides in cultured cerebellar Purkinje cells attenuated long term depression induction (14), a process recently shown to depend on internalization events (16 -18). However, blocking PICK1-GluR2 interaction had no effect on long term depression in hippocampal neurons (19). PICK1 interaction with the dopamine transporter has been implicated in the proper targeting of the latter in midbrain dopaminergic neurons (12). Thus, PICK1 could have different functions depending on the cerebral region and the specific complement and nature of PICK1 binding partners. In accordance with its biological effects, immunocytochemical data demonstrated that PICK1 is concentrated at excitatory synapses in hippocampal cultured neurons (7,8,10). Moreover, the presence of PICK1 immunoreactivity in synaptic vesicle fractions as well as in postsynaptic density fractions (7,8) strongly suggests that PICK1 is expressed in both presynaptic and postsynaptic domains. PICK1 distribution has been shown to be dynamically regulated upon PKC activation, which triggered mobilization of PICK1 to excitatory synaptic sites (15).
The molecular determinants within PICK1 important for synaptic clustering and receptor coclustering are unknown. Although the three-dimensional structure of PICK1 is not known, its primary structure contains motifs that predict the existence of different structural domains including a PDZ domain, a coiled-coil domain, and an acidic region constituted of a stretch of glutamic and aspartic acid residues. PDZ and coiled-coil domains are typically involved in protein-protein interaction and/or protein dimerization (20 -22). Indeed, the * This work was supported by the Pew Charitable Trust and National Institutes of Health Grant NS39286 (to A. M. C.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  putative carboxylate binding loop in the PICK1 PDZ domain has been shown to bind the C terminus of PKC, GluR2/3, and mGluR7a (8 -11, 23). To examine the role of the different structural domains of PICK1 in the mechanisms underlying PICK1 synaptic localization and mGluR7a coclustering, several Myc epitope-tagged mutant PICK1 constructs were generated to analyze their distribution in transfected hippocampal cultured neurons. In addition, their ability to induce coclusters with mGluR7a was tested when coexpressed in fibroblast cells. We show that the three domains correspond to functional signals for PICK1 synaptic localization and mGluR7a coclustering; the PDZ and coiled-coil domains are both required, whereas the acidic region plays an inhibitory role in these processes. Our data suggest that synaptic aggregation and receptor coclustering depend on PICK1 binding to a target membrane receptor, e.g. mGluR7a, by a PDZ-mediated interaction and on PICK1 oligomerization through the coiled-coil domain.

EXPERIMENTAL PROCEDURES
cDNA Constructs-All the PICK1 cDNA constructs used for mammalian cell expression described in this study contained a Myc-epitope tag of 12 amino acids originally inserted into the pRK5 vector (modified from Genentech) to produce N-terminally tagged proteins. PICK1-WT and PICK1 mutated in its PDZ domain (mutation of Lys 27 and Asp 28 to alanines: PICK1-KD/AA) in pPC86 (8) were subcloned into the SalI/ NotI sites of the Myc-pRK5 vector. PICK1 deleted of its coiled-coil domain (PICK1-⌬CC) and of its acidic region (PICK1-⌬DE) were obtained by a two-step polymerase chain reaction strategy using appropriate primers, and the resulting polymerase chain reaction products were inserted into the SalI/Eco47III sites and Eco47III/NotI sites of Myc-pRK5-PICK1-WT for PICK1-⌬CC and PICK1-⌬DE construction, respectively. These mutants were then subcloned into pPC97 and pPC86 (24) for yeast expression. Several other constructs were generated for the yeast two-hybrid assay. Fragments corresponding to each domain of PICK1 were synthesized by polymerase chain reaction amplification to generate the PDZ (encoding amino acids 20 -110), PDZ-KD/AA (encoding amino acids 20 -110 with Lys 27 and Asp 28 substituted to alanines), coiled-coil (CC, encoding amino acids 139 -166), and acidic (DE, encoding amino acids 380 -390) fragments. All of these fragments were inserted into the SalI/NotI sites of pPC97. Construction of GW1-mGluR7a and pPC97-CT7 (encoding the C-terminal tail of mGluR7a) was described previously (10).
Neuron Culture and Transfection-Briefly, embryonic day-18 rat hippocampi were dissected and dissociated by trypsin and trituration. Before plating, neurons were transfected by a lipid-mediated gene transfer method using the Effectene kit (Qiagen). Approximately 4 ϫ 10 5 cells were incubated with 1 g of DNA for 2 h at 37°C in the presence of 8 l of Enhancer and 15 l of Effectene transfection reagents and were then plated onto poly-L-lysine-coated coverslips in fresh medium at a density of 14,000 cells/cm 2 . Cells were maintained in serum-free N2.1 medium suspended above a glial feeder layer cultured from newborn rat forebrain (25). For transfection after plating, 9 days in vitro (DIV) neurons on coverslips were removed from the culture dish and incubated in the transfection mixture containing the plasmid DNA for 90 min at 37°C. Coverslips were then placed back into the culture dish. At 15 DIV, the cells were fixed to proceed to the immunostaining.
Neuron Immunostaining and Quantification-Neurons were fixed in 4% paraformaldehyde, 4% sucrose in PBS for 15 min and permeabilized for 5 min in 0.25% Triton X-100. For experiments involving N-methyl-D-aspartate receptor NR1 immunostaining, neurons were fixed in methanol for 10 min at Ϫ20°C. The cells were blocked for 30 min in 10% bovine serum albumin in PBS and incubated for 2 h at 37°C in primary antibodies diluted in PBS containing 3% bovine serum albumin. The primary antibodies used were as follows: mouse monoclonal anti-Myc (clone 9E10) and anti-GAD6 were used at 1:40 and 1:2 dilutions, respectively (Developmental Studies Hybridoma Bank); the rabbit antisynapsin antibody was used at 1:10,000 (Sigma); the mouse anti-NR1 was used at a concentration of 0.1-0.3 g/ml (PharMingen); and the rabbit anti-PICK1 (1:400) and rabbit anti-GluR1 (1:2000) were generous gifts from Dr. R. Huganir (Johns Hopkins University, Baltimore, MD) and Dr. R. McIlhinney (Oxford, United Kingdom), respectively. The cells were then incubated with appropriate Texas Red-or fluorescein isothiocyanate-conjugated secondary antibodies (Jackson Laboratories), and the coverslips were mounted on glass slides for microscopy analysis. Each coverslip was systematically scanned on a Zeiss Axios-kop with an ϫ25 lens, and images of each transfected cell were acquired with a ϫ 63 1.4 numerical aperture objective using a Photometrics Sensys cooled charge-coupled device camera and Metamorph software. The colocalization of recombinant Myc-PICK1-WT with synapsin, GluR1, and NR1 as well as PICK1-KD/AA, PICK1-⌬CC, and PICK1-⌬DE with synapsin was quantified using Metamorph software as described previously (26). The fluorescence intensity and size of the endogenous and recombinant Myc-PICK1-WT clusters were quantified on 60 and 150 clusters, respectively, using Metamorph software. For each Myc-PICK1 construct, between 30 and 60 transfected neurons were analyzed from 5 independent experiments. The images for presentation were prepared for printing with Adobe Photoshop.
CV1 Cell Transfection and Immunostaining-Fibroblast CV1 cells were cotransfected with mGluR7a-GW1 and Myc-PICK1-WT-pRK5 or Myc-PICK1-KD/AA-pRK5, or Myc-PICK1-⌬CC-pRK5, or Myc-PICK1-⌬DE-pRK5 cDNAs using the Effectene kit (Qiagen). After 48 h of transfection, the cells were fixed in 4% paraformaldehyde/4% sucrose in PBS for 15 min, permeabilized for 5 min in 0.25% Triton X-100 in PBS, blocked for 30 min in 10% bovine serum albumin in PBS, and incubated for 2 h at 37°C in a mixture of rabbit anti-mGluR7a 1:2000 (generously provided by Dr. R. Shigemoto, Japan) and mouse anti-Myc 1:40 antibodies. The cells were then incubated with the appropriate secondary antibodies, and the coverslips were mounted on glass slides for microscopy analysis. For each experiment (n ϭ 3), coverslips were scanned, and all cotransfected cells were scored as exhibiting a diffuse or clustered mGluR7a/Myc-PICK1 staining.
Yeast Two-hybrid Assay-To test the interaction between two protein constructs, the corresponding cDNAs were cotransformed in the yeast strain PJ69-4A harboring the HIS3, ADE2, and lacZ reporter genes (27), and positive clones were selected on plates lacking leucine, tryptophan, and histidine. The positive clones were further confirmed by growing in a medium lacking leucine, tryptophan, histidine, and adenine and were then tested in four independent experiments for ␤-galactosidase activity using a colorimetric assay (28). Positive clones in this assay were selected by a ␤-galactosidase activity of at least five times above the control values obtained from untransformed PJ69-4A yeast.
The different PICK1 constructs PICK1-KD/AA, PICK1-⌬CC, and PICK1-⌬DE cloned into pPC97 (which contains the GAL4 binding domain) were tested for their interaction with PICK1-WT and the C-terminal tail of mGluR7a (CT7), both cloned into pPC86 (which contains the GAL4 DNA activation domain). Interaction of specific fragments of PICK1 inserted in pPC97 (PDZ, PDZ-KD/AA, CC, and DE fragments) were also tested with PICK1-WT in pPC86.

Recombinant Myc-PICK1 Is Selectively Clustered at Excitatory Synapses in Cultured Hippocampal
Neurons-To determine which domain of PICK1 is required for the synaptic localization and clustering properties, several Myc epitopetagged wild-type and mutant protein constructs of PICK1 were generated and expressed in hippocampal cultured neurons by a lipid-mediated transfection. Initial experiments were conducted to determine whether the exogenously expressed recombinant wild-type Myc-PICK1 localized appropriately in neurons. Neurons were transfected either the day of plating or at 9 DIV. The distribution of the recombinant Myc-PICK1 was analyzed at 15 DIV by immunofluorescence using a Myc monoclonal antibody. Myc-PICK1 expressed by transfection was distributed in clusters along dendrites and, although to a lesser extent, along axons as well (Fig. 1). In addition, some diffuse labeling was often detected, most prominently in large dendrites close to the cell body. Double labeling with the synaptic marker synapsin showed that most of the PICK1-containing clusters in both dendrites and axons were synaptic (Fig. 1). A similar distribution was observed in transfected pyramidal cells and interneurons, which represent the two types of cells present in this hippocampal culture model (differentiated based on both morphological features and glutamic acid decarboxylase immunostaining), suggesting no cell type-specific targeting of PICK1. No difference in either Myc-PICK1 distribution or intensity of immunoreactivity was observed between neurons transfected the day of plating or at 9 DIV. To compare the PICK1 expression level between transfected and untransfected neurons, quantitative analyses were performed in the same culture after PICK1 staining to label endogenous as well as exogenous PICK1. Although the intensity of PICK1 immunoreactivity was clearly much higher in a few neurons, likely corresponding to the transfected ones, double labeling with the Myc antibody was performed to confirm that these brighter PICK1-expressing neurons corresponded to the Myc-PICK1transfected neurons. The fluorescence intensity and size of the recombinant Myc-PICK1 clusters determined by PICK1 immunostaining were on average four and six times higher, respectively, than for the endogenous protein.
To determine whether the transfected PICK1 was selectively associated with excitatory synapses, as reported for the endogenous protein, double immunostainings were performed with markers for excitatory and inhibitory synapses (Fig. 2). Double immunolocalization of recombinant Myc-PICK1 and the GluR1 subunit of the AMPA receptor showed a good overlap of the two proteins, although some clusters of Myc-PICK1 lacked concentrations of GluR1 and vice versa. Quantitative analysis indicated that 65% of transfected Myc-PICK1 clusters colocalized with GluR1. Similarly, 54% of recombinant PICK1 clusters were associated with the N-methyl-D-aspartate receptor subunit NR1-positive synapses. By contrast, no colocalization was observed between PICK1-positive clusters and glutamic acid decarboxylase (GAD)-positive synapses, indicating that recombinant PICK1 was specifically targeted to excitatory synapses and excluded from inhibitory ones. These data indicate that even with a high level of protein expression, the recombinant Myc-PICK1 expressed by transfection in hippocampal cultured neurons was specifically concentrated at a subset of excitatory synapses. Therefore, this transfected neuron system could be exploited to elucidate the function of the different structural domains of PICK1 in synaptic clustering properties.
Both the PDZ and Coiled-coil Domains Are Necessary for PICK1 Synaptic Clustering and for mGluR7a Receptor Coclustering-Analysis of the PICK1 amino acid sequence suggests the presence of three distinct structural domains: a PDZ domain (amino acids 20 -110), a coiled-coil domain (amino acids 139 -166), and an acidic region containing 10 acidic amino acids, either aspartic or glutamic acid residues (amino acids 380 -390) (8,23). To determine the involvement of each of these domains in the synaptic localization of PICK1, several constructs of PICK1 were generated by deletion or mutation, and the resulting PICK1 mutants were expressed by transfection in hippocampal-cultured neurons to study their distribution by immunofluorescence. A mutant of the putative carboxylatebinding loop of the PDZ domain was generated by substitution of the Lys-Asp residues into Ala-Ala residues (PICK1-KD/AA). This mutation has been shown to disrupt the interaction of PICK1 with protein kinase C (23), the AMPA glutamate receptor subunits GluR2/3 (8,9), and the metabotropic glutamate receptor mGluR7a (10,11). Mutants of the coiled-coil domain and the acidic region were constructed by deletion of the whole coiled-coil domain (PICK1-⌬CC) or acidic region (PICK1-⌬DE), respectively. Similar to what was observed for the wild-type protein, all the mutated PICK1 variants were expressed in both dendrites and axons without any detectable differences. However, marked differences were observed with regard to their synaptic clustering ability (Fig. 3). The mutants PICK1-KD/AA and PICK1-⌬CC did not form clusters but were diffusely distributed throughout the processes, although GluR1, NR1, and synapsin staining remained unaffected, suggesting that the synapses were normal (not shown). To compare the ability to form synaptic clusters among the different Myc-PICK1 constructs, quantitative analysis was performed on neurons double-labeled with the synaptic marker synapsin. The number of PICK1 synaptic clusters was recorded and normalized for 100 m of dendrites (Fig. 4). PICK1-WT formed an average of 13 synaptic clusters/100 m of dendrites, whereas only 3 synaptic clusters/100 m of dendrites were observed in neurons transfected with PICK1-KD/AA or PICK1-⌬CC. The mutant PICK1-⌬DE showed a significant increase in the number of synaptic clusters compared with the wild-type protein, with an average of 21 clusters/100 m of dendrites. Importantly, the average Myc-PICK1 immunofluorescence intensity measured in neuronal cell bodies was similar for the various constructs (Fig. 4), indicating that the differential distribution observed for the PICK1 mutants was not caused by a different protein expression level but rather to a different targeting mechanism. These data indicate that both the PDZ and coiled-coil domains are critical for PICK1 synaptic clustering and that the acidic region plays an inhibitory role in this process.
We previously reported that PICK1 and the mGluR7a receptor expressed alone in fibroblast cells were distributed diffusely throughout the cell but that coexpression of PICK1 and mGluR7a receptor induced coclustering of these two proteins (10). To test whether the structural domains of PICK1, which are important for the synaptic clustering in neurons, were also functionally important for the mGluR7a coclustering activity, we analyzed the coclustering efficacy for each Myc-PICK1 mutant coexpressed with mGluR7a in fibroblast CV1 cells (Fig. 5). The number of cells exhibiting coclustering was expressed as a percentage of total cotransfected cells. Both PICK1-WT and PICK1-⌬DE induced mGluR7a receptor coclustering, with a significantly more pronounced effect for PICK1-⌬DE. By contrast, the mutation of the PDZ domain in PICK1-KD/AA and the deletion of the coiled-coil domain in PICK1-⌬CC completely abolished mGluR7a receptor coclustering. These data demonstrate the importance of the PDZ and coiled-coil domains in mGluR7a coclustering efficacy and establish the existence of common features between PICK1 synaptic clustering and mGluR7a coclustering activity.
The PDZ and Coiled-coil Domains Are Involved in PICK1 Self-association-The data presented above showed that the three distinct structural domains of PICK1 play an important role in the synaptic localization and coclustering activity. It has been shown previously that PICK1 possessed the ability to Double labeling with GluR1 or NR1 shows that Myc-PICK1 immunoreactivity colocalizes well with these receptor subunits, which are only expressed at excitatory postsynaptic sites. NR1 staining associated with dendrites from an untransfected cell within the same field can be seen. Double labeling with GAD, a marker for GABAergic inhibitory synapses, shows no overlap between Myc-PICK1 immunoreactivity and GADcontaining nerve terminals, indicating that Myc-PICK1 is excluded from inhibitory synapses. Scale bar, 10 m. tion of PICK1, several PICK1 constructs were generated to test their ability to bind PICK1 in the yeast two-hybrid system. First, PICK1-KD/AA, PICK1-⌬CC, and PICK1-⌬DE were used as baits to test their interaction with PICK1-WT. All these constructs efficiently interacted with PICK1-WT in the yeast two-hybrid system (Fig. 6). An explanation for these results is that more than one domain could bind to PICK1 and that mutating or deleting only one of them was not sufficient to disrupt PICK1 self-association. To test this possibility, we asked whether each fragment corresponding to the PDZ, coiledcoil, or acidic domain fused to the GAL4 binding domain could bind to PICK1 fused to the GAL4 activation domain in the yeast two-hybrid system. Both the PDZ and coiled-coil domain showed an interaction with PICK1, whereas the acidic domain showed no detectable interaction (Fig. 6). It has been shown previously that the PDZ mutant form of PICK1-KD/AA was no longer able to interact with the PICK1-binding partners PKC (23), GluR2/3 (8), and mGluR7a (10,11). To examine whether this mutation also interfered with the ability of the PDZ domain to mediate PICK1 self-association, we tested the binding of the fragment PDZ-KD/AA to PICK1 in the yeast two-hybrid assay. An interaction was found, suggesting that the PICK1 PDZ domain contains two distinct modules involved in proteinprotein interaction, one involved in the self-association and the other in binding other proteins. To examine whether PICK1 mutants could still associate with the mGluR7a receptor, we tested the binding of the C-terminal tail of mGluR7a (CT7) fused to the GAL4 activation domain with the different PICK1 constructs fused to the GAL4 DNA binding domain. As shown in Fig. 6, PICK1-⌬CC and PICK1-⌬DE interacted as efficiently as PICK1-WT with CT7. By contrast, PICK1-KD/AA did not interact with CT7. DISCUSSION This study shows that the three structural domains of PICK1 deduced from its primary sequence played an important role in PICK1 synaptic aggregation and mGluR7a receptor coclustering activity. PICK1 mutated in the PDZ domain or lacking the coiled-coil domain was no longer localized at synapses and lost the ability to cluster the mGluR7a receptor. We found that mutation of the predicted carboxylate binding loop of the PDZ domain abolished the direct interaction with mGluR7a recep- tor, while the ability of PICK1 to self-associate was maintained. On the other hand, deletion of the coiled-coil domain suppressed the other potential site of PICK1 self-association but did not affect binding with mGluR7a. These results suggest that synapse localization and receptor coclustering depend on both PICK1 self-association through the coiled-coil domain and binding to a target membrane receptor, e.g. the mGluR7a receptor, by a PDZ-mediated interaction. In addition, we showed that the acidic region of PICK1 plays an inhibitory role in PICK1 clustering properties.
The PDZ mutant protein PICK1-KD/AA was not able to cocluster mGluR7a in heterologous cells or to localize at synapses in hippocampal neurons. Because this PICK1 mutant form did not bind mGluR7a (10,11), the absence of coclustering activity was expected. The impairment of synaptic clustering in neurons was more surprising. Given that mutation of the two amino acid residues Lys 27 -Asp 28 to Ala-Ala has been shown to specifically disrupt the interaction between PICK1 and several of its presynaptic and postsynaptic target proteins such as GluR2/3 (8,9) and mGluR7a (10,11), the simplest explanation for the diffuse neuronal distribution of the PICK1 PDZ mutant is that the target receptor plays an active role in the synaptic recruitment and/or stabilization of PICK1. In Drosophila photoreceptors, the PDZ protein INAD acts as a scaffold protein to assemble a signaling complex involved in the phototransduction initiated by light activation of rhodopsin (29). This signaling complex contains phospholipase C, the transient receptor potential (TRP) ion channel, and eye-PKC and is discretely localized in rhabdomeres of the photoreceptor cell. The proper distribution of this complex depends on INAD, as demonstrated in inad null mutant flies (29). In turn, the rhabdomeric localization of INAD depends on the interaction with TRP as demonstrated in trp null or trp mutant flies (30,31). Thus, INAD is required for anchoring the signaling complex in the rhabdomeres, but in turn the INAD target-interacting protein TRP is required to localize INAD to the rhabdomeres. It has also been shown in transfected neuron cultures that the PDZ domain protein PSD-95 could be localized in axons as a result of interactions with the membrane protein shaker-type K ϩ channel Kv1.4 (32) and that transfected Kv1.4 also depended on the interaction with PSD-95 to be selectively localized in axons (32) and in postsynaptic sites (33). Our data suggest the existence of a similar dual dependence to properly localize the PDZ protein PICK1 and its target binding proteins. We previously showed that the mGluR7a receptor lacking the PICK1 PDZ binding site failed to localize at presynaptic sites (10). Conversely, our present data show that the PDZ mutant PICK1 lacking the inter-action site with its binding partners mGluR7a and GluR2/3 was no longer concentrated at presynaptic or postsynaptic sites. This interdependent relationship could enhance PICK1receptor complex stability at synaptic sites and could prevent the build-up of excess PICK1 molecules at synapses without bound receptor.
The deletion of the coiled-coil domain also abolished the synaptic localization of PICK1. Coiled-coil domains correspond to peptidic motifs mediating protein-protein interaction typically involved in dimerization of a wide variety of proteins (20,21). Here we showed by a yeast two-hybrid assay that the PICK1 coiled-coil domain can mediate PICK1 self-association. Our data suggest thus that PICK1 homooligomerization through the coiled-coil domain is central to the mechanism of both synaptic aggregation and receptor coclustering in fibroblast cells. Dimerization or multimerization of other proteins interacting with membrane receptors or channels has been shown to play a critical role in their clustering functions. For example, rapsyn-induced nicotinic acetylcholine receptor aggregation in heterologous systems required rapsyn self-association through tetratricopeptide repeat domains (34). Similarly, the multimerization of PSD-95 by its N-terminal part was necessary to induce K ϩ channel Kv1.4 clustering in heterologous cells (35).
In addition to the coiled-coil domain, we found that the PDZ domain of PICK1 can also mediate self-association. This finding is similar to the self-association and heteromultimerization of GRIP and the related protein ABP via a subset of their PDZ domains (36) but is more surprising for the single PDZ domain protein PICK1. The PDZ domain-mediated self-association of PICK1 was not abolished by mutations that abolish receptor ligand binding. However, the requirement of the coiled-coil domain for coclustering of PICK1 with mGluR7a suggests that binding of the PDZ domain to a receptor ligand renders this domain unable to simultaneously mediate self-association. Thus, it is tempting to speculate that PDZ domain-mediated self-association might be a mechanism to regulate availability of the PDZ domain to a receptor ligand.
Contrasting with what was observed for the PDZ and coiledcoil domains, deletion of the acidic region located in the Cterminal part of PICK1 notably increased the extent of PICK1 synaptic aggregation as well as its ability to cluster mGluR7a. Thus, this acidic region played a negative regulatory role in these processes. Because acidic amino acid sequences have been found in some proteins to be involved in inter-and intramolecular interactions to modulate the function of proteins (37,38), we tested whether the fragment containing the acidic region could bind to PICK1 in the yeast two-hybrid system. No interaction was evidenced, suggesting that the mechanism by which the acidic region affected PICK1 localization and mGluR7a coclustering did not involve intramolecular interactions. However, we can not rule out the participation of yet unknown proteins in intermolecular interactions with the PICK1 acidic domain. An attractive possibility is that the PICK1 acidic region may directly bind Ca 2ϩ . Stretches of aspartic and glutamic acid residues have been shown in several other proteins such as calreticulin (39) and calsequestrin (40) to bind Ca 2ϩ . Ca 2ϩ binding can lead to conformational changes and modulation of interactions with protein partners, as demonstrated for calmodulin (41).
Although associated with the vast majority of excitatory synapses, transfected PICK1-WT and PICK1-⌬DE were not associated with all of them. This result implies that mechanisms other than the intrinsic structural features of PICK1 play a role in the regulation of PICK1 synapse targeting and/or stabilization. It has been shown recently that PKC activation FIG. 6. Interaction of PICK1 mutants with PICK1-WT and the C-terminal tail of mGluR7a in the yeast two-hybrid system. Interactions were tested by cotransformation of yeast with full-length PICK1 fused to the GAL4 DNA activation domain and with various PICK1 constructs or fragments fused to the GAL4 binding domain. Positive clones leading to expression of the three reporter genes histidine, adenine, and ␤-galactosidase are indicated by ϩ. Only the fragment corresponding to the PICK1 acidic region (DE fragment) does not interact with PICK1-WT. When the interaction of the different PICK1 mutants was tested with the C-terminal tail of mGluR7a (CT7), only the PDZ mutant PICK1-KD/AA does not interact with CT7 (Ϫ). in cultured neurons enhanced PICK1 synaptic localization, suggesting that phosphorylation events are involved in PICK1 neuronal subcellular distribution (15). Although there was no direct demonstration that this effect correlated with PKC-induced PICK1 phosphorylation, previous studies showed that PICK1 could be phosphorylated in vitro by PKC (6) in accordance with the presence of several consensus phosphorylation sites for PKC in PICK1. Thus phosphorylation, perhaps by modifying coiled-coil-mediated oligomerization or PDZ domain associations, could regulate PICK1 distribution and function.
Based on our data, the simplest model for PICK1 synaptic aggregation and its ability to cocluster a target membraneassociated receptor requires two specific elements: binding to the target membrane receptor by a PDZ-mediated interaction and PICK1 self-association through the coiled-coil domain. This model takes into account the fact that the PICK1-interacting receptors exist as dimeric or heteromeric forms, as shown for mGluR7a (10) and the AMPA receptor (42), respectively, thus allowing for rafting of PICK1 dimers and multimeric receptors with multiple PICK1-binding sites. In addition, the PICK1 acidic region represents an inhibitory module in these processes, possibly by binding with other protein partners or divalent cations. PDZ proteins play a central role in synaptic organization by their involvement in neurotransmitter receptor clustering and signaling complex assembly. Changes in the molecular composition of synapses are thought to underlie some aspects of synaptic plasticity. In this study, we defined three structural signals within PICK1 regulating its synaptic localization and receptor coclustering activity. These motifs could correspond to molecular substrates involved in synaptic development and plasticity.