Interaction between SAP97 and PSD-95, Two Maguk Proteins Involved in Synaptic Trafficking of AMPA Receptors*

Synapse-associated protein 97 (SAP97) and postsynaptic density 95 (PSD-95) are closely related membrane-associated guanylate kinase homologs (Maguks) implicated in the synaptic targeting and anchoring of α-amino-5-methyl-3-hydroxy-4-isoxazolepropionic acid (AMPA)-selective glutamate receptors. Prompted by accumulating evidence for an oligomeric nature of Maguks, we examined the potential of SAP97 and PSD-95 to form heteromeric complexes. SAP97 and PSD-95 coimmunoprecipitated from rat brain detergent extracts and subsequent glutathione S-transferase pull-down and immunoprecipitation experiments showed that the interaction is mediated by binding of the N-terminal segment of SAP97 (SAP97NTD) to the Src homology 3 domain of PSD-95 (PSD-95SH3). In cultured hippocampal neurons, expression of green fluorescent protein-tagged PSD-95 triggered accumulation of SAP97 in synaptic spines, which was totally inhibited by coexpression of PSD-95SH3. Furthermore, overexpression of green fluorescent protein-PSD-95 induced dendritic clustering of GluR-A subunit-containing AMPA receptors, which was strongly inhibited by cotransfection with SAP97NTD and PSD-95SH3 constructs. Our results demonstrated a direct interaction between SAP97 and PSD-95 and suggested that this association may play a functional role in the trafficking and clustering of AMPA receptors.

Trafficking of AMPA receptors is under intensive study because of its believed central role in synaptic plasticity (9 -12). Both SAP97 and PSD-95 participate in synaptic targeting of AMPA receptors, although the underlying molecular mechanisms are still poorly understood. SAP97 interacts directly with AMPA receptor subunit GluR-A (GluR1) (13,14), whereas PSD-95 is able to associate with AMPA receptors in an indirect manner via stargazin and related transmembrane AMPA receptor regulatory proteins, TARPs (15,16). Overexpression of PSD-95 (17) and SAP97 (18,19) drives GluR-A-containing AMPA receptors to synapses in transfected neurons, and conversely, RNA interference knockdown of endogenous SAP97 and PSD-95 inhibits surface expression and synaptic delivery of AMPA receptors (19).
Biochemical experiments indicate that Dlg proteins may form oligomers and multimers. SAP97 exists as dimers and possibly higher oligomers stabilized by N-terminal interactions (19,20). PSD-95 and PSD-93 have been reported to form disulfide-linked multimers in the brain, including heteromeric complexes (21,22). More recently, an alternative mechanism involving palmitoylation of the N-terminal cysteines implicated earlier in disulfide formation has been proposed for PSD-95 multimerization (23). In epithelia, SAP97 has been reported to associate with CASK/mLin-2, a distantly related Maguk, via L27 domains in both proteins (24). Recent NMR and x-ray diffraction studies provide a detailed structural explanation for the ability of L27 domains to mediate formation of heterodimers (25,26).
In the present study, we examined the possibility that the two Dlg proteins implicated in AMPA receptor trafficking, SAP97 and PSD-95, may directly interact. We demonstrated that SAP97 and PSD-95 associate via binding of the N-terminal segment of SAP97 to the SH3 domain of PSD-95 and provided evidence for a role of this interaction in the synaptic transport of GluR-A-containing AMPA receptors.

EXPERIMENTAL PROCEDURES
Materials-N-terminally GFP-tagged SAP97 in pEGFP-C1 vector and C-terminally GFP-tagged PSD-95 in pGW1-CMV vector were generous gifts from Dr. Craig Garner (Stanford University, Stanford, CA) and Dr. David Bredt (University of California, San Francisco, CA), respectively. Antiserum raised against the N-terminal residues 1-100 of SAP97 (anti-SAP97 N ) has been described (Cai et al. (14)). Monoclonal antibody against PSD-95 was from Sigma. SAP97 and PSD-95 antibodies were specific for their respective Maguks in immunoblotting against recombinant SAP97, SAP102, PSD-95, and PSD-93 (see supplemental data, Fig. S1). Polyclonal IgG antibodies against Myc tag and GFP were obtained from AbCam (Cambridge, UK). Anti-His tag monoclonal antibody was from Qiagen, Hilden, Germany. Cy3-conjugated anti-mouse and anti-rabbit IgGs were from Jackson ImmunoResearch Europe, Cambridgeshire, UK. Anti-mouse-IgG-Alexa Fluor 488 was a product of Molecular Probes, Eugene, OR. Horseradish peroxidase-conjugated anti-mouse and anti-rabbit IgGs were from Amersham Biosciences. Rabbit antiserum against GluR-A C-terminal domain was raised in the * This study was supported by grants from the Academy of Finland (Grant 202892) and National Technology Agency of Finland (TEKES) and by the Finnish Graduate School in Neurosciences. 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 U.S.C. Section 1734 solely to indicate this fact. □ S The on-line version of this article (available at http://www.jbc.org) contains four supplemental figures. 1 To whom correspondence should be addressed. Laboratory Animal Facility of Viikki Biocenter, University of Helsinki following standard procedures. Briefly, primary immunization was performed with 400 g of purified GST fusion protein of GluR-A residues 827-907 in complete Freund's adjuvant. This was followed by three booster immunizations (at 3, 5, and 7 weeks after the primary immunization), each with 200 g of protein in Freund's incomplete adjuvant. In immunofluorescence microscopy, GluR-A antiserum but not the cognate preimmune serum produced a bright somatodendritic staining of cultured hippocampal pyramidal cells, and in Western blots of transfected HEK293 cells, the antiserum was specific for the GluR-A subunit (see supplemental data, Fig. S2). DNA Constructs-For expression of GST fusion proteins, pGEX-BNH, a modified version of pGEX4-T3 (Amersham Biosciences), was used as a vector (14). GST fusion constructs for full-length of SAP97 and its N-terminal residues 2-100 (N1) and for the PDZ1-3 domains of SAP97 (residues 213-549) and PSD-95 (residues 54 -399) have been described previously (14). The following derivatives of SAP97 and PSD-95 were amplified by PCR from plasmid templates encoding the full-length molecules and subcloned in pGEX-BNH: the entire N-terminal segment of SAP97 (NTD; residues 2-223), SH3-guanylate kinase segment of SAP97 (SH3-GK; residues 565-911), SAP97 lacking the N-terminal segment (SAP97⌬NTD; residues 224 -911), full-length PSD-95 (residues 1-724), SH3-guanylate kinase segment of PSD-95 (SH3-GK; residues 432-724), SH3 domain of PSD-95 (residues 432-502), guanylate kinase domain of PSD-95 (GK; residues 531-724), and PSD-95 lacking the N-terminal segment (⌬N; residues 54 -724). SAP97 constructs used in this study carry the I1b insert at the N-terminal segment (27).
For mammalian cell expression, DNA fragments encoding the above described segments of SAP97 and PSD95 and PCR-generated SAP97 fragment coding for residues 101-223 (N2) were subcloned into pcDNA3.1(-) (Invitrogen) modified to carry sequences encoding an N-terminal c-Myc tag and/or a C-terminal His 6 tag and into pEGFP-C1 (Clontech) for an N-terminal GFP tag. (For an outline of SAP97 and PSD-95 constructs, see supplemental data, Fig.S3.) Full-length SAP102 and PSD-93 were amplified by PCR from rat brain cDNA using appropriate primers and cloned into pcDNA3.1 derivative for expression as N-terminally Myc-tagged proteins. The correctness of the DNA sequences of all PCR products was verified by dideoxy sequencing.
GST Pull-down Assays-GST fusion proteins were expressed in Escherichia coli BL21 and purified using glutathione-Sepharose (Amersham Biosciences) according to the manufacturer's instructions. Proteins were extracted from adult rat brain homogenates or from transfected HEK293 cells in 50 mM Tris-HCl, pH 8.0, 140 mM NaCl, 1 mM EDTA, 1 mM Na 3 V0 4 , 1 mM NaF, and 1 mM phenylmethylsulfonyl fluoride (TNE buffer) containing 1% Triton X-100, centrifuged for 20 min at 20,000 ϫ g, and incubated overnight at ϩ4°C with 10 g of GST fusion protein bound to glutathione-Sepharose beads in a 1-ml total volume. Sepharose beads were collected by centrifugation, washed five times with 10 mM Tris-HCl, pH 7.4, 140 mM NaCl (TBS), containing 1% Triton X-100 and then washed twice with TBS, and finally eluted in SDS sample buffer and resolved by electrophoresis. After SDS-PAGE, the proteins were transferred to nitrocellulose, stained briefly with Ponceau S, and blocked by 5% nonfatted dry milk in TBS. Blots were incubated with anti-Myc (1:1000), anti-GFP (1:10000), or anti-His (1:200) for 16 h at ϩ4°C. Thereafter, the blots were washed with TBS and incubated (1 h, room temperature) with the anti-rabbit or anti-mouse IgG horseradish peroxidase conjugate (1:5000) and developed by using ECL Plus system (Amersham Biosciences).
Cell Culture and Transfections-HEK293 cells were grown in Dulbecco's modified Eagle's medium supplemented with 2 mM glutamine, 10% (v/v) fetal calf serum, penicillin (50 units/ml), and streptomycin (50 units/ml) at 37°C under 5% CO 2 . HEK293 cells were transfected with SAP97 and PSD-95 expression plasmids by using calcium phosphate coprecipitation. The cells were harvested 36 -48 h after transfection in TNE buffer and used for immunoprecipitations as described below.
Primary hippocampal neurons were obtained from E17 mouse embryos. Hippocampi were dissected and digested with 0.25% trypsin for 15 min in Hanks' balanced salt solution. Cells were dissociated by back-and-forth pipetting with a fire-polished Pasteur pipette and plated at a density of 5 ϫ 10 4 cells/cm 2 on poly-D-lysine-coated 13-mm round glass coverslips in 24-well plates in serum-free Neurobasal medium supplemented with B27. After 4 days, half of the medium was replaced by glial cell-conditioned Neurobasal/B27. The cells were transfected after 10 days in vitro. The medium was changed to Dulbecco's modified Eagle's medium/10 mM MgCl 2 without serum (transfection medium). For each well, a total of 1-2 g of plasmid DNA in 16 l was mixed with 1.5 l of 2.5 M CaCl 2 and combined with 17.5 l of 42 mM HEPES, pH 7.06, containing 274 mM NaCl, 10 mM KCl, 1.4 mM Na 2 HPO 4 , and 15 mM D-glucose (2ϫ HEPES-buffered saline), and the solution was added dropwise to the cells. For cotransfections, equal amounts of each plasmid were used. After a fine precipitation had formed (20 -40 min), the transfection medium was removed and replaced by Neurobasal/B27. The cells were analyzed 7 days later as described below.
Immunoprecipitations-Detergent extracts were prepared from HEK293 cells or from cerebella of adult male Wistar rats by homogenizing the cells or tissue in 10 volumes of TNE buffer, after which Triton X-100 was added to a final concentration of 1% (w/v), and the suspension was mixed at ϩ4°C for 2 h followed by centrifugation at 20,000 ϫ g for 20 min. One-ml aliquots of supernatant were incubated with the appropriate IgG (2 g) or antiserum (2 l) overnight at 4°C, after which 20 l of GammaBind G-Sepharose (Amersham Biosciences) was added, and the incubation was continued with gentle rotation for an additional 2 h. GammaBind-Sepharose beads were washed five times with TNE and twice with PBS and then finally suspended in 20 l of SDS sample buffer, heated at ϩ95°C for 5 min, and centrifuged briefly. The supernatants were analyzed by SDS-PAGE followed by immunoblotting as described above for GST pull-downs. For control purposes, preimmune sera or a nonrelevant IgGs were used to replace the primary antibody in the precipitations.
Immunostainings and Fluorescence Microscopy-Hippocampal neurons were fixed using warm 4% paraformaldehyde, 4% sucrose in PBS for 20 min and permeabilized in 0.2% Triton X-100 in PBS for 10 min, rinsed with PBS, and incubated for 2 h with 4% bovine serum albumin, 2% goat serum, and 0.2% Triton X-100 (blocking solution; in PBS). The cells were then incubated with the primary antibodies/antisera overnight at ϩ4°C in the blocking solution. For colocalization studies, SAP97 N antiserum (1:100) and anti-PSD-95 monoclonal antibody (1:400) were used. GluR-A CTD antiserum (1:400) was used for staining clusters of GluR-A-containing AMPA receptors in neurons. The stained cells were washed with 0.2% Triton X-100/PBS (3 ϫ 10 min) and incubated with appropriate Cy3-(red) or Alexa Fluor 488 (green)-conjugated secondary antibodies (1:400). After mounting in glycerol-gelatin, epifluorescence was visualized using the ϫ40 objective at maximum resolution in an Olympus AX70 microscope equipped with fluorescein isothiocyanate (for Alexa Fluor 488 and GFP) and TRITC (for Cy3) filters and a charge-coupled device camera. The digital images were analyzed by using ImagePro 4.0 (Media-Cybernetics, Silver Spring, MD). Spines expressing endogenous SAP97 were counted from three largest caliber dendrites in 12 GFP-PSD-95 transfected or untransfected neurons. For quantitation of GluR-A clustering, 10 -20 labeled neurons were chosen randomly from 3-4 independent transfection experiments. The intensity of clusters, defined as discrete spots with an average fluorescence of Ն10-fold higher than the background, was measured from one or two largest proximal dendrites. Student's unpaired t test was used for statistical analysis.

SAP97 Associates with PSD-95 in Vivo and in Vitro-
In immunoprecipitation analyses of rat brain proteins interacting with glutamate receptors, we consistently noticed coprecipitation of SAP97 and PSD-95. Anti-SAP97 N , an antiserum raised against the unique N-terminal sequence (residues 1-100) of SAP97 (14) but not the cognate preimmune serum, precipitated PSD-95, present as a 95/80-kDa PSD-95 doublet in immunoblots (Fig. 1A). Likewise, a 120-kDa SAP97 band was observed in immunoprecipitates generated by anti-PDS-95 IgG (Fig.  1A). The antibodies were specific for their respective Dlg proteins, as direct anti-PSD-95 blot of rat brain did not detect any 120-kDa species, nor was the 95/80-kDa PSD-95 band labeled by direct anti-SAP97 N blot (Fig. 1A, input lanes). These findings indicate that SAP97 and PSD-95 are present in the same molecular complexes in rat brain.
Coimmunoprecipitation results prompted us to examine the possibility that PSD-95 and SAP97 may actually directly interact to form heteromeric Maguk complexes. For this purpose, HEK293 cells were transfected for expression of Myc-tagged SAP97 (Myc-SAP97) and GFP-tagged PSD-95 (GFP-PSD-95), either singly or together, and cellular detergent extracts were analyzed for the presence of a binary complex. A 115-kDa GFP-PSD-95 band was observed in anti-Myc immunoprecipitates, whereas a 120-kDa Myc-SAP97 species was present in anti-GFP immunoprecipitates prepared from cotransfected cells. Untransfected or singly transfected cells did not produce either of the these bands, indicating that in cotransfected cells, SAP97 and PSD-95 exists as a complex (Fig. 1, B and C). Because SAP97 is also endogenously expressed in HEK293 cells, appearing as a 120-kDa species in anti-SAP97 N immunoblots (result not shown), we next studied whether PSD-95 would associate with the native SAP97. Consistent with such an association, anti-SAP97 N , but not the cognate preimmune serum, brought down GFP-PSD-95 from transfected cells (Fig. 1D).
Mapping the PSD-95 Binding Site in SAP97-To identify the structural determinants responsible for binding of SAP97 to PSD-95, N-terminally Myc-tagged polypeptide fragments representing different parts of SAP97 were expressed in HEK293 cells with Histagged PSD-95, and the association was analyzed by immunoprecipitations from cell extracts. As shown in Fig. 2A, full-length SAP97 as well as SAP97⌬SH3-GK immunoprecipitated together with PSD-95, whereas three SAP97 fragments devoid of the N-terminal segment, but covering the rest of the polypeptide, SAP97⌬NTD, PDZ1-3, and SH3-GK, were not precipitated by anti-His antibody. Consistent with this, SAP97 NTD (residues 1-223) alone showed association with PSD-95, whereas a shorter N-terminal fragment, SAP97 N1 , comprising residues 1-100, including the L27 domain, did not coprecipitate with PSD-95 ( Fig. 2A). Similar results were obtained when the same detergent extracts were immunoprecipitated with anti-Myc antibody; immunoblots with anti-His show association of PSD-95 with full-length SAP97, SAP97⌬SH3-GK and with SAP97 NTD (Fig. 2B). These results demonstrate that the N-terminal segment of SAP97 is both necessary and sufficient for association with PSD-95.
Mapping the SAP97 Binding Site in PSD-95-To identify the structures in PSD-95 that SAP97 NTD interacts with, we used an approach similar to the one described above. HEK 293 cells were now transfected for expression of GFP-tagged full-length SAP97 together with Myctagged PSD-95 domains, and the cell lysates were analyzed by immunoprecipitation. When anti-Myc antibody was used to analyze GFP immunoprecipitates, deletion of the PSD-95 N terminus did not have any effect on association with SAP97, suggesting that N-terminal interactions similar to those described for homo-oligomerization of SAP97 and PSD-95 do not mediate association of SAP97 with PSD-95 (Fig. 3A). Remarkably, the SH3 segment of PSD-95 showed association with GFP-SAP97, whereas the PDZ1-3 and GK domains were not precipitated together with SAP97 (Fig. 3A). Surprisingly, considering the association of SH3 segment, PSD-95 SH3-GK did not coprecipitate with SAP97 (Fig.  3A). The reason for this failure is likely to be technical because when the  same extracts were subjected to anti-His immunoprecipitation followed by anti-GFP detection, both PSD-95 SH3 and PSD-95 SH3-GK were clearly present in the immunoprecipitates (Fig. 3B). In all other respects, the results were not affected by the selection of the precipitating versus detecting antibody (Fig. 3B). Based on this analysis, the SH3 segment of PSD-95 is sufficient for association with SAP97 in transfected HEK293 cells.
GST Pull-down Analysis of SAP97-PSD-95 Interaction-To corroborate the immunoprecipitation results, we used GST pull-down assays to analyze the interaction. First, in good agreement with immunoprecipitation experiments, GFP-tagged PSD-95 was retained by GST fusions of SAP97 and SAP97 NTD but not by SAP97 N1 or SAP97 SH3-GK3 (Fig. 4A). Interestingly, a weak interaction between GFP-PSD-95 with the GST fusion of SAP97 PDZ1-3 segment was consistently noticed in GST pull-downs (Fig. 4A). The specificity and characteristics of this presumably low affinity interaction were not pursued further, however, in this study. Furthermore, GFP-tagged PSD-95 SH3 bound to GST fusion of SAP97 NTD but did not show any interaction with SAP97 lacking the NTD (Fig. 4B). In conclusion, in vitro interaction studies indicate that the N-terminal segment of SAP97 and the SH3 domain of PSD-95 are necessary and sufficient for an heteromeric interaction between these Dlg proteins in mammalian cells.

SAP97-PSD-95 Interaction in Cultured
Neurons-To obtain information on the physiological role of the interaction between SAP97 and PSD-95, we first examined the cellular distributions of the two Maguks in cultured hippocampal neurons by immunofluorescence microscopy. Strong somatodendritic labeling was observed for both proteins, PSD-95 also exhibiting prominent punctate staining in spine-like structures along the dendrites. In contrast, SAP97 immunoreactivity was more diffuse and confined outside of spines, although a small number of putative spines were stained for both SAP97 and PSD-95, in agreement with previous studies (18,19,28) (Fig. 5A).
Interestingly, transfection of the neurons with GFP-tagged PSD-95 plasmid caused a marked change in the cellular distribution of endogenous SAP97, leading to accumulation of endogenous SAP97 in synaptic spines and coclustering with GFP-PSD-95 (Fig. 5B), consistent with an interaction between the two molecules. PSD-95-driven clustering of SAP97 is indicative of a direct or indirect interaction between these molecules. To characterize this finding further, neurons were transfected for expression of GFP-PSD-95 with or without PSD-95 SH3 , and the numbers of dendritic spines labeled by anti-SAP97 N antiserum were counted. GFP-PSD-95 increased the clustering of SAP97 to spines by a factor of 3-4 in comparison with untransfected controls, and this effect was totally eliminated by PSD-95 SH3 (Table 1, Fig. 6). The number of GFP-PSD-95 fluorescent spines was not affected by coexpression of the SH3 domain ( Table 1).

Involvement of SAP97-PSD-95 Interaction in AMPA Receptor
Trafficking-Because both SAP97 and PSD-95 have been implicated in the synaptic trafficking of AMPA receptors, we next studied the possibility that SAP97-PSD-95 interaction is involved in this process. In agreement with earlier reports (17,29), overexpression of GFP-tagged PSD-95 in cultured hippocampal neurons led to an increased clustering of GluR-A-immunoreactivity in spine-like structures (Fig. 7A). The intensity of GluR-A clusters on the proximal dendrites of GFP-positive cells was on average 2.2 times higher than in the neighboring nontransfected cells (Table 2; Fig. 7A). Cotransfection with SAP97 NTD abolished the GFP-PSD-95-induced clustering of GluR-A AMPA receptors entirely ( Fig. 7B and Table 2). Furthermore, cotransfection with PSD-95 SH3 also inhibited the PSD-95 induced clustering, albeit as completely as was observed with SAP97 NTD (Fig. 7C and Table 2). When SAP97 NTD or PSD-95 SH3 were expressed alone as GFP fusion proteins, neither of them had any significant influence on the intensity of GluR-A clusters ( Table 2; see also supplemental data, Fig. S4). These results are consistent with the hypothesis that SAP97-PSD-95 interaction is involved in PSD-95-induced clustering of GluR-A AMPA receptors, and therefore, this interaction may be involved in the synaptic trafficking of AMPA receptors.

DISCUSSION
The results of the present study indicated that in rat brain and in transfected cells, PSD-95 and SAP97 can associate as heteromeric complexes. There are previous reports on Dlg hetero-oligomerization, both involving PSD-95. PSD-93/Chapsyn-110 is coimmunoprecipitated with PSD-95 from rat brain, and biochemical and mutagenic analyses indicate that this interaction is mediated by the cysteine-containing N termini of the two proteins (21,22). In transfected cells, recombinant PSD-95 has been reported to form a complex with SAP102, the mechanism involving the SH3-guanylate kinase segments of the two proteins (30). Our present demonstration that the N-terminal part of SAP97 and the SH3 segment of PSD-95 mediate SAP97-PSD-95 interaction has  illustrated yet another mechanism that can give rise to heteromeric Dlg complexes. Interestingly, an analogous intramolecular interaction has been previously identified in SAP97, where it may regulate the ability of the SH3-GK segment to interact with other proteins (31,32). Accordingly, SAP97 is postulated to exist in at least two alternative states; intimate interactions between the SH3 and GK domains stabilize a "closed" complex, in which binding of a guanylate kinase-associated protein, GKAP, to the GK domain is blocked, whereas association of the N-terminal domain with the SH3 segment liberates the GK domain for GKAP binding (31). The former, closed complex is likely to correspond to the recently determined crystallographic structure of the SH3-GK segment of PSD-95 (33,34). Our present results indicated that under physiological conditions prevailing in cultured cells, SAP97 NTD can engage the SH3 domain in a tight association also in an intermolecular fashion, resulting in a heterodimeric complex with PSD-95. Conse-quently, we predict that by recruiting the SH3 domain of PSD-95, SAP97 NTD may be able to promote association of GKAP to the guanylate kinase domain of PSD-95 as well. The potential mechanistic consequences of the interaction between SAP97 NTD and PSD-95 SH3 are not, however, limited to regulation of GKAP binding. It is obvious that due to steric reasons, the interaction may well influence (or, in a mutual way, be influenced by) many other protein interactions of SAP97 and PSD-95.
Both PSD-95 and SAP97 are believed to participate in the trafficking of AMPA-type glutamate receptors, probably by interacting directly or indirectly with AMPA receptors and thereby providing links to transport motors, cytoskeletal elements, and/or regulatory kinases and phosphatases, although detailed mechanisms are still poorly understood (9,10). Earlier studies indicate that activity-dependent synaptic incorporation of GluR-A/B receptors, a process that shares properties with PSD-95-induced accumulation of GluR-A to synapses (17,29,35), differs mechanistically from a constitutive GluR-B/C-specific pathway (36 -38). Notably, synaptic insertion of GluR-A receptors requires the activity of Ca 2ϩ /calmodulin-dependent protein kinase II and the presence of the C-terminal Type I PDZ motif in GluR-A (38). SAP97 is considered as a strong candidate to serve an essential function in this process because its overexpression promotes synaptic delivery of GluR-A (18,    FEBRUARY 17, 2006 • VOLUME 281 • NUMBER 7 19), it is driven to synapses by calmodulin-dependent protein kinase II activity (39,40), and is so far the only PDZ protein identified to bind the C terminus of GluR-A with Type I specificity (13,14). PSD-95 does not appear to bind directly AMPA receptors, and its functional role in AMPA receptor trafficking is believed to be largely mediated by stargazin and related TARPs (15,16,29). Interestingly, TARPs bind to all AMPA receptor subunits showing little or no subunit specificity (15), implying further interactions with additional subunit-specific factors, including SAP97, to impart selectivity to the transport processes. We found that in cultured hippocampal neurons, SAP97, which is normally not significantly present in spines, is strongly driven to spines by overexpression of GFP-tagged PSD-95. Remarkably, coexpression of PSD-95 SH3 abolished this effect entirely, suggesting that direct interaction with PSD-95 may be involved in the synaptic targeting of SAP97. The effect of SAP97 NTD on this phenomenon could not be analyzed because by our SAP97 antibody does not distinguish between the endogenous SAP97 and the recombinant N-terminal domain. Earlier, it had been shown that PSD-95 triggers trafficking of GluR-A-subunit containing AMPA receptors to synaptic spines (17,41). We found that GFP-PSD-95-induced accumulation of GluR-A in synaptic clusters was strongly inhibited by cotransfection with SAP97 NTD and PSD-95 SH3 constructs in cultured hippocampal neurons. These findings were consistent with a role of SAP97-PSD-95 interaction in synaptic transport of GluR-A receptors. Clearly, however, the observed inhibition of GluR-A clustering by SAP97 NTD and PSD-95 SH3 constructs did not necessarily indicate SAP97-PSD-95 complex as an essential step in the synaptic incorporation of GluR-A. An alternative explanation, often to be considered when using "dominant negative" constructs to specifically interfere with biological functions, is that SAP97 NTD and PSD-95 SH3 may actually interfere with other physiological activities of these proteins in AMPA receptor trafficking, unrelated to their mutual interaction. In principle, this might occur via neutralization of essential protein interactions that involve the N terminus of SAP97 or the SH3 domains of both PSD-95 and SAP97. At present, we cannot exclude this possibility, and further work, including mutations specifically eliminating binding of SAP97 to PSD-95 but leaving other interactions of the responsible protein domains intact, is clearly needed to resolve this issue.

SAP97-PSD-95 Interaction
Based on the findings discussed above, we suggest that SAP97 and PSD-95 interact at or near synaptic sites and that this interaction contributes to the synaptic clustering of GluR-A AMPA receptors. The existence of a synaptic complex between SAP97 and PSD-95 is supported by a recent report of association of SAP97 with immunopurified PSD-95 complexes prepared from rat postsynaptic densities (42). Importantly, the same study demonstrated that GluR-A associates with PSD-95 complexes to the same relative extent as SAP97, suggesting that association of GluR-A with postsynaptic PSD-95 complexes is mediated by SAP97 (42). Association of GluR-A with SAP97-PSD-95 complexes exclusively at synaptic sites, in a detergent-insoluble postsynaptic complex, would also be consistent with a general failure to demonstrate coimmunoprecipitation of GluR-A with PSD-95 (43). 3 The available data are insufficient for any detailed mechanistic models, but in principle, SAP97-PSD-95 interaction may either facilitate GluR-A clustering/ synaptic incorporation or help stabilize the clustered synaptic receptors. Regarding the latter possibility, it was recently reported that endocytosis of synaptic GluR-A receptors is dependent on a ternary interaction between GluR-A, SAP97, and myosin VI (44). Based on the present results, it can be speculated that SAP97 N-terminal domain may interact alternatively with PSD-95, thereby stabilizing GluR-A on the synaptic surface, or with myosin VI (45), resulting in endocytosis of the receptor.
In conclusion, our results have demonstrated an interaction between SAP97 and PSD-95, two Dlg homologs involved in synaptic clustering of

TABLE 2 Intensities of GluR-A immunoreactive clusters in transfected neurons
Clustering of GluR-A was analyzed by immunofluorescence and quantified as described under "Experimental Procedures." Cluster intensities in GFP-positive cells are indicated as percentage of control using cluster intensities measured from neighboring untransfected cells as a 100% reference (including a total of 127 clusters in 20 neurons; 100 Ϯ 8.4%). Statistical significance of the results was analyzed by using unpaired t test to calculate the indicated two-tailed p values. NS, not significant (p Ͼ 0.05); NA, not applicable.