Interaction of the Tyrosine Kinase Pyk2 with the N -Methyl- D -aspartate Receptor Complex via the Src Homology 3 Domains of PSD-95 and SAP102*

The protein-tyrosine kinase Pyk2/CAK (cid:1) /CADTK is a key activator of Src in many cells. At hippocampal synapses, induction of long term potentiation requires the Pyk2/Src signaling pathway, which up-regulates the activity of N -methyl- D -aspartate-type glutamate receptors. Because localization of protein kinases close to their substrates is crucial for effective phosphorylation, we investigated how Pyk2 might be recruited to the N -methyl- D -aspartate receptor complex. This interac- tion is mediated by PSD-95 and its homolog SAP102. Both proteins colocalize with Pyk2 at postsynaptic dendritic spines in the cerebral cortex. The proline-rich regions in the C-terminal half of Pyk2 bind to the SH3 domain of PSD-95 and SAP102. The SH3 and guanylate kinase homology (GK) domain of PSD-95 and SAP102 interact intramolecularly, but the physiological signifi-cance of this interaction has been unclear. We show that Pyk2 effectively binds to the Src homology 3 (SH3) domain of SAP102 only when the GK domain is removed from the SH3

SH3 and GK domains of PSD-95 and its homologs form intramolecular and, to some degree, intermolecular interactions (20 -24). A point mutation in the Drosophila protein Dlg (a homolog of PSD-95) results in the substitution of a conserved leucine in the SH3 domain to a proline; this mutation leads to the loss of septate junction formation and overproliferation of the imaginal disc (25). Replacing the homologous leucine (Leu-460) with a proline in the SH3 domain of PSD-95 inhibits the interaction of the mutant SH3 domain with the GK domain (20,22). Collectively these findings indicate the physiological importance of the SH3-GK domain interaction, but a molecular function of this interaction remains to be established.
Tyrosine phosphorylation of NR2B (26,27) is augmented on LTP induction (28,29), and several observations suggest that up-regulation of NMDA receptor activity by Src-mediated tyrosine phosphorylation is an important step in the induction of LTP. 1) Src increases NMDA receptor activity (30 -32) by elevating its peak current (33), 2) the activity of Src is up-regulated during LTP (34), 3) inhibition of Src can prevent LTP (30), and 4) an increase in synaptic transmission on Src activation occludes subsequent LTP induction (34).
It is well established that Src can be activated by another tyrosine kinase, Pyk2/CAK␤/CADTK (35)(36)(37)(38). Pyk2 is stimulated upon Ca 2ϩ influx, activation of PKC by phorbol esters, or stimulation of G q -linked receptors such as the muscarinic acetylcholine receptor M1 or the metabotropic glutamate receptor mGluR1 (35, 37, 39 -48). Neither Ca 2ϩ nor PKC appear to directly regulate Pyk2 in vitro (39); how Ca 2ϩ or PKC activates Pyk2 in vivo is unknown. When stimulated, Pyk2 autophosphorylates itself on tyrosine 402. The SH2 domain of Src binds to this phosphorylated tyrosine of Pyk2, resulting in activation of Src (35,36). Recent evidence indicates that similar to Src, Pyk2 is necessary for the induction of LTP, likely because it increases the activity of Src (49).
Precise localization of protein kinases is a critical requirement for fast and specific signaling via phosphorylation of a variety of substrates (50 -52). For example, efficient regulation of AMPA receptors requires that PKA be localized in close proximity to protein kinase A anchor proteins (53). Similarly, Ca 2ϩ -calmodulin-dependent protein kinase II is recruited to the postsynaptic site via multiple, highly regulated interactions with the NMDA receptor (54 -58). We now show that Pyk2 binds to PSD-95 and SAP102 in the NMDA receptor complex, which also contains Src. More specifically, the prolinerich region in the C-terminal half of Pyk2 directly interacts with the SH3 domains of PSD-95 and SAP102. This binding is inhibited by the intramolecular interaction of the GK domain with the SH3 domain in SAP102 by regulating Pyk2 binding to the SH3 domain.

EXPERIMENTAL PROCEDURES
Materials-Glutathione-Sepharose was purchased from Amersham Biosciences and protein A-Sepharose from Sigma. Antibodies against Pyk2 and pcDNA3 plasmids for the expression of hemagglutinin-tagged full-length rat Pyk2 and FLAG-tagged human CRNK (residues 692-1009 of Pyk2) (59,60) in HEK293 cells were generously provided by Drs. L. M. Graves and H. S. Earp (University of North Carolina, Chapel Hill). We also purchased monoclonal antibodies against the C terminus of Pyk2 from BD Transduction Laboratories. Mammalian pCMV expression vectors of Myc-tagged full-length rat PSD-95 and rat SAP102 and bacterial expression vectors for the production of GST fusion proteins containing the PDZ1-3 (residues 57-405), PDZ2 (residues 156 -248), and GK (residues 534 -723) domains of PSD-95 and the SH3-GK domains (residues 537-647) of SAP102 (13,61,62) were graciously supplied by Dr. C. C. Garner (Stanford University, Palo Alto, CA), and plasmids for the production of GST fusion proteins of the PDZ3 (residues 302-402; originally in pGAD; subcloned into the EcoRI site of pGEX4T-1) and SH3 (residues 431-500) domains of PSD-95 were graciously supplied by Dr. M. Sheng (Howard Hughes Medical Institute, Massachusetts Institute of Technology, Boston) (63,64). Expression vectors for GST-CRNK (residues 671-1009 in pGEX2TK) and GST-⌬N-CRNK (residues 876 -1009) were generously made available by Drs. T. Sasaki (Sapporo Medical University, Sapporo, Japan), and M. D. Schaller (University of North Carolina, Chapel Hill). Other reagents were of standard biochemical quality and from established providers.
Expression Constructs for GST Fusion Proteins of the SH3 and GK Domains of SAP102-A vector for expressing GST-SH3 fusion protein carrying residues 537-647 of rat SAP102 was produced by removing the GK domain from the original SAP102 GST-SH3-GK vector (a modified pGEX2T vector (13)) with NdeI and EcoRI. Overhangs were filled in with Klenow fragment, followed by blunt-end ligation. Positive clones were identified by colony PCR. The vector for expressing the GST-GK fusion protein covering residues 658 -837 of full-length SAP102 was obtained through amplification of the GK domain by PCR using primers containing EcoRI restriction sites (5Јprimer, 5Ј CAA GAA GAA TTC TAT GCC AGG CCT GTG ATC ATC TTG G 3Ј; 3Јprimer, 5Ј CCA AAT GAA TTC TCA AGA CTG GTC CTC AAT GAT TTG TTT G 3Ј). The PCR product was digested with EcoRI and inserted into EcoRI-digested pGEX4T-1 vector before transformation into Escherichia coli. Clones containing pGEX4T-1 vectors with the GK domain inserted in the correct orientation were identified by colony PCR. All new and old vectors used in this study were confirmed by DNA sequencing.
Interaction Assays with GST Fusion Proteins-GST and GST fusion proteins of the PDZ1-3, PDZ2, PDZ3, SH3, and GK domains of PSD-95 or SH3, GK, and SH3-GK domains of SAP102, along with CRNK and ⌬N-CRNK, were expressed from pGEX vectors (13) in the E. coli strain BL21 (Novagen) as described (18). Equal amounts of the different GST fusion proteins or GST (ϳ25 g) were absorbed onto glutathione-Sepharose (ϳ20 l), and the residues were incubated for 2 h at 4°C with lysates (prepared with RIPA buffer) from HEK293 cells transiently overexpressing Pyk2, CRNK, PSD-95, or SAP102. Typically, the aliquots of the HEK293 lysates contained 300 g of total protein as determined with the BCA assay (Pierce). Resins were washed with 500 mM NaCl, 10 mM Tris-Cl, pH 7.4, with 0.1% Triton X-100 in TBS (150 mM NaCl, 10 mM Tris-Cl, pH 7.4) and with TBS alone and then analyzed by immunoblotting as described (18,55). Lysate containing 5-10 g of protein was loaded onto gels as a positive control. Pyk2 was detected in these experiments with an antibody against the C-terminal portion of the kinase (residues 833-997; purchased from BD Transduction Laboratories). The presence of equal amounts of the GST fusion proteins was routinely verified by Ponceau S staining (Sigma) or probing with an anti-GST antibody (18).
Immunofluorescence-Four male Sprague-Dawley rats (200 -350 g; Charles River) were deeply anesthetized with sodium pentobarbital (60 mg/kg intraperitoneally), and intracardially perfused with heparinized saline followed by freshly depolymerized 0.5% paraformaldehyde in phosphate buffer (PB, 0.1 M, pH 7.4). Coronal 50-m sections were cut on a Vibratome and collected in cold PB. For single immunofluorescence, sections were incubated overnight in antibody to Pyk2 (1:1,000). Immunoreactivity was then visualized by donkey anti-rabbit conjugated to fluorescein isothiocyanate (FITC; Jackson ImmunoResearch). Because Pyk2, SAP102, and NR2A/B antibodies were raised in rabbit, tyramide signal amplification (TSA) was used for one of them, and subsequent conventional fluorescent staining was used for one of the other antibodies for double-labeling (72)(73)(74). After preincubation in 10% normal donkey serum, sections were incubated with Pyk2 antiserum at a dilution not recognized by a conventional fluorochrome-conjugated secondary antibody (1:20,000) and then reacted with biotinylated secondary antibody. Biotin was localized by a FITC-TSA direct kit (PerkinElmer Life Sciences). After preincubation in 1% normal donkey serum, SAP102 antiserum (1:1,000; Chemicon) or NR2 antiserum (1: 1,000; Chemicon) was applied overnight. Immunoreactivity was visualized with donkey anti-rabbit antibody, conjugated to CY-3 (Jackson ImmunoResearch). Sections were then processed for synaptophysin (1:1,000; Clone SVP-38; Sigma) or PSD-95 (1:1,000; Clone 16; Transduction Laboratories). After overnight incubation in primary antibody, immunoreactivity was visualized with donkey anti-mouse antibody conjugated to Cy-5 (Jackson ImmunoResearch). Control sections were processed as above, except that primary or secondary antibodies were omitted or replaced by nonspecific control antisera. To control for possible cross-reaction between the first primary antibody and the second secondary antibody, tissue was processed as above, except that the second primary antibody was omitted. Omission of primary or secondary antibodies eliminated specific labeling. Furthermore, consistent results were obtained after reversing the first and the second primary antibodies, using first SAP102 or NR2 (1:20,000) with FITC-TSA and then Pyk2 (1:1,000) with Cy-3-conjugated secondary antibody.

Coimmunoprecipitation of Pyk2 and Src with NMDA Recep-
tors-Because kinases are often localized in close proximity to their targets, we hypothesized that kinases involved in tyrosine phosphorylation of the NMDA receptor may be associated with this receptor. In fact, an earlier report (31) suggests that Src, which can up-regulate NMDA receptor activity, binds to the NMDA receptor complex. To test whether Pyk2, an activator of Src, associates with the NMDA receptor complex, rat forebrain samples were solubilized with deoxycholate and used for immunoprecipitations with antibodies against NR2A and -2B. Previously, we showed (18,55,65) that the interactions between the NMDA receptor and its associated proteins including PSD-95, SAP102, and Ca 2ϩ -calmodulin-dependent protein kinase II are well preserved under these conditions. Subsequent immunoblotting revealed that Pyk2 specifically and reproducibly coprecipitated with the NMDA receptor complex (Fig. 1A). We also observed copurification of Src with the Pyk2-NMDA receptor complex, especially if a mixture of NR2A and 2B antibodies was used for immunoprecipitation of the NMDA receptor (Fig. 1B). Coprecipitation of Pyk2 and Src with the NMDA receptor was specific because precipitations with nonimmune control antibodies did not result in isolation of Pyk2 or Src.
Coimmunoprecipitation of Pyk2 with PSD-95 and SAP102-PSD-95 and its homologs SAP102 and PSD-93 mediate the association of the NMDA receptor with various signaling proteins including neuronal nitric-oxide synthase and Ras-activating SynGAP (75)(76)(77). Therefore, we tested whether Pyk2 might be associated with PSD-95 or SAP102. Immunoprecipitation with antibodies against SAP102 resulted in coprecipitation of Pyk2 as detected with the antibody against the N-terminal 80 residues of Pyk2 (Pyk2-N; Fig. 2A, lanes 6 and 7). Probing of total membrane extract with the Pyk2-N antibody demonstrated the specificity of this antibody for Pyk2, which possesses a molecular mass of about 115 kDa (lane 1). Immunoprecipitation with both Pyk2-N and Pyk2-C also yielded the 115-kDa polypeptide that is immunoreactive for Pyk2-N (lanes 3 and 4, respectively). Pyk2 copurified with PSD-95 (lane 11), as well. To test whether Pyk2 might bind to SAP102 or PSD-95 after extraction with detergents rather than being associated with those two proteins in vivo, we solubilized brain mem- branes with 1% SDS at 60°C and subsequently added an excess of Triton X-100, which forms mixed micelles with SDS, thereby neutralizing it. This treatment dissociates the different components of the NMDA receptor complex including PSD-95 and SAP102 (9,55,62,65). Under these conditions Pyk2 did not coprecipitate with either SAP102 (lanes 8 and 9) or PSD-95 (lane 13). These observations indicate that the interactions between Pyk2 and PSD-95 or SAP102 do not form during the extraction procedure and are likely to occur in vivo. These interactions were further confirmed by immunoprecipitation of Pyk2 with either Pyk2-N or Pyk2-C and subsequent detection of SAP102 (Fig. 2B) and PSD-95 (data not shown) in immunocomplexes by immunoblotting.
Pyk2 Interacts with PSD-95 and SAP102 in HEK293 Cells in the Absence of the NMDA Receptor-To evaluate whether the association between Pyk2 and PSD-95 or SAP102 depends on the presence of the NMDA receptor or is due to a direct interaction, we expressed Pyk2 with either PSD-95 or SAP102 in the heterologous HEK293 cell line. Immunoprecipitation of Pyk2 with either Pyk2-N or Pyk2-C from cell lysates resulted in coprecipitation of PSD-95 and SAP102 (Fig. 3, upper panels). The specificities of these coprecipitations were demonstrated by the lack of PSD-95 and SAP102 immunoreactivity in immunoprecipitations with control IgG. Because HEK293 cells are non-neuronal and do not express NMDA receptors, these results suggest that Pyk2 interaction with PSD-95 and SAP102 does not require the NMDA receptor and might be direct.
Colocalization of Pyk2 with SAP102 and PSD-95-Multiple immunofluorescence staining was performed to compare the cellular and subcellular distribution of Pyk2 with that of SAP102 and PSD-95. Pyk2 antibodies stained a population of neurons distributed throughout the cerebral cortex (Fig. 4). Some of these neurons can be identified as pyramidal neurons (Fig. 4B, arrows); others were non-pyramidal cells of various shapes and sizes (Fig. 4B, arrowheads). Staining was typically patchy within dendrites and somatic cytoplasm, excluding the nucleus. At higher magnification, Pyk2 immunoreactivity was obvious in many dendritic spines (Fig. 4C). Numerous Pyk2 immunoreactive puncta were observed throughout the neuropil (Fig. 4D1, arrowheads). Such a distribution pattern often reflects a synaptic distribution. Double labeling with the presynaptic marker synaptophysin confirmed that many of the Pyk2positive puncta were at synaptic sites (e.g. see Fig. 4F, 1, 3, and 4, arrows).
SAP102-immunopositive neurons were also widely distributed in the cerebral cortex in a pattern that matched Pyk2 staining (Fig. 4D2). Double labeling confirmed that almost all SAP102-positive neurons were also positive for Pyk2 and vice versa (Fig. 4D3). In the neuropil, punctate colocalization was prominent (Fig. 4D3, arrowheads). Many of these puncta are likely to be synapses as indicated by double labeling with synaptophysin (e.g. Fig. 4, F2-4, arrows). Furthermore, Pyk2 and SAP102 immunoreactivity codistributed along dendrites, concentrating at dendritic spines (Fig. 4E1) juxtaposed to presynaptic synaptophysin staining (Fig. 4F). These results indicate that Pyk2 and SAP102 are colocalized not only in neuronal somata and dendrites but also at postsynaptic sites, where they may be associated with NMDA receptors.
NR2 and PSD-95 antibodies stained a large population of pyramidal and non-pyramidal neurons throughout the cortex. Triple immunofluorescence labeling showed cellular colocalization of Pyk2 with NR2 and PSD-95 (Fig. 5). This distribution was patchy in somata and proximal dendrites; some of these patches exhibited all three antigens, but others were immunopositive only for NR2 (Fig. 5, lower right). In the neuropil, most Pyk2-positive puncta colocalized with NR2 and PSD-95 staining. Likewise, Pyk2-positive spines were also triple-la-beled (Fig. 5, arrowheads in insets). However, many PSD-95 and some NR2-positive puncta were negative for Pyk2. These findings indicate that Pyk2 is colocalized with PSD-95 and NMDA receptors at many, although perhaps not all, postsynaptic spines where these proteins might be associated with each other.
The Proline-rich Region of Pyk2 Binds to the SH3 Domain of PSD-95-To define the interaction site of Pyk2 in PSD-95, we expressed GST fusion proteins of the PDZ1-3, PDZ2, PDZ3, SH3, and GK domains of PSD-95 in E. coli. Equal amounts of these fusion proteins were immobilized on glutathione-Sepharose, washed, and incubated with cell lysates of transfected HEK293 cells that ectopically expressed Pyk2, after lysates were pre-cleared with GST alone. Subsequent immunoblotting demonstrated that Pyk2 specifically bound to the SH3 domain of PSD-95 (Fig. 6A) and SAP102 (Fig. 6B). We obtained the same results when brain extracts were used as a source of native Pyk2 in similar pull-down experiments (data not shown). Immunoblots were stained with Ponceau S (Fig. 6, lower panels) and re-probed with anti-GST antibodies (not shown) to ensure that comparable amounts of each GST fusion protein were present in the various samples.
Pyk2 contains two segments that are proline-rich and constitute potential interaction sites for SH3 domains. These segments are toward the C terminus of Pyk2 (residues 713-720 and 855-862). We used the C-terminal Pyk2 fragment spanning residues 671-1009 for interaction assays. This fragment included the two proline-rich regions together with flanking sequences on both ends. It lacked the N-terminal region of Pyk2, which contains a homologous band 4.1 domain and the central catalytic domain. This catalytically inactive fragment, called CRNK (CADTK or CAK␤-related non-kinase) or Pyk2related non-kinase, is expressed as a truncated version of Pyk2 in brain and other tissues (59,78,79). Analogous to a similar C-terminal fragment of the related focal adhesion kinase FAK, which is also expressed in vivo and inhibits certain functions of FAK (80 -82), CRNK is thought to interfere with those signal- ing mechanisms of Pyk2 that depend on the interaction of the C-terminal domain of Pyk2 with corresponding binding partners (59,78,79). It is, therefore, not only of biochemical but also of physiological interest to determine whether the SH3 domain of PSD-95 binds to CRNK. Like Pyk2, we expressed CRNK in HEK293 cells and used lysates of these cells for pull-down experiments with the PSD-95 derived GST-SH3 fusion protein. CRNK bound to this fusion protein but not to GST alone (Fig. 6B) indicating a specific interaction with the SH3 domain of PSD-95.
To test whether the proline-rich segments of Pyk2 are responsible for PSD-95 and SAP102 binding, we expressed GST fusion proteins of CRNK (residues 671-1009) and ⌬N-CRNK (residues 876 -1009, lacking the two proline-rich segments) as GST fusion proteins. Lysates of HEK293 cells ectopically expressing full-length PSD-95 or SAP102 were used for pull-down experiments with GST fusion proteins of CRNK, ⌬N-CRNK, and the SH3 and GK domains of PSD-95 or SAP102. CRNK but not ⌬N-CRNK specifically interacted with full-length PSD-95 and SAP102 in these in vitro binding assays (Fig. 7). These results indicate that the proline-rich segments mediate the Pyk2-PSD-95 interaction. We consistently observed much weaker binding of SAP102 to CRNK than PSD-95 and therefore analyzed the availability of the SH3 domains in full-length PSD-95 and SAP102 for binding. Biochemical and crystallographic studies demonstrated that the SH3 and GK domains of PSD-95 and its homologs form intramolecular interactions (20 -24). Intermolecular binding of the SH3 domain of one  SAP102 (B). This Pyk2 antibody was from Transduction Laboratories and recognizes full-length Pyk2 (A and B) as well as CRNK (C) but does not cross-react with the GST moiety of our fusion proteins (in contrast to the Pyk2-N and Pyk2-C antibodies, which had been produced against GST fusion proteins of the corresponding Pyk2 regions (68)). Binding to PSD-95 was mediated by the C-terminal region of Pyk2 (CRNK; C). Blots were stained with Ponceau S (lower panels) before blocking and probing to show that comparable amounts of GST fusion proteins were present on each resin sample. Similar results were obtained in several other experiments.
PSD-95 molecule with the GK domain of another PSD-95 molecule can occur, but the intramolecular interaction is favored and occludes exogenous SH3 from binding to the GK domain and vice versa. The GST-SH3 and GST-GK fusion proteins specifically bound full-length PSD-95 on a weak basis, suggesting that, under our conditions, there is some access for the SH3 domain to its binding site on the GK domain, and vice versa, in full-length PSD-95. The interaction of GST-SH3 with SAP102 was weaker, and binding of GST-GK to SAP102 was not detectable. These results raise the possibility that the intermolecular interaction between the SH3 and GK domains of SAP102 may limit the interaction between Pyk2 and the SH3 domain.
Intramolecular GK Domain Association with the SH3 Domain Inhibits Pyk2 Binding-To test whether the presence of the GK domain within the same polypeptide influences the interaction of the SH3 domain of SAP102 with Pyk2, we performed pull-down experiments with GST fusion proteins carrying the SH3 or GK domain separately or in tandem. Whereas the GST fusion protein carrying only the SH3 domain of SAP102 pulled down Pyk2 from HEK293 cell lysates, a similar amount of a GST fusion protein with the combined SH3-GK domain did not (Fig. 8). The GK domain by itself did not mediate precipitation of Pyk2 in these pull-down experiments. These experiments indicate that SAP102 binds via its SH3 domain to Pyk2 and that the GK domain can inhibit this interaction. DISCUSSION Our results indicate that Pyk2, like Src, is associated with the NMDA receptor complex (Fig. 1) via the SH3 domains of PSD-95 and SAP102. The Pyk2 interaction with PSD-95 and SAP102 appears to be rather stable, as it can be detected after immunoprecipitation from brain (Fig. 2), a procedure that takes several hours, and also from HEK293 cells ectopically expressing these proteins (Fig. 3). Re-association does not appear to play a role during these immunoprecipitations; after dissociation of the Pyk2-PSD-95 and Pyk2-SAP102 complexes with SDS and neutralization of SDS with Triton X-100, Pyk2 does not rebind to PSD-95 or SAP102 (Fig. 2). An in vivo interaction of Pyk2 with PSD-95 and SAP102 is further supported by our immunofluorescence microscopy studies. Pyk2 colocalizes with SAP102 and PSD-95 at synapses, which were identified by triple labeling with the presynaptic marker protein synaptophysin (Fig. 4) or with antibodies against the NR2A and -2B subunits of the NMDA receptor complex (Fig. 5).
Pyk2 did not interact with the different PDZ domains of PSD-95 in pull-down experiments (Fig. 6), consistent with the fact that the C terminus of Pyk2 (Pro-Pro-Ala-Glu) does not conform to a classic PDZ domain binding motif. Rather, it specifically bound to the SH3 domains of PSD-95 and SAP102. An ϳ300-residue-long C-terminal fragment of full-length Pyk2 containing two proline-rich regions (713-720 and 855-862) was sufficient for binding to the SH3 domain. This portion of Pyk2, known as CRNK, is expressed in parallel with full-length Pyk2 in several tissues including brain (59,78,79), similar to the homologous fragment FAK-related non-kinase of the related focal adhesion kinase FAK (80 -82). Signaling pathways that depend on the interaction of the C-terminal domain of Pyk2 with corresponding binding partners can be inhibited by CRNK (59,78,79). Therefore, CRNK may regulate the Pyk2 interaction with PSD-95 and SAP102 and thereby the efficiency of NMDA receptor regulation by Pyk2. In support of this speculation, the dominant negative Pyk2 mutant K457A (35,39) prevents up-regulation of NMDA receptor peak currents by endogenous Pyk2 (49), suggesting the mutant blocks access of the endogenous wild-type Pyk2 to the SH3 domains of PSD-95 and SAP102 in the NMDA receptor complex.
Only a few binding partners for the SH3 domain of PSD-95 and its homologs have been identified, including the KA2 subunit of the kainate-type glutamate receptor (83), A kinase anchor protein 150 (84), and huntingtin (85). In addition to these interactions, the SH3 and GK domains of PSD-95 and its homologs can undergo intramolecular and, to some extent, intermolecular interactions (20 -24). Mutations that disrupt this intramolecular association in the PSD-95 homolog Dlg result in overproliferation of the imaginal disc in Drosophila (25). These observations suggest that the regulation of proteinprotein interactions between the SH3 and GK domains has a critical function in vivo. Recent experiments indicate that disruption of the SH3-GK domain interaction interferes with the channel clustering activity of PSD-95 (22); however, the precise mechanism underlying this phenomenon is unclear.
The linker domain between the SH3 and GK domain has been implicated as a potential binding site for ligands that might induce conformational changes of the SH3-GK module (21,86,87). This region is also involved in targeting Dlg to septate junctions in epithelial cells (88) and to synapses at the neuromuscular junction (89) of Drosophila. Our present results indicate that the SH3 domain of SAP102, and perhaps also PSD-95, primarily binds Pyk2 when the SH3 domain is not engaged in an intramolecular interaction with the GK domain. We hypothesize that specific regulatory mechanisms, such as association of the GK or linker domain with a binding partner or a post-translational modification, like phosphorylation of SAP102 and PSD-95, control the SH3-GK interaction and, thereby, Pyk2 binding to the SH3 domain.
Several PKC phosphorylation sites have been identified on the NMDA receptor (65, 90 -92) that influence receptor function. For example, phosphorylation of the C1 cassette of the NR1 C terminus may regulate the distribution of the NMDA receptor on the cell surface (91,93). In addition, stimulation of PKC generally results in up-regulation of NMDA receptor activity (94 -99). However, recent findings suggest that potentiation of NMDA receptors is not always due to direct phosphorylation of the receptor subunits by PKC (100). Elegant work by MacDonald and co-workers (33,49) demonstrated that Pyk2 and Src act downstream of PKC to increase NMDA receptor activity. Earlier studies indicated that NR2A and -2B are phosphorylated on tyrosine residues in vivo (26,27,46,101) and that NMDA receptor activity is up-regulated by Src (30 -32,102). Furthermore, the increase in NMDA receptor activity by PKC was prevented by specific inhibitors of Src (33). These observations suggest that intermediary kinases, such as Pyk2 and Src, are involved in the PKC signaling cascade resulting in the phosphorylation of the NMDA receptor. This model is supported by coimmunoprecipitation of Src and the NMDA receptor ( Fig. 1; see also Ref. 34) reflecting the formation of a Src-NMDA receptor complex for efficient and specific phosphorylation of the receptor. However, additional experiments are necessary to characterize further the interaction between Src and the NMDA receptor and the phosphorylation of NR2A and -2B by Src.
How does PKC control Src activity? Stimulation of PKC by phorbol esters or G q -coupled receptors results in activation of Pyk2 (35,38,39,46). Up-regulation of NMDA receptor activity by PKC can be inhibited by injection of a purified, dominant negative Pyk2 mutant into hippocampal pyramidal neurons (49). Furthermore, injection of purified wild-type Pyk2 increases NMDA receptor currents, and this effect is blocked by inhibition of Src. In contrast, the Src-mediated increase in NMDA receptor activity (31)(32)(33) is not antagonized by dominant negative Pyk2 (49). These results place PKC upstream of Pyk2 and Src downstream of Pyk2 in that signaling pathway (49). Because the induction of NMDA receptor-dependent LTP requires Src (34) and Pyk2 (49), we hypothesize that LTP may depend on the coassembly of Pyk2 and Src at the NMDA receptor. Because Ca 2ϩ influx also leads to the activation of Pyk2 independent of PKC stimulation (39,46), anchoring of the kinase at the NMDA receptor may place Pyk2 at a strategically ideal location where NMDA receptor-mediated Ca 2ϩ influx into the cytosol may most effectively contribute to Pyk2 activation. As a result, Pyk2 autophosphorylation could lead to the association and activation of Src (35). However, it is yet unknown how Ca 2ϩ influx or PKC activation induces Pyk2 activation in vivo.
As an integral component of the NMDA receptor complex, Pyk2 plays a critical role in the regulation of NMDA receptor activity and function ( Fig. 1) (49). We demonstrated that Pyk2 binds to the SH3 domain of PSD-95 and SAP102 and that the interaction of Pyk2 with the SH3 domain is inhibited by the GK domain. Our biochemical studies indicate a functional role of the SH3-GK domain interaction in PSD-95 homologs, the regulation of Pyk2 binding. Our results will aid further studies concerning the precise regulation of the NMDA receptor by PKC, Pyk2, and Src signaling and the role of PSD-95 and its homologs in synaptic functions that depend on the targeting of these kinases.