Molecular mechanisms regulating the differential association of kainate receptor subunits with SAP90/PSD-95 and SAP97.

Recent studies have demonstrated that kainate receptors are associated with members of the SAP90/PSD-95 family (synapse-associated proteins (SAPs)) in neurons and that SAP90 can cluster and modify the electrophysiological properties of GluR6/KA2 kainate receptors when co-expressed in transfected cells. In vivo, SAP90 tightly binds kainate receptor subunits, while SAP97 is only weakly associated, suggesting that this glutamate receptor differentially associates with SAP90/PSD-95 family members. Here, green fluorescent protein (GFP)-tagged chimeras and deletion mutants of SAP97 and SAP90 were employed to define the molecular mechanism underlying their differential association with kainate receptors. Our results show that a weak interaction between GluR6 and the PDZ1 domain of SAP97 can account for the weak association of GluR6 with the full-length SAP97 observed in vivo. Expression studies in HEK293 cells and in vitro binding studies further show that although the individual Src homology 3 and guanylate kinase domains in SAP97 can interact with the C-terminal tail of KA2 subunit, specific intramolecular interactions in SAP97 (e.g. the SAP97 N terminus (S97N) binding to the Src homology 3 domain) interfere with KA2 binding to the full-length molecule. Because receptor subunits are known to segregate to different parts of the neuron, our results imply that differential association of kainate receptors with SAP family proteins may be one mechanism of subcellular localization.

(SH3), and a catalytically inactive guanylate kinase (GUK) domain (3)(4)(5). Each domain has been shown to be a site of protein-protein interaction, allowing SAP90/PSD-95 family members to interact with ligand-and voltage-gated ion channels, cell adhesion molecules, and cytoskeletal proteins as well as proteins involved in intracellular signaling pathways (2,6). In isolation, analogous domains from each family member have been shown to exhibit similar binding specificities, yet in vivo the full-length molecules are differentially distributed and selectively associate with specific subclasses of ionotropic glutamate receptors. For example, SAP90/PSD-95, SAP102, and chapsyn-110/PSD-93 are all found highly concentrated at and tightly bound to the postsynaptic density (PSD) of type 1 glutamatergic synapses, where they are known to associate with subunits of the N-methyl-D-aspartate receptors and kainate receptors (7)(8)(9)(10). In contrast, SAP97, which is less tightly associated with these synapses, interacts preferentially with ␣-amino-3-hydroxy-5-methyl-isoxazole-4-propionic acid (AMPA) receptors (11). Differences between SAP97 and the other SAPs are also observed in transfected HEK293 (9,12). While SAP90/ PSD-95 is readily able to cluster Shaker K ϩ channels at the plasma membrane of HEK293 cells (9), SAP97 traps these channels intracellularly in an endoplasmic reticulum-like compartment (12,13). Given their striking structural similarities, these observations raise the important issue of how SAP family members differentially select their binding partners.
Recent studies point to at least two types of mechanisms governing the selection process. The first appears to involve small differences in the sequences flanking the conserved PDZ, SH3, and GUK domains. For example, sequences N-terminal to the first PDZ domain in SAP90/PSD-95 and SAP97 (S90N and S97N, respectively) have been found responsible for the differential localization of these two proteins in neurons (14,15) and nonneuronal cells (16) as well as the ability of SAP90/PSD-95 to cluster ion channels in HEK293 cells (13). The second mechanism appears to involve a series of intramolecular interactions between the individual domains (17), which in turn regulate both GKAP binding to the GUK domain of SAP97 (18) and Kv1.4 clustering by SAP90/PSD-95 (19).
Typically, kainate receptors are heteromeric receptors composed of various combinations of GluR5, GluR6, GluR7, KA1, or KA2 subunits (20,21). In a recent study, we found that SAP90 binds the kainate receptor subunits, GluR6 and KA2 (10). In contrast, the SAP97 fails to interact with KA2 and only weakly binds GluR6 (10). Studies on heteromeric GluR6/KA2 receptors also revealed that SAP90/PSD-95 can associate via its PDZ1 domain with the C-terminal ETMA sequence of the GluR6 subunit and the C terminus of KA2 subunits via its SH3 and GUK domains (10). Importantly, these interactions cause reduced desensitization of kainate receptors (10).
To understand better the molecular bases of SAP binding partner specificity, we have evaluated what structural features in SAP97 are functionally important for interfering with its association with kainate receptor subunits in vivo. Our results show that GluR6 association with SAP97 is feeble due to a weak interaction between the PDZ1 and GluR6 subunit. In contrast, we also find that although both the SH3 and the GUK domain in SAP97 can interact with KA2, a combination of intrinsic features in SAP97 interfere with its ability to interact with KA2. Specifically, intramolecular interactions between the N terminus of SAP97 and its SH3 domain play a dominant role in preventing its binding to the kainate receptor KA2 subunit.

EXPERIMENTAL PROCEDURES
Mammalian DNA Expression Vectors-The subcloning of cDNAs encoding SAP97 and SAP90 into eukaryotic expression vectors with an N-terminal green fluorescent protein (GFP) tag has been described previously (16). SAP90 with a C-terminal GFP tag was created by cloning the coding region of SAP90 into the pEGFP-N1 vector (CLON-TECH). The SAP97 molecules in which specific subdomains such as the I3 (⌬I3), S97N (⌬S97N), PDZ1-3 (⌬PDZ1-3), and SH3(⌬SH3) domains are deleted have been described (16). GFP-PDZ1 S90 and GFP-PDZ1 S97 were constructed by subcloning polymerase chain reaction-amplified DNA fragments directionally into the EcoRI-SalI sites of pEGFP-C2. Single amino acid changes in the SH3 domain of SAP90 were generated by the QuickChange protocol (Stratagene). The cDNAs encoding KA2 and GluR6 (kindly provided by Dr. S. Heinemann) were subcloned into the pcDNA3 expression vector (Invitrogen).
Protein Preparation, Immunoblotting, and Immunoprecipitation-For immunoprecipitation experiments, fresh adult rat brains were homogenized and centrifuged at 4°C for 1 h at 165,000 ϫ g. Insoluble pellets were dissolved in radioimmune precipitation buffer with 2% SDS. Solubilized extracts were passed several times through a 25-gauge needle, diluted to a final SDS concentration of 0.33%, and centrifuged at 10,000 ϫ g for 5 min. Transfected cell extracts were prepared in a similar fashion, except 0.1% SDS was used in the solubilization buffer. Immunoprecipitations were performed by incubating extracts with 5 g of anti-KA2 (Upstate Biotechnology, Inc., Lake Placid, NY), 5 g of anti-GluR1 (Upstate Biotechnology), or 5 g of anti-GFP (Quantum Biotechnology) antibodies for 2 h at 4°C, followed by incubation with 100 l of protein G-Sepharose (Amersham Pharmacia Biotech) for 12-16 h and four washes in radioimmune precipitation buffer. Western blot analysis of the depleted cell extracts showed no detectable anti-KA2, anti-GluR1, or anti-GFP immunoreactivity, suggesting that immunoprecipitation was complete. Bound proteins were eluted from the beads by boiling and separated by SDS-PAGE. Gels were then blotted and immunostained as previously described (10). Primary antibodies used for Western blotting included mouse monoclonal anti-SAP90 (1: 250; Transduction Laboratories), rabbit polyclonal anti-GST (1:1000; Sigma), mouse monoclonal anti-SAP97 (1:2000; Stressgen), mouse monoclonal anti-T7 (1:10,000; Novagen), rabbit polyclonal anti-GluR1 (1 g/ml; Upstate Biotechnology), rabbit polyclonal anti-KA2 (1 g/ml; Upstate Biotechnology), and rabbit polyclonal anti-GluR6 (1 g/ml; Upstate Biotechnology).
Construction and Purification of Fusion Proteins-H 6 KA2(c-term) expressed in pTrcHisB (Invitrogen) was prepared as described by Garcia et al. (10). H 6 S97N was prepared by subcloning the polymerase chain reaction fragment containing amino acids 1-104 of SAP97 in pRSETC vector (Invitrogen). Synthesis of recombinant proteins in TOP10 cells (Invitrogen) was induced by 1 mM isopropyl-␤-D-thiogalactopyranoside for 12-14 h at 37°C. Cells were then harvested, resuspended in 50 mM phosphate buffer (pH 8.0) with 300 mM NaCl and 8 M urea, lysed by French press as directed by the manufacturer (SLM Instruments, Inc.), and centrifuged at 70,000 ϫ g for 25 min at 20°C. Supernatant containing H 6 KA2(c-term) was bound to Ni 2ϩ -nitrilotriacetic acid beads (Qiagen) for 1 h at room temperature with mixing and then loaded into a column. After washing, bound H 6 KA2(c-term) was eluted using an imidazole step gradient from 50 to 500 mM and then dialyzed in a series of buffers with decreasing urea concentrations (from 6 to 1 M) to renature the protein and remove the imidazole.
In Vitro Binding Assay-For each sample, 800 pmol of H 6 KA2(cterm) or H 6 S97N were incubated with 200 pmol of GST fusion proteins in radioimmune precipitation buffer (1% Nonidet P-40, 0.5% sodium deoxycholate, 100 g/ml phenylmethylsulfonyl fluoride, 10 g/ml aprotinin, 10 g/ml leupeptin, and 10 g/ml pepstatin A in phosphatebuffered saline) containing 10 mM EDTA and 40 mM imidazole. After 6 -8 h of continuous rotation at 4°C, 25 l of Ni 2ϩ -nitrilotriacetic acid-Sepharose beads were added. Samples were then incubated for 1 h and washed four times, and bound proteins were eluted with 50 l of 500 mM imidazole. The eluted proteins were separated by 12% SDS-PAGE, transferred to nitrocellulose, probed for GST, and visualized by chemiluminesence as directed by the manufacturer (Amersham Pharmacia Biotech). In competition experiments, 200 pmol of H 6 KA2 were incubated with 50 pmol of GST fusion proteins and 0 -1600 pmol of H 6 S97N in radioimmune precipitation buffer (1% Nonidet P-40, 0.5% sodium deoxycholate, 100 g/ml phenylmethylsulfonyl fluoride, 10 g/ml aprotinin, 10 g/ml leupeptin, and 10 g/ml pepstatin A in phosphate-buffered saline) containing 10 mM EDTA and 40 mM imidazole. The samples were processed as above, blotted, and probed using anti-KA2 antibodies (0.05 g/ml; Upstate Biotechnology). To compare the bindings of SH3 S90 and SH3 S90 (W-A) to H 6 KA2 and H 6 S97N, 400 pmol of H 6 KA2(c-term) or H 6 S97N were incubated with 200 pmol of GST-SH3 fusion proteins in phosphate-buffered saline with 100 g/ml phenylmethylsulfonyl fluoride, 10 g/ml aprotinin, 10 g/ml leupeptin, 10 g/ml pepstatin A, and 40 mM imidazole. After 6 -8 h of continuous rotation at 4°C, 25 l of glutathione-Sepharose beads were added, and samples were incubated and washed as above, with elution in 50 l of 10 mM glutathione.

Differential Interaction of SAP90 and SAP97 with the GluR6 and KA2 Subunits of Kainate
Receptors-In our previous study, co-immunoprecipitations from rat brain extracts showed that the GluR6 and KA2 subunits of kainate receptors associated with SAP90 (10). These experiments also showed a weak association between GluR6 and SAP97 and no association between KA2 and SAP97. Given the striking sequence homology between SAP90 and SAP97, this result was unexpected and raised the possibility that either SAP97 and kainate receptors do not coexist in the same subcellular compartments or that the structure of SAP97 is subtly different from SAP90, preventing its association with kainate receptors. To evaluate the first possibility, GluR6 or KA2 subunits were co-expressed with GFP-tagged SAP90 or SAP97 in HEK293 cells. After solubilization, immunoprecipitation was performed with anti-KA2 or anti-GFP antibodies, and the immunoprecipitated products, separated by SDS-PAGE, were probed with antibodies against SAP90, SAP97, or GluR6 ( Fig. 1). Similar to our previous results using rat brain extracts, both the GluR6 and KA2 subunits interacted with SAP90, while GluR6 interacted weakly with SAP97 ( Fig. 1, left) and no detectable interaction was seen between KA2 and SAP97 (Fig. 1, right). These results indicate that even when SAP97 and kainate receptors are co-expressed in HEK293 cells, they do not interact, suggesting that structural differences between SAP90 and SAP97 may be responsible for their selective interaction with kainate receptors.
Binding of PDZ1 Domain of SAPs to GluR6 Determines Their Interaction-Previously, we showed that SAP90 binding to GluR6 is mediated via its PDZ1 domain (10). Although sequence alignment of the PDZ1 domain of SAP90 and SAP97 indicates about 90% identity, subtle change(s) in the amino acid sequence(s) of PDZ1 could be responsible for the selective interaction of SAPs with GluR6. To evaluate this possibility, we co-expressed GFP-tagged PDZ1 domains from SAP90 or SAP97 (GFP-PDZ1 S90 and GFP-PDZ1 S97 ) with the GluR6 subunit in HEK293 cells. Co-immunoprecipitations reveal that the GluR6 subunit binds well to GFP-PDZ1 S90 and weakly to GFP-PDZ1 S97 ( Fig. 2A, left). This demonstrates that the amino acid sequence of the PDZ1 domain of SAPs is critical for its selective binding to GluR6.
SAP97 has also been shown to interact with the GluR1 subunit of AMPA receptors (11), and it has been speculated that the interaction is mediated via PDZ domains (11,22). In order to show that GFP-PDZ1 S97 forms a functional domain, we co-expressed it with the GluR1 subunit. Immunoprecipitation with the anti-GluR1 antibody reveals that GluR1 binds well to the GFP-PDZ1 S97 (Fig. 2B, right). Taken together, these data suggest that the PDZ1 domain of SAPs is an important determinant for their differential interaction with GluR6.
The C Terminus of KA2 Can Interact with the SH3 and GUK Domains of SAP97-SAP90 binds to the cytoplasmic C-terminal tail of KA2 via its SH3 and GUK domains (10). Although the SH3 domains are not well conserved between SAP90 and SAP97, GUK domains share a high degree of sequence identity (23). We were therefore curious whether the SH3 and GUK domains of SAP97 could interact with C terminus of KA2. This was addressed by utilizing a "pull down" assay with several segments from the C-terminal half of SAP97 and SAP90, including the SH3 (GST-SH3 S97 and GST-SH3 S90 ), GUK (GST-GUK S97 and GST-GUK S90 ), or contiguous SH3 and GUK domains (GST-SG97 and GST-SG90) fused to GST and the C terminus of KA2 expressed with a histidine (H 6 ) tag (H 6 KA2). The purity of each recombinant purified protein used in this assay was assessed by SDS-PAGE. Each exhibited a single band after staining with Coomassie Brilliant Blue except for the GST-SH3 S90 and GST-SH3 S97 proteins (Fig. 3, lower gel), which are subject to some degradation in Escherichia coli. In our pull down assay, GST fusion proteins were incubated in solution with H 6 KA2 and then Ni 2ϩ beads to bind out H 6 KA2 and any interacting GST fusion proteins. Bound proteins were eluted, immunoblotted, and probed with an anti-GST antibody.
Based on the relative amounts of the GST fusion protein detected after immunoblotting (Fig. 3, upper panel), the SH3 domain of SAP97 (SH3 S97 ) appears to have a higher affinity for KA2 than the SH3 domain of SAP90 (SH3 S90 ). The GUK domains of both SAP97 (GUK S97 ) and SAP90 (GUK S90 ) bind equally well to KA2. Given that the SH3 and GUK domains of SAPs bind to KA2 in a similar manner, we would conclude that the inability of KA2 to bind full-length SAP97 is not due to deficiency in SH3 and/or GUK binding. Previous studies (17,18) raise the possibility that the inability to bind KA2 may involve sequences that either directly flank the SH3 or GUK domains, sequences upstream of the SH3 domain, or intramolecular interactions either between the SH3 and GUK domains or with other regions in SAP97. To address whether an SH3-GUK intramolecular interaction affected KA2 binding, we assessed KA2 binding to the GST-fused SH3-GUK proteins from both SAP90 and SAP97 (GST-SG90 and GST-SG97). KA2 was able to bind SG90 and SG97, albeit at a slightly lower level than SH3 alone, indicating that this intramolecular interaction is not important for the differential binding of KA2 to SAP90 versus SAP97. These data suggest that some other sequence elements unique to SAP97 must interfere with KA2 binding. N Terminus of SAP97 Prevents Its Association with KA2-A comparison of the deduced amino acid sequences of SAP97 and SAP90 reveals that the N-terminal region preceding the first PDZ domain (S97N and S90N, respectively) and the sequence situated between the SH3 and GUK domains (U5 region) are the most divergent (18,23). For example, the amino terminus of SAP97 is 187 amino acids in length and, based on molecular modeling, can be folded into two fibronectin-like domains, while the region containing 63 amino acids in the N terminus of SAP90 has an unordered structure (18). Similarly, the U5 region of SAP97 is longer than that found in SAP90 and can have short insert sequences of 1-33 amino acids that arise from alternative splicing. The most predominantly expressed form of SAP97 used in these studies contains the I3 insert (18). We therefore hypothesized that differences in the primary structure of SAP97 might govern its unusual behavior. To test this possibility, we assessed the ability of several GFP-tagged SAP97 deletion constructs (Fig. 4A) to interact with KA2 subunits expressed in HEK293 cells. GFP-tagged SAP97 deletion mutants included constructs in which the N terminus (SAP97⌬S97N), PDZ1-3 (SAP97⌬PDZ1-3), SH3 (SAP97⌬SH3), I3 (SAP97⌬I3), or GUK (SAP97⌬GUK) domains were deleted. Full-length chimeras in which the S90N was replaced by the S97N (S97N/ S90PDZ1-GUK) or the GUK S90 domain for the GUK S97 domain in SAP97 (SAP97N-I3/S90GUK) were also tested. A GFP tag that does not affect SAP97's biological activity (16) was used to follow the protein expression levels and to monitor transfection efficiencies.
The results from Fig. 4B show that SAP97 constructs with deleted PDZ, I3, or GUK domains do not bind to immunoprecipitated KA2, suggesting that inhibitory elements are absent in these regions. In contrast, deleting the S97N region was found to permit KA2 binding, indicating that this domain inhibits the interaction with KA2. Interestingly, a construct lacking the SH3 domain showed a weak interaction with KA2. Since KA2 can interact with both the SH3 and GUK domains of SAP97, weak association of SAP97⌬SH3 with KA2 is likely to be via the GUK domain. As anticipated, the lack of interaction between chimeric SAP97N-I3/S90GUK and KA2 indicates that conservative GUK domain replacement has no effect on SAP97-KA2 binding. These observations suggest that the S97N exerts an inhibitory affect on the interaction of SAP97 with KA2.
How Does S97N Inhibit KA2 Binding?-The finding that the N terminus of SAP97 prevents its association with KA2 suggests that an intramolecular interaction between the N terminus and SH3 and/or GUK domains of SAP97 may prevent KA2 interaction. Given that the C terminus of KA2 interacts with SH3 via PXXP motifs (10), S97N, which also has several proline-rich sequences containing PXXP motifs, could potentially bind to its own SH3 domain. We tested this by employing an in vitro binding assay using the histidinetagged recombinant N terminus of SAP97 (H 6 S97N) and GST-fusion proteins. As described above, binding to H 6 S97N was done in solution, and the ability of GST-fusion proteins to interact with H 6 S97N was identified using Western blots. We find that H 6 S97N binds to the SH3 S97 and SG97 domains but not to the GUK S97 domain (Fig. 5).
Since the N terminus of SAP97 appears to block its binding to KA2 by intramolecular interaction with its SH3 domain, we speculated that replacement of the N terminus of SAP90 with S97N would result in a chimera (S97N/S90PDZ1-GUK) that is structurally similar to SAP97 and would not associate with KA2. Surprisingly, our results show that the S97N/S90PDZ1-GUK chimera can interact with KA2 (Fig. 4B). To investigate this unexpected finding, we compared the binding of S97N to SH3 and SG proteins from SAP90 and SAP97. As shown in Fig.  5, H 6 S97N binds much better to SH3 and SG proteins from SAP97 than SAP90, indicating that a weaker intramolecular interaction between the S97N and SH3 may allow KA2 binding to the S97N/S90PDZ1-GUK.
KA2 can interact with both the SH3 and the GUK domains of SAP97 in vitro (Fig. 3B). However, in full-length SAP97, both of the binding sites appear to be occluded. The higher affinity binding site on the SH3 domain appears to be masked by an intramolecular association between S97N and SH3, while the lower affinity binding site on the GUK domain seems to be blocked by interaction between the SH3 and GUK domains. This theory is further supported by the binding of KA2 to the SAP97 with either its N terminus or SH3 domain deleted (Fig.  4B). Together, the data suggest that intramolecular interactions between the S97N-SH3 and SH3-GUK domains of SAP97 are strong enough to prevent any KA2 binding to SAP97.
N Terminus Can Compete with KA2 for SH3 Binding-Our results indicate that in vivo S97N binding to SH3 may prevent KA2 association with SAP97. In other words, the effective intramolecular binding of SH3 to the S97N appears to be stronger than intermolecular KA2 binding. To evaluate this, we performed an in vitro in-solution competition assay. Recombinant GST-SH3 S97 or GST-SG97 (50 pmol) were incubated with 200 pmol of H 6 KA2 and varying amounts (0 -1600 pmol) of H 6 S97N. After overnight incubation, glutathione beads were added, and specific binding of H 6 KA2 to the GST-fusion proteins was assessed. As shown in Fig. 6A, increasing concentrations of H 6 S97N can progressively compete H 6 KA2 binding to SH3 S90 and SG97. Since the intramolecular binding between the SH3 and GUK domains slightly lowers the binding of KA2 to the SH3 domain, results from this experiment consistently show that binding of H 6 KA2 is stronger to SH3 S97 , as compared with SG97 (Figs. 3 and 6A). As a result, lower concentrations of H 6 S97N are required to compete off H 6 KA2 binding to SG97 than SH3 S90 (Fig. 6A). Our data support the hypothesis that this type of competition exists in vivo and that intramolecular interactions (between S97N and SH3 S97 ) interfere with intermolecular interactions (between KA2 and SH3 S97 ). However, these results do not indicate whether H 6 S97N and H 6 KA2 bind to the same site or distinct but overlapping sites on SH3.
We reported earlier (10) that H 6 KA2 binding to SH3 S90 could be blocked by a tryptophan to alanine mutation (at position 470) in SH3, which has also been reported to inhibit other SH3-mediated protein-protein interactions (24). The potential importance of this site was tested by in vitro binding of GST-SH3 or GST-SH3(W-A) to H 6 KA2 or H 6 S97N. The bound complexes were pulled down by Ni 2ϩ beads and probed for GST. We found that H 6 S97N binds to the mutated SH3 as well as wild type (Fig. 6B, compare lanes 5 and 6). In contrast, there was minimal binding of H 6 KA2 to mutated SH3 (Fig. 6B, compare  lanes 2 and 3), indicating that tryptophan is a critical residue for SH3 interaction with KA2 but not with S97N. Because the tryptophan is required for most SH3-ligand interactions (24), this suggests that the S97N-SH3 interaction is atypical and that S97N may block the KA2 interaction via steric hindrance.

Differential Interaction of SAP97 with Kainate Receptor Subunits-
Previous studies indicate that the GluR6 and KA2 subunit of kainate receptors interact strongly with SAP90 and SAP102 (10). However, while GluR6 weakly associates with SAP97, KA2 does not. These unusual findings prompted us to investigate the underlying mechanisms. Since SAP90 and SAP97 are homologous, a weak interaction between SAP97 and GluR6 or no interaction between SAP97 and KA2 could be explained if the kainate receptors and SAP97 had different subneuronal localizations. However, expression of GluR6 and SAP97 or KA2 and SAP97 in a heterologous system gave similar results, suggesting that the level of interaction of SAP97 with kainate receptors is not related to their subcellular distribution. Instead, our data suggest that SAP97 is structurally impaired to interact with kainate receptors and specifically indicate that the PDZ1 domain of SAP97 differs from that of SAP90 in its ability to interact with GluR6: the binding between PDZ1 S90 and GluR6 is much higher than binding between PDZ1 S97 and GluR6, thus providing an explanation for the reduced interaction between GluR6 and wild-type SAP97. In addition, the GluR1 subunit of AMPA receptors have been hypothesized to interact with PDZ domains of SAP97 (11,22). We show here that PDZ1 S97 can bind to the GluR1 subunit.
Earlier reports have shown that the SH3 and the GUK domains mediate KA2 interaction with SAP90 (10). The sequence of the GUK domain is well conserved between SAP90 and SAP97, as compared with the SH3 domain. Our results indicate that neither the GUK domain nor the SH3 domain of SAP97 is functionally impaired in their capacity to interact with KA2 yet reveal that KA2 does not have access to these sites in full-length SAP97.
Intramolecular Binding between the S97N and the SH3 Domain of SAP97 Prevents Intermolecular Binding of KA2 and SAP97-Although SAP90 and SAP97 are homologous, SAP97 contains an extended N terminus and an insertion in the U5 region between the SH3 and GUK domains. Based on our data indicating that SAP97 lacking S97N binds well to KA2 and that in vitro the S97N can bind its own SH3 domain, we have identified S97N as the negative element preventing the interaction between KA2 and SAP97. Recent biochemical and structural studies have revealed that S97N association with its own SH3 and U5 regions modulates GKAP binding (18). Our findings indicate that an interaction between the SAP97 N terminus and SH3 domain occludes binding to KA2. However, since KA2 interacts with both the SH3 and GUK domains, it should be able to interact with SAP97 via the GUK domain. Nonetheless, SAP97 does not interact with KA2 in vivo or in vitro. One possible reason could be that the affinity of KA2 for the GUK domain is about 5-fold less than for the SH3 domain. Alternatively, an SH3-GUK intramolecular interaction may block the GUK binding site for KA2. Based on our results, a SH3-GUK interaction appears to be an important factor, since the SAP97 with deleted SH3 can bind to KA2 with lower affinity.
The sequence of the SH3 domain is not well conserved between SAP97 and SAP90. Interestingly, the N terminus of SAP97 binds SH3 597 and SG97 better than SH3 590 and SG90. These in vitro binding data are consistent with the co-immunoprecipitation results from HEK cells transfected with KA2 and a SAP90/SAP97 chimera with an N-terminal replacement of SAP90 with SAP97. The chimeric protein is similar in structure to SAP97 but binds to KA2. Since S97N binds much better to SG97 than SG90, the inhibition by S97N may not be strong enough to ward off KA2. Competition studies show that in vitro S97N can more easily compete KA2 binding to SG97 than SH3 S97 . From these results, it appears that the presence of a GUK domain contiguous with its SH3 domain has some inhibitory effects on KA2 interaction, possibly due to intramolecular SH3-GUK interactions.
Interaction between the S97N and the SH3 Domain Is Atypical-SH3 domains are known to interact with proline-rich sequences containing the consensus sequence PXXP (25,26). However, binding of a SAP SH3 domain to its GUK domain is an exception (17)(18)(19), since the GUK does not have any such motif. Our results show that binding of the N terminus of SAP97 to its SH3 may also be atypical. Although the N terminus contains proline-rich sequences and has PXXP motifs, it interacts well with the SH3 containing a W-A mutation, which normally blocks any PXXP-mediated binding to SH3 domains. It is conceivable that the N terminus may still bind to SH3 via its PXXP motifs, but the tryptophan residue is not important. Alternatively, some other sequences at the N terminus maybe required for its interaction with SH3. Nevertheless, our results show that intramolecular interactions between the N terminus of SAP97 and SH3 can deny access of KA2 to its potential binding site (Fig. 7).
These effects appear to be similar to the ones described for Src and Tec family kinases (27,28), where intramolecular FIG. 4. N terminus of SAP97 prevents its association with KA2. A, a schematic diagram shows various deletion and chimeric GFP-SAP97 constructs used to identify regions responsible for inhibition of KA2 binding. N terminus, PDZ1-3, SH3, I3, or GUK domains were deleted in SAP97⌬S97N, SAP97⌬PDZ1-3, SAP97-⌬SH3, SAP97⌬I3, or SAP97⌬GUK constructs, respectively. The S97N/S90PDZ1-GUK chimera is SAP90 with S90N replaced with S97N, and the SAP97N-I3/ S90GUK chimera is SAP97 with the GUK S97 domain replaced with the GUK S90 domain. B, the N terminus of SAP97 blocks its interaction with KA2. KA2 was immunoprecipitated from the HEK293 cell extracts expressing GFP-SAP fusion proteins and KA2, blotted and probed with anti-GFP antibodies. SAP97 with SH3 or Nterminal deletions and the S97N/ S90PDZ1-GUK chimera interact with KA2. WCE, whole cell extract; IP, immunoprecipitation; WB, Western blot. interactions between the SH3 and catalytic domains cause autoinhibition. However, once an extrinsic ligand binds the SH3 domain, the intermolecular interaction interferes with the intramolecular interaction, freeing the catalytic domain and making it active. It is conceivable that a high affinity ligand for the N terminus of SAP97 may exist in vivo to shift SAP97 into a more open state, exposing its SH3 domain and making it available to bind KA2. Potentially, any resultant binding might be too transient to be detected by standard co-immunoprecipitation or immunocytochemical assays. Alternatively, SAP97 intramolecular interactions could be part of a basic molecular mechanism used by cells to allow the differential association of SAPs with binding partners. FIG. 6. SAP97 N terminus can compete with KA2 for SH3 binding. A, increasing concentrations of the N terminus of SAP97 can block the interactions between KA2 and the SH3 domain. 50 pmol of GST-SH3 S97 or -SG97 fusion proteins were incubated overnight with 200 pmol of H 6 KA2 and varying amounts of H 6 S97N (0 -1600 pmol). Glutathione beads were added, and GST-fusion proteins and any associated proteins were eluted, run on SDS-PAGE, and Western blotted for H 6 KA2 using anti-KA2 antibodies. Higher concentrations of H 6 S97N are required to compete off H 6 KA2 binding to GST-SH3 S97 than to GST-SG97. The GST lane does not show any binding to H 6 KA2, indicating the specificity of interactions. B, interaction between the SAP97 N terminus and SH3 domain is atypical. Equimolar amounts of GST-SH3 S90 and GST-SH3 S90 (W-A) were incubated with H 6 KA2 or H 6 S97N overnight. The samples were processed as in Fig. 4. H 6 KA2 binds only to the wild-type SH3 S90 , while H 6 S97N binds almost equally well to both the wild-type SH3 S90 and the mutant SH3 S90 . WB, Western blot. FIG. 7. Mechanism regulating the differential interaction of the KA2 subunit with SAPs. SAP90 can cluster KA2 receptors by interacting with the SH3 and GUK domains. Although KA2 can potentially bind to the SH3 and GUK domains of SAP97 with, respectively, high and lower affinities, the SH3 site is occluded by binding of the SAP97 N terminus, and the GUK site is occluded by SH3 binding. Effectively, these intramolecular interactions block SAP97 association with KA2.