Postsynaptic density 95 (PSD-95) serine 561 phosphorylation regulates a conformational switch and bidirectional dendritic spine structural plasticity

Postsynaptic density 95 (PSD-95) is a major synaptic scaffolding protein that plays a key role in bidirectional synaptic plasticity, which is a process important for learning and memory. It is known that PSD-95 shows increased dynamics upon induction of plasticity. However, the underlying structural and functional changes in PSD-95 that mediate its role in plasticity remain unclear. Here we show that phosphorylation of PSD-95 at Ser-561 in its guanylate kinase (GK) domain, which is mediated by the partitioning-defective 1 (Par1) kinases, regulates a conformational switch and is important for bidirectional plasticity. Using a fluorescence resonance energy transfer (FRET) biosensor, we show that a phosphomimetic mutation of Ser-561 promotes an intramolecular interaction between GK and the nearby Src homology 3 (SH3) domain, leading to a closed conformation, whereas a non-phosphorylatable S561A mutation or inhibition of Par1 kinase activity decreases SH3-GK interaction, causing PSD-95 to adopt an open conformation. In addition, S561A mutation facilitates the interaction between PSD-95 and its binding partners. Fluorescence recovery after photobleaching imaging reveals that the S561A mutant shows increased stability, whereas the phosphomimetic S561D mutation increases PSD-95 dynamics at the synapse. Moreover, molecular replacement of endogenous PSD-95 with the S561A mutant blocks dendritic spine structural plasticity during chemical long-term potentiation and long-term depression. Endogenous Ser-561 phosphorylation is induced by synaptic NMDA receptor activation, and the SH3-GK domains exhibit a Ser-561 phosphorylation-dependent switch to a closed conformation during synaptic plasticity. Our results provide novel mechanistic insight into the regulation of PSD-95 in dendritic spine structural plasticity through phosphorylation-mediated regulation of protein dynamics and conformation.

The ability for synapses to undergo dynamic rearrangements is important for cognitive processes such as learning and memory (1)(2)(3)(4)(5). On a cellular level, synaptic plasticity is reflected by changes in the number and size of dendritic spines, which are mushroom-shaped protrusions on neurons that receive most of the excitatory synaptic inputs in the CNS (6). Dendritic spines enlarge during long-term potentiation (LTP), 4 which is a longlasting increase in synaptic strength (7). Enlarged spines usually contain larger postsynaptic density (PSD), an electron-dense protein complex located underneath the postsynaptic membrane. Larger spines are also associated with increased surface expression of AMPA-type glutamate receptors, which mediate basal synaptic transmission at glutamatergic synapses (8). By contrast, dendritic spines shrink during long-term depression (LTD), which is a long-lasting decrease in synaptic strength. These shrunken spines have smaller PSD and reduced surface expression of AMPA-type glutamate receptors, leading to weaker synaptic transmission (9 -11).
Synaptic plasticity requires the dynamic reorganization of the PSD protein scaffold. One of the key scaffolding proteins at the postsynapse is PSD-95, which is a membrane-associated guanylate kinase (MAGUK) family protein (12). PSD-95 plays a key role in synaptic plasticity as its overexpression blocks LTP and facilitates LTD (13), whereas knockdown of PSD-95 decreases LTD (14,15). At the structural level, PSD-95 contains three PDZ domains on the N terminus followed by a Src homology 3 (SH3) domain and a guanylate kinase (GK) domain. The SH3 and GK domains form an intramolecular interaction with each other (16 -18). This intramolecular interaction is important for PSD-95 function (19,20), and previously identified genetic mutations in discs large (Dlg), the Drosophila homolog of PSD-95, disrupt this interaction (16,21). However, it is unclear how the SH3-GK interaction is regulated under physi-ological conditions and how this interaction may contribute to synaptic plasticity.
We recently identified a phosphorylation site in the GK domain of PSD-95, Ser-561, that is phosphorylated by the microtubule affinity-regulating kinase (MARK)/partitioningdefective 1 (Par1) family of kinases (22). Using pulldown and fluorescence resonance energy transfer (FRET) assays, we now show that Ser-561 phosphorylation regulates a conformational switch in PSD-95 by promoting the intramolecular interaction between the SH3 and GK domains. In addition, we show that a non-phosphorylatable S561A mutation increases the interaction between PSD-95 and its binding partners. Using fluorescence recovery after photobleaching (FRAP) imaging, we further show that a phosphomimetic S561D mutation increases the dynamics of PSD-95 at the synapses, whereas the S561A mutant shows increased stability at the synapse. Moreover, we show that the S561A mutant is resistant to chemical LTP-and LTD-induced spine structural plasticity. Finally, we show that endogenous Ser-561 phosphorylation is induced by synaptic N-methyl-D-aspartic acid (NMDA) receptor activation, and the SH3-GK domains exhibit a Ser-561 phosphorylation-dependent switch to a closed conformation during chemical LTP. Taken together, we propose a model whereby synaptic NMDA receptor activation induces the phosphorylation of Ser-561 in PSD-95, which promotes the SH3-GK interaction and reduces its interaction with other synaptic binding partners. This leads to increased dynamics of PSD-95 at the synapse and allows for rearrangement of the PSD. In the absence of Ser-561 phosphorylation, PSD-95 is stable at the synapse, which renders the spines resistant to chemical LTP-and LTD-induced structural changes.

Ser-561 phosphorylation promotes SH3-GK interaction
Our recent studies show that Par1 phosphorylates PSD-95 on Ser-561 (22). The same conserved site in Drosophila Dlg is phosphorylated by dPar1 and regulates neuromuscular junction formation (23). Ser-561 is located in the GK domain of PSD-95 (Fig. 1a). The GK domain forms an intramolecular interaction with the adjacent SH3 domain that is important for PSD-95 function (16,19). We hypothesize that Ser-561 phosphorylation regulates SH3-GK interaction and thus modulates PSD-95 function. To test this hypothesis, we first performed a GST pulldown assay by incubating GST-SH3 glutathione beads with lysates from HEK293 cells expressing GFP-GK constructs. Beads were then washed, and the resulting protein complex was resolved by SDS-PAGE and Western blotting. Significantly less GFP-GK S561A was pulled down with GST-SH3 as compared with wild-type GFP-GK, whereas more GFP-GK S561D was pulled down as compared with its wild-type counterpart (Fig. 1,  b and c). This suggests that the Ser-561 phosphorylation regulates the interaction between SH3 and GK domains.
To further examine this intramolecular interaction, we constructed a single-chain FRET biosensor (CFP-SH3-GK-YFP) (Fig. 1a). Because the SH3 domain and the GK domain interact, we predicted that YFP and CFP will come close enough for FRET to occur. If the SH3-GK interaction is disrupted, it should result in lower FRET efficiency. Indeed, when the wild-type probe was expressed in HEK293 cells, it showed a strong FRET signal (Fig. 1, d and e). Because the very C terminus of the GK domain forms a ␤-sheet that folds back to the SH3 domain (17,18), we wanted to confirm that the FRET biosensor can detect changes in intramolecular SH3-GK interactions. It has been reported that an L460P mutation in PSD-95 disrupts the intramolecular SH3-GK interaction (16). We found that the L460P mutant FRET probe exhibits significantly reduced FRET efficiency as compared with the WT probe (supplemental Fig. S1), showing that the FRET biosensor can detect changes in intramolecular SH3-GK interactions. To determine the effects of the Ser-561 phosphorylation site, we expressed non-phosphorylatable and phosphomimetic mutants of the FRET probe. We found that the S561A mutant probe showed a significant decrease in FRET signal as compared with the WT probe, whereas the phosphomimetic S561D mutant showed an enhanced FRET signal (Fig. 1, d  and e). Treatment of cells with a Par1 inhibitor resulted in a significant decrease in FRET signals in the WT FRET probe (supplemental Fig. S2), suggesting that Par1-mediated phosphorylation of Ser-561 is necessary for this intramolecular interaction. Overall expression levels of FRET probes were similar for all constructs (supplemental Fig. S3).
We next aimed to further confirm that the FRET signals were a result of intramolecular interactions between the SH3 and GK domains and not from intermolecular interactions. Although intramolecular SH3-GK interactions are the preferred mode of interaction, intermolecular SH3-GK interactions do occur (16). To distinguish between these two possibilities, we performed FRET spectroscopy on serial dilutions of lysates expressing different FRET constructs. At high concentrations, FRET signals can potentially come from both intra-and intermolecular interactions; however, the contribution from intermolecular interactions should drop significantly with dilution. A CFP-YFP fusion protein was used as a control in which the FRET signals should primarily come from intramolecular interactions. As shown in supplemental Fig. S4, FRET ratios in the CFP-YFP fusion protein gradually decreased upon dilution. FRET ratios in the WT and S561D probes decreased at a similar or slower rate than the CFP-YFP fusion protein, indicating that the FRET signals in these two constructs come primarily from intramolecular interactions. By contrast, FRET ratios in the S561A mutant probe rapidly decreased upon dilution, indicating that the FRET signals of the S561A mutant at high concentrations come from intermolecular interactions. This is consistent with previous studies showing that disrupting intramolecular SH3-GK interactions will facilitate intermolecular SH3-GK interactions (16).
Taken together, these results suggest that Ser-561 phosphorylation promotes the intramolecular interaction between the SH3 and GK domains. When Ser-561 is phosphorylated, PSD-95 shows a "closed" conformation with stronger SH3-GK interaction. If Ser-561 is dephosphorylated, PSD-95 switches to an "open" conformation with weaker SH3-GK interaction.

S561A mutation of PSD-95 increases its interaction with synaptic binding partners
The GK domain binds several synaptic proteins that may influence downstream signaling cascades, including guanylate PSD-95 Ser-561 phosphorylation and structural plasticity kinase-associated protein (GKAP) (24), mPins (also known as LGN, named after a series of repeats (Leu-Gly-Asn) found in its sequence) (25), and spine-associated RapGAP (SPAR) (26,27). We examined whether Ser-561 phosphorylation regulates PSD-95 interaction with these proteins using a GST pulldown assay. We purified GST-SH3-GK constructs with or without the Ser-561 mutation. For GKAP, whole-brain lysates were incubated with the purified GST proteins. For SPAR and LGN, we used whole-cell lysates of HEK cells expressing the respective proteins. We then analyzed bound proteins by Western blotting with antibodies targeting the above mentioned molecules. We found that PSD-95 interaction with SPAR, GKAP, Figure 1. Effects of Ser-561 phosphorylation mutants of PSD-95 on SH3-GK interaction. a, a schematic representation of the structure of PSD-95 and the single-chain FRET probes (CFP-SH3-GK-YFP with or without mutation). b, GST pulldown assay. HEK293 cells expressing GFP-GK or GFP-GK S561A/D were incubated with GST-SH3 beads. GST beads were used as a control. The bound proteins were analyzed by Western blotting (WB) with a GFP antibody. c, quantification for the GST pulldown assay revealed that significantly less GFP-GK S561A was pulled down with GST-SH3 as compared with wild-type GFP-GK, whereas more GFP-GK S561D was pulled down as compared with the wild-type counterpart. Error bars in the scatter plot represent S.D. n ϭ 3 independent experiments. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001 by one-way ANOVA with Tukey's post hoc test. d, representative FRET images of the CFP-SH3-GK-YFP and mutants in HEK293 cells. CFP-YFP fusion was used as a positive control, and cotransfected CFP and YFP were used as a negative control. FRET efficiency images were generated using the sensitized emission module of Olympus Fluoview after background subtraction and correction for donor and acceptor bleedthrough and for expression levels of the biosensor. Scale bar, 10 m. e, quantification of FRET efficiency using the acceptor photobleaching method. Significantly decreased FRET efficiency in the FRET construct with S561A mutation and significantly increased FRET efficiency in the S561D mutant were observed. Boxes denote 25th and 75th percentiles, lines within boxes denote medians, and whiskers denote minimum and maximum values. 50 cells per group from three independent experiments were analyzed. ****, p Ͻ 0.0001 by one-way ANOVA with Tukey's post hoc test.

PSD-95 Ser-561 phosphorylation and structural plasticity
and LGN were affected by Ser-561 phosphorylation. The S561A mutant showed increased interaction with these proteins as compared with the wild-type version (Fig. 2, a-f), which was confirmed to be phosphorylated after incubation with wholecell lysates (supplemental Fig. S5) or whole-brain lysates (data not shown). In addition, we found that the S561A mutant also shows increased interaction with endogenous PSD-95 (supplemental Fig. S6), which is consistent with previous reports that disrupting the intramolecular SH3-GK interaction facilitates intermolecular interactions between PSD-95 molecules (16). Taken together, these results suggest that dephosphorylation of Ser-561 facilitates the interaction of the GK domain with its binding partners.

Ser-561 phosphorylation regulates PSD-95 dynamics
Because Ser-561 affects the interaction between PSD-95 and its synaptic binding partners, we wondered whether this would lead to altered PSD-95 dynamics at the synapse. To test this possibility, we used FRAP imaging. Day in vitro (DIV) 24 hippocampal neurons expressing different PSD-95-GFP constructs were imaged live using confocal microscopy. Single spines were photobleached using 100% laser power, and the recovery of PSD-95 fluorescence was observed over time ( Fig.  3). Wild-type PSD-95 recovery reached about 50% after 45 min. By contrast, the recovery of the S561A mutant was less than 10%, whereas the S561D mutant recovered to almost 80%. No significant baseline photobleaching was observed during the imaging period (supplemental Fig. S7). This suggests that Ser-561 phosphorylation increases the dynamics of PSD-95 at the synapse. The increased stability of the S561A mutant at the synapse is consistent with the observed increase in the interaction of PSD-95 with multiple synaptic binding partners.

S561A mutation of PSD-95 inhibits bidirectional spine structural plasticity
Because spine plasticity involves dynamic rearrangements of the PSD scaffold and Ser-561 phosphorylation affects PSD-95 dynamics, we hypothesized that Ser-561 phosphorylation regulates structural plasticity. To test this hypothesis, we used a bicistronic lentiviral vector to deplete endogenous PSD-95 and re-express a GFP-tagged PSD-95. Viral titer was controlled so that the exogenous PSD-95 was expressed at levels equivalent to the endogenous PSD-95 (Fig. 4a). Because PSD-95 overexpression blocks LTP, this strategy allows us to dissect the role of Ser-561 phosphorylation without interference from the effect of PSD-95 overexpression. mRFP was then transfected into the neurons to visualize cell morphology.
To examine the effects of Ser-561 phosphorylation on synaptic plasticity, we first induced chemical LTP in cultured hippocampal neurons by glycine treatment (200 M in Mg 2ϩ -free medium; 3 min). As reported previously, glycine-induced LTP shares many features with stimulus-induced LTP that occurs in the CA1 region of the hippocampus, including activation of synaptic NMDA receptors, increased GluA1 insertion, a requirement for Ca 2ϩ /calmodulin-dependent protein kinase II activation, and dendritic spine enlargement (28 -30). Consistent with previous studies (29,30), dendritic spines showed a significant enlargement in control neurons expressing a replacement wild-type PSD-95-GFP. By contrast, although the PSD-95 S561A mutant did not cause any significant changes in spine size under control conditions, neurons expressing the replacement S561A mutant were resistant to glycine-induced spine enlargement (Fig. 4, b-d). This suggests that Ser-561 phosphorylation is important for spine enlargement during chemical LTP.
Next, we wanted to examine whether Ser-561 phosphorylation is also needed for LTD. To test this possibility, we treated cultured hippocampal neurons with NMDA for 10 min, which is known to induce LTD (31). As expected, control neurons expressing a replacement wild-type PSD-95-GFP showed shrinkage of dendritic spines. By contrast, neurons expressing a replacement PSD-95 S561A showed no significant changes in spine size (Fig. 4, e-g), which suggests that LTD also requires Ser-561 phosphorylation. Taken together, these results indicate that Ser-561 phosphorylation is necessary for bidirectional spine structural plasticity.

Synaptic NMDA receptor activation induces Ser-561 phosphorylation and conformational change in PSD-95
We next aimed to examine potential conformational changes in the SH3-GK domain during synaptic plasticity in live hippocampal neurons. Hippocampal neurons were transfected with the SH3-GK FRET biosensor (WT or S561A) and imaged live at DIV11. Chemical LTP was induced using glycine in Mg 2ϩ -free medium as described above. As expected, in WT FRET biosensor-expressing neurons, we observed a gradual increase in dendritic spine volume upon glycine treatment (Fig.  5a). S561A biosensor-expressing neurons did not show a significant increase in dendritic spine volume upon glycine induction, probably due to a dominant-negative effect of the mutant biosensor. FRET signals in the WT biosensor were significantly increased following glycine-induced chemical LTP as compared with the biosensor with the S561A mutation, suggesting that the SH3-GK domains show a Ser-561 phosphorylation-dependent switch to a closed conformation during synaptic plasticity (Fig. 5, a and b). The increase in FRET signals was observed throughout the spines and the dendritic shafts ( Fig.  5a) likely because of the long time course of the FRET increase, the global nature of the glycine stimulation, and the absence of a synaptic targeting domain in the FRET biosensor. Interestingly, although the S561A biosensor showed an overall lower FRET efficiency, both the WT and S561A biosensors exhibited an increase in FRET efficiency during the initial 25 min after glycine induction. After 25 min, the WT biosensor FRET efficiency continued to increase, whereas the S561A biosensor efficiency started to decrease (Fig. 5b). This suggests that there could be an additional Ser-561 phosphorylation-independent transient conformational change induced by NMDA receptor activation.
Previous studies have confirmed the physiological phosphorylation of Ser-561 in mammalian brains by mass spectrometry (32). However, it remains unknown whether and how this phosphorylation is regulated by synaptic activity. To examine this, we developed a phosphospecific antibody against pSer-561 in PSD-95. The antibody preferentially immunoprecipitated WT PSD-95 as compared with PSD-95 S561A (supplemental Fig.   PSD-95 Ser-561 phosphorylation and structural plasticity PSD-95 Ser-561 phosphorylation and structural plasticity S8a) and failed to immunoprecipitate PSD-95 in calf intestinal alkaline phosphatase (CIP)-treated samples (supplemental Fig.  S8b), showing that it is specific for phosphorylated Ser-561. To see whether Ser-561 phosphorylation is regulated by synaptic NMDA receptor activity, we treated hippocampal neurons with picrotoxin, a GABA A receptor antagonist, and 4-aminopyridine (4-AP), a potassium channel antagonist, a combination that is known to activate synaptic NMDA receptors (33)(34)(35). Consistent with our previous studies, the MARK/Par1 kinases are stimulated by synaptic NMDA receptor activation (36) (Fig.  5c). Remarkably, Ser-561 phosphorylation of PSD-95 was significantly increased by synaptic NMDA receptor activation (Fig. 5, c and d), showing that endogenous phosphorylation of the Ser-561 site is regulated by synaptic NMDA receptors.

Discussion
PSD-95 is perhaps one of the most extensively studied molecules at the postsynaptic density. It is well-known for its involvement in synaptic maturation and plasticity. However, despite numerous studies showing a role for PSD-95 in synaptic plasticity, there is still a lack of detailed mechanistic insight into how PSD-95 is regulated both structurally and functionally during activity-dependent plasticity. Because PSD-95 is a scaffolding protein that lacks any enzymatic activity, one of the likely modes of regulation is through posttranslational modifications that change protein-protein interactions and possibly cause conformational changes in PSD-95. Although several phosphorylation sites have been identified in PSD-95 (37-42), how these sites regulate PSD-95 structure and function during plasticity is still not well-understood.
In this study, we show that Ser-561 phosphorylation of PSD-95 regulates its dynamics and conformation and is important for bidirectional spine structural plasticity. We show that Ser-561 regulates the intramolecular interaction between the SH3 and GK domains of PSD-95. Pulldown experiments and FRET imaging show that Ser-561 phosphorylation promotes the interaction between SH3 and GK domains, leading to a closed conformation of PSD-95, whereas the non-phosphorylatable S561A mutant shows decreased SH3-GK interaction and an open conformation. We further show that the S561A mutant exhibits increased binding to proteins that normally interact with the GK domain, including GKAP, SPAR, and LGN/mPins. These data suggest that dephosphorylation of Ser-561 facilitates protein-protein interactions between the GK domain of PSD-95 and its interacting partners.
In addition, we show that Ser-561 phosphorylation regulates PSD-95 dynamics. Using FRAP imaging, we show that, compared with wild-type PSD-95, the S561A mutant is more stably incorporated at the synapse, whereas the phosphomimetic

PSD-95 Ser-561 phosphorylation and structural plasticity
S561D mutant is more dynamic. This is consistent with our data showing that the S561A mutant exhibits increased interaction with its synaptic binding partners. Moreover, using chemical LTP and LTD protocols, we show that the S561A mutant is resistant to chemical LTP-or LTD-induced changes in spine size, suggesting that Ser-561 phosphorylation is necessary for spine structural plasticity. Finally, we show that endogenous Ser-561 phosphorylation is induced by synaptic NMDA receptor activation, and the SH3-GK domains exhibit a Ser-561 phosphorylation-dependent switch to a closed conformation during glycine-induced LTP. Taken together, these data indicate that Ser-561 phosphorylation is important for bidirectional dendritic spine structural plasticity. Because of the global nature of the chemical LTP and LTD induction methods, it would be of interest to determine the role of the Ser-561 phosphorylation of PSD-95 during plasticity at the single-spine level.
What might be the underlying mechanism by which Ser-561 phosphorylation regulates PSD-95 protein-protein interaction and conformation? Interestingly, Ser-561 is not one of the few residues known to be required for SH3-GK intramolecular binding (17). It is, however, within the GMPbinding site of the GK domain known to mediate the interactions with proteins like GKAP, LGN, and MAP1A (43,44). It is likely that phosphorylation of Ser-561 will block the GMP-binding site, causing weakened interactions between PSD-95 and its GK domain-interacting partners. Furthermore, we speculate that the Ser-561 phosphorylation may allosterically alter SH3-GK interactions, leading to conformational changes in PSD-95. Consistent with our hypothesis, previous studies have shown allosteric control of SH3-GK interactions that involves residues distant from the SH3-GK-binding interface (45).

PSD-95 Ser-561 phosphorylation and structural plasticity
Taken together, our data show a novel role for NMDA receptor-dependent Ser-561 phosphorylation of PSD-95 in regulating its dynamics and conformation and promoting dendritic spine plasticity. Together with our recent data showing that MARK/Par1 is activated downstream of NMDA receptors through a PKA-dependent mechanism (36), we propose the following model (Fig. 5e). When Ser-561 of PSD-95 is dephosphorylated, the SH3-GK domains show an open conformation.

PSD-95 Ser-561 phosphorylation and structural plasticity
In addition, dephosphorylation of Ser-561 facilitates interaction between PSD-95 and its GK domain binding partners, which makes PSD-95 stably incorporated at the synapse. Upon plasticity-inducing stimuli, NMDA receptors are activated, which leads to the downstream activation of MARK/ Par1 and the phosphorylation of Ser-561 in PSD-95. Ser-561 phosphorylation leads to an increase in SH3-GK interaction and switches PSD-95 to the closed conformation. Furthermore, it reduces PSD-95 interaction with its binding partners and increases PSD-95 dynamics. This allows PSD-95 to temporarily leave the PSD scaffold, which permits dynamic rearrangement of the PSD. Our data are consistent with the observation that PSD-95 temporarily leaves the PSD during synaptic plasticity (37). Together, our studies provide novel, mechanistic insight into the regulation of PSD-95 during dendritic spine structural plasticity and may shed light on the regulation of other MAGUK family proteins because the serine site is conserved across multiple members of the MAGUK family (23).

DNA constructs
GW1-PSD-95-GFP construct was a generous gift from Dr. Ann Marie Craig (University of British Columbia). The S561A and S561D mutants of PSD-95 were generated by site-directed mutagenesis using GW1-PSD-95-GFP as the template. For construction of GST and GFP fusion proteins of PSD-95 fragments, the domains (SH3, aa 412-509; GK, aa 510 -724; SH3-GK, aa 412-724) were PCR-amplified using rat PSD-95 in GW1 as the template and subcloned into EcoRI-NotI sites (SH3 and SH3-GK) in pGEX-4T-1 vector or NotI-EcoRI sites (GK) in pKvenus vector. For construction of single-chain FRET probe, the domains (SH3-GK, aa 412-724) were PCR-amplified using rat PSD-95 in GW1 as the template and subcloned into BamHI-EcoRI sites of pKseFRET vector (46). For simultaneous knockdown of endogenous PSD-95 and re-expressing the PSD-95-GFP mutants, PSD-95 shRNA (14) was cloned into the pLVTHM vector after the H1 RNA promoter. PSD-95 constructs with silent mutations that make them resistant to the shRNA were cloned into the same vector after the EF1␣ promoter. pLV-venus-LGN construct was a generous gift from Drs. Yongliang Huo and Ian G. Macara (Vanderbilt University). Myc-SPAR construct was a generous gift from Dr. Daniel Pak (Georgetown University).
For chemical LTP (cLTP) and chemical LTD (cLTD), hippocampal neurons were infected at 0 DIV by lentivirus expressing the bicistronic PSD-95 constructs and transfected at 13 DIV with mRFP to visualize spines. For cLTP, hippocampal neurons were treated with glycine (Sigma-Aldrich; 200 M) in Mg 2ϩ -free Hanks' balanced salt solution (HBSS; Sigma-Aldrich) containing 2 mM CaCl 2 for 3 min on DIV17 and then transferred to HBSS with Ca 2ϩ but without any added glycine for 60 min. For cLTD, hippocampal neurons on the coverslips were treated with NMDA (Sigma-Aldrich; 50 M) in Mg 2ϩ -free HBSS containing 1.8 mM CaCl 2 and 1 M glycine for 10 min on DIV21 and then transferred to the same solution without any added NMDA for 60 min. Neurons were then fixed in 4% paraformaldehyde with 4% sucrose in PBS for 15 min at room temperature. For synaptic NMDA receptor stimulation, neurons were treated with 4-AP (Sigma; 1 mM final) and picrotoxin (Sigma; 10 M final) for 10 min and lysed as described previously (36). Par1/MARK inhibitor (MRT00207148) was a generous gift from Dr. Janet Brownlees (Medical Research Council Technology, London, UK) and was used at 10 M final concentration.

GST pulldown assay, immunoprecipitation, and Western blotting
For GST pulldown assay and co-immunoprecipitation, HEK293 cells expressing different plasmid constructs or wholebrain tissue were lysed on ice in buffer containing 25 mM Hepes, 150 mM NaCl, 10 mM MgCl 2 , 1% Nonidet P-40, and 10 mM DTT and supplemented with protease inhibitor mixture (Sigma-Aldrich), phosphatase inhibitor mixture (Sigma-Aldrich), 10 mM ␤-glycerophosphate, and 10 mM NaF. Lysates were cleared by centrifugation at 13,000 ϫ g for 10 min at 4°C. GST fusion proteins were purified using glutathione-Sepharose beads (GE Healthcare) according to the manufacturer's instructions. Beads with bound GST proteins were incubated with lysates from HEK cells or whole-brain lysates for 3 h at 4°C. For coimmunoprecipitation, cell lysates were incubated with anti-GFP antibody (2 g; Invitrogen A11122) for 1 h at 4°C followed by incubation with 20 l of Dynabeads Protein G, which was preblocked with 5% BSA in lysis buffer. Beads were washed three times with lysis buffer. Bound proteins were eluted with 3ϫ sample buffer and subjected to SDS-PAGE and Western blot analysis.
For detecting endogenous phosphorylation of the Ser-561 site, we developed a rabbit polyclonal phosphospecific antibody against pSer-561 of PSD-95 (AbMart) using the phosphopeptide PDKFGpSCVPHT. pSer-561 antibody (10 g) was used for immunoprecipitation from hippocampal lysates as described above, and the immunoprecipitate was analyzed by Western blotting with PSD-95 antibody. For the CIP control, lysates of HEK293 cells expressing PSD-95 constructs were treated with CIP for 1 h at 37°C before immunoprecipitation with the pSer-561 antibody.
For Western blot analysis, the primary antibodies used were rabbit anti-Myc antibody

FRET imaging and spectroscopy
FRET imaging was performed on an Olympus FV1000 confocal microscope with a LUMPLFLN 60ϫ water immersion lens (numerical aperture, 1.00). A 405 nm laser was used with a 405/440 dichroic mirror to excite CFP. CFP emission was collected in one channel spanning 480 -495 nm, and YFP emission was collected in 535-565 nm. Live-cell FRET images were acquired using the sensitized emission module in Olympus Fluoview software. FRET efficiency images were generated using the sensitized emission module of Olympus Fluoview software after background subtraction and correction for donor and acceptor bleed-through and for expression levels of the biosensor. For FRET efficiency measurements in live cells, images of each channel were acquired before and after photobleaching of YFP. The following formula was used to calculate the FRET efficiency of each construct after background subtraction. For time-lapse FRET imaging in live neurons, ratio imaging of the FRET channel over the CFP channel was performed to visualize FRET efficiency changes over time using the ratio imaging module in the Olympus Fluoview software. FRET/CFP ratios in dendritic spines were quantified and normalized to the FRET/CFP ratio at the 0-min time point.
For FRET spectroscopy, HEK293 cells expressing different FRET constructs were lysed. Emission spectra of the lysates between 450 and 550 nm were measured with 430-nm excitation using a BioTek Synergy H1 multimode plate reader.

FRAP
FRAP experiments was performed at 37°C with an Olympus FV1000 confocal microscope with a LUMPLFLN 60ϫ water immersion lens (numerical aperture, 1.00). Hippocampal neurons were transfected with GFP-tagged PSD-95 constructs. PSD-95 clusters at the spine head were bleached using a 488 nm laser set at 100% laser power. Time-lapse images were collected before and after photobleaching using 5% laser power, and the fluorescent intensities of the bleached spot were quantified using ImageJ and plotted as a function of time.

Image quantification
Dendritic spine morphology was quantified by two individuals who were blind to the experimental conditions, and the results were averaged. Dendritic spines were defined as stubby or mushroom-shaped protrusions and contacted by presynaptic terminals. Spine length and width were measured using ImageJ. Spine length was defined as the length from the tip of the spine head to the point where the spine joins the dendrite. Spine width was defined as the maximal width of the spine head perpendicular to the long axis of the spine neck. 1500 -1800 spines from at least 15 neurons for cLTP and 1200 -1600 spines from at least 15 neurons for cLTD were measured.

Statistical analysis
Statistical analysis was performed using GraphPad Prism. For all Western blot data, scatter plots are shown with mean Ϯ S.D. The number of independent biological replicates is indicated in the figure legends. Quantification of imaging data are shown as either mean Ϯ S.D. or box-whisker plots. Boxes denote 25th and 75th percentiles, lines within the boxes denote medians, and whiskers denote minimum and maximum values. The number of neurons/dendritic spines quantified is indicated in the figure legends. Data were analyzed by one-way or twoway ANOVA with Tukey's post hoc test for multiple group comparisons. Student's t test was used for comparison between two groups. One-sample t test was used for Western blot data in which the control levels are normalized to 1. p Ͻ 0.05 is considered statistically significant.