Phosphorylation of Synaptic GTPase-activating Protein (synGAP) by Ca2+/Calmodulin-dependent Protein Kinase II (CaMKII) and Cyclin-dependent Kinase 5 (CDK5) Alters the Ratio of Its GAP Activity toward Ras and Rap GTPases*

Background: synGAP inactivates Ras and Rap at synapses. Results: Phosphorylation of synGAP by CaMKII increases Rap1 GAP activity more than HRas GAP activity; phosphorylation by CDK5 has the opposite effect. Conclusion: Phosphorylation by CaMKII and CDK5 alters the ratio of Rap1 and HRas GAP activities. Significance: Phosphorylation of synGAP by CaMKII and CDK5 can alter the balance of synaptic functions regulated by Ras and Rap. synGAP is a neuron-specific Ras and Rap GTPase-activating protein (GAP) found in high concentrations in the postsynaptic density (PSD) fraction from the mammalian forebrain. We have previously shown that, in situ in the PSD fraction or in recombinant form in Sf9 cell membranes, synGAP is phosphorylated by Ca2+/calmodulin-dependent protein kinase II (CaMKII), another prominent component of the PSD. Here, we show that recombinant synGAP (r-synGAP), lacking 102 residues at the N terminus, can be purified in soluble form and is phosphorylated by cyclin-dependent kinase 5 (CDK5) as well as by CaMKII. Phosphorylation of r-synGAP by CaMKII increases its HRas GAP activity by 25% and its Rap1 GAP activity by 76%. Conversely, phosphorylation by CDK5 increases r-synGAP's HRas GAP activity by 98% and its Rap1 GAP activity by 20%. Thus, phosphorylation by both kinases increases synGAP activity; CaMKII shifts the relative GAP activity toward inactivation of Rap1, and CDK5 shifts the relative activity toward inactivation of HRas. GAP activity toward Rap2 is not altered by phosphorylation by either kinase. CDK5 phosphorylates synGAP primarily at two sites, Ser-773 and Ser-802. Phosphorylation at Ser-773 inhibits r-synGAP activity, and phosphorylation at Ser-802 increases it. However, the net effect of concurrent phosphorylation of both sites, Ser-773 and Ser-802, is an increase in GAP activity. synGAP is phosphorylated at Ser-773 and Ser-802 in the PSD fraction, and its phosphorylation by CDK5 and CaMKII is differentially regulated by activation of NMDA-type glutamate receptors in cultured neurons.

synGAP is tightly associated with the postsynaptic plasma membrane and binds to the PSD-95, Discs-large, ZO1 (PDZ) domains of PSD-95 (1,2), which positions it in close proximity to the NMDA-type and AMPA-type glutamate receptors (NMDARs and AMPARs), respectively. Loss of one copy of the synGAP gene produces cognitive dysfunction in mice (7) and humans (8), indicating that the functions of synGAP are critically important for normal synaptic regulation. Influx of Ca 2ϩ through NMDARs and binding of ligands to TrkB (BDNF/neurotrophin-3 growth factors receptor B) receptors can activate Ras and Rap postsynaptically (3,9,10). Over time, Ras and Rap modulate synaptic strength in opposite directions. Active Ras increases insertion (exocytosis) of AMPARs at the synapse, whereas active Rap increases their removal (endocytosis) from the synapse (10). Here, we investigated whether phosphorylation of individual residues on synGAP could alter the ratio of its Ras and Rap GAP activities such that the balance of these two activities, and thus the balance of active Ras and Rap, could be changed under different physiological conditions.
We previously showed that activation of NMDARs on cultured CNS neurons leads to phosphorylation of synGAP by CaMKII (1,9,11), and we identified several major sites on partially purified, membrane-bound synGAP where phosphorylation by CaMKII increases its Ras GAP activity (11). Here we show, with purified and soluble recombinant synGAP lacking the first 102 N-terminal residues (r-synGAP), that phosphorylation by CaMKII increases its HRas and Rap1 GAP activities by 25 and 76%, respectively, but does not affect its Rap2 GAP activity.
We found that phosphorylation of r-synGAP by CDK5 increases its HRas and Rap1 GAP activities by 98 and 20%, respectively. Thus, phosphorylation of synGAP by CaMKII accelerates the rate of inactivation of Rap1 more potently than the rate of inactivation of HRas, whereas phosphorylation by CDK5 has the opposite effect. We show that Ser-773 and Ser-802 are both phosphorylated in a mouse PSD fraction. Furthermore, phosphorylation by the two kinases is dynamically regulated by NMDA receptors in cultured neurons. These results mean that differential phosphorylation of synGAP by CaMKII and CDK5 may alter the proportions of activated Ras and Rap in synapses with consequent effects on cellular processes regulated by the two GTPases.

EXPERIMENTAL PROCEDURES
Cloning, Expression, and Purification of r-synGAP-Soluble recombinant synGAP comprising residues 103-1293 (r-syn-GAP) was purified from Escherichia coli. Pena et al. (6) had previously shown that removal of the N-terminal 102 residues and several hundred residues from the C terminus enabled purification of a soluble active fragment of synGAP comprising residues 103-725 (sr-synGAP). Gene fragments encoding rat r-synGAP and sr-synGAP (AF048976) were amplified by PCR with PfuUltra II fusion HS (Agilent, catalog no. 600670) and cloned by the polymerase incomplete primer extension-ligation-independent cloning method (18,19) into pET-47b(ϩ) (EMD Millipore, catalog no. 71461) in-frame with an N-terminal His 6 tag and a PreScission protease cleavage site. The DNA sequence was verified by Sanger sequencing (Laragen). The plasmids were transformed into the Rosetta2(DE3) E. coli strain (EMD Millipore, catalog no. 71397) for protein expression. Single colonies were grown overnight at 37°C in 5 ml of Miller lysogeny broth (LB; EMD Millipore, catalog no. 71753-6) supplemented with 50 g/ml kanamycin and 34 g/ml chloramphenicol. Cultures grown overnight were diluted 1:500 into LB and grown at 37°C for 4 -6 h to an A 600 of 1.0. Cultures were then chilled to 18°C, and protein expression was induced with 0.2 mM isopropyl 1-thio-␤-D-galactopyranoside for 24 h. Bacterial pellets were harvested by centrifugation and resuspended in Lysis/Wash Buffer (20 mM Tris, 500 mM NaCl, 10 mM tris(2carboxyethyl)phosphine (TCEP; Gold Biotechnology, catalog no. TCEP25), 5 mM MgCl 2 , 60 mM imidazole, 2 mM ATP, 1 mM PMSF (Sigma, catalog no. P76726), 0.2% Tergitol TM -type Nonidet P-40, and Complete EDTA-free protease inhibitor (Roche Applied Science, catalog no. 5056489001), pH 7.0) supplemented with 25 units/ml benzonase (EMD Millipore, catalog no. 71206) and 10 kilounits/ml ReadyLyse (Epicenter, catalog no. R1810M). Cells were lysed with an ML-110 microfluidizer (Microfluidics), and insoluble material was removed by centrifugation at 30,000 ϫ g. r-synGAP in the supernatant was bound to 1.5 ml of Talon metal affinity resin per g of cell pellet (Clontech, catalog no. 635503) by batch absorption for 90 min. The resin was poured into a gravity-fed column and washed with 10 column volumes of Lysis/Wash Buffer. r-synGAP was eluted with Lysis/Wash Buffer (Ϫ2 mM ATP) supplemented with 10 mM EDTA and 250 mM imidazole. Fractions containing syn-GAP were pooled and concentrated by ultrafiltration through a 150-kDa cutoff filter (Thermo Scientific, catalog no. 89923) for r-synGAP, and a 9-kDa cutoff filter (Thermo Scientific, catalog no. 89885A) for sr-synGAP. The presence of 0.2% Tergitol TM was necessary to prevent loss of synGAP on the filter. Concentrated samples of r-synGAP (0.25-1 mg/ml) were then exchanged into Storage Buffer by ultrafiltration (20 mM Tris, 500 mM NaCl, 10 mM TCEP, 5 mM MgCl 2 , 1 mM PMSF, 0.2% Tergitol TM -type Nonidet P-40, and Complete EDTA-free protease inhibitor, pH 7.0), flash-frozen in liquid nitrogen, and stored at Ϫ80°C. Concentrated samples of sr-synGAP were further purified on a HiLoad 26/600 Superdex 200 preparation grade column (GE Healthcare, catalog no. 28-9893-36) equilibrated with Storage Buffer. Fractions containing sr-synGAP were pooled, concentrated by ultrafiltration to 2 mg/ml, flashfrozen in liquid nitrogen, and stored at Ϫ80°C. All mutant forms of r-synGAP were purified by the same procedure as for wild type r-synGAP.
Cloning, Expression, and Purification of HRas, Rap1, and Rap2-Gene fragments coding for full-length rat HRas (AAA42009.1), human Rap1B (CAB46488), and human Rap2A (CAA31052.1) were amplified by PCR and inserted into pGEX-6P-1 expression vectors (GE Healthcare, catalog no. 28-9546-48) at a site bound by EcoRI and XhoI recognition sites and in-frame with an N-terminal GST tag and a PreScission Protease cleavage site. The sequence of each plasmid was verified, and they were transformed into Rosetta2 E. coli cells (EMD Millipore, catalog no. 71402-4) for protein expression. Single colonies were grown overnight at 37°C in 5 ml of LB supplemented with 100 g/ml carbenicillin and 34 g/ml chloramphenicol. Overnight cultures were diluted 1:500 into LB supplemented with 2% glucose and grown at 37°C for ϳ4-6 h to an A 600 of 1.0. Cultures were then chilled to 22°C, and protein expression was induced with 0.1 mM isopropyl 1-thio-␤-Dgalactopyranoside for 4 h. Bacterial pellets were harvested by centrifugation, resuspended in GTPase Lysis/Wash Buffer (50 mM Tris, 100 mM NaCl, 5 mM TCEP, 5 mM MgCl 2 , 1 mM EDTA, 1 mM PMSF, 30 M GDP, 0.1% Triton X-100, Complete protease inhibitor, pH 7.5) supplemented with 25 units/ml benzonase, and 10 kilounits/ml ReadyLyse, and lysed by fluidization. Insoluble material was removed by centrifugation at 30,000 ϫ g. The expressed protein in the supernatant was bound to 0.3 ml of glutathione-agarose resin per g of cell pellet (Thermo Scientific, catalog no. 16100) by batch absorption for 3 h. The resin was poured into a gravity-fed column and washed with 20 column volumes GTPase Lysis/Wash Buffer. GTPases were eluted with Cleavage Buffer (50 mM Tris, 100 mM NaCl, 5 mM TCEP, 5 mM MgCl 2 , 1 mM EDTA, 30 M GDP, 0.1% Triton X-100, and 50 mM glutathione, pH 8.0). Fractions containing the GTPases were pooled and concentrated by ultrafiltration through a 9-kDa cutoff filter. Concentrated samples of GTPase were mixed with 20 units/ml PreScission Protease (GE Healthcare, catalog no. 27-0843-01) and cleaved for 18 h at 4°C to release GST. Samples were exchanged into SEC Buffer (20 mM Tris, 100 mM NaCl, 5 mM TCEP, 5 mM MgCl 2 , 1 mM PMSF, 30 M GDP, and 10% glycerol, pH 8.0) and concentrated by ultrafiltration before being fractionated on a HiLoad 26/600 Superdex 75 preparation grade column (GE Healthcare, catalog no. 28-9893-33) equilibrated with SEC buffer. Fractions containing GTPases were pooled, concentrated by ultrafiltration to greater than 1 mg/ml, flash-frozen in liquid nitrogen, and stored at Ϫ80°C. After SEC, all three GTPases were purified to Ͼ95% purity, with yields of ϳ25%. The intrinsic k cat values of purified GTPases, measured in our assay, were 0.0008, 0.0001, and 0.00008 s Ϫ1 , for HRas, Rap1, and Rap2, respectively. Site-directed Mutagenesis of CDK5 Phosphorylation Sites in r-synGAP-Ten r-synGAP mutants were prepared to study the effects of specific CDK5 phosphorylation sites on synGAP activity. They included single mutations S773A, T775A, S802A, S773D, T775D, and S802D; double mutations S773A/T775A, S773D/T775D, and S773D/S802D; and the triple mutation S773A/T775A/S802A. Oligonucleotides encoding these mutations were designed with the QuikChange II primer design program (Agilent) and were synthesized by Integrated DNA Technologies (IDT). Mutations of synGAP were made with the QuikChange II XL site-directed mutagenesis kit (Agilent, catalog no. 200522).

SDS-PAGE and Assessment of Protein Purity and Yield-We
Site-directed Mutagenesis of CaMKII Phosphorylation Sites in r-synGAP-A mutant r-synGAP (CTM-Plus) was constructed based on the CTM-mutant, published earlier (11), by adding mutations of newly identified CaMKII phosphorylation sites to determine their effects on synGAP activity. Serine and threonine residues at sites 750, 751, 756, 764, 765, 1058, 1062, 1064, 1093, 1095, 1099, 1123, and 1125 in synGAP were all replaced with alanine residues. A pair of complementary oligonucleotides encoding serine to alanine mutations of 750, 751, 756, 764, and 765 was designed with DNA2.0 Gene Designer (version 1.1.4) (DNA2.0) and IDT Oligo Analyzer software and synthesized by IDT. Mutagenesis of r-synGAP was performed with a QuikChange II XL site-directed mutagenesis kit, as described previously (20), to generate the CTM-half. Next, a pair of large oligonucleotides encoding mutations to alanine at sites 1058, 1062, 1064, 1093, 1095, 1099, 1123, and 1125 was designed and synthesized as described above. The second oligonucleotide pair was inserted into the CTM-half cDNA by the polymerase incomplete primer extension-ligation-independent cloning method (18,19) to generate the CTM-Plus cDNA. CTM-Plus was transformed into One-shot TOP10 chemically competent cells (Life Technologies, catalog no. C4040-06) for propagation, and the DNA sequence was verified.
Stoichiometry and Rate of r-synGAP Phosphorylation by CaMKII and CDK5-Phosphorylation of r-synGAP by CaMKII (purified as described in Ref. 21) or by CDK5/p35 (EMD Millipore, catalog no. 14-477M) was carried out in a 250-l reaction mixture containing 50 mM Tris-HCl, pH 8.0, 10 mM MgCl 2 , 0 (CDK5) or 0.7 mM (CaMKII) CaCl 2 , 0.4 mM EGTA, 30 or 500 M (as indicated) [␥-32 P]ATP (100 -375 cpm/pmol), 0 (CDK5) or 3.375 M (CaMKII) calmodulin, 10 mM DTT, 286 nM r-syn-GAP, and 3.1 nM rat brain CaMKII or 110 nM CDK5/p35. Carrier-free ␥-labeled ATP (6000 Ci/mmol, PerkinElmer Life Sciences, catalog no. BLU002Z/NEG002Z) was diluted to the desired specific activity by addition of unlabeled ATP (Sigma, catalog no. A2383-25G). Phosphorylation was initiated by addition of ATP to a reaction mixture prewarmed to 30°C. At the indicated time points, reaction aliquots were removed and quenched by the addition of 0.25 volume of ice-cold 4ϫ LDS Sample Buffer containing 10% ␤-mercaptoethanol. Samples were fractionated by SDS-PAGE as described above. Gels were rinsed in water, wrapped in Saran Wrap, and exposed to a Stor-age Phosphor screen (GE Healthcare, catalog no. 63-0034-79) for 2-4 h to locate phosphorylated proteins. To quantify the amount of phosphate incorporated, the density of 32 P bands was determined by scanning with a Typhoon LA 9000 Phos-phorImager (GE Healthcare) followed by analysis with Image-Quant TL software (GE Healthcare, catalog no. 29000737). Relative densities were converted to picomoles of phosphate by comparison with densities from standard amounts of [␥-32 P]ATP spotted onto filter paper and imaged simultaneously. The stoichiometry of phosphorylation was calculated by dividing the picomoles of incorporated phosphate by the amount of synGAP (3.8 pmol) loaded per lane.
GTPase Assays-The GTPases were loaded with [␥-32 P]GTP by nucleotide exchange as described previously (15,22,23). The GTPases pre-loaded in this way undergo one round of hydrolysis of the bound GTP at fixed, slow intrinsic rates as follows: ϳ4 -8 ϫ 10 Ϫ4 s Ϫ1 (Ras) and ϳ0.5-1 ϫ 10 Ϫ4 s Ϫ1 (Rap1 and -2). When synGAP is added to an assay under these conditions, it acts as a catalyst, accelerating the rate of the single hydrolysis reactions. The acceleration of hydrolysis of GTP by HRas, Rap1, or Rap2 by synGAP was assayed by measuring the rate of release of [␥-32 P]phosphate as described previously (24). Briefly, phosphorylated or nonphosphorylated synGAP (250 -500 nM) was mixed with increasing concentrations of each GTPase bound to radioactive [␥-32 P]GTP at 25°C in 6.5 l of Assay Buffer (25 mM Tris, pH 8.0, 100 mM NaCl, 15 mM MgCl 2 , and 5 mM DTT) as described previously (15). After 10 min, the assay was quenched with activated charcoal in phosphate buffer to adsorb proteins GTP and GDP. Following removal of the charcoal, free [␥-32 P]phosphate was quantified by scintillation counting in Beta Blend scintillant (MP Biomedicals, catalog no. 0188245001). Initial rates of GTP hydrolysis for each concentration of GTPase with and without synGAP were determined after fitting the data by linear regression. The rates were plotted against the corresponding concentrations of GTPase with Prism software (version 6.0d, GraphPad Software, La Jolla, CA), and the data were fit to a hyperbolic curve by nonlinear regression. The Michaelis-Menten equation was used to calculate k cat and K m values.
Mass Spectrometry of Phosphorylated r-synGAP-Mass spectrometry of r-synGAP was performed by the Proteome Exploration Laboratory at the California Institute of Technology. r-synGAP was phosphorylated for 2 and 10 min in the presence of 30 M ATP as described above, except with unlabeled ATP. All liquid chromatography-mass spectrometry (LC-MS) experiments were performed on an EASY-nLC (Proxeon Biosystems, now Thermo Scientific) connected to a hybrid LTQ-FT (Thermo Scientific) equipped with a nano-electrospray ion source (Proxeon Biosystems, now Thermo Scientific) essentially as described previously (25), with the modifications indicated below. Peptides were separated on a 15-cm reverse phase analytical column (75 m inner diameter) packed in-house with 3 m ReproSil-Pur C18AQ beads (Dr. Maisch GmbH, catalog no. r13.aq.) with a 60-min gradient at a flow rate of 350 nl/min. The gradient was run from 0 to 40% solvent B (97.8% acetonitrile, 0.2% formic acid, 2% water) in 40 min, increased to 100% B in 1 min, and run at 100% B for 19 min. The mass spectrometer was operated in data-dependent mode to automatically switch between full scan MS and MS2 acquisition. Survey full scan mass spectra were acquired in ion cyclotron resonance, over a range of 400 -1800 m/z, following accumulation of 1,000,000 ions, with a resolution of 100,000 at 400 m/z. The top five most intense ions from each survey scan were isolated and, after the accumulation of 5000 ions, fragmented in the linear ion trap by collision-induced dissociation (collisional energy 35% and isolation width 2 Da). Additionally, an MS3 neutral loss scan was acquired to detect loss of phosphoric acid (32.67, 49.00, or 97.97 m/z) in the top five most intense peaks in the MS2. Precursor ion charge state screening was enabled, and all singly charged and unassigned charge states were rejected. The dynamic exclusion list was set with a maximum retention time of 90 s and a relative mass window of 10 ppm, and early expiration was enabled.
Tandem mass spectra were converted to mgf files with the use of ReAdW4Mascot2 (peptide.nist.gov/metrics). MS3 spectra were separated from MS2 spectra, and accurate masses were determined for MS3 precursor spectra by the method of Timm et al. (26). All MS/MS samples were analyzed with the use of Mascot (version 2.2.06, MatrixScience). We constructed a sequence database containing the sequence of synGAP and common contaminants, including keratins and trypsin (262 entries). Trypsin was specified as the digestion enzyme, and up to two missed cleavages were allowed. We set mass tolerances of 10 ppm for parent ions and 0.50 Da for fragment ions. Carbamidomethylation of cysteine (ϩ57.0215 Da) was specified in Mascot as a fixed modification, and oxidation of methionine (ϩ15.9949 Da), dehydration of serine and threonine (Ϫ18.0106 Da, only MS3 spectra), and phosphorylation of serine, threonine, and tyrosine (ϩ79.9663 Da, only MS2 spectra) were specified as variable modifications. Mascot results were loaded into Scaffold (version 3.00.06, Proteome software) and filtered at a protein probability of 99% with a minimum of two peptides and a peptide probability of 95%. We calculated false localization rates of phosphorylation sites using the Mascot Delta-score as described by Savitski et al. (27). In this algorithm, peptides with only one possible phosphorylation site are given a false localization rate of zero. Summaries of data and annotated spectra for all phosphorylation sites are included in the supplemental material.
Detection of Phosphorylated Ser-773 and Ser-802 in the PSD Fraction by Mass Spectrometry-Aliquots of PSD fraction from mice (65-125 g), prepared as described below, were incubated with 0 or 30 M ATP for 10 min in 120 l of the reaction mix described above. In some experiments, exogenous CDK5/p35 (1-2 units) and/or 100 M roscovitine (Sigma, catalog no. R7772-5MG) were added. Proteins were fractionated by SDS-PAGE and stained with Gel Code Blue, and the 140 -160-kDa bands containing synGAP were excised and digested with trypsin in the gel as described previously (28). synGAP samples were analyzed on an Orbitrap Fusion (Thermo Scientific) equipped with a nano-electrospray ion source coupled to an EASY-nLC. Peptides were separated on a 25-cm reverse phase analytical column (75 m inner diameter) packed in-house with 3 m ReproSil-Pur C18AQ beads with a 100-min gradient at a flow rate of 350 nl/min. The gradient ran from 2% solvent B (80% acetonitrile, 0.2% formic acid, 19.8% water) to 40% solvent B in 90 min, then increased to 100% in 1 min, and held at 100% for 9 min. A parallel reaction-monitoring strategy (29) was employed to monitor the levels of phosphorylation on synGAP at sites Ser-773 and Ser-802. Precursor ions were isolated in the quadrupole, fragmented by HCD, and sent to the Orbitrap measurement (isolation width 1 Da, normalized collision energy of 35%, AGC target of 100,000 ions with a maximum injection time of 50 ms). The peptides monitored for phosphorylation at site Ser-773 included LPSPTK (m/z 361.6779, z ϭ 2), GLNSSMDMARLPSPTK (m/z 595.6072, z ϭ 3), GLNSSMD-MARLPSPTK (m/z 606.2705, z ϭ 3), and LPSPTKEKP-PPPPPGGGK (m/z 620.9922, z ϭ 3). The peptide monitored for phosphorylation at site Ser-802 was SSPAYCTSSSDITEPEQK (m/z 1033.9166, z ϭ 2). Special caution was taken to identify the correct chromatographic peak based on fragment ions unique to phosphorylation at Ser-802 and not Tyr-805, Thr-807, Ser-808, Ser-809, or Ser-180. Three additional nonmodified peptides from synGAP were used to control for loading amounts. These peptides included VIQNLANFSK (m/z 567.3193, z ϭ 2), SASGDTVFWGEHFEFNNLPAVR (m/z 827.3925, z ϭ 3), and EFAEYVTNHYR (m/z 714.8308, z ϭ 2). Raw data were imported into Skyline (30) for manual validation and extraction of peak areas. Statistical analysis of the variation in phosphorylation levels was carried out using MSstats (31).
Preparation of the PSD Fraction from Mouse Forebrain-The PSD fraction was prepared as described previously (32) by a modification of the method of Carlin et al. (33). We have been unable to extract synGAP from the PSD fraction without destroying its GAP activity.
Preparation and Treatment of Cortical Neuronal Cultures-Cultures of cortical neurons were prepared from fetal mice as described previously (9,34). After 4 days in vitro cultures were treated with 5 M cytosine arabinoside (Sigma, catalog no. C6645) to halt the growth of non-neuronal cells. For treatments, cultures (days in vitro [13][14] were washed in HEPES/ control salt solution (HCSS: 20 mM HEPES, 120 mM NaCl, 5 mM KCl, 5.5 mM glucose, 0.8 mM MgCl 2 , pH 7.4) and then exposed to 25 M NMDA (Tocris, catalog no. 114) dissolved in HCSS (2-4 wells for each condition) for 0, 0.25, 1, 3, and 5 min. In some experiments, cultures were pretreated prior to exposure to NMDA with 30 M roscovitine (Sigma, catalog no. R7772-5MG), 5 M cyclosporin A (inhibits calcineurin; Sigma catalog no. 30024), 1 M okadaic acid (inhibits PP2A strongly and PP1 weakly at this concentration; EMD, catalog no. 459620), or 500 nM tautomycin (inhibits PP1; EMD catalog no. 580551), as stated. For experiments in which CDK5 activity was inhibited with roscovitine, control wells were treated with an equal amount of vehicle (DMSO at ϳ0.12%). After treatment, cells were lysed, and protein concentrations were determined by the bicinchoninic acid method with BSA as the protein standard.

RESULTS
Purification of r-synGAP-It was previously shown that a shortened form of synGAP (sr-synGAP) comprising residues 103-725 can be expressed in soluble form in E. coli and that this construct displays GAP activity against Rap with a potency similar to other Rap-GAPs. In contrast, its GAP activity against Ras is considerably less potent than that of other RasGAPs (6). sr-synGAP includes the N-terminal pleckstrin homology, C2, and RasGAP domains, but it lacks residues 726 -1293 that contain a predicted disordered domain containing the regulatory CaMKII phosphorylation sites (11), a short predicted coiled coil domain, and the C-terminal PDZ domain-binding motif (1,2). To enable study of the regulatory roles of phosphorylation of purified synGAP by CaMKII and other protein kinases, we  FEBRUARY 20, 2015 • VOLUME 290 • NUMBER 8 expressed and purified a nearly full-length soluble synGAP fragment (r-synGAP; residues 103-1293) (Fig. 1B). We also purified sr-synGAP for comparison (Fig. 1A). We found that both of these shortened forms of synGAP were soluble when expressed in bacteria at low temperature (e.g. 18°C), but they required addition of ATP and Mg 2ϩ to release bound chaperone (GroEL) and to permit effective binding of the recombinant protein to Talon metal affinity columns during purification (data not shown). A combination of immobilized metal affinity chromatography and SEC was used to purify the synGAP constructs as described under "Experimental Procedures." sr-syn-GAP was purified to Ͼ95% purity, with typical yields in excess of 75%. r-synGAP was purified by immobilized metal affinity chromatography alone to ϳ70% purity, with yields in excess of 85%. r-synGAP eluted in the void volume of SEC columns with no improvement in purity, suggesting that the presence of the disordered domain and/or the coiled-coil domain may predispose r-synGAP to multimerization. Fig. 1C is a domain map of synGAP showing the boundaries of the two recombinant forms and the positions of previously identified phosphorylation sites.

Distinct Regulation of Ras and Rap GAP Activities of synGAP
Intrinsic GAP Activity of r-synGAP-synGAP has been shown to stimulate the GTPase activity of Rap much more potently than that of Ras (5, 6). sr-synGAP, containing the pleckstrin homology, C2, and Ras GAP domains, was found to stimulate the intrinsic GAP activity of Rap1 by 4 orders of magnitude, to a k cat of 0.8 s Ϫ1 , and that of HRas by 2 orders of magnitude (6).
We attempted to measure the enzymatic constants for the acceleration of GAP activity of HRas, Rap1, and Rap2 by limiting amounts of r-synGAP or sr-synGAP ( Fig. 2 and Table 1). In our assay, both synGAP constructs stimulated the rate of GTP hydrolysis by HRas by 2 orders of magnitude ( Fig. 2A), consistent with previous findings (6). Stimulation of GTPase activity by sr-synGAP varied approximately linearly with HRas concentration in the range that we were able to test (10 -700 M). Therefore, we were not able to estimate k cat or K m values. At 700 M HRas, k obs in the presence of sr-synGAP was 0.13 s Ϫ1 . In . Data are mean Ϯ S.E. (n ϭ 3-6 for r-synGAP and n ϭ 1-2 for sr-synGAP). The k obs was calculated as moles of phosphate released per s/mol of synGAP when synGAP is present, and as moles of phosphate released per s/mol of GTPase concentration in the absence of synGAP. Note that the concentration of synGAP is much less than that of the GTPases, so the increase in k obs with GTPase concentration reflects increased binding of the GTPases to limiting synGAP. See Table 1 for estimates of k cat and K m . •, ϩsynGAP; OE, ϪsynGAP.
contrast, stimulation of HRas GTPase by r-synGAP varied nonlinearly with HRas concentration and could be fit by a hyperbolic curve, permitting estimation of k cat ϭ 0.12 Ϯ 0.05 s Ϫ1 and K m ϭ 640 Ϯ 400 M. The error of these calculations is high; thus, they can only be considered estimates.
Stimulation of GTPase activity of Rap1 by both sr-synGAP and r-synGAP followed a hyperbolic curve with increasing Rap1 concentrations (Fig. 2B). For sr-synGAP, we calculated k cat ϭ 0.47 Ϯ 0.06 s Ϫ1 , which is in reasonable agreement with the value of 0.8 s Ϫ1 reported by Pena et al. (6), and K m ϭ 200 Ϯ 55 M. r-synGAP stimulated the GTPase activity of Rap1 to a k cat of 3.8 Ϯ 1.1 s Ϫ1 , which is greater than the stimulation by sr-syn-GAP. However, the affinity of r-synGAP for Rap1 appeared considerably lower (K m ϭ 1800 Ϯ 600 M) than that of sr-synGAP. The fold stimulations of Rap1 GTPase activity by sr-synGAP and r-synGAP were 4300 and 35,000, respectively. These results suggest that the carboxyl portion of r-synGAP (residues 726 -1293) alters the interaction of synGAP with Rap1, producing a lower affinity, but a greater maximum stimulation of its GTPase activity.
Stimulation of GTPase activity of Rap2 by sr-synGAP was linear over the range of Rap2 concentrations we were able to test (15-400 M), again precluding estimation of k cat and K m values (Fig. 2C). r-synGAP stimulated Rap2 GTPase to approximately the same rate as did sr-synGAP. However, for r-syn-GAP, the plot of k obs against the Rap2 concentration was slightly hyperbolic, permitting rough estimates of k cat ϭ 6.8 Ϯ 3 s Ϫ1 and K m ϭ 3100 Ϯ 1500 M. This represents a stimulation of GTP hydrolysis in the presence of r-synGAP of ϳ87,000-fold over the rate of hydrolysis by Rap2 alone (Table 1). Because we could not calculate a k cat for sr-synGAP, it is not clear whether the carboxyl-half of r-synGAP alters its interaction with Rap2 as it appears to with Rap1.
Our measured K m values are relatively high compared with many other enzymes, indicating a relatively low affinity of syn-GAP for the GTPases. However, other dual specificity GAPs and Rap GAPs have K m values in the range of tens to a few hundred micromolars (15,22). A high K m value can be advantageous because it permits a linear response of catalytic rate over a wide range of substrate concentrations (35). The importance of this property in spine synapses is not clear because the effective concentrations of activated Ras or Rap in the confined space of the postsynaptic density after different in vivo stimuli are not known.
Stoichiometry and Rate of r-synGAP Phosphorylation by CDK5-Preliminary mass spectrometric analysis of synGAP in the PSD fraction had revealed phosphorylation at Ser-773 within a motif recognized by CDK5 (data not shown). Therefore, we tested whether purified CDK5 could phosphorylate purified r-synGAP in vitro. Under the conditions of our assay (Fig. 3A), phosphorylation of r-synGAP by CDK5 was approximately linear with time for 10 min, after which r-synGAP contained ϳ0.3 to 0.4 mol of phosphate/mol. By 60 min, the stoichiometry approached an asymptote of 1 mol of phosphate/mol of r-synGAP.
Phosphorylation of 0.3 M r-synGAP by CDK5/p35 occurred at a similar rate (0.03 mol of phosphate/mol of substrate/min) as phosphorylation of the CDK5 substrate histone H1 at a concentration of 4.3 M (data not shown). These results indicate that r-synGAP is at least as good a substrate for CDK5 as histone H1.
Identification of Sites in r-synGAP Phosphorylated by CDK5-We used sensitive proteomic mass spectrometric methods to identify all sites on r-synGAP phosphorylated by CDK5. r-syn-GAP was incubated with CDK5/p35 for 2 or 10 min and quenched with LDS stop buffer as described under "Experimental Procedures." Following separation by SDS-PAGE, phosphorylated r-synGAP was reduced, alkylated, and digested in the gel with trypsin as described under "Experimental Procedures." Tryptic peptides were fractionated by nano-LC and analyzed on a hybrid mass spectrometer equipped with a linear ion trap in combination with Fourier transform ion cyclotron resonance. Ions collected in the linear ion trap were fragmented by collision-induced dissociation (MS2) as described under "Experimental Procedures." If a neutral loss of 32.67, 49.00, or 97.97 Da was observed in MS2, an MS3 neutral loss scan was acquired to detect the presence of dehydroalanine at phosphorylation sites, resulting from the ␤-elimination of H 3 PO 4 from phosphoserine. Following completion of data collection and analysis via Mascot, as described under "Experimental Procedures," results were stringently filtered to a protein probability of 99%, a two peptide minimum, and a peptide probability of 95% in Scaffold as described under "Experimental Procedures." The predicted false localization rate of phosphorylation site identifications was then assessed for each phosphopeptide using the Mascot Delta score as described previously (27). Peptide sequence coverage of r-synGAP was in excess of 84%, with 90% coverage of the disordered domain (91.9% theoretical maximum coverage of r-synGAP calculated by ExPASy Peptide Mass Program). We found no phosphorylated sites on r-syn-GAP before it was incubated with CaMKII or CDK5. We confirmed that Ser-773 and Ser-802 are phosphorylated by CDK5, and we detected three additional sites (Ser-728, Thr-775, and Ser-842) ( Table 2 and supplemental data). Of the identified CDK5 phosphorylation sites, only Ser-773 and Ser-802 are located in a consensus motif for phosphorylation by CDK5 ((S/T)P). Ser-773 is contained in a favored CDK5 consensus sequence with a Lys at the ϩ3 position (SPX(R/K)). Although Ser-773 had a predicted phosphorylation false localization rate of less than 1%, Ser-802 had a predicted rate greater than 5% because it has a neighboring serine. However, close inspection of mass spectra from peptides that are predicted to be phosphorylated at Ser-802, and high mascot scores (ϳ65-85) for their MS2 and MS3 spectra, confirmed the presence of phosphorylation at Ser-802. Because of its close proximity to Ser-773, phosphorylation at Thr-775 may serve the same regulatory function as Ser-773, and/or its phosphorylation may result from promiscuity of the enzymatic phosphorylation. We chose to focus our efforts on Ser-773 and Ser-802 because of their localization within CDK5 consensus sequences and their phosphorylation in synGAP within the PSD (see below). However, we note that Ser-728 and Ser-842 may also be legitimate CDK5 phosphorylation sites.
Stoichiometry and Rate of r-synGAP Phosphorylation by CaMKII-In preparation for comparison of the effects of phosphorylation of r-synGAP by CDK5 and CaMKII, we examined the time course of phosphorylation of r-synGAP by CaMKII. As we found previously for full-length synGAP (11), r-synGAP is phosphorylated rapidly and to a high stoichiometry by CaMKII (Fig. 3B). The most rapid phosphorylation occurs within 2 min, after which r-synGAP contains ϳ6 mol of phosphate/mol, indicating that there are many phosphorylation sites recognized by CaMKII. After 10 min, the rate reaches an asymptote at a stoichiometry of ϳ9 mol of phosphate/mol. Because our GAP assays are inhibited by concentrations of ATP in excess of 30 M, we compared the stoichiometry of r-synGAP phosphorylation by CaMKII at 30 and 500 M ATP. Reaction mixtures containing 30 M ATP followed the same reaction profile as reactions with 500 M ATP; however, the overall stoichiometry plateaued at ϳ6.5 mol of phosphate incorporated per pmol of r-synGAP (data not shown). The stoichiometry and time course of phosphorylation of r-synGAP by CaMKII in 30 M ATP were nearly identical to those of phosphorylation " Labeled synGAP was quantified as described for A. C, phosphorylation by CaMKII of individual sites on r-synGAP. r-synGAP (286 nM) was phosphorylated as described in B, with cold ATP in place of [␥-32 P]ATP. Following fractionation by SDS-PAGE, r-synGAP was transferred to PVDF membranes and probed with anti-phospho-synGAP-Ser-764/5 and anti-phospho-synGAP-Ser-1123, as described under "Experimental Procedures." Staining was quantified on a Li-Cor imager as described under "Experimental Procedures." Intensity of staining for phosphorylated Ser-765 and Ser-1123 was normalized for each to the value at 10 min. All data are mean Ϯ S.E. (n ϭ 4).

TABLE 2 Identification of CDK5 phosphorylation sites in r-synGAP
Residues marked with an X are CDK5 phosphorylation sites in r-synGAP identified as described under "Experimental Procedures." Phosphorylated residues are categorized according to the length of the phosphorylation reaction (2 and 10 min) and their detection in MS2 or MS3 spectra. The sites highlighted with an asterisk have phosphorylation false localization rates less than 1% (predicted as described under "Experimental Procedures").

MS3 MS2 MS3
Ser-728 X* X* X* X* Ser-773 X* X* X* X* Thr-775 X X X X Ser-802 X X X X Ser-842 X* X X reactions carried out with endogenous synGAP and CaMKII in the PSD fraction (11). Phosphorylation of r-synGAP by CaMKII (Fig. 3B) proceeded more rapidly and to a higher stoichiometry than phosphorylation by CDK5 (Fig. 3A), despite the fact that reaction mixtures for CaMKII contained a 100:1 molar ratio of r-syn-GAP (substrate) to kinase, whereas those for CDK5 contained a 2:1 molar ratio of r-synGAP to kinase.
To compare the rate of phosphorylation of two previously identified sites on r-synGAP by CaMKII, we used phosphositespecific antibodies developed against phosphorylated peptides containing Ser-765 and Ser-1123 (11). We measured phosphorylation of the two sites by Western blotting as described under "Experimental Procedures" (Fig. 3C). Ser-1123 was phosphorylated very rapidly, to ϳ70% of maximum in the first 2 min and reaching a plateau by 10 min. Ser-765 was phosphorylated more slowly at a linear rate reaching a maximum at 10 min.
Identification of Sites in r-synGAP Phosphorylated by CaMKII-We identified all sites of phosphorylation of r-synGAP by CaMKII using the same sensitive proteomic mass spectrometric methods described above. We confirmed the presence of six CaMKII phosphorylation sites previously identified by our laboratory (Ser-751, Ser-765, Ser-1093, Ser-1095, Ser-1099, and Ser-1123) ( Table 3 and supplemental data). We also identified 10 additional phosphorylation sites in r-synGAP (Ser-737, Ser-821, Ser-825, Ser-827, Ser-882, Ser-883, Thr-897, Ser-1150, Ser-1210, and Ser-1283) ( Table 3 and supplemental data). All of these sites were detected at both the 2-and 10-min time points and had false localization rates of less than 1%. Phosphorylation was detected at sites Ser-750, Ser-756, and Ser-764, primarily at 10 min; however, the phosphorylation false localization rates at these sites were Ͼ5%. Because sites Ser-737, Ser-751, Ser-765, Ser-1093, Ser-1095, Ser-1099, Ser-1123, Ser-1150, Ser-1210, and Ser-1283 had false localization rates under 1%, we believe that these 10 serines are all candidates for sites of phosphorylation of r-synGAP by CaMKII. We detected a large number of phosphorylation sites between residues 800 and 900 suggesting that this region may be exposed to the protein surface and/or particularly disordered. Either circumstance could lead to promiscuous phosphorylation of serine and threonine residues in this region.
Effect of CDK5 Phosphorylation on GAP Activity of r-synGAP-We measured GAP activity of r-synGAP before and after phosphorylation by CDK5 for 1, 2, 5, and 10 min as described under "Experimental Procedures" (Fig. 4, A-C). Phosphorylation by CDK5 increased the HRas (Fig. 4A) GAP activity of r-synGAP substantially (98%) and the Rap1 (Fig. 4B) GAP activity slightly (20%), but it did not increase Rap2 GAP activity (Fig. 4C). The effect of CDK5 on HRas reached significance after 2 min and plateaued at 5 min. The effect on Rap1 was nearly maximal (15%) after 1 min. These data show that phosphorylation of r-synGAP by CDK5 accelerates its rate of inactivation of HRas considerably more potently than its inactivation of Rap1.
Effect of CaMKII Phosphorylation on GAP Activity of r-synGAP-We measured GAP activity of r-synGAP before and after phosphorylation by CaMKII as described under "Experimental Procedures" (Fig. 4, D-F). In contrast to CDK5, phosphorylation by CaMKII increased the HRas (Fig. 4D) GAP activity of r-synGAP slightly (25%) and the Rap1 (Fig. 4E) GAP activity of r-synGAP substantially (76%). Like CDK5, it did not increase its Rap2 GAP activity (Fig. 4F). CaMKII phosphorylation of r-synGAP produced a gradual increase in Ras GAP activity, plateauing at 10 min. In contrast, Rap1 GAP activity increased to near its maximum level after 1 min (61%) of phosphorylation by CaMKII. Thus, phosphorylation of r-synGAP by CaMKII accelerates its rate of inactivation of Rap1 more potently than its rate of inactivation of HRas.
Phosphorylation by CDK5 or CaMKII Alters the Ratio of HRas GAP and Rap1 GAP Activities of r-synGAP-The intrinsic rate of hydrolysis of GTP by HRas (ϳ0.0004 -0.0008 s Ϫ1 ) (36) is ϳ7-10 times faster than that of Rap1 (ϳ0.00005-0.0001 s Ϫ1 ) (37). However, unphosphorylated r-synGAP and endogenous synGAP (5) stimulate Rap's GTP hydrolysis much more potently than that of Ras (5, 6). In our experiments, unphosphorylated r-synGAP caused Rap1 to be inactivated at a rate ϳ3.9 times faster than HRas (Fig. 5). Phosphorylation by CDK5 or CaMKII shifts this ratio significantly. Phosphorylation by CDK5 for 10 min decreases it to ϳ2.2 by preferentially increasing stimulation of HRas GAP activity. Conversely, phosphorylation by CaMKII increases the ratio to ϳ5.5 by preferentially increasing stimulation of Rap1 GAP activity.
Regulation of GAP Activity of r-synGAP Mutants-To determine which CDK5 phosphorylation sites are responsible for the effects of phosphorylation on GAP activity, we created mutations in r-synGAP at sites Ser-773 and Ser-802, as well as at

Identification of CaMKII phosphorylation sites in r-synGAP
Residues marked with an X are CaMKII phosphorylation sites in r-synGAP identified as described under "Experimental Procedures." Phosphorylated residues are categorized according to the length of the phosphorylation reaction (2 and 10 min) and detection in MS2 or MS3 spectra. The sites highlighted with an asterisk have phosphorylation false localization rates less than 1% (predicted as described under "Experimental Procedures").

Distinct Regulation of Ras and Rap GAP Activities of synGAP
combinations of other phosphorylation sites. The sites were mutated to alanine to prevent their phosphorylation (phosphodeficient) or to aspartic acid to mimic the effect of phosphorylation by adding a negative charge (phosphomimetic). Mutants were purified by the same procedure as wild type r-synGAP as described under "Experimental Procedures." We first measured the HRas and Rap1 GAP activities of the mutants (Fig. 6). As expected, mutations T775A and S802A did not affect the GAP activity of r-synGAP. Unexpectedly, mutation S773A and the double mutation S773A/T775A resulted in significantly reduced stimulation of GAP activity compared with the wild type construct. This effect was most pronounced for Rap2 GAP activity (Fig. 6C), least pronounced for Rap1 GAP activity (Fig. 6B), and intermediate for HRas GAP activity (Fig.  6A). In contrast, the triple mutation S773A/T775A/S802A produced an increase in HRas and Rap2 GAP activities (Fig. 6, A   FIGURE 4. Effect of CDK5 or CaMKII phosphorylation on GAP activity of r-synGAP. r-synGAP was phosphorylated for 1-10 min in the presence of CDK5/p35 or CaMKII as described under "Experimental Procedures." In the phosphorylation reactions, the synGAP, CDK5, and CaMKII concentrations were 2.5, 2.1, and 3.2 times higher, respectively, than in the reactions shown in Fig. 3. GAP activities of control and phosphorylated r-synGAP (250 nM) were then measured by incubation with [␥-32 P]GTP-loaded GTPases for 10 min at 25°C as described under "Experimental Procedures." A, HRas GAP activity; B, Rap1 GAP activity; and C, Rap2 GAP activity following phosphorylation by CDK5. D, HRas GAP activity; E, Rap1 GAP activity; and F, Rap2 GAP activity following phosphorylation by CaMKII. Initial rates were measured at 10 min; GTPase activity was linear for 25 min (data not shown). Data are mean Ϯ S.E. (A-C, n ϭ 10; D-F, n ϭ 9). The statistical significance of differences from the GTPase activity of unphosphorylated r-synGAP (No Kinase) was determined by ordinary one-way ANOVA (uncorrected Fisher's LSD). **, p Ͻ 0.01; ***, p Ͻ 0.001; ****, p Ͻ 0.0001. Dotted lines indicate the levels of GTPase activity in the presence of unphosphorylated synGAP. and C) and did not affect Rap1 GAP activity (Fig. 6B). These data suggest that mutation of Ser-773 to alanine disrupts folding of synGAP in a way that reduces catalysis; however, mutation of Ser-802 to alanine abrogates this effect and even stimulates catalysis. These two phosphorylation sites flank a 6-residue polyproline motif (Fig. 1C), which can form a relatively rigid rod-like structure. It is possible that mutation of serines, or their phosphorylation, on either side of this motif have a particularly powerful effect on the position of the disordered domain in relation to the GAP catalytic domain and thus alter GAP activity.
The CTM-Plus mutant of r-synGAP, described under "Experimental Procedures," in which serines and threonines at positions 750, 751, 756, 764, 765, 1058, 1062, 1064, 1093, 1095, 1099, 1123, and 1125 are all mutated to alanine, closely resembles mutant S773A in its enzymatic behavior. This suggests that bulk mutagenesis of serines and threonines to alanine in the disordered domain alters protein folding in a way similar to the mutation S773A.
Single and double phosphomimetic mutations S773D, T775D, and S773D/T775D completely inhibited HRas and Rap2 GAP activities, suggesting that phosphorylation of Ser-773 or Thr-775 may be inhibitory. However, the inhibition of Rap1 GAP activity by these mutations was small. In contrast, the S802D mutant and the double mutant S773D/S802D increased all three GAP activities above that of wild type, HRas (ϳ25%), FIGURE 5. Alteration of the ratio of HRas GAP and Rap1 GAP activities of r-synGAP after phosphorylation by CDK5 or CaMKII. Data from Fig. 4 are re-plotted to emphasize the change in ratio of HRas GAP and Rap1 GAP activities after phosphorylation of r-synGAP for 10 min by CDK5 or CaMKII. The data for unphosphorylated r-synGAP from Fig. 4 were combined. The numbers above each bracket are the ratios of the means of Rap1 GAP activity to HRas GAP activity for each pair of conditions. Plotting the data for HRas and Rap1 on the same axis emphasizes r-synGAP's greater stimulation of GTP hydrolysis by Rap1 compared with HRas. Note that, in the absence of synGAP, Rap1 has a lower rate of GTP hydrolysis than HRas (Table 1). Rap1 (ϳ20%), and Rap2 (30 and 17%, respectively). This result suggests that phosphorylation of Ser-802 may enhance GAP activity.
When compared with Fig. 4, the data indicate that phosphomimetic mutants of r-synGAP at these three sites do not accurately reproduce the effects of phosphorylation by CDK5. However, they do suggest that phosphorylation at Ser-773/Thr-775 might inhibit GAP activity of r-synGAP, whereas phosphorylation at Ser-802 might increase it. They also suggest that phosphorylation of both sites at the same time would produce a net increase in GAP activity.
We further tested this hypothesis by phosphorylating the S773D r-synGAP mutant with CDK5. Before phosphorylation, this mutant has no HRas GAP activity and reduced Rap1 GAP activity. We predicted that CDK5 would phosphorylate Ser-802 in the mutant and that this would increase HRas and Rap1 GAP activities. As predicted, HRas GAP activity was increased by 220% and Rap1 GAP activity by 9% after phosphorylation (Fig.  7). In comparison, phosphorylation of WT synGAP with CDK5 increased HRas GAP activity 98% and Rap1 GAP activity 20%, supporting the conclusion that phosphorylation of Ser-802 may account for a major portion of the differential regulation of HRas and Rap GAP activities by CDK5.
We also predicted that phosphorylation of the S802A mutant by CDK5 would result primarily in phosphorylation at Ser-773, causing a decrease in HRas and Rap1 GAP activities, as we had observed for the S773D mutant. Again, as predicted, phosphorylation of the S802A mutant with CDK5 caused a 61% decrease in HRas GAP activity and a 9% decrease in Rap1 GAP activity, supporting the inhibitory role of phosphorylation at Ser-773.
To determine whether additional, noncanonical CDK5 phosphorylation sites play a role in regulating GAP activity, we phosphorylated the S773A/T775A/S802A triple mutant in the presence of CDK5, which should cause phosphorylation at one or both of the noncanonical sites, Ser-728 and Ser-842, which had been detected by mass spectrometry ( Table 2). The phosphorylation produced a 57% increase in HRas GAP activity and no significant change in Rap1 GAP activity. These results suggest that noncanonical phosphorylation sites do indeed play a role in regulation of HRas GAP activity by CDK5. They may act together with Ser-802 to regulate HRas and Rap1 GAP activities by a combinatorial mechanism. An understanding of the structural basis of this complex regulation of GAP activity by phosphorylation of the disordered domain will require additional structural studies.
Phosphorylation of Ser-773 and Ser-802 in the PSD Fraction from Mouse Brain-To test whether CDK5 phosphorylates synGAP in vivo in the PSD, we looked for phosphorylation of Ser-773 and Ser-802 on endogenous synGAP in the PSD fraction from mouse brain. We incubated mouse PSD fractions in the absence and presence of Mg 2ϩ /ATP, exogenous CDK5/ p35, and roscovitine, as described under "Experimental Procedures". Proteins were fractionated by SDS-PAGE, and the 140 -160-kDa bands containing synGAP were excised for tryptic digestion. Coupled liquid chromatography and hybrid quadrupole-Orbitrap mass spectrometric analysis of the digests confirmed that Ser-773 and Ser-802 are phosphorylated in isolated PSDs without addition of exogenous kinase (Fig. 8, A and B) and are phosphorylated concurrently when exogenous CDK5/p35 is added to the PSD (Fig. 8, C and D). Addition of Mg 2ϩ /ATP to the PSD did not significantly change the amount of phosphorylation at either site; however, simultaneous addition of purified CDK5/p35 and Mg 2ϩ /ATP increased the phosphorylation of both sites substantially, indicating that both sites are accessible to phosphorylation by CDK5 in the PSD. Addition of the CDK5 inhibitor roscovitine with exogenous CDK5/p35 prevented additional phosphorylation at both sites, indicating that CDK5 is the primary kinase responsible for phosphorylation of Ser-773 and Ser-802 in these experiments.
NMDA Receptor-dependent Activity Differentially Regulates Phosphorylation of synGAP at Ser-773 (CDK5) and at Ser-765 and Ser-1123 (CaMKII)-We used a commercially available antibody that recognizes CDK5 substrates phosphorylated on FIGURE 7. Effect of CDK5 phosphorylation on the GAP activities of selected alanine and aspartic acid mutants of r-synGAP. Wild type and mutant r-synGAP were phosphorylated for 10 min in the presence of CDK5/ p35 as described in Fig. 4, A and B. Initial rates of GAP activities for control and phosphorylated r-synGAP (250 nM) were then measured by incubation with [␥-32 P]GTP-loaded GTPases for 10 min at 25°C. GTPase activity was linear for 25 min (data not shown). A, HRas GAP activity; and B, Rap1 GAP activity. Proteins are named according to their specific point mutations as in Fig. 7. Data are mean Ϯ S.E. (n ϭ 6 -7). The statistical significance of differences between GTPase activity of unphosphorylated and phosphorylated r-synGAP was determined by ordinary one-way ANOVA (uncorrected Fisher's LSD). *, p Ͻ 0.05; **, p Ͻ 0.01; ****, p Ͻ 0.0001. serine residues within a CDK5 consensus sequence to study phosphorylation of synGAP by CDK5 in living neurons. We found that it binds to a band of the size of synGAP on immunoblots of proteins from the PSD fraction and to r-synGAP after phosphorylation by purified CDK5 (Fig. 9A). We also verified that the antibody binds to synGAP after its immunoprecipitation from a brain membrane fraction with an anti-syn-GAP antibody, as described under "Experimental Procedures" (Fig. 9B). The antibody was raised against an antigen matching the consensus sequence SPX(R/K). It binds to phosphorylated mutant synGAPs containing T775A or S802A, but it does not bind to a phosphorylated mutant containing both S773A and T775A, in which Ser-802 is available for phosphorylation (Fig.  9A). These data indicate that the antibody recognizes only phosphorylated site Ser-773, which contains a lysine at position ϩ3 in addition to proline at ϩ1. Ser-802 has a proline at ϩ1, but it does not have a basic residue at ϩ3. We were unable to find a commercial antibody that recognizes phosphorylated Ser-802.
Activation of NMDA receptors on cultured neurons allows influx of Ca 2ϩ and activation of CaMKII. It has also been shown previously to cause a transient activation of CDK5, followed by down-regulation through degradation of the CDK5 activator p35 (38). Furthermore, prolonged NMDAR activation activates calpain, an activator of CDK5 (39,40). Calcium influx through NMDA receptors also activates Ras via the guanine nucleotide exchange factor, RasGRF1, and can activate Rap1, possibly via activation of Epac2 or CalDAGGEF1 (9, 41-45). Thus, it is plausible that the relative levels of phosphorylation of synGAP by CDK5 and CaMKII, regulated by NMDAR activity, might have an important role in controlling the absolute level, ratio, and duration of NMDAR-dependent Ras and Rap activities at the synapse.
To investigate whether NMDAR activity can regulate phosphorylation of synGAP by CDK5, we measured the time course of phosphorylation of Ser-773 following treatment of cultured cortical neurons with 25 M NMDA for 15 s to 5 min. For comparison, we also measured phosphorylation of the CaMKII sites, Ser-765 and Ser-1123 (11). Their levels of phosphorylation were measured by immunoblotting neuronal extracts with phospho-site-specific antibodies (11). We detected steadystate phosphorylation of CDK5 sites and of the CaMKII sites in neuronal cultures that were not treated with NMDA. After treatment with 25 M NMDA, phosphorylation of Ser-765 was elevated at 15 s and peaked at ϳ1 min, and it returned to baseline at 5 min after addition of NMDA (Fig. 9C). As we showed previously (9), phosphorylation at Ser-1123 following treatment with NMDA was biphasic, increasing at 15 s, falling below baseline after a 5-min treatment (Fig. 9C), and further below baseline by 10 min (data not shown). In contrast, phosphorylation of the CDK5 site Ser-773 decreased steadily following NMDAR activation (Fig. 9C). Thus, the three sites have distinct patterns of regulation during NMDAR activation.
Calcineurin and Protein Phosphatase 2A Differentially Regulate Dephosphorylation of synGAP-To dissect the factors influencing the dynamic phosphorylation patterns of the three sites, we determined which phosphatase activities regulate their phosphorylation at steady state and after NMDAR activation. We tested for involvement of three well known phospha-tases, protein phosphatase 1 (PP1), protein phosphatase 2A (PP2A), and calcineurin (CaN). We used cyclosporin A (CsA, 5 M) as a highly specific inhibitor of CaN and okadaic acid (OKA, 1 M) to inhibit both PP1 and PP2A. At the extracellular concentration that we used (1 M), OKA is a more potent inhibitor of PP2A (30). Finally, we used 0.5 M tautomycin, a PP1specific inhibitor, to differentiate between the actions of PP1 and PP2A. Tautomycin had little effect on any of the sites (data not shown); therefore, it is likely that most of the effects of OKA resulted from inhibition of PP2A. The inhibitors were applied to cultured cortical neurons for 30 min followed by application of 25 M NMDA for 0, 0.25, 1, 3, or 5 min. We measured phosphorylation of each of the sites in cortical extracts as described under "Experimental Procedures." Treatment of cortical cultures with 1 M OKA resulted in an increase in phosphorylation of Ser-765 at steady state and during the NMDA treatment, indicating that PP2A is involved in maintaining the steady-state level of phosphorylation of Ser-765 under our conditions (Fig. 10A). In contrast, 5 M CsA had no effect on its phosphorylation either at steady state or transiently following NMDA treatment, indicating that CaN does not regulate Ser-765.
Treatment with 1 M OKA, had a small effect on phosphorylation of Ser-1123 at early times after treatment with NMDA (0.25 and 1 min) but no significant effect at steady state or at later time points (5 min) (Fig. 10B). In contrast, treatment with 5 M CsA increased phosphorylation of Ser-1123 at steady state and also after NMDAR activation with the highest increases occurring at the longer times. The data suggest that, in the absence of inhibitors, NMDA treatment activates CaMKII, FIGURE 9. Changes in phosphorylation of synGAP at Ser-773, Ser-765, and Ser-1123 following activation of NMDA receptors in cultured primary neurons. A, commercial anti-CDK5 substrate antibody binds to bands corresponding to synGAP on immunoblots of the PSD fraction and to purified WT and mutant (T775A and S802A) r-synGAP variants phosphorylated with CDK5/p35. The antibody does not bind to a phosphorylated mutant in which Ser-773 is mutated to alanine. B, commercial antibody recognizes synGAP after it is immunoprecipitated from a brain membrane fraction with an anti-synGAP antibody as described under "Experimental Procedures." In this experiment, anti-PSD-95 Ab was used as a positive control because synGAP binds to PSD-95. Anti-GluR2 Ab was used as a negative control to show that the immunoprecipitation was specific. PSD-95 was present in the immunoprecipitate and GluR2 was not, as expected (data not shown). C, cultured cortical neurons were treated with 25 M NMDA for 0.25, 1, 3, or 5 min. Phosphorylation of synGAP at Ser-765, Ser-773, and Ser-1123 was measured by immunoblot as described under "Experimental Procedures." Dotted line, Ser-765, n ϭ 9 -11 independent experiments; Dashed line, Ser-1123, n ϭ 9 -11 independent experiments; solid line, Ser-773, n ϭ 5-7 independent experiments. Levels of phosphoproteins were normalized to intensities of staining for MAP-2. Significance was determined by one sample t test compared with 100%. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001.
leading to rapid phosphorylation of Ser-1123, and it also activates CaN, leading to slower dephosphorylation of Ser-1123. The result is a peak of phosphorylation at 15 s, followed by a gradual fall to below baseline at 5 min. Thus, in contrast to Ser-765, CaN appears to be the primary phosphatase regulating dephosphorylation of Ser-1123.
In contrast to the CaMKII sites, phosphorylation of the CDK5 site Ser-773 slowly decreased during treatment with 25 M NMDA (Fig. 10C), suggesting that phosphatases play a role in NMDA receptor-dependent regulation of this site. Indeed, inhibition of CaN with 5 M CsA changed the response to a small but significant increase in phosphorylation of Ser-773 at all times other than steady state. These results mean that activation of CaN during treatment with NMDA is responsible for the gradual dephosphorylation of synGAP at Ser-773. It has previously been shown that NMDA treatment can transiently (within 1 min) activate CDK5 activity in cultured neurons (24). Such an increase can only be seen in our data after 1 min of NMDA treatment in the presence of CsA (Fig. 10C). Treatment with 1 M OKA also caused an increase in phosphorylation at Ser-773 at steady state and at early time points, indicating that PP2A dephosphorylates Ser-773 at steady state. Phosphorylation of synGAP at specific sites was measured by immunoblotting as described in Fig. 9C. We verified that the effects of OKA in this experiment were a result of inhibition of PP2A by performing complementary experiments with the PP1-specific inhibitor tautomycin, which had very little effect (data not shown). A, phosphorylation at Ser-765 was significantly increased at all time points (including steady state) in OKA-treated cultures compared with controls in which no inhibitor was added. In contrast, treatment with CsA had no significant effect on phosphorylation at Ser-765. B, phosphorylation at Ser-1123 was significantly increased at 1 and 3 min in OKA-treated cultures, and at all time points in CsA-treated cultures, compared with controls. C, phosphorylation of Ser-773 was increased at all time points in the presence of CsA, peaking at 1 min; treatment with OKA increased phosphorylation at steady state and from 15 s to 3 min after NMDA treatment, compared with control. Significance in A-C was determined by t test, n ϭ 3-5 independent experiments. D, left, steady-state level of phosphorylation of Ser-773 in cortical cultures is regulated, in part, by CDK5 and PP2A. Cortical cultures were treated with 1 M OKA for 30 min in the absence or presence of 30 M roscovitine, which was added 5 min before OKA to block CDK5 prior to inhibition of PP2A. OKA alone increased the steady-state level of phosphorylation of Ser-773 ϳ2.3-fold. Addition of roscovitine resulted in a significantly smaller increase, supporting the conclusion that CDK5 phosphorylates Ser-773 in neurons independently of treatment with NMDA (OKA/vehicle (veh): 231 Ϯ 18%, OKA/roscovitine (Ros): 183 Ϯ 4%). Pretreatment with 100 M roscovitine for 30 min did not further decrease the phosphorylation of Ser-773 (data not shown). Significance was determined by one-sample t test compared with 100% and by t test, n ϭ 3-5 independent experiments. Right, transient change in phosphorylation of Ser-773 during activation of NMDA receptors is regulated by CDK5 and CaN. Cortical cultures were treated with NMDA in the absence or presence of 5 M CsA for 30 min and in the absence or presence of 100 M roscovitine, which was added 30 min before CsA to block CDK5 prior to inhibition of CaN. Addition of NMDA decreased phosphorylation of Ser-773 to 75% of control in the absence of CsA or roscovitine. The presence of CsA during the treatment with NMDA increased phosphorylation of Ser-773 ϳ1.5-fold. This effect of CsA was almost completely blocked by simultaneous treatment with roscovitine. Vehicle, 74 Ϯ 6%; CsA/vehicle, 121 Ϯ 2%; CsA/ roscovitine, 92 Ϯ 7%. Significance was determined by ANOVA with post hoc Tukey-Kramer multiple comparisons test, n ϭ 3 independent experiments. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001. FEBRUARY 20, 2015 • VOLUME 290 • NUMBER 8

JOURNAL OF BIOLOGICAL CHEMISTRY 4923
To confirm involvement of CDK5 in the dynamic regulation of phosphorylation of Ser-773, we performed similar experiments in which we incubated neurons with OKA or with OKA and roscovitine, a specific inhibitor of CDK5 (46). The increase in steady-state phosphorylation of Ser-773 in OKA was significantly reduced, although not back to control levels, by pretreatment with 30 M roscovitine (Fig. 10D), verifying that at least a portion of phosphorylation of Ser-773 at steady state is due to CDK5 activity. Pretreatment with a higher concentration of roscovitine (100 M) did not further decrease the phosphorylation (data not shown), suggesting that another kinase (such as GSK3␤) may also phosphorylate Ser-773 in neurons. We also confirmed the involvement of CDK5 in the increased phosphorylation of Ser-773 at 1 min after activation of NMDA receptors in the presence of CsA (Fig. 10C). We treated neurons with 5 M CsA alone or with CsA plus 100 M roscovitine for 30 min and then applied 25 M NMDA for 1 min. The increase in phosphorylation of Ser-773 in the presence of CsA is significantly reduced by pretreatment with 100 M roscovitine (Fig.  10D). This result means that phosphorylation of Ser-773 by CDK5 is stimulated by activation of NMDA receptors, and CDK5 accounts for most of the increase in phosphorylation of Ser-773 observed when CaN is inhibited.

DISCUSSION
Biological Role of synGAP-Mutations in synGAP1 are a major cause of nonsyndromic cognitive disability in humans and are particularly common (8 -10%) in patients with associated epilepsy or autism (8,47,48). Most of the pathogenic mutations characterized to date cause haploinsufficiency. In mice, heterozygous deletion of synGAP, which produces haploinsufficiency, leads to precocious spine formation during development (49,50), disruption of synaptic plasticity and learning (7,9), and an increased level of surface AMPA receptors in neurons (51). The deleterious effects of loss of just one copy of synGAP suggest that its functions are rate-limiting for processes that are essential for normal synaptic function.
synGAP accelerates inactivation of the small GTPases Ras and Rap (1, 2, 5, 6). These two GTPases are central regulators in dendrites; in particular, they contribute to regulation of exocytosis and endocytosis of AMPA receptors near the synapse (10). A prolonged increase in active Ras in neurons leads to increased AMPAR exocytosis and thus to an increase in total surface AMPA receptors, whereas prolonged activation of Rap causes an increase in AMPAR endocytosis and a decrease in surface AMPA receptors. The steady-state levels of activity of the two GTPases are set by the balance of production of the active forms by exchange of GDP for GTP and inactivation by GTP hydrolysis. Unphosphorylated synGAP catalyzes inactivation of Rap much more potently than inactivation of Ras (Fig. 2) (6). Thus, its presence would be expected to reduce endocytosis of AMPA receptors relative to their exocytosis by decreasing the steady-state level of active Rap to a greater extent than the steady-state level of active Ras.
Physiological Role of Changes in the Relative Levels of Ras and Rap GAP Activity of synGAP-Here, we have shown that phosphorylation of synGAP by CaMKII increases its ratio of Rap1 GAP to HRas GAP activities. We predict that, in neurons, this would shift the steady-state balance of AMPAR trafficking near the synapse toward exocytosis by decreasing the level of active Rap1 more than active HRas. The net result would be an increase in surface AMPA receptors. Conversely, phosphorylation of r-synGAP by CDK5 decreases its ratio of Rap1 GAP to HRas GAP activities. We predict that this would shift the balance of trafficking near the synapse toward endocytosis of AMPA receptors and lead to a decrease in the number of surface AMPA receptors. Thus, phosphorylation of synGAP could act like a rheostat to create transient changes in the number of surface AMPA receptors or to gradually adjust their steadystate level.
Although r-synGAP increases Rap2 GAP activity, this activity is not altered by any of the phosphorylation events reported here. Both Rap1 and Rap2 are found in pyramidal neurons in the hippocampus and cortex (Allen Mouse Brain Atlas); but, the differences in their functions are not yet clear. There is a precedent for differential regulation of the two forms of Rap. It has been reported that the rasGEF family of proteins catalyze nucleotide exchange for Rap2 but not for Rap1 (52).
Potential Role of Phosphorylation by CaMKII-CaMKII is concentrated in spines and in the PSD (53). It is an early target of Ca 2ϩ /calmodulin that is formed in the spine when Ca 2ϩ flows through NMDA receptors (45). Active CaMKII phosphorylates Ser-1123 on synGAP at a rapid rate, nearly maximally after 2 min in our assay (Fig. 3C). The increase in Rap1 GAP activity of synGAP is also maximal after 2 min of phosphorylation (Fig. 4E), suggesting that it is produced by phosphorylation of Ser-1123 or of another rapidly phosphorylated site. We predict that this rapid increase in Rap1 GAP activity would cause a rapid decrease in active Rap1, leading to a relatively rapid decrease in AMPAR endocytosis. The result would be an increase in surface AMPA receptors.
In neurons, we have shown that activation of NMDA receptors produces a rapid increase in phosphorylation of Ser-1123, followed by a slower decrease in phosphorylation mediated by the Ca 2ϩ /calmodulin-dependent phosphatase, CaN (Figs. 9C and 10B). Thus, we predict that the level of active Rap1 would be decreased transiently relative to active HRas, producing a transient increase in the rate of exocytosis of AMPA receptors near the synapse. This rate would decrease gradually as CaN dephosphorylates Ser-1123 and brings the rate of inactivation of Rap1 back down.
It is well established that activation of CaMKII in the spine triggers an increase in synaptic strength produced by enhanced trapping of surface AMPA receptors at the postsynaptic site (45,54). A slightly slower increase in insertion of surface AMPA receptors near the synapse has been observed after induction of long term potentiation (55). We suggest that phosphorylation of synGAP by CaMKII helps to mediate this transient increase in AMPAR exocytosis.
Potential Role of Phosphorylation by CDK5-In contrast to CaMKII, CDK5 activity is not regulated directly by Ca 2ϩ influx or by synaptic activity. Instead, its catalytic subunit is activated by binding to either of two small subunits termed p35 and p39 (14). The p35 and p39 subunits are degraded constitutively; thus, the cellular level of CDK5 activity is maintained by synthesis of the catalytic and regulatory subunits, which are induced by growth factors, including NGF and BDNF (14). CDK5 is essential for normal brain development and function, but its regulatory roles are generally homeostatic, in contrast to the many transient and activity-dependent regulatory roles of CaMKII (56).
The rate of phosphorylation of synGAP by CDK5 is considerably slower than the rate of phosphorylation by CaMKII (Fig.  3, A and B). A relatively slow catalytic rate has also been observed for CDK5 with other important neuronal substrates (16). The absence of rapid activation mechanisms and its relatively slow catalytic rate are consistent with a homeostatic role. We postulate that phosphorylation of synGAP by CDK5 may contribute to the homeostatic down-regulation of surface AMPA receptors termed synaptic scaling (57). The mechanisms that contribute to synaptic scaling are complex. Several mechanisms act in parallel to tightly couple the rate of neuronal activity to the total level of surface synaptic AMPA receptors in neurons (57). CDK5 and polo-like kinase 2 (PLK2) have been shown to be necessary for homeostatic reduction of surface AMPA receptors following elevated neuronal activity (58). The two kinases act in concert to phosphorylate spine-associated Rap-specific GAP protein (SPAR), a constitutive synaptic Rap GAP, and tag it for degradation. Loss of SPAR activity increases the steady-state level of active Rap, thus increasing endocytosis of surface AMPA receptors. Our results suggest that CDK5 phosphorylation can also decrease the level of surface AMPA receptors by selectively increasing the Ras GAP activity of syn-GAP, leading to a reduction of AMPAR exocytosis.
Regulation of synGAP by CDK5 appears to be more complex than its regulation by CaMKII. We have found that two sites on synGAP, Ser-773 and Ser-802, are phosphorylated in isolated PSDs and are phosphorylated concurrently when exogenous CDK5 is added to PSDs (Fig. 8). However, mutations of these two sites suggest that the functional effects of their phosphorylation are quite different (Figs. 6 and 7). Mutations of Ser-773 to alanine (phospho-deficient) or to aspartic acid (phosphomimetic) inhibit HRas GAP and Rap2 GAP activities substantially, and Rap1 GAP activity slightly. This finding means that Ser-773 is crucial for GAP activity itself. In contrast, mutation of Ser-802 to alanine has no effect on GAP activity, although mutation to aspartic acid increases all GAP activities and overrides the effects of mutation of Ser-773. It is interesting that these two sites are located on opposite sides of a 6-residue polyproline motif (Fig. 1C) that can form a relatively stiff rod along the protein backbone. It is possible that residues on either side of the polyproline motif play an important role in positioning the downstream portion of the predicted disordered domain with respect to the catalytic domains of synGAP.
We identified a phospho-specific antibody that allowed us to monitor phosphorylation of Ser-773 in cultured neurons; unfortunately, we were unable to monitor phosphorylation of Ser-802. We found that activation of NMDA receptors in cultured neurons triggers slow dephosphorylation of Ser-773 mediated by CaN (Figs. 9C and 10C). If Ser-802 is also dephosphorylated by CaN, this would reduce HRas GAP activity and help to shift the balance of Ras and Rap activity in favor of active Ras, consistent with our rheostat hypothesis. However, confir-mation of the role of Ser-802 will await new reagents or methods.
The large number of serine and threonine phosphorylation sites that we identified for CDK5 (at least 5) and CaMKII (at least 10) are located in clusters within the predicted disordered domain of synGAP (see Fig. 1C). This finding is consistent with the proposal of Holt et al. and Collins (59,60) that regulation of protein function by serine/threonine phosphorylation often depends on relatively nonspecific mechanisms that alter protein-protein interactions. These authors found that phosphorylation sites in yeast proteins regulated by CDK1 are often located in clusters that shift position in rapidly evolving disordered domains. We have focused on the functions of two sites phosphorylated by CDK5 (Ser-773 and Ser-802) that are located next to proline residues. However, it is possible, perhaps even likely, that the other three sites can also influence regulatory interactions of the disordered domain with the Ras GAP domain. This possibility is supported by the finding shown in Fig. 7A that phosphorylation of sites Ser-728 and Ser-842 can contribute to an increase in HRas GAP activity in addition to phosphorylation of Ser-802.
Significance of Differential Patterns of in Vivo Regulation of Phosphorylation Sites on synGAP-The distinct patterns of regulation of the three phosphorylation sites, Ser-765, Ser-802, and Ser-1123, during synchronous activation of NMDA receptors is striking (Fig. 9C). The patterns arise from differential regulation of each site by phosphatases and by kinases (Fig. 10). This precise regulation lends credence to the importance of synGAP in mediating activity-dependent and homeostatic synaptic plasticity.
In summary, phosphorylation of r-synGAP by CaMKII and CDK5 has opposite effects on the ratio of Rap1 GAP and HRas GAP activities of synGAP. We postulate that phosphorylation of synGAP by these two kinases contributes to regulation of the number of surface AMPA receptors in and near spines during synaptic potentiation (CaMKII) and during synaptic scaling (CDK5).