Synapsin I Is a Major Endogenous Substrate for Protein L-Isoaspartyl Methyltransferase in Mammalian Brain*

The accumulation of potentially deleterious l-isoaspartyl linkages in proteins is prevented by the action of protein l-isoaspartyl O-methyltransferase, a widely distributed enzyme that is particularly active in mammalian brain. Methyltransferase-deficient (knock-out) mice exhibit greatly increased levels of isoaspartate and typically succumb to fatal epileptic seizures at 4–10 weeks of age. The link between isoaspartate accumulation and the neurological abnormalities of these mice is poorly understood. Here, we demonstrate that synapsin I from knock-out mice contains 0.9 ± 0.3 mol of isoaspartate/mol of synapsin, whereas the levels in wild-type and heterozygous mice are undetectable. Transgenic mice that selectively express methyltransferase only in neurons show reduced levels of synapsin damage, and the degree of reduction correlates with the phenotype of these mice. Isoaspartate levels in synapsin from the knock-out mice are five to seven times greater than those in the average protein from brain cytosol or from a synaptic vesicle-enriched fraction. The isoaspartyl sites in synapsin from knock-out mice are efficiently repaired in vitro by incubation with purified methyltransferase and S-adenosyl-l-methionine. These findings demonstrate that synapsin I is a major substrate for the isoaspartyl methyltransferase in neurons and suggest that isoaspartate-related alterations in the function of presynaptic proteins may contribute to the neurological abnormalities of mice deficient in this enzyme.

PIMT catalyzes conversion of isoaspartyl sites to normal Asp-Xaa linkages in synthetic peptides (7)(8)(9) and has been shown to restore function to isoaspartate-containing forms of proteins such as calmodulin (10) and the Escherichia coli phosphocarrier protein HPr (11). Pharmacological inhibition of PIMT activity in rat PC12 cells leads to reversible accumulation of isoaspartyl sites in a variety of proteins, including tubulin (12) and histone H2B (13).
Studies with PIMT-deficient mice further support a repair function for PIMT (14,15). Whereas proteins in extracts from wild-type or heterozygous PIMT mice contain only trace levels of isoaspartate, proteins in extracts from knock-out mice show greatly increased isoaspartate levels in extracts from all tissues examined, with the most significant accumulation occurring in extracts of brain. Although they appear generally normal at birth, PIMT knock-out mice develop enlarged brains and typically succumb to fatal epileptic seizures between 28 and 70 days after birth. More recently, transgenic (knock-in) mice that selectively express PIMT only in neurons have been described (16). Even though the transgenic mice continue to accumulate high levels of isoaspartate in tissues other than brain, many of them survive for 1 or more years and appear physically and behaviorally normal. These findings suggest that neuronal function is uniquely sensitive to accumulation of isoaspartate or that PIMT itself plays a special role in neuronal function independent of its putative role as a repair enzyme.
Synapsin I is an abundant neuronal protein localized primarily on the cytoplasmic surface of small presynaptic vesicles at both inhibitory and excitatory synapses (17,18). Synapsin I appears to play an important role in controlling the availability or "ready state" of small synaptic vesicles in response to regulatory signals mediated by several protein kinases, including protein kinase A and Ca 2ϩ /calmodulin-dependent protein kinase type II. Before it was established that PIMT has a specificity for isoaspartyl sites in proteins (a time when this enzyme was known simply as an abundant brain enzyme catalyzing formation of protein carboxyl methyl esters), we reported that purified bovine synapsin I serves as a highly efficacious substrate for this methyltransferase in vitro, exhibiting an apparent K m in the low M range (19). Subsequently, we found that purified bovine synapsin I is unusually susceptible to spontaneous formation of isoaspartyl sites during incubation at pH 7.4 and 37°C, conditions designed to model in vitro protein "aging" (20,21).
In this study, we investigated the possibility that synapsin I is an important target for PIMT action in vivo. Consistent with our in vitro aging studies, we found that synapsin I accumulates near-stoichiometric levels of isoaspartate in PIMT knock-out mice and that this accumulation greatly exceeds that of the average protein in brain cytosol or in a synaptic vesicle-enriched fraction. The accumulation of isoaspartyl sites in synapsin is significantly reduced in transgenic (knock-in) mice that selectively express PIMT in neurons, and the degree of reduction seems to correlate with the overall health of the mice. We demonstrate that the isoaspartyl sites in synapsin I isolated from PIMT-deficient mice can be effectively eliminated by PIMT-dependent methylation in vitro. These findings show that synapsin I is a major physiological substrate for PIMT and suggest that the neurological deficits seen in PIMT knockout mice may be due, at least in part, to the accumulation of dysfunctional synaptic proteins. 3 H]L-methionine (AdoMet; 15 Ci/ mmol) and [ 35 S]ATP␥S (65 Ci/mmol) were purchased from PerkinElmer Life Sciences. Where indicated, [methyl-3 H]AdoMet was adjusted to lower specific activity by dilution with unlabeled AdoMet (Sigma) that had been purified on carboxymethylcellulose (22). Recombinant rat PIMT was expressed in E. coli and purified as described previously (23). Bovine synapsin I was purified to homogeneity from frozen brain as described (20). The batch of bovine synapsin used here as a methylation standard contained 0.2 mol of isoaspartate/mol of synapsin (accumulated during long-term storage) as determined by two different PIMT-dependent methylation assays (24,25). Mammalian protease inhibitor mixture (catalog no. P3840) and protein A-Sepharose CL-4B were from Sigma.

Materials-S-Adenosyl-[methyl-
Mice-Heterozygous (PIMT ϩ/Ϫ ) founder pairs (from which wildtype, heterozygous, and knock-out mice were bred) were kindly provided by Dr. Edward Kim (Gladstone Institute of Cardiovascular Disease, San Francisco, CA) in conjunction with the laboratories of Dr. Stephen Young (of the same institution) and Dr. Steven Clarke (UCLA) (14). The presence of the neo gene and the presence/absence of the PIMT gene were detected by PCR of tail DNA using the following prim-ers: PIMT, 5Ј-GCAGCGACGGCAGTAACAGC-3Ј (forward) and 5Ј-ACCCTCTTCCCATCCACATC-3Ј (reverse); and neo, 5Ј-GCAC-GAGGAAGCGGTCAGCCCATTC-3Ј (forward) and 5Ј-CGCATCG-AGCGAGCACGTACTCGG-3Ј (reverse). Mice were killed at ϳ4 -6 weeks of age by cervical dislocation, and brains were rapidly removed and frozen at Ϫ80°C. As an additional source of wild-type PIMT mice, frozen brains from 8 -10-week-old Swiss Webster mice were purchased from Pel-Freez Biologicals (catalog no. 55004-2).
Line 29 transgenic mice (expressing PIMT under the control of the neuron-specific enolase promoter) have been described previously (16). The relevant properties of the transgenic mice used in this study are summarized in Table 1.
Purification of Synapsin I from Mouse Brain-Synapsin I from mouse brain was purified by a modification of the method of Huttner et al. (26). Individual preparations utilized one to four brains weighing 0.3-0.4 g each. All procedures were carried out at 0 -4°C. Thawed brains were homogenized in 6 volumes of buffer A (25 mM Tris-Cl (pH 7.8), 0.25 M sucrose, 2 mM EDTA, 15 mM ␤-mercaptoethanol, and 100 M phenylmethylsulfonyl fluoride) using an Ultra-Turrax Model STD Tissumizer (Tekmar). Homogenates were centrifuged for 45 min at 24,000 ϫ g, and pellets were resuspended in 5 ml buffer B (10 mM sodium phosphate (pH 7.2), 2 mM EDTA, 15 mM ␤-mercaptoethanol, 100 M phenylmethylsulfonyl fluoride, and 2 g/ml pepstatin A). The pH of each sample was adjusted to 3.0 by dropwise addition of 0.5 M HCl. Samples were kept on ice for 15 min and then centrifuged for 15 min at 24,000 ϫ g. Supernatants were recovered, adjusted to pH 7.2 by dropwise addition of 0.3 M Na 3 PO 4 , and centrifuged again for 15 min at 24,000 ϫ g. Supernatants were then used for cation-exchange chromatography on 3.5 ϫ 1-cm columns of Whatman CM-52 equilibrated with buffer B without pep-FIGURE 1. Mechanism of spontaneous isoaspartate formation and enzymatic repair in proteins. The formation of isoaspartyl sites (dashed arrows) starts with spontaneous dehydration of an Asp-Xaa linkage or deamidation of an Asn-Xaa linkage to produce a metastable succinimide that has a typical half-life of 2-4 h at physiological pH and temperature. The succinimide undergoes a spontaneous hydrolysis that generates an unequal mixture of two products; typically, 15-30% of the product is a normal L-aspartyl linkage (left solid arrow), with the remaining 70 -85% forming an abnormal L-isoaspartyl-Xaa linkage (right dashed arrow). The isoaspartate repair pathway is indicated by the solid arrows. PIMT rapidly converts L-isoaspartyl sites to ␣-carboxyl-O-methyl esters (downward pointing solid arrow). At physiological pH and temperature, the ␣-carboxyl-O-methyl esters have a typical half-life of only a few minutes, undergoing spontaneous demethylation to reform the same succinimide from which they originated. The final stage of repair occurs when a portion of the succinimide spontaneously hydrolyzes to generate a normal L-aspartyl site. Although the repair pathway is inefficient and requires multiple cycles for substantial net repair, its component reactions are very rapid compared with the damage reactions that generate isoaspartyl sites; thus, PIMT is able to maintain a very low steady-state level of isoaspartyl protein sites in vivo. The repair pathway shown here is also the basis for in vitro assays of isoaspartate levels in proteins. If [methyl-3 H]AdoMet is used as the methyl donor, isoaspartate levels in proteins can be quantitated by measuring [ 3 H]methyl incorporation into the protein of interest or by measuring the production of [ 3 H]methanol. With unlabeled AdoMet, isoaspartate can be quantitated by HPLC to measure the production of S-adenosyl-L-homocysteine. statin A. After loading samples (2.8 -3.8 ml), columns were washed with 2 ml of the same buffer. Synapsin was then eluted with 2 ml of 0.08 M NaCl in buffer B, followed by 2 ml of 0.15 M NaCl in buffer B. Fractions enriched in synapsin were identified by SDS-PAGE, pooled, and concentrated. In later preparations, a 1:100 dilution of protease inhibitor mixture was used in place of phenylmethylsulfonyl fluoride and pepstatin.
Purification of Calmodulin-Calmodulin was purified to homogeneity from bovine and mouse brains by the method of Gopalakrishna and Anderson (27) as modified by Potter et al. (28). Individual preparations from mouse brain utilized three pooled brains/preparation. Subcellular Fractionation-Subcellular fractionation of mouse brain tissue was based on the methods of Jones and Matus (29) and Huttner et al. (30). Approximately 1.5 g of tissue was homogenized in 9 volumes of 5 mM Na-HEPES (pH 7.6), 0.5 mM EDTA, 10% (w/v) sucrose, 0.1 mM dithiothreitol, and protease inhibitor mixture. Homogenization was performed by nine strokes in a Potter-Elvehjem homogenizer with a rotating Teflon pestle. Homogenates were centrifuged for 20 min at 800 ϫ g. The pellet (P1), representing a crude nuclear fraction, was processed further by passage through a 25-gauge needle and centrifugation for 10 min at 3000 ϫ g (13). The supernatant (S1) was centrifuged for 20 min at 9000 ϫ g, and the new supernatant (S2) was centrifuged for 2 h at 100,000 ϫ g to generate the cytosolic fraction. The 9000 ϫ g pellet (P2) was resuspended in the homogenization buffer and spun again for 20 min at 9000 ϫ g. The resulting pellet (representing the crude synaptosomal fraction) was hypotonically lysed in 5 mM Na-HEPES (pH 7.6), 0.5 mM EDTA, 0.1 mM dithiothreitol, and protease inhibitor mixture (1.25 ml of buffer/g of tissue). The suspension was kept on ice for 30 min and homogenized by six strokes in the Potter-Elvehjem homogenizer.
From this step, synaptic vesicles were partially purified as described (30). Lysed synaptosomes were centrifuged for 20 min at 25,000 ϫ g. The resulting supernatant (corresponding to the LS1 fraction of Huttner et al. (30)) was centrifuged for 2 h at 165,000 ϫ g. The pellet (LP2) was resuspended in 0.2 ml of buffer and drawn six times through a 25-gauge needle.
Analysis of Isoaspartate in Mouse Brain Proteins (Synapsin I and Calmodulin) by SDS-PAGE and 3 H Fluorography-3 H methylation of purified proteins was carried out at pH 6.0 -6.2 for 10 min at 30°C in reactions containing either the phosphate/citrate/EDTA buffer of Kim et al. (31) or 75 mM K-MES, 2.5 M PIMT, and 20 -50 M [methyl-3 H]AdoMet (5000 -7000 dpm/pmol). Reaction volumes and synapsin or calmodulin content were varied as described in the appropriate figure legend. Reactions were terminated by addition of the 3-fold concentrated SDS-PAGE sample buffer (pH 6.8) of Laemmli and Favre (32). After heating at 50°C for 5 min, samples were subjected to SDS-PAGE on NuPAGE Novex 4 -12% gradient gels using the MES running buffer (Invitrogen). Electrophoresis was carried out at 100 V at 4°C to minimize methyl ester hydrolysis. Tritium fluorography of salicylate-impregnated gels was carried out as described previously (12).
Western Blot Analysis of Synapsin I in Homogenates and Vesicle-enriched Fractions of Mouse Brain-Proteins were separated by SDS-PAGE on Novex NuPAGE 4 -12% gradient gels with MES running buffer following the manufacturer's recommendation and transferred onto nitrocellulose. The membrane was exposed to a 1:2000 dilution of anti-synapsin primary antibody (catalog no. MAB355, Chemicon International, Inc.) for 1 h, followed by a 1-h exposure to a 1:2500 dilution of alkaline phosphatase-conjugated goat anti-mouse secondary antibody (Bio-Rad), with appropriate wash steps between exposures. Immunoreactive bands were visualized with a 5-bromo-4-chloro-3-indolyl phosphate/nitro blue tetrazolium color development substrate kit from Promega Corp.
In Vitro Repair of Isoaspartyl Sites-Repair reactions were carried out in Pierce Slide-A-Lyzer microdialysis tubes (7-kDa cutoff) in a 50-l mixture composed of repair buffer (20 mM Na-HEPES (pH 7.8), 1 mM EGTA, 1 mM dithiothreitol, and protease inhibitor mixture), 20 M AdoMet, 5 M PIMT, and ϳ9 g of synapsin. This mixture was dialyzed against 40 ml of repair buffer containing 20 M AdoMet for 7-8 h at 37°C. Samples were then dialyzed against the same buffer minus AdoMet for 14 -15 h at 30°C to remove the AdoMet and to allow hydrolysis of any residual methyl esters and cyclic imide intermediates (7,33). The samples were dialyzed for an additional 7-18 h at 4°C against 75 mM K-MES (pH 6.2) without AdoMet, with one change of external buffer during that time. Samples were recovered from dialysis and 3 H-methylated for analysis by SDS-PAGE and 3 H fluorography as described above.
Analysis of Isoaspartyl Sites Formed during in Vitro Aging of Bovine Synapsin I-Synapsin I was purified to homogeneity from frozen bovine cerebral cortex as described by Bennett et al. (34). In vitro aging was carried out by incubating the purified synapsin (1.0 mg/ml) in 50 mM K-HEPES (pH 7.4), 1.0 mM EDTA, and 5% (w/v) glycerol for 30 days at 37°C. Under these conditions, synapsin I accumulates 1.4 mol of isoaspartyl sites/mol of synapsin (20). Identification of the major isoaspartyl sites was carried out by the same procedures described in detail for our analyses of in vitro aged recombinant human growth hormone (35), bovine brain calmodulin (28), and recombinant human tissue plasminogen activator (36). Briefly, the aged synapsin was reduced, alkylated, and digested with trypsin. The tryptic peptides were separated by reversed-phase HPLC, and fractions containing isoaspartyl peptides were identified by post-column 3 H methylation catalyzed by recombinant PIMT. The sequences of the isoaspartyl peptides and the exact positions of the isoaspartyl sites within these peptides were determined (when possible) by a combination of mass spectrometry and Edman sequencing.
ATP Binding-The general approach to measure ATP binding was to preincubate purified synapsin I with [ 35 S]ATP␥S, to immunoprecipitate the synapsin with anti-synapsin antibody bound to protein A-Sepharose, and then to quantitate the amount of [ 35 S]ATP␥S in the pellet by liquid scintillation counting. Validation of the ATP binding assay was first carried out using synapsin I purified as described above from commercial mouse brain. Before testing ATP binding, the synapsin was cleared of nonspecific protein A binding by combining 2 volumes of synapsin I (0.1 mg/ml in buffer B), 1.5 volumes of binding buffer (50 mM Na-HEPES (pH 7.4), 25 mM NaCl, 2 mM CaCl 2 , 0.02% Tween, and 1% (v/v) protease inhibitor mixture), and 1 volume of packed protein A-Sepharose beads that had been previously equilibrated with binding buffer. After rotating the suspension for 2.5 h at 4°C, the beads were pelleted by brief centrifugation. The supernatant, containing the cleared synapsin I, was used directly for ATP binding. In the first step of the binding assay, 60 l of cleared synapsin I (2.4 g) was mixed with 10 l of water, ATP (10 mM), or GTP (10 mM) and allowed to sit for 5 min on ice. 10 l of [ 35 S]ATP␥S (2.0 M) was added, and the incubation was continued for 15 min. In the second step, this mixture was transferred to a 1.8-ml microcentrifuge tube containing 20 l of packed protein A beads to which 3.5 g of anti-synapsin I antibody (product no. VAP-SV060, Stressgen) had been previously adsorbed. After rotating the suspension for 3.5 h at 4°C, the beads were pelleted and subjected to two brief washes, each with 800 l of cold binding buffer. Bound [ 35 S]ATP␥S was extracted from the pellet with 200 l of 8.8% (v/v) formic acid and counted in 1.0 ml of Liquiscint (National Diagnostics). Measurement of ATP binding to synapsin I purified from mouse PIMT ϩ/Ϫ and PIMT Ϫ/Ϫ brains was carried out as described above, except that, in the first step of the binding assay, we used 60 l of cleared synapsin I (2.7 g), 10 l of [ 35 S]ATP␥S (1.0 M), and 10 l of water.
Protein Determination-With the exception of purified mouse brain synapsin, protein concentrations were determined by the method of Lowry et al. (37) after precipitation with 7% (w/v) trichloroacetic acid with bovine serum albumin as the standard. The concentration of synapsin I in purified preparations from mouse brain was determined by comparing the density of the synapsin bands on Coomassie Bluestained SDS-polyacrylamide gels with standards of purified bovine synapsin I (covering the range of 0.2-0.8 g/lane) run on the same gels. NIH Image software was used to quantitate band intensities from digital images of the stained gels.

RESULTS
Synapsin I Is a Major Isoaspartate-forming Protein in the Brains of PIMTdeficient Mice-We chose to study the status of synapsin I in PIMT-deficient mice because of its high affinity for PIMT (19), its tendency to form isoaspartate rapidly in vitro at physiological pH and temperature (20), and its important role in modulating synaptic transmission (17). Multiple smallscale synapsin I purifications from each of three PIMT genotypes (ϩ/ϩ, ϩ/Ϫ, and Ϫ/Ϫ) were performed to achieve an accurate representation of isoaspartate levels. To assess isoaspartate content, synapsin was 3 H-methylated by PIMT under conditions known to label accessible sites with an efficiency of ϳ80% (24). The reaction mixture was then subjected to SDS-PAGE, and 3 H-methylated protein bands were visualized by fluorography. All four synapsin preparations from PIMT Ϫ/Ϫ mice showed similar high levels of methyl-accepting capacity, whereas synapsin from PIMT ϩ/ϩ and PIMT ϩ/Ϫ mice was virtually devoid of any significant methyl-accepting capacity ( Fig. 2A). By comparing the image densities of 3 H-labeled synapsin I bands in lanes 7-10 with those of an 3 H-methylated bovine synapsin I standard (0.2 mol of isoaspartate/mol of synapsin) in lanes 1 and 2, we calculated that synapsin I from the knock-out mice contained 0.9 Ϯ 0.3 mol of isoaspartate/mol of synapsin. No significant isoaspartate was detected in synapsin from wild-type or heterozygous mice (lanes 3-6) even when fluorography of the gel in Fig. 2A was extended from 30 h to 7 days.
Synapsin I is a trimer consisting of a single Ia subunit and two slightly smaller Ib subunits. The two subunit polypeptides result from differential splicing of a single gene (38). It appears from Figs. 2 (A and D) and 4 that both subunits accumulate isoaspartyl sites in vivo.
Given the dramatic differences in isoaspartate content seen in Fig. 2A, we wondered whether synapsin I expression is altered in PIMT-deficient mice. It seemed possible, for example, that accumulation of damaged synapsin molecules might lead to a compensatory increase in the total amount of synapsin produced. Equivalent amounts of total protein from brain homogenates or synaptic vesicles from PIMT ϩ/ϩ and PIMT Ϫ/Ϫ mice were subjected to SDS-PAGE and then blotted and probed with anti-synapsin I monoclonal antibody (Fig. 2B). No difference was observed between genotypes in either homogenates or crude synaptic vesicles. Quantitation of synapsin I yields from the eight different purifications ( Fig. 2A, lanes 3-10), representing all three genotypes, also indicated no significant difference in expression (data not shown).
Calmodulin Is Not a Major Isoaspartate-forming Protein in the Brains of PIMT-deficient Mice-We have shown previously that calmodulin, an abundant brain protein that plays several key roles in neuronal function, accumulates isoaspartate during in vitro aging at a rate similar to that observed with synapsin I (10,28). Isoaspartyl sites accumulate mainly in the Ca 2ϩ -binding domains, significantly decreasing calmodulin's ability to activate Ca 2ϩ /calmodulin-dependent protein kinase type II. Much of this lost activity can be restored by incubating the damaged calmodulin with PIMT and AdoMet.
To determine whether calmodulin is a substrate for PIMT in vivo, we carried out a study with the PIMT knock-out (KO) mice similar to that shown in Fig. 2A. As shown in Fig. 2C (lanes 5 and 6), calmodulin from the KO mouse brains exhibited a low but measurable level of [ 3 H]methyl-accepting capacity, whereas calmodulin from the wild-type littermates did not. The average amount of calmodulin loaded in lanes 3-6 of Fig. 2C (1.5 g) was twice that of the synapsin loaded in lanes 3-10 of Fig. 2A. This was done because calmodulin (a small, highly acidic protein) binds Coomassie Blue R-250 rather poorly, and we wanted to visually confirm the levels of calmodulin that were loaded. To compensate for this 2-fold higher loading, fluorographic exposure times for Fig. 2C were roughly half of those for Fig. 2A. Densitometric analysis of the 4.0-day exposure in lanes 5 and 6 (KO-calmodulin) and the 0.7day exposure in lane 1 (standard, aged bovine brain calmodulin containing 0.9 mol of isoaspartate/mol of calmodulin) allowed us to estimate a The middle and lower panels show 3 H fluorograms of the same gel at the indicated exposure times (in days (d)). Lanes 1 and 2 contain 0.75 and 2.25 g, respectively, of a synapsin I standard (aged synapsin from bovine brain) containing 0.2 mol of isoAsp/mol of synapsin. B, shown is a Western blot of synapsin I from mouse brain homogenates (2.5 g/lane) and partially purified synaptic vesicles (0.35 g/lane). C, calmodulin was purified from the brains of mice with the indicated genotypes (lanes 3-6), methylated with [methyl-3 H]AdoMet, and subjected to SDS-PAGE. Each of these lanes represents an individual calmodulin preparation (three brains/preparation) from mice of the indicated genotype. The uppel panel shows the Coomassie Blue staining used to verify equal loading of ϳ1.5 g/lane. The middle and lower panels show 3 H fluorograms of the same gel at the indicated exposure times. Lanes 1 and 2 contain 0.75 and 1.5 g, respectively, of a calmodulin standard (aged calmodulin (calmod) from bovine brain) containing 0.9 mol of isoAsp/mol of calmodulin. D, shown is isoaspartate in synapsin I from PIMT knock-in transgenic mice. Protein samples containing 0.7 g of synapsin I were methylated, separated by gel electrophoresis, and processed as described for A. Lanes 2-5 contain synapsins purified from PIMT knock-in mice with the indicated genetic backgrounds. The brains used for these preparations are described in Table 1; synapsin preparations A-D were used for lanes 2-5, respectively. Lanes 6 and 7 contain non-transgenic controls similar to those in lanes 7 and 3, respectively, in A. Lanes 1 and 8 (Ag) contain 0.7 g of the same aged bovine synapsin standard shown in lane 1 in A. stoichiometry of 0.0035 Ϯ 0.0005 mol of isoaspartate/mol of KO-calmodulin. This molar ratio is only 0.4% (0.0035/0.9) of that estimated for KO-synapsin. When normalized to mg of protein, the isoaspartate content of KO-calmodulin is 1.8% of that of KO-synapsin (Fig. 3).
These results demonstrate that, unlike synapsin I, calmodulin is not a major target for PIMT in vivo. This low level of isoaspartate accumulation in calmodulin may reflect stabilization of susceptible Asx residues by calcium binding (an effect that has been demonstrated in vitro (39)) and/or its association with numerous regulatory target proteins. The preferential degradation of isoaspartyl calmodulin demonstrated in Xenopus oocytes (40) may also help to explain its very limited accumulation in the PIMT-KO mouse.
Isoaspartate Levels in Synapsin I from Transgenic Mice Expressing PIMT Only in Neurons-In addition to brain, isoaspartyl proteins have been shown to accumulate in the hearts, lungs, kidneys, livers, and erythrocytes of PIMT-deficient mice (14), an observation reflecting the ubiquitous tissue distribution of this enzyme as well as the widespread susceptibility of proteins to isoAsp formation. Selective expression of PIMT in the neurons of transgenic mice with a PIMT Ϫ/Ϫ background has been reported (16). The median life span of these transgenic knock-in mice is 213 days, a dramatic increase over the 44-day average life span of PIMT Ϫ/Ϫ mice. Moreover, the phenotype of these mice appears normal or nearly normal throughout their prolonged life span. The transgenic expression results strongly suggest that PIMT activity is especially important to neurons and that the levels of PIMT activity in neurons largely determine the phenotype.
If synapsin I is a major neuronal target for PIMT action in vivo, then we would expect isoAsp levels to be lower in synapsin from transgenic animals compared with PIMT Ϫ/Ϫ mice from which the transgenic animals were derived. Fig. 2D shows that this is indeed the case. Lane 4 shows the methyl-accepting capacity of synapsin I isolated from a pool of two brains obtained from knock-in mice with a PIMT Ϫ/Ϫ background ( Table 1). Methylation of this synapsin is detectable in the autoradiogram, but the level was substantially lower than that seen in syn-apsin from PIMT Ϫ/Ϫ mice without the transgene (lane 6). The results obtained from a single knock-in mouse brain also with neuron-specific expression in a PIMT Ϫ/Ϫ background are shown in lane 5. This brain was processed separately from the brains shown in lane 4 because of the animal's more "sickly" phenotype. As expected, synapsin I from this mouse had a methyl-accepting capacity intermediate between the synapsins shown in lanes 4 and 6. The intermediate level of isoaspartate seen in lane 5 mostly likely results from a lower efficiency of neuronal PIMT expression in this particular sickly mouse compared with the mice shown in lane 4. This is consistent with the large variation in PIMT expression seen in the brains of the line 29 PIMT Ϫ/Ϫ transgenic mice (16). The results shown in Fig 2D. 4 support the idea that isoaspartate levels in synapsin vary inversely with the level of PIMT activity in neurons. Fig. 2 show that synapsin I accumulated near-stoichiometric levels of isoaspartate in the absence of PIMT in vivo, whereas calmodulin exhibited a barely detectable increase in isoaspartate. To determine how these proteins compare with the "average" brain protein, subcellular fractions were prepared from both PIMT ϩ/ϩ and PIMT Ϫ/Ϫ mice, and isoaspartate content was measured via methyl-accepting capacity. As shown in Fig. 3, the level of isoaspartate in synapsin I (expressed as nmol/mg of protein) was at least five times higher than that in the average protein from brain homogenate, cytosol, nuclei, or synaptic vesicles. Although there may be individual proteins within these

. In vitro repair by PIMT of the isoaspartate sites in aged bovine synapsin I and in synapsin I from PIMT-KO mice.
Left panels, aged bovine synapsin I standard (9.0 g) was subjected to in vitro repair as described under "Experimental Procedures" in the presence or absence of PIMT and unlabeled AdoMet. After repair, residual isoaspartyl sites were 3 H-methylated under standard conditions in a 12-18-l volume and then subjected to SDS-PAGE and autoradiography. The plus and minus symbols indicate the reagents present or absent during the repair reaction only; the gels show the results of the 3 H methylation reactions used to evaluate the effectiveness of the repair. The complete lack of [ 3 H]methyl incorporation into synapsin I from the reaction containing AdoMet and PIMT indicates that the complete repair reaction eliminated virtually all the isoaspartate sites in the aged synapsin I standard. Right panels, synapsin I partially purified from PIMT-KO mice was subjected to repair, and the isoaspartate content was evaluated as described for the left panels. Again, the isoaspartate content in this synapsin I was eliminated when PIMT was present in the repair reaction.

TABLE 1 Properties of transgenic (PIMT knock-in) mice used in this study
All of the mice included below expressed a PIMT transgene under the control of a neuron-specific enolase promoter as described by Lowenson et al. (16). The genetic background relevant to the endogenous PIMT gene of these mice is given in the Genotype column. Ϫ/Ϫ Sickly a a The preparation D mouse had a lower weight and lower motor activity than the preparation C mice. The preparation D mouse also had mild seizures. fractions that have isoaspartate levels similar to or even higher than synapsin I (see below), it is clear that synapsin I contains considerably more isoaspartate than does the average brain protein. Calmodulin presents a dramatic contrast to synapsin. Although calmodulin readily forms isoaspartate in vitro in the absence of Ca 2ϩ , it accumulates considerably less isoaspartate in vivo than does the average brain protein.

PIMT Repairs Isoaspartyl Sites in Synapsin I from PIMT-KO Mice-Pre-
vious studies have demonstrated that purified PIMT can catalyze the in vitro repair of isoaspartyl peptide bonds in synthetic peptides (7-9) and age-damaged proteins (10,11). For this study, we employed an improved repair protocol in which the isoaspartyl protein is incubated with PIMT and unlabeled AdoMet while being dialyzed against a pH 7.8 buffer containing a reservoir of additional AdoMet. Performing the reaction under dialysis allows for continuous dilution of the methylation product S-adenosyl-Lhomocysteine (a potent inhibitor of PIMT) while simultaneously maintaining a nearly constant supply of AdoMet. After repair, the analyte is extensively dialyzed against fresh AdoMet-free buffer to ensure complete hydrolysis of any methyl esters or cyclic imide forms of the protein that could block subsequent assessment of unrepaired isoaspartyl sites. In the final phase of this protocol, any remaining isoaspartyl sites are quantitated by incubating the analyte with fresh PIMT and [methyl-3 H]AdoMet at pH 6.2 (a pH that stabilizes isoaspartyl methyl esters), followed by SDS-PAGE and fluorography. Development and optimization of this repair protocol (not shown) were carried out using the protein phosphacan, an isoaspartate-rich component of the extracellular matrix of brain (41).
A bovine synapsin I standard containing 0.2 mol of isoaspartate/mol of protein was used to evaluate the effectiveness of this new repair protocol, and the results are shown in Fig. 4 5 versus lane 6). These results demonstrate that the isoaspartyl sites formed in mouse synapsin I in vivo are capable of efficient repair in vitro by purified recombinant rat PIMT, providing further support to the conclusion that synapsin I is a physiologically important target for PIMT action in neurons.
To be certain that the lack of [ 3 H]methyl incorporation in Fig. 4 (lanes 1 and 5) reflects repair rather than persistence of unlabeled methyl groups acquired during the repair phase, we carried out the following control reaction. Samples of aged bovine synapsin I control were repaired using [methyl-3 H]AdoMet and then dialyzed overnight under the same conditions used in the standard repair reaction. After dialysis, we measured the amount of [ 3 H]methyl remaining on the synapsin. The residual methyl incorporation amounted to Ͻ2% of the isoaspartyl sites subjected to the repair reaction.
Location of Isoaspartyl Sites Formed during in Vitro Aging of Bovine Brain Synapsin I-It was of interest to determine whether the accumulation of isoaspartate in synapsin I occurs at one major site or at several different sites and how the location/distribution of sites relates to the domain structure of synapsin I. To answer this question, we used a strategy that has proved effective in locating the major sites of isoaspartate accumulation that arise during in vitro aging of recombinant human growth hormone (35), bovine brain calmodulin (28), and recombinant human tissue plasminogen activator (36). Because this type of analysis requires mg quantities of purified protein, it was not possible for us to analyze synapsin from the KO mouse brain. As an alternative, we choose to analyze the isoaspartyl sites formed during in vitro aging of synapsin I purified from bovine brain. We believe that this is a valid alternative because isoaspartate formation is an nonenzymatic intramo-lecular process driven by the sequence and local flexibility of the polypeptide chain. Moreover, as noted below, the bovine and mouse synapsin I sequences are nearly identical.
Synapsin I was purified to homogeneity from bovine brain, and a portion was aged by incubation at pH 7.4 for 28 days at 37°C as described under "Experimental Procedures." "Aged" and "control" (non-aged) synapsin samples were digested with trypsin, and the resulting mixtures of peptides were 3 H-methylated by PIMT to label the major isoaspartyl sites. Immediately after this pre-column methylation, the peptides were separated by reversed-phase HPLC. The UV profile of eluted peptides monitored at 214 nm revealed little difference between digests of the aged and control samples (Fig. 5A). In contrast, the pattern of 3 H-methylated peptides was clearly different for the two samples (Fig.  5B). The aged synapsin digest contained a complex of isoaspartate-rich peptides that eluted between 50 and 70 min. A much lower level of isoaspartate was seen in the same region of the control. The 3 H methylation in the control samples represents any isoaspartate that was present in synapsin in vivo combined with additional isoaspartate that arose during purification and tryptic digestion. The peak of the 3 H label seen at 8 -10 min in Fig. 5B (upper and lower panels) is unrelated to synapsin because it was present at the same level in a mock trypsin digest that lacked synapsin (data not shown).
The methylation assay used in Fig. 5B is convenient for determining the retention times of the major isoaspartyl peptides because it requires only one pre-column methylation reaction, but it suffers from two limitations. First, the efficiency of isoaspartate labeling is limited to a level of 80% or less because of the inherent instability of the methyl esters (24); and second, the integrity of isoaspartyl peptides is altered by the ability of PIMT to convert isopeptide bonds to succinimides and to normal aspartyl peptide bonds, actions that would complicate attempts to localize major isoaspartyl sites within the protein sequence. To avoid these concerns, we isolated isoaspartate-rich tryptic peptides from aged synapsin in an HPLC run that did not employ pre-column methylation (Fig.  5C). Fractions from the 50 -70-min region of the run were collected, and 10% of each fraction was analyzed for isoaspartate using a [ 3 H]methanol recovery assay that measures isoaspartate content with an efficiency of close to 100% (24). The remaining material in the fractions from the three regions richest in isoaspartate were pooled as indicated in Fig. 5C and concentrated. The identities of the isoaspartyl peptides within these pools and the locations of the isoaspartyl sites within these peptides were determined by a combination of mass spectrometry and Edman sequencing as described in detail (28,35,36).
The results of this analysis are presented in Fig. 6 and Table 2. Four isoaspartyl peptides were identified. Two of these peptides each contained a single isoaspartyl site that arose via deamidation of an asparagine residue. One peptide (positions 382-403) contained a single isomerized aspartyl residue. The fourth peptide (positions 282-299) contained a nest of two or three closely spaced sites that arose from isomerization of aspartyl residues. Of the five to six total isoaspartyl sites we identified, all but one were located in domain C.
Synapsin I from PIMT-deficient Mice Exhibits Normal ATP Binding-It was of interest to determine whether the isoaspartyl synapsin I from PIMTdeficient mice is altered with regard to any of its presumed functions. Finding a functional deficit presents itself as a potentially daunting task given that synapsin is reported to interact directly with numerous components of the presynaptic terminal, including neurotransmitter vesicles, the cytoskeleton, several protein kinases, and ATP.
To assess a potential loss of function in isoaspartyl synapsin I from PIMT Ϫ/Ϫ mice, we decided to measure ATP binding. This decision was based on three considerations. 1) As shown above, in vitro aging of bovine synapsin I indicates that isoaspartate formation occurs mainly in domain C. 2) Domain C has a number of important functions that are mediated through its ability to bind ATP, vesicles, and cytoskeletal proteins. 3) ATP binding can be studied with a high degree of precision and sensitivity, important considerations given the limited amount of purified synapsin we were able to obtain from the mutant mouse brains.
We used a protocol for ATP binding modeled after the studies of Hosaka and Südhof (42). Briefly, we incubated mouse synapsin with 100 or 200 nM [ 35 S]ATP␥S and then immunoprecipitated the synapsin with anti-synapsin polyclonal antibody to measure bound ligand. The results of this study are shown in Fig. 7. In Fig. 7A, we used synapsin isolated from commercially obtained frozen mouse brain to optimize and validate our ATP binding assay. Specificity was demonstrated by the ability of excess unlabeled ATP (but not GTP) to displace binding of the labeled ATP (Fig. 7A). Fig. 7B shows that there was no significant difference in ATP binding between isoaspartate-rich synapsin obtained from PIMT Ϫ/Ϫ mice and isoaspartate-free synapsin obtained from PIMT ϩ/Ϫ mice. Hosaka and Südhof reported that ATP binds to recombinant synapsin domain C with a K d of 120 nM. From the data in Fig. 7B, where binding to mouse brain synapsin was performed in the presence of 100 nM [ 35 S]ATP␥S, we calculated that 45.6% of the immunoprecipitated synapsin I molecules had the ATP analog bound. Thus, the level of ATP binding we observed with synapsin I purified from mouse brain is consistent with the ATP affinity previously reported for recombinant domain C of synapsin Ia (42).

DISCUSSION
The ability of synapsin I to serve as a highly effective substrate for PIMT in vitro was first reported in 1983 (19), 1 year before it was reported that this methyltransferase (formerly known as protein carboxyl methyltransferase) is selective for L-isoaspartyl sites (43,44). We subsequently showed that bovine synapsin I exhibits an extraordinary susceptibility to spontaneous isoAsp formation when incubated at pH 7.4 and 37°C (20,21). The recent availability of PIMT-deficient mice made it possible to test the idea that synapsin I undergoes isoAsp formation in vivo.
In this work, we utilized the PIMT-KO mouse to provide a convincing case that synapsin I generates isoaspartyl sites in vivo (Fig. 2) to a substantially higher level than found in the average neuronal protein (Fig. 3) and that neuronal PIMT serves to prevent the accumulation of isoaspartyl synapsin I (Figs. 2D and 4) by the same methylationdemethylation-succinimide repair pathway (Fig. 1) previously demon-FIGURE 6. Locations of major isoaspartyl sites in aged bovine synapsin I. A, shown is the ClustalW alignment of bovine (Bov; Swiss-Prot accession number P17599) and mouse (Mus; accession number O88935) synapsin I. The isoAsp-containing tryptic peptides identified in this study are underlined, and the locations of the isoAsp residues within these peptides are indicated in boldface with a caret. Differences between the Ia and Ib isoforms of synapsin I are evident in the two sets of sequence lines starting at reside 661. B, the domain structure of synapsin I is shown as originally described by Sü dhof et al. (38). Phosphorylation sites for protein kinase A (PKA) and Ca 2ϩ /calmodulin-dependent protein kinase type II (CaMKII) are indicated along the top. Major sites of isoAsp accumulated during in vitro aging of bovine PIMT are shown along the bottom. strated in vitro with purified PIMT and synthetic isoaspartyl peptides (7,8). It is unclear whether or not the isoaspartyl form of synapsin I is functionally defective. ATP binding, taken as one indicator of the structural integrity of the large domain C, is apparently normal in synapsin I from PIMT-deficient mice, even though the level of isoAsp is nearly stoichiometric (0.9 Ϯ 0.3 mol of isoAsp/mol of synapsin I). It is important to note, however, that synapsin I exhibits numerous other interactions, the integrity of which have not yet been tested.
We first reported the accumulation of isoaspartyl synapsin I in PIMTdeficient mice in 2001 (45). In 2002, Shimizu et al. (46) reported findings that partly overlap with this study. These workers purified synapsin I from PIMT Ϫ/Ϫ and PIMT ϩ/ϩ mice of a different origin (15) and demonstrated that the PIMT-dependent [ 3 H]methyl-accepting capacity (isoAsp levels) in brain proteins in PIMT-deficient mice was 10 -20 times that in wild-type mice. Although the synapsin used in that study was substantially pure as determined by SDS-PAGE, their assessment of methylation capacity was based on 3 H methylation of the bulk-purified synapsin. The 3 H-methylated synapsin was not analyzed by SDS-PAGE and fluorography, making it impossible to draw firm conclusions about which proteins or protein bands in their synapsin preparation were the primary carriers of isoAsp. If these authors attempted to perform fluorography on the 3 H-methylated synapsins after SDS-PAGE, they may have been foiled by the instability of isoaspartyl methyl esters, which generally do not survive SDS-PAGE at the high pH and temperatures encountered in the commonly used Tris/glycine buffer systems (47,48). In our study, we characterized the 3 H-methylated synapsins after separation at 4°C in the BisTris-based NuPAGE Novex gel system (Invitrogen), which utilizes a substantially lower pH compared with the Tris/glycine-based systems. SDS-PAGE at lower temperatures and pH allowed us to demonstrate that PIMT-dependent methylation is exactly coincident with both the synapsin Ia and 1b protein staining bands (Figs. 2 and 4) and also allowed us, by utilizing a well characterized bovine synapsin standard, to estimate the stoichiometry of the mouse synapsin I methylation.
Another area of overlap between our study and the work of Shimizu et al. (46) pertains to the ability of a PIMT transgene to rescue PIMT Ϫ/Ϫ mice from early-onset fatal epilepsy. In our study, we used the transgenic mice previously described by Lowenson et al. (16), in which PIMT is expressed under the control of a neuron-specific enolase promoter. Because of this highly specific neuronal expression, the results shown in Fig. 2D demonstrate that isoAsp repair in synapsin I occurs via direct interaction of synapsin with neuronal PIMT. In the parallel work of Shimizu et al., transgenic PIMT was expressed in multiple tissues under the control of a prion promoter. This provides a weaker case for a direct relationship between neuronal PIMT expression and synapsin integrity.
The idea that PIMT repairs synapsin in vivo is strengthened further by our in vitro demonstration that purified PIMT can efficiently eliminate isoaspartyl sites in synapsin purified from PIMT Ϫ/Ϫ mice (Fig. 4). We and others have previously demonstrated similar in vitro repair of model peptides (7-9), age-damaged calmodulin (10), and a spontaneously deamidated form of the HPr phosphocarrier protein of E. coli (11). In these previous studies, incubations of 24 h or longer were required to repair a majority of isoAsp sites. In this work, we were able to achieve nearly complete repair in 7-8 h of incubation. The improved kinetics are due mainly to the use of continuous dialysis to maintain a favorable AdoMet/S-adenosyl-L-homocysteine ratio during the repair reaction, a process involving multiple cycles of PIMT-catalyzed methylation and spontaneous demethylation (Fig. 1). The efficient repair is also a reflection of the high affinity (K m Ϸ 0.1 M) with which isoaspartyl synapsin binds to PIMT (19).
Isoaspartate formation has been shown to markedly reduce the function of proteins such as calmodulin (10), the HPr phosphocarrier protein of E. coli (11), and others (49). Given the high stoichiometry of isoaspartyl sites in synapsin from PIMT-deficient mice, it is possible that altered synapsin function may contribute significantly to the phenotype of these mice. Consistent with this idea, there are several elements of similarity in the phenotypes of mice that are deficient in PIMT and mice that are deficient in synapsin I. Most notable among these similarities is the propensity of both types of mice to epileptic seizure (14,50) and subtle abnormalities in synaptic physiology and synaptic vesicle distribution in hippocampal neurons (51)(52)(53)(54).
We do not propose that damage to synapsin in the PIMT-KO mice can account for all or even most of the neurological problems of these mice. This seems unlikely for several reasons. First, there are numerous proteins that accumulate isoAsp sites in the brains of PIMT-deficient We determined that there are at least two and possibly three isoaspartyl sites in this peptide. Of the three sites listed, we are uncertain as to the relative contribution of each individual site.

FIGURE 7.
Binding of ATP to synapsin I purified from mouse PIMT ؊/؊ brain is normal. A, the conditions for immunoprecipitation of synapsin I in the presence of [ 35 S]ATP␥S were optimized with wild-type synapsin partially purified from mouse brains purchased from Pel-Freez Biologicals. The specificity of the ATP/synapsin interaction was confirmed by the presence of 1 mM unlabeled ATP or GTP. B, shown is a comparison of ATP binding by synapsin I partially purified from PIMT ϩ/Ϫ versus PIMT Ϫ/Ϫ mice. Data represent the average of two assays for each condition, and error bars represent S.D. Con, control.
mice, including tubulin (12,55), histone H2B (13), and many others that have not yet been identified (14,15). Second, it is well established that there are numerous genetic lesions that can independently lead to epilepsy and/or subtle alterations in synaptic transmission. Third, there are distinct differences in the phenotypes of synapsin I-and PIMT-deficient mice; for example, the seizures seen in synapsin I-deficient mice are much less prominent, of later onset, and generally not fatal. Finally, there is recent evidence that insulin-dependent signaling pathways are altered significantly in PIMT Ϫ/Ϫ mice, a observation that may be directly related to the enlarged brains of these mice (56). Although synapsin I damage by itself is unlikely to account for the abnormalities in PIMT-deficient mice, it is clear that synapsin I is a major target for PIMT in neurons. At a minimum, this should bring attention to the possibility that isoaspartate-related damage to a limited ensemble of synaptic proteins could explain the neuropathology of these mice. Organs such as heart and liver appear to function normally in PIMT-deficient mice, even though these organs accumulate high levels of isoaspartyl proteins, implying that isoAsp-related damage to most proteins can be dealt with via protein turnover (16). The unique sensitivity of neuronal function to isoAsp accumulation suggests that the associated neuropathology may arise from damage to a limited number of neuron-specific proteins with a critical sensitivity to isoAsp accumulation that cannot be abrogated by protein turnover.
Introduction of an L-isoaspartyl linkage into a self-protein or peptide has the capacity to break immune tolerance (57). Recently, auto-antibodies to synapsin Ia have been detected in the sera of patients with lupus erythematosus and rheumatoid arthritis, providing a potentially important clue to the molecular basis of neuropsychiatric syndromes that are found in some patients with autoimmune disease (58). The possible relationship between isoAsp formation in brain proteins and autoimmune-related neurological disorders would seem to warrant further study.