Polymorphic Variants of Human Protein l-Isoaspartyl Methyltransferase Affect Catalytic Activity, Aggregation, and Thermal Stability

Protein l-isoaspartyl methyltransferase (PIMT/PCMT1), a product of the human pcmt1 gene, catalyzes repair of abnormal l-isoaspartyl linkages in age-damaged proteins. Pcmt1 knock-out mice exhibit a profound neuropathology and die 30–60 days postnatal from an epileptic seizure. Here we express 15 reported variants of human PIMT and characterize them with regard to their enzymatic activity, thermal stability, and propensity to aggregation. One mutation, R36C, renders PIMT completely inactive, whereas two others, A7P and I58V, exhibit activity that is 80–100% higher than wild type. G175R is highly prone to aggregation and has greatly reduced activity. R17S and R17H show markedly enhanced sensitivity to thermal denaturation. Based on previous studies of moderate PIMT variation in humans and mice, we predict that heterozygosity for R36C, G175R, R17S, and R17H will prove detrimental to cognitive function and successful aging, whereas homozygosity (if it ever occurs) will lead to severe neurological problems in the young.

Protein L-isoaspartyl methyltransferase (PIMT/PCMT1; product of the pcmt1 gene) 2 functions to repair abnormal isoaspartyl (isoAsp) sites in proteins that occur via isomerization of the peptide bond linking asparagine or aspartate residues to their carboxyl-flanking neighbor (1)(2)(3). isoAsp formation, like oxidation, is an extremely common type of spontaneous protein damage that occurs under mild conditions in vivo and in vitro and is known to alter protein function and antigenicity. PIMT utilizes the cofactor S-adenosyl-L-methionine (AdoMet) to transfer a methyl group onto the ␣-carboxyl group of isoAsp sites leading to a succinimidyl intermediate that subsequently hydrolyzes to L-Asp and L-isoAsp (Fig. 1). Continuing cycles of PIMT action efficiently repair L-isoAsp sites in vitro (4 -9), and reduction of PIMT activity in cultured cells or KO mice dramatically increases the level of isoAsp-containing proteins (10 -13). PIMT is widely distributed in mammalian tissues but is particularly rich in the CNS (14 -17). A critical need for PIMT in brain is evident from the overt neurological phenotype of PIMT-deficient mice; increased brain size, abnormal neuroanatomical and electrophysiological properties of hippocampal cells along with reduced cognitive function (18), atypical open field behavior (19), and fatal epileptic seizures beginning at 4 weeks of age (11,12). A proteomic study utilizing the PIMT-KO mouse revealed over 30 neuronal targets for PIMT, including synapsins I and II, CRMP2 (collapsin response mediator protein 2), dynamin 1, ␣and ␤-tubulin, and creatine kinase B, among others (20). Based on the phenotype of the PIMT-KO mouse, a major reduction of PIMT repair activity in humans would likely contribute to neurodevelopmental disorders, including epilepsy, in the young. Moreover, current evidence suggests that even moderate decrements in PIMT activity accelerate age-related changes in cognition and even lifespan. Brain extracts of PIMT ϩ/Ϫ mice have 50 -55% of the PIMT activity found in WT (ϩ/ϩ) mice. Between 8 months and 2 years, male ϩ/Ϫ mice accumulate six times more isoaspartyl damage than ϩ/ϩ males (21) and have a median lifespan that is ϳ6 weeks shorter (22). Interestingly, female ϩ/Ϫ mice accumulate approximately three times more isoaspartyl damage over the same time period but show no apparent reduction in lifespan. In WT mice, PIMT activity shows a modest decrease over the lifespan (21), but in postmortem human brain samples, the highest PIMT activities are found in the eldest (23). This latter observation suggests that a higher inherited PIMT activity in humans may promote longevity.
There is one polymorphism, at position 119 in human PIMT, in which both alleles (Val and Ile) are very common (see Table  1). Using erythrocyte cell extracts from humans of known genotype, it was previously found that PIMT activity is slightly higher and more thermally stable in Ile/Ile than Val/Val individuals, with heterozygotes exhibiting an intermediate level of activity and stability (24,25). In contrast, extracts from Val/Val individuals were superior in binding isoaspartyl-damaged pro-teins obtained from PIMT-KO mouse brain, as indicated by a 30% lower K m value in the methylation kinetics (25). This latter study also presented intriguing evidence for Val/Ile hybrid vigor at position 119. By comparing PIMT genotypes in populations of younger versus healthy older Ashkenazi Jews, it was found that the frequency of Val/Ile heterozygotes in the healthy older population (mean age, 87) was significantly higher (p ϭ 0.049) than predicted by Hardy-Weinberg equilibrium. It thus appears that even modest changes in the profile of PIMT activity can have profound implications for human health.
Ongoing genome-wide association studies are generating massive amounts of data that are progressively uncovering the molecular basis of various diseases. Non-synonymous SNPs in protein coding regions account for a substantial fraction of all disease-related genetic mutations. Knowing how a given SNP actually affects the function of the protein in which it resides is an invaluable complement to such studies. For these reasons, we decided to express and purify known polymorphic variants of human PIMT and then characterize them with regard to their catalytic activity and thermal stability. Although it was not an original goal of this study, the findings reported here also provide some insight on the propensity of these variants for spontaneous aggregation, a characteristic that is also likely to affect PIMT activity in vivo.

Results
Expression and Aggregation of PIMT Variants-As of July 2015, the NCBI SNP database (dbSNP) contained 14 non-syn-onymous variants in the coding regions of the human pcmt1 gene ( Table 1). The minor allele frequency data in Table 1 was updated as of December 15, 2016.
As described under "Experimental Procedures," we used standard protocols to express and purify WT PIMT and all 14 of these variants. When subjected to SDS-PAGE, all of the purified enzymes exhibited a major band near the expected mass (28 -30 kDa), but a pair of lighter bands was also evident at 65-70 kDa (Fig. 2). These higher mass bands were barely detectable for the WT and many of the mutants but very strong for G175R and moderately strong for several others. For those mutants that showed a relatively high level of high mass doublet, the relative abundance of these bands varied somewhat from prep to prep (Fig. 3A). We postulated that these bands could be due to 1) incomplete reduction of disulfide bonds during sample preparation for SDS-PAGE; 2) the formation of highly stable aggregates of PIMT during expression, purification, or storage; or 3) the presence of Escherichia coli proteins caused by incomplete purification. The results shown in Fig. 3B argue against incomplete disulfide reduction as a source of high mass bands. A 3-fold increase in DTT concentration during heating in sample buffer had no discernable effect on the staining patterns of either WT PIMT or the mutants with prominent high mass bands.
To test the "stable aggregates" hypothesis, we subjected WT and selected mutant PIMTs to Western blotting using an anti-T7 antibody (Fig. 4A). A T7 antigen sequence is located Under physiological conditions, deamidation of asparagine residues or dehydration of aspartic acid residues results in the formation of a metastable intermediate succinimide that spontaneously hydrolyzes to form a mixture of normal L-aspartyl and atypical L-isoaspartyl linkages. PIMT, using AdoMet as a methyl donor, selectively methylates the isoaspartyl ␣-carboxyl group to form a highly labile methyl ester. Spontaneous demethylation occurs within minutes to reform the original succinimide, with release of methanol as a by-product. This succinimide is now the starting point for further cycles of repair, resulting in nearly complete conversion of the isoaspartyl ␤-linkages to normal aspartyl ␣-linkages. Dashed arrows indicate the damage reactions, and solid lines indicate the repair pathway.
between the His tag and the N terminus of expressed proteins with our plasmid and serves as a useful mark to monitor expression. The results show clear evidence of putative dimer and higher mass aggregates for all samples on the gel, with WT PIMT showing the lowest level. The possibility that the T7 antibody reacts with certain endogenous E. coli proteins was considered, but this was ruled out by the results shown in Fig. 4B where a T7 blot was carried out on crude extracts from E. coli hosting an expression plasmid for the R17H PIMT variant. A strong T7 signal that correlates with R17H expression is seen at 29 kDa, but there is no discernable binding of the T7 antibody to any endogenous E. coli protein.
We conclude here that some SNPs of human PIMT, as exemplified best by G175R, are highly prone to spontaneous aggregation. The prominent bands at 60 -65 kDa suggest the presence of PIMT dimers, and the stability of the aggregates suggests they arise either from non-reducible covalent crosslinking or highly stable non-covalent interactions akin to those associated with proteins such as ␤-amyloid and synuclein.
Enzymatic Activity of PIMT Variants-The methyltransferase activity of the WT human PIMT produced in this study was essentially identical to two previous reports from our lab wherein we expressed and purified recombinant PIMT starting from rat cDNA. The recombinant rat PIMT characterized by David and Aswad (26) was expressed without any tags and had a specific activity of 16.8 nmol/min of methyl transfer per mg of enzyme using the assay originally described by Aswad and Deight (27). The rat PIMT used by Zhu et al. (20) was expressed with a His/T7 tag (using the same plasmid as in the present study) and had a specific activity of 12.6 -16.3 nmol/min of methyl transfer per mg of enzyme. The specific activity of our human WT PIMT is the same as that reported by Zhu et al., and the activity remained within this range after complete removal of the His tag by thrombin cleavage (Fig. 5). Because the His/T7 tag does not affect WT PIMT activity, all of our activity and stability data were obtained with the N-terminal tags intact.
We used the methyltransferase assay detailed under "Experimental Procedures" to compare the activity of the 14 variants to WT PIMT (Fig. 6A). Except for position 119 in the PIMT sequence, all the variants are quite rare. The NCBI database considers valine at position 119 to be the WT, so this form was used as the basis of comparison in our study. Activity of three

Non-synonymous SNPs in the coding region of the pmct1 gene listed in the dbSNP as of July 2015
The amino acid positions in all the SNP databases are off by ϩ59 because they assume an unused start codon upstream of the real start. Note that the status of these data was updated in the dbSNP in December 2016.  variants is quite remarkable. R36C exhibits no detectable activity, whereas A7P and I58V have activity that is 80 -100% higher than that of WT. Activity is reduced in G175R and F72L by ϳ40 and 20%, respectively. The activity loss in G175R is likely related to its marked state of aggregation (Figs. 2-4). However, as explained under "Experimental Procedures," the specific activity data in Fig. 6 was normalized to the relative amount of protein that runs at 28 kDa, not the total amount of protein in the enzyme assay. Had we used total protein in the specific activity calculations, the activity of G175R would be only 20% of the WT.

SNP ID
Thermal Stability-Previous studies with human RBC extracts from genotyped individuals have shown a marked difference in thermal stability between the two common variants (Val versus Ile) in position 119 of PIMT. David et al. (24) found that preheating for 15 min at 52°C resulted in activity losses of 80 and 20%, respectively. In a subsequent study with a different subject population, DeVry and Clarke (25) obtained qualitatively similar results. Fig. 6B shows the relative heat stability of our PIMT variants as judged by the ratio of activity in preheated (15 min at 52°C) versus control (unheated) samples for each mutant. The horizontal solid line indicates the heated/control FIGURE 3. The presence of high mass protein bands in certain mutants varies between expression preps and is not due to insufficient reduction of disulfide bonds prior to electrophoresis. A, demonstration of moderate prep to prep variation in the staining pattern of WT PIMT and three mutants whose preparations show relatively high levels of putative aggregation. B, demonstration that the level of putative aggregates is not altered by more aggressive sample preparation prior to SDS-PAGE. The first lane in each pair was loaded with 5 g of protein prepared according to our standard procedure (see "Experimental Procedures"), which involves heating for 10 min at 70°C in sample buffer containing 50 mM DTT. In the second lane of each pair, heating was carried out for 5 min at 95°C with 150 mM DTT. ratio obtained with WT PIMT (0.91), whereas the dashed lines indicate two standard deviations of this ratio. Four of the mutants (R17H, R17S, A33S, and G175R) exhibited a loss of activity that was well below the range seen for the WT. It is interesting that, with the exception of G175R, these mutants had basal specific activities (Fig. 6A) nearly identical to that of the WT.
We were surprised to see that neither the WT (Val-119) nor the Ile-119 variant lost much activity upon heating under these conditions. This prompted us to reinvestigate the thermal stability of all our PIMT variants by differential scanning fluorimetry (DSF). This method uses real time quantitative PCR instrumentation to obtain precise melting temperatures (T m ) via increased binding of a soluble fluorophor as a protein unfolds and exposes hydrophobic regions that are normally buried in the native protein (28). Our DSF results (Fig. 7) indicate that only 3 of the 14 variants (F72L, P162S, and K124R) have T m values similar to the WT. Ile-119 has a T m that is 1.6°C higher, whereas all the remaining variants have significantly lower T m values. The Ile-119 result is consistent with its enhanced thermal stability as observed in the two human RBC extract studies mentioned above. R17H and R17S have by far the lowest T m values, consistent with our activity data in Fig. 6B. We were unable to determine a T m for G175R because the DSF curve was highly abnormal. This is no doubt due to its high state of preexisting aggregation.

Discussion
Aggregation-Among the 14 SNPs in this study, we found several that differ significantly from the WT with respect to aggregation, enzymatic activity, and thermal stability. All variants of PIMT exhibited a pair of bands in the 65-70-kDa region, but these were barely detectable for the WT and six of the mutants. Aggregation was most prominent for G175R, accounting for ϳ50% of the total protein. Based on molecular weight, it is likely that one or both of the 65-70-kDa bands are dimers. The resistance of these aggregates to heating in SDS-PAGE sample buffer suggests that they are formed either by an unusual covalent cross-link or, more likely, by strong non-covalent interactions of the type seen in ␤-amyloid, ␣-synuclein, and numerous other proteins. The results shown in Figs. 3 and 4 tend to rule out intermolecular disulfide bonding or lack of purity as contributors to the high mass bands.
We do not believe that the His/T7 tag plays any role in the small degree of aggregation seen in WT PIMT or in those mutants that have similar low levels of aggregation. This small degree of aggregation was seen previously in PIMT purified from bovine brain (27) and in rat recombinant PIMT expressed with or without the same tags (20,26). In our experience, aggregation of PIMT is mainly due to its lack of solubility at concentrations above 1.5 mg/ml. It is tempting to speculate that the aggregation prone variants seen here (e.g. G175R) are also prone to aggregation in vivo, but this is uncertain because of differences in the microenvironment and the presence of the tags.
Enzyme Activity-The results of the activity assays were extremely interesting and highlighted by the complete absence of activity in R36C. Losses of ϳ40 and 20% were also observed for G175R and F72L, respectively. Equally interesting, and unexpected, is the observation that the two of the variants, A7P and I58V, exhibit methyltransferase activity that is almost double that of the WT. The three-dimensional structure of human PIMT with S-adenosyl-L-homocysteine (AdoHcy) bound has been determined to a resolution of 1.6 Å (Protein Data Bank code 1I1N). The guanidino nitrogens of Arg-36 are ϳ6 Å from the ␣-carboxyl of the AdoHcy. Given the buried location of this arrangement and its presumed low dielectric environment, a guanidino-carboxyl electrostatic interaction could be critical for positioning AdoMet for its role as the methyl donor. Gly-175 is located on the enzyme surface in the turn of a loop that separates two ␤-sheet regions of the C-terminal domain. Substitution with an arginine might produce distortion of this region by interactions with the side chains of nearby amino acids such as Gln-198. The phenyl side chain of Phe-72 is exposed to the surface where it might help anchor the two N-terminal helices, a function that could be compromised by the leucine substitution.
Based on the previous studies of human PIMT activity in RBC extracts from genotyped individuals (24,25), we expected that the specific activity of the common Ile-119 variant (essentially another wild type) would be ϳ8% higher than WT, but we saw no apparent difference. We suspect the difference seen with the RBC extracts is related to the fact that human RBCs have an average lifetime in vivo of ϳ115 days (29). Any activity loss during this prolonged "in vivo incubation" is expected to be greater in the more labile (Val-119) of the two forms. The same explanation has previously been espoused by DeVry and Clarke (25), who also observed higher activity in the Ile-199 variant in human RBC extracts. The in vivo incubation effect, in addition to the relative complexity of the extracts, might also explain why preheating at 52°C for 15 min causes much greater loss of PIMT activity in RBC extracts than it did with our recombinant PIMTs.
The unusual enhancement of enzyme activity seen with A7P and I58V was surprising but might have a simple explanation. Because rare and random mutations typically disrupt protein function, we propose that these mutations decrease the binding affinity of PIMT for AdoHcy (S-adenosyl-L-homocysteine), a product of the methylation reaction that competes for AdoMet binding and binds ϳ20 times stronger (30). A more rapid dissociation of AdoHcy from the active site of these variants could thus increase the turnover rate of the enzyme. Because AdoHcy and AdoMet use the same binding site, these mutations might also worsen AdoMet binding; however, our assays were done at near saturating levels of AdoMet, which would minimize any effects on actual AdoMet binding in the assay. This explanation is supported by the fact that Ile-58 is situated just a few Angstroms from the AdoHcy carboxyl group in the crystallized enzyme (Protein Data Bank code 1I1N). Ala-7 is quite distant from the active site, but substitution with a proline would likely cause a significant change in conformation of the N-terminal region with possible long range effects.
Subsequent to obtaining all the data for this paper, we became aware of several computational algorithms for predicting the effects of non-synonymous SNPs on protein function. Table 2 shows the results for our 14 variants using two distinct but related algorithms: SIFT (Sort Intolerant from Tolerant) and fathmm (functional analysis through hidden Markov models). Both algorithms compare sequences of the subject protein over a range of different organisms to determine the tolerance to variation at the SNP sites being queried. We were impressed with how well both predictions correlated with our experimental activity data. R36C is the only mutation predicted to be damaging by both SIFT and fathmm. G175R was the only other mutation seen as damaging by fathmm but with a less severe score than R36C. G175R also had the second lowest SIFT score and barely missed the cutoff for damaging.
A markedly lower thermal stability relative to WT was evident for both mutations at Arg-17. The side chain of Arg-17 thus appears to play an important role in stabilizing the tertiary structure of PIMT. DSF revealed lower T m values for seven other of the rare PIMT variants (a common finding in rare SNPs) and a higher T m for the common Ile-119 variant.
As reviewed in the introduction, there is reason to believe, from both mouse and human studies, that a major deficit in human PIMT activity would lead to severe neurological dysfunction in the young and that moderate reductions would contribute to reduced lifespan and quality of life in the aged. As data gleaned from genome-wide association studies continues to grow, studies such as this will help delineate which mutations contribute to disease states at the level of protein function. From the current study, we predict that heterozygosity for R36C, G175R, R17H, or R17S would be detrimental to successful aging, whereas homozygosity (should it ever occur) would produce devastating neuropathology.

Experimental Procedures
Expression and Purification of PIMT Variants-Custom expression plasmids for all PIMT variants were based on the pET28a(ϩ) vector and were purchased from GenScript (Piscataway, NJ). DNA coding for the human WT PIMT sequence (pcmt1 gene) was synthesized with codon optimization for expression in E. coli and inserted between the BamHI and XhoI sites. Plasmids for SNP variants were generated from the parent WT plasmid using PCR, and the integrity of all plasmid inserts was verified by DNA sequencing.
Expression and purification of PIMT variants was carried out in our laboratory using E. coli BL21(DE3) as described previously (20). The purified WT PIMT had a specific activity of ϳ16,000 pmol/min/mg using ␥-globulin as the methyl acceptor (27). Proteins were stored at Ϫ80°C at concentrations ranging from 0.3 to 1.2 mg/ml in buffer consisting of 25 mM Tris-Cl, pH 7.0, 150 mM NaCl, 5 mM 2-mercaptoethanol, and 10% (w/v) glycerol. Protein concentrations were determined in triplicate by a microtiter plate version of the Bradford assay (31) with reagent and protocol from Bio-Rad and bovine serum albumin as the standard. To assess the effect of the His tag on enzyme activity, WT PIMT was treated with immobilized thrombin using the regents and protocol provided with the thrombin Clean-Cleave TM kit from Sigma. After cleavage, the treated enzyme was dialyzed against PIMT storage buffer, and the protein concentration was redetermined. Complete cleavage in a 2-h incubation was verified by SDS-PAGE. Cleaved and uncleaved versions of WT PIMT were then assayed for activity in the same experiment.
Electrophoresis and Western Blotting-SDS-PAGE was carried out on 1-mm NOVEX NuPAGE 4 -12% gels using the MES running buffer (Thermo Fisher). The gels were stained with 0.05% Coomassie Blue R-250. For Western blotting, unstained proteins were transferred to a PVDF membrane (0.45 m) that was subsequently blocked with 5% nonfat milk and probed with a 1/10,000 dilution of HRP-conjugated goat anti-T7 antibody (Bethyl Labs, Montgomery, TX). Immunostaining as carried out with 3,3Ј,5,5Ј-tetramethylbenzidine (TMB) using the 1-Step TM TMB-Blotting Substrate reagent from Thermo Fisher. PageRuler TM -prestained protein standards were also from Thermo Fisher.
PIMT Activity Assays-PIMT methyl transferase activity was measured in triplicate samples under initial rate conditions using bovine ␥-globulin as the methyl acceptor. Prior to each set of assays, PIMT samples (in reaction buffer plus ␥-globulin) were split into two groups. One group was subjected to a preheat at 52°C for 15 min prior to the assay, whereas the other group remained on ice. Assays were carried out in a final volume of 50 l containing 1) 100 mM BisTris-Cl (pH 6.4), 2) 5.0 mg/ml bovine ␥-globulin (from a stock of 25 mg/ml in 10 mM HCl), 3) 0.05 g of PIMT, and 4) 10 M S-adenosyl-[methyl-3 H]L-methionine ([ 3 H]AdoMet, 10,000 dpm/pmol). Components 1-3 were combined and preincubated in a water bath at 30°C for 2.0 min. The [ 3 H]AdoMet was added, and the reaction was allowed to proceed for 10.0 min, followed by the addition of 750 l of ice-cold 7% (w/v) trichloroacetic acid to terminate the reaction and precipitate the protein. Pellets were recovered by centrifugation, washed once with 800 l of water, and then dissolved in 100 l of 0.2 M NaOH containing 2% (v/v) MeOH. 50 l of this solution was quickly transferred to folded filter paper lodged in the cap of a 4-dram glass vial prefilled with 2.5 ml of Liquiscint counting fluid, and the vial was then rapidly capped. After incubation at 40°C for 1 h to allow hydrolysis of the protein methyl esters and transfer of [ 3 H]methanol to the scintillation fluid, the cap was replaced, and the vial was counted by liquid scintillation. The protein values used in the specific activity calculations refer only to the fraction of protein that runs at 28 -30 kDa. This was determined by integrating the protein stain in each lane of Fig. 2.
Differential Scanning Fluorimetry-The thermal stability of PIMT variants was determined using an Mx3005P quantitative PCR machine (Agilent Technologies). Each 40-l sample contained 20 l of PIMT (0.29 -0.58 mg/ml in storage buffer, pH 7.0) and 40 M Sypro Orange dye. Fluorescence (excitation and emission wavelengths of 492 and 610 nm, respectively) was recorded over the range of 25-95°C with a temperature gradient of 1°C/min. All variants were tested in triplicate on the same plate, and all edge wells were avoided. T m values were extracted as the inflection point of the fluorescence curves.
Calculations and Statistics-Each activity-stability experiment consisted of triplicate samples of WT heated (Wh), WT unheated (Wu), mutant heated (Mh), and mutant unheated (Mu) all assayed in parallel to minimize the effects of potential day to day variation in the specific activity of the [ 3 H]AdoMet (an inherently unstable compound) and water bath temperatures. To calculate the activity of each mutant relative to WT, we calculated, within a given experiment, all nine possible ratios of the individual samples (Wu1/Mu1, Wu1/Mu2 . . . Wu3/Mu3). From these nine ratios we calculated the average and standard deviation. For the preheat stability data, we calculated the average and standard deviation of all nine ratios of heated/unheated for each enzyme tested. For the DSF data, p values were obtained using the two-tailed, heteroscedastic t test.
Author Contributions-C. J., B. C., J.-L. B., K. N., E. H., M. M., and J. K., conducted the experiments. D. W. A., C. J., and B. C. analyzed the data. D. W. A. conceived and designed the project and wrote the paper.

TABLE 2 Computational predictions for the effects of SNPs on PIMT function
Predictions were calculated using online versions of SIFT (Sort Intolerant From Tolerant) and fathmm (Functional Analysis Through Hidden Markov Models). TOL means tolerated, and DAM means damaging or non-tolerated. Scores of less than 0.05 or Ϫ3.00 are considered damaging for SIFT and fathmm, respectively. The two variants that we found to be most damaged (R36C and G175R) are indicated with boldface.