|Home | About | Journals | Submit | Contact Us | Français|
Protein L-isoaspartate O-methyltransferase (PIMT) repairs isoaspartate residues in damaged proteins, and it contains a Val–Ile polymorphismin in α5, ~13 Å from its active site. Val119 has lower activity and thermal stability but increased affinity for endogenous substrates. Studies suggest that heterozygosity for Val/Ile favors efficient isoaspartate repair. We have performed multiple molecular dynamics simulations of 119I and 119V PIMT. Both V119 and I119 interact with the same residues throughout all of the simulations. However, the larger Ile altered the orientations of α5 and β5, both of which have co-substrate binding residues on their distal ends. I119 increases the flexibility of several residues, loosening up the S-adenosylmethionine (SAM)-binding site. These subtle changes are propagated towards the isoaspartate-docking site via residues common to both active sites. The increased mobility in 119I PIMT reorients α3, resulting in a salt-bridge network at the substrate-binding interface that disrupts several key side-chain interactions in the isoaspartate site. In contrast, 119V PIMT remains quite rigid with little change to the co-substrate binding site, which could hinder SAM's binding and release, accounting for the decreased activity. These results shed light on the molecular basis behind the decreased activity and increased specificity for endogenous substrates of 119V PIMT relative to the 119I variant. 119I PIMT catalyzes the methylation reaction but may have difficulties recognizing and orienting specific substrates due to its distorted substrate-binding site. Heterozygosity for both the Ile and Val alleles may provide the best of both worlds, allowing the fast and specific methylation of damaged proteins.
Asparagine residues undergo spontaneous deamidation to form L-isoaspartate, kinking the neighboring main chain structure through the addition of an extra methylene group in the polypeptide backbone (Najbauer et al., 1996; Aswad et al., 2000; Reissner and Aswad, 2003; Shimizu et al., 2005). This process is exacerbated by heat shock and oxidative stress, and it is a major source of protein damage. The accumulation of isoaspartate residues has been shown to affect the function of several cellular proteins, including histone H2B (Young et al., 2001, 2005), Bcl-x (Deverman et al., 2002), protein kinase A (Pepperkok et al., 2000), collagen type 1 (Lanthier and Desrosiers, 2004) and ribosomal protein S11 (David et al., 1999). Isoaspartate residues have also been found in long-lived proteins such as eye-lens β-crystallin (Voorter et al., 1988; Fujii et al., 1994) and myelin basic protein (Fisher et al., 1986; Shapira et al., 1988). The buildup of isoaspartate residues in amyloid Aβ peptide and erythrocytic membrane-proteins, including ankyrin, have been associated with Alzheimer's Disease (Shimizu et al., 2000, 2005) and Down's syndrome (Galletti et al., 2007), respectively.
Protein L-isoaspartate O-methyltransferase (PIMT, EC.18.104.22.168) is a ubiquitous housekeeping protein that catalyzes the first step of isoaspartate repair by transferring a methyl group from co-substrate S-adenosylmethionine (SAM) to the α-carboxyl group of the L-isoaspartate residue (Clarke, 2003; Reissner and Aswad, 2003; Shimizu et al., 2005). PIMT is highly expressed in developing neural tubes, the hippocampus and cerebral cortex, and it is essential for maintaining proper CNS function (Diliberto and Axelrod, 1976; Mizobuchi et al., 1994; Shirasawa et al., 1995; Lanthier et al., 2002). PIMT knockout mice accumulate 4–8-fold more isoaspartate-containing proteins in the brain relative to wild-type mice, leading to atypical neurotransmission, developmental defects and fatal seizures within 2 months of birth (Kim et al., 1997; Zhu et al., 2006b). In addition, overexpression of PIMT in drosophila melanogaster has been shown to increase the organism's lifespan by 32–39% (Chavous et al., 2001).
PIMT has a common valine → isoleucine polymorphism at residue 119 (Ingrosso et al., 1989; Tsai and Clarke, 1994). The 119V and 119I alleles occur with approximately equal frequencies in Caucasian populations, whereas 119I (the ancestral allele) is dominant in Asian and African populations (Tsai and Clarke, 1994; DeVry and Clarke, 1999). The specific activity of 119I PIMT is ~20% higher than that of the 119V protein in red blood cells (David et al., 1997; DeVry and Clarke, 1999). The activity of 119I PIMT is also more thermo-tolerant and less sensitive to oxidative stress than that of 119V PIMT (David et al., 1997). Although there appears to be no difference in the affinities of 119I and 119V PIMT for exogenous peptides (Lowenson and Clarke, 1991), the 119V protein has an ~30% higher affinity for endogenous substrates found in the brain cytosol of PIMT knockout mice (DeVry and Clarke, 1999). In addition, Lowenson and Clarke (1991) showed that despite the higher specific activity of 119I PIMT, the accumulation rate of damaged proteins is governed by substrate affinity. These data suggest that combining the 119I protein's increased activity and stability with the 119V variant's improved substrate affinity may result in the most effective repair of isoaspartate-mediated protein damage. Indeed, a study of PIMT allele frequencies in the Ashkenazi Jewish population showed that 65% of the healthy older population were heterozygous for the V and I alleles compared with just 45% of a younger ethnically matched control group (DeVry and Clarke, 1999). These data suggest that heterozygosity for the V and I alleles promotes longevity and healthy aging, and is advantageous over homozygosity for either variant. Homozygosity for the 119V PIMT allele has recently been shown to decrease risk for spina bifida (Zhu et al., 2006a).
Nine crystal structures of PIMT from human (Ryttersgaard et al., 2002; Smith et al., 2002), drosophila melanogaster (Bennett et al., 2003), thermatoga maritime (Skinner et al., 2000) and pyrococcus furiosus (Griffith et al., 2001) have been solved. The proteins share >33% sequence identity across species and show very high structural homology, underscoring the importance of the PIMT protein in survival. The structure of human PIMT is shown in Fig. 1. It consists of a central 7-stranded β-sheet (organized as 5↑4↑3↑6↑7↑8↓9↑) sandwiched between two sets of α-helices. The co-substrate, SAM, interacts with residues along the C-terminal ends of β-strands 5, 4, 3 and 6. Once bound, residues in β2 and α8 bury SAM so that only its CE methyl group, positioned in the center of the isoaspartate-binding site, is visible (Fig. 1).
PIMT interacts with a diverse group of protein substrates with Km values that can vary 50 000-fold (Lowenson and Clarke, 1991). However, the Vmax for isoaspartate methylation varies only 2–3-fold among substrates, suggesting that PIMT contains a structurally accommodating substrate-binding site to facilitate the reaction (Lowenson and Clarke, 1991). The damaged proteins dock to a large interface on the protein surface defined by residues in α3, β-strands 1, 2, and 9 and α8 (Fig. 1). This interface facilitates substrate binding so that the isoaspartate residue is positioned adjacent to the SAM CE group with a minimum of rearrangement. Residue 119 is located in the third turn of helix α5 near the SAM-binding site and on the opposite face of the protein from the isoaspartate-docking site (Fig. 1). Interestingly, only crystal structures of the ‘less stable’ 119V PIMT variant are available in the PDB.
We are interested in how changes in side-chain interactions within the polymorphic site caused by the introduction of one methyl group at residue 119 can be propagated throughout the protein to alter both the SAM-binding site and the substrate-binding interface. Therefore, we ran multiple long molecular dynamics (MD) simulations of the 119V and 119I PIMT variants at 37°C to examine the effects of the polymorphism on protein structure and dynamics.
Studies have reported that the I119V PIMT polymorphism affects both the activity and protein stability of PIMT. Simulations of the 119I and 119V apoproteins were therefore essential to study the effects of the polymorphic residue on PIMT structure and dynamics. The starting structure for the simulations was a 1.5 Å crystal structure of the 119V variant of human protein L-isoaspartate O-methyltransferase [PDB entry 1I1N, residues 2–225 (Smith et al., 2002)] bound with S-adenosylhomocysteine (SAH). SAH is completely buried in the crystal structure (Fig. 1B) forming several hydrogen bonds and van der Waals contacts with residues in α4, α5, α6, α8 and β2-β6. The presence of SAH in the simulations would likely stabilize the protein through its extensive contacts, masking any structural effects caused by the polymorphism. To date, no structures of the 119I PIMT protein bound with SAM have been published and no information regarding the effect of I119 on the structure of the SAM-binding site is available. This lack of data makes it very difficult to generate an accurate starting structure for a simulation of 119I holoprotein. For these reasons SAH was removed prior to energy minimization and no simulations of PIMT bound with SAH were performed. The 119I PIMT variant was generated by replacing valine 119 with isoleucine in the 1I1N crystal structure, and minimizing the torsional, electrostatic and van der Waals energies of the resultant structure in vacuo (Levitt, 1990).
MD simulations of 119V and 119I PIMT were performed with the in lucem molecular mechanics (ilmm) simulation package (Beck, 2004–2009), using protocols and a potential energy function that have been described elsewhere (Levitt et al., 1995, 1997; Beck and Daggett, 2004). The simulations included all hydrogen atoms and explicit flexible three-centered waters (Levitt et al., 1997). The proteins were solvated in a periodic rectangular box with walls extending at least 10 Å from all protein atoms. Simulations performed at 25 and 37°C had solvent densities set to the experimental densities of 0.997 and 0.933 g/ml, respectively (Kell, 1967). The box volume was held fixed and the NVE microcanonical ensemble employed once the solvent densities were set. A 10 Å force-shifted, non-bonded cut off was used and updated every two steps (Beck et al., 2005). A time step of 2 fs was used in all calculations. All simulations were run for a minimum of 45 ns, with structures saved every 1 ps for analysis. One 25°C and three 37°C simulations were performed for each protein, for a total simulation time of 360 ns.
Average values for Cα-rmsd, contact distances, and solvent-accessible surface area (SASA) were calculated using structures from the last 5 ns (5000 structures) of each simulation to ensure that the analyses included only structures from equilibrated systems in which both the energies and properties had reached convergence. SASA was calculated using in-house software implementing the NACCESS algorithm (Hubbard and Thornton, 1993). A contact distance was defined as a C–C atom distance within 5.4 Å, or any other heavy atom distance within 4.6 Å between two non-neighboring residues. Cα-atom root-mean-square fluctuations (rmsf) values for the PIMT crystal structure [PDB entry 1I1N (Smith et al., 2002)] were calculated using crystallographic B-values (B) via the equation: Cα-rmsf = (3B/(8π2))1/2. Cα−rmsf values were calculated relative to the average structure over the last 10 ns of each simulation.
The data described in Tables I and andIIII are based on sets of three independent simulations for each of 119V and 119I PIMT at 37°C, and all errors are standard deviations in the average values of the ensembles for each property reported. Statistical significance of the data was determined using the Student's t-test (Hoel, 1954).
We performed multiple MD simulations of the apo versions of 119V PIMT [PDB entry 1I1N (Smith et al., 2002)] and a 119I variant created from the crystal structure of 119V PIMT in water. All simulations were at least 45 ns in duration, yielding 0.4 µs of total simulation time. In general, both of the PIMT variant structures were stable throughout the simulations at 37°C, reaching average Cα-atom root-mean-squared-deviations (rmsd) values of 2.9 Å (Table I). The proteins expanded slightly during the simulations, increasing their overall SASAs from 9937 to ~10 800 Å2. This expansion did not greatly affect the overall methyltransferase structure, which was maintained throughout the simulations. However, both the amino and carboxy termini of the proteins became very flexible, undergoing displacements of ~2.3 and 1.6 Å, respectively, from their starting positions. The flexibility of the carboxy terminus and α8 in particular, is expected as α8 completely buries SAM in the crystal structure of the PIMT holoprotein (Fig. 1), and α8 presumably must move to allow SAM binding and SAH release. Interestingly, the solvent exposures of both the SAM-binding sites and the isoaspartate-binding interfaces decrease during the simulations of the PIMT apoproteins relative to those in the crystal structure starting structures (Table I).
Figure 2 shows the average Cα-atom root-mean-square fluctuations (rmsf) for each residue in the 119I and 119V PIMT proteins. The rmsf reflects the degree of main chain fluctuations about the mean structure over the simulation. The Cα-rmsf values for the two proteins are similar to each other and follow the pattern of the crystallographic B-values from the structure of 119V PIMT. However, the overall magnitude of the Cα-fluctuations is much greater than that of the B-values. The low B-values are most likely due to the stabilizing presence of SAH in the crystal structure, which locks the protein in one conformation. The Cα-rmsf values of 119I PIMT peak in β1 and β2 (residues 50–59), the α4–β4 loop (residues 100–104), α5 (residues 112–126) which includes the polymorphic residue 119, the α5-β5 loop (residues 130–135) and at residues 220–225 in the protein's C-terminus (Fig. 2). β-Strands 1 and 2, along with α-helices 5 and 8 contain residues that contribute to SAM-binding. In contrast, the V119 PIMT structures have greater Cα-rmsf values in the protein's N-terminus, α6 (residues 142–149), and the in loops between β6–α7 (residues 162-165), and β8–β9 (residues 197–201). Interestingly, the Cα-rmsf values for helix α5 and residue 119 are lower than the crystallographic B-values for the crystal structure (Fig. 2). This indicates that the backbone of this region remains in a relatively fixed position during the simulations, maintaining the structure of the active-site pocket despite the absence of SAH. Side-chain dynamics must be examined in order to fully understand how residue 119 affects substrate binding.
Residue 119 is located in the third turn of helix α5 near the SAM-binding site and on the opposite face of the protein from the isoaspartate-docking site (Figs 1 and and3).3). Both V119 and I119 form main chain hydrogen bonds with residues V115 and R123 within α5, and hydrophobic contacts with residues in α5 (D116, S118, V122, R123), β5 (L137), and with L130 in the α5–β5 loop (Fig. 3, Table I). Overlaying structures from the final ns of each of the 119V and 119I PIMT simulations show very little difference in the orientations of α5, which reaches a displacement of 0.6 Å relative to the 119V and 119I PIMT starting structures. However, the larger Ile is less accommodating of changes in its local environment caused by the normal dynamics of the solvated protein. Residues V115, L130, V135 and L137 are more mobile in the 119I PIMT simulations (Fig. 2), and the distances between the side chains of these residues and 119I increase during the simulations (Table I). This disruption in packing allows for the reorientation of β5 away from α5, and a significant increase in the solvent exposure of the polymorphic site in 119I PIMT, relative to that in 119V PIMT (Table I, Fig. 3).
Both α5 (D109, H110, L114) and β5 (D141) contain SAM-binding residues on their distal ends (Fig. 3). The slight structural changes caused by the larger Ile residue are propagated along α5 and β5, increasing the flexibility of these SAM-binding residues, whereas slightly distorting the nearby α6, and moving G142 and R143 away from the active site (Figs 3 and and4).4). R143 forms a salt bridge with D217 of α8 in 119V PIMT (Table II). The movement of α6 disrupts the R143-D217 salt bridge in the 119I PIMT simulations, reorienting the C-terminal helix (α8) in relation to the active site and altering the orientation of SAM-binding residue Q221 (Figs 44–6, Table II). The packing around I119 and the increased mobility of α8 lengthen distances between several residues, including S88-D109, D109-Q221 and Q221-H110, that define the SAM-binding site structure (Table II, Fig. 4). These changes lead to a general loosening of the SAM-binding site, and a significant increase in the solvent exposure of the active site in the 119I PIMT simulations (Table I). It is possible that the inherent flexibility of the SAM-binding site in the 119I PIMT apoprotein simulations facilitates co-substrate binding, allowing the active site residues to rearrange around SAM to provide its best fit. In contrast, the SAM binding-site structures from the crystal structure of the 119V PIMT protein bound with SAM and the simulations of the 119V apoprotein are virtually identical (Table I, Fig. 4). Residues within the active site of 119V PIMT are more rigid (Fig. 4), with several pairs of SAM-binding residues forming closer interactions than those seen in the 119I protein. On the other hand, the observed rigidity of the 119V active site could detrimentally affect PIMT activity by requiring SAM to enter the active site in a specific orientation, making co-substrate binding more difficult.
PIMT interacts with isoaspartate-containing proteins through a large interface defined by α4, β1, β2, β9 and α8 (Fig. 1). This binding surface essentially positions the damaged isoaspartate residue adjacent to the CE group of SAM, facilitating its methylation. This substrate-binding interface in positioned on the opposite face of the protein from the polymorphic site (Fig. 1), yet the V119I mutation affects the interface structure, decreasing the protein's affinity for endogenous substrates (DeVry and Clarke, 1999). The SAM- and isoaspartate-binding sites share several common residues [I52 (β1), T57 (β2), S59 (β2), V213 (β9)] (Figs 1 and and4).4). β1 and β2 are slightly more flexible in the 119I PIMT simulations (Figs 2 and and5),5), and the orientations of residues I52, T57 and S59 differ greatly between the 119V and 119I PIMT proteins (Fig. 4). The movement of β-strands 1 and 2 alters the orientations of residues within α3, enabling a small salt-bridge network between D73 (α3), R177 (β8) and D195 (β9) to form along the bottom of the substrate-binding interface that is absent in both the PIMT crystal structure and the simulations of the 119V protein (Table II, Fig. 5). This network strains the orientation of α3 and β9, forcing them apart distal from the salt bridges and opening up the substrate-binding site (Table I, Figs 5 and and6).6). In addition, the reorientation of β9 increases the mobility of α8, facilitating its cyclic opening and closing of the SAM-binding site to promote co-substrate binding and release. Although, the solvent exposure of the isoaspartate-docking site does not differ significantly between the two proteins, interactions between residue pairs that define the interface, such as T57-V213 and M63-Y212, are lost in the 119I simulations (Table II, Fig. 5). This distortion of the substrate-docking site may account for the 119I protein's decreased affinity for endogenous substrates (Lowenson and Clarke, 1991; DeVry and Clarke, 1999). However, it is likely that the isoaspartate-binding sites of both proteins are disrupted in these simulations due to the absence of SAM.
The activity of 119V PIMT is less tolerant to increases in temperature (52°C) and oxidative stress than that of 119I PIMT (David et al., 1997). However, all of the 119V structures within a given simulation were very similar and the I119V mutation did not appear to destabilize the protein at 37°C. In fact, the global secondary structures from throughout the 119V simulations do not differ significantly from that of the 119V PIMT crystal structure. Further studies will be necessary to fully understand how the V119I polymorphism affects the thermodynamic properties of PIMT.
The MD simulations described here reveal the molecular basis behind the decreased activity and increased specificity for endogenous substrates of 119V PIMT relative to the 119I variant. V119 is buried within the polymorphic site, whereas the larger Ile displaces α5, β5 and α6. This displacement increased the flexibility of several nearby residues common to both the SAM- and isoaspartate-binding sites, effectively loosening up the SAM-binding site and distorting the isoaspartate interface. As a result, α8, which caps the SAM-binding site in the crystal structure, becomes much more mobile, promoting SAM:SAH exchange through the cyclic opening and closing of the SAM-site. In contrast, 119V PIMT remained more compact at 37°C, with the SAM-site less exposed to solvent than in the 119I protein whereas key residues involved in isoaspartate docking appeared poised for substrate binding. The results suggest that the decreased activity of 119V PIMT is due to a slower rate of SAM:SAH exchange compared with the 119I variant. However, once SAM is bound, the isoaspartate interface of 119V PIMT is primed to accept a specific substrate and position the isoaspartate residue towards SAM enabling a fast methylation reaction. In contrast, 119I PIMT catalyzes the methylation reaction but may have difficulties recognizing and orienting specific substrates due to its distorted substrate-binding site. Heterozygosity for both the Val and Ile alleles would provide the best of both worlds, allowing the fast (I) and specific (V) methylation of damaged proteins.
This work was supported by the National Institutes of Health (GM50789 to V.D.), the Canadian Institutes of Health Research (DRA MD-75910 to K.R.) and the Microsoft External Research Program (www.microsoft.com/science).
We thank Amanda Jonsson for running the simulations. Figures were produced using the USCF Chimera package from the Computer Graphics Laboratory, University of California, San Francisco (supported by NIH P41 RR-01081) (Pettersen et al., 2004).
Edited by Jane Clarke (Board Member), Alan Fersht (Senior Editor).