The Bifunctional Active Site of S-Adenosylmethionine Synthetase

S-Adenosylmethionine (AdoMet) synthetase catalyzes the biosynthesis of AdoMet in a unique enzymatic reaction. Initially the sulfur of methionine displaces the intact tripolyphosphate chain (PPPi) from ATP, and subsequently PPPi is hydrolyzed to PPi and Pibefore product release. The crystal structure of Escherichia coli AdoMet synthetase shows that the active site contains four aspartate residues. Aspartate residues Asp-16* and Asp-271 individually provide the sole protein ligand to one of the two required Mg2+ ions (* denotes a residue from a second subunit); aspartates Asp-118 and Asp-238* are proposed to interact with methionine. Each aspartate has been changed to an uncharged asparagine, and the metal binding residues were also changed to alanine, to assess the roles of charge and ligation ability on catalytic efficiency. The resultant enzyme variants all structurally resemble the wild type enzyme as indicated by circular dichroism spectra and are tetramers. However, all have k cat reductions of ∼103-fold in AdoMet synthesis, whereas the MgATP and methionine K m values change by less than 3- and 8-fold, respectively. In the partial reaction of PPPihydrolysis, mutants of the Mg2+ binding residues have >700-fold reduced catalytic efficiency (k cat/K m ), whereas the D118N and D238*N mutants are impaired less than 35-fold. The catalytic efficiency for PPPi hydrolysis by Mg2+ site mutants is improved by AdoMet, like the wild type enzyme. In contrast AdoMet reduces the catalytic efficiency for PPPi hydrolysis by the D118N and D238*N mutants, indicating that the events involved in AdoMet activation are hindered in these methionyl binding site mutants. Ca2+ uniquely activates the D271A mutant enzyme to 15% of the level of Mg2+, in contrast to the ∼1% Ca2+ activation of the wild type enzyme. This indicates that the Asp-271 side chain size is a discriminator between the activating ability of Ca2+ and the smaller Mg2+.

ϩ PP i ϩ P i enzyme-bound REACTION 1 The structural and mechanistic properties of AdoMet synthetase have been extensively studied in the Escherichia coli metK enzyme, which is a tetramer of identical 383 residue subunits (5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16). The crystal structure of E. coli AdoMet synthetase has been solved by Takusagawa and co-workers as the apo-enzyme (13)(14)(15), and in the presence of ADP and P i (which arose from enzymatic ATP hydrolysis in the crystals), as well as with the bound products PP i and P i . Recent NMR results have provided a proposed binding site of the product AdoMet (16). All of the polar active site residues are conserved in the more than 30 reported AdoMet synthetase sequences. A number of naturally occurring human variants have been recently discovered in patients with familial hypermethioninemia (17)(18)(19); only one of the variations is at an active site residue, an arginine whose crucial role in the E. coli enzyme has been recently characterized (12).
The active site of the enzyme resides between two subunits, such that residues from both subunits make contributions (Fig.  1). The phosphate groups bind at the bottom of a deep cavity with the adenosyl group nearer the entrance. Each active site binds two divalent metal ions (e.g., Mg 2ϩ ) and a monovalent cation (K ϩ ), all of which are required for maximal AdoMet synthetic activity (5)(6)(7)(8)(9)(10). Thus, AdoMet synthetase belongs to a diverse group of enzymes that use two or more metal ions to facilitate phosphoryl group transfers (21). Within the active site are four aspartate residues, two from each subunit (a residue from the second subunit is denoted by *, e.g. Asp-16*, Asp-118, Asp-238*, and Asp-271).
Crystallographic data show that the Asp-16* and Asp-271 side chains each provide the sole protein ligands to one of the two Mg 2ϩ ions; both Mg 2ϩ ions are also ligated to three phosphoryl groups ( Fig. 1) (13)(14)(15). Both Asp-16* and Asp-271 provide bidentate ligands to a Mg 2ϩ ; bidentate coordination of Mg 2ϩ to a carboxylate helps to provide precise positioning of the cation (22). In the absence of substrates, AdoMet synthetase binds a single divalent cation per active site, while two bind in the presence of phosphate containing ligands (6). None of the available crystallographic or spectroscopic data have resolved which site is occupied in the absence of phosphoryl group containing ligands (6 -8). The side chains of aspartates 118 and 238* are implicated in the NMR model as binding the methionine group; Asp-118 interacts with the sulfur and Asp-238* with the ␣-amino group (16). In addition to the overall reaction, AdoMet synthetase catalyzes the Mg 2ϩ -dependent hydrolysis of PPP i to PP i and P i , a reaction that mimics the second part of the synthetase reaction (5,(23)(24)(25). The rate of PPP i hydrolysis is greatly stimulated by AdoMet; the physical mechanism of this activation is as yet unknown but is probably due to local conformational adjustments. The features of AdoMet or the enzyme that contribute to this activation are unclear, although the sulfonium group is important since neither the adenosylhomocysteine nor methylthioadenosine fragments activate the rate of PPP i hydrolysis (5).
In the present studies, the active site aspartate residues have been changed to asparagine to remove their electrostatic contributions to binding and catalysis while maintaining steric and (altered) hydrogen bonding capabilities. In addition, the metal binding aspartates have been changed to alanine to evaluate the role of steric effects in discrimination among cations of different sizes. The properties of these mutants provide insight into the important side chain properties for metal ion binding and into the basis of AdoMet activation of PPP i hydrolysis.
Site-directed Mutagenesis-Oligonucleotides used in mutagenesis were prepared in the Core Facility at the Fox Chase Cancer Center. Mutants were constructed using the Quikchange kit (Stratagene) on plasmid pT7K (11), which has the E. coli metK gene inserted between the PstI and EcoRI sites of plasmid pT7-6. Following transformation and selection for ampicillin resistance, plasmid DNA was extracted using the Wizard Plus Minipreps DNA Purification System (Promega Corp., Madison, WI). Mutants were identified directly by DNA sequencing on an ABI automated sequencer. In all cases complete DNA sequences confirmed that only the desired mutations were introduced.
Expression, Purification, and Characterization of Mutant AdoMet Synthetases-Plasmids were transformed into strain RSR15(DE3). Cultures were grown in LB media containing 50 g/ml carbenecillin. Following overnight growth, 0.1 mM isopropyl-1-thio-␤-D-galactopyranoside was added to the cultures for 30 min prior to harvesting. Mutant proteins accounted for ϳ10 -20% of the total cellular protein.
A standard purification protocol was used to isolate both wild type and mutant AdoMet synthetases (11). Steps consisted of ammonium sulfate fraction followed by successive chromatography on the hydrophobic interaction matrix phenyl-Sepharose HR (Amersham Pharmacia Biotech), hydroxylapatite CHT-I (Bio-Rad), and the weak anion exchange resin aminohexyl-Sepharose (EAH-4B, Amersham Pharmacia Biotech). Purifications were monitored by electrophoresis on 10 -15% gradient gels containing SDS. All AdoMet synthetases were electrophoretically homogeneous. Following purification, AdoMet synthetases were analyzed for oligomerization state by native gel electrophoresis on 8 -25% gradient gels (11). Secondary structure was assessed by circular dichroism spectra recorded on an Aviv model 62A spectropolarimeter. Samples (0.3 mg/ml protein in 25 mM Tris/HCl, 25 mM KCl, pH 8.0) were placed in 1-mm path length cells, spectra were recorded from 200 to 260 nm, and spectra were corrected for buffer contributions.
In AdoMet synthetase activity assays using CaCl 2 in place of MgCl 2 , CaCl 2 was present at 40 mM. It was not possible to perform tripolyphosphatase activity assays in the presence of Ca 2ϩ due to the low solubility of the Ca 2ϩ complexes.
In substrate tests for 2-hydroxy-4-methylthiobutyric acid and 2-keto-4-methylthiobutyric acid, potential products were separated from substrates by chromatography on a Mono-S cation exchange column developed in 0.6 N HCl since the products would be uncharged at neutral pH and would not be retained in the filter binding assay. Reactions contained 10 mM ATP and 10 mM methionine or analog. Product formation was assessed from absorbance at 254 nm; the analogs yielded Ͻ1% of the amount of product obtained with methionine.
Substrate saturation data were evaluated using the kinetic equations of Cleland (28) as implemented in the program Scientist (Micro-Math, Inc.), or the Enzfitter program (Elsevier Biosoft).

Characterization of AdoMet Synthetase Mutants-
The six mutants constructed in this study are physically indistinguishable from the wild type enzyme. All the proteins behaved identically to the wild type enzyme through a purification procedure that exploits a variety of physical properties (in contrast, previously studied mutants at Cys-89 yielded rather differently behaving active dimeric enzymes; Ref. 11). These six mutant AdoMet synthetases are similar to the wild type enzyme in secondary structure as judged by indistinguishable circular residues denoted by * come from a different subunit than the other residues. The unlabeled Mg 2ϩ ligands are probably water molecules that complete an octahedral coordinating sphere but water molecules are not included in this crystal structure. This figure was prepared with MOL-SCRIPT (20). dichroism spectra, and are tetrameric according to native polyacrylamide gel electrophoresis. However, the mutations have profound effects on the catalytic behavior of the enzyme (Tables  I-III and illustrated graphically in Fig. 2).
AdoMet Synthesis Activity-In the presence of Mg 2ϩ as activator, the k cat for AdoMet synthesis by each of the six mutants is reduced by ϳ3 orders of magnitude with respect to the wild type enzyme (Table I). In contrast, there are only 3-and 8-fold alterations in the K m values for MgATP and methionine, respectively. The kinetic properties of the D16A mutant show a 2-fold improved k cat and different K m values from the D16N mutant, being 4-fold diminished for ATP and 3-fold increased for methionine. In contrast, the kinetic parameters for the D271A and D271N mutants are not significantly different. In all six cases, the production of AdoMet is linear with time from within 10% of a single turnover, demonstrating that an early step in the mechanism, likely AdoMet formation, remains ratelimiting. There is no obvious pattern in k cat , K m , or k cat /K m alterations with respect to the proposed role of the residues or the subunit from which the residues originate. It should be noted that K m values, at least for the wild type enzyme, are complex kinetic constants and may not closely reflect true binding constants (29).
Since residues Asp-16* and Asp-271 are implicated in Mg 2ϩ binding, the concentration dependence of Mg 2ϩ activation of the AdoMet synthetase reaction was assessed. The Mg 2ϩ concentration dependences for the D16*N and D271N mutants showed 15-and 10-fold increases in the activator constant (K a ) for Mg 2ϩ , while 11-fold increases are found for both Asp 3 Ala mutants. The Mg 2ϩ activator constants are also increased 8and 10-fold for the D118N and D238*N mutants, respectively.
Tripolyphosphatase Activity-AdoMet synthetase also catalyzes the hydrolysis of added tripolyphosphate, and the reaction rate is stimulated by AdoMet. The kinetic parameters for PPP i hydrolysis by the variants under consideration are summarized in Table II. In the absence of AdoMet, there is a dramatic effect on k cat /K m of mutations at the metal ligand positions Asp-16* and Asp-271. We could not detect PPP i hydrolysis by the D16*N or D16*A mutants, Ͻ0.1% of the wild type activity. For the D271N mutant, the K m for PPP i increased ϳ1000-fold so that it was not possible to saturate the enzyme; however, the extrapolated k cat for the D271N mutant is close to the wild type value and to the value for the D271A mutant, which has a 120-fold elevated K m value. For the D118N and D238*N mutants, the K m values for PPP i did not greatly change from the wild type level, while k cat values decreased 11-fold and actually increased 5-fold, respectively.
In common with the wild type enzyme, AdoMet dramatically enhances k cat /K m for the D16*N, D16*A, D271A, and D271N enzymes, although the AdoMet concentrations required are larger than the wild type levels. For the D271N and D271A mutants AdoMet causes ϳ200-fold and 20-fold reductions in the K m of PPP i to approximately the wild type value; the k cat value appears to decrease 1.5-2-fold, although since the value in the absence of AdoMet is an extrapolation this may not be significant. In the presence of AdoMet, the PPP i K m for the D16*N mutant is ϳ6-fold larger than the wild type value while the K m for the D16*A mutant is 77-fold larger. The k cat values at saturating AdoMet for each of the Asp-16* and Asp-271 mutants are substantially larger than the rate of the overall reaction, consistent with an early rate-limiting step in the overall reaction.
The AdoMet effects on PPP i hydrolysis for methionine binding site mutants are quite different. AdoMet has no effect on k cat for D238*N, although the K m for PPP i does increase 14-fold yielding a less efficient enzyme. Measurements at subsaturating levels of PPP i allowed determination of a half-maximal effect value (K a ) for AdoMet of 0.4 M, 4-fold lower than the wild type enzyme; this parallels the 5-fold reduced K m for methionine in the overall reaction. For the D118N mutant, the K a for AdoMet increased ϳ28-fold, as the K m for methionine   (5). Uncertainties in k cat (moles of product formed/mol of enzyme active site/s) are within 15%, and uncertainties in K m or K a are within 20%.
while the k cat actually decreased 4-fold.
As a complementary approach to mutagenesis, we evaluated interactions of the wild type enzyme with two methionine analogs that lack the ␣-amino group, i.e. 2-hydroxy-4-methylthiobutyric acid and 2-keto-4-methylthiobutyric acid. Neither was a substrate, and concentrations up to 10 mM did not inhibit the rate of AdoMet formation at 0.1 mM methionine, illustrating the importance of the amino group in the affinity for the substrate.
Ca 2ϩ Activation of Metal Binding Mutants-Since previous studies suggested that one of the two metal sites was more tolerant of replacement of Mg 2ϩ by the larger Ca 2ϩ ion (7), we tested Ca 2ϩ activation of the isosteric Asp 3 Asn mutants and the side chain size reduced Asp 3 Ala mutants. Of all the mutants, activity could only be detected for the D271A enzyme (Table III). For the D271A enzyme, the k cat is only 7-fold reduced from Mg 2ϩ , in contrast to the wild type enzyme, where k cat is reduced 78-fold in the presence of Ca 2ϩ . The K a for Ca 2ϩ is 20 mM for the D271A mutant, and 15 mM for the wild type enzyme. The K m values for ATP are similar and are 8-and 5-fold larger than with Mg 2ϩ for the mutant and wild type enzyme. For the wild type enzyme, the methionine K m is 53fold increased in the presence of Ca 2ϩ ; this may reflect that in the presence of Mg 2ϩ the K m may be decreased from the dissociation constant due to contributions from steps subsequent to binding which would be reduced in the slower Ca 2ϩ -activated reaction (29). 2 In the D271A mutant the K m for a methionine is 2.6-fold larger in the presence of Ca 2ϩ than Mg 2ϩ , which may more clearly reflect the changes in dissociation constants. In all, the poor ability of Ca 2ϩ to activate the wild type enzyme appears to be largely due to the bulk of the Asp-271 side chain. DISCUSSION The mutants reported here were constructed to clarify the roles of the conserved active site aspartate residues, two of which are known from crystallography to be involved in Mg 2ϩ binding and two that are proposed to be involved in methionine binding based on NMR and molecular modeling. All six mutants were tetrameric with no obvious change in secondary structure, which facilitates interpretation of the results of the kinetic studies illustrated in Fig. 2. For the Asp-16*, Asp-238*, and Asp-271 mutants, the rate of PPP i hydrolysis in the pres-ence of AdoMet is much greater than the rate of the overall AdoMet synthesis reaction, showing that an early step in the reaction has been preferentially impaired. For the D118N mutant, the k cat values for the overall and PPP i ase reactions are indistinguishable at ϳ1% of wild type levels, showing that both reactions have been significantly impeded. Nevertheless, the time course for a single turnover of the overall reaction by D118N does not show a burst of enzyme-bound AdoMet formation, which confirms that the AdoMet synthesizing step has indeed been hindered. The preferential impairment of AdoMet formation is in common with the four other residues that we have studied by mutagenesis: the K ϩ -binding glutamate-42 residue (10), the reactive cysteines 89 and 239 (11) and the active site arginine 244 (12). All of the mutations have had predominant effects on AdoMet synthesis despite varying locations in the protein structure. 3 The D118N mutation is the only mutation thus far found in which both AdoMet synthesis and AdoMet activated PPP i hydrolysis are comparably reduced, and this is due to AdoMet inhibition, rather than activation, of the PPP i hydrolysis rate.
The lack of a dramatic effect on the concentration dependence for Mg 2ϩ activation of the Asp-16* and Asp-271 mutations was initially surprising. However, a survey of the literature finds precedent in studies of inorganic pyrophosphatase, alkaline phosphatase, and T7 RNA polymerase, all of which transfer phosphoryl groups and contain several metal ions at the active site (30 -34). For pyrophosphatase, which is functionally closest to the PPP i ase activity of AdoMet synthetase, Asp 3 Asn mutations in the metal binding sites caused no change in the concentration dependence of Mg 2ϩ activation but large   (31,32). In the case of T7 RNA polymerase, single Asp 3 Asn mutations decreased the affinity of the Mn 2ϩ probe only 2-5-fold, but inactivated the enzyme (33). In contrast, a single Asp 3 Asn mutation in alkaline phosphatase dramatically decreased Mg 2ϩ affinity, with the result that Mg 2ϩ no longer bound at that site but activated by slowly replacing Zn 2ϩ at an adjacent site (34). For AdoMet synthetase, the lack of a dramatic alteration in the concentration dependence of Mg 2ϩ activation may reflect that the majority of the ligands arise from the phosphate groups rather than the protein so that reduction in the affinity of the free enzyme for the cation may be masked in functional measurements. Nevertheless, the D16*A, D16*N, D271A, and D271N mutations do dramatically impair the PPP i hydrolytic efficiency, particularly in the absence of AdoMet, while AdoMet causes partial restoration of the catalytic efficiency. These results indicate that the natural coordination geometry about each metal ion is well tuned to facilitate both AdoMet formation and the PPP i hydrolysis reaction.
The Ca 2ϩ activation results for the Asp-16* and Asp-271 mutants appear to resolve the uncertainty as to which metal binding site is occupied in the absence of polyphosphate ligands. The relative ability of Ca 2ϩ and Mg 2ϩ to activate the D271A mutant is 15-fold greater than this ratio for the wild type enzyme; the other mutant enzymes, including D16*A, have no detectable activity in the presence of Ca 2ϩ . Thus, Ca 2ϩ , with its large ionic radius (0.99 Å) compared with Mg 2ϩ (0.65 Å) can more easily fit into the void created by the D271A mutation. Since Ca 2ϩ complexes often have more than six ligands (unlike Mg 2ϩ ) (35,36), the additional space may be filled by Ca 2ϩ coordinated waters. This result suggests that the site involving Asp-16* tolerates Ca 2ϩ rather well, and inspection of the crystal structure shows that the environment of Asp-16* is less restricted by other protein groups than is the environment of Asp-271. Thus, it appears that the substrate independent site that selectively binds VO 2ϩ is near Asp-271, which allows Ca 2ϩ to bind at Asp-16* and activate the enzyme in the mixed VO 2ϩ /Ca 2ϩ complex to a ϳ100-fold higher level than the Ca 2ϩ /Ca 2ϩ complex (7). This is consistent with spectroscopic studies showing a lysine amino group ligand to VO 2ϩ (8), and analysis of the crystal structure, which shows the amino group of Lys-265 within 3 Å of the Mg 2ϩ ligated to Asp-271; there is no lysine side chain near the Mg 2ϩ ligated to Asp-16*. Thus, it appears that the divalent metal ion that binds in conjunction with the substrate migrates to the bottom of the active site near Asp-16*. Although the available crystal structures show both metal ions coordinated to all three phosphoryl groups in complexes with PP i plus P i , or ADP plus P i (13)(14)(15), molecular models indicate that doubly tridentate coordination of a triphosphate chain in ATP or PPP i is geometrically unlikely. Thus, the metal ion coordination scheme in the substrate complexes may not be well represented by the available structures of the product complexes. It is possible that a metal ion coordinated water is the source of the oxygen that hydrolyzes the polyphosphate chain, and thus the different properties of Ca 2ϩ and Mg 2ϩ coordinated water could be significant. However, since Ca 2ϩ impairs the AdoMet forming step (7), there is an early event that is less efficient with the larger cation.
The location of the methionine binding site of the enzyme proposed in the NMR-based docking model is consistent with the results of this study (16). The increases in K m for methionine and K a for AdoMet in the D118N mutant are consistent with the loss of an energetically favorable electrostatic interaction between the sulfur and the carboxylate, the larger effect being seen with the positively charged sulfur of AdoMet. Fa-vorable electrostatic interactions between sulfur atoms with a full or partial positive charge have been documented in studies of the crystallographic literature (37) and supported by theoretical studies (38,39). The decreases in K m and K i values for methionine and AdoMet in the D238*N mutant may reflect the inability of the mutant to exploit binding energy in catalysis as manifest in the failure of AdoMet to activate PPP i hydrolysis; thus, favorable binding interactions are reflected in the affinity. The present results indicate that interactions with both the sulfonium and ␣-amino groups of AdoMet are important in activation, consistent with studies using AdoMet analogs (5). The inability of AdoMet to activate the k cat for PPP i hydrolysis in the D118N mutant is consistent with the requirement of enzyme interaction with the sulfonium center since neither the fragments adenosylhomocysteine nor methylthioadenosine activate (5). Furthermore, the methionine analogs having the ␣-amino group replaced by an ␣-hydroxy or an ␣-keto group are neither substrates nor inhibitors of the wild type enzyme, showing the importance of specific recognition of the ␣-amino group in substrate recognition. In contrast, methionine methyl ester is a good substrate, indicating that interactions between the carboxylate and the enzyme are less important (6).
The active site of the enzyme contains contributions from two subunits; however, the results of these and previous mutagenesis studies indicate that the residues from a particular subunit do not have a uniformly greater or lesser contribution to any property measured. Further studies combining mutant enzymes, AdoMet analogs, and structural data are ongoing in order to pinpoint the structural basis of the AdoMet induced activation of PPP i hydrolysis. It is apparent from the sum of the observations of this study that the catalytic efficiency in AdoMet synthesis arises from a highly coordinated active site, which can easily be perturbed to impair the enzyme in unexpected, albeit informative, ways.