Identification of a Highly Diverged Class ofS-Adenosylmethionine Synthetases in the Archaea*

S-Adenosylmethionine is the primary alkylating agent in all known organisms. ATP:l-methionineS-adenosyltransferase (MAT) catalyzes the only known biosynthetic route to this central metabolite. Although the amino acid sequence of MAT is strongly conserved among bacteria and eukarya, no homologs have been recognized in the completed genome sequences of any archaea. In this study, MAT has been purified to homogeneity from the archaeon Methanococcus jannaschii, and the gene encoding it has been identified by mass spectrometry. The peptide mass map identifies the gene encoding MAT as MJ1208, a hypothetical open reading frame. The gene was cloned in Escherichia coli, and expressed enzyme has been purified and characterized. This protein has only 22 and 23% sequence identity to the E. coli and human enzymes, respectively, whereas those are 59% identical to each other. The few identical residues include the majority of those constituting the polar active site residues. Each complete archaeal genome sequence contains a homolog of this archaeal-type MAT. Surprisingly, three bacterial genomes encode both the archaeal and eukaryal/bacterial types of MAT. This identification of a second major class of MAT emphasizes the long evolutionary history of the archaeal lineage and the structural diversity found even in crucial metabolic enzymes.

enzymatic reactions (7). It has been speculated that only ATP is involved in more different metabolic processes (1).
The only known route of AdoMet synthesis is catalyzed by ATP:L-methionine S-adenosyltransferase (MAT, or S-adenosylmethionine synthetase) (EC 2.5.1.6) (8 -10). MAT activity has been described in members of the bacteria, eukarya, and archaea, and the enzyme has been purified to homogeneity from organisms of the first two of these domains (11)(12)(13)(14)(15). Sequences of MAT from eukarya and bacteria reveal that it is a strongly conserved protein typically having 380 -400 residues, with, for example, 59% sequence identity between either of the two human isozymes (which are 84% identical to each other) and that from Escherichia coli.
Despite the conservation of MAT in the bacterial and eukaryal domains, no genes encoding MAT have been identified in the published archaeal genome sequences. A previous report described the partial purification and characterization of MAT activity in the crenarchaeon Sulfolobus solfataricus, but no sequence data were reported (16). The lack of recognizable archaeal homologs of MAT raised the possibility that archaea use an alternative methyl donor in vivo, a plausible hypothesis given the variety and abundance of euryarchaeal methyl-carrying cofactors (17) and the chiral and covalent instability of AdoMet at high temperatures and physiological pH (18).
In order to identify the enzyme that catalyzes AdoMet formation in archaea, we have purified the protein from Methanococcus jannaschii, the first archaeon for which the complete genome sequence was determined (19). We have used mass spectrometry of tryptic peptides to identify the gene that encodes it. Although the protein sequence is distantly related to the bacterial and eukaryotic MAT proteins, we demonstrate that recombinantly produced M. jannaschii MAT protein is functionally similar to those enzymes. Comparison with the crystal structure of the E. coli enzyme (20) indicates that the active site residues are among the few residues conserved across both classes of MAT enzymes.

MATERIALS AND METHODS
Reagents were purchased from Sigma unless noted. AdoMet was obtained from Research Biochemicals International. L-[methyl-14 C]Methionine was purchased from NEN Life Science Products. Ecoscint scintillation fluid was purchased from National Diagnostics.
AdoMet synthetase activity was determined by the [ 14 C]AdoMet filter binding method (12). Routine assays were performed at 58°C in the presence of 0.1 mM L-[methyl-14 C]methionine (1.9 mCi/mmol), 10 mM ATP in 50 mM Hepes⅐K ϩ at pH 8.0 with 50 mM KCl, and 20 mM MgCl 2 . The phosphate forming activity was determined by quantifying orthophosphate production (21,22). Assays of the tripolyphosphate hydrolytic activity contained 50 mM Hepes⅐K ϩ at pH 7.8 with 10 mM KCl and 10 mM MgCl 2 . The enzyme was stable for more than 1 h at 58°C.
Cell Growth-M. jannaschii JAL-1 cells were grown at 85°C using a H 2 /CO 2 /H 2 S gas mixture in a 12-liter reactor at the University of Illinois Fermentation Facility. Anaerobic growth medium was modified from previously described formulations (23,24 MAT Purification from M. jannaschii-Since we anticipated that this purification would not be repeated, no attempt was made to optimize any step. At each step, a portion of the protein was used in pilot chromatographic trials; therefore, it was not meaningful to calculate an overall yield. Purification was monitored by SDS gel electrophoresis on a Pharmacia Phast system using 8 -25% gradient gels. Cells (5 g) were suspended in 20 ml of 0.1 M Tris⅐Cl, 1 mM EDTA, 0.03 mM phenylmethylsulfonyl fluoride, 0.1% 2-mercaptoethanol, pH 8.0. Cells were lysed by a single pass through a French press at 15,000 p.s.i., followed by centrifugation at 15,000 ϫ g for 30 min. The pellet was discarded.
The protein was loaded onto a 2.6 ϫ 16-cm column of phenyl-Sepharose HR equilibrated in 1 M ammonium sulfate, 10 mM Tris⅐Cl, 1 mM EDTA, 10% glycerol, 1 mM dithiothreitol, pH 8.0. The column was eluted at 2.5 ml/min with 160 ml of starting buffer followed by a 750-ml gradient to a final buffer lacking ammonium sulfate. Enzyme eluted near 0.5 M ammonium sulfate. The fractions containing activity were pooled and dialyzed into buffer A (50 mM Tris⅐Cl, 50 mM KCl, 1 mM EDTA, 1 mM dithiothreitol, pH 8.0). The protein was then fractionated by ion exchange chromatography on a 2.6 ϫ 16-cm column of Q-Sepharose equilibrated with buffer A and eluted at 4 ml/min with 150 ml of buffer A followed by a 650-ml gradient to 1 M KCl in buffer A. Enzyme activity eluted at ϳ0.3 M KCl. The fractions containing activity were pooled and dialyzed to equilibrate with buffer A. The protein was further purified by chromatography on a 1.6 ϫ 10-cm column of hydroxylapetite (type CHT-1; Bio-Rad). The column was eluted at 4 ml/ min with 50 ml of buffer A followed by a 200-ml gradient of 0 -0.1 M potassium phosphate in buffer A. Enzyme eluted at ϳ75 mM phosphate. Fractions containing activity were concentrated using Centriprep concentrators (Amicon), and 1-ml aliquots were chromatographed on a 1.6 ϫ 40 cm Superdex-200 gel filtration column that was equilibrated and eluted at 0.5 ml/min with buffer A. The column was calibrated with molecular mass markers (Bio-Rad) to allow estimation of the native size of the enzyme.
Final purification was obtained by chromatofocusing on a Mono-P column (0.5 ϫ 20 cm). Protein was dialyzed into 25 mM imidazole, 1 mM dithiothreitol, pH 6.8. The sample was then loaded onto the column, which was equilibrated with the same buffer. After washing with 10 ml of equilibration buffer, the column was eluted with a solution of a 1:8 dilution of Polybuffer (Amersham Pharmacia Biotech), pH 4.0. Enzyme activity eluted at pH 4.2. The protein showed a single band after silver staining a 8 -25% gradient SDS gel and a 8 -25% nondenaturing gel. The total protein was estimated as 5 g using the Coomassie Plus protein reagent (Pierce) with bovine serum albumin as a standard.
Mass Spectrometry-Mass spectrometry was performed in the Fanny Rippel Biotechnology Facility at the Fox Chase Cancer Center. Approximately 1 g of purified protein was electrophoresed on an 8 -25% Phast gel. The Coomassie-stained protein band was excised, reduced, alkylated, and proteolytically cleaved with trypsin. Peptide mass maps comprising the ensemble of tryptic peptides were obtained by matrixassisted laser desorption ionization-time of flight mass spectrometry (Voyager DE, Perseptive Biosystems) (25)(26)(27).
Cloning-The MJ1208 gene was amplified from M. jannaschii genomic DNA by polymerase chain reaction using the primers MJ1208FWD (GCATATGAGAAACATAATTGTAAAAAAATTAG) and MJ1208REV (GGATCCTTAGAATGTAGTTACTTTTCC). The NdeI and BamHI sites introduced into the gene through the primers were used to clone the DNA into the pET19b vector (Novagen Inc.). The recombinant plasmid (pMJ1208-1) was transformed into E. coli BL21 (DE3) for protein expression.
Expression of M. jannaschii MAT in E. coli-E. coli strain BL21(DE3) pMJ1208-1 was grown in LB medium containing 50 g/ml carbenecillin to an absorbance at 600 nm of 0.6. MAT expression was induced by the addition of 0.1 mM isopropyl-␤-D-thiogalactopyranoside, and the culture was allowed to grow for 5 h. Cells were harvested by centrifugation and stored at Ϫ80°C.
Purification of Recombinant MAT from E. coli-Cells were suspended in 50 mM Tris⅐Cl, 300 mM NaCl, 0.03 mM phenylmethylsulfonyl fluoride, pH 8.0, at a density of 4 ml of buffer/g of cells. Cells were lysed by a sonication for three 30-s cycles, followed by centrifugation at 15,000 ϫ g for 30 min. The pellet was discarded. Imidazole was added to the supernatant to a final concentration of 5 mM. The solution was purified by chromatography on a nickel-chelate column (2 ϫ 8 cm; Novagen). The column was eluted at 0.75 ml/min with 50 mM Tris⅐Cl, 300 mM NaCl, 5 mM imidazole followed by a 150-ml gradient of 5-500 mM imidazole. The eluted protein was concentrated using a Centriplus device (Amicon) and chromatographed on a 1.6 ϫ 40-cm Superdex-200 gel filtration column in 50 mM Tris⅐Cl, 50 mM KCl. The enzyme was exchanged into 50 mM Hepes⅐(CH 3 ) 4 N ϩ , pH 8.0, before use.
Phylogenetic Inference-Amino acid sequences for archaeal-type MAT proteins were obtained from the nonredundant protein data base at NCBI: M. jannaschii (spPQ58605), Methanobacterium thermoautotrophicum ⌬H (spPO27429), M. thermoautotrophicum str. Marburg (spPP26498), Archaeoglobus fulgidus VC-16 (spPO30186), Pyrococcus sp. OT3 (dbjPBAA30943.1), Pyrococcus abyssi (embPCAB49302.1), Aeropyrum pernix (dbjPBAA80596.1), and Aquifex aeolicus (giP2983673). Sequence data from partial genome sequences was obtained from TIGR (available on the World Wide Web) for Chlorobium tepidum, the University of Utah (available on the World Wide Web) for Pyrococcus furiosus, Caltech (available on the World Wide Web) for Pyrobaculum aerophilum, and the University of Oklahoma (available on the World Wide Web) for Streptococcus pyogenes. Sequence data for Methanococcus maripaludis JJ were from unpublished data. 2 Taxa are identified using Swiss-Prot five-letter tags. Amino acid sequences were aligned using the CLUSTALW (version 1.7.4) program (28). From the alignment, 426 positions deemed to be confidently aligned were analyzed by protein maximum parsimony methods using a heuristic search algorithm (PAUP* version 4 beta 2; D. Swofford, Sinauer Associates, Inc). The 500 shortest trees were evaluated by maximum likelihood criteria using the PROTML program (version 2.2) in the MOLPHY package (29) with the JTT model for amino acid substitutions. The TreeCons program (version 1.0) (30) was used to standardize and exponentially weight the trees using maximum likelihood scores and the Kishino-Hasegawa test for significance (p Յ 0.01). The CONSENSE program (J. Felsenstein, PHYLIP (phylogeny inference package), version 3.5c, Department of Genetics, University of Washington) constructed a consensus tree from the weightings.

RESULTS AND DISCUSSION
Purification of AdoMet synthetase from M. jannaschii yielded a protein that chromatographed as an 86-kDa species on gel filtration and gave a single band upon SDS gel electrophoresis with an apparent molecular mass of 44 kDa. Thus, the native protein appears to be a dimer. The specific activity of the purified enzyme was 0.26 mol of AdoMet formed/min/mg of protein at 58°C. A portion of the protein was eluted from a SDS gel and digested with trypsin, and the resultant peptides were analyzed by mass spectrometry. The masses of the peptides were compared with those predicted from the entire M. jannaschii genome. All of the masses were found in the open reading frame containing the MJ1208 gene (Table I). MJ1208 encodes a 406-amino acid protein of 45,252 daltons; we designate this gene as MAT. The peptides identified covered 65% of the protein (264 residues) ranging from residue 3 to 397.
The MJ1208 gene was cloned into an E. coli expression system with an N-terminal decahistidine tag and an inducible T7 RNA polymerase promoter. The recombinant enzyme was then purified using nickel-chelate chromatography and gel filtration. The enzyme is quite thermally stable, losing no activity during incubation for 30 min at 85°C, conditions under which the E. coli MAT has a half-life of less than 1 min (31). The enzyme activity continued to increase at least 90°C.
Functional properties of MAT were assessed at 70°C. No activity was detectable in the absence of Mg 2ϩ , and maximal activity was observed at 10 mM MgCl 2 when ATP was present at 5 mM, indicating that MgATP is probably the true substrate and suggesting involvement of additional Mg 2ϩ beyond the Mg 2ϩ ⅐ATP complex as is the case for the E. coli and human isozymes (14,20). MAT activity was stimulated 10-fold by the addition of KCl to a reaction mix to which only large organic amines (Tris ϩ , (CH 3 ) 4 N ϩ ) had been added as monovalent counterions. Half-maximal activation was observed at 25 mM KCl, which is ϳ10-fold larger than observed with the E. coli and eukaryotic enzymes (9,12). Apparent K m values of 0.25 and 0.14 mM were found for ATP and methionine, respectively. Table II compares steady state kinetic properties of the archaeal enzyme with reported values for bacterial and eukaryal MAT. There are no dramatic functional differences between the archaeal and bacterial enzymes, whereas the eukaryal enzymes have significantly different properties. The two yeast isozymes show cooperative kinetic behavior for both ATP and methionine. The mammalian enzymes have substantially different apparent K m values for methionine that vary with isozyme, and the rat liver isozyme also has a much lower V max . Remarkably, the active site residues are all conserved in the functionally distinct bacterial-and eukaryal-type enzymes.
All characterized MATs from eukarya and bacteria catalyze a two-step reaction in which reaction of ATP and methionine initially yields the AdoMet and tripolyphosphate (PPP i ); the PPP i remains enzyme-bound and is cleaved to pyrophosphate and orthophosphate before product release (9, 10). The phosphorous-containing products of the MAT reaction catalyzed by recombinant M. jannaschii MAT were also P i and PP i , since P i was produced from ATP in the presence of methionine, and the rate of P i production tripled in the presence of inorganic pyrophosphatase. M. jannaschii MAT hydrolyzed added PPP i to P i and PP i . The hydrolytic rate was stimulated by 1.8-fold by AdoMet with half-maximal activation at 19 M. In the presence of saturating AdoMet, the K m for PPP i was 4.5 M, and the maximal rate of P i production was 1.4-fold greater than the maximal rate observed in the MAT reaction. No hydrolysis of 0.1 mM PP i was observed (Ͻ5% of the rate of PPP i ), differentiating the enzyme from pyrophosphatase. In these properties, this protein is remarkably analogous to the eukaryal and bacterial enzymes. The largest functional difference lies in the ϳ10-fold higher concentration of KCl required for activation, commensurate with higher intracellular K ϩ levels expected for the marine microorganism, M. jannaschii.
Phylogenetic Relationship of the Archaeal-type MAT Proteins-Data base searches revealed the presence of this highly diverged M. jannaschii form of MAT in all completed archaeal genome sequences, three phylogenetically diverse bacterial ge-nomes and no eukaryal genomes. This observation raises questions about its evolutionary history and origin. A phylogenetic reconstruction of this history (Fig. 1) is congruent with the archaeal small subunit ribosomal RNA phylogeny (33) and clearly separates the archaeal sequences from the three bacterial sequences. These results suggest that the archaeal-type MAT was present in the archaeal common ancestor and was vertically inherited throughout the lineage.
In contrast, the three bacterial members of the archaeal-type MAT family, found in A. aeolicus, C. tepidum, and S. pyogenes, were probably introduced to their respective genomes by lateral transfer. Curiously, the bacterial members share a unique common ancestor, implicating a single transfer event from the archaeal lineage. All three bacteria also contain bacterial-type MAT genes; this duplicity implies that the archaeal-type MAT genes were recruited for a novel function or for differential regulation of MAT activity, as observed in the S. cerevisiae, human, and plant MAT isozymes (34 -36). Because we cannot confidently root this tree using the bacterial-type MAT sequences, the origin of these transferred proteins remains uncertain.
The extreme sequence divergence of the archaeal-type MAT proteins from the bacterial/eukaryal-type MAT proteins cannot be entirely due to requirements for thermal stability. The bacterial-type MAT protein from the hyperthermophilic bacterium A. aeolicus (95°C optimum growth temperature) is 50% identical to that of the mesophilic E. coli (37°C). Furthermore the sequence of the archaeal-type MAT protein from the hyperthermophilic M. jannaschii (87°C) is 72% identical to that of the mesophilic archaeon M. maripaludis (37°C).
Functional Alignment of M. jannaschii MAT to the Crystal Structure of the E. coli Homolog-An alignment of the M. jannaschii MAT sequence on the E. coli structure was facilitated by the present of short sequences in which active site residues were conserved at similar spacing in the two sequences (Fig. 2). In this alignment all of the insertions and deletions in the M. jannaschii protein fall between elements of regular secondary structure in the E. coli protein and are in surface loop regions. Relative to the E. coli MAT, this protein lacks one surface helix at its carboxyl terminus but has an amino-terminal extension; the latter is also present in numerous eukaryal MAT isozymes. All known MATs are homodimers or tetramers. This association is important for catalytic activity, since the E. coli MAT crystal structure shows both subunits of the dimer contributing functional groups to the interfacial active site (20).
Based on the alignment of the E. coli MAT sequence with all known archaeal-type MAT sequences, we observe that predicted active site residues are disproportionately conserved. Polar amino acids within 3 Å of the ligands observed in the crystal structure of the E. coli MAT (crystallized with ADP and  PO 4 ) are categorically conserved, suggesting that the two different classes of MAT share a common mechanism. A constellation of amino acids contacting Mg 2ϩ and polyphosphate is evolutionarily conserved in the archaeal-type MAT sequences: His-14*, Asp-16*, Lys-165*, Arg-244* and Lys-265 (position numbers refer to the E. coli amino acid sequence and an asterisk indicates amino acids contributed by the second subunit of the homodimeric protein). Two other essential residues in the active site, phosphate-binding Lys-245* and Mg 2ϩ -binding Asp-271, are both replaced by Gly. These replacements are particularly surprising because Lys-245 3 Met and Asp-271 3 Ala mutations in the E. coli MAT reduce activity by nearly 10 5and 10 3 -fold, respectively (37). It is noteworthy that in a linear sequence alignment there are no nearby residues that can replace the missing functional groups, suggesting that the surrogate residues arise from elsewhere in the linear sequence but are brought into the active site in the three-dimensional structure.
In contrast, amino acids responsible for binding the adenine moiety of the substrate and product are less stringently conserved. In the crystal structure of the E. coli enzyme, the side chain amide of Gln-98 hydrogen-bonds to the exocyclic N-6 amine of ATP. Thus, the replacement of Gln-98 with a hydrophobic amino acid (Leu) may explain why ITP is a substrate for the M. jannaschii enzyme (K m ϭ 1.4 mM, V max ϭ 2% of ATP) but neither a substrate nor inhibitor of the E. coli enzyme. Similarly, Lys-269 may recognize the N-3 atom of ATP; Lys-269 3 Met mutations in the E. coli MAT reduce enzyme activity by 10 3 -fold. The replacement of Lys-269 with His in the archaealtype MAT, however, may be a conservative substitution, with no detrimental effect on activity. Least conserved are the residues responsible for K ϩ activation and dimerization of the E. coli MAT. The carboxylate of Glu-42, which is required for K ϩ activation of the E. coli enzyme (38), is replaced by more hydrophobic residues (Leu or Met), and Ser-263 is sometimes replaced by Ala. The predicted hydrogen bond between Ser-80 and Ser-80* would be also lost. These changes in interfacial amino acids suggest that the interactions between M. jannaschii subunits may differ from those of the E. coli subunits.
Of the additional 68 residues that are conserved in the sequence alignment, 13 (17%) are surface residues in the E. coli structure, and the rest appear to be randomly distributed throughout the protein, suggesting that the majority of the matches may be coincidental. This analysis begs the possibility that the protein fold of the archaeal-type MAT is different from the bacterial enzyme; however, the apparent constraints of active site residues distributed through more than 250 residues in the linear sequence makes the possibility of a common topology seem plausible.
Conclusions-This study has identified the methionine adenosyltransferase gene from the hyperthermophilic archaeon M. jannaschii. This sequence has close homologs in the other archaea for which genome sequences have been reported. In contrast, the sequence is highly diverged from the majority of the bacterial and eukaryal MAT proteins, which are similar to  each other. We anticipate that this archaeal class of MAT proteins will bolster future work in elucidating the complete mechanism of MAT catalysis. The identification of the sequence of M. jannaschii MAT has also led to the finding that several bacteria, including the pathogenic organism S. pyogenes, contain both bacterial-and archaeal-type MAT genes. The existence of two diverged classes, bacterial/eukaryal and archaeal versions, of a critical metabolic enzyme is reminiscent of previous observations for thymidylate synthase and dihydropteroate synthase (39). In aggregate, these differences emphasize the profound evolutionary events that accompanied the formation of modern archaeal metabolism. The apparent conservation of active site amino acids between these two diverse classes of MAT focuses our attention on functionally important residues that could not have been identified from alignments of bacterial and eukaryal forms due to the high level of overall sequence identity among those enzymes. Determination of the three dimensional structure of the archaeal enzyme will be of substantial interest.