Methanococcus jannaschii Uses a Pyruvoyl-dependent Arginine Decarboxylase in Polyamine Biosynthesis*

The genome sequence of the hyperthermophilic methanogen Methanococcus jannaschii contains homologs of most genes required for spermidine polyamine biosynthesis. Yet genomes from neither this organism nor any other euryarchaeon have orthologs of the pyridoxal 5′-phosphate-dependent ornithine or arginine decarboxylase genes, required to produce putrescine. Instead, as shown here, these organisms have a new class of arginine decarboxylase (PvlArgDC) formed by the self-cleavage of a proenzyme into a 5-kDa subunit and a 12-kDa subunit that contains a reactive pyruvoyl group. Although this extremely thermostable enzyme has no significant sequence similarity to previously characterized proteins, conserved active site residues are similar to those of the pyruvoyl-dependent histidine decarboxylase enzyme, and its subunits form a similar (αβ)3 complex. Homologs of PvlArgDC are found in several bacterial genomes, including those of Chlamydiaspp., which have no agmatine ureohydrolase enzyme to convert agmatine (decarboxylated arginine) into putrescine. In these intracellular pathogens, PvlArgDC may function analogously to pyruvoyl-dependent histidine decarboxylase; the cells are proposed to import arginine and export agmatine, increasing the pH and affecting the host cell's metabolism. Phylogenetic analysis of Pvl- ArgDC proteins suggests that this gene has been recruited from the euryarchaeal polyamine biosynthetic pathway to function as a degradative enzyme in bacteria.

Polyamines are ubiquitous organic cations that have been broadly implicated in modulating ion channels, stimulating cell proliferation, and facilitating protein synthesis (1)(2)(3). Acting alongside Mg 2ϩ , polyamines bind to macromolecular polyanions, especially RNA and DNA, stabilizing higher order structures (3). Although polyamine function and biosynthesis have been extensively studied in bacteria and eukaryotes, they are poorly understood in the Archaea, where extremophilic organisms may rely on polyamines to stabilize nucleic acid conformations at high temperatures (4,5).
Cells make putrescine either by decarboxylating L-ornithine or by decarboxylating L-arginine and subsequently hydrolyzing the agmatine product to release urea (13). Gene complementation experiments suggest that the two pathways are functionally interchangeable (14). Remarkably, M. jannaschii has no recognizable homolog of either pyridoxal 5Ј-phosphate (PLP) 1 -dependent enzyme: L-ornithine decarboxylase or L-arginine decarboxylase (15,16). Yet this organism has a homolog of agmatine ureohydrolase (17), suggesting that putrescine biosynthesis proceeds by the L-arginine decarboxylation pathway.
In this paper, we describe a new class of L-arginine decarboxylase (EC 4.1. 1.19) that uses a pyruvoyl group rather than a PLP cofactor to decarboxylate L-arginine to form agmatine. Similar to previously characterized pyruvoyl-dependent decarboxylases, the recombinantly expressed MJ0316 protein selfcleaves to form a 12-kDa (␣) subunit with an amino-terminal pyruvoyl group and a 5-kDa (␤) subunit (18,19). The resulting subunits associate to form a highly thermostable and thermoactive (␣␤) 3 complex. Although the full MJ0316 protein has no significant sequence similarity to any previously characterized protein, the predicted active site of the MJ0316 protein resembles that of the well studied pyruvoyl-dependent histidine decarboxylase (20).

EXPERIMENTAL PROCEDURES
Materials-All reagents and synthetic precursors were purchased from Sigma unless otherwise specified.
Cloning and Recombinant Expression of the MJ0316 Gene in Escherichia coli-The M. jannaschii gene at locus MJ0316 (encoding the protein submitted to the Swiss Protein Data base under Swiss-Prot accession number Q57764) (15) was amplified by PCR from genomic DNA using oligonucleotide primers synthesized by Invitrogen. Primer MJ0316-Fwd (5Ј-GGTCATATGAATGCTGAGATAAAC-3Ј) introduced an NdeI restriction site at the 5Ј-end of the amplified DNA, whereas MJ0316-Rev (5Ј-GATCGGATCCTTATTTATACCACAATGC-3Ј) introduced a BamHI site at the 3Ј-end. DNA fragments were ligated into compatible sites in plasmids pET19b (Novagen) or pT7-7 (21) as described previously (9). Recombinant plasmids were transformed into E. coli BL21-CodonPlus(DE3)-RIL (Stratagene). Sequences of cloned DNA were confirmed by dye-terminator cycle sequencing (University of Iowa DNA Facility). Transformed E. coli cells were grown in complex growth medium, and recombinant protein expression was induced with 28 mM ␣-lactose (9). Cells were harvested 4 h after induction by centrifugation (6,000 ϫ g, 10 min) and were frozen at Ϫ30°C.
Purification of the PvlArgDC Protein-Recombinant pyruvoyldependent arginine decarboxylase (PvlArgDC) protein was purified from soluble cell-free extract by heat treatment and chromatography. E. coli cells (1.1 g, wet weight) expressing PvlArgDC from the pT7-7derived plasmid were suspended in 6 ml of buffer A (20 mM Tris/HCl (pH 7.6)) and lysed by sonication. Soluble cell-free extract was obtained after centrifugation at 16,000 ϫ g for 10 min at 25°C. Native E. coli proteins were denatured by heating the soluble extract at 70°C for 20 min, and insoluble material was removed by centrifugation at 25°C (16,000 ϫ g, 10 min). Heat-soluble cell-free extract (5 ml) was applied to a DEAE-Sepharose FF column (16 mm ϫ 5.2 cm; Amersham Biosciences) equilibrated in buffer A. Protein was eluted with a 30-ml linear gradient from 0 to 1 M NaCl in buffer A at a flow rate of 0.5 ml/min. Fractions (1.0 ml) containing ArgDC activity (which eluted from 0.1 to 0.4 M NaCl) were pooled and sealed in an M r 8000 cut-off membrane (Spectrum). Protein was dialyzed against buffer B (20 mM bis-Tris/HCl (pH 6.5)) for 10 h at 4°C and then concentrated in the membranes using polyethyleneglycol 8000.
Concentrated protein was loaded onto a UNO Q strong anion exchange column (7 ϫ 35 mm; BioRad) equilibrated in buffer B, and protein was eluted with a 25-ml linear gradient from 0 to 1 M NaCl in buffer B at a flow rate of 1.0 ml/min. Fractions containing activity (which eluted from 0.2 to 0.3 M NaCl) were pooled and concentrated in a Centricon-10 ultrafiltration unit (Millipore Corp.). Concentrated protein was applied to a Sephacryl S-200HR size exclusion column (16 mm ϫ 60 cm; Amersham Biosciences) equilibrated in buffer C, which contained 150 mM NaCl and 50 mM Hepes adjusted to pH 7.2 with NaOH. Chromatography was performed in buffer C at a flow rate of 0.5 ml/min, and ArgDC activity eluted maximally at 57 ml. Fractions containing protein were pooled and concentrated in an N 2 -pressurized stirred cell with a YM10 ultrafiltration membrane (Millipore) and then stored at 4°C. Protein purity was evaluated by SDS-PAGE with diamine or nondiamine (Bio-Rad) silver staining. Protein sizes were analyzed using SDSpolyacrylamide gels with either a Tris/glycine buffer system (12% T, 2.7% C acrylamide) or a Tris/Tricine buffer system (12% T, 3.3% C acrylamide). Protein concentrations were measured using the BCA total protein assay (Pierce) with bovine serum albumin as a standard. Recombinant protein expressed with an amino-terminal polyhistidine tag (PvlArgDC His ) was purified by nickel affinity chromatography (9).
Qualitative Analysis of Amino Acid Decarboxylase Activity-To screen compounds as substrates for the M. jannaschii PvlArgDC, amino acids were incubated with enzyme, and the products were analyzed by thin layer chromatography (TLC). Purified PvlArgDC enzyme (1 g) was incubated with 10 mM amino acid and 50 mM Mes/NaOH (pH 6.0) in a volume of 20 l at 70°C for 20 min. Reaction mixtures (2 l) were spotted onto a silica gel 60 TLC plate (5 ϫ 10 cm; EM Science) and developed in a solvent system containing acetonitrile/water/formic acid (88%) at a volumetric ratio of 80:20:10. Plates were dried, sprayed with ninhydrin reagent (0.2% (w/v) in ethanol), and then heated to visualize amines and amino acids. Amino acids tested as substrates included L-argininamide, L-arginine, D-arginine, N G -methyl-L-arginine, L-arginine methyl ester, L-aspartate, L-canavanine, DL-␣-⑀-diaminopimelate, L-citrulline, L-glutamate, L-glutamine, L-histidine, L-lysine, and L-ornithine. Stock solutions of amino acids were dissolved in water and adjusted to neutral pH with HCl or NaOH.
To confirm the identities of decarboxylated amino acids, trifluoroacetyl derivatives of reaction products were analyzed by gas chromatography-mass spectrometry (GC-MS). Reactions (100 l) containing 8 g of PvlArgDC, 50 mM NH 4 OAc, and 10 mM L-arginine, N G -methyl-Larginine, or L-canavanine were incubated at 80°C for 30 min. Reactions were evaporated to dryness by heating under a stream of N 2 before the addition of ϳ100 l of trifluoroacetic anhydride (50% (v/v)) in methylene chloride. Reactions were heated in sealed vials for 12 h at 50°C before evaporation to dryness under a stream of N 2 . Trifluoroacetyl deriva-tives, dissolved in methylene chloride, were analyzed using a VG-70 -70EHF GC-MS operating at 70 eV equipped with an HP-5 column (0.32 mm ϫ 30 m) programmed from 80 to 280°C at 8°C/min.
Quantitative Measurement of Arginine Decarboxylase Activity-One unit of ArgDC activity releases 1 mol/min CO 2 from L-arginine (22). Standard activity assays included 50 mM Mes/NaOH (pH 6.0), 50 mM KCl, 0.5 mM EDTA, 1 mM DL-dithiothreitol (DTT), and enzyme in a volume of 100 l. Buffer salts and enzyme were preincubated at 70°C for 10 min before the addition of 10 mM L-arginine HCl and 135 nCi of L-[1-14 C]arginine (55 mCi/mmol) (American Radiolabeled Chemicals). After a 5-10-min incubation at 70°C, the reactions were terminated by the addition of 100 l of 4 M HCl, and 14 CO 2 was collected on 8-mm diameter circles of Whatman No. 1 filter paper soaked with 10 l of a saturated Ba(OH) 2 solution (23) and placed in the top of a 1.5-ml polypropylene microcentrifuge tube. Acidified reactions were incubated at 70°C for 15 min before filters were removed to vials for liquid scintillation counting in 1 ml of ScintiVerse BD fluid (Fisher).
Temperature, pH, and Cofactor Effects on PvlArgDC Activity-ArgDC activity of the purified enzyme was optimized by varying reaction conditions including temperature, pH, and potential cofactors. All reactions were carried out in enzymatic activity-limited conditions. Effects of reaction temperature were studied in standard assays initiated by the addition of enzyme to reaction mixtures pre-equilibrated in water or sand baths. Reactions were incubated at temperatures from 23 to 90°C for 10 -12 min and then terminated with HCl. To further test its thermostability, 10 g of PvlArgDC (in 20 l of 50 mM Hepes/NaOH (pH 7.1)) was sealed in a melting point capillary tube and then incubated at 125°C for 30 min.
Kinetic Analyses of ArgDC Activity-Initial rates of PvlArgDC-catalyzed decarboxylase activity were measured at various concentrations of L-arginine to infer kinetic parameters. All reactions were preincubated at the reaction temperature before the addition of L-arginine. Assays (100 l) contained 2 g of PvlArgDC, pH buffer, 50 mM KCl, 1 mM DTT, 0.5 mM EDTA, 0.25-15 mM L-arginine, and 50 -150 nCi of L-[1-14 C]arginine (55 mCi/mmol). The pH buffers used were 50 mM trisodium citrate/HCl (pH 4.8), 50 mM Mes/NaOH (pH 6.0) or 50 mM Hepes/NaOH (pH 7.2). Initial rate data were fitted to the hyperbolic Michaelis-Menten-Henri equation, and kinetic parameters were calculated using the Levenberg-Marquardt method of nonlinear least squares regression using equal weighting (SigmaPlot 2000; SPSS Science).
Analytical Size Exclusion Chromatography-Protein subunit structure was analyzed using a Superose 12HR column (1 ϫ 30 cm; Amersham Biosciences). The column was operated and calibrated as described previously (9). Eluted protein was detected by its absorbance at 280 nm and ArgDC activity.
MALDI-TOF Mass Spectrometry of ArgDC-A saturated solution of 3,5-dimethoxy-4-hydroxycinnamic acid matrix was prepared in 50% (v/v) acetonitrile with 0.1% (v/v) trifluoroacetic acid. Purified ArgDC (0.5 l of a 3.9 mg/ml solution) was mixed with 0.5 l of the matrix solution on a stainless steel target (Shimadzu). Dried, crystalline samples were washed with 5 l of 0.1% trifluoroacetic acid and dried under a stream of N 2 . Samples were analyzed in a Kratos KOMPACT SEQ MALDI-TOF instrument (Shimadzu) operated in the linear mode at 20 kV. For each sample, 50 spectra were collected by scanning across the sample, and average masses were identified using the combined sample set. Ions of sodium, matrix, neurotensin, gramicidin D, bovine pancreas insulin chain B (oxidized), and horse heart cytochrome c were used for mass calibration.
Chemical Synthesis of Modified Pyruvoyl Peptide-t-Butoxycarbonyl-L-isoleucine was coupled with ␣-amino-␥-butyrolactone⅐HBr in the presence of triethylamine and N-ethoxycarbonyl-2-ethoxy-1,2-dihydroquinoline in a dimethylformamide and tetrahydrofuran solvent mixture to form the N-t-butoxycarbonyl-L-isoleucine-homoserine lactone peptide (25). The resulting dipeptide was deprotected with 3 M HCl in ethyl acetate (26) and then coupled, using N-ethoxycarbonyl-2-ethoxy-1,2dihydroquinoline, with the ONBH oxime derivative of pyruvic acid. The oxime was formed by the condensation of ONBH hydrochloride with pyruvic acid in pyridine. The resulting peptide was purified by preparative TLC using ethyl acetate as the solvent. The peptide derivative had max at 268 nm and had a partition coefficient between ethyl acetate and water of 26. 1 H NMR and COSY analysis of the peptide was consistent with the assigned structure and confirmed that the peptide consisted of four isomers resulting from the presence of the racemic homoserine lactone and the cis-and trans-oximes. Mass spectral analysis by electron impact (70 eV) gave a M ϩ of 434 m/z. The peptide was also analyzed by liquid chromatography electrospray ionization mass spectrometry (LC-ESI-MS) on a LCQDecaXP instrument (Thermo-Finnigan). Analysis of a sample dissolved in water and separated on a C 18 column (Phenomenex) showed a peak with a MH ϩ ion at 453.1 m/z, indicating that the lactone had undergone ring opening. Tandem mass spectrometry identified a b 2 -type ion (333.8 m/z) and an a 2 -type ion (305.9 m/z), both derived from the 453.1 m/z molecular ion.
Derivatization and Identification of a Pyruvoyl Group-containing Peptide from PvlArgDC-To a sample containing 300 g of PvlArgDC in 100 l of 40 mM Hepes/NaOH (pH 7.2) was added 0.5 mol of ONBH (Fig. 1). The reaction was incubated at 75°C for 30 min. The sample was cooled to room temperature before the addition of 25 mol of NH 4 HCO 3 and 93 mol of 2-mercaptoethanol. Protein was incubated under an atmosphere of N 2 for 12 h before the sample was evaporated to dryness under a stream of N 2 . Protein was resuspended in 200 l 70% (w/v) formic acid with a small amount of cyanogen bromide (27). After a 12-h incubation at room temperature, the sample was evaporated to dryness under a stream of N 2 , suspended in 200 l of water, and extracted with 150 l of ethyl acetate. The organic phase, containing the ONBH-Pvl-Ile-Hse peptide (Fig. 1), was evaporated to dryness under a stream of N 2 , and the residue was resuspended in water for LC-ESI-MS analysis as described above.
Identification of ArgDC Homologs, Sequence Alignment, and Phylogenetic Inference-The translated sequence of the M. jannaschii gene MJ0316 (Swiss-Prot accession number Q57764) was used in iterative queries of the nonredundant protein data base at the National Center for Biotechnology  Amino acid sequences were aligned automatically using the Clust-alW program (version 1.82) (29). From the alignment of 17 protein sequences, 165 positions were deemed to be confidently aligned. These were analyzed by protein maximum likelihood methods using the ProML program (30) with the JTT amino acid substitution model. Bootstrap proportions were calculated using Seqboot, ProML, and Con-sense programs (30) to create and evaluate 500 resampled alignments. An alternative tree was inferred using the Protdist and Neighbor programs (30) to analyze the same alignment.

Identification, Expression, and Purification of M. jannaschii
PvlArgDC-The complete genome sequence of M. jannaschii encodes a single member of the PLP-dependent family of enzymes that decarboxylate amino acids at the C-␣ position, which includes the canonical arginine decarboxylase enzyme (15,31). The protein encoded by this locus (MJ1097) is most similar to bacterial diaminopimelate decarboxylases. We have demonstrated that the recombinantly expressed MJ1097 enzyme decarboxylates DL-␣-⑀-diaminopimelate (to produce L-lysine) but does not decarboxylate L-arginine or L-ornithine. 2 Therefore, M. jannaschii must have a noncanonical means of producing putrescine that is required for spermidine biosynthesis.
A cluster of genes in the M. jannaschii genome encodes all of the proteins required for polyamine biosynthesis from metabolic intermediates. S-Adenosylmethionine decarboxylase (locus MJ0315) (11,12), a homolog of spermidine synthase (MJ0313), a homolog of agmatine ureohydrolase (MJ0309) (17), and a previously uncharacterized open reading frame (MJ0316) are conserved in most euryarchaeal genomes. This group of genes that are required for polyamine biosynthesis suggested that the MJ0316-encoded protein could be an arginine decarboxylase. This protein sequence has no significant similarity to characterized proteins in sequence databases; nor do homologs of MJ0316 share any conserved lysine residue required to bind a PLP. However, members of this family of uncharacterized proteins have conserved serine-serine residues similar to the pyruvoyl group-forming cleavage site of some histidine decarboxylase enzymes (32,33).
E. coli cells that expressed the M. jannaschii MJ0316 gene from a strong promoter produce large amounts of recombinant protein, comprising ϳ10% of the cells' total soluble protein.
Protein purification by heating (to denature most native E. coli proteins), anion exchange, and gel filtration chromatography removed most contaminating proteins and nucleic acids. Table  I shows the purification of 8 mg of MJ0316 protein from 1.1 g (wet mass) recombinant E. coli. The enzyme is soluble at concentrations of at least 4 mg/ml, and it retained 80% activity when stored for 8 months at 4°C in 50 mM Hepes/NaOH (pH 7.5). Diluted in phosphate-buffered saline (pH 7.4), the purified protein had a single absorbance maximum at 279 nm and an extinction coefficient ⑀ 280 ϭ 1.7 ml/mg/cm.
Amino Acid Decarboxylase Activity of MJ0316 Protein-When MJ0316 protein was incubated at 70°C with 10 mM L-arginine or analogs (N G -methyl-L-arginine and L-canavanine), the enzyme produced the respective decarboxylated amino acids. Trifluoroacetyl derivatives of the L-arginine reaction product analyzed by GC-MS resolved into six peaks that are isomers of the tri-to pentatrifluoroacetyl derivatives of agmatine. Molecular and fragment ions, especially  [CH 2 NHCOCF 3 ] ϩ 126 m/z, were characteristic of ␣-amino trifluoroacetyl derivatives of both reaction product and a known sample of agmatine. Derivatives of enzyme reaction product formed from N G -methyl-L-arginine included the di-to pentatrifluoroacetyl derivatives of N G -methylagmatine. Molecular and fragment ions of N G -methylagmatine-trifluoroacetyl derivatives were 14 m/z units higher than peaks from their respective agmatine-trifluoroacetyl derivatives. Finally, the ditrifluoroacetyl derivative of decarboxylated L-canavanine was identified by its molecular ion (M ϩ ϭ 324 m/z), and a tetratrifluoroacetyl derivative (M ϩ ϭ 516 m/z) was inferred from fragment ions. No decarboxylation products were detected in TLC analyses of reactions including enzyme and D-arginine, L-aspartate, L-citrulline, L-histidine, L-lysine, or L-ornithine. Because the MJ0316 protein specifically decarboxylates L-arginine using a pyruvoyl group (described below), this enzyme is the first described PvlArgDC.
Optimization of Enzyme Activity-In standard assays of L-[1-14 C]arginine decarboxylase activity, purified PvlArgDC had a specific activity of 2.5 units/mg and required no exogenous cofactors for full activity. The purified enzyme contained no detectable PLP (measured by UV-visible absorbance spectroscopy), and the addition of 0.1 mM PLP had no effect on the enzyme's ArgDC activity. Nor did additions of 5 mM thiamine pyrophosphate, 1 mM 2-mercaptoethanol, 1 mM dithionite, 5 mM MgCl 2 , 25-200 mM KCl, 50 mM NaCl, or 50 mM NH 4 Cl affect activity. Although neither 0.5 mM EDTA nor 1 mM DTT enhanced activity, both were included in standard reactions, along with 50 mM KCl, to minimize potential variation in activity.
Assays of ArgDC activity at varying pH showed that Pvl-ArgDC has maximum specific activity at pH 6.0, although the enzyme has significant activity over the range of physiologically relevant pH values (Fig. 2). The PvlArgDC enzyme is both thermoactive and thermostable; specific activity increases at higher temperatures up to 90°C (the practical limit for the ArgDC assay). Enzyme heated in an autoclave for 20 min at 121°C retained 50% activity when assayed at 70°C. When heated for 30 min at 125°C in a sealed glass capillary tube, it retained 16% activity.
Kinetic Analyses of ArgDC Activity-To measure the kinetic properties of PvlArgDC, initial reaction rates were calculated from measurements of 14 CO 2 released at various L-[1-14 C]arginine concentrations. Although the agmatine reaction product inhibits ArgDC activity (discussed below), product accumulation did not significantly affect initial rate measurements. Therefore, ArgDC activity data were fit to the Michaelis-Menten-Henri first-order rate equation. At pH 6.0 and 70°C, kinetic parameters (and their S.E.) were K m ϭ 4.7 Ϯ 0.6 mM and V max ϭ 4.7 Ϯ 0.3 units/mg (Fig. 3). At pH 6.0 and 83°C, the enzyme's turnover number increases to 2.7 s Ϫ1 , with K m ϭ 7.1 Ϯ 1.0 mM and V max ϭ 9.0 Ϯ 0.6 units/mg. The pyruvoyl-dependent histidine decarboxylase enzyme undergoes a pH-dependent conformational change that significantly reduces ac-tivity and increases the K m by several orders of magnitude at neutral pH (34,35). However, no comparable change was observed in the PvlArgDC, for which K m ϭ 3.8 Ϯ 0.5 mM and V max ϭ 4.3 Ϯ 0.2 units/mg at pH 4.8 and K m ϭ 8.3 Ϯ 1.2 mM and V max ϭ 3.0 Ϯ 0.2 units/mg at pH 7.2 (assayed at 70°C).
Inhibitors of ArgDC Activity-Substrate and product analogs were tested for their ability to inhibit PvlArgDC activity at 70°C in reactions containing 5 mM L-arginine. The most effective inhibitors of ArgDC activity were arginine analogs: Largininamide and L-arginine methyl ester substantially abolished ArgDC activity at 5 mM concentrations. Other basic amino acids, L-citrulline, L-ornithine, L-histidine, L-lysine, and D-arginine, were not inhibitory at 10 mM concentrations. Gua-  nidine and methylguanidine reduced activity by at least 50% at 5-10 mM concentrations. Low concentrations of urea (10 -20 mM) had no effect on activity, although 4 M urea reduced activity by 80%. PvlArgDC activity was also inhibited by agmatine at a 50% inhibitory concentration of 1.4 Ϯ 0.1 mM, in the presence of 5 mM L-arginine. Polyamines putrescine, spermidine, spermine, and cadaverine did not affect activity (at 10 mM concentrations). As expected for a pyruvoyl-dependent enzyme, reagents that react with ketones to form covalent adducts were potent inhibitors. Preincubation with 10 mM phenylhydrazine (32), 5 mM methoxyamine, 5 mM o-phenylenediamine (36), or 5 mM ONBH reduced activity by more than 90%.
Subunit Composition of PvlArgDC-Pyruvoyl-dependent proenzymes self-cleave at a serine residue to form a large subunit (␣) that contains the nascent pyruvoyl group and a small subunit (␤) that includes the amino-terminal residues of the proenzyme (37). Purified PvlArgDC analyzed by MALDI-TOF mass spectrometry (three experiments with separate calibrations) identified three ions: 12,356 Ϯ 57 m/z, 5436 Ϯ 11 m/z, and 5301 Ϯ 14 m/z (Fig. 4). The largest ion corresponds to the predicted ␣Ϫsubunit (expected M r ϭ 12,272). Derivatization with ONBH shifts this ion by 145 m/z, close to the expected increase of 152 m/z (Fig. 4). The smaller ions correspond to the predicted ␤Ϫsubunit (expected M r ϭ 5435) and the aminoterminal methionine-cleaved ␤Ϫsubunit (expected M r ϭ 5294). MALDI-TOF analysis of PvlArgDC His , containing an aminoterminal polyhistidine tag, formed 12,246 and 8054 m/z ions, which correspond to the ␣Ϫsubunit and ␤ His Ϫsubunit (expected M r ϭ 8192). No ion corresponding to the MJ0316 proenzyme (expected M r ϭ 17,710) was detected. These results indicate that cleavage of the MJ0316 proenzyme between amino acids Ser 52 and Ser 53 forms active PvlArgDC.
Purified PvlArgDC preparations show three distinct bands on an SDS-polyacrylamide gel; these proteins had apparent molecular masses of 44, 17, and 13 kDa (Fig. 5, lane 1). Migrations of the two largest bands change relative to denatured protein standards depending on the gels' acrylamide and crosslinker compositions. A diffuse band, with an apparent molecular mass of 7 kDa, was not detected by diamine silver staining but was visible by nondiamine silver nitrate staining (38), presumably due to the lack of cysteine residues in the ␤-subunit of PvlArgDC. Separation on a more highly cross-linked SDS-polyacrylamide gel resolved three bands with apparent molecular masses of 57, 25, and 11 kDa (Fig. 5, lane 2).
The two fast migrating bands (5-7 and 11-13 kDa) correspond to the ␤and ␣-subunits of PvlArgDC, respectively. The two slow migrating bands may correspond to nondenatured forms of PvlArgDC, similar to those observed for the pyruvoyldependent L-aspartate-␣Ϫdecarboxylase (39). Besides being remarkably thermostable, PvlArgDC retains full activity in the presence of 1% (w/v) SDS at 70°C; therefore, the enzyme may not denature during standard SDS-PAGE sample preparation procedures. This enzyme also resists proteolytic cleavage by trypsin, pepsin, carboxylpeptidase A, or papain, although the slowly migrating bands were cleaved in samples treated with ␣-chymotrypsin or proteinase K. When PvlArgDC was heated with ONBH or methoxyamine prior to electrophoresis, the high FIG. 5. SDS-PAGE analysis of PvlArgDC. Shown are nondiamine silver-stained gel (12% T, 2.7% C acrylamide) of PvlArgDC (lane 1) and silver diamine-stained gel (12% T, 3.3% C) of PvlArgDC using a Tris/ Tricine buffer system (lanes 2-4). PvlArgDC incubated alone at 80°C for 20 min forms two high molecular weight bands (lane 2) that substantially dissociate into the low molecular weight band in identical samples incubated with 2.5 mM ONBH (lane 3) or 2.5 mM methoxyamine (lane 4). PvlArgDC samples (10 g) and protein standards (not shown) were heated in SDS-PAGE loading buffer at 100°C for 5 min before electrophoresis.

FIG. 4. MALDI-TOF mass spectra of PvlArgDC (bottom spectrum) and ONBH-derivatized PvlArgDC (top spectrum).
Both spectra have ions 5311 m/z and 5440 m/z that correspond to the ␤-subunit formed from the amino-terminal residues of the proenzyme during protein cleavage. A 12,379 m/z ion in the PvlArgDC spectrum corresponds to the ␣-subunit, which contains the active site pyruvoyl group. Treatment with ONBH forms an oxime derivative of the ␣-subunit (broken line in the inset) that is 145 m/z units higher than the underivatized form (solid line in the inset). Incomplete proteolysis by E. coli methionyl aminopeptidase gives rise to both ␤-subunit and methionine-cleaved ␤-subunit ions that are invariant in both spectra. All ions are identified by their average mass/ charge values. molecular weight forms dissociated into low molecular weight products, demonstrated by increased staining of the ␣-subunit band (Fig. 5, lanes 3 and 4).
From a Sepharose 12HR size exclusion column, both Pvl-ArgDC and PvlArgDC His eluted with Stokes radii of 30 Å, corresponding to an apparent molecular mass of 55,900 Da. This elution profile suggests that PvlArgDC has a (␣␤) 3 subunit composition.
CNBr Cleavage and Peptide Identification-To confirm the site of protein cleavage deduced from MALDI-TOF MS measurements, the ONBH derivative of PvlArgDC was cleaved with CNBr, and the resulting derivative of the pyruvoyl-isoleucylhomoserine lactone peptide (Fig. 1) was analyzed by LC-ESI-MS. The ion profile of the molecular ion MH ϩ ϭ 453 m/z eluting from a reversed phase column was consistent with that of synthetic peptide (maximum intensity at 21.1 min).
Sequence Analysis of ArgDC Homologs and Phylogenetic Inference-In an alignment of the M. jannaschii PvlArgDC amino acid sequence with several homologous sequences, most amino acid positions are not conserved (Fig. 6). However, residues near the site of protein cleavage (Ser 52 -Ser 53 ) are conserved in PvlArgDC homologs and correspond to conserved residues near the cleavage site of the Lactobacillus 30a pyruvoyl-dependent histidine decarboxylase (PvlHisDC-Ser 81 -Ser 82 ) (40). A position containing conserved hydrophobic residues (Leu 31 ) corresponds to the substrate binding pocket lid of histidine decarboxylase (PvlHisDC-Ile 59 ) (20,41). The PvlArgDC amino acid Phe 34 corresponds to PvlHisDC-Tyr 62 . A Y62F mutant of Pvl-HisDC retains activity but has a higher K m than wild-type enzyme (41). Finally the PvlArgDC Asp 35 aligns with Pvl-HisDC-Asp 63 , which has been proposed to form an ion pair with the imidazolium side chain of that enzyme's L-histidine substrate (41). In the arginine decarboxylase, the aspartyl side chain could form an equivalent ion pair with the guanidino group of arginine and agmatine.
Outside of the active site region, no residues are conserved between PvlArgDC and PvlHisDC sequences. Two positions (PvlHisDC-Asp 53 and -Asp 198 ) implicated in the pH regulatory mechanism of histidine decarboxylase cooperativity are notably absent from PvlArgDC sequences (35,42).
The phylogeny of PvlArgDC homologs was inferred using maximum likelihood criteria from a short alignment of 17 highly diverged sequences (Fig. 7). The distribution of homologs from all euryarchaeal genomes is consistent with vertical inheritance of the gene from an ancestor of the euryarchaeal lineage. Three paralogs in A. fulgidus and two paralogs in M. barkeri resulted from separate gene duplication events in each lineage. One of the A. fulgidus paralogs is severely truncated and may be a pseudogene (not shown in Fig.  7). The bacterial homologs of PvlArgDC are particularly interesting; the phylogenetic tree suggests at least three independent lateral gene transfer events. The homolog in the hydrothermal vent bacterium C. hydrogenoformans is most similar to genes from hydrothermal vent-dwelling archaea. The gene in the moderately thermophilic green sulfur bacterium C. tepidium resembles genes from the Archaeoglobus/Methanosarcina lineage. Finally, highly diverged PvlArgDC homologs in the bacterial pathogens P. gingivalis and Chlamydia spp. may be products of an ancient gene transfer and a rapid rate of protein evolution. The phylogeny of PvlArgDC homologs in- ferred by neighbor-joining methods is substantially similar, except for the placement of Chlamydia spp. and P. gingivalis homologs.

DISCUSSION
Amino acids are not only substrates for protein synthesis and foodstuffs for heterotrophic organisms but are also the feed stock for numerous cofactor biosyntheses. Some of the simplest putative neurotransmitters and molecular signals (agmatine, ␥-aminobutyrate, dopamine, ethanolamine, histamine, taurine, and tyramine) are all formed by eukaryotic PLP-dependent amino acid decarboxylases. In methanogenic archaea, tyramine is a precursor of the cofactor methanofuran (43), ␤-alanine may be a pantothenate precursor (44), and agmatine is required for polyamine biosyntheses. For the decarboxylation of L-histidine and now L-arginine, both PLP-dependent and pyruvoyl-dependent decarboxylases have been identified. Because modern organisms, including M. jannaschii, use PLP as a cofactor in numerous amino acid reactions, it is surprising that pyruvoyl-dependent amino acid decarboxylases are so widespread.
By analogy with other pyruvoyl-dependent enzymes, the Pvl-ArgDC proenzyme self-cleaves by nonhydrolytic serinolysis at a serine-serine peptide bond to form an ␣-subunit (with a catalytic pyruvoyl group) and a ␤-subunit (with substrate-binding residues) (45). Yet none of the previously identified families of pyruvoyl-dependent enzymes appear to be related. Both Pvl-ArgDC and PvlHisDC form trimers, and both share a similar set of conserved residues near the catalytic site of protein cleavage and pyruvoyl group formation. Although these features suggest that the two genes share a common ancestor, the lack of significant sequence similarity outside of this active site region requires a structural model of PvlArgDC to test the relationship.
An arginine decarboxylase from the oat plant is also activated by proteolytic cleavage (46). Unlike PvlArgDC, that enzyme requires a separate protease to cleave the recombinant protein, and it has significant sequence similarity to the PLPdependent E. coli biosynthetic arginine decarboxylase SpeA. Therefore, the oat enzyme is not homologous to PvlArgDC.
Both pyruvoyl-dependent enzymes and their PLP-dependent analogs have similar reaction mechanisms (19), and both groups have comparable kinetic properties (Table II). Compared with the SpeA biosynthetic ArgDC from E. coli, Pvl-ArgDC has a similar turnover number but a much higher K m . In contrast to the SpeA enzyme, PvlArgDC is not feedbackinhibited by putrescine or spermidine end products (47). Therefore, the high K m of PvlArgDC could be advantageous for an anaerobic, autotrophic organism with no apparent means to oxidatively degrade or recycle polyamines.
In a related comparison, the pyruvoyl-dependent histidine decarboxylase of Lactobacillus sp. 30a has similar kinetic properties to its PLP-dependent analog from Morganella morganii (Table II). However, only the activity of the pyruvoyl-dependent enzyme is exquisitely regulated by pH (34,48). Although most proton-consuming decarboxylation reactions are faster at acidic pH, low pH specifically stabilizes an ion pair between Asp 198 of the PvlHisDC ␣-subunit and Asp 53 of an adjacent ␤-subunit to align active site residues (35). This ion pair is not conserved in PvlArgDC.
In contrast to the PvlHisDC proteins, PvlArgDC homologs are found in many microbes, including all euryarchaea and several bacterial lineages. Homologs in the intracellular pathogens Chlamydia spp. are particularly interesting, because numerous pathogenic microorganisms use host cell L-arginine as an energy or nitrogen source. Yet genomes of Chlamydia spp. have no recognizable genes for arginine (or agmatine) catabolism other than PvlArgDC. Instead, in those genomes, the PvlArgDC genes are adjacent to genes encoding basic amino acid permease transporters that are similar to the arginine/ ornithine antiporter (arcD from Pseudomonas aeruginosa) and the putrescine/ornithine antiporter (potE from E. coli). Therefore, Chlamydia spp. could take up arginine (and a proton) and decarboxylate it to produce and export agmatine. In addition to depleting the host cell's arginine pool, such a system would raise the extracellular pH and introduce significant amounts of agmatine, an inhibitor of nitric-oxide synthase and a suppressor of apoptosis (49,50). The genome of Clostridium perfringens encodes a similar amino acid permease gene adjacent to a homolog of PvlHisDC (51). As observed in Lactobacillus sp. 30a, C. perfringens may import histidine and excrete the decarboxylated product, raising the extracellular pH (52,53).
Geneticists have long appreciated the importance of microbial operons in co-regulating gene expression (54). Yet comparative genome analyses have unveiled clusters of functionally related genes whose breadth exceeded expectations (55). Combined with a comparative analysis of gene distribution in organisms of known phenotype, gene cluster analysis led us to identify the novel arginine decarboxylase described here. Furthermore, the same type of analysis led us to propose a completely different physiological function for PvlArgDC homologs in pathogenic bacteria.
This finding of a pyruvoyl-dependent arginine decarboxylase revisits the redundancy in nature of having both PLP and pyruvoyl-dependent amino acid decarboxylases (19). An attractive model for evolutionarily primitive decarboxylation reactions, pyruvoyl cofactors are formed by acyl rearrangement reactions that are conserved in protein cleavage and splicing mechanisms (56). Yet PLP is pervasive in modern organisms,  (57); SpeA, a constitutively expressed biosynthetic arginine decarboxylase from E. coli (47,58); PvlHisDC from Lactobacillus sp. 30a (59); and HisDC from M. morganii (48).
b One unit of activity decarboxylates 1 mol of amino acid per min. c The structures described for pyruvoyl-dependent enzymes refer to paired (␣␤) subunits.
and PLP-dependent enzymes should be less susceptible to costly substrate-mediated transamination side reactions that inactivate amino acid decarboxylases.