S-Adenosylmethionine Decarboxylase from the Archaeon Methanococcus jannaschii: Identification of a Novel Family of Pyruvoyl Enzymes

Abstract

Polyamines are present in high concentrations in archaea, yet little is known about their synthesis, except by extrapolation from bacterial and eucaryal systems. S-Adenosylmethionine (AdoMet) decarboxylase, a pyruvoyl group-containing enzyme that is required for spermidine biosynthesis, has been previously identified in eucarya and Escherichia coli. Despite spermidine concentrations in the Methanococcales that are several times higher than in E. coli, no AdoMet decarboxylase gene was recognized in the complete genome sequence of Methanococcus jannaschii. The gene encoding AdoMet decarboxylase in this archaeon is identified herein as a highly diverged homolog of the E. coli speD gene (less than 11% identity). The M. jannaschii enzyme has been expressed in E. coli and purified to homogeneity. Mass spectrometry showed that the enzyme is composed of two subunits of 61 and 63 residues that are derived from a common proenzyme; these proteins associate in an (αβ)₂ complex. The pyruvoyl-containing subunit is less than one-half the size of that in previously reported AdoMet decarboxylases, but the holoenzyme has enzymatic activity comparable to that of other AdoMet decarboxylases. The sequence of the M. jannaschii enzyme is a prototype of a class of AdoMet decarboxylases that includes homologs in other archaea and diverse bacteria. The broad phylogenetic distribution of this group suggests that the canonical SpeD-type decarboxylase was derived from an archaeal enzyme within the gamma proteobacterial lineage. Both SpeD-type and archaeal-type enzymes have diverged widely in sequence and size from analogous eucaryal enzymes.

Polyamines are ubiquitous small molecules that are found at high concentrations in most organisms (7, 16, 30). They appear to stabilize molecular complexes, such as ribosomes, and to facilitate protein-nucleic acid interactions. Deregulation or inhibition of polyamine synthesis broadly affects the eucaryal cell cycle, making these enzymes attractive targets for drug therapy against cancer and parasitic infections (15).

Mesophilic species of the Methanococcales group of methanogenic euryarchaea contain substantial concentrations of three common polyamines: putrescine (0.5 to 3.4 μmol/g), 1,3-diaminopropane (0.3 μmol/g), and spermidine (29 to 40 μmol/g) (27). Putrescine, the diamine precursor to spermidine, is produced either by the single enzyme ornithine decarboxylase or by the consecutive actions of arginine decarboxylase and agmatine ureohydrolase. Spermidine synthase catalyzes the transfer of a propylamine moiety to putrescine, producing spermidine, a triamine.

The most unusual feature of spermidine biosynthesis is the nature of the propylamine donor, decarboxylated S-adenosylmethionine (AdoMet) (35). AdoMet plays a central role in the metabolism of all known organisms (31). AdoMet is best known as an activated methyl donor used by cells to modify DNA, RNA, proteins, cofactors, lipids, etc. In addition, eucaryotes, some bacteria, and the crenarchaeon Sulfolobus solfataricus have been shown to contain AdoMet decarboxylase (AdoMetDC), which produces S-adenosyl(5′)-3-methylthiopropylamine, the cosubstrate used by spermidine synthase (6, 8, 14, 21, 30, 35).

Similar to the mechanisms of many other amino acid decarboxylases, in the AdoMetDC reaction a Schiff base intermediate forms with the substrate, activating the α-carbon for nonoxidative decarboxylation (9, 14, 35, 39). Whereas most amino acid decarboxylases use a pyridoxal 5′-phosphate cofactor, AdoMetDC uses a covalently attached pyruvoyl group in an analogous fashion, forming a Schiff base with AdoMet during catalysis (12, 22, 31). The active AdoMetDC enzymes are derived from a precursor protein that autocatalytically cleaves at a specific internal serine to generate two polypeptides, a conventional protein from the N terminus (denoted β) and a protein from the C-terminal region (denoted α) that has a pyruvoyl moiety as the N-terminal group (reviewed in reference 12). This cleavage proceeds through an ester intermediate and is related to a broader group of protein-processing reactions at internal serines, including intein processing (17, 23).

The amino acid-decarboxylating family of pyruvoyl group-containing enzymes includes AdoMetDC, Lactobacillus l-histidine decarboxylase, E. coli l-aspartate-1-decarboxylase, and E. coli phosphatidylserine decarboxylase (12, 22, 34). Although these enzymes have similar reaction mechanisms, their protein sequences are dissimilar and are probably not homologous. In most cases, the cofactor pyridoxal 5′-phosphate has been shown to catalyze the same decarboxylations, either in the context of an analogous enzyme or in solution, paired with a metal cation (34, 36). Crystallographic studies show unrelated topologies for the pyruvoyl group containing histidine decarboxylase, aspartate-1-decarboxylase, and human AdoMetDCs (2, 10, 20). Although eucaryal AdoMetDCs are highly similar to one another and previously identified bacterial AdoMetDCs are quite similar among themselves, these two groups are dissimilar to each other and may not be homologous. Eucaryal AdoMetDCs cleave to give ca. 31- and 8-kDa α and β chains forming an (αβ)₂ heterodimer; these enzymes are often activated by putrescine (22, 29, 30). In contrast, the only well-characterized bacterial AdoMetDC, the SpeD protein from Escherichia coli, contains 17.7- and 12.4-kDa α and β chains in an (αβ)₄ holoenzyme and requires a divalent metal ion, such as Mg²⁺, for activity (9, 14, 32, 35, 38). The only archaeal AdoMetDC previously studied, from S. solfataricus, was reported to be a 32-kDa monomeric protein that also contained a catalytically active pyruvoyl group (6); no gene or protein sequence was reported. The S. solfataricus enzyme was not activated by either metal ions or polyamines.

The discrepancy between a high concentration of spermidine in the Methanococcales without identification of an AdoMetDC found in the complete genome sequence of the hyperthermophilic Methanococcus jannaschii (5) led us to search that genome for distant homologs. In a previous identification of the archaeal AdoMet synthetase (MAT), the discovery of a distant homolog not only resolved the route of AdoMet synthesis in M. jannaschii but also advanced understanding of MAT structure and mechanism through identification of an enzyme widely diverged from the eucaryal and most bacterial forms (11). In this study we have used profiling similarity searches to identify the M. jannaschii gene that encodes a new class of AdoMetDC. Our in vitro characterization of purified, recombinantly expressed MJ0315 protein confirms that the M. jannaschii protein is a thermostable AdoMetDC and identifies the proenzyme's cleavage site. This work extends a recent report that both the MJ0315 gene and a homologous Bacillus subtilis gene (ytcF) can complement a speD mutation in E. coli. B. subtilis ytcF mutants lack spermidine, the production of which can be restored by the introduction of DNA containing MJ0315, the E. coli speD gene, as well as ytcF itself (28). Homologs of the M. jannaschii and B. subtilis enzymes have been identified in diverse archaeal and bacterial genomes. These results describe a third class of AdoMetDC, evolutionarily distinct from the well-studied eukaryotic and proteobacterial classes.

MATERIALS AND METHODS

Sequence identification.

The PSI-BLAST program (3) at the National Center for Biotechnology Information (NCBI) (http://www.ncbi.nlm.nih.gov) was used to search the nonredundant-protein database for homologs of the E. coli SpeD protein (SWISS-PROT ID, P09159). The BLOSSUM 62 amino acid substitution matrix was used for scoring with default parameters (gap existence cost, 11; per residue gap cost, 1; λ ratio, 0.87). The low complexity filter was disabled, and the E-value threshold for inclusion of sequences in each iteration was 0.001. Profile searching was iterated until convergence was reached. The profile-hidden Markov modeling program HMMER (version 2.1.1; S. Eddy, Washington University School of Medicine) produced a model from aligned bacterial and archaeal AdoMetDCs that was used to search for additional homologs in the archaeal genome sequences.

AdoMetDC activity assays.

Reagents were purchased from Sigma unless otherwise noted. AdoMet was purchased from Research Biochemicals Inc.

AdoMetDC was assayed by measuring the production of ¹⁴CO₂ from [carboxy-¹⁴C]AdoMet (Moravek Biochemicals) (14). Routine reaction mixtures contained 40 μM AdoMet (0.9 mCi/mmol) in 50 mM HEPES · KOH–50 mM KCl–0.1 mM EDTA (pH 7.4) in a volume of 200 μl. Reactions were stopped by addition of an equal volume of 4 N HCl; the CO₂ evolved was trapped on a filter paper containing saturated Ba(OH)₂ or 1 M hyamine hydroxide and quantified by scintillation counting. Assays were generally performed at 70°C.

Enzyme purification.

E. coli strain AMJAZ62, which contains the putative M. jannaschii AdoMetDC gene (MJ0315) cloned into pUC18, was purchased from the American Type Culture Collection as part of the collection from The Institute for Genome Research. The M. jannaschii DNA fragment cloned in this plasmid encodes the complete MJ0315 gene flanked by the carboxy-terminal coding regions of the adjacent MJ0314 and MJ0316 genes. This strain was grown overnight at 37°C in Luria-Bertani medium containing 50 μg of carbenicillin/ml. Attempts to increase enzyme expression by addition of 0.1 mM isopropyl-β-d-thiogalactopyranoside (IPTG) were unsuccessful. While expression levels could probably be enhanced by subcloning MJ0315 into a more sophisticated expression system, that was not required or attempted for the present purposes.

All purification steps were conducted at room temperature. In a typical preparation, 30 g (wet weight) of cells obtained from an 8-liter growth was suspended in 120 ml of 50 mM Tris-Cl–50 mM KCl–30 μM phenylmethylsulfonyl fluoride– 1 mM dithiothreitol (pH 8.0). The cells were lysed by a single pass through a French press at 15,000 lb/in². The lysate was centrifuged for 30 min at 25,000 × g to remove insoluble material.

Methylglyoxal bis-(guanylhydrazone) (MGBG) is an inhibitor of all AdoMetDCs where it has been tested, and immobilized MGBG has been used to purify these enzymes from various sources (4, 8, 14, 21). An affinity resin of MGBG-Sepharose was prepared by reacting 0.2 M MGBG with epoxy-activated Sepharose 6B (Pharmacia) for 18 h at pH 11 (a modification of the protocol described in reference 4), followed by extensive washing with 1 M KCl and then with water before use.

The cell lysate supernatant was passed through a 2.5- by 10-cm column of MGBG-Sepharose that was equilibrated with 50 mM Tris-Cl–50 mM KCl–1 mM EDTA–1 mM dithiothreitol (pH 8.0) (buffer A); the E. coli enzyme does not bind to this resin in the absence of Mg²⁺ (14). The column was washed first with buffer A until the absorbance at 280 nm of the eluate was the same as that of the buffer, followed by washing with buffer A containing 1 M KCl. After the absorbance at 280 nm of the eluate returned to that of the buffer, the column was reequilibrated with buffer A. AdoMetDC was eluted by washing with buffer A containing 3 mM MGBG. Fractions containing enzyme activity were pooled and concentrated to 7 ml by using a Centricon-10 device (Amicon). Aliquots of 2 ml were further purified by gel filtration on a 1.6- by 60-cm Superdex-75 column that was eluted with buffer A. The column was calibrated with protein standards from Bio-Rad to allow estimation of the native molecular mass.

Polyacrylamide gel electrophoresis was conducted on a Pharmacia Phast System using 8 to 25% gradient gels for native electrophoresis and “High Density” (20%) gels when separations were performed in the presence of sodium dodecyl sulfate.

Metal ion analyses were performed at the Chemical Analysis Laboratory of the University of Georgia by using the inductively coupled argon plasma emission method. The enzyme was exchanged into 25 mM Tris-Cl (pH 8.0) before analysis; analyses of the buffer were also performed for reference.

For kinetic measurements the enzyme was exchanged into 50 mM HEPES · KOH (pH 7.5)–10 mM EDTA, and reactions contained this buffer with other additions as noted.

MALDI-TOF mass spectrometry.

The purified protein (2 mg/ml) was diluted 30-fold in water, and 0.5 μl was applied to the matrix-assisted laser desorption ionization (MALDI) target along with 0.5 μl of sinapinic acid (20 mg/ml in 50% CH₃CN) and allowed to dry. Mass spectra were recorded on a Reflex III MALDI-time of flight (TOF) mass spectrometer (Bruker Instruments) operated in reflected-ion mode with a nitrogen laser (337 nm). External mass calibrations were performed using recombinant human insulin as the standard. Measured masses were matched to those of the derived amino acid sequence using the PAWS program (http://www.proteometrics.com).

Protein sequence alignments.

Amino acid sequences for AdoMetDC proteins were obtained from the nonredundant-protein database at NCBI: Aeropyrum pernix (dbj|BAA78988.1 and dbj|BAA79610.1), Aquifex aeolicus (gb|AAC06577.1), Archaeoglobus fulgidus (gb|AAB89640.1), B. subtilis (emb|CAB14861.1), E. coli (sp|P09159), M. jannaschii (sp|Q57763), Pyrococcus abyssi (emb|CAB50680.1) Pyrococcus horikoshii OT3 (dbj|BAA31119.1), S. solfataricus (emb|CAB57763.1 and emb|CAB57715.1), and Thermotoga maritima (gb|AAD35739.1). Sequence data from partial genome sequences were obtained from websites for Bacillus anthracis and Thiobacillus ferrooxidans (http://www.tigr.org), Clostridium acetobutylicum (http://www.genomecorp.com), Clostridium difficile (http://www.sanger.ac.uk), Francisella tularensis (http://www.medmicro.mds.qmw.ac.uk/ft), Nitrosomonas europaea and Prochlorococcus marinus (http://www.jgi.doe.gov), Pseudomonas aeruginosa (http://www.pseudomonas.com), Pyrococcus furiosus (http://www.genome.utah.edu), and Pyrobaculum aerophilum (http://genome.caltech.edu/pyrobaculum). Twenty-two amino acid sequences of AdoMetDC homologs were aligned using the CLUSTAL W (version 1.7.4) program (33). These alignments were manually edited using the AE2 alignment editor (T. Macke, Ribosomal Database Project). The GCG software package (version 10.0; Genetics Computer Group, Madison, Wis.) was used for additional sequence manipulations.

Phylogenetic inference.

From the alignment of 22 proteins, 117 positions were deemed to be confidently aligned. These were analyzed with previously described (37) protein maximum-parsimony methods using a heuristic search algorithm (PAUP* 4.0 beta 2; D. Swofford, Sinauer Associates, Inc.). The two copies of AdoMetDC in each crenarchaeon were constrained to be paralogs, as inferred by fastDNAml analysis of aligned nucleotide sequences (18). The 1,000 shortest trees were evaluated by maximum-likelihood criteria using the PROTML program (version 2.2) in the MOLPHY package (1) with the JTT model for amino acid substitutions. Bootstrap percentages for each node in the tree were estimated by the resampling estimated log-likelihood (RELL) method (13) using the PROTML program to compare the 1,000 most parsimonious trees. The CONSENSE program (PHYLIP [phylogeny inference package] version 3.5c, 1993; J. Felsenstein, Department of Genetics, University of Washington, Seattle) constructed a consensus tree from the RELL weightings. Phylogenetic trees were viewed and edited with the TreeView program (version 1.5.2) (19).

RESULTS AND DISCUSSION

Identification of the gene encoding AdoMetDC in M. jannaschii.

A candidate for the gene encoding M. jannaschii AdoMetDC was identified based on four iterations of PSI-BLAST (3) using the sequence of the E. coli speD-encoded proenzyme (264 amino acids) (32) to search the nonredundant-protein database at NCBI. The search identified the M. jannaschii gene MJ0315, which is annotated as encoding a 135-amino-acid hypothetical protein (5). A reciprocal search, using MJ0315 as the query sequence, identified the E. coli SpeD protein. Comparison of E. coli SpeD with the MJ0315 homolog shows that the two proteins are highly divergent: only 30 aligned residues are identical, even after the introduction of large gaps into the alignment (Fig. 1). Despite this limited similarity and the much shorter length of MJ0315 than that of E. coli SpeD, amino acids predicted to be important for proenzyme cleavage and pyruvoyl group formation are conserved (SHIXXHTYPE). This motif includes serine-110, the pyruvoyl precursor in the E. coli sequence. Furthermore, the proposed cysteine-containing signature sequence [TCGX(4–6)KAL], the only sequence conserved between eucaryotic and bacterial AdoMetDCs (22), is present C terminal to the pyruvoyl precursor serine residue in each of these sequences. A subsequent search of all predicted M. jannaschii proteins, using the aligned SpeD and MJ0315 homologs to construct a profile-hidden Markov model, revealed no additional homologs. Similar analysis using an alignment of eucaryal AdoMetDCs also failed to identify archaeal homologs. Since functionally important residues appeared to be conserved in these otherwise dissimilar sequences, we predicted that MJ0315 would encode an AdoMetDC.

FIG. 1.

Alignment of the proenzyme sequence of M. jannaschii AdoMetDC (MJ0315) with archaeal homologs from A. aeolicus, S. solfataricus, B. subtilis, C. difficile, P. aeruginosa, and SpeD homologs from E. coli, P. aeruginosa, and C. acetobutylicum. The alignment is truncated at the end of the longest archaeal sequence; SpeD sequences contain approximately 57 additional amino acids at the C terminus. Identical residues are shaded black, while highly conserved positions are shaded grey. The pyruvoyl group precursor serine and catalytic cysteine residues are indicated with an arrow.

Purification of M. jannaschii AdoMetDC from E. coli.

An E. coli strain containing MJ0315 cloned into pUC18 was assayed for AdoMetDC activity at 70°C in the presence of 10 mM EDTA. Under these conditions, the Mg²⁺-dependent, chromosomally encoded E. coli enzyme is inactive (35), while the recombinant M. jannaschii protein was hypothesized to be active, based on reported properties of the S. solfataricus enzyme (6). Substantial AdoMetDC activity was observed in the recombinant cell extract, in contrast to a control strain lacking MJ0315. Thus, purification of the recombinant M. jannaschii AdoMetDC was initiated.

MGBG is a general inhibitor of AdoMetDCs, although it has higher affinity for the eucaryotic enzymes (21). Immobilized MGBG has been used to purify eucaryotic and E. coli AdoMetDCs (4, 8, 14, 21, 24). Recombinant M. jannaschii AdoMetDC was purified from E. coli by affinity chromatography on MGBG-Sepharose and subsequent gel filtration as described in Materials and Methods. Approximately 0.2 mg of enzyme per g of cells was obtained. Native polyacrylamide gradient gel electrophoresis revealed a single band migrating at approximately 30 kDa. Denaturing sodium dodecyl sulfate-polyacrylamide gel electrophoresis showed two bands migrating at 6 to 8 kDa. The molecular mass of the native protein was estimated as 35 kDa by gel filtration chromatography. Thus, the enzyme appears to have an (αβ)₂ composition. The UV-visible absorption spectrum did not show the presence of any nonprotein chromaphores, indicating the absence of pyridoxal phosphate, as anticipated; mass spectrometry verified the absence of covalently attached cofactors (see below).

The enzyme was analyzed for the presence of tightly bound metal ions. The results showed no significant levels of metals (<0.1 equivalent per αβ heterodimer of Ca, Co, Cr, Cu, Fe, K, Mg, Mn, Mo, Ni, or Zn). Thus, the thermal stability of the enzyme does not appear to arise from a stabilizing metal ion, and a tightly bound metal ion apparently does not function in place of the Mg²⁺ required by the E. coli enzyme.

Localization of the proenzyme cleavage site by mass spectrometry.

MALDI-TOF mass spectrometry of the purified AdoMetDC revealed the presence of two polypeptides with neutral monoisotopic masses of 6,793.1 and 6,990.2 Da (Fig. 2). The larger species corresponds to that predicted for the C-terminal 61 residues (6,991.6 Da) and is consistent with polypeptide chain cleavage N terminal to a serine that is converted to a pyruvoyl group yielding a mass reduction of 17 Da due to loss of NH₃. The smaller mass corresponds to the N-terminal region beginning with a methionine (6,794.5 Da). Comparison of the mass spectrometry data to values predicted from the protein sequence shows consistency with values for a 124-amino-acid MJ0315 proenzyme, in which the pyruvate moiety is formed from serine residue 64 of the proenzyme. Therefore the recombinant enzyme is 11 amino acids shorter than that predicted in the original M. jannaschii genome annotation (5) and recent analysis (28). Whereas previously published annotations described TTG as the gene's initiator codon, mass spectral data of the recombinant protein implicate an ATG initiator. An unusually long canonical Shine-Dalgarno sequence (underlined) precedes this ATG initiator (in italics): TTTGGAGGTGAAAGCATG. No Shine-Dalgarno sequence was observed upstream of the originally proposed TTG initiator. This evidence, combined with sequence data deduced for other archaeal-type AdoMetDCs, suggests that this ATG is the relevant translation start site and that a protein of the same length is expressed in M. jannaschii and E. coli. We denote the N-terminal and C-terminal polypeptides as β (63 residues) and α (61 residues) to be consistent with the nomenclature adopted for other pyruvoyl group-containing decarboxylases. Our purification of cleaved M. jannaschii AdoMetDC recombinantly expressed in E. coli suggests an autocatalytic cleavage mechanism; this model is consistent with recent evidence for partial cleavage after in vitro transcription and translation (28).

FIG. 2.

Mass spectrum of M. jannaschii AdoMetDC analyzed by MALDI-TOF. Two subunits are formed by autocatalytic cleavage of the MJ0315 proenzyme: the β subunit (63 amino acids) has a neutral isotopic mass of 6,793.1 Da, and the pyruvoyl group-containing α subunit (61 amino acids) has a mass of 6,990.2 Da.

Characterization of the purified enzyme.

M. jannaschii AdoMetDC retained full activity after 3 h at 70°C. Thus the enzyme is substantially more thermostable than its substrate. The enzyme activity at a subsaturating AdoMet concentration of 0.04 mM was not altered by the addition of 10 mM MgCl₂, 10 mM CaCl₂, 10 mM EDTA, 10 mM putrescine, or 50 mM KCl, showing that these species do not alter either the K_m or V_max. This is analogous to the properties reported for the enzyme from S. solfataricus (6). These features distinguish the enzyme from both the AdoMetDC from E. coli, which requires divalent metal ions for activity, and the polyamine-activated eucaryotic enzymes.

The kinetic properties of AdoMetDC from M. jannaschii are compared to those from other sources in Table 1. At a comparable temperature, the K_m values of the enzymes are similar; however, the K_m for the M. jannaschii enzyme increases ca. fivefold at the near-physiological temperature of 70°C. The V_max value of the M. jannaschii enzyme's near-physiological temperatures is ca. 10-fold lower than those reported for the E. coli and rat enzymes at their physiological temperatures; the even lower specific activity of the S. solfataricus preparation may reflect the difficulty in purifying this low-abundance protein from its native source. MGBG competitively inhibits the M. jannaschii enzyme with affinity comparable to that for the E. coli enzyme; both the E. coli and M. jannaschii enzymes have lower affinity for MGBG than do the mammalian enzymes (24).

TABLE 1.

Comparison of kinetic properties of purified AdoMetDCs from various organisms

Organism	Source or reference	Temp (°C)	Vmax (μmol/ min/mg)	Km (AdoMet) (mM)	Activator	Ki (MGBG) (μM)
M. jannaschii	This work	70	0.69	0.60	None	NDa
		25	0.023	0.12	None	14
S. solfataricus	6	70	0.012	0.11	None	33
E. coli	14	37	6.8	0.06	Mg2+	20
Rat	39	37	3.6	0.1	Putrescine	<1b

^aND, not determined. 

^bSee reference 24. 

Phylogeny of archaeal AdoMetDC.

The existence of a new class of AdoMetDCs potentially explains the lack of identifiable speD genes in the complete genome sequences of some bacteria known to produce spermidine (7, 28). Therefore, MJ0315 and E. coli SpeD sequences were used to identify additional homologs in the sequence databases. An archaeal-type. AdoMetDC was identified in each complete archaeal genome sequence, except for Methanobacterium thermoautotrophicum (Fig. 1 and 3). This result is consistent with data showing no detectable spermidine in that methanogen (27). Each crenarchaeal genome, including that for S. solfataricus, encodes two AdoMetDC paralogs. Many diverse bacteria also have homologs of this archaeal-type enzyme, including A. aeolicus, T. maritima, B. subtilis, P. marinus, T. ferrooxidans, and N. europaea. This surprisingly broad phylogenetic distribution indicates that the canonical bacterial-type AdoMetDC, defined by E. coli SpeD, is a derived form of the more widespread enzyme described here. The short lengths of these proteins and their low degree of similarity complicate phylogenetic analysis of the AdoMetDCs. As the number of sequenced AdoMetDC genes increases, we expect that the quality of the inferred phylogeny will improve substantially. It may also become possible to resolve the relationship between the SpeD and archaeal types of AdoMetDCs. In the interim, a consensus phylogeny (Fig. 3) defines several groups of archaeal AdoMetDCs. Most bacterial sequences group together, including two paralogs present in B. anthracis. Crenarchaeal sequences, with two paralogs in each genome, are broadly related to euryarchaeal M. jannaschii and A. fulgidus sequences in addition to several bacterial sequences. Sequences from the euryarchaeal Pyrococcus spp. tenuously group with bacterial members.

FIG. 3.

Consensus phylogenetic tree of archaeal-type AdoMetDC sequences inferred by protein maximum-likelihood analysis. This tree is arbitrarily rooted. Bootstrap probabilities for each node are estimated by the RELL method (13) and reported where substantial. The scale bar represents 10 substitutions per 100 amino acid positions.

This phylogeny does not unambiguously resolve the history of all archaeal AdoMetDC genes. Nevertheless, it is not inconsistent with a model of vertical inheritance in which an ancestral sequence was inherited by both archaea and bacteria. The SpeD-type protein, identified in E. coli, Salmonella spp., Klebsiella pneumoniae, P. aeruginosa, Xanthomonas campestris, and Yersinia pestis, probably evolved within the gamma proteobacterial lineage through several insertions at the C-terminal ends of both α and β subunits (Fig. 1).

This evolutionary history is complicated by probable instances of horizontal gene transfer. P. aeruginosa contains both SpeD and archaeal proteins. While C. difficile has only an archaeal AdoMetDC, C. acetobutylicum has only a speD gene instead. The genome sequence of the proteobacterium Shewanella putrefaciens encodes a eucaryal AdoMetDC. These observations suggest that the AdoMetDC gene is modular: it functions independently of its cellular environment and is easily transferable. Its proenzyme structure, in which both subunits are encoded and spliced by a single gene, probably facilitated such transfer events.

Functional implications of AdoMetDC diversity.

This M. jannaschii enzyme represents a new class of AdoMetDCs that is the most evolutionarily divergent and widespread form among the three known classes. Despite the substantial divergence between the two types of prokaryotic AdoMetDC, α subunits of both the E. coli SpeD and archaeal types of enzymes share several highly conserved motifs. The first motif (SHIXXHTYPE) includes the pyruvoyl precursor serine and splice site (4, 32, 38); no functions have been proposed for the other residues conserved within this motif. MJ0315 is only the second prokaryotic AdoMetDC for which the site of pyruvoyl formation has been experimentally demonstrated. The second motif (TCG) contains a cysteine that has been proposed to act as a nucleophile in catalysis and inactivating side reactions (9, 39, 40). Other residues proposed to be important for splicing or catalysis in the eucaryal enzyme are not conserved. A lysine residue that precedes the first motif in E. coli SpeD and an LKAL peptide that follows the second motif in both SpeD and eucaryal types of enzymes are unrecognized in most archaeal-type enzymes. Assuming a common catalytic mechanism and active site structure in SpeD and archaeal enzymes, the alignment of these two divergent classes of AdoMetDC focuses attention on a handful of key residues for future site-directed mutagenesis experiments.

The recently determined three-dimensional structure of the human AdoMetDC demonstrates that the active site is located far from the interface between the two αβ heterodimers and that it contains residues from the both the α and β subunits (10). The topology of each αβ unit contains an internal structural duplication in which residues 4 to 164 and 172 to 329 form two separate domains that have the same topology; these domains associate in a novel fold. However, there is no detectable sequence homology between the two structural units (10). The sizes of these domains are intriguingly comparable to the size of a single αβ unit of the M. jannaschii enzyme. Prediction of the secondary structure of the M. jannaschii enzyme using the PredictProtein server (25) indicates that the sequence appears to have ca. 30% each of helical and sheet components but appears to be a domain rather than a globular structure. Even though primary sequence alignments of eucaryal- and archaeal or SpeD types of AdoMetDCs do not demonstrate homology, these observations suggest that the eucaryal enzymes could have evolved from duplication and fusion of a gene that encoded an archaeal AdoMetDC. Since the M. jannaschii enzyme is an (αβ)₂ protein, it is possible that the archaeal type of enzyme active site is formed from the interaction of two αβ heterodimers in a topology resembling that of the eucaryal enzyme. It has not been determined if the AdoMetDC paralogs in the crenarchaea and B. anthracis form a heterodimer or act as isozymes. Determination of the three-dimensional structural relationships of these enzymes is of substantial interest.

The M. jannaschii AdoMetDC is the first pyruvoyl group-containing enzyme identified in a member of the Euryarchaeota. As observed for the archaeal MAT (11), the archaeal-type AdoMetDC is highly divergent from its previously recognized homologs. In contrast, spermidine synthases are highly conserved enzymes, easily recognized in all three domains of life. Causes of this disparity in evolutionary rates are not obvious, although horizontal transfer of AdoMetDC genes may scramble phylogenies and hasten the evolutionary tempo. The lower affinity of SpeD and archaeal AdoMetDCs for some inhibitors suggests that horizontal transfer is a potential antibiotic resistance mechanism that could frustrate the use of AdoMetDC inhibitors as antiprotozoan drugs.

The ubiquity of pyruvoyl-dependent AdoMetDCs prompts an important evolutionary question about amino acid decarboxylases. Why have both pyridoxal phosphate (PLP)-dependent and pyruvoyl-dependent amino acid decarboxylases persisted throughout evolution? Compared to PLP catalysis, pyruvoyl electrophilic catalysis may be a more primitive mechanism that requires no exogenous cofactor. Both pyruvoyl- and pyridoxal 5′-phosphate-dependent decarboxylases have similar catalytic mechanisms and have been identified in phylogenetically diverse organisms. These two cofactors sometimes overlap in specificity: two classes of histidine decarboxylase have been identified; one uses pyruvate and the other PLP. Yet despite observed PLP catalysis of AdoMet decarboxylation in vitro (36), no PLP-dependent AdoMetDC has been identified. Although the persistence of both cofactors is currently inexplicable, future studies of AdoMetDC catalytic mechanisms and side reactions may clarify their comparative advantages and limitations.