Paramecium bursaria Chlorella Virus-1 Encodes an Unusual Arginine Decarboxylase That Is a Close Homolog of Eukaryotic Ornithine Decarboxylases*

Paramecium bursaria chlorella virus (PBCV-1) is a large double-stranded DNA virus that infects chlorella-like green algae. The virus encodes a homolog of eukaryotic ornithine decarboxylase (ODC) that was previously demonstrated to be capable of decarboxylating l-ornithine. However, the active site of this enzyme contains a key amino acid substitution (Glu for Asp) of a residue that interacts with the δ-amino group of ornithine analogs in the x-ray structures of ODC. To determine whether this active-site change affects substrate specificity, kinetic analysis of the PBCV-1 decarboxylase (PBCV-1 DC) on three basic amino acids was undertaken. The kcat/Km for l-arginine is 550-fold higher than for either l-ornithine or l-lysine, which were decarboxylated with similar efficiency. In addition, α-difluoromethylarginine was a more potent inhibitor of the enzyme than α-difluoromethylornithine. Mass spectrometric analysis demonstrated that inactivation was consistent with the formation of a covalent adduct at Cys347. These data demonstrate that PBCV-1 DC should be reclassified as an arginine decarboxylase. The eukaryotic ODCs, as well as PBCV-1 DC, are only distantly related to the bacterial and plant arginine decarboxylases from their common β/α-fold class; thus, the finding that PBCV-1 DC prefers l-arginine to l-ornithine was unexpected based on evolutionary analysis. Mutational analysis was carried out to determine whether the Asp-to-Glu substitution at position 296 (position 332 in Trypanosoma brucei ODC) conferred the change in substrate specificity. This residue was found to be an important determinant of substrate binding for both l-arginine and l-ornithine, but it is not sufficient to encode the change in substrate preference.

ODC is a pyridoxal 5Ј-phosphate (PLP)-dependent enzyme that catalyzes the initial step in the polyamine biosynthetic pathway, the decarboxylation of L-ornithine to produce putrescine (4). The polyamines putrescine, spermidine, and spermine are ubiquitous compounds that are essential for cell growth and differentiation in many organisms. They have demonstrated roles in cell cycle, apoptosis, cancer, embryonic development, immune system functions, and neurochemistry (5)(6)(7)(8)(9). Overexpression of polyamine biosynthetic enzymes leads to mammalian cell transformation (10) or to growth perturbations in plants (11), whereas knockout of the polyamine biosynthetic enzymes causes polyamine auxotrophy (e.g. yeast (12), Trypanosoma brucei (13), and Leishmania donovani (14,15)) and is lethal at early embryonic stages in mice (16,17). Inhibitors of polyamine biosynthetic enzymes have anti-proliferative effects and have been utilized as anticancer and anti-parasite agents (18,19).
Two distinct structural classes have evolved within the PLPdependent enzymes to catalyze the decarboxylation of basic amino acids (20,21). One family contains only bacterial decarboxylases, and these are structural homologs of aspartate aminotransferase. These include the constitutive and inducible bacterial ODCs and lysine decarboxylases as well as the biosynthetic (constitutive) arginine decarboxylase (ADC). An xray structure of a bacterial ODC from this class has been solved (22). The second family contains enzymes from both eukaryotic and prokaryotic organisms that are structurally related to alanine racemase. These include eukaryotic ODCs, plant ADC, and a number of bacterial enzymes that encompass a wide range of substrate specificities (arginine, ornithine, lysine, diaminopimelate, and carboxynorspermidine decarboxylases) (20). The x-ray structures of several eukaryotic ODCs (mouse, T. brucei, and human) (23)(24)(25)(26)(27) and of bacterial diaminopime-late decarboxylase have been solved (28). The N-terminal domain forms a ␤/␣-barrel, and the C-terminal domain is folded into a ␤-barrel structure. The active sites are formed at the dimer interface between the N-terminal domain of one subunit and the C-terminal domain of the other (Fig. 1A). Although all enzymes in the family share a number of essential active-site residues (24, 29 -32), eukaryotic ODCs are very distantly related to ADC and diaminopimelate decarboxylase and share only ϳ15% overall sequence identity based on pairwise comparisons. However, a bacterial enzyme from this fold type has been described that has activity on both lysine and ornithine (Lys/Orn decarboxylase (DC)) and that shares extensive sequence similarity (ϳ35%) with eukaryotic ODCs (33).
The decarboxylase identified in chlorella virus (PBCV-1 DC) contains nearly 40% sequence identity to the family of eukaryotic ODCs (3). Phylogenetic analysis indicates that PBCV-1 DC branches with the eukaryotic ODCs and with the bacterial Lys/OrnDCs (Fig. 2). Consistent with this grouping, the recombinant PBCV-1 enzyme was characterized and found to have activity on L-ornithine as a substrate. However, it was inhibited more strongly by ␣-difluoromethylarginine (DFMA) than by ␣-difluoromethylornithine (DFMO), suggesting that it might have unusual substrate specificity.
Alignment of the primary amino acid sequence of this enzyme with the x-ray structures for mammalian and trypano-some ODCs reveals a key amino acid substitution in the substrate-binding pocket that is predicted to alter the substrate specificity ( Fig. 1). The structures of T. brucei ODC (TbODC) in complex with several substrate and product analogs demonstrate that the ␦-nitrogen of L-ornithine forms salt bridge interactions with Asp 361 from the C-terminal domain and with Asp 332 from the N-terminal domain across the subunit boundary ( Fig. 1A) (24,26,27). These two residues are conserved in all functional ODCs that have been described.
To investigate the substrate preference of PBCV-1 DC, we performed steady-state kinetic analysis on a range of basic amino acids. We show here that, although PBCV-1 DC contains detectable activity with L-ornithine, its activity with L-arginine is significantly higher; therefore, this enzyme should be reclassified as an ADC. The fact that it is closer in amino acid sequence to the enzymes with specificity for L-ornithine and L-lysine than to the ADCs from bacteria and plants suggests that PBCV-1 DC represents a new activity within the clade of the ornithine-specific enzymes.

EXPERIMENTAL PROCEDURES
Materials-Amino acids, polyamines, and a carbon dioxide kit were purchased from Sigma. The AccQ-Fluor reagent kit for labeling amino acids was purchased from Waters (Milford, MA). L-[1- 14  Modeling and Sequence Analysis-The program InsightII (Accelrys Inc., San Diego, CA) was used to display the active site of TbODC for the wild-type enzyme in complex with putrescine (Protein Data Bank code 1f3t) (27). The distance tree was built using the PHYLIP package (34) from the alignment of the displayed sequences constructed with the PCMA program (35). PSI-BLAST was used to identify eukaryotic ODC homologs in the GenBank TM Data Bank (Version 2.2.8) and to generate a local alignment. A total of 53 eukaryotic sequences that had published sequence data were included. (Hypothetical proteins were excluded from the analysis.) Protein Expression and Purification-PBCV-1 DC was produced in Escherichia coli from the expression plasmid for PBCV-1 DC, termed pODCTM9, which was described previously (3). This vector directs the expression of N-terminally His-tagged PBCV-1 DC from a pET-15b plasmid containing the cDNA corresponding to open reading frame A207R present in the PBCV-1 genome. Two sources of PBCV-1 were used in these studies. 1) A protein with the wild-type sequence was used in the 14 CO 2 release assay (see Table I) and for DFMO and DFMA inactivation studies (see Fig. 4, A and B); and 2) a protein containing a Thr-to-Ala substitution at position 142 was used for the spectrophotometric analysis and as the background for the E296D mutant (see Figs. 3 and 5 and Table II). This change was generated during cloning, but has no effect on the activity or substrate specificity of the enzyme. PBCV-1 DC, PBCV-1 DC E296D, TbODC, and TbODC D332E were expressed and purified as described previously for TbODC using Ni 2ϩagarose column chromatography and Sephadex 200 (36).
Radiolabel Release-based Enzymatic Assays-ODC and ADC activities were determined by measuring the release of 14 CO 2 from the enzymatic decarboxylation of L-[1-14 C]ornithine or L-[U-14 C]arginine (37). The 14 CO 2 released was trapped in a center well containing Hyamine hydroxide and counted by liquid scintillation spectrometry as described previously (37). Kinetic constants were measured under two conditions using either 25 mM Tris-HCl (pH 8.2) at 37°C or 50 mM CAPSO (pH 9) at 42°C. All reaction mixtures contained 0.04 mM PLP and 2.5 mM dithiothreitol. The concentration of L-ornithine was varied from 5 to 250 mM and that of L-arginine from 0.05 to 5 mM. Protein was determined by the Bradford assay using bovine serum albumin as a standard.
The molar ratios of CO 2 formed from the labeled substrate were confirmed to be equivalent to the product by analyzing the products by reverse-phase HPLC using a standard system for separation of polyamines (38) with a Canberra Packard A140 Radiomatic detector as described previously (39). L-[2,3-3 H]Ornithine was used to detect the formation of putrescine, and L-[U-14 C]arginine was used to detect the formation of agmatine. Authentic [U-14 C]agmatine and [2,3-3 H]putrescine were used as markers.
HPLC Analysis of Reaction Products-The products of an enzymatic reaction with L-arginine and wild-type PBCV-1 DC and TbODC D332E were analyzed by HPLC using an AccQ-Tag kit (Waters) in 5% sodium tetraborate and labeling reagent (6-aminoquinolyl-n-hydroxysuccinimidyl in acetonitrile) as described previously (24). PBCV-1 DC or TbODC D332E was incubated with L-arginine (20 mM for wild-type PBCV-1 DC and 170 mM for the TbODC mutant) in 15 mM KPO 4 , 1 mM dithiothreitol, and 0.15 mM PLP at 37°C for 2.5, 5.0, and 10.0 min. Enzyme concentrations of 475 and 950 nM for PBCV-1 DC and 100 and 200 M for TbODC D332E were used in the reactions. Enzymatic reactions were terminated by the addition of trichloroacetic acid to a final concentration of 6%. Labeled samples (5 l) were injected onto an AccQ-Tag column using the buffers and gradient described previously (24). The column was calibrated with known amounts of the following derivatized reagents: agmatine (retention time ϭ 27.8 min), cadaverine (40.1 min), L-ornithine (34.8 min), putrescine (44.5 min), and L-arginine (21.1 min).
Inactivation of PBCV-1 DC with DFMO or DFMA-The His-tagged PBCV-1 DC protein in 50 mM NaH 2 PO 4 (pH 8), 300 mM NaCl, 250 mM imidazole, 0.04 mM PLP, 2.5 mM dithiothreitol, and 0.5 mg/ml bovine serum albumin (3) was dialyzed to remove dithiothreitol and imidazole before passing over a Talon metal affinity column (Clontech, Palo Alto, CA) to remove bovine serum albumin. After elution and further dialysis, an aliquot of PBCV-1 DC protein (104 g in 50 mM NaH 2 PO 4 (pH 8), 0.04 mM PLP, and 0.5 mM dithiothreitol) was incubated in the absence or presence of 10 mM DFMO or 1 mM DFMA for 2 h at 37°C in a total volume of 250 l. Aliquots (5 l) of each reaction were removed at the times indicated and diluted accordingly to monitor inactivation of the protein.
Preparation of Samples of PBCV-1 DC Protein for MALDI-TOF Analysis-MALDI analyses were performed at the Mass Spectrometry/Proteomics Core Facility of the Pennsylvania State College of Medicine. Reactions containing 4.2 g (ϳ100 nmol) of untreated or DFMO-or DFMA-inactivated PBCV-1 DC protein in a total volume of 100 l were subjected to tryptic or endoproteinase Glu-C digestion using a protease/ protein ratio over the range 1:84 to 1:21 (w/w) as indicated. Samples were digested with modified trypsin in 50 mM NH 4 HCO 3 (pH 8) containing 10% (v/v) acetonitrile for 16 h, stopped by the addition of 4 l of glacial acetic acid, and stored frozen until analyzed by MALDI-TOF. Samples were digested with Glu-C in 25 mM NH 4 HCO 3 (pH 7.8) containing 10% (v/v) acetonitrile for 2-16 h at 25°C, stopped by the addition of glacial acetic acid, and stored frozen until analysis by MALDI-TOF using an Applied Biosystems 4700 Proteomics Analyzer. Digested samples were evaporated and resuspended three times in 200 l of deionized water to remove volatile digestion buffers (NH 4 HCO 3 ), which can interfere with subsequent binding to strong cation exchange resins. The final resuspension was evaporated to ϳ10 l, and then 0.11 volume of 1.0% trifluoroacetic acid was added to bring the final trifluoroacetic acid concentration to 0.1%. Using a BioHit Multipipettor, ZipTip SCX tips were equilibrated with 3 ϫ 10 l of 0.1% trifluoroacetic acid, and then the sample aliquots were pipetted up and down across the resin 15 times to bind and concentrate peptides. The bound peptides were washed with 5 ϫ 10 l of 0.1% trifluoroacetic acid to remove salts, followed by elution of the bound peptides by carefully pipetting 2 l of freshly prepared 5% NH 4 OH and 30% MeOH up and down four to five times. The droplet containing the eluted peptides was deposited onto a polished stainless steel MALDI target plate (Applied Biosystems) and allowed to dry. After drying, each spot was overlaid with 0.6 l of freshly prepared matrix (5 mg/ml recrystallized ␣-cyanohydroxycinnamic acid, 2 mg/ml NH 4 H 2 PO 4 , 50% acetonitrile, and 0.1% trifluoroacetic acid).

Analysis of PBCV-1 DC Substrate Preference and Activity-
Decarboxylation of L-ornithine or L-arginine by PBCV-1 DC was measured by following 14 CO 2 release from L-[1-14 C]ornithine or L-[U-14 C]arginine (Table I). In studies carried out in Tris-HCl (pH 8.2) at 37°C, the protein was highly active on L-arginine, with a K m of 0.45 mM and a k cat of 15 s Ϫ1 . It also had detectable activity on L-ornithine; however, although the k cat was only slightly lower, the K m was ϳ400 times higher (K m ϭ 180 mM) (Table I), giving an overall difference in substrate preference of 550-fold for k cat /K m . These results contrast with a previous report of a K m of 0.78 mM for L-ornithine (3). The previously reported assays were conducted in CAPSO (pH 9) at 42°C. The assays were therefore repeated using these conditions, and the ability to decarboxylate L-ornithine was slightly improved (K m ϭ 46 mM) (Table I). However, even under these conditions, the protein still preferred L-arginine as a substrate, with a K m 2 orders of magnitude lower than for L-ornithine. The products of the reaction were identified by HPLC (38,39) after the protein had acted upon either L-[U-14 C]arginine or L-[2,3-3 H]ornithine. As expected, [ 14 C]agmatine was found in stoichiometric amounts with 14 CO 2 when L-[U-14 C]arginine was decarboxylated, and [ 3 H]putrescine was formed in equivalent amounts when L-[2,3-3 H]ornithine was the substrate.
The substrate preference of PBCV-1 DC was also characterized using a coupled NADH spectrophotometric assay. Steadystate decarboxylation was measured for L-arginine, L-ornithine, and L-lysine over a wide range of concentrations (Fig. 3). The relative substrate preferences for L-arginine over L-ornithine were similar to those observed by the 14 CO 2 release assay; however, the k cat /K m for both substrates was 8-fold lower under the conditions of the spectrophotometric assay (Table II). L-Lysine was also a substrate for the reaction and was decarboxylated with similar efficiency compared with L-ornithine (Fig. 3). This result is in contrast to the ODCs from both mouse and T. brucei, for which the k cat /K m for L-lysine is 300-fold lower than for L-ornithine (31,37). However, the similar catalytic efficiency of PBCV-1 DC for L-ornithine and L-lysine is reminiscent of the bacterial Lys/OrnDC from Selenomonas ruminantium (33).
Inhibition of PBCV-1 DC with DFMO and DFMA-A previ-ous report indicated that DFMA is a more potent inhibitor of PBCV-1 DC compared with DFMO (3). To confirm this finding, we undertook similar inhibition experiments with DFMA and DFMO (Fig. 4A). PBCV-1 DC was irreversibly inactivated by incubation with either DFMO or DFMA. Consistent with previous reports, DFMA inactivated the enzyme more rapidly than DFMO; after a 30-min incubation with 1 mM DFMA, only 6% of the activity remained, whereas 10 mM DFMO reduced the activity to only 26% during the same incubation time. By 1 h, only 0.8% of the activity remained with DFMA, yet 17% of the activity was still present with DFMO. Mouse ODC is inactivated by DFMO, with the formation of a covalent S- ((2-(1-pyrroline))methyl)cysteine adduct at Cys 360  (major product, 90%) and a minor product (ϳ10%) at Lys 69 (40,41). Although previous studies have shown that DFMA is an irreversible inhibitor of bacterial and plant ADCs (42), the adduct formation site has not been identified. However, based on the inactivation of mouse ODC, the likely sites of interaction in native PBCV-1 DC are Cys 347 and Lys 71 , which correspond to Cys 360 and Lys 69 in mouse ODC.
Endoproteinase Glu-C digestions of untreated PBCV-1 DC or of PBCV-1 DC inactivated by DFMO or DFMA were analyzed by linear MALDI-TOF mass spectrometry in positive ion mode using PBCV-1 DC and Glu-C self-digested peaks for internal calibration of the spectra. The spectra of the inactivated enzyme had new peptide fragments with masses corresponding to adducts on the Lys 321 -Glu 361 peptide (KSVPTPQLLRDVPD-DEEYVPSVLYGCTCDGVDVINHNVALPE) containing Cys 347 . The presence of this peptide in the Glu-C digest and the absence of shorter fragments cleaved at the internal DDEEY sequence are probably due to the digestions being carried out in NH 4 CO 3 buffer, which increases the specificity of Glu-C but restricts activity and the known inactivity of Glu-C to cut at clusters of acidic residues (43). As shown in Fig. 4B, these masses (m/z 4681.0 and 4740.8) were not observed in the control digests, whereas all three digests contained a peak slightly smaller than the theoretical value (4599.2 average mass) of the equivalent unmodified fragment. (It should be noted that, within the limited resolution of these linear spectra, the potential contributions of the unmodified peptide and either of two similar mass Glu-C self-digested peptides to the observed peak cannot be separated definitively.) The new peptide that appeared after DFMO inactivation (m/z 4681.0) is 81.8 Da heavier than the unmodified Lys 321 -Glu 361 peptide, which, within experimental error, agrees with the theoretical 81.1-Da difference previously observed for the S- ((2-(1-pyrroline))methyl)cysteine adduct formed by DFMO at Cys 360 of mouse ODC (41). In the digest of the protein inactivated by DFMA, the new peptide observed at m/z 4740.8 is 141.6 Da larger than the unmodified Lys 321 -Glu 361 peptide. This value is within experimental error of the theoretical adduct mass of 140.2 Da, which would occur with the analogous linear DFMA adduct to a Cys residue in the peptide. This result is consistent with DFMA inactivation occurring similarly to DFMO inactivation of ODC, with decarboxylation of the DFMA-PLP Schiff base leading to the loss of a fluoride ion to generate a reactive electrophilic conjugated imine that binds covalently to Cys 347 . Subsequent elimination of the second fluoride anion followed by an internal trans-aldimination reaction with Lys 71 would generate the adduct. This adduct cannot cyclize in the same way as the adduct derived from DFMO because of the replacement of the terminal amino group with a guanidino group. The Lys 321 -Glu 361 peptide does contain another Cys residue at position 345. Because we were unable to produce interpretable tandem mass spectrometric spectra of the modified peptides, addition to this site cannot be ruled out; but, based on the existing data on the structure and inactivation of mammalian and trypanosome ODC by DFMO (24 -26, 41), Cys 347 is the likely site of attachment.
Assuming that the peaks in the DFMO-and DFMA-inactivated protein spectra at m/z 4597.9 and 4598.6, respectively, do represent unmodified peptide, then not all of the PBCV-1 DC was modified at this site. Although MALDI analysis is not quantitative, the relative sizes of the modified and parent peaks are similar, which suggests that only about half the protein was altered at Cys 347 , in contrast to the almost total loss of enzyme activity observed with DFMA (Ͼ99% inhibition) and with DFMO (Ͼ90% inhibition). One explanation for this discrepancy is that interaction with another site such as Lys 71 occurs to a larger extent compared with inactivation of mouse ODC with DFMO. However, we were unable to identify by mass spectrometry a putative DFMO or DFMA adduct to peptides containing Lys 71 after digestion by Glu-C or trypsin (where an adduct at Lys 71 would prevent tryptic cleavage at that site). Alternative explanations could be (a) that only one of the two PBCV-1 DC subunits making up the homodimer needs to be modified to cause loss of catalytic activity; (b) that analytical work-up resulted in the preferential loss of some of the modified peptide; (c) that, although we did not find any other peptides that were significantly different between control and DFMO-or DFMA-inactivated PBCV-1 DC in the analysis of either Glu-C or trypsin digests, modification of an additional residue at the active site occurs (the sequence coverage was ϳ50% in the Glu-C digests, so modifications of the unrepresented proteolytic fragments would not have been observed); and (d) that some proportion of the recombinant protein extract may be enzymatically inactive prior to the start of the reaction and thus unable to react with DFMO or DFMA. It is also possible that a significant portion of the ''unmodified'' peak represents Glu-C self-digested fragments.
Sequence Analysis of PBCV-1 DC Active-site Changes-The amino acid sequence of PBCV-1 DC was aligned with the sequences of ODC family members, including those for which structural information is available. This allowed amino acid changes to be mapped onto the three-dimensional structure of the active site complexed with the putrescine product (Fig. 1). The only residues within 4.5 Å of the putrescine side chain that are variable between mouse and T. brucei ODCs and PBCV-1 DC are Tyr 331 (replaced by Phe) and Asp 332 (replaced by Glu). When a wider range of ODCs from various species are considered, the residue at position 331 may be either Tyr or Phe; thus, the difference in this position for PBCV-1 DC is a previously observed variation.
Asp 332 is highly conserved in the ODC family and plays an important role in substrate binding and catalysis (Fig. 1) (44). The equivalent position in PBCV-1 DC is residue 296, which is Glu, suggesting that this substitution is a key determinant in the change in substrate specificity observed for PBCV-1 DC (Fig. 1B). To determine whether this sequence substitution occurs in any other ODC, analysis of the GenBank TM Data Bank (Version 2.2.8) was conducted. Asp 332 is invariant in the eukaryotic ODCs examined in the Data Bank; the analysis included 53 sequences that were linked to publications (hypothetical proteins were not included in the analysis). The prokaryotic Lys/OrnDC identified in S. ruminantium also contains an Asp residue at this position. Several sequences for antizyme a Data for these enzymes were also collected by HPLC analysis of agmatine formation using AccQ-Tag labeling as described under ''Experimental Procedures,'' and similar results were obtained.
b Data were taken from Refs. 30 and 32. inhibitor (human, NP_680479; mouse, NP_061215; and rat, NP_072107), an inactive ODC homolog that regulates ODC activity (45), contain the D332E substitution. However, the analysis suggests that D332E substitution in PBCV-1 DC is unique among the currently known eukaryotic ODC-like sequences that encode active enzymes. Thus, because PBCV-1 DC is the only member of the eukaryotic ODC branch that has a preference for L-arginine and the only member containing the Asp-to-Glu substitution, it appears likely that the D296E alteration is one of the structural determinants responsible for the change in substrate specificity. Kinetic Analysis of TbODC and PBCV-1 DC Mutants-To investigate the impact of the active-site difference at position 332 on substrate specificity differences between the ODCs and PBCV-1 DC, mutational analysis was undertaken. Asp 332 was mutated to Glu in TbODC (TbODC D332E), and the equivalent substitution was conducted for PBCV-1 DC (PBCV-1 DC E296D). For the wild-type enzymes, the substrate selectivity of TbODC for L-ornithine was more stringent than that of PBCV-1 DC for L-arginine; and thus, the conversion would potentially be more difficult. Spectrophotometric and HPLC analyses of the substrate specificity of both wild-type and mutant enzymes demonstrated that these mutations decreased the activity on both L-arginine and L-ornithine (Table II). Significant increases in the K m for the preferred substrate were observed with both TbODC D332E and PBCV-1 DC E296D compared with the wild-type enzymes. In addition, with PBCV-1 DC, the k cat values decreased for both L-ornithine and L-arginine. The k cat effects were not observed in the background of the T. brucei sequence. Although the mutant enzymes were both much less active than the wild-type enzymes, a small change (20-fold) in the k cat /K m ratio upon comparison of L-arginine and L-ornithine occurred in both backgrounds, which relaxed the substrate preference for the alternative amino acid (Fig. 5). The data demonstrate that the residue at position 332 is an important determinant for substrate binding in both TbODC and PBCV-1 DC. However, this substitution alone is insufficient to produce the observed substrate specificity change, and clearly additional amino acid changes are required.

DISCUSSION
The PLP-dependent decarboxylases that belong to the ␤/␣barrel fold are from both bacterial and eukaryotic origins, and they include enzymes capable of decarboxylating a range of basic amino acid substrates (20). Within this fold class, the eukaryotic ODCs and a group of bacterial enzymes with dual specificity for L-ornithine and L-lysine (e.g. S. ruminantium Lys/OrnDC) (33) share high sequence similarity, whereas the enzymes with specificity for L-arginine and other basic amino  acids share very low sequence similarity with the ornithinespecific enzymes (Fig. 2). Thus, PBCV-1 DC is an anomaly because it shares greater sequence identity with the ODCs, whereas it strongly prefers L-arginine as a substrate. The preference for L-arginine over L-ornithine and L-lysine is reflected in the 550-fold higher k cat /K m observed for L-arginine at all pH values and temperatures studied. DFMA is a more potent inhibitor of PBCV-1 DC than DFMO, although both compounds form adducts at the same site. Thus, our data clearly indicate that PBCV-1 DC should be reclassified as an ADC.
The x-ray structures of ODC provide insight into amino acid variations that are likely to be important for changing substrate specificity. The only residue in direct contact with the substrate that differs between ODCs and PBCV-1 DC is Asp 332 , which is Glu in PBCV-1 DC. This residue interacts directly with substrate, and site-directed mutagenesis established that it is an important residue for substrate binding in both ODCs and PBCV-1 DC. However, the residue at position 332 is not sufficient on its own to determine substrate preference. Swapping the residue at position 332 between Asp and Glu did not significantly increase the relative activity of either TbODC or PBCV-1 DC on the less preferred substrate. These results demonstrate that amino acid changes at residues distant from the active site are also necessary to switch the substrate preference. A number of studies have shown that amino acid residues that are distant from the active site and that do not contact ligand directly are often important in the change of function between homologous proteins (e.g. Refs. 46 -49). Indeed, we have previously demonstrated that ODC amino acid residues in the dimer interface that are distant from the active site are important for enzyme activity (44).
PBCV-1 is the first virus known to encode polyamine biosynthetic enzymes (1). In addition to encoding PBCV-1 DC, the virus contains the enzymes necessary to produce putrescine from agmatine (agmatine iminohydrolase and N-carbamoylputrescine amidohydrolase), and it also contains homospermidine synthase, but not spermidine synthase, suggesting that the end product of the pathway in the virus is homospermidine (50). It is interesting to speculate on the evolutionary driving force for the substrate specificity of PBCV-1 DC for L-arginine. In algae, including chlorella isolates, amino acid analyses demonstrated that L-arginine is one of the most abundant amino acids, whereas L-ornithine is one of the least abundant (51). Thus, the lack of L-ornithine in these cells suggests that the switch in substrate specificity of PBCV-1 DC to L-arginine was required for efficient production of putrescine.
The high amino acid sequence identity (40%) between PBCV-1 DC and the ODCs suggests that PBCV-1 DC is more closely related to enzymes with ornithine specificity than to bacterial ADCs in the ␤/␣-fold class, despite their common substrate preference. A potentially important link in this evolution is the observation that a group of bacterial enzymes that also share strong sequence similarity with the eukaryotic ODCs have dual specificity for L-lysine and L-ornithine (33). PBCV-1 DC groups most closely to these bacterial enzymes in phylogenetic analysis, and these sequences form a bridge between the eukaryotic ODCs and the more distantly related ADCs from bacteria and plants (Fig. 2). Interestingly, like the bacterial enzymes, PBCV-1 DC has equal activity on both Llysine and L-ornithine, although it is distinct in preferring L-arginine. These observations suggest that PBCV-1 DC may have been acquired from a bacterial source or from a common ancestor that contained a dual specificity Lys/OrnDC and that it has evolved a minimum set of amino acid substitutions to switch specificity without adversely affecting activity. Interestingly, the PBCV-1 homospermidine synthase was also proposed to be more closely related to bacterial enzymes than to plant enzymes (50), suggesting that the entire polyamine biosynthetic pathway in the virus may have been acquired from bacteria.
The role of putrescine and homospermidine in viral pathogenesis remains unclear because the host cell also makes both polyamines. However, the virus is known to inhibit protein translation in the host (1,2), and altered levels of polyamines could play a role in this inhibition. Eukaryotic initiation factor 5A plays an essential role in translation of selective messages required for cell division and proliferation (52). This factor is synthesized as an inactive precursor that is modified to its active form by the covalent attachment of hypusine via two enzymatic steps. The first step is catalyzed by deoxyhypusine synthase and utilizes spermidine as the donor substrate. Recently, it was reported that this same enzyme can catalyze the reverse reaction using putrescine as an acceptor, producing homospermidine and unmodified eukaryotic initiation factor 5A (53). Putrescine levels increase significantly in chlorella cells after infection with PBCV-1 (50), and this increase could potentially alter the levels of hypusine-modified eukaryotic initiation factor 5A, thereby affecting host cell translation.