A Nudix Enzyme Removes Pyrophosphate from Dihydroneopterin Triphosphate in the Folate Synthesis Pathway of Bacteria and Plants* □ S

triphosphate (DHNTP) is the second step in the pterin branch of the folate synthesis pathway. There has been controversy over whether this reaction requires a specific pyrophosphohydrolase or is a metal ion-dependent chemical process. The genome of Lactococcus lactis has a multicistronic folate synthesis operon that includes an open reading frame ( ylgG ) specifying a putative Nudix hydrolase. Because many Nudix enzymes are pyrophosphohydrolases, YlgG was expressed in Escherichia coli and characterized. The recombinant protein showed high DHNTP pyrophosphohydrolase activity with a K m value of 2 (cid:1) M , had no detectable activity against deoxynucleoside triphosphates or other typical Nudix hydrolase substrates, required a physiological level ( (cid:1) 1 m M ) of Mg 2 (cid:2) , and was active as a monomer. Essentially no reaction occurred without enzyme at 1 m M Mg 2 (cid:2) . Inactivation of ylgG in L.

In favor of an enzymatic mechanism, Suzuki and Brown (9) partially purified a small (17 kDa) Escherichia coli protein able to catalyze the release of pyrophosphate from DHNTP but not nucleoside triphosphates. The enzyme was accordingly named DHNTP pyrophosphohydrolase. The K m for DHNTP was 11 M, and activity was Mg 2ϩ -dependent. No evidence was presented showing this activity to be necessary for folate synthesis, however. More recent work in Bacillus subtilis has questioned the requirement for an enzyme, because chemical conversion of DHNTP to DHNP was observed at pH 8 and 37°C in the presence of Mg 2ϩ or Ca 2ϩ (10). Unfortunately, the Mg 2ϩ and Ca 2ϩ concentrations studied were unphysiologically high (5-12 mM), with typical cytosolic levels of free Mg 2ϩ and Ca 2ϩ being ϳ1 mM and ϳ0.1 M, respectively, in both bacteria and plants (11)(12)(13)(14).
The mechanism of the DHNTP 3 DHNP step is of considerable interest. First, if there is a specific DHNTP pyrophosphohydrolase, it could be a novel target for antibacterial drug discovery in the folate synthesis pathway (8). Second, modestly higher fluxes to folate can be engineered in bacteria (15) and plants (16,17) by greatly increasing the production of DHNTP, but it is not yet clear whether this DHNTP is efficiently converted to DHNP. If it is not, engineering the overexpression of a DHNTP pyrophosphohydrolase could be used to correct the problem.
The annotated genome sequence of the lactic acid bacterium Lactococcus lactis (18) offers a clue to the existence of a specific DHNTP pyrophosphohydrolase, for it reveals a folate biosynthesis gene cluster in which known folate genes flank a gene (ylgG) encoding an unknown protein from the Nudix hydrolase family. Nudix (nucleoside diphosphate linked to x) hydrolases are typically small proteins (15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25), contain a characteristic motif (GX 5 EX 7 REUXEEXGU, where U is a hydrophobic residue), and many have nucleoside triphosphate pyrophosphohydrolase activity (19). Furthermore, ylgG was shown to be cotranscribed with three of the flanking folate genes in a multicistronic operon (15, * This work was supported in part by the Florida Agricultural Experimental Station, by an endowment from the C. V. Griffin, Sr., Foundation, and by Grant 2005-35318-15228 from the National Research Initiative Competitive Grants Program of the United States Department of Agriculture. This work was approved for publication as Journal Series number R-10630. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 20) (Fig. 1B). Genomic evidence thus identifies YlgG as a candidate for the elusive DHNTP pyrophosphohydrolase.
In this study, we set out to find whether recombinant L. lactis YlgG has specific DHNTP pyrophosphohydrolase activity, and whether ylgG deletants are defective in pteridine and folate synthesis. When this proved to be the case, we sought homologs of YlgG among the Nudix proteins of Arabidopsis (21), and we showed that the closest homolog has DHNTP pyrophosphohydrolase activity.

EXPERIMENTAL PROCEDURES
Chemicals and Reagents-Pteridines were obtained from Schircks Laboratories (Jona, Switzerland). The Biomol Green TM phosphate assay reagent was from Biomol (Plymouth Meeting, PA). Ni 2ϩ -nitriloacetic acid superflow resin was from Qiagen (Valencia, CA). PD-10 columns were from Amersham Biosciences. Oligonucleotides were from MWG Biotech (High Point, NC). Reduced nicotinamide mononucleotide (NMNH) was a gift of G. Magni (University of Ancona, Italy). Other chemicals were from Sigma or Fisher.
Inactivation of ylgG by Double Crossover Recombination-The bacterial strains and plasmids used are listed in Table I. E. coli was grown at 37°C in TY medium (22), and L. lactis was grown at 30°C in M17 medium (23) containing 0.5% glucose (GM17 medium) or in chemically defined medium (CDM), prepared as described (24) except that folic acid was omitted. E. coli was transformed by heat shock and L. lactis by electroporation (25). For E. coli, kanamycin and erythromycin were used at final concentrations of 50 and 200 g/ml, respectively. PCRs were performed with Pfx polymerase (Invitrogen). The pre-integration vector for inactivation of ylgG was constructed by using pORI280 (26) and an amplified linear DNA fragment (2287 nucleotides) consisting of 450 bp of the 3Ј-end of folP, the entire ylgG gene, and the entire folC gene, by using primers folPNcoI-F and folCXbaI-R, which contained an NcoI site and an XbaI site, respectively (Table II). The amplicon was digested with NcoI and XbaI and cloned between the matching sites in pORI280, giving plasmid pORI280-folP-ylgG-folC. This plasmid was transformed to E. coli and harvested. The final integration vector, containing an in-frame deletion of almost the entire ylgG gene, was obtained by PCR amplification using pORI280-folP-ylgG-folC as a template and primers ylgGPstI-F and ylgGPstI-R (Table II). Both primers contained a terminal PstI site. The amplicon was digested with PstI and ligated, generating plasmid pORI280-folP-folC, which was transformed to E. coli and isolated for subsequent introduction into L. lactis strain NZ9000 (27). L. lactis transformants in which plasmids had integrated via single crossover were grown overnight in GM17 with 2 g/ml erythromycin and plated on GM17 agar containing 2 g/ml erythromycin and 80 g/ml 5-bromo-4-chloro-3-indoyl-galactopyranoside (X-gal). The orientation of the single crossover event was tested by PCR amplifications directly on cell material from a blue colony using the primers pORIcntr-F (sequence present downstream of multiple cloning site of pORI280) and folP-F (sequence present in folP but not the integration vector) (Table II). Subsequently, cells that contained the plasmid integrated over the short folP-flanking region were grown in medium with-out antibiotics and re-inoculated in the same medium to a density of ϳ1-10 cells/ml. After growth for ϳ30 -40 generations, dilutions of the culture were plated on agar containing X-gal. After 48 h white colonies were selected, and the presence or absence of ylgG was determined by PCR amplification from cell material by using primers folP-F and folCXbaI-R (Table II).
Gene Cloning and Protein Expression in E. coli-The ylgG gene was amplified from a pNZ8048 derivative (28) using KOD polymerase (Novagen, Madison, WI) and the primers 5Ј-CATGCACCATGGATGAG-GATTTGATTTCT-3Ј (forward) and 5Ј-CCCGCTCGAGTATGTTCT-TATAACGATA-3Ј (reverse), containing NcoI and XhoI sites, respectively. Using the NcoI site for cloning changed the second amino acid of YlgG from Asn to Asp. A full-length cDNA (Arabidopsis Biological Resource Center U50379) served as template to amplify the At1g68760 coding sequence, using the primers 5Ј-ACCCACATGTCGA-CAGGAGAAGCG-3Ј (forward) and 5Ј-CCCGCTCGAGGTCTCCAC-CACCATGAGT-3Ј (reverse), containing AflIII and XhoI sites, respectively. The ylgG and At1g68760 amplicons were digested with appropriate enzymes and cloned between the NcoI and XhoI sites of pET-28b (Novagen), which added a hexahistidine tag to the C terminus. The sequence-verified constructs were electroporated into E. coli BL21-CodonPlus®-RIL (DE3) cells (Stratagene, La Jolla, CA). For enzyme production, cells were grown at 37°C in 1 liter of LB medium containing 100 g/ml ampicillin and 20 g/ml chloramphenicol until the A 600 reached 0.5. Isopropyl-D-thiogalactopyranoside was then added (final concentration 1 mM) and incubation continued for 16 h at 25°C.
Protein Isolation and Molecular Mass Determination-E. coli cells from a 1-liter culture were harvested by centrifugation, resuspended in 20 ml of 50 mM Tris-HCl, pH 8.0, containing 0.1 M KCl (buffer A), and broken in a Mini-BeadBeater (Biospec Products, Bartlesville, OK) using 0.1-mm zirconia/silica beads. Protein purification by Ni 2ϩ -affinity chromatography under native conditions followed the manufacturer's protocol (Qiagen). All steps were carried out at 0 -4°C. Purified proteins were desalted on a PD-10 column equilibrated in buffer A. For YlgG, a second purification step was carried out using a Waters 626 high pressure liquid chromatography system. Proteins were loaded on a MonoQ HR 5/5 column (Amersham Biosciences) equilibrated in buffer A containing 0.1 M NaCl. Proteins were eluted with a linear gradient of 0.1-0.5 M NaCl (0.2 ml/min), and fractions with DHNTP pyrophosphohydrolase activity were pooled. Purified proteins were stored at 4°C for up to 6 months, which did not cause significant activity loss. Protein concentration was determined by Bradford's method (29) using bovine serum albumin as standard. The native molecular mass of YlgG was estimated using a calibrated Superdex 200 HR 10/30 column (Amersham Biosciences) equilibrated with 50 mM HEPES-KOH, pH 8.0, containing 150 mM NaCl, using a flow rate of 0.3 ml/min. The standards were carbonic anhydrase, bovine serum albumin, ␤-amylase, apoferritin, and thyroglobin.
Preparation of DHNTP-Recombinant E. coli GTP cyclohydrolase I (30) was used to prepare DHNTP. The reaction was run in 1.5 ml of 50 mM HEPES-KOH containing 0.1 M KCl (buffer B) plus 1 mM GTP, 10 mM ␤-mercaptoethanol, and 30 g of purified GTP cyclohydrolase I for 2 h at 37°C. DHNTP formation was followed by absorption at 330 nm until all GTP was converted; the molar extinction coefficient of DHNTP is 6300 M Ϫ1 cm Ϫ1 at 330 nm (31). The solution was then deproteinized using a Centricon YM-10 unit (Millipore, Billerica, MA). Aliquots containing 30 -80 nmol of DHNTP were frozen in liquid N 2 and stored at Ϫ80°C. After thawing, aliquots were used at once and not re-frozen.
Enzyme Assays-Assays were routinely made in 50 -200-l reaction mixtures at 30°C for 5-180 min in buffer B containing the specified concentrations of substrates and divalent metal ions (as chloride salts). The 0.1 M KCl in this buffer did not significantly affect enzymatic reactions. DHNTP pyrophosphohydrolase activity was assayed by fluorometric HPLC. Reactions were stopped by adding oxidizing solution (0.5% I 2 , 1% KI, w/v, in 1 M HCl, 10 l per 100-l reaction mix) and incubating for 1 h at 4°C. Then 10 l of 2% sodium ascorbate and 4.1 l of 1 M Na 2 HPO 4 , 1 M NaOH were added to destroy excess I 2 and adjust the pH to 6.0, and denatured protein was pelleted. Separation of neopterin and its phosphates was basically as described (4), using a Waters 2695 instrument (Waters, Milford, MA). Briefly, samples (20 -50 l) were injected on a C 18 Synergi 4 Fusion-RP 80 column (4 m, 4.6 ϫ 250 mm, Phenomenex, Torrance, CA) and eluted isocratically with 10 mM sodium phosphate buffer, pH 6.0, at 1 ml/min. Peaks were detected using a Waters 2475 Multi Fluorescence Detector (excitation 350 nm, emission 450 nm) and quantified relative to neopterin monophosphate (Schircks). A neopterin diphosphate standard was prepared from DHNTP by acid phosphatase treatment and oxidation. Pyrophosphate was quantified using the Biomol Green TM phosphate detection reagent after addition of 1-2 units of inorganic pyrophosphatase (Sigma) to enzymatic assays (32). To measure NADH hydrolysis, samples (20 l) were injected on a C 18 symmetry column (3.5 m, 4.6 ϫ 250 mm, Waters) and eluted in 50 mM sodium phosphate, pH 6.0, containing 8 mM tetrabutylammonium bisulfate using a linear gradient from 12 to 30% methanol. The NMNH peak was quantified fluorometrically (excitation 340 nm, emission 460 nm) relative to authentic NMNH. Hydrolysis of nucleoside triphosphates and their 2Ј-deoxy derivatives was assayed by pyrophosphate release as described above. Reactions were stopped by adding either excess EDTA (2-3 times the quantity of M 2ϩ ions) or the Biomol Green TM reagent. Hydrolysis of other substrates (Ap4A, Ap3A, NAD, ADP-ribose, UDP-glucose, and coenzyme A) was assayed by treating the reaction products with shrimp alkaline phosphatase (Roche Applied Science) as described (32) and determining phosphate with the Biomol Green TM reagent. Kinetic data were analyzed by direct fitting to the Michaelis-Menten equation using nonlinear regression.
Folate and Pteridine Analysis-Total folate was quantified using the Lactobacillus casei microbiological assay (33) after enzymatic deconjugation, as described previously (15). Pteridines were extracted from L. lactis cells in buffer C (0.2 M sodium citrate, pH 4.5, containing 5 mM ␤-mercaptoethanol) and analyzed by HPLC. Briefly, cells of a 50-ml overnight culture (30°C) were harvested by centrifugation, resuspended in buffer C, and broken in a Mini-BeadBeater. Samples were centrifuged, and the supernatant was incubated for 10 min at 37°C plus or minus 3 units of wheat germ acid phosphatase (Sigma). Acid hydrolysis conditions were used to avoid formation of cyclic DHNP, which is not attacked by phosphatase (4). Reactions were stopped by adding 15 l of oxidizing solution per 100-l sample and then incubated for 1 h at 4°C. Then 15 l of 2% sodium ascorbate and 6.1 l of 1 M Na 2 HPO 4 , 1 M NaOH were added, and the samples were centrifuged and analyzed by HPLC as described above except that the flow rate was 1.5 ml/min.

RESULTS
Inactivation of ylgG-The ylgG gene was inactivated in L. lactis strain NZ9000⌬ylgG by double crossover recombination using modifications of a published method (26). The NZ9000⌬ylgG strain, which lacks 95% of the ylgG gene, had the same growth rate as the wild type on minimal (CDM) medium ( Fig. 2A, inset). However, the mutant had intracellular folate levels 3-fold lower than the wild type and, unlike the wild type, secreted almost no folate to the medium ( Fig. 2A). The complementation of ylgG into the NZ9000⌬ylgG strain restored folate level to that of the wild type (not shown). These data indicate that YlgG mediates a step in folate biosynthesis. HPLC analysis showed this step to be dephosphorylation of DHNTP, for mutant cells accumulated a pteridine with the    (Fig. 3A) was expressed in E. coli with a C-terminal hexahistidine tag. The expression level was insufficient to detect the recombinant YlgG in the total soluble protein fraction, but a polypeptide close to the predicted size (20.4 kDa, including the His tag) was readily isolated from this fraction by Ni 2ϩ affinity and anion exchange chromatography (Fig. 4A). The purified YlgG constituted ϳ0.2% of total soluble protein.
Enzyme assays showed that purified YlgG had high DHNTP pyrophosphohydrolase activity, i.e. that it cleaved DHNTP to give DHNP and pyrophosphate in a 1:1 ratio, with negligible production of free phosphate (Fig. 4B). This activity was dependent on Mg 2ϩ and reached ϳ70% of its maximal value at 1 mM (Fig. 4C). Because 1 mM is approximately the physiological level of free Mg 2ϩ (11,12), this concentration was adopted for subsequent work. Ca 2ϩ at 1 mM did not activate YlgG; Zn 2ϩ and Mn 2ϩ could not be tested as they caused rapid chemical breakdown of DHNTP (9). At 1 mM Mg 2ϩ , chemical conversion of DHNTP to DHNP was almost undetectable (Ͻ0.5% of the maximal rate of the enzymatic reaction in Fig. 4C). This chemical conversion increased when Mg 2ϩ or Ca 2ϩ concentration was raised to 10 -12 mM as reported (10), but at these concentrations other major breakdown products were also formed (not shown).
Kinetic characterization of YlgG (Table IV) gave a K m value for DHNTP (2.1 M) that is among the lowest reported for Nudix enzymes and gave a fairly typical k cat value (0.21 s Ϫ1 at 30°C) (34). YlgG had no detectable activity with GTP, ATP, TTP, CTP, the corresponding deoxynucleoside triphosphates, 8-oxo-dGTP, NADH, NAD ϩ , diadenosine tri-or tetraphosphate, ADP-ribose, UDP-glucose, or coenzyme A (the detection limit was Յ1% of the activity with DHNTP). Consistent with this lack of activity, the YlgG amino acid sequence (Fig. 3A) does not have the motifs associated with specificity for NADH, diadenosine polyphosphates, ADP-ribose, or coenzyme A (35)(36)(37)(38). The native molecular mass of recombinant YlgG was estimated to be 24.2 kDa on the basis of its elution from a calibrated size exclusion column. This value is close to the calculated mass of the His-tagged polypeptide (20.4 kDa), indicating that the native enzyme is a monomer.
Arabidopsis Homologs of YlgG-BLAST searches of the Arabidopsis genome revealed 20 proteins with a canonical Nudix motif (see supplemental Table I). Most of these share homology with YlgG only in the area of the Nudix box, but five (shown in Fig. 3B) have some degree of overall homology. The overall homology is greatest for the At1g68760 gene product, which is 34% identical and 44% similar to YlgG in a 73-residue region spanning the Nudix box (Fig. 3A). Phylogenetic analysis confirmed that this protein is more closely related to YlgG than the other four (Fig. 3B). Moreover, the At1g68760 protein is of similar size (16.4 kDa) to YlgG and, like other early enzymes in plant folate biosynthesis (4, 6), has no obvious organellar targeting sequence. Like YlgG, the At1g68760 protein lacks the motifs associated with preference for NADH, diadenosine polyphosphates, ADP-ribose, or coenzyme A (35)(36)(37)(38). Finally, the At1g68760 protein is by far the closest Arabidopsis homolog (33% identity, 51% similarity) of the Nudix moiety of a pre- FIG. 3. Alignment and phylogenetic analysis of the deduced amino acid sequences of YlgG and selected Arabidopsis Nudix proteins. A, alignment of YlgG (GenBank TM AAN64308) with its closest Arabidopsis homolog, the At1g68760 gene product. Identical residues are shaded in black, and similar residues are shaded in gray. Dashes are gaps introduced to maximize alignment. The Nudix motif is overlined. B, unrooted phylogram, generated using the neighbor-joining method and 1,000 bootstrap iterations of YlgG, the At1g68760 protein, and other Arabidopsis proteins sharing homology with YlgG outside the Nudix box. Asterisks show branch points with bootstrap values Ͼ70%. Phylogenetic analysis was carried out using PHYLIP at the Institut Pasteur server (bioweb.pasteur.fr). Sequences were aligned using Multalin (prodes.toulouse.inra.fr/multalin/multalin.html).

TABLE III
Effect of ylgG inactivation on intracellular pteridine levels in L. lactis Cultures (50 ml) were grown in CDM medium to an A 600 value of 3. One-half of the pteridine extract of the cell pellet was treated with acid phosphatase (see "Experimental Procedures"); the other half was not. Pteridines were then oxidized to their aromatic forms and quantified by fluorometric HPLC. Values are means of three determinations Ϯ S.E. and are expressed as intracellular concentrations. The internal cell volume of a 50-ml culture at A 600 ϭ 3 was estimated as 93 l (taking an A 600 value of 1 as equivalent to 0.2 mg of protein ml Ϫ1 and L. lactis internal cell volume as 3.1 l mg Ϫ1 protein). Characterization of the At1g68760 Gene Product-The At1g68760 gene product with or without a C-terminal His tag was abundantly expressed in E. coli (Fig. 5A). Pilot assays with desalted extracts indicated that the tag made no difference to the enzymatic characteristics of the protein, so all further work was done on the tagged protein after purification to homogeneity by Ni 2ϩ affinity chromatography (Fig. 5A). Like YlgG, the At1g68760 protein had DHNTP pyrophosphohydrolase activity, producing DHNP and pyrophosphate in a 1:1 ratio but essentially no free phosphate (Fig. 5B). Also like YlgG, activity required Mg 2ϩ (Fig. 5C), and this could not be replaced by Ca 2ϩ (1 mM). Unlike YlgG, however, in the presence of 1 mM Mg 2ϩ the At1g68760 protein was far more active against dGTP, 8-oxo-GTP, dTTP, and other nucleoside and deoxynucleoside triphosphates than against DHNTP (Table V). Activities with other Nudix substrates were negligible (Table V).  To further explore this striking contrast with the substrate specificity of YlgG, we evaluated the kinetic properties of the At1g68760 protein with DHNTP or dGTP as substrate in the presence of 1 mM Mg 2ϩ (Table VI). The K m value for DHNTP was fairly similar to that for dGTP, but the k cat value was 500-fold lower, so that the catalytic efficiency was far higher for dGTP.
The At1g68760 gene product was reported by Dobrzanska et al. (21) to have high NADH pyrophosphatase activity and was designated AtNUDT1 (we will use this name henceforth). This activity against NADH was, however, characterized using an Mn 2ϩ concentration of 5 mM, which is ϳ10 3 -fold greater than typical levels of Mn 2ϩ in growing cells (39). Low activity was also found in the presence of 5 mM Mg 2ϩ (21). We therefore measured NADH pyrophosphatase activity at various Mn 2ϩ and Mg 2ϩ concentrations (Fig. 5D). This activity fell drastically as Mn 2ϩ concentration approached the usual physiological range, and at 5 M Mn 2ϩ was only 0.001% that at 5 mM (Fig.  5D). With Mg 2ϩ , activity was insignificant except at the highest concentrations tested (Fig. 5D).

DISCUSSION
Our genetic and biochemical results with L. lactis identify the YlgG protein as a specific DHNTP pyrophosphohydrolase that participates in the folate pathway. We therefore propose that the gene encoding this protein be redesignated folQ. The Mg 2ϩ requirement, substrate specificity, K m value, and native molecular mass of YlgG are very like those reported for the partially purified DHNTP pyrophosphohydrolase of E. coli (9). As E. coli and L. lactis are not closely related, the occurrence of similar DHNTP pyrophosphohydrolases in both species implies that specific DHNTP-cleaving Nudix enzymes are common to a wide range of bacteria.
Although deleting ylgG reduced folate production markedly in L. lactis, it did not eliminate it and did not affect growth. The residual folate synthesis in the deletion mutant is not unexpected, because broad spectrum phosphatases can attack DHNTP (40), and the massive accumulation of DHNTP seen in the mutant could potentially drive significant flux through nonspecific, normally minor, enzymatic dephosphorylation reactions. The accumulation of dihydroneopterin diphosphate found in the YlgG-deficient cells supports this possibility, for it indicates that iterative removal of single phosphate residues has inefficiently replaced removal of pyrophosphate as the main route from DHNTP to DHNP.
Two lines of evidence show that metal ion-catalyzed chemical dephosphorylation of DHNTP (10) is certainly not a major process under normal in vivo conditions and is probably not even a significant minor one. First, the large build up of DHNTP in deletant cells indicates that most of the DHNTP 3 DHNP flux is normally mediated by YlgG. Second, the rate of chemical dephosphorylation was negligible in in vitro assays made at physiological pH and Mg 2ϩ levels. The proposal that chemical dephosphorylation is important (10) was based solely on biochemical experiments involving unphysiological Mg 2ϩ or Ca 2ϩ levels and is unsupported by genetic or other in vivo evidence.
Cleavage of pyrophosphate from DHNTP is a new addition to the list of reactions shown to be mediated by Nudix enzymes (19,36,41,42). The Mg 2ϩ response observed for YlgG is very similar to those of Nudix nucleoside triphosphate pyrophosphohydolases such as E. coli MutT (the archetype of this enzyme class), which requires two divalent cations, one coordinated by the nucleoside triphosphate and the other by conserved residues in the Nudix motif and elsewhere (43). The low K m value observed (2.1 M) for YlgG is consistent with the estimated in vivo concentration of DHNTP in wild type L. lactis cells (0.33 M).
It is intriguing that the DHNTP concentration in wild type L. lactis is 10-fold higher than that of DHN, because this suggests that the rate of DHNTP dephosphorylation limits the pool size of DHN, and hence flux from DHN to folates. If this is the case, increasing YlgG expression could potentially enhance folate production in L. lactis strains engineered to overproduce DHNTP (15).
DHNTP is an intermediate in the biosynthetic pathway of tetrahydrobiopterin as well as of folate, and organisms such as pseudomonads, cyanobacteria, and fungi have both pathways (44 -47). If these organisms use a DHNTP pyrophosphohydrolase, then this enzyme would be the one that commits pteridine moieties to folate synthesis. This differs from the usual situation in which the committing role is played by GTP cyclohydrolase I, the enzyme that produces DHNTP (1, 10). Regardless of whether DHNTP pyrophosphohydrolase has a committing role, it would seem to be a poor target for antibacterial drug discovery because inactivating it, in L. lactis at least, did not totally suppress folate synthesis and had no impact on growth.
The previously known bacterial folate synthesis enzymes all have homologs in plants (2)(3)(4)(5)(6)(7), and the same is true for YlgG, although the homology in this case is weaker than usual. More generally, bacterial Nudix enzymes often share sequence features outside the Nudix box with eukaryotic enzymes having the same activity (35,38,47,48). This pattern of cross-kingdom structural similarity between functionally equivalent Nudix hydrolases in itself makes the original identification of At-NUDT1 as an NADH pyrophosphatase (21) seem suspect, for this protein lacks the SQPWPFPXS motif found downstream of the Nudix box in other NADH pyrophosphatases (35,38). (This motif is, however, present in another Arabidopsis protein, the At5g20070 gene product that has yet to be characterized, see Ref. 38.) This suspicion about the function of AtNUDT1 was corroborated by biochemical evidence, for we found that the activity against NADH requires unphysiological levels of Mn 2ϩ and that DHNTP and nucleoside and deoxynucleoside triphosphates are far better substrates than NADH in the presence of physiological Mg 2ϩ levels. A similar switch in substrate preference when Mg 2ϩ replaces Mn 2ϩ has been reported for another Nudix hydrolase (49).
The catalytic efficiency (k cat /K m ) of AtNUDT1 for DHNTP is ϳ10 4 -fold lower than that of YlgG and, unlike YlgG, AtNUDT1 is not at all specific for DHNTP, deoxynucleoside and nucleoside triphosphates being much better substrates. These findings, added to the probability that cytosolic deoxynucleoside and nucleoside triphosphate levels greatly exceed that of DHNTP (17,50), raise the question of whether AtNUDT1 could account for DHNTP hydrolysis in vivo or whether another, more efficient, DHNTP pyrophosphohydrolase remains to be discovered. We cannot at this point exclude the existence of another enzyme, but two arguments suggest there is no need to invoke one. First, the DHNTP 3 DHNP flux is so small that it might well be sustained even by an inefficient enzyme operating in the presence of a large excess of better substrates. (This flux can be estimated from data on folate turnover in plants as Յ15 pmol h Ϫ1 g Ϫ1 fresh weight (51,52), which is ϳ10 6 -fold smaller than the respiratory ADP 3 ATP flux (53), for example.) Second, there are precedents for Nudix hydrolases mediating key minor reactions while awash in alternative substrates that they are concurrently cleaving. MutT-type Nudix enzymes are thought to sanitize nucleotide pools by removing oxidized, mutagenic derivatives such as 8-oxo-dGTP. These enzymes usually prefer the oxidized derivatives but, like At-NUDT1, also attack normal deoxynucleoside and nucleoside triphosphates, which are far more abundant in vivo (54 -57). Because the substrate range of AtNUDT1 is like that of MutTtype hydrolases (54), it could be a bifunctional DHNTP pyrophosphohydrolase/sanitizing enzyme. If so, the metabolic inefficiency involved in hydrolyzing large amounts of normal nucleotides as well as DHNTP would add no extra energy cost because such wastage would already be part of the vital sanitizing function of the enzyme.