General Enzymatic Screens Identify Three New Nucleotidases in Escherichia coli BIOCHEMICAL CHARACTERIZATION OF SurE, YfbR, AND YjjG*

To find proteins with nucleotidase activity in Escherichia coli , purified unknown proteins were screened for the presence of phosphatase activity using the general phosphatase substrate p -nitrophenyl phosphate. Proteins exhibiting catalytic activity were then assayed for nucleotidase activity against various nucleotides. These screens identified the presence of nucleotidase activity in three uncharacterized E. coli proteins, SurE, YfbR, and YjjG, that belong to different enzyme superfamilies: SurE-like family, HD domain family (YfbR), and haloacid dehalogenase (HAD)-like superfamily (YjjG). The phosphatase activity of these proteins had a neutral pH optimum (pH 7.0–8.0) and was strictly dependent on the presence of divalent metal cations (SurE: Mn 2 > Co > >

DNA and RNA synthesis requires a continuous and balanced supply of the four deoxyribonucleotides and four ribonucleo-tides. A network of biosynthetic and catabolic enzymes regulates the size of each pool with the enzymes ribonucleotide reductase, nucleoside kinases, and nucleotidases playing the main roles (1). By opposing the phosphorylation of nucleosides by kinases, intracellular 5Ј-nucleotidases participate in substrate cycles that regulate the cellular levels of ribo-and deoxyribonucleoside monophosphates and thereby all ribo-and deoxyribonucleotide pools (1,2).
Nucleoside monophosphate phosphohydrolases or nucleotidases (EC 3.1.3.5 and EC 3.1.3.6) are phosphatases that specifically dephosphorylate nucleoside monophosphates to nucleosides and inorganic phosphate. Seven mammalian 5Јnucleotidases have been identified through cloning and biochemical characterization. These enzymes differ in tissue specificity, subcellular localization, primary structure, and substrate specificity (2). All act on nucleoside monophosphates producing free nucleosides and P i . Each natural nucleotide can be the substrate for several nucleotidases, because they have overlapping specificities. The ubiquitous ecto-5Ј-nucleotidase, eN, 1 is anchored to the surface of the plasma membrane, and AMP is considered to be its major physiological substrate (3). Five other nucleotidases, including dNT-1, occur in the cytosol and one (dNT-2) occurs in mitochondria. dNT-1 and dNT-2 show a preference for the dephosphorylation of dUMP and dTMP (4). The "high K m -nucleotidase" cN-II prefers GMP and IMP (5), the cytosolic nucleotidases cN-IA and cN-IB have a preference for dephosphorylation of AMP (6,7), and PN-1 (or cN-III) is most active with CMP and UMP (8). For bovine cN-II, the formation of a covalent phosphoenzyme intermediate has been demonstrated, and the phosphate group is localized to the N-terminal motif DXDX(T/V) (9). This short motif can also be found in other 5Ј-nucleotidases and is related to a similar motif in the haloacid dehalogenase (HAD)-like superfamily of enzymes (10). Recently, the structure of human dNT-2 (a mitochondrial deoxyribonucleotidase) was solved and revealed a structural relationship to the HAD superfamily, members of which use an aspartyl nucleophile as their common catalytic strategy (11). Based on these structures, the mononuclear metal center (Mg 2ϩ ) and the N-terminal DXDX(T/V) motif are proposed to be the main players in the catalytic mechanism.
In contrast to the well characterized mammalian nucleotidases, the field of prokaryotic nucleotidases remains unexplored. Previous studies demonstrated the presence of 5Ј-nucleotidase activity in crude extracts (probably in the periplasmic fraction) from Proteus vulgaris, Haemophilus influenzae, and Staphylococcus aureus (12). The membranebound nucleotidase from Bacillus cereus was purified and shown to be active toward ribo-and deoxyribonucleoside 5Јand 3Ј-monophosphates (13). The soluble periplasmic 5Ј-nucleotidase UshA from E. coli was purified and shown to be able to hydrolyze all nucleoside 5Ј-mono-, di-, and triphosphates, as well as UDP-glucose (14). The crystal structure of this protein revealed the presence of a dimetal center and a catalytic dyad Asp-His in the active site indicating that its catalytic mechanism is different from that of the human 5Ј-nucleotidase dNT-2 (15). The major physiological role of the E. coli UshA is the degradation of exogenous UDP-glucose and 5Ј-nucleotides to nucleosides, glucose 1-phosphate, and phosphate for utilization by the cell (16). However, in the KEGG PATHWAY data base of metabolic pathways (www.genome.ad.jp/kegg/pathway.html), this periplasmic protein UshA is mentioned as a nucleotidase responsible for the intracellular conversion of nucleoside 5Јmonophosphates to corresponding nucleosides in E. coli.
To identify intracellular nucleotidases in E. coli, we screened purified unknown proteins for phosphatase activity using a general phosphatase substrate, p-nitrophenyl phosphate (pNPP). Proteins that showed phosphatase activity against pNPP were further screened with a set of natural phosphatase substrates (70 phosphorylated compounds), including various nucleotides. E. coli SurE (253 amino acids), YfbR (199 amino acids), and YjjG (225 amino acids) demonstrated phosphohydrolase activities against general substrate pNPP and nucleotides. These uncharacterized proteins show no sequence similarity to each other (15.5-18.3% identity) and belong to different phosphohydrolase families: SurE-like, HD domain (YfbR), and haloacid dehalogenase (HAD)-like hydrolases (YjjG). Biochemical characterization of these proteins demonstrated that they have different substrate specificities and are likely to be involved in different reactions of intracellular nucleotide metabolism.

MATERIALS AND METHODS
Gene Cloning and Protein Purification-Purified uncharacterized E. coli proteins used for screening were provided by the Ontario Centre for Structural Proteomics (www.uhnres.utoronto.ca/proteomics/). The genes encoding these proteins were amplified by PCR from E. coli DH5␣ genomic DNA and cloned as previously described (17) into the NdeI and BamHI sites of a modified form of pET15b (Novagen) in which a tobacco etch virus protease cleavage site replaced the thrombin cleavage site and a double stop codon was introduced downstream from the BamHI site. The E. coli YfbR was cloned from the E. coli K12 W3110 genomic DNA into the archive vector pCA24N (Genobase data base, ecoli.aistnara.ac.jp). To prepare protein samples for screening, recombinant proteins containing an N-terminal His 6 fusion tag were expressed in 1-liter E. coli BL21(DE3) cultures and affinity-purified on nickel-nitrilotriacetic acid resin (Qiagen) as previously described (17). Briefly, cell lysates were loaded onto small (1.5 ϫ 4 cm) nickel columns equilibrated with loading buffer (50 mM HEPES-K, pH 7.5; 0.5 M NaCl; 5 mM imidazole), washed with 10 volumes of washing buffer (loading buffer containing 30 mM imidazole), and eluted with 250 mM imidazole in loading buffer. For biochemical characterization, the proteins were overexpressed in 4-liter cultures of the E. coli BL21(DE3) and purified to over 95% of homogeneity using a combination of affinity chromatography on nickel-nitrilotriacetic acid resin (Qiagen) and anion-exchange chromatography on Q-Sepharose (Amersham Biosciences).
General Phosphatase Screens-Purified E. coli proteins were screened for phosphatase activity using general phosphatase substrate p-nitrophenyl phosphate (pNPP) in 96-well microplates. The reaction mixture (0.2 ml) contained 50 mM HEPES-K buffer (pH 7.5), 5 mM MgCl 2 , 0.5 mM MnCl 2 , 0.5 mM NiCl 2 , 40 mM pNPP, 0.1-1.0 g of purified protein. After addition of protein samples, the plates were incubated for 1-3 h at 37°C before taking the reading at 410 nm. Positive controls with 1 g of calf intestinal phosphatase from Sigma were included in all tests.
Natural Phosphatase Screens-To screen purified proteins for phos-phatase activity against natural substrates, 70 phosphorylated compounds from Sigma (various nucleotides and phosphorylated sugars, amino acids, and organic acids) were used as individual substrates (one compound/well) or as a mix of several related compounds (nucleoside 5Ј-mono-, di-, or triphosphates, nucleoside 3Ј-monophosphates). Screening was performed in 96-well microplates using 160-l reaction mixtures containing 50 mM HEPES-K (pH 7.5), 0.1 mM substrate, 5 mM MgCl 2 , 0.5 mM MnCl 2 , 0.5 mM NiCl 2 , and 2-5 g of protein. After 30 -60 min of incubation at 37°C, the reaction was terminated by the addition of 40 l of Malachite Green reagent (18), and after 5 min the production of P i was measured at 630 nm. Enzymatic Assays-Phosphatase activity against pNPP was measured in a reaction mixture (1 ml) containing 50 mM HEPES-K buffer (pH 7.0, for SurE; pH 8.0, for YfbR; and pH 7.5, for YjjG), 0.1-10 mM metal (as indicated in the figure legends), 4 -8 mM pNPP, and 1-8 g of protein. After 10 -20 min of incubation at 37°C, the reaction was stopped by the addition of 0.2 ml of 6 N NaOH, and the concentration of the p-nitrophenol product was determined from the absorbance at 410 nm. The pH dependence of phosphatase activity toward pNPP (4 mM) or natural substrates (0.1 mM) was determined in the presence of optimal metal (5 mM Mg 2ϩ or 0.5 mM of other metals) and 1-8 g of purified protein. Assays were performed in MES-K buffer between pH 5.5 and 6.5, in HEPES-K buffer between pH 7.0 and 8.0, in CHES-K buffer between pH 8.5 and 10.0, in CAPS-K buffer between pH 10.0 and 11.0. The metal dependence (metal profiles) of the phosphatase activity with pNPP or natural substrates was determined at optimal pH using various metals (5 mM Mg 2ϩ or 0.5 mM of other metals), substrate (4 mM pNPP or 0.1 mM of natural substrate), and 1 g of protein.
Phosphatase (nucleotidase) activity toward selected natural substrates (substrate profiles) was assayed in 0.8-ml reaction mixtures containing 50 mM HEPES-K buffer (pH 7.0, for SurE; pH 8.0, for YfbR; pH 7.5, for YjjG), 0.1-0.5 mM substrate, 0.1-0.5 mM of metal (as indicated), 1-8 g of protein. After 20 -30 min incubation at 37°C, the reaction was terminated by the addition of 0.2 ml of Malachite Green reagent (18), and the production of P i was measured at 630 nm. One unit of activity is defined as 1 mol of P i /min. For determination of the K m and V max , the phosphatase assays contained substrates at concentrations of 0.005-2.0 mM. Kinetic parameters were determined by non-linear curve fitting from the Lineweaver-Burk plot using GraphPad Prism software (version 4.00 for Windows, GraphPad Software, San Diego, CA, www.graphpad.com).
Phosphotransferase activity of purified proteins was assayed qualitatively by TLC analysis of reaction products using cellulose plates in solvent A (saturated ammonium sulfate/3 M sodium acetate/isopropyl alcohol; 80:6:2) (19). The reaction mixtures (20-l final volume) were the same as for nucleotidase assays (described above), except that higher substrate (donors) concentrations (10 mM) were used and nucleoside acceptors (guanosine, cytidine, uridine, and thymidine; 10 mM final concentration) were added. The reaction products and nucleoside/ nucleotide standards were visualized under UV light.
Preparation of Polyphosphate Substrates-Polyphosphate mix (polyP-type 75, polyP-type 35, and polyP-type 15, all from Sigma) was fractionated on a HiTrap Q column (XL, 1 ml, Amersham Biosciences) using stepwise elution by the buffer (25 mM HEPES-K, pH 7.6) containing increasing concentrations of NaCl (0.05-1.0 M). The P i concentration in fractions was determined after acid hydrolysis (1 M HCl, 7 min in boiling water bath) by the method of Chen (20). The chain length of polyP substrates was assessed using polyacrylamide gel electrophoresis in 20% polyacrylamide gel with 7 M urea. polyP standards were generated by a partial acid hydrolysis (0.1 M HCl, 75°C with different time intervals) of polyP 750 , and the chain length was determined by counting the bands after staining the gel with 0.05% toluidine blue.

Screening of Purified Proteins for Phosphatase Activity-
Over 300 purified unknown proteins from E. coli were screened for phosphatase activity against general phosphatase substrate p-nitrophenyl phosphate (pNPP). These screens identified the presence of phosphatase activity in several E. coli proteins, including SurE (Fig. 1A), YfbR, and YjjG (data not shown). Secondary catalytic screens with natural phosphatase substrates (70 compounds) and three metals (Mg 2ϩ , Mn 2ϩ , and Ni 2ϩ ) revealed high phosphatase activity of these three proteins against a range of nucleoside 5Ј-or 3Ј-monophosphates ( Fig. 1B; results are shown for SurE) indicating that the E. coli proteins SurE, YfbR, and YjjG are nucleotidases. These pro-teins were purified in large scale and further characterized. Gel-filtration experiments demonstrated that, in solution, all three proteins exist mostly in an oligomeric state. YjjG showed the presence of dimers (main fraction, molecular mass ϭ 52.8 Ϯ 2.19 kDa), monomers (molecular mass ϭ 35.7 Ϯ 0.77 kDa), and possibly tetramers (molecular mass Ͼ 100 kDa). The native molecular mass of the main fractions of YfbR and SurE was higher than 100 kDa suggesting at least tetrameric organization of these proteins. Gel-filtration analysis of these proteins also revealed the presence of the SurE monomers (molecular mass ϭ 41.6 Ϯ 0.42 kDa) and the YfbR trimers (molecular mass ϭ 71.3 Ϯ 5.26 kDa).
E. coli SurE-The SurE family is an evolutionarily conserved group of proteins with homologues (or domains) found in eubacteria, archaea, and eukaryotes. The E. coli SurE was annotated as an acid phosphatase on the basis that its Yarrowia lipolytica homologue (PHO2; 21.5% sequence identity to the E. coli SurE) complemented mutations in two of the major acid phosphatases of Saccharomyces cerevisiae (23). However, in our work purified E. coli SurE hydrolyzed pNPP with highest activity at neutral pH (pH 7.0; data not shown). E. coli SurE also showed different metal requirement with highest activity and affinity to Mn 2ϩ followed by Co 2ϩ Ͼ Ni 2ϩ Ͼ Mg 2ϩ (Fig. 2). With natural phosphatase substrates, the E. coli SurE showed significant activity toward nucleoside 5Ј-or 3Ј-monophosphates FIG. 1. Screening of purified proteins for phosphatase activity. A, general phosphatase screen with pNPP as substrate. 47 proteins from E. coli were purified under native conditions (as described under "Experimental Procedures") and 10 g (in duplicate) were loaded into microplate wells containing 200 l of phosphatase reaction mixture (described under "Experimental Procedures"). A positive control was set up using 2 g of calf intestinal phosphatase from Sigma, and 10 l of elution buffer were used for the negative control. The picture was taken after 1-h incubation at 37°C. Positive reactions (indicated by the bracket) were obtained with the E. coli SurE protein. B, screening of the E. coli SurE for phosphatase activity with natural substrates. 70 phosphorylated compounds were added (one compound/well or as a mix of several related compounds) to 96-well microplate containing the phosphatase reaction mixture (described under "Experimental Procedures") without enzyme (rows A, C, E, and G) or with 2 g of SurE (rows B, D, F, and H) and incubated for 1 h at 37°C. The reactions were stopped by the addition of Malachite Green reagent, which in the presence of free phosphate produced a strong green color. Positive results were obtained with three mixtures: 5Ј-NMPs (A1, no SurE; B1, ϩSurE), 3Ј-NMPs (A5, no SurE; B5, ϩSurE), and 5Ј-dNMPs (G10, no SurE; H10, ϩSurE).  3A). The protein hydrolyzed both purine and pyrimidine ribo-and deoxyribonucleotides effectively and had a neutral pH optimum (pH 7.0 -7.2) (Fig. 3B). With all substrates, the protein showed saturation kinetics with high affinity to 3Ј-AMP, 5Ј-GMP, 5Ј-dGMP, 5Ј-AMP, and 3Ј-CMP (Table I). The observed K m values (0.10 -0.37 mM) are in the range typical of "high K m " soluble 5Ј-nucleotidases from mammals (5). As with pNPP, hydrolysis of natural substrates was greatly stimulated by a divalent metal cation with the same four metals showing similar efficiency (Fig. 2). With natural substrates, the protein had even higher affinity for metals with apparent K D at least 2-fold lower than in the presence of pNPP (Table II).
The only non-nucleotide natural substrate hydrolyzed by the E. coli SurE was found to be polyphosphate (Fig. 4). With this substrate, the protein showed exopolyphosphatase activity (P i release) that was also dependent on the divalent metal cation. However, the protein showed highest activity in the presence of Mg 2ϩ followed by Co 2ϩ and Zn 2ϩ (Fig. 4A) suggesting some difference between the nucleotidase and polyphosphatase active sites. With polyphosphate, the E. coli SurE also showed saturation kinetics but with sigmoidal saturation curve (Fig.  4B), instead of classic hyperbolic curve, with a Hill coefficient n H of 1.86 Ϯ 0.41 indicating positive cooperativity in polyphosphate binding.
The kinetic parameters of the E. coli SurE with polyP 12-13 as a substrate are presented in Table I. The affinity of the E. coli SurE to polyP [12][13] was slightly lower than that (K m ϭ 3.9 M) of the yeast exopolyphosphatase (24). Two other E. coli proteins with exopolyphosphatase activity have been already purified and characterized, the principal exopolyphosphatase PPX and the second exopolyphosphatase GPP, a guanosine pentaphosphate phosphohydrolase, which also possesses an exopolyphosphatase activity (25,26). Both proteins showed high activity and affinity to the long-chain polyphosphate substrate polyP 500 . In contrast to these proteins, the E. coli SurE exhibited a clear preference for the short chain-length substrates with highest activity toward the PolyP 20 -25 substrate (Fig. 4C).
E. coli YfbR-The E. coli YfbR is a hypothetical protein with homologues in bacteria, archaea, eukaryotes, and humans. In general phosphatase screens, the purified E. coli YfbR demonstrated high catalytic activity toward pNPP. The protein had a slightly alkaline pH optimum (pH 8.0) and an absolute requirement in a divalent metal cation for activity (Fig. 5A). Co 2ϩ was the most efficient metal followed by Mn 2ϩ and Cu 2ϩ . YfbR showed relatively low affinity to Co 2ϩ (apparent K D ϭ 0.8 Ϯ 0.05 mM) and to pNPP (K m ϭ 2.09 Ϯ 0.13 mM) (Tables I and II). Screens with natural phosphatase substrates identified the presence of phosphatase activity in the E. coli YfbR against canonical deoxyribonucleoside 5Ј-monophosphates (Fig. 5B). The protein showed strict specificity toward deoxyribonucleoside 5Ј-monophosphates and did not hydrolyze ribonucleotides or deoxyribonucleoside 3Ј-monophosphates. YfbR had similar levels of activity (less than 2-fold difference) against all tested deoxyribonucleoside 5Ј-monophosphates (Fig. 5B). As with pNPP, the hydrolysis of deoxyribonucleoside monophosphates was strictly dependent on the presence of a divalent metal cation with Co 2ϩ being the most efficient metal. With natural substrates (5Ј-dAMP), YfbR exhibited 8-fold higher affinity to Co 2ϩ than with pNPP as a substrate (Table II). With all substrates, the protein showed saturation kinetics with highest affinity (and lowest activity) to dTMP, whereas highest activity was observed with dAMP (Table I). Similar to eukaryotic low K m 5Ј-nucleotidases (27,28), the K m values for all natural substrates were in the micromolar range (Table I). Like the Helicobacter pylori 5Ј-nucleotidase HppA (29), the E. coli YfbR hydrolyzed natural substrates with lower rate than that for pNPP. However, the affinity to natural substrates was significantly higher (Table I). Similar rates of catalytic activities were also reported for 5Ј-nucleotidases from pig thyroid and from mouse liver (30,31).
E. coli YjjG-The E. coli YjjG is annotated as a hypothetical protein with many homologues in different bacteria, archaea, and eukaryotes. Purified E. coli YjjG demonstrated high catalytic activity in general phosphatase screens with pNPP as a substrate (data not shown). The activity had a neutral pH optimum (pH 7.5) and a strict metal requirement (Mg 2ϩ Ͼ Mn 2ϩ Ͼ Co 2ϩ ) (Fig. 6A). The protein showed highest activity in the presence of Mg 2ϩ and had highest affinity to Co 2ϩ . Analysis of metal saturation curves revealed the presence of two plateaus suggesting the presence of two metal binding sites with different affinity or that the protein is oligomeric in solution. Our gelfiltration experiments (data not shown) support the second suggestion. The high affinity site had highest affinity to Mn 2ϩ (K D ϭ 10 Ϯ 1.0 M) followed by Co 2ϩ (K D ϭ 16.2 Ϯ 1.4 M) and Mg 2ϩ (220 Ϯ 30.0 M), whereas the low affinity site showed highest affinity to Co 2ϩ (Co 2ϩ Ͼ Mn 2ϩ Ͼ Mg 2ϩ ). The affinity of YjjG to pNPP was significantly lower than that of SurE or YfbR (Table I).
When screened with natural phosphatase substrates, the E. coli YjjG demonstrated high phosphatase activity toward three nucleoside 5Ј-monophosphates, UMP, dUMP, and dTMP, and low activity against TDP, IMP, UDP, GMP, dGMP, AMP, dAMP, and 6-phosphogluconate (Fig. 6B). YjjG was strictly specific to substrates with 5Ј-phosphates and showed no activity against nucleoside 2Ј-or 3Ј-monophosphates. For the hydrolysis of natural substrates, the protein had an absolute metal requirement with Mn 2ϩ being the most effective metal for the hydrolysis of nucleotides and Ni 2ϩ for the hydrolysis of 6-phosphogluconate (Fig. 6C). Thus, YjjG showed different metal profiles for the hydrolysis of pNPP and nucleotides suggesting that different amino acid residues are involved in the binding of metals and substrates. With natural substrates, YjjG (like SurE and YfbR) showed increased affinity to a divalent metal cation (Table II). With all substrates, YjjG demonstrated saturation kinetics with similar levels of activity and affinity (Table I). In general, the catalytic activity and efficiency of YjjG were significantly higher that those of SurE and YfbR. The kinetic parameters of this protein place it into the "high K m " (millimolar range) group of 5Ј-nucleotidases.
Inhibition by Nucleotides-Eukaryotic nucleotidases can be activated or inhibited by nucleoside di-or triphosphates. For example, the cytosolic high K m 5Ј-nucleotidase (cN-II) from human placenta is activated by dATP, ATP, GTP, and ADP, whereas the human ecto-5Ј-nucleotidase (eN) and the cytosolic low K m 5Ј-nucleotidase from rat brain are inhibited by ATP and ADP (3,32). The three nucleotidases from E. coli differed in their sensitivities to various nucleoside di-and triphosphates (Table III). YjjG was insensitive to all checked ribo-and deoxyribonucleoside di-and triphosphates (Table III). SurE was inhibited by various ribo-or deoxyribonucleoside 5Ј-triphosphates but was insensitive to nucleoside diphosphates. The hydrolysis of 3Ј-AMP by SurE was inhibited in a dose-depend-ent manner by ATP with 50% reduction caused by 0.65 mM ATP (K i ϭ 24.1 M) (Fig. 7C). YfbR was inhibited by both ribo-and deoxyribonucleoside di-and triphosphates (Table III) and was equally sensitive to low concentrations of ATP (K i ϭ 1.65 M), dATP (K i ϭ 1.39 M), and ADP (K i ϭ 1.30 M) (Fig. 7, A and B). ATP acted as a competitive inhibitor of the hydrolysis of dAMP by YfbR and induced a significant increase in K m for this substrate (0.56 mM in the presence of 10 M ATP). The inhibi-   tion by ATP was released at high substrate concentrations (data not shown).
Two groups of eukaryotic nucleotidases (cN-II and cN-III; high K m nucleotidases) possess also phosphotransferase ac-tivity and can catalyze phosphotransfer from a nucleotide donor to nucleoside acceptor (33)(34)(35). However, no phosphotransferase activity was found in the E. coli SurE, YfbR, and YjjG using various substrates as phosphate donors and guanosine, cytidine, uridine, or thymidine as phosphate acceptors (data not shown). DISCUSSION We have discovered three nucleotidases in E. coli with different substrate specificities and affinities, metal cofactor pref-   (A and B) by ATP (B and C) or ADP (A). Reaction mixtures contained: C, 2.5 mM 3Ј-AMP, 0.05 mM Mn 2ϩ , 0.04 g of SurE; A and B, 0.5 mM dAMP, 0.5 mM Co 2ϩ , 7.5 g of YfbR. Experimental details were as described under "Experimental Procedures." erences, and sensitivities to ATP/ADP. SurE is a broad specificity 5Ј(3Ј)-nucleotidase and polyphosphatase, which can hydrolyze various ribo-and deoxyribonucleotides and use different divalent metal cations (Co 2ϩ , Mn 2ϩ , Mg 2ϩ , and Ni 2ϩ ) for activity. YfbR is a Co 2ϩ -dependent 5Ј-nucleotidase with a strict specificity to deoxyribonucleotides. YjjG is an Mn 2ϩ -dependent 5Ј-nucleotidase specific to 5Ј-UMP, 5Ј-dUMP, and 5Ј-dTMP. These proteins also exhibit different affinities to substrates and sensitivity to the inhibition by nucleoside di-and triphosphates. Database analysis (Swiss-Prot) of their protein sequences showed no evidence for the presence of the N-terminal signal peptide suggesting that these nucleotidases are intracellular proteins.
These three E. coli nucleotidases belong to different enzyme superfamilies with different conserved motifs: the SurE family with characteristic N-terminal DD motif, the HD domain proteins with a conserved HD motif, and HAD-like hydrolases with the N-terminal DXD motif. This suggests that these proteins use different catalytic mechanisms to catalyze the same reaction. The molecular mechanisms of this catalytic reaction have been characterized only for HAD-like phosphohydrolases (36,37), which cleave covalent bonds of phosphorylated substrates by nucleophilic attack of the motif I Asp on the phosphorus of the substrate, resulting in the formation of a phosphoenzyme intermediate. In HAD-like phosphohydrolases, the metal cofactor (usually Mg 2ϩ ) is involved in the coordination of the nucleophilic side-chain carboxylate of catalytic Asp and the phosphoryl group of the substrate (36,37). Recent structural studies of SurE proteins from T. maritima and P. aerophilum revealed the presence in the active site of mononuclear metal center chelated by several conserved residues, including the strictly conserved N-terminal DD pair (17,38,39). The threedimensional structure of the N-terminal fragment of the bifunctional Rel/Spo homologue (the HD domain protein) from Streptococcus dysgalactiae identified several conserved residues (His-53, His-77, Asp-78, and Asp-144) involved in the coordination of Mn 2ϩ (40). However, the molecular mechanisms of catalysis by SurE and HD domain proteins are still unknown.
Like all six intracellular eukaryotic nucleotidases, the E. coli YjjG is a member of the haloacid dehalogenase (HAD)-like hydrolase superfamily (10) that comprises a large superfamily of hydrolytic enzymes with over 2800 entries in EMBL-EBI data base. These proteins have low overall sequence similarity (Ͻ29%) that is centered on four short catalytic motifs. The vast majority of HAD-like hydrolases have unknown function, while members with known function catalyze one of five activities: haloacid dehalogenase, phosphonohydrolase, phosphatase, phosphoglucomutase, or ATPase. HAD-like phosphatases catalyze the metal-dependent dephosphorylation of small molecules (phosphoserine and phosphoglycolate) or proteins (transcription factor "Eyes absent" in Drosophila) (41)(42)(43). Several members of this sub-group have been characterized biochemically and structurally: phosphoserine phosphatase from human and from Methanococcus jannaschii (41,44), 3-deoxy-D-mannooctulosonate-8-phosphate phosphatase from E. coli and H. influenzae (45,46), phosphoglycolate phosphatase from Thermoplasma acidophilum (42), and human mitochondrial 5Јdeoxyribonucleotidase dNT2 (47). Like the E. coli YjjG, the human dNT2 preferentially dephosphorylated dUMP, dTMP, and UMP; however, these proteins share only 20.9% overall sequence identity.
The E. coli YfbR is a member of the HD domain superfamily of phosphohydrolases named after the conserved HD motif (48). This superfamily includes a variety of uncharacterized proteins and domains associated with nucleotidyltransferases and heli-cases from bacteria, archaea, and eukaryotes and is predicted to have a metal-dependent phosphohydrolase activity. To date, only three HD domain proteins have been characterized biochemically: E. coli dGTPase, the RelA proteins from E. coli and Streptomyces equisimilis, and the E. coli tRNA-nucleotidyltransferase. The E. coli dGTPase (EC 3.1.5.1) catalyzes the Mg 2ϩ -dependent hydrolysis of dGTP to deoxyguanosine and tripolyphosphate (49,50). The major role of the RelA proteins from E. coli and S. equisimilis is the breakdown of (p)ppGpp by a Mn 2ϩ -dependent (p)ppGpp pyrophosphohydrolase activity (51)(52)(53). Very recently, we demonstrated that the HD domain of the E. coli tRNA-nucleotidyltransferase has 2Ј,3Ј-cyclic phosphodiesterase, 2Ј-nucleotidase, and phosphatase activities (54). In this work, we showed that the E. coli YfbR is a Co 2ϩ -dependent 5Ј-deoxyribonucleotidase. By its substrate specificity, this protein shows some resemblance to the human cytosolic 5Ј(3Ј)-deoxyribonucleotidase dNT1, which dephosphorylated 5Ј-, 3Ј-, or 2Ј-deoxyribo-and ribonucleotides (4). However, the E. coli YfbR is the first nucleotidase that is strictly specific to 5Ј-deoxyribonucleotides and shows no activity against 5Ј-ribonucleotides or 3Ј-deoxyribonucleotides.
Our biochemical studies of the E. coli SurE and the previous data on two SurE proteins from the thermophilic bacterium Thermotoga maritima (17,38) and from the archaebacterium Pyrobaculum aerophilum (39) clearly demonstrated that the annotation of SurE proteins as acid phosphatases is not accurate. Acid phosphatases (EC 3.1.3.2) comprise a large group of nonspecific phosphohydrolases capable to hydrolyze a broad range of phosphorylated sugars, amino acids, nucleoside mono-, di-, and triphosphates (BRENDA data base, www.brenda.uni-koeln.de/). In contrast to nonspecific acid phosphatases, SurE proteins from E. coli, T. maritima, and P. aerophilum showed strict specificity to nucleoside 5Ј(3Ј)-monophosphates and, therefore, should be annotated as 5Ј(3Ј)-nucleotidases. The substrate specificity of the E. coli SurE (3Ј-AMP, 5Ј-dGMP, 5Ј-GMP, 3Ј-CMP, and 5Ј-AMP) shows no similarity to that of any known eukaryotic nucleotidase. Therefore, SurE proteins represent a new group of nucleotidases.
The physiological function of intracellular nucleotidases is not well understood. In eukaryotes, these enzymes, together with nucleoside kinases, regulate the cellular concentration of ribo-and deoxyribonucleoside monophosphates, and, therefore, control all ribo-and deoxyribonucleotide pools (1,2,55,56). Because the intracellular nucleotidases show overlapping substrate specificities, it is difficult to relate a particular enzyme to the maintenance of a specific nucleotide pool. Two deoxyribonucleotidases, dNT-1 and dNT-2, are especially active on thymidine and deoxyuridine monophosphates and are involved in the regulation of precursors for nuclear and mitochondrial DNA synthesis, respectively (2). They reverse the phosphorylation of thymidine and deoxyuridine made by cytosolic thymidine kinase TK1 and mitochondrial TK2, producing regulatory substrate cycles (57). The mammalian nucleotidase cN-IA was shown to operate on AMP and probably on dCMP (6,58,59), whereas cN-II on IMP and GMP (57,59).
In E. coli, several Nudix hydrolases (dGTPase, dATPase, dCTPase, and dTTPase), which produce the deoxynucleoside monophosphates from their corresponding triphosphates, have been biochemically characterized and proposed to be involved in the regulation of the nucleotide pools (60 -63). However, no information is presently available about the regulation of NTP/ dNTP pools in prokaryotes by substrate cycles. In E. coli, there are at least two intracellular nucleoside kinases (encoded by tdk and udk) that can phosphorylate thymidine, deoxyuridine, uridine, and cytidine (64, 65) and might be involved in the regulatory nucleotide substrate cycles. Our work identified three possible metabolic counterparts for these kinases in E. coli. The YjjG nucleotidase showed high specificity for the 5Ј-phosphates of thymidine, deoxyuridine, and uridine (Table I) and, therefore, might be involved in the pyrimidine nucleotide substrate cycles. When compared with human dNT-1 and dNT-2, the E. coli YfbR showed even more narrow specificity to deoxyribonucleoside 5Ј-monophosphates, suggesting that this protein might be involved in the regulation of all dNTP pools in E. coli. Similarly, the E. coli SurE might be involved in the regulation of dNTP and NTP pools, because this protein has a broad specificity to various deoxyribo-and ribonucleoside 5Јmonophosphates (Fig. 3A). The presence of a 3Ј-nucleotidase activity in SurE suggests that it might also be involved in the turnover of 3Ј-mononucleotides produced by numerous intracellular RNases (T1, T2, and F) during the degradation of various RNAs (BRENDA data base, www.brenda. uni-koeln.de/) (66 -68). Moreover, SurE is the first known polyphosphatase with a preference for short-chain-length substrates (Fig. 4C). Two other E. coli exopolyphosphatases (PPX and GPP) preferentially hydrolyze long-chain substrates (polyP 500 -750 ) (24,25). Therefore, three E. coli polyphosphatases (SurE, PPX, and GPP) complement each other, and, together, they can effectively degrade a wide range of intracellular polyphosphates. Previous genetic experiments showed that the E. coli SurE plays significant physiological role in stress response and is required for the survival in stationary growth phase (69,70). Our work demonstrated that the E. coli SurE is a multifunctional protein exhibiting 5Ј-nucleotidase, 3Ј-nucleotidase, and exopolyphosphatase activities and suggests that one or more of these activities are necessary for E. coli survival during the stationary phase.
Thus, our work established that, like eukaryotes, E. coli cells and evidently other prokaryotes produce multiple intracellular nucleotidases with different substrate specificities, with which they can dephosphorylate a wide range of 2Ј-, 3Ј-, and 5Јdeoxyribo-and ribonucleotides. Because the E. coli nucleotidases belong to different protein superfamilies, they are likely to use different catalytic mechanisms and, therefore, they represent an attractive model of how different catalytic motifs evolved to catalyze the same reaction. Future studies will be directed toward further screening and structural/functional characterization of new E. coli nucleotidases.