A New Family of Phosphotransferases with a P-loop Motif*

In most Gram-positive bacteria, catabolite repression is mediated by a bifunctional enzyme, the histidine-containing protein kinase/phosphatase (HprK/P). Based either on its primary sequence or on its recently solved three-dimensional structure, no straightforward homology with other known proteins was found. However, we showed here that HprK/P exhibits a restricted homology with an unrelated phosphotransferase, the phosphoenolpyruvate carboxykinase. This includes notably two consecutive Asp residues from the phosphoenolpyruvate carboxykinase active site, whose equivalent residues were mutated in Bacillus subtilis HprK/P. Characterization of the corresponding mutants emphasizes the crucial role of these Asp residues in the HprK/P functioning. Furthermore, superimposition of HprK/P and phosphoenolpyruvate carboxykinase active sites supports the view that both enzymes bear significant resemblance in their overall mechanism of functioning showing that these two enzymes constitute a new family of phosphotransferases.

In bacteria, carbon catabolite repression is a key regulatory mechanism allowing a rapid regulation of many gene expressions in response to the availability of metabolizable carbon sources (1). Hence, recent global gene expression analysis in Bacillus subtilis showed that about 10% of all genes are submitted to glucose regulation (2). In low GϩC Gram-positive bacteria, this phenomenon is essentially controlled by a reversible phosphorylation of the histidine-containing protein (HPr) 1 on its Ser-46 residue, which mediates its interaction with a transcriptional regulator, the catabolite control protein A (3). A HPr-like protein, Crh (for catabolite repression HPr), was discovered in B. subtilis, and it was shown to be involved as well in carbon catabolite repression by a similar phosphorylation of its Ser-46 residue (4). The long search for the enzyme responsible for HPr or Crh phosphorylation ended lately thanks to the B. subtilis sequencing program, with the identification of its encoding hprK gene (5,6). Genomic programs have since shown that this gene is present not only in most Gram-positive bac-teria but also in some pathogenic Gram-negative species. Because disruption of the hprK gene led to severe growth defects in Staphylococcus xylosus (7), the corresponding proteins in pathogenic bacteria might be potential targets for new antimicrobial therapies.
Biochemical characterization of the B. subtilis enzyme revealed that it is a homo-oligomer that displays positive cooperativity especially for the binding of its allosteric activator, the fructose 1,6-bisphosphate (8). Also, this enzyme is bifunctional, capable of a slow dephosphorylation of Ser(P)-HPr or Ser(P)-Crh, and has therefore been called HPr kinase/phosphatase (HprK/P) (9). Concerning its functioning at the molecular level, HprK/P appeared unrelated to the classical eukaryotic protein kinases since it does not possess the typical signatures found in this family (5,6). Instead, it contains a Walker A-motif (or P-loop) commonly found in ATPases or GTPases but also in some kinases such as adenylate kinase (10,11). The recently solved threedimensional structure of Lactobacillus casei HprK/P confirms that this loop adopts a conformation similar to that found in adenylate kinase, but it failed to identify any closer homologues based on a global comparison of structures (12).
Nevertheless, we show here that HprK/P bears a significant, although restricted structural homology with another apparently unrelated phosphotransferase, the phosphoenolpyruvate carboxykinase (PEPCK), which catalyzes the decarboxylation and phosphorylation of oxaloacetate to form phosphoenolpyruvate (13). This local structural homology is supported by biochemical evidence, indicating that both enzymes phosphorylate their respective substrate according to a similar catalytic mechanism. Overall, this suggests that both enzymes have evolved from a common ancestor.

MATERIALS AND METHODS
Site-directed Mutagenesis-Plasmid pAG4 (5) was used as a template for introducing point mutations into the hprK by PCR (14), and the presence of the correct mutations were verified by DNA sequencing.
Protein Purification-HprK/P wild-type, or mutants and the protein substrate Crh were purified as previously described (4,5).
Limited Trypsin Digestion-6 g of HprK/P from wild type or mutants were incubated in the presence of 40 mM NaCl, 10 mM Tris/HCl, pH 8.0, and 0.6 g of trypsin (Promega). The reaction mixture was incubated for 15 min at 37°C. The digestion was stopped by adding an equal volume of electrophoresis loading buffer to the assay mixtures and by heating for 5 min at 80°C before applying the samples to a 15% sodium dodecyl sulfate polyacrylamide electrophoresis gel.
* This work was supported by the CNRS Program Physique et Chimie du Vivant and the Ministère de la Recherche Actions Concertées Incitatives-jeunes-chercheurs (to A. G.). 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.
Surface Plasmon Resonance-The interaction between HprK/P wild type or mutants and Crh was monitored on a BIAcore TM 1000 upgraded biosensor system (Biacore AB, Uppsala, Sweden) after covalent coupling of Crh to the surface of CM5 sensorchip through its amino group and according to the manufacturer's instructions. HprK/P was injected at a flow rate of 20 l/min using a running buffer containing 20 mM Tris-HCl, pH 8.0, 5 mM MgCl 2 , 0.005% surfactant P-20.
Fluorescence Measurements-All experiments were performed using a Photon Technology International Quanta Master I spectrofluorometer as described previously (8). The N-methylanthraniloyl (mant) derivatives of nucleotides were purchased from Molecular Probes.

RESULTS AND DISCUSSION
Homology between HprK/P and PEPCK-While searching for putative critical residues in HprK/P, we noticed that close to the Walker A motif (GX 4 GKS), two consecutive Asp residues were absolutely conserved in all HprK/P sequences known to date (see Fig. 1A for a few representative sequences). We then checked if a similar pattern could be present in other proteins and, therefore, used the tool PattInProt available on the Internet at npsa-pbil.ibcp.fr (15) to scan different data bases with the following signature, G-X (4)-G-K-[TS]-X (10, 20)-D-D (using a Prosite syntax). Quite interestingly, this search picked up the well characterized PEPCK, which transforms the oxaloacetate into phosphoenolpyruvate by a process involving sequential decarboxylation and phosphorylation steps (13). Partial sequence alignment around the P-loop of both PEPCK and HprK/P revealed a local homology within a short stretch of residues close to the C-terminal part of both P-loops and notably including two consecutive Asp residues (Fig. 1A). Additionally, a His residue located a few residues before both P-loops could be aligned too, but whether this was fortuitous or not was unclear at that time. Biochemical experiments supported by the Escherichia coli enzyme three-dimensional structure show that these two Asp residues fulfil a crucial function in PEPCK as ligands of two different cations required for enzyme activity (13, 16 -18). Hence, the first Asp residue (Asp-268) interacts indirectly through a water molecule with the so-called metal nucleotide (Mg 2ϩ here) classically found in all nucleotide-binding proteins containing a P-loop (19). The second Asp residue (Asp-269) interacts both indirectly with the metal nucleotide through another water molecule and directly with a Mn 2ϩ ion bound to the active site (Fig. 1B). A direct role in the phosphoryl transfer reaction has been proposed for this second cation (18).
Role of Asp-176 and Asp-177 Residues in HprK/P-To analyze the role of these two conserved Asp residues in HprK/P, each of them was replaced by an Ala residue in the B. subtilis enzyme. After overexpression and purification of the corresponding mutant proteins as for the wild-type enzyme, sensitivity to limited trypsin digestion was first assessed. Fig. 2A shows that a similar pattern of digestion was obtained for the  (32). Blue frames were drawn when more than 50% of the residues were conserved, with white characters in red boxes for strict identity and red characters in yellow boxes for similarity. B, the three-dimensional structure of the corresponding part of the E. coli PEPCK was drawn using Weblab viewer. The structure of the bound ATP and the position of magnesium (blue) and manganese (green) ions are shown. The side chain of the two conserved Asp residues, highlighted in the partial alignment of panel A by pink triangles, Asp-268 and Asp-269, are also represented. Nter and Cter, N and C termini, respectively. two mutants as compared with that of the wild-type protein, although the D176A mutant appeared to be somewhat more resistant. Therefore, the overall structure of the protein is most likely not affected by the introduced mutations.
The kinetic properties of each mutant were then compared with those of the wild-type enzyme. The phosphatase activity was totally abolished for both mutants since neither of them was capable of hydrolyzing the phospho-O-Crh (Fig. 2B). Also, no kinase activity could be detected under standard conditions (i.e. using limited amount of enzyme to maintain an initial rate of phosphorylation) whether or not the allosteric activator FBP was present (Fig. 2C). When an excess amount of enzyme was used, a slight phosphorylation was detected for the D176A mutant but only in the presence of FBP (Fig. 2D). By contrast, the D177A mutant remained unable to phosphorylate Crh, even under these conditions. Altogether, these results indicate that the Asp-176 and Asp-177 residues are involved both in the kinase and phosphatase activities of HprK/P, and each mutation, especially the D177A, drastically alter these two enzymatic activities.
Binding of FBP and Substrates-We then verified that the absence of activity for each mutant was due either to a lack of effector or substrate binding or to a lack of cooperativity between the monomers otherwise required for an efficient catalysis (8). The fluorescence properties of the unique Trp residue in the B. subtilis HprK/P sequence (Trp-235), ideally located to sense the conformational modifications associated with the binding of either FBP or nucleotides, was used for this purpose. As previously shown (8), the binding of FBP to the wild-type HprK/P occurred with a positive cooperativity mechanism whose kinetic parameters are reported in Table I. They were essentially unaltered by each mutation, showing that the cooperativity between the enzyme monomers is conserved in the mutant proteins. Either ATP or GTP has been shown to be an efficient phosphate donor for the kinase activity catalyzed by HprK/P (6), and so the affinities for these nucleotides were measured for the two mutants. It was found that they were consistently higher for the D176A mutant as compared with those of the wild-type protein, especially that for the guanylic nucleotides (K D for GTP about 6-fold lower). By contrast, the D177A mutant exhibited roughly the same affinities for nucleotides as those found for the wild-type enzyme. Concerning the binding of the protein substrate Crh, the affinity for this substrate monitored by surface plasmon resonance was again slightly increased for the D176A mutant by about 3-fold, whereas it was unmodified in the D177A mutant as compared with the wild type (Table I). Overall, these results indicate that the lack of kinase activity in the two mutants is neither due to a strongly reduced affinity for one of the substrates, nucleotide or protein, nor to an altered cooperativity within homo-oligomers.
Superimposition of PEPCK and HprK/P Active-site Residues-The first three-dimensional structure was recently solved for the L. casei HprK/P, and the use of a global structural homology approach revealed only some similarity to cytidylate or adenylate kinase around the P-loop motif (12). Although PEPCK was not picked up by this search, Fig. 3A shows that there is a good superimposition of the two sequence stretches used in Fig. 1A. Hence, the structural organization of this part of each nucleotide-binding site folds in a similar way. The fold of the PEPCK nucleotide-binding site has been previously proposed to be unique (20) among the P-loop-containing proteins, which are classified in the SCOP data base (21). In particular, this was the first structure containing a P-loop where the core of the ␤-sheet was formed with a mixed antiparallel and parallel ␤-strands. This is exemplified here with the central ␤-strand following the His residue (in blue) antiparallel to its two flanking ␤-strands, the one preceding the P-loop (in the back) and the one ended by the two consecutive Asp residues (in the front). The same topology and ordering are found for the three ␤-strands of HprK/P, whereas the major difference observed between PEPCK and HprK/P partial structures concerns the ␣-helix that follows the P-loop. It is shorter in PEPCK so as to accommodate the shorter sequence between the P-loop and the two Asp residues (see Fig. 1A). These conserved Asp residues occupy the same spatial position in the two structures relative to the Lys residues of both P-loops, although the relative orientations of their lateral chains are somewhat different (Fig. 3A). It should be noted, however, that the HprK/P structure was solved with no bound nucleotide (12), and the nucleotide co-ordinates were added from the PEPCK structure. Quite interestingly, His residues preceding the Ploop and aligned in Fig. 1A are also superimposed in the two structures. In PEPCK, this residue interacts either with the ␥-phosphate of ATP in the structure solved in the presence of ATP and a Mg 2ϩ ion (22) or with Mn 2ϩ in the structure obtained in the presence of both ATP and Mg 2ϩ and Mn 2ϩ ions (18). Although binding of two cations as a requirement for kinase activity of HprK/P is not yet determined, the difference in the cation requirement for the binding of either nucleotide or the HPr substrate, the latter occurring in the absence of nucleotide, definitely points to this possibility. 2 Trp-253 Residue Is Very Near the Nucleotide-binding Site-Assuming that the ATP structure found in PEPCK is essentially the same in HprK/P, which seems reasonable at least for the triphosphate moiety for most P-loop containing proteins  2, 4, and 6, respectively, but the assay medium contained 10 mM FBP in addition. (19), the location of each HprK/P Trp residues might be inferred relative to the bound nucleotide. Within a monomer of L. casei, the unique Trp is located about 15 Å away from the base moiety of the nucleotide bound to this subunit (not shown). However, the structure of the L. casei enzyme revealed a hexameric association with three subunits on top of three others (12). Hence, the Trp residue of one of the neighboring subunits is possibly located at about 6.5 Å of the ribose of the nucleotide bound to the vicinal subunit (Fig. 3B). B. subtilis HprK/P also functions as a homo-oligomer (8), and each subunit contains a unique Trp residue (Trp-235) equivalent to the Trp residue of L. casei. Considering a single subunit, the Trp residue would be located about 15 Å away from the base of the nucleotide bound to the same subunit. However, considering the hexameric nature of the enzyme, a Trp residue from one subunit would be located at about 6.5 Å of the nucleotide bound to the vicinal subunit. Therefore, the modification of fluorescence of the Trp residue from one monomer is most likely affected by the binding of nucleotide to the neighboring monomer (distance of 6.5 Å) rather than the binding to the same monomer (distance of 15 Å) (Table I). In addition, we showed that binding of the mant-ADP or -GDP provoked a very high energy transfer between the mant moiety and the Trp residues (8). Substituting a mant moiety to the 2Ј (or 3Ј) of the ribose in Fig. 3B would bring the latter even closer to the Trp residue than 6.5 Å (not shown), which would explain the high energy transfer observed previously.
Perturbation in the Nucleotide Surroundings of the D176A Mutant-Binding of fluorescent nucleotide analogues, mant-ATP or -GTP, were also measured with the two mutants, especially since these analogues were both shown previously to efficiently replace their respective unmodified nucleotides in phosphorylation experiments (8). The additions of increasing concentrations of mant-GTP to the wild-type enzyme progressively quenched the Trp fluorescence emission peak (centered around 340 nm). It gave rise concomitantly to a new peak of fluorescence (centered around 430 nm), revealing a fluorescence resonance energy transfer between Trp residues and the mant moiety (Fig. 4, panel A). In this case too, the height of this new peak shows that there is a very high efficiency of energy transfer, similar to that observed previously with mant-ADP. When the same experiment was performed in the presence of equimolar amounts of Mg 2ϩ , a similar efficiency of energy transfer was reached with the wild-type enzyme (Fig. 4, panel  D). However, much lower concentrations of mant-GTP were required in this case to saturate the nucleotide-binding site, as indicated by the fitted K D values obtained from the plot of the fluorescence resonance energy transfer intensity in the function of the mant-GTP concentrations (8) (K D ϭ 12.28 Ϯ 0.79 M and 1.54 Ϯ 0.1 M in the absence and in the presence of Mg 2ϩ , respectively). The D177A mutant behaved likewise, with an affinity for mant-GTP much higher in the presence as compared with in the absence of Mg 2ϩ (K D ϭ 8.53 Ϯ 1.28 M and 0.68 Ϯ 0.07 M, respectively) and with efficiencies of energy transfers similar to those observed for the wild-type enzyme (panels C and F). At the opposite, a quite spectacular effect was seen on the efficiencies of energy transfer with the D176A mutant. In the absence of Mg 2ϩ (panel B), a much higher energy transfer was seen with this mutant as compared with the wild type, whereas a reversed effect was obtained in the presence of Mg 2ϩ , as the energy transfer with the mutant became considerably attenuated (panel E). Accordingly and in comparison to the wild-type enzyme, the maximal quenching of the Trp fluorescence for the D176A mutant was higher in the absence of Mg 2ϩ and lower in its presence. However, the affin-TABLE I Binding of FBP, nucleotides, and Crh to HprK/P wild type or mutants The binding of FBP and nucleotides was monitored by the increased of fluorescence of the Trp residues, as described previously (8). The binding of Crh was monitored by using the surface plasmon resonance approach (see "Materials and Methods"). Sensorgrams were analysed using BIAevaluation 2.0 software, and kinetics constants were obtained by fitting curves to a single-site binding model (A ϩ B ϭ AB). The quality of fit was assessed by inspecting the 2 values (Ͻ2) and the random distribution of the residuals.  3. Superimposition of the three-dimensional structures of the partial nucleotide-binding site of PEPCK and HprK/P. A, E. coli PEPCK structure (dark gray) was taken from Fig. 1B and that of L. casei HprK/P (olive green) was obtained from the structure recently solved (12). Three-dimensional structures were drawn using Weblab viewer with the residues from E. coli PEPCK (His-232, Lys-254, Asp-268, and Asp-269), colored in blue, and residues from HprK/P (His-140, Lys-161, Asp-178, and Asp-179), colored in green. B, some residues of L. casei HprK/P partial structure are highlighted: Ser-162 residue of the P-loop, which is H-bonded to the Asp-178 residue (numbering of L. casei, which corresponds to Asp-176 of B. subtilis), and the Trp-237 residue from a neighboring subunit from the hexameric structure. The residues and ATP are colored by element, with the color coding for the different atoms as follows: gray, carbon gray; blue, nitrogen; red, oxygen; yellow, phosphate. The structure of ATP is taken from the PEPCK structure shown in panel A. Nter and Cter, N and C termini, respectively.
ities for mant-GTP were not markedly altered in this mutant, since its fitted K D values were 4.9 Ϯ 0.47 M and 0.85 Ϯ 0.22 M in the absence or in the presence of Mg 2ϩ , respectively. Besides, similar effects were observed using the mant-ATP analogue instead, although the level of transfer was lower for each enzyme in the absence of Mg 2ϩ , especially for the D176A mutant (not shown). Only equimolar concentrations of Mg 2ϩ could be used here since it was shown previously that using an excess of Mg 2ϩ over the nucleotide concentration provoked a progressive precipitation of the enzyme, which precluded fluorescence titrations on the long run (8). Nonetheless, adding 1 mM Mg 2ϩ in excess of the highest concentration of mant-GTP used in Fig. 4 (panels D-F) did not change significantly the efficiency of energy transfer for the three enzymes. This suggests that the effect observed with the D176A mutant in the presence of Mg 2ϩ is not due to a reduced affinity for this cation. Rather, the binding of nucleotides or analogues might be slightly modified consequent to the mutation. It is noteworthy that a point mutation of the equivalent Asp residue into an Asn residue in E. coli PEPCK somehow uncoupled the two enzyme activities; it eliminated virtually all the phosphorylation of the intermediate, enolate of pyruvate, thereby unveiling an oxaloacetate decarboxylase activity. The latter activity was nevertheless stimulated by the addition of ATP, showing that this mutant still bound this nucleotide but in a distorted conformation unable to promote the phosphorylating activity (23). In the D176A HprK/P mutant, the altered conformation of the bound nucleotide might also explain the dramatic effect seen on the energy transfer efficiencies given the spatial proximity of the Trp residue and the mant moiety. For distances between two chromophores shorter than 10 Å, the Förster theory (dipoledipole interaction) is no longer valid to explain the mechanism of energy transfer (24,25), and instead a Dexter mechanism is involved (electron-exchange interaction) (26). In this latter case, although no useful relationship is known to correlate the transfer efficiency to the distance separating two chromophores (27), the rate of transfer is strongly influenced by this distance, since it decreases by approximately 1 order of magnitude for a 1.3-Å increment in separation (26). Additionally, the transfer efficiencies might also be affected by the relative orientation of the two chromophores (27). Therefore, a subtle modification in distance and/or orientation between the Trp residue and the mant moiety caused by the D176A mutation are sufficient to account for the perturbations observed in the energy transfer efficiencies. The equivalent Asp residue (Asp-178) in L. casei interacts with the Ser-162 residue of the P-loop through an H-bond (Fig. 3B). Such an interaction is observed in all nucleotide-binding proteins bearing a P-loop except adenylate kinase; this interaction occurs between the equivalent P-loop residue (always a Ser or a Thr) and an Asp residue that interacts with the metal nucleotide and is, therefore, functionally equivalent to the Asp-178 residue (19). In B. subtilis, the D176A mutation will prevent the formation of this H-bond, and one might tentatively propose that this would slightly distort the nucleotide-binding site. This would thus lead to an altered conformation of the bound nucleotide precluding the phosphorylation of the protein substrate.
Asp-177 Residue Is the Catalytic Base-Regarding the Asp-177 residue and given that its substitution by an Ala residue totally abolishes the substrate phosphorylation, it might participate directly as a catalytic base to abstract a proton from the attacking hydroxyl group of the Ser-46 residue of HPr (or Crh). A similar catalytic role has been proposed for the Asp-37 residue of the chloramphenicol phosphotransferase, another Ploop-containing phosphotransferase (28), and it is noteworthy that in both enzyme structures these Asp residues occupy a similar spatial position (not shown).
The postulated role for these two Asp residues agrees well with the very recently solved three-dimensional structure of a complex between L. casei HprK/P and the protein substrate HPr, obtained in the presence of Ca 2ϩ ions. 3 First, this structure reveals that the Asp-178 residue (i.e. Asp-176 in B. subtilis) interacts with two residues directly liganded to the Ca 2ϩ ion, the Ser-162 residue of the P-loop and the Glu-204 residue. This Ca 2ϩ ion occupies the same position as the Mg 2ϩ ion in other P-loop-containing proteins, for instance in the PEPCK, both cations usually behaving as similar co-factors for most of these proteins (29,30). Thus, mutation of this Asp residue would destabilize the correct network of interactions with the bound cation. Second, the Asp-179 residue (i.e. Asp-177 in B. subtilis) is hydrogen-bound to the hydroxyl hydrogen of the Ser-46 residue of HPr, which makes this Asp residue the ideal candidate to deprotonate this Ser residue before phosphorylation proceeds.
The two Asp residues highlighted in this study also play a critical role in the dephosphorylation process, suggesting that both the kinase and phosphatase activities occur at the same catalytic site and that the phosphatase activity would probably function as a reversal mechanism of the phosphorylation pathway.
In conclusion, although HprK/P and PEPCK catalyze different overall reactions, their nucleotide-binding sites share a similar topology. This allows a conservation of some crucial active site residues that are likely to play a similar role in the catalytic pathway of each enzyme. Altogether, this suggests that these two enzymes have evolved from a common ancestor. Interestingly, both enzymes also appear to be able to dephosphorylate their respective substrates, since in the conditions used for the crystallization, the PEPCK enzyme has been reported to form ATP and pyruvate from ADP and phosphoenolpyruvate (18).