Regulation of the Activity of Lactate Dehydrogenases from Four Lactic Acid Bacteria*

Background: Lactate dehydrogenases (LDHs) are key metabolic enzymes in lactic acid bacteria (LAB). Results: The effects of fructose 1,6-bisphosphate, phosphate, pH, and ionic strength on enzyme activity differ for six LDHs from four LAB. Conclusion: The regulation of LDH activity differs among LAB. Significance: These results have implications for understanding enzyme evolutionary adaptation, for quantitative comparative modeling, and for biotechnological application of LAB. Despite high similarity in sequence and catalytic properties, the l-lactate dehydrogenases (LDHs) in lactic acid bacteria (LAB) display differences in their regulation that may arise from their adaptation to different habitats. We combined experimental and computational approaches to investigate the effects of fructose 1,6-bisphosphate (FBP), phosphate (Pi), and ionic strength (NaCl concentration) on six LDHs from four LABs studied at pH 6 and pH 7. We found that 1) the extent of activation by FBP (Kact) differs. Lactobacillus plantarum LDH is not regulated by FBP, but the other LDHs are activated with increasing sensitivity in the following order: Enterococcus faecalis LDH2 ≤ Lactococcus lactis LDH2 < E. faecalis LDH1 < L. lactis LDH1 ≤ Streptococcus pyogenes LDH. This trend reflects the electrostatic properties in the allosteric binding site of the LDH enzymes. 2) For L. plantarum, S. pyogenes, and E. faecalis, the effects of Pi are distinguishable from the effect of changing ionic strength by adding NaCl. 3) Addition of Pi inhibits E. faecalis LDH2, whereas in the absence of FBP, Pi is an activator of S. pyogenes LDH, E. faecalis LDH1, and L. lactis LDH1 and LDH2 at pH 6. These effects can be interpreted by considering the computed binding affinities of Pi to the catalytic and allosteric binding sites of the enzymes modeled in protonation states corresponding to pH 6 and pH 7. Overall, the results show a subtle interplay among the effects of Pi, FBP, and pH that results in different regulatory effects on the LDHs of different LABs.

Despite high similarity in sequence and catalytic properties, the L-lactate dehydrogenases (LDHs) in lactic acid bacteria (LAB) display differences in their regulation that may arise from their adaptation to different habitats. We combined experimental and computational approaches to investigate the effects of fructose 1,6-bisphosphate (FBP), phosphate (P i ), and ionic strength (NaCl concentration) on six LDHs from four LABs studied at pH 6 and pH 7. We found that 1) the extent of activation by FBP (K act ) differs. Lactobacillus plantarum LDH is not regulated by FBP, but the other LDHs are activated with increasing sensitivity in the following order: Enterococcus faecalis LDH2 < Lactococcus lactis LDH2 < E. faecalis LDH1 < L. lactis LDH1 < Streptococcus pyogenes LDH. This trend reflects the electrostatic properties in the allosteric binding site of the LDH enzymes. 2) For L. plantarum, S. pyogenes, and E. faecalis, the effects of P i are distinguishable from the effect of changing ionic strength by adding NaCl. 3) Addition of P i inhibits E. faecalis LDH2, whereas in the absence of FBP, P i is an activator of S. pyogenes LDH, E. faecalis LDH1, and L. lactis LDH1 and LDH2 at pH 6. These effects can be interpreted by considering the computed binding affinities of P i to the catalytic and allosteric binding sites of the enzymes modeled in proto-nation states corresponding to pH 6 and pH 7. Overall, the results show a subtle interplay among the effects of P i , FBP, and pH that results in different regulatory effects on the LDHs of different LABs.
Lactobacillus plantarum, Streptococcus pyogenes, Enterococcus faecalis, and Lactococcus lactis are Gram-positive microorganisms belonging to the phylogenetic order Lactobacillales. These bacteria have their natural habitats in rather different environments (L. lactis, milk; L. plantarum, plants; E. faecalis, feces; S. pyogenes, skin and mucosal membranes) and interact differently with human beings. L. lactis and L. plantarum are bacteria of major importance for use in the food industry. E. faecalis is an important commensal of the human gut, a food contaminant, and a facultative pathogen. S. pyogenes is an exclusively human pathogen causing diseases like tonsillitis, pharyngitis, scarlet fever, and necrotizing fasciitis (1)(2)(3)(4)(5). Despite their different lifestyles, all four species predominantly gain energy by homolactic acid fermentation. The free energy generated during homolactic acid fermentation is 2 mol of ATP/1 mol of glucose. The crucial enzyme in this pathway is LDH, 4 which is responsible for catalyzing the reversible reduction of pyruvate to lactate. This reaction serves solely to balance the redox potential by oxidation of NADH ϩ H ϩ to NAD ϩ .
The LAB genomes differ in the number and type of LDH enzymes that they encode. The L. plantarum strain WCFS1 genome encodes at least two LDHs catalyzing the production of L-lactate (lp_0537, L-LDH1 and lp_1101, L-LDH2) and one producing D-lactate (lp_2057, D-LDH) (6). (The LDH locus and corresponding protein short names as they appear in the Uni-Prot database are given in parentheses.) The genome of the S. pyogenes M49 strain NZ131 encodes for one L-LDH (spy49_0904c, L-LDH) and one D-LDH (spy49_0919, D-LDH) (7,8). In the E. faecalis strain V583 genome, two L-lactate dehydrogenases (ef_0255, L-LDH1 and ef_0641, L-LDH2) are encoded (9,10). The E. faecalis genome additionally encodes a D-LDH involved in vancomycin resistance by incorporation of D-lactate instead of D-alanine in the cell wall (9,11). The L. lactis strain MG1363 genome possesses three L-LDH genes (llmg_1120, L-LDH1; llmg_0392, L-LDH2; and llmg_1429, L-LDH) (12). This study is restricted to L-LDH enzymes, which we will refer to simply as LDH.
The activation of L-LDHs extracted from L. lactis and E. faecalis was studied by Crow and Pritchard (13), but they did not distinguish between the two types of enzyme in each organism. Gaspar et al. (14) later characterized the kinetics of the individual L. lactis LDH1 and LDH2 isoenzymes. However, knowledge of the kinetics and allosteric regulation of the respective isoenzymes from L. plantarum, S. pyogenes, and E. faecalis remains lacking.
The crystal structure of a bacterial LDH is shown in Fig. 1. The biological assembly is a homotetramer. Each monomer has one active site, and the tetramer has two allosteric sites, each situated at the interface between two monomers. Fructose 1,6bisphosphate (FBP) has been shown to allosterically regulate some LDHs, including L. lactis LDHs, and to bind at these allosteric sites. The mechanism of LDH regulation by FBP has been defined to be allosteric because of a sequential intersubunit rearrangement of the LDH tetramer accompanied by local intrasubunit conformational changes (15,16). However, it has also been shown that the L. plantarum LDH1 is not allosterically regulated, and this has been ascribed to the presence of an aspartic acid residue in the allosteric site (17).
To provide a detailed understanding of the kinetics and regulation of LAB LDHs, we undertook a study combining both experimental and computational approaches. We determined the kinetic parameters and the allosteric regulation of the heterologously expressed LDH isoenzymes from L. plantarum, S. pyogenes, and E. faecalis at pH 6 and pH 7 in the presence of varying concentrations of FBP, inorganic phosphate (P i ), and sodium chloride (NaCl). Structural models of the LDH enzymes and comparative computational analyses of their binding properties were used to interpret the experimental data. This analysis allowed us to propose a unified model of the regulatory mechanisms of LAB LDHs.

EXPERIMENTAL PROCEDURES
Strains and Culture Conditions-Escherichia coli DH5␣ strains harboring recombinant plasmid DNA were cultivated in lysogeny broth (LB) medium containing 100 mg/liter ampicillin at 37°C in shaking cultures. S. pyogenes M49 591, E. faecalis V583, and L. plantarum WCFS1 strains were grown in Todd Hewitt broth supplemented with 0.5% yeast extract or brainheart infusion medium at 37°C as standing cultures.
Construction of Recombinant Plasmids-The chromosomal DNA of E. faecalis V583 and L. plantarum WCFS1 was isolated according to the Qiagen Blood and Tissue kit (Qiagen, Hilden, Germany) and used as the template for PCR amplification of the ldh genes with the Phusion TM High Fidelity PCR kit (Finzymes). The resulting PCR fragments were ligated into the pASK-IBA2 vector (IBA GmbH, Göttingen, Germany) system via BamHI and SalI restriction sites. The recombinant vectors were transformed in E. coli DH5␣ cells. Correct insertion of the PCR products was confirmed by plasmid sequencing. The construction of the expression plasmid for the S. pyogenes M49 591 ldh gene has been described previously (18).
Expression and Purification of Proteins-For heterologous expression of the isoenzymes, recombinant E. coli strains were grown in 500 ml of LB medium at 37°C under vigorous shaking. At an optical density of about 0.4, expression was induced by addition of anhydrotetracycline (0.2 g/ml). Cells were harvested after overnight shaking at 22°C, and pellets were stored at Ϫ80°C. For purification of the Strep-tagged proteins, cell pellets (from 500 ml of culture) were thawed and suspended in 8 ml of buffer W (100 mM Tris-HCl, pH 8.0, 1 mM EDTA, 150 mM NaCl), and cell disruption was achieved by the FastPrep method with acid-washed glass beads. Five cycles of 30 s each at a speed of 6.0 m⅐s Ϫ1 were applied. In between, samples were cooled on ice for at least 2 min. Cell debris was removed by centrifugation. Clear supernatants were loaded on Strep-Tactin-Sepharose columns (5 ml volume) prewashed with buffer W (3 times column volume). Unbound proteins were washed from the column with buffer W (5-7 times column volume). The protein carrying a Strep tag was eluted from the column in fractions of 0.5 ml each in buffer E. Buffer E is buffer W supplemented with 2.5 mM desthiobiotin. Elution fractions were checked for purified protein with SDS-PAGE. Fractions containing pure protein were pooled and dialyzed overnight in 100 mM MES buffer to prepare them for enzymatic assays.
Protein Concentration Measurements-The protein concentrations in the purified recombinant isoenzyme fractions were determined by the Bradford method using the Bio-Rad Protein Assay kit.
Enzymatic Assays-The LDH activity was determined in the forward and reverse reactions by measuring the rate of NADH oxidation for the forward reaction and the rate of NAD reduction for the reverse reaction at 380 nm using an extinction coefficient for NADH of 1.244 mM Ϫ1 cm Ϫ1 . All assays were carried out at 30°C in 60-l reaction volumes in 384well format plates, and the time course of NADH absorbance was monitored with a BMGLabtech NOVOstar plate reader (automated fluorescence/fluorescence polarization/absorbance reader; BMGLabtech, Offenburg, Germany). The assays were automated so that all the reagents in the reaction buffer were in 45 l, the enzyme was in 5 l, and the substrate was in 10 l. The assay setup was similar to that used previously (19). Full sets of experiments could be run simultaneously for each isoenzyme with its two substrates and at two pH values (pH 6 and pH 7) for the forward and reverse reactions, allowing quick production and reliable comparison of data. For measuring the activity of LDH, a standard assay mixture of 100 mM MES/ KOH, pH 7, 5 mM MgCl 2 , 3 mM FBP, 1 mM NADH was used and incubated with a suitable concentration of the enzyme at 30°C. Then 10 mM pyruvate was added to start the reaction. For determining the kinetic parameters of the different isoenzymes at pH 6 and pH 7 for the forward and reverse reactions and the effect of allosteric activators, the following measurements were conducted: 1) the kinetic parameters in the absence and presence of FBP at a fixed concentration of NADH/NAD using various concentrations of pyruvate/lactate for the K m of pyruvate/ lactate and at a fixed concentration of pyruvate/lactate using various concentrations of NADH/NAD for the K m of NADH/ NAD and 2) the effect of FBP, P i , and ionic strength (NaCl) on the LDH activity by conducting assays at substrate concentrations equal to the K m values and not at saturated concentrations. K act values were determined whenever an allosteric activation was found by measuring the activity of the enzyme at various concentrations of the activator. The effect of activators was also checked in the absence/presence of other activators. To determine whether or not the activation by P i is solely due to the change in ionic strength, measurements were made with different concentrations of NaCl.
All measurements were based on at least duplicate determinations of the reaction rates, and for all assays, control experiments were run in parallel to check and correct for any unwanted background activity. The data obtained were analyzed with KineticsWizard software (20). Data fitting was performed with the default settings according to the Michaelis-Menten equation.
Comparative Modeling-Modeling of the structures of the LDHs from L. plantarum (UniProt identifier P56512), S. pyogenes (P56259), E. faecalis (Q839C1 and Q838C9 for LDH1 and LDH2, respectively), and L. lactis (Q01462 and Q9CII4 for LDH1 and LDH2, respectively) was performed with MODELLER8v.2 (21). 5 The crystal structure of the LDH from Bacillus stearothermophilus (Protein Data Bank code 1LDN) determined (22) at 2.5-Å resolution and shown in Fig. 1 was used as a template. The sequence alignment of the target LDHs shows a high sim- 5 Although the selected Q01462 sequence for modeling L. lactis LDH1 is from strain IL1403, it shows 100% identity to the LDH1 from L. lactis strain MG1363. The sequence Q9CII4 (strain IL1403) for modeling L. lactis LDH2 has 97% identity to the LDH2 from L. lactis strain MG1363; there are no differences in charged residues within 10 Å of the FBP or the pyruvate binding sites.

TABLE 1 Kinetic parameters measured for the LAB LDH enzymes
Assays were carried out in a buffer consisting of 100 mM MES, 5 mM MgCl 2 at pH 6 and pH 7 at 30°C. The kinetic parameters were measured in the absence and presence of 3 mM FBP. All components in the reaction mixture were incubated at 30°C for 5 min before starting the reaction. Control experiments were run in parallel. Kinetic parameters were determined as described under "Experimental Procedures." The error value is defined as a percentage (%) of the measured value. n/a, not applicable as L. plantarum LDH1 is not FBP-regulated; N/A, not applicable as measurement was not possible. -, no measurements were performed.

Regulation of Lactate Dehydrogenase Activity
ilarity of approximately 70% to the template (supplemental Fig.  S1). Although the 1LDN structure is a homotetramer, only two monomers with an interfacial allosteric binding site for FBP were selected for the template. This homodimer contains one complete allosteric site and two complete active sites. Thus, for the purposes of this study, all LDHs were modeled in their homodimeric form. To ensure that differences in the modeled structures stemmed from their sequence only, we applied the protocol described by Gabdoulline et al. (23). This protocol uses the "automodel.very_fast" settings of MODELLER8v.2 and differs from the default procedure by not randomizing the initial coordinates and by performing a maximum of 50 steps of energy minimization without molecular dynamics refinement. After modeling, the LDH structures were resuperimposed on the template. Computation of Protein Electrostatic Potentials-Polar hydrogen atoms were added at pH 7 to the protein structures by optimizing the hydrogen bond network with the WHAT IF 5.0 program (24). Electrostatic potentials were computed with the University of Houston Brownian Dynamics program (25). Partial atomic charges and atomic radii were assigned from the optimized potential for liquid simulations (OPLS) parameter set (26). Relative dielectric constants of 2 and 78 were assigned to the protein interior and solvent, respectively. Potentials were computed at an ionic strength of 150 mM on a cubic grid of 110 3 points with a 1-Å spacing. To obtain the electrostatic potential at pH 6, His-171 in the allosteric binding site and His-178 in the catalytic binding site were doubly protonated in all proteins except for L. plantarum LDH1. In L. plantarum LDH1, only the catalytic His-178 was doubly protonated as residue 171 is an aspartic acid. Here, we use the sequence numbering of L. lactis LDH1 for all LABs (supplemental Fig. S1).
Comparison of the Allosteric Binding Sites-The Protein Interaction Property Similarity Analysis (PIPSA) tool allows comparison of protein interaction fields (in this work electrostatic potentials) in the intersection of the "skins" of two superimposed structures (27). The "skin" region is defined as the volume remaining after exclusion of the region within the protein surface accessible to a probe of radius ϭ 3 Å and the region outside the protein surface accessible to a probe of radius ϩ ␦ (␦ ϭ 4 Å). The skin thus has a thickness of ␦. The comparison was restricted to the skin within a sphere of 15-Å radius centered on the geometric center of FBP as positioned in the allosteric binding site of the crystal structure 1LDN.
To compare electrostatic potentials (⌽), pairwise Hodgkin similarity indices (SI a,b ) were computed. For two proteins, a and b, SI a,b is given by . SI a,b ϭ 1 if the potentials are identical, SI a,b ϭ 0, if they are uncorrelated, and SI a,b ϭ Ϫ1 if they are anticorrelated. Corresponding pairwise electrostatic distances were defined as D a,b ϭ ͱ ͑2 Ϫ 2SI a,b ͒.

Computation of Energetically Favorable Binding Sites for
Phosphate and Carboxylate Groups-GRID software (29) was used to identify energetically favorable binding sites for P i and carboxylate (COO) probes in the catalytic and allosteric sites. The P i probe was used to investigate binding of phosphate ions and FBP, whereas the COO probe was used for pyruvate. The binding of these probes in the catalytic site was assessed to examine the competition between P i and pyruvate in the absence and presence of FBP and the consequent impact on the enzyme activity. Calculations were performed for the unliganded LDH models as well as for the LDHs with FBP positioned in the allosteric binding sites as in the template 1LDN structure. The computation of binding energies (E) was carried out at both pH 6 and pH 7.
Algorithm to Computationally Estimate the Effect of P i -To estimate whether P i has an activatory or an inhibitory effect on the enzymes, the computed probe binding energies were compared with those for the LDH from L. plantarum whose activity is known to be unaffected by P i . The binding energies of the P i probe in the allosteric binding site (AS) and the COO probe in the catalytic binding site (CS) of LDH from L. plantarum were FIGURE 1. The crystal structure of the LDH from B. stearothermophilus (22). The homotetrameric quaternary structure and the homodimeric structure, which was used as a template for modeling the LAB LDH structures, are shown in A and B, respectively. The catalytic and allosteric binding sites are indicated. The axes along the allosteric sites and the active sites are defined as P and Q axes, respectively (31). FBP molecules are shown in sphere representation in the allosteric binding sites. NAD, shown in orange, and oxamate (a pyruvate analog), shown in blue, are bound in the catalytic binding sites.
defined as E AS,threshold and E CS,threshold , respectively. For the other LDH enzymes, the activity was considered to be enhanced by P i if the binding energy of the P i probe in the allosteric binding site (E Pi ) was ՅE AS,threshold . When the binding energy of P i in the catalytic site (E Pi ) was ՆE CS,threshold , no inhibition by P i was anticipated. The magnitude of activation and inhibition by P i was assessed by computing and analyzing the following energy differences: A strong enhancement of enzyme activity (designated by letter A) was assigned when ⌬E AS Յ Ϫ2 kcal/mol. The letter a designates weak activation, which was assigned when Ϫ2 Յ ⌬E AS Յ 0 kcal/mol. Strong inhibition (I) was assigned when both ⌬E CS Ͻ 0 and ⌬E CS pr Յ Ϫ2 kcal/mol. Weak inhibition (i) was assumed  when both ⌬E CS Ͻ 0 and Ϫ2 Յ ⌬E CS pr Յ 0 kcal/mol. For ⌬E CS Ն 0, no competition between P i and COO in the catalytic site and thus no inhibition were expected (¬I).
The following relations were used to deduce the overall effect of the presence of P i considering the possibility of P i binding at the allosteric site and the catalytic site: A∧¬I 3 A; a∧¬I 3 a; ▫∧¬I 3 ¬E (no effect); A∧i 3 a∨¬E; ▫∧i 3 i; A∧I 3 ¬E∨i; ▫∧I 3 I. Here, we use the common logical connectives (30) ∧ for "and," ∨ for "or," ¬ for negation "not," and ▫ for the cases when the allosteric binding site is occupied by FBP and thus not accessible for P i .

RESULTS AND DISCUSSION
Protein Expression and Purification-We heterologously expressed the LDH1 isoenzymes of L. plantarum WCFS1, S. pyogenes M49 591, and E. faecalis V583 and the LDH2 isoenzyme of E. faecalis V583 in E. coli DH5␣ using the pASK-IBA2 vector and purified the enzymes by affinity chromatography. The calculated molecular masses of the isoenzymes carrying the Strep tag were 39.1 kDa for the LDH1 of L. plantarum, 37.2 kDa for the LDH of S. pyogenes, 37.4 kDa for LDH1, and 36.33 kDa for the LDH2 of E. faecalis. The success of the overexpression and purification of the isoenzymes was checked by SDS-PAGE (supplemental Fig. S2).
Enzyme Kinetics-The purified recombinant proteins were produced to study the kinetics of the isoenzymes and to determine the Michaelis constants (K m ) for NADH, pyruvate, NAD, and lactate at pH 6 and pH 7. Furthermore, the allosteric regulation of the isoenzymes by P i and FBP and the effect of ionic strength (NaCl) were investigated. K act values were measured when applicable.
As shown in Table 1, for each of the four LDH enzymes for which measurements were made, the K m values for pyruvate, NADH, lactate, and NAD were similar at pH 6 and at pH 7. Similar results have been reported for the L. lactis LDH1 enzyme, whereas the L. lactis LDH2 was found to have higher affinity for its substrates at pH 6 than at pH 7 (14). As discussed in the next sections, the trends in activation/inhibition by P i of each enzyme except L. lactis LDH2 were also similar at pH 6 and pH 7.
Allosteric Mechanisms of LDH Activation-From crystallographic studies, LDH was determined to mainly form a tetramer with two FBP binding sites and four active sites (31,32). The axes along the FBP binding sites and the active sites were defined as P and Q axes, respectively (see Fig. 1). In the absence of FBP, two conserved positive residues, Arg and His, on the P axis interface can generate repulsion between subunits that in some bacterial organisms leads to a disassembly of the tetramer into dimers (16,33).
Iwata et al. (15) determined the crystal structure of a 1:1 mixture of the inactive T-state and active R-state of Bifidobacterium longum LDH (Protein Data Bank code 1LTH). From this crystal structure, they proposed the following mechanism of LDH activation triggered by FBP binding. Addition of FBP neutralizes the repulsion of two dimers and stabilizes the tetrameric oligomerization of LDHs by hydrogen bond formation to Arg-173 localized on the P axis interface. A different rotametric state of Arg-173 in the complex with FBP then initiates a conformational change affecting the active site. Oxamate binding in the active site seems to complete the allosteric rearrangement from the inactive to the active state of the enzyme. It is accompanied by a helix (57-73 residues) sliding at the Q axis interface and subsequent replacement of His-68 on the next subunit by Arg-171 localized in the active site. The activation of the enzyme is thus mediated by the quaternary structural change.
Cameron et al. (16) proposed another mechanism of LDH activation. They found the LDH from B. stearothermophilus to be in an equilibrium of dimeric and tetrameric oligomers (Protein Data Bank codes 1LDB and 1LDN) in which FBP binding promotes tetramerization. The crystallographic analysis revealed an allosteric mechanism that depended on intrasubunit conformational changes. In this model, upon FBP binding, Arg-173 rearranges such that a subsequent helix shift causes improved positioning of two adjacent charged residues, Arg-171 and Asp-168, in the active site of the same subunit, thereby affecting the substrate binding affinity.
Our LDH models are based on the tetrameric crystal structure of LDH from B. stearothermophilus (Protein Data Bank code 1LDN) with FBP, NAD ϩ , and oxamate bound, thereby imitating the activated R-form of the enzyme. Although the key residues that take part in the conformational rearrangement are conserved among the LAB LDHs (apart from L. plantarum), the allosteric mechanism by which conformational changes upon FBP or P i binding are propagated can be different. A detailed analysis of the LDH dynamics would be necessary to investigate the LAB LDH allosteric mechanisms. However, this is not the aim of the present work, which is focused on the question of whether and how the binding of FBP and P i in the allosteric and active sites (and therefore the regulation) varies among LDHs from a set of LABs that have adapted to different environments.
Activation by FBP-The LDH1 from L. plantarum has previously been described to be non-allosteric due to the presence of an aspartate residue in the effector binding site where the FBPregulated LDHs have a histidine (17). Consistently, the activity of the LDH1 from L. plantarum was unaffected by the absence or presence of FBP at both pH values tested. The experimental results on activation by FBP are given in Table 2 and supplemental Table S1. In contrast to the L. plantarum LDH1, the activity of the S. pyogenes LDH decreased to 1.2% at pH 6 and 0.6% at pH 7 in the absence of FBP compared with its activity in the presence of 3 mM FBP. The effect of FBP on the activity of the E. faecalis LDH enzymes was less pronounced than that for S. pyogenes. In the absence of FBP, the E. faecalis LDH1 activity (relative to that in the presence of 3 mM FBP) decreased to 12.7% at pH 6 and 19.4% at pH 7, and that of LDH2 only decreased to 30.8 and 37.3% at pH 6 and pH 7, respectively. Thus, for activation by FBP, the S. pyogenes LDH resembles the LDH1 from L. lactis (14), the E. faecalis LDH2 resembles the LDH2 from L. lactis, and the E. faecalis LDH shows an intermediate activation level.
FBP is a negatively charged molecule, and therefore, to assess whether FBP activation could be related to the ability of the enzyme to bind FBP, we computed and compared the electrostatic potentials of the six LDH enzymes. Our findings reveal a distinct electrostatic potential for LDH1 from L. plantarum at its allosteric binding site (Fig. 2). In contrast to the other LDH enzymes, which display positive patches at the regions complementary to the negatively charged phosphate moieties of FBP, the allosteric binding site in LDH1 from L. plantarum is rather negatively charged. The negative electrostatic potential in the allosteric binding site might hinder or alter the mode of FBP binding. This supports the experimental evidence that LDH1 from L. plantarum in contrast to the other target LDHs studied is not FBP-regulated. By comparison of the electrostatic potentials in the vicinity of the FBP binding site using PIPSA (see "Experimental Procedures"), we observed the highest similarity between L. lactis LDH2 and E. faecalis LDH2 with an SI of 0.999, showing that the potentials are almost identical (Fig. 3). The L. lactis LDH1, S. pyogenes LDH, and E. faecalis LDH1 enzymes demonstrated high similarity with pairwise SI values over 0.94. As anticipated, the electrostatic potential of L. plantarum LDH1 was the most distinct, revealing a lower similarity to other enzymes, with pairwise SI values of about 0.8 (Fig. 3).
Activation/Inactivation by P i -It has been noted previously (13,14) that the presence of phosphate can affect the activation of L. lactis LDH. We hypothesized that P i can affect LDH activity by binding at the allosteric site and/or at the catalytic site and that this binding can result in net activatory or inhibitory effects. We therefore tested experimentally and computationally the effects of P i on the LDH activity in the absence and presence of FBP. Energetically favorable positions for the P i probe in the allosteric binding site in the absence of FBP and in the catalytic binding site when FBP occupies the allosteric site are exemplified in Fig. 4. The experimental results are shown in Fig. 5, and the computational and experimental results are summarized and compared in Table 3. For experimentally measured absolute values, see supplemental Table S2. Here, we discuss each of the six LDH enzymes in turn in the order given in Fig. 5. As shown in Fig. 5A and consistent with the lack of FBP regulation, L. plantarum LDH1 activity is not significantly affected by the addition of P i at concentrations up to 200 mM. Therefore, the binding energies at which P i and COO probes (the latter representing the carboxylate group of the pyruvate substrate) were computed to bind in the allosteric and catalytic binding sites, respectively, were taken as energy thresholds for deducing the effect of P i on the activity of the other LDHs from the probe binding energy calculations (see "Algorithm to Computationally Estimate the Effect of P i " under "Experimental Procedures").
In contrast to L. plantarum LDH1, S. pyogenes LDH is weakly activated ( Table 3) by P i concentrations above 50 mM in the absence of FBP. The activity of the S. pyogenes LDH at pH 7 is about 5 times higher at 200 mM P i , and at pH 6, 200 mM P i leads to a 2.5-fold increase in activity (Fig. 5B). However, because of the intrinsically low activity of S. pyogenes LDH when effector is absent (0.8 and 0.6 mol/min/mg of protein at pH 6 and pH 7, respectively), the activation by addition of P i is considered to be weak. Computationally, at both pH values, we observed a favorable binding of the P i probe at the allosteric binding site (⌬E AS ϭ Ϫ3 and Ϫ2 kcal/mol) and no competition with a COO probe in the catalytic binding site (⌬E CS ϭ ϩ1 and 0 kcal/mol), thereby indicating activation by P i . In the presence of FBP, no significant P i -dependent changes (¬E) in the activity of the S. pyogenes LDH were observed at either pH 6 or pH 7.
The activity of the E. faecalis LDH1 is enhanced about 3-fold upon addition of 200 mM P i in the absence of FBP (Fig. 5C). The computational results suggest activation of the E. faecalis LDH1 by P i , although this is weaker than for S. pyogenes LDH due to less favorable binding of the P i probe in its allosteric binding site (⌬E AS ϭ Ϫ1 and 0 kcal/mol). In the presence of FBP, the addition of P i does not alter the activity of E. faecalis LDH1 (Fig. 5C), and this is also predicted from the computations (Table 3).
According to Gaspar et al. (14), the addition of 100 mM P i in the absence of FBP at pH 6 enhances the activity of L. lactis TABLE 3 Computed and measured effects of P i on LDH enzymes in the presence or absence of FBP at pH 6 and pH 7 The computational results reflect the analysis of the P i probe binding energies in the catalytic binding sites and, if FBP is absent, in the allosteric site of each LDH homodimer. The values were analyzed by comparison with the binding energies computed for L. plantarum LDH1, which are assigned as thresholds, and are therefore expressed as the energy differences ⌬E CS and ⌬E AS , respectively. In the presence of FBP, competition between the P i and COO probes in the catalytic binding site is described by ⌬E CS pr . For definitions of these quantities and their analysis, see "Experimental Procedures." A, strong activation; a, weak activation; I, strong inhibition; i, weak inhibition; ¬I, no inhibition; ¬E, no effect; n/a, not available. The logical operator V means "or." a For comparison with Gaspar et al. (14), only the activity measurements performed in the range from 0 to 100 mM P i are given.  (29). Thus, for probe positions with equivalent locations for the oxygen atom engaged in hydrogen bonding to the protein, the P i probe can appear to be less buried than the COO probe; this results in greater screening of the favorable electrostatic term and therefore less favorable binding of the P i probe than the COO probe. This artifact in electrostatic screening only arose for the E. faecalis LDH2 catalytic site as it has a Thr (where the other LDHs have Ala) within 4 Å of an Arg N2/N1 atom, which makes H-bonds to the probes. Therefore, for this case, we imposed a 3 kcal/mol correction on ⌬E CS pr . d Computational results suggest a weak competition of the P i probe and the COO probe in LDH2 from L. lactis. No inhibition was observed experimentally up to 50 mM P i . However, at higher P i concentration (100 mM), inhibition like that observed for LDH1 from L. lactis at pH 7 might be expected. LDH1 by about 3-fold to 0.3% of the activity in the presence of 3 mM FBP. At pH 7, the activity is reduced by a factor of 4 to 0.7% of the activity in the presence of 3 mM FBP. Because the overall activity of the enzyme in the absence of FBP is very low, these variations with P i concentration are not significant (Fig.  5D). Therefore, the P i effect in Table 3 is designated as weakly activatory or insignificant (a ∨¬E) and weakly inhibitory or insignificant (i ∨¬E) at pH 6 and pH 7, respectively. Computationally, at pH 6, we observed a strong binding of the P i probe in the allosteric binding site and a weak competition with the COO probe in the catalytic binding site, thereby resulting in a weak activation or no P i effect. At pH 7, the lack of effect of P i is more pronounced due to less favorable binding of the P i probe in the allosteric binding site. In the presence of FBP, a weak inhibitory effect of P i was observed both experimentally and computationally. The activity of the LDHs from L. lactis has been previously studied under different environmental conditions (13 Gaspar et al. (14), in the absence of FBP, the activity of L. lactis LDH2 was enhanced from 22 to 75% at pH 6, whereas at pH 7, the initially low activity of 4% remained the same after adding 100 mM P i (Fig. 5E). This observation is consistent with the computational results. In the presence of FBP at pH 7, a weak inhibition (i) was experimentally measured as well as computationally assigned. However, at pH 6, no significant effect of P i (¬E) was observed experimentally, whereas the computations suggest weak inhibition (i). We speculate that because the experiments were restricted to only 50 mM P i inhibition might indeed occur close to 100 mM as was measured by Gaspar et al. (14) for L. lactis LDH1 (pH 7).
Measurements for E. faecalis LDH2 in the absence of FBP activity showed an ϳ4-fold inhibitory effect (i) upon addition of P i at both pH 6 and pH 7. In the presence of FBP, a strong inhibition (I) of 6-and 8-fold for pH 6 and pH 7, respectively, was observed (Fig. 5F). The inhibitory effects were predicted computationally except for E. faecalis LDH2 at pH 6 in the absence of FBP for which a weak activation or no P i effect instead of weak inhibition was expected. P i Versus NaCl-To elucidate whether the observed activatory or inhibitory effect of P i in the absence of FBP was due to differences in ionic strength, we also measured enzyme activities with increasing NaCl concentrations for three LDHs at different pH values as control cases. The relative values are shown in Fig. 6, and the corresponding absolute values are presented in supplemental Table S3.
As can be seen in Fig. 6, NaCl did not influence the activity of S. pyogenes (pH 6) and E. faecalis (pH 7) LDHs. In contrast, for E. faecalis LDH2 (pH 6), addition of salts (P i and NaCl) had an inhibitory effect (Fig. 6C). The inhibition upon addition of P i presumably resulted from the stronger positive electrostatic potential of the enzyme at the phosphate binding positions (Fig.  2) and the high affinity of P i to the catalytic and allosteric binding sites. Computation of P i binding energies suggests more favorable binding to the catalytic site (Ϫ18 kcal/mol), thereby explaining a rapid inhibition at low P i concentrations (up to 5 mM). At higher P i concentration, binding to the allosteric binding site may also take place (Ϫ15 kcal/mol). This might have an activatory effect and thus suppress the strong inhibitory effect. The competition between inhibitory and activatory effects might lead to either a weak activation or no effect on enzyme activity upon P i addition (Table 3). A different concentration dependence was observed upon addition of NaCl, suggesting a distinct mechanism. E. faecalis LDH2 probably demonstrates higher sensitivity to salt screening due to its stronger electro-FIGURE 6. Comparison of the impact of P i and NaCl on the activity of LDH enzymes. The activities of S. pyogenes LDH at pH 6 (A), E. faecalis LDH1 at pH 7 (B), and E. faecalis LDH2 at pH 6 (C) were measured 1) at constant 50 mM P i and with NaCl concentrations ranging from 0 to 200 mM (red) and 2) at constant 50 mM NaCl and with P i concentrations ranging from 0 to 200 mM (black). The measured activities are shown relative to those at 50 mM P i and 50 mM NaCl, which were defined as 100% for both sets of experiments. All experiments were conducted at 0 mM FBP. Error bars represent the S.E. static potential, especially at the allosteric binding site. Chloride ions at the allosteric binding site might hinder the approach of P i to the allosteric binding site, thereby preventing activation of the enzyme upon P i addition.
Concluding Remarks-Our results show that the LAB LDH enzymes from the four species studied differ in their regulation by FBP, P i , ionic strength, and pH. These differences have presumably evolved as the organisms have adapted to their different biological environments. With our investigations, we show that the experimentally measured differences can be explained by the computed binding properties of the allosteric and catalytic binding sites of the respective isoenzymes. The subtle interplay between the activatory and inhibitory effects of substrate, FBP, and P i binding is illustrated in Fig. 7. FBP activates all the enzymes except the L. plantarum LDH1 for which mutation of a histidine residue to aspartic acid at the dimer interface hinders binding of FBP. It should be noted that this single point mutation affects both sides of the allosteric site because of its location at the homodimer interface and therefore results in a rather negatively charged allosteric binding site. In the absence of FBP, P i can have a (weaker) activatory effect by binding at the allosteric site of LDH enzymes. However, it can also be inhibitory by binding at the catalytic site and competing with substrate binding. The balance between these two effects differs for the LDHs studied, resulting in P i being inhibitory in some cases and activatory in others. Calculations (not shown) also suggest that the activatory effect of P i might be compromised by binding of substrate in the allosteric binding site, thereby competing with P i and reducing the activatory role of P i . Levering et al. (18) have shown by experiment and kinetic modeling that L. lactis and S. pyogenes differ with respect to phosphate regulation of their central metabolism. Insufficient phosphate supply resulted in inhibition of glycolysis that was more severe in S. pyogenes than in L. lactis. This may result from the tendency for P i to activate the S. pyogenes LDH but to inactivate the corresponding L. lactis LDH1 (see Table 3). Our results show the need to explicitly account for the effects of P i on LDHs in kinetic models of LAB central metabolism. Moreover, the use of kinetic measurements and protein structure-based modeling provides a mechanistic understanding of the LAB LDHs, and this combined approach should be generally useful for studying other allosterically regulated enzymes.