Mevalonate 5-diphosphate mediates ATP binding to the mevalonate diphosphate decarboxylase from the bacterial pathogen Enterococcus faecalis

The mevalonate pathway produces isopentenyl diphosphate (IPP), a building block for polyisoprenoid synthesis, and is a crucial pathway for growth of the human bacterial pathogen Enterococcus faecalis. The final enzyme in this pathway, mevalonate diphosphate decarboxylase (MDD), acts on mevalonate diphosphate (MVAPP) to produce IPP while consuming ATP. This essential enzyme has been suggested as a therapeutic target for the treatment of drug-resistant bacterial infections. Here, we report functional and structural studies on the mevalonate diphosphate decarboxylase from E. faecalis (MDDEF). The MDDEF crystal structure in complex with ATP (MDDEF–ATP) revealed that the phosphate-binding loop (amino acids 97–105) is not involved in ATP binding and that the phosphate tail of ATP in this structure is in an outward-facing position pointing away from the active site. This suggested that binding of MDDEF to MVAPP is necessary to guide ATP into a catalytically favorable position. Enzymology experiments show that the MDDEF performs a sequential ordered bi-substrate reaction with MVAPP as the first substrate, consistent with the isothermal titration calorimetry (ITC) experiments. On the basis of ITC results, we propose that this initial prerequisite binding of MVAPP enhances ATP binding. In summary, our findings reveal a substrate-induced substrate-binding event that occurs during the MDDEF-catalyzed reaction. The disengagement of the phosphate-binding loop concomitant with the alternative ATP-binding configuration may provide the structural basis for antimicrobial design against these pathogenic enterococci.

Multidrug-resistant microorganisms have been found in humans and food animals due to the overuse of antibiotics (1). In the United States, a surveillance of antibacterial resistance from the Centers for Disease Control and Prevention reports the deaths and the cost resulting from infectious diseases and emphasizes the urgency to control drug-resistant bacteria (2). This critical list includes vancomycin-resistant enterococci (VRE), 2 which cause a range of enterococcal infections such as bacteremia, urinary tract infections, intra-abdominal and pelvic infections, central nervous system infections, skin and skin structure infections, and infective enterocarditis (2)(3)(4).
Enterococci are Gram-positive and facultative anaerobes colonized in the gastrointestinal tract. The first clinical isolates of vancomycin-resistant enterococci were reported in Europe in 1988 (5,6). The VRE bacteria are intrinsically insensitive to detergents and antibiotics (clindamycin, cephalosporins, and aminoglycosides) and keep gaining drug resistance to other newly invented antibiotics, such as linezolid and daptomycin (7)(8)(9)(10)(11). The emergence of drug-resistant enterococci results in difficulties of treatment of VRE infections (12)(13)(14). For the protection of patients in health care settings, new approaches and new antimicrobial agents against enterococcal infections are urgently needed.
In 2000, Wilding et al. (15) reported that the mevalonate pathway is pivotal in the Gram-positive bacterial pathogens, enterococci, staphylococci, and streptococci. This pathway includes six enzymes, acetyl-CoA acetyltransferase, HMG-CoA synthase, HMG-CoA reductase, mevalonate kinase, phosphomevalonate kinase, and mevalonate diphosphate decarboxylase (EC 4.1.1.33), that ultimately produce IPP. IPP then serves as a building block for polyisoprenoid synthesis in living organisms ranging from bacteria to humans (16 -18). In bacteria, isoprenoid production is also involved in the biosynthesis of bacterial cell wall and electron carriers in the respiratory chain (16, 19 -21).
The five enzymes in the mevalonate pathway (except acetyl-CoA acetyltransferase) have been identified to be critical for bacterial growth (15). Experiments with the mammalian This work was supported by the Purdue Center for Cancer Research and National Institutes of Health Grant P30 CA 023168. The authors declare that they have no conflicts of interest with the contents of this article. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. This article contains supplemental Fig. S1 and Tables S1 and S2. The atomic coordinates and structure factors ( enzyme have shown that the ATP-dependent decarboxylation of MVAPP catalyzed by MDD is a rate-limiting step in IPP synthesis (22), implicating MDD as a suitable drug target for the disease treatment. Interestingly, in Streptococcus pneumoniae, an in vitro experiment has indicated that a higher level of MVAPP can also inhibit mevalonate kinase (23). This implies that the mevalonate pathway is sophisticatedly regulated by its downstream products and strengthens the case for MDD as a promising drug target for treatments of enterococcal infections.
In this research, we have studied MDD from Enterococcus faecalis (MDD EF ) from functional, biophysical, and structural points of view to decipher the substrate-binding mechanism of MDD EF . MDD proteins trigger the irreversible ATP-dependent decarboxylation of MVAPP to produce IPP in the last step of the mevalonate pathway ( Fig. 1) (24). Several residues (aspartate, lysine, serine and arginine) in the active site of MDD proteins have been identified as being involved in substrate binding and enzyme catalysis (25)(26)(27). In 2012, Barta et al. (27) reported several crystal structures of the wild-type and mutant forms of MDD from Staphylococcus epidermidis (MDD SE ) with or without substrates/analogues binding to the active-site cleft. From their structural models, they proposed that MVAPP is the first substrate binding to the deeper pocket in the active site of MDD SE , where Ser-106, Arg-144, Ser-192, and Arg-193 hold MVAPP in position. Next, ATP sits in the ATP-binding pocket interacting with Ser-94 and Asn-96, and the phosphate tail is clamped by Ser-106, Ser-107, and the phosphate-binding loop (Ala-101, Gly-103, and Leu-104), followed by enzyme catalysis (27). The substrate-binding order of MDD isolated from chicken liver was examined with results that suggest avian MDD performs a sequential ordered bi-substrate mechanism with MVAPP as the first substrate (28,29). However, no kinetic evidence was presented that prokaryotic MDDs act on substrates in the same manner,. A case in point is the enzyme mechanism of proteins in the GHMP kinase family (galactokinases, homoserine kinases, mevalonate kinases, and phosphomevalonate kinases), which vary due to different biological sources (27). In this work, we aimed to obtain a crystal structure of MDD EF bound with ATP alone to investigate structural differences between the MDD EF -ATP structure and those complex structures of MDD SE (supplemental Table S1). With that, we could better understand how the two substrates initially bind to the active site, which may aid in our rational drug design in vitro and in silico.

MDD EF protein expression and purification
Protein expression and purification conditions are detailed under "Experimental procedures." Overexpressed His-tagged MDD EF in Escherichia coli (BL21) was purified via nickelaffinity chromatography ( Fig. 2A). Eluted protein fractions were examined by SDS-PAGE (Fig. 2B). Fractions containing MDD EF were collected for TEV treatment to remove the fusion N-terminal His tag from MDD EF (Fig. 2C) ("Experimental procedures"). Finally, we obtained a His tag-removed MDD EF (less than 5% impurities, Fig. 2C, flow-through) with a protein yield of 80 -120 mg/liter of culture and adjusted the protein concentration to 8 -10 mg/ml.

Crystal structure of MDD EF -ATP and the apo form of MDD EF
The detailed procedures for obtaining co-crystals of MDD EF bound with ATP (MDD EF -ATP, PDB accession number 5V2L) and apo-MDD EF (MDD EF , PDB accession number 5V2M), data collection, and structure refinement are summarized under "Experimental procedures." The data collection and refinement statistics of both structures are listed in Table 1. According to the CATH protein structure classification database analysis, the overall structure of the MDD family of proteins contains ␣ and ␤ secondary structure elements (␣ and ␤ class) (Fig. 2D), and the folding architecture is characterized as a two-layer sandwich (30). The protein family annotation (Pfam) (31) indicates MDD proteins belong to the GHMP kinase (galactokinase, homoserine kinase, mevalonate kinase, and phosphomevalonate kinase) superfamily, which catalyzes ATP-dependent reactions. The overall structures of MDD EF and MDD EF -ATP superimpose well with an r.m.s.d of 0.174 Å. In MDD EF , the phosphate-binding loop (97-104) and residues from 183 to 192 are missing. Similarly, in MDD EF -ATP, residues from 184 to 189 cannot be determined. However, the phosphate-binding loop can be observed (Fig. 2D). In the literature, the phosphatebinding loop in MDD is known for ATP binding (27). In our crystal structure of MDD EF -ATP, the phosphate-binding loop does not interact with the ATP molecule (Fig. 3A), although ATP binds to the ATP-binding pocket via the interactions between ATP and the conserved residues (Ser-93, Asn-95, and Ser-105 (Fig. 3B).

Structural differences between MDD EF -ATP and MDD SE -FMVAPP-ATP␥S infer the pivotal role of MVAPP in ATP binding
By comparing our crystal structure of MDD EF -ATP with the tertiary crystal structure of MDD SE (MDD SE -FMVAPP-ATP␥S, PDB code 4DPT) (27), three major differences were found in the phosphate-binding loop (Fig. 3C) and the ATP/ ATP␥S configurations (Fig. 3D). First, the phosphate-binding loop in MDD EF -ATP remains in an open conformation and does not bend down to form contacts with ATP, if compared with the phosphate-binding loop in MDD SE -FMVAPP-ATP␥S (Fig. 3C). The largest distance moved by the backbone

Induced substrate-binding mechanism of MDD EF
atoms in the phosphate-binding loop is 11.1 Å (C␣ of Ala-100 in MDD EF -ATP and C␣ of Ala-101 in the MDD SE complex structure). Second, the ␥-thiophosphate group of ATP␥S in MDD SE -FMVAPP-ATP␥S is reported to form contacts (hydrogen bonding) with the backbone amide of two residues, Ser-107 and Ala-108 in Helix ␣2 of MDD SE (106 -121) (supplemental Fig.  S1A), which are located in the active site; instead, the ␥-phosphate of ATP in our structure of MDD EF -ATP is facing out and does not interact with any residues in the molecule ( Fig. 3D and supplemental Fig. S1B). Third, the ␣and ␤-phosphates in these two structures are located and oriented differently and thus result in the distinct orientations of the phosphate tails in these structures (Fig. 3D). The adenosine group of ATP in our MDD EF -ATP binds to the ATP-binding pocket with a temperature factor similar to the average B-factor of the protein (Table  1). However, the phosphate tail of ATP does not interact with any residues, consistent with its higher temperature factor (B-factor Ͼ60), suggesting ATP does not initially bind to MDD EF in its favored position. These findings lead us to hypothesize the prerequisite binding of MVAPP to MDD EF may be important for forming the ATP-binding pocket. To test the hypothesis, we utilized enzymatic and thermodynamic approaches to determine the substrate-binding order of MDD EF and to examine whether the binding of MVAPP enhances binding of ATP.

Kinetics reveals a compulsory ordered bi-substrate mechanism of MDD EF with MVAPP binding first
The crystal structure of MDD EF -ATP described above implied that the binding of MVAPP may affect ATP binding. The previous work on avian MDD suggests the substrate-binding order of MDD from chickens belongs to an ordered bi-substrate mechanism with MVAPP binding first (28). From our crystal structure of MDD EF -ATP, ATP can bind to the ATP-binding pocket by itself, even if in an unproductive form (Fig. 3, B and D). Therefore, the question arises whether the substrate-binding order is conserved in both eukaryotic and prokaryotic MDD proteins. To address this issue, we performed kinetics studies intro-

Induced substrate-binding mechanism of MDD EF
ducing ATP␥S as a competitive inhibitor of ATP to determine the enzyme mechanism of MDD EF . We utilized the coupled enzyme method (pyruvate kinase/lactate dehydrogenase) to determine the enzymatic activity of MDD EF at varying concentrations of the two substrates and inhibitors (25,29,32). The detailed procedures are described under "Experimental procedures." MDD triggers MVAPP decarboxylation coupled with ATP consumption. MVAPP and ATP are its two substrates. To differentiate whether the reaction belongs to either the ping-pong or the sequential bi-substrate mechanism, we have performed enzymatic reactions at varying concentrations of both substrates of MDD EF , MVAPP, and MgATP ("Experimental procedures") ( Fig. 4, A and C), and we analyzed the kinetic data by the Lineweaver-Burk method (Fig. 4, B and 4D). Each 1/v (reciprocal reaction velocity) was plotted as a function of either 1/[MVAPP] (reciprocal of varying concentrations of MVAPP) (Fig. 4B) or 1/[MgATP] (reciprocal of varying concentrations of MgATP) (Fig. 4D). The data were best represented by a sequential bi-substrate model using the general equation for a bi-substrate enzyme catalysis system as rep- The patterns in Fig. 4, B and D, show that the kinetic curves in the double-reciprocal plots converge to the x axis. We could then confirm that the enzyme mechanism of MDD EF belongs to a sequential bi-substrate mechanism instead of a ping-pong mechanism or a rapid-equilibrium sequential ordered mechanism. From the curve fitting, we obtain the following kinetic parameters: V max ϭ 16.1 Ϯ 0.3 mol/min/mg; k cat ϭ 9.8 Ϯ 0.2 s Ϫ1 ; K mMVAPP ϭ 39.7 Ϯ 2.8 M; and K mMgATP ϭ 166 Ϯ 12 M (Table 2).
To determine whether substrates bind to MDD EF in an ordered or random manner, we introduced ATP␥S into the reaction buffer as a dead-end inhibitor of MDD EF and then determined the substrate-binding order based on the inhibitory patterns of ATP␥S (supplemental Table S2) (28,33). Two sets of inhibitory assays were conducted. The first set of experiments was performed at a constant MVAPP concentration of 40 M and varying concentrations of MgATP and ATP␥S (see under "Experimental procedures"). The kinetic data from the inhibitory assays were best represented by a modified competitive inhibition model as shown in Equation 2,  5A) and a Lineweaver-Burk plot (Fig. 5B). The values of V max , K mMgATP , and K IATP␥S were derived from Equation 2 and are summarized in Table 2.
The second set of experiments was performed at a constant MgATP concentration of 200 M and varying concentrations of MVAPP and ATP␥S ("Experimental procedures"). The data set was best represented by a modified uncompetitive inhibition model as shown in Equation 3, where V max is the maximum velocity; [A] is the varying concentrations of MVAPP; K A is K mMVAPP ; K IA is K dMVAPP ; K IA /K A is 0.8, which was derived from the enzymatic reactions described above; [I] is the varying concentration of ATP␥S, and K I is K IATP␥S (33).
The kinetics data are shown as a Michaelis-Menten plot (Fig.  5C) and a Lineweaver-Burk plot (Fig. 5D). The values of V max , K mMVAPP , and K IATP␥S were also derived from Equation 3 and are summarized in Table 2. According to the inhibitory patterns with the addition of ATP␥S in the enzymatic reactions under the designed conditions, we suggest that the enzymatic reaction of MDD EF belongs to a sequential ordered bi-substrate mechanism with MVAPP binding first, and we find the sub-

Induced substrate-binding mechanism of MDD EF
strate-binding mechanism of MDD to be conserved between the eukaryotic and prokaryotic species in the family of MDD proteins.

Isothermal titration calorimetry (ITC) results suggest an MVAPP-induced ATP-binding mechanism of MDD EF
From the kinetics study on MDD EF , MVAPP is identified as the first substrate in the reaction mechanism. Our complex structure of MDD EF -ATP shows that in the absence of MVAPP, ATP binds the ATP-binding pocket in an unusual configuration (Fig. 3, B and D), and the phosphate-binding loop does not bend down to interact with the ATP molecule (Fig. 3C). These results imply the binding of MVAPP may help form a binding site for ATP in its catalytically favored configuration. In other words, the binding of ATP might be enhanced by the prerequisite binding of MVAPP. To test this, we employed thermodynamic approaches and utilized ITC (34) to determine the dissociation constants of the first substrate, MVAPP, the second substrate, MgATP, and the ATP analogues, ATP␥S and AMPPCP.
We have determined the K d values of substrates or ligands (MVAPP, MgATP, AMPPCP, and ATP␥S). The raw data for titrations of MDD EF are shown in Fig. 6. All the derived thermodynamic parameters, K d (1/K a ), are listed in Table 3. The K dMVAPP value is 20.4 Ϯ 9.3 M (Fig. 6A); the K dATP value is 288 Ϯ 36 M (Fig. 6B); and the K dATP␥S value is 215 Ϯ 8 M (Fig.  6C). Next, we determined K dATP␥S under the condition in which MDD EF was pre-incubated with MVAPP. MDD EF was incubated with 1 mM MVAPP for 30 min to ensure that the MVAPP-binding pocket of MDD EF was nearly fully occupied (about 97.8%), according to the dissociation constant at equilibrium as shown in Equation 4, where [MDD EF ] total is the total concentration of MDD EF ; [MVAPP] total is the total concentration of MVAPP; and x is the concentration of the MDD EF -MVAPP complex.
Under these conditions, the K d value between ATP␥S and MDD EF -MVAPP was determined to be 25.4 Ϯ 5.5 M (Fig. 6D), which is 10-fold less than the K d value between ATP␥S and MDD EF alone, supporting our proposed model in which the pre-binding of MVAPP would strengthen ATP␥S binding. Our

Discussion
GHMP family kinases employ a variety of mechanisms and enzyme order. Previously, the substrate-binding mechanism of MDD from chickens was determined to belong to a sequential bi-substrate mechanism with MVAPP as the first substrate (28); from our study, we determined prokaryotic MDD from E. faecalis also binds MVAPP first during the enzymatic reaction. MDD is characterized as a member GHMK of the kinase family (26). For the galactokinase from pig liver, MgATP is the first substrate of the galactokinase, followed by the binding of galactose (35); however, in E. coli, the galactokinase enzyme performs a random bi-substrate mechanism (36). For the homoserine kinase in E. coli, MgATP is preferred to bind to the enzyme first (37). For the mevalonate kinase from hog liver, a sequential ordered bi-substrate mechanism is performed with mevalonate binding first (38). The enzyme mechanism of the phosphomevalonate kinase from pigs belongs to a sequential ordered bi-substrate mechanism with phosphomevalonate binding first (39). In the case of the MDD, the enzyme mechanism may be conserved among the family of proteins; however, in other cases in the GHMP kinase family, the enzyme mechanisms may vary in species. Although in this study we aim at

Induced substrate-binding mechanism of MDD EF
elucidating the intriguing observations from the structure of MDD EF -ATP and the structural basis of MVAPP for assisting ATP binding to MDD EF , alternative substrates or product inhibition strategies at both saturated and unsaturated concentrations of MVAPP and MgATP (40) are also the ways to comprehensively investigate the overall enzyme mechanism of MDD EF .
MDD proteins have been known to trigger the ATP-dependent decarboxylation of MVAPP, but with a requirement for metal ions (Mg 2ϩ ) to perform catalysis under physiological conditions (29). In our crystal structure of MDD EF -ATP, the ATP molecule does not have magnesium binding to the ␣and ␤-phosphate group. In the crystallization condition, we have 5 mM magnesium chloride and 10 mM ATP in the solution. The buffer condition is 50 mM sodium acetate at pH 4.6, and the protonation state of ATP is mainly HATP 3Ϫ under this condition (41). The dissociation constant between HATP 3Ϫ and magnesium is about 1.58 mM (log K a ϭ 2.79 Ϯ 0.15) (42), and the MgATP concentration under this condition is about 4 mM, suggesting the concentration of free HATP 3Ϫ is about 6 mM. The reason for not seeing magnesium-bound ATP in the crystal structure of MDD EF -ATP could be due to the competitive binding of free HATP 3Ϫ to the ATP-binding pocket of MDD EF . It is also possible that a higher average temperature factor of the phosphate tail of ATP results in poorly defined electron density in this region so that we cannot identify a magnesium ion in this structure. In 2012, Barta et al. (27) reported experimental findings on the substrate-binding order of MDD SE . However, they do not identify any metal-binding sites in their crystal structures of MDD SE with substrate analogues. Based on the observations above, we suggest the current structure models of MDD proteins in the PDB do not necessarily represent the final conformation of MDD proteins upon substrate binding and enzyme catalysis. This could also be a problem for structurebased drug design, which is heavily dependent on the structural details in atomic models of macromolecules. To study the structure and function of the MDD family of proteins remains an important but challenging field.   )) and the standard deviation of ⌬G () is calculated from the equation: ϭ RT͌ln(1 ϩ (v/m 2 )), where m is the mean of the association constant (K a ), and v is the variance of K a derived from each ITC experiment. b The mean of T⌬S ( 1-2 ) is derived from the difference between the means of ⌬G ( 1 ) and ⌬H ( 2 ) and the standard deviation ( 1-2 ) of T⌬S is derived from the equation: 1-2 ‫؍‬ ͌( 1 2 /n ϩ 2 2 /n), where 1 is the standard deviation of ⌬G and 2 is the standard deviation of ⌬H.

Induced substrate-binding mechanism of MDD EF
The open conformation of the phosphate-binding loop and the unusual configuration of ATP in our crystal structure of MDD EF -ATP suggest the importance of MVAPP binding in forming the productive ATP-binding site. We proposed that MVAPP could induce ATP binding to MDD EF , and the thermodynamic results are in agreement with our proposal. In addition to the examination of changes of K dATP␥S in the presence of excess amounts of MVAPP in the solution, we also tested whether this enhanced binding occurs in the titration experiments using AMPPCP. The K dAMPPCP value was 271 Ϯ 43 M (Table 3, row e); however, in the presence of MVAPP, we could not detect heat changes and so could not obtain the thermodynamic parameters in this case (Table 3, row f). If instead MDD EF is pre-incubated with excess amounts of AMPPCP (10-fold, MDD EF -AMPPCP complex formation rate: 77.4%) and then the mixture is titrated with MVAPP ( Table 3, row g), the K dM -VAPP value in this case is 18.7 Ϯ 7.4 M, which is a similar value as from the MDD EF -MVAPP titration (Table 3, row a), indicating AMPPCP may have little effect on MVAPP binding. We have also done a preliminary test on ATP␥S. MDD EF was preincubated with excess amounts of ATP␥S (10-fold, MDD EF -ATP␥S complex formation rate: 81.1%) and then the mixture was titrated with MVAPP (data not shown). The thermogram shows an unusual two-process binding event and reaches saturation with no further heat change after the 6th injection, suggesting a tight binding event occurs. K dMVAPP was roughly determined to be 76 nM (data not shown). Although there are no other experimental results to explain the two-process binding, a quick saturation after the 6th injection indicates that MVAPP binds MDD EF -ATP␥S much tighter than MVAPP binds MDD EF alone (Table 3, row a). It also implies that AMP-PCP might not be able to access a catalytically-favored conformation for "locking" MVAPP in the active site, which might need, for example, an aid from the conformational changes of MDD EF upon substrate binding. The conserved phosphatebinding loop of the MDD family of proteins functions to interact with the phosphate tail of ATP. It has been shown that AMPPCP has poor inhibitory activity (K i Ͼ 1 mM) against MDD from chickens (28) and MDD EF in our preliminary enzymatic study (data not shown). AMPPCP has a replacement of the bridging oxygen with the methyl group (CH 2 -) between the ␤and ␥-phosphorus atoms. We therefore propose that the ␤-,␥bridging oxygen of ATP serves as a key checkpoint for the molecular recognition of ATP by MDD EF during the enzymatic reaction.
To summarize our results, we suggest that an induced substrate-binding mechanism of MDD proteins occurs during the enzymatic reaction (Fig. 7), in which the binding of the first substrate, MVAPP, to MDD will trigger the conformational changes of the enzyme to accommodate ATP binding to the ATP-binding pocket in its catalytically favored position, followed by enzyme catalysis and product release. The observation that ATP can bind by itself but in a catalytically unfavorable position may also provide an alternative configuration of the MDD EF enzyme that provides a new structural basis for antibacterial design for pathogenic enterococci.

Cloning, overexpression, and purification of the recombinant form of MDD EF
A gene fragment encoding MDD from E. faecalis (MDD EF ) was amplified via PCR and subcloned into the expression plasmid pET30a (43). Upon confirmation of the DNA sequence, the construct was transformed into E. coli BL21 (DE3). Transformed cells were cultured in LB broth supplemented with kanamycin (50 mg/ml) at 37°C to an A 600 nm of 1.0. Protein expression of MDD EF was induced by addition of isopropyl 1-thio-␤-D-galactopyranoside (0.1 mM) for another 4 h at 37°C. Cells were harvested by centrifugation at 8000 rpm, resuspended in binding buffer (50 mM sodium phosphate, pH 7.4, 300 mM NaCl, and 10 mM imidazole), and lysed to homogeneity by a French press. His-tagged MDD EF protein was soluble in the supernatant after centrifugation and trapped on a Ni 2ϩ -NTA column followed by the elution with increasing concentrations of elution buffer (50 mM sodium phosphate, pH 7.4, 300 mM NaCl, and 300 mM imidazole). Eluted fractions were pooled and desalted against dialysis buffer (25 mM Tris-HCl, pH 8.0, 100 mM NaCl, and 10 mM MgSO 4 ) twice, the first time with ␤-mercaptoethanol (2-ME) (20 mM) and the second time without 2-ME. The N-terminal His tag was removed from MDD EF by treatment with recombinant TEV protease in dialysis buffer containing 1 mM DTT and 0.5 mM EDTA overnight at room temperature for 4 h followed by a final dialysis procedure at 4°C in the dialysis buffer without DTT and EDTA. His-tagged TEV and residual His-tagged MDD EF were removed by passing the protein mixture through a nickel affinity resin. The artificial sequence NA remains at the N terminus of MDD EF after the TEV treatment. The purified MDD EF protein solution was concentrated to 8 -10 mg/ml by ultrafiltration and stored at Ϫ20°C.

Protein crystallization
The co-crystal of MDD EF in complex with ATP (MDD EF -ATP) grew under the crystallization condition (10 mM ATP, 26% PEG 3350, 50 mM sodium acetate, pH 4.6, and 5 mM MgCl 2 ) by sitting drop vapor diffusion. ATP was added to the

Induced substrate-binding mechanism of MDD EF
protein solution to a final concentration of 10 mM. 1 l of MDD EF -ATP solution was mixed with 1 l of the reservoir. Crystals formed after a 2-day equilibrium period. Cryo-protectant (26% PEG 3350, 17% PEG 400, 50 mM sodium acetate, pH 4.6, and 5 mM MgCl 2 ) was prepared and gradually added into the protein/reservoir mixture to prevent ice formation when flash-freezing the crystal in liquid nitrogen.
A stable condition (1.6 M ammonium sulfate, 50 mM sodium acetate, pH 4.6) for growing crystals of the apo-form of MDD EF was obtained and optimized from a commercialized crystallization screening kit (Qiagen, class II, condition A2: 0.1 M sodium acetate, pH 4.5, 2 M ammonium sulfate). Protein solution (8 to 9 mg/ml) was mixed with the reservoir solution at a 1:1 ratio and equilibrated by vapor diffusion of sitting drops at 20°C for 2 days. In a 96-well plate, there were about 20 wells containing single cuboid crystals with a size of 0.2 to 0.3 mm. Cryo-protectant (25% glycerol, 1.6 M ammonium sulfate, 50 mM sodium acetate, pH 4.6) was prepared and gradually added into the protein/reservoir mixture to prevent ice formation when flashfreezing the crystal in liquid nitrogen.

Data collection and structure determination, refinement analysis, and refine statistics
The diffraction data from crystals of MDD EF co-crystallized with ATP (MDD EF -ATP) and an apo-form of MDD EF (MDD EF ) were collected at the 23-ID-D and 19-ID-BM beamlines at the Advanced Photon Source (APS) at Argonne National Laboratory in Chicago. The HKL2000 software was used for data integration, data reduction, and data scaling (44), generating a scale pack reflection file. The space group of the crystal was determined as P2 1 (45,46). The software in CCP4, scalepack2mtz, was then used to convert the scale pack reflection file (.sca) to an MTZ format (.mtz) with R-free flag assigned (5-5.5%) (47).
One molecule was expected in an asymmetric unit by analyzing the Matthews coefficient and the water content in each case (MDD EF -ATP: cell volume,: 370710.5 Å 3 , molecular mass: 36,489.5 Da, Matthews coefficient: 2.54, % solvent: 51.6; MDD EF : cell volume: 370,279.3 Å 3 , molecular mass: 36489.5 Da, Matthews coefficient: 2.54, % solvent: 51.6). Phenix.phaser was used to solve the phases and determine the structure of MDD EF . MDD EF and MDD SE share 60% sequence identity. A modified search model (polyalanine) generated from MDD from S. epidermidis (MDD SE , PDB code 3QT5) using CHAIN-SAW in CCP4 (48) was used to produce initial phases for the co-crystal structure of MDD EF -ATP in molecular replacement with phenix.phaser (45). In each step of molecular replacement, only one solution was found. The rotation function with loglikelihood gain greater than 0, the translation function with Z-score greater than 8, and no violations after the packing analysis also suggest the success of the molecular replacement approach. The phases for determining the crystal structure of the apo-form of MDD EF were produced by molecular replacement with Phenix.phaser (45) using a polyalanine-modified MDD EF -ATP structure (described above) as a search model. After phase determination, the structure refinement was performed in phenix.refine (45). During structure refinement, strategies for refining the model geometry (XYZ coordinates), atomic positions (Real space), and atomic B-factors (individual B-factors) were chosen, and after the refinement, the missing side chains of the residues in the structural model were manually rebuilt in the graphics program Coot (49) based on observation of the electron density map (2F o Ϫ F c ) and the difference map (F o Ϫ F c ). The simulated-annealing (Cartesian) option was also employed in the first few runs of structure refinement. In phenix, we used the phenix.composit_omit_map software to generate the omit map and used the maps to evaluate the refined structural models. Target function optimization was also chosen for refining the weight between X-ray data and the structural model (optimize X-ray/stereochemistry weight, optimize X-ray/ADP weight). Water molecules were built by either phenix.refine or Coot and inspected in the Coot interface.
The crystallographic information file (.cif) and the PDB format file (.pdb) of ATP were generated using phenix. eLBOW (45). After a few runs of structure refinement without ligands, ATP was manually placed and fitted into the weighted difference electron density maps in Coot (49). Geometry and rotamer outliers were inspected and evaluated after each structure refinement. Finally, ATP was omitted by setting the ligand occupancy to zero or simply removing it from the structure for calculating a simulatedannealing-ligand-omit map in phenix.composite_omit_map, and the ligand-omit map and its corresponding structure were examined in PyMOL (50).

Enzyme kinetics of wild-type MDD EF
Enzymatic activity of MDD EF was determined using an ATP/ NADH enzyme-coupled assay. In this assay, ADP produced from the MDD reaction is converted back to ATP by pyruvate kinase, coupled with the conversion of phosphoenolpyruvate to pyruvate, and the pyruvate is then converted to lactate by lactate dehydrogenase, coupled with the oxidation of NADH to NAD ϩ . The oxidation of NADH to NAD ϩ results in an absorbance decrease at 340 nm. Thus, the ADP production can be detected by measuring the decrease in A 340 nm . The MDD enzymatic activity would be proportional to the slope of the continuously declining value of A 340 nm . We have utilized this method to determine the enzymatic activity of MDD EF under different conditions for obtaining kinetic parameters of MDD EF . Each reaction was performed at 30°C under the buffer condition (100 mM HEPES, pH 7.0, 100 mM KCl, 10 mM MgCl 2 , 0.2 mM NADH, 0.4 mM phosphoenolpyruvate, 4 units of pyruvate kinase, 4 units of lactate dehydrogenase, and 100 nM MDD EF ) (27). Initial velocity of each reaction was determined at a range of concentrations of MgATP (50,100,150,200, 400, 600, 800, and 1000 M) and MVAPP (10,15,25,50,100,150,200, and 300 M). Final enzymatic parameters, K m and V max , were determined by fitting kinetic data to a sequential bi-substrate mechanism model using SigmaPlot12.5/Enzyme Kinetic Module 1.3 (Systat Software, Inc.).

Inhibition assays
ATP␥S was used as a competitive inhibitor of ATP for determining the inhibition kinetics of MDD EF . Different fixed concentrations of ATP␥S (0, 100, 200 and 400 M) were added into the reactions versus varying MgATP (50,100,200, and 400 M) and fixed MVAPP (40 M) or varying MVAPP (10, 20, 40, 80 M) and fixed MgATP (200 M). Assays were performed under the conditions as described above. The kinetic data with a fixed MVAPP concentration were fit with a competitive inhibition model; the kinetic data with a fixed MgATP concentration were fit with an uncompetitive inhibition model using SigmaPlot as described previously.

ITC experiments
The preparation of TEV-treated MDD EF was described above. The protein solution was dialyzed against the same buffer as used in the enzymatic reactions described previously (100 mM HEPES, pH 7.0, 100 mM KCl and 10 mM MgCl 2 ). All the buffer solutions in ITC experiments were filtered through an 0.45-m filter and degassed for 1 h at room temperature. The protein concentration was adjusted to 100 M (260 l). Each ligand (MVAPP, ATP, AMPPCP, and ATP␥S) was prepared in the same dialysis buffer to avoid buffer mismatch. The concentration of each titrant was optimized in different experiments based on the experimental designs simulated using the MicroCal Origin 7.0 software package, and the final concentration of each ligand was adjusted to 2 or 3 mM. The ITC instrument, MicroCal iTC200, was employed for isothermal titrations in this study, and the reference cell was filled with double-distilled H 2 O containing 0.01% sodium azide. The experimental temperature was set at 25°C. Each experimental profile was composed of the addition of an initial aliquot of 0.4 l, followed by 22 aliquots of 1.8 l of the substrate or ligand solution. The time interval between two consecutive injections was 180 s. The data were further processed with NITPIC (51) and analyzed using an one-site model in SEDPHAT (52).