The Structure of the Pantothenate Kinase·ADP·Pantothenate Ternary Complex Reveals the Relationship between the Binding Sites for Substrate, Allosteric Regulator, and Antimetabolites*

Pantothenate kinase catalyzes the first step in the biosynthesis of coenzyme A, the major acyl group carrier in biology. In bacteria, regulation of pantothenate kinase activity is a major factor in controlling intracellular coenzyme A levels, and pantothenate analogs are growth-inhibiting antimetabolites. We have extended the structural information on Escherichia coli pantothenate kinase by determining the structure of the enzyme·ADP· pantothenate ternary complex. Pantothenate binding induces a significant conformational change in amino acids 243–263, which form a “lid” that folds over the open pantothenate binding groove. The positioning of the substrates suggests the reaction proceeds by a concerted mechanism that involves a dissociative transition state, although the negative charge neutralization of the γ-phosphate by Arg-243, Lys-101, and Mg2+ coupled with hydrogen bonding of the C1 of pantothenate to Asp-127 suggests different interpretations of the phosphoryl transfer mechanism of pantothenate kinase. N-alkylpantothenamides are substrates for pantothenate kinase. Modeling these antimetabolites into the pantothenate active site predicts that they bind in the same orientation as pantothenate with their alkyl chains interacting with the hydrophobic dome over the pantothenate pocket, which is also accessed by the β-mercaptoethylamine moiety of the allosteric regulator, coenzyme A. These structural/biochemical studies illustrate the intimate relationship between the substrate, allosteric regulator, and antimetabolite binding sites on pantothenate kinase and provide a framework for studies of its catalysis and feedback regulation.

Coenzyme A (CoA) 1 is the major acyl group carrier in living systems and is synthesized by a universal series of enzymatic steps beginning with vitamin B 5 (1,2). Escherichia coli is capable of de novo pantothenate biosynthesis (1) but can also import pantothenate from its environment via a sodiumdependent symport process (3,4). All of the genes and enzymes involved in the biosynthetic pathway have been identified in E. coli (1,2). The pathway is initiated by pantothenate kinase (PanK) (ATP:D-pantothenate 4Ј-phosphotransferase, EC 2.7.1.33), the product of the coaA gene (5). Cysteine is next added to the phosphopantothenate and is rapidly decarboxylated to form 4Ј-phosphopantetheine by a bifunctional polypeptide (Dfp, renamed CoaBC) (6). The adenine group is added by 4Ј-phosphopantetheine adenylyltransferase (CoaD) (7,8), and the 3Ј-ribose phosphate is added by the dephospho-CoA kinase (CoaE) (9).
PanK is a key regulatory enzyme of CoA biosynthesis. E. coli PanK is a homodimer of 36 kDa subunits, and the amino acid sequence contains an A-type ATP-binding consensus sequence, GXXXXGKS. PanK exhibits highly positive cooperative ATP binding and mediates a sequential ordered mechanism with ATP as the leading substrate (10). PanK activity is inhibited by non-esterified CoA and to a lesser extent by its thioesters, which competitively interfere with ATP binding (10,11). Recently, the crystal structure of PanK was determined in complex with either ATP or CoA (12). These structures revealed that ATP and CoA bound to the enzyme in distinctly different ways; however, their phosphate binding sites overlapped at Lys-101, explaining the kinetic competition between the CoA regulator and the ATP substrate. There is no structural information on the nature of the pantothenate binding site on the enzyme and its relationship to the cavities occupied by ATP and the CoA allosteric regulator.
All of the steps in CoA biosynthesis are encoded by widely expressed, essential genes, making them attractive targets for antibacterial drug discovery (13). E. coli PanK belongs to a family of bacterial enzymes that is distinct from the eukaryotic counterparts that carry out the same reaction (14 -16). Although these data suggest that targeting PanK may be a viable strategy to attack a subset of bacterial species, one potential drawback to this approach is that such inhibitors may not be broad spectrum, because there are different, uncharacterized PanK isoforms in important pathogens, such as Staphylococcus aureus and Pseudomonas aeruginosa (17). Several CoA analogs are inhibitors of key CoA-utilizing enzymes in vitro (18), and one successful antibacterial strategy is the design of CoA antimetabolites based on pantothenate (19,20). These compounds are phosphorylated by PanK and converted by CoaD and CoaE to ethyldethia-CoA, which is thought to interfere with cell growth by blocking a group of enzymes that utilize CoA and its thioesters (20). HoPan is another pantothenate antagonist that has been used to treat mental disorders in humans (21)(22)(23), but its effectiveness as an antibacterial agent has not been evaluated. N5-Pan and N7-Pan are much larger molecules than pantothenate (Fig. 1), and understanding the structural basis for the ability of these bulky molecules to function as PanK substrates is important for the design of other analogs. The goal of this study was to characterize the structure of the pantothenate binding site on PanK to determine its relationship to the leading substrate, ATP, the allosteric regulator, CoA, and the antimetabolites. Our data provide new insights into the catalytic mechanism, CoA feedback regulation, and design of pantothenate antimetabolites.

EXPERIMENTAL PROCEDURES
Protein Preparation-For crystallization experiments, the coaA gene encoding E. coli PanK (5) minus the first eight amino acids (residues 9 -316) was subcloned into pET21a and transformed into the E. coli BL21 (DE3) strain. The cells were grown at 37°C in Luria Bertani broth supplemented with 100 g/ml ampicillin to an absorbance of 0.6 at 600 nm and then induced with 1 mM isopropyl-1-thio-␤-D-galactopyranoside for 3 h. The harvested cell pellets from 2 liters of bacterial culture were resuspended in 150 ml of disruption buffer (30 mM HEPES, pH 7.5, 30 mM NaCl, 1 mM dithiothreitol, 1 mM EDTA, and 1 mM phenylmethylsulfonyl fluoride) and lysed with a French Pressure cell. The cell lysate was centrifuged at 30,000 ϫ g for 30 min. The supernatant was further centrifuged at 150,000 ϫ g for 60 min. After filtration, the crude lysate was applied to DEAE resin and eluted with a linear gradient of increasing salt concentration. Fractions containing PanK were pooled, concentrated, equilibrated to 1 M ammonium sulfate, and applied to phenyl-Sepharose 6B resin (Amersham Biosciences). PanK was eluted with a linear gradient of decreasing salt concentration. Fractions containing PanK were pooled, concentrated, and applied to Sephadex-75 gel filtration chromatography (Amersham Biosciences). The purified protein was concentrated to 30 mg/ml, dialyzed into storage buffer (20 mM Tris-Cl, pH 8.0, 1 mM dithiothreitol, and 1 mM EDTA), flash-frozen in liquid nitrogen, and stored at Ϫ80°C. For kinetic experiments, E. coli PanK was overexpressed from a pET-15b plasmid with an N-terminal His tag and purified with nickel-nitrilotriacetic acid-agarose and gel filtration chromatography as described previously (12). Protein concentrations were determined by the method of Bradford (24) using ␥-globulin as a standard.
Crystallization-Co-crystals of PanK, ADP, and pantothenate were grown by the hanging drop vapor diffusion method. Prior to crystallization, PanK was incubated overnight with 30 mM ADP, 30 mM pantothenate, and 30 mM magnesium nitrate. Hanging drops containing equal volumes of the protein and reservoir solutions were equilibrated at 18°C against the reservoir solution (11% polyethylene glycol 3,350 and 200 mM sodium citrate, pH 8.2). After several days, monoclinic crystals appeared and were allowed to grow for 1 week. The crystals were transferred to 50/50 Paratone-N/mineral oil for cryodata collection.
Data Collection and Structure Determination of the PanK⅐ADP⅐ Pantothenate Ternary Complex-Single wavelength native data were measured from a single crystal of the ternary complex with a MAR165 CCD detector operating at 100 K at the SER-CAT beamline 22-ID at the Advanced Photon Source of the Argonne National Laboratory. The data set was indexed, integrated, and scaled with the programs DENZO and SCALEPACK (26). The crystals belong to space group C2 with unit cell dimensions a ϭ 181.2 Å, b ϭ 181.7 Å, c ϭ 47.7 Å, and ␤ ϭ 104.8°. The data collection statistics are summarized in Table I.
The structure of the ternary complex was determined by the molecular replacement method using one subunit of a dimer of the PanK⅐AMP-PNP binary complex (12) as a search model. The crossrotation and translation functions and Patterson correlation refinement were calculated using the program CCP4 (27). Data between 10.0 and 4.0 Å were used with the rotation function and generated four outstanding solutions corresponding to four PanK molecules in the asymmetric unit. Multiple iterations of applying the translation function to each solution allowed determination of the orientations and positions of the other three molecules. Each of the four PanK molecules forms a biologically functional dimer with a symmetrically related molecule. The model of all four molecules was subjected to rigid body refinement using data between 6.0 and 4.0 Å in the program CNS (28). The high resolution restriction was set at 4.0 Å to allow for a possible large positional shift in the structure. The low resolution limit was set at 6.0 Å because intensities of the reflections below this resolution are severely affected by the diffraction of solvent molecules in the crystal. The R-factor at this stage was high (49%), indicating that the conformation of the PanK⅐ADP⅐pantothenate ternary complex was significantly different from that of the search model, the AMP-PNP-bound protein. Solvent contributions to low resolution intensities were considered during the first refinement. After subsequent simulated annealing (28) using data between 20.0 and 2.2 Å resolution, the quality of the structure was significantly improved: the working R-factor was 34.8%; the free Rfactor, 36.7%. The F o Ϫ F c difference map showed continuous densities for both ligands, ADP and pantothenate, that were absent from the search model, confirming the correct molecular replacement solution. Several iterations of model building in the program O (29) and refinement in the program CNS (28) further lowered both R-factors. The electron density map of the ADP and pantothenate ligands is shown in Fig. 2A. The final refinement statistics are shown in Table I. Because of slight differences between the four subunits, non-crystallographic symmetry restraints were applied to only 53% of the total residues for each subunit. In the final model, insufficient electron density led to substitution of several side chains with alanine: in the first molecule, residues Glu-33, Glu-44, Arg-71, Asn-83, Arg-86, Arg-119, Leu-178, Asp-185, Lys-246, Lys-264, and Glu-268; in the second molecule, residues Lys-40, Asn-83, Gln-85, Arg-86, Arg-119, Arg-120, Leu-178, Asp-185, Lys-246, and Glu-268; in the third molecule, residues Glu-33, Lys-40, Asn-83, Arg-86, Arg-119, Lys-137, and Leu-178; in the fourth molecule, residues Glu-33, Lys-40, Gln-85, Arg-86, Arg-119, Arg-120, Lys-137, Asp-185, Asp-256, Lys-264, and Glu-268. Also, several residue backbones could not be properly fit into the calculated maps: in the first molecule, residues Ser-26-Pro-28, Thr-30, Gly-84, and Gln-192; in the second molecule, residues Asp-25-Thr-30; in the third molecule, residues Asp-25-Pro-28, Gly-84, Gln-85, Ile-87, His-118, and Arg-120; in the fourth molecule, residues Ser-26-Pro-28 and Val-191.
Assay of PanK-The standard PanK assays contained D-[1-14 C]pantothenate (45 M; specific activity 55 mCi/mM), ATP (100 M), MgCl 2 (10 mM), Tris-HCl (0.1 M, pH 7.5), and 100 ng of the purified PanK in a total volume of 40 l (10). The mixture was incubated at 37°C for 10 min. The reaction was stopped by adding 4 l of 10% (v/v) acetic acid to the mix. Then 40 l of the mixture was deposited onto a Whatmann DE81 ion-exchange filter disk that was washed in three changes of 1% acetic acid in 95% ethanol (25 ml/disk; 20-min wash) to remove unreacted pantothenate. 4Ј-phosphopantothenate was quantitated by counting the dried disk in 3 ml of scintillation solution. A second assay was used to determine the utilization of the pantothenate analogs by PanK. The reaction mix contained pantothenate or pantothenate analog at the concentrations indicated in the figures, [␥-32 P]ATP (0.25 mM; specific activity 1 Ci/mM), MgCl 2 (10 mM), Tris-HCl (0.1 M, pH 7.5), and 100 ng of PanK. After incubation at 37°C for 20 min, the reaction was stopped by adding 4 l of 0.5 M EDTA. A 10-l aliquot of the reaction mixture was spotted onto an activated silica gel H plate, which was developed with butanol/acetate/water (5:2:4, v/v/v). The dried plate was exposed to a storage phosphor screen, and the phosphorylated product was quantitated using ImageQuant 5.2 software (Amersham Biosciences).
Synthesis of Pantothenate Analogs-The synthetic procedures were adapted from the procedure of Strauss and Begley (20). Sodium pantothenate (4.5 g, 18.6 mM) was dissolved in 20 ml of methanol and passed through an Amberlite IR-120 (H ϩ ) ion exchange column. The column was subsequently eluted with 200 ml of methanol until neutral. The elution solvent was then removed under high heat and vacuum on a rotary evaporator to complete dryness. The resulting free acid obtained in this procedure (4 g, 18.2 mM) was a viscous, straw-colored oil, which was then redissolved in 10 ml of dry dimethylformamide. The 10-ml solution was split into two 5-ml fractions and placed into 100-ml reaction vessels. To each solution an additional 10 ml of dry dimethylformamide was added. For the synthesis of N5-Pan, amylamine (1.16 ml, 10 mM) and diphenylphosphoryl azide (2.24 ml, 10 mM) were added to the first pantothenate solution. For the synthesis of N7-Pan, heptylamine (1.48 ml, 10 mM) and diphenylphosphoryl azide (2.24 ml, 10 mM) were added to the second pantothenate solution. The reaction mixtures were then cooled to 0°C, and triethylamine (1.39 ml, 10 mM) was added to each. The reactions were stirred at 0°C for 2 h, followed by stirring at room temperature overnight. The reaction volumes were reduced by rotary evaporation with high vacuum to remove dimethylformamide. Each of the products was then purified by flash column chromatography on silica gel using a linear elution gradient of 100:0 to 95:5 chloroform/ methanol. The fractions containing each of the products were collected and concentrated in vacuo, yielding both products as white powders (1.6 g, 60% yield, and 2.1 g, 71% yield, respectively).
HoPan was synthesized by dissolving 4-aminobutyric acid (4.5 g, 43.6 mM) in methanol with a excess of diethylamine (6.3 ml, 60.9 mM). The mixture was warmed for 5 min, and the solvent and excess diethylamine was removed in vacuo. The resulting diethylamine salt was redissolved in methanol (100 ml), and (R)-pantolactone (5.7 g, 43.8 mM) was added. The reaction mixture was warmed at 60°C for 8 h and the solvent subsequently removed in vacuo. The residue was dissolved in water and applied to an Amberlite IR-120 H ϩ ion exchange column. The column was washed with water until neutrality. The eluate was then extracted twice with dichloromethane to remove the unreacted pantolactone, and the aqueous layer was evaporated to dryness to yield HoPan (9.2 g, 92% yield).
Molecular Modeling-To investigate the binding conformation of the N5-Pan ligand, a molecular dynamics simulation was performed using the Dynamics function available with the Sybyl 6.9 molecular design software package (Tripos). The active site of the PanK enzyme was defined as all residues located within 7 angstroms of the pantothenate ligand in the crystal structure. The pantothenate ligand was extracted from the structure and modified into the N-pentyl ligand, which was then merged back into the active site. Thus, the starting conformation for the ligand in this dynamics run was essentially the same as the crystal structure, with the exception of the N-pentyl tail. PM3 charges were loaded onto the ligand, and Kollman charges were used for the enzyme. Waters that appeared to make key hydrogen bond interactions within the active site or that simulated surrounding water molecules on the surface of the enzyme were included in the simulation; all others were ignored. For this dynamics run, only the atoms and bonds of the ligand were allowed to move; all others in the enzyme and water molecules were held rigid. Thus, although the ligand was able to move and respond to the steric and electrostatic fields of the enzyme, the enzyme and water molecules could not move in return.
This Dynamics simulation ran for 100 ps using the NTV ensemble, with steps every 1 fs and snapshots every 5 fs. The first 5 ps were at 100°K, the next 5 were at 300 K, the next 5 were at 700 K, and the last 85 were at 300 K. For the purposes of this Dynamics simulation, the Sybyl default values of 100 maximum iterations, 0.0001 tolerance, and 100-fs coupling were used. The system was raised to 700 K purposefully to give the ligand the necessary energy to pass through high energy conformations to achieve a lower energy conformation. The resulting solution is an average conformation of the simulation.

RESULTS
Nucleotide Binding Site-The three-dimensional fold of PanK in the PanK⅐ADP⅐pantothenate ternary complex is almost identical to that of the protein in previously published structures (Fig. 2B) (12). In the PanK⅐ADP⅐pantothenate structure, ADP is bound in a groove formed by residues from the P-loop and the loop connecting strands 10/11 (Fig. 2C), similar to the location of the ATP analog in the PanK⅐AMP-PNP binary complex (Fig. 2D) (12). A comparison of the nucleotide binding site for the PanK⅐ADP⅐pantothenate and PanK⅐AMP-PNP structures shows that the significant differences are the interaction of ADP with the Arg-243 side chain and the presence/ absence of a bound Mg 2ϩ ion. In the ternary complex, the side chain of Arg-243 is displaced relative to its position in the PanK⅐AMP-PNP structure in the direction of the nucleotide, and the guanidinium group of Arg-243 interacts with nonbridging oxygens of the ADP ␣and ␤-phosphates (Fig. 2C). The same guanidinium group of Arg-243 in the AMP-PNP structure forms a salt-bridge interaction with non-bridging oxygen of ␥-phosphate of AMP-PNP (12) (Fig. 2D). In the PanK⅐AMP-PNP structure, Mg 2ϩ is coordinated by the nucleotide ␤and ␥-phosphates and the side chains of Ser-102 and Glu-199, suggesting the role of Mg 2ϩ in ATP binding (Fig. 2D). However, the PanK⅐ADP⅐pantothenate structure lacks Mg 2ϩ (Fig. 2C). A F o Ϫ F c difference map of ADP bound at the nucleotide binding site shows full density of the bound ADP, indicating that ADP is stably bound with high occupancy (Fig. 2A). This finding suggests that Mg 2ϩ , although necessary for ATP binding and the phosphoryl transfer reaction, is not a requirement for ADP binding.
Pantothenate Binding Site-The PanK⅐ADP⅐pantothenate structure offers the first view of the pantothenate substrate bound at the active site and identifies key residues conferring pantothenate substrate specificity. The pantothenate binding site is located at the distal end of a large surface groove starting c R free is equivalent to R work except that 5% of the total reflections were set aside for an unbiased test of the progress of the refinement. on one side of the protein, where the nucleotide binds (12), and continuing across the protein (Fig. 2C). Above the groove, the pantothenate binding site is bounded by ␣-helices H and I, which provide important contacts with the substrate. The pantothenate molecule is surrounded by hydrophilic residues, providing multiple hydrogen bonding opportunities (Fig. 3). The C1 hydroxyl group of pantothenate forms a hydrogen bond with the side chain carboxyl of Asp-127. This interaction illustrates the role of Asp-127 in deprotonation of the C1 hydroxyl group, facilitating the phosphoryl transfer reaction as we suggested based on the PanK⅐AMP⅐NP structure (12). The 2Ј-hydroxyl group of pantothenate is hydrogen-bonded to the side chain imidazole of His-177, which explains how the enzyme preferentially interacts with the R enantiomer of pantothenate. The carbonyl oxygen of pantothenate is indirectly hydrogen-bonded to the hydroxyl group of Tyr-175 via a water molecule. The other half of the pantothenate molecule is oriented by hydrogen bond interactions of the carboxyl group of pantothenate with the side chains of Tyr-240 and Asn-282, a hydrogen bond between the amide nitrogen of pantothenate and Asn-282, and a hydrogen bond between the amide nitrogen of pantothenate and the backbone carbonyl of Lys-145 (not shown). Also, steric allowance for the two pantothenate methyl groups is provided by the absence of a side chain in Gly-146 (not shown). All of the residues that interact with pantothenate are invariant in 13 prokaryotic PanK (CoaA) sequences, whereas only Asp-127 and Tyr-240 are conserved in the different eukaryotic PanK sequences (for sequence alignments, see Refs. 12,14).
Induced Fit Binding of Pantothenate-The comparison of the PanK⅐AMP-PNP structure to the PanK⅐ADP⅐pantothenate structure shows a large movement of the helix-H/loop region, corresponding to amino acids 243-263, which acts as a lid that closes over the pantothenate binding site (Fig. 4). The most significant conformational changes involve Glu-249 and Phe-259. In the PanK⅐AMP-PNP structure, the carboxyl side chain of Glu-249 occupies the pantothenate binding groove in approximately the same location as will be occupied by the incoming pantothenate (12). The binding of the carboxyl group of pantothenate in the groove displaces the carboxyl side chain of Glu-249 to the surface of the molecule (Fig. 4A). Glu-249 and three adjacent residues (Arg-248, Gly-250, and Ala-251) become part of helix-H, extending its length by one turn longer compared with the PanK⅐AMP-PNP structure. In the PanK⅐AMP-PNP structure, the side chain of Phe-259 interacts with the aliphatic portion of the Glu-249 side chain (12). When Glu-249 moves to the surface of the molecule to become a part of helix-H, Phe-259 loses its interaction with Glu-249 and moves close to helix-H to form a hydrophobic interaction with the side chain of Phe-247 of helix-H (Fig. 4B). These structural rearrangements caused by pantothenate binding modify the active site, which is necessary to accommodate pantothenate and bring together aromatic residues Phe-244, Tyr-258, Phe-259, and Tyr-262 to form a hydrophobic pocket. This constellation of residues extends over the top of the carboxyl group of pantothenate but does not interact with pantothenate itself. A similar structural rearrangement of the helix-H/loop region with the assembly of aromatic residues that forms a hydrophobic pocket is observed in the PanK⅐CoA structure, where the thiol group of CoA was surrounded by aromatic resides in a hydrophobic pocket (12). The structural rearrangement of the helix-H/loop region in the PanK⅐CoA structure is facilitated by the phosphopantetheine moiety of CoA mimicking pantothenate binding to the enzyme.
Reactivity of N5-Pan and N7-Pan-Previous work established that N5-Pan in the presence of a mixture of PanK (CoaA), 4Ј-phosphopantetheine adenylyltransferase (CoaD), and dephospho-CoA kinase (CoaE) was converted to the CoA analog, ethyldethia-CoA (20). Our experiments focused on examining the interactions between N5-Pan or N7-Pan (Fig. 5) and PanK. Both pantothenamides were effective inhibitors of PanK, exhibiting IC 50 s of 60 M under our standard assay conditions (Fig. 5A). These inhibitors were competitive with respect to pantothenate (not shown); therefore, we determined the experimental K m for these compounds as substrates for PanK. The pantothenate K m was 41 M, essentially the same as reported previously (10). Both N5-Pan and N7-Pan were also excellent substrates for PanK, exhibiting K m values of 140 and 124 M, respectively (Fig. 5B). Our K m values are the same as reported previously using a direct radiochemical assay (10) but are higher that those reported using a two-enzyme-coupled assay (20). The reason for this discrepancy is not clear, but in all cases the apparent binding of the analogs to PanK was 3-fold less than that of the authentic substrate, illustrating that N5-Pan and N7-Pan are efficiently utilized by PanK.
Relationship between the Binding Sites for Pantothenate, CoA, and N5-Pan-Structural comparison of the PanK⅐ADP⅐ pantothenate and PanK⅐CoA complexes shows that even though there are significant changes in the orientation of several structural elements, the space occupied by pantothenate in the ternary complex is essentially the same as that occupied by the pantothenate moiety of CoA in the PanK⅐CoA complex (Fig. 6). The structural basis for the utilization of the pantothenamides by PanK was explored using the Sybyl program to dock N5-Pan into the PanK⅐ADP⅐pantothenate structure. The fact that N5-Pan and N7-Pan are bulkier than pantothenate yet serve as efficient PanK substrates is unusual because most small molecule substrates are bound with high specificity. A superposition of the highest scoring docking solution of N5-Pan in the PanK⅐ADP⅐pantothenate structure shows that the analog's hydroxyl group that accepts the phosphate from ATP resides in the same location as the pantothenate hydroxyl group, whereas the extended hydrophobic chain of N5-Pan interacts with the hydrophobic surface that extends over the carboxyl group of pantothenate that is lined by the aromatic residues of the helix-H/loop (Fig.  6, A and B). The same hydrophobic residues form the binding groove for the ␤-mercaptoethylamine moiety of the CoA allosteric regulator (Fig. 6C). Thus, similar structural features of PanK form a flexible binding site that accommodates the pantothenate substrate, the CoA feedback inhibitor, and the pantothenamide antimetabolites.
HoPan was neither an inhibitor nor a substrate for E. coli PanK (Fig. 5A). Also, HoPan does not have any detectable antibacterial activity against E. coli (data not shown). These experiments indicate that HoPan does not form a productive complex with PanK. The extension of the carboxyl group of pantothenate by one carbon in HoPan results in the loss of hydrogen bond interactions with both Tyr-240 and Asn-282. The interaction with Tyr-240 on helix-H is important to the conformational change that reorganizes the structure for pantothenate binding, and the inability of HoPan to make this contact may prevent the substrate binding pocket from forming. Docking solutions using HoPan and the PanK⅐ADP⅐ pantothenate structure with the pantothenate removed result in binding predictions that do not place the hydroxyl group in a position to be phosphorylated (data not shown). These structural considerations explain why this pantothenate analog does not interact with bacterial PanK. DISCUSSION The geometry of the ligands in the PanK⅐ADP⅐pantothenate ternary complex points to a concerted mechanism for phosphoryl transfer. An associative mechanism (an axial bond distance of 1.617 Å and, thus, the sum of the axial bond orders approaching 2.0) involves a pentavalent phosphorane intermediate, whereas the dissociative mechanism (an axial bond distance Ͼ3.3 Å, the sum of the axial bond orders of 0.0) proceeds via a trigonal planar metaphosphate intermediate (30). In practice, most enzymes fall between these two extremes, based on the reaction coordinate distance used to calculate the fractional bond order (31). In concerted mechanisms, one observes a dissociative transition state when the sum of the axial bond orders is less than unity, an S N 2 mechanism when the sum of the axial bond orders in the transition state is close to unity, or an associative transition state when the sum of the axial bond orders is more than unity (30). In the structure of the PanK⅐ADP⅐pantothenate complex, the distance between the hydroxyl nucleophile of pantothenate (the incoming oxygen atom) and the ␤-phosphate oxygen atom of ADP (a leaving oxygen atom) is 5.1 Å. This distance translates into an average bond distance of transferred phosphorus to the entering and leaving oxygens of 2.55 Å. Using Pauling's equation (31,32) and considering the value of 1.617 Å as the single covalent bond distance between phosphorus and oxygen (33), the calculated fractional bond order is 0.08 for the axial bonds between transferred phosphorus to the entering and leaving oxygens and thereby the sum of the axial bond orders of 0.16, suggesting a concerted mechanism with the dissociative transition state. A small torsional rotation alone of the pantothenate hydroxyl group would shorten the distance between the hydroxyl oxygen and the ␤-phosphate oxygen, thereby increasing the sum of the axial bond orders to 0.36, which is still typical for enzymatic reactions with the dissociative transition state (30). Taken together, the bond distance measurements suggest that PanK catalyzes phosphoryl transfer via a concerted mechanism that involves dissociative characteristics.
A metaphosphate transition state will have a charge of Ϫ1, whereas a phosphorane group in an associative transition state will have a charge of Ϫ3 (25). The structures of PanK⅐ADP⅐ pantothenate and PanK⅐AMP-PNP show that positively charged residues (Lys-101 and Arg-243) and a Mg 2ϩ ion interact with the ␥-phosphate, and the presence of these positive charges is consistent with a requirement to neutralize multiple negative charges that develop in the associative transition state. In addition, Asp-127 interacts with the C1 hydroxyl of pantothenate to increase its nucleophilicity, promoting an S N 2like attack on the ␥-phosphate of ATP (Fig. 7). In our proposed mechanism, pantothenate is bound and oriented by hydrogen bonding interactions with Tyr-240, Asn-282, Tyr-175, and His-177. Asp-127 then activates the C1 hydroxyl oxygen of pantothenate, which attacks the ␥-phosphate of ATP. The in-line transfer of the phosphate occurs with charge stabilization pro-vided by Lys-101, Arg-243, and Mg 2ϩ . Therefore, charge neutralization of the ␥-phosphate and the putative role of Asp-127 as a general base leave room for other interpretations of the phosphoryl transfer mechanism of PanK. Further experiments will be required to confirm that PanK catalyzes phosphoryl transfer via a concerted mechanism with a dissociative transition state.
The PanK⅐ADP⅐pantothenate ternary complex structure also offers an explanation for how the bulky pantothenamide antimetabolites are capable of being phosphorylated by PanK. N5-Pan and N7-Pan are effective inhibitors of bacterial cell growth and are thought to act through their conversion by PanK to phosphorylated intermediates that are subsequently utilized to produce inactive CoA analogs (see the Introduction). These findings, coupled with our observation that the affinity of PanK for the analogs is comparable with the authentic substrate pantothenate (Fig. 5), are unusual because enzymes that utilize small molecules are highly selective for their substrates. The flexibility of the pantothenate binding site may account for the ability of PanK to phosphorylate the pantothenamides. The pantothenate binding site is also part of the binding site for CoA, and despite the structural differences between pantothenate and the allosteric regulator, the binding of CoA induces a similar conformational change in PanK. Furthermore, the pantothenate moiety of bound CoA has the same location and orientation as the bound pantothenate substrate, whereas the ␤-mercaptoethylamine moiety of CoA extends into the hydrophobic dome over the pantothenate binding pocket (Fig. 6B). Therefore, pantothenamide binding is predicted to induce similar conformational changes in the pantothenate binding site. A modeling solution of N5-Pan with the PanK⅐ADP⅐pantothenate structure (with the pantothenate removed) supports the pro- posal that the hydroxyl of the pantothenamide is correctly oriented for phosphorylation by binding in the same conformation as the pantothenate substrate. The hydrophobic tail of the antimetabolite interacts with the hydrophobic domain (Fig.  6B), which binds the ␤-mercaptoethylamine moiety of CoA (Fig.  6C) (12). This model suggests that the design of future pantothenamide derivatives with hydrophobic groups that more strongly interact with the aromatic residues of the hydrophobic roof over the pantothenate pocket will increase the affinity and specificity of the inhibitor/substrates.