Probing Ligand-binding Pockets of the Mevalonate Pathway Enzymes from Streptococcus pneumoniae*

Diphosphomevalonate (Mev·pp) is the founding member of a new class of potential antibiotics targeting the Streptococcus pneumoniae mevalonate (Mev) pathway. We have synthesized a series of Mev·pp analogues designed to simultaneously block two steps in this pathway, through allosteric inhibition of mevalonate kinase (MK) and, for five of the analogues, by mechanism-based inactivation of diphosphomevalonate decarboxylase (DPM-DC). The analogue series expands the C3-methyl group of Mev·pp with hydrocarbons of varying size, shape, and chemical and physical properties. Previously, we established the feasibility of a prodrug strategy in which unphosphorylated Mev analogues could be enzymatically converted to the active Mev·pp forms by the endogenous MK and phosphomevalonate kinase. We now report the kinetic parameters for the turnover of non-, mono-, and diphosphorylated analogues as substrates and inhibitors of the three mevalonate pathway enzymes. The inhibition of MK by Mev·pp analogues revealed that the allosteric site is selective for compact, electron-rich C3-subsitutents. The lack of reactivity of analogues with DPM-DC provided evidence, counter to the existing model, for a decarboxylation transition state that is concerted rather than dissociative. The Mev pathway is composed of three structurally and functionally conserved enzymes that catalyze consecutive steps in a metabolic pathway. The current work reveals that these enzymes exhibit significant differences in specificity toward R-group substitution at C3 and that these patterns are explained well by changes in the volume of the C3 R-group-binding pockets of the enzymes.

Streptococcus pneumoniae, the primary cause of bacterial meningitis and pneumonia, kills over 4000 people daily worldwide and disproportionately affects children and the elderly (1,2). Antibiotic resistance is a major problem in fighting this organism, with multiple-drug resistance rates as high as 95% in some regions (3). Although vaccines targeting the 7 or 23 most prevalent strains have shown success in reducing disease incidence in developed countries (4), there is a continual need for new antibiotic strategies to combat unvaccinated strains, which are rapidly filling the biological niches left by the vaccine (5).
The mevalonate (Mev) 3 pathway is essential for the survival of S. pneumoniae in lung and serum (6). The bacterium uses this pathway to convert Mev to isopentenyl diphosphate, the "building block" of the isoprenoids: a class of 25,000 unique molecules having a wide range of biological functions. The pathway consists of three GHMP family kinases: Mev kinase (MK), phosphomevalonate kinase (PMK) and diphosphomevalonate decarboxylase (DPM-DC) ( Fig. 1) (7). Mutations that knock out genes in this pathway kill the organism in vivo, suggesting that S. pneumoniae cannot obtain the necessary precursors or downstream products from the host (6). In principle, each of the three enzymes is an antibiotic target, because inhibition of any of them prevents the production of isopentenyl diphosphate. We have shown that S. pneumoniae MK is potently (K i ϭ 500 nM) allosterically inhibited by diphosphomevalonate (Mev⅐pp), the third compound in the pathway, whereas the human MK homologue is not (8). Thus, the allosteric site provides an opportunity to selectively target the bacterium. The S. pneumoniae MK crystal structure revealed a pore at the subunit interface with excellent charge-and shape-complementarity to Mev⅐pp that may be the allosteric site.
In an effort to enhance the inhibitory properties of Mev⅐pp for use as an antibiotic, we have built a series of ten Mev analogues (9) in which the C 3 -methyl group has been altered to other hydrocarbon substituents ( Table 1). Five of the analogues have linear and branched alkyl groups at C 3 that act as structural probes for the MK allosteric site and the active sites of the three GHMP kinases. The remaining five analogues resemble the alkyl series but also have the potential to act as mechanismbased inhibitors of DPM-DC.
In the DPM-DC mechanism, the C 3 -hydroxyl of Mev⅐pp is phosphorylated by ATP, generating p⅐Mev⅐pp, which ionizes, leaving a carbocation on C 3 ; decarboxylation is thought to follow rapidly (10). Abeles and coworkers have provided evidence with the mammalian enzyme that a carbocation must form during turnover by showing that substitution of the C 3 -methyl group with a hydrogen or a fluoromethyl group causes the reaction to halt after phosphorylation, implying that the electrondonating methyl group is required to stabilize the carbocation inductively (10). To exploit the potential carbocation formation for DPM-DC inhibition, we replaced the C 3 -methyl group with substituents that could undergo resonance with the carbocation, thereby setting up an electrophilic site for nucleophilic attack by the enzyme (Fig. 2). Because the inhibitors are structurally related to Mev⅐pp, we expected that they might also inhibit MK through binding at the allosteric site. By targeting two steps in the same pathway, such molecules are expected to have enhanced antibiotic potency.
We now report the chemical and chemoenzymatic synthesis of the Mev⅐pp analogue series, the kinetic parameters of the Mev, Mev⅐p, and Mev⅐pp analogues as substrates of their three respective Mev pathway GHMP kinases, and the inhibition of MK and DPM-DC by Mev⅐pp analogues, noting their potential as antibiotics and the substrate-binding features of the enzyme family that they reveal.
Chemical Nomenclature-In the description of kinetic experiments, the substrate analogues (in free acid form) are referred to by boldface numbers that correspond to the struc-tures given in Table 1. Where appropriate, numbers are appended with "⅐p" or "⅐pp" to indicate the number of phosphates on the C 5 -hydroxyl. For example, 6⅐pp refers to the 5Ј-diphosphorylated form of the vinyl analogue (see Table 1). A group of analogues of a particular phosphorylation state is referred to using a boldface X in place of the number (i.e. X, X⅐p, or X⅐pp).
Enzymatic Synthesis of Stereochemically Pure Substrates of the Mevalonate Pathway-A series of (R,S)-mevalonolactone analogues ((R,S)-X) were synthesized as racemic mixtures as described (9). Approximately 25 mg of each lactone was treated with five equivalents of KOH at 37°C for 1 h to convert it to the corresponding carboxylate. The solution was then adjusted to pH 7.0 with 1 M HCl and diluted with Hepes/K ϩ (50 mM, pH 7.0) and MgCl 2 (1.0 mM) to a final concentration of ϳ240 mM in 750 l. Concentration of the R-isomer was determined by enzymatic assay (see below).
Stereochemically pure mono-and diphosphorylated Mev analogues ((R)-X⅐p and (R)-X⅐pp (X ϭ 1, 3, 6, and 7)) were enzymatically synthesized from racemic Mev analogues (R,S)-X in successive steps catalyzed by MK and PMK. The first phosphorylation reaction contained S. pneumoniae MK (0.350 M), (R,S)-X (ϳ7.5 mM), ATP (4.5 mM), PK (10 units/ml), PEP (10 mM), 2-ME (20 mM), MgCl 2 (5.5 mM), and Hepes/K ϩ (50 mM, pH 8.0) After 72 h at room temperature, 98% of the starting material had been converted to the (R)-X⅐p analogue. The second phosphorylation was initiated by the addition of PMK (0.28 M) and PEP (1.0 mM) to the reaction mixture, and this reaction achieved 99% conversion to (R)-X⅐pp analogue after ϳ24 h. Because (R,S)-9 was known to be a slow substrate for S. pneumoniae MK, the production of (R)-9⅐pp was facilitated using a one-pot reaction that included S. aureus MK, which is weakly inhibited by Mev⅐pp, and PMK. The reaction was completed in 36 h. (R)-9⅐p was synthesized as described above, but starting from (R)-9. (R)-X compounds (X ϭ 1, 3, 6, 7, and 9) were generated by removing the pyrophosphoryl moiety from two-thirds of the purified (see below) (R)-X⅐pp using alkaline phosphatase. The reaction mixtures contained: alkaline phosphatase (2.5 units/ml), (R)-X⅐pp (2.4 mM), Tris/HCl (50 mM, pH 8.5) at 37°C (98% conversion to product was reached in 1 h). The (R)-X analogues were found to lactonize upon purification (see below) and were saponified to the carboxylates as described above.
Reaction progress and analogue concentration were monitored by enzymatic assay (12). MK, PMK, and DPM-DC produce ADP in the presence of their respective substrates; ADP formation was measured spectrophotometrically at 339 nm by stoichiometrically coupling (1:1) ADP production to NADH oxidation using the well established PK/LDH-coupled assay . For each mole of (R,S)-X in the initial reaction, 0.5 mol of product was produced, representing a biologically relevant (R)-isomer. ADP (2.0 mM) was added at the end of synthesis reactions to completely convert PEP to pyruvate, facilitating purification of the phosphorylated Mev species.
Determination of Initial Rate Constants-Initial rate constants of MK, PMK, and DPM-DC reactions with substrate analogues were determined by a progress curve method. Complete MK reaction progress curves were obtained by monitoring absorbance at 398 nm. Products were removed by including PMK and DPM-DC in the reaction mixture to prevent possible product inhibition; therefore, three molar equivalents of NADH are oxidized per MK substrate. (R)-1, -3, -6, -7, and -9 were previously shown to be substrates for MK (9).  (3,15,15), (6, 1.5, 1.0), (7, 2.0, 3.0), and (9, 3.0, 2.0). The maximum concentrations of (R)-1, -3, -6, -7, and -9 that could reasonably be achieved in the assay were too low relative to their K m to obtain precise K m values; consequently, only V/K values were determined (see Table 1).
(R,S)-2, -4, -5, -8, and -10 proved to be extremely poor substrates for MK; product was not detected under the following conditions: (  The dissociative model of Mev⅐pp decarboxylation (shown for the 9⅐pp analogue) begins with phosphorylation of the C 3 -OH by ATP. Phosphate then ionizes, leaving a C 3 carbocation, which either rapidly decarboxylates to form the double-bonded product (left resonance form) or, in the case of analogues 6⅐pp through 10⅐pp, rearranges and is quenched by a nucleophile on the protein surface (right resonance form), forming a covalent adduct. Opening the cyclopropyl ring of the 9⅐pp analogue results in a homoallyl carbocation that is stabilized, relative to the ring-intact carbocation, as a result of the release of the ϳ27 kcal/mol ring strain (29,30).
PMK reaction progress curves were measured by monitoring absorbance at 339 nm. The product was removed by including DPM-DC in the reaction mixture to prevent possible product inhibition; therefore, two molar equivalents of NADH are pro- At the end of each reaction, additional equivalents of the analogue and NADH were added to the cuvette, and a second progress curve was obtained; upon comparison to the first curve, it was found to be identical, indicating the absence of product inhibition at the concentrations of product generated by the reaction.
Progress Curve Analysis-Values of k cat and K m for each enzyme with its substrates were determined by statistically fit-ting complete reaction progress curves to the integrated Michaelis-Menten equation (Equation 1), using a FORTRAN77 program of Cleland (15) adapted by one of us (S. T. L.) for this purpose (available upon request). Prior to analysis, the linear background ATP hydrolysis was subtracted from the curve, and the data were manually truncated to remove the lag in NADH oxidation due to coupling enzymes. Absorbance values were normalized to the minimum absorbance value and converted to substrate concentrations (S t ) using the extinction coefficient of NADH and equivalents of ADP produced in each reaction (see above). The initial substrate concentration (S 0 ) was taken from the corrected, normalized absorbance at t ϭ 0.
Fitted parameters k cat and K m (p n ) and associated errors from two or three independent progress curves of each enzyme-substrate pair were averaged (16). Mean parameter values (p) and their associated errors ( p ) were determined using Equations 2 and 3. Weighting factors (a n ) were obtained from Equation 4, where the variance ( n 2 ) is the square of the standard deviation ( n ) of the least-squares fitted parameter.
S. pneumoniae DPM-DC Homology Model with Mev⅐pp-A homology model of the S. pneumoniae enzyme was generated in Swiss PDB Viewer (17). from the crystal structure of S. pyogenes DPM-DC (70% sequence identity, PDB: 2GS8). The fifteen active-site residues used to position Mev⅐pp in the model are completely conserved in the two sequences. Mev⅐pp was manually positioned using the criteria of Byres et al. (18), which were used to position Mev⅐pp in the apo Trypanosoma brucei DPM-DC structure (PDB: 2HKE). To facilitate comparison of the two structures, the T. brucei structure was aligned with the S. pneumoniae homology model, using PyMOL (19). The Mev⅐pp ligand was then positioned using the criteria below to determine placement of the terminal phosphate, the C 1 -carboxylate, and the C 3 -hydroxyl. The terminal phosphate of Mev⅐pp in the T. brucei model was positioned where a sulfate is bound by four highly conserved residues in the crystal structure; two of the four residues coordinating this sulfate are conserved in S. pneumoniae DPM-DC (Lys-22 and Gly-140), the third is conservatively replaced (Lys-74 for Arg-77), and Arg-186 replaces the fourth residue, Thr-200. In the T. brucei structure, the helix containing Arg-77 is positioned closer to the sulfate than the corresponding helix in S. pneumoniae, such that the corresponding Lys-74 is 9 Å from the sulfate. It is therefore possible that Arg-186 functionally replaces Lys-74. With these considerations, the terminal phosphate of Mev⅐pp was placed within hydrogen bond distance (2.5-4.0 Å) of Lys-22, Gly-140, and Arg-186 in the S. pneumoniae model, guided by the structural alignment. The C 1 -carboxylate, which in the T. brucei model is coordinated by Arg-149, was positioned within 4 Å of the corresponding Arg-144 and the conserved Ser-141 side chain in the S. pneumoniae model. This Ser residue is conserved in GHMP kinases that bind Mev, Mev⅐p, and Mev⅐pp and forms a hydrogen bond with the C 1 -carboxylate in all three ligand-bound crystal structures (PDB: 2HFU, 2OI2, and 3GON). The C 3 -hydroxyl group, which is the nucleophile that attacks the ATP ␥-phosphate, is positioned in the T. brucei model near Asp-293, a highly conserved residue in GHMP kinases that assists phosphoryl transfer by deprotonating the C 3 -hydroxyl (18). In the S. pneumoniae model, the C 3 -hydroxyl was positioned within 3.5 Å of Asp-276 and points toward the conserved ATP-binding pocket. To make these contacts, the C 2 -C 3 bond angle of Mev⅐pp was rotated ϳ120°r elative to its conformation in the MK crystal structure. This change puts the C 1 -carboxylate antiperiplanar to the C 3 -hydroxyl that departs during decarboxylation, making the optimal geometry for a concerted elimination reaction, one of the possible mechanisms for this enzyme. To determine the inhibition mechanism and kinetic constants for 6⅐pp and 9⅐pp inhibition, initial rates were measured for a matrix of conditions comprising three inhibitor concentrations and three or four fixed-variable substrate concentrations with the other substrate held constant. When Mev was varied, the assay contained MK (2.5 nM), Mev Diphosphomevalonate Decarboxylase Electrophilic Inactivation-Inactivation of S. pneumoniae DPM-DC by putative X⅐pp electrophiles (X ϭ 6-10) was tested as a function of turnover. Reactions were initiated by the addition of the diphosphorylated analogues and allowed to proceed until Ͼ1000 enzymeequivalents of product had formed. Inactivation was measured as a percent reduction in velocity after 1000 turnovers at saturating substrate. The 1 H NMR spectrum of the reaction mixture was taken at t 0 and every 7 min until Ͼ95% conversion of product was reached. 1 H-13 C HSQC and HMBC spectra were taken on the final product to identify carbon functional groups and assign the chemical shifts of cyclopropyl protons.
All NMR experiments were performed at 25°C on a Bruker DRX 600-MHz spectrometer equipped with a 5-mm inverse triple resonance probe. One-dimensional proton spectra were collected with 64 scans of 64,000 points over 20 ppm and a total recycle delay of 6 s. The residual water in the spectrum was removed using presaturation of the HOD signal. Spectra were processed with an exponential line broadening of 1 Hz, and the proton chemical shifts were referenced to internal 3-trimethylsilyl propionate. Two-dimensional 1 H- 13

Synthesis of Mev, Mev⅐p, and Mev⅐pp
Analogues-The initial steps of isoprenoid biosynthesis are catalyzed by three GHMP family kinases (MK, PMK, and DPM-DC), which convert Mev to isopentenyl diphosphate, through the metabolic intermediates Mev⅐p and Mev⅐pp. Because our goal was to use Mev analogues as prodrugs to generate Mev⅐pp analogues in the cell, knowledge of the substrate selectivities of these enzymes is helpful in designing potential inhibitors. Toward this end, racemic mixtures of ten Mev analogues (1-10, Table 1) were synthesized chemically (9), each with a different substituent replacing the C 3 -methyl of mevalonate. Analogues 1, 3, 6, 7, and 9 (i.e. those with small planar substituents) were enzymat- ically phosphorylated to the mono-(X⅐p) and diphosphorylated (X⅐pp) forms and purified by anion-exchange chromatography (see "Experimental Procedures"). Because MK and PMK only act on the (R)-isomer (20), the mono-and diphosphorylated products are stereochemically pure. Pure (R)-Mev analogues were obtained by treating the (R)-Mev⅐pp analogues (synthesized enzymatically) with alkaline phosphatase and purifying as above. The monophosphorylated forms of five of the analogues (2, 4, 5, 8, and 10, indicated by a dot (•) in Table 1) could not be produced enzymatically (9), and were not pursued further. However, racemic mixtures of the diphosphorylated forms of these analogues were synthesized chemically starting from the corresponding lactones and used to probe the structural constraints of the allosteric pocket of MK and the active site of DPM-DC (for a full description see supplemental material). Thus, of the 30 possible analogues (i.e. the non-, mono-, and diphosphorlyated forms of each of the 10 analogue backbones), 25 were synthesized, purified, and tested as substrates and allosteric inhibitors of the Mev pathway enzymes.
Mevalonate Analogues as Substrates of GHMP Kinases-The ability of the Mev pathway GHMP kinases to accept alternative substrates was probed using the 25 Mev analogues listed in Table 1. Kinetic parameters for each enzyme-substrate pair were determined by analysis of complete reaction progress curves. Reactions were monitored using absorbance or fluorescence by stoichiometrically coupling (1:1) the production of ADP to the reduction of NADH, employing the well established pyruvate kinase/lactate dehydrogenase (PK/LDH) coupled assay system (12). The activity of PK regenerates the nucleotide, so the steady-state concentration of ATP is essentially fixed throughout the reaction. For MK and PMK, phosphorylated products were removed by the addition of downstream Mev pathway enzymes, obviating possible product inhibition. For DPM-DC, progress curves generated in the presence and absence of decarboxylated products had identical curvatures, indicating an absence of product inhibition under assay conditions (data not shown). Therefore, product versus time curves were fit directly to the integrated Michaelis-Menten equation (Equation 1 (21)) to extract k cat and K m ( Table 1).
Substrate selectivity varied widely across the three GHMP kinases. MK and PMK showed similar trends in catalytic efficiency, accepting small, planar substituents (6) and excluding larger, branched substituents (2, 4, 5, 8, and 10). However, PMK generally exhibited lower K m and higher k cat /K m values than MK for its substrates (see Table 1), indicating a relaxation of substrate selectivity in the second step of the pathway. DPM-DC was the least discriminating of the three enzymes, accepting all ten analogs as substrates. DPM-DC kinetic parameters for 1⅐pp, 6⅐pp, and 7⅐pp were similar to those for Mev⅐pp, indicating excellent tolerance of a variety of small substituents. For larger substituents, branching played an important role in determining substrate selectivity, with unbranched compounds favored over branched ones. Branched analogs (2⅐pp, 4⅐pp, and 10⅐pp) showed greater than 14-fold decreased affinity compared with the native substrate, whereas unbranched and cyclic analogues (3⅐pp and 9⅐pp, respectively) had K m values similar to that of Mev⅐pp. The binding pocket is apparently limited to three-carbon substituents, shown by the strongly reduced k cat /K m values for 4⅐pp and 5⅐pp. The substrate selectivity trend in DPM-DC is fundamentally different from that of MK and PMK, suggesting that substrate recognition is altered in this family member. This is perhaps not surprising, given that the Mev moiety is being phosphorylated on a different hydroxyl group in DPM-DC, requiring that the geometry of the binding pocket relative to ATP be substantially different from the other enzymes.
Structural Determinants of Substrate Selectivity-To ascertain whether active-site structural factors might explain the substrate selectivity of the Mev pathway enzymes, the active sites of all three enzymes were compared in the vicinity of the  (18), including positioning the C 3 -hydroxyl to attack the ␥-phosphate of ATP via interactions with conserved resides known to activate chemistry in GHMP kinases, the presence of an ordered sulfate interacting with conserved active-site residues (this sulfate was presumed to occupy the binding pocket for the terminal phosphate of Mev⅐pp) and an Arg that is well positioned in all DPM-DC structures to interact with the C 1 -carboxlyate of Mev⅐pp. In addition, we note that, in the three GHMP kinase structures that contain bound Mev, Mev⅐p, or Mev⅐pp, the C 1 -carboxylate interacts with a conserved serine. This same Ser is present in each of the seven DPM-DC structures, and positioning Mev⅐pp according to the Byres et al. rationale places the carboxylate within contact distance of that Ser. These conserved anchor points were used to position Mev⅐pp in our S. pneumoniae model (see Fig. 3C), which was generated from the crystal structure of S. pyogenes DPM-DC (PDB: 2GS8) using Swiss PDB Viewer (17). The sequence of the S. pyogenes DPM-DC is 70% identical to that of the S. pneumoniae enzyme.
A comparison of these models highlights differences in the active site near the C 3 -methyl group across these enzymes (Fig. 3). The narrow selectivity seen in MK seems likely due to the fact that the C 3 -methyl points into a shallow depression anchored by the side chain of His-20, a highly conserved MK residue (Fig. 3A) that appears to be fixed in its orientation by a hydrogen-bonded network that includes the main-chain carbonyls of Gly-256 and His-20, the amide of Val-23, and a crystallographic water (Fig. 3A). This configuration places the C 3 -methyl in van der Waals contact with residues lining the C 3 -binding pocket and leaves little room to admit larger R-groups.
Similar to the interactions seen in the MK⅐Mev structure, the C 3 -methyl group of Mev⅐p resides in a shallow, "tight" binding pocket in PMK. In this case, the C 3 -methyl is directed toward a tyrosine side chain (Tyr-16) that occupies a position analogous to that of His-20 in MK (Fig. 3B). However, instead of a hydrogen-bonded network, Tyr-16 is embedded in a hydrophobic pocket composed of Glu-15, Met-216, Val-217, Ile-220, Ile-224, Leu-265, Ile-269, and Ala-293. Although these structures offer little definitive evidence to explain the somewhat relaxed selectivity of PMK over MK, it is interesting to consider that a hydro-phobic pocket may have greater flexibility that one linked by hydrogen bonds (23,24).
In contrast to MK and PMK, our model of the DPM-DC⅐Mev⅐pp complex directs the C 3 -methyl group into a deep, narrow hydrophilic cavity that contains four crystallographic water molecules in the apo structure (Fig. 3C). Inspection of the seven DPM-DC homologues in the PDB revealed that the structure of the C 3 -cavity and the residues that define it are remarkably well conserved across prokaryotes and eukaryotes. The volume of the cavity is capable of accommodating two-carbon R-groups without steric strain, three-carbon R-groups will likely result in a "snug" fit, whereas larger substituents are unlikely to access the pocket well. This simple analysis of the C 3 -pocket accessibility predicts the behavior of the enzyme toward the C 3 analogues well (see Table 1) and supports the validity of the model. It is interesting to consider that this expanded cavity, which extends the substrate repertoire of the enzyme at C 3 , may provide the cell with the means to decarboxylate other, as yet unidentified, ␤-hydroxy carboxylic acids.
Mev⅐pp Analogues as Allosteric Inhibitors of MK-The ten diphosphorylated analogues (Table 1) were designed to probe the structure of the allosteric site of MK. A preliminary study was carried out in which all ten compounds were screened at 250 M for MK inhibition at fixed concentrations of substrates. Compounds that inhibited MK activity by Ͻ1% at this level (2⅐pp, 4⅐pp, 5⅐pp, and 10⅐pp) were not considered further. For compounds exhibiting detectable inhibition at 250 M, a titration was carried out to determine the IC 50 (Table 2 and "Experimental Procedures"). MK inhibitors fell into three broad groups based on their IC 50 values relative to that of Mev⅐pp: ϳ50-fold higher (6⅐pp and 9⅐pp), ϳ250-fold higher (1⅐pp, 7⅐pp,  and 8⅐pp), and ϳ900-fold higher (3⅐pp). Although we expect that this inhibition represents binding of a ligand at the allosteric site, it is formally possible that the analogue (X⅐pp), which resembles the acceptor (X) bound to the ␤and ␥-phosphates of ATP, could act as a bisubstrate inhibitor, in which case it is predicted to compete with the substrates. To distinguish between these possibilities, the inhibition mechanism of the two best inhibitors, 6⅐pp and 9⅐pp, was determined in a classic initial rate study in which the concentration of one substrate (Mev or ATP, four values spanning the K m values) was varied against the concentration of inhibitor (three values spanning the K i ) (see supplemental Fig. S1). A noncompetitive reversible inhibition model provided the best fit to each data set (Table 3). Pure noncompetitive inhibition by 6⅐pp and 9⅐pp against both substrates, as indicated by identical K is and K ii values (within error), is diagnostic of an allosteric inhibition mechanism (13,25). These results indicate that X⅐pp inhibition is due to binding at the allosteric site.
The X⅐pp analogues are relatively weak inhibitors, suggesting that the allosteric binding pocket does not tolerate substitution of the Mev⅐pp C 3 -methyl group by larger moieties. The shape of the pocket is difficult to infer, because the IC 50 values do not correlate with the size or flexibility of the C 3 substituents. For example, among analogues with two-carbon substituents, the . The Mev⅐pp C 3 -methyl points into a large water-filled cavity composed of nine conserved residues. Images were generated by using PyMOL (19).  vinyl analogue (6⅐pp) inhibits MK 4-to 5-fold better than either the ethyl (1⅐pp) or ethynyl (7⅐pp) analogues, suggesting that some subtle combination of size, geometry, and electrophilicity makes the vinyl group preferred. Analogues with three-carbon substituents appear to inhibit differently on the basis of shape: linear 8⅐pp outperforms the more flexible 3⅐pp. Surprisingly, the vinyl and cyclopropyl (9⅐pp) analogues inhibit equally well, despite the addition of a carbon in 9⅐pp and the exclusion of all other branched substituents. This result suggests that the region of the allosteric site that binds the C 3 substituent is highly sensitive to the geometry of the group and that favorable electronic interactions between the enzyme and the -orbitals in the vinyl group and the cyclopropyl ring may contribute to binding affinity.
This study revealed that our prodrug strategy for MK inhibition will be limited to Mev analogues with small C 3 substituents. Fluorinated Mev⅐pp analogues, which are isosteric with Mev⅐pp, are inhibitors of mammalian DPM-DC and should be excellent inhibitors of S. pneumoniae MK (26,27).
Mev⅐pp Analogues as Probes of DPM-DC Reaction Mechanism-Five analogues (6⅐pp through 10⅐pp) were designed to covalently inactivate DPM-DC by forming a highly reactive carbocation capable of covalent attachment at the active site (see the introduction and Fig. 2). These compounds did not detectably inactivate the enzyme under the initial rate conditions used to measure their kinetic constants as substrates of DPM-DC (see "Mevalonate Analogues as Substrates of GHMP Kinases"). To assess whether inactivation was too slow to detect in the initial rate experiments, the DPM-DC reaction conditions were adjusted (see "Experimental Procedures") so that the concentration of the analogue would remain saturating during Ͼ1000 turnovers of the enzyme. With each analogue, the rates of reaction before and after 1000 turnovers were identical, within experimental error, indicating that Յ4% of the enzyme was inactivated during the measurement. Thus, either the analogues are decarboxylated (like Mev⅐pp) or the carbocation reacts with water to form non-decarboxylated products.
To determine the fate of the analogues, we focused on the cyclopropyl derivative (9⅐pp), which we expected to inactivate DPM-DC via opening of the strained three-membered ring (28). When a tertiary carbocation forms adjacent to a cyclopropyl ring, it is expected to rearrange into a homoallyl cation (see Fig. 2) that is stabilized due to release of 27 kcal/mol ring strain (29,30). If the homoallyl cation were to react with water (rather than the protein) at the active site, the reaction would produce a ring-opened primary alcohol that can be readily identified by NMR spectrometry due to the large chemical shift differences between cyclopropyl protons and their linear counterparts (31). An NMR experiment that monitored the cyclopropyl proton resonances during product formation revealed that the substrate peaks are shifted only slightly downfield (ϳ0.3 ppm) in the product. The sum of the integrated intensities of the substrate and product resonances remained fixed throughout the experiment (Fig. 4). An HSQC experiment identified the product resonances (0.5 and 0.68 ppm, Fig. 4A) as those of cyclopropyl ring protons (Fig. 4B). Furthermore, the cyclopropyl protons in the product showed long range correlations to carbons at 107.6 ppm and 148.3 ppm (two-dimensional 1 H-13 C HMBC (32); data not shown), indicating that a double bond is adjacent to the cyclopropyl ring, which confirmed that the product results from decarboxylation. These NMR data rule out a ring-opened product and support the analogue undergoing a decarboxylation that resembles that of the native substrate, Mev⅐pp.
These results call into question the extent of carbocation formation in the DPM-DC transition state. If a full carbocation had formed, we expected to observe rearrangement of the cyclopropyl group. If no nucleophile is in close proximity to the carbocation (the protein surface or water), we might expect to see only intact cyclopropyl products; however, the presence of four crystallographic water molecules in the large pocket adjacent to C 3 and the proximity of the Asp-276, Lys-18, Ser-185, and Met-189 side chains and the Tyr-19 carbonyl as potential nucleophiles argue against this possibility (Fig. 3C). Alternatively, carbocation formation may be minimal or absent, in which case elimination of the carboxylate and the phosphate is concerted rather than dissociative. Abeles' work on the effects of altered electron induction at C 3 with the mammalian enzyme shows clearly that DPM-DC chemistry is sensitive to such changes and supports the development of a positive charge at C 3 in the transition state (10). Further, by replacing C 3 with a positively charged amine, he created an analogue that mimicked the structure and charge characteristics of a dissociative transition state. The affinity of this analogue (0.75 M) was only 20-fold higher than that of the substrate (33), which, while supportive of a dissociative character in the transition state, is perhaps more consistent with development of partial rather than complete positive charge at C 3 . Although studies that correlate the extent of positive charge formation with degree of ring opening in cyclopropyl ring systems do not yet exist, our results, which demonstrate no detectible ring opening, are consistent with only slight positive charge formation in the transition state. Using kinetic isotope effects, it may be possible to assess the extent of positive charge development on C 3 at the transition state.
Targeting the Human DPM-DC-Exclusive of their effects on bacterial enzymes, the Mev⅐pp analogues might also inhibit the human DPM-DC, because it is not clear whether the corresponding human and bacterial enzymes have diverged to the point where they could be targeted orthogonally, as is the case for MK. Inhibitors of the human Mev pathway (statins and bisphosphonates) are used clinically to reduce cholesterol biosynthesis, increase bone density, and decrease cell proliferation in cancer (34). Down-regulation of the human Mev pathway by statins has recently been linked to disruption of replication by hepatitis C virus (35,36) and human immunodeficiency virus (37), enhancement of anticancer drugs (38), and anti-inflammatory effects in the lung and airways (39,40). These findings suggest that DPM-DC inhibitors may find additional uses in the treatment of non-infectious diseases, which obviates the need for selective DPM-DC inhibition.
Conclusions-Twenty-five novel Mev analogues have been tested as substrates and inhibitors of three enzymes that comprise the Mev pathway in S. pneumoniae. Although the MK allosteric binding pocket admits certain analogues, it is highly selective for Mev⅐pp. Substrate selectivity of the enzymes varies considerably across the pathway, with MK providing the most stringent selection. The active-site structures of the Mev pathway GHMP kinases provide a rationale for the substrate selectivity of these enzymes toward substitution at C 3 . The C 3 R-group pocket in DPM-DC is considerably larger than that of MK or PMK, and the volume of this cavity correlates well with selectivity toward R-group substitution. The high conservation of this pocket indicates that it has been evolutionarily maintained, and that the ability to decarboxylate substrates that vary at the C 3 R-group may provide an important metabolic function. Finally, results using analogues designed to act as mechanism-based inhibitors of DPM-DC suggest that ionization of phosphate and decarboxylation of the p⅐Mev⅐pp intermediate occurs in a concerted fashion with little carbocation development at C 3 .