Feedback Inhibition of Deoxy-d-xylulose-5-phosphate Synthase Regulates the Methylerythritol 4-Phosphate Pathway*

Background: The methylerythritol phosphate (MEP) pathway is required for the biosynthesis of plastid-derived isoprenoids from plants. Results: Deoxyxylulose-5-phosphate synthase (DXS) was cloned from Populus trichocarpa, and metabolic regulation was tested. Conclusion: Both isopentenyl diphosphate and dimethylallyl diphosphate inhibit DXS by competing with thiamine pyrophosphate. Significance: Prediction of isoprene emission from trees and bioengineering of MEP pathway will be aided by these results. The 2-C-methyl-d-erythritol 4-phosphate (MEP) pathway leads to the biosynthesis of isopentenyl diphosphate (IDP) and dimethylallyl diphosphate (DMADP), the precursors for isoprene and higher isoprenoids. Isoprene has significant effects on atmospheric chemistry, whereas other isoprenoids have diverse roles ranging from various biological processes to applications in commercial uses. Understanding the metabolic regulation of the MEP pathway is important considering the numerous applications of this pathway. The 1-deoxy-d-xylulose-5-phosphate synthase (DXS) enzyme was cloned from Populus trichocarpa, and the recombinant protein (PtDXS) was purified from Escherichia coli. The steady-state kinetic parameters were measured by a coupled enzyme assay. An LC-MS/MS-based assay involving the direct quantification of the end product of the enzymatic reaction, 1-deoxy-d-xylulose 5-phosphate (DXP), was developed. The effect of different metabolites of the MEP pathway on PtDXS activity was tested. PtDXS was inhibited by IDP and DMADP. Both of these metabolites compete with thiamine pyrophosphate for binding with the enzyme. An atomic structural model of PtDXS in complex with thiamine pyrophosphate and Mg2+ was built by homology modeling and refined by molecular dynamics simulations. The refined structure was used to model the binding of IDP and DMADP and indicated that IDP and DMADP might bind with the enzyme in a manner very similar to the binding of thiamine pyrophosphate. The feedback inhibition of PtDXS by IDP and DMADP constitutes an important mechanism of metabolic regulation of the MEP pathway and indicates that thiamine pyrophosphate-dependent enzymes may often be affected by IDP and DMADP.

Isoprenoids, also known as terpenoids, constitute one of the largest groups of metabolites and are widely dispersed in all living organisms including both prokaryotes and eukaryotes (1). They represent the largest group of secondary metabolites, having more than 35,000 members including sterols, carotenoids, dolichols, ubiquinones, plastoquinones, prenylated proteins, cytokinins, gibberellic acid, and abscisic acid (1,2). They are involved in various biological processes including photosynthesis, respiration, and regulation of growth and development (3). Isoprenoids also play significant roles in attracting pollinators and seed dispersers, defense against different biotic and abiotic stresses, intracellular signal transduction, vesicular transport within the cell, and construction of cellular and organelle membranes (1)(2)(3)(4)(5). In addition, some isoprenoids have commercial applications as flavors and fragrances, pigments, polymers, and drugs (6).
Despite the great diversity of structure and function, only two metabolic precursors, isopentenyl diphosphate (IDP) 3 and its isomer dimethylallyl diphosphate (DMADP) are used as the building blocks for all the isoprenoids (7,8). The well known acetate/mevalonate pathway was thought to be the only biochemical pathway for the synthesis of IDP (8,9). However, incorporation of 13 C-labeled precursors into polyterpenoids revealed that an alternative, mevalonate-independent pathway known as the 2-C-methyl-D-erythritol 4-phosphate (MEP) pathway also leads to the biosynthesis of IDP in bacteria (8,10,11). Further studies showed that the MEP pathway is also present in green algae (12) and higher plants (13)(14)(15)(16)(17)(18). Both the acetate/mevalonate and MEP pathways are present in plants; however, they are localized in the cytosol and the chloroplast, respectively (4,14).
The MEP pathway starts with the synthesis of 1-deoxy-Dxylulose 5-phosphate (DXP) from glyceraldehyde 3-phosphate (GAP) and pyruvate. The reaction is catalyzed by the enzyme called 1-deoxy-D-xylulose-5-phosphate synthase (DXS), which requires a divalent cation, Mg 2ϩ or Mn 2ϩ , and uses thiamine pyrophosphate (TPP) as a cofactor (19 -24). In the next step, DXP is converted to the branched compound MEP, the first committed intermediate of this pathway, by 1-deoxy-D-xylulose-5-phosphate reductoisomerase (DXR) using NADPH as a cofactor and a divalent cation, Mg 2ϩ or Mn 2ϩ (25). MEP is ultimately converted to a mixture of IDP and DMADP, precursors of all higher isoprenoids (26 -29). The isomerization of DMADP and IDP is catalyzed by the enzyme isopentenyldiphosphate isomerase (28).
The absence of the MEP pathway in humans and its presence in different eubacteria, various apicomplexan parasites, and photosynthetic eukaryotes make it an attractive target for drug discovery and herbicides (7,20,30). Some isoprenoids also serve as important targets for biotechnological applications because of their nutritional and medicinal benefits (7,31,32). Isoprene, the most abundantly produced isoprenoid in plants, significantly affects atmospheric chemistry (33). In this regard, a mechanistic model predicting isoprene emission from plants will be of great importance in atmospheric chemistry. A mechanistic model requires a deep understanding of the regulation of the MEP pathway.
Gene expression studies have demonstrated a potential regulatory role for DXS in the synthesis of DMADP/IDP (7), but metabolic regulation of MEP pathway is not yet understood completely. Wolfertz et al. (34,35) used deuterium-labeled deoxyxylulose 5-phosphate to show that the carbon flux through the MEP pathway is under strong metabolic regulation. They suggested that a feedback inhibition of DXS enzyme by the metabolites of this pathway downstream of DXP, especially DMADP, plays a critical role in this regulation. However, no direct evidence has been shown so far that can explain the tight control of the carbon flux through the MEP pathway.
Here we describe the cloning and characterization of the recombinant DXS (PtDXS) protein from Populus trichocarpa. We also report a rapid and convenient high performance liquid chromatography-tandem mass spectrometry (HPLC-MS/MS)based assay for PtDXS. This assay was used to study the activity of PtDXS in the presence of different metabolites of the MEP pathway. Our results show that IDP and DMADP, the last metabolites of the MEP pathway, significantly inhibit PtDXS by competing with TPP. Computational analysis shows that both of these metabolites bind with the enzyme in a manner similar to that of TPP. The inhibition of PtDXS by IDP and DMADP constitutes an important regulatory mechanism of the MEP pathway where the very last metabolite of the pathway regulates the activity of the very first enzyme of the pathway.

EXPERIMENTAL PROCEDURES
Cloning-The cDNA encoding the mature PtDXS was amplified by PCR using the primers 5Ј-G GAA TTC CAT ATG GCA TCA CTA TCA GAA AGA GGA GAG-3Ј and 5Ј-CG GGA TCC TTA TGA TGA CAT AAT CTC CAG AGC-3Ј. The PCR product was digested with the restriction enzyme BamHI to completion followed by a partial digestion with the restriction NdeI as the coding DNA of PtDXS contained an NdeI site. The partially digested PCR product was ligated with a laboratorymade vector, pET17b-HR, digested with the same two restriction enzymes. The ligation mixture was transformed into E. coli strain DH5␣. The clones that contained the complete cDNA were selected by agarose gel electrophoresis of the isolated DNA constructs. The resultant overexpression plasmid construct was designated as pET17bHR/PtDXS for the production of PtDXS with a tobacco etch virus protease-cleavable His 6 tag at the N terminus. An overexpression plasmid construct for the production of PtDXS with a tobacco etch virus protease-cleavable His 10 tag at the N terminus was engineered by PCR-based site-directed mutagenesis using pET17bHR/PtDXS as the template. The mutagenesis primers were 5Ј-G CAT CAC CAT CAC CAT CAT CAC CAT CAC CAT AGC GGT ACC GAG AAC CTG TAC TTC-3Ј and 5Ј-GAA GTA CAG GTT CTC GGT ACC GCT ATG GTG ATG GTG ATG ATG GTG ATG GTG ATG C-3Ј. The resulting construct was designated as pET17b10HR/PtDXS. An overexpression plasmid construct for the production of PtDXS with a tobacco etch virus protease-cleavable His 6 tag at both the N and the C termini was engineered in two steps. First, the overexpression construct pET17bHR/PtDXS was digested with BamHI and EcoRI and ligated with a synthetic duplex DNA consisting of the oligos 5Ј-GA TCC GAG AAC CTG TAC TTC CAG GGT CAC CAC CAC CAC CAC CAC TAA-3Ј and 5Ј-AA TT TTA GTG GTG GTG GTG GTG GTG ACC CTG GAA GTA CAG GTT CTC G-3Ј. The clones with the inserted DNA were selected by EcoRI digestion as the insertion destroyed the EcoRI site. Then the stop codon at the end of the PtDXS-coding sequence was converted to Ser by PCR-based site-directed mutagenesis using the primers 5Ј-CTG GAG ATT  ATG TCA TCA TCA GGA TCC GAG AAC CTG TAC-3Ј and  5Ј-GTA CAG GTT CTC GGA TCC TGA TGA TGA CAT AAT CTC CAG-3Ј. The resultant DNA construct was designated as pET17bHR3ЈHR/PtDXS. The N-terminal His tag and tobacco etch virus protease site were removed by PCR-based site-directed mutagenesis using pET17bHR3ЈHR/PtDXS as the template. The two primers for the mutagenesis were 5Ј-CT  TTA AGA AGG AGA TAT ACC ATG GCA TCA CTA TCA  GAA AGA GGA GAG-3Ј and 5Ј-CTC TCC TCT TTC TGA  TAG TGA TGC CAT GGT ATA TCT CCT TCT TAA AG-3Ј. The resultant DNA construct was designated as pET17b3ЈHR/ PtDXS. For all the plasmid constructs, the presence of the correct PtDXS-coding sequence and the absence of any undesired mutation were confirmed by DNA sequencing.
Overexpression and Purification-The E. coli strain BL21(DE3)pLysS was used to overexpress the various forms of PtDXS. A liter of LB medium containing 100 g/ml ampicillin and 20 g/ml chloramphenicol was inoculated with colonies of fresh transformants of an overexpression construct and incubated at 37°C with vigorous shaking (225 rpm) until A 600 reached 1. The culture was then cooled with ice to room temperature, induced with 0.5 mM isopropyl 1-thio-␤-D-galactopyranoside (final concentration), and further incubated with vigorous shaking at room temperature for ϳ18 h. The E. coli cells were harvested by centrifugation and resuspended in 10 ml of cold buffer A (50 mM sodium phosphate, 300 mM NaCl, 10% glycerol, pH 8.0)/g of bacterial paste. MgCl 2 was added to a final concentration of 20 mM followed by the addition of DNase I and EDTA-free inhibitors. The cells were lysed with a French press. The cell debris was removed by centrifugation at ϳ27,000 ϫ g for 20 min. Ammonium sulfate was added to the supernatant in small quantities to 45% saturation under gentle stirring. After 30 min, the suspension was centrifuged at ϳ27,000 ϫ g for 20 min. The pellet was redissolved in the same volume of cold buffer A. The solution was centrifuged, and the ammonium sulfate precipitation was repeated with the supernatant. The pellet of the second ammonium sulfate precipitation was redissolved in cold buffer A, and the solution was centrifuged. The supernatant was dialyzed against 2 liters of the same buffer for 3 h. The dialyzed protein solution was mixed with Ni-NTA resin with gentle shaking for 1 h and loaded onto a column. The column was washed with 10 mM imidazole in buffer A until A 280 of the effluent was Ͻ0.05 and eluted with a 10 -250 mM linear imidazole gradient in buffer A. Fractions containing PtDXS were identified by SDS-PAGE and pooled. PtDXS was precipitated with ammonium sulfate (45% saturation), and the pellet was redissolved in a minimal volume of a cold buffer (50 mM Tris-HCl, 10% glycerol, 1 mM DTT, pH 7.5). The protein solution was dialyzed against 1 liter of the same buffer and centrifuged at ϳ27,000 ϫ g for 20 min. The supernatant was dispensed into microtubes, frozen in liquid nitrogen, and stored at Ϫ80°C. All protein purification procedures were carried out at 4°C unless specified otherwise.
Coupled Enzyme Assay-The steady-state kinetic constants of PtDXS were measured using a DXR-coupled assay. The assay components in a buffer containing 100 mM HEPES, pH 7.5 included pyruvate, GAP, 5 mM MgCl 2 , 1 mM TPP, 50 mM NADPH, 4 M recombinant Acinetobacter baumannii DXR (AbDXR; laboratory-made), and 0.5 M PtDXS. The reaction was initiated by the addition of GAP at room temperature. For measuring the K m of pyruvate, GAP was fixed at 0.2 or 0.5 mM, and pyruvate varied in the range of 0.05-1.2 mM. For measuring the K m of GAP, pyruvate was fixed at 2 or 5 mM, and GAP varied in the range of 10 -175 M. The kinetic constants were evaluated by non-linear least square fitting of the data to the Michaelis-Menten equation using the program Origin.
Enzymatic Synthesis of DXP and [ 13 C 2 ]DXP-DXP/[ 13 C 2 ]DXP were synthesized enzymatically from pyruvate/2,3-[ 13 C 2 ]pyruvate, and GAP (produced in situ) using E. coli DXS (EcDXS) as described (36) with some modifications. The reaction mixture was prepared by dissolving D-fructose 1,6-bisphosphate (406 mg; 25 mM) and pyruvate (or 2,3-[ 13 C 2 ]pyruvate) (ϳ220 mg; 50 mM) in ϳ39 ml of 50 mM Tris-HCl, pH 7.5 with 1 mM DTT, 5 mM MgCl 2 , and 0.5 mM TPP. The reaction mixture also contained recombinant Staphylococcus aureus fructose-bisphosphate aldolase (2.4 M; laboratory-made), yeast triose-phosphate isomerase (0.04 M; laboratory-made), and EcDXS (1.5 M; laboratory-made). The reaction was then carried out at 37°C for ϳ24 h. The enzymes were removed from the reaction mixture by ultrafiltration through YM10 (Millipore) membrane. The filtrate was then loaded on Dowex 1X8 column (40 ml; chloride form), which was equilibrated with water. The column was washed with 100 ml of water after collecting the flowthrough. DXP was then eluted from the column with 100 ml of 1% NaCl solution. The fractions containing DXP were lyophi-lized to obtain solid DXP. The solid was then dissolved in ϳ3 ml of water. The solution was desalted with a Sephadex G10 column by eluting with water. The concentration of the pure DXP solution was obtained from NMR.

Preparation of PtDXS Assay Mixture for LC-MS/MS-based Assay-
The activity of the purified PtDXS enzyme was studied using an LC-MS/MS-based assay. A mixture of dihydroxyacetone phosphate (DHAP) and triose-phosphate isomerase from rabbit muscle was used to maintain a constant supply of GAP in the reaction mixture. The ratio of equilibrium concentration of DHAP and GAP at the temperature of the reaction mixture (37°C) was calculated to be 18:1. The assay mixture contained 40 mM Tris-HCl buffer at pH 8.0, 10 mM MgCl 2 , 1 mM dithiothreitol (DTT), 100 M TPP, 1 unit/ml rabbit muscle triosephosphate isomerase, and 0.25 M PtDXS in a total volume of 100 l. The reaction was initiated with a mixture of DHAP and pyruvate. The concentrations used to study the K m of DHAP and pyruvate were 0 -197 M DHAP in the presence of 5 mM pyruvate and 0 -1 mM pyruvate in the presence of 260 M DHAP, respectively. The concentrations used to study the K m of TPP were 0 -1 mM of TPP in the presence of 260 M DHAP and 500 M pyruvate. The reaction was incubated at 37°C for 5 min and then quenched by freezing in liquid nitrogen followed by the addition of 400 l of ice-cold acetonitrile, keeping the frozen reaction mixture on dry ice. Then the reaction mixture was thawed on ice followed by the addition of 2 M [ 13 C 2 ]DXP as an internal standard for the mass spectrometry. The assay mixture was then centrifuged at 28,000 ϫ g for 10 min, and the supernatant was stored at Ϫ80°C until further analysis.
Inhibition Studies-Different metabolites of the MEP pathway, namely MEP, 4-(cytidine 5Ј-diphospho)-2-C-methyl-Derythritol, 2-C-methyl-D-erythritol 2,4-cyclodiphosphate, 4-hydroxy-3-methylbut-2-enyl diphosphate, IDP, and DMADP were tested to study their effect on PtDXS activity. All the metabolites except IDP and DMADP were purchased from Echelon Biosciences Inc. (Salt Lake City, UT). IDP and DMADP were purchased from Isoprenoids, LC (Tampa, FL). The assay mixture was prepared as mentioned before with 400 M metabolites. The reaction was initiated with a mixture of DHAP (210 M) and pyruvate (200 M). Similarly, the assay mixtures for obtaining the Michaelis-Menten plot at different inhibitor concentrations were prepared as mentioned before at different TPP concentrations in the presence of 0, 100, and 1000 M IDP, keeping the concentration of DHAP and pyruvate fixed at 210 and 200 M, respectively. The reaction was initiated with a mixture of DHAP and pyruvate. To calculate the K i of DMADP and IDP, the assay was done in the presence of 25 M TPP (ϳK m for TPP) and 0 -3 mM metabolites. The reaction was initiated with a mixture of DHAP (210 M) and pyruvate (200 M). The assay was then carried out as described before.
LC-MS/MS of the PtDXS Assay Mixture-The assay mixture was analyzed by LC-MS/MS to separate the product (DXP) from the substrate (DHAP). Liquid chromatography was performed using a Merck SeQuant TM ZIC-pHILIC (50 ϫ 2.1 mm; 5 m; 200 Å; polymeric beads; poly(etherether ketone)) column (The Nest Group, Inc., Southborough, MA) fitted to two LC-20AD HPLC pumps and an SIL-HTc autosampler (Shimadzu, Kyoto, Japan) as described (37). The rest of the instru-mental setup for the mass spectrometer coupled to this chromatography system was done as described (37). The assay mixture was filtered through a Whatman syringeless filter device (Mini-Uniprep TM polytetrafluoroethylene filter medium), and 5 l of the sample was injected into the column. The analyte was eluted with a binary gradient consisting of 50 mM ammonium acetate at pH 10 and acetonitrile (composition as shown in Fig. 1, inset) at a flow rate of 0.15 ml/min. Mass spectrometry was performed as described (37). Multiple reaction monitoring mode was used to acquire the precursor/product ion pairs for DXP, [ 13 C 2 ]DXP, and DHAP. The mass pairs used for scanning these compounds were (in atomic mass units) 213/ 97, 215/97, and 169/97 for DXP, [ 13 C 2 ]DXP, and DHAP, respectively. The optimized declustering potential used for acquiring the mass spectra was Ϫ30 V for DXP and [ 13 C 2 ]DXP and Ϫ20 V for DHAP. The optimized collision energy used for acquiring the mass spectra was Ϫ20 V for DXP and [ 13 C 2 ]DXP and Ϫ15 V for DHAP.
Quantification and Data Analysis-To obtain better quantification of the analytes, an internal standard was used in all the samples to allow a correction for the ionization efficiency. Standard DXP samples containing 2 M [ 13 C 2 ]DXP as an internal standard were run to obtain a calibration curve. The calibration curve was used to quantify the amount of DXP produced in the assay samples. After correction for the dilution factor, the amount of DXP produced in the reaction mixture was used to calculate the specific activity of the PtDXS enzyme. The kinetic constants of PtDXS enzyme for different substrates and the IC 50 curves for IDP and DMADP were obtained by fitting the experimental data with non-linear regression using the program Origin. The kinetic constants were evaluated using the Michaelis-Menten equation. Calculation of the IC 50 values was done using a logistic equation (38), where v is the percent activity, v min is the minimum percent activity, v max is the maximum percent activity, [I] is the concentration of the inhibitor, and H is the Hill coefficient. Calculation of the K i of the inhibitors was done using the Cheng-Prusoff equation (39) as follows.
The IC 50 (40) as the template. The crystal structure of DrDXS, not that of EcDXS (Protein Data Bank code 2O1S), was chosen as the template for the homology modeling because DrDXS has only one segment (residues 199 -243) with no electron density, whereas EcDXS has two segments (residues 183-238 and 292-317) with no electron density (40). Each monomer of the modeled homodimeric enzyme contains one Mg 2ϩ ion and one coenzyme TPP. The Mg 2ϩ ion is coordinated with four oxygen atoms, two from pyrophosphate of TPP and one each from Asp-145 and Asn-174. Two water molecules were placed near each Mg 2ϩ ion based on the crystal structure of EcDXS so that the Mg 2ϩ ion is coordinated with six oxygen atoms. The structural model was then refined by molecular dynamics using the AMBER program package (version 10) (41). The modeled homodimeric protein was solvated with ϳ32,900 transferable intermolecular potential with three point charges (TIP3P) water molecules in a rectangular box with the edges at least 12 Å from the protein. The system was neutralized using 12 Na ϩ ions. Glu-377 was considered protonated as a hydrogen donor to form a hydrogen bond with N1 of the pyrimidine ring of TPP as in the crystal structures of EcDXS and DrDXS (40). Water molecules in the system were minimized first using a combination of steepest descent (15,000 steps) and conjugated gradient (5,000 steps) methods with protein and ligand heavy atoms restrained with a force constant of 100 kcal mol Ϫ1 Å Ϫ2 . The whole system was then minimized using a combination of steepest descent (5,000 steps) and conjugated gradient (5,000) methods without any positional restraint for any atoms except those interacting with the Mg 2ϩ ions and forming the hydrogen bond between TPP and Glu-377. The restraints for the Mg 2ϩ coordination and the hydrogen bond between TPP and Glu-377 were also enforced during the subsequent heating and equilibration steps. The minimized system was heated from 0 to 300 K in 500,000 steps in 1 ns at constant volume and equilibrated at 300 K and constant pressure for 4 ns. The system was further simulated without any distance restraint for 1.5 ns. The minimization, heating, and equilibration simulations were carried out using the Sander module in AMBER 10 (41) with the ff99SB force field.
The force field parameters for the coenzyme TPP were derived using the AMBER antechamber program (41). The PMEMD module in AMBER 10 (41) was used for the subsequent production simulation. The particle mesh Ewald method (42) was used to evaluate long range electrostatic interactions. The nonbonded cutoff was 10 Å. All bonds to hydrogen atoms were constrained in the simulations with the SHAKE algorithm (43), permitting a time step of 2 fs. Temperature was controlled with Langevin dynamics. The data of the 1.5-ns simulation were analyzed using the PTRAJ module in the program AmberTools (41). The refined structural model of PtDXS in complex with TPP and Mg 2ϩ was used to dock IDP and DMADP molecules. TPP was mutated to IDP and DMADP using the xleap program in AmberTools (41). The force field parameters for IDP and DMADP were derived using the same procedure as for the coenzyme TPP. The models of the inhibitor complexes were neutralized with Na ϩ ions, solvated with explicit water molecules, and minimized using essentially the same procedure as for the TPP complex. All structural illustrations were drawn with the program PyMOL (Schrödinger, LLC).

RESULTS
Cloning, Overexpression, and Purification-To produce PtDXS in E. coli, the cDNA encoding the mature PtDXS was Feedback Inhibition of Deoxyxylulose Synthase JUNE 7, 2013 • VOLUME 288 • NUMBER 23 cloned into a homemade overexpression vector derived from the commercial vector pET17b from Novagen under the control of an isopropyl 1-thio-␤-D-galactopyranoside-inducible T7 promoter. A large quantity of PtDXS can be produced using this overexpression construct in the E. coli strain BL21(DE3)pLysS at 37°C (not shown), but more soluble PtDXS protein could be obtained at room temperature (data not shown). The recombinant protein contained a six-histidine tag at its N terminus to facilitate its purification using Ni-NTA resin, but the His tag did not help the purification as the majority of the protein molecules did not bind to Ni-NTA resin. To address this issue, the His tag was extended to 10 histidine residues, but the longer His tag did not improve the binding. Then a six-histidine tag was engineered at the C terminus of the protein, and the N-terminal His tag was removed as it did not help the purification. The C-terminal His tag helped the purification of the protein. However, the purified protein could be easily degraded. Ammonium sulfate precipitation effectively removed protease contamination. Two steps of ammonium sulfate precipitation before the Ni-NTA chromatography and one step of ammonium sulfate precipitation after the Ni-NTA chromatography yielded a stable protein preparation. The average yield for the purified protein was ϳ14 mg/liter of E. coli culture.
Steady-state Kinetic Analysis-Steady-state kinetic parameters of the recombinant PtDXS were measured by a DXR-coupled enzyme assay. AbDXR was used as a coupling enzyme in the assay as this enzyme is stable. The assay was validated by varying the concentrations of PtDXS and AbDXR. AbDXR at 4 M was deemed sufficient as the reaction rate doubled as the PtDXS concentration doubled, and increasing the AbDXR concentration further did not increase the reaction. The kinetic data are summarized in Table 1. The K m values for pyruvate and GAP (87.8 and 18.5 M, respectively) were both higher than those of Mycobacterium tuberculosis DXS (40 and 6.1 M, respectively) (44) but significantly smaller than those of Rhodobacter capsulatus DXS (440 and 68 M, respectively) (45).

Development of LC-MS/MS-based Assay for PtDXS
Enzyme-The effect of different MEP pathway metabolites on the activity of PtDXS enzyme was studied by measuring the amount of DXP produced by the in vitro reaction of the enzyme in the presence of those metabolites. The chromatogram of the assay mixture is shown in Fig. 1. Manual addition of substrates to the assay mixture followed by its quenching in liquid nitrogen involves a time lag. This leads to the production of small amount of DXP at 0 min. Study of the time course of the PtDXS enzymatic reaction showed that it was linear for the initial 10 min (data not shown). Samples were collected at 5 min to calculate the specific activity of the enzyme. Table 1 shows the kinetic constants of PtDXS enzyme for different substrates measured by this method.
Effect of pH on PtDXS Activity-The activity of the PtDXS enzyme was monitored at different pH values using Bistris propane buffer. The useful pH range for this buffer is 6.3-9.5. The activity of the enzyme was found to be highly sensitive to the pH of the assay mixture (Fig. 2). The enzyme did not show substantial activity below pH 6.5 and above pH 8.5. The highest activity of this enzyme was obtained at pH 8.0. This is typical for a chloroplastic enzyme.
Effect of Different Metabolites of the MEP Pathway Using LC-MS/MS Method-DMADP was selected for testing based on previous suggestions of its role in metabolic regulation of the MEP pathway (34,35). The rest of the intermediates of the MEP pathway were screened for potential effects on PtDXS activity    (Fig. 3A). The intermediate 4-(cytidine 5Ј-diphospho)-2-Cmethyl-D-erythritol 2-phosphate could not be tested due to its instability. Fig. 3B shows the effect of MEP on PtDXS activity. Because the molecular weight of MEP is the same as the internal standard ([ 13 C 2 ]DXP) used for the LC-MS/MS-based assay, we could not normalize the data for MEP with respect to the internal standard. We compared the effect of MEP relative to the control instead of using absolute values because the ion suppression effect could not be eliminated in this case. Analysis of variance followed by Bonferroni posttest indicated that IDP alone showed statistically significant inhibition (p ϭ 0.0036).
Mechanism of Inhibition-The next goal was to study the mode of inhibition of IDP on PtDXS enzyme. The effect of each of the substrates on the inhibition of IDP on PtDXS enzyme was tested. In each case, the concentration of a particular substrate was doubled as compared with the control, keeping the concentration of the inhibitor (400 M) the same as that in the control. It was observed that the extent of inhibition was reduced in the presence of a higher concentration of TPP (Fig. 4). The use of a higher concentration of pyruvate and DHAP did not affect the inhibition significantly. A two-way analysis of variance followed by Bonferroni posttest indicated that doubling the concentration of each metabolite had no effect on the rate in the absence of inhibitor, and the inhibition by IDP is significantly reduced (p Ͻ 0.01) only in the presence of twice the amount of TPP. This suggests that IDP acts as a competitive inhibitor of TPP. The inhibitory effect of IDP on PtDXS enzyme was tested by varying the concentration of IDP in the presence of different concentrations of TPP and fixed concentrations of pyruvate and DHAP. The Michaelis-Menten plot of PtDXS activity at different TPP concentrations showed that the activity was decreased in the presence of a higher concentration of IDP (Fig.  5). This also indicates that IDP acts as a competitive inhibitor of TPP. Regression of the experimental data points was done using the method of least squares. The Michaelis-Menten plot for the activity of the enzyme at different substrate concentrations with varying inhibitor concentrations could not be fitted well by assuming the equation for standard competitive inhibition kinetics. A better curve fitting was obtained by incorporating a Hill coefficient in the rate equation (Fig. 5). The H value obtained from the least square regression was 0.68.    JUNE 7, 2013 • VOLUME 288 • NUMBER 23

JOURNAL OF BIOLOGICAL CHEMISTRY 16931
The activity of PtDXS enzyme was studied over a broad range of concentrations of IDP and DMADP in the presence of ϳK m concentration of TPP (Fig. 6). The K i values of IDP and DMADP for PtDXS enzyme were found to be 65.4 Ϯ 4.3 and 81.3 Ϯ 10.5 M, respectively. The IC 50 curves of IDP and DMADP were not easily modeled by standard competitive inhibition kinetics. We applied the logistic equation (as mentioned under "Experimental Procedures") to the experimental data points using two different approaches: fitting using a fixed value of H ϭ 1 or a fixed value of v min ϭ 0. A better fitting of the IC 50 curve was obtained using the second approach. This approach resulted in H values less than 1. The Hill coefficients of the inhibitor binding as obtained from the non-linear curve fitting using Origin were H ϭ 0.69 Ϯ 0.03 for IDP and H ϭ 0.61 Ϯ 0.06 for DMADP. These H values are consistent with both the inhibitors exhibiting negative cooperativity of binding for the enzyme; in other words, binding of one inhibitor to one member of a dimer reduces the binding of a second inhibitor mole-cule to the other member of the dimer. One part of the IDP and DMADP molecules that they have in common with TPP is the diphosphate group. Therefore, the effect of pyrophosphate was tested (Fig. 6, inset). Sodium pyrophosphate did not show any effect on PtDXS activity even at a concentration of 1 mM. This suggests that the inhibitory effect of IDP and DMADP on PtDXS activity is not due to a nonspecific effect of the diphosphate part of the molecules.
Computational Modeling-To understand substrate and inhibitor binding, a three-dimensional atomic structure of PtDXS in complex with TPP and Mg 2ϩ was built by homology modeling based on the crystal structures of DrDXS and EcDXS (40) and refined by molecular dynamics simulations. The modeled structure was stable with an average root mean square fluctuation of 0.75 Å (Fig. 7A). Significant fluctuations are located mainly in the region that was not observed in the crystal structure of DrDXS (residues 199 -243), the template used for the homology modeling, and where PtDXS has an eight-residue insertion. The core structure of PtDXS aligns well with the crystal structures of EcDXS and DrDXS (Fig. 7B). The interactions among the enzyme, the coenzyme TPP, and the metal ion Mg 2ϩ are illustrated in Fig. 8A and are essentially the same as in the crystal structures of EcDXS and DrDXS (40). The Mg 2ϩ ion is coordinated with Asp-145 and Asn-174 of the TPP-binding motif of GDGX 25-30 N (46), the pyrophosphate moiety of TPP, and two water molecules. The Mg 2ϩ coordination is stable during the production phase of the molecular dynamics simulation without any restraint. The coenzyme TPP is anchored at the active site mainly by its pyrophosphate and pyrimidine moieties (Fig. 8A) as in EcDXS and DrDXS. In addition to the interaction with the Mg 2ϩ ion, the pyrophosphate moiety of TPP is hydrogen-bonded to the side chains of His-73 and Lys-291 and the  main-chain amides of Gly-146 and Ala-147. In addition to many van der Waals interactions, the pyrimidine ring of TPP is hydrogen-bonded to Gly-114, Ser-116, and Glu-377. The protonation of Glu-377 is crucial for the interaction with N1 of the pyrimidine ring. The interactions of IDP or DMADP with the enzyme are very similar to those of TPP (Fig. 8B). The binding of IDP or DMADP to the enzyme is mainly through the interactions of its pyrophosphate moiety, which is very similar to the binding of TPP (Fig. 8C), but several van der Waals interactions are also predicted.

DISCUSSION
To study potential feedback regulation in the MEP pathway required a kinetic study of DXS in the presence of different MEP pathway metabolites. The most common assay for this enzyme involves the measurement of radioactivity incorporated into the product DXP from radiolabeled pyruvate (19 -22, 47). However, this assay involves a laborious separation of the precursors from products. Another useful method to study DXS activity involves a coupled spectrophotometric assay exploiting the consumption of NADPH by DXR enzyme, which uses DXP as a substrate (48). We found some ambiguities in early results because of potential effects of the tested metabolites on DXR. A fluorometric assay for DXS was developed using a fluorescent derivative of the product DXP (49). This assay suffered from a lack of selectivity. Another assay for DXS has been reported using HPLC-based separation of derivatized DXP with a fluorophore using fluorescence detection (50). This assay still involved an additional step for derivatization of the product. Recently, an assay based on circular dichroism has been reported for DXS (51). This assay appeared to be extremely important for studying the mechanistic behavior of DXS, illustrating detailed insights about different TPP-bound intermediates involved in the DXS-catalyzed reaction. Enzymatic synthesis of DXP from pyruvate and GAP by yeast transketolase has been successfully monitored by an HPLC-electrospray ionization-MS/MS-based technique (52). Here we report another DXS assay in which DXP produced in the enzymatic reaction is measured by LC-MS/MS. This method is well suited for studying inhibitors of DXS activity.
Recombinant PtDXS enzyme from P. trichocarpa exhibited a K m for pyruvate of 87.8 Ϯ 3.2 and 119.2 Ϯ 14.2 M by coupled assay and LC-MS/MS-based assay, respectively ( Table 1). The K m for GAP obtained from the coupled assay (18.5 Ϯ 0.7 M) was higher than that obtained from the LC-MS/MS-based assay (5.9 Ϯ 0.9 M). This could be due to the consumption of GAP in the coupled assay and in situ production of GAP from DHAP and triose-phosphate isomerase in the LC-MS/MS-based assay. The k cat values (ϳ0.5 s Ϫ1 ) obtained from different measurements by the coupled assay were higher than the values obtained from the LC-MS/MS-based assay (ϳ0.2 s Ϫ1 ) ( Table  1). Use of a substantially lower concentration of TPP (100 M) in the LC-MS/MS-based assay compared with 1 mM TPP in the coupled assay may have caused this variation. The k cat value of ϳ0.5 s Ϫ1 for PtDXS is lower than the reported k cat value of ϳ1.9 s Ϫ1 from R. capsulatus (45).
The feeding experiment with deuterium-labeled deoxyxylulose 5-phosphate by Wolfertz et al. (35) indicated that a feedback regulation controls the carbon flux through the MEP pathway but did not provide evidence for a specific mechanism. Our results show that DMADP and IDP, the very last metabolites of the MEP pathway, inhibit PtDXS, the first enzyme of this pathway. The inhibitors compete with thiamine pyrophosphate for binding with the enzyme. The K i values for the inhibitor binding are 65.4 Ϯ 4.3 M for IDP and 81.3 Ϯ 10.5 M for DMADP. It is interesting that inhibitors can compete with TPP, which is generally thought to be an integral part of the enzyme. The  JUNE 7, 2013 • VOLUME 288 • NUMBER 23 results reported here indicate that IDP and DMADP have the potential to inhibit other TPP-dependent enzymes.

Feedback Inhibition of Deoxyxylulose Synthase
The absolute physiological concentrations of chloroplastic IDP and DMADP are not known. Non-aqueous fractionation to measure the chloroplastic DMADP pool of kudzu leaves has estimated a range of ϳ0.25 to ϳ3.5 mM (34). Measurement of the chloroplastic DMADP pool by postillumination isoprene emission estimated a concentration of ϳ43 M in oak leaves (53,54). Metabolic profiling studies by LC-MS/MS estimated the chloroplastic IDP/DMADP pool to be ϳ30 M in poplar leaves (37). Considering the variability in the measurement of the metabolites by these methods, the physiological concentration for DMADP/IDP can certainly be assumed in the range of K i of these metabolites for PtDXS. Therefore, the inhibition of PtDXS by IDP and DMADP under physiological concentration constitutes a significant feedback regulation of the MEP pathway, which can play an important role in regulating the amount of carbon lost by plants as isoprene.
The apparent cooperativity of inhibitor binding causes PtDXS activity to be very sensitive to low concentrations of inhibitor but relatively less sensitive when the inhibitor concentration is above the K i . The crystal structures of DXS from E. coli and D. radiodurans show that the enzyme exists as dimer (40). Therefore, the inhibitor cooperativity would ensure some DXS activity even in the presence of high levels of DMADP and IDP. This could be important because DXP is also the substrate for thiamine and pyridoxol synthesis (21,22,55). If DXS were too effectively shut off by IDP and DMADP it could interfere with thiamine and pyridoxol synthesis.
Wolfertz et al. (35) observed that a very large reduction in DXS activity could be seen with little increase in the rate of isoprene synthesis. This work was done with eucalyptus, and the enzyme kinetics of eucalyptus isoprene synthase has recently been published (56). Eucalyptus isoprene synthase has a K m(DMADP) of 0.16 mM, k cat of 0.195 s Ϫ1 , and substrate inhibition (K is ) of 0.9 mM. The inhibition of DXS by IDP and DMADP alone would not be sufficient to reduce DXS activity to such an extent that could explain the constant overall rate of isoprene emission. The complex kinetics of isoprene synthase also plays a significant role in this case. A combination of inhibition of DXS by IDP and DMADP and substrate inhibition of isoprene synthase by DMADP can satisfactorily account for the constant overall rate of isoprene emission using physiologically realistic assumptions (not shown).
Computational modeling shows that the binding of IDP/ DMADP to PtDXS is mainly through the pyrophosphate moiety. Fig. 8, A and B, show that the oxygen atoms of the pyrophosphate moiety of both TPP and IDP have polar interactions with Lys-291, His-73, Gly-146, and Ala-147 residues of PtDXS. The pyrophosphate oxygen atoms from both TPP and IDP interact with Asn-174 and Asp-145 residues of PtDXS through the Mg 2ϩ ion. However, the polar interaction of TPP through the pyrimidine N1 and N5 atoms with Glu-377 and Ser-116, the nitrogen atom of the NH 2 group of pyrimidine C4 with Gly-114, and the thiazolium sulfur atom with Ser-178 of PtDXS is absent for the binding of IDP with the enzyme. The van der Waals interactions of the carbon chain of IDP/DMADP with Leu-179, Ala-352, Gly-146, and Ala-147 orient the pyrophosphate group in the appropriate position for binding with the enzyme. This interaction is important for binding of the molecule with the enzyme as pyrophosphate alone does not show any inhibitory effect (Fig. 6, inset).
Several studies indicated that DXS could have a role in the regulation of the MEP pathway. The regulation of gene transcription and translation by different developmental and environmental cues is most evident for DXS enzyme (57)(58)(59). Posttranscriptional regulation of DXS was observed by the level of the end product of the pathway (59). Higher or lower accumulation of isoprenoid end products in transgenic species having an over-or underexpressed DXS gene, respectively, was observed in both bacteria (60 -62) and plants (63)(64)(65). DXS gene expression is strongly regulated with different developmental stages and strongly correlated with carotenoid accumulation in tomato fruits (58). The DXS gene expression pattern under the influence of various exogenous elicitors in Ginkgo biloba is strongly correlated with ginkgolide accumulation (66). All of this evidence suggests a regulatory role for DXS in the MEP pathway. Recently, a feed-forward activation of IspF enzyme (2-C-methyl-D-erythritol-2,4-cyclodiphosphate synthase) by MEP and a feedback inhibition of IspF-MEP complex by a downstream isoprenoid farnesyl diphosphate have been reported in E. coli (67). This explains a new regulatory mechanism that modulates the synthesis of one of the key intermediates of the MEP pathway. It has also been shown that the steps using reducing power (steps involving DXR, hydroxymethylbutenyl-diphosphate synthase, and hydroxymethylbutenyldiphosphate reductase enzymes) act as strong regulatory points of the MEP pathway under different environmental conditions (37). This suggests the presence of a regulatory mechanism at the middle of the pathway. Our work shows the connection between the beginning and the end of the pathway. It suggests that the beginning of the pathway can control the flow of carbon through the pathway, coordinating with the signals provided by the end of the pathway.
In conclusion, PtDXS activity was monitored in the presence of different MEP pathway metabolites. Only IDP and DMADP were found to have significant inhibitory effect on PtDXS activity. Both IDP and DMADP compete with TPP for binding with the enzyme. The inhibitors also exhibit negative cooperativity for binding with the enzyme. Computational modeling shows that IDP/DMADP use similar polar and non-polar contacts as TPP for binding with the enzyme. Inhibition of PtDXS by IDP and DMADP shows a potentially important metabolic regulation within the MEP pathway that plays a significant role in controlling the carbon flow through this pathway. Beyond its role in regulation of the MEP pathway, the competition among IDP, DMADP, and TPP could affect nearly any other TPP-dependent reaction, depending on the relative affinities for these compounds.