Structure of 2C-methyl-d-erythritol 2,4- cyclodiphosphate synthase: An essential enzyme for isoprenoid biosynthesis and target for antimicrobial drug development

Abstract

The crystal structure of the zinc enzyme Escherichia coli 2C-methyl-d-erythritol 2,4-cyclodiphosphate synthase in complex with cytidine 5′-diphosphate and Mn²⁺ has been determined to 1.8-Å resolution. This enzyme is essential in E. coli and participates in the nonmevalonate pathway of isoprenoid biosynthesis, a critical pathway present in some bacterial and apicomplexans but distinct from that used by mammals. Our analysis reveals a homotrimer, built around a β prism, carrying three active sites, each of which is formed in a cleft between pairs of subunits. Residues from two subunits recognize and bind the nucleotide in an active site that contains a Zn²⁺ with tetrahedral coordination. A Mn²⁺, with octahedral geometry, is positioned between the α and β phosphates acting in concert with the Zn²⁺ to align and polarize the substrate for catalysis. A high degree of sequence conservation for the enzymes from E. coli, Plasmodium falciparum, and Mycobacterium tuberculosis suggests similarities in secondary structure, subunit fold, quaternary structure, and active sites. Our model will therefore serve as a template to facilitate the structure-based design of potential antimicrobial agents targeting two of the most serious human diseases, tuberculosis and malaria.

The enzyme 2C-methyl-d-erythritol 2,4-cyclodiphosphate (MECP) synthase participates in the biosynthesis of the isomers isopentenyl pyrophosphate (IPP) and dimethylallyl pyrophosphate. These compounds are the universal five-carbon precursors of isoprenoids, a diverse and large family of natural products including sterols, dolichols, triterpenes, ubiquinones, and plastoquinone, and also components of macromolecules such as the prenyl groups of prenylated proteins and isopentenylated tRNAs (1–3). Isoprenoids contribute to many functions, including electron transport in respiration and photosynthesis, hormone-based signaling, the regulation of transcription, and posttranslational processes that control lipid biosynthesis, meiosis, apoptosis, protein cleavage, and degradation. In addition, certain isoprenoids constitute an important structural component of cell membranes.

Isoprenoid biosynthesis depends on the production of IPP, which occurs in mammals, higher plants, fungi, and certain bacteria through the mevalonate (MVA) pathway (4–7). This pathway begins with the conversion of acetyl-CoA to 3-hydroxy-3-methylglutaryl-CoA, followed by reduction, phosphorylation, and decarboxylation to generate IPP, some of which is then isomerized to dimethylallyl pyrophosphate. The MVA pathway was considered ubiquitous until recently, when a nonmevalonate-dependent route, the 1-deoxy-d-xylulose 5-phosphate (DOXP) or 2C-methyl-d-erythritol 4-phosphate (MEP) pathway, was discovered in chloroplasts, algae, cyanobacteria, eubacteria, and apicomplexa (2, 6, 7).

Five distinct steps of the DOXP/MEP pathway have been elucidated (6–10). The pathway starts with the condensation of pyruvate and glyceraldehyde 3-phosphate to produce DOXP, which is then converted to MEP in reactions catalyzed by DOXP synthase (11, 12) and DOXP reductoisomerase (13–15), respectively. MEP is linked with CTP to produce 4-diphosphocytidyl-2C-methyl-d-erythritol (CDP-ME) and pyrophosphate in a reaction catalyzed by 4-diphosphocytidyl-2C-methylerythritol synthetase (16, 17). An ATP-dependent 4-(cytidine 5′-diphospho)-2C-methylerythritol kinase (18) then phosphorylates CDP-ME to produce CDP-ME-2-phosphate (CDP-ME2P). In the fifth stage, directly relevant to the present study, CDP-ME2P is converted to MECP and CMP (Fig. 1) by MECP synthase, an enzyme that, in E. coli, is encoded by the ispF gene (9, 10, 19–21). The stages by which MECP is converted to IPP are not yet clear, but further details of the DOXP/MEP pathway are expected to be unveiled in due course.

Figure 1.

The reaction catalyzed by MECP synthase.

A major impetus to the study of the DOXP/MEP biosynthetic pathway and the component enzymes is that it is absent from humans and found in many pathogens (6, 7, 10, 22). In addition to E. coli, the species that utilize the DOXP/MEP pathway include two of the world's most serious pathogenic microorganisms, Mycobacterium tuberculosis and Plasmodium falciparum, the causal agents for tuberculosis and cerebral malaria, respectively. Several components of the DOXP/MEP pathway are essential in eubacteria, including the three enzymes that catalyze the production of MECP starting from MEP and CTP (9, 23). Such genetic validation of the enzymes as therapeutic targets is complemented by the chemical validation of DOXP reductoisomerase (15, 22). This enzyme is inhibited by the potent antimicrobial fosmidomycin, which has proven efficacious in the treatment of rodent malaria (15). Fosmidomycin has a short half life in serum and is unsuitable as a human treatment, but this study has drawn attention to the enzymes in the DOXP/MEP pathway as exciting targets for the development of novel antimicrobial compounds (22, 24).

We are investigating the structure–reactivity relationships of MECP synthase and now report the crystal structure of the recombinant E. coli enzyme. The structure was determined by using multiwavelength anomalous dispersion measurements on a selenomethionine (SeMet) derivative and refined to 1.8-Å resolution. The protein fold, with a distinct topology, and trimeric quaternary structure are described; sequence and structural comparisons with functional and fold homologues are also presented. Details of protein–ligand interactions involving CDP and two metal ions provide insight into the specificity and mechanism of the enzyme.

Experimental Procedures

Sample Preparation and Crystallization.

The E. coli ispF gene was amplified from genomic DNA by PCR and cloned into the pET15b expression vector (Novagen). The resulting construct was sequenced to confirm the integrity of the product and heat-shock transformed into E. coli strains BL21 (DE3) and B834 (DE3) (Novagen) for protein expression. The latter is a Met⁻ strain and was used to prepare a SeMet derivative that was purified by using established protocols (25).

MECP synthase was incubated with 2 mM manganese chloride and 2 mM CDP for 2 h before crystallization by the vapor-diffusion hanging drop method. Orthorhombic plates in space group P2₁2₁2 grew over 4 days, to a maximum size of 0.05 × 0.2 × 0.4 mm, in drops assembled from 3 μl of protein solution (6 mg⋅ml⁻¹ 50 mM sodium chloride/50 mM Tris⋅HCl, pH 7.7) and 1 μl of a reservoir comprising 10–20% polyethylene glycol 2000, 0.1 M ammonium sulfate, and 0.1 M sodium acetate, pH 4.4–5.0.

Data Collection, Structure Solution, and Refinement.

Details are presented in Table 1. Crystals were maintained at 100 K in the presence of 2-methyl-2,4-pentane diol as a cryoprotectant and characterized in house and on station PX9.6 at the Synchrotron Radiation Source Daresbury Laboratory (Daresbury, U.K.), with measurements subsequently carried out on ID29 at the European Synchrotron Radiation Facility (Grenoble, France). A single crystal of the SeMet protein was used for a three-wavelength anomalous dispersion data collection by using monochromoter settings derived from an Se K-edge x-ray absorption near-edge structure scan. The unit cell lengths are a = 54.20, b = 114.85, and c = 87.75 Å, and a trimer of approximate mass 51 kDa constitutes the asymmetric unit.

Table 1.

Data collection and refinement statistics for SeMet MECP synthase

	λ1 (peak)	λ2 (inflection)	λ3 (remote)
Wavelength, Å	0.9786	0.9788	0.9184
Resolution range, Å	30–1.8	30–1.8	30–1.8
Measurements	176021	180420	193317
Unique reflections	48312	48628	50289
Coverage overall, %	93.9 (71.0)	94.3 (71.3)	97.4 (89.5)
I/σI	16.9 (1.6)	18.7 (1.4)	16.9 (2.0)
Rsym, %	7.2 (25.4)	6.2 (30.5)	6.7 (32.0)
Ranom, %	5.7	4.1	4.5
Riso, %	3.7		5.8
Rwork/Rfree, %			18.0/21.3
rms deviation bonds, Å/angles, °			0.018/1.9
Wilson B/average B, Å2			25.0/31.4

Numbers in parentheses refer to a high-resolution bin of ≈0.1 Å in width. 

Data were processed, scaled (DENZO/SCALEPACK; ref. 26), and input to the program SOLVE (27), treating the inflection point data (λ₂) as the native data set. Six Se positions were identified and provided phases to 2-Å resolution with an overall figure of merit of 0.48. Density modification (28) by solvent flattening and 3-fold noncrystallographic symmetry (NCS) averaging with phase extension into the high-energy SeMet data set (λ₃) raised the figure of merit to 0.65 at 1.8-Å resolution and resulted in an electron density map of excellent quality. The map was interpreted automatically (29) and a model of 434 residues constructed. The remaining amino acids and ligands were then included by using the program o (30) after rounds of refinement (31, 32) interspersed with map inspection. The careful placement of solvent molecules, metal ions, and a sulfate concluded the analysis. The progress of the refinement was monitored with the use of R_free (33), and an example of the electron density is provided in Fig. 2. The resulting model comprises three subunits, each with 157 amino acids, three CDP molecules, three Zn²⁺ and three Mn²⁺ ions, a sulfate, and 301 waters. A residue from the N-terminal tag is modeled and labeled as Glu-0. The final three residues of each subunit are disordered and are not included. Several residues on the surface of the protein are poorly ordered, and all or some of their side chain atoms have been assigned zero occupancy (Glu-15, Glu-28, Lys-29, His-34, Asp-63, and Lys-69 on each subunit). Figures were prepared by using o (30), molscript (34), RASTER3D (35), alscript (36), topdraw (available from C.S.B.), and promotif (37).

Figure 2.

The omit  F_o  −  F_c  α_c difference density for CDP, metal ions, and three solvent molecules in the active site. F_o are observed, F_c the calculated structure factors, and α_c phases derived from the model. The map is contoured at 3σ (purple) and 15σ (cyan). A ball-and-stick representation of CDP colored according to atom type is used where red, pink, and blue spheres represent oxygen, phosphorus, and nitrogen atoms, respectively; black sticks, the bonds to carbon; gray spheres, the metal ions; and blue sticks identify the ligand coordination. Four metal ion coordinating side chains are shown in a similar way to CDP.

Results and Discussion

Architecture of the Subunit and the Functional Trimer.

The MECP synthase subunit is a single α/β domain comprising 156 residues, of which approximately 36% are in α-helices, 35% in β strands, and 4% in two short 3₁₀ sections (Fig. 3). One side of the monomer has a flat four-stranded β-sheet with strand order 1 6 4 5; strand 6 is arranged antiparallel to the others. There are five α-helices with α1, α2, α3 following β1 and a short hairpin (β2-turn-β3), α4 lying between β4 and β5 with a turn of 3₁₀ helix on either side (θ1 and θ2), and α5 between β5 and β6. Helices α2 and α5 are only one turn in length. Most of the polypeptide segments linking elements of secondary structure are short, with the exception of the two loops formed between β3–α1 and α2–α3, which contribute to the active site (see below).

Figure 3.

Fold, topology, and sequence of E. coli MECP synthase. (a) Stereo Cα trace color-ramped from blue (N terminus) to red (C terminus). Every 10th Cα is depicted by a black sphere and labeled. (b and c) Ribbon and topology diagrams of a monomer (β-sheet, purple; α-helix, gold; θ-helix, red). (d) Sequence alignment of MECP synthase from E. coli, M. tuberculosis, and part of the P. falciparum protein. The secondary structure elements based on the E. coli enzyme structure are colored and labeled as in b. Residues whose identity is strictly conserved in all three sequences are boxed in black, conservative substitutions in gray, gray triangles mark Zn²⁺ ligands, and an open triangle the Mn²⁺-binding Glu-135. Circles mark CDP-binding residues (open and filled to distinguish subunits).

The crystal structure presents a compact homotrimer in the asymmetric unit, which forms an extended trigonal prism approximately 45 Å (axial) and 56 Å (equatorial) in size. The average surface accessible area per subunit is approximately 7,750 Å (3). Because the total for the trimer is about 16,080 Å (3), this means that a surface area equivalent to that of one subunit is buried on oligomerization. Consistent with the relatively large area utilized at each subunit interface, we note that the enzyme forms a highly stable trimer that persists in SDS/PAGE and in matrix-assisted laser desorption ionization time-of-flight mass spectrometry (data not shown). Because residues from a pair of subunits form the active site (see below), we conclude that this stable oligomer represents the functional enzyme. The structure of the subunits and their interactions with ligands are well conserved in the asymmetric unit. For example, the rms deviation between 157 Cα atoms shared by the subunits is less than 0.4 Å.

The trimer is formed by using a single type of subunit–subunit interface with almost all of the intersubunit hydrogen bonds involving side-chain atoms. The elements of secondary structure that contribute to this interface are β1 and the C-terminal section of α1 interacting with β6/β4 and β4/β5 sections of an adjacent subunit, the α1-loop-α2 sections and the N terminus of α4. There is also a self association of α5. The assembly creates a triangular β prism at the core of the homotrimer formed by the antiparallel alignment of three pairs of β1 and β6 strands around the 3-fold noncrystallographic symmetry (NCS) axis. An interesting pattern of side chains and their interactions occurs around the NCS axis and makes a significant contribution to the stability of the trimer. The β prism is fluted at the “top” end (designated by the C-terminal end of strands 1, 4, and 5; Fig. 4), where a sulfate is observed. The anion is positioned by three arginines (Arg-142), which in turn are held by a salt-bridge interaction with three glutamic acids (Glu-144). Beneath these polar residues is a solvent-filled cavity lined by a series of hydrophobic residues, six phenylalanines (three Phe-7 and Phe-139 combinations), three valines, and three isoleucines (Val-9 and Ile-99). Further down the β prism are three interacting glutamates (Glu-149), which must be protonated, held in place by interactions with histidines (His-5), all of which are buried by formation of the trimer. The region between these polar residues and the bottom surface of the trimer is filled by the aliphatic side chains of Met-1, Ile-3, Val-151, and Leu-153 provided by each subunit.

Figure 4.

Chain trace of the MECP synthase trimer viewed parallel to the 3-fold axis (chain A, purple; chain B, blue; chain C, gold). The hydrophobic intersubunit cavity is shown as a red semitransparent surface, a sulfate, and those water atoms in the cavity are shown.

The Active Site and Interactions with CDP and Two Metal Ions.

The homotrimer carries three active sites located in a cleft formed by two subunits, each of which contributes residues that interact with CDP. The active site is principally formed by the C-terminal section of α1, the turn leading into and the N-terminal region of α1 together with the short α2 of one subunit. These segments interact with one of the metal ions (Zn), the ribose and diphosphate of CDP, whereas the base mainly interacts with the N terminus of α4 and the C terminus of β5 from the partner subunit. In the subsequent discussion, we use a prime (′) to signify a residue from the partner subunit.

The cytosine is positioned in an aliphatic pocket created by the side chains of Ala-100′, Lys-104′, Met-105′, Leu-106′, Ala-131′, and Thr-133′, and forms four hydrogen bonds with main chain atoms (Fig. 5). The base amine N4 donates hydrogen bonds to the carbonyl groups of cis-Pro-103′ and Ala-100′, whereas N3 and O2 are acceptors for such interactions with the amides of Met-105′ and Leu-106′, respectively. The ribose hydroxyl groups form direct hydrogen-bonding interactions with the carboxylate of Asp-56 and the amide of Gly-58 (Fig. 5) and, in addition, solvent-mediated interactions with Asp-46 Oδ2 and the carbonyl of Ala-131′ are also observed (not shown). The side chain of Asp-56 is held in place by interactions with the amides of Gly-58, Lys-59, and Ala-131′, and it provides an anchor for the ribose. One α-phosphate phosphoryl oxygen interacts with Thr-133′ by accepting two hydrogen bonds from the amide and hydroxyl groups, whereas the other free α-phosphate oxygen atom coordinates Mn²⁺. The β-phosphate provides oxygen ligands for both Zn²⁺ and Mn²⁺ and a solvent-mediated interaction with Thr-132′ Oγ (not shown).

Figure 5.

The substrate-binding site of MECP synthase (stereo view). Residues contributed from subunit A with yellow bonds and subunit B with purple bonds. Selected hydrogen-bonding interactions are shown as blue dashed lines. The proposed ME2P-binding site is indicated.

CDP represents a fragment of the physiological substrate CDP-ME2P. The position of the β phosphate in our structure infers that the ME2P-binding site is a cavity, created by the two loops that lead into α1 and α3, with α3 providing the floor. The hydrophobic side chains of Ile-57, Phe-61, and Leu-76 form the base of the cavity, which is lined by several main chain functional groups (not shown). The sequence conservation and position of three residues around this putative ME2P-binding site, His-34, Asp-63, and Lys-69 (Fig. 3; see below), suggest that they may interact with the substrate. An obvious role for the basic residues would be that of binding the ME2P phosphate. In all of the sequences available for MECP synthase (over 45; data not shown), the pentapeptide sequence starting at His-34 in the E. coli enzyme is very highly conserved (Fig. 3). The position of this segment at the side of the active site suggests a key role in substrate binding. Although the electron density for the main chain in this part of the structure is well defined, density for the side chains of His-34, Asp-63, and Lys-69 was not convincing enough for the inclusion of all atoms in the refined model. The thermal parameters in these two loops are approximately 10–25 Ų higher than the average thermal parameter for all atoms of 31.4 Ų and are the most flexible parts of the polypeptide chain.

The structure now reveals the basis of the dependence of enzyme activity on the presence of divalent cations and loss of activity when EDTA is present (19–21). In the complex, there are two metal ions ≈5.7 Å apart that adopt different coordination geometries (Figs. 2 and 5). On the basis of using MnCl₂ in the crystallization conditions, both ions were at first refined as Mn²⁺. Subsequent analysis, including the use of atomic absorption spectroscopy, anomalous dispersion measurements near the Zn²⁺ K-edge, and refinement of complexes crystallized in the absence of Mn²⁺ (data not shown), indicates that the ion with tetrahedral geometry is in fact Zn²⁺. The coordination sphere of this ion involves interaction with two histidines (Nδ1 His-10, Nɛ2 His-42), the Asp-8 carboxylate Oδ2, and a CDP β-phosphate oxygen. The amino acid ligands are positioned in well-defined elements of secondary structure, β1 and α1, respectively, and presumably hold the metal tightly so it can interact with and anchor the substrate against one side of the active site. The other ion has been assigned as Mn²⁺ and adopts an octahedral environment by using oxygen ligands contributed from three water molecules, both CDP α and β phosphate groups, and Oɛ1 of Glu-135 from a subunit different from that binding the Zn²⁺.

The CDP–enzyme complex structure suggests that both metal ions contribute to the precise alignment of the substrate α and β phosphates and also act as Lewis acids to polarize the phosphates. An associative in-line mechanism is most likely to occur with, as the first step, nucleophilic attack by the ME2P phosphate of CDP-ME2P on the β phosphate (Fig. 1). This reaction would generate a pentacoordinate transition state, also stabilized by metal ion coordination, which in the second stage then collapses to release CMP and the cyclodiphosphate products. The enzyme requires both divalent cations for catalysis to occur, one of which is always present in the active site (i.e., Zn²⁺), the other presumably brought in with the substrate.

Structural Homologues.

Searches for structural homologues were carried out with dali (38), ce (39), and dejavu (40). The results were interesting and varied depending on the program used. We found it useful to first amalgamate the six most significant structural homologues, then to divide them into two groups: (i) three distant sequence homologues from bacteria, and (ii) three examples of structurally similar domains with different topology.

Group i consisted of a putative enzyme called Yjgf (PDB code 1QU9; ref. 41), acyl carrier protein synthase (1F7L; ref. 42), and chorismate mutase (2CHS; ref. 43). Structural homology covers greater than 70% of the protein length, and similarities in topology are observed. An active-site cleft formed by pairs of subunits has in each case been identified. We note, however, that the overlay of the MECP synthase trimer with these homologues indicates that each enzyme has a distinct active site, and that any evolutionary relationship would therefore be distant. The group ii protein structures include tubulin (1TUB; ref. 44), the bacterial cell division protein FtsZ (1FSZ; ref. 45), and a mammalian Migratory Inhibition Factor (1FIM; ref. 46). The structural homology with tubulin and FtsZ covers only ≈25% of their full length but nevertheless defines an intact domain with no known catalytic function that uses the flat face of the β-sheet to interface with the rest of the molecule. Notably, both the first group of proteins and the Migratory Inhibition Factor share the trimeric quaternary structure of MECP synthase with, in structural terms, a well-conserved β prism core.

The MECP synthase subunit appears to represent a stable fold used as a domain of larger structures and is particularly useful for a homotrimer quaternary structure. This small subunit is unable to provide a complete catalytic structure by itself but acquires this property by oligomerization to form an enzyme active site with contributions from pairs of subunits.

Functional Homologues.

The E. coli MECP synthase shares approximately 35% sequence identity with the catalytic domain of the P. falciparum homologue (18) and 40% with the enzyme from M. tuberculosis. The three enzymes collectively share a sequence identity of about 20% (Fig. 3). Amino acid conservation extends throughout the sequence and is well maintained in elements of secondary structure and in particular in those regions of the protein that are involved in trimer assembly. For example, 9 of the 12 residues discussed earlier as being important to the β prism trimer core are either strictly conserved or homologous.

The sequence conservation also extends to those key residue side chains shown to interact with CDP and the metal ions. The recognition and interactions of E. coli MECP synthase with CDP and metal ions involve some 22 residues discussed previously. Of these, 13 are absolutely conserved in the E. coli, M. tuberculosis, and P. falciparum sequences, and a further seven are strictly conserved in at least two of the sequences and homologous in the other. The two residues of the 22, which differ in the three sequences, are Lys-59 and Leu-106, and in both cases they use the main chain amide to donate a hydrogen bond into the active site, so the nature of the side chain is of reduced significance.

The level of sequence conservation suggests a close similarity in the overall structures of MECP synthase from these three organisms. That the conservation extends to those residues forming the active site and interacting with the ligands indicates that the three-dimensional model resulting from our analysis provides a high-resolution template for a rational structure-based approach to aid the search for inhibitors of this enzyme and subsequently of isoprenoid biosynthesis. Success in this venture, which may be aided by using CDP as a lead compound, has the potential to provide new and improved therapies against a range of bacterial and parasitic infections, including two of the most serious of human diseases, tuberculosis and malaria.

While this manuscript was under review, two lower-resolution structures of this enzyme were reported (47, 48). There is good overall agreement between the structures, although details of the metal ion coordination, in particular in the lowest resolution structure (48), are less well defined.

Abbreviations

MECP

2C-methyl-d-erythritol 2,4-cyclodiphosphate

IPP

isopentenyl pyrophosphate

DOXP

1-deoxy-d-xylulose 5-phosphate

MEP

2C-methyl-d-erythritol 4-phosphate

CDP-ME2P

4-diphosphocytidyl-2C-methyl-d-erythritol 2-phosphate

SeMet

selenomethionine