The structure of glucose‐1‐phosphate thymidylyltransferase from Mycobacterium tuberculosis reveals the location of an essential magnesium ion in the RmlA‐type enzymes

Abstract

Tuberculosis, caused by the bacterium Mycobacterium tuberculosis, continues to be a major threat to populations worldwide. Whereas the disease is treatable, the drug regimen is arduous at best with the use of four antimicrobials over a six‐month period. There is clearly a pressing need for the development of new therapeutics. One potential target for structure‐based drug design is the enzyme RmlA, a glucose‐1‐phosphate thymidylyltransferase. This enzyme catalyzes the first step in the biosynthesis of l‐rhamnose, which is a deoxysugar critical for the integrity of the bacterium's cell wall. Here, we report the X‐ray structures of M. tuberculosis RmlA in complex with either dTTP or dTDP‐glucose to 1.6 Å and 1.85 Å resolution, respectively. In the RmlA/dTTP complex, two magnesium ions were observed binding to the nucleotide, both ligated in octahedral coordination spheres. In the RmlA/dTDP‐glucose complex, only a single magnesium ion was observed. Importantly, for RmlA‐type enzymes with known three‐dimensional structures, not one model shows the position of the magnesium ion bound to the nucleotide‐linked sugar. As such, this investigation represents the first direct observation of the manner in which a magnesium ion is coordinated to the RmlA product and thus has important ramifications for structure‐based drug design. In the past, molecular modeling procedures have been employed to derive a three‐dimensional model of the M. tuberculosis RmlA for drug design. The X‐ray structures presented herein provide a superior molecular scaffold for such endeavors in the treatment of one of the world's deadliest diseases.

Short abstract

PDB Code(s): 6B5E,; 6B5K

Introduction

Mycobacterium tuberculosis, a facultative intracellular pathogen, is the causative agent of tuberculosis. According to the World Health Organization, M. tuberculosis resulted in the deaths of 1.8 million people worldwide in 2015 alone.1 Although the number of tuberculosis deaths fell by 22% between 2000 and 2015, there is still cause for concern given the appearance of multidrug‐resistant and extensively drug‐resistant bacterial strains.2

The cell wall of M. tuberculosis, known to be essential for viability, consists of three components: the peptidoglycan, the arabinogalactan, and the outermost layer composed of mycolic acids.3 The arabinogalactan is connected to the peptidoglycan by a disaccharide composed of α‐l‐rhamnosyl‐(1→3)‐α‐d‐N‐acetylglucosaminosyl‐1‐phosphate.4 The precursor to the rhamnosyl moiety is dTDP‐l‐rhamnose, which is synthesized as outlined in Scheme 1.5 l‐rhamnose is found not only in M. tuberculosis, but also in a variety of pathogenic bacteria including but not limited to Salmonella enterica, Shigella flexneri, and various Escherichia coli strains.6 As can be seen in Scheme 1, four enzymes, RmlA, RmlB, RmlC, and RmlD, are required for the production of dTDP‐l‐rhamnose. Given that l‐rhamnose is not found in humans, these enzymes have served as potential targets for drug design.6, 7

Scheme 1.

Pathway for the biosynthesis of dTDP‐rhamnose

The focus of this investigation is on RmlA, which catalyzes the first step in the pathway, namely the conversion of glucose‐1‐phosphate and dTTP to dTDP‐glucose and pyrophosphate. Thus far the enzymes from Pseudomonas aeruginosa, S. enterica, and E. coli have been biochemically and structurally characterized.8, 9, 10 These enzymes are known to be allosterically regulated by dTDP‐rhamnose,11 and the allosteric binding pockets have been well characterized.8, 9, 10

Given that the rmlA gene has been shown to be essential for mycobacterial growth, it is somewhat surprising that the crystal structure of the enzyme from M. tuberculosis has never been described.12 Indeed, two recent reports focusing on possible inhibitors of the enzyme from M. tuberculosis relied solely upon molecular modeling procedures.13, 14 Here we report a functional and structural investigation of M. tuberculosis RmlA. Importantly, for the first time we demonstrate the manner in which enzymes in this superfamily coordinate a magnesium ion through their nucleotide‐linked sugar products.

Results and Discussion

Overall architecture of M. tuberculosis RmlA

The first structure determined in this investigation was that of RmlA in complex with dTTP. The crystals utilized for the X‐ray analysis belonged to the space group I4 with two subunits in the asymmetric unit. RmlA functions as homotetramer with 222‐symmetry. Thus, one of the twofold axes of the tetramer was coincident to a crystallographic dyad. The structure was determined to 1.6 Å resolution and refined to an overall R‐factor of 19.6%. Tyr29, in each subunit, adopts ϕ, ψ angles of ∼68^° and −81^°, respectively, which are outside of the allowed regions in the Ramachandran plot. The electron densities for Tyr29 in each subunit were unambiguous, however. It is located in a tight turn connecting β‐strands two and three of the subunits.

Shown in Figure 1(A) is a ribbon representation of the tetramer with the crystallographic axis indicated by the red line. The tetramer has overall dimensions of ∼77 Å × 78 Å × 84 Å. Subunit 1 in the asymmetric unit extends from Met1 to Glu286 with breaks in the polypeptide chain backbone between Phe147 to Leu154 and Ala159 to Pro163. Subunit 2 extends from Met1 to Glu286 with breaks in the polypeptide chain backbone between Glu146 to Ser155 and Glu157 to Ser165. The α‐carbons for the two subunits in the asymmetric unit superimpose with a root‐mean‐square deviation of 0.2 Å.

Figure 1.

Overall structure of M. tuberculosis RmlA. Shown in (A) is a ribbon representation of the RmlA tetramer, which displays 222‐symmetry. One of the twofold axis of the tetramer, indicated by the red line, packed along a crystallographic dyad thereby reducing the contents of the asymmetric unit to half a tetramer. Electron density corresponding to the bound dTTP in the active site of subunit 1 is shown in stereo in (B). The electron density map was calculated with (F_o‐F_c) coefficients and contoured at 3σ. The ligands were not included in the X‐ray coordinate file used to calculate the omit map, and thus there is no model bias. Shown in (C) is the electron density corresponding to the bound dTDP ligand in the allosteric binding pocket (subunit 1). The electron density map was calculated as described in (B) and contoured at 3σ. As can be seen, the density is weaker most likely due to the lower occupancy of the ligand. This figure and figures 2, 3, 4, and 6 were prepared with PyMOL.27

There was clear electron density in each subunit for a dTTP molecule in the active site and a dTDP molecule in the allosteric binding pocket. The active site and the allosteric binding pocket, within a subunit, are separated by ∼20 Å. Representative electron density for the dTTP ligand and its associated magnesium ions, as observed in subunit 1, is presented in Figure 1(B), whereas that for the dTDP molecule is displayed in Figure 1(C).

A ribbon representation of subunit 1 is depicted in Figure 2(A). The overall molecular architecture of the subunit is built around 14 β‐strands and 11 α‐helices. Pro16, which lies between the first β‐strand and the first α‐helix adopts the cis conformation. It is not close to the active site but rather is situated at a subunit:subunit interface. The predominant feature of the subunit is a seven‐stranded mixed β‐sheet with the active site positioned at the end of the β‐sheet. There are three additional anti‐parallel β‐sheets (one containing three β‐strands and the other two composed of two β‐strands). The allosteric binding pocket is surrounded primarily by α‐helices [Fig. 2(A)].

Figure 2.

Structure of the M. tuberculosis RmlA subunit. A ribbon representation of the RmlA subunit 1 is presented in (A). The bound ligands are drawn in stick representations, and the magnesium ions are depicted as black spheres. Shown in (B) is the region of the active site surrounding the dTTP/Mg²⁺ complex. Water molecules are depicted as red spheres. Potential hydrogen bonding interactions, within 3.2 Å, are indicated by the dashed lines. Presented in (C) are the observed interactions between the dTDP ligand and the protein in the allosteric binding pocket.

A close‐up view of the active site is displayed in Figure 2(B). The side‐chain of Gln80 plays a key role in anchoring the thymine ring into the active site cleft. Arg13 and Lys23 interact through their side‐chains with the γ‐ and α‐phosphoryl groups of the dTTP ligand, respectively. One of the magnesium ions, referred to as MgA, is octahedrally coordinated by an α‐phosphoryl oxygen, the side‐chains of Asp108 and Asp222 and three water molecules. The average ligand:metal bond length is 2.1 Å. The second magnesium ion, MgB, is octahedrally coordinated by an α‐, a β‐, and a γ‐phosphoryl oxygen and three waters. Again, the average ligand:metal bond length is 2.1 Å. Interestingly, in the structure of the E. coli RffH, only one magnesium ion was observed binding in the enzyme/dTTP complex (corresponding to MgA in the M. tuberculosis RmlA).15 The E. coli RffH is also a glucose‐1‐phosphate thymidylyltransferase and demonstrates 59% and 74% amino acid sequence identity and similarity, respectively, to the M. tuberculosis RmlA.

Shown in Figure 2(C) is a close‐up view of the allosteric binding pocket. Most of the ligand:protein hydrogen bonding interactions are provided by backbone amide groups (Gly113, Gly115, Gly117, Thr118, Ser119, and Ser248). There are two side‐chains, Thr118 and Ser119, that participate in hydrogen bonding interactions with the pyrophosphoryl group.

Structure of the M. tuberculosis RmlA/dTDP‐glucose complex

The second structure determined in this investigation was that of the enzyme in complex with its product, dTDP‐glucose. The crystals used in this analysis belonged to the space group P2₁2₁2₁ with two tetramers in the asymmetric unit. The α‐carbons for the individual subunits superimpose with root‐mean‐square deviations of between 0.2 Å (subunits 3 and 4) to 0.4 Å (subunits 5 and 8). There are little structural perturbations upon binding the dTDP‐glucose product. Indeed, the α‐carbons for the RmlA/dTTP and the RmlA/dTDP‐glucose subunits superimpose with a root‐mean‐square deviation of 0.3 Å.

The observed electron density corresponding to the bound ligand and a magnesium ion in subunit 1 is shown in Figure 3(A). A close‐up view of the active site is presented in Figure 3(B). In this complex, only one magnesium ion is observed binding, namely that coordinated by Asp108 and Asp222. In addition to these amino acid residues, the other ligands to the metal include an α‐ and a β‐phosphoryl oxygen and two water molecules. This type of metal binding is similar to that previously observed for the Salmonella typhi glucose‐1‐phosphate cytidylyltransferase with bound CDP‐glucose.16 The pyranosyl moiety of the dTDP‐glucose is anchored into the active site by the side‐chain of Glu158, the backbone amide group of Gly143, and the carbonyl oxygen of Val169. As in the first RmlA/dTTP structure, dTDP is observed in the allosteric binding pocket.

Figure 3.

Close‐up stereo views of the active site with bound dTDP‐glucose. The observed electron density in subunit 1 is presented in (A). The map was calculated and contoured as described in figure legend 1. Potential hydrogen bonding interactions between the protein and the dTDP‐glucose ligand are depicted as dashed lines in (B). The bound magnesium ion is represented by the gray sphere whereas ordered water molecules are depicted as red spheres.

RmlA‐type enzymes are known to be magnesium‐dependent.17 Surprisingly, of all the entries for RmlA‐type enzymes in the Protein Data Bank, not one has a bound nucleotide‐linked sugar with its associated magnesium ion. As such this investigation represents the first direct observation of the manner in which a magnesium ion is coordinated to the RmlA product. In one molecular dynamics study of the M. tuberculosis RmlA, a model was produced in order to aid in the development of novel inhibitors.14 Coordinates for the M. tuberculosis RmlA model were deposited with the Protein Model Database (PM0076036). The α‐carbons for the “RmlA model” and the complex determined in this investigation superimpose with a root‐mean‐square deviation of 0.9 Å. Shown in Figure 4(A) is a superposition of the ribbon drawings for the two models. Overall the molecular architectures are similar. The issue, however, arises in the active sites. The modeling never took into account the position of the magnesium ion, and specifically that Asp222 serves as a ligand to the metal. As can be seen in Figure 4(B), the net result is that in the theoretical model, Asp222 is splayed away from the active site. This type of detail has profound consequences for structure‐based drug design.

Figure 4.

Comparison of the experimentally determined RmlA model presented here with that based on the homology modeling previously described.13 A superposition of the ribbon representations for the two models in stereo is presented in (A) with the experimentally determined model highlighted in violet. Overall the folds are similar as would be expected. One of the major differences, however, is the positioning of the loop delineated by Trp 220 to Phe 226. As can be seen in (B), in the theoretical model the Asp 222 side‐chain is positioned away from the active site.

Kinetic analysis of M. Tuberculosis RmlA

The kinetic parameters for M. tuberculosis RmlA were determined by a coupled spectrophotometric assay as described in the Materials and Methods. Shown in Figure 5(A) and 5(B) are plots of initial velocities versus either dTTP or glucose‐1‐phosphate concentrations, respectively. As can be seen, RmlA demonstrates classical Michaelis‐Menten kinetics. The observed K _M values of 10.6 ± 0.7 μM and 32.7 ± 2.9 μM for dTTP and glucose‐1‐phosphate, respectively, are similar to those values reported for the enzyme from E. coli.10 M. tuberculosis RmlA displays catalytic efficiencies of 7.5 × 10⁵ M⁻¹s⁻¹ and 2.2 × 10⁵ M⁻¹s⁻¹ for dTTP and glucose‐1‐phosphate, respectively.

Figure 5.

Plots of initial velocities versus substrate concentrations. Shown in (A) is a graph of the initial rate versus dTTP concentration. The initial rate versus glucose‐1‐phosphate concentration is shown as a graph in (B).

The kinetic mechanisms for the E. coli thymidylyltransferase and the S. typhi cytidylyltransferase have been previously reported to be sequential‐ordered Bi Bi.10, 18 A superposition of the coordinates for the M. tuberculosis RmlA/dTTP complex onto those of the M. tuberculosis RmlA/dTDP‐glucose is presented in Figure 6. Most likely the binding of the glucose moiety in dTDP‐glucose mimics what would occur for the substrate, glucose‐1‐phosphate. If that is, indeed, the case, then the phosphoryl group of glucose‐1‐phosphate would be ∼3.1 Å from the α‐phosphorus of dTTP. It can thus be envisioned how the substrate, glucose‐1‐phosphate, is in line for a direct attack at the α‐phosphorus of the dTTP molecule. The pyrophosphate group, with its associated magnesium ion, would function as an ideal leaving group.

Figure 6.

Superposition of the ligands in the RmlA/dTTP and the RmlA/dTDP‐glucose complexes. The dTDP‐glucose is highlighted in violet bonds whereas the dTTP is colored in cyan.

In summary, the structure of the M. tuberculosis thymidylyltransferase has now been determined to high resolution and the coordination geometry around the catalytically required magnesium ions defined. The high‐resolution models presented herein thus provide new and important data for the subsequent design of novel therapeutic inhibitors. Treatment for tuberculosis typically requires extensive use of isoniazid, rifampicin, ethambutol, and pyrazinamide for two months followed by an additional four months of isoniazid and rifampicin.19 Without the development of shorter and more effective treatments, tuberculosis will remain a worldwide disease responsible for both significant illness and economic devastation.

Materials and Methods

Cloning, expression, and purification

The gene encoding RmlA (rv0334) from M. tuberculosis H37Rv was synthesized by Integrated DNA Technologies using codons optimized for protein expression in E. coli. The gene was cloned into a pET31b vector (Novagen) using NdeI and XhoI restriction sites to yield a construct with a C‐terminal His₆ tag. The pET31b‐rmlA plasmid was utilized to transform Rosetta2(DE3) E. coli cells (Novagen). The cultures were grown in lysogeny broth supplemented with ampicillin and chloramphenicol (100 mg/L and 25 mg/L concentrations, respectively) at 37°C with shaking until an optical density of 0.9 was reached at 600 nm. The flasks were cooled in cold water, and the cells were induced with 1 mM isopropyl β‐d‐1‐thiogalactopyranoside and allowed to express protein at 21°C for 20 h.

The cells were harvested by centrifugation and frozen as pellets in liquid nitrogen. The pellets were subsequently disrupted by sonication on ice in a lysis buffer composed of 50 mM sodium phosphate, 20 mM imidazole, 10% glycerol, and 300 mM sodium chloride, pH 8.0. The lysate was cleared by centrifugation, and RmlA with a C‐terminal His‐tag was purified at 4°C utilizing nitrilotriacetic acid resin (Qiagen) according to the manufacturer's instructions. All buffers were adjusted to pH 8.0 and contained 50 mM sodium phosphate, 300 mM sodium chloride, and imidazole concentrations of 25 mM for the wash buffer and 250 mM for the elution buffer. Following purification, the protein was dialyzed against 10 mM Tris‐HCl (pH 8.0) and 200 mM NaCl at 4°C, and concentrated to 9 mg/mL based on the calculated extinction coefficient of 0.71 (mg/mL)⁻¹ cm⁻¹.

Protein crystallization and X‐ray structural analyses

Crystallization conditions were surveyed by the hanging drop method of vapor diffusion using a sparse matrix screen developed in the Holden laboratory. Initial experiments were conducted with the enzyme in complex with either 10 mM MgCl₂ and 5 mM dTTP (Sigma‐Aldrich) or 10 mM MgCl₂ and 10 mM dTDP‐glucose at room temperature. The dTDP‐glucose ligand was prepared as previously described.20 X‐ray diffraction quality crystals grown in the presence of dTTP reached a maximal size of approximately 0.6 × 0.2 × 0.2 mm after 1 week. Specifically, they were grown by mixing 1:1 the protein sample at 9 mg/mL containing 10 mM MgCl₂ and 5 mM dTTP with a precipitant solution composed of 18–22% poly(ethylene glycol) 8000 and 100 mM MOPS (pH 7.0). The crystals belonged to the space group I4 with unit cell dimensions of a = b = 96.4 Å, and c = 151.9 Å. The asymmetric unit contained two subunits. Prior to X‐ray data collection, the crystals were transferred step‐wise into a solution composed of 25% poly(ethylene glycol) 8000, 250 mM NaCl, 100 mM MOPS (pH 7.0), 10 mM MgCl₂, 5 mM dTTP, and 15% ethylene glycol.

X‐ray diffraction quality crystals of the RmlA/dTDP‐glucose complex grew to a maximal size of approximately 0.8 mm x 0.2 mm x 0.4 mm after one to two weeks. They were grown by mixing 1:1 the protein sample at 9 mg/mL containing 10 mM MgCl₂ and 10 mM dTDP‐glucose with a precipitant solution composed of 20% poly(ethylene glycol) 3350, 200 mM KCl, and 100 mM CHES (pH 9.0). These crystals belonged to the space group P2₁2₁2₁ with unit cell dimensions of a = 72.3 Å, b = 111.3 Å, and c = 290.4 Å. The asymmetric unit contained eight subunits. Prior to X‐ray data collection, the crystals were transferred step‐wise into a solution composed of 25% poly(ethylene glycol) 3350, 250 mM KCl, 250 mM NaCl, 100 mM CHES (pH 9.0), 10 mM MgCl₂, 10 mM dTDP‐glucose, and 15% ethylene glycol.

X‐ray data sets were collected at the Structural Biology Center beamline 19‐BM at a wavelength of 0.9794 Å (Advanced Photon Source). The X‐ray data were processed and scaled with HKL3000.21 Relevant X‐ray data collection statistics are provided in Table 1. The structure of the RmlA/dTTP complex was solved via molecular replacement with the software package PHASER22 using PDB entry code 1H5S (E. coli RmlA)10 as a search probe. Iterative cycles of model building with COOT23 and refinement with REFMAC24 reduced the R _work and R _free to 19.5% and 22.7%, respectively, from 50 to 1.60 Å resolution. The structure of RmlA in the presence of dTDP‐glucose was solved via molecular replacement using the RmlA/dTTP complex coordinates as a search probe. Iterative cycles of model building with COOT and refinement with REFMAC reduced the R _work and R _free to 17.1% and 21.8%, respectively, from 50 to 1.85 Å resolution. Relevant refinement statistics are listed in Table 2.

Table 1.

X‐ray Data Collection Statistics

	RmlA/dTTP Complex	RmlA/dTDP‐glucose Complex
Resolution limits	50–1.60 (1.66–1.60)b	50.0–1.85 (1.92–1.85)
Space Group	I4	P212121
Unit Cell a (Å) b (Å) c (Å)	96.4 96.4 151.9	72.3 111.3 290.4
Number of independent reflections	89386 (8778)	195396 (18536)
Completeness (%)	98.5 (97.0)	97.7 (93.8)
Redundancy	5.9 (3.7)	6.5 (4.5)
avg I/avg σ(I)	55.7 (5.9)	45.8 (14.9)
R sym (%)a	4.9 (19.8)	10.3 (17.9)
Wilson B factor (Å2)	20.0	19.9

^a R _sym = (∑|I − $\overline{\mathrm{I}}$|/∑ I) x 100.

^bStatistics for the highest resolution bin.

Table 2.

Refinement Statistics

	RmlA/dTTP Complex	RmlA/dTDP‐glucose Complex
Resolution limits (Å)	50–1.60	50–1.85
a R‐factor (overall)%/no. reflections	19.6/89386	17.4/195366
R‐factor (working)%/no. reflections	19.5/84972	17.1/185587
R‐factor (free)%/no. reflections	22.7/4414	21.8/9779
Number of protein atoms	4217	17583
Number of heteroatoms	466	2003
Average B values
Protein atoms (Å2)	32.4	26.8
Ligands (Å2)	28.9	23.6
Solvent (Å2)	33.3	29.4
Weighted RMS deviations from ideality
Bond lengths (Å)	0.013	0.013
Bond angles (°)	1.7	1.7
General planes (°)	0.007	0.007
Ramachandran regions (%)b
Most favored	98.3	98.8
Additionally allowed	1.5	0.8
Generously allowed	0.2	0.4

^a R‐factor = (Σ|F_o ‐ F_c|/Σ|F_o|) x 100 where F _o is the observed structure‐factor amplitude and F _c. is the calculated structure‐factor amplitude.

^bDistribution of the Ramachandran angles according to PROCHECK.26

Kinetic analyses

Inosine, microbial xanthine oxidase, and human purine nucleoside phosphorylase were purchased from Sigma Aldrich.

RmlA activity was monitored spectrophotometrically by following an increase in absorbance at 290 nm concomitant with xanthine formation in a modified coupled assay first reported by Forget et al.25 Reactions were monitored with a Beckman DU 640B spectrophotometer. Briefly, pyrophosphate released by RmlA is hydrolyzed to free phosphate by inorganic pyrophosphatase. Purine nucleoside phosphorylase then reacts with inosine and phosphate to yield hypoxanthine. Finally, hypoxanthine is oxidized to xanthine via xanthine oxidase, and its formation is monitored at 290 nm with an extinction coefficient of 12200 M⁻¹ cm⁻¹.

100 μL reactions were set up at room temperature with 50 mM Tris‐HCl (pH 7.5), 1 mM inosine, 10 U/mL of E. coli inorganic pyrophosphatase, 1 μM purine nucleoside phosphorylase, and 1.5 U/mL of xanthine oxidase. The required inorganic pyrophosphatase with a C‐terminal tag was purified in the laboratory according to standard procedures. The K _M of the enzyme for dTTP was determined by holding the glucose‐1‐phosphate concentration constant at 400 μM and varying the dTTP concentrations from 1 to 200 μM. The K _M of the enzyme for glucose‐1‐phosphate was determined by holding the dTTP concentration constant at 100 μM and varying the glucose‐1‐phosphate concentrations from 1 to 250 μM.

As a control, all reactions were initially set up without RmlA to monitor the background rate due to phosphate contamination in the starting materials. Reactions were initiated with the addition of RmlA to a final concentration of 0.5 μg/mL. All data were fitted by initial velocity Michaelis – Menten kinetics to the following equation: v = (V _max[S])/(K _M + [S]). The k _cat values were calculated according to the equation k _cat = V _max/[E]_t. Relevant kinetic parameters are listed in Table 3.

Table 3.

Kinetic Parameters

Substrate	K M (μM)	k cat (s−1)	k cat/K M (M−1s−1)
dTTP	10.6 ± 0.7	7.9 ± 0.2	(7.5 ± 0.5) × 105
Glucose‐1‐phosphate	32.7 ± 2.9	7.3 ± 0.2	(2.2 ± 0.2) × 105

Conflict of Interest Statement

The authors have no competing financial interests.