AnPRT (anthranilate phosphoribosyltransferase), required for the biosynthesis of tryptophan, is essential for the virulence of Mycobacterium tuberculosis (Mtb). AnPRT catalyses the Mg2+-dependent transfer of a phosphoribosyl group from PRPP (5′-phosphoribosyl-1′-pyrophosphate) to anthranilate to form PRA (5′-phosphoribosyl anthranilate). Mtb-AnPRT was shown to catalyse a sequential reaction and significant substrate inhibition by anthranilate was observed. Antimycobacterial fluoroanthranilates and methyl-substituted analogues were shown to act as alternative substrates for Mtb-AnPRT, producing the corresponding substituted PRA products. Structures of the enzyme complexed with anthranilate analogues reveal two distinct binding sites for anthranilate. One site is located over 8 Å (1 Å=0.1 nm) from PRPP at the entrance to a tunnel leading to the active site, whereas in the second, inner, site anthranilate is adjacent to PRPP, in a catalytically relevant position. Soaking the analogues for variable periods of time provides evidence for anthranilate located at transient positions during transfer from the outer site to the inner catalytic site. PRPP and Mg2+ binding have been shown to be associated with the rearrangement of two flexible loops, which is required to complete the inner anthranilate-binding site. It is proposed that anthranilate first binds to the outer site, providing an unusual mechanism for substrate capture and efficient transfer to the catalytic site following the binding of PRPP.
The worldwide emergence of extensively drug-resistant Mycobacterium tuberculosis (Mtb), the causative agent of TB (tuberculosis) , has been suggested to be one of the most profound challenges facing global health . Treatment of TB requires a multidrug approach  and thus there is a desperate need for new and effective anti-TB drugs that function via novel modes of action .
Recent studies have shown that host CD4+ T-cells employ tryptophan starvation in an attempt to eradicate M. tuberculosis infection . This innate immune response is effective against organisms that are natural tryptophan auxotrophs, but as the tryptophan biosynthetic pathway can normally overcome this starvation in M. tuberculosis, the host immune response is rendered ineffective. Drugs that inhibit the tryptophan biosynthetic pathway could therefore work synergistically with the host immune response to starve M. tuberculosis of this essential amino acid. A tryptophan auxotrophic mutant of the virulent M. tuberculosis strain H37Rv, with a deletion of the gene encoding the enzyme that catalyses the second committed step in tryptophan biosynthesis, AnPRT (anthranilate phosphoribosyltransferase) (also known as trpD; EC 220.127.116.11), has a decreased ability to multiply in murine macrophages and is essentially avirulent in both immunocompetent and immunocompromised mice . AnPRT, like the other enzymes involved in tryptophan biosynthesis, is absent from mammals, so inhibitors of this enzyme may have limited mammalian toxicity, strengthening it further as an attractive target for anti-TB drug design.
AnPRT is a member of the PRT (phosphoribosyltransferase) family of enzymes characterized by their common substrate PRPP (5′-phosphoribosyl-1′-pyrophosphate) , which acts as the phosphoribosyl donor in a substitution reaction with a nitrogenous, and generally aromatic, base . At least four distinct structural types of PRT have been identified, with AnPRT the sole representative of a type III PRT . Type I PRTs are the most common, and enzymes in this class act on purines and pyrimidines as part of salvage pathways (e.g. adenine PRT, hypoxanthine PRT or uracil PRT) or in the case of orotate PRT as part of the de novo pyrimidine biosynthetic pathway . Quinolate PRT and nicotinate PRT are type II PRTs, and ATP PRT, which catalyses the first step of histidine biosynthesis, is the only type IV PRT identified to date.
The displacement of PPi from PRPP by a nitrogenous base that is catalysed by PRTs takes place with inversion of the stereochemistry at C1 of the ribose ring. Reactions of this type can be either SN1-like, in which PPi release precedes nucleophilic attack on PRPP by the base, generating a short-lived oxocarbenium ion intermediate, or SN2-like, where the nucleophilic attack and PPi loss occur in the same reaction step. KIE (kinetic isotope effect) experiments with type I orotate PRT are consistent with a transition state that has extensive ribo-oxocarbenium ion character [10–13]. Structures of type I PRT enzymes in complex with substrates and transition state analogues show that C1 of the ribosyl ring adopts various positions away from PPi and towards the nitrogenous base, as would be consistent with an SN1 mechanism [9,14,15]. It has been proposed that type I PRTs guide catalysis by binding PRPP in an orientation that favours its breakdown into PR (5′-phosphoribosyl) and PPi [16,17].
AnPRT catalyses the transfer of PR from PRPP to anthranilate. AnPRT is structurally distinct from the other PRT types, and instead has a similar structural fold, quaternary structure and active-site location to that of type II NPs (nucleoside phosphorylases) [18,19]. Type II NP enzymes catalyse glycosyl transfer reactions that result in the production of ribose 1-phosphate, which is the near-reverse reaction of that catalysed by AnPRT [18,19]. To date, most of the NPs characterized using KIEs are of type I and have been shown to have SN1-like transition states. The only type II NP characterized by KIEs, human thymidine phosphorylase, has been proposed to have an SN2 transition state [20,21]. Thus it is not clear whether AnPRT catalyses an SN1 or an SN2 reaction (Figure 1). Further evidence about the catalytic mechanism of AnPRT will aid in inhibitor design.
The biosynthesis of tryptophan from anthranilate
We have previously determined crystal structures of Mtb-AnPRT with and without PRPP (PDB codes 2BPQ, 1ZVW and 3QR9) [22,23]. The enzyme crystallizes as a homodimer with one active site in each subunit. Each monomer comprises two domains: a smaller α-helical N-terminal domain, and a larger C-terminal domain containing both α-helices and β-sheets. A hinge region, consisting of the α4–β1, β3–α8 and α9–β4 connecting loops, links the two domains. Dimer formation occurs through interactions between the helical N-terminal domains, as is also observed for the AnPRT enzymes from Acinetobacter baylyi (PDB code 4GTN), Sulfolobus solfataricus (Sso) , Pectobacterium carotovorum (PDB codes 1KGZ and 1KHD) , Thermus thermophilus (PDB codes 1V8G and 2ELC) and Xanthomonas campestris (PDB code 4HKM). The PRPP substrate has been shown to bind in the C-terminal domain of the enzyme, at the bottom of a deep cleft, with the PPi group buried the most deeply. When compared with the ligand-free protein structure, two loops (the β1–α5 and β2–α6 loops) can be identified in the PRPP- and Mg2+-bound structure that move out of the PRPP-binding site in order to accommodate the PRPP substrate. Lee et al.  used in silico docking to predict the binding site of the enzyme's second substrate, anthranilate, and surprisingly predicted two sites despite a 1:1 reaction stoichiometry with PRPP (Figure 1). Two similar anthranilate-binding sites (S1 and S2) were observed directly in Sso-AnPRT (PDB codes 1ZYK and 2GVQ) , and a third anthranilate-binding site most distal to the PRPP (S3) has been proposed to be part of a substrate capture mechanism that was also utilized in the binding of Mtb-AnPRT inhibitors . Structures of Mtb-AnPRT in complex with bi-anthranilate-like inhibitors show binding across both S1 and S2 and S2 and S3 [23,26].
In the present study, we investigated the Mtb-AnPRT catalytic mechanism using compounds that have been reported recently to have antimycobacterial activity both in vitro and in a mouse model of disease . The present study reveals that Mtb-AnPRT exhibits a broad substrate tolerance, and analysis of the crystal structures of the enzyme in complex with these alternative substrates has illuminated some details of the reaction mechanism.
Unless otherwise stated, all chemicals were obtained from Sigma–Aldrich. Anthranilate was obtained from Merck Schuchardt. 4-Fluoroanthranilate was obtained from Sigma–Aldrich, whereas 3-, 5- and 6-fluoroanthranilate and 3- and 4-methylanthranilate were obtained from Fluorochem.
The Mtb-AnPRT protein was expressed and purified as described previously . The Escherichia coli (Eco) trpFC gene had been cloned previously into the expression vector pProEX-HTb, which adds an N-terminal His6 tag and rTEV (recombinant tobacco etch virus) site. This plasmid was transformed into E. coli BL21(DE3) cells containing plasmids (pBB528 and pBB541) for co-expression of chaperonins (GroES and GroEL) .
The E. coli fusion protein of PRAI (5′-phosphoribosyl anthranilate isomerase) and InGPS (indole glycerol phosphate synthase) was purified as described for the Mtb-AnPRT protein with the omission of the size-exclusion chromatography step. For the NMR analyses, the proteins were dialysed against 150 mM NH4HCO3 (pH 8.0).
Protein concentrations were determined using a Nanodrop ND-1000 spectrophotometer (Thermo Scientific) and molar absorption coefficient values of 34 615 and 40 715 litres·mol−1·cm−1 for Mtb-AnPRT and Eco-PRAI–InGPS respectively.
A coupled assay where PRA (5′-phosphoribosyl anthranilate) formation is detected by the formation of InGP (indole glycerol phosphate) was used to determine the enzyme kinetic parameters of Mtb-AnPRT as described previously .
The final concentrations of the following components were included in the coupled assay: PRPP (0–65 μM), MgCl2 (1 mM), Mtb-AnPRT (0.1 μM), Eco-PRAI–InGPS (1.7 μM), anthranilate or anthranilate analogue (0–10 μM) and Tris buffer (50 mM Tris/HCl and 150 mM NaCl, pH 8.0). Reactions were initiated with the addition of Mtb-AnPRT, with initial rates of reaction determined by least-squares fit of the initial rate data. Raw data were fitted to the ternary mechanism equation:
using GraFit 5 (Erithacus Software).
The commercially available EnzChek® Pyrophosphate Assay Kit (Invitrogen) was used to determine the kinetic parameters pertaining to alternative substrate turnover, as not all compounds were turned over by Eco-PRAI–InGPS. Along with PRA, Mtb-AnPRT produces PPi in a 1:1 stoichiometric ratio with anthranilate. In this assay, the PPi produced by the Mtb-AnPRT reaction is converted first into Pi by inorganic pyrophosphatase (1:2), which is then reacted with the substrate MESG (2-amino-6-mercapto-7-methylpurine ribonucleoside), which has an absorbance maximum of 330 nm, by PNP (purine nucleoside phosphorylase) in a 1:1 ratio. The products of this reaction are ribose 1-phosphate and 2-amino-6-mercapto-7-methylpurine; the latter can be monitored spectrophotometrically at 360 nm. Inorganic pyrophosphatase, PNP and Eco-PRAI–InGPS were all added in excess to ensure that the AnPRT-catalysed reaction was the rate-limiting step. The molar absorption coefficient for 2-amino-6-mercapto-7-methylpurine production from anthranilate and PRPP was determined to be 26200 litres·mol−1·cm−1 under assay conditions using known concentrations of anthranilate.
All assays were measured with a Varian Cary 100 UV-VIS spectrophotometer using quartz cuvettes thermally equilibrated to 22°C. The final concentrations of the following components were included in the coupled assay: PRPP (120 μM), MgCl2 (1 mM), Mtb-AnPRT (0.1 μM), Eco-PRAI–InGPS (6 μM), anthranilate/anthranilate analogue (0–1 mM), MESG (200 μM), PNP (1 unit), inorganic pyrophosphatase (0.03 unit) and reaction buffer (50 mM Tris/HCl and 1 mM MgCl2, pH 7.5, containing 100 μM sodium azide) made up to 1 ml. Reactions were initiated with the addition of Mtb-AnPRT, with initial rates of reaction determined by least-squares fit of the initial rate data. Apparent Km values were obtained by fitting the data to the Michaelis–Menten equation:
whereas substrate inhibition data were fitted to the substrate inhibition equation:
using GraFit 5.
Solutions for determination of the mode of epimerization of PRA via MS contained 2 mM PRPP, 2 mM anthranilate, 1 mM MgCl2 and 20 mM NaCNBH3 made up to 2 ml with Tris buffer. Solutions were mixed by inversion and 0.2 μM Mtb-AnPRT was added. After a 5 min reaction time, solutions were spun through Vivaspin 10000 Da molecular-mass cut-off concentrators to remove the enzyme present, flash-frozen with liquid nitrogen and left to freeze dry (FreeZone 2.5, Labconco) for 4–5 days. Dried products were then resuspended in ultrapure water and analysed via UHR-TOF (ultra-high-resolution TOF) MS/MS (Bruker maXis 3G).
Solutions for analysis of Mtb-AnPRT/Eco-PRAI–InGPS reaction products contained 1.2 mM PRPP, 400 μM anthranilate/fluoroanthranilate or 2 mM methylanthranilate, 2 mM MgCl2, 0.6 μM Mtb-AnPRT, 6 μM Eco-PRAI–InGPS made up to 500 μl with ultrapure water for anthranilate/fluoroanthranilate or 5 mM NH4HCO3 for methylanthranilate. Reactions with anthranilate or fluoroanthranilate were initiated with Mtb-AnPRT immediately before analysis via MS, whereas reactions for methylanthranilate were incubated with all substituents present for 1 h before analysis.
Assay solutions for analysis of Mtb-AnPRT-catalysed reactions by 1H-NMR spectroscopy contained 2–2.5 mM PRPP, 0.5–1 mM anthranilate or anthranilate analogue, 0.2–0.8 μM Mtb-AnPRT, 2 mM MgCl2, 0.25 mM trimethylsilyl propanoic acid, 80–90% (v/v) 2H2O and 10 mM NH4HCO3 (pH 8). The 1D 1H-NMR experiments were acquired at 25°C on a DRX-400 spectrometer (Bruker), or an AV-600 spectrometer (Bruker) fitted with a cryoprobe, and using presaturation of the water resonance . A 1D 1H-NMR spectrum of each assay solution was collected before addition of Mtb-AnPRT, and then after addition of the enzyme at 100–300 s intervals, with averaging over 16–64 transients. In experiments with Eco-PRAI–InGPS, this enzyme was added at a concentration of 2.5 μM to an NMR sample in which PRA or PRA analogue had been formed in situ using the assay conditions detailed above, without removal of Mtb-AnPRT.
Crystallization of native Mtb-AnPRT has been described previously . For soaking experiments and data collection, native crystals were removed from the crystallization drop with a nylon loop. The native Mtb-AnPRT crystals were soaked in a mixture of the mother liquor with 5 mM PRPP, 2.5 mM MgCl2 and the alternative substrate to a final concentration of 1 mM or anthranilate to a final concentration of 5 nM for various periods of time (Supplementary Table S1 at http://www.biochemj.org/bj/461/bj4610087add.htm). The soaked crystals were quickly pulled through a drop of cryoprotectant containing the mother liquor and 26% (v/v) glycerol before being flash-cooled in liquid nitrogen.
Data collection, structure solution and refinement
Data were collected at the Australian Synchrotron, Beamline MX2, within 72 h. The crystals of the Mtb-AnPRT–anthranilate analogue complexes diffracted to a maximum resolution that varied between 1.89 and 2.43 Å (1 Å=0.1 nm). X-ray diffraction data were indexed, integrated and scaled with the autoPROC suite , XDS  and SCALA  or with the HKL2000 suite of programs . Chain A of the original Mtb-AnPRT–PRPP complex structure (PDB code 1ZVW) , without ligands, was used as the search model for structure determination by molecular replacement using MOLREP , and as the initial model for refinement of the AnPRT–anthranilate analogue complex structures. Refinement and model building was performed with Coot  and Refmac5 , or autoBUSTER , with restraints for the anthranilate analogues generated by using the PRODRG server . Restraints on bond lengths and angles were based on the ideal values of Engh and Huber  and model quality was assessed using MolProbity . Full data collection and refinement statistics are given in Supplementary Table S1. Figures illustrating structural details were prepared using PyMOL (Schrödinger; http://www.pymol.org), with maps generated using the FFT program  in the CCP4 suite .
The crystallographic co-ordinates and structure factors have been deposited in the RCSB PDB (Supplementary Table S1).
Mtb-AnPRT catalyses a sequential reaction and is inhibited by high levels of its substrate anthranilate
To examine the kinetic mechanisms of the reaction, initial rate measurements were taken at various concentrations of both substrates. The kinetic profile is entirely consistent with a sequential mechanism for this enzyme (Figure 2A). In addition, for assays performed at higher concentrations of the substrate anthranilate, a significant fall in the enzyme activity was observed, indicating that the enzyme is subject to significant substrate inhibition (Figure 2B and Table 1).
Plots for determination of the catalytic mechanism and anthranilate kinetic parameters
Fluorine- and methyl-substituted analogues of anthranilate act as alternative substrates for Mtb-AnPRT
In the process of screening a targeted library of 165 compounds for inhibitors of Mtb-AnPRT, we discovered several compounds, including 5MA (5-methylanthranilate) and orthanilate, that are alternative substrates for the enzyme . The fluoroanthranilates inhibit mycobacterial growth in a tryptophan-dependent manner  and we established that 4FA (4-fluoroanthranilate) was also an alternative substrate of Mtb-AnPRT. These results prompted us to investigate a series of anthranilate analogues, with fluorine- or methyl-substitutions on the aromatic ring, as substrates for Mtb-AnPRT (Table 1, and Supplementary Figures S1 and S2 at http://www.biochemj.org/bj/461/bj4610087add.htm). The anthranilate analogue with C6 fluorine substitution has been identified previously as inhibiting Pseudomonas aeruginosa pathogenesis in mice , and both C5 and C6 fluoroanthranilate analogues have shown promising bactericidal activity against M. tuberculosis in the absence of tryptophan . Orthanilate, which carries a sulfonyl group in place of the carboxy group of anthranilate, was also tested in the present study. The ability of Mtb-AnPRT to process these alternative substrates was probed using a combination of UV-based assays, MS and 1H-NMR spectroscopy.
In order to assess the ability of Mtb-AnPRT to process each anthranilate analogue, a coupled UV-based assay for PPi production was used. PPi production by Mtb-AnPRT was observed in the presence of all alternative substrates and PRPP, and kinetic parameters were determined for these anthranilate analogues (Table 1).
Fluorine substitution around the aromatic ring of anthranilate had a relatively minor effect on the rate of catalysis, with a decrease in kcat values of 25–40%. The apparent Km for the fluorinated analogues, however, is highly dependent on the site of substitution, with fluorine substitution on C3 or C5 resulting in an 18- and 6-fold increase in Km respectively, whereas fluorine substitution on either C4 or C6 produces similar Km values to that for anthranilate. The biggest decrease in kcat/Km is observed for the alternative substrate 3FA, where the electron-withdrawing fluorine substituent is found adjacent to the nucleophilic amino group. Ki values for substrate inhibition by these analogues are again highly dependent on the site of substitution, but show a different trend, with fluorine substitution on C3 and C4 resulting in a 24- and 8-fold increase in Ki respectively compared with anthranilate, but substitution on either C5 or C6 having relatively little effect. Of the anthranilate analogues tested in the present study, 6FA has kinetic parameters that are most similar to those of anthranilate.
Methylation of the anthranilate ring has a more significant effect on the rate of catalysis, with this substitution at C3, C5 or C6 resulting in a ~3-fold reduction in kcat, and substitution at C4 yielding a 10-fold reduction. All of the methylated analogues have much poorer affinity for the enzyme, with apparent Km values that are increased by at least two orders of magnitude relative to that of anthranilate. Substrate inhibition was not observed with these analogues, although this analysis was limited by the concentrations of methylated analogue required.
1H-NMR spectroscopy confirms the production of substituted PRA products from anthranilate analogues
To confirm that the hydrolysis of PPi in the presence of the anthranilate analogues and PRPP indeed results in product formation, the enzymatic production of substituted PRA analogues by Mtb-AnPRT was followed using 1H-NMR spectroscopy. Anthranilate, and its substituted analogues, were each incubated with an excess of PRPP and Mtb-AnPRT, and 1H-NMR spectra were recorded at 100–300 s intervals as the reactions proceeded. In the reaction with the native substrate anthranilate, there was a clear consumption of PRPP over time, as demonstrated by the loss of signal at 5.80 p.p.m. due to its anomeric proton, and a corresponding gain of signals between 6.8 and 7.8 p.p.m. and between 5.3 and 5.7 p.p.m., which are due to PRA (Figure 3, and Supplementary Figure S3 and Supplementary Table S2 at http://www.biochemj.org/bj/461/bj4610087add.htm). As observed previously , PRA is unstable and was observed to break down to anthranilate and ribose, with a peak concentration of PRA observed after approximately 12 min. The anthranilate released again forms a substrate for Mtb-AnPRT, allowing the PRA formation reaction to proceed until all of the PRPP is consumed (Figure 3A).
Progress curves for Mtb-AnPRT-catalysed reactions followed by 1H-NMR spectroscopy
Unexpectedly, two anomeric protons (at 5.70 and 5.31 p.p.m.) and two sets of aromatic proton peaks were identified corresponding to the product PRA. This indicates that both α- and β-epimers at the C1 position of the ribose moiety of PRA are present in the product mixture. The 5.70 and 5.31 p.p.m. resonances were assigned to the α- and β-epimers of PRA respectively on the basis of chemical shifts and coupling constants reported for related compounds [44,45]. From the relative integrals of the two anomeric proton peaks, the ratio of the α- to the β-epimer is approximately 7:3 at equilibrium under the conditions of the assay.
As with other PRT enzymes [46,47], it is reasonable to expect that the Mtb-AnPRT reaction proceeds with inversion of stereochemistry at the C1 position on the ribose ring of PRPP. Crystal structures of several different AnPRT enzymes [19,22,25] provide supporting evidence for this claim, as it can be seen that anthranilate can only approach from the β-position, thereby only creating β-PRA. It is therefore likely that the α-PRA observed is formed after the reaction due to rapid epimerization of β-PRA in solution, a phenomenon that has been identified in other PRA-like compounds such as phosphoribosyl amine [44,48]. Epimerization is likely to occur via intramolecular rearrangement through an imine intermediate (Supplementary Figure S4 at http://www.biochemj.org/bj/461/bj4610087add.htm). The products of the Mtb-AnPRT reaction, formed in the presence of NaCNBH3 to trap the proposed iminium ion intermediate, were analysed using MS. Evidence for the reduced imine was indeed observed, with m/z peaks present at 352.0782 and 374.0603 Da, corresponding to the masses of the reduced imine and its sodium salt respectively.
The Mtb-AnPRT-catalysed turnover of PRPP in the presence of all of the anthranilate analogues was followed using 1H-NMR spectroscopy. For all of these alternative substrates, excluding 3MA, the concurrent formation of the corresponding PRA analogue was detected (Supplementary Figure S5 and Supplementary Table S2 at http://www.biochemj.org/bj/461/bj4610087add.htm). As observed for native PRA, the fluorine- and methyl-substituted PRAs accumulate as a mixture of α- and β-epimers in an approximate 7:3 ratio in the assay solution.
Fluorinated and sulfonylated analogues of PRA were observed to be significantly more stable than PRA itself. As illustrated for 4FA in Figure 3(B), the stability of the product, 4F-PRA (4-fluoro-PRA), results in a more distinct biphasic progress curve during monitoring of the Mtb-AnPRT-catalysed reaction by NMR spectroscopy. Initially, the reaction progresses quickly until 4FA is consumed, and then it continues at a lower rate governed by the breakdown of 4F-PRA to release 4FA, until the excess PRPP is consumed. Conversely, the inability to detect the PRA analogue resulting from turnover of 3MA may reflect a shorter half-life of 3M-PRA (3-methyl-PRA) relative to PRA itself. NMR spectroscopy was also used to monitor directly the degradation of PRA and two substituted PRA analogues in solution (Supplementary Figure S6 at http://www.biochemj.org/bj/461/bj4610087add.htm).
The substituted PRA analogues were also shown to be substrates for the next two enzymes of the tryptophan biosynthesis pathway from E. coli PRAI and InGPS, which are expressed as a single protein with two catalytic domains . The formation of fluorinated and methylated InGP analogues was observed by MS, UV spectroscopy and 1H-NMR spectroscopy (Supplementary Figures S7–S9 and Supplementary Table S3 at http://www.biochemj.org/bj/461/bj4610087add.htm) in the presence of Eco-PRAI–InGPS. In the 1H-NMR experiments, it was observed that both epimers of PRA or 4F-PRA were consumed in line with the predicted rapid interconversion of these isomers. Unlike the other PRA analogues studied, sulfonyl-PRA could theoretically undergo PRA isomerization catalysed by PRAI to yield CdRP [1-(o-carboxyphenylamino)-1′-deoxyribulose-5′-phosphate], but not the decarboxylation catalysed by InGPS to form the indole InGP. Consistent with this, the formation of InGP was not observed on incubation of sulfonyl-PRA with Eco-PRAI–InGPS. However, the formation of a small amount of an uncharacterized compound or compounds was indicated by the appearance of a new set of 1H-NMR peaks (Supplementary Figure S8 at http://www.biochemj.org/bj/461/bj4610087add.htm).
Structures of Mtb-AnPRT in complex with fluorine- and methyl-substituted anthranilates define two distinct substrate-binding sites
Structures of Mtb-AnPRT soaked with the anthranilate analogues for various time periods prior to crystal harvesting and freezing were determined to gain structural insight into the binding of these alternative substrates. The lower turnover numbers of the enzyme with these analogues has allowed substrate-bound complexes to be visualized, which are difficult to access with the natural substrate anthranilate. The structures of these complexes reveal a consistent pattern where S3, the substrate-capture site, is occupied first, after which the substrate moves through the tunnel to a position adjacent to PRPP, the catalytic site (S1). The existence of more than one binding site for anthranilate has previously been inferred from modelling  and inhibition [23,26] studies and by analogy with results for the Sso-AnPRT enzyme . Notably, however, no direct structural information for Mtb-AnPRT is yet available showing the substrate anthranilate bound in a catalytically relevant mode.
The structure of Mtb-AnPRT soaked with PRPP, Mg2+ and 4FA for 6 min most clearly defines the two major sites S1 and S3 (Figure 4), and is used here as the reference structure to describe the active site and understand the reaction mechanism. In this complex, both sites are fully occupied by the 4FA substrate and the same interactions are made between enzyme and substrate in both molecules of the Mtb-AnPRT dimer. The enzyme structure is unchanged from its ligand-free state by substrate binding, and no domain movements are necessary to bring the anthranilate substrate into contact with the PRPP as was proposed for Sso-AnPRT . Superposition of the 4FA complex on to the substrate-free and liganded (with PRPP, Mg2+ and inhibitors) Mtb-AnPRT structures previously described  gave RMSDs of 0.22–0.88 Å for all Cα atoms.
Structure of Mtb-AnPRT soaked with PRPP, Mg2+ and 4FA for 6 min
The pocket within which the substrate binds in its catalytically competent S1 position is bounded by the PRPP ribose ring and residues from the β1–α5 and β2–α6 loops that close over PRPP (Val106, Gly107, Thr108, His136 and Asn138), together with Ala179, Arg193, Gly206 and Pro207. The binding mode of the 4FA molecule at S1 is shown in Figure 4(C). It has several interactions with the surrounding structure including hydrogen bonds between its amino group and the carbonyl oxygen of Gly107 and the C2 hydroxy group of PRPP, and hydrogen bonds made between its carboxy group, the side-chain amide group of Asn138, the η-nitrogen of Arg193 and several water molecules. The other notable interaction involves the aromatic ring, which is sandwiched between the Cα of Gly206 and the Cβ of Asn138. In this orientation, the amino nitrogen is 4.3 Å from the C1 atom of the PRPP ribose ring, in position to attack this atom from the β position. The 4-fluoro-substituent of 4FA is well accommodated, making hydrophobic contacts with Val106 Cγ2 (3.2 Å) and with the side chain of His136 (3.2–3.4 Å). The same binding mode is seen for the 5FA structure, where the fluorine moiety is situated ~3.5 Å from the centre of the His136 imidazole ring. Structures of Mtb-AnPRT with 6FA or 6MA bound also show binding at S1, where the C6-substituent projects unimpeded along the substrate tunnel. The accommodation of 4FA, 5FA and 6FA at S1 is consistent with the similarity of their kinetic parameters to that of the native substrate, anthranilate. Crystal structures of complexes with 3MA, 4MA and 5MA did not show binding at S1, and may be indicative of steric clashes amplified by the larger methyl substitutions. Likewise, 3FA accommodated in this site would result in unfavourable electrostatic interactions between the fluorine substituent and the Gly107 backbone carbonyl.
The second binding site of 4FA, which we designate the substrate-capture site, is located ~8 Å from S1 (distance measured between the centres of the two rings), in a position at the entrance to the tunnel that extends into the active site. This site is different from a more buried site (S2) that was predicted by Mtb-AnPRT modelling  and which is occupied by anthranilate in Sso-AnPRT structures , but is coincident with the outer inhibitor-binding site (S3) seen for many inhibitor complexes of Mtb-AnPRT [23,26]. That it differs in detail from the second anthranilate site found for Sso-AnPRT  is not surprising given that the residues involved are mostly non-conserved. In S3 (Figure 4D), the substrate is held in position by stacking between Pro180 on one side and the side chains of Tyr186 and Ala190 on the other, and by hydrogen bonds between its carboxy group, Arg194 and two waters, and between its amino group, a water molecule and the carbonyl oxygen of His183.
Early binding events as deduced from the substrate complex structures
All the crystal structures of Mtb-AnPRT complexed with the fluorine- and methyl-substituted anthranilates show binding at S3, even for the very shortest soaking times. This clearly points to binding at this site as the earliest event in substrate binding. We illustrate this for the 4FA substrate after a 6 s soak (Figure 5). For this complex, the binding mode in S3 is already well defined for both Mtb-AnPRT molecules of the dimer, with the substrate in the same orientation as seen in the longer 6-min 4FA-soaked complex. When the crystal structures of the other substrate complexes are compared, the substrate is always bound in S3 with its anthranilate ring stacked between Pro180 and Tyr186 as for the 4FA complex, and with its carboxy group hydrogen-bonded to Arg194. The substrate's aromatic ring can, however, be flipped to give two different modes of interaction by the amino group, hydrogen-bonded either to the backbone oxygen of His183 (4FA, 5FA, 4MA, 5MA and 6MA) or to Oδ1 of Asn138 (3FA, 6FA and 3MA). It is possible that the fluorine or methyl substituents force the substrate to adopt one of two orientations in S3 to minimize unfavourable interactions with the surrounding residues. These interactions and orientations of the substrates at S3 could contribute to the variation in kinetic parameters seen between substrates with substituent groups at different positions.
Structures of Mtb-AnPRT soaked with PRPP, Mg2+ and either 6MA for 30 min or 4FA for 6 s
The movement of anthranilate-like substrates from S3 into the catalytic S1 is more complex than it appears at first sight. At S3, the amino group for a number of the alternative substrates is oriented towards His183, hydrogen-bonding to the backbone carbonyl oxygen. The ring must be flipped 180°, however, to orient the amino group towards the PRPP molecule in the substrate pocket. This could happen by normal ‘breathing’ of the protein structure. However, the structure of the 6MA complex after a 30 min soak shows positive electron density at a third site, equivalent to the previously described S2 , which is located between S1 and S3 (Figure 5E). S2 is formed in a mostly hydrophobic pocket between the side chains of Met86, Ala179, His183, Tyr186 and His136. The electron density at this site is more poorly defined, and cannot be accurately modelled, suggesting that it is a transient site that lacks anchor points for the substrate molecule.
Mtb-AnPRT soaked with PRPP, Mg2+ and anthranilate for 4 s shows electron density present in the substrate-capture site alone. The observation that anthranilate can bind in this outer site without PRPP or Mg2+ present provides some insight into the order of substrate-binding events, and is consistent with the recent report by Evans et al.  that a bi-anthranilate-like inhibitor can bind in both S2 and S3 in the absence of PRPP or Mg2+. Previous comparisons of the Mtb-AnPRT PRPP-bound and ligand-free structures  show clear rearrangement of the flexible β1–α5 and β2–α6 loops upon PRPP/Mg2+ binding, and these rearrangements are likely to be necessary to construct the inner catalytic binding site for anthranilate. These flexible loops remain disordered in the 4-s anthranilate soak structure, indicating that binding to the outer site S3 does not require prior PRPP/Mg2+ binding or lead to the rearrangement of the loops. Binding of anthranilate at S3 may precede PRPP/Mg2+ binding, with the enzyme effectively sequestering this substrate until it can proceed to the inner catalytic site S1 to enable reaction with PRPP.
PRTs are present in the biosynthetic pathways of tryptophan and histidine, as well as playing a role in nucleotide-salvage and -synthesis pathways, and as such are important enzymes for bacterial metabolism. Their potential as antibacterial drug targets makes the understanding of their mechanisms of significant interest. The present study provides insight into the reaction mechanism of Mtb-AnPRT, and includes the first crystal structures of Mtb-AnPRT in complex with both PRPP and anthranilate analogues. These structures provide a clearer picture of how anthranilate is transported from the bulk solvent to the catalytic pocket for a productive reaction with PRPP.
Mtb-AnPRT catalyses a sequential reaction and catalytic activity is inhibited by high concentrations of the substrate anthranilate. Mtb-AnPRT is remarkably tolerant of substitution on the phenyl ring of anthranilate. Fluorine- and methyl-substituted analogues of anthranilate were accepted as alternative substrates, and substituted PRA products were identified by NMR and MS studies. Unsurprisingly, with its lower steric demands, fluorine substitution was more efficiently accommodated than the introduction of a methyl group. The bactericidal activity of 5FA and 6FA against M. tuberculosis demonstrated by Zhang et al.  is only observed in the absence of tryptophan, consistent with these compounds acting by targeting tryptophan biosynthesis. This activity could plausibly originate from the toxicity of fluorinated intermediates generated by the turnover of these compounds by Mtb-AnPRT and subsequent pathway enzymes.
In crystal structures of Mtb-AnPRT reported previously [22,23], it has been observed that residues from the flexible loops β1–α5 and β2–α6 that close over the PRPP molecule after binding into the substrate pocket form the contacts for anthranilate binding in S1. This observation, coupled with our findings of a sequential mechanism, strongly indicates that anthranilate can only productively bind at the inner site S1 after PRPP has bound. The crystal structures solved in the present study strengthen this claim further, as 4FA can clearly be seen to bind to the residues from the flexible loops at S1, namely Asn138 and Gly107. It can also be observed that the above-mentioned flexible PRPP-binding loops are in the ‘open’ position in the ligand-free structure and closed in the 4FA/PRPP/Mg2+-bound and PRPP/Mg2+-bound structures, indicating that the enzyme is only primed for efficient anthranilate binding at the catalytic S1 once PRPP is bound. In contrast, the AnPRT enzymes from S. solfataricus and Saccharomyces cerevisiae have both been reported to exhibit random sequential Bi Bi mechanisms [25,50]. For Sso-AnPRT, this discrepancy is surprising given the similarity of the substrate-binding modes between the two structures.
The closure of flexible loops over a substrate-binding pocket in Mtb-AnPRT is not unique among the PRT family. Substrate binding instigates conformational change in both type I and type II PRTs, either by flexible loop rearrangement (type I) or by relative domain rotation (type II). In both cases, the importance of enclosing the active site has been correlated with protecting the hypothesized highly reactive carbocation intermediate. Previous experiments on orotate PRT, ATP-PRT and hypoxanthine PRT using KIEs strongly support the idea of an SN1-like reaction [12,13]. The observation that none of the fluoroanthranilate analogues in the present study significantly adversely affects the overall kcat value of Mtb-AnPRT could indicate that the nucleophilicity of the amino group is not of great significance to the overall reaction rate, consistent with either a dissociative SN1-like mechanism for AnPRT or rate-limiting substrate-binding or release steps.
The substituted analogues allow direct observation of two distinct binding sites for anthranilate in Mtb-AnPRT: an inner catalytic site (S1) and an outer substrate-capture site (S3). Additionally, some evidence for passage of anthranilate between the outer S3 and the catalytic S1 is seen via an intermediate binding site (S2), in a manner consistent with previously observed inhibitor-binding studies . The role of S3 in trapping anthranilate and the flexibility of the loops associated with the active site may have a direct role in the productive use of highly reactive PRPP in the transferase reaction. If anthranilate is first captured in the outer S3 and PRPP then binds, leading to concomitant formation of the higher-affinity inner catalytic site by loop rearrangements, efficient shuttling of anthranilate to the catalytically relevant S1 is achieved. This sequence of events allows anthranilate to efficiently react with the developing oxocarbenium ion intermediate derived from PRPP and minimizes exposure of this highly reactive intermediate to water.
Although potentially facilitating efficient formation of PRA by AnPRT, the outer S3 may also be somewhat of an Achilles heel for the enzyme. These studies reveal that the substrate anthranilate significantly inhibits this enzyme, and it is tempting to speculate that this inhibition may arise from non-productive binding of anthranilate to S3 as anthranilate is transferred to the productive catalytic S1. It is noteworthy that bactericidal 5FA and 6FA differ from the other alternative substrates in their ability to mimic the anthranilate substrate inhibition, and this contrasting kinetic behaviour is suggestive of different affinities of the two anthranilate sites for these analogues. Moreover, several inhibitors of this enzyme have been shown to utilize S2 and S3 [23,26], suggesting that this unique allosteric feature of this enzyme may be fruitfully targeted for inhibition, and may be the basis for new anti-TB therapeutics.
indole glycerol phosphate
indole glycerol phosphate synthase
kinetic isotope effect
purine nucleoside phosphorylase
5′-phosphoribosyl anthranilate isomerase
Shaun Lott, Tammie Cookson, Alina Castell, Esther Bulloch and Emily Parker conceived and designed the experiments. Tammie Cookson, Alina Castell, Esther Bulloch and Francesca Short performed the experiments. Tammie Cookson, Alina Castell, Esther Bulloch and Emily Parker analysed the data. Emily Parker, Tammie Cookson, Alina Castell, Esther Bulloch, Genevieve Evans and Edward Baker wrote the paper. Alina Castell, Genevieve Evans, Tammie Cookson, Esther Bulloch, Emily Parker, Edward Baker and Shaun Lott revised the paper before acceptance.
We gratefully acknowledge Dr Joseph Mire, Dr Inna Kriger and Professor Jim Sacchettini (Department of Biochemistry and Biophysics, Texas A&M University, College Station, TX, U.S.A.) for the phenotypic high-throughput screens that identified 4FA, initial tryptophan biosynthesis inhibition experiments involving 4FA and for providing us with a small amount of 4FA compound for initial experiments. The Mtb-AnPRT expression plasmid was originally supplied by Dr Farah Javid-Majd (Department of Biochemistry and Biophysics, Texas A&M University). We also acknowledge Dr Jackie Kendall, Dr Julie Spicer and Professor Bill Denny (Auckland Cancer Society Research Centre, Auckland, New Zealand) for providing the compounds for the initial targeted screen against Mtb-AnPRT, and for the 5MA, 6MA and orthanilate compounds tested in the present study. This work was performed as part of the M. tuberculosis Structural Genomics Consortium (http://www.webtb.org).
The research was funded by grants from the Health Research Council of New Zealand, the Foundation for Research, Science and Technology of New Zealand and the Tertiary Education Commission of New Zealand, through the Maurice Wilkins Centre.
These authors contributed equally to this work.
Present address: Karolinska Institutet, Department of Microbiology, Tumor and Cell Biology, 171 77 Stockholm, Sweden.
Present address: Division of Molecular Microbiology, College of Life Sciences, University of Dundee, Dow Street, Dundee DD1 5EH, U.K.
The structural co-ordinates reported for Mycobacterium tuberculosis anthranilate phosphoribosyltransferase complexed with anthranilate or anthranilate analogues have been deposited in the PDB under codes 4N5V, 4N8Q, 4N93, 4OWM, 4OWN, 4OWO, 4OWQ, 4OWS, 4OWU and 4OWV.