Transition State Structure and Inhibition of Rv0091, a 5′-Deoxyadenosine/5′-methylthioadenosine Nucleosidase from Mycobacterium tuberculosis

Abstract

5′-Methylthioadenosine/S-adenosylhomocysteine nucleosidase (MTAN) is a bacterial enzyme that catalyzes the hydrolysis of the N-ribosidic bond in 5′-methylthioadenosine (MTA) and S-adenosylhomocysteine (SAH). MTAN activity has been linked to quorum sensing pathways, polyamine biosynthesis, and adenine salvage. Previously, the coding sequence of Rv0091 was annotated as a putative MTAN in Mycobacterium tuberculosis. Rv0091 was expressed in Escherichia coli, purified to homogeneity, and shown to be a homodimer, consistent with MTANs from other microorganisms. Substrate specificity for Rv0091 gave a preference for 5′-deoxyadenosine relative to MTA or SAH. Intrinsic kinetic isotope effects (KIEs) for the hydrolysis of [1′-³H], [1′-¹⁴C], [5′-³H₂], [9-¹⁵N], and [7-¹⁵N]MTA were determined to be 1.207, 1.038, 0.998, 1.021, and 0.998, respectively. A model for the transition state structure of Rv0091 was determined by matching KIE values predicted via quantum chemical calculations to the intrinsic KIEs. The transition state shows a substantial loss of C1′–N9 bond order, well-developed oxocarbenium character of the ribosyl ring, and weak participation of the water nucleophile. Electrostatic potential surface maps for the Rv0091 transition state structure show similarity to DADMe-immucillin transition state analogues. DADMe-immucillin transition state analogues showed strong inhibition of Rv0091, with the most potent inhibitor (5′-hexylthio-DADMe-immucillinA) displaying a K_i value of 87 pM.

Graphical abstract

Mycobacterium tuberculosis is the causative agent of tuberculosis (TB), a leading cause of death among infectious diseases.¹ In 2014, 9.6 million cases of TB were reported, and 1.5 million people died of TB-related infections. Accordingly, the continued rise of multidrug-resistant, extensively drug-resistant, and totally drug-resistant strains of M. tuberculosis has become a major public health concern.² These challenges highlight the pressing need for new antitubercular therapies with novel targets and mechanisms of action.

5′-Methylthioadenosine/S-adenosylhomocysteine nucleosidases (MTANs) are involved in S-adenosylmethionine (SAM)-related pathways, including methylation, polyamine biosynthesis, adenine salvage, the biosynthesis of quorum sensing autoinducers and radical reactions involving SAM.^3,4 MTANs catalyze the hydrolysis of the N-ribosidic bond in 5′-methylthioadenosine (MTA), S-adenosylhomocysteine (SAH), and 5′-deoxyadenosine (5′-dAdo; Figure 1). The absence of MTANs in human metabolism makes this protein class an appealing target for antibacterials that function through blocking the quorum-sensing pathway or SAM recycling.

Figure 1.

The MTAN reaction. MTANs catalyze the hydrolysis of the N-ribosidic bond of 5′-methylthioadenosine (MTA), S-adenosylhomocysteine (SAH), and 5′-deoxyadenosine (5′-dAdo) to produce methylthioribose (MTR), S-ribosylhomocysteine (SRH), and 5′-deoxyribose (5′-dR), respectively. MTAN: 5′-methylthioadenosine nucleosidase.

Previously, we reported models for the transition state (TS) structures of MTANs from Escherichia coli,⁵ Streptococcus pneumonia,⁶ and Neisseria meningitidis⁷ using a combination of experimental kinetic isotope effect (KIE) measurements and computational methods. Intrinsic KIEs report on bond vibrational changes between the ground state (GS) and TS of chemical reactions and can be used as boundary constraints on quantum chemical calculations to model the structure of reactants at the TS. Such analyses provide information that can be leveraged in the design of chemically stable analogues that mimic features of the TS structure.^4,8–10 TS analogues designed for the MTAN reaction function as extremely potent inhibitors with K_d values in the picomolar (pM) and femtomolar (fM) range.^11–13 Importantly, these inhibitors have been shown to disrupt quorum sensing pathways in E. coli and Vibrio cholerae¹⁴ and inhibit growth in Helicobacter pylori by blocking menaquinone biosynthesis.¹⁵

The Rv0091 protein from M. tuberculosis was annotated previously as a putative bifunctional MTA/SAH nucleosidase.¹⁶ Herein, we report the expression, purification, and kinetic characterization of Rv0091. We demonstrate that the preferred substrate for Rv0091 is 5′-dAdo, and the protein exhibits less efficient nucleosidase activity with MTA and SAH substrates. A model of the Rv0091 TS structure for MTA hydrolysis was produced by matching KIE values predicted via quantum chemical calculations to a family of intrinsic KIEs. This model reveals that the Rv0091 TS structure and DADMe-immucillin TS analogues share similar electrostatic distributions. The potential of DADMe-immucillin TS analogues to modulate Rv0091 activity was investigated via in vitro inhibition assays.

RESULTS AND DISCUSSION

Oligomeric State and Substrate Specificity for Rv0091

A synthetic gene was designed for Rv0091 (NCBI GenBank: CCP42816.1) and purchased from DNA2.0 Inc. in a pJexpress414 expression vector. Rv0091 was heterologously expressed in E. coli with an N-terminal His₆ tag and purified to homogeneity via Ni²⁺-affinity chromatography. The purified protein was analyzed via SDS-PAGE to reveal a protein band of a molecular weight consistent with the predicted monomer size for Rv0091 (29.4 kDa; Figure 2). To determine the oligomeric state of Rv0091, the purified protein was treated with glutaraldehyde to enable intersubunit cross-linking.¹⁷ SDS-PAGE analysis of the cross-linked protein revealed a band consistent with a dimeric oligomerization state (Figure 2). These data are consistent with previous reports for MTANs from other organisms, which have been characterized as homodimers in X-ray crystallography.^14,18–20

Figure 2.

SDS-PAGE analysis of purified monomeric and cross-linked Rv0091. Rv0091 was purified to homogeneity, and SDS-PAGE analysis revealed a band consistent with the predicted mass (29.4 kDa) of the monomeric protein. The pure protein was treated with 0.1% to 2% glutaraldehyde to allow for intersubunit cross-linking. A band consistent with a dimeric oligomerization state was observed for the cross-linked protein. Lane 1, protein molecular weight standards; lane 2, Rv0091 + 0.1% glutaraldehyde; lane 3, Rv0091 + 0.5% glutaraldehyde; lane 4, Rv0091 + 2% glutaraldehyde; lane 5, Rv0091 before cross-linking.

Recently, a 5′-methylthioadenosine phosphorylase (MTAP) was identified in M. tuberculosis²¹ and M. smegmatis.²² MTAPs catalyze the phosphorolysis of MTA to adenine and 5′-methylthioribose-1′-phosphate. Prior to these reports, MTAPs were believed to be absent from bacteria, and the biological reason why M. tuberculosis and M. smegmatis expresses both MTAP and MTAN is currently unknown. We explored the substrate specificity of Rv0091 by determining the kinetic constants with MTA, SAH, and 5′-dAdo.²³ These experiments indicate that the preferred substrate for Rv0091 is 5′-dAdo, which displayed a specificity constant (k_cat/K_M) 7-fold greater than MTA and 700-fold greater than SAH (Table 1). MTA was reported as the preferred substrate for M. tuberculosis MTAP,²¹ and the addition of phosphate to the Rv0091 reaction mixture did not enhance the rate of MTA hydrolysis. These data indicate Rv0091 does not exhibit MTAP activity and suggests that Rv0091 primarily functions as a 5′-dAdo nucleosidase.

Table 1.

Kinetic Parameters for Rv0091 with 5′-dAdo, MTA, and SAH^a

substrateb	kcat (s−1)	KM (µM)	kcat/KM (M−1 s−1)
5′-dAdo	0.47 ± 0.02	10.9 ± 1.5	43.4 × 103
MTA	0.023 ± 0.001	3.7 ± 0.4	6.3 × 103
SAH	0.004 ± 0.001	63.1 ± 7.0	0.06 × 103

^aParameters obtained using a luciferase-based assay²³ at 25 °C.

^bAdenosine and 2′-deoxyadenosine showed less substrate activity than MTA and were not further characterized.

5′-dAdo, 5′-deoxyadenosine; MTA, 5′-methylthioadenosine; SAH, S-adenosylhomocysteine.

The catalytic efficiencies (k_cat/K_M) for Rv0091 with 5′-dAdo and MTA (Table 1) are lower than those of E. coli MTAN (k_cat-MTA/K_M-MTA = 2.3 × 10⁶ M⁻¹ s⁻¹; k_cat-5′dAdo/K_M-5′dAdo = 0.8 × 10⁶ M⁻¹ s⁻¹).²³ The low catalytic efficiency of enzymes from M. tuberculosis, such as Rv0091, has been attributed to the slow doubling time of M. tuberculosis (24 h) as compared to E. coli (20 min).²⁴

Determination of Intrinsic KIEs for MTA Hydrolysis

KIEs for the hydrolysis of MTA by Rv0091 were measured via the competitive radiolabel approach^9,10 using MTA substrates with isotopic labels incorporated at sensitive or remote positions (Table 2). Isotope effects on enzymatic reactions measured via internal competition provide V/K KIEs, which report on all steps from substrate binding, up to and including the first irreversible chemical step.²⁵ The most valuable information for interrogating TS structure is derived from intrinsic KIEs, which reflect the chemical step alone.²⁶ For a given isotope x, the intrinsic KIE on an enzymatic reaction (^xk) can be extracted from the ^xV/K KIE using Northrop’s equation (eq 1),²⁵ when forward commitment (C_f), reverse commitment (C_r), and the equilibrium isotope effect (^xK_eq) are known.

$${}^{\mathrm{x}}V/K=\frac{{}^{\mathrm{x}}k+C_{\mathrm{f}}+C_{\mathrm{f}}·{}^{\mathrm{x}}{K}_{\text{eq}}}{1+C_{\mathrm{f}}+C_{\mathrm{r}}}$$ (1)

Table 2.

Kinetic Isotope Effects for the Hydrolysis of MTA by Rv0091

heavy MTA	light MTA	type of KIE	V/K KIEs	intrinsic KIEsc	calculated KIEsd
[1′-14C]	[5′-3H2]	primary	1.035 ± 0.004b	1.038 ± 0.005	1.039
[1′-3H]	[5′-14C]a	α-secondary	1.194 ± 0.009	1.207 ± 0.010	1.209
[9-15N, 5′-14C]	[5′-3H2]	primary	1.019 ± 0.006b	1.021 ± 0.007	1.023
[7-15N, 5′-14C]	[5′-3H2]	β-secondary	0.998 ± 0.004b	0.998 ± 0.005	0.985
[5′-3H2]	[5′-14C]	remote	0.998 ± 0.008	0.998 ± 0.009	0.990

^aThe KIE on the remote [5′-¹⁴C] label is assumed to be unity.

^bExperimental values were corrected for the remote [5′-³H₂] KIE using the expression KIE = KIE_obs × [5′-³H₂] KIE.

^cIntrinsic KIEs were determined by correcting V/K values using eq 2 assuming C_f = 0.079.

^dCalculated values (Gaussian 09, RB3LYP/6-31g(d) theory) corresponds to those for the final TS model shown in Figure 4.

As the Rv0091 reaction was irreversible under the conditions used to measure KIEs, C_r is assumed to be zero. As such, eq 1 is reduced to eq 2 whereby ^xk can be extracted from the ^xV/K KIE using C_f alone:

$${}^{\mathrm{x}}V/K=\frac{{}^{\mathrm{x}}k+C_{\mathrm{f}}}{1+C_{\mathrm{f}}}$$ (2)

The V/K and intrinsic KIEs for the hydrolysis of MTA by Rv0091 are reported in Table 2. Experimental KIEs were corrected for remote effects to provide the V/K KIEs, and intrinsic values were determined using eq 2 where C_f = 0.079. The value of C_f for Rv0091 was measured experimentally using the substrate trapping method developed by Rose,²⁷ as previously described (Figure S1).⁶

The primary C1′ KIE for N-ribosyltransferases is sensitive to atomic motion along the reaction coordinate and provides information for differentiating S_N1 and S_N2 reactions. For a fully dissociative S_N1 reaction, with complete loss of bond order to the leaving group and no participation of the incoming nucleophile at the TS, a primary 1′-¹⁴C KIE near unity would be predicted.⁹ For S_N1 mechanisms in which partial bond order to either the leaving group or the nucleophile is retained at the TS, 1′-¹⁴C KIEs would range from 1.01 to 1.03. By contrast, primary 1′-¹⁴C KIEs for associative S_N2 mechanisms range from 1.08 to 1.13. The 1′-¹⁴C KIE measured for MTA hydrolysis by Rv0091 was 1.038 ± 0.005 (Table 2), which suggests either the N9 of the adenine leaving group and/or the oxygen of the nucleophilic water (Ow) retain partial bond order with the C1′ position at the TS. The primary 9-¹⁵N KIE reports on the loss of C1′–N9 bond order and rehybridization of the N9 position. A value of 1.021 ± 0.007 was obtained for the primary 9-¹⁵N KIE on the Rv0091 reaction (Table 2), which indicates that the C1′–N9 bond undergoes significant, but not full, loss of bond order at the TS.

The α-secondary 1′-H KIE for N-ribosyltransferases reports directly on changes in hybridization at the C1′ position and reflects the degree of oxocarbenium character retained by the ribose ring in the TS structure.²⁸ A highly dissociative TS, with considerable oxocarbenium character, will typically display 1′-³H KIEs in the range of 1.06 to 1.22.⁹ By contrast, an associative TS with less oxocarbenium character will display 1′-³H KIE values ranging from 1.00 to 1.03. The 1′-³H KIE measured for hydrolysis of MTA by Rv0091 was 1.207 ± 0.010 (Table 2), which suggests a TS structure with significant oxocarbenium character. Finally, the β-secondary 7-¹⁵N KIE, which reports on the protonation state at the N7 position of adenine, was measured as 0.998 ± 0.005. Previous studies suggest that intrinsic 7-¹⁵N KIEs near unity for the MTAN reaction indicate protonation of the N7 position at the TS.¹¹

Computational Models of the Rv0091 TS Structure

To investigate the mechanism of MTA hydrolysis by Rv0091, we developed a model of the Rv0091 TS structure using intrinsic KIEs as experimental boundaries for density functional theory (DFT) calculations. As there is no structural information available for the conformation of MTA bound to Rv0091, the input geometry for the TS structure search was taken from the coordinates for 5′-methylthiotubercidin bound to E. coli MTAN (PDB: 1NC1).²⁰ A family of energy-minimized TS structures was generated by constraining the C1′–N9 bond distance along the reaction coordinate as detailed below. Predicted KIEs were calculated (Gaussian 09, RB3LYP/6-31g(d) theory)²⁹ from the scaled vibrational frequencies of the optimized structures for MTA in the GS and at the TS. The optimized structure of MTA in the GS was generated without constraints using water as an implicit solvent (polarizable continuum model) and was identical for all KIE calculations.

Previous studies on the TS structures of MTAN variants from E. coli,⁵ S. pneumonia,⁶ and N. meningitides⁷ found MTA hydrolysis proceeds through a D_N*A_N process wherein C1′–N9 bond cleavage and C1′–Ow bond formation occur in a stepwise manner (Figure 3). Accordingly, initial efforts to model the Rv0091 TS structure considered two possible D_N*A_N manifolds: (1) a D_N^‡*A_N mechanism in which cleavage of the C1′–N9 ribosidic bond to form an oxocarbenium intermediate is rate-limiting and (2) a D_N*A_N^‡ mechanism in which the oxocarbenium intermediate is in equilibrium with MTA and nucleophilic attack of water is rate-limiting (Figure 3).

Figure 3.

General mechanisms for the hydrolysis of MTA by Rv0091. The Rv0091 reaction is represented as two possible mechanisms in which the transformations are considered as their elementary steps. For MTA hydrolysis, the A_N step refers to the association of the water nucleophile and the D_N step refers to dissociation of the adenine leaving group. If the reaction proceeds through a concerted bimolecular TS, the mechanism is termed A_ND_N. If dissociation of the leaving group precedes association of the nucleophile and the reaction proceeds through a stepwise process via discrete transition states, the reaction is termed D_N*A_N. For D_N*A_N processes, a superscript “‡” is used to denote the rate-limiting step, e.g., D_N^‡*A_N or D_N*A_N^‡. MTA, 5′-methylthioadenosine; MTR, 5′-methylthioribose.

To examine the D_N^‡*A_N mechanism, TS structures were generated by fixing the C1′–N9 bond of MTA at increasing lengths from 1.6 to 3.0 Å in 0.2 Å increments. Predicted KIEs were calculated and TS structures yielding close matches to the intrinsic values were refined by varying the C1′–N9 bond along the reaction coordinate in 0.01 Å steps. The TS structure providing the closest match to the intrinsic KIEs was identified at a C1′–N9 distance of 2.78 Å, although this model matched only the primary 1′-¹⁴C KIE within experimental error (Figure S2). The influence of charge accumulation on the leaving group was investigated by a parallel set of calculations to those detailed above in which the N7 position of adenine was protonated at the TS. Protonation at the N7 position activates the adenine for scission of the C1′–N9 bond by neutralizing negative charge build-up on the leaving group and is a common TS feature among N-ribosyltransferases,¹⁰ including E. coli⁵ and N. meningitides⁷ MTANs. In this TS structure search, the best match to intrinsic KIEs was found at a C1′–N9 distance of 2.42 Å (Figure S3). Relative to the TS structures in which adenine was modeled as an anionic leaving group, the TS structure including N7 protonation provided a better match to intrinsic KIEs. The primary 1′-¹⁴C and 9-¹⁵N KIEs were well within experimental error of the intrinsic values, and the α-secondary 1′-³H and β-secondary 7-¹⁵N were just outside these boundaries (Figure S3).

The feasibility of a D_N*A_N^‡ mechanism was also considered, in which the oxocarbenium intermediate is in equilibrium with MTA prior to the rate-limiting addition of water (Figure 3). For D_N*A_N^‡ mechanisms, observed KIEs are the product of the equilibrium isotope effect (EIE) for the first step and the KIE for the second step (Figure 3).²⁸ The EIEs for oxocarbenium formation were calculated for both neutral (N7 protonated) and anionic (N7 deprotonated) forms of adenine (Figure S4). Predicted KIEs were then calculated for TS structures involving nucleophilic attack of water on the oxocarbenium intermediate with C1′–Ow bond distances from 1.7 to 2.7 Å. Interestingly, the predicted 9-¹⁵N EIEs for both neutral and anionic forms of adenine were within experimental error of the intrinsic KIEs (Figure S4). However, the product of the EIEs and KIEs (for the nucleophilic attack of water) did not agree with the intrinsic values for the 1′-¹⁴C or 1′-³H KIEs (Figure S4).

The preliminary D_N^‡*A_N model for MTA hydrolysis by Rv0091 (Figure S3) was refined through the addition of low bond order to the nucleophilic water. The TS structure providing the best match to intrinsic values (Table 2) had a C1′–N9 distance of 2.45 Å and a C1′–Ow distance of 2.70 Å (Figure 4). Interestingly, the C1′–Ow distance of 2.70 Å in this TS model is similar to the distance between the 1′ position of the aminoribitol ring and the bound active site water (2.75 Å) in the binary complex of 5′-methylthio-DADMe-immucillinA (MT-DADMe-ImmA) and E. coli MTAN (PDB: 1Y6Q; Figure S5).³⁰ The slight discrepancy between the predicted 7-¹⁵N KIE and corresponding intrinsic value (Table 2) likely results from hydrogen bonding interactions with Asp220, which is predicted by homology modeling to interact with the N7 position of adenine,³¹ but was not included in this TS model. This final model is most consistent with a dissociative A_ND_N TS displaying significant loss of C1′–N9 bond order, low bond order to the incoming nucleophile, and substantial oxocarbenium character of the ribosyl ring. The C1′–N9 distance of 2.45 Å corresponds to a 0.74 loss of bond order at the TS, and the nucleophilic water shares a bond order of 0.074 with C1′ (Table S1). Oxocarbenium character of the ribosyl ring at the TS is supported by a short O4′–C1′ bond (1.42 Å in MTA versus 1.27 Å at the TS), which corresponds to an increase of 0.468 in bond order relative to the GS (Table S1). In addition, natural bond orbital (NBO) analysis indicates the relative positive charge increases by 0.182 for C1′ and 0.178 for O4′ at the TS (Table S2).

Figure 4.

Calculated model for the Rv0091 TS structure. TS structure model of MTA hydrolysis catalyzed by Rv0091 was generated by matching KIEs predicted for calculated transition states (RB3LYP/6-31g(d) theory) to intrinsic KIEs. Features of the Rv0091 TS structure include a C1′–N9 bond distance of 2.45 Å, a C1′–Ow bond distance of 2.70 Å, protonation at the N7 position of adenine, and significant oxocarbenium character. TS, transition state; MTA, 5′-methylthioadenosine; KIEs, kinetic isotope effects; Ow, water.

Previous reports indicate the MTAN reaction for other species proceeds via highly dissociative (E. coli⁵ and S. pneumonia⁶) and early (N. meningitides⁷) D_N*A_N transition states (Figure 3). As such, inclusion of the nucleophilic water in the Rv0091 TS model (Figure 4) distinguishes it from that of other MTAN variants, though dissociative A_ND_N mechanisms are common among N-ribosyltransferases.^28,32–35 A feature of note in the Rv0091 TS model is the Ow–C1′–N9 angle is 150.4° rather than the 180° (Figure 4), which would be predicted for optimal orbital overlap in a pure S_N2 reaction. Accordingly, the partial bond orders present in this model dictate a nonlinear TS for the most favorable orbital overlap. The term “nucleophile-assisted S_N1” has been invoked to describe S_N1 mechanisms in which low bond order must be included to the nucleophile in order to match predicted KIEs with experimental values.³⁶ As the C1′–Ow bond order is low (0.074; Table S1) in the Rv0091 TS model, it is plausible that the active site water could be preassociated in the TS structure via stabilizing interactions with the oxocarbenium species, but not directly participating in a concerted displacement reaction. Preassociation of a nucleophilic water within electron-reorganization distance of the C1′ reactive center is consistent with mass spectrometry experiments carried out with MT-DADMe-ImmA and E. coli MTAN, where the enzyme, inhibitor, and nucleophilic water remain associated in the gas phase.³⁷

TS Analogues Are Picomolar Inhibitors of Rv0091

The first- and second-generation immucillins were developed as TS analogue inhibitors for N-ribosyltransferase reactions.^38–40 We sought to compare the electrostatics of the Rv0091 TS structure with those of 5′-methylthio-immucillinA (MT-ImmA; a first generation analogue) and MT-DADMe-ImmA (a second generation analogue; Figure 5). EPS maps extrapolated from single-point energy calculations (RB3LYP/6-31g(d) theory)²⁹ are presented in Figure 5 for the Rv0091 TS model, and optimized structures of MT-ImmA and MT-DADMe-ImmA. Comparison of the EPS maps reveals that the geometry and charge localization observed in the Rv0091 TS (Figure 5a) is more similar to that of MT-DADMe-ImmA (Figure 5b) than to MT-ImmA (Figure 5c). Consistent with these models, the K_i for MT-DADMe-ImmA with Rv0091 (1.5 nM) was 56-fold more potent than that of MT-ImmA (K_i = 85 nM).

Figure 5.

EPS maps for Rv0091 TS structure and immucillin TS analogues. EPS maps (red = partial negative; blue = partial positive) were extrapolated from single-point energy calculations (RB3LYP/6-31g(d) theory)²⁹ and visualized in GaussView 5.0 (isovalue = 0.04). (a) Energy-minimized structure and EPS map for the TS structure of MTA hydrolysis by Rv0091. (b) Energy-minimized structure and EPS map for MTDADMe-ImmA. (c) Energy-minimized structure and EPS map for MT-ImmA. EPS, electrostatic potential surface; TS, transition state; MTA, 5′-methylthioadenosine; MT-DADMe-ImmA, 5′-methylthio-DADMe-immucillinA; MT-ImmA, 5′-methylthioimmucillin-A.

DADMe-ImmA TS analogues inhibit MTAN activity in the picomolar to femtomolar range in other bacterial MTANs and display increased binding affinity when functionalized with hydrophobic groups at the C5′ position.^12,15 Crystallographic studies of DADMe-ImmA TS analogues with MTANs from Salmonella enterica¹⁸ and H. pylori¹⁵ revealed an elongated hydrophobic binding pocket proximal to the C5′ substituent of the bound immucillin scaffolds. Consistent with these findings, the most potent inhibition of Rv0091 was observed with DADMe-ImmA inhibitors bearing long alkyl groups at the C5′ position (Figure 6). For example, hexyl thioether 3 inhibited Rv0091 activity with a K_i of 87 pM, whereas methyl thioether 1 and butyl thioether 2 both displayed K_i values in the low nanomolar range (Figure 6).

Figure 6.

Inhibition of Rv0091 with DADMe-ImmA TS analogues. Inhibition of Rv0091 activity was monitored at 305 nm via a coupled assay with xanthine oxidase⁴³ in which the adenine product was converted to 2,8-dihydroxyadenine. The K_i values were calculated using the Morrison equation for tight-binding inhibitors.⁴⁴

Overall, DADMe-ImmA TS analogues inhibit Rv0091 less potently than they do other MTANs. For example, the K_i values for MT-DADMe-ImmA with E. coli MTAN and Rv0091 are 2 pM³¹ and 1.5 nM (Figure 6), respectively. This difference in binding affinity for MT-DADMe-ImmA is anticipated from the lower catalytic activity of Rv0091 relative to E. coli MTAN. Catalytic rate enhancement has been correlated with the affinity of TS analogues for their enzyme targets by applying the Wolfenden approximation.^4,41,42 As such, the attenuated binding of DADMe-ImmA inhibitors with Rv0091 is consistent with the slow turnover of Rv0091 indicated by k_cat. In addition, the DADMe-ImmA scaffold was designed to mimic N-ribosyltransferase reactions with highly dissociative TS structures.^38–40 The crystal structure of MT-DADMe-ImmA in binary complex with E. coli MTAN reveals a distance of 2.8 Å between the N1′ position of the 3′-hydroxypyrrolidine ring and C9 of the 9-deazaadenine moiety.³⁰ This inhibitor geometry is a good match to the TS structure of E. coli MTAN wherein adenine is 3.0 Å from C1′.³¹ By comparison, the Rv0091 TS structure is somewhat earlier with a C1′–N9 distance of 2.45 Å (Figure 4). Moreover, the protonated aminoribitol scaffold of the DADMe-ImmA analogues (Figure 6) is designed to mimic the oxocarbenium species resulting from near full dissociation of adenine. The earlier TS of Rv0091 (relative to E. coli MTAN) would result in less electropositive charge at the C1′ position, which may not be as accurately mimicked by the DADMe-ImmA scaffold. The EPS maps for the Rv0091 TS structure (Figure 5a) and MT-DADMe-ImmA (Figure 5b) reveal greater positive charge is localized at the 1′ position of the inhibitor. Accordingly, differences in TS geometry and electrostatics between E. coli MTAN and Rv0091 may also contribute to the lower potency observed for DADMe-ImmA inhibitors with the Rv0091 protein.

The TDR M. tuberculosis gene function library classifies Rv0091 as a nonessential gene.^45,46 This is consistent with the assignment of Rv0091 as a putative MTAN with roles in quorum sensing and SAM recycling. With the function of Rv0091 now defined as a 5′-dAdo hydrolase, the in vivo activity of Rv0091 is more likely coupled to radical SAM reactions that act to generate radical reaction centers, with 5′-dAdo and methionine as products. Radical SAM enzymes are recognized by the CX₃CX₂C motif,⁴⁷ and 21 of these proteins have been annotated in the M. tuberculosis genome.⁴⁸ Preliminary growth experiments with M. tuberculosis indicate that MT-DADMe-ImmA (3) and butylthio-DADMe-immucillinA (4) have no effect on growth of cultured cells.

CONCLUSION

The Rv0091 protein from M. tuberculosis, previously annotated as a putative MTAN, exists as a dimer in solution and displays substrate specificity for 5′-dAdo relative to MTA and SAH. Thus, the protein is not crucial in quorum sensing pathways but more likely plays a role in reactions involving SAM radical enzymes. Intrinsic KIE measurements and quantum chemical calculations were used to establish a model for the TS structure of Rv0091. The TS for Rv0091 differs from the transition states of previously reported MTANs. The Rv0091 TS is characterized by a significant decrease in C1′–N9 bond order, substantial oxocarbenium character of the ribosyl ring, and weak participation of the water nucleophile. DADMe-ImmA TS analogues resemble the Rv0091 TS structure and inhibit Rv0091 activity with K_i values in the range of 10⁻⁹ to 10⁻¹¹ M. The most potent inhibitor of Rv0091 activity identified in this study was hexyl-thio-DADMe-immucillinA, which exhibited a K_i of 87 pM. Studies to investigate the in vivo role of Rv0091 in mycobacteria, as well as the potential for DADMe-ImmA TS analogues to function as modulators of virulence in M. tuberculosis, are ongoing.

METHODS

Full details for all experimental methods are provided in the Supporting Information.