Transition State Analogue Inhibitors of 5′-Deoxyadenosine/5′-Methylthioadenosine Nucleosidase from Mycobacterium tuberculosis

Abstract

Mycobacterium tuberculosis 5′-deoxyadenosine/ 5′-methylthioadenosine nucleosidase (Rv0091) catalyzes the N-riboside hydrolysis of its substrates 5′-methylthioadenosine (MTA) and 5′-deoxyadenosine (5′-dAdo). 5′-dAdo is the preferred substrate, a product of radical S-adenosylmethionine-dependent enzyme reactions. Rv0091 is characterized by a ribocation-like transition state, with low N-ribosidic bond order, an N7-protonated adenine leaving group, and an activated but weakly bonded water nucleophile. DADMe-Immucillins incorporating 5′-substituents of the substrates 5′-dAdo and MTA were synthesized and characterized as inhibitors of Rv0091. 5′-Deoxy-DADMe-Immucillin-A was the most potent among the 5′-dAdo transition state analogues with a dissociation constant of 640 pM. Among the 5′-thio substituents, hexylthio-DADMe-Immucillin-A was the best inhibitor at 87 pM. The specificity of Rv0091 for the Immucillin transition state analogues differs from those of other bacterial homologues because of an altered hydrophobic tunnel accepting the 5′-substituents. Inhibitors of Rv0091 had weak cell growth effects on M. tuberculosis or Mycobacterium smegmatis but were lethal toward Helicobacter pylori, where the 5′-methylthioadenosine nucleosidase is essential in menaquinone biosynthesis. We propose that Rv0091 plays a role in 5′-deoxyadenosine recycling but is not essential for growth in these Mycobacteria.

Graphical abstract

5′-Methylthioadenosine/S-adenosylhomocysteine nucleosi-dases (MTANs) are absent in human metabolism; however in bacteria, they are involved in pathways related to S-adenosylmethionine (SAM) recycling, quorum sensing, meth-ylation, and menaquinone and polyamine biosynthesis.^1–3 The MTANs catalyze N-ribosidic bond hydrolysis of 5′-methyl-thioadenosine (MTA), S-adenosylhomocysteine (SAH), and 5′-deoxyadenosine (5′-dAdo). Mycobacterium tuberculosis Rv0091 was originally annotated as a MTAN; however, recent studies demonstrate that it functions primarily as a 5′-dAdo nucleosidase.⁴ 5′-dAdo is generated together with methionine as products of radical SAM-dependent enzyme reactions (Scheme 1).⁵ The M. tuberculosis genome encodes 21 open reading frames annotated as radical SAM enzymes based on the sequence analysis identifying an iron–sulfur CX₃CX₂C motif that has been linked to radical SAM function.⁶ Accumulation of 5′-dAdo is reported to block the activity of some SAM-related enzymes as a product inhibitor.³

Scheme 1.

Reaction Catalyzed by Radical SAM Enzymes for the Reductive Cleavage of SAM with [4Fe-4S]⁺ in the Presence of Substrate [AH] To Give 5′-dAdo and Methionine

Tuberculosis (TB) is one of the most difficult bacterial infections because of the spread of multidrug resistant strains. TB affects one-third of the world’s population and is resistant to many of the current antibacterial therapies.⁷ TB poses a worldwide threat, and new antibiotics are needed to treat the causative agent, M. tuberculosis. Here we explore the transition state analogue specificity of Rv0091 MTAN and test several inhibitors against cultured organisms as potential antibacterials.

Transition state analogues capture the geometry and electrostatic features of an enzymatic transition state in chemically stable structures. The analogues have the potential to bind to the enzyme much tighter than the substrate, with an upper limit of the factor equal to the catalytic enhancement imposed by the enzyme, relative to the Michaelis complex.^8–10 A perfect transition state analogue is impossible to design, as the actual transition state structure consists of non-equilibrium bond lengths and charge distributions that cannot be captured in a stable molecule. The Immucillins and DADMe-Immucillins are stable transition state analogues of N-ribosyltransferases developed via transition state analysis using kinetic isotope effects (KIEs) and computational chemistry.^11–13 Several bacterial MTANs are inhibited by these compounds with potent dissociation constants (femtomolar to picomolar range).^14–17

Previously, the transition state structure of M. tuberculosis Rv0091 was analyzed and found to resemble the DADMe-Immucillins.⁴ The DADMe-Immucillins contain a 5′-thio group to mimic the MTAN substrate MTA. A family of the DADMe-Immucillins inhibited Rv0091 with dissociation constants with picomolar to nanomolar affinity. 5′-dAdo was previously shown to be the preferred substrate of Rv0091; therefore, we anticipated transition state analogues similar to the transition state for 5′-dAdo to have the best affinities. In this study, we incorporate features of 5′-dAdo into transition state analogue design. Inhibition constants are established for several 5′-deoxyalkyl-DADMe-Immucillin-A compounds. Chemical synthesis and inhibition for the novel compound 5′-deoxy-DADMe-ImmA (4) are described. The 5′-deoxyalkyl- and 5′-thio-containing DADMe-Immucillin-A compounds (MTA and SAH substrate features) displayed dissociation constants in the picomolar to low nanomolar range. The 5′-deoxyalkyl-DADMe-Immucillin inhibitors show optimal inhibitory activity with short 5′-substituents, while 5′-thioalkyl-DADMe-Immu-cillins bind most tightly with longer 5′-substituents. This inhibitory pattern is not found in other bacterial MTANs and is therefore unique to Rv0091. Molecular modeling of inhibitors into the catalytic site of Rv0091 indicates an altered binding mode for 5′-substituents compared to those of other bacterial MTANs.

Bacterial growth assays show that Rv0091 inhibitors have weak effects on the growth of M. tuberculosis and Mycobacterium smegmatis. In contrast, the same inhibitors displayed potent antibacterial activity against Helicobacter pylori species. The essential nature of MTAN in menaquinone synthesis for H. pylori is well-documented.^17,18 Thus, the function of Rv0091 MTAN is not essential for laboratory growth in these Mycobacteria.

MATERIALS AND METHODS

General Experimental Approach for Inhibitor Synthesis

Compounds 5–20 were synthesized as previously described.^17,19–23 Air sensitive reactions were performed under argon. Organic solutions were dried over anhydrous MgSO₄, and the solvents were evaporated under reduced pressure. Anhydrous and chromatography solvents were obtained commercially and used without any further purification. Thin layer chromatography (TLC) was performed on glass or aluminum sheets coated with 60 F254 silica gel. Organic compounds were visualized under ultraviolet light or via the use of a dip of ammonium molybdate (5 wt %) and cerium(IV) sulfate·4H₂O (0.2 wt %) in aqueous H₂SO₄ (2 M), one of I₂ (0.2%) and KI (7%) in H₂SO₄ (1 M), or one of 0.1% ninhydrin in EtOH. Chromatography (flash column) was performed on silica gel (40–63 μm) or on an automated system with a continuous gradient facility. 1H nuclear magnetic resonance (NMR) spectra were measured in CDCl₃, CD₃OD, DMSO-d₆ (internal Me₄Si, δ 0), or methanol-d₄ and ¹³C NMR spectra in CDCl₃ (center line, δ 77.0) or CD₃OD (center line, δ 49.0). Assignments of ¹H and ¹³C resonances were based on two-dimensional (¹H–¹H DQF-COSY, ¹H–¹³C HSQC, or HMBC) and DEPT experiments. Positive electrospray mass spectra were recorded on a Q-TOF tandem mass spectrometer.

(3S,4R)-tert-Butyl-3-(bromomethyl)-4-hydroxypyrroli-dine-1-carboxylate (2) Synthesis

Methanesulfonyl chloride (0.45 mL, 5.7 mmol) was added dropwise to a stirred solution of tert-butyl 3-hydroxy-4-(hydroxymethyl)pyrrolidine-1-carbox-ylate (1) (1.13 g, 5.2 mmol) and 2,6-dimethylpyridine (1.2 mL, 10.3 mmol) in acetone (20 mL) and the mixture left to stir for 24 h. The resulting suspension was filtered to remove salt, and then lithium bromide (2.25 g, 25.9 mmol) was added to the filtrate and the mixture refluxed for 3 h, at which time the reaction was deemed complete. The crude reaction mixture was absorbed onto silica gel (5 g) and concentrated in vacuo and the resulting residue purified by chromatography (20% → 40% → 100% EA/PE solvent) to afford title compound 2 (832 mg, 57%) as a syrup: ¹H NMR (500 MHz, CDCl₃) δ 4.06 (brs, 1H), 3.67 (brs, 2H), 3.47−3.36 (m, 2H), 3.28−3.18 (m, 2H), 2.50 (brs, 1H), 1.46 (s, 9H); ¹³C NMR (125 MHz, CDCl₃) δ 154.6, 79.9, (73.3, 72.5), (52.6, 52.3), (48.8, 48.3), (48.2, 47.6), 32.6, 28.5; HRMS (ESI) m/z calcd for C₁₀H₁₈NO₃BrNa⁺ 302.0368, observed 302.0364.

(3R,4S)-4-Methylpyrrolidin-3-ol (3) Synthesis

A mixture of 2, triethylamine (1.85 mL, 13.2 mmol), and Perlman’s catalyst (150 mg, 1.0 mmol) in ethanol (20 mL) was stirred under an atmosphere of hydrogen for 3 h. The crude reaction mixture was then filtered through Celite and concentrated in vacuo and the residue dissolved in methanol (2 mL) and concentrated HCl (2 mL) and concentrated in vacuo. The resulting syrup was dissolved in additional concentrated HCl (2 mL) and concentrated in vacuo. The residue was dissolved in MeOH, absorbed onto silica, and concentrated in vacuo, and the resulting solid was purified by chromatography (30% 7N NH₃ in a MeOH/CHCl₃ solvent) to afford title compound 3 (501 mg, 97%) as an oil: ¹H NMR (500 MHz, CDCl₃) δ 4.24 (brs, 1H), 3.65−3.55 (m, 2H), 3.25 (d, J = 15.0 Hz, 1H), 3.08−3.04 (m, 1H), 2.39 (brs, 1H), 1.09 (d, J = 10.0 Hz, 3H); ¹³C NMR (125 MHz, CDCl₃) δ 75.5, 50.8, 50.2, 39.9, 15.1; HRMS (ESI) m/z calcd for C₅H₁₂NO [MH]⁺ 102.0919, observed 102.0914.

(3R,4S)-1-[(9-Deazahypoxanthin-9-yl)methyl]-3-hy-droxy-4-methylpyrrolidine (4) Synthesis

Formaldehyde (0.41 mL, 5.4 mmol, 37 mass % aqueous) was added to a stirred suspension of 3 (500 mg, 4.9 mmol) and 9-deazaadenine (77 mg 5.4 mmol) in a mixture of water (2.5 mL) and ethanol (5 mL) and then left to stir at room temperature for 48 h. The crude reaction mixture was absorbed onto silica and concentrated in vacuo and the solid residue purified by silica gel chromatography to afford title compound 4 (647 mg, 53%) as a solid: ¹H NMR (500 MHz, methanol-d₄) δ 8.18 (s, 1H), 7.49 (s, 1H), 3.87−3.74 (m, 2H), 3.37 (s, 1H), 3.05 (dd, J = 9.6, 7.7 Hz, 1H), 2.82 (dd, J = 10.4, 6.5 Hz, 1H), 2.70 (dd, J = 10.4, 4.3 Hz, 1H), 2.17 (dd, J = 9.6, 7.9 Hz, 1H), 2.04 (ht, J = 7.2, 3.5 Hz, 1H), 1.05 (d, J = 6.9 Hz, 3H); ¹³C NMR (125 MHz, methanol-d₄) δ 152.1, 151.0, 147.0, 130.1, 115.1, 112.6, 79.0, 62.1, 61.0, 49.2, 43.0, 17.8; HRMS (ESI) m/z calcd for C₁₂H₁₈N₄O [MH]⁺ 248.1511, observed 248.1508.

Expression and Purification of 5′-Deoxyadenosine/5′-Methylthioadenosine Nucleosidase

Enzyme expression and purification were performed as described previously.⁴ Briefly, Escherichia coli BL21 Star (DE3) plysS cell lines were employed for expression. A 25 mL initial culture in LB medium (Gibco) supplemented with ampicillin (100 μg/mL) and chloramphenicol (100 μg/mL) was incubated overnight at 37 °C. The following day, 6 mL of culture was added to 1 L of fresh LB-ampicillin (100 μg/mL) and incubated at 37 °C to an OD₆₀₀ of 0.6–0.8. Cells were induced with 0.5 mM isopropyl D-thiogalactoside (IPTG) (Goldbio) and incubated overnight at 28 °C. Cells were harvested, and the protein was purified by Ni-NTA (Qiagen) chromatography with a 30 to 250 mM imidazole buffer containing 20 mM Tris-HCl (pH 7.4) and 300 mM NaCl. Fractions containing purified protein (>90% as determined by a sodium dodecyl sulfate gel) were pooled, exchanged into 50 mM HEPES (pH 7.4) and 10% glycerol, and stored at −80 °C.

Dissociation Constants

Dissociation constants (K_i values) of Rv0091 inhibitors were determined as previously described.²⁴ Briefly, the 5′-dAdo nucleosidase reaction mixture was coupled with 1 unit of xanthine oxidase (Sigma-Aldrich) in the presence of 1 mM 5′-dAdo and varying concentrations of inhibitor in 50 mM HEPES (pH 7.4) at 25 °C. Formation of 2,8-dihydroxyadenine was monitored at 305 nm for 2 h. The reactions were initiated by adding 10 nM enzyme. The K_i values were obtained from the rates with and without inhibitors (v_i/v_o) that were fitted using the Morrison quadradic equation for tight-binding inhibitors (eq 1) in GraphPad Prism.²⁵ The K_M for 5′-dAdo is 10.9 μM, and [E], [I], and [S] are the enzyme, inhibitor, and substrate concentrations, respectively.⁴

$$v=v_{\mathrm{o}}\left\{1-\left[[{\mathrm{E}}_{\mathrm{T}}]+[\mathrm{I}]+K_{\mathrm{i}}\left(1+\frac{[\mathrm{S}]}{K_{\mathrm{M}}}\right)\right]\right\}-\frac{\sqrt{[{\mathrm{E}}_{\mathrm{T}}]+[\mathrm{I}]+K_{\mathrm{i}}{\left(1+\frac{[S]}{K_{\mathrm{M}}}\right)}^2-4[{\mathrm{E}}_{\mathrm{T}}][\mathrm{I}]}}{2[{\mathrm{E}}_{\mathrm{T}}]}$$ (1)

Antibacterial Testing

M. smegmatis strain mc²155 and M. tuberculosis strain mc²6230 (nonpathogenic; pantothenate auxotroph) were cultured in Middlebrook 7H9 liquid medium (Difco) to an OD₆₀₀ of 0.6 and plated on Middlebrook 7H10 agar plates as described previously.^26,27 Bacterial growth inhibition was assessed by a drug-diffusion method. Test compounds were added to the center of discs, placed on plates inoculated with M. smegmatis and M. tuberculosis, and allowed to grow for 96 h at 37 °C.

Antibacterial Activity Assay with H. pylori

H. pylori was grown under microaerophilic conditions (5% O₂, 10% CO₂, and 85% N₂) at 37 °C in brain heart infusion medium (Oxoid) with 10% fetal bovine serum (Gibco) and DENT supplement (Oxoid). HT-DADMe-ImmA and tetracycline (Sigma-Aldrich) stocks were made in culture medium, and minimum inhibitory concentration (MIC) values were determined by adding 50 μL of varied drug concentrations in a 96-well plate format at a 2× concentration in culture medium. Cultured H. pylori (50 μL) was added to the medium with the inhibitor. After incubation for 72 h under microaerophilic conditions at 37 °C, the OD at 600 nm was determined on a SpectraMax plate reader and MIC₅₀ values were obtained using a nonlinear regression curve fit using GraphPad Prism.

Homology Modeling

An in silico model of Rv0091 for structure prediction was performed with I-TASSER, an online homology platform for automated protein structure prediction without any additional templates.²⁸ The structural comparisons of the Rv0091 model to MTAN structures that were closely homologous were performed with the secondary structure matching (SSM) algorithm in COOT.²⁹ The superposition of the binding site of Rv0091 was done in comparison to H. pylori MTAN [HpMTAN, Protein Data Bank (PDB) entry 4FFS] and E. coli MTAN (EcMTAN, PDB entry 4WKC).

The Rv0091 model was analyzed by molecular docking with AutoDock Vina for binding with the transition state analogues BT-DADMe-ImmA (13) and HT-DADMe-ImmA (14).³⁰ Coordinates for BT-DADMe-ImmA were obtained from BT-DADMe-ImmA bound to HpMTAN (PDB entry 4FFS). Coordinates for HT-DADMe-ImmA (14) were generated and optimized using the eLBOW function of the PHENIX system for macromolecular structures.³¹ Molecular docking was performed using a grid box of the whole protein. The potential binding site for BT-DADMe-ImmA and HT-DADMe-ImmA was identified by applying the binding mode of the inhibitors to the Rv0091 model. A second round of docking was done with the smaller grid (14 Å × 14 Å × 14 Å), centered on the binding site. The search completeness was evaluated by the default value E = 8 during docking with AutoDock Vina. The binding site analysis and comparisons were done with the published structures of HpMTAN (4FFS) and EcMTAN (4WKC).^17,32 The figures were made by using PyMOL (The PyMOL Molecular Graphics System, version 1.3, Schrödinger, LLC).

RESULTS AND DISCUSSION

DADMe-Immucillin Mimics of the 5′-Substituent of Rv0091 Substrates

The transition state structure of M. tuberculosis Rv0091 revealed a ribocation-like ribose ring with low C1′–N9 bond order, protonation of the adenine leaving group, and weak participation of the water nucleophile (Scheme 2).⁴ The 5′-alkylthio-DADMe-Immucillin inhibitors resemble the Rv0091 transition state structure while incorporating features of the MTA substrate. MT-DADMe-ImmA (11) incorporates the 5′-methylthio group of MTA and exhibits an inhibition constant of 1.5 ± 0.4 nM (Figure 2 and Table 1). Increased hydrophobicity induced by modifying the 5′-substituent group to a 5′-hexylthio (14) improved the dissociation constant to 87 ± 12 pM.⁴

Scheme 2. Reaction and Transition State Catalyzed by Rv0091 To Give Adenine and 5-Methylthioribose (MTR)^a.

^aThe reaction goes through a transition state characterized by a ribocation character, protonation of the adenine leaving group, weak but significant participation of the nucleophilic water, and advanced loss of the C1′–N9 bond.⁴

Figure 2.

DADMe-Immucillin and Immucillin-based inhibitors of Rv0091. 4, 7, 9–15, 19, and 20 are enantiopure compounds, and 5, 6, 8, and 16–18 are racemic mixtures of 3-hydroxy-4-alkylpyrrolidine. Abbreviated names for some of these compounds are provided in the text.

Table 1.

Inhibition of Rv0091 with DADMe-Immucillin Analogues

compound	Ki (nM)a	compound	Ki (nM)a
4	0.64 ± 0.01	13	1.3 ± 0.14
5	1.5 ± 0.1	14	0.087 ± 0.0124
6	18.9 ± 1.1	15	8.2 ± 0.3
7	0.74 ± 0.05	16	6.7 ± 0.3
8	1.64 ± 0.20	17	2.8 ± 0.2
9	>300	18	13.5 ± 1.2
10	5.0 ± 1.1	19	85 ± 114
11	1.5 ± 0.44	20	NIb
12	13.9 ± 2.4

^aDissociation constants are obtained from the xanthine oxidase coupled assay at 25 °C. The K_i value was calculated by using eq 1 for analysis of tight-binding inhibitors.²⁵

^bNot inhibited up to 400 μM.

Enzyme catalytic theory posits that mimics of the transition state for the most proficient substrate will provide the tightest binding transition state analogue.⁸ Values of k_cat/K_M for 5′-dAdo, MTA, and S-adenosylhomocysteine (SAH) with Rv0091 are reported to be 4.4 × 10⁴, 0.6 × 10⁴, and 0.006 × 10⁴ M⁻¹ s⁻¹, respectively. Therefore, transition state analogues of 5′-dAdo would be anticipated to have the highest affinity. The transition state structures of MTANs from E. coli and Streptococcus pneumoniae showed MTA hydrolysis to be a late dissociative process^14,15 and that of Neisseria meningitides to be an early S_N1 mechanism.³³ Analogues designed to mimic the early dissociative transition state are the Immucillins, characterized by a cationic N4′-imino group and a single bond between the ribocation mimic and the 9-deazapurine mimic of the leaving group. An example is MT-ImmA [19 (Figure 2)]. The DADMe-Immucillins incorporate a methylene bridge between the ribocation mimic and the adenine leaving group analogue to more accurately mimic the transition state separation of these two groups. Their geometry resembles the fully dissociated C1′–N9 bond distance of 3.0 Å at the transition state.^13,16 The Rv0091 transition state has a C1′–N9 distance of 2.45 Å and a weakly bonded nucleophilic water, making it intermediate between the early and late N-ribosyltransferase transition states.⁴ The geometric preference for early or late transition state mimics was tested with the MTA transition state mimics, MT-ImmA (19) and MT-DADMe-ImmA (11). The K_i values were 85 nM for 19 and 1.5 nM for 11, demonstrating the DADMe-Immucillin analogues more closely mimic the Rv0091 transition state, consistent with the transition state analysis.⁴ Here we further explore the DADMe-ImmA analogues but extend their structural features to 5′-dAdo, the preferred substrate.

The catalytic activity of Rv0091 with 5′-dAdo as a substrate is relatively inefficient. Like many enzymes from M. tuberculosis, it displays a relatively low catalytic efficiency (k_cat/K_M) of 4.4 × 10⁴ M⁻¹ s⁻¹; however, this is 7-fold better than for MTA and 730-fold better than for SAH. 5′-Alkylthio-DADMe-Immucil-lin-A inhibitors mimic the MTA substrate at the transition state and gave K_i constants in the picomolar to low nanomolar range.⁴ Transition state analogues of the 5′-dAdo substrate were obtained by the synthesis of 5′-deoxy-DADMe-ImmA (4) and its analogues (Scheme 3). The most similar analogue of the 5′-dAdo transition state (4) was a potent inhibitor with a dissociation constant of 0.64 nM (Figure 1).

Scheme 3. Synthesis of 5′-Deoxy-DADMe-Immucillin-A (4)^a.

^aReagents: (a) (i) MsCl, 2,6-dimethylpyridine, acetone, room temperature; (ii) NaBr, 57% for two steps; (b) (i) Pd(OH)₂, H₂ (g), EtOH, room temperature; (ii) cHCl, room temperature, 97% yield for two steps; (c) 9-deazaadenine, 37% aqueous formaldehyde, H₂O, EtOH, room temperature, 53% yield.

Figure 1.

Inhibition of Rv0091 by 5′-deoxy-DADMe-Immucillin-A (4). The reaction rate and inhibition were monitored at 305 nm for the conversion of 5′-dAdo to 2,8-dihydroxyadenine in a reaction coupled to xanthine oxidase as described in Materials and Methods. The K_i value was calculated from the fit to the Morrison equation for tight-binding inhibitors.²⁵ The excess substrate (1.0 mM) relative to its K_m value of 10.9 μM alters the apparent K_i by a factor of 93 for this competitive inhibitor. The true K_i of 0.64 nM is shown.

Crystal structures of Salmonella enterica³⁴ and H. pylori¹⁷ MTANs contain hydrophobic tunnels from the 5′-position of bound nucleosides, reaching to the solvent. This structure can accommodate a variety of hydrophobic 5′-substituents. We increased the hydrophobicity at the 5′-substituent position; as previously described, 5′-thio-DADMe-ImmA analogues (11–15) showed favorable binding properties for this change. This is also the case for the 5′-thio-DADMe-ImmA analogues with Rv0091 MTAN. For example, HT-DADMe-ImmA (14) bound 17 times tighter than MT-DADMe-ImmA (11) did (Table 1).⁴ However, the hydrophobicity of the 5′-alkyl substituent did not result in improved inhibitor potency for Rv0091 (Figure 2 and Table 1). The more hydrophobic 5′-substituent compounds (5, 7, and 8) had inhibition constants slightly weaker than those of the shorter 5′-substituted 4. Compound 9 containing the longest, but hydrophilic, substituent (PEG) was the weakest inhibitor, with a K_i of >300 nM. In comparison, 15, which contains a 5′-thio PEG substituent, had a better potency of 8.2 ± 0.3 nM. These results suggest that the 5′-binding site of Rv0091 is unlike those of other MTANs in that it prefers 5′-alkyl substituents over 5′-thio substituents if the 5′-substituent is short. When the substituent is longer, the 5′-thio substituents are preferred.

Preference for (3R,4S) Stereochemistry in the DADMe-Immucillin Inhibitors

The stereochemistry of the 3-hydroxy-4-alkylpyrrolidine moiety with respect to inhibitor potency was explored. trans-Racemate compound 5 was compared to cis-racemate 6 and was shown to be 13-fold more potent, demonstrating that the trans stereochemistry is preferred at the 3-hydroxy-4-alkylpyrrolidine position. To evaluate the preference of (3R,4S) over (3S,4R) stereo-chemistry, the enantiopure (3R,4S)-MT-DADMe-ImmA (11) was compared to the enantiomer (3S,4R)-MT-DADMe-ImmA (20). 11 inhibits Rv0091 with a good potency of 1.5 nM, while compound 20 had no activity at concentrations of ≤400 μM. This observation is consistent with previous findings where (3R,4S) inhibitors displayed potencies better than those of their enantiomeric counterparts toward other MTAN nucleosidases, including that from E. coli.³⁴

The affinities of these transition state analogues for Rv0091 (87 pM for the best compound) and other bacterial MTANs, such as E. coli MTAN (femtomolar range for best compounds),^20,35 establish a reduced affinity for Rv0091. This difference can be attributed to relative catalytic efficiencies of the E. coli and M. tuberculosis enzymes. Wolfenden and Snider^9,10 postulated that the binding of transition state analogues is associated with enzyme catalytic efficiency. E. coli MTAN has a high catalytic efficiency of 4.7 × 10⁷ M⁻¹ s⁻¹ for MTA, approximately 1000-fold more efficient than Rv0091 for 5′-dAdo.⁴ The E. coli MTAN is also more efficient for 5′-dAdo than Rv0091 is even though it is not a preferred substrate. Consequently, transition state analogues mimicking both MTA and the 5′-dAdo substrate are known to bind more tightly for the E. coli enzyme by several orders of magnitude. This binding is also consistent with previous observations for the inhibitor profiles of the slow S. pneumonia MTAN (SpMTAN).³⁶

Molecular Modeling and Putative Binding Site Residue Identification

The crystal structure of Rv0091 from M. tuberculosis is not available. The binding interactions of active site residues for Rv0091 were probed by docking the transition state analogue inhibitors BT-DADMe-ImmA and HT-DADMe-ImmA to explain the difference in 5′-substituent interactions from the specificity known for H. pylori and E. coli MTANs. The in silico model of Rv0091 was generated by the I-TESSER server (Figure S1). The best structural model of Rv0091 was obtained with a C score of 0.43 (the ideal C score is between −5 and 2; a score of 0.43 is an excellent match) and a TM score of 0.77 ± 0.10. The model has a structural fold similar to those seen in other MTANs with the exception of an extended loop (from Leu148 to Pro162) unique to Rv0091 (Figures S1 and S2). The sequence of Rv0091 is 29 and 32% identical with those of H. pylori and E. coli MTANs, respectively (Figure S2). Comparison of the Rv0091 model with H. pylori and E. coli MTANs gave root-mean-square deviations between 0.96 and 1.03 Å without and with the Rv0091 external loop, respectively.

Crystal structures of H. pylori and E. coli MTAN in complex with BT-DADMe-ImmA have been published.^24,32 Autodoc-Vina was used for the in silico docking of BT-DADMe-ImmA and HT-DADMe-ImmA to the binding site of Rv0091. The best fit of the interactions between the inhibitor and Rv0091 was selected from the top-ranked cluster by binding affinities. Active site superimposition of BT-DADMe-ImmA bound to HpMTAN and EcMTAN with Rv0091 modeled with both BT-DADMe-ImmA and HT-DADMe-ImmA shows the pyrrolidine and deazaadenine inhibitor groups bind in the same orientation in all the MTANs (Figure 3 and Figure S3). The catalytic site geometries of 5′-alkylthio inhibitor groups differ considerably for Rv0091. Active site tunnel residues Ile52 (in HpMTAN) and Ile50 (at the same position in EcMTAN) have nonpolar interactions with the 5′-alkylthio group of BT-DADMe-ImmA. This interaction is disrupted by the larger Met52 residue in this position in Rv0091 MTAN. The side chain of the Met52 creates a steric clash with the 5′-alkylthio group of the inhibitors to displace it relative to its geometry in EcMTAN and HpMTAN (Figure 3).

Figure 3.

Stereoview superimposition of the active sites of H. pylori and E. coli MTAN in complex with BT-DADMe-ImmA compared with Rv0091. The interactions of binding site residues of Rv0091 modeled with BT-DADMe-ImmA (yellow, panel A) and with HT-DADMe-ImmA (green, panel B) are highlighted. The BT-DADMe-ImmA compounds of H. pylori and E. coli are colored blue and pink, respectively.

Inhibitors modeled into Rv0091 show the 3′-hydroxyl group to be hydrogen-bonded to Glu197. N1 and N3 of the 9-deazaadenine ring have hydrogen bonding interactions with the backbone nitrogen of Leu176 and Glu195, respectively. The N6 amino and N7 of the inhibitors form hydrogen bond interactions with both the peptide bond NH group of Leu176 and the side chain hydroxyl group of Asp220 (Figure 3 and Figure S3). These groups are conserved among Rv0091, HpMTAN, and EcMTAN, supporting the major divergence of Rv0091 being related to the hydrophobic region in the catalytic site that accommodates the 5′-alkylthio groups. The dissociation constant for BT-DADMe-ImmA is 1.3 nM, which is 15 times weaker binding than that of HT-DADMe-ImmA (K_d = 87 pM). HT-DADMe-ImmA has two additional carbons at the 5′-alkylthio group that make additional hydrophobic contacts with the enzyme and contribute to the higher affinity of HT-DADMe-ImmA. From the model, the butylthio group of BT-DADMe-ImmA predicts van der Waals interactions with eight amino acids, Ile11, Met52, Met196, Pro119, Tyr113, Val104, Tyr175, and Phe230. These along with two additional contacts with Glu108 and Val105 are predicted for the hexylthio group of HT-DADMe-ImmA, consistent with its tighter binding.

Antibiotic Activity against Mycobacterium Species and H. pylori

The antibacterial activity of DADMe-Immucillin compounds was investigated in a disc-diffusion assay against M. tuberculosis mc²6230 and M. smegmatis mc²155 strains.³⁷ HT-DADMe-ImmA (14) is the most potent inhibitor toward Rv0091 and displayed a small zone of clearance at 1000 μg/mL in both Mycobacterium species (Figure 4A,B). Kanamycin as an antibiotic control displayed large zones of clearance at 100 μg/ mL. Other DADMe-Immucillin compounds showed results similar to those of HT-DADMe-ImmA (14), indicating that the function of Rv0091 is not essential for the growth of Mycobacteria. This finding is in agreement with previous gene deletion experiments in M. tuberculosis, where Rv0091 was found to be nonessential for growth or infection.^38,39 Unlike Mycobacteria, MTAN has an essential function in menaquinone synthesis in H. pylori.¹⁷ We tested HT-DADMe-ImmA (14) in a cultured cell assay using H. pylori with tetracycline as a control. HT-DADMe-ImmA (14) exhibited antibacterial activity against H. pylori with an IC₅₀ of 13.0 ± 1.8 ng/mL (equivalent to 35 ± 5 nM), which is 6 times more potent than tetracycline as an antibiotic (Figure 4C).

Figure 4.

Antibacterial activity. Transition state analogue HT-DADMe-ImmA (14) and an antibiotic control were tested for growth inhibition against (A) M. smegmatis mc²155 and (B) M. tuberculosis mc²6230 in a disc-diffusion assay by growth for 96 h at 37 °C on agar plates. (C) Inhibition of H. pylori (J99) growth at 37 °C in a liquid culture by HT-DADMe-ImmA (14) and tetracycline. Growth was evaluated at OD = 600 nm, and MIC₅₀ values were calculated by nonlinear regression using GraphPad Prism 7.

Genomic Context of Rv0091

MTAN in E. coli has been reported to be involved in SAH and MTA recycling and in quorum sensing pathways.² BT-DADMe-ImmA (13) inhibited production of the quorum sensing signaling molecule auto-inducer (AI-2) but was not toxic to E. coli or Vibrio cholerae growth, providing support for its role in quorum sensing.¹ Comparison of the gene maps for Rv0091 (NCBI gene 886953), EcMTAN (NCBI gene 948542), and HpMTAN (NCBI gene 900251) indicates no organized operon function. In E. coli, the open reading frame for MTAN is adjacent to btuF, expressing a periplasmic B₁₂ binding protein, and dgt, known to express a dGTP triphosphohydrolase (Figure S4). In H. pylori, the adjacent Hp0090 is a malonyl-CoA:acyl carrier protein transacylase. Near the Rv0091 locus in M. tuberculosis, a nearby open reading frame, the sequence of amino acids 32–146 (of 197) of Rv0089 is 41% identical with that of the E. coli BioC protein, a SAM-dependent malonyl-acyl carrier protein methyltransferase, an early step in biotin synthesis.⁴⁰ In E. coli, the BioC protein is expressed from the bioC gene within the bioABFCD operon.⁴¹ On the basis of the local gene organization near Rv0091, no insight is provided to its biological function, supporting a role in removing the 5′-dAdo product from the 21 potential radical SAM proteins in M. tuberculosis.

CONCLUSIONS

Transition state analogues with geometries and charges thatresemble those of the transition state for hydrolysis of 5′-dAdo by Rv0091 were synthesized and found to bind with picomolar to nanomolar dissociation constants. (3R,4S)-3-Hydroxy-4-alkylpyrrolidine stereochemistry is preferred for these compounds, consistent with findings for related bacterial MTANs.³⁴ Comparison of 5′-alkylthio-DADMe-ImmA inhibitors that mimic the MTA substrate provided insight into the relative specificity for 5′-dAdo and 5′-methylthio adenosine substituents for Rv0091. Short chain 5′-substituents are preferred for the 5′-deoxyalkyl inhibitors, while longer hydrophobic chains displayed tighter binding for the 5′-alkylthio inhibitors. In both cases, substitution of polar atoms into the 5′-substituent, i.e., CH₂ to O, reduces the potency of the inhibitors. In silico docking of BT-DADMe-ImmA (13) and HT-DADMe-ImmA (14) with a model of Rv0091 showed similar binding to E. coli and H. pylori MTAN of the pyrrolidine and deazaadenine moieties, but they differed at the 5′-alkyl or 5′-alkylthio groups. This difference is due to a methionine steric clash near the 5′- substituent in the catalytic site of Rv0091. Isoleucine is present at this position in the MTANs from E. coli or H. pylori. Furthermore, the weak antibacterial activity of these inhibitors against Mycobacteria indicates that Rv0091 functions in a nonessential pathway for growth.

ABBREVIATIONS

MTANs

5′-methylthioadenosine/S-adenosylhomocysteine nucleosidases

MTA

5′-methylthioadenosine

SAH

S-adenosyl-homocysteine

5′-dAdo

5′-deoxyadenosine

PEG

polyethylene glycol

SAM

S-adenosylmethionine

KIE

kinetic isotope effect