Mechanism of MenE Inhibition by Acyl-Adenylate Analogues and Discovery of Novel Antibacterial Agents

Abstract

MenE is an o-succinylbenzoyl-CoA (OSB-CoA) synthetase in the bacterial menaquinone biosynthesis pathway and is a promising target for the development of novel antibacterial agents. The enzyme catalyzes CoA ligation via an acyl-adenylate intermediate, and we have previously reported tight-binding inhibitors of MenE based on stable acyl-sulfonyladenosine analogues of this intermediate, including OSB-AMS (1) which has an IC₅₀ value of ≤ 25 nM for the Escherichia coli MenE. Herein, we show that OSB-AMS reduces menaquinone levels in S. aureus, consistent with its proposed mechanism of action, despite the observation that the antibacterial activity of OSB-AMS is ~1000-fold lower than the IC₅₀ for enzyme inhibition. To inform the synthesis of MenE inhibitors with improved antibacterial activity, we have undertaken a structure–activity relationship (SAR) study stimulated by the knowledge that OSB-AMS can adopt two isomeric forms in which the OSB side chain exists either as an open-chain keto acid or a cyclic lactol. These studies revealed that negatively charged analogues of the keto-acid form bind, while neutral analogues do not, consistent with the hypothesis that the negatively-charged keto-acid form of OSB-AMS is the active isomer. X-ray crystallography and site-directed mutagenesis confirm the importance of a conserved arginine for binding the OSB carboxylate. Although most lactol isomers tested were inactive, a novel difluoroindanediol inhibitor (11) with improved antibacterial activity was discovered, providing a pathway toward the development of optimized MenE inhibitors in the future.

Introduction

Novel antibiotics are urgently needed to treat infections caused by drug-resistant strains of pathogens such as multi- and extensively drug-resistant Mycobacterium tuberculosis and methicillin-resistant Staphylococcus aureus (MRSA)^[1–4]. Approximately one-third of the world's population is infected with active or latent M. tuberculosis and the mortality rate is growing rapidly in both developed and developing countries due to drug-resistant strains^[5–7]. Furthermore, community-acquired MRSA is the cause of more than 7 million hospitalizations from skin and soft tissue infections annually in the US alone^{[8, 9]}. Treatments of these infections are estimated to cost $10–20 billion annually in the US^[10].

Lipoquinones such as ubiquinone (CoQ), phylloquinone, and menaquinone are biological quinones that play pivotal roles in all living organisms^[11–16]. CoQ is the primary redox-active electron carrier in the electron transport chain in humans and most aerobically growing Gram-negative bacteria^{[17, 18]}, while phylloquinone plays this role in photosynthetic organisms^[19]. Menaquinone is the sole electron transport chain quinone in Gram-positive bacteria, Mycobacteria spp., and all anaerobically growing bacteria regardless of Gram-variability^[20–22]. Although menaquinone also serves as a co-factor to many vitamin K-dependent enzymes in humans such as γ-glutamyl carboxylases^{[18, 23]}, de novo biosynthesis of menaquinone only occurs in certain prokaryotes while humans obtain menaquinone almost exclusively through diet (ingestion of phylloquinone^[24]) or from gut flora. Thus, enzymes in the menaquinone biosynthetic pathway are attractive therapeutic targets for treating diseases caused by menaquinone-producing pathogens.

Due to the unique and essential role of menaquinone in bacterial growth, virulence, and survival, inhibitors of menaquinone biosynthesis have been developed for use as chemical tools to interrogate pathway essentiality and as a prelude to novel therapeutic discovery. Organisms such as M. tuberculosis and S. aureus synthesize menaquinone from chorismate via a pathway that includes at least nine distinct enzymes (Scheme 1)^{[16, 25–27]}. Inhibitors of MenD^{[28, 29]}, MenC^[30], MenE^[31], MenB^{[32, 33]}, and MenA^[34–37] have been reported. These pharmacological studies validate genetic approaches that have demonstrated the essentiality of men genes in bacteria including M. tuberculosis^{[38, 39]} and Bacillus subtilis^{[25, 40]}, and suggest that inhibitors of menaquinone biosynthesis may lead to therapeutics that target both replicating as well as latent populations of M. tuberculosis^{[34, 35, 41, 42]}.

Scheme 1. Classical de novo menaquinone biosynthesis pathway.

This pathway consists of at least nine enzymes that catalyze the formation of menaquinone from chorismate^{[12, 13, 16]}. The fifth enzyme, MenE, is an acyl-CoA ligase which ligates CoA to OSB via an OSB-AMP intermediate. (OSB = o-succinylbenzoate; AMP = adenosine monophosphate; DHNA-CoA = 1,4-dihydroxy-2-napthoyl-CoA)

Our own efforts have focused on MenE, the o-succinylbenzoyl-CoA synthetase that belongs to the ANL (acyl-CoA synthetase, nonribosomal peptide synthetase adenylation domain, luciferase)^[43] family and catalyzes the ligation of CoA to o-succinylbenzoate (OSB) via an ordered Bi Uni Uni Bi Ping-Pong mechanism in the presence of a divalent cation and ATP (Scheme 1)^{[31, 44]}. In the first half-reaction, MenE catalyzes the reaction of OSB and ATP to form a tightly-bound OSB-AMP intermediate coupled with release of inorganic pyrophosphate (PP_i). In the second half-reaction, CoA attacks the mixed anhydride intermediate to form OSB-CoA with release of AMP^{[31, 44–47]}. Genetic knockout and genome-wide transposon bioinformatic experiments have demonstrated that MenE is essential for the survival, growth, and virulence of Mycobacteria spp., Bacillus spp., Lactococcus lactis, and other pathogenic and nonpathogenic bacteria^{[38, 40, 48–50]}. MenE homologs are also proposed to be essential in photosynthetic organisms such as Synechocystis spp. and Arabidopsis^{[51, 52]}.

We have previously described the development and characterization of MenE inhibitors based on acyl-AMS (5'-O-(N-acylsulfamoyl)adenosine) analogues known to inhibit adenylate-forming enzymes^{[44, 47]}. We reported that OSB-AMS (1) is a tight-binding inhibitor of the S. aureus, M. tuberculosis, and E. coli MenE enzymes, with low-nM K_i^app values with respect to OSB^[44]. We noted that a free carboxylate on the OSB side chain is required for potent MenE inhibition, and reported that OSB-AMS exists in equilibrium between an open chain keto-acid and a closed lactol form (Figure 1). To guide future modifications of the scaffold, we have now evaluated which form of the inhibitor is the active pharmacophore using analogues of OSB-AMS that mimic either the open keto-acid or closed lactol isomers in enzyme inhibition assays, binding assays, and X-ray crystallographic studies. We have also analyzed the mechanism of inhibition of these OSB-AMS analogues, and evaluated their antimicrobial activities against pathogenic bacteria.

Figure 1. OSB-AMS (1) analogs.

Open keto-acid and closed lactol isomers of OSB-AMS (1) are in equilibrium at physiological pH. We synthesized analogues (3–11) of both conformations to test the active pharmacophore. Compounds 3–8 are keto-acid analogues which favor the open-ring conformation and compounds 9–11 are lactol analogues which favor the close-ring structure. Compound 2 is a meta-carboxylate analogue to test the importance of the ortho position of the carboxylate. Synthesis of these analogues can be found in the supplemental section.

Materials and Methods

Sequence Alignment

To determine the sequence homology between the MenE enzymes from different organisms, the amino acid sequences of MenE from E. coli (K-12, ecMenE), S. aureus (RN4220, saMenE), and M. tuberculosis (Erdman, mtMenE) were aligned using INRA Multalin^[53] and Alignment Annotator^[54]. Overall, ecMenE shares 29% and 30% sequence identity with saMenE and mtMenE respectively, while mtMenE shares 26% sequence identity with saMenE. The same procedures were performed to align the acyl-CoA synthetase superfamily.

Site-directed mutagenesis

High fidelity Phusion polymerase (NEB) was used for standard PCR amplification and mutagenesis. Based on the nucleotide sequence of ecMenE (from K-12), the following primers were designed, synthesized, and used for the mutation of ecMenE:

Mutation	Forward 5'–3'	Reverse 5'–3'
R195K	GGAATTATGTGGAAGTGGTTATACGC	GCGTTAAACCACTTCCACATAATTCC
R195Q	GGAATTATGTGGCAGTGGTTATACGC	GCGTATAACCACTGCCACATAATTC

Circular Dichroism

CD experiments were performed using a Chirascan CD spectrometer. MenE was diluted to 10 μM in 350 μL 20 mM NaHPO₄ buffer pH 7.4 containing 150 mM NaCl and 1 mM MgCl₂. Far-UV wavelength (190–260 nm) spectra were collected in triplicate using a 1 mm cuvette in 1 nm increments. Data were collected using Chirascan software and reprocessed using MATLAB. The mean residue ellipticity (deg cm² dmol⁻¹ residue⁻¹) was calculated using the following equation:

$$\Theta =\text{Ellipticity}∕(\text{path lengh}\cdot [\text{protein}]\cdot \mathrm{n})$$

where the path length is in mm, protein concentration is in M, and n is the number of peptide bonds.

Enzyme assays

Enzyme inhibition studies were performed in 20 mM NaHPO₄ buffer pH 7.4 containing 150 mM NaCl and 1 mM MgCl₂ using a MenE-MenB coupled assay in which MenE is rate limiting^{[44, 47]}. IC₅₀ values were determined in reaction mixtures containing OSB (60 μM), ATP (240 μM), CoA (240 μM), mtMenB (2.5 μM) and varying inhibitor concentrations (5–250 μM). Reactions were initiated by addition of ecMenE (25 nM) and the production of DHNA-CoA was monitored at 392 nm (ε₃₉₂ 4000 M⁻¹ cm⁻¹). To determine k_cat and K_M, OSB was varied from 10–250 μM, while K_i^app was determined by varying both inhibitor and OSB from 5–250 μM and 60–350 μM, respectively.

X-ray crystallography

Co-crystallization trials used the ecMenE R195K mutant (~15 mg/mL) and ~200 μM OSB-AMS and initial crystallization conditions were obtained by sitting-drop vapor diffusion with Hampton Research HR2-110 buffer A6. The optimized growth condition was found to be 0.2 M MgCl₂ (6·H₂O), 0.1 M Tris HCl pH 8.5, 18–22% w/v PEG 4,000, with hanging-drop vapor diffusion at 298.15 K. Diffraction data were collected at 100 K at the X6A beamline at Brookhaven National Laboratories using an ADSC Q270 CCD detector. Data were indexed, integrated, and scaled using HKL2000^[55], and the structure was solved by molecular replacement using MolRep^[56]. The search model consisted of the large N-terminal domain of saMenE (PDB: 3IPL) and the small C-terminal domain of 4-chlorobenzoate:CoA synthetase (PDB: 1T5D). The model was refined through successive rounds of manual model building using COOT^[57] and restrained refinement using REFMAC5^[58]. The ligand was added directly to the difference Fourier map only after the refinement converged. Ligand restraints were generated using the PRODRG server^[59]. The data collection and refinement statistics are provided in Table S1 and Table S2, respectively.

Isothermal titration calorimetry

Isothermal titration calorimetry was performed at 22 °C using a MicroCal VP-ITC instrument. A 1 mM solution of OSB-AMS (1) (dissolved in 20 mM NaHPO₄ buffer pH 7.4 containing 150 mM NaCl and 1 mM MgCl₂ at 22 °C) was titrated in 4 μL increments into the 1.8 mL cell containing a 25 μM solution of ecMenE (wild-type, R195K, or R195Q) in the same buffer. The data were fit to a single binding site model with the Origin software package.

Fluorescence binding assays

ITC experiments with difluoroindanediol 11 did not show a measurable change in enthalpy, and thus ITC was unable to quantify the binding of this compound to the enzyme. Instead, to determine the K_d for 11, we used a direct binding assay in which the change in the intrinsic tryptophan fluorescence of ecMenE was monitored (see Figure S1 for the binding isotherm). A solution of 50 μM 11 was titrated into 300 nM ecMenE in 20 mM NaHPO₄ buffer pH 7.4 containing 150 mM NaCl and 1 mM MgCl₂ at 25 °C. The solution was stirred continuously and fluorescence measurements were made with a Quanta Master fluorimeter using excitation and emission wavelengths of 280 and 332 nm, respectively. Slit widths were optimized to 4 and 2 nm for excitation and emission, respectively. Data were corrected for the inner filter effect and then fit to the following equation using MATLAB:

$$\frac{\Delta F_i}{\Delta F_{\mathit{max}}}=\frac{\left(\right[E]+[I]+K_d)-\sqrt{{\left(\right[E]+[I]+K_d}^2)-4[E\left]\right[I]})}{2\left[E\right]}$$

Antibacterial activity

Minimum inhibitory concentrations (MIC) were determined using visual growth inspection of cells grown in transparent 96-well plates. E. coli, B. subtilis (ATCC 6051), S. aureus (ATCC BAA-1762), and M. tuberculosis (H37Rv) were grown to mid-log phase (OD₆₀₀ of 0.6–0.8) in LB, Miller Hinton, synthetic broth, or 7H9 with 0.5% Glycerol, 0.05% Tween, and 10%OADC media at 37 °C in an orbital shaker. A final inoculum concentration of 1–2 × 10⁶ cells per well was treated with inhibitor at final concentrations ranging from 0.5 – 500 μg/mL. The MIC was defined as the minimum concentration at which a well showed no obvious growth by visual inspection (MIC-99). Growth rescue studies were performed by supplementing minimal media (synthetic broth) with 10 μM menaquinone-4 (MK4) and following the same procedure.

Cytotoxicity

To obtain insight into the potential cytotoxicity of our MenE inhibitors, we evaluated the in vitro cytotoxicity using Vero monkey kidney cells. Briefly, 10⁵ cells/well were aliquoted into 96-well culture plates in serum rich media. The cells were incubated for 24–36 h at 37 °C in 5% CO₂. The media was then aspirated and replaced with 200 μL of serum-free fresh media. Cells were incubated for 5 h at 37 °C in 5% CO₂, after which compounds dissolved in serum-free cell media were added, giving a concentration range of 0.97 – 250 μg/mL. The cells were incubated for 24 – 36 h at 37 °C in 5% CO₂. Cell death was assessed by incubating 20 μL of cell suspension from each well with 20 μL Trypan Blue for 5 min. The ratio of viable/dead cells was determined using a hemocytometer where stained cells were scored as dead and non-stained cells were scored as viable. The cytotoxic concentration was defined as the minimum inhibitor concentration which gave ~90% cell death.

Quantification of menaquinone levels in S. aureus

The effect of MenE inhibitors on menaquinone levels in S. aureus was determined as follows. Cultures of S. aureus ATCC BAA-1762 (5 mL synthetic broth medium with 10% glucose) were incubated overnight in a 37 °C shaker in the presence or absence of OSB-AMS below its MIC (15.6 μg/mL). The Blight and Dyer (1959) lipid extraction protocol was used to isolate the menaquinone-containing fraction from the cells^[60]. Briefly, 0.75 mL of 1:2 (v/v) CHCl₃:MeOH was added to 0.2 mL of culture. The mixture was vortexed thoroughly, and 0.25 mL of CHCl₃ was added followed by further vortexing after which 0.25 mL of H₂O was added. The mixture was then vortexed and centrifuged at 500 × g for 5 minutes at rt. The bottom phase was recovered, transferred to a glass vial and 200 μL was analyzed by APCI LC-MS/MS in positive ion mode using a Thermo TSQ Quantum Access (Thermo-Fisher) Triple Quadrupole Mass Spectrometer. Menaquinone levels were quantified using standard curves based on menaquinone-4 (MK4) or menaquinone-9 (MK9) (Sigma). Samples were introduced into the mass spectrometer by flow injection at 100 μL/min with 2:1 MeOH/CHCl₃ as the solvent. Multiple Reaction Monitoring (MRM) was performed at 30 eV and MK4, MK5 and MK6 were quantified using the standard curve for MK4 whereas MK7, MK8, and MK9 were quantified using MK9. Ubiquinone-4 (CoQ4) was used as an internal standard and experiments were performed in triplicate.

Results and Discussion

Synthesis of OSB-AMS Analogues

OSB-AMS has a carboxylate group ortho to the succinyl substituent that can undergo cyclization through attack of the carboxylate on the succinyl ketone to generate a lactol isomer (Figure 1). To explore the importance of this isomeric equilibrium for enzyme inhibition and to investigate SAR toward the development of optimized MenE inhibitors, we synthesized one analogue in which the carboxylate and succinyl substituents were meta to one another (MSB-AMS, 2) as well as a series of keto-acid (3–8) and lactol analogues (9–11) of OSB-AMS (Figure 1, Scheme S1).

Synthesis of MSB-AMS proceeded by a Suzuki coupling of previously reported vinyl bromide 12^{[44, 47, 61]} and aryl boronic acid 13, followed by ozonolysis and deprotection of the t-butyl ester to give the acyl side chain 15. The acyl side chain was then coupled to protected AMS scaffold 16, prepared as previously described^{[44, 47, 61]}, followed by global deprotection to give MSB-AMS (2). OSB-AMS analogues 3–9 were synthesized analogously using this general method of initial synthesis of the acyl side chain, subsequent coupling to the protected AMS scaffold 16, and global deprotection.

Alternative synthetic strategies (Scheme S1f, g) were necessary to access the lactam (10) and difluoroindanediol (11) analogues due to reactivity associated with their individual structures. Briefly, the lactam analogue 10 was synthesized from TBS-protected MeOSB-AMS analogue 19 (Scheme S1f), accessed via previously reported synthetic steps^{[44, 47, 61]}. This intermediate was treated with anhydrous ammonia followed by global deprotection to yield lactam 10. Difluoroindanediol 11 was synthesized by alkylation of difluoroindandione 20 followed by deprotection to give the acyl side chain 22. This acyl side chain was then coupled to the protected AMS scaffold 16 to give intermediate 23 before subsequent reduction, and global deprotection to afford difluoroindanediol 11.

Inhibition of MenE

We next determined IC₅₀ values for inhibition of E. coli MenE (ecMenE) by compounds 1–11 using a MenE-MenB coupled assay (Table 1). MSB-AMS (2), as well as the oxazole (5), nitro (6), boronic acid (7), and trifluoroethanol (8) analogues of the keto acid form, and the lactone (9) and lactam (10) analogues of the lactol form, showed no inhibition of ecMenE up to a concentration of 100 μM. In contrast, the tetrazole analogue 3 inhibited ecMenE with an IC₅₀ of 2.2 ± 0.4 μM, while the squarate analogue 4 showed more potent inhibition with an IC₅₀ of 0.17 ± 0.05 μM. Interestingly, the difluoroindanediol analogue 11 was also an effective inhibitor with an IC₅₀ of 1.5 ± 0.1 μM and a K_d of 120 ± 23 nM was determined using a fluorescence perturbation binding assay (Figure S1). These results indicate that analogues of both the open and closed isomers are able to inhibit MenE. Crucially, analogues 3 and 4 have acidic protons on the ortho substituent (tetrazole 3 pK_a = 3.4, squaric acid 4 pK_a = 1.3) such that these groups carry a negative charge at neutral pH like the natural OSB substrate. These data suggest that enzyme inhibition requires a negative charge in the inhibitor close to the position normally occupied by the OSB carboxylate group. Difluoroindanediol 11 has a pK_a of 11.5 and may also be deprotonated when bound to the enzyme. Furthermore, we believe that the net neutral charge of the nitro analogue 6 substantially decreases its ability to hydrogen bond or form ionic interactions despite being a respectable carboxylate bioisostere when interacting with a metal cofactor^{[62, 63]}. Finally, we speculate that the boronic acid analogue (7) (pK_a > 13) forms a boronate complex with the ortho-ketone on the succinate chain, raising the pK_a of the boronic acid such that the inhibitor is uncharged at neutral pH.

Table 1.

In vitro enzyme inhibition of OSB-AMS analogues using ecMenE

Inhibitor1		IC50 (μM)2	pKa3
1 (OSB-AMS)		0.025 ± 0.005	4
2 (MSB-AMS)		> 100	4
3 (tetrazole)		2.2 ± 0.4	3.4
4 (squarate)		0.17 ± 0.05	1.3
5 (oxazole)		> 100	NA
6 (nitro)		> 100	NA
7 (boronic acid)		> 100	> 14
8 (trifluoroethanol)		> 100	11.5
9 (lactone)		> 100	NA
10 (lactam)		> 100	NA
11 (difluoroindanediol)		1.5 ± 0.1	11.5

¹MSB-AMS 2 is an isomeric meta- substituted analogue. Analogues 3–8 are mimics of the keto-acid isomer of OSB-AMS while analogues 9–11 are mimics of the lactol isomer.

²IC₅₀ values for inhibition of E. coli MenE (ecMenE). All IC₅₀ measurements were performed in triplicate.

³pK_a of the OSB analogue. NA, not applicable.

Role of R195 in OSB-CoA formation and inhibition of MenE

We previously used computational docking to build a model of OSB-AMS (1) bound to saMenE^[44]. Using this approach, we identified R222 within 4.7 Å of the OSB carboxylate in the active site of saMenE. Sequence alignment reveals that R222 is conserved in other MenE homologs and corresponds to R90 in mtMenE and R195 in ecMenE (Figure 2). To explore the role of this conserved residue, we replaced R195 in ecMenE with Lys or Gln by site-directed mutagenesis. Circular dichroism spectra of these mutants showed no significant alteration in the secondary structure (Figure S2). We then analyzed the change in catalytic efficiency (k_cat/K_M) of the mutant enzymes compared to wild-type ecMenE (Table 2) using the MenE-MenB coupled assay^[64]. These studies revealed that the k_cat/K_M decreased by 15-fold for R195K MenE compared to wild-type MenE, while the R195Q mutant had no detectable activity up to an OSB concentration of 240 μM. Further analysis demonstrated that the effect of the R195K mutation on activity was primarily a result of a 16-fold increase in K_M for OSB while k_cat for product formation was unchanged (Table 2).

Figure 2. Sequence alignment of MenE homologs from pathogenic bacteria.

E. coli (K-12), S. aureus (MRSA, Rosenbach), M. tuberculosis (H37Rv) were aligned using INRA hierarchical clustering^[53] and Alignment Annotator^[54]. Blue arrow indicates the conserved arginine identified in the active site of saMenE (R222) by docking OSB-AMS into the crystal structure of apo saMenE^[44]. The homologous residue in ecMenE is R195 and R90 in mtMenE. Red arrows indicate relevant residues interacting with OSB-AMS seen in the ecMenE crystal structure and conserved in saMenE and mtMenE.

Table 2.

Catalytic parameters and ITC data for the interaction of OSB and OSB-AMS with wt and mutant ecMenE

ecMenE	$K_M^{\mathit{OSB}}$ (μM)1	kcat (min−1)1	kcat/KM (μM−1min−1)1	$K_d^{\mathrm{OSB}-\mathrm{AMS}}$ (nM)2	ΔH (kcal/mol)2	ΔG (kcal/mol)2	ΔΔG (kcal/mol)2
wt	1 ± 0.02	46 ± 0.12	46 ± 0.023	44 ± 11	−2.0 ± 0.1	−10.0
R195K	16 ± 1.4	47 ± 0.33	3 ± 0.2	394 ± 36	−2.5 ± 0.20	−8.8	1.2
R195Q		Not Active		4500 ± 112	−3.1 ± 0.10	−7.3	2.7

¹Kinetic parameters were obtained using the MenE-MenB coupled assay^{[12, 64]}. Measurements were performed in 20 mM NaHPO₄ buffer pH 7.4 containing 150 mM NaCl and 1 mM MgCl₂ at 25°C. k_cat and K_M values are reported for wild-type and R195K MenE. Replacement of R195 with lysine resulted in a 15-fold decrease in k_cat/K_M. No activity could be detected for the R195Q mutant even at an OSB concentration of 240 μM. All measurements were performed in triplicate.

²Binding of OSB-AMS to ecMenE determined by ITC. A 1 mM solution of inhibitor (dissolved in 20 mM NaHPO₄ buffer pH 7.4 containing 150 mM NaCl and 1 mM MgCl₂ at 25°C) was titrated in 4 μL increments into the 1.8 mL cell containing 25 μM solution of ecMenE in the same buffer. The data were fit to a single binding site model. Measurements were made in triplicate. Replacement of R195 with Lys and Gln results in a ~ 10 and 100-fold increase in K_d of the inhibitor for the enzyme, respectively.

To investigate the role of R195 in enzyme inhibition, we evaluated the binding of OSB-AMS to ecMenE R195K and R195Q by isothermal calorimetry and demonstrated that replacement of R195 with Lys or Gln decreased binding affinity of the inhibitor to ecMenE by 9- or 102-fold, respectively (Table 2, Figure S3).

X-ray crystal structure of the ecMenE (R195K)·OSB-AMS complex

To underpin our efforts to develop potent MenE inhibitors and extend the modeling studies with saMenE, we set out to obtain the X-ray structure of MenE in complex with OSB-AMS (1). These efforts were most successful with the R195K mutant of ecMenE resulting in a 2.4 Å resolution structure of R195K ecMenE co-crystallized with OSB-AMS (PDB: 5C5H). The structure was solved by molecular replacement using the structures of saMenE and 4-chlorobenzoate:CoA ligase (CBAL) from Alcaligenes sp. AL3007 (PDB: 3IPL and 1T5D, ~29% sequence identity) as search models. Data collection and refinement statistics are given in Table S1 and Table S2, respectively.

MenE is a member of the adenylate-forming enzyme superfamily in which ATP is used to activate a carboxylate for subsequent attack of a nucleophile. One of the best characterized members of this family is CBAL, which has been extensively studied by Gulick, Dunaway-Mariano, and colleagues^[65–67]. Both MenE and CBAL are comprised of a larger N-terminal domain and a smaller C-terminal domain, and structures of CBAL in complex with an adenylate intermediate as well as CoA thioester product analogue reveal that ligand binding causes the two domains to move relative to each other as the reaction proceeds. Domain alternation reconfigures the active site from a conformation that catalyzes acyl-adenylate formation to one that facilitates CoA binding and thioester formation^[68–70].

In Figure 3, we show the structure of the OSB-AMS:ecMenE·complex overlayed with that of apo saMenE (PDB: 3IPL). These structures differ in the relative orientations of domains 1 and 2. However both structures are representative of the adenylate-bound conformation observed for CBAL (PDB: 3CW8), in which G408 in region A8 (399-GRVDDMIISG-408) is removed from the active site whereas K492 in region A10 (486-PKNALNK-492) is located in the active site^[65–67]. The corresponding residues in ecMenE (saMenE) are G358 (G402) and K437 (K483) (Figure S4), and in Figure 3 it can be seen that G358/G402 are located away from the MenE binding site, whereas K437 is close to the bound OSB-AMS in ecMenE. Note that K483 is disordered in the structure of saMenE.

Figure 3. Overlaid structures of OSB-AMS: R195K ecMenE and apo saMenE.

Structure overlay of the OSB-AMS:ecMenE complex with apo saMenE (3IPL.pdb). These structures differ in the relative orientation of large domain 1 (silver) and small domain 2 (beige for E. coli and cyan for S. aureus), but represent the adenylate-bound conformation in which G358/G402 (yellow) in the A8 core motif is removed from the active site whereas the K437/K483 (red) is located in the active site. G358 and K437 are residues from E. coli MenE. G402 and K483 are residues from S. aureus. K483 is disordered in the S. aureus structure.

OSB-AMS binding site and keto-acid isomeric form of the inhibitor

Residues that interact with OSB-AMS (1) are highlighted in Figure 4 and Figure S5 and include T142, H186, S188, K195 (R195), S222, T272, D336, R350, and K437, which are all conserved in E. coli, S. aureus, and M. tuberculosis MenE (Figure 2). Residues T142 (Motif 1, A3, P-loop), T272 (Motif II, A5), D336 (Motif III, A7), R350 (A8, hinge) and K437 (A10) are components of the conserved sequence motifs in the adenylate-forming enzyme superfamily (Figure S4) and are, thus, involved in the general chemical reaction that leads to acyl-adenylate formation. Residues S188, K195 (R195), S222 and T277 are clustered around the OSB portion of OSB-AMS and likely confer substrate specificity upon MenE.

Figure 4. X-ray structure of OSB-AMS: R195K ecMenE showing interactions with OSB-AMS (A).

The overall structure of ecMenE:OSB-AMS is shown with the larger N-terminal (domain I) and the smaller C-terminal domain (domain II) highlighted by transparent surface representations in blue and green, respectively. The ligand is provided in ball-and-stick representation. (B) The structure of the bound ligand, OSB-AMS, is shown in the active site. The ligand with green carbon atoms in ball-and-stick representation and side chains that hydrogen bond with the ligand are shown with grey carbon atoms. (C) A schematic of OSB-AMS in the ecMenE active site. The ecMenE sidechains interacting with OSB-AMS are shown in black, the ligand is shown in red, and hydrogen bonding interactions are illustrated with dashed lines.

The electron density of the OSB-AMS ligand is well-defined and consistent with the keto-acid isomer rather than the lactol isomer (Figure S6). In addition, the OSB carboxylate interacts with K195 via a water-mediated ionic bridge comprised of two conserved water molecules (Figure 4). We posit that R195 in wild-type EcMenE also participates in this water mediated interaction, although a direct interaction with the OSB carboxylate cannot be ruled out. In either case, the X-ray structure is consistent with our model of OSB-AMS bound to saMenE^[44] as well as the site-directed mutagenesis studies reported above. In particular, the experimentally observed change in binding free energy (ΔΔG) of OSB-AMS to ecMenE is consistent with the removal of one (R195K) or two (R195Q) water-mediated hydrogen bond interactions with the ligand, suggesting that the R195 guanadinium group in wild-type ecMenE makes two interactions with the OSB carboxylate moiety. These studies further support the notion that the OSB substrate binds to MenE as its open-chain keto acid isomer.

Antibacterial activity and cytotoxicity

To assess the antibacterial activity of the biochemically-active MenE inhibitors OSB-AMS (1), tetrazole 3, squarate 4, and difluoroindanediol 11, we determined their ability to inhibit growth of B. subtilis, MRSA, M. tuberculosis and E. coli using a 96-well plate inhibitor screening assay (Table 3). E. coli was included as a control since this Gram-negative organism does not produce menaquinone under aerobic conditions^[71] and, as expected, none of the compounds inhibited growth of E. coli up to 500 μg/mL concentration. Meanwhile, OSB-AMS (1) had MIC values of 62.5, 31.25, and 125 μg/mL against B. subtilis, MRSA, and M. tuberculosis, respectively. Notably, while tetrazole 3 and squarate 4 showed little or no antibacterial activity (MIC ≥ 250 μg/mL), difluoroindanediol 11 had MIC values of 31.25, 15.6, and 15.6 μg/mL against B. subtilis, MRSA and M. tuberculosis, respectively, 2- to 8-fold more potent than OSB-AMS. One possibility is that the negatively charged sidechain of OSB-AMS (1), tetrazole 3, and squarate 4 limits their bacterial permeability and, hence, their antimicrobial activities, despite their potent biochemical activities (Table 1). In contrast, difluoroindanediol 11 has a sidechain pK_a of 11.5 (presumed to correspond to the secondary alcohol) and is neutral at physiological pH, which may afford improved cell permeability and, hence, increased antibacterial activity, despite its 60-fold less potent biochemical inhibition of ecMenE compared to OSB-AMS. We note that the acyl sulfamate is expected to be negatively charged in all cases (pK_a ~ 1.5–2.0).

Table 3.

Antimicrobial and cytotoxic activity of lead OSB-AMS analogues.

Inhibitor	MIC (μg/mL)1				MIC + MK4 (μg/mL)2		Cytotoxicity (μg/mL)3
	E. coli	B. subtilis	MRSA	M. tuberculosis	MRSA	M. tuberculosis	Vero
1	> 250	62.5	31.25	125	> 250	> 250	125
3	> 250	250	> 500	ND	> 500	ND	ND
4	> 250	250	> 500	ND	> 500	ND	ND
11	> 250	31.25	15.6	15.6	> 250	> 250	> 250

¹MIC values were obtained against E. coli (K-12), B. subtilis (ATCC 6057), methicillin-resistant S. aureus (ATCC BAA-1762), and M. tuberculosis (H37Rv). Inoculum levels for each MIC measurement ranged from 1×10⁶ to 2 ×10⁶ cells/mL. All MICs were performed in technical and experimental triplicate. ND, not determined.

²MICs performed with exogenous 10 μg/mL MK4 added to the synthetic growth media.

³Cytotoxicity values were obtained against Vero (monkey kidney epithelial) cells. Measurements were performed in technical and experimental triplicate.

Finally, the antibacterial activity of the four MenE inhibitors was assessed in the presence of menaquinone-4 (MK4, 10 μM). All bacteria that were sensitive to the MenE inhibitors were rescued by supplementation with MK4, supporting the target specificity and mechanism of action of the inhibitors (Table 3). The cytotoxicity of OSB-AMS and difluoroindanediol 11 in Vero cells was 125 and >250 μg/mL, respectively (Table 3). Thus, difluoroindanediol 11 exhibits useful selectivity for antibacterial activity over mammalian cell cytotoxicity.

Effect of OSB-AMS on menaquinone levels in S. aureus

To provide direct insight into the mode of action of the MenE inhibitors, we analyzed the effect of OSB-AMS on menaquinone levels in S. aureus by LC-MS/MS (Figure 5). S. aureus produces a series of menaquinones that differ in the number of isoprene units that comprise the side chain. In our analysis, menaquinone-8 (MK8) was the major species with significant quantities of MK7 and MK9. Treatment of S. aureus with OSB-AMS resulted in a ~3–5 fold decrease in the levels of all menaquinones, consistent with the antibacterial activity of this compound resulting from a direct effect on menaquinone biosynthesis.

Figure 5. Effect of OSB-AMS on menaquinone levels in MRSA.

Menaquinone levels were quantified by LC-MS/MS using standard curves generated with MK4 and MK9 (see experimental methods section for more details). A distribution of MKs are present in untreated MRSA with MK8 the most abundant, consistent with other reports^[72]. Treatment with 15.6 μM OSB-AMS (half-MIC) results in a ~60% decrease in MK levels consistent with MenE inhibition. Error bar shown as experimental triplicates.

Discussion

Menaquinone is a lipid-soluble, redox-active cofactor in the bacterial respiratory chain of many pathogens including S. aureus and M. tuberculosis. Enzymes involved in menaquinone biosynthesis are attractive targets for antibacterial drug discovery, and we have previously reported rationally-designed inhibitors of MenE, an ATP-dependent ligase that catalyzes the formation of OSB-CoA in the pathway. Our acyl-sulfonyladenosine inhibitors are stable, non-hydrolyzable analogues of the acyl-adenylate intermediate that is formed during the MenE catalyzed reaction, and the parent inhibitor OSB-AMS (1) (Figure 1) is a nanomolar inhibitor of the MenE enzymes from E. coli, S. aureus, and M. tuberculosis^[44]. However, OSB-AMS has relatively weak antibacterial activity, possibly due to issues with cell penetration/efflux, and consequently we have undertaken a series of studies herein to explore the mode of action of OSB-AMS and to generate SAR for the design of novel MenE inhibitors with improved antibacterial activity.

As expected, OSB-AMS had no effect on E. coli grown aerobically (MIC > 500 μg/mL), as this organism only utilizes menaquinone under anaerobic conditions. However, OSB-AMS also has only moderate antibacterial activity against menaquinone-dependent bacteria, with MIC values of 30–125 μg/mL against S. aureus, M. tuberculosis and B. subtilis, in contrast to its potent biochemical inhibition of MenE. Despite this, there is evidence that OSB-AMS does directly affect menaquinone biosynthesis in these organisms: sub-MIC concentrations of OSB-AMS reduce the level of menaquinone in S. aureus, and the antibacterial activity of OSB-AMS can be complemented by the addition of MK4 to the media. We speculated that the disconnect between enzyme inhibition and antibacterial activity might be a consequence of poor cell penetration and/or efflux, and thus we synthesized analogues of OSB-AMS to explore further the relationship between MenE biochemical inhibition and cellular antimicrobial activity. Because OSB-AMS can exist in two isomeric forms (Figure 1), a specific goal of our studies was to determine the bioactive isomer of our lead compound.

Previous modeling studies using the apo structure of saMenE identified a putative interaction between a conserved arginine (R222 in saMenE, R195 in ecMenE) and the OSB carboxylate. Site-directed mutagenesis supports the importance of R195 for both substrate ligation and MenE inhibition: replacement of R195 in ecMenE with Lys resulted in a 16-fold increase in K_m for OSB, while an R195Q mutant had no detectable catalytic activity. In addition, the affinity of OSB-AMS for ecMenE decreased 9- and 102-fold for the R195K and R195Q ecMenE mutants, respectively, supporting a direct interaction between K195 and the OSB carboxylate. Further weight to this conclusion can now be provided by the structure of OSB-AMS bound to the R195K ecMenE mutant reported here, which is the first liganded structure of a MenE enzyme. Our structure reveals a water-mediated interaction between K195 and the OSB-AMS carboxylate, which we posit is also present in enzyme–inhibitor and enzyme–substrate complexes involving wild-type MenE. Finally, the importance of the ortho-carboxylate group was also evaluated by synthesizing the meta-substituted congener MSB-AMS (2). This analogue did not affect MenE activity up to a concentration of 250 μM.

In addition to substantiating the importance of R195, the X-ray structure of the ecMenE(R195K)·OSB-AMS complex also provides insight into the active isomeric form of the inhibitor: electron density of the ligand is consistent with the keto-acid isomer of OSB-AMS (Figure S6). To further explore the importance of the keto-acid versus lactol isomers of OSB-AMS for enzyme inhibition, we synthesized a series of OSB-AMS analogues that mimic either of the two isomeric species. OSB-AMS has an IC₅₀ value of 25 nM for ecMenE, and we discovered analogues of both the keto-acid and lactol isomers that inhibit the enzyme with IC₅₀ values in the 0.2–2 μM range. Given the key interaction between the OSB carboxylate and R(K)195 in the active site of MenE revealed by structural studies and site-directed mutagenesis, we speculate that the presence of a negatively charged substituent in tetrazole 3 and squarate 4 is likely critical for inhibition of MenE. In addition, although the pK_a of difluoroindanediol 11 is 11.5, it is possible that one of the hydroxyl groups could be deprotonated when bound in the MenE active site and interact with R(K)195. However, the corresponding open-chain trifluoroethanol analogue 8 also has a side chain pK_a of 11.5 but does not inhibit MenE, suggesting an alternative scenario in which difluoroindanediol 11 may adopt a distinct binding mode compared to OSB-AMS. The difluoroindanediol lactol analogue 11 is particularly interesting as it exhibits promising antibacterial activity (MIC 15–30 μg/mL) despite inhibiting ecMenE 60-fold less potently than OSB-AMS. This may result from improved cell penetration as the difluoroindanediol is expected to be neutral at physiological pH. Taken together, our studies provide critical mechanistic understanding that will drive future efforts to develop MenE inhibitors with improved enzyme inhibition and cellular activity.