In Vitro Validation of Acetyltransferase Activity of GlmU as an Antibacterial Target in Haemophilus influenzae

Background: A search was initiated to identify inhibitors of the acetyltransferase domain of GlmU that could be exploited as starting points for new antimicrobials.

Results: Sulfonamide inhibitors were identified that upon chemical modification displayed antimicrobial activity mediated via GlmU.

Conclusion: Enzymatic inhibition of GlmU can lead to antimicrobial activity.

Significance: For the first time, GlmU was validated as an antimicrobial target in vitro.

Abstract

GlmU is a bifunctional enzyme that is essential for bacterial growth, converting d-glucosamine 1-phosphate into UDP-GlcNAc via acetylation and subsequent uridyl transfer. A biochemical screen of AstraZeneca's compound library using GlmU of Escherichia coli identified novel sulfonamide inhibitors of the acetyltransferase reaction. Steady-state kinetics, ligand-observe NMR, isothermal titration calorimetry, and x-ray crystallography showed that the inhibitors were competitive with acetyl-CoA substrate. Iterative chemistry efforts improved biochemical potency against Gram-negative isozymes 300-fold and afforded antimicrobial activity against a strain of Haemophilus influenzae lacking its major efflux pump. Inhibition of precursor incorporation into bacterial macromolecules was consistent with the antimicrobial activity being caused by disruption of peptidoglycan and fatty acid biosyntheses. Isolation and characterization of two different resistant mutant strains identified the GlmU acetyltransferase domain as the molecular target. These data, along with x-ray co-crystal structures, confirmed the binding mode of the inhibitors and explained their relative lack of potency against Gram-positive GlmU isozymes. This is the first example of antimicrobial compounds mediating their growth inhibitory effects specifically via GlmU.

Introduction

The increasing medical need for new antibacterial agents resulting from primary resistance of emerging pathogens and acquired resistance among established pathogens has been well documented (1). To address this need, a two-pronged approach has emerged. Existing scaffolds inhibiting well established targets have been optimized further to circumvent resistance mechanisms, e.g. ceftaroline (2). Alternatively, novel targets have been explored because presumably their antibacterial effectiveness has not been eroded by accumulation of mechanism-based resistance mutations in clinical strains. Inhibitors of new antibacterial targets have started to enter the initial phases of clinical testing (3).

GlmU is a bifunctional enzyme involved in the synthesis of UDP-N-acetylglucosamine (UDP-GlcNAc). UDP-GlcNAc is a key intermediate in the synthesis of peptidoglycan, which is essential for growth for nearly all eubacteria. GlmU contains two separate catalytic sites, an acetyltransferase and a uridyltransferase site, responsible for yielding N-acetylglucosamine 1-phosphate and UDP-GlcNAc in consecutive steps (4). As synthesis of UDP-GlcNAc is essential for peptidoglycan formation and, in Gram-negative species, synthesis of lipopolysaccharides, inhibition of GlmU leads to a loss of cell wall integrity, resulting in cell death (5, 6). Quinazolines and aminopiperidines are the first reported inhibitors of the uridyl transferase domain of Haemophilus influenzae GlmU (Protein Data Bank codes 2WOV and 2WOW) (7). Inhibitors of the acetyltransferase of Escherichia coli GlmU have also been identified (8).

This report describes the identification of novel sulfonamide inhibitors of the acetyltransferase of E. coli GlmU that are competitive with acetyl-CoA. A subsequent iterative chemistry effort improved the biochemical potency of these inhibitors and afforded compounds with antimicrobial activity against a strain of H. influenzae that lacks efflux via the AcrB-TolC efflux pump (9). Mode-of-action studies showed that the compound acts via GlmU, thus providing for the first time in vitro validation of the target, i.e. showing that chemical inhibition of GlmU results in inhibition of bacterial growth.

EXPERIMENTAL PROCEDURES

Strains

Bacterial strains used in this study for both susceptibility testing and as a source of genomic DNA for cloning work were Streptococcus pneumoniae NCTC7466, Staphylococcus aureus RN4220 (10), E. coli ATCC 27325, and H. influenzae ATCC51907. The latter two were parental strains of E. coli ΔtolC::Tn10 and H. influenzae acrB::cat, respectively.

Construction of Expression Vectors

All glmU genes were amplified by PCR using genomic DNA isolated from the respective pathogens (Wizard Genomic Prep, Promega, Madison WI) as templates and the following primer pairs: E. coli, 5′-ACGTCATATGTTGAATAATGCTATGAGC-3′ and 5′-ACGTGAATTCTCACTTTTTCTTTACCGGACG-3′; H. influenzae, 5′-AACTACATATGACAAAAAAAGCATTAAGTGCGG-3′ and 5′-GTCTTGCTGCCTTTTGTTCG-3′; S. aureus, 5′-ACTACATATGCGAAGACCACGCGATAATTTTGGC-3′ and 5′-ACTAAGTCGACGGATTATCCTTTTATCCTAGCC-3′; and S. pneumoniae, 5′-ACTAACATATGTCAAATTTTGCCATTATTTTAGCAGCG-3′ and 5′-ACTAAGTCGACTCATCACTGGTTCTTAGGATGATGAGG-3′. Amplifications were performed using High Fidelity PCR Master (Roche Applied Science, Indianapolis, IN), and PCR products were purified using a QuickStep 2 PCR purification kit (EDGE Biosystems, Gaithersburg, MD). The 1.4-kb E. coli glmU PCR product was digested with NdeI/EcoRI and cloned into similarly digested pMAL-p2x vector (New England Biolabs) to generate pLH734. E. coli glmU was subcloned from pLH734 as an NdeI/SalI fragment into NdeI/XhoI-digested pZT-73.3 (11) to generate pBA738. The 1.7-kb H. influenzae glmU PCR product was cloned into pCR4-TOPO (Invitrogen) to generate pBA742 and then subcloned as an NdeI/EcoRI fragment from pBA742 into similarly digested pET30a (Novagen, Madison, WI) to generate pBA750. The 1.4-kb S. aureus glmU PCR product was cloned into pCR4-TOPO to generate pBA986. The glmU gene was isolated from pBA986 by SalI digest followed by partial NdeI digest, and the fragment was cloned into NdeI/XhoI-digested pZT7–3.3 to generate pBA987. The 1.4-kb S. pneumoniae glmU PCR product was digested with NdeI/SalI and cloned into NdeI/XhoI-digested pZT7–3.3 to generate pBA989. DNA sequences of the cloned glmU genes were confirmed by sequencing on an ABI PRISM 3100 DNA sequencer (Applied Biosystems) using Big Dye Terminator cycle sequencing kit (Applied Biosystems). Computer analyses of DNA sequences were performed with Sequencher (Gene Codes Corp., Ann Arbor, MI).

GlmU Overexpression

E. coli GlmU

pBA738 was transformed into E. coli HMS174(DE3) (Novagen, Madison, WI) and plated on Luria-Bertani (LB)³ agar containing 10 μg/ml tetracycline (Fisher Scientific). After overnight growth at 37 °C, several transformants were inoculated into 3 liters of LB broth containing 10 μg/ml tetracycline and grown at 37 °C with aeration to mid-logarithmic phase (A₆₀₀ = 0.5) after which isopropyl β-d-thiogalactopyranoside (Acros Organics, Geel, Belgium) was added to a final concentration of 1 mm. After 2 h at 37 °C, the cells were harvested by centrifugation at 5000 × g for 15 min at 25 °C. Cell paste was stored at −20 °C, and protein expression and solubility were checked by SDS-PAGE.

H. influenzae GlmU

pBA750 was transformed into E. coli BL21(DE3) (Novagen) and plated on LB agar containing 25 μg/ml kanamycin (Acros Organics). After overnight growth at 37 °C, several transformants were inoculated into 3 liters of LB broth containing 25 μg/ml kanamycin and grown at 37 °C with aeration to mid-logarithmic phase (A₆₀₀ = 0.5), and then isopropyl β-d-thiogalactopyranoside was added to a final concentration of 1 mm. After 3 h at 37 °C, the cells were harvested by centrifugation at 5000 × g for 15 min at 25 °C. Cell paste was stored at −20 °C, and protein expression and solubility were checked by SDS-PAGE.

S. aureus GlmU

pBA987 was transformed into E. coli BL21(DE3) (Novagen) and plated on LB agar containing 10 μg/ml tetracycline. After overnight growth at 37 °C, several transformants were inoculated into 3 liters of LB broth containing 10 μg/ml tetracycline and grown at 30 °C with aeration to mid-logarithmic phase (A₆₀₀ = 0.5), and then isopropyl β-d-thiogalactopyranoside was added to a final concentration of 1 mm. After 3 h at 30 °C, the cells were harvested by centrifugation at 5000 × g for 15 min at 25 °C. Cell paste was stored at −20 °C, and protein expression and solubility were checked by SDS-PAGE.

S. pneumoniae GlmU

pBA989 was transformed into E. coli HMS174(DE3) (Novagen) and plated on LB agar containing 10 μg/ml tetracycline. After overnight growth at 37 °C, several transformants were inoculated into 3 liters of LB broth containing 10 μg/ml tetracycline and grown at ambient temperature with aeration to mid-logarithmic phase (A₆₀₀ = 0.5), and then isopropyl β-d-thiogalactopyranoside was added to a final concentration of 1 mm. After 18 h of induction at ambient temperature, the cells were harvested by centrifugation at 5000 × g for 15 min at 25 °C. Cell paste was stored at −20 °C, and protein expression and solubility were checked by SDS-PAGE.

Purification of GlmU Isozymes

The frozen cell pastes from 3 liters of cell culture were suspended in 50 ml of lysis buffer consisting of 50 mm Tris-HCl (pH 8.0), 2 mm EDTA, 2 mm DTT, 10% (v/v) glycerol, 50 mm NaCl, and 1 protease inhibitor mixture tablet (Roche Diagnostics). Cells were disrupted by French press at 18,000 psi twice at 4 °C, and the crude extract was centrifuged at 25,000 rpm (45Ti rotor, Beckman-Coulter, Brea, CA) for 30 min at 4 °C. The supernatant was applied at a flow rate of 2.0 ml/min onto a 20-ml Q-Sepharose HP (HR16/10) column (GE Healthcare) pre-equilibrated with Buffer A, consisting of 50 mm Tris-HCl (pH 8.0), 2 mm EDTA, 2 mm DTT, 50 mm NaCl, and 10% (v/v) glycerol. The column was then washed with Buffer A, and the protein was eluted by a linear gradient from 0 to 1 m NaCl in Buffer A. Fractions containing GlmU were pooled, and solid (NH₄)₂SO₄ was added to a final concentration of 1 m. The sample was applied at a flow rate of 2.0 ml/min onto a 20-ml Phenyl Sepharose HP (HR16/10) column pre-equilibrated with Buffer B consisting of 50 mm Tris-HCl (pH 8.0), 2 mm EDTA, 2 mm DTT, 10% (v/v) glycerol, and 1 m (NH₄)₂SO₄. The column was washed with Buffer B, and the protein was eluted by a linear gradient from 1 to 0 m (NH₄)₂SO₄ in Buffer B. Fractions containing GlmU were pooled, and solid (NH₄)₂SO₄ (0.4 g/ml) was added to precipitate all the proteins, mixed, and kept on ice for 1 h. The sample was centrifuged at 20,000 rpm for 30 min at 4 °C, and the pellet was then dissolved in 5 ml of Buffer A. The 5-ml sample was applied at a flow rate of 1.5 ml/min onto a 320-ml Sephacryl S-200 (HR 26/60) column (GE Healthcare Life Sciences) pre-equilibrated with Buffer C consisting of 50 mm Tris-HCl (pH 8.0), 2 mm EDTA, 2 mm DTT, 10% (v/v) glycerol, and 150 mm NaCl. The fractions containing GlmU were pooled and dialyzed against 1 liter of storage buffer consisting of 50 mm Tris-HCl (pH 8.0), 2 mm EDTA, 2 mm DTT, 150 mm NaCl, and 20% (v/v) glycerol. The protein was characterized by SDS-PAGE analysis and analytical LC-MS. The protein was stored at −80 °C. The same procedure was used to purify H. influenzae GlmU, S. aureus GlmU, and S. pneumoniae GlmU.

IC₅₀ Measurements

Compounds were dissolved in dimethyl sulfoxide (DMSO) and 2-fold serial dilutions were prepared with DMSO, with a top concentration of 10 mm. Two μl of compound dilutions were added to clear polystyrene 96-well plates (Corning, Corning, NY). Enzyme reactions were performed in a 100-μl volume containing 225 μm acetyl-CoA (Sigma), 225 μm d-glucosamine 1-phosphate (Sigma), and 110 pm E. coli GlmU, 2.7 nm H. influenzae GlmU, 260 pm S. pneumoniae GlmU, or 6 nm S. aureus GlmU in assay buffer consisting of 50 mm MOPS-NaOH (pH 7.35), 75 mm potassium acetate, 10 mm MgCl₂, and 0.005% Tween 20. Reactions were quenched after 30 min with 50 μl of 1.5 mm Ellman's reagent (Sigma) in 0.1 m sodium phosphate (pH 7.2). The absorbance at 412 nm was measured after 5 min with a Spectramax plate reader (Molecular Devices, Sunnyville, CA). The baseline absorbance for 100% inhibition was measured for reactions inhibited by pretreatment of the enzyme for 30 min with 200 μm γ-maleimidobutyric acid (Sigma). Because the background absorbance, which is due to the presence of CoA thiol as a contaminant of the acetyl-CoA, was suppressed by reaction with γ-maleimidobutyric acid, the absorbance was corrected by supplementation with 12 μm 4-nitrophenol (Sigma). Percent inhibition was calculated using the equation: % inhibition = 100(1 − (A₄₁₂ − Min)/(Max − Min)), where A₄₁₂ is the absorbance in the test well, Min is the absorbance of the fully inhibited reaction, and Max is the absorbance of the uninhibited reaction. The compound concentration resulting in 50% inhibition (IC₅₀) was calculated by nonlinear least-squares regression to the equation: % inhibition = 100[I]^b/(IC₅₀ + [I]^b), where [I] is the inhibitor concentration, and b is the Hill slope (typically 1.0).

Kinetic Mode of Inhibition

Enzyme activity measurements for analysis of the kinetic mode of inhibition were performed in assay buffer in 384-well clear polystyrene assay plates (Corning). The reaction volume of 50 μl contained 2% (v/v) DMSO from the inhibitor stock solution. Reactions contained 0–4 μm inhibitor, 0–1400 μm d-glucosamine 1-phosphate, 60–1000 μm acetyl-CoA, and 100 pm E. coli GlmU. After 30 min, the reactions were quenched with 25 μl of 3 mm Ellman's reagent in 100 mm sodium phosphate (pH 7.2). The absorbance at 412 nm was measured after 8 min with a Spectramax plate reader. The averages and standard deviations were calculated for 4 replicate measurements. The data were fit globally by nonlinear least squares regression to kinetic rate equations using the program Grafit (Erithacus Software).

The kinetic mechanism of the acetyltransferase activity of GlmU has not been elucidated. Therefore, the data from the above experiment were fit to rate equations for a set of plausible mechanisms, taking into account the observation of substrate inhibition by both acetyl-CoA and d-glucosamine 1-phosphate. The purpose of the fitting was to determine whether the inhibitor was likely to be competitive with acetyl-CoA or d-glucosamine 1-phosphate or both, rather than to support a particular kinetic mechanism for the enzyme. The rate equations allowed for competitive inhibition by the inhibitor with both of the substrates with different potencies and exclusive of simultaneous binding with the substrate inhibitor occupying the same substrate binding site. The steady-state ordered bisubstrate rate equation is as follows (12).

The Ping Pong bisubstrate rate equation is as follows.

V is the reaction rate. V_max is the theoretical maximal reaction rate when the enzyme is saturated with substrates. [A] and [B] are the concentrations of substrates A and B, respectively. K_mA and K_mB are the Michaelis constants for substrates A and B, respectively. K_ia is the dissociation constant of substrate A. K_siA and K_siB are the substrate inhibition constants for substrates A and B, respectively. K_I1 and K_I2 are the inhibition constants for the inhibitor when binding to the enzyme without bound substrate and the enzyme with substrate A bound, respectively, in the steady-state ordered kinetic mechanism. In the ping-pong mechanism, K_I2 is the inhibition constant for the inhibitor binding to the enzyme after its reaction with substrate A and dissociation of the first product.

Susceptibility Testing

Minimum inhibitory concentrations (MIC) were determined using broth microdilution according to the guidelines of the Clinical and Laboratory Standards Institute (13). In the case of the GlmU overexpression strains and compound-resistant strains, the differences in MIC were too small to be measured reliably with 2-fold serially diluted compounds. Instead, arithmetic dilutions were used either starting at a maximum concentration of 100 μg/ml in steps of 10 μg/ml or 200 μg/ml in steps of 20 μg/ml maintaining a final DMSO concentration of 2% (v/v).

H. influenzae GlmU Overexpression Strains

glmU was amplified from H. influenzae acrB::cat genomic DNA using 5′-AACTACATATGACAAAAAAAGCATTAAGTGCGG-3′ and 5′-ACTAAGTCGACTTATTTTTTCTTTATTGGTCGTTGCC-3′. The 1.7-kb fragment was digested with NdeI and SalI and cloned in similarly digested pASK5 (14), yielding pBA756. A TetAR cassette was generated by amplification of pASK6 (14) using 5′-TGCTGCAGTTAAGACCCACTTTCACATTTAAGTTGTTTTTCTAATCCGCAAATGATCAATTCAAGGCC-3′ and 5′-TTGCCCGGGAAGCTTGCATGCTGCAGGCTATTTACCGCGGCTTTTTATTGAGC-3′, and the resulting 2.1-kb product was digested with PstI and cloned into similarly digested pBA756, which replaces the chloramphenicol resistance marker with a tetracycline resistance marker. Using 5′-CGTACGGACGTCAAGCGTATTCAGCAAGCG-3′ and 5′-CTCGCAAAGCTTGCCAAATGGCTTGTTGAAATGCG-3′, the Omp-GlmU-TetAR was amplified and transformed into H. influenzae acrB::cat. Similarly, the control cassette was amplified from pASK6 and transformed into H. influenzae acrB::cat. This yielded strains H. influenzae ARC2347 and ARC2348, respectively.

NMR

All NMR spectra were acquired at 298 K with a 600 MHz NMR instrument (Bruker, Billerica, MA) with an AVANCE III console and a triple-resonance cryogenic probe. In the WaterLOGSY experiment (15), the first water-selective 180° Sinc pulse was 6 ms long, and a weak rectangular pulse field gradient was applied during the mixing time (1.8 s). A gradient recovery time of 2 ms was introduced after the mixing time. Water suppression was achieved by the excitation sculpting scheme (16), and the water-selective 180° Sinc shape pulse was 3 ms long. The data were collected with a sweep width of 9157 Hz, 0.45 s of aquisition time, and 1.8 s for the relaxation delay. 64 scans were recorded for each experiment, requiring 5 min per spectrum. The data were zero-filled to 32,768 complex points and multiplied by an exponential function (line broadening 3 Hz) prior to Fourier transformation.

Binding Studies

Binding of substrates and an inhibitor to GlmU was investigated with isothermal titration calorimetry (MicroCal VP-ITC, GE Healthcare). E. coli GlmU was dialyzed into buffer consisting of 50 mm MOPS-NaOH (pH 7.35), 75 mm potassium acetate, 10 mm magnesium chloride, and 2% DMSO. Substrates acetyl-CoA and d-glucosamine 1-phosphate, and compound 2 were dissolved in identical buffer. For titrations, 500 μm acetyl-CoA was injected into 15 μm GlmU, 200 μm compound 2 was injected into 18.8 μm GlmU, and 500 μm d-glucosamine 1-phosphate was injected into 25 μm GlmU. Collected data were normalized and fit to a one-site binding model using Origin software (version 7).

Crystallization and Structure Solution

Crystals were grown at ambient temperature by the hanging drop method. The protein in 50 mm Tris-HCl, pH8.0, 150 mm NaCl, 2 mm EDTA, 2 mm Tris(2-carboxylethyl)phosphine was equilibrated against a reservoir solution containing 19–22% (w/v) PEG3350, 100 mm propionic acid/cacodylate/Bis-trispropane system (pH 5.5), and 400 mm ammonium sulfate. Complexes were obtained by co-crystallization of the protein with compound 3 to a final concentration of 4 mm. Crystals were flash frozen, and diffraction data were collected to 1.9 Å in a home source. Refinement was performed using refmac in CCP4 and Coot.

Isolation of Resistant Mutants

Mutants of H. influenzae acrB::cat were isolated as described previously (17). In short, 100 μl of freshly prepared Haemophilus test medium (13) containing 10⁴–10⁸ cfu from an overnight culture were plated on Haemophilus test medium agar plates containing filter-sterilized hematin (Sigma), filter-sterilized nicotinamide adenine dinucleotide (Sigma), and various 2-fold serial dilutions of compound 5, ranging from 16–128 μg/ml, were all added immediately prior to pouring the plates. Resistant colonies appeared after incubation for 72 h at 37 °C with 5% (v/v) CO₂ at the plate MIC compound concentration of 32 μg/ml at a rate of 10⁻⁶–10⁻⁷. None were found at higher compound concentrations, suggesting a resistance rate of <10⁻⁸. Resistant colonies were purified on plates identical to those from which they were isolated. About half of the colonies grew up again and were subsequently stored in Haemophilus test medium containing 50% (v/v) glycerol at −80 °C prior to further analysis.

Inhibition of Macromolecule Biosynthesis

Measurements of inhibition of the incorporation into growing cells of various radiolabeled metabolic precursors was performed according to Hilliard et al. (18), with changes as described below. Specific radioactivities and final concentrations were as follows: 315 Ci/mol l-[U-¹⁴C]leucine, final concentration of 5 μCi/ml; 28 Ci/mmol l-[3,4(n)-³H]valine, final concentration of 5 μCi/ml; 50 Ci/mol N-acetyl-d-[1-¹⁴C]glucosamine, final concentration of 1 μCi/ml; 58 Ci/mol [2-¹⁴C]acetic acid, final concentration of 5 uCi/ml plus 50 mg/liter unlabeled sodium acetate; 25 Ci/mmol [³H]uridine, final concentration of 5 μCi/ml plus 120 mg/liter unlabeled uridine (Sigma); and 25 Ci/mmol [methyl-³H]thymidine, final concentration of 5 μCi/ml plus 5 mg/liter unlabeled thymidine (Alfa Aesar, Ward Hill, MA). Radiolabeled precursors were from GE Healthcare. H. influenzae was harvested when in exponential phase (A₆₀₀ = 0.4, 250 ml of culture in a 1-liter flask, shaking at 100 rpm, 30 °C), centrifuged at ambient temperature for 10 min at 3500 × g, and resuspended at the same cell density in fresh, prewarmed Haemophilus test medium. A volume of 40 μl was added to wells of a prewarmed 96-well plate containing 20 μl of 2-fold serially diluted compound in Mueller Hinton broth (12) (plus DMSO (2% v/v final) plus 40 μl of Mueller Hinton broth containing the radiolabeled precursor. The incorporation was stopped with one volume of ice-cold stopping reagent (20% (w/v) trichloroacetic acid with Bacto^TM casamino acids (Bacto, Liverpool, UK), uridine, thymidine, N-acetylglucosamine, and sodium acetate, each at 1% (w/v)) after 0, 30, and 60 min of incubation, during which incorporation progressed linearly. Plates were stored at 4 °C overnight after which well contents were transferred using a Filtermate harvester (Packard, Downers Grove, IL) onto Packard GF/B 96-well Unifilters. Wells and filters were washed 10 times with 200 μl of water to remove unincorporated precursors. These filters were dried, and 40 μl of scintillation fluid (Microscint 20, PerkinElmer Life Sciences, Waltham, MA) was added to each well, and radioactivity was measured with a TopCount scintillation counter (Packard).

Compounds

The chemical synthesis of compounds 1-5 was performed in-house and will be presented elsewhere. Compound purity was determined by LC/MS shortly before biological testing and was >90%.

RESULTS

A high throughput screen of the AstraZeneca compound collection was performed using E. coli GlmU by detecting inhibition of its acetyltransferase activity, resulting in the identification of compound 1 (Table 1). Its inhibitory activity was retained upon resynthesis of the compound. The compound was similarly potent against H. influenzae GlmU but did not inhibit GlmU of Gram-positive pathogens S. pneumoniae and S. aureus at concentrations <200 μm.

TABLE 1.

Biochemical and antimicrobial properties of sulfonamide inhibitors of GlmU acetyl transferase

No antimicrobial activity, i.e. MIC > 64 μg/ml, was observed with wild type strains of S. aureus (Sau), S. pneumoniae (Spn), H. influenzae (Hin), and E. coli (Eco) and E. coli tolC::Tn10.

Using E. coli GlmU, the mode of inhibition of compound 1 was investigated with steady-state kinetics. Kinetic constants were determined by non-linear least-squares global fitting of the data from the mode of inhibition experiment to steady-state ordered bisubstrate rate equations in which either acetyl-CoA or d-glucosamine 1-phosphate is the first substrate to bind (Table 2). Both substrates caused substrate inhibition. The equations allowed for competition by compound 1 with both substrates with different K_I values. The results for the Ping Pong bisubstrate model fitting are not shown because the fitting to the ordered models was significantly better (p < 0.01). The difference in χ² values between the fits to the two ordered models is not statistically significant (p = 0.22). The results show that compound 1 is predominantly competitive with acetyl-CoA. The fit of the data to the ordered bisubstrate rate equation with acetyl-CoA as the first substrate is shown in Fig. 1. The Michaelis constants for acetyl-CoA and d-glucosamine 1-phosphate are similar to apparent K_m values reported in the literature for E. coli GlmU acetyltransferase. Mengin-Lecreulx and van Heijenoort (4) reported apparent K_m values of 600 and 150 μm for acetyl-CoA and d-glucosamine 1-phosphate, respectively. Gehring et al. (19) reported K_m values of 320 and 250 μm for acetyl-CoA and d-glucosamine 1-phosphate, respectively.

TABLE 2.

Results of best-fit nonlinear least squares regression of the kinetic mode of inhibition data for the inhibitor to the steady-state ordered rate equation including substrate inhibition by both substrates and competition by compound 1 with both substrates

The data were fit to models in which acetyl-CoA or d-glucosamine 1-phosphate is substrate A, i.e. the first substrate to bind. The difference in χ² values for the two regressions did not reach statistical significance (p = 0.22).

Kinetic constant	Substrate A
	Acetyl-CoA	d-Glucosamine 1-phosphate
Vmax (ΔA412)	1.49 ± 0.03	1.48 ± 0.03
Kia (μm)	24 ± 2	25 ± 3
KmA (μm)	398 ± 15	290 ± 2
KmB (μm)	311 ± 11	423 ± 16
KsiA (μm)	2050 ± 190	965 ± 40
KsiB (μm)	881 ± 36	2400 ± 280
KI1 (μm) (substrate A competitive)	0.79 ± 0.02	7.1 ± 0.5
KI2 (μm) (substrate B competitive)	none	0.86 ± 0.02
χ2	0.48	0.44

FIGURE 1.

A, the steady-state ordered bisubstrate kinetic mechanism in which acetyl-CoA is the first substrate, both substrates cause substrate inhibition, and the inhibitor competes with acetyl-CoA is one plausible kinetic mechanism for the E. coli GlmU acetyltransferase activity based on the data herein. B, comparison of the experimental data to the best-fit non-linear least squares regression for the mechanism in A.

Isothermal titration calorimetry was carried out to assess the binding of substrates acetyl-CoA, glucosamine 1-phosphate, and a representative inhibitor to the free form of the enzyme. Because the solubility of compound 1 was insufficient, a more soluble but equipotent analog, compound 2, was used (Table 1). Acetyl-CoA bound to GlmU with K_d = 15 μm and a stoichiometry of 0.6, whereas no evidence of binding of d-glucosamine 1-phosphate was observed (upper limit of detection ∼ 30 μm) (data not shown). Compound 2 bound to GlmU with K_d = 1.9 μm and a stoichiometry of 0.6, very similar to that of acetyl-CoA (Fig. 2). The K_d of acetyl-CoA agrees well with the K_ia from steady-state kinetics with acetyl-CoA as the first substrate to bind (Table 2). Likewise, the binding of compound 2 to free GlmU supports the kinetic conclusion of inhibitor competition with acetyl-CoA. This result was confirmed by WaterLOGSY NMR using H. influenzae GlmU with more potent compounds 4 and 5 (Fig. 3). Reduction of acetyl-CoA signals in the presence of compound 4 or 5 clearly demonstrates that both compounds compete with acetyl-CoA binding (Fig. 3). In addition, greater acetyl-CoA signal reduction in the presence of compound 5 by comparison with compound 4 indicates compound 5 is a stronger acetyl-CoA competitor than compound 4, which is consistent with the IC₅₀ data of these compounds against H. influenzae GlmU (Table 1).

FIGURE 2.

Isothermal titration calorimetry studies of compound 2 binding to E. coli GlmU (K_d = 1.9 μm). Observed heats evolved over time are shown in the top, and the corresponding fitted one-site binding curve is shown at the bottom.

FIGURE 3.

One-dimensional WaterLOGSY spectral comparison of 200 μm acetyl-CoA binding to 10 μmH. influenzae GlmU with (bottom) and without (top) the addition of 200 μm compound 4 (left) or 5 (right). Signals of proton a, b, and c from acetyl-CoA are labeled and compared in the WaterLOGSY spectra. The signals in the dashed boxes are from the compounds. The NMR samples were prepared in 50 mm HEPES (pH 7.5), 5 mm MgCl₂, 5 mm DTT, 0.1 mm EDTA, and 5% D₂O. The spectra were acquired at 298 K.

An x-ray crystallographic system was developed with E. coli GlmU that allowed inhibitors to be co-crystallized with the protein. The overall structure of the protein chain was identical between the acetyl-CoA-bound and inhibitor-bound forms. Clear electron density could be observed for bound compound 3, allowing all atoms to be assigned (Fig. 4B). The inhibitor is bound in each of the three acetyl-CoA binding sites formed by the interfaces of two of the three chains of the trimer and the carboxyl-terminal loop of the remaining monomer (Fig. 4A) (20). This is in agreement with the NMR studies. The aliphatic carboxylate group is buried deeper in the pocket, whereas the sulfonamide side chain faces the bulk solvent.

FIGURE 4.

A, a view of the inhibitor binding pocket in the E. coli GlmU trimer. B, electron density (2F_o − F_c map at 1.2σ) of compound 3 in complex with GlmU. C, binding interactions of compound 3 and acetyl-CoA. Compound 3 has been depicted in thick yellow sticks, and acetyl-CoA is shown in thin gray lines. The three monomers are colored pink, gray, and cyan. The residues mutated in the resistant strains are highlighted in darker shades of the respective monomer color. Hydrogen bonds are depicted as dotted lines.

An iterative chemistry effort led to compound 5 that displayed a 300-fold lower IC₅₀ against the GlmU isozymes from E. coli and H. influenzae (Table 1). Measurable inhibition could also be detected against the Gram-positive GlmU isozyme from S. pneumoniae but not S. aureus. Routine testing of antimicrobial activity showed MIC > 64 μg/ml against any wild type strains and an E. coli strain lacking the TolC subunit of the major efflux transporter AcrABTolC. However, growth of an H. influenzae strain missing AcrB was sensitive and showed an MIC of 32 μg/ml (Table 1).

To determine the mode of action against H. influenzae acrB::cap, inhibition of incorporation of radiolabeled precursors into cellular macromolecules was measured at various concentrations of compound 5. Incorporation of N-acetylglucosamine into peptidoglycan and acetate into lipids was inhibited by 50% at 16–32 μg/ml compound 5, which is similar to the MIC value (Table 1). Inhibition of incorporation of thymidine into DNA, uridine into RNA, and valine and leucine into protein, required at least 64 μg/ml of compound 5 (Fig. 5).

FIGURE 5.

Compound 5 preferentially inhibits N-acetylglucosamine and acetic acid incorporation into macromolecules of H. influenzae acrB::cat. Inhibition of synthesis of protein (leucine and valine incorporation; not shown), RNA (uridine incorporation; circles), fatty acids (acetate incorporation; triangles), peptidoglycan (N-acetylglucosamine incorporation; squares), and DNA (thymidine incorporation; not shown) was measured as described under “Experimental Procedures.” The incorporation rates of added precursors into uninhibited cells were 21, 65, 1100, 7600, 3.1, and 35 μmol/h/A₆₀₀, respectively. Inhibition of leucine, valine, thymidine, and uridine all showed similar modest inhibition and stayed below 50% even at the highest compound concentration tested. For clarity, only uridine is shown as a representative of these four. Data are averages of two independent measurements; S.D. are ∼10%.

A GlmU overexpression strain of H. influenzae acrB, ARC2347, was constructed that was isogenic to its parent, ARC2348, but contained a copy of glmU under control of the ompP1 promotor. A comparison of glmU and omp1 transcript levels in H. influenzae acrB::cat suggested that expression of glmU under control of ompP1 would result in a relatively small, 8-fold, increase in GlmU (data not shown). As a result, the MIC value against ARC2347 for compounds 4 and 5 increased by 50% relative to the parental strain (133 ± 27 μg/ml versus 80 ± 18 μg/ml (n = 6, p = 0.003) and 57 ± 8 μg/ml versus 38 ± 4 μg/ml (n = 7, p = 0.0001), respectively).

Three strains of H. influenzae acrB::cat resistant to compound 5 were isolated with a frequency of 10⁻⁶–10⁻⁷ at a compound concentration that equaled the MIC value 32 μg/ml. None were found at higher compound concentrations suggesting a relatively small increase in MIC. Arithmetic dilution of compound in broth confirmed an increase in MIC from 30 μg/ml to 50 (1 strain) or 60 μg/ml (2 strains). The glmU gene and its promoter of all three strains were sequenced and revealed a single mutation in the glmU genes in the two more resistant strains. Both mutations were located in the region encoding the acetyltransferase domain (Fig. 5). One mutation resulted in a W449G substitution (E. coli numbering (21)), which removed a stabilizing π-stacking interaction between the tryptophan side chain and the aromatic sulfonamide side chain. The other mutation, T420I, sterically alters the orientation of Trp-449 and thus indirectly alters the same π-stacking interaction. In Gram-positive GlmU isozymes, leucine is present in this position (Fig. 6), which likely prevents the π-stacking interaction, and thus efficient binding of the inhibitors, in similar fashion.

FIGURE 6.

Amino acid sequences of the carboxyl-terminal end of four GlmU isozymes, numbered according to the sequence of the E. coli isozyme. Thr-420 and Trp-449, which when mutated to Ile and Gly respectively led to resistance of H. influenzae acrB::cat against compound 5, are highlighted.

DISCUSSION

In a search for chemical starting points to exploit GlmU acetyltransferase for the development of new antibacterials, a high-throughput screen of AstraZeneca's library was performed. To maximize success both quality and diversity of the screened compound set and sensitivity of the assay are important. Because the aim was to identify inhibitors to either the acetyl CoA or the d-glucosamine 1-phosphate binding site, both the kinetic mechanism and the K_m values for each substrate were determined. Following the method of Cheng and Prusoff (22), two equations were derived to calculate the IC₅₀/K_i ratio for the GlmU acetyltransferase assay for dead-end inhibitors competitive either with acetyl-CoA or glucosamine 1-phosphate. The basis for the derivations was the steady-state ordered bisubstrate rate equation given above, with acetyl-CoA as the first substrate to bind and substrate inhibition by both substrates. The kinetic constants from Table 2 and concentration of 225 μm for both substrates were used for the calculations. The IC₅₀/K_i ratio for the high-throughput screen and IC₅₀ assay were 2.6 for acetyl-CoA competitive inhibitors and 3.6 for glucosamine 1-phosphate competitive inhibitors. Thus, the high-throughput screen was only slightly biased toward acetyl-CoA competitive inhibitors at the expense of glucosamine 1-phosphate competitive inhibitors. The 1.9 μm K_d of compound 2 measured by isothermal titration calorimetry and corresponding IC₅₀ of 3.3 μm are consistent, within experimental error, with the IC₅₀/K_i ratio for an acetyl-CoA-competitive inhibitor.

X-ray crystallography confirmed binding of compound 3 in the acetyl-CoA binding site. The carboxamide hydrogen bonds with the peptide backbone of Tyr-387 and Ser-405 and the sulfonyl oxygens hydrogen bond with the Ala-423 backbone and the guanidine side chain of Arg-440. This binding mode is similar to that of the pantothenate moiety of CoA (23). The iterative chemistry effort led to 300-fold improvement in IC₅₀ against the Gram-negative isozymes and introduced activity against the isozyme of S. pneumoniae but not S. aureus. This is mainly due to the extension of the amide substituent with an aliphatic carboxylate, which mimics the β-mercaptoethylamine extension of the pantothenate moiety of CoA and reaches to Ala-380 and Asn-377. The aromatic side chain on the sulfonamide is stabilized by π-stacking with Trp-449. The importance of this interaction is illustrated by the W449G mutant that confers resistance to compound 5 and by weaker activity against S. pneumoniae GlmU. S. pneumoniae GlmU activity could be measured only after the initial hit was extended with larger groups in this region, thus providing additional interactions which compensated for the absence of π-π interactions. Sequence analysis also suggested that S. aureus GlmU contains a shorter carboxyl-terminal end, lacking residues forming this part of the pocket (Fig. 6). This may explain why no activity was observed against S. aureus GlmU with this scaffold.

The compounds with improved biochemical potency displayed antimicrobial activity in a strain of H. influenzae devoid of its main efflux pump. This activity was shown to be consistent with inhibition of GlmU acetyltransferase in the cell. Incorporation of N-acetylglucosamine is particularly sensitive to compound 5, as can be expected when inhibiting a key step in the synthesis of UDP-GlcNAc (Fig. 5). More surprising was the increased sensitivity of acetate incorporation. We hypothesize this reflects inhibition of lipopolysaccharide synthesis. The LpxA protein, UDP-N-acetylglucosamine acyltransferase, catalyzes the first step in lipid A synthesis by transferring β-hydroxymyristate from the acyl transfer protein to UDP-GlcNAc (24). Inhibition of UDP-GlcNAc synthesis could reduce turnover of the acyl-acyl transfer protein pool and thus decrease incorporation of acetic acid.

Antimicrobial activity against wild type strains of H. influenzae and a strain of E. coli lacking its AcrB-TolC efflux system is absent, suggesting poor permeation of the compounds into the cell. This could be due to the polar nature of the most potent compounds (logD_7.4 of −0.5 and −0.3 for compounds 4 and 5, respectively), arising from the carboxylate moiety and resulting in poor membrane permeability. Lowering the pH of the medium from 7.3 to 5.5 to increase protonation and improve membrane permeability reduced the MIC of compound 5 against E. coli tolC::Tn10 (data not shown). This result suggests a path for improving the antimicrobial activity of this series of GlmU acetyltransferase inhibitors.

In summary, an earlier report (8) is confirmed here that the GlmU acetyltransferase is “drug-able,” in that relatively small, specific biochemical inhibitors can be found. Various binding studies, including x-ray crystallography, complemented enzymology studies that both guided iterative chemistry plans and explained the mostly Gram-negative spectrum of the biochemical activity. Compounds with single digit nanomolar IC₅₀s showed antimicrobial activity in H. influenzae acrB::cat that could be measured in standard susceptibility assays (13). Importantly, this antimicrobial activity of these inhibitors was shown to be caused by inhibition of GlmU acetyltransferase transferase. Although clinical drug candidates acting via GlmU are still a long time away, this work for the first time provides in vitro validation of this enzyme activity as an antibacterial target.