Design, Synthesis and Crystal Structures of 6-Alkylidene-2’-Substituted Penicillanic Acid Sulfones as Potent Inhibitors of Acinetobacter baumannii OXA-24 Carbapenemase

Abstract

Class D β-lactamases represent a growing and diverse class of penicillin inactivating enzymes that are usually resistant to commercial β-lactamase inhibitors. As many such enzymes are found in multi-drug resistant (MDR) Acinetobacter baumannii and Pseudomonas aeruginosa, novel β-lactamase inhibitors are urgently needed. Five unique 6-alkylidene-2’-substituted penicillanic acid sulfones (1, 2, 3, 4, and 5) were synthesized and tested against OXA-24, a clinically important β-lactamase that inactivates carbapenems and found in A. baumannii. Based upon the roles Tyr112 and Met223 play in the OXA-24 β-lactamase, we also engineered two variants (Tyr112Ala and Tyr112Ala,Met223Ala) to test the hypothesis that the hydrophobic tunnel formed by these residues influences inhibitor recognition. IC₅₀ values, against OXA-24, and two OXA-24 β-lactamase variants ranged from 10 ± 1 (4 vs. WT) to 338 ± 20 nM (5 vs. Tyr112Ala, Met223Ala). Compound 4 possessed the lowest K_i (500 ± 80 nM vs. WT) and 1 possessed the highest inactivation efficiency (k_inact/K_i = 0.21 ± 0.02 μM^-1s^-1). Electrospray ionization mass spectrometry revealed a single covalent adduct, suggesting the formation of an acyl-enzyme intermediate. X-ray structures of OXA-24 complexed to four inhibitors (2.0-2.6 Å) reveal there is formation of stable bicyclic aromatic intermediates with their carbonyl oxygen in the oxyanion hole. These data provide the first structural evidence that 6-alkylidene-2’-substituted penicillin sulfones are effective mechanism-based inactivators of class D β-lactamases. Their unique chemistry makes them developmental candidates. Mechanisms for class D hydrolysis and inhibition are discussed, and a pathway for the evolution of the BlaR1 sensor of Staphylococcus aureus to the class D β-lactamases is proposed.

INTRODUCTION

β-Lactamase enzymes (E.C. 3.5.2.6) are one of the most important mechanisms of resistance to β-lactam antibiotics in Gram-negative bacteria (1-10). As the prevalence of resistant pathogens is increasing in our health care institutions, this formidable challenge is creating significant concern among the medical and scientific community (2,11). Currently, there are more than 870 unique naturally occurring β-lactamases (2). Based on protein sequence similarities, four major β-lactamase classes are described (classes A, B, C, and D) (12-15). The class A, C, and D enzymes use serine as the active site, reactive nucleophile. Class B enzymes are metallo-β-lactamases that use one or two Zn⁺² atoms to catalyze the hydrolysis of the β-lactam bond (4). In general, serine β-lactamases inactivate β-lactams by following a two step reaction mechanism:

$$\mathrm{E}+\mathrm{S}\underset{k_{­1}}{\overset{k_1}{\rightleftharpoons }}\mathrm{E}:\mathrm{S}\stackrel{k_2}{\to }\mathrm{E}­\mathrm{S}\underset{{\mathrm{H}}_{2}O}{\overset{k_3}{\to }}\mathrm{E}+\mathrm{P}$$ Eq 1

Here, E represents the β-lactamase, S is the β-lactam substrate, E:S is the Henri-Michaelis complex, E-S, the acyl-enzyme, and P is the inactive product.

The most rapidly growing and diverse group of β-lactamases are the class D enzymes (16). As a group class D β-lactamases, also called “oxacillinases”, hydrolyze penicillins, cephalosporins, and carbapenems (Figure 1, Panel A). Unlike class A enzymes, class D β-lactamases are typically resistant to inhibition by clavulanate, sulbactam, and tazobactam. In order to preserve our current β-lactam antibiotics, potent inhibitors of the class D oxacillinases are urgently needed (17,18).

Figure 1.

Chemical structures of substrates, commercially available inhibitors (sulbactam, tazobactam and clavulanic acid), and 6-alkylidene-2’-substituted penicillanic acid sulfone compounds used in this study. The chemical structure of penicillin G is used as a model for all penicillins. In like manner, cephaloridine and imipenem are represented as models for cephalosporins and carbapenems.

Among the many experimental β-lactamase inhibitors that were developed, Chen et al. designed the 6Z-(α-pyridylmethylidene) penicillin sulfone series and showed them to be potent compounds with the ability to inactivate a wide spectrum of serine β-lactamases (19). Buynak, et al. later showed that the incorporation of a 2′ substitution (Figure 1, compound 1, Panel B) could both improve synergy and augment the inhibitory spectrum of this series (20-25). Based upon a consideration of the potency of 1 and related compounds against class A β-lactamases compared to clavulanate, sulbactam and tazobactam, a reaction mechanism was hypothesized.

To elucidate this mechanism, an analysis of SHV-1 inactivated by 1 was undertaken (26). A complex of 1 with SHV-1 (PDB 3D4F) showed that the inhibitor’s C6 (heteroaryl)alkylidene group plays a critical role in the formation of a planar bicyclic aromatic intermediate. Moreover, the pyridyl nitrogen of the C6 substituent nucleophilically adds to the intermediate imine, leading to the formation of a bicyclic aromatic species (indolizine). Furthermore, the crystal structure showed that the acyl-enzyme ester carbonyl of the intermediate is resonance stabilized by the conjugated π system and the carbonyl group of the intermediate is positioned outside the oxyanion hole. The decreased deacylation rates of this species are likely due to resonance stabilization, the location of the carbonyl at an increased distance from the hydrolytic water, and improperly positioned for nucleophilic attack by the enzyme’s backbone nitrogens.

A recent investigation involving the synthesis and microbiological evaluation of more than 100 substituted analogs of the general 6-(pyridylmethylidene)penam sulfone inhibitory subtype revealed that many of these derivatives, particularly those incorporating 2′ substitutions, were also potent submicromolar inhibitors of the class D OXA-24 carbapenemase (27). In addition, many showed synergy with carbapenems against resistant microorganisms. These findings are especially noteworthy as the commercially available inhibitors are ineffective against class D β-lactamases. A comparative susceptibility and kinetic study of OXA-1, -10, -14, -17 and 24 β-lactamases with 1 and related compounds established the potential efficacy of these inhibitors. This analysis was consistent with a branched pathway model for inhibition (Scheme 1 and ref. (28)) and served as a springboard for more in-depth analyses.

Scheme 1.

Based upon the analyses of the C2/3-substituted penicillin and cephalosporin sulfone series against OXA β-lactamases, the following model was proposed (28). In this model, E:I represents the formation of the pre-acylation complex and E-I, the acyl-enzyme species. The acyl-enzyme (E-I) can proceed to hydrolysis (E + P₁) or undergo rearrangement to a transiently inhibited species (E-I^T). The E-I^T intermediate may then return to E-I, proceed to hydrolysis (E + P₂), or form an inactivated acyl-enzyme (E-I*). The rate constants, k, describing each of these steps are also represented.

Here we present the design, synthesis and characterization of novel potent 6Z-(α-pyridylmethylidene)penam sulfone inhibitors (2, 3, 4, and 5) against OXA-24 β-lactamase. These inhibitors were complexed with OXA-24 (also known as OXA-40). A previous report by Santillana et al. showed that carbapenem substrate specificity is largely determined by a hydrophobic barrier that is established through an arrangement of the Tyr112 and Met223 side chains (29). Tyr112 and Met223 residues define a tunnel-like entrance to the active site. To test the contribution of these important residues to the inhibitor profile the Tyr112Ala, and Tyr112Ala/Met223Ala β-lactamases were also assayed in inhibition studies. We show for the first time that 6-alkylidene-2′-substituted penicillin sulfone inhibitors are effective mechanism-based inactivators for this challenging class of β-lactamases; their unique reaction chemistry makes then suitable lead compounds for further development.

MATERIALS AND METHODS

Chemical syntheses

The design, synthesis and evaluation of 1 were previously reported (25,26). Compounds 2, 3, 4, and 5 were synthesized as part of a drug discovery effort (27). The chemical structures of the compounds studied and details of each synthesis are illustrated in Figure 1 Panel B and Schemes 1, 2, and 3 (Supplementary Information, SI) and discussed below.

Scheme 2.

Proposed mechanism of OXA-24 β-lactamase inhibition by 5 based upon the intermediates hypothesized to be formed in the model illustrated in Scheme 2. Panel A and Panel B.

Scheme 3.

Proposed mechanism for the formation of the BlaR1 acyl-enzyme and requirement for additional proton

Genetic constructs and host strains

bla_OXA-24 (also named bla_OXA-40) was isolated from a strain of A. baumannii RYC 52763/97 (30). The wild type, WT, bla_OXA-24 and mutated genes were subcloned into pGEX-6p-1 (BamHI and EcoRI restriction sites) to generate a fusion protein between GST and the OXA-24 lacking the signal peptide. The recombinant β-lactamase was then purified to homogeneity using the GST Gene Fusion System (Amersham Pharmacia Biotech, Europe GmbH). The mature purified β-lactamases lacking the GST fusion protein appeared on sodium dodecyl sulphate-polyacrylamide gels as a band of approximately 29 kDa (≥95% purity).

bla_OXA-24 and the mutated bla_OXA-24 genes [bla_{OXA-24-Tyr112Ala}, bla_{OXA-24-Met223Ala}, and bla_{OXA-24-Tyr112Ala, Met223Ala}] were directionally subcloned into the pAT-RA plasmid (rifampin resistance) at the SmaI and EcoRI restriction sites under the control of the bla_CTX-M-14 β-lactamase gene promoter. Once the correct constructs were confirmed by DNA sequencing, the different plasmids were electroporated into the carbapenem-susceptible clinical strain, A. baumannii JC7/04 (29).

Antibiotic susceptibility

Antibiotic susceptibility profiles were determined in cation-adjusted Mueller-Hinton Broth by microdilution testing following Clinical Laboratory Standards Institute (CLSI) criteria (31). Five different inhibitor compounds, 1, 2, 3, 4, and 5 were tested for their capacity to inhibit A. baumannii strain JC7/04 that possessed the different bla_OXA-24 genes. The inhibitors were tested at two concentrations, 4 and 16 μg/mL, in the presence of either imipenem (Merck) or meropenem (AstraZeneca). As a comparator inhibitor, tazobactam was used (Wyeth) at two different concentrations, 4 and 16 μg/mL. MICs were determined in the presence of 50 μg/mL of rifampin (Sigma). The MICs reported are the result of three independent experiments.

Kinetic parameters

The half maximum inhibitory concentrations (IC₅₀s) were determined using two different approaches. In a first set of experiments, we used 3-[(3-carboxy-4-nitrophenyl)sulfanylmethyl]-8-oxo-7- [(2-thiophen-2-ylacetyl)amino]-5-thia-1-azabicyclo[4.2.0]oct-2-ene-2-carboxylic acid, CENTA (Figure 1, Panel A), (Calbiochem, EMD, San Diego, CA) at 25 μM as an indicator substrate (32). We incubated the different inhibitors and β-lactamases for 10 min (complete inactivation) at 37°C before measurement of the remaining enzymatic activity (32). The β-lactamases were used in the reactions at 0.13, 0.12, and 0.072 nM concentrations for OXA-24 WT, the OXA-24 doubly substituted enzyme, and OXA-24 Tyr112Ala, respectively. These determinations were performed at 25°C, in a Nicolete Evolution 300 spectrophotometer (Thermo Electron Corporation, Waltham, MA) with quartz cuvettes of optical path length 1 cm, (each determination was measured in triplicate).

In addition, we also used nitrocefin, NCF (Figure 1, Panel A) (100 μM) as the indicator substrate, incubating the enzymes and inhibitors at room temperature using 3.5 nM OXA-24 and 7 nM of each variant. Residual velocities were determined after 10 minutes. The data were plotted as 1/velocity (1/ν) as a function of inhibitor concentration (I), fitted to a linear equation, and the value of IC₅₀ determined by dividing the y-intercept by the slope of the line.

Steady state kinetic parameters were determined using an Agilent™ 8453 Diode Array spectrophotometer (33). The kinetic determinations were performed at room temperature in 50 mM sodium phosphate supplemented with a saturating concentration of sodium bicarbonate (20 mM) (34). First, the kinetic parameters, V_max and K_m, were obtained with non-linear least squares fit of the data (Henri-Michaelis equation) using Enzfitter™ (Biosoft Corporation, Ferguson, MO):

$$\nu =V_{max}[\mathrm{S}]/(K_{\mathrm{m}}+[\mathrm{S}])$$ Eq 2

We determined the K_i for the inhibitors by measuring initial steady state velocities in the presence of a constant concentration of enzyme, (E; 7 nM) with increasing concentrations of inhibitor competed against 100 μM of the indicator substrate nitrocefin (NCF). The competition assay between the inhibitor, I, and substrate, S, in the reaction can be represented by Scheme 1. Assuming this competitive mode of inhibition, initial velocity (ν₀) measurements immediately after mixing yield a K_i which closely approximates K_m, and is represented by the following equation:

$${\nu }_0=(V_{max}\ast [\mathrm{S}])/(K_{\mathrm{m}}\ast (1+\mathrm{I}/K_{\mathrm{i}}+[\mathrm{S}])$$ Eq 3

The data were plotted as 1/ν as a function of inhibitor concentration, fitted to a linear equation, and the value of K_i determined by dividing the y-intercept by the slope of the line. The K_i (observed) value was corrected to account for the affinity of NCF for the OXA β-lactamases (35).

$$K_{\mathrm{i}}\left(\mathrm{corrected}\right)=K_{\mathrm{i}}\left(\mathrm{observed}\right)/[1+([\mathrm{S}]/K_{\mathrm{m}\mathrm{NCF}})]$$ Eq 4

The first-order rate constant for enzyme and inhibitor complex inactivation, k_inact, was measured directly by monitoring the reaction time courses in the presence of inhibitors (I) 1, 2, 3, 4, and 5. A fixed concentration of enzyme (E; 7 nM), NCF, and increasing concentrations of I were used in each assay. The k_obs were determined using a non-linear least squares fit of the data to Equation 5 using Origin 7.5^®:

$$A=A_{\mathit{0}}+{\nu }_{f}t+({\nu }_{\mathit{0}}­{\nu }_f)[1­\exp (­k_{\mathrm{obs}}t)]/k_{\mathrm{obs}}$$ Eq 5

Here, A is absorbance, ν₀ (expressed in variation of absorbance per unit time) is initial velocity, ν_f is final velocity, and t is time. Each k_obs was plotted versus I and fit to determine k_inact (33,34).

Electrospray ionization Mass spectrometry (ESI-MS)

ESI-MS of the intact OXA β-lactamases inactivated by 1, 3, 4, and 5 was performed on an Applied Biosystems (Framingham, MA) Q-STAR XL quadrupole-time-of-flight mass spectrometer equipped with a nanospray source as described previously (33,34). Experiments were conducted by first desalting the reaction mixtures using C₁₈ ZipTips (Millipore, Bedford MA) according to the manufacturer’s protocol. The protein sample was then diluted with 50% acetonitrile/ 0.1% trifluoroacetic acid to a concentration of 10 μM. This protein solution was then infused into the mass spectrometer at a rate of 0.5 μL/min and data were collected for 2 min. Spectra were deconvoluted using the Applied Biosystems (Framingham, MA) Analyst program.

Protein purification and crystallization

From purified OXA-24 β-lactamase tetragonal crystals were grown by the hanging drop vapor diffusion method in a crystallization solution containing 0.1 M sodium acetate, 28% PEG 2000 MME buffered with 0.1M HEPES (pH 7.5). Diffraction-quality crystals were obtained by mixing equal volumes of the crystallization solution with protein at a concentration of 6 mg/ml. Bipyramidal shaped crystals grew in a period of 5-8 days reaching their maximal dimensions (0.15 × 0.15 × 0.10 mm). The crystals belong to space group P4₁2₁2 with one molecule in the asymmetric unit (Table 1). The first trials to co-crystallize the β-lactamase in complex with several of the selected inhibitors were unsuccessful, for these crystals the inhibitor was poorly defined in the electron density maps. To solve this, crystals were subjected to different soaking times while varying the concentration of the inhibitor, which was intended to stabilize the crystals and to minimize cracking. Final optimised conditions were determined at incubation times as short as 5 min in crystallization conditions containing also 3mM of the corresponding inhibitor. Furthermore, 3 was incubated with the enzyme for 15 min at 10 mM concentration to test time-dependent inhibition (3b).

Table 1.

Data Collection and Refinement Statistics

Data Collection
	Native	2	3	3b	5
Space Group	P41212	P41212	P41212	P41212	P41212
Cell dimensions [a, b, c (Å)]	102.4, 102.4, 84.1	102.4, 102.4, 84.2	102.8, 102.8, 83.1	103.1, 103.1, 85.8	102.5, 102.5, 84.5
Resolution (Å)a	45.8 – 1.9 (2.0 -1.9)	45.8 – 2.0 (2.11 – 2.0)	72.7 – 2.10 (2.21 – 2.10)	32.6 – 2.60 (2.67 – 2.6)	45.8 – 2.0 (2.11 - 2.0)
Rmerge (%)	7.3 (67.7)	7.6 (54.0)	10.1 (62.1)	11.0 (30.1)	7.8 (56.4)
I/σ(I)	9.6 (1.1)	9.0 (1.4)	6.8 (1.1)	12.1 (2.1)	9.2 (1.4)
Completeness (%)	100 (100)	99.9 (99.8)	99.0 (99.0)	93.4 (59.2)	99.9 (99.1)
Redundancy	9.6 (9.5)	9.6 (9.8)	7.3 (4.4)	6.0 (2.1)	9.5 (9.5)
Refinement
Resolution (Å)	45.8 – 1.97	45.8 – 2.0	54.7 – 2.10	32.6 – 2.60	45.8 – 2.0
Number of reflections	30,384	30,828	26,265	13,058	30,940
Rwork/Rfree (%)	19.7/24.4	22.2/25.7	23.2/25.5	17.4/21.1	22.1/24.8
No. atoms
Protein	1939	1939	1939	1958	1939
Ligand	0	27	29	29	29
PEG (SO4)	0	0	13	4	13
Solvent	158	112	79	105	116
B-factors
Protein	33.1	34.9	33.5	27.2	33.3
Ligand	0	67.5	49.6	55.8	65.9
PEG / SO4	0	0	46.7	81.1	60.0
Solvent	40.1	39.4	37.3	28.3	40.0
Rmsd deviations
Bond lengths (Å)	0.006	0.008	0.007	0.011	0.009
Bond angles (°)	1.2	1.3	1.2	1.39	1.5
PDB code	3G4P	3FV7	3FYZ	3MBZ	3FZC

^aHighest resolution shell is shown in parentheses.

Data collection, structure determination and refinement

Crystals were transferred into a crystallization solution containing 15% (v/v) PEG 400 for cryoprotection before immersion into liquid N₂ for data collection. X-ray diffraction data for 1, 2, 3, and 5 inhibitor complexes were collected at the European Synchrotron Radiation Facility (ESRF, Grenoble) beamlines ID14-EH1, ID14-EH2 and ID14-EH4 using single frozen crystals (100 K). Inhibitor 3b data was collected at beamline X-29 at National Synchrotron Light Source Brookhaven National Laboratory Upton, NY. Diffraction images were indexed and integrated with MOSFLM (36). Data for the 3 complex were processed using HKL2000 (37). Data scaling, merging and reduction were carried out with programs of the CCP4 suite (38). Relevant statistics are presented in Table 1.

The structure of native OXA-24, crystallized at pH 7.5, was determined by difference Fourier techniques using the protein atomic coordinates of the original OXA-24 β-lactamase crystallized at pH 4.5 (PDB entry 2JC7) (29). Carbamylation of Lys84 was clearly visible in the electron density maps and could be built with confidence using COOT (Emsley and Cowtan, 2004) (39). Moreover, this new crystal form shows the absence of the sulfate ion in the active site of the enzyme, making it a better target that could help address the structural studies of the selected inhibitors.

The model was subjected to several rounds of refinement with the program REFMAC (40) whereas interactive model building used COOT (39). Water molecules were modelled according to residual density profiles and geometrical requirements for hydrogen bonding. The crystallographic R-factor of the model is 19.7% for all unique reflections from 8.0 to 1.97 Å resolution (R_free = 24.4%).

The structures of OXA-24 in complex with four different penicillin sulfone based inhibitors were solved by Fourier synthesis employing the coordinates of the native enzyme. Refinement of these structures was carried out with CNS (41) and REFMAC (40). After rigid-body refinement and model fitting, the position of the inhibitors was clearly defined in the active binding site, covalently attached to the active serine Ser81, from the electron density maps. Topology and parameter values for each of the ligands, 1, 2, 3 and 5, were generated using the Dundee PRODRG2 server (42). Several rounds of refinement were combined with model rebuilding in COOT after inspection of electron density maps. All residues are in the most-favored and additionally allowed regions of the Ramachandran plot. A summary of refinement statistics is presented in Table 1. Structural figures were done using Chimera (43) or Pymol (44). Electrostatic potential surfaces were calculated with GRASP (45).

RESULTS AND DISCUSSION

Inhibitor Design and Chemical Syntheses

With the knowledge of the mechanism of inhibition depicted in Scheme 1 and previous results regarding structure activity relationships, SARs, due to modification of the 2′β position (25), we synthesized compounds 1-5 according to the following plan.

As shown in Scheme 1 (SI), the carboxylic acid and amino functionalities of the commercially available 6-aminopenicillanic acid (6-APA) were sequentially protected by treatment with diphenyldiazomethane and allyl chloroformate, respectively, and the sulfur was oxidized to the corresponding sulfoxide with mCPBA. Utilizing the method of Kamiya (46), the protected penam sulfoxide 6 was heated in the presence of mercaptobenzothiazole to generate an intermediate sulfenic acid, which was trapped in situ to generate the disulfide 7. The thiazolidine was regenerated by treatment with silver acetate in the presence of chloroacetic acid to stereoselectively produce the 2′β functionalized penam 8, together with the cepham 9. After deprotecting the C6 and C7 amines, this mixture was converted to the corresponding diazo compounds, then treated with a catalytic amount of rhodium octanoate in the presence of excess propylene oxide to generate the corresponding 6-oxopenicillinates and 7-oxocephalosporinates, 12 and 13, respectively (47). Reaction with α-pyridylmethylenetriphenylphosphorane selectively produced the olefins of the Z-geometry. Separation of the penam and cephem isomers was achieved by oxidation of the mixture with mCPBA, which quickly oxidizes the cephalosporin to the sulfone 17, while leaving most of the more hindered sulfur of the penicillin at the sulfoxide oxidation state 16. Subsequent oxidation of the penicillin sulfoxide 16 to the corresponding sulfone, removal of the chloroacetate protecting group, and activation of the 2′β alcohol as the p-nitrophenyl carbonate produced intermediate 20. Intermediate 20 was then reacted with the appropriate amine and the benzhydryl ester removed to produce the inhibitors 4 (compound 22) and 5 (compound 24).

As shown in Scheme 2 (SI) , the functionalized pyridyl moiety was prepared from commercially available 2,4-lutidine, which undergoes a selective oxidation of the C4 methyl group on treatment with potassium permanganate (48). Generation of the corresponding acid chloride and reaction with allyl alcohol produces ester 26, which was converted to the corresponding pyridine-N-oxide 27 on treatment with mCPBA. Treatment of this oxide with phosphorus oxychloride produced chloromethyl pyridine 28 (49), which was subsequently converted to ylide 29. Reaction of this ylide with the 6-oxopenicillinate 31 produced a benzyhydryl 6Z-(α-pyridylmethylidene) penicillinate 32, which was converted to the corresponding sulfone and deprotected to produce acid 34. This acid was then converted to the unstable acyl azide, which was immediately thermolyzed in methanol to effect the Curtius rearrangement to the isocyanate, which was then trapped to produce carbamate 35. Deprotection of the benzhydryl ester with TFA-anisole produced inhibitor 2.

Lastly, the “east-west” dual-functionalized inhibitor 3 was synthesized as shown in Scheme 3 (SI). Thus the functionalized 6-oxopenicillinate 12, (mixture with the corresponding cepham 13), was treated with ylide 29, to produce a mixture of 6-alkylidenepenam, 37, and 7-alkylidenecepham, 38, as shown. As before, treatment with 1.5 eq of mCPBA affected the oxidation to the corresponding cepham sulfone and, primarily, the penam sulfoxide, which were separable. Further oxidation of the sulfoxide 39 to the sulfone 41, followed by removal of the chloroacetate and the allyl ester protecting groups produced hydroxyacid 43. Treatment of this material with excess p-nitrophenyl chloroformate simultaneously converted the acid to the p-nitrophenyl ester and the alcohol to the p-nitrophenyl carbonate. Reaction with ammonia in dioxane produced the amidocarbamate intermediate 45, which was subsequently deprotected to produce inhibitor 3 (compound 46).

Microbiological Studies

In order to be an effective partner inhibitor for clinical use, the mechanism-based β-lactamase inactivator must effectively and rapidly penetrate the outer membrane of Gram-negative bacteria in sufficient concentrations to lower MICs into the susceptible range (28). To establish a comparison, we employed the carbapenems that are used clinically (imipenem or meropenem) with the inhibitors at two different concentrations, 4 and 16 μg/ml. We also used tazobactam at the same concentrations as a comparator β-lactamase inhibitor. In order to ensure an appropriate contrast, we also expressed each of the different constructs in an isogenic host, the carbapenem-susceptible A. baumannii JC7/04 strain (29). In contrast to laboratory strains of E. coli, this model system ensures a more realistic appraisal of the efficacy of the compounds against the pathogen containing the bla_OXA-24 (50).

Against A. baumannii JC7/04 without OXA-24 expressed, the meropenem MICs are 1.0 μg/ml, well within the susceptible range (31). In the genetic background where bla_OXA-24 in A. baumannii strain JC7/04 is expressed, high-level carbapenem resistance is observed (Table 2a and 2b, meropenem and imipenem MIC = 32 μg/ml). When the inhibitor tazobactam, at 4 μg/ml, was combined with meropenem or imipenem, we did not detect a reduction in MICs (only slight and no significant inhibition with imipenem). This is consistent with the clinical observation that β-lactam-tazobactam combinations are not effective against carbapenem resistant isolates (51,52). As each inhibitor possesses a β-lactam scaffold, we first tested each inhibitor without a partner antibiotic. Our results showed that 1-5 do not possess any intrinsic antibiotic activity against A. baumannii JC7/04 with pAT-RA or pAT-RA with bla_OXA-24 (MIC ≥ 64 μg/ml). With respect to the inhibitory activity of the compounds, 1 or 5 at 4 μg/ml combined with meropenem we showed a noticeable reduction in MICs (32 to 4 μg/ml). Overall, MICs decreased greater when the inhibitors were tested at a concentration of 16 μg/ml (Table 2b). This enhanced, “dose-dependent” effect was slightly more pronounced for meropenem than imipenem. Overall, 1 combined with imipenem or meropenem reduced MICs better than any other combination.

Table 2.

a. MIC values of the carbapenem-susceptible A. baumannii strain JC7/04 transformants

	MIC (μg /ml)a
	*pAT-RA	pAT-RA/OXA-24 (WT)	pAT-RA/OXA-24 (Y112A,M223A)
Meropenem	1	32	2
Meropenem/tazobactam	0.5	32	2
Meropenem/ 1	1	4	2
Meropenem/ 2	1	16	2
Meropenem/ 3	0.5	16	2
Meropenem/ 4	0.5	16	1
Meropenem/5	0.5	4	1
Imipenem	1	32	4
Imipenem/tazobactam	0.5	16	4
Imipenem/1	0.5	4	4
Imipenem/ 2	0.5	32	4
Imipenem/ 3	0.5	8	4
Imipenem/ 4	0.25	8	4
Imipenem/ 5	0.5	8	4
b. MIC values of the carbapenem-susceptible A. baumannii strain JC7/04 transformants
	MIC (μg /ml)b
	*pAT-RA	pAT-RA/OXA-24 (WT)	pAT-RA/OXA-24 (Y112A,M223A)
Meropenem	1	32	2
Meropenem/tazobactam	0.5	32	2
Meropenem/ 1	1	1	0.5
Meropenem/ 2	1	16	1
Meropenem/ 3	0.5	4	0.5
Meropenem/4	0.5	8	1
Meropenem/5	0.5	4	1
Imipenem	1	32	4
Imipenem/tazobactam	0.25	16	2
Imipenem/1	0.25	1	1
Imipenem/2	0.25	16	4
Imipenem/3	0.25	2	1
Imipenem/4	0.25	4	2
Imipenem/5	0.5	8	2

^aTazobactam, 1, 2, 3, 4, and 5 tested at 4μg/ml;

^*pAT-RA plasmid pAT-RA in A. baumannii without bla _OXA-24 gene. The MICs of tazobactam, 1, 2, 3, 4, 5 without meropenem or imipenem were > 64 μg/ml.

^bTazobactam, 1, 2, 3, 4, and 5 tested at 16 μg/ml;

^*pAT-RA plasmid pAT-RA in A. baumannii without bla _OXA-24 gene.

Table 3.

Kinetic Parameters of Inhibition

Compound		Ki (μM)	kinact (s-1)	kinact/Ki (μM-1s-1)
1	OXA-24 WT	0.70 ± 0.05	0.15 ± 0.01	0.21 ± 0.02
2	OXA-24 WT	2.3 ± 0.3	0.05 ± 0.01	0.022 ± 0.005
3	OXA-24 WT	1.6 ± 0.3	0.25 ± 0.02	0.16 ± 0.02
4	OXA-24 WT	0.50 ± 0.08	0.09 ± 0.01	0.18 ± 0.03
5	OXA-24 WT	4.1 ± 0.6	0.08 ± 0.01	0.020 ± 0.003

Tazobactam K_i was determed by Drawz, et al and was 271 ± 37 μM (ref).

We next measured the activity of the carbapenems against the variants possessing the Tyr112Ala, Met223Ala substitution (Table 2b). As is shown by MICs, imipenem and meropenem resistance is reduced for the strain possessing the doubly substituted enzyme (WT is 32 μg/ml and the doubly substituted enzyme = 2 or 4 μg/ml). This supports the observation that the two residues, Tyr112Ala and Met223Ala, also play a critical role in resistance to carbapenems (29). The MICs were slightly reduced when each of the inhibitors was combined with meropenem or imipenenem against the doubly substituted enzyme (especially at 16 μg/ml).

Kinetic parameters

In the steady state experiments summarized in (Table 1 (SI), the OXA-24 β-lactamase and variants studied were purified to greater than 95% homogeneity. The WT enzyme, OXA-24 β-lactamase, hydrolyzed NCF with k_cat/K_m values 19.2 ± 2.6 μM^-1s^-1 (Table 1, SI). This robust activity was similar to the hydrolysis of NCF by OXA-1 (28,33) and ranks with the high catalytic efficiency demonstrated by class A β-lactamases towards penicillins and certain class C enzymes towards cephalosporins (53-55). Categorically, OXA enzymes are regarded as carboxy-penicillinases and amino-penicillinases (16). Since OXA-24 is characterized as a carbapenemase, we next measured the catalysis of imipenem as a reference carbapenem. In keeping with susceptibility testing (MICs = 32 μg/ml), the k_cat/K_m value was 1.7 ± 0.2 μM^-1s^-1 (Table 1, SI). Interestingly, the k_cat of imipenem was much lower when compared to the k_cat of NCF. This finding is consistent with the observation made by Queenan and Bush that OXA carbapenemases have a high apparent affinity for carbapenems, but a low turnover (10).

The single (OXA-24Tyr112Ala) and doubly substituted enzymes (OXA-24Tyr112Ala, Met223Ala) demonstrated less robust activity (14% activity of the WT vs. NCF; 14-24% of the WT vs. imipenem). Each of the variants showed higher K_m values for NCF (OXA-24 Tyr112Ala, K_m = 238 μM and OXA-24 Tyr112Ala Met 223Ala K_m = 143 μM).

In order to enrich our understanding of the mechanism of inhibition, two approaches were used to assess the efficacy of the inhibitors against OXA-24 and the OXA-24 variants. As discussed previously, the appropriate biochemical correlates that translate into effective β-lactamase inhibition in the clinical setting are complex and multi-factorial (cell penetration, pharmacodynamics, pharmacokinetics, etc.) (17). These considerations are fundamental to the assessment of each inhibitor as OXA-24 is resistant to the commercially available inhibitors and is expressed in the Acinetobacter spp. background.

To begin, we first measured IC₅₀s at 10 minutes using CENTA, as this parameter informs us of the relative effectiveness of an inhibitor. The IC₅₀ values for all four inhibitors against the OXA-24 β-lactamase ranged from 127 ± 42 nM (1) to 237 ± 7 nM (5) (Table 2a, SI). We also noted a slight increase in the IC₅₀ values with respect to the four inhibitors when tested against the singly and doubly substituted enzyme, but this increase was not enough to confer inhibitor resistance (≤ 1 μM).

In a similar manner to CENTA, we determined IC₅₀s with NCF. As the chemical properties and affinities of CENTA are different than NCF (both are indicator substrates), this served as a confirmation of the affinities of the compounds for OXA-24 and variants. As is evident from the data in Table 2b (SI), by using NCF we find the inhibitors demonstrated a 10 fold greater affinity (lower IC₅₀s) than with CENTA.

To more precisely identify the correlates of inactivation and inhibition, we next determined the K_i and the inactivation efficiency (k_inact/K_i) of each inhibitor (Table 3 and Table 3 (SI)). We examined the activity of each of these inhibitors against the WT enzyme and the two variants.

Notably, 4 and 1 showed the lowest K_i for WT followed by 3, 2, and 5. Compound 1 possessed the highest inactivation efficiency (k_inact/K_i = 0.21 ± 0.02 μM^-1s^-1) (Table 3). With regards to the two variant OXA β-lactamases, we see for the five inhibitors an overall lowering of inactivation efficiencies (k_inact/ K_i) due to decreases in k_inact and increases in K_i (Table 3, SI).

We note that there is a difference between the kinetic parameters obtained measuring IC₅₀ and K_i measurements vs. microbiological values. We reconcile this difference by positing that the ability and rate of each of the inhibitors to penetrate across the outer membrane of this strain of A. baumannii may be different (50). Factors such as permeability coefficients, diffusion rates, presence or absence of specific porins may play an important role here and merit further studies (56,57). The design of 1 attempts to enhance transport across the cell membrane (26).

ESI-MS

To establish the nature of the inactivation products, ESI-MS was performed with a Q-Star quadrupole time-of-flight mass spectrometer equipped with a nanospray source. We inactivated OXA-24 β-lactamase with each of the inhibitors. The incubation (inactivation) time was 15 min (900 s). The deconvoluted spectra are presented (Figure 1, SI) and our results are summarized in Table 4 (SI). The ESI-MS measurements (29,071 ± 3 amu) were in agreement with the theoretical mass of the OXA-24 β-lactamase, which is 29,073 (this includes the five additional amino acids at the N-terminus as a result of the cloning procedure). The preparative method did not permit us to identify a mass increase consistent with the addition of a CO₂ group (carboxylation at Lys84) to the β-lactamase.

Covalent attachment of each inhibitor to the OXA-24 β-lactamase and variant enzymes was demonstrated in each of the spectra. During the time period studied, we did not find evidence of the fragmentation of the inhibitors, as was seen in the inactivation of TEM-1, CMY-2, SHV-1, the Arg244Ser variant of SHV-1, and the Ser130Gly variant of SHV-1 with tazobactam and clavulanate (58-64). These observations were further rationalized after examination of the mechanisms of inactivation.

Crystallography

We next solved the crystal structures of OXA-24 complexed with different penicillin sulfone based inhibitors (1, 2, 3 and 5). This offered us an unprecedented opportunity to compare the inactivation mechanisms of different inhibitors of the same chemical series with different K_is. Structures were solved by Fourier synthesis employing the coordinates of the native enzyme. After rigid-body refinement and model fitting, the position of the inhibitors was clearly defined in the active site covalently attached to Ser81. The overall fold of the complexes were similar to the native OXA-24 structure (29).

When complexed to inhibitors, the conformations of the active site are very similar to that observed in the native enzyme, although several changes can be appreciated. The side chains of Tyr112 and Arg261 re-orient slightly (~ 0.6 Å) to accommodate the sulfinic and carboxylate groups of the inhibitors (Figure 2).

Figure 2.

Stereoview of the superposition at the active binding site of native OXA-24/40 (gray) and 1 (yellow), 2 (cyan), 3 (orange), 5 (green) complexes. The secondary structure of the enzyme is in gray. For clarity, only side-chain residues implicated in binding are represented. In the active site major changes are not observed upon inhibitor binding, but several modifications are seen in active site residues directly involved in inhibitor accommodation.

The initial omit electron density maps strongly indicated the formation of an acyl-enzyme intermediate containing a bicyclic aromatic ring system composed of two fused aromatic rings, a five-membered ring fused to a six-membered pyridine ring (Figures 3A-E). The electron density and structures of 3 soaked at different time points (3, 3b, Figures 3A, B, D) indicate that the bicyclic inhibitor intermediate is quite stable during the 5-15 min time period. This aromatic indolizine system points towards the solvent region and sterically blocks the entry to the catalytic cleft on the left side of the tunnel-like cavity. Contrary to what was found in SHV-1 and SHV-2 class A β-lactamases complexed with 1 (26), the carbonyl oxygen of the intermediates of 2, 3, and 5 is located in the oxyanion hole forming hydrogen bonds to the backbone amide of Trp121 (2.9 Å), in a similar manner to that found in the structure of the class C GC1 β-lactamase with the reaction product of the sulfone DVR-II-41S (65). The side chains of the ring-derived system extending from C2 in the three inhibitors also show other significant contacts with the enzyme.

Figure 3.

Binding mode of the substituted penicillin sulfone inhibitors in the active site of OXA-24 b-lactamase. (A) initial F_o-F_c omit maps for 3b contoured at 3.0 σ. Final 2F_o-F_c electron density maps for (B) 3b at 2.6 Å resolution. (C) 3 at 2.1 Å resolution in stereoview, (D) 2 at 2.0 Å resolution and (E) 5 at 2.0 Å resolution. Contour levels are at 1.0 σ. The maps show a clear density for an intermediate containing an indolizine moiety as a result of the formation of a five-membered ring fused to the pyridine ring extending from the 6-alkylidene substituted penicillin sulfone inhibitors.

We stress that the network of interactions which clearly contributes to the stability of the intermediate is strictly conserved for the inhibitors 3, 2, and 5 (Figures 4A-C). The sulfinate anion is in close proximity to the guanidinium group of Arg261 establishing strong salt-bridge interactions (2.67 Å and 3.02 Å). Besides this electrostatic interaction, OXA-24 still also makes strong hydrogen bonds to the hydroxyl groups of Ser128 (3.04 Å) and Ser219 (2.52 Å), two highly conserved residues at the active site. The orientation adopted by the sulfinate anion emulates the positioning of the carboxylate moiety of the antibiotic, as was reported in other acyl-enzyme complexes with meropenem (66,67). As a consequence of this orientation the carboxylate group folds back over the tunnel establishing a strong hydrogen bond with the hydroxyl group of Tyr112 (2.8 Å) one of the key residues that conform the tunnel-like entrance to the active site in OXA-24 (Figures 2A and 2B, SI) (29). Apart from this common interaction network, the remaining contacts between the intermediate and the enzyme vary depending on the chemical nature of the inhibitor. Thus, the effect of the substituents at the pyridylmethylidene moiety in the stabilization of the acyl-enzyme intermediate can only be described for compounds 2 and 3. Whereas the carbamoyl group of compound 3 is in close contact with the hydroxyl group of Thr111 (3.06 Å), the methoxycarbamoyl substituent on compound 2 has not revealed additional interactions with the enzyme (Figures 4A and B).

Figure 4.

Active site and inhibitor interactions. Detailed interactions in the active site of OXA-24 with (A) 2, (B) 3 and (C) 5. The indolizine aromatic ring is visibly anchored into the tunnel-like cavity of the binding site through its conjugated acyl group covalently attached to the catalytic serine residue Ser81. The secondary structure of the enzyme is in blue. The side chains of the ligand-binding residues and inhibitors are represented in ball-and-stick mode. Selected interacting residues are labelled and hydrogen bonds are indicated by dotted lines.

On the other hand, the more complex tailored substituents at position C2 of the penicillin sulfone ring do not show favorable interactions to provide an additional stabilization of the intermediate with exception of the 3,4-(dihydroxyphenyl)acetate of 1. Unfortunately, the electron density maps for the 1 inhibitor do not show a clearly defined position for the catechol moiety of the inhibitor and also the population of the central core is not full, perhaps indicating disorder or partial occupation of the site (data not shown). However, it can be said that the bulky group at C2 runs almost parallel to the right side of the active site by stacking the catechol moiety with various residues on the hydrophobic surface of the enzyme (Figure 2A, SI). Stacking of 1’s catechol moiety has also been observed in its complex with SHV-1 (26).

The combined results imply that the stable acyl-enzyme complexes from the 6-alkylidene-2’-substituted penicillin sulfone inhibitors are formed by a sequence of events in which initial nucleophilic attack of Ser81 O^γ on the carbonyl of the penicillin ring releases the β-lactam nitrogen lone pair, thus enabling the opening of the neighboring sulfone ring. Another rearrangement provided by the pyridyl nitrogen bonded to the former C5 of the resultant imine leads to the formation of the crystallographic observed indolizine system.

In each of the inhibitor complexes, the π system of the acyl-enzyme carbonyl is orthogonal to the π system of the bicylic indolizine. Thus, it is clear that the hydrolytic stability of the covalent ester linkage is not due to a resonance interaction with the nitrogens (i.e. β-aminoacrylate or ‘enamine’) or due to interaction of the ester carbonyl with the aromatic system. What is stabilizing these acyl-enzymes toward hydrolysis? Notably absent in all structures is a hydrolytic water molecule proximal to both the carboxylated lysine and to the acyl-enzyme carbonyl carbon. It has been observed that, unlike the corresponding class A β-lactamases, a crystallographically observable hydrolytic water molecule is not seen in the class D apoenzymes, and it is assumed that such a water must diffuse into the site after formation of the acyl-enzyme (68). However, in the case of these inhibitors, the active site is occupied with the conformationally rigid bicyclic indolizine moiety, prospectively preventing the entry of extraneous water, and thus stabilizing the acyl-enzyme (Figures 2B and 2C, SI).

Based upon previous insights proposed by Golemi et al., a likely mechanism for the inactivation of OXA-24 β-lactamase is shown in Scheme 2 (69). An interesting feature of the mechanism is that, relative to the normal hydrolytic process (Panel B), the inhibitor provides the enzyme with one extra proton, which is lost from C5 of the inhibitor in order to achieve aromaticity (Y → Z, Panel A). This would then leave the carboxylated lysine as its conjugate acid and thus unable to activate water for hydrolysis of the acyl-enzyme. Note that, in the case of the penam substrate (bottom), the carbamic acid can be deprotonated by the proximal amine (previously N4 of the penam), while, in the case of the inhibitor, N4 is rendered significantly less basic due to its conjugation with the indolizine and interaction with the acyl-enzyme carbonyl.

This mechanism also illustrates the ability of the OXA-24 carboxylated lysine to repetitively cycle between its conjugate acid and base forms and acyl- and apo state, thus differentiating these catalytic class D β-lactamases from the highly homologous sensor domain of the BlaR1 protein from Staphylococcus aureus. In the case of BlaR1, the formation of the acyl-enzyme (and consequent formation of the carboxylysine conjugate acid) results in decarboxylation, thus removing the proximal base and fixing the sensor in the acylated (‘on’) state as shown in Scheme 3 (70). Through an unknown mechanism, this acylated form results in a conformational change of the transmembrane domain and relays a signal to the cytoplasm leading to the proteolytic degradation of a repressor protein, thus derepressing the blaZ gene and leading to production of more β-lactamase. In contrast to the catalytic action of the class D β-lactamases, the BlaR1 protein is irreversibly acylated by penicillins, presumably as a consequence of the lysine decarboxylation which occurs upon acylation (and formation of the conjugate acid of the carbamic acid) (71,72).

Ab initio quantum chemical studies have shown that the activation energy for the decarboxylation of carbamic acids is lowered by 44 kcal mole^-1 by the presence of one molecule of water in the transition state as shown in Scheme 3 (73). As shown in Figure 5, comparison of the respective active sites reveals a key structural difference between the BlaR1 sensor and class D β-lactamases is the presence of a highly conserved hydrophobic residue at the β-lactamase position 130 (OXA-24 numbering, Val or Ile, corresponds to position 120 using DBL consensus numbering), in contrast to a hydrophilic residue (Asn in case of S. aureus, or Thr, in the case of Bacillus licheniformis (74)) at corresponding BlaR1 position. Additionally, as shown, two proximal water molecules were found in the 1XA1 BlaR1 apoenzyme prospectively providing the precise water molecule involved in this decarboxylation (75).

Figure 5.

Comparison of the active site carboxylysine environment of OXA-24 (left, 3G4P) with BlaR1 (right, 1XA1).

CONCLUSIONS

OXA carbapenemases are becoming one of the most clinically important resistance determinants found in A. baumannii (16). Most clinical isolates of A. baumannii are resistant to narrow-spectrum and extended-spectrum cephalosporins due to the expression of the chromosomally-encoded AmpC β-lactamase, called ADC, Acinetobacter Derived Cephalosporinase (ADC) (51,76). An examination of the current clinical experience shows that OXA-23, -24, -48, -51, -69, and -72 seem to be very prevalent among outbreaks of carbapenem resistant A. baumannii (16). At present, there are only two carbapenem hydrolyzing class D enzymes (OXA-48 and OXA-24) crystal structures that are available for study.

As shown by Santillana et al. the crystal structure of OXA-24 β-lactamase revealed several unique features (29). In contrast, the structure of OXA-48 at 1.9 Å was overall similar to the crystal structure of OXA-10, a class D β-lactamase that does not hydrolyze imipenem (77). The present work captures for the first time the crystal structure of novel 6-alkylidene-2’-substituted penicillanic acid sulfones as active site inhibitors of OXA-24. This work significantly extends the previous studies with 1 in class A enzymes to a class of β-lactamases that has escaped inhibition by commercial inhibitors (26).

Five important insights are obtained from this analysis. Firstly, with all inhibitors, we see the formation of a stable un-fragmented adduct with the carbonyl oxygen positioned deep in the oxyanion hole. Normally, one would expect that this conformation would facilitate hydrolysis as opposed to inhibition. Yet, the unique reaction chemistry followed by these C6 substituted compounds, formation of a bicyclic aromatic ring that may impede approach of an attacking water molecule. This adds an additional enhancing feature to these molecules (a theme reminiscent of DVR-II-41S, see reference 65). This is in stark contrast to what was seen with SHV-1 β-lactamase (26) because of the positioning of the intermediate carbonyl oxygen.

Secondly, the negatively charged sulfinate group on each of these compounds is positioned to interact via a stabilizing salt bridge with the residue Arg261. The strong H bonds to Ser128 and 219 also add to this stability.

Thirdly, the carboxylate of the inhibitor, folding over to make an H bond to Tyr112, further enhances the stability of the complex.

Fourthly, a new and previously unappreciated feature of these penicillin sulfones may be their capability to provide an extra proton to the active site thus generating the conjugate acid of the carboxylysine, and interrupting the catalytic process.

Lastly, a comparison of the environment surrounding the carboxylysine of the class D β-lactamases with that of the highly homologous BlaR1 sensor protein reveals that the class D β-lactamases have a significantly more hydrophobic environment. The more hydrophilic environment of the carboxylysine of BlaR1 may provide an explanation for its rapid decarboxylation, since it is known that the transition state for decarboxylation of carbamic acids is significantly lowered by incorporating one molecule of water in the process. Relative to ongoing catalysis, this decarboxylation also requires at least one additional proton thus potentially providing a mechanism to induce the large conformational shift needed for intracellular signaling.

Taken together, our findings point to an effective strategy to inhibit not only this OXA carbapenemase but also other serine based enzymes that inactivate β-lactam antibiotics, and provide several intriguing hypotheses to be explored with respect to the highly homologous BlaR1 sensor protein. A comparative analysis against other OXA carbapenem hydrolyzing and extended-spectrum OXA β-lactamases enzymes is warranted as the pathways to evolution of these enzymes are different (77). The quest to find inhibitors active against a wide range of carbapenem hydrolyzing enzymes will remain a persistent challenge as the diversity in class D grows.