Exploring the potential of boronic acids as inhibitors of OXA‐24/40 β‐lactamase

Abstract

β‐lactam antibiotics are crucial to the management of bacterial infections in the medical community. Due to overuse and misuse, clinically significant bacteria are now resistant to many commercially available antibiotics. The most widespread resistance mechanism to β‐lactams is the expression of β‐lactamase enzymes. To overcome β‐lactamase mediated resistance, inhibitors were designed to inactivate these enzymes. However, current inhibitors (clavulanic acid, tazobactam, and sulbactam) for β‐lactamases also contain the characteristic β‐lactam ring, making them susceptible to resistance mechanisms employed by bacteria. This presents a critical need for novel, non‐β‐lactam inhibitors that can circumvent these resistance mechanisms. The carbapenem‐hydrolyzing class D β‐lactamases (CHDLs) are of particular concern, given that they efficiently hydrolyze potent carbapenem antibiotics. Unfortunately, these enzymes are not inhibited by clinically available β‐lactamase inhibitors, nor are they effectively inhibited by the newest, non‐β‐lactam inhibitor, avibactam. Boronic acids are known transition state analog inhibitors of class A and C β‐lactamases, and are not extensively characterized as inhibitors of class D β‐lactamases. Importantly, boronic acids provide a novel way to potentially inhibit class D β‐lactamases. Sixteen boronic acids were selected and tested for inhibition of the CHDL OXA‐24/40. Several compounds were identified as effective inhibitors of OXA‐24/40, with K_i values as low as 5 μM. The X‐ray crystal structures of OXA‐24/40 in complex with BA3, BA4, BA8, and BA16 were determined and revealed the importance of interactions with hydrophobic residues Tyr112 and Trp115. These boronic acids serve as progenitors in optimization efforts of a novel series of inhibitors for class D β‐lactamases.

Short abstract

PDB Code(s): 5TG4; 5TG5; 5TG6; 5TG7;

Abbreviations

CENTA

cephalothin nitrothiobenzoic acid

CHDL

carbapenem hydrolyzing class D β‐lactamase

MDR

multidrug resistant

MIC

minimum inhibitory concentration

PBP

penicillin binding protein

Introduction

β‐lactam antibiotics are heavily relied upon by clinicians to treat bacterial infections. However, due to over‐prescription and misuse, bacterial resistance is threatening their efficacy. An enormous clinical threat, Acinetobacter baumannii has been termed a multidrug‐resistant (MDR) strain of bacteria due to its ability to resist cephalosporins and, most recently and concerning, carbapenems.1, 2, 3 Outbreaks of MDR Acinetobacter baumannii are reported worldwide, primarily as nosocomial infections, but also affecting injured United States service members in Afghanistan and the Iraq‐Kuwait region.4 Acinetobacter strains employ many mechanisms of resistance, including mutations in penicillin binding protein, decreased membrane permeability, and expression of β‐lactamase enzymes, which hydrolyze the defining β‐lactam ring [Fig. 1(A)] of the antibiotic.5, 6 β‐lactamases are grouped into four different classes: A, B, C, and D, all of which (besides class B metalloenzymes) use a serine‐based mechanism for lactam hydrolysis.7 Acinetobacter spp. strains have been found to express many clinically significant and threatening β‐lactamases, including the class A TEM‐1 and TEM‐2, class C ACE‐1, ARI‐1, and ADC‐7, and the class D OXA‐1, OXA‐23, and OXA‐24/40.6, 8, 9 Over time, Acinetobacter strains have evolved to possess a threatening militia of β‐lactamases.

Figure 1.

Structures of β‐lactamase ligands. A: Penicillin, a β‐lactam antibiotic. β‐lactam ring is highlighted in red. B: Clavulanic acid, a β‐lactamase inhibitor. C: Boronic acid inhibitor. D: Boronic acids are reversible, competitive transition state analog inhibitors. Boronic acids inhibit by binding to the catalytic serine through a reversible, covalent bond.

One way to combat β‐lactamase‐mediated resistance is through the use of β‐lactamase inhibitors.10 These inhibitors, such as clavulanic acid, sulbactam, and tazobactam [Fig. 1(B)], are used clinically in combination with a β‐lactam antibiotic to treat resistant bacterial infections. However, these inhibitors share the same β‐lactam core structure found in the β‐lactam antibiotics. Bacteria rapidly evolve resistance to these structurally similar molecules by recruiting or modifying pre‐existing mechanisms. The overuse of β‐lactams has promoted the spread of these mechanisms to formerly susceptible strains of bacteria.

To date, almost 500 known class D β‐lactamases, or OXAs, are identified (http://www.lahey.org/studies/), and these enzymes are responsible for much of the β‐lactam resistance in Acinetobacter spp. Unfortunately, the class D β‐lactamases are generally not inhibited by these current clinical β‐lactamase inhibitors.11 Even more concerning is a subgroup of the class D enzymes that efficiently hydrolyze the newer, more potent carbapenem antibiotics, such as doripenem. These carbapenem‐hydrolyzing class D β‐lactamases (CHDLs), of which Acinetobacter baumannii OXA‐24/40 is a representative member6, 12 are a clinically important target to inhibit.

The discovery of a novel non‐β‐lactam inhibitor is crucial for maintaining the efficacy of β‐lactam antibiotics. Several non‐β‐lactam based inhibitors have previously been studied, including boronic acids,13 phosphonates,14 hydroxamates,15 and diazabicyclooctanones.10, 16 The feasibility of this type of therapy has been realized with the FDA approval in 2015 of Avycaz®, which combines the expanded‐spectrum cephalosporin ceftazidime with the diazabicyclooctanone inhibitor, avibactam. Avibactam has also been reported to inhibit the narrow‐spectrum class D enzyme OXA‐10 and the CHDL OXA‐48.17, 18 Additionally, a cyclic boronic acid, vaborbactam (formerly RPX7009), has been shown to inhibit certain class A, C, and D β‐lactamases. Currently, vaborbactam, in combination with the β‐lactam meropenem, is demonstrating success in clinical trials as Carbavance®.19

Boronic acids have long been known to inhibit both class A and C β‐lactamases with K_i values in the nM range [Fig. 1(C)].20, 21 Inhibition of class D β‐lactamases, such as OXA‐24/40, by boronic acids is far less characterized, with only several being reported [Fig. 1(D)].22, 23, 24 An intriguing approach to inhibiting β‐lactamases, boronic acids act as competitive inhibitors, forming a tetrahedral intermediate by binding to the catalytic serine through a reversible, dative covalent bond. The bound inhibitor mimics the tetrahedral structure of the high energy intermediate formed during the mechanism of β‐lactam hydrolysis.

To explore the potential of boronic acids as class D inhibitors, a panel of commercially available boronic acids were selected and assayed for inhibition against OXA‐24/40. Chosen compounds displayed a variety of functional groups with which to search for new scaffolds to inhibit OXA 24/40. Several of the boronic acids demonstrated inhibition of OXA‐24/40, with K_i values in the clinically meaningful range. The X‐ray crystal structures of OXA‐24/40 in complex with four boronic acids were determined to atomic resolution and provide a better understanding of the important binding determinants between OXA‐24/40 and boronic acid inhibitors. Ultimately, boronic acids present intriguing, non‐β‐lactam based inhibition of the CHDL OXA‐24/40 and potentially other class D enzymes.

Results

Kinetics

Boronic acids are known to inhibit class A and C β‐lactamases with high affinity.20, 21, 25, 26, 27 The covalent bond between the boron of the inhibitor and the catalytic serine contributes significantly to the binding affinity, but this interaction alone is not sufficient for high affinity to the β‐lactamase. The addition of functional groups beyond the boronic acid group can also impact binding affinity, with many enhanced affinity inhibitors making extensive hydrogen bonding interactions within the active sites, as observed in the structures of class A and C enzymes in complex with boronic acids.28 In contrast to classes A and C, the class D β‐lactamases, including the CHDL OXA‐24/40, possess mostly non‐polar active sites.29 The lack of polar “hot spots” is a challenge to designing soluble, high affinity boronic acid inhibitors to specifically complement the OXA‐24/40 active site.

Sixteen commercially available boronic acids that displayed a wide variety of functional groups, as well as offered the potential to complement a hydrophobic active site, were selected for kinetic analysis. The most hydrophobic boronic acids (BA5, BA7, and BA9) were not soluble and could not be tested for inhibition of OXA‐24/40.

The thirteen remaining compounds were initially screened for inhibition of OXA‐24/40 in kinetic assays. The K_i values ranged from as low as 28 μM to >2 mM (Table 1). In general, arylboronic acids with large substituents para to the boronic acid group were not good inhibitors of OXA‐24/40 (BA6, BA8, BA11, BA12, BA13). All had K_i values between 288 μM to 2 mM, with the exception of BA8, which had a somewhat more encouraging K_i of 97 μM.

Table 1.

Plate Reader Kinetic Data for OXA‐24/40 with Boronic Acid Inhibitors

Identity	Structure	Ki (μM)	Identity	Structure	Ki (μM)
BA1		>2000	BA9		NM
BA2		353	BA10		243
BA3		29.8	BA11		>2000
BA4		28.3 (5.03 ± 0.64)a	BA12		913
BA5		NMb	BA13		681
BA6		288	BA14		309
BA7		NM	BA15		39.5
BA8		97.3	BA16		475 (206.0 ± 58.8)

^a K_i values in parentheses were determined using data measured on an Agilent Cary‐60.

^bNM, not measured.

Meta substituted compounds appeared to be better accommodated in the OXA‐24/40 active site with the three best inhibitors of OXA‐24/40 containing substituents at this position (BA3, BA4, and BA15). Each of these compounds had K_i values ranging from 30 to 40 μM (Table 1). BA3 (K_i 29.8 μM) is reminiscent of the third generation cephalosporin, cephalothin. In addition to an amide group that is similar to the one found in the R1 side chains of β‐lactams, BA3 contains a furan in place of the thiophene ring found in cephalothin. Interestingly, the amide group in BA3 is flipped (i.e., carbonyl proximal to the aryl ring) in orientation relative to the R1 amide of β‐lactams. BA15 (K_i 39.5 μM) contains two small substituents (a nitro and carboxylate group), both meta to the boronic acid and both charged. The low K_i value was somewhat surprising given the presence of charges and its small size.

Overall, of the 13 compounds tested, BA4 was the best inhibitor identified in our initial screens (K_i 28.3 μM). Interestingly, BA4 is a benzoxaborole, distinguishing it from the rest of the arylboronic acids tested. BA16 contains exactly the same substituent as BA4 but is the arylboronic acid version of it. BA16 bound ∼17‐fold worse to OXA‐24/40, with a K_i of 475 μM.

Given the promising initial results of BA4 in the screen, a full kinetic analysis was performed on BA4 and BA16 for comparison and confirmation (Fig. 2). The K_i value of BA4 was determined to be 5.03 μM and that of BA16 was 206 μM (Table 1). The same trend observed on the plate reader was confirmed in the full analysis (i.e. BA4 binds better than BA16 to OXA24/40). However, the full analysis suggests that BA4 binds ∼40‐fold better than BA16, suggesting a key role for the benzoxaborole in OXA‐24/40 binding.

Figure 2.

Kinetic characterization of OXA‐24/40 inhibition by the benzoxaborole BA4. Data measured on an Agilent Cary‐60 UV–vis spectrophotometer.

Structural analysis

To study the structural basis for the observed inhibition, the X‐ray crystal structures of OXA‐24/40 in complex with several of the boronic acids (BA3, BA4, BA8, and BA16) were determined to resolutions ranging from 2.28 to 1.44 Å (Table 2). All four of the complexes crystallized in the same space group (P 4₁2₁2), with similar cell parameters, and each complex contained one monomer in the asymmetric unit. Analysis of the OXA‐24/40 inhibitor complexes indicated quality models. Using MolProbity, the Ramachandran plots revealed that 96.3% of all residues in the BA3 complex were in the favored region, 97.7% in the BA4 complex, 98.4% in the BA8 complex, and 98.0% in the BA16 complex. For all complexes, 100% of all residues were in the allowed regions of the Ramachandran plots, with no outliers.30

Table 2.

Crystallographic Summary for OXA‐24/40/Boronic Acid Complexes

	OXA‐24/BA3	OXA‐24/BA4	OXA‐24/BA8	OXA‐24/BA16
Cell constants (Å; °)	a = b = 103.08	a = b = 102.42	a = b = 102.97	a = b = 102.35
	c = 86.44	c = 85.05	c = 86.73	c = 85.44
	α = β = γ = 90	α = β = γ = 90	α = γ = 90	α = β = γ = 90
Space group	P 41 21 2	P 41 21 2	P 41 21 2	P 41 21 2
Resolution (Å)	103.08–2.28 (2.290–2.282)a	85.05–1.78 (1.782–1.776)	102.97–1.75 (1.756–1.750)	102.35–1.44 (1.446–1.441)
Unique reflections	21,572 (206)	41,363 (414)	45,159 (460)	82,068 (805)
Total observations	170,474 (1685)	300,064 (3085)	328,802 (3528)	654,438 (5183)
R merge (%)	9.9 (87.0)	9.4 (92.1)	8.7 (70.7)	6.2 (95.7)
Completeness (%)b	99.0 (100)	93.7 (96.3)	95.0 (96.2)	100.0 (100.0)
<I/σ I>	15.0 (2.3)	12.4 (2.2)	12.0 (2.2)	19.7 (2.0)
Resolution range for refinement (Å)	103.08–2.28	72.42–1.78	72.81–1.75	72.37–1.44
Number of protein residues	244	244	244	245
Number of water molecules	84	199	172	342
RMSD bond lengths (Å)	0.020	0.011	0.009	0.025
RMSD bond angles (°)	2.24	1.53	1.85	2.28
R‐factor (%)	20.4	16.7	17.8	11.2
R free (%)c	25.2	19.2	20.7	13.2
Average B‐factor, protein atoms (Å2)	46.84	27.26	32.37	19.04
Average B‐factor, inhibitor atom (Å2)	65.50	28.39	48.51	25.68

^aValues in parentheses are for the highest‐resolution shell.

^bFraction of theoretically possible reflections observed.

^c R _free was calculated with 5% of the reflections set aside randomly.

Similarities between the complexes

In all four complexes, initial F _o – F _c difference maps contoured at 3 σ showed unambiguous electron density indicating the presence of the boronic acid inhibitors bound in the OXA‐24/40 active site. Additionally, this electron density was observed connecting the boron atom of the inhibitors to the Oγ atom of the catalytic serine, Ser81, consistent with a covalent attachment of the inhibitor to the enzyme. The density also exhibited the expected tetrahedral geometry around the boron upon binding to Ser81Oγ. In all of the complexes, the boronic acid O1 atom is bound within the oxyanion hole, stabilized by interactions with the backbone nitrogens of Trp221 and Ser81, as well as the main chain carbonyl oxygen of Trp221. The O2 atom of the boronic acids, including the ring bound O2 oxygen of the benzoxaborole BA4 are stabilized by a hydrogen bond with the Oγ atom of Ser128. In each of the complexes, except the one with the benzoxaborole BA4, the inhibitor O2 atom is involved in a water mediated bridge to Arg261. The general base in the reaction (Lys84) is fully carbamylated in all structures. The hydrophobic bridge, composed of residues Tyr112 and Met223 and unique to OXA‐24/40, is intact in each of the structures and occludes the active site. Omit maps were calculated for each inhibitor in the final models, confirming their conformations in the OXA‐24/40 active site (Fig. 3).

Figure 3.

Stereoview of OXA‐24/40 complexes with boronic acids. F _o – F _c omit maps (gray; contoured at 3.0 σ) surrounding the inhibitor and 2F _o – F _c electron density maps (green; contoured at 1.0 σ) surrounding conserved active site residues. A: BA3; B: BA8; C: BA4; D: BA16. Carbon atoms are colored orange for BA3, white for BA8, cyan for BA4, and magenta for BA16. Nitrogen atoms are blue, oxygens red, sulfurs yellow, and borons pink. This and all subsequent figures were made with PyMOL.54

Unique characteristics of OXA‐24/40/BA3 complex

Beyond the canonical hydrogen bond interactions with the boronic acid hydroxyl groups, the aryl ring of the inhibitor BA3 is observed to make van der Waals interactions with the side chains of Trp115 and Leu168 [3.6 Å; Fig. 4(A)]. Val130 is observed in two different rotamers in this complex, likely due to a close contact with the inhibitor aryl ring in one of the conformations (3.1 Å). The meta substituent side chain of BA3 contains an amide group, and the oxygen atom hydrogen bonds with the backbone nitrogen of Met223. A water molecule (Wat42), bound in a pocket formed by the main chain atoms of Leu168, Met223, and Gly224, makes an additional hydrogen bond to the amide oxygen atom of BA3 (2.8 Å). The remainder of the side chain does not make hydrogen bonds with OXA‐24/40 but may be involved in several van der Waals interactions with hydrophobic residues Leu168, Val169, and Gly224. The furan ring at the distal end of the inhibitor sits at the edge of the active site and does not appear to be making any significant interactions.

Figure 4.

Active site interactions between OXA‐24/40 and boronic acid inhibitors. A: BA3; B: BA8; C: BA4; D: BA16. Hydrogen bonds are indicated with dashed yellow lines and represent distances between 2.5 and 3.2 Å. Carbon atoms of the OXA‐24/40 residues are colored green. Boronic acid carbons atoms are colored as described in Figure 3. The side chain of Met223 was removed for optimal visualization of the bound inhibitor.

Unique characteristics of OXA‐24/40/BA8 complex

BA8 contains an intriguing four‐membered nitrogen‐containing ring, reminiscent of the lactam ring found in β‐lactamase substrates. Additionally, BA8 was the best of the inhibitors containing a para substituent. The aryl ring of the inhibitor makes van der Waals interactions with the side chains of Trp115 and Leu168, as well as with Val130 [3.5–3.8 Å; Fig. 4(B)]. The tetrahedral geometry at the sulfone group in the side chain places the four‐membered ring near Val130 (∼4.2–4.6 Å).

Unexpectedly, a second inhibitor molecule was observed noncovalently bound in nearby area of the OXA‐24/40 active site (Supporting Information Fig. S1). A sulfone oxygen atom of the second BA8 molecule hydrogen bonds with the O2 atom of the serine‐bound BA8 inhibitor molecule (3.1 Å), and the other sulfone oxygen interacts with Arg261 (3.3 Å) and a water molecule (Wat16; 2.7 Å). The aryl ring of the second BA8 makes van der Waals interactions with Trp221 (3.7 Å), and the four‐membered ring interacts similarly with Trp115 (3.8 Å). The boronic acid functional group is planar in geometry and is oriented toward the bulk solvent. The electron density surrounding this group suggests that it may adopt several conformations through rotation about the boron‐carbon bond. As modeled, one of the boronic acid hydroxyl groups hydrogen bonds with the main chain nitrogen of Gly258, as well with two water molecules.

Unique characteristics of OXA‐24/40/BA4 complex

The most kinetically and structurally intriguing inhibitor is BA4, the benzoxaborole. Despite the constraints induced by the oxaborole functionality, the inhibitor's O1 atom makes the expected interactions within the oxyanion hole, and the ring bound O2 oxygen also forms a hydrogen bond with Ser128 [2.7 Å; Fig. 4(C)]. The aryl ring of the benzoxaborole is observed to make van der Waals interactions with carbon atoms in the side chains of Trp115 and Leu168 (∼3.8 Å). Additionally, the aryl ring of the inhibitor is oriented such that it is involved with edge‐to‐face aromatic interactions with the side chains of Tyr112 and Trp115 (Fig. 5). The centroid‐centroid distance between the inhibitor aryl ring and that of Tyr112 and Trp115 measures 6.7 and 5.8 Å, respectively. The angle between the planes of the aryl rings of the inhibitor and Tyr112 is 86º, and the same interaction between the inhibitor and Trp115 is 57º.

Figure 5.

Superposition of BA4 and BA16. Edge‐to‐face trimer aromatic interactions between the benzoxaborole BA4, Tyr112, and Trp115 are indicated with purple dashed lines between the centroid atom positions of the aryl rings.

The side chain attached to the benzoxaborole makes two water‐mediated hydrogen bonding interactions with the inhibitor amide group found there. The amide oxygen of BA4 hydrogen bonds to Wat104, and the water makes hydrogen bonds with the main chain oxygen atom of Leu168 and the main chain nitrogen of Met223and Gly224 (2.8–3.1 Å). The amide nitrogen of BA4 makes a single hydrogen bond to Wat74 (3.0 Å).

Finally, several non‐polar interactions appear to stabilize the binding of the tert‐butyl group to OXA‐24/40. This distal part of the molecule makes van der Waals interactions with Leu168, Val169, and Met223 (3.6–3.8 Å).

Unique characteristics of OXA‐24/40/BA16 complex

To investigate the drastic difference in inhibition between the benzoxaborole BA4 (K_i 5.03 μM) and the related arylboronic acid BA16 (K_i 206 μM), the crystal structure of OXA‐24/40 in complex with BA16 was determined [Fig. 4(D)]. Similar to BA4, the aryl ring of BA16 makes van der Waals interactions with carbon atoms in the side chains of Trp115 (3.4 Å) and Leu168 (3.8 Å).

Aromatic interactions were also analyzed. The centroid‐centroid distance between the inhibitor aryl ring and that of Tyr112 and Trp115 measures 8.0 and 6.0 Å, respectively. The angle between the planes of the inhibitor aryl ring and the side chain aryl ring of Tyr112 is 88°, and the angle between the inhibitor plane and Trp115 plane is 23°. Interestingly, Val130 is observed in two different rotamers in this complex, presumably due to a close contact with the inhibitor aryl ring in one of the conformations (2.8 Å). The amide nitrogen of the inhibitor side chain hydrogen bonds to Wat132, which makes a water‐mediated interaction to main chain nitrogen and oxygen atoms of Met223, as well as with the main chain oxygen of Leu168 (2.8–3.2 Å). The distal tert‐butyl group is also involved with van der Waals interactions with the side chain of Val169 (3.4 Å).

Microbiology

Antimicrobial susceptibility test results revealed that each of the boronic acids tested (BA3 and BA4) failed to restore susceptibility to E. coli DH10B cells harboring OXA‐24/40. The MIC's for imipenem, and the zone sizes for ampicillin, were not reduced when BA3 or BA4 were added. For imipenem, the MIC's were unchanged at 128 μg/mL when the boronic acid was added at 4 μg/mL. For ampicillin, the zone remained unaffected at 6 mm. The lack of microbiological efficacy can be attributed to either poor penetration into the bacterial periplasm or poor stability of the compounds.

Discussion

Boronic acids have been extensively characterized as covalent, reversible inhibitors of class A and C β‐lactamases.13, 23, 25 Most of those reported are arylboronic acids. In many of these cases, optimization efforts for these targets resulted in improved affinity, with K_i values reported in the low nanomolar to sub‐nanomolar range.20, 21, 31 The paucity of information regarding boronic acids as class D β‐lactamase inhibitors prompted us to explore their potential as inhibitors for a particularly worrisome target, the carbapenemase OXA‐24/40 from A. baumannii.

Overall, we show that boronic acids offer the potential to be developed as class D β‐lactamase inhibitors. Specifically, two inhibitors (BA3 and BA4) were identified as novel leads, with K_i values for OXA‐24/40 of 5 and 30 μM, respectively (Table 1). The X‐ray crystal structures of OXA‐24/40 in complexes with our two best inhibitors (BA3 and BA4), as well as two lower affinity compounds (BA8, K_i 97 μM; BA16, K_i 475 μM) provide insight on key binding determinants for improving boronic acids as class D β‐lactamase inhibitors.

In every structure, the boronate group of the inhibitor bound in a canonical manner that has been observed in other β‐lactamase/boronic acid complexes, with the boron atom covalently bound to the catalytic serine, the O1 hydroxyl bound in the presumed oxyanion hole, and the O2 hydroxyl interacting with a water molecule.28, 32

We were especially interested in observing how these inhibitors complement the predominantly hydrophobic active site of OXA‐24/40, which is a signature feature of the class D β‐lactamases, since this information would be important for further optimization of this class of novel inhibitors. Beyond the conserved boronate core structure, each inhibitor contained a side chain substituent that fits under the hydrophobic bridge composed of Tyr112 and Met223 and does not disrupt this notable structural feature of OXA‐24/40. The side chain substituent of the two best inhibitors (BA3 and BA4) is positioned meta to the boronate group. The distal nonpolar portions of these side chains occupy essentially the same area and complement a hydrophobic pocket composed of Leu168, Val169, and Gly224 at the edge of the binding site (Fig. 4). BA8, with its para side chain substituent, adopts a different trajectory from the other inhibitors, which prevents the distal portion from binding in the hydrophobic pocket. However, the distal part of the BA16 side chain, which is the same as the side chain of BA4, also binds in hydrophobic pockets, yet displays quite a poor affinity (K_i 475 μM). Therefore, we conclude that this pocket may aid in inhibitor binding but is not a major reason for the observed differences in affinity.

Further analysis of our structures indicated that the arylboronic acid may not be the optimal scaffold for inhibition of OXA‐24/40. The best inhibitor identified was a benzoxaborole (BA4), which exhibited a K_i value of 5 μM for OXA‐24/40. The benzoxaborole appears to offer a more effective inhibition scaffold for class D β‐lactamases. Anacor initially reported oxaboroles as β‐lactamase inhibitors,33 and recent work from Astra‐Zeneca further explored the ability of these compounds as to act as broad‐spectrum serine β‐lactamase inhibitors, reporting inhibition of representative members of classes A (TEM‐1, CTX‐M, KPC‐2), C (AmpC, P99), and D enzymes (OXA‐10, including carbapenemases OXA‐24/40 and OXA‐48).24

In our studies, the benzoxaborole (BA4) showed improved inhibition, a 40‐fold lower K_i value, than its arylboronic acid analog (BA16). Analysis of the two OXA‐24/40 complexes provides insight into the observed functional differences in binding affinity. Superposition of the structures highlights two main differences in the conformation of the inhibitors. First, the side chain carbamate groups are flipped ∼180° relative to each other [Fig. 4(B,D)]. In the case of BA4, the carbonyl oxygen is observed to hydrogen bond to a water molecule (Wat104). Wat104 makes three hydrogen bonds with protein atoms: two main chain amide nitrogens acting as hydrogen bond donors and a main chain amide oxygen acting as an acceptor. The oxygen of the BA4 inhibitor acts as the second hydrogen bond acceptor with Wat104. In contrast, the side chain of BA16 orients its carbamate nitrogen toward a water in an analogous position (Wat138). Since the inhibitor nitrogen would act as a hydrogen bond donor, this atom does not ideally complement the water molecule. This provides a possible explanation for the higher K_i value of BA16 for OXA‐24/40.

More convincing evidence for the observed differences in binding affinity can be elucidated from interactions with the benzoxaborole moiety itself. In contrast to the conformation of BA16, the aryl portion of the benzoxaborole BA4 adopts a different angle due to the constraints induced by its being fused to the oxaborole ring (Fig. 5). The benzoxaborole is rotated 38° from the position of the arylboronic acid ring and is oriented away from Val130. In this position, the aryl ring of the benzoxaborole makes favorable CH‐π interactions with the aliphatic methyl group of Val130 [Fig. 4(C)]. Whereas the aryl ring of BA16 clashes with one of the two observed conformers of Val130 [Fig. 4(D)]. The electron density surrounding the aryl ring of BA16 shows indications that this ring may “wobble” in the active site, given that the volume of electron density is larger around the top of the ring (atoms C1 and C2) as compared to the bottom of the ring. Presumably, this wobble is due to the clash with Val130 in one conformation and may provide some rationale for the lower binding affinity of BA16 as compared with BA4.

Not only is the benzoxaborole scaffold optimally constrained to make stabilizing CH‐π interactions with Val130, but this scaffold also allows BA4 to nicely complement the largely hydrophobic OXA‐24/40 active site by positioning the aryl ring to favorably interact with active site residues, Tyr112 and Trp115. Aromatic‐aromatic interactions are important for contributing to and stabilizing the tertiary structures of proteins, as well as aiding in protein recognition of different medicinal agents.34, 35, 36 Analysis of the aromatic interactions with the side chains of Tyr112 and Trp115 shows a difference between the OXA‐24/40 complex with BA4 and BA16 (Fig. 5). The centroid‐centroid distances between the inhibitor aryl ring and that of Trp115 remain essentially the same between the BA4 (5.8 Å) and BA16 (6.0 Å) complexes. However, the distance to the centroid of Tyr112 has increased to 8.0 Å in the arylboronic acid complex, 1.3 Å greater than observed in the benzoxaborole complex, and outside of the range for favorable aromatic interactions, which are ideally 4.5–7.0 Å.34 The geometry of aromatic‐aromatic interactions observed in protein structures is ideally between 50 and 90°, with an average of 57°.34 In the benzoxaborole complex, this angle is 57°, as measured between the aryl rings of the inhibitor and Trp115, indicating an ideal edge‐to‐face interaction. By comparison, this angle in the arylboronic acid complex is 23°, showing that the two rings are oriented in an edge‐to‐edge manner that is not as optimal for binding.

Edge‐to‐face aromatic interactions are most often observed in networks, that is, groups of three or more residues, within proteins.34, 37 In the OXA‐24/40 complex with the benzoxaborole BA4, the aryl ring of the inhibitor completes such a network. An edge‐to‐face interaction exists within the OXA‐24/40 protein between Tyr112 and Trp115 (centroid‐centroid distance, 4.7 Å; angle between planes, 64°), and the aryl ring of the benzoxaborole completes an ideal edge‐to‐face network by making aromatic interactions with both Tyr112 and Trp115 (Fig. 5). The analogous arylboronic acid BA16 lacks an interaction with Tyr112 entirely and is involved with edge‐to‐edge interactions with Trp115, instead of the more favored edge‐to‐face arrangement. The observed aromatic‐aromatic, as well as the CH‐π, interactions provide strong evidence in explaining the higher binding affinity of BA4.

Identification of a broad‐spectrum inhibitor of the class D β‐lactamases would be beneficial in combating β‐lactamase mediated resistance. Of utmost clinical relevance are the CHDLs, which possess the ability to destroy the most recent class of β‐lactam antibiotics, the carbapenems. Benzoxaboroles may present an optimal scaffold for inhibition of other CHDLs. Trp115 and Val130 are conserved in class D β‐lactamases and would be expected to make similar aromatic and CH‐π interactions with the aryl ring of the benzoxaborole. Our structural insights will require a better understanding of the chemical properties that can lead to improved cell penetration. In our antibiotic susceptibility assays, cell penetration was minimal; we speculate that perhaps the hydrophobic nature of the compounds may interfere with cell penetration.

Superposition of the OXA‐24/40/benzoxaborole complex with CHDLs OXA‐23 (PDB 4K0X38), OXA‐48 (PDB 3HBR39), OXA‐51 (PDB 4ZDX,40 5KZH41), and OXA‐58 (PDB 4OH042) shows that Trp115 is oriented almost identically in each structure and poised to make similar edge‐to‐face interactions with the benzoxaborole. Tyr112 is only found in OXA‐24/40, where it comprises one of the residues that form a hydrophobic bridge over the active site. OXA‐48 lacks a bridge and contains an isoleucine in place of a tyrosine at this position. An edge‐to‐face trimer would not be possible for OXA‐48, and benzoxaboroles may not bind as well to OXA‐48. A bridge is present in the other CHDLs (OXA‐23, OXA‐51, and OXA‐58), but Tyr112 is replaced by a phenylalanine in these enzymes. The phenylalanine is positioned such that it could make edge‐to‐face interactions with the benzoxaborole in much the same way as Tyr112.

Two differences in apo OXA‐51 would likely prevent benzoxaboroles from binding well. An additional tryptophan (Trp222) that occludes the active site would clash with the benzoxaborole, but this residue has been shown to rotate out of the active site upon ligand binding.43 However, CH‐π interactions in OXA‐51 could not be made due to replacement of Val130 with the larger isoleucine side chain, which would clash with the benzoxaborole aryl ring.

Benzoxaboroles appear to present an ideal scaffold for inhibition of OXA‐24/40, and potentially certain other class D β‐lactamases. Edge‐to‐face aromatic and CH‐π interactions between enzyme and inhibitors offer a sensible way to optimize binding affinity by complementing a hydrophobic active site, such as those found in the class D β‐lactamases. Given the lack of class D β‐lactamase inhibitors, the next challenge is to improve bacterial permeability or stability of these novel compounds for development into clinical inhibitors.

Materials and Methods

Protein expression and purification

OXA‐24/40 was expressed and purified using previously described methods.44 Protein designated for crystallization was concentrated to ∼10 mg/mL using a Centricon centrifugal ultrafiltration device (10 kDa molecular weight cutoff; Amicon) and used fresh. Protein destined for kinetics was concentrated to ∼5 mg/mL and stored at −80°C for later use. The concentration of OXA‐24/40 was determined by measuring A ₂₈₀ using an extinction coefficient of 43,430 M⁻¹ cm⁻¹.

Kinetics

The sixteen selected boronic acids (Alfa Aesar) were dissolved in dimethyl sulfoxide to a concentration of 50 mM and stored at −20°C. The boronic acids were tested at concentrations ranging from 0 to 2 mM (0, 0.05, 0.1, 0.2, 0.5, 1, 2 mM) in sodium cacodylate buffer (50 mM sodium cacodylate, 0.01% triton, 25 mM sodium bicarbonate, pH 7.4) against OXA‐24/40 (66 nM) with its synthetic substrate, cephalothin nitrothiobenzoic acid (CENTA; 20 μM) at 25°C. Using the Synergy H1 Hybrid Multi‐Mode Microplate Reader, all concentrations of the inhibitor were tested without preincubation of the inhibitor with enzyme, and the absorbance was determined at 405 nm every 10 s for 5 min. Initial rates ( $v_0$) were calculated, and velocities from at least three trials were averaged and plotted as a function of substrate concentration.

IC₅₀ values were determined through nonlinear regression of the dose‐response curve in Microsoft Excel using SDAS with Eq. (1), where $v_u$ represents uninhibited enzymatic activity. K_i values were calculated from the IC₅₀ value and the K _m of CENTA (15 μM) using the Cheng − Prusoff equation [Eq. (2)].45

A full kinetic analysis was conducted on BA4 and its derivative, BA16, on a Cary 60 Spectrophotometer using the OXA‐24/40 (66 nM) and CENTA (20 μM) in buffer (50 mM NaH₂PO₄, 25 mM NaHCO₃, pH 7.4) at 25°C (Fig. 2). Determination of the K_i value was again conducted using the Cheng‐Prusoff equation.

$$v_0=v_u- \frac{v_u\left[I\right]}{{IC}_{50}+\left[I\right]}$$ (1)

$$K_i=\frac{{IC}_{50}}{1+ \frac{\left[S\right]}{K_m}}$$ (2)

Crystal growth and structure determination

OXA‐24/40 was crystallized by hanging drop vapor diffusion in a 10 μL drop containing 10 mg/mL β‐lactamase mixed 7:3 (v/v) with 100 mM Tris HCl, pH 8.5 and 2.0M ammonium sulfate hanging over 500 μL of well buffer. Crystals were observed within 2 days, after which they were harvested, soaked for 8–13 min in 5% sucrose, 25 mM sodium bicarbonate, and 200–1,000 μM boronic acid and then flash cooled in liquid nitrogen. Each of the diffraction data sets was measured from a single crystal at 100 K at the LS‐CAT sector of the Advanced Photon Source at Argonne National Laboratory (Argonne, IL). Reflections were indexed, integrated, and scaled using autoPROC.46 The OXA‐24/40 structures were determined initially by molecular replacement with Phaser47 using the model of OXA‐24/40 (PDB 3PAE44 with all ligand and water molecules removed) as the initial phasing model. Refinement and electron density map calculations were carried out with REFMAC548 in the CCP4 program suite.49 Manual rebuilding of the model was accomplished with Coot.50 Optimization of the model was performed with PDB_REDO.51, 52 The coordinates and structure factors have been deposited in the Protein Data Bank as 5TG4 for the structure of OXA‐24/40 in complex with BA3, 5TG5 for the structure of OXA‐24/40 in complex with BA4, 5TG6 for the structure of OXA‐24/40 in complex with BA8, and 5TG7 for the structure of OXA‐24/40 in complex with BA16.

Microbiology

Determination of minimum inhibitory concentrations (MICs) and the disk diffusion assays were each performed according to the Clinical and Laboratory Standards Institute (CLSI) guidelines53 as previously described.26 The MICs were determined in cation‐adjusted medium Mueller‐Hinton (MH) and employed an E. coli construct of OXA‐24/40. For the imipenem/boronic acid combinations, the substrate concentrations were varied while the inhibitors BA3 and BA4 were tested at a constant concentration of 4 μg/mL. For the disk diffusion assays, E. coli OXA‐24/40 bacterial liquid culture grown overnight was diluted using MH broth to a McFarland Standard OD₆₀₀ of 0.224. Bacteria were streaked onto a plate composed of MH agar and a disk containing 100 μg of compound and 10 μg of ampicillin was added. The plates were incubated overnight at 37°C, and zone sizes were measured after 18 h.

Supporting information

Supporting Information