Active-Site Plasticity Is Essential to Carbapenem Hydrolysis by OXA-58 Class D β-Lactamase of Acinetobacter baumannii

Abstract

Carbapenem-hydrolyzing class D β-lactamases (CHDLs) are a subgroup of class D β-lactamases, which are enzymes that hydrolyze β-lactams. They have attracted interest due to the emergence of multidrug-resistant Acinetobacter baumannii, which is not responsive to treatment with carbapenems, the usual antibiotics of choice for this bacterium. Unlike other class D β-lactamases, these enzymes efficiently hydrolyze carbapenem antibiotics. To explore the structural requirements for the catalysis of carbapenems by these enzymes, we determined the crystal structure of the OXA-58 CHDL of A. baumannii following acylation of its active-site serine by a 6α-hydroxymethyl penicillin derivative that is a structural mimetic for a carbapenem. In addition, several point mutation variants of the active site of OXA-58, as identified by the crystal structure analysis, were characterized kinetically. These combined studies confirm the mechanistic relevance of a hydrophobic bridge formed over the active site. This structural feature is suggested to stabilize the hydrolysis-productive acyl-enzyme species formed from the carbapenem substrates of this enzyme. Furthermore, our structural studies provide strong evidence that the hydroxyethyl group of carbapenems samples different orientations in the active sites of CHDLs, and the optimum orientation for catalysis depends on the topology of the active site allowing proper closure of the active site. We propose that CHDLs use the plasticity of the active site to drive the mechanism of carbapenem hydrolysis toward efficiency.

INTRODUCTION

Carbapenems, such as imipenem and meropenem, are β-lactam antibiotics with a wide spectrum of activity. Their clinical introduction was driven by the need to overcome β-lactam resistance as a result of their efficacy in inhibiting most β-lactamases. However, as a result of selective pressure, β-lactamases acquired the ability to recognize and hydrolytically degrade the carbapenems. β-Lactamases hydrolyze the β-lactam bond of these antibiotics. Four subclasses of the β-lactamases are identified, based on their amino acid sequences: A, B, C, and D. Classes A, C, and D are β-lactamases with a serine active site, and class B β-lactamases are metalloenzymes. Importantly, the progressively increasing ability of all four classes of β-lactamases to hydrolyze carbapenems has been documented (1, 2).

Class D β-lactamases are also known as oxacillinases (OXA β-lactamases) due to their ability to hydrolyze the oxacillin subclass of the penicillins (3). The OXA β-lactamases were first observed to be a resistance mechanism against imipenem in the mid-1980s, coincidentally in the same year imipenem was introduced for clinical use (4, 5). Carbapenem-hydrolyzing class D β-lactamases (CHDLs) are directly associated with outbreaks of carbapenem-resistant Acinetobacter baumannii around the world. CHDLs have emerged as a major problem in the treatment of multidrug-resistant A. baumannii, for which carbapenems were previously the drugs of choice (6–8). CHDLs now are commonly found in all Acinetobacter species, thus increasing the risk of their being spread to other organisms. Indeed, CHDLs are also found in the Enterobacteriaceae, Klebsiella pneumoniae, and Pseudomonas aeruginosa (9). OXA-23 was the first CHDL to be identified (10) and, to date, there are close to 200 variants of class D β-lactamases reported to possess carbapenemase activity (9). Their genes are either plasmid borne or contained on chromosomes, and they are believed to be either recently acquired or native to A. baumannii, respectively. Sequence alignments indicate that CHDLs share 18% sequence identity with the class D oxacillinases and 40 to 90% sequence identity with each other (9).

CHDLs are divided into 12 distinct subgroups. The most widely spread carbapenemases in A. baumannii are OXA-23-like, OXA-40-like (prior name for OXA-24), OXA-51-like, and OXA-58-like β-lactamases (9). Another important CHDL is OXA-48. This enzyme is nonnative to A. baumannii and was isolated from K. pneumoniae in 2001 (11). Since then, OXA-48 and its variants have also been isolated in A. baumannii and other Enterobacteriaceae (12). Of all the CHDLs, only the structures of OXA-24, OXA-23, OXA-146 (OXA-23 like), and OXA-48 have been solved, either alone or in complex with another carbapenem (13–17). Additionally, the structure of OXA-58 alone (referred to as apo-OXA-58) was recently reported (14). Structural analyses of these CHDLs show that they share key structural features with the class D β-lactamases, such as OXA-1 and OXA-10, including the carboxylated state of the conserved catalytic residue Lys-86 (OXA-58 numbering), which was first identified as being essential to the class D β-lactamase catalytic mechanism in class D β-lactamases (18).

Structural analyses and mutagenesis studies of CHDLs suggest that two structural elements in CHDLs are responsible for the gain in carbapenemase activity by these enzymes: a shorter connecting loop of the β6- and β7-strands that form one side of the active site (19), and a hydrophobic bridge over the active-site cleft (13, 15, 17). The second structural feature was observed in the crystal structures of OXA-23 and OXA-24, and its presence was predicted in the homology model of the OXA-58 structure (20). It was not observed in the crystal structure of OXA-58 solved by Smith et al. (14) or in the structure of OXA-48 (16). Therefore, OXA-58 and OXA-48 were proposed to have different hydrolytic mechanisms toward carbapenems (14).

In this study, we evaluated the structural elements that contribute to the catalysis of carbapenems by OXA-58 by solving the structure of this enzyme in an acyl-enzyme complex with the 6α-hydroxymethyl penicillin derivative (6αHMP) as a carbapenem structural mimetic (Fig. 1). Analysis of the OXA-58–6αHMP complex structure revealed that the structural changes that take place on the two sides of the active-site cleft lead to the formation of a hydrophobic bridge over the OXA-58 active site. Two rotamers of the acyl-enzyme species are thus formed. We also carried out kinetic studies on several OXA-58 variants to probe the significance of the residues homologous to those that are involved in the formation of the hydrophobic bridge in OXA-23 and OXA-24. These findings shed new light on the relationship between the plasticity of OXA-58 in particular, and CHDLs in general, with respect to the optimum binding mode to allow the hydrolysis of carbapenems.

FIG 1.

Chemical structures of 6αHMP, 6αHOP, and imipenem (representative of carbapenems) and its acylated species with a β-lactamase.

MATERIALS AND METHODS

Chemicals and reagents.

All reagents were American Chemical Society (ACS) grade and were purchased from either Merck Millipore, Sigma, Fluka, or Bio-Rad. Growth media were purchased from HiMedia Laboratories (Mumbai, India) or VWR (Canada). Chromatography media and columns were purchased from GE Healthcare (Canada). Crystallization screens (Crystal Screen I & II, PEG/Ion I & II, Index, Salt, and Crystal Screen Cryo) were procured from Hampton Research (USA).

Protein crystallization.

OXA-58 was produced and purified as described earlier (20). Purified OXA-58 was concentrated to 20 mg/ml in 10 mM sodium phosphate (pH 7.0) buffer and 100 mM NaCl. Crystals of OXA-58 were obtained using the vapor diffusion method in 96-well sitting drop plates (Hampton Research, Inc., Aliso Viejo, CA) at 293 K. For the initial crystallization screening, small drops were prepared by mixing 1 μl of protein solution with the same volume of well solution and equilibrated against 50 μl of well solution. All of the listed crystallization kits were evaluated. We obtained diffraction quality crystals of OXA-58 using a well solution containing 0.1 M Tris-HCl (pH 8.0) buffer and 28% (wt/vol) polyethylene glycol 3350 (PEG 3350).

To obtain crystals of OXA-58 in the acyl-enzyme state, we soaked crystals of native protein in a solution of 10 mM 6α-hydroxymethyl penicillin (6αHMP) or 6α-hydroxyoctyl penicillin (6αHOP) in 0.1 M Tris-HCl (pH 8.0) buffer and 30% (wt/vol) PEG 3350 for 15 min at room temperature.

Data collection and processing.

The diffraction data sets for native and β-lactam-bound OXA-58 crystals were collected on an in-house X ray setup with a MAR 345 imaging plate detector mounted on a Bruke Nonius Microstar-H rotating anode generator. For all crystal structures, the diffraction data were processed and scaled using HKL-2000 (21). The native protein crystal belonged to the P2₁2₁2₁ space group with the unit cell parameters a = 37.07, b = 67.03, and c = 93.51. The 6αHMP-bound OXA-58 crystals belonged to the P2₁ space group with the unit cell parameters a = 36.94, b = 65.65, and c = 191.59, and β = 91.7°. The data collection statistics are given in Table 1.

TABLE 1.

Data collection and refinement statistics

Data collection statistics	Value(s) for:
	OXA-58 (apo structure)	OXA-58–6αHMP complex	OXA-58 (pseudoapo structure)
Wavelength (Å)	1.54	1.54	1.54
Resolution range (Å)	33.52–2.37 (2.46–2.37)	36.93–2.60 (2.69–2.60)a	34.79–2.29 (2.37–2.29)
Space group	P212121	P21	P21
Unit cell a, b, c (Å); β (°)	37.07, 67.03, 93.51	36.94, 65.65, 191.59; 91.7	37.02, 65.10, 191.96; 91.28
No. of total reflections	29,760 (1,605)	70,515 (4,578)	100,971 (7,493)
No. of unique reflections	9,500 (721)	21,116 (1,637)	37,470 (2,741)
Multiplicitya	3.1 (2.2)	3.3 (2.8)	2.8 (2.7)
Completeness (%)	95.7 (73.1)	87.3 (69.3)	91.23 (74.8)
Avg I/sigma(I)	11.26 (2.75)	8.45 (2.30)	9.15 (2.65)
Wilson B-factor	34.76	39.21	33.11
Rmerge	0.11 (0.36)	0.15 (0.43)	0.12 (0.42)
Rmeas	0.13	0.15	0.14
CC1/2	0.99 (0.77)	0.97 (0.68)	0.99 (0.78)
CCa	0.99 (0.93)	0.99 (0.90)	0.99 (0.94)
Rwork	0.19 (0.30)	0.21 (0.31)	0.19 (0.30)
R-free	0.24 (0.36)	0.26 (0.36)	0.24 (0.36)
No. of nonhydrogen atoms	2,060	8,126	8,146
Macromolecules	1,943	7,784	7,758
Ligands		60
Water	117	282	388
No. of protein residues	243	975	968
RMS (bonds) (Å)	0.014	0.019	0.016
RMS (angles) (°)	1.74	1.80	1.65
Ramachandran plot (%)
Favored regions	98	96	99
Outlier regions	0	0.41	0
Clash score	0.77	0.38	1.29
Avg B-factor	44.0	54.2	44.5
Macromolecules	44.2	54.1	44.7
Ligands		56.2
Solvent	39.9	55.7	44.5

^aStatistics for the highest-resolution shell are shown in parentheses.

Structure modeling and refinement.

The initial phases for OXA-58 were obtained by molecular replacement with MolRep in the CCP4 suite (22, 23) using a monomer of the carbapenemase OXA-24 structure (Protein Data Bank [PDB] code 2JC7, chain A) as a search model (24). The rigid body refinement was followed by iterative cycles of restrained coordinate and atomic displacement parameter refinement, including translation-libration-screwmotion (TLS) refinement with Refmac5 (22) and Phenix.Refine (25). The repetitive cycles of manual model rebuilding based on σA-weighted 2F_o−F_c and F_o−F_c maps were performed using COOT (26). The F_o−F_c difference map for the 6αHMP-bound form of OXA-58 clearly showed density for 6αHMP above the 3σ level, allowing the 6αHMP structure to be modeled into the density. Water molecules were added in peaks that were simultaneously >3σ in the F_o−F_c difference map and >1σ in the 2F_o−2F_c map. Several rounds of refinement and model rebuilding were performed until the best possible model was achieved.

Programs used.

The stereochemical attributes of the refined model were validated using the program MolProbity (28). The refinement statistics are presented in Table 1. All protein structure figures were prepared with University of California, San Francisco (UCSF) Chimera (Resource for Biocomputing, Visualization, and Informatics, UCSF) (29). DynDom was used to determine the hinge regions and calculate the rigid body domain movements involved in the conformational changes between apo and 6αHMP-bound OXA-58 (30).

Enzyme kinetics.

OXA-58 F114A, F113A, M225T, M225A, F113Y, and F114I clones were generated by Mutagenex, Inc. (Hillsborough, NJ). OXA-58 and its variants were produced and purified as described earlier (20). The steady-state kinetics parameters k_cat and K_m for all of the mutant enzymes, with imipenem and penicillins as substrates, were determined using isothermal titration calorimetry (ITC) analysis in a MicroCal iTC-200 instrument (GE Healthcare), in accordance with the protocol first developed by Poduch and colleagues (31) and adopted for OXA-58 by Verma et al. (20).

Protein accession numbers.

The atomic coordinates and the structure factors for OXA-58 with and without ligand bound were deposited in the Protein Data Bank (PDB) (27) with the accession numbers 4Y0O, 4Y0T, and 4Y0U.

RESULTS

Determination of OXA-58 crystal structure.

The 6-hydroxyalkyl penicillin derivatives have been described as good probes of the deacylation mechanism of serine β-lactamases (32–34). The stereochemistry at the C-6 position may direct the hydroxyalkyl group of the penicillin to either the α-face or β-face of the β-lactam. Depending which face of the carbonyl of the acyl-enzyme species is approached by the deacylating water molecule, one stereoisomer of the 6α- and 6β-hydroxyalkyl penicillin derivatives will be a substrate, and the other stereoisomer will be an inhibitor. In the case of OXA-58, Verma et al. (20) showed that 6βHMP was a substrate for OXA-58, while the 6αHMP stereoisomer behaved as a competitive inhibitor. The 6αHOP behaved as an inactivator of OXA-58 (19). Carbapenems are 6α-hydroxyalkyl β-lactam derivatives. Hence, we set to understand the binding modes of 6α-hydroxyalkyl penicillin derivatives in order to understand the structural elements required for the hydrolysis of carbapenems by CHDLs in general and OXA-58 in particular.

We attempted to solve the crystal structure of OXA-58 bound to 6αHOP (Fig. 1). The diffraction profile of the OXA-58 crystal after being soaked in a 6αHOP solution for 15 min revealed the space group of the crystal to be P2₁. The crystal diffracted to 2.29 Å resolution, and the OXA-58 structure obtained from this crystal revealed four protein molecules per asymmetric unit. However, no ligand, either covalent or noncovalent, was found in the active site. The final refined structural model of OXA-58 contained 968 amino acid residues, 388 water molecules, and eight bicarbonate molecules.

Soaking the OXA-58 crystals with 6αHMP for 15 min resulted in the formation of a stable acyl-enzyme species. The crystal of the OXA-58–6αHMP complex diffracted to 2.60 Å resolution and showed the same space group (P2₁) and thus four molecules per asymmetric unit, as observed for the OXA-58 crystal soaked with 6αHOP. The refined structural model of the OXA-58–6αHMP complex contained 975 amino acid residues, 282 water molecules, four 6αHMP molecules, and eight bicarbonate molecules.

Overall OXA-58 structure in unbound and bound states.

Brief soaking of the OXA-58 crystal with 6αHOP led to a reduction in the symmetry of the space group of the crystal, from P2₁2₁2₁ obtained for the OXA-58 alone to P2₁. An identical result was observed following soaking of the OXA-58 crystal with 6αHMP, but in this case, a stable acyl-enzyme species was formed. We know that both penicillin derivatives interact with OXA-58, albeit with different dissociation constants, with a K_i of 83 ± 3 μM (mean ± standard deviation) for 6αHMP and a K_i of 404 ± 10 μM for 6αHOP (20). Thus, the decrease in the space group symmetry of the OXA-58 crystal soaked in the solution of either compound must be presumed to result from their binding to the enzyme (whether covalently or noncovalently), effecting a structural change to the protein. An example of ligand binding reducing the space group symmetry of the protein crystal, which is linked to the induction of conformational changes in the protein structure, is found in the literature: the flavin adenine dinucleotide (FAD)-linked quinone reductase 2 enzyme gave a reduction in the space group symmetry and structural change upon the binding of chloroquinone (35).

Here, we refer to the native OXA-58 crystal structure as apo-OXA-58. Our apo-OXA-58 and that of Smith et al. (14) superimpose with a backbone root mean square deviation (RMSD) of 0.31 Å among 243 atom pairs (Fig. 2). The crystal structure of OXA-58 solved from soaking with 6αHOP is referred to as pseudoapo-OXA-58. The apo and pseudoapo structures superimpose, with an RMSD of 0.36 Å among 242 atom pairs.

FIG 2.

Superposition of crystal structures of apo-OXA-58 (PDB code 4Y0O)and pseudoapo-OXA-58 (PDB code 4Y0T) and the native crystal structure of OXA-58 reported by Smith et al. (14) (PDB code 4OH0). (A) Ribbon representation of apo (blue), pseudoapo (green), and 4OH0 (orange) structures of OXA-58. The α3/α4- and β6/β7-loops are indicated by an arrow. (B) Close-up view of the α3/α4-loop and residues Phe-113 and Phe-114 in apo (blue) and pseudoapo (green) structures. (C) Close-up view of the β6/β7-loop in apo (blue) and pseudoapo (green) structures.

The noncovalent binding of 6αHOP to OXA-58 caused two noticeable changes in structure, localized in the loops that connect α3- and α4-helices and β6- and β7-strands (Fig. 2). The largest effect on these loops was noted for chain B. The α3/α4-loop harbors residues Phe-113 and Phe-114, and the β6/β7-loop harbors Met-225. The homologues of Phe-114 in OXA-23 (Phe-110) and OXA-24 (Tyr-112) form a hydrophobic bridge with a methionine residue (Met-221 in OXA-23 and Met-223 in OXA-24) located in the β6/β7-loop (15, 17). In OXA-48, the homologues of Phe-114 and Met-225 (Ile-102 and Thr-244, respectively) do not engage in hydrophobic interactions (16). Phe-113 of OXA-58 has no homologue in OXA-23, OXA-24, or OXA48; instead, a polar residue is found at this position in both enzymes (Ser-109, Thr-111, and Asp-101, respectively).

The conformational change in the α3/α4-loop of pseudoapo-OXA-58 was associated with an inward movement by as much as 1.79 Å (measured for chain B) in comparison to the apo-OXA-58 structure (Fig. 2B). The inward movement in the α3/α4-loop was also associated with the rearrangement of the side chains of Phe-113 and Phe-114 and shifts by 0.97 Å and 0.86 Å of their respective Cα atoms, in comparison to the apo structure (Fig. 2B). The transient binding of 6αHOP also led to an overall 0.66-Å inward movement of the β6/β7-loop (homologues of β4/β5 in OXA-24/OXA-23 and β5/β6 in OXA-48) (Fig. 2C). The conformational changes in the α3/α4- and β6/β7-loops led to an overall 2-Å decrease in the distance between the Cα atoms of Phe-113/Met-225 and Phe-114/Met-225, from 14.45 Å and 15.56 Å in the apo structure to 12.76 Å and 14.18 Å in the pseudoapo structure. This inward movement exemplifies the key structural plasticity in the active site of OXA-58.

The above-mentioned conformational changes in the α3/α4- and β6/β7-loops were more pronounced in the OXA-58–6αHMP complex, especially in the β6/β7-loop (Fig. 3). We measured a 1.03-Å inward shift of the β6/β7-loop (calculations made for chain A). Notably, this movement was associated with a significant reorientation of the Met-225 side chain. In the OXA-58–6αHMP complex, Met-225 is pointed toward Phe-113 and not Phe-114, as seen in the apo-OXA-58, OXA-23–meropenem, and OXA-24–doripenem complexes (Fig. 3). The movements of both loops brought the side chains of Phe113/Phe114 and Met-225 closer: the distances between the Cα atoms of Phe-113/Met-225 and Phe-114/Met-225 changed from 14.45 Å and 15.56 Å in the apo complex to 12.06 Å and 13.67 Å, respectively, in the bound complex. In OXA-23, the distance between the Cα atoms of Phe-110 and Met-221 is 13.89 Å, and in the OXA-24 structure, the corresponding distance between Tyr-112 and Met-223 is 14.27 Å. The homolog residues of Phe-114 and Met-225 engage in hydrophobic interactions in OXA-23 and OXA-24, leading to the formation of the tunnel-like topology of their active sites (15, 17). In the case of the OXA-58–6αHMP complex, the side chains of the above-mentioned residues come also within the limits of van der Waals interactions (4.09 Å for Phe-113/Met-225 and 5.30 Å for Phe-114/Met-225 [distances measured between the C-ζ of the phenyl ring and the sulfur atom of methionine]).

FIG 3.

Stereo diagram of superposition of crystal structures of apo (blue), pseudoapo (green), and OXA-58–6αHMP complex (chain A, in red; PDB code 4Y0U). The depicted side chains and 6αHMP are shown in a stick representation. The F_o-F_c omit electron density maps, contoured at the 3σ level, indicating the position and conformation of the 6αHMP in OXA-58 are displayed in green. Water molecules are shown as green spheres in the apo structure (W1, W2, and W3) and as orange spheres in the pseudoapo structure (W1′ and W3′).

Analysis of the apo and 6αHMP-bound OXA-58 structures (chain A) by the DynDom server (30) identified two rigid domains in the protein. The first encompassed 214 residues from 42 to 103 to 124 to 276, and the second domain consisted of 20 residues from 104 to 123. The two rigid regions were connected through a hinge composed of the flexible residues 103 and 104 (…EI…) and 122 and 123 (…TL…) (see Fig. S1 in the supplemental material). The binding of 6αHMP to OXA-58 induced an interdomain closure motion of 8.3° around a rotation axis close to the flexible residues. This finding further exemplifies the plasticity of the OXA-58 active site and draws our focus to the domain closure that occurs upon covalent binding of the β-lactam.

Active site of OXA-58 in the native and bound states.

β-Lactam binding, either noncovalent or covalent, did not alter significantly the conformation of the key catalytic residues in the active site (Fig. 3). In the pseudoapo and 6αHMP-bound OXA-58 structures, we observed that the carboxyl group of the carboxylated Lys-86 was stabilized by hydrogen bonding to Trp-169 (3.04 Å), the O-γ of Ser-83 (2.57 Å), and a water molecule (Fig. 3). This network is similar to the hydrogen bond network seen in the OXA-58 structure (the apo-OXA-58 structure in this study or the structure solved by Smith et al. [14]). In both apo/pseudoapo structures, the O-γ of Ser-83 formed a hydrogen bond with the O-γ of Ser-130 of the SNV motif, with the O-γ of Ser-130 also hydrogen bonding with N-ζ of Lys-220 of the K(S/T)G motif. Formation of the acyl-enzyme species brought about rotation of the O-γ atom of Ser-130 by 76° toward the N atom of 6αHMP.

The apo-OXA-58 structure contains two water molecules (W1 and W2) in the active site, as was also observed in the Smith et al. (14) structure of OXA-58 (Fig. 3). W1 is positioned in the oxyanion hole by hydrogen bonding to the backbone amide protons of Trp-223 and Ser-83. W2 bridges the carboxylated Lys-86 and Ser-83. A third water molecule, W3, which hydrogen bonds to Trp-169 and carboxylated Lys-86, was seen in the apo, pseudoapo, and 6αHMP-bound structures. The transient binding of 6αHOP led to the displacement of W2 (it is not observed in the pseudoapo structure). Formation of the acyl-enzyme species led to the displacement of both W1 and W2, but not W3, from the active site (Fig. 3).

Binding mode of 6αHMP to the active site of OXA-58.

Notably, we observed that the hydroxymethyl group of 6αHMP occupies two different conformations in the active site of OXA-58 (see Fig. S2 in the supplemental material). This group occupies the same position in chains A, C, and D (Fig. 4A). In contrast, in chain B, the hydroxyl group is rotated by 60° (counterclockwise) from the first position, toward the back of the active site (Fig. 4B and 5). We compared the OXA-58 6αHMP structure (chains A and B) with the structures of the deacylation-deficient OXA-24 variant (Lys84Asp) bound to doripenem (PDB code 3PAE) and to that of the deacylation-deficient wild-type OXA-23 bound to meropenem (PDB code 4JF4 [Fig. 5]). The wild-type OXA-23 was rendered deacylation deficient by crystallization at low pH (13). Superposition of the OXA-23–meropenem, OXA-24–doripenem, and OXA-58–6αHMP structures shows that the hydroxymethyl group of each carbapenem side chain is about 65° (clockwise) away from the position of the 6αHMP hydroxymethyl group seen in chains A, C, and D and closer to the entrance of the active site (Fig. 5). Each carbapenem molecule bound to OXA-23 or OXA-24 adopts the Δ²-pyrroline tautomer (Fig. 1) in the respective crystal structure of its acyl-enzyme species. These acyl-enzyme species represent the hydrolysis-productive species that OXA-24 and OXA-23 form with doripenem and meropenem, respectively, during catalysis. The fact that the hydroxyl group in 6αHMP is positioned away from this orientation, toward the back of the active site, may explain the stability of this acyl-enzyme species of 6αHMP. At this conformation, the hydroxyl group of 6αHMP forms a hydrogen bond with the carboxylated Lys-86 (Fig. 5), which may dampen the basicity of carboxylated Lys-86 toward the deacylating water molecule. Moreover, the conformation of 6αHMP in chain B may also provide a physical barrier to the approach of the deacylating water molecule. Incidentally, the 6αHMP rotamer seen in chains A, C, and D of OXA-58 is similar to that seen in 6αHMP bound to TEM-1 (PDB code 1TEM) (see Fig. S3 in the supplemental material) (33). In the TEM-1–6αHMP complex, it was noted that the hydroxyl group of 6αHMP displaced the structurally conserved deacylating water molecule, and that hydrogen bonding with Glu-166 may prevent the glutamate from activating an incoming water molecule.

FIG 4.

Close-up view of the active-site structural elements in chain A (A) and chain B (B) of the OXA-58–6αHMP complex. The 6αHMP is shown in a stick representation. The F_o-F_c omit electron density maps, contoured at the 3σ level and indicating the position and conformation of the 6αHMP in chains A and B of OXA-58–6αHMP, are displayed in green.

FIG 5.

Superposition of active-site structural elements of chains A (red) and B (gray) of OXA-58–6αHMP complex, OXA-23–meropenem complex (PDB code 4JF4, yellow), and OXA-24–doripenem (PDB code 3PAG, magenta) complex. The ligands are shown in a stick representation and color-coded as the respective protein ribbon color: 6αHMP bound in chain A is presented in red, 6αHMP bound in chain B is presented in gray, meropenem is shown in yellow, and doripenem is shown in magenta. The arrow labeled “1” indicates the shift (60°) in position of the hydroxymethyl groups in 6αHMP bound to chain B of OXA-58 with respect to the same group in chain A. The arrow labeled “2” indicates the shift in positions (65°) of the hydroxymethyl groups in meropenem (green) bound to OXA-24 and in doripenem (red) bound to OXA-23 with respect to the same group in chain A.

Both conformations of the 6αHMP side chain in OXA-58 are stabilized through hydrophobic interactions between the 6αHMP side-chain methylene carbon and the side chains of the conserved residues Val-132 (4.18 Å apart) and Leu-170 (2.67 Å apart) (Fig. 5). Of note, these two residues are in closer proximity to the 6αHMP side-chain methylene carbon in OXA-58, in comparison to their homolog residues in OXA-23 (13), OXA-24 (36), and OXA-48 (37) (Fig. 5; see also Fig. S4 in the supplemental material). The apparent movement of the Ω-loop so as to be positioned closer to the 6αHMP side chain may be due to an intrinsic ability of the Ω-loop, where Leu-170 is located, to interact with the alkyl group of the carbapenem side chain so as to enable efficient catalysis (see below).

As a result of the inward movement of the α3/α4-loop in OXA-58–6αHMP, Phe-114 is within hydrophobic interaction contact with the methyl groups at the C-3 position of 6αHMP (4.37 Å). These methyl groups are also in van der Waals contact with the Ile-260 side chain (3.92 Å), located at the very back of the active site. In addition, Trp-117 located in the α3/α4-loop is in van der Waals contact (4.14 Å) with the plane of the thiazolidine ring of 6αHMP. Overall, these interactions create a sandwich effect with respect to the β-lactam (Fig. 6). The sandwich effect, which acylated carbapenem molecules are also likely to experience, together with the hydrophobic bridge formed over the active site, may effect preferential stabilization of the Δ²-tautomer of carbapenems (Fig. 1). Stabilization of the Δ²-tautomer of the carbapenem will put the side chain at the C-3 carbon of carbapenem away from the side chains of Phe-114 and Phe-113 and avoid steric hindrance among them; as a result, this will lead to more stable acyl-enzyme species.

FIG 6.

Molecular surface representation of the binding pocket of 6αHMP in the active site of OXA-58–6αHMP (chain A in red). The side chains of Trp-117, Phe-113/114, Met-225, and Ile260 are shown in balls and sticks. 6αHMP is represented as spheres and colored cyan (C atoms), red (oxygen atoms), and yellow (sulfur atoms). The snapshots are of same molecule rotated 180° horizontally.

Kinetic characterization of wild-type and OXA-58 variants.

We substituted other amino acids for the Phe-113, Phe-114, and Met-225 residues (Table 2) in order to explore their effects on catalysis. Homologues of Phe-114 and Met-225 in OXA-23 and OXA-24 contribute to the tunnel-like topology of the active site in the native structure of these enzymes. Phe-113, a position that is occupied by Ser-109 and Thr-111, respectively, in OXA-23 and OXA-24, seems to play a role in OXA-58 catalysis, as the Met-225 is oriented toward Phe-113 in the acyl-enzyme species.

TABLE 2.

Steady-state kinetic parameters for OXA-58 and variants^a

Kinetic data by antibiotic	WT	M225A	F113Y	F114A	F113A	M225T	F114I
Penicillin G
kcat/Km (μM−1 s−1)	15 ± 3	19 ± 4	20 ± 2	7 ± 1	4 ± 1	4.1 ± 0.5	0.58 ± 0.06
kcat (s−1)	106 ± 10	141 ± 29	137 ± 9	124 ± 8	37 ± 7	42 ± 3	5.7 ± 0.5
Km (μM)	7 ± 1	7.3 ± 0.7	6.9 ± 0.7	17 ± 2	10.3 ± 2.9	10 ± 1	9.7 ± 0.2
Ampicillin
kcat/Km (μM−1 s−1)	4.6 ± 0.9	1.6 ± 0.3	1.6 ± 0.3	1.5 ± 0.4	0.83 ± 0.09	0.52 ± 0.08	0.13 ± 0.02
kcat (s−1)	97 ± 14	75 ± 8	108 ± 4	188 ± 25	39 ± 1	38 ± 4	6.80 ± 0.03
Km (μM)	21 ± 3	46 ± 9	66 ± 11	127 ± 25	48 ± 5	72 ± 9	54 ± 9
Carbenicillin
kcat/Km (μM−1 s−1)	1.7 ± 0.4	3.3 ± 0.5	1.1 ± 0.3	0.6 ± 0.3	0.31 ± 0.05	0.25 ± 0.07	0.06 ± 0.02
kcat (s−1)	272 ± 13	102 ± 15	167 ± 15	173 ± 64	44 ± 4	41 ± 4	14.8 ± 0.7
Km (μM)	160 ± 40	31 ± 2	153 ± 47	268 ± 65	142 ± 16	161 ± 39	248 ± 66
Amoxicillin
kcat/Km (μM−1 s−1)	2.8 ± 0.5	3.5 ± 0.7	2.9 ± 0.8	1.4 ± 0.1	0.61 ± 0.04	0.7 ± 0.1	0.10 ± 0.02
kcat (s−1)	62 ± 10	97 ± 15	79 ± 7	204 ± 14	24 ± 1	32 ± 4	8.2 ± 0.7
Km (μM)	22 ± 2	28 ± 4	27 ± 7	145 ± 5	40 ± 2	49 ± 4	80 ± 10
Oxacillin
kcat/Km (μM−1 s−1)	3.5 ± 0.4	3.2 ± 0.4	1.1 ± 0.3	0.05 ± 0.02	0.11 ± 0.02	0.036 ± 0.005	0.0110 ± 0.0002
kcat (s−1)	137 ± 14	179 ± 14	85 ± 6	59 ± 19	17 ± 3	17 ± 2	3.69 ± 0.04
Km (μM)	39 ± 3	56 ± 4	74 ± 21	1400 ± 500	156 ± 1	462 ± 24	335 ± 6
Imipenem
kcat/Km (μM−1 s−1)	0.5 ± 0.2	0.52 ± 0.09	0.48 ± 0.08	0.10 ± 0.03	0.09 ± 0.02	0.090 ± 0.007	0.011 ± 0.001
kcat (s−1)	1.2 ± 0.2	0.63 ± 0.02	0.9 ± 0.1	0.442 ± 0.008	0.11 ± 0.01	0.19 ± 0.01	0.024 ± 0.001
Km (μM)	2.3 ± 0.5	1.2 ± 0.2	2.0 ± 0.2	4 ± 1	1.1 ± 0.2	2.1 ± 0.2	2.2 ± 0.2

^aEach progress curve was fitted to the nonlinear equation to obtain the k_cat and K_m. The means and standard deviations shown were calculated from the results from three individual experiments. The enzyme concentration varied from 5 nM for the wild-type (WT) enzyme to 200 nM for the F113L variant.

The steady-state kinetics for the wild-type and OXA-58 variants were determined for imipenem and a number of penicillins. The results are summarized in Table 2. The mutation of Phe-113 or Phe-114 to Ala had a negative impact, by as much as 5-fold, on the efficiency of imipenem turnover (Table 2). This mutation affected both the binding affinity for imipenem (K_m; 1.7-fold reduction) and the turnover rate constant (k_cat; 3-fold reduction). Of the penicillins assessed, the greatest impact observed for the F113A and F114A variants was on the turnover of oxacillin (K_m increases of 4- and 36-fold, respectively, and k_cat decreases of 8- and 2.2-fold, respectively). The mutation of Phe to Ile at position 114 had the most deleterious impact on the activity of OXA-58 against imipenem, a 50-fold reduction mainly expressed on k_cat. A similar effect on the k_cat was also measured for the penicillins, with a reduction in k_cat of as much as 18-fold for penicillin G and 37-fold for oxacillin (Table 2). The effect of the F114I substitution on the activity of OXA-58 might result from steric clashes between the isoleucine side chain and the Phe-113 side chain, with a loss of the optimal orientation for the interaction Phe-113 with Met-225. Moreover, the deleterious effects of this mutation point to the potential role of Phe-114 in properly orienting the Phe-113 side chain. This role is supported by the smaller effect measured for the Phe-to-Tyr mutation at position 113.

The effect of the Met-to-Thr substitution at position 225 on the activity of OXA-58 mirrors that of the Phe-to-Ala substitution at position 113. These observations suggest that threonine, with its shorter and branched side chain and having a polar group attached to the C-β, may not allow interaction with Phe-113. In contrast, the M225A mutation, involving a reduction in the side-chain mass, has a minor effect on enzyme activity. This outcome might be due to the flexibility of the α3/α4-loop, which can move closer to the β6/β7-loop to compensate for the loss in carbon chain length at this position.

DISCUSSION

The crystal structure of the OXA-58–6αHMP complex illuminates two important observations. First, the α3/α4- and β6/β7-loops move toward each other, leading to the formation of a hydrophobic bridge across of the active site. Analysis of apo and 6αHMP-bound OXA-58 structures using the DynDom server (30) revealed that binding of 6αHMP to OXA-58 induces an interdomain closure motion of 8.3°. The second important observation is the ability of the hydroxymethyl group of 6αHMP to occupy two positions in the active site of OXA-58. The more-populated rotamer (found in three out of four chains) is about 65° away from the rotamer of meropenem or doripenem observed in structures of the OXA-23–meropenem and OXA-24–doripenem complexes (Fig. 5). The less-populated rotamer of 6αHMP is about 120° away from these carbapenem rotamers (Fig. 5). These findings directly impact our understanding of the structural factors critical for carbapenem catalysis by class D β-lactamases.

The inward movements that the α3/α4 and β6/β7-loops experience in the OXA-58–6αHMP complex lead to an ∼2-Å decrease in the distance between the Cα atoms of the Phe-113/Phe-114 and Met-225. This decrease is significant, considering that the side chains of these residues are brought into van der Waals contact with each other. Any further decrease would lead to steric clashes among these residues. Moreover, the conformational changes in the α3/α4- and β6/β7-loops are also associated with rearrangement of the Phe-113/Phe-114 and Met-225 side chains. Notably, Met-225, which in the apo and pseudoapo structures of OXA-58 is oriented toward Phe-114 (as in the cases of OXA-23 and OXA-24), is repositioned so as to point toward Phe-113. As a result, the side chains of Phe-113/Phe-114 and Met-225 are at the proper distance and orientation to establish close hydrophobic interactions. This interaction coincides with formation of a hydrophobic bridge above the active site, like a lid over the acyl-enzyme species, and it exemplifies the plasticity of the OXA-58 active site (Fig. 7). A side-by-side comparison of the molecular surfaces of the OXA-58–6αHMP, pseudoapo, and apo structures indicates that the active-site closure is likely to be initiated upon Michaelis-Menten complex formation and is further established with the acyl-enzyme species formation (Fig. 7).

FIG 7.

Snapshots of active-site closure in OXA-58 induced by 6αHMP. The Phe-113, Phe-114, and Met-225 side chains implicated in the formation of the hydrophobic bridge over the active-site cleft are shown in orange, cyan, and magenta, respectively. Shown are surface diagrams depicting the active-site pocket of the apo structure (A), the pseudoapo structure (B), and the OXA-58–6αHMP complex (C). (D) Side-by-side comparison of the overall structures of apo structure (left) and OXA-58–6αHMP complex (right) depicted in molecular surface diagrams.

The active-site plasticity of OXA-58 allows extensive hydrophobic interactions between 6αHMP and the side chains of Phe-114, Trp-117, and Ile-260 during catalytic turnover. One must infer that β-lactams bind snugly to the active site of OXA-58, and this snugness correlates (in the case of carbapenem substrates) to preferential stabilization of the Δ²-tautomer of the ring-opened carbapenem (Fig. 1) and persistent occupation of the oxyanion hole by the carbonyl oxygen of the acyl-enzyme species. These features play an important role in driving the reaction toward the formation of productive acyl-enzyme species (20, 36).

The above-mentioned hypothesis is well supported by the effects that substitutions at Phe-113, Phe-114, and Met-225 had on the activity of OXA-58. The replacement of Phe with Ile at position 114 had a substantial and deleterious effect on the hydrolysis of imipenem and penicillins, predominantly on the k_cat values (50-fold decrease for imipenem). This substitution may affect the formation of the hydrophobic bridge through a lack of proper orientation of Phe-113 and removal of the sandwiching effect that the β-lactam experiences while bound to the active site (see above). In all, the loss of the snugness in β-lactam binding may lead to lower acylation rates and/or formation of nonproductive acyl-enzyme species; both scenarios will result is smaller k_cat values. The formation of nonproductive acyl-enzyme species may result from a lack of locking of the carbonyl O atom of acyl-enzyme species into the oxyanion hole, and/or hindrance of the carbonyl carbon from the incoming deacylating water molecule. Both scenarios have been reported for the TEM-1–imipenem complex. In that case, the carbonyl oxygen of acyl-enzyme species did not remain bound into the oxyanion hole, and the hydroxyl moiety of the imipenem side chain displaced the deacylating water molecule (PDB code 1BT5) (38).

The replacement of Phe-113 or Phe-114 with Ala led to a decrease in the catalytic efficiency of the enzyme for imipenem and penicillins as a result of a negative effect on both the k_cat and K_m values. However, the replacement of Phe with Tyr at position 113 had very little effect on hydrolysis of all the tested β-lactams, indicating the importance of an aryl ring at this position to interact with Met-225. In addition, the replacement of Met-225 with Ala also had little effect on hydrolysis and might reflect the plasticity of both the loops, α3/α4 and β6/β7, in compensating for a shorter side chain at this position.

The hydrophobic bridge formed in the OXA-58–6αHMP structure was also noted in the native structures of OXA-24 and OXA-23 (13, 15). This hydrophobic bridge remains intact in their respective structures as acyl-enzyme species (13, 36). The native structure of OXA-48 does not show this hydrophobic bridge (16). However, a recent structure of an OXA-48–avibactam acyl-enzyme complex showed a partial closure of the active site as a result of inward movement of the α3/α4-loop (37). Incomplete closure of the active site in OXA-48 is not an indication that OXA-48 is an outlier among CHDLs using this structural feature for the catalysis of carbapenem hydrolysis, or that it brings no functional benefit to catalysis by these enzymes. In fact, the native structure of OXA-48 shows that the active site is more crowded with hydrophobic residues (16). In light of our findings, the crowdedness of the active site may be just a different way to meet the requirement for a tight-fit binding of carbapenem in the active site of this enzyme in comparison to OXA-58, OXA-23, and OXA-24, which may use adjustable active-site closure to meet this requirement.

Despite the tight fit of the 6αHMP acyl enzyme in the OXA-58 active site, we observe that the hydroxymethyl group of 6αHMP occupies two different conformations in the active site. Both conformations of the 6αHMP side chain are stabilized through hydrogen bonding between the hydroxyl group of 6αHMP and either Trp-223 (chains A, C, and D) or carboxylated Lys-86 (chain B) (Fig. 5) and hydrophobic interactions between the 6αHMP side-chain methylene carbon atom and the side chains of the conserved residues Val-132 (4.18 Å apart) and Leu-170 (2.67 Å apart). Docquier et al. (16) proposed that Val-132 and Leu-170 play a role in the optimal positioning of the hydroxyethyl side chain of carbapenems for catalysis by enabling the deacylating water molecule to approach the carbon of the carbonyl of the acyl-enzyme species. It is peculiar that the Ω-loop, which harbors Leu-170, is positioned closer to the active site of OXA-58 than to OXA-23 and OXA-24. This structural feature seems to have an effect on the contact that Leu-170 establishes with the methylene carbon atom of 6αHMP and might be a reflection of an intrinsic property of the Ω-loop to interact with the side chains of substrates, with the purpose of positioning them correctly for retaining the deacylating water molecule.

A comparison of the two rotamers of a 6αHMP hydroxymethyl side chain with the hydroxyethyl side chains of meropenem and doripenem, seen in their respective complexes with OXA-23 (PDB code 4JF4) (13) and OXA-24 (PDB code 3PAG) (36) (Fig. 5), show that the rotamer of 6αHMP seen in chain B is further away from the entrance of the active site and is less populated; also, the rotamer seen in the meropenem/doripenem bound to OXA-24–OXA-23 complex is closer to the entrance of the active site and is the only rotamer seen in these complexes. The rotamer of 6αHMP seen in chains A, C, and D occupies an intermediate conformation between the two rotamers mentioned above and has an occurrence that is at a 3:1 ratio with the rotamer seen in chain B (Fig. 5). Looking at these three rotamers (two rotamers of 6αHMP and the virtually identical rotamers of meropenem and doripenem), a sequence of key events that may take place during carbapenem catalysis emerges. The first rotamer (chain B) may correspond to the acyl-enzyme complex formed from the Michaelis-Menten complex, while the second rotamer (chain A) may represent an intermediate conformation obtained while the hydroxyethyl side-chain samples the conformational space for optimal orientation for catalysis in the active site of CHDLs. The third rotamer (seen in OXA-23–meropenem and OXA-24–doripenem complexes) may show the optimal location of the hydroxyethyl side chain for the last step of catalysis, deacylation. A deacylation step in β-lactam catalysis by OXA β-lactamases requires that a water molecule be in hydrogen bond distance from the carboxylysine active-site residue (Lys-86) and the carbonyl moiety of the acyl-enzyme species. The proposed flow of catalytic events suggests that class D β-lactamases may have acquired carbapenemase activity by selecting a rotamer conformation that allows not simply retention of the deacylating water molecule but its retention in the position between the carboxylysine and the carbonyl of the acyl-enzyme complex, as required for catalytic deacylation.

Our study demonstrates that the active site of OXA-58 has the flexibility to mold around the ligand during catalysis. This feature may be shared and optimized by other CHDLs to provide a tight-fit binding for the carbapenem substrates. In addition, our study suggests that the OXA-58 active site enables the carbapenem substrates to sample and assume an optimum conformation for catalysis. These structural features may enable the formation of a hydrolysis-productive acyl-enzyme species, thereby identifying the essential catalytic difference between the class D β-lactamases capable of carbapenem hydrolysis (the CHDLs) and the class D β-lactamases incapable of carbapenem hydrolysis. Last, it is plausible that in the presence of antibiotic pressure, these structural elements are modulated to achieve catalytic efficiency in carbapenems.