Interactions of Oximino-Substituted Boronic Acids and β-Lactams with the CMY-2-Derived Extended-Spectrum Cephalosporinases CMY-30 and CMY-42

Abstract

CMY-30 and CMY-42 are extended-spectrum (ES) derivatives of CMY-2. ES characteristics are due to substitutions of Gly (CMY-30) and Ser (CMY-42) for Val211 in the Ω-loop. To characterize the effects of 211 substitutions, we studied the interactions of CMY-2, -30, and -42 with boronic acid transition state inhibitors (BATSIs) resembling ceftazidime and cefotaxime, assessed thermal stability of the enzymes in their free forms and in complexes with BATSIs and oximino-β-lactams, and simulated, using molecular dynamics (MD), the CMY-42 apoenzyme and the CMY-42 complexes with ceftazidime and the ceftazidime-like BATSI. Inhibition constants showed that affinities between CMY-30 and CMY-42 and the R1 groups of BATSIs were lower than those of CMY-2. ES variants also exhibited decreased thermal stability either as apoenzymes or in covalent complexes with oximino compounds. MD simulations further supported destabilization of the ES variants. Val211Ser increased thermal factors of the Ω-loop backbone atoms, as previously observed for CMY-30. The similar effects of the two substitutions seemed to be due to a less-constrained Tyr221 likely inducing concerted movement of elements at the edges of the active site (Ω-loop–Q120 loop–R2 loop/H10 helix). This inner-protein movement, along with the wider R1 binding cleft, enabled intense vibrations of the covalently bound ceftazidime and ceftazidime-like BATSIs. Increased flexibility of the ES enzymes may assist the productive adaptation of the active site to the various geometries of the oximino substrates during the reaction (higher frequency of near-attack conformations).

INTRODUCTION

CMY-2-type cephalosporinases are widespread plasmidic class C β-lactamases closely related to chromosomal AmpC of Citrobacter freundii. They are highly active against some penicillins (e.g., benzylpenicillin and ampicillin), older cephalosporins, and cephamycins, while their efficiency against expanded-spectrum cephalosporins (ESCs) with an oximino side chain, such as ceftazidime (CAZ) and cefotaxime (CTX), is relatively low. Recently, various class C β-lactamases, including CMY-2 variants, with enhanced activity against ESCs have been characterized (1–8).

According to kinetic analysis of ESC hydrolysis by CMY-2, the antibiotics rapidly acylate the catalytic serine, but the acyl enzyme is hydrolyzed at low rates, a feature common among class C β-lactamases (9). Structural studies with Escherichia coli AmpC (AmpC^{E. coli}) have suggested that the “inhibitory” behavior of ESCs is due to steric clashes of the bulky oximino substituent with the side chains of the conserved Val211 (Ω-loop) and Tyr221 (H7 helix), which restrain the movement of the R1 oximino group, thus hindering formation of the deacylation tetrahedral high-energy intermediate (10). Efficient hydrolysis of ESCs by AmpC mutants in which the Val211-Tyr221 surface has been disrupted (e.g., in the GC1 enzyme [Ala208-Val209-Arg210 Ω-loop duplication], ACC-4 and CMY-30 [Val211Gly], and the Tyr221Gly laboratory mutant of AmpC^{E. coli}) (6, 7, 11–13) do not contradict the above notion. Yet molecular dynamics simulations of CMY-2 and CMY-30 indicated that the extended-spectrum (ES) properties conferred by Gly-for-Val211 substitution were associated with complex changes in the CMY structure, such as increased mobility of elements forming the active site, including the Ω-loop (14). Also, the recent characterization of a Val211Ser variant of CMY-2 (CMY-42) with ES activity (2) further suggested that apart from abolition of steric clashes between the 211 side chain and the oximino substituents of ESCs, additional mechanisms are involved in the increasing turnover rates against the latter drugs.

We attempted here to study the mechanisms conferring ES properties to CMY-30 and CMY-42. For this purpose, the interactions between the enzymes and boronic acid transition state inhibitors (BATSIs) carrying the oximino side chains of CAZ and CTX (15) were studied. The effect of the Val211Gly and Val211Ser substitutions on stability of the free enzymes and their complexes with BATSIs and oximino-β-lactams was assessed by determining the respective melting temperatures. Also, CMY-42, in its free form and in covalent complexes with CAZ and the CAZ-like BATSI, was simulated using molecular dynamics. Data revealed common structural and functional features likely associated with the ES activity of CMY-30 and CMY42.

MATERIALS AND METHODS

Bacterial strains and plasmids.

E. coli MC100 (ΔampC) was utilized as the host strain in cloning experiments. The strain was transformed with the pB-cmy-2, pB-cmy-30, and pB-cmy-42 derivatives of the high-copy-number vector pBC-SK (+/−) and with pA-cmy-2, pA-cmy-30, and pA-cmy-42, derived from the pACYC184 low-copy-number vector. All plasmids but pA-cmy-42 have been described previously (2, 6). pA-cmy-42 was derived from pA-cmy-2 (6) by site-directed mutagenesis using the QuikChange mutagenesis kit (Stratagene) and the mutagenic primers CMY-V211S-F and CMY-V211S-R (2). β-Lactam susceptibility of E. coli clones harboring pACYC184 derivatives encoding CMY-type enzymes under isogenic conditions was determined by the Etest (bioMérieux).

Enzyme expression and purification.

The pB-cmy-carrying strains overexpressing the CMYs were used for enzyme purification as previously described (6). Briefly, cells from overnight cultures in 3 liters of LB were harvested by centrifugation and resuspended in Tris-HCl (20 mM, pH 8.3). Periplasmic proteins were released by sonication. Cell lysates were clarified by centrifugation, concentrated by ultrafiltration, and loaded on Q-Sepharose columns (Bio-Rad). β-Lactamase-containing effluents were subjected to buffer exchange by gel filtration (10DG desalting columns; Bio-Rad) and 20 mM NaP_i (pH 7). Preparations were loaded on S-Sepharose, and the bound β-lactamases were eluted with a 0 to 0.5 M NaCl gradient. Fractions exhibiting β-lactamase activity (as determined by a nitrocefin assay) were pooled, dialyzed against 100 mM NaCl–50 mM KP_i (pH 7.0), and concentrated by ultrafiltration. The protein concentration was determined by the Bradford method. Purity of the enzyme preparations was higher than 95% as determined by SDS-PAGE. The yield of the CMY-30 enzyme was approximately 15% lower than those of CMY-2 and CMY-42.

The exact masses of CMY-2, CMY-30, and CMY-42 were determined using matrix-assisted laser desorption–ionization time of flight (MALDI-TOF) mass spectrometry on an Autoflex mass spectrometer (Bruker Daltonics). Before injection, the samples were desalted by four cycles of concentration dilution in 5% acetonitrile (ACN)–0.1% trifluoroacetic acid (TFA) using Amicon centrifugal filters (3-kDa cutoff; Millipore). For spectrum acquisition, 1 μl of each sample was spotted on a steel MALDI target, mixed (on target) with Matrix-II solution (0.8% a-cyano-4-hydroxycinnamic acid in 50% ACN and 0.1% TFA; Sigma-Aldrich), and then allowed to dry. The determined molecular masses of the three enzymes differed slightly from the expected ones of the mature proteins (39,853.4 versus 39,854.6 Da for CMY-2, 39,804.2 versus 39,812.5 Da for CMY-30, and 39,843.8 versus 39,842.6 Da for CMY-42).

Synthesis of boronic acid transition state inhibitors.

The achiral boronic acid transition state analogues of CAZ and CTX, 1 and 3 (Fig. 1), respectively, were synthesized as previously described (15, 16). Enantioselective synthesis of the corresponding chiral boronic acid transition state analogues 2 and 4 (Fig. 1) was performed as described in the supplemental material.

Fig 1.

Thermal stabilities of CMY β-lactamases in their free forms and in complexes with oximino compounds. (A) DSF melting curves of the free forms of CMY-2, CMY-30, and CMY-42. The melting temperature (T_m) of each enzyme was obtained by estimation of the turning point of the curve of relative fluorescence intensity versus temperature. T_m^CMY-2 = 55.3 ± 0.17°C; T_m^CMY-30 = 52.3 ± 0.10°C; T_m^CMY-42 = 54.0 ± 0.14°C. (B to F) Effects of three concentrations of each oximino compound (10, 50, and 100 μM) on the melting temperatures of CMYs expressed as differences of T_ms (ΔΤ_m = T_m^CMY/compound − T_m^CMY). The structures of aztreonam (B), glycyl boronic acid (BA) analogue of CAZ (C) (compound 1), phenylalanyl BA analogue of CAZ (D) (compound 2), glycyl BA analogue of CTX (E) (compound 3), the phenylalanyl BA analogue of CTX (F) (compound 4), and CAZ and CTX (G) are also shown. Points on the graphs represent mean values of three independent measurements; bars indicate standard deviations. The oximino compounds stabilized all the CMY enzymes, with the effects on ES variants being significantly weaker. The BATSIs appeared to increase the stability from 50 to 100 μM, although according to the measured K_is and the concentration of the enzymes, the stabilization effects should be constant as saturation was reached. This was most likely due to the strong binding of SYPRO to hydrophobic areas of proteins, such as the active site, thus competing with BATSIs. This would cause the binding constants of the latter compounds to appear higher than the measured K_i values during these experiments.

Inhibition studies.

Inhibition constants (K_is) characterizing the interaction of BATSIs with CMY-2, -30, and -42 were determined by measuring initial velocities of cephalothin hydrolysis in the absence and presence of each inhibitor at various concentrations. Reactions were carried out in 50 mM KP_i (pH 7) at 25°C. Each enzyme was used at a final concentration of 0.4 nM. Absorbance differences were monitored using a Hitachi U-2001 UV/Vis spectrophotometer. Results were fitted to equation 1, characterizing “mixed” inhibition (17) using the Prism software program (GraphPad Software Inc.).

$$V=\frac{V_{\max }·\left[S\right]}{\left[S\right]·\left(1+\frac{\left[I\right]}{\alpha ·K_i}\right)+K_m·\left(1+\frac{\left[I\right]}{K_i}\right)}$$ (1)

In all cases, the value of the α parameter obtained by nonlinear regression was essentially infinite (α > 10¹³), indicating a competitive mode of inhibition (17). When K_i was <1,000 · [E₀], inhibition constants were determined by nonlinear regression analysis using the Morrison equation (equation 2) (17, 18) with the apparent inhibition constant (K_i^app) being that of competitive inhibition (equation 3).

$$\frac{Vi}{Vo}=1-\frac{\left(\left[E\right]+\left[I\right]+K_i^{\text{app}}\right)-\sqrt{{\left(\left[E\right]+\left[I\right]+K_i^{\text{app}}\right)}^2-4\left[E\right]\left[I\right]}}{2\left[E\right]}$$ (2)

$$K_i^{\text{app}}=K_i·\left(1+\frac{\left[S\right]}{K_m}\right)$$ (3)

The chiral boronic acid analogues of CAZ and CTX (compounds 2 and 4) exhibited time-dependent inhibition, as has previously been observed with other BATSIs during interaction with other AmpCs (16) and SHV-type β-lactamases (19). Therefore, the respective competition experiments were preceded by a 10-min preincubation period using cephalothin at a concentration of 15 μM. K_i values were determined using equations 2 and 3.

Thermal stability experiments.

The relative thermal stabilities of the three enzymes in their free and complex forms were estimated by measuring their melting temperatures (T_m) through differential scanning fluorimetry (20, 21). Thermal denaturation was carried out by increasing the temperature from 28 to 75°C (2°C/min) in a MiniOpticon instrument (Bio-Rad) in the presence of 5× SYPRO orange (Invitrogen). Assays were performed in 25 mM NaPi (pH 7)–150 mM NaCl using 3 μΜ of each enzyme in a final volume of 20 μl. Three concentrations (10, 50, and 100 μΜ) of each oximino compound were tested with boronic acids analogues being assayed after a 15-min preincubation at 25°C. Melting temperatures were determined based on the sigmoid curve of differences in fluorescence (caused by the binding of SYPRO orange to the hydrophobic areas of each enzyme exposed during denaturation) as a function of temperature. Reversibility of thermal denaturation was assessed by heating the enzymes at 60°C for 5 min and then measuring the residual activity after cooling the preparations on ice. The low residual activity (<20%) suggested an irreversible thermal denaturation for all three enzymes.

Molecular simulations.

The CMY-42 apoenzyme, ceftazidime acyl enzyme, and enzyme complex with the CAZ boronic acid analogue (deacylation transition-state analogue) were simulated for 5 ns using molecular dynamics simulations, as was done previously with CMY-2 and CMY-30 (14). Details of the computational procedures followed are described in the supplemental material.

RESULTS AND DISCUSSION

Turnover rates of ESCs and inhibitory activity of BATSIs.

CMY-30 (Gly211) and CMY-42 (Ser211) hydrolyzed oximino-β-lactams more efficiently than CMY-2 (Val211) due to higher turnover rates (k_cat) that likely reflected a faster deacylation (Table 1). CMY-30 operated at higher rates than CMY-42 against CTX and aztreonam (ATM), though the turnover rates of CAZ were similar. The increased hydrolysis of oximino-β-lactams by these ES variants was also depicted in the resistance profiles of E. coli strains expressing the enzymes under isogenic conditions (Table 1).

Table 1.

Interactions of CMY-type cephalosporinases with compounds carrying oximino R1 side chains

Compound	CMY-2 (Val211)				CMY-30 (Gly211)				CMY-42 (Ser211)
	Km or Ki (μΜ)	kcat (s−1)	kcat/Km (μΜ−1 · s−1)	MIC (μg/ml)	Km or Ki (μΜ)	kcat (s−1)	kcat/Km (μΜ−1 · s−1)	MIC (μg/ml)	Km or Ki (μΜ)	kcat (s−1)	kcat/Km (μΜ−1 · s−1)	MIC (μg/ml)
Ceftazidimea	0.02 ± 0.003	0.01 ± 0.002	0.5	128	0.14 ± 0.02	0.4 ± 0.05	2.9	>512	0.30 ± 0.08	0.50 ± 0.02	1.7	>512
Cefotaximea	0.005 ± 0.001	<0.01	<2	32	0.30 ± 0.02	1.7 ± 0.1	5.7	192d	0.08 ± 0.03	0.20 ± 0.03	2.5	256d
Aztreonama	0.002 ± 0.001	<0.001	<0.5	16	0.017 ± 0.004	0.02 ± 0.005	1.2	96	0.006 ± 0.002	0.003 ± 0.001	0.5	48
1	0.27 ± 0.06	NAb	NA	NDc	0.65 ± 0.12	NA	NA	ND	0.64 ± 0.11	NA	NA	ND
2	0.08 ± 0.01	NA	NA	ND	0.30 ± 0.03	NA	NA	ND	0.40 ± 0.03	NA	NA	ND
3	2.2 ± 0.3	NA	NA	ND	7.1 ± 1.5	NA	NA	ND	4.2 ± 0.9	NA	NA	ND
4	0.53 ± 0.04	NA	NA	ND	1.7 ± 0.1	NA	NA	ND	0.80 ± 0.07	NA	NA	ND

^aKinetic constants were taken from reference 2.

^bNonapplicable.

^cNot determined.

^dThe discrepancy between cefotaxime MICs and hydrolytic efficiencies of CMY-30 and CMY-42 was likely due to the smaller amount of the former enzyme in the periplasm of the respective clone.

Apparent affinities of CMY-30 and CMY-42 for oximino substrates were lower than those of CMY-2 (Table 1). However, K_m constants can be affected by hydrolysis rates of the ester. Therefore, inhibition by glycyl- and phenylalanyl-boronic acids bearing the R1 side chains of CAZ (identical to R1 of ATM) and CTX was studied in order to accurately determine affinities of CMYs for these structures. Inhibition constants showed that CMY-2 exhibited higher affinity than CMY-30 and CMY-42 for all the BATSIs tested, although differences were less pronounced than those seen with oximino-β-lactams (Table 1). CMY-30 and CMY-42 bound the glycyl-boronic acid analogue of CAZ (compound 1) with similar affinities that were 2.4 times weaker than that determined for CMY-2. CMY-42 had the lower affinity for the chiral boronic acid analogue of CAZ (compound 2), followed by CMY-30 and CMY-2 (Table 1). The boronic acid analogues of CTX exhibited weaker interactions with the three enzymes than their CAZ-like counterparts, probably due to the absence of a carboxyl group in the R1 side chain. CMY-30 had the lower affinity for both the glycyl and chiral analogues of CTX (compounds 3 and 4, respectively). CMY-42 interacted more tightly with these compounds than CMY-30, while its inhibition was weaker than that of CMY-2 (Table 1).

Overall, inhibition experiments using BATSIs indicated that Val211Gly and Val211Ser substitutions reduced affinities for the R1 side chains of ESCs and aztreonam. Also, differences in the inhibitory activity between the CAZ-like compounds 1 and 2 showed that addition of a meta-carboxybenzyl group (simulating the carboxylated dihydrothiazine ring of ESCs) strengthened interactions mostly with CMY-2, while the corresponding differences between the CTX-like compounds 3 and 4 suggested that the meta-carboxybenzyl group increased affinity for the three CMYs to similar degrees. These findings indicated that 211 substitutions may indirectly affect ionic interactions between the CAZ carboxylates and positively charged active side residues.

Thermal stabilities of free enzymes and complexes.

Melting temperatures showed that the ES variants were less stable than CMY-2. Val211Gly had the strongest effect on thermal stability, lowering the T_m by 3°C, while Val211Ser reduced the T_m by 1.3°C (Fig. 1A). Additionally, CMY-30 and CMY-42 were stabilized by ΑΤΜ and boronic acid analogues of CAZ and CTX to a lesser extent than CMY-2 (Fig. 1B to F). CAZ and CTX did not affect the T_ms of CMY-30 and CMY-42 at any of the concentrations tested, most likely due to increased hydrolysis, while they increased the T_m of CMY-2 at concentrations of 100 μM (3.2 and 3.9°C, respectively). ATM stabilized all three enzymes to a greater extent, with CMY-2 being highly affected, followed by CMY-42 and CMY-30 (Fig. 1B). The effects of oximinoboronic acids on enzymes' thermal denaturation exhibited a pattern similar to that observed with ATM. Again, CMY-30 was stabilized to a lower degree than CMY-42 which, in turn, was less affected than CMY-2 (Fig. 1C to F). Notably, the chiral boronic acid analogues exhibited a weaker effect on enzymes' stabilization than the glycyl analogues, although their binding with the CMYs was stronger. Considering the structural difference of these molecules (Fig. 1C to F), it can be assumed that the presence of a large carboxylated substitute analogous to the dihydrothiazine ring and the R2 side chain of oximinocephalosporins counteracted the stabilization effect of the oximino R1 side chains.

The reduced thermal stability of CMY-30 and CMY-42 probably resulted from an increased mobility of the polypeptide backbones. The ES variants also became even more unstable than CMY-2 when interacting with oximino compounds. The effect of β-lactams on thermal denaturation of each enzyme may be dependent on hydrolysis rates, the magnitude of dissociation constants, and/or the degree of flexibility of the covalent acyl enzyme complex (the last two factors are also applied for boronic acids). In the cases of CAZ and CTX interactions with CMY-30 and CMY-42, the lack of β-lactam stabilization effects was likely due to elimination of the substrate before the T_m was reached. However, given that ATM and BATSIs affected the T_ms of CMY-30 and CMY-42 less even at saturating concentrations (100 μM), the likely reason for the observed phenomena was the increased flexibility of the acyl and boronate enzyme backbones. Thus, it can be supported that the complexes of the ES variants were less stable than the respective complexes of CMY-2. These differences were more pronounced than those observed for the apoenzymes.

Increased catalytic efficiency is correlated with low thermal stability in a wide range of enzymes, including β-lactamases (13, 22–24). Data presented here suggested that the activity-for-stability “trade-off” may also apply to the 211 ES variants of CMY-2.

Molecular dynamics of apoenzymes and complexes.

The lower stability of CMY-30 and CMY-42 and their reduced affinity for the R1 side chains of oximino-β-lactams compared to findings for CMY-2 indicated the complexity of effects induced by mutations at position 211. We proposed in a previous study that the ES properties of CMY-30 were possibly linked to intense correlated movements of active site elements assisting collapse of acyl enzymes (14). Here, to better understand the hydrolytic behavior of the 211 variants, we studied by molecular dynamics simulations the complexes of CAZ and the achiral CAZ-like BATSI with CMY-42.

Estimation of backbone atoms' thermal factors (which reflect the degree of atomic fluctuations during the trajectories and hence can be used as a measurement of a polypeptide's chain flexibility) revealed common effects of the Val211Gly and Val211Ser substitutions. The most pronounced of those effects concerned the Ω-loop and its neighboring elements. In simulations of free enzymes, both CMY-30 and CMY-42 exhibited higher vibrations in the 210–213 area of the Ω-loop, while the 205–208 segment was less motile than that of CMY-2 (Fig. 2A, upper panel). The Ω-loop in CAZ acyl enzymes exhibited lower vibrations in the ES variants, but the thermal factors of the adjoining the H7 helix (containing Tyr221) were increased compared to the respective adduct of the parental enzyme (Fig. 2A, middle panel). The boronate enzyme complexes exhibited features similar to those of the free enzymes regarding Ω-loop vibrations (Fig. 2A, lower panel). Furthermore, the 211 mutants exhibited high backbone flexibility in the loop connecting B2 and B2a β-sheets in the majority of the simulated species, with vibrations in CMY-42 being more intense. The Val211Ser substitution affected the vibrations of the H10 helix and the R2 loop, but as was observed for Val211Gly (14), the effects were diverse depending on the simulated species (Fig. 2A).

Fig 2.

(A) Backbone atoms' b-factor profiles of CMY apoenzymes, CAZ acyl enzymes, and complexes with the achiral CAZ-like BATSI (compound 1). In each graph, the profile of CMY-2 (black) is compared to those of the ES variants (gray) (CMY-30, left panel; CMY-42, right panel). The common effects of the Val211Gly and Val211Ser substitutions on vibrations of the Ω-loop are apparent. At the top of each column, the enzyme's secondary structure is also shown. (B) Ribbon diagrams of the average structures of CAZ acyl enzymes as determined by the 5-ns trajectories. Elements forming the active site are highlighted. Segments of the Ω-loop and H7 helix are colored according to the effects the two substitutions in the various simulations (orange, 211 to 221, increased vibrations; green, 205 to 208, decreased thermal factors). CAZ, Tyr221, and residues in position 211 are shown as sticks. The side chain of Ser211 in CMY-42 is relocated away from Tyr221 at a position allowing interaction with Glu61 (right panel). At this position, Ser211 also formed frequent hydrogen bonds with the amine of the aminothiazole ring of CAZ.

In line with the results of thermal stability experiments, the b-factor profiles showed that the ES variants had a more flexible backbone than CMY-2 (Fig. 2A), with the Ω-loop probably playing an important role in folding/unfolding equilibria. Examining the most common structures generated by the 5-ns trajectories, a likely reason for the observed phenomena in the Ω-loop and H7 helix emerged: in CMY-30 and CMY-42, the side chain of Tyr221 had higher degrees of freedom than in CMY-2, where its interaction with Val211 restricted its movement. The loosening of Tyr221 in CMY-30 was due to the absence of a 211 side chain, while in CMY-42, this was induced by the shift of the Ser211 side chain toward B3 (i.e., in the opposite direction from the Tyr221-containg H7 helix), where it interacted with the carboxylate of Glu61 through its hydroxyl group (Fig. 2B).

Interpretation of the observed atomic fluctuations through covariance analysis revealed similar motion patterns in the two ES enzymes. First, in CMY-30 and CMY-42 apoenzymes, the Ω- and Q120 loop moved in the same direction (correlated movement) and in a concerted manner with the H10 helix-R2 loop, while in CMY-2 this movement was significantly less intense (Fig. 3, upper panel; see also Fig. S2 in the supplemental material). In CAZ acyl enzyme simulations, the above concerted movement was weakened in CMY-42 compared to that in CMY-30. Nevertheless, in both enzymes the movements of the covalently bound CAZ were correlated with those of the Ω-loop, the H7 helix, and the side chain of Tyr221 and were more intense than those for the CMY-2 acyl enzyme (Fig. 3, middle panel). The bound CAZ-like BATSI was also highly motile in the ES enzymes compared to CMY-2, with the respective movements again being correlated with the vibrations of the aforementioned structural elements (Fig. 3, lower panel).

Fig 3.

Porcupine plots depicting the principal motions in the three simulated species of each CMY enzyme. Drawing of plots was carried out with the aid of the VMD software program and the Tcl scripting language using the two extreme projections of the first eigenvector (corresponding to >20% of the observed atomic fluctuations) extracted from covariance analysis applied on the stable phase of each simulation. In apoenzyme simulations, the main motion in the area of the CMY-2 active site concerned an in-and-out vibration of the R2 loop. In contrast, in the CMY-30 and CMY-42 apoenzymes, the Ω-loop/H7 helix, Q-120 loop, and R2 loop/H10 helix exhibited increased correlated movement (i.e., at the same direction). In CAZ acyl enzymes (middle panel), this correlated movement was apparent only in CMY-30. Nevertheless, the bound compounds vibrated to a higher degree in both ES variants than in CMY-2. Motions of the oximino R1 side chain of CAZ in the active site of CMY-30 and CMY-42 were in concert with those of the terminal segment of the Ω-loop and the H7 helix that contains Tyr221. Vibrations of the covalently bound CAZ-like BATSI (lower panel) were also higher in the ES variants, while in CMY-2 the compound was relatively rigid. The fact that the side chain of Tyr221 in these complexes did not perform the motions observed in acyl enzymes may indicate that the dihydrothiazine ring of CAZ exerted tensions on this residue through the R1 side chain. In ES variants, these tensions resulted in the movement of Tyr221 due to detachment from residue 211 In the CMY-2 acyl enzyme, the force applied on the Tyr221-Val211 surface caused an intense movement of the side chain of Val211, thus retaining its ability to hinder the movement of R1.

The above data showed that CMY-30 and CMY-42 share common dynamic features distinguishing them from parental CMY-2. Substitution of either Gly or Ser for Val211 did the following: (i) increased flexibility of the enzyme's backbone, especially of the Ω-loop, (ii) intensified the correlated movement of the Ω-loop, the Q120 loop, and the R2 loop/H10 helix, and (iii) increased vibrations of ceftazidime and its deacylation transition state boronic acid analogue in covalent enzyme complexes. Destabilization of bound ceftazidime was most probably facilitated by the existence of a wider space in the R1 binding cleft that resulted from the removal of the side chain at 211 (CMY-30) or the relocation of the serine side chain (CMY-42), which enabled movement of the R1 side chain. Additionally, the increased motility of the Ω-loop–Q120 loop–R2 loop/H10 helix system could also contribute to the observed vibrations of ceftazidime by modulating the movement of the antibiotic's dihydrothiazine ring.

Implications of molecular dynamics for turnover rates.

The observed destabilization of the active site during simulations of ES enzymes' complexes may lead to changes in affinity of CMY-30 and CMY-42 for oximino compounds. Analysis of salt bridge formation between the two carboxylate groups of ceftazidime (i.e., the carboxyl groups attached to the oximino side chain and C-3 of the dihydrothiazine ring) and the guanidinium groups of Arg204 and Arg349 flanking the active site revealed that in CMY-2 these interactions were stronger (Fig. 4A). In the vast majority of structures generated during the simulation of CMY-2/CAZ acyl enzyme, the dihydrothiazine carboxylate was in salt bridge formation distance from the side chain of Arg349. In ES acyl enzymes, such interactions were either less frequent (CMY-30) or absent (CMY-42) (Fig. 4A, left panel). This was apparently due to the displacement of dihydrothiazine from the Arg349-containing α/β domain. An additional reason for the loosening of the above interaction may be the interference of the R1 carboxylate, which may occasionally form salt bridges with Arg349 (Fig. 4A, middle panel) as a result of the motions of the diethyl-methyl-carboxyl group attached to the oximino group of R1. The dynamic features of the R1 side chain of CAZ in the Gly211 and Ser211 enzymes, along with the altered vibrations of the Ω-loop, also resulted in the weakening of the R1 carboxylate's interactions with Arg204 (Fig. 4A, right panel). Notably, it has been reported that P99 laboratory mutants possessing Val, Phe, or Ser instead of Arg204 exhibited higher activity against CAZ (25). It can thus be hypothesized that the loosening of the Arg204 interaction with R1 of CAZ observed in the ES CMY variants could be implicated in the increased turnover rates. Further investigations should be carried out in order to elucidate the role of this residue.

Fig 4.

(A) Distribution of distances between the two CAZ carboxyl groups and the side chains of Arg349 and Arg204 that flanked the active site (black, CMY-2; gray, CMY-30; red, CMY-42). The values correspond to distances between the center of COO⁻ and the guanidinium groups. “Frequency” denotes the absolute number of structures. Superimposed structures of the three acyl enzymes are also shown (carbon atoms are colored green in CMY-2, purple in CMY-30, and orange in CMY-42). The C-3 carboxylate of CAZ interacted strongly with Arg349 in CMY-2, while in ES, acyl enzyme displacement of dihydrothiazine reduced the frequency of salt bridge formations (left panel). Arg349 in ES variants interacted with the carboxylate of the oximino R1 side chain during a short period of the simulations (middle panel) due to rotational movements of the diethyl-methyl-carboxyl group. These movements, in conjunction with the relocation of the side chain of Arg204 in CMY-30 and CMY-42, caused the abolishment of the R1 COO⁻-Arg204 salt bridge that was frequent in CMY-2 (right panel). (B) Superimposition of the most frequent conformers of CAZ acyl enzymes with the average structures of the CAZ-like BATSI complexes of each enzyme. The carbon atoms in the CAZ and BA complexes are colored orange and purple, respectively. The higher likeness of the common atoms of CAZ and its deacylation transition state analogue is apparent in the ES enzymes, especially CMY-42. The oxygen of the BA (O2) that is analogous to that of the tetrahedral deacylation adduct is positioned at a higher distance from the nitrogen of CAZ dihydrothiazine (N1) in ES variants, suggesting a less obstructed formation of these species. The displacement of Tyr150 away from the dihydrothiazine ring in CMY-30, which left a wider space in the β-face of the esteric bond, is also shown. (C) Simulation snapshots showing the structure and dynamics of the water environment in the vicinity of the esteric bond between CAZ and Ser64. Water molecules occupying hydration sites are colored based on their residence times that were calculated through integration of the autocorrelation function of contacts of the solvent's oxygen atoms and the groups of protein and CAZ using a cutoff of 3.6 Å (green, τ = 0 to 50 ps; orange, τ = 50 to 1,000 ps). The abundance of hydration sites and their high residence times in ES enzymes denotes the less obstructed entrance of water molecules in the β-face of the esteric bond.

The increased flexibility of the Gly211 and Ser211 enzymes, along with the higher vibrations of the bound oximino compounds, could also affect the catalytic process by increasing deacylation rates. Superimposition of the most frequent structures of CAZ acyl enzyme with the boronate complexes (analogous to the deacylation transition state) of each enzyme showed that the active site's heavy atoms of the two kinds of species shared a higher resemblance in ES enzymes than in CMY-2 (see Table S1 in the supplemental material). Similar observations were made when the common atoms of CAZ and the boronic acid analogue were compared; in fact, in CMY-42, their positions were nearly identical (Fig. 4B). We propose that ceftazidime in ES acyl enzymes more frequently adopts conformations that are nearer the transition state, and therefore the hydrolytic attack can proceed at higher rates.

The higher possibility for “near-attack conformations” in CMY-30 and CMY-42 was also indicated by the distance of the boronic acid oxygen (representing the oxygen of the attacking water in the tetrahedral deacylation intermediate) from the dihydrothiazine's ring nitrogen. In CMY-2, the two atoms were close to each other, suggesting that the tetrahedral intermediate cannot be formed due to a steric clash between the open β-lactam ring nitrogen and the oxygen of the attacking water, while in CMY-30 and CMY-42, the positioning of the nitrogen atom would permit formation of the deacylation high-energy adduct (Fig. 4B; see also Table S1 in the supplemental material).

Analysis of the water environment around the β-face of the esteric bond further depicted the more efficient hydrolytic attack in the ES acyl enzymes. In CMY-2, the presence of a water molecule in a position that would allow nucleophilic attack on the CAZ-Ser64 bond was a rare event, while in CMY-30 and CMY-42, the deacylation-competent hydration site was frequently occupied by a water molecule (Fig. 4C). Moreover, the other hydration sites in the vicinity of the hydrolytic water also exhibited higher residence times in the ES enzymes. Entrance of water molecules in the area around the β-face of the esteric bond in CMY-2 was hindered by dihydrothiazine and the side chain of Tyr150. In CMY-30, the vibrations of the dihydrothiazine ring combined with the higher motility of Tyr150 seemed to allow access of water molecules into the area of the acylated serine. In CMY-42, water access was facilitated primarily by the altered position of the dihydrothiazine ring (see Movies S1 to S3 in the supplemental material).

Concluding remarks.

The emergence of CMY-30 and CMY-42 underscores the potential of one of the commonest plasmid-borne cephalosporinases, CMY-2, to significantly enhance its hydrolysis spectrum toward expanded-spectrum β-lactams through single point mutations. Data presented in this study indicate that the conserved Val211 is an important determinant of substrate specificity. Indeed, its replacement by either glycine or serine led to a wide range of functional and structural changes, such as a reduced affinity for oximino moieties, increased turnover rates of expanded-spectrum β-lactams, and decreased thermal stability of the enzyme both in its free form and in covalent complexes with oximino-substituted compounds. It seems that the core of the observed functional alterations is the disengagement of the side chain of Tyr221, leading to increased flexibility of the backbone in the vicinity of residue 211. This in turn triggers a concerted movement of elements in the outer edges of the active site (Ω-loop, Q120 loop, R2 loop/H10 helix). It is remarkable that in CMY-30 and CMY-42, a similar series of events was induced by an amino acid lacking a side chain (Gly) and a polar one (Ser). It would be interesting to search for similar characteristics in other cephalosporinases with ES activity. Notwithstanding the possible existence of common patterns associated with enhanced ESC hydrolysis among ES AmpCs, it seems that conformational flexibility of the CMY-2-type enzymes plays a pivotal role in the interaction with oximino-β-lactams. The dynamic nature of this interaction has also been evidenced in structural and kinetic studies with other AmpCs (26, 27).