Metallo-β-lactamases withstand low Zn^II conditions by tuning metal-ligand interactions

Abstract

A number of multiresistant bacterial pathogens inactivate antibiotics by producing Zn^II-dependent β-lactamases. We show that metal uptake leading to an active dinuclear enzyme in the periplasmic space of Gram-negative bacteria is ensured by a cysteine residue, an unusual metal ligand in oxidizing environments. Kinetic, structural and affinity data show that such Zn^II-Cys interaction is an adaptive trait tuning the metal binding affinity, thus enabling antibiotic resistance at restrictive Zn^II concentrations.

The efficacy of β-lactam antibiotics is being challenged by the worldwide dissemination of genes encoding metallo-β-lactamases (MβLs).^{1, 2} These hydrolases are able to confer multiresistance to β-lactam antibiotics in many pathogenic and opportunistic bacteria, leading to an urgent need for MβL inhibitors or new generations of β-lactam antibiotics. Most of the clinically relevant targets are in mobile genetic elements and belong to subclass B1.² These enzymes bind up to two Zn^II equivalents, giving rise to a tetrahedral site (M1), with 3 His and a bridging hydroxide (the nucleophile in the hydrolysis reaction) as metal ligands, and a trigonal-bypiramidal site (M2), with a Cys, His and Asp ligand set, completed by two solvent molecules.^{3, 4} Assessing the mechanistic role and essentiality of the M1 and M2 sites in B1 lactamases has proven a difficult task. Some reports suggest that the species with one metal ion located at the M1 site (mono-M1) or at the M2 site (mono-M2) are active,^5–9 while others conclude that only the dinuclear forms are active.¹⁰ Crystal structures obtained for monometallic surrogates disclose a metal ion bound to the M1 site.^11–13 However, such models may not reflect the mononuclear enzymes in vivo, since the thiolate of the Cys ligand (Cys221) appears oxidized, precluding metal binding to the M2 site.^12,13

Cysteine ligands are uncommon in catalytic zinc sites, even more in aerobic oxidizing environments like the periplasm or the extracellular space.¹⁴ Actually, most proteins from the broad MβL Superfamily devoid of β-lactamase activity display a conserved Asp residue in this position, despite exhibiting the same characteristic protein fold.¹⁵ Besides, the fact that the Cys221Asp mutation in the B1 lactamase IMP-1 results in a wild-type like catalytic efficiency makes the absence of the Asp ligand even more puzzling.¹⁶

We obtained the Cys221Asp mutant of BcII, a prototypical B1 enzyme. The activity of this variant in the presence of excess Zn^II paralleled or even surpassed that of wild-type BcII (Supplementary Results, Supplementary Table 1), resembling the situation reported for IMP-1.¹⁶ Instead, BcII-Cys221Asp was inactive when the reaction medium was not supplemented with Zn^II. The hydrolytic activity of the mutant was recovered upon addition of large amounts of the metal ion, in contrast with the wild-type enzyme, which was fully active at low Zn^II concentrations (Supplementary Fig.1).

These results suggest that MβL variants with an Asp ligand might not be able to withstand restrictive Zn^II concentrations. To test this hypothesis, we examined the in vivo performance of wild-type BcII and BcII-Cys221Asp in E. coli cells expressing and secreting these proteins to the periplasmic space. Both proteins accumulated to comparable levels in the periplasm, as revealed by Western blot analysis (Supplementary Fig.2). However, the minimum inhibitory concentrations of ampicillin and cefotaxime were substantially lower for the mutant (cf. 32 and 2 µg/ml respectively, whereas wild-type values were 512 and 32 µg/ml). Addition of EDTA to the growth medium abolished antibiotic resistance in cells expressing BcII-Cys221Asp, while cells expressing the wild-type enzyme proved to be less sensitive to metal limiting conditions (Fig. 1). The antibiotic-sensitive phenotype in cells expressing BcII-Cys221Asp was reverted by addition of extra Zn^II to the growth medium, thus enhancing resistance (Fig.1). In line with these results, the lactamase activity measured with periplasmic extracts was more sensitive to Zn^II availability in the case of BcII-Cys221Asp (Supplementary Fig. 3). We conclude that the in vivo performance of BcII-Cys221Asp is conditioned by Zn^II availability, indicating a decreased metal binding affinity in this mutant.

Figure 1. The capacity to confer resistance of BcII-Cys221Asp enzyme to E. coli cells is impaired under low Zn^II conditions.

The antibiotic sensitivity of E. coli JM109 expressing BcII-Cys221Asp and BcII in the periplasmic space were measured as a function of the Zn^II availability in the growth medium. Colony spots result from serial dilutions of cells transformed with the pKP vector (negative control), pKP-BcII (expressing wild-type BcII), and pKP-BcII- Cys221Asp (expressing the Cys221Asp mutant). Each strain was challenged with ampicillin (32 µg/ml) and cefotaxime (2 µg/ml), under different Zn^II concentrations; namely, low Zn^II (5 µM EDTA), no added Zn^II, and excess Zn^II (250 µM ZnSO₄). These experiments were repeated three times with independent cultures for each condition tested. The length of the white scale bar corresponds to 20 mm.

Dissociation constants for Zn^II were estimated by competition with the chromophoric chelator 4-(2-pyridylazo)-resorcinol (PAR) (Supplementary Fig. 4). For the wild type enzyme, two binding events with similar K_d values in the low-nanomolar range were evident (Supplementary Table 2). Instead, BcII-Cys221Asp displayed markedly different affinities for the two binding sites (K_d1= (9 ± 1.7) nM and K_d2= (267 ± 71) nM), i.e., binding of the second Zn^II equivalent is impaired in the mutant. On the other hand, recovery of the lactamase activity of the apoenzymes upon addition of Zn^II depends on a single binding event, with K_act values of (88 ± 7) nM for wild type BcII, and (7305 ± 365) nM for BcII-Cys221Asp (Supplementary Fig. 1 and Supplementary Methods). These values represent Zn^II dissociation constants in the presence of substrate. Although for wild type and Cys221Asp mutant the affinity for the second metal equivalent is slightly impaired in the presence of substrate compared to the affinity in the resting state, this effect is much more drastic in the Cys221Asp mutant.

The different K_d values in BcII-Cys221Asp allow us to assess the effect of each metal binding site on the enzyme activity. For this purpose, we analyzed the activity of BcII-Cys221Asp at different metal/enzyme ratios at enzyme concentrations high enough to warrant stoichiometric conditions. We employed a rapid mixing device coupled to a spectrophotometer, which allows detection of reaction rates spanning several orders of magnitude. No activity was detected at metal/enzyme ratios ≤ 1, while addition of excess Zn^II gave rise to a fully active enzyme (Supplementary Fig. 5). We conclude that the mononuclear species is not active in BcII-Cys221Asp, and thus the impaired resistance at limiting Zn^II conditions is due to an inability to assemble a functional dinuclear site.

In order to characterize the mono- and dinuclear species at the molecular level, we obtained crystal structures for the mono- and di-Zn^II forms of BcII-Cys221Asp, solved at 1.71 Å and 1.58 Å resolution, respectively (Fig. 2 and Supplementary Table 3). The di-Zn^II species could be obtained only upon soaking the crystals with 20 mM Zn^II. In both cases, the M1 site shows a distorted tetrahedral geometry analogous to those witnessed for available BcII structures, whereas the Asp221 residue conformation depends on the M2 site metal occupancy. In mono-Zn^II BcII-Cys221Asp (Fig. 2a), Asp221 appears salt bridged to the Arg121 guanidinium group, whereas the carboxylate sidechain is rotated 64° in the di-Zn^II variant, coordinated in a bidentate-chelating fashion to the M2 site Zn^II (Fig. 2b). The strong salt-bridge interaction in the mono-Zn^II variant traps Asp221 in a conformation disfavoring metal binding to the M2 site, and is expected to elicit the sequential metal binding mode. On the other hand, the structure of the fully loaded BcII-Cys221Asp metal site closely resembles those of native di-Zn^II B1 enzymes (Supplementary Fig. 6), in line with the high activity determined for this enzyme species. These structures are consistent with the otherwise counterintuitive lower binding affinity provided by an Asp ligand, and reveal that the inactive mononuclear species of BcII-Cys221Asp accumulating at low Zn^II concentrations is mono-M1. Therefore, we conclude that metal binding to the M2 site acts as an activity switch.

Figure 2. Structures of the mono-Zn^II and di-Zn^II binding sites of BcII-Cys221Asp enzyme.

The metal-binding sites of mono-Zn^II (a), and di-Zn^II (b) BcII-Cys221Asp were solved at 1.71 Å and 1.58 Å resolution, respectively. Electron density maps (2F_O – F_C) are shown for Asp221 and Arg121 sidechains, contoured at 1.5 σ. Numbers indicate Zn^II ions. Dotted lines indicate relevant electrostatic interactions. An acetate ion (ACE) was modeled bound to the di-Zn^II form. Red spheres depict metal-bound water molecules.

The different binding affinities in BcII-Cys221Asp enable the study of the M1 mononuclear species and to assess its mechanistic behavior, in contrast to wild-type BcII.¹⁷ Coordination changes in the metal site during enzymatic turnover can be followed via a photodiode array coupled to a stopped-flow device in the Co^II-substituted enzyme. Imipenem hydrolysis by di-Co^II BcII-Cys221Asp proceeds with the accumulation of an intermediate with a strong absorption band at 390 nm (Fig. 3a), which resembles that reported for the same reaction mediated by Co^II-wild type BcII.⁹ When the same experiment was performed in the presence of 0.7 equivalents of Co^II, no imipenem hydrolysis occurred and the spectral features of the mono-M1 species remained unperturbed, revealing that the mononuclear variant, aside from being inactive, is unable to bind imipenem (Fig. 3b).The final spectrum of the reaction catalyzed by the dinuclear enzyme is identical to that of the mono-M1 species (Fig. 3b and Supplementary Fig. 7), suggesting that the M2 site is depleted during catalysis, giving rise to the inactive mono-Co^II variant. These mechanistic studies reveal that the metal ion at the M2 site is essential to stabilize the reaction intermediate, and therefore to render an active enzyme.

Figure 3. The monoCo^II BcII-Cys221Asp is inactive and unable to bind the imipenen substrate.

The hydrolysis of imipenem mediated by di-Co^II BcII-Cys221Asp (a), and mono-Co^II BcII-Cys221Asp (b) was detected via photodiode-array stopped-flow kinetic measurements. Spectra represent absorbance for a time span of 100 milliseconds (from pink to black). Insets show the same data as difference spectra (ΔA) taking the spectrum obtained after 100 milliseconds as reference. The initial spectrum in (a) exhibits an absorption band at 390 nm due to formation of a reaction intermediate, which decays in time together with ligand-field bands in the 600–650 nm region. The spectrum of the mono-Co^II form (peak at 540 nm) (b) remained unchanged after 50 seconds. Reaction medium was 100 mM Hepes pH 7.5, 200 mM NaCl, at 7 °C, with 163 µM for both mono- and di-Co^II species and an initial imipenem concentration of 2.5 mM.

Some authors have suggested that the role of the M2 site is essential for lowering the pK_a of the bridging water molecule.¹⁰ Both mono- and di-Zn^II Cys221Asp display Zn^II-H₂O distances consistent with a bound hydroxide, suggesting that the M1 site suffices for nucleophile activation. This evidence further supports the proposal that the M2 site is essential for substrate binding, C-N bond cleavage and intermediate stabilization in the mechanism of MβLs.^{9, 18} B2 lactamases are active as mononuclear enzymes, with the only metal ion located at the M2 site, further supporting its essentiality for catalysis and antibiotic resistance.¹⁹

Though we were unable to definitively discriminate between apo-, mono-, and dimetallic proteins as directly purified from cells (Supplementary Figure 8), the multiple lines of evidence reported herein support the conclusion that BcII and the clinically relevant B1 MβLs are likely to be active in vivo as dimetallic enzymes. An Asp221 residue decreases the M2 site metal-binding affinity, precluding the accumulation of the functional dinuclear species at sub-stoichiometric metal concentrations. Instead, the native Cys221 residue warrants an active dinuclear enzyme in these conditions, enabling activation of B1 lactamases even at low Zn^IIconcentrations.²⁰

Bacteria are known to concentrate Zn^II ions in the cytosol even under metal limiting conditions.²¹ However, Zn^II bioavailability in the periplasm of Gram-negative bacteria depends largely on its extracellular levels, being particularly challenged by the presence of glutathione and selective ion pumps which deplete the Zn^II content in the periplasm.²² Because MβLs are translocated to the periplasmic space in a Zn(II)-free unfolded form²³, they are expected to have evolved molecular mechanisms to refold in Zn(II)-deficient environments, ensuring survival of the host when exposed to β-lactam antibiotics. This study shows that natural selection has favored a Cys221 ligand in B1 lactamases to warrant formation of the dinuclear species in these conditions, despite being an unusual metal ligand for such a catalytic zinc site functional in aerobic environments. In addition, this work suggests that the active form of B1 lactamases contains a Zn^II ion in the M2 site, and inhibitor design should target this metallated species.

Abbreviations

MβL

metallo-β-lactamase

LMCT

ligand-to-metal charge-transfer

EDTA

ethylene-diamine-tetraacetic acid

PAR

chelator 4-(2-pyridylazo)-resorcinol.