Abstract
In most bacteria, β-lactam antibiotics inhibit the last cross-linking step of peptidoglycan synthesis by acylation of the active-site Ser of D,D-transpeptidases belonging to the penicillin-binding protein (PBP) family. In mycobacteria, cross-linking is mainly ensured by L,D-transpeptidases (LDTs), which are promising targets for the development of β-lactam-based therapies for multidrug-resistant tuberculosis. For this purpose, fluorescence spectroscopy is used to investigate the efficacy of LDT inactivation by β-lactams but the basis for fluorescence quenching during enzyme acylation remains unknown. In contrast to what has been reported for PBPs, we show here using a model L,D-transpeptidase (Ldtfm) that fluorescence quenching of Trp residues does not depend upon direct hydrophobic interaction between Trp residues and β-lactams. Rather, Trp fluorescence was quenched by the drug covalently bound to the active-site Cys residue of Ldtfm. Fluorescence quenching was not quantitatively determined by the size of the drug and was not specific of the thioester link connecting the β-lactam carbonyl to the catalytic Cys as quenching was also observed for acylation of the active-site Ser of β-lactamase BlaC from M. tuberculosis. Fluorescence quenching was extensive for reaction intermediates containing an amine anion and for acylenzymes containing an imine stabilized by mesomeric effect, but not for acylenzymes containing a protonated β-lactam nitrogen. Together, these results indicate that the extent of fluorescence quenching is determined by the status of the β-lactam nitrogen. Thus, fluorescence kinetics can provide information not only on the efficacy of enzyme inactivation but also on the structure of the covalent adducts responsible for enzyme inactivation.
Peptidoglycan is an essential constituent of bacterial cell walls since it prevents cell swelling and lysis by mechanically sustaining the osmotic pressure of the cytoplasm 1. This protective function depends upon synthesis and maintenance during the entire cell cycle of the net-like peptidoglycan macromolecule, which completely surrounds the bacterial cell. Peptidoglycan is made of glycan strands cross-linked by short peptide stems. In most bacteria, the cross-linking step is performed by D,D-transpeptidases, which are the essential targets of β-lactam antibiotics and are often referred to as penicillin-binding proteins (PBPs) 2. In mycobacteria 3–4 and in Clostridium difficile 5–6, the cross-links found are mainly (70% to 80%) formed by a second class of enzymes, the L,D-transpeptidases (LDTs). Since LDTs are not inhibited by β-lactams belonging to the penam class, such as ampicillin, 7–8, these enzymes are responsible for high-level resistance to these drugs in mutants of Enterococcus faecium 9–10 and Escherichia coli 11 selected in laboratory conditions.
PBPs and LDTs are structurally unrelated 12–15 and proceed through different catalytic mechanisms for activation of Ser and Cys nucleophiles 16, which are part of Lys-Ser 17 and Cys-His-Asp 18 catalytic diad and triad, respectively. PBPs and LDTs also differ by the structure of the stem peptide used as an acyl donor, a pentapeptide for PBPs 17 and a tetrapeptide for LDTs 16, except for Enterococcus faecalis Ldtfs 19. PBPs cleave the D-Ala4-D-Ala5 peptide bond of the acyl donor, hence the D,D designation for these transpeptidases, and link the carbonyl of D-Ala4 to the side-chain amino group of the 3rd residue thereby generating 4→3 cross-links 17 (Supplementary Fig. S1). In contrast, LDTs cleave the L-Lys3-D-Ala4 bond of the donor (L,D designation) and form 3→3 cross-links 16.
The first substrate of the transpeptidation reaction, the acyl donor, forms an acylenzyme intermediate that subsequently reacts with the second substrate, the acceptor, leading to the final cross-linked product. For PBPs, nucleophilic attack of the carbonyl of D-Ala4 by the active-site Ser results in an acylenzyme containing an ester bond and release of D-Ala5 17. For the acylenzyme formed by LDTs, a thioester bond connects the carbonyl of L-Lys3 to the γ sulfur of the active-site Cys 16. β-lactams are structure analogues of the D-Ala4-D-Ala5 extremity of stem pentapeptides and act as suicide substrates of the PBPs 20. Nucleophilic attack of the carbonyl of the β-lactam ring by the active-site Ser of PBPs leads to formation of an acylenzyme, which is only hydrolyzed very slowly, leading in practice to “irreversible” inactivation in the time scale of a bacterial generation (Supplementary Fig. S1) 2. LDTs are also acylated by β-lactams although efficacious enzyme inactivation and antibacterial activity occur only for a single class of β-lactams, the carbapenems, such as imipenem 7–8. β-lactams of the cephem (cephalosporin) class, such as ceftriaxone, are only active at high concentrations since acylation is slower and the resulting thioester bond is prone to hydrolysis 7. The kinetic parameters are even less favorable for penams (ampicillin and penicillin) leading to equilibrium between the acylated and unacylated (functional) forms, which accounts for the lack of antibacterial activity 7.
Fluorescence kinetics revealed that the acylation reaction of LDTs by β-lactams of the carbapenem class comprises two limiting steps (Fig. 1A and B). In the first step, nucleophilic attack of the β-lactam carbonyl by the negatively charged sulfur atom of the catalytic Cys was thought to lead to formation of a covalent sulfur-carbon bond and to the development of a negative charge on the drug 21. Initially, this negative charge was proposed to be located on the β-lactam oxygen 21. This hypothesis was based on analogies with PBPs, which contain an oxyanion hole for stabilization of the negative charge developing on the β-lactam oxygen 17. However, hybrid potential simulation of the acylation reaction indicated that the path to an oxyanion is energetically disfavored in the case of LDTs and that an oxyanion could not correspond to a stabilized reaction intermediate 22. Rather, hybrid potential simulation indicated that the most energetically favored reaction path involves a concerted mechanism leading to the concomitant formation of the thioester bond, rupture of C-N β-lactam bond, and formation of an amine anion 22. In the second step, the final acylenzyme is formed by protonation of the amine anion. The origin of this proton remains elusive 18, 22. Formation of a non-covalent complex is not considered in this reaction scheme as the affinity of the Ldtfm for carbapenems is very low, in the order of 60 to 80 mM 7.
Fluorescence kinetics proved useful to assess the efficacy of inactivation of LDTs from Mycobacterium tuberculosis 23–25 and contributed to the choice of the drugs that were tested in a phase II clinical trial 26. The assays were also useful in evaluating series of synthetic carbapenems 27 and non-β-lactam LDT inhibitors 28. However, the basis for the variation in fluorescence intensity detected during the two steps of the acylation reaction remained elusive. Initially, we developed fluorescence kinetics since determination of the crystal structure of Ldtfm revealed a Trp residue at the entrance of the catalytic cavity 15, which appeared ideally located for fluorescence quenching upon drug binding due to changes in its environment 21. However, this naïve explanation was subsequently found to be insufficient to account for the various behaviors observed for acylation of LDTs by representatives of the β-lactam classes 7, 22-25. This prompted us to investigate here the basis for variations in fluorescence intensity occurring upon acylation of LDTs using the well-characterized L,D-transpeptidase Ldtfm from E. faecium as a model. We show that variations in fluorescence intensity are not due to modification of the environment of Trp residues but to the formation of a β-lactam-derived fluorescence quencher during the acylation reaction.
RESULTS
Trp to Phe and His to Ala substitutions do not abolish the biphasic behavior of fluorescence kinetics observed for Ldtfm inactivation by imipenem
We have previously proposed that the acylation of Ldtfm by carbapenems is a two-step reaction involving reversible formation of an amine anion (EIAn, step 1) followed by irreversible formation of an acylenzyme (EI*, step 2) (Fig. 1A) 21–22. Fluorescence kinetics also display two phases enabling the determination of kinetic parameters for the two steps of the reaction (Fig. 1B) 21. In the first phase, fluorescence intensity decreases because EIAn is rapidly formed to the detriment of the free enzyme (E) and the fluorescence intensity of EIAn is lower than that of E. In the second phase, fluorescence intensity increases since the acylenzyme (EI*) is formed to the detriment of EIAn (acylation step) and the fluorescence intensity of EI* is greater than that of EIAn. Our first aim was to determine whether variations in the fluorescence intensity could be correlated to the modification of the environment of a specific Trp residue of Ldtfm. To address this question, Trp residues of the catalytic domain of Ldtfm were replaced by Phe residues in various combinations. For the sake of simplicity, the six Trp residues at positions 355, 385, 410, 415, 425, and 434 of Ldtfm were designated as residues a to f, respectively. Fluorescence kinetics were determined for Ldtfm derivatives lacking 2 to 5 Trp residues (Fig. 1C). Fluorescence kinetics remained bi-phasic in all cases, except for the Ldtfm derivative only retaining the “a” residue (i.e. Trp355). Thus, the variations in the fluorescence intensity associated with the two steps of the acylation reaction do not depend upon the modification of the environment of a specific Trp residue of Ldtfm. This prompted us to investigate fluorescence quenching by His residues of Ldtfm. Fluorescence kinetics obtained with the His353Ala and His440Ala derivatives of Ldtfm indicated that neither residue is essential for biphasic fluorescence kinetics (Fig. 1C). His at position 421 could not be analyzed since this residue is part of the catalytic triad and is essential for Ldtfm activity.
Impact of Trp to Phe and His to Ala substitutions on the efficacy of the inactivation reaction and on the relative fluorescence intensity of the three enzyme forms
Fitting simulations to experimental data for four concentrations of imipenem was used to determine the inactivation kinetic parameters (Supplementary Fig. S2). None of the Trp residues was essential for the acylation of Ldtfm by imipenem although important variations were observed in the k1 (formation of the amine anion) and k2 (acylation step) parameters (Fig. 2A and 2B, respectively). The Trp to Phe substitutions produced large variations in the intrinsic fluorescence of the three forms of the enzyme (apo Ldtfm, amine anion, and acylenzyme; Fig. 2C). The formation of the acylenzyme (EI*) was confirmed by mass spectrometry for all Ldtfm derivatives (mass increment of 299.3 Da; data not shown).
Conservation of Trp residues in Ldtfm homologues
The Trp residues were poorly conserved in L,D-transpeptidases from Gram-positive and Gram-negative bacteria (Supplementary Fig. S3), as expected from the substantial high residual acylation activity of Ldtfm derivatives containing Trp to Phe substitutions (above, Fig. 2A and B). However, residue e (Trp425) was relatively conserved in L,D-transpeptidases from M. tuberculosis, being present in the Mt1, Mt2, Mt3, and Mt4 enzymes as well as in the corresponding orthologues of M. abscessus. These enzymes are known to display bi-phasic fluorescence kinetics upon inactivation by carbapenems 23–24, 29 and (unpublished results). Since Trp425 was the only conserved Trp residue in the LDTs from mycobacteria we investigated the impact of a Trp to Phe substitution at this position in Ldtfm. As shown in Supplementary Fig. S4A and B, derivatives of Ldtfm lacking all Trp residues except Trp425 or lacking only this residue displayed biphasic kinetics. These results indicated that the biphasic nature of fluorescence kinetics does not depend upon any conserved Trp residue in the L,D-transpeptidase protein family.
Fluorescence kinetics with a single ectopic Trp residue at position 383
All six Trp residues of Ldtfm were replaced by Phe residues. As expected, the variant devoid of Trp residues was very weakly fluorescent (data not shown). No modification of this weak fluorescence was observed upon acylation by imipenem. This observation ruled out the formal possibility that the catalytic Cys acylated by imipenem could be a fluorophore. Replacement of Tyr at position 383 by a Trp residue in this Ldtfm derivative restored biphasic fluorescence kinetics (Supplementary Fig. S4C). This result confirmed that biphasic fluorescence kinetics are not dependent upon the presence of any of the six Trp residues of Ldtfm in their original positions.
Investigation of the opened β-lactam ring linked to the catalytic Cys residue as a quencher
Since the variations in the fluorescence observed during the two steps of the acylation reaction could not be accounted for by modification of the environment of any specific Trp residue, our next objective was to investigate whether quenching could result from energy transfer from Trp residues to the Cys-β-lactam covalent adduct generated upon enzyme inactivation. To investigate this possibility, we first examined acylation of Ldtfm by the chromogenic cephalosporin nitrocefin. Fluorescence kinetics with this compound displayed a monophasic behavior with a large decrease (72%) in the fluorescence intensity (Fig. 3A). In the Ldtfm-nitrocefin acylenzyme, the negative charge developing on the β-lactam nitrogen is stabilized by a mesomeric effect that prevents protonation of the amine 30. Thus, the amine anion formed as an end product with nitrocefin (Fig. 3A) or as an intermediate with imipenem (Fig. 1) was associated with extensive fluorescence quenching. As found for nitrocefin, acylation of Ldtfm by ceftriaxone led to a monophasic decrease in the fluorescence intensity (Fig. 3B). Acylation of Ldtfm by ceftriaxone involves a concerted mechanism leading to the elimination of the side chain and formation of a conjugated imine 7. The extent of fluorescence quenching observed for the acylenzyme formed with ceftriaxone (40%; imine) or by the intermediate formed with imipenem (45%; amine anion) were more important than that observed for the final Ldtfm-imipenem adduct (20%) suggesting that quenching was decreased by protonation of the β-lactam nitrogen.
Fluorescence quenching associated with inactivation of β-lactamases by β-lactamase inhibitors
In order to enrich the structural diversity of molecules investigated in the fluorescence quenching assay, we aimed to test β-lactamase inhibitors. Most β-lactams could not be tested with β-lactamases since the fluorescence assay requires relatively high enzyme concentrations (typically 10 µM) leading to rapid hydrolysis of the drugs even if the turnovers are low. For this reason, we focused on inactivation of BlaC from Mycobacterium tuberculosis by clavulanate since the corresponding acylenzyme is not prone to hydrolysis at the timescale of our experiments 31. Acylation of BlaC by clavulanate led to a large decrease in the fluorescence intensity (37%) (Fig. 3C). Previous studies have shown that acylation of BlaC by clavulanate leads to formation of imine and trans-enamine tautomers (depicted in Fig. 3C) following drug decarboxylation 32. Since Ldtfm and BlaC contain Cys and Ser as the active-site nucleophile, respectively, these results indicate that fluorescence quenching is observed with acylenzymes containing both thioester and ester bonds.
Avibactam offered the possibility to test a non-β-lactam inhibitor that inhibits BlaC by formation of a carbamoylenzyme 33. Formation of the adduct (Supplementary Fig. S5) was not associated with any significant modification of the fluorescence intensity of the enzyme (Fig. 3D). Protonation of the amino sulfate nitrogen of avibactam in the avibactam-BlaC adduct was thus associated with limited quenching as found for the protonation of the β-lactam nitrogen in the final Ldtfm-imipenem and BlaC-clavulanate adducts (Fig. 1 and 3C).
Acylation of Ldtfm and BlaC by faropenem
Faropenem, a β-lactam belonging to the penem class, is poorly hydrolyzed by BlaC and rapidly inactivates L,D-transpeptidases 24. Thus, faropenem provided the opportunity to study the acylation of the two enzyme types, i.e. a β-lactamase (BlaC) and an L,D-transpeptidase (Ldtfm), by the same β-lactam. Acylation of BlaC by faropenem led to extensive fluorescence quenching (44%) (Fig. 3E). The resulting acylenzyme contained an unprotonated β-lactam nitrogen. In contrast, acylation of Ldtfm by faropenem, which resulted in the rupture of the C5-C6 bond and the loss of a large portion of the drug including the β-lactam nitrogen 24, was associated with a moderate decrease in the fluorescence intensity (17%) (Fig. 3F).
Acylation of BlaC by the meropenem carbapenem
Although BlaC displays moderate carbapenemase activity, previous studies have shown that a BlaC-meropenem adduct is the predominant enzyme form upon incubation of the enzyme with low drug concentrations 34–35. As expected, the control experiment depicted in Supplementary Fig. S6 shows that the concentration of meropenem remains saturating for at least 80 min during hydrolysis of meropenem (40 – 200 µM) by BlaC (10 µM). Fluorescence kinetics at a shorter time scale (6 min) showed a transitory decrease in fluorescence intensity (40%) (Fig. 3G). This observation suggests that formation of an amine anion initially leads to an important fluorescence quenching. Then, fluorescence returned to the initial level suggesting that protonation of the amine anion fully suppressed quenching. Thus, formation of an amine anion following nucleophilic attack of the carbonyl carbon of carbapenems by Ldtfm (Fig. 1) and BlaC (Fig. 3G) similarly led to extensive fluorescence quenching despite the presence of different catalytic residues (Cys versus Ser) and different modes of activation of these nucleophiles involving His and Lys residues.
DISCUSSION
The main aim of the present study was to determine the basis for the variations in the fluorescence intensity observed upon acylation of LDTs by β-lactam antibiotics. Previous analyses of the D,D-transpeptidase R61 (PBP family) have shown that acylation of the enzyme by penicillin G leads to a fluorescence quenching, which specifically depends upon one of two Trp residues of the protein 36–37. This residue, W233, is essential for the D,D-carboxypeptidase activity of the protein, as determined by the release of the terminal D-Ala residue from a model peptide ending in D-Ala-D-Ala 37. The crystal structure of R61 in complex with a peptide mimicking the acyl donor revealed hydrophobic interactions between the side-chain of W233 and methylene groups of this substrate 38. Likewise, the structure of a penicillin-R61 complex revealed that W233 is in hydrophobic interaction with the phenylacetyl side chain of the drug 39 indicating that fluorescence quenching may depend upon this direct interaction. In contrast, we show here that the variations in fluorescence during acylation of LDTs by β-lactam antibiotics require at least one Trp residue but that the position of this residue is not determinant. This was established by deleting Trp residues from Ldtfm in various combinations (Fig. 1 and 2), by introducing a single Trp residue in an ectopic position of Ldtfm (Fig. S4), and by showing that large variations in fluorescence intensity occur in distantly related LDTs and unrelated β-lactamases that do not share any conserved Trp residue (Fig. S3). Since fluorescence quenching was not mainly due to changes in the environment of Trp residues, we investigated the possibility that the nucleophilic attack of the β-lactam ring could itself generate a fluorescence quencher. If this is the case, the fluorescence profile should depend upon the structural features of the drug bound to the active-site residue. Accordingly, extensive quenching was not mainly determined by the size of the side chains of β-lactams or by the nature of the enzyme nucleophile (Cys versus Ser in Ldtfm and BlaC, respectively) (Fig. 3). Acylation of Ser or Cys per se was not sufficient for fluorescence quenching since a modest or no decrease in fluorescence intensity was observed for the final acylation products of Ldtfm by faropenem, of BlaC by avibactam, and of both enzymes by carbapenems. By elimination, these results pointed to the status of the β-lactam nitrogen as the only source of variation that could account for the main variations in fluorescence intensity. A large fluorescence quenching occurred if the β-lactam nitrogen was negatively charged, as transiently found during acylation of Ldtfm and BlaC by carbapenems. A large fluorescence quenching also occurred if the nitrogen atom was engaged in a double bond as observed for the acylenzymes formed by Ldtfm with ceftriaxone or nitrocefin and by BlaC with faropenem or clavulanate. In agreement with the critical role of the double bond, fluorescence quenching was not observed for acylation of BlaC by avibactam, which results in the formation of a carbamoyl-enzyme with a protonated nitrogen, or for acylation of Ldtfm by faropenem, which results in the formation of an acylenzyme lacking the nitrogen atom following rupture of the C5-C6 bond. These results indicate that fluorescence kinetics can provide information not only on the efficacy of enzyme inactivation but also on the structure of the covalent adducts, including for example the presence of Δ1- or Δ2-pyrroline ring in enzyme-carbapenem adducts. Our results also indicate that nitrocefin provides a very sensitive assay to titrate the active site of enzymes that do not hydrolyze this cephalosporin. This property could be exploited to identify β-lactams and non-β-lactam inhibitors that reversibly or irreversibly bind to LDTs.
EXPERIMENTAL SECTION
Enzyme production and purification
The catalytic domain of Ldtfm (residues 341 to 466) and soluble BlaC (residues 39 to 306) were produced in E. coli and purified by metal affinity and size exclusion chromatography as previously described 30. Synthetic genes were purchased for production of Ldtfm derivatives with Trp to Phe and His to Ala substitutions (GeneCust). All experiments were performed with a fixed concentration of Ldtfm and BlaC (10 µM).
Spectrofluorimetry
Fluorescence kinetics were studied in 100 mM sodium phosphate (pH 6.0) at 20°C by using a stopped-flow apparatus (RX-2000, Applied Biophysics) coupled to a spectrofluorimeter (Cary Eclipse; Varian). Excitation was performed at 224 nm with a 5 nm slit and a 2 mm optical path. Fluorescence emission was determined at 335 nm with a 5 nm slit and a 10 mm optical path. Fluorescence data were collected exactly in the same conditions for comparison of the relative fluorescence of wild-type Ldtfm and derivatives with Trp to Phe and His to Ala substitutions. These conditions included the photomultiplier voltage, which was set to 600 V. Kinetic constants were determined using the DynaFit software (BioKin Ltd) 40 for each Ldtfm derivative by simultaneously fitting the fluorescence data for the various inhibitor concentrations to differential equations derived from the reaction scheme depicted in Fig. 1A.
Mass spectrometry
Formation of the proposed Ldtfm-β-lactam and BlaC-β-lactam adducts in the conditions described in the text was systematically checked by mass spectrometry (data not shown). Enzymes and β-lactams were incubated for appropriate time periods at 20°C. Five μl of acetonitrile and 1 μl of 1% formic acid were extemporaneously added. Injection in the mass spectrometer (Qstar Pulsar I; Applied Biosystem) was performed at a flow rate of 0.05 ml/min (acetonitrile, 50%, water, 49.5%, and formic acid, 0.5%; per volume). Spectra were acquired in the positive mode, as previously described 8. Detailed analyses of the mass spectra of the β-lactam adducts considered in the current study have been previously published 7–8, 24, 29, 41.
ANCILLARY INFORMATION
Supporting Information
Supplementary figures S1 to S6.
Author Contribution
ST, ZE contributed equally to this work. MA and JEH contributed equally to this work.
Acknowledgement
We thank L. Dubost and A. Marie for technical assistance in the collection of mass spectra. This work was supported by the French National Research Agency (grant MycWall, ANR-17-CE18-0010) and the Fondation pour la Recherche Médicale (grant ECO20160736080 to ZE).