|Home | About | Journals | Submit | Contact Us | Français|
Antimicrobial resistance is a global issue currently resulting in the deaths of hundreds of thousands of people a year worldwide. Data present in the literature illustrate the emergence of many bacterial species that display resistance to known antibiotics; Acinetobacter spp. are a good example of this. We report here that Acinetobacter radioresistens has a Baeyer-Villiger monooxygenase (Ar-BVMO) with 100% amino acid sequence identity to the ethionamide monooxygenase of multidrug-resistant (MDR) Acinetobacter baumannii. Both enzymes are only distantly phylogenetically related to other canonical bacterial BVMO proteins. Ar-BVMO not only is capable of oxidizing two anticancer drugs metabolized by human FMO3, danusertib and tozasertib, but also can oxidize other synthetic drugs, such as imipenem. The latter is a member of the carbapenems, a clinically important antibiotic family used in the treatment of MDR bacterial infections. Susceptibility tests performed by the Kirby-Bauer disk diffusion method demonstrate that imipenem-sensitive Escherichia coli BL21 cells overexpressing Ar-BVMO become resistant to this antibiotic. An agar disk diffusion assay proved that when imipenem reacts with Ar-BVMO, it loses its antibiotic property. Moreover, an NADPH consumption assay with the purified Ar-BVMO demonstrates that this antibiotic is indeed a substrate, and its product is identified by liquid chromatography-mass spectrometry to be a Baeyer-Villiger (BV) oxidation product of the carbonyl moiety of the β-lactam ring. This is the first report of an antibiotic-inactivating BVMO enzyme that, while mediating its usual BV oxidation, also operates by an unprecedented mechanism of carbapenem resistance.
Bacteria classified within the genus Acinetobacter play an increasingly important role in the pathogenesis of human diseases. Among the members of this genus, Acinetobacter baumannii is the species that is the most frequently isolated from humans (and is found in endotracheal aspirate, blood, and perianal and wound secretions) where it is known to manifest multidrug resistance (MDR) (1). Acinetobacter species are widely distributed in nature and can be found in soil, water, sewage, and a variety of foodstuffs (1, 2). They are present in the hospital environment, especially intensive care units (3, 4), as they are inhabitants of healthy human skin and are part of the normal flora; can be isolated from dry surfaces and equipment; and can easily survive for many days or weeks, even under dry conditions (1, 2). The increasing rates of recovery of MDR A. baumannii strains in nosocomial settings are a frightening reality (5, 6), and the combination of their environmental resilience and their wide range of resistance determinants renders them successful nosocomial pathogens (6). These MDR strains often spread to cause outbreaks throughout entire cities, countries, and continents (7,–10). The importation of MDR strains from areas with high rates of antimicrobial resistance to areas with an historically low rate has been demonstrated. United Kingdom and U.S. military and nonmilitary personnel returning from operations in Iraq and Afghanistan harbored infections caused by multiresistant A. baumannii strains (11,–13). The resistance of A. baumannii to the carbapenem imipenem, which is the drug of choice for the treatment of serious infections caused by this species, leads to difficult-to-treat nosocomial infections (7, 14, 15). Carbapenem resistance mostly results from the expression of acquired carbapenem-hydrolyzing oxacillinases in A. baumannii. The blaOXA-23 gene, which confers drug resistance in A. baumannii, codes for the oxacillinase OXA-23, one of the five Ambler class D β-lactamases (16). Outbreaks of OXA-23-producing A. baumannii strains have been reported worldwide and may represent an emerging threat (17). Poirel and coworkers found that Acinetobacter radioresistens, a commensal bacterial species that is identified on the skin of hospitalized and healthy patients, is the source of the blaOXA-23 gene (18).
We recently identified a novel A. radioresistens Baeyer-Villiger monooxygenase (Ar-BVMO) in A. radioresistens strain S13 (19). Baeyer-Villiger monooxygenases (BVMOs) are flavin-containing enzymes that mediate specific Baeyer-Villiger (BV) oxidations on the carbonyl moiety of substrates. Ketones are converted into the corresponding esters and lactones by the Baeyer-Villiger reaction. BVMO enzymes are abundant in bacterial, fungal, and plant genomes, but they are absent in animal and human genomes. All characterized BVMOs contain a flavin cofactor that is crucial for catalysis, while NADH or NADPH is required as an electron donor. Most reported BVMOs are soluble proteins, in contrast to many other monooxygenase systems, which are often found to be membrane bound. BVMOs are divided into three different types according to their general characteristics (20). Type I enzymes contain a tightly bound FAD cofactor, are NADPH dependent, and possess two Rossmann folds for dinucleotide binding and a conserved Baeyer-Villiger motif, a fingerprint sequence [FXGXXXHXXXW(P/D)] involved in catalysis (21). Type II enzymes do not present the Baeyer-Villiger sequence motif and utilize flavin mononucleotide as a coenzyme and NADH as a cosubstrate, whereas type 0 enzymes use flavin adenine dinucleotide (FAD) and NAD(P)H and lack the BVMO fingerprint motif (22). A heme-containing BVMO belonging to the cytochrome P450 superfamily has also been reported (23). Earlier studies had already suggested the Baeyer-Villiger activity of other eukaryotic cytochrome P450 enzymes (24). This finding indicates that during evolution several different enzymes evolved into Baeyer-Villiger monooxygenases.
Most biochemical and biocatalytic studies have been performed with type I BVMOs (22). This is partly due to the fact that they represent relatively uncomplicated monooxygenase systems. These monooxygenases are typically soluble and composed of only one component. Expression systems have been developed for a number of type I BVMOs, while no recombinant expression has been reported for type II BVMOs. All these enzymes are known to perform various catalytic activities on different compounds even partially sharing their substrate profile. BVMOs are also known to perform oxygenation on heteroatom-containing compounds. The first evidence of soft nucleophile-containing substrate oxidation was reported with a cyclohexanone monooxygenase (CHMO) from an Acinetobacter sp. that showed 4-tolyl ethyl sulfide conversion to the corresponding (S)-sulfoxide with a modest enantioselectivity (25). BVMOs may also have medical relevance, as in the case of ethionamide monooxygenase (EtaA), a type I BVMO capable of converting a range of ketones into the corresponding esters (26). Except for catalyzing Baeyer-Villiger oxidations, the enzyme is also able to oxidize the sulfide moieties of several antitubercular thioamide drugs (26, 27). The oxidized drugs appear to be highly toxic to mycobacteria, indicating that ethionamide monooxygenase acts as a prodrug activator.
Phylogenetic analysis placed the sequence of Ar-BVMO together with that of Mycobacterium tuberculosis EtaA (19), with only a distant relation to the sequences of other known class I BVMO proteins being found. In vitro experiments carried out with the purified enzyme confirmed that this novel BVMO is indeed capable of typical Baeyer-Villiger reactions as well as oxidation of the prodrug ethionamide (19). Qian and Ortiz de Montellano (27) demonstrated that ethionamide is also oxidized by human flavin-containing monooxygenase 3 (FMO3). Human flavin-containing monooxygenase (FMO) enzymes are a superfamily of flavoprotein monooxygenases involved in the detoxification of a wide range of xenobiotics containing a soft nucleophile, usually nitrogen or sulfur (28). Since Ar-BVMO has been shown to oxidize ethionamide, we wondered whether this enzyme could play a role in the detoxification of xenobiotics and recognize synthetic compounds, such as drugs, typically recognized by human FMO3. We took into consideration two synthetic drugs, danusertib and tozasertib, shown to be N-oxygenated by human FMO3 (29, 30) and certainly not present in the natural environment of this bacterium. Species belonging to the Acinetobacter genus have been reported to be among the major superbugs resistant to antibiotic treatment (31). The presence in their genomes of genes coding for monooxygenases similar to those involved in drug metabolism in humans becomes highly relevant when tackling the threat that they pose.
We report here that the overexpression of Ar-BVMO in Escherichia coli renders it resistant to imipenem and show that while this enzyme mediates its usual Baeyer-Villiger oxidation on the carbonyl moiety of the β-lactam ring, it also operates by an unprecedented mechanism of carbapenem resistance. In the future, if appreciable carbapenemase activity is attributed to this monooxygenase under physiological conditions, it could represent a new target for drug design in the battle against carbapenem resistance in Acinetobacter.
Cells of Escherichia coli strain BL21(DE3) (Invitrogen) transformed with the expression vector plasmid pT7 harboring the gene coding for Acinetobacter radioresistens S13 Baeyer-Villiger monooxygenase (Ar-BVMO) (19) and untransformed BL21 cells were used in this study.
FAD, acetonitrile, NADPH, imipenem, methanol, and salts were purchased from Sigma-Aldrich (Milan, Italy). Tozasertib and danusertib were purchased from Aurogene (Rome, Italy).
Amino acid sequences from completely sequenced genomes of proteobacteria were retrieved from the GenBank database (http://www.ncbi.nlm.nih.gov). Probing of the database by BLAST analysis (32) was performed using the default parameters of the BLASTP option of the BLAST program. The ClustalW program (33) was used to perform pairwise and multiple-amino-acid-sequence alignments.
Preliminary multiple-sequence alignments of amino acid sequences were generated with the program ClustalW using the default gap penalties. The selection of characters eligible for the construction of a phylogenetic tree was optimized by comparing entire sections of all the available BVMO alignments with comprehensive inventories of significant binary alignments obtained by probing of the protein database by BLAST analysis with representative bacterial sequences. The structure and organization of the BVMOs were deduced from the data present in GenBank from the microorganisms available in data banks. Probing of the protein databases was performed by BLAST analysis with the BLASTP program, and sequences with an E value of 0.0, 100% query coverage, and a threshold of 99% sequence identity were chosen. For construction of the phylogenetic tree, a representative of each class of BVMO was selected to demonstrate its relationship to the Ar-BVMO.
The phylogenetic tree was constructed using the distance matrix (DM) method and the MEGA (version 4) program (34), after multiple-sequence alignment and truncation of all sequences to the same length. Distances according to the Kimura two-parameter model (35) and clustering with the neighbor-joining method (36) were determined using bootstrap values based on 1,000 replications.
The expression and purification of Ar-BVMO were performed as previously reported (19). After Ni-nitrilotriacetic acid affinity chromatography, the heterologous expression of Ar-BVMO yielded 8 mg of purified protein per liter of culture. The concentration of the pure Ar-BVMO was determined by spectroscopy with the peak absorbance at 450 nm and the extinction coefficient of free FAD (ε450 = 11,300 M−1 cm−1).
N-oxygenation of the anticancer drugs danusertib and tozasertib was evaluated by incubating a mixture of 0.3 μM purified enzyme in 50 mM potassium phosphate buffer (pH 7.4), 0.5 mM NADPH, and increasing amounts of substrates (6, 12.5, 25, 50, 75, 100, 150, 200, 300, and 400 μM) in a final volume of 200 μl. Incubations were carried out at 37°C for 15 min and were terminated by the addition of 100 μl of ice-cold methanol. The aqueous supernatant was centrifuged at 2,000 × g for 10 min and was subjected to high-pressure liquid chromatography (HPLC). Kinetic parameters related to the turnover were determined through Michaelis-Menten kinetics using SigmaPlot software.
HPLC was performed on an Agilent quaternary pump HPLC system (Agilent Technologies) equipped with an analytical C18 column (4.6 by 150 mm; particle size, 5 μm). The method to separate tozasertib from tozasertib N-oxide and danusertib from danusertib N-oxide used a mobile phase of 78% methanol and 22% distilled water at a flow rate of 1.0 ml/min. The formation of tozasertib N-oxide was monitored at a wavelength of 250 nm, while the formation of the product danusertib N-oxide was monitored at 295 nm (29, 30). The kinetic analysis of N- or S-oxygenation was carried out using a nonlinear regression analysis program (SigmaPlot).
The imipenem sensitivity of (i) E. coli BL21 cells transformed with pT7-Ar-BVMO and induced with IPTG (isopropyl-β-d-thiogalactopyranoside), (ii) E. coli BL21 cells not induced and, (iii) negative-control E. coli BL21 cells transformed with pT7 (carrying another gene ) and induced with IPTG was determined by the disk diffusion method (38) according to the latest updated guidelines (2014) of the European Committee on Antimicrobial Susceptibility Testing (EUCAST; www.eucast.org) using standard antibiotic disks (Oxoid) and Mueller-Hinton agar (MHA) plates (Sigma). There were five replicates for each condition tested. Inhibition zones were measured as the mean diameter and were categorized as susceptible (S), intermediate (I), and resistant (R) according to EUCAST standards.
The effect of Ar-BVMO activity on the antimicrobial properties of imipenem was subsequently assessed by a microbiological assay. The inactivation reaction mixture contained 3 mM imipenem, 2 mM NADPH, and 100 μM purified Ar-BVMO in 50 mM phosphate buffer, pH 7.4, in a total volume of 0.5 ml. The reaction was allowed to proceed for 30 min at 37°C. A control reaction was performed under the same conditions described above but without the addition of NADPH. Fifty-microliter aliquots of the two reaction mixtures were used for agar diffusion plate tests (39) in order to evaluate the effect of metabolized imipenem on the growth of sensitive E. coli BL21 cells. MHA plates were prepared using sterile MHA. Upon solidification of the agar medium, sensitive nontransformed BL21 cells and transformed but noninduced BL21 cells were spread over the plate surface at a concentration equivalent to a 0.5 McFarland standard. Then, a well was made in the center of each plate using a sterile Pasteur pipette. After adding 50 μl of the Ar-BVMO–imipenem reaction mixture with and without NADPH, an uninoculated agar plug larger than the well was placed over it to prevent volatilization of the enzymatic reaction mixture into the headspace of the petri dish (39). The diameter of the growth inhibition zone was measured after 24 h from the time of application of the test substance and incubation at 37°C. There were five replicates for each condition tested.
Enzyme activity was measured spectrophotometrically by monitoring the substrate-dependent decrease in the NADPH concentration at 340 nm (ε 340 nm = 6.22 mM−1 cm−1). The reaction mixtures contained 50 mM Tris-HCl pH 8.0, 200 μM NADPH, and imipenem over a range of concentrations of from 5 to 150 μM, and air-saturated buffers were used. Steady-state kinetic parameters were determined by fitting the initial rates, calculated from the first 200 s of the reaction, to the standard Michaelis-Menten equation using the SigmaPlot program. Experiments were conducted in triplicate.
The chromatographic separations were run on a Phenomenex Luna C18 column (150 by 2.0 mm; particle size, 3 μm; Phenomenex, Torrance, CA, USA). The injection volume was 10 μl, and the flow rate was 0.2 ml min−1. An isocratic mobile phase composition was adopted: 5:95 acetonitrile-aqueous formic acid (0.05%). An LTQ-Orbitrap hybrid mass spectrometer (MS; Thermo Scientific, Bremen, Germany) equipped with an atmospheric pressure interface and an electrospray ionization ion source was used. The liquid chromatography (LC) column effluent was delivered into the ion source using nitrogen as the sheath and auxiliary gas. The needle voltage was set at a value of 4.5 kV. The heated capillary temperature was maintained at 270°C. The acquisition method used was previously optimized in the tuning sections for the analyte ion (capillary, magnetic lenses, and collimating octapole voltages) in order to achieve the maximum sensitivity. Spectra were acquired in the positive high-resolution MS (HRMS) and tandem MS (MS/MS) mode with precursor ions at m/z 300 and 316, a normalized collision energy of 25%, and a mass range of 320 to 80 m/z.
The phylogenetic tree of Ar-BVMO was inferred from the sequences selected by the DM method, as described in Materials and Methods. As shown in Fig. 1, the clade comprising Mycobacterium tuberculosis EtaA, Ar-BVMO, and A. baumannii EtaA belongs to a separate group distant from the other BVMO proteins. Furthermore, Ar-BVMO shares 100% amino acid sequence identity with the ethionamide monooxygenase of MDR A. baumannii strains (GenBank accession numbers EXB35767.1, EXB73644.1, EXE14728.1, and KCX39290.1).
We had previously tested in our laboratories different therapeutic drugs as possible substrates of this human hepatic enzyme, including danusertib and tozasertib (28, 29). Tozasertib (VX-680) and danusertib (PHA-739358) are inhibitors of aurora kinases, which have been identified to be a potential target in anticancer therapy. As a starting point, the possibility that these two inhibitors act as the substrates for the Ar-BVMO enzyme was tested by investigating whether any enzymatic products could be detected. After each set of reactions at 37°C in the presence of NADPH, the product(s) was separated by HPLC, and the data obtained are shown in Fig. 2. As can be seen in Fig. 2, the expected N-oxides were detected with shorter retention times of 3.7 min for danusertib N-oxide (Fig. 2A) and 4.1 min for tozasertib N-oxide (Fig. 2C), in agreement with the data previously published by our group (29, 30). The catalytic activity of Ar-BVMO toward two kinase inhibitors, danusertib and tozasertib, was investigated. Enzymatic reactions were carried out with the purified recombinant Ar-BVMO in the presence of NADPH and increasing amounts of each drug, as reported in Materials and Methods. Kinetic parameters for the N-oxygenation of danusertib (Fig. 2B) and tozasertib (Fig. 2D) were determined by nonlinear regression analysis, with the calculated Km values being 352 ± 58 μM and 54.3 ± 8.3 μM, respectively. The corresponding values for the maximum rate of metabolism (Vmax) were also calculated for danusertib and tozasertib and were 8.1 and 7.4 nmol/min/mg of protein, respectively. These results confirm that the bacterial enzyme is indeed able to metabolize these two synthetic drugs.
Once it was proven that Ar-BVMO is capable of metabolizing the synthetic drugs tested, we turned our attention to antibiotics as another class of synthetic drugs directly relevant to the MDR effects seen in Acinetobacter spp. Initially, in order to understand whether the expression of Ar-BVMO in imipenem-sensitive E. coli BL21 cells could confer resistance to imipenem (i.e., whether there was an enzymatic reaction of Ar-BVMO with imipenem resulting in a product no longer active as an antibiotic), a disk diffusion assay was performed following EUCAST guidelines (www.eucast.org). According to the EUCAST clinical breakpoint table (version 5.0; valid from 1 January 2015) for Enterobacteriaceae, the zone diameter breakpoints for imipenem are ≥22 mm for sensitivity and <16 mm for resistance, corresponding to MIC values of 2 and 8 mg/liter of imipenem, respectively, for sensitive and resistant members of the family Enterobacteriaceae, including E. coli, which is the most important model organism of this family. As shown in Fig. 3, the average diameters of the inhibition zones for E. coli BL21 cells (Fig. 3A, top), the same E. coli strain transformed with pT7-Ar-BVMO but not induced (Fig. 3B, top), and the same E. coli strain transformed with pT7 (negative control) and induced with IPTG (Fig. 3C, top) were measured to be 25.5 mm, 24.0 mm, and 28.0 mm, respectively. The results clearly show that the E. coli BL21 cells are sensitive to imipenem (zone diameter breakpoint, ≥22 mm) but there is no visible inhibition zone when E. coli BL21 cells are transformed and the expression of Ar-BVMO is induced by IPTG; i.e., the cells became resistant to imipenem (zone diameter breakpoint, <16 mm) (Fig. 3D, top). These data provide preliminary but encouraging evidence that the overexpression of Ar-BVMO in imipenem-sensitive cells may result in antibiotic resistance.
The effect of Ar-BVMO activity on the antimicrobial properties of imipenem was further studied by a microbiological assay (40). The reaction product(s) of purified Ar-BVMO and imipenem was loaded in a well in the MHA plates where imipenem-sensitive BL21 cells and BL21 cells carrying pT7-Ar-BVMO whose gene expression was not induced were grown at 37°C for 24 h. As shown in Fig. 3B and andDD (bottom), imipenem was not active after the reaction with Ar-BVMO since nontransformed BL21 cells and BL21 cells transformed but not induced exhibited inhibition zones of 12.0 mm and 0.5 mm, respectively. When the enzymatic reaction mixture without NADPH (negative control) was deposited in the agar well, the diameters of the inhibition zones were 24.0 mm (Fig. 3A, bottom) and 24.5 mm (Fig. 3C, bottom) for BL21 cells that were not transformed and BL21 cells that were transformed but not induced, respectively. These data also support the idea that the enzymatic reaction of Ar-BVMO converts imipenem into an inactive compound no longer possessing its original antibacterial property.
Subsequent to collection of the in vivo data, a spectroscopic method was used to measure the activity of the purified Ar-BVMO in the presence of imipenem. In order to determine the steady-state kinetic parameters for Ar-BVMO turnover, an NADPH consumption assay which has previously been described in the literature as an efficient system for evaluation of Km (41) was used. This method relies on the spectrophotometric monitoring of the consumption of the cosubstrate NADPH at 340 nm in the presence of increasing concentrations of imipenem. The experimental data from the NADPH consumption assay were fitted to a Michaelis-Menten curve, used for the determination of a Km value of 12.3 ± 0.8 μM (Fig. 4) and a Vmax value of 468.2 ± 9.6 nmol/min/mg of protein. These values are in the same range as those reported for known substrates of other BVMO enzymes (41), indicating that imipenem could be a real substrate of Ar-BVMO.
In order to identify the product(s) of the Ar-BVMO reaction with imipenem, LC-MS experiments were carried out. Imipenem (100 μM) was incubated in phosphate buffer (50 mM, pH 7.4) at room temperature and monitored by LC–high-resolution mass spectrometry at different reaction times. Ar-BVMO and NADPH concentrations were 0.45 and 200 μM, respectively. The incubation mixture was simply 10-fold diluted using the elution mobile phase and injected into the LC-MS instrument. The imipenem peak was eluted at 4.7 min. The main metabolite was detected as a double peak (5.0 and 5.5 min) due to a mixture of isobaric products. The accuracy of Orbitrap high-resolution mass determination allowed us to deduce that the elemental composition of the two compounds corresponded to C12H18N3O5S with 1.05 ppm of error (MH+ ion at 316.0962 m/z). A classical Baeyer-Villiger asymmetric oxygen insertion can explain this finding, as shown in Fig. 5.
An analogous BV catalytic synthetic reaction was described for a similar lactam (bicyclo[3.2.0]hept-3-en-1-one) (42). In order to elucidate the structure, an MS/MS study was performed. High-resolution MS/MS spectra of imipenem and the corresponding Ar-BVMO oxidized products are shown in Fig. 6A and andB,B, respectively. The multistage mass spectrometry study of the imipenem precursor ion (300.0867 m/z) allows the fragmentation pathway shown in Fig. 7 to be hypothesized. The main fragmentation pathways are the (consecutive) loss of water and carbon dioxide. The elimination of the amidino-ethylthio lateral chain is noteworthy. High-resolution mass defect data were fundamental to differentiation of the loss of CO2 and NH-CH-NH2, both of which have nominal masses of 44 Da but with different significant digits. The interpretation of the MS/MS spectrum of the Ar-BVMO metabolites is reported in Fig. 8. A double CO2 loss characterizes the spectrum. The intense ion due to the loss of the amidino-ethylthio lateral chain confirms that oxygen insertion occurred on the lactamic ring. A possible intramolecular hydrogen bond justifies both the proposed formation of a new lactone cyclic structure and the chromatographic behavior of the two metabolites which were retained more than the parent compound on the reverse-phase column. The successive loss of water, ammonia, and ethylamidine molecules could explain all the fragment ions obtained. The same fragmentation pathway is valid in the interpretation of the MS/MS spectrum of the ion corresponding to the asymmetrical isomer reported in Fig. 5.
The World Health Organization (WHO) has recently classified antibiotic resistance to be one of the three greatest threats to human health. The β-lactams are one of the most important and frequently used classes of antibiotics in medicine and are essential in the treatment of serious Gram-negative bacterial infections. The past 2 decades have seen substantial increases in the utilization of carbapenems, such as imipenem and meropenem. Predictably, this increase in carbapenem consumption has been accompanied by the emergence of carbapenem-resistant Gram-negative bacterial pathogens (43). The initial and seemingly unstoppable success of antibiotics has been countered by an escalation of resistance mechanisms in bacteria, with the emergence of many genera of bacteria that are resistant to all antibiotics (44,–46). The genus Acinetobacter epitomizes this trend and deserves close attention. Acinetobacter spp. display mechanisms of resistance to all existing antibiotic classes as well as a prodigious capacity to acquire new determinants of resistance (3, 47, 48).
Furthermore, Acinetobacter spp. appear to be well suited for genetic exchange and are among a unique class of Gram-negative bacteria that are described to be naturally transformable (49). The increasing rates of recovery in the clinic of multidrug-resistant A. baumannii strains are a frightening reality (6). A previously published study describing the genome sequences of both susceptible (strain SDF) and resistant (strain AYE) isolates of A. baumannii has shed light on the abundance of the resistance genes found in this organism (50). Fournier and coworkers identified a resistance island in the AYE MDR isolate that contained a cluster of 45 resistance genes (50). Among the key resistance genes identified were those coding for β-lactamases, various aminoglycoside-modifying enzymes, and tetracycline efflux pumps. A significant fraction of the open reading frames was located in putative alien islands, indicating that the genome acquired a large amount of foreign DNA. More than 30 years ago it was demonstrated that Acinetobacter spp. could acquire antimicrobial resistance factors through conjugation of plasmids (51). The genome of A. baumannii contains a gene coding for an ethionamide monooxygenase that not only is phylogenetically part of the same group to which Ar-BVMO belongs but also shares 100% amino acid sequence identity with Ar-BVMO. As mentioned earlier, human FMO3, Ar-BVMO, and M. tuberculosis EtaA can metabolize a common substrate, ethionamide. Based on this, we investigated the ability of Ar-BVMO to use as a substrate two potent and selective aurora kinase inhibitors, danusertib and tozasertib, which are used for the treatment of solid tumors and hematopoietic cancers, respectively, and metabolized by human FMO3 in liver microsomes (52, 53).
In vitro experiments with recombinant Ar-BVMO demonstrated that both danusertib and tozasertib are monooxygenated by Ar-BVMO and have Km values comparable to those obtained with recombinant human FMO3 (29). The reason why a prokaryotic flavin monooxygenase should metabolize substrates that are typically metabolized by the eukaryotic FMO enzymes is not clear. Since both proteins, together with EtaA, belong to the same class of flavoproteins and are involved in detoxification processes, it can be speculated that during evolution some FMO proteins became BVMOs, providing the acquiring organism new metabolic capabilities. This is not the first example of a bacterial FMO that shares the same substrate with a human flavoprotein. Chen and coworkers (54) showed that an FMO from Methylocella silvestris is able to oxidize trimethylamine, which is a typical substrate of eukaryotic FMO enzymes. It can be assumed that the A. baumannii enzyme would be able to monooxygenate the two anticancer drugs, but further experiments are required to confirm this. It could be hypothesized that genetic exchange between the progenitor (A. radioresistens) and its recipient of clinical relevance (A. baumannii) may have occurred in humans (18), leading to the acquisition of BVMO by the latter species. Both A. radioresistens and A. baumannii are identified on the human skin, especially in hospitalized patients; it is therefore possible that BVMO gene exchange may have occurred at this location (4).
In order to provide support for the hypothesis of lateral gene transfer, the genetic context of Ar-BVMO (GenBank accession number GG705134.1) was evaluated and compared to that of EtaA of Acinetobacter baumannii strains 1461402 (GenBank accession number JEWP01000001), 230853 (GenBank accession number JEWW01000001), 983759 (GenBank accession number JEXO01000002.1), and 263903-1 (GenBank accession number JMNM01000001). The genes upstream and downstream of Ar-BVMO not only were found to be identical to those in the four A. baumannii strains but also were found to be in the same position as those of the four A. baumannii strains. In addition, the GC content (55) of the ethionamide monooxygenase genes of the four A. baumannii strains was compared with the GC content of the A. radioresistens genome. The GC content was calculated using Endmemo software, and the values were 42.65% and 42.69% for the A. baumannii EtaA gene and Ar-BVMO of A. radioresistens, respectively. These data do not support horizontal gene transfer, unless a large portion of the genome was transferred from A. radioresistens to A. baumannii, and therefore, a more-in-depth analysis is required, but that was outside the scope of this work.
The presence of a flavin monooxygenase in an MDR bacterium that has 100% sequence identity with Ar-BVMO (i.e., the proteins are identical) with prodrugs as the substrates strongly suggests that the EtaA could be used by A. baumannii to metabolize/detoxify different types of synthetic drugs, including antibiotics, in exactly the same way as Ar-BVMO.
In this work, we have demonstrated that sensitive E. coli cells overexpressing Ar-BVMO became resistant to the antibiotic imipenem. An NADPH consumption assay showed that imipenem is a substrate of Ar-BVMO, and an agar well diffusion assay proved that the enzyme destroys the antibacterial property of the antibiotic. Like all other β-lactams, imipenem inhibits bacterial cell wall synthesis by binding to and inactivating penicillin binding proteins. Antibiotic resistance can occur via three general mechanisms: prevention of the interaction of the drugs with the target, efflux of the antibiotic from the cell, and direct destruction or modification of the compound. The most prevalent mechanism of resistance to β-lactams is the production of β-lactamases. Among the β-lactamases, metalloenzymes are distinguished by having a zinc ion required for enzymatic activity (56). Sometimes referred to as carbapenemases, they are able to hydrolyze many β-lactam antibiotics, including carbapenems (57). In the present work, we demonstrate for the first time that a bacterial flavin monooxygenase is able to metabolize imipenem, thereby modifying the parent compound so that it is no longer active. Two molecules with evidence of the transformation of the lactam ring, according to the Baeyer-Villiger oxygen insertion mechanism, were identified by LC-HRMS analysis. This is not the first report of a flavin monooxygenase catalyzing antibiotic inactivation, since Yang and coworkers have previously reported that TetX, a flavin-dependent monooxygenase from bacteria of the genus Bacteroides, regioselectively hydroxylates tetracycline when it is used as a substrate (40). However, they could not capture the product and suggested that the resulting hydroxylated product was an unstable compound that underwent a nonenzymatic decomposition (40).
Future efforts will be focused on the screening of other carbapenem members, and if under physiological conditions appreciable carbapenemase activity is attributed to this monooxygenase enzyme, it could represent a new target for drug design in the battle against carbapenem resistance in Acinetobacter.
This work was supported in part by the Progetto Ateneo-San Paolo 2012 (Italy) awarded to S.J.S.