|Home | About | Journals | Submit | Contact Us | Français|
A novel multidrug resistance phenotype mediated by the Cfr rRNA methyltransferase is observed in Staphylococcus aureus and Escherichia coli. The cfr gene has previously been identified as a phenicol and lincosamide resistance gene on plasmids isolated from Staphylococcus spp. of animal origin and recently shown to encode a methyltransferase that modifies 23S rRNA at A2503. Antimicrobial susceptibility testing shows that S. aureus and E. coli strains expressing the cfr gene exhibit elevated MICs to a number of chemically unrelated drugs. The phenotype is named PhLOPSA for resistance to the following drug classes: Phenicols, Lincosamides, Oxazolidinones, Pleuromutilins, and Streptogramin A antibiotics. Each of these five drug classes contains important antimicrobial agents that are currently used in human and/or veterinary medicine. We find that binding of the PhLOPSA drugs, which bind to overlapping sites at the peptidyl transferase center that abut nucleotide A2503, is perturbed upon Cfr-mediated methylation. Decreased drug binding to Cfr-methylated ribosomes has been confirmed by footprinting analysis. No other rRNA methyltransferase is known to confer resistance to five chemically distinct classes of antimicrobials. In addition, the findings described in this study represent the first report of a gene conferring transferable resistance to pleuromutilins and oxazolidinones.
The bacterial ribosome is the site of protein synthesis and the target for many chemically diverse classes of antimicrobial agents. The antimicrobial drugs target important functional centers of the ribosome and most often bind to rRNA. Recently, a new phenicol and clindamycin resistance phenotype was found to be caused by an RNA methyltransferase designated Cfr. A detailed analysis by drug footprinting studies and matrix-assisted laser desorption-ionization time of flight/tandem mass spectrometry showed that Cfr adds an additional methyl group at position A2503 of 23S rRNA (9). Since A2503 is located in close proximity to the overlapping ribosomal binding sites of phenicols and clindamycin, it was concluded that the Cfr-mediated methylation confers resistance to these two classes of antimicrobial agents by interfering with the positioning of the drugs (9).
The cfr gene was first discovered in 2000 during a surveillance study for florfenicol resistance among staphylococci from animals. It was initially detected on the 17.1-kb multiresistance plasmid pSCFS1 from a bovine strain of Staphylococcus sciuri (24) and has also been found in bovine strains of Staphylococcus simulans (6). In addition to cfr, the pSCFS1 plasmid carries the rRNA methylase gene erm(33), the aminocyclitol phosphotransferase gene spc, and the ABC transporter gene lsa(B), which confer resistance to macrolide-lincosamide-streptogramin B (MLSB) antibiotics, spectinomycin, and lincosamides, respectively. The cfr gene was recently detected on the 35.7-kb plasmid, pSCFS3, from a porcine Staphylococcus aureus strain, together with the chloramphenicol/florfenicol exporter gene fexA (8). Cloning of the cfr gene and expression in Escherichia coli revealed that Cfr conferred resistance not only in the original gram-positive hosts but also in gram-negative bacteria. Comparison with other protein sequences deposited in the databases showed that the Cfr protein is not related to other known resistance-conferring rRNA methyltransferases but rather to the Radical SAM superfamily (9), which includes a wide range of enzymes from a diverse set of bacteria involved in protein radical formation, isomerization, sulfur insertion, anaerobic oxidation, and unusual methylations (26).
As the Cfr-mediated methylation of position A2503 of 23S rRNA confers resistance to chloramphenicol and florfenicol (phenicol drugs) and clindamycin (a lincosamide drug) (9), it may also affect binding of other drugs to the ribosomal peptidyl transferase center. Therefore, we assayed strains harboring the cfr gene for decreased susceptibility to a number of important antimicrobial drugs that are known to bind close to A2503 at the peptidyl transferase center. These included pleuromutilins, oxazolidinones, and streptogramin A antibiotics. The effect of cfr on drug susceptibility was investigated both in gram-negative E. coli and gram-positive S. aureus strains with plasmids lacking and carrying the cfr gene. Moreover, drug binding to Cfr-methylated ribosomes was investigated by footprinting studies.
A 3,594-bp BglII fragment carrying the cfr gene from plasmid pSCFS3 (EMBL database accession number AJ879565) was inserted into the pBluescript II SK(+) cloning vector (Stratagene, Amsterdam, The Netherlands). The recombinant plasmid, designated pBglII, was transformed into E. coli recipient strains HB101 or AS19 (25) by the CaCl2 method (23). To investigate the effects of cfr, a deletion variant of pBglII was constructed by BamHI digestion and subsequent religation of the largest fragment. This variant, designated pBamHI, had a 1,664-bp deletion which included the 5′ terminal 532 bp of the cfr gene and its regulatory region. Transformation of the staphylococcal plasmids pSCFS1 carrying the cfr gene (24), pSCFS3 harboring the cfr plus fexA genes (7), or pSCFS2 carrying only the fexA gene (8) into the recipient strain S. aureus RN4220 (14) was achieved by polyethylene glycol-mediated protoplast transformation (23).
The original E. coli recipient strain, AS19, E. coli AS19 transformants that carried either the empty cloning vector or the recombinant vector pBglII or pBamHI, as well as the original S. aureus recipient strain, RN4220, and S. aureus RN4220 transformants carrying plasmid pSCFS1, pSCFS2, or pSCFS3, were comparatively investigated for their MICs to the antimicrobial agents listed in Table Table1.1. The determination of MICs by broth macrodilution or broth microdilution was performed according to guideline M31-A2 of the Clinical and Laboratory Standards Institute (formerly NCCLS) (12) using S. aureus ATCC 29213 and E. coli ATCC 25922 as quality control strains. All MIC determinations were performed at least three times. The MICs of the test strains and the quality control strains were validated according to the data presented in Clinical and Laboratory Standards Institute documents M31-S1 (13) and M100-S14 (1).
E. coli strains carrying an intact cfr gene or not (9) were grown in LB broth to an optical density of 0.3 at 450 nm and harvested by centrifugation. The cells were washed and resuspended in TMN buffer (50 mM Tris-HCl at pH 7.8, 10 mM MgCl2, 100 mM NH4Cl). Then, the cells were lysed by sonication, and cell debris was removed by centrifugation (twice for 12 min at 6,000 rpm). Ribosomes were collected from the supernatants by centrifugation at 18,000 rpm for 16 h at 4°C in a Beckman Ti50 rotor. The pellet was resuspended in TMN buffer and stored at 80°C.
E. coli 70S ribosomes (2.5 pmol) were incubated with 0.5, 2, or 10 μM tiamulin (a gift from Novartis), 0.2, 0.5, or 2 μM valnemulin (a gift from Novartis), 0.5, 5, or 50 μM virginiamycin M1 (Sigma-Aldrich), or with no antimicrobial agent in modification buffer (50 mM HEPES-OH, pH 8.0, 10 mM MgCl2, 100 mM KCl, 5 mM dithiothreitol) for 30 min at 37°C. The ribosome complexes (12.5 μl) were modified with 12.5 μl CMCT [1-cyclohexyl-3(2-morpholinoethyl)-carbodiimide metho-p-toluene sulfonate] (42 mg/ml in modification buffer) for 20 min at 37°C. The reactions were terminated by precipitating the ribosomes with ethanol. The ribosomes were recovered by centrifugation, resuspended in 0.3 M sodium acetate, and extracted with phenol and chloroform. rRNA was precipitated with ethanol, resuspended in water, and monitored by primer extension analysis with avian myeloblastosis virus reverse transcriptase (Finnzymes). The 5′-32P-labeled deoxyoligonucleotide primer Ec2654 (5′-TCCGGTCCTCTCGTACT-3′), complementary to nucleotides 2654 to 2670 of E. coli 23S rRNA, was used. The cDNA extension products were separated on 8% polyacrylamide sequencing gels. The positions of the stops were visualized by autoradiography and identified by referencing to dideoxy sequencing reactions on 23S rRNA that were electrophoresed in parallel. Reverse transcriptase stops one nucleotide before the corresponding nucleotide in the sequencing lanes.
The expanded view of the drug binding site (Fig. (Fig.1B)1B) was created by aligning the coordinates of the four antibiotic-50S ribosomal subunit cocrystal structures relative to the RNA surrounding the peptidyl transferase center in the molecular modeling package MolMol (10). The image showing the slice plane (Fig. (Fig.1A)1A) was generated using VMD (5). A molecular surface of the RNA surrounding the drug-binding cavity was generated and the surface area of nucleotide A2503 colored red. The surface was cut by a slice plane to show the internal components of the binding site. The image was postprocessed in a drawing package to emphasize the location of the slice plane.
The effect of Cfr on the antimicrobial susceptibility of bacterial strains was investigated. The rationale for drug selection was based on current knowledge of antibiotic binding sites on the ribosome derived from footprinting, mutational, cross-linking, and crystallographic data. The drugs are all of clinical or veterinary importance and bind at the peptidyl transferase center in close proximity to 23S rRNA nucleotide A2503, the position targeted by the Cfr methyltransferase. In addition to the previously investigated phenicols (florfenicol and chloramphenicol) and clindamycin (a member of the lincosamide group), we have investigated the drugs described below. Oxazolidinones (in this study represented by linezolid) and streptogramins are clinically important antimicrobial agents for the control of infections in humans caused by gram-positive cocci. Both the natural origin and the clinical use of streptogramins are as drug mixtures composed of an A and B component that act synergistically. Only the A component (as, for example, virginiamycin M1) interacts at the peptidyl transferase center, while the B component targets the macrolide binding site on the ribosome. The availability of these drugs as separate compounds sets some limits on their investigation. As the streptogramin A compound dalfopristin is not available, the only commercially available streptogramin A, virginiamycin M1, was included in this study as a representative of this class. The pleuromutilin antibiotics tiamulin and valnemulin are used in the treatment of economically important infections of the intestinal and respiratory tracts of pigs caused by Brachyspira spp., Mycoplasma spp., and Actinobacillus pleuropneumoniae, as well as respiratory tract infections in poultry due to Mycoplasma spp. Additional pleuromutilin derivatives are currently being developed for use in humans.
Antimicrobial susceptibility testing was conducted with both Cfr-positive (+Cfr) and Cfr-negative (−Cfr) S. aureus RN4220 (14) and E. coli AS19 (25) to investigate whether differences in antimicrobial susceptibility were detectable in gram-positive and gram-negative host bacteria. As E. coli has a relatively low susceptibility to many drugs, the hyperpermeable strain E. coli AS19 was used to emphasize the observed effects. All test strains were exposed to 10 antimicrobial drugs and two drug mixtures at various concentrations, and the MIC results are shown in Table Table11.
The presence of the Cfr methyltransferase substantially reduces the susceptibilities to all antimicrobials known to bind close to A2503. The relative MIC difference between strains lacking or harboring the Cfr methyltransferase varied from 4- to ≥4,096-fold (Table (Table11 and Fig. Fig.1C).1C). Although the overall pattern was the same for both E. coli and S. aureus (Table (Table1),1), the magnitude of the MIC increases for the pleuromutilins differed distinctly between the two host organisms. For E. coli, the tiamulin and valnemulin MICs increased 128- and 8-fold, respectively, whereas the corresponding MIC increases for S. aureus are ≥2,048- and ≥4,096-fold. According to the approved clinical breakpoints for the antimicrobial agents that are available, strains with the Cfr methyltransferase exhibit MICs that allow their classification as either resistant (e.g., S.aureus RN4220::pSCFS1 for chloramphenicol, linezolid, quinupristin-dalfopristin, and clindamycin; S. aureus RN4220::pSCFS3 for chloramphenicol and clindamycin) or borderline susceptible (e.g., S. aureus RN4220::pSCFS3 for linezolid and quinupristin-dalfopristin) (see footnote c of Table Table11).
As a control that the Cfr methyltransferase only affects sensitivity to a subset of antimicrobials, susceptibility testing was also performed with the macrolides erythromycin, acetylisovaleryltylosin, and telithromycin. Macrolides bind in the peptide exit tunnel in a cleft adjacent to the peptidyl transferase center (15), but they are not in direct contact with nucleotide A2503 of 23S rRNA. As expected from the different ribosomal binding site of macrolides, the Cfr-mediated methylation of A2503 did not affect erythromycin, acetylisovaleryltylosin, and telithromycin MICs for E. coli and S. aureus (Table (Table1).1). S. aureus RN4220::pSCFS1 is erythromycin resistant because pSCFS1 also carries the inducibly expressed MLSB resistance gene erm(33), for which erythromycin is an excellent inducer. The same strain has low MICs for acetylisovaleryltylosin and telithromycin, since 16-membered macrolides, such as tylosin derivatives, and ketolides are not efficient inducers. The high MICs for chloramphenicol and florfenicol observed for S. aureus RN4220::pSCFS3 are most likely due to a synergistic effect of the two different resistance mechanisms including target site modification via Cfr and phenicol efflux via FexA (7). The fourfold-higher MIC for quinupristin-dalfopristin of S. aureus RN4220::pSCFS1 than that of S. aureus RN4220::pSCFS3 is presumably the result of a synergistic effect of Cfr-mediated streptogramin A resistance and low-level expression of the pSCFS1-borne macrolide-lincosamide-streptogramin B resistance erm(33) gene in the absence of an inducer.
A reduced level of chloramphenicol, florfenicol, and clindamycin binding to Cfr-modified ribosomes was established previously using chemical footprinting (9). Here we show that it is also true for the pleuromutilin drugs tiamulin and valnemulin and the streptogramin A drug virginiamycin M1 (Fig. (Fig.2).2). A reduction in drug binding to ribosomes can be monitored by changes in the chemical modification pattern of rRNA at the drug binding site. In the chemical footprints of tiamulin, valnemulin, and virginiamycin M1, nucleotides U2506, U2584, and U2585 of 23S rRNA are protected from CMCT modification (16, 19). Binding of antimicrobials to ribosomes isolated from E. coli strains lacking or harboring the Cfr methyltransferase was assayed by CMCT modification of ribosome-drug complexes, followed by primer extension analysis with reverse transcriptase and gel electrophoresis. Autoradiograms of the tiamulin, valnemulin, and virginiamycin M1 footprints on ribosomes are shown in Fig. Fig.2.2. The protections observed at U2506 and U2584/U2585 in the presence of the drugs are significantly decreased in ribosomes from +Cfr cells compared to those from −Cfr cells, indicating that drug binding is reduced in +Cfr cells. The prominent additional band present in the +Cfr samples is from the reverse transcriptase stop caused by the Cfr-mediated methylation at nucleotide A2503. The drug mixtures virginiamycin (virginiamycin S and M1) and quinupristin-dalfopristin are composed of streptogramin B and A components, respectively. These bind synergistically to overlapping sites on the ribosome, and therefore an evaluation of their relative binding by footprinting is not feasible. Since the footprinting data show that binding of virginiamycin M1 to +Cfr relative to −Cfr ribosomes is reduced (Fig. (Fig.2B),2B), it is believed that the attenuated binding affinities of the streptogramin A components of the mixtures are responsible for the observed MIC increases for virginiamycin and quinupristin-dalfopristin in strains expressing the cfr gene (Table (Table1).1). Taken together with our previous data (9), the footprinting data demonstrate that the binding of phenicol, lincosamide, pleuromutilin, and streptogramin A drugs to ribosomes modified by Cfr is diminished.
The exact position of single representatives from the aforementioned four groups of antimicrobial agents bound to a bacterial 50S ribosomal subunit is available from X-ray crystallography. The crystal structures of bound chloramphenicol, clindamycin, tiamulin, and dalfopristin (4, 20, 21) show that they bind to overlapping sites at the peptidyl transferase site at the entrance of the ribosomal exit tunnel (Fig. 1A and B). The corresponding data for clindamycin and virginiamycin M1 with the archaeon Haloarcula marismortui show the same binding sites with only minor differences in orientation and interactions and thus support the bacterial sites (27). The target of Cfr methylation, nucleotide A2503, comprises a considerable part of the exposed surface area in the peptidyl transferase cavity (shown in red in Fig. Fig.1B),1B), and chloramphenicol, clindamycin, tiamulin, and dalfopristin bind in close proximity to this surface area. A simple model of resistance is suggested, in which the modification at A2503 interferes with drug binding. The methylation either disturbs binding directly or indirectly by causing a shift in the position of A2503, which in turn leads to decreased drug binding.
Various lines of evidence suggest that oxazolidinones also target the peptidyl transferase center. Most of the 23S rRNA resistance mutations known to confer linezolid resistance are located near the site of peptide bond formation (11). In vivo cross-linking data show that oxazolidinones cross-link to nucleotide A2602, a position that is also cross-linked to sparsomycin in the presence of a P-site-bound tRNA (2). Like sparsomycin, oxazolidinones do not produce a footprint on empty ribosomes. However, a weak linezolid footprint has been reported on a ribosome-mRNA-tRNA complex (30). In addition, an X-ray structure of an oxazolidinone-50S complex shows that this drug binds in the A-site of the peptidyl transferase center of the ribosome (J. Ippolito, Z. Kanyo, B. Wimberly, D. Wang, E. Skripkin, J. Devito, B. Freeborn, J. Sutcliffe, E. Duffy, and F. Franceschi, Abstr. 45th Intersci. Conf. Antimicrob. Agents Chemother., abstr. F-1254, 2005). The observed increases in linezolid MICs (Table (Table1)1) in strains expressing the Cfr methyltransferase can thereby be rationalized in the same manner as for the drugs described above. It is thus expected that bacterial strains expressing Cfr will show decreased susceptibility to all drugs belonging to these five groups of antimicrobial agents. Therefore, we suggest naming the phenotype caused by the Cfr methyltransferase PhLOPSA for Phenicol, Lincosamide, Oxazolidinone, Pleuromutilin, and Streptogramin A resistance.
A frightening scenario can be foreseen if the appearance and spread of the Cfr methyltransferase parallels the situation observed with the Erm methyltransferases and combined resistance to MLSB. The Erm family consists of approximately 40 different classes of methylases (http://faculty.washington.edu/marilynr/), all acting at position A2058 of 23S rRNA and causing MLSB resistance (18). Mutations in 23S rRNA can also cause the MLSB phenotype (28), but these mutations are not transferable by horizontal gene transfer. In contrast, erm methylase genes have been identified in a wide range of gram-positive and gram-negative bacteria with the transposon-borne erm(B), erm(F), and erm(A) genes, as well as the plasmid-borne erm(C) gene having the broadest host range (18). About 10 years ago, the number of reports of Erm-mediated MLSB resistance rose suddenly and it was discovered that this type of resistance appeared worldwide in a number of different bacteria. The high incidence of resistance was probably caused by the extensive use of macrolides for treatment of bacterial infections in humans and animals and by their use as growth promoters in the farming industry. As the acquired resistance is not detrimental to bacteria, it can persist for a long time, which in turn promotes its spread.
The cfr gene has been identified on structurally related multiresistance plasmids from animal staphylococci and can, in principle, be easily disseminated among staphylococci. However, surveillance studies in Germany have identified only 6 cfr-carrying staphylococcal strains during the past 17 years (7). The low prevalence of the cfr gene might result from a comparatively low selective pressure imposed by the PhLOPSA drugs on animal staphylococci. Of the relevant drug classes, only lincosamides are approved for several indications in food and companion animals in the European Union. Florfenicol and pleuromutilins are approved only for selected infections in cattle and/or swine. Moreover, the use of chloramphenicol has been prohibited in food animals in the European Union since 1994. Although previously used as growth promoter, neither virginiamycin nor other streptogramin antibiotics are currently approved in the European Union for use on animals, and the oxazolidinone drug class is licensed exclusively for human use worldwide.
As very few new antimicrobial agents appear on the market, the fact that Cfr confers resistance to five different classes of relevant antimicrobial agents warrants attention. For the drugs with approved clinical breakpoints with Staphylococcus spp., the elevated MICs observed in strains expressing Cfr classifies them as resistant or borderline susceptible (Table (Table1).1). In the case of the streptogramins, the Cfr-mediated decrease in susceptibility to the A component, in combination with other resistance mechanisms, such as Erm-mediated resistance to the B component as seen with the erm(33) gene present together with cfr on plasmid pSCFS1, could seriously affect drug efficiency. The oxazolidinone linezolid is a last-resort antimicrobial agent for the control of gram-positive bacterial pathogens against which no other antimicrobials are effective anymore. Therefore, transferable resistance or largely decreased susceptibility to oxazolidinones is of particular concern.
The detection of the plasmid-borne resistance gene cfr is, to our knowledge, the first report of transferable resistance to pleuromutilins and oxazolidinones. In Brachyspira isolates with reduced susceptibility to tiamulin, mutations in ribosomal protein L3 and 23S rRNA genes have been identified (17). The reports of linezolid resistance published to date have also described only mutations in 23S rRNA or ribosomal protein L4 genes (11, 29). Thus, it was believed that the likelihood of transferable resistance to either pleuromutilins or oxazolidinones based on these mechanisms was rather low, and it has been claimed that there are no mechanisms of cross-resistance to linezolid (3). However, this assumption must be revised on the basis of the findings presented in this study.
We conclude that expression of the Cfr methyltransferase confers a PhLOPSA resistance phenotype. As the Cfr-modified nucleotide A2503 abuts the overlapping binding sites of PhLOPSA drugs and drug binding to Cfr-modified ribosomes is impaired, we infer that the resistance is caused by perturbation of the drug-binding site on the ribosome.
We thank L. H. Hansen and Vera Nöding for excellent technical assistance and Pfizer for providing linezolid.
This work was supported by The Danish National Research Foundation, the European Commission's 5th Framework Program (grant QLK2-CT-2002-00892), and the Deutsche Forschungsgemeinschaft (SCHW 382/6-2 and SCHW 382/6-3).