|Home | About | Journals | Submit | Contact Us | Français|
Tolerance is a poorly understood phenomenon that allows bacteria exposed to a bactericidal antibiotic to stop their growth and withstand drug-induced killing. This survival ability has been implicated in antibiotic treatment failures. Here, we describe a single nucleotide mutation (tol1) in a tolerant Streptococcus gordonii strain (Tol1) that is sufficient to provide tolerance in vitro and in vivo. It induces a proline-to-arginine substitution (P483R) in the homodimerization interface of enzyme I of the sugar phosphotransferase system, resulting in diminished sugar uptake. In vitro, the susceptible wild-type (WT) and Tol1 cultures lost 4.5 and 0.6 log10 CFU/ml, respectively, after 24 h of penicillin exposure. The introduction of tol1 into the WT (WT P483R) conferred tolerance (a loss of 0.7 log10 CFU/ml/24 h), whereas restitution of the parent sequence in Tol1 (Tol1 R483P) restored antibiotic susceptibility. Moreover, penicillin treatment of rats in an experimental model of endocarditis showed a complete inversion in the outcome, with a failure of therapy in rats infected with WT P483R and the complete disappearance of bacteria in animals infected with Tol1 R483P.
Bacteria have evolved at least two mechanisms that allow them to escape the killing effects of bactericidal antibiotics: resistance and tolerance. Resistant bacteria can typically grow in the presence of a drug at a concentration greater than the one inhibiting the growth of the majority of the other strains of the same species. Therefore, in vitro, resistant bacteria exhibit an increased MIC to the drug. However, they remain susceptible to drug-induced killing when they are exposed to concentrations exceeding their new increased MIC (35). In contrast, the MIC for tolerant bacteria is unchanged from the MIC for the majority of the other strains of the same species, but they have a considerably increased ability to survive killing, even at concentrations exceeding the drug MIC by several orders of magnitude (23). Hence, bactericidal drugs act as mere bacteriostatic agents toward tolerant bacteria.
Tolerance was first reported in autolysin-defective laboratory mutants of Streptococcus pneumoniae in 1970 (52). This was followed by the isolation of β-lactam-tolerant clinical strains of Staphylococcus aureus in 1974 (3). Since then, antibiotic tolerance has been implicated in therapeutic failures in cases of endocarditis (18, 24, 28), meningitis (34), osteomyelitis (50), pharyngitis (58), and bacteremia (41). Furthermore, tolerant bacteria are believed to represent a reservoir of treatment survivors that can potentially develop further resistance (25, 55), yet in contrast to antibiotic resistance, the molecular mechanisms underlying tolerance are poorly understood. Moreover, the contribution of tolerance to treatment failure remains difficult to assess, mainly due to a lack of proper detection techniques (22).
Previous work in our laboratory showed that tolerant mutants of Streptococcus gordonii—including the Tol1 mutant evaluated in the present study—were deregulated in the expression of the arginine deiminase (arc) operon (10), yet arc was not the primary cause of tolerance but, rather, a marker of tolerance, since its inactivation did not abolish the drug survival phenotype (10). In S. gordonii as well as in other bacterial species, arc is under the control of the carbon catabolite repression system (CCR), a global regulatory mechanism that allows bacteria to use the most efficient carbohydrate available for their growth (9, 17). One of the trans-acting factors of CCR is the protein CcpA. We observed that CcpA was necessary for the complete expression of the tolerance phenotype of Tol1 both in vitro and in vivo (5), yet the ccpA gene of Tol1 was not mutated and its expression was unchanged compared to that of the killing susceptible wild-type (WT) parent. Thus, CcpA was not the tolerance effector itself, indicating that another as yet unidentified tolerance mutation had to be located elsewhere.
Indeed, the tolerance phenotype of Tol1 was transformable and could be mapped on a 200-kb SmaI fragment of its chromosome, with backcross experiments indicating a transformation rate compatible with the existence of a point mutation (10). Here, we show that a single nucleotide mutation in ptsI, the gene coding for enzyme I (EI) of the S. gordonii sugar phosphotransferase system (PTS) (Fig. (Fig.1)1) (57), confers antibiotic tolerance both in vitro and in a rat model of experimental endocarditis. The ptsI tolerance mutation induces a proline-to-arginine substitution in the carboxy-terminal domain of the protein, which is responsible for its homodimerization. This substitution decreases its activity and subsequently reduces the level of carbohydrate uptake.
The bacterial strains, plasmids, and primers used in this study are described in Table Table1.1. The streptococci were grown at 37°C in brain heart infusion (BHI) broth (Difco) without aeration or on Columbia agar (Becton Dickinson) supplemented with 3% human blood. Growth was followed both by measurement of the optical density at 600 nm (OD600) with an Ultrospec 500 pro spectrophotometer (Amersham Biosciences) and by determination of the viable CFU counts on agar plates. Escherichia coli strains were grown at 37°C in Luria-Bertani (LB) broth (Difco) or on LB agar. When appropriate, the following antibiotics were added to the medium at the indicated concentrations: streptomycin, 100 μg/ml; erythromycin, 5 μg/ml; and ampicillin, 75 μg/ml. The bacterial stocks were stored at −80°C in either BHI or LB broth supplemented with 10% (vol/vol) glycerol.
The MICs were determined by a standard macrodilution method (62). Time-kill curves were determined with exponentially growing streptococcal cultures by adding an antibiotic concentration equal to 500× the MIC (i.e., 2 μg/ml for penicillin G and imipenem) at an OD600 of 0.2 (corresponding to about 1 × 108 CFU/ml), as described previously (5). This concentration was chosen because it is readily achieved in the serum of humans during standard penicillin therapy (2) and was previously used in our laboratory for the selection of tolerant mutants (5, 10).
An S. gordonii ptsI deletion mutant was generated by the PCR ligation mutagenesis technique (32) with primer pair ptsI_L5 and ptsI_L3 and primer pair ptsI_R5 and ptsI_R3 to amplify the upstream and the downstream portions of the ptsI gene, respectively.
Plugs of streptococcal genomic DNA, macrorestriction with SmaI, pulsed-field gel electrophoresis (PFGE), and gel extraction of DNA fragments were performed by use of a published method (10).
A library of a 200-kb tolerance-conferring fragment of S. gordonii Tol1 was generated by extracting it from PFGE gels, followed by digestion with DraI, ligation of the restriction fragments into pJDC9, and transformation into E. coli DH5α. The inserts from randomly chosen clones were sequenced and aligned with the S. gordonii Challis NCTC 7868 chromosomal sequence (available at http://www.tigr.org/) restricted in silico with SmaI. This allowed the identification of a 200.222-kb segment, which was compatible to the length of the original PFGE fragment.
To seek the tolerance mutation contained in this fragment, we refined the library by designing 21 pairs of primers amplifying 13-kb overlapping regions of the entire 200-kb segment (data not shown). The PCR products were pooled in groups of amplicons, which were used to transform competent S. gordonii WT arcB::luc cultures to tolerance by a previously described enrichment technique (10). In this setting, tolerant derivatives would arise after fewer enrichment cycles in cultures transformed with a DNA fragment harboring the putative tolerance mutation. For each pool, the acquisition of tolerance was selected for by successive enrichment cycles, each of which consisted of the exposure of exponentially growing bacteria to 500 times the penicillin G MIC for 20 h, followed by three washes by centrifugation and the addition of antibiotic-free medium, which allowed the regrowth of survivors. The tolerance phenotype was assessed by determining the loss of viability by determination of the CFU counts. Moreover, since the recipient cells contained a luciferase insert in the arc operon, arc deregulation, measured by luminescence experiments, was used as an additional control for conversion to tolerance (10).
The introduction of single nucleotide mutations in the S. gordonii WT or Tol1 chromosome was realized by using a published method (46). In brief, 1.5-kb PCR-amplified DNA fragments harboring either the wild-type or tolerant ptsI sequence were cotransformed with pBS714, a lacG insertion vector based on a published plasmid (6). Emr transformants were selected and screened by using the mismatch amplification mutation assay principle (39). Correct clones were subsequently sequenced to confirm the nucleotide change.
The structure of the carboxy-terminal domain of the EI was predicted by using Protein Data Bank template 2BG5, which corresponds to the carboxy-terminal domain of the Thermoanaerobacter tengcongensis EI (36). The model was built by homology by using the Modeler program (42). To analyze the contribution of Pro 483 to the stability of the dimer, the binding free energy (ΔG) contribution of this residue was evaluated by using the FoldX program (45), after 500 steps of energy minimization, and by using an r-dependent dielectric constant and the Charmm program (8). The contribution of all residues of monomer 2 within 10 Å of Pro 483 from monomer 1 was assessed.
The wild-type and the mutated ptsI genes were amplified from the WT and Tol1 genomic DNA by using primers Histag_PtsI_5 and Histag_PtsI_3, respectively, and were subcloned into expression vector pJFFruA_H6, from which the original fruA gene insert was removed. The constructs were entirely sequenced, named pJFPtsI483P_H6 (wild-type sequence) and pJFPtsI483R_H6 (tolerant sequence), and transformed into E. coli TH074, which lacks a functional ptsI gene. This yielded strains E. coli TH074(pJFPtsI483P_H6) and E. coli TH074(pJFPtsI483R_H6). An E. coli TH074 strain transformed with pJFFruAH_6_H6 and a plasmid harboring the wild-type E. coli EI gene (pJFPtsI) were used as a negative control and a positive control, respectively. The expression of the recombinant EI was controlled by electrophoresis of crude extracts on 18% sodium dodecyl sulfate-polyacrylamide gels stained with Coomassie brilliant blue, which showed ca. 64-kDa bands after induction with 0.2 mM isopropyl-β-d-thiogalactopyranoside (IPTG) (data not shown).
Cell lysates were prepared from E. coli TH074 transformed with the EI-encoding plasmids and were assayed for EI activity in the presence of purified phosphotransferase protein (HPr; 277 μM), IIAGlc (4.7 μM), IICBGlc (3.9 μM), phosphoenolpyruvate (PEP; 1 mM), and [U-14C]glucose (0.5 mM), as described previously (19, 30, 33). After 30 min of incubation at 37°C, the free and the phosphorylated sugars were separated by anion-exchange chromatography and the level of radioactive [U-14C]glucose-6-phosphate was determined by liquid scintillation counting.
S. gordonii cells were grown to an OD550 of ca 0.3 in BHI broth, harvested by centrifugation at 4°C, washed two times with phosphate-buffered saline (PBS)-5 mM MgCl2, and resuspended in 1 ml of the same buffer (final OD550 range, 20 to 25). Then, 220 μl of cell suspension was added to 440 μl of buffer and the mixture was incubated for 10 min at 35°C. As a control used to account for the nonspecific adsorption and/or the uptake of [14C]glucose by bacteria at 4°C and a high dilution, 100 μl of this suspension was diluted in 10 ml stopping buffer (PBS and 5 mM MgCl2 at 4°C), and then 2 μl of 20 mM [14C]glucose (600 dpm/nmol) was added. The uptake reaction was started by the addition of 11 μl of 20 mM [14C]glucose (600 dpm/nmol) to the preincubated cell suspension. Aliquots of 100 μl were removed at various times and diluted in 10 ml stopping buffer. Cells were collected by filtration through Whatman GF glass fiber filters. Radioactivity was determined by liquid scintillation counting. For measurement of the level of [14C]glucose uptake in the presence of penicillin, cells were exposed to 2 μg/ml of the antibiotic for either 20, 40, or 60 min before they were processed as described above.
Permission for experimentation with living animals in the present work was granted by the State Veterinary Office of the Canton de Vaud (permission 879.5). Sterile aortic vegetations were produced in female Wistar rats according to a published procedure (26). Twenty-four hours later, groups of animals were inoculated by intravenous (i.v.) challenge with 0.5 ml of saline containing 107 CFU of exponential-phase streptococci. This inoculum was chosen because it consistently produced 100% infection with each strain in titration experiments (data not shown). Penicillin treatment was started 16 h after inoculation. Groups of animals infected with either of the test organisms were treated for 2 days with procaine penicillin (300,000 U/kg of body weight), given subcutaneously every 12 h. This regimen produced peak and trough antibiotic levels in the serum of rats which approximated the concentrations in humans during i.v. penicillin therapy (18). Control rats were killed at the time of treatment onset in order to measure the severity of valve infection at the start of therapy. The treated rats were killed 12 h after the trough level of the last antibiotic dose was achieved. The cardiac vegetations were weighed and plated for determination of the numbers of CFU. Vegetations with negative cultures were given a value of 2 log10 CFU/g, the lower limit of detection, in subsequent calculations for statistical analysis. The median bacterial titers in the vegetations were compared by the nonparametric Mann-Whitney unpaired test. Differences were considered significant when the P value was <0.05 by using two-tailed significance levels.
We previously observed that an uncharacterized ca. 200-kb SmaI fragment of the Tol1 mutant chromosome harbored a putative tolerance mutation that could be transformed back into the kill-susceptible parent (10). Newly available S. gordonii data from The Institute for Genomic Research (TIGR) allowed us to identify the 200-kb genomic sequence (see Materials and Methods). We therefore transformed WT S. gordonii with pools of PCR products covering the entire Tol1 fragment.
The bacteria were then cycled with penicillin for enrichment in tolerant transformants. Cultures transformed with the first pool of PCR products (pool 1) converted to tolerance after two enrichment cycles, and cultures transformed with whole chromosomal DNA of the Tol1 mutant converted to tolerance after three such cycles. In both cases, conversion to tolerance was accompanied by the deregulation of arc expression (data not presented), thus confirming previous observations (10). In contrast, none of the cultures transformed with the other PCR pools or with the chromosomal DNA of the WT converted to tolerance within this frame of time, and their levels of arc expression remained unchanged.
Consequently, the region covered by pool 1 was further sequenced and compared with the sequence available at TIGR. Pool 1 encompassed a 48.838-kb region along which there was only a single nucleotide change between the Tol1 mutant and the S. gordonii strain sequenced by TIGR. As a control, the sequence from the nontolerant WT parent was 100% identical to the TIGR sequence. Hence, the mutation was exclusively present in the Tol1 strain.
The mutation of Tol1 (named tol1) was identified as a C-to-G substitution located at nucleotide 1448 from the start codon of the bacterial ptsI gene (GenBank/EMBL/DDBJ database accession number DQ182633). The ptsI gene codes for the EI of the phosphoenolpyruvate:sugar phosphotransferase system (PTS), which phosphorylates all PTS sugars (Fig. (Fig.1)1) (37, 48). At the amino acid level, the mutation induces a proline-to-arginine substitution at position 483 of the protein.
The tol1 mutation was introduced into the WT parent by the pseudolinkage technique, yielding strain WT P483R. Reciprocally, the WT sequence of ptsI was restored in the Tol1 mutant by the same means, yielding strain Tol1 R483P. To ensure that the incorporation of the plasmid vector used for pseudolinkage (plasmid pBS714; see Materials and Methods) did not interfere with subsequent experiments, transformants that had incorporated the plasmid but not the PCR fragment harboring the desired sequence were used as controls.
The MICs of the two β-lactams tested (penicillin MIC, 0.008 μg/ml; imipenem MIC, 0.004 μg/ml) were unaltered in both the transformants and the empty plasmid controls. Table Table22 presents the results of time-kill experiments performed with these two drugs. Killing of the untransformed WT recipient and the WT P483R tolerant transformant clearly showed that the tol1 mutation conferred β-lactam tolerance to the WT recipients, which became nearly 10,000 times less sensitive to killing by penicillin and imipenem. Conversely, the restoration of the parent ptsI sequence in the transformant Tol1 R483P resulted in the complete loss of its original tolerance phenotype. Chromosomal insertion of the empty plasmid did not affect drug-induced killing, as was observed for the MIC experiments (data not shown).
Since the tol1 mutation was located in the ptsI gene, we determined whether tol1 resulted in an altered function of its EI product. E. coli strain TH074, which lacked a functional ptsI gene, was complemented in trans with a ptsI gene cloned from the S. gordonii WT or the Tol1 mutant. An E. coli ptsI gene was used as a positive control. Cell extracts were assayed for EI-dependent glucose phosphotransferase activity in a heterologous system complemented with purified E. coli PTS subunits. The S. gordonii WT EI was active, although much less so (2%) than the E. coli EI. The Tol1 mutant of EI exhibited no activity in this in vitro assay (Fig. (Fig.2A2A).
To further characterize the difference between the wild-type and Tol1 mutants, we measured the rate of glucose uptake by starved S. gordonii cells harboring either the WT or the mutated tolerant allele (WT P483R) and the influence of the duration of penicillin exposure on sugar transport (Fig. (Fig.2B).2B). Our results clearly show a reduction in glucose uptake that is proportional to the duration of antibiotic exposure only in the WT, whereas in WT P483R, the rate of uptake constantly remained low. It can therefore be concluded that in the parent strain, glucose transport is fully functional and is reduced by antibiotic exposure, whereas in strains harboring the tolerance mutation, glucose transport is already at a low level which is almost not influenced by penicillin exposure.
The next logical step was therefore to completely inactivate the EI gene in wild-type S. gordonii and assess its phenotype. Unexpectedly, the ΔptsI mutant exhibited a dramatic reduction in growth rate in BHI medium. Whereas the WT doubling time was 44 ± 3 min, the ΔptsI mutant had a doubling time greater than 135 min. In contrast to the ΔptsI mutant, the tolerant S. gordonii P483R mutant had a doubling time of 49 ± 3 min, which is only a few minutes longer than that of the WT. The reason for the reduced-growth phenotype of the ΔptsI mutant is not known. One possibility is catabolite repression of a metabolic pathway that is required for growth on BHI medium. Another is the defective activation of a pathway that is positively controlled by the uptake of PTS sugars (60). It is likely that the EI P483R point mutation, unlike the EI deletion, has enough residual activity to alleviate catabolite repression and thus enable growth on BHI medium. Of note, EI has an enzyme flux control coefficient of nearly zero, and glucose metabolism depends on EI only at very low EI concentrations (59). The small difference in growth rates in BHI medium between the wild type and the P483R mutant cannot account for the diminished killing effects of the antibiotics (54). Of note, the ΔptsI mutant was not killed by penicillin, which can be explained by its very slow growth rate (data not shown).
The EI protein consists of two major N-terminal and C-terminal domains. The C-terminal domain is required for homodimerization, which is critical for EI autophosphorylation and further kinase activity (21). Modeling of the tertiary structure of EI (Fig. (Fig.3)3) indicates that the proline at residue 483 is involved in the homodimerization interface formed by the two carboxy-terminal domains during EI autophosphorylation. This is consistent with the model proposed for T. tengcongensis, in which this amino acid is 1 of the 32 conserved residues of the 45 residues making up the dimer interface (36).
To analyze the contribution of Pro 483 to the stability of the homodimer, the binding free energy contribution of this residue was evaluated. The contribution of all residues of monomer 2 within 10 Å of Pro 483 from monomer 1 was assessed. Our data show that Pro 483 might play an important role in stabilizing the dimer formation, as the total binding free energy component associated with this residue is about 8.72 kcal/mol (Table (Table3).3). Since most of this binding free energy is coming from van der Waals interactions with hydrophobic residues on the other monomer, the replacement of Pro 483 by Arg is likely to disrupt some of these interactions and to induce a conformation change that reduces EI activity.
The effect of acquiring or losing the tol1 mutation was tested in rats with experimental endocarditis treated with penicillin (Fig. (Fig.4).4). Both WT mutant P483R and the Tol1 mutant in which the wild-type codon was restored (Tol1 R483P) were equally able to colonize damaged valves after bacterial challenge (they had similar 90% infectious doses of 105 CFU) and were equally able to multiply in the vegetations after colonization (they displayed similar bacterial counts in the vegetations at the start of therapy). As observed in previous studies (5, 18), treatment with penicillin successfully cured rats infected with the kill-susceptible WT parent within 3 days (Fig. (Fig.4B).4B). In contrast, the penicillin treatment failed when the WT parent had acquired the tol1 mutation (strain WT P483R in Fig. Fig.4A).4A). In symmetry, infection with the originally isolated Tol1 mutant was difficult to treat (Fig. (Fig.4B)4B) (5, 18), whereas restoration of the R483 mutation to the P483 wild type in the Tol1 strain (strain Tol1 R483P in Fig. Fig.4A)4A) fully restored susceptibility to therapy and decreased the median bacterial titers to levels under the limit of detection in 100% of the rats. Thus, in accordance with the results of the in vitro time-kill curve studies, the tol1 mutation also conferred tolerance in vivo and had a very significant impact on the treatment outcome.
We previously observed that a putative tolerance mutation was located in the chromosome of an S. gordonii tolerant mutant (10). The results presented in the current study indicate that the tol1 mutation consists of a single C-to-G nucleotide change in the gene ptsI, which codes for EI of the PTS pathway (21).
The PTS pathway mediates the uptake and phosphorylation of selected carbohydrates and regulates cellular metabolism in response to their availability (15, 37, 48). In addition, it interacts with a variety of sensing, signaling, and regulatory pathways, including CCR (15). EI is the conductor of the PTS cascade and is therefore involved in all PTS activities, from carbohydrate transport to signaling (Fig. (Fig.1).1). It is one of the most conserved bacterial proteins without a close homolog in animals. EI is a homodimer of two ca. 64-kDa subunits consisting of two flexibly linked domains. The amino-terminal domain contains the phosphorylation site (His 191) and the HPr-binding site. The carboxy-terminal domain contains the PEP-binding site and the dimerization contact interface (7).
In strains with the tol1 mutation, a nonpolar/neutral amino acid, proline, is replaced by a polar/basic amino acid, arginine. The mutation at residue 483 in the carboxy-terminal domain is located in the dimer interface. The crystal structure of this domain has recently been described in T. tengcongensis (36). Sequence alignments reveal that this proline is conserved in nearly all (n > 100) of the sequenced PtsI gram-positive bacterial homologs that we were able to retrieve from the current NCBI and TIGR databases, with the only exception being Bacillus anthracis, in which a leucine replaces the proline. Moreover, both our structure-based alignments and computer modeling indicate that the mutation is situated in an important residue of the (α/β)8 triosephosphate isomerase barrel (61), which is involved in the dimerization interface of the enzyme (36).
The exact relationship between a possible monomer-dimer equilibrium of EI and phosphotransferase activity is still unclear (36, 49), but mutations in residues in the dimerization interface have been shown to reduce the dimerization and the subsequent autophosphorylation capability of EI (7), which is consistent with its reduced activity in the in vitro functional test reported herein. Indeed, when we measured the end result of the PTS phosphorylation cascade (i.e., the uptake of glucose) in strains harboring either the WT or the tolerant allele, a clear difference in uptake was shown, with strains harboring the mutation having a lower uptake rate. Interestingly, when we quantified sugar uptake after penicillin exposure, we found a clear decrease in the level of transport, proportional to the exposure time, in the WT strain. In contrast, transport in the tolerant strains constantly remained at a low level.
During active replication, nontolerant bacteria are susceptible to drug-induced killing. On the other hand, when bacteria grow at a slower pace or enter the stationary phase because of a depletion of carbohydrate sources, their susceptibility to β-lactam-induced killing decreases abruptly even in the absence of the tolerance mutation, a state which is referred to as phenotypic tolerance (56).
How, then, would the tol1 mutation affect the system? First, it would affect the system by lowering the activity of EI and, as a consequence, sugar uptake. In the wild-type bacteria, penicillin exposure leads to a shutdown of sugar transport and a halt in bacterial growth. However, to be complete, this process takes 60 min of penicillin exposure. During all this time, WT strains remain metabolically active and therefore susceptible to the killing effect of the drug. On the other hand, because tolerant strains exhibit a lower glucose uptake rate as a consequence of the EI mutation and are therefore closer to the metabolically dormant state of bacteria in the stationary phase of growth, their sensitivity to the killing effect, which depends on actively replicating cells, is drastically decreased. Tolerance mediated by this kind of mutation is closer to the phenomenon of phenotypic tolerance (56) or to the phenotype of persistence and phenotypic heterogeneity (1, 16, 31) than to alterations of direct effectors of cell integrity, such as the autolytic enzymes (53). This kind of tolerance mutation might also be more common than alterations of direct effectors of cell integrity because it provides a selective advantage in response to more than one type of noxious stress. Interestingly, various genes linked to bacterial carbohydrate metabolism have been shown to be involved in the tolerance phenomenon (5, 40, 63).
Second, the phosphorylation state of both the EI and the EII complex and the equilibrium of glycolytic intermediates such as fructose 1,6-bisphosphate or free versus bound trans-acting factors (i.e., HPr and CcpA) are involved in the regulation of other cellular processes (13, 21, 44, 47). In the course of bacterial growth, the ratios between the various phosphorylation states of these proteins vary, depending on the carbohydrates present in the environment and the phase of growth of the microorganism (12). As well, the PTS is linked to CCR through HPr phosphorylated on a serine 46 residue and bound with CcpA. CCR switches off the catabolism of alternative nutrients and regulates a number of genes (9, 15, 51). The altered availability of these intermediates would make the bacterium more responsive to noxious stresses, including starvation and cell wall inhibition, by providing some kind of survival signal that activates or represses appropriate sets of genes.
The tolerance mutations occurring in such a global system must be subtle. A drastic inactivation of a regulatory pathway may impede cell growth altogether, while the wild-type situation does not confer sufficient tolerance. This is reminiscent of the fitness cost of antibiotic-resistant bacteria, which are less competitive than their wild-type counterpart bacteria in the absence of selective pressure by the antibiotic (14). Bacteria that lose the ability to efficiently import PTS sugars are severely hindered in their growth (29, 60). Indeed, when we inactivated the PTS in wild-type S. gordonii by deleting the ptsI gene, the strains presented a dramatic reduction in growth rate. To gain further insight into the mechanisms underlying the tolerance phenomenon, we are in the process of transferring the P483R mutation to other genetic bacterial backgrounds and screening tolerant clinical isolates for the presence of mutations in their PTS systems.
Because of the major impact that a single nucleotide tolerance mutation had on the antibiotic treatment outcome at a minimal cost to bacterial growth and because of the potential role of tolerance in the acquisition of further resistance mutations, enzyme I and the related sugar phosphotransferase system may represent new targets for antimicrobial intervention.
The present work was supported by grants 3235-62′698 (to A.B.) and 3100A0-105247 (to B.E.) from the Swiss National Science Foundation, a grant from the Société Académique Vaudoise (to A.B.), and a grant from the Roche Research Foundation (to I.A.).
We are grateful to Alexander Tomasz and Axel Hartke for helpful suggestions on the manuscript. We thank Vladimir Lazarevic for the gift of plasmid pBS714 and Stéphanie Rosset, Marlyse Giddey, Stéphane Piu, and Jacques Vouillamoz for outstanding technical assistance.
Published ahead of print on 26 October 2009.