|Home | About | Journals | Submit | Contact Us | Français|
Pathogen evolution and subsequent phenotypic heterogeneity during chronic infection are proposed to enhance Staphylococcus aureus survival during human infection. We tested this theory by genetically and phenotypically characterizing strains with mutations constructed in the mismatch repair (MMR) and oxidized guanine (GO) system, termed mutators, which exhibit increased spontaneous-mutation frequencies. Analysis of these mutators revealed not only strain-dependent increases in the spontaneous-mutation frequency but also shifts in mutational type and hot spots consistent with loss of GO or MMR functions. Although the GO and MMR systems are relied upon in some bacterial species to prevent reactive oxygen species-induced DNA damage, no deficit in hydrogen peroxide sensitivity was found when either of these DNA repair pathways was lost in S. aureus. To gain insight into the contribution of increased mutation supply to S. aureus pathoadaptation, we measured the rate of α-hemolysin and staphyloxanthin inactivation during serial passage. Detection of increased rates of α-hemolysin and staphyloxanthin inactivation in GO and MMR mutants suggests that these strains are capable of modifying virulence phenotypes implicated in mediating infection. Accelerated derivation of altered virulence phenotypes, combined with the absence of increased ROS sensitivity, highlights the potential of mutators to drive pathoadaptation in the host and serve as catalysts for persistent infections.
It is becoming increasingly recognized that pathogen evolution is dynamic and can dictate virulence, adaptation, and persistence within the host during infection. Unclear in this pathoadaptation process are the genetic determinants and host microenvironments that shape the genomic landscape upon which natural selection occurs. It is now appreciated that major changes in virulence potential in many pathogens can be enacted by a low number of mutations to the genome that alter expression or function of central virulence factors (1).
In the bacterial pathogen Staphylococcus aureus, this evolutionary theme is at the forefront of adaptation that enables disease persistence or drug resistance in the host (2). S. aureus is among the most frequently encountered pathogens in the clinic and community and is now causing more mortality in the United States than HIV (3). It exists as either a commensal on skin and mucous membranes of the anterior nares or a versatile pathogen capable of causing a wide spectrum of infections. To manifest such a wide array of diseases, S. aureus tightly expresses a large repertoire of virulence factors through a complex network of virulence regulators, with the most well known being the two-component quorum sensor accessory gene regulator (agr).
While the contribution of these virulence factors during acute infection is well described, the expression of factors that promote persistence during chronic infection is more obscure. For instance, although agr does play an important role in orchestrating the production of toxins and exoproteins that aid in invasion and systemic infections (4), for unknown reasons agr dysfunctional mutants are selected for during chronic infection (5). An “insurance hypothesis” suggesting that bacterial diversity promotes an increased capacity for withstanding external stress was proposed to explain the S. aureus phenotypic variants observed in chronic osteomyelitis and cystic fibrosis (CF) airway infection (5–7). Based on that hypothesis, one could envision that S. aureus mutators, defined as strains with increased mutation frequencies up to 100-fold greater than those of parental strains, might be capable of accelerating diversity when under stress by constitutively increasing the mutation supply rate.
Genetic analysis of mutators demonstrates that inactivation of at least two DNA repair pathways, (i) the mismatch repair (MMR) system and (ii) the oxidized guanine (GO) system, is responsible for this phenotype. In the majority of mutators analyzed from different bacterial species, mutation of MMR alleles mutS and mutL seems to be most frequent. In Escherichia coli, MutS recognizes the mismatch and provides a dock for MutL, which orchestrates mismatch excision through interaction with (i) single-stranded endonuclease MutH and (ii) DNA helicase UvrD. Cleavage of only the unmethylated daughter strand by MutH followed by unwinding of the DNA from the nick by UvrD provides an opportunity for excision of the mismatch by exonucleases ExoI and RecJ and resynthesis of the correct sequence by DNA polymerase III (PolIII) and a β sliding clamp (8). Yet prior analysis showed that S. aureus MMR lacks a mutH homolog and a role in homologous recombination (9). This divergence from the classical MMR mechanism in E. coli and the inability to detect loss-of-function mutations in MMR for several S. aureus mutators (9, 10) suggest that mutations to the GO system may also be important for induction of mutation in S. aureus.
The GO system is most thoroughly described from genetic analysis of E. coli. The GO system N-glycosylases MutM and MutY represent the first step of base excision repair responsible for correcting mutations from 7,8-dihydro-8-oxo-deoxyguanosine (8-oxo-dG) by cleaving the N-glycosidic bond between a base and deoxyribose (11). Since 8-oxo-dG is a promutagenic DNA lesion that displays increased affinity for dA, GC→TA transversions can result following another round of replication if the mutation is not repaired (12, 13). To prevent this event, MutM is responsible for excision of 8-oxo-dG, while MutY excises adenine when selectively paired with 8-oxo-dG (12, 13). MutT, the third member of the GO system, functions as an 8-oxo-dGTP Nudix (nucleoside diphosphate linked to an X moiety) phosphohydrolase that cleanses the nucleotide pools and prevents the incorporation of 8-oxo-dG into the nascent DNA (11). In Pseudomonas aeruginosa, detection of hypermutators with mutations in the GO system suggests that this DNA repair system is important in maintaining genomic integrity (14). Although detection of mutators with similar mutations in S. aureus is currently lacking, it is likely that an increase in the 8-oxo-dG level following exposure to reactive oxygen species (ROS), such as superoxide (O2−·), hydroxyl radical (·OH), and hydrogen peroxide (H2O2), may be capable of overburdening this DNA repair mechanism in strains with a competent GO system.
During infection, exposure to innate immune system-derived ROS may be among the most frequent events experienced by S. aureus. Upon entering the host, the bacteria encounter neutrophils and macrophages that exhibit a respiratory oxidative burst capable of damaging pathogen-associated DNA and protein. To counter this immune response, S. aureus elicits an oxidative stress response to prevent or treat the damage induced by ROS. Despite knowledge of this oxidative stress response, less is known about the DNA repair mechanisms S. aureus uses to cope with the ROS-induced DNA damage that occurs when prevention fails. Although ROS are exceptional in killing bacteria, failure to complete this objective may result in promutagenic events leading to pathogen adaptation. Under these circumstances, the inability of mutators to repair 8-oxo-dG-associated DNA lesions may provide a better chance of survival.
Previous studies showed that S. aureus mutators were detected from (i) the CF airway, (ii) endocarditis, and (iii) a sequenced collection of USA300 isolates (SRX007710) (2, 15–17). In nearly all cases, coexistence of antimicrobial resistance was detected among strains exhibiting a mutator phenotype, suggesting that this was an important pressure favoring selection of this phenotype. Unexplored, and perhaps less obvious than antimicrobial resistance in these mutators, was whether mutations enhancing fitness in response to the host microenvironment during an infection (i.e., nutrient starvation, hypoxia, immune response) were also genetically linked to the mutator phenotype. To begin addressing the knowledge gap that currently exists on the impact of increased mutation frequencies in S. aureus, we chose to elucidate the role of the GO and MMR systems in preventing ROS-induced promutagenic and cytotoxic DNA damage and the evolution in virulence potential that results when these DNA repair systems are missing in S. aureus.
Here, we constructed mutants in the GO and MMR systems and showed that inactivation of these DNA repair systems enhances mutation frequency and shifts the predominant mutation type in a manner consistent with similar GO and MMR mutations reported from other bacterial species. Furthermore, we provide evidence that GO system glycosylases MutM and MutY are involved in but not exclusively responsible for the removal of 8-oxo-dG-associated base pairs from duplex DNA. In support of the idea that 8-oxo-dG lesions are more promutagenic than cytotoxic, we found that there is no increased sensitivity to H2O2 when the GO or MMR systems are lacking in S. aureus. To determine the effect of mutation with respect to altering virulence in S. aureus, we examine the inactivation rate of α-hemolysin (αH) and staphyloxanthin production during serial passage in GO and MMR mutants. The accelerated alteration of these virulence phenotypes in S. aureus mutators suggests that this lifestyle may be beneficial for the production of persistent infection.
Bacterial strains utilized in this work are listed in Table S1 in the supplemental material. Bacteria were cultivated in Trypticase soy broth (TSB) or agar (TSA) (Becton, Dickinson, Sparks, MD) with shaking (250 rpm) at 30°C, 37°C, or 42°C when appropriate. For antibiotic selection, the following antibiotics and concentrations were employed for S. aureus and E. coli: chloramphenicol (Clm) at 10 μg/ml, ampicillin at 100 μg/ml, and kanamycin at 30 μg/ml.
All mutants were constructed by allelic replacement with an in-frame, unmarked deletion allele in the chromosome of S. aureus NCTC8325 (8325) and HG003. For ΔmutM and ΔmutY mutants, approximately 1-kb flanks were amplified using primers listed in Table S2 in the supplemental material and ligated together using a SacII site incorporated into the upstream forward and downstream reverse primer sequences. The deletion allele was then cloned into E. coli-S. aureus shuttle vector pKOR1 in place of the gateway cloning cassette using BP clonase II, as described previously (18). Upon confirmation of the proper insertion in pKOR1 by DNA sequencing, the vector was electroporated into restriction modification-negative, intermediate host strain S. aureus RN4220 (4220) (19). Mutation constructs were then transduced with ϕ11 or ϕ80α into S. aureus 8325 and HG003. Mutants were selected for by two passages at 42°C and one passage at 37°C and plating 100-μl 10-fold serial dilutions onto TSA containing 1 μg/ml anhydrotetracycline. Large colonies were picked and screened for plasmid loss by patching to TSA and TSA-Clm. Colonies with a Clm-sensitive phenotype were screened for the deletion by genomic DNA isolation and PCR using upstream forward and downstream reverse primers. All mutants were confirmed by DNA sequence analysis. When a mutation was constructed in a polycistronic operon, semiquantitative reverse transcription-PCR (RT-PCR) transcript analysis was performed on alleles 3′ to the mutated allele to rule out the possibility that the mutation produced a polar effect.
For ΔmutS, ΔmutL, and ΔmutS ΔmutL (ΔmutSL) mutants, the same procedure was followed as described above except that deletion alleles were cloned into the multiple-cloning site of pWedge (see Fig. S1 in the supplemental material), a modified pKOR1 vector. Instability of the gateway cloning cassette and subsequent false-positive colony results upon cloning pKOR1 into E. coli DH5α motivated us to modify this vector. We did so by amplifying the multiple-cloning site from pBluescript II KS (−) (Agilent Technologies/Stratagene, Santa Clara, CA) using primers listed in Table S2 in the supplemental material. The amplicon was digested with ApaI and Acc65i (New England BioLabs, Ipswich, MA) and cloned into pKOR1 that was digested with same enzymes. DNA sequencing using primers listed in Table S2 in the supplemental material (e.g., pWedge-MCS-F and -R) confirmed the sequence and orientation of the multiple-cloning site in pWedge.
Strains utilized for the E. coli ΔmutT complementation assay were constructed by introducing mutT homologs in trans from low-copy-number-vector pWSK29 (20). Homologs of mutT or mutT were amplified from 8325 or E. coli MG1655, respectively, using primers listed in Table S2 in the supplemental material. The amplicons were digested with NdeI and XhoI and cloned in-frame into the multiple-cloning site of pET16b (Novagen). Upon confirmation of the proper insertion in pET16b by DNA sequencing, all insertions, except for mutT, flanked 5′ by a ribosomal binding site and an N-terminal 10× His tag, were excised with XbaI and BamHI and cloned into pWSK29 digested with same two enzymes. Since a BamHI site existed that was internal to mutT, this insertion, flanked 5′ by a ribosomal binding site and an N-terminal 10× His Tag, was amplified using the primers listed in Table S2 in the supplemental material, digested with XbaI and ApaI, and cloned into pWSK29 cut with the same two enzymes. Following confirmation of all DNA insertions in pWSK29, all pWSK29 constructs were transformed into the E. coli ΔmutT strain.
Single colonies were selected and cultured at 37°C in 1 ml TSB (S. aureus) or Luria-Bertani (LB) broth (E. coli) for 16 to 18 h. After reaching stationary phase, cultures were diluted appropriately and spread onto TSA without or with rifampin (Rif) (100 μg/ml). Following 24 h of growth at 37°C, bacteria were enumerated and the mutation frequency was calculated by dividing the number of rifampin-resistant (Rifr) colonies by the total count of viable cells from the same culture. The median mutation frequency was calculated to avoid bias from “jackpot” cultures that exhibit a high Rifr colony count due to a resistance-conferring mutation during an early stage of growth. The Mann-Whitney test was employed to determine if the differences in mutation frequencies observed between parent and mutant strains were statistically significant.
Rifr mutants obtained following a PFA were picked into a 96-well microtiter plate (Costar 3595; Corning, Lowell, MA) containing 200 μl Tris-EDTA (TE) buffer (Fisher Scientific, Pittsburgh, PA) supplemented with 0.125 mg/ml lysostaphin (Sigma, St. Louis, MO) and 0.25 mg/ml RNase (Sigma, St. Louis, MO). Following cell lysis at 37°C for 30 min, genomic DNA was harvested using a 96-well genomic DNA purification kit (Qiagen, Valencia, CA) and quantified using a NanoDrop spectrophotometer (Thermo Scientific, Wilmington, DE). A 702-bp fragment in rpoB corresponding to the rifampin resistance-determining region (RRDR) was amplified from the purified genomic DNA using primers listed in Table S2 in the supplemental material. Amplifications were carried out in a 12.5-μl volume containing 0.2 μM (each) primer, 1× Accuprime buffer II (Life Technologies, Grand Island, NY), 10 to 20 ng of template DNA, and 1 U of Accuprime High Fidelity Taq polymerase (Life Technologies). Reaction mixtures were denatured at 95°C for 2 min and amplified for 30 cycles at 95°C for 30 s, 50°C for 30 s, and 68°C for 1 min, followed by one final extension at 72°C for 10 min. Detection of the desired amplicon size was performed by electrophoresis of 2 μl of the PCR mixture on a 1.0% agarose gel.
Amplicons were purified with shrimp alkaline phosphatase (U.S. Biological, Boston, MA) and exonuclease I. Sequencing reactions were performed with a BigDye Terminator sequencing kit (Applied Biosystems, Carlsbad, CA) per the recommendations of the manufacturers using M13 sequencing primers (see Table S2 in the supplemental material). DNA sequencing was performed on an ABI 3130XL sequencer according to the manufacturer's protocol. To detect mutations, DNA sequences were aligned to the corresponding region of rpoB sequenced from our −80°C stock of S. aureus 8325 using MacVector software (MacVector, Cary, NC).
Strains to be tested were struck from −80°C stock onto TSA and incubated overnight at 37°C. Single colonies were picked and cultured overnight to the stationary phase in 3 ml TSB with shaking (250 rpm) at 37°C. Overnight cultures were diluted 1/100 in 3 ml TSB and cultured for 3 h to the late exponential phase (optical density at 600 nm [OD600] = 0.8). The late-exponential-phase culture was diluted 1/10 in phosphate-buffered saline (PBS) containing the appropriate amount of H2O2 (Sigma) to reach the desired H2O2 concentration. H2O2 killing reactions were performed for 1 h at 25°C, and then the reaction mixtures were quenched by addition of Micrococcus lysodeikticus catalase (Sigma) (1,700 U/ml). Bacteria were serially diluted in PBS by log increments and appropriate dilutions plated on TSA and enumerated following incubation for 24 h at 37°C. Results are representative of a single experiment that was performed in at least triplicate.
For measurement of the H2O2-inducible mutation frequency, strains tested were exposed to H2O2 as described above, except that bacteria were diluted 1/10 in 3 ml TSB following H2O2 treatment and subcultured at 37°C with shaking (250 rpm) to the stationary phase (~16 h). Spontaneous-mutation frequency was then determined as described above.
The assay was performed as described by Gonzalez et al. (21). The oligonucleotides used in this study are shown in Table S2 in the supplemental material. DNA substrates for the MutM activity assay containing 8-oxo-dG were created by end-labeling oligonucleotides with [γ-32P]ATP (PerkinElmer, Waltham, MA) at the 5′ end using T4 polynucleotide kinase (Invitrogen, Carlsbad, CA) per the manufacturer's directions. Unincorporated nucleotides were removed from labeled DNA using a QIAquick nucleotide removal kit (Qiagen, Chatsworth, CA) and eluted in TE buffer. Labeled strands of 8-oxo-dG were annealed to Comp-A, -C, -G, or -T oligomers in a 1:2 molar ratio by heating to 90°C for 2 min and then cooling to room temperature over 3 h. DNA substrates for the MutY activity assay containing base A and/or Comp-A, -C, -G, or -T oligomers were created by end-labeling oligonucleotides with [γ-32P]ATP (PerkinElmer, Waltham, MA) at the 5′ end using T4 polynucleotide kinase (Invitrogen, Carlsbad, CA) per the manufacturer's directions and annealed to 8-oxo-dG oligomer in a 1:2 molar ratio by heating to 90°C for 2 min and then cooling to room temperature over 3 h.
Crude protein extracts were prepared from S. aureus cells grown overnight in 150 ml TSB, as described by Gonzalez et al. (21). Briefly, a cell pellet from an overnight TSB culture was weighed and was resuspended in an equal volume of cold 10 mM Tris HCl (pH 7)–50 mM NaCl–1 mM EDTA. The resuspended cells were added to vials containing 1 g of glass beads and bead beaten for 2 min twice at 4°C. The lysate was formed into pellets using an Eppendorf 5415D centrifuge (8,000 rpm, 5 min) at 4°C for supernatant collection. A second round of purification was performed on the supernatant fraction by spinning at 4°C on an ultracentrifuge (Beckman Optima XL-100K) (32,000 rpm, 1 h). The purified crude protein extract was then dialyzed twice overnight in 2 liters of prechilled 10 mM Tris-HCl–1 mM EDTA (pH 7.5) at 4°C. A Bradford protein assay (Pierce, Rockford, IL) was performed in triplicate according to the manufacturer's protocols to obtain the desired protein concentration. Purified protein extracts were supplemented with 0.1 mg/ml phenylmethanesulfonyl fluoride and frozen at −80°C in 50-μl aliquots.
Reactions were initiated by mixing 25 μg of crude extracts from the wild type (WT) (8325 or HG003) or the GO mutants (ΔmutM and ΔmutY mutants), DNA substrates (10 pmol), and assay buffer (200 mM Tris-HCl [pH 7.8], 100 mM EDTA, 2.5 M KCl) in a volume of 50 μl. Reaction mixtures were incubated at 37°C for 30 min before the addition of 10 μl of dye (98% [vol/vol] formamide, 10 mM EDTA, 0.025% bromophenol blue [wt/vol], and 0.025% [wt/vol] xylene cyanol). To prevent base reannealing with the DNA substrate following glycosylase excision, reaction mixtures were combined with 5 μl of 0.4 M NaOH and incubated at 90°C for 2 min. Samples were loaded onto a 20% denaturing polyacrylamide gel and subjected to electrophoresis for 2.5 h at 60 W. Images were analyzed by phosphorimaging in a Bio-Rad FX phosphorimager (Bio-Rad, Hercules, CA).
Experiments were performed according to a method used by Somerville et al. (22). Briefly, the bacterial strains tested were struck on TSA and cultured overnight at 37°C. A single colony for each strain was picked and subcultured at 37°C with shaking (250 rpm) to the stationary phase (~16 h) and diluted 1/200 into 3 ml TSB in triplicate. Bacteria were cultured to the late stationary phase (~20 to 24 h) and subcultured by diluting 1/200 into 3 ml TSB daily. Bacteria were serially diluted in PBS by log increments and plated on TSA supplemented with 5% washed rabbit erythrocytes (Hemostat Laboratories, Dixon, CA). These plates were incubated for either 16 h (8325) or 20 h (HG003), and total viable bacteria and those with an α-hemolysin-positive (αH+) phenotype were enumerated. A colony was defined as α-hemolysin negative when no hemolysis clearing could be discerned around the colony. Counting of viable and α-hemolysin-positive bacteria from the stationary phase was performed daily for either 1 week (8325) or 10 days (HG003).
To determine if mutations in agr had caused the α-hemolysin-negative phenotype, three α-hemolysin colonies were chosen (one for each replicate) for the 8325, ΔmutM, ΔmutY, ΔmutS, ΔmutL, and ΔmutSL strains after the last day of passage. Genomic DNA extraction was performed using a DNeasy blood and tissue kit (Qiagen, Valencia, CA) as described above (see the DNA sequence analysis of Rifr mutants), and a ~5-kb region encompassing the entire agr locus was amplified using primers listed in Table S2 in the supplemental material. Amplicon sequencing was performed as described above, and mutations were detected by aligning DNA sequences to the corresponding region of agr sequenced from our −80°C stock of S. aureus 8325.
Following colony enumeration for the α-hemolysin phenotype in HG003, bacteria were incubated for another 24 h at 37°C for enumeration of white colonies. White colonies were identified by eye as those with no detectable yellow coloring, a result of staphyloxanthin inactivation.
Exposure to ROS is a frequent event during host colonization and infection, and yet much remains unknown about how S. aureus responds to ROS-induced DNA damage. In many organisms, the GO system is a primary DNA repair system utilized for processing oxidative DNA lesions associated with 8-oxo-dG. To identify GO system alleles, the annotated genome sequence of strain 8325 was searched for terms (e.g., glycosylase, phosphohydrolase) associated with GO system function. Single genes for mutM (formamidopyrimidine glycosylase; saouhsc_1796) and mutY (A/G-specific glycosylase; saouhsc_2005) and five homologs of mutT (saouhsc_429, saouhsc_1593, saouhsc_1782, saouhsc_1913, and saouhsc_2791) were identified. Significant homology was observed throughout nearly the entire amino acid sequence when S. aureus MutM (33% identical; 51% similar) and MutY (44% identical; 64% similar) were compared by BLASTP to orthologs in Bacillus subtilis 168 (see Fig. S2a and S2b in the supplemental material). Although a full-length alignment was not produced from the MutT comparison (38% identical; 58% similar) (see Fig. S2c in the supplemental material), conservation of amino acids previously defined as essential for Nudix phosphohydrolase functionality was present in all of the MutT homologs of 8325 (see Fig. S2d in the supplemental material).
To determine if any of these 8325 MutT homologs possessed MutT functionality, we examined the ability of each homolog to reduce the heightened mutation frequency observed in E. coli ΔmutT. We reasoned that if MutT existed in S. aureus, it would be capable of complementing an E. coli ΔmutT because MutT is reported to function independently to cleave 8-oxo-dG from the intracellular nucleotide pools (11). In trans complementation of E. coli ΔmutT using a phenotypic fluctuation assay (PFA) with each of the five S. aureus mutT homologs demonstrated that only saouhsc_02791 was capable of reducing the median mutation frequency associated with loss of mutT function (Table 1). Despite inactivation of saouhsc_2791 by allelic exchange in 8325, no increase in mutation frequency was observed when this mutant was tested by PFA (data not shown).
To eliminate the possibility that any of the other mutT homologs were performing this function in S. aureus, we inactivated three (i.e., saouhsc_429, saouhsc_1593, and saouhsc_1913) of the other five mutT homologs in 8325. Again, no increase in mutation frequency was detected in any of these mutants by PFA (data not shown). Although, we cannot definitively conclude that the lone mutT homolog (i.e., saouhsc_1782) yet to be inactivated is providing mutT functionality to S. aureus, such data would not be consistent with the results of the E. coli ΔmutT complementation experiment (see above). Based on our current data, we suspect that the function of MutT is shared among the mutT homologs (see above) or may be encoded by another allele that is novel to S. aureus. Recent descriptions of multiple MutT-like alleles in E. coli (i.e., Orf135, Orf17, RibA) lend credence to the former possibility (23–25). Alternatively, S. aureus might invest more in the clearance of 8-oxo-dG from the DNA rather than the prevention of its incorporation. Such a scenario may provide compensation for the loss of MutT function, which would be unmasked only in the presence of ROS stress or when combined with mutations in alleles involved with the clearance of 8-oxo-dG-associated lesions (i.e., MutM or MutY). Our inability to identify a ΔmutT mutator phenotype in S. aureus led us to focus our remaining work in this report on the GO system-associated DNA glycosylases mutM and mutY.
Deletion mutants of mutM and mutY were constructed in strains 8325 (rsbU tcaR) and HG003 (isogenic to 8325 but rsbU+ and tcaR+) by allelic replacement and were tested for mutation frequency using PFA (Table 2). Each of the mutants constructed had no defect in growth rate compared to WT (data not shown). Statistically significant increases in median mutation frequency were observed for ΔmutM (3-fold) and ΔmutY (11-fold) strains (Table 2). A similar trend (i.e., mutY > mutM > WT) in mutation frequency was observed for GO mutants constructed in HG003 (Table 3).
We also chose to investigate the impact of the MMR system in the repair of oxidative DNA lesions in strain 8325 because MMR was previously reported to contribute to removal of 8-oxo-dG in E. coli (26). MMR alleles mutS and mutL were inactivated and analyzed in 8325 and HG003 as described for the GO system mutants. In 8325, deletion of MMR pathway alleles provoked an increase in the median mutation frequency of ΔmutS (18-fold), ΔmutL (24-fold), and ΔmutSL (13-fold; double mutant) strains (Table 2). A similar trend (i.e., mutL ≈ mutS ≈ mutSL > WT) in mutation frequency was observed for MMR mutants constructed in HG003 (Table 3).
The increases in mutation frequency displayed by S. aureus MMR strain 8325 and HG003 mutants were modest and more than a log lower than those seen with similar mutations constructed in E. coli (27). This difference may be a result of increased DNA replication fidelity in S. aureus, although no reports currently exist showing higher fidelity with Gram-positive replicative polymerase PolIIIC. Rather, it may suggest that an extra layer of postreplicative DNA repair exists as a backup to MMR in S. aureus, a role that may be fulfilled by other DNA repair systems. Support for this possibility is demonstrated by the 100-fold increase in mutation frequency for ΔmutL when in strain 4220 compared to 8325 (9). 4220 is a UV- and nitrosoguanidine-mutagenized derivative of 8325 and serves as an intermediate cloning strain harboring mutations in nucleotide excision repair (NER) alleles: uvrB (1,195-bp deletion) and uvrC (P331S) (19, 28). Based on these findings, it is interesting that an MMR and NER interaction might impact mutation frequency in S. aureus. Indeed, a physical interaction between MMR and NER proteins in Saccharomyces cerevisiae and E. coli supports the possibility of this notion (29). Unlike E. coli and B. subtilis, only uvrA and uvrB are known to be regulated by the SOS response in S. aureus (30), and this might underscore a possible mechanistic difference of S. aureus NER.
Although a similar trend (i.e., mutL ≈ mutS > mutSL > mutY > mutM > WT) in mutation frequency was observed for mutations constructed in strains 8325 and HG003, the mutation frequency was reduced 4-fold in HG003 compared to 8325 (Table 4). HG003 is a double-repair mutant of 8325 with restored functionality in (i) rsbU, a positive activator of stress-responsive sigma factor B (sigB), and (ii) tcaR, a cell wall-associated regulator and positive activator of protein A (spa) (31). We hypothesized that repair in one or both of these alleles caused the reduction in mutation frequency observed with HG003. To test this hypothesis, we determined the mutation frequency for 8325 and HG003 and for variants of 8325 with single repairs in either rsbU (HG001) or tcaR (HG002). Results from this experiment demonstrated that only the rsbU-repaired strain (HG001) exhibited the reduced mutation frequency (~3-fold) detected in HG003, which suggests that repair of rsbU is solely responsible for the reduced mutation frequency observed in HG003. In support of this finding, an increased mutation frequency was also detected in sigB mutants of S. aureus SH1000 (8325-derivative strain) isolated from prolonged growth in a biofilm (32). Analysis of the S. aureus sigB regulon yielded no obvious expressional changes in DNA repair candidates to explain the increased mutation frequency observed in sigB mutants. Rather, the reduced expression of O2−· scavenger sodM (superoxide dismutase) in sigB functional strains suggests that the increased endogenous production O2−· might be a potential cause of the increased mutation frequency in sigB mutants.
To indirectly investigate the role 8-oxo-dG may play in promoting mutations in GO and MMR mutants, we sequenced and quantified the single nucleotide polymorphism (SNP) profile observed in the rifampin resistance-determining region (RRDR) of rpoB for Rifr mutants generated in each of the GO and MMR mutants tested in the PFA (Table 1). Rifampin serves as an inhibitor of RNA polymerase, and most of the mutations generating rifampin resistance localize to its binding site in RRDR of rpoB (33). In the first experiment, 5 individual colonies of the parental and each mutant strain were picked from PFA plates, grown as independent cultures, and replated on 5 plates each (5 plates of the parental strain and 5 separate plates for each of the mutant strains). Ten colonies from each plate were picked and grown in liquid media for isolation of genomic DNA and sequencing of rpoB. In this way, a total of 48 colonies (8 to 10 colonies from each plate) of the parental and each mutant strain were sequenced (288 total rpoB sequences). To address the possibility that our results would be biased by sequencing sibling colonies from the same plate, we repeated the experiment in the following way. Fifteen individual colonies of the parental and each mutant strain were picked from PFA plates, grown as independent cultures, and replated on 15 plates (15 plates of the parental strain and 15 separate plates for each of the mutant strains). A single colony from each plate was picked and grown in liquid media for isolation of genomic DNA and sequencing of rpoB. In this way, a total of 15 colonies (single colonies from each plate) of the parental and each mutant strain were sequenced (90 rpoB total sequences). The results from both experiments were nearly identical, with mutations found in each colony and at the same nucleotide positions. The data presented below are from the first experiment.
As expected, GC→TA transversion was found more frequently in the ΔmutM (61%) and ΔmutY (79%) strains than in strain 8325 (18.4%) (Table 5), suggesting that these alleles are involved in the prevention of 8-oxo-dG-associated mutations (12, 13). These increases were offset by reductions in GC→AT transition, the dominant mutational type in 8325 (69.4%). To a lesser extent, GC→TA transversions were found in 8325, suggesting that 8-oxo-dG might still overwhelm the capability of the cell to cope with this promutagenic DNA substrate during in vitro aerobic growth.
Alternatively, we detected a sharp increase in AT→GC transitions for the ΔmutS (43.8%), ΔmutL (41.7%), and ΔmutSL (19.1%) strains compared with strain 8325 (2%) (Table 5). This spike in AT→GC transitions is similar to that reported for ΔmutS and putative MMR alleles ΔrecD2 and ΔyycJ mutants in the Gram-positive bacterium Bacillus anthracis (34, 35). Unlike ΔmutM and ΔmutY strains, GC→AT transversion remained the most common mutation in ΔmutS (50%), ΔmutL (56.3%), and ΔmutSL (66%) strains compared to strain 8325 (69.4%) (Table 5). The lack of GC→TA transversions in all MMR mutants argues against a primary role of MMR in removing 8-oxo-dG mutations from the double-stranded DNA (dsDNA) of S. aureus.
There are several explanations beyond that involving 8-oxo-dG to account for the Rifr rpoB mutational signatures observed in WT and MMR mutants. The most obvious culprit is DNA replication errors that escape the proofreading activity of DNA PolIIIC or are induced by error-prone translesion synthesis (TLS) polymerases (i.e., DinB, UmuD2C, Pol) encoded in S. aureus (discussed below).
Apart from DNA replication errors, the high frequency of GC→AT and AT→GC mutations in WT and MMR mutants, respectively, may be explained by DNA deamination following exposure to reactive nitrogen species (RNS), such as nitrous oxide (NO·) derivatives peroxynitrite (ONOO-) and nitrous anhydride (N2O3). RNS-induced deamination of cytosine to uracil, guanine to xanthine, or adenine to hypoxanthine promotes selective base pairing with adenine, thymine, or cytosine, respectively, during a further round of DNA replication. Although several DNA repair proteins exist to prevent these mutations from occurring, it is now known that the MMR system in B. subtilis contributes to the removal of mutations associated with base deamination (36). Although S. aureus endogenously produces NO· for protection from ROS (37), it is currently unknown whether NO· from this source or from phagocytes exerts a cost for maintenance of DNA integrity.
Sequence analysis of the rpoB RRDR identified mutational hot spots that were overrepresented in GO and MMR mutants. The penchant for GC→TA transversion in ΔmutM and ΔmutY strains was focused primarily at two loci of rpoB: nucleotides (nt) 1261 (468: N→K) and 1289 (477: A→D) (Fig. 1 and Fig. 2). Interestingly, the frequencies at which the sites were mutated were dissimilar, depending on the GO system glycosylase that was inactivated. For instance, the ΔmutM strain stood similar chances of a GC→TA transversion at both locations, while the ΔmutY strain seemed to prefer the GC→TA transversion at nt 1289. Although the reason for the difference between ΔmutM and ΔmutY strains in hot spot selectivity is unknown, it is known that the GC→TA transversion found at both sites provides high-level resistance to rifampin (MIC > 128 μg/ml). Interestingly, nt 1261 appears as a conserved hot spot of mutation in ΔmutY and ΔmutY ΔmutM mutants in B. anthracis (nt 1402) and, to a lesser extent, in E. coli (nt 1537) (Fig. 1 and Table 3) (35, 38).
Besides carrying a hot spot for a GC→AT transition mutation similar to that of strain 8325 at nt 1300 (481: H→Y), two other AT→GC transition mutation sites at nt 1262 (468: Q→R) and 1301 (481: H→L) were prevalent among the MMR mutants (ΔmutS, ΔmutL, and ΔmutSL strains). The appearance of an AT→GC transition signature only after inactivation of MMR suggests that the MMR system is responsible for removal of this particular DNA replication error in S. aureus. Again, these results revealed a strong similarity to rpoB transversion hot spots found in ΔmutS mutants for B. anthracis (nt 1403 and 1422) and, to a lesser extent, for E. coli (nt 1538 only) (Fig. 1 and Table 3). Interestingly, at both AT→GC transition sites, the template T is followed by G (5′-TG-3′), which is at odds with the preference for G preceding the template T (5′-GT-3′) found in E. coli MMR mutants (39). However, this difference may reflect the use of the lacI reversion system rather than the rpoB reporter system employed in this study.
Given that replication errors are likely to contribute to the AT→GC or GC→AT transitions observed in MMR mutants, defining the polymerases that contribute to those errors is also of value. Analysis of replication errors in E. coli indicates that SOS response-inducible TLS polymerase dinB does not contribute significantly to these replication errors (40). S. aureus also encodes dinB and polV (umuD) homologs, which are induced by the SOS response (30). It is currently unclear what role TLS and replicative polymerase polIIIC play in the incorporation of oxidized DNA precursors and error-prone bypass of template strand oxidative lesions, such as 8-oxo-dG, in S. aureus and in Gram-positive bacteria in general. It is likely that both replicative and TLS polymerases contribute to 8-oxo-dG incorporation, given that both were shown to perform this function in E. coli during antibiotic stress (41).
Glycosylase activity of mutM and mutY was confirmed by isolating whole-cell protein extracts from ΔmutM and ΔmutY strains in strain HG003 and assaying for cleavage of 8-oxo-dG-paired mismatched DNA substrates. A loss or reduction in cleavage activity with 8-oxo-dG/C, 8-oxo-dG/G, and 8-oxo-dG/T substrates was observed in the ΔmutM strain (Fig. 3A). Interestingly, MutM demonstrates more preferential activity for cleavage of 8-oxo-dG/G substrates than for 8-oxo-dG/C or 8-oxo-dG/T. Similar results were also observed for whole-protein extracts from the ΔmutM strain in 8325 (data not shown). Weak cleavage activity in 8-oxo-dG paired with A (8-oxo-dG/A) is consistent with the idea that removal would be promutagenic when 8-oxo-dG exists in the template strand (i.e., 8-oxo-dG/A→T/A). Similar data obtained from MutM/Fpg glycosylase analysis in Mycobacterium smegmatis (42), E. coli (43), and Lactococcus lactis (44) corroborate this viewpoint. In contrast to these other organisms, residual cleavage of 8-oxodG/C and 8-oxodG/T in the presence of ΔmutM protein lysates indicates that other proteins are also involved in excision of 8-oxo-dG from these DNA lesions in S. aureus. Specifically, endonucleases III and VIII from E. coli (nth, nei) and humans (hNTH1, hNEIL1) exhibit the ability to remove 8-oxo-dG from double-stranded DNA templates (45, 46). The existence of two nth homologs and induction of one of these nth homologs during H2O2 stress suggest that these genes may also play a critical role in the removal of 8-oxo-dG from S. aureus (47).
Alternatively, cleavage of 8-oxo-dG/A mispairs from the DNA template could be attributed entirely to mutY (Fig. 3B). Protein extracts from the ΔmutY strain demonstrated specific activity for 8-oxo-dG/A pairing only (data not shown), suggesting that MutY repair is exclusive to 8-oxo-dG/A lesion and does not repair lesions repaired by MutM (i.e., 8-oxo-dG/C, 8-oxo-dG/G, or 8-oxo-dG/T).
Previous work has shown that S. aureus is susceptible to killing by ROS in the form of H2O2 at relatively low concentrations (~7.5 to 10 mM) (47–51). Based on our data demonstrating that mutM and mutY are partially responsible for preventing 8-oxo-dG-associated mutation in S. aureus (Fig. 3), we reasoned that increasing the overall burden of 8-oxo-dG upon exposure to ROS may be cytotoxic and thus enhance cell sensitivity. Accordingly, we hypothesized that ΔmutS, ΔmutL, and ΔmutSL mutants would lack enhanced sensitivity to ROS because of the absence of an 8-oxo-dG-associated mutational signature (i.e., GC→TA transversion) in rpoB (Table 5). Additionally, loss of the GO or MMR function might also be cytotoxic if they were to contribute to repair of ROS-induced DNA strand breaks or blocking lesions. To measure sensitivity to ROS, we performed H2O2 sensitivity assays with ΔmutM, ΔmutY, ΔmutS, ΔmutL, and ΔmutSL strain HG003 mutants. H2O2 was chosen because of its well-known ability to release DNA-reactive ·OH following oxidation of ferrous iron (Fe2+) in the Fenton reaction. Bacteria (1 × 108 to 2 × 108 CFU/ml) were treated for 1 h with 5 and 10 mM H2O2 and then plated for survivors. To our surprise, exposure to H2O2 at each concentration tested failed to demonstrate a significant potentiation in the sensitivity of ΔmutM and ΔmutY mutants (Fig. 4). As we expected, ΔmutS, ΔmutL, and ΔmutSL mutants exhibited little difference in H2O2 sensitivity at either concentration tested. These observations were confirmed by H2O2 disk diffusion assays (data not shown). These results corroborate previous data from Chang et al. (47) which demonstrated that exposure of S. aureus to 10 mM H2O2 for 10 and 20 min did not result in transcriptional changes of the GO or MMR genes or sigB. However, in their experiments, increased transcription (including that of uvrA, uvrB, lexA, and recA) was observed in DNA repair proteins, suggesting that these DNA repair and protection mechanisms bypass the function of GO and MMR in DNA repair and enable S. aureus to resume growing after exposure to H2O2 while DNA damage is being repaired.
Such a result may also suggest that 8-oxo-dG-associated mutations produced in ΔmutM and ΔmutY upon exposure to ROS are more likely to be promutagenic rather than cytotoxic. To test for this possibility, we measured the frequencies of ΔmutM, ΔmutY, ΔmutS, ΔmutL, and ΔmutSL mutations in strain 8325 following exposure to low (10 mM), high (50 mM), and very high (250 mM) doses of H2O2 using the same methodology described above for H2O2 sensitivity testing and phenotypic fluctuation assays. Again, we did not observe any substantial increase in mutation frequency following exposure to any of the H2O2 concentrations tested (see Table S3 in the supplemental material). In fact, exposure to very high concentrations of H2O2 slightly reduced the fold difference in mutation frequency observed with MMR mutants but produced a ca. 2-fold increase in 8325. This result, combined with the comparable levels of H2O2 killing across all strains tested, suggests that neither the GO nor MMR system is required for limiting mutations following acute H2O2 exposure. To verify this possibility, we examined whether H2O2 treatment imposed any difference on the Rifr mutational profile by sequencing the RRDR of rpoB for Rifr mutants pretreated with 50 mM H2O2. Similar rpoB mutational profiles were observed for Rifr mutants in the presence and absence of 50 mM H2O2 pretreatment (data not shown). This result implies that bacterial survivors from the H2O2 sensitivity assay experience few ·OH-induced mutations within at least rpoB RRDR or that the mutations observed following H2O2 treatment represent the same mutations as those experienced during aerobic growth.
The results of the ROS sensitivity analysis described above suggest that GO and MMR mutants are adept at protecting the DNA from Fenton reaction-induced ·OH radicals. This result may be attributable to the lack of cytotoxic DNA single-stranded breaks or 3′ blocking lesions created when MutM independently or MutY in concert with Nth, Nfo, and XthA removes 8-oxo-dG or -A, respectively, from the DNA upon exposure to H2O2. Support for this possibility is provided by the increased survival of exonuclease and endonuclease mutants of N. meningitidis or Salmonella enterica serovar Typhimurium under conditions of ROS or RNS stress, respectively, following inactivation of mutM (52, 53). Alternatively, the lack of ROS-induced cytotoxic DNA lesions could also be due to effective cleansing of increased 8-oxo-dG from the nucleotide pools by one or multiple MutTs.
However, protection resulting from the oxidative stress response might take place before ROS have a chance to interact with nucleotide pools or DNA in GO and MMR mutants. In S. aureus, the oxidative stress response is repressed by perR and becomes derepressed upon exposure to ROS, allowing the expression of several proteins involved in oxidative stress resistance (i.e., KatA, AhpC, and Fur) (49). Among all the ROS stress defense proteins expressed during this response, MrgA, a ferritin-like Dps homolog, may be especially important. Derepression of mrgA transcription by PerR in response to H2O2 exposure was shown to induce nucleoid compaction and increase survival, presumably because Fe2+ scavenging near the DNA by MrgA prevented interaction with H2O2 and release of DNA-damaging ·OH radicals (54).
Subversion of the oxidative stress response might occur more readily during infection, as fewer cells are exposed to high ROS concentrations of professional phagocytes following an oxidative burst. It is important that this continuous and chronic H2O2 dosing of a few cells was at odds with the acute H2O2 exposure of a large number of cells employed in our experiments and might be another reason why we failed to detect any differences in killing or induced mutagenesis in our GO or MMR mutants. Based on the shortcomings of this acute H2O2 exposure, it is possible that ROS killing kinetics or induced mutation frequencies might differ in GO and MMR mutants in cell culture or animal infection assays.
Environmental stresses such as those encountered during infection or as a result of treatment with antibiotics may also account for elevated levels of ROS or inactivation of the GO and/or MMR pathways that we have not addressed in our experiments. As one example, three classes of antibiotics (β-lactams, quinolones, and aminoglycosides) have been shown to potentiate cell death as a result of elevated levels of reactive ·OH (41, 55) which oxidize the guanine nucleotide pool, resulting in an increase in 7,8-dihydro-8-oxo-deoxyguanosine (8-oxo-dG). The GO system N-glycosylases MutM and MutY participate in the first step for correcting the 9-oxo-dG mutation by a base excision repair process that produces single-stranded breaks in the DNA. Accumulation of these single-stranded breaks in the MMR and GO genes upon exposure to excessive levels of oxidative stress may inhibit transcription of these genes by RNA polymerase. The possibility of this scenario is supported by work from Foti et al. (41) in which they showed that the production of single-stranded breaks within close proximity in both DNA strands following GO repair effectively led to the death of E. coli bacteria due to oxidative stress induced by exposure to bactericidal antibiotics. Similar experiments with S. aureus are required to confirm a similar role for single-stranded nucleotide breaks.
To gain insight on the evolutionary impact of a mutator phenotype on S. aureus virulence modulation, we monitored for the loss of golden pigment (staphyloxanthin) in HG003 colonies during in vitro passage. Daily passage and enumeration of white colonies for 5 days in strain HG003 and GO and MMR mutants demonstrated significant increases in the number of white colonies for the ΔmutS, ΔmutL, and ΔmutSL mutants by day 5 of passage (Fig. 5B). This intense change in colony color from gold to white for the ΔmutS, ΔmutL, and ΔmutSL mutants, shown in Fig. 5A, likely represents a dispensable phenotype for the in vitro environment employed in this experiment. During an infection, staphyloxanthin functions as a shield from ROS released by host immune cells (56). Subsequently, staphyloxanthin production significantly increases survival in a mouse skin infection and sepsis model but not in NADPH oxidase-deficient chronic granulomatous disease mice.
To determine whether this lost virulence factor production was specific to staphyloxanthin or a more general phenomenon, we studied the rate of α-hemolysin inactivation during in vitro passage of WT strains (8325 and HG003) and GO and MMR mutants. We monitored the rate of α-hemolysin inactivation because (i) rabbit erythrocyte lysis is specific for α-hemolysin and (ii) the α-hemolysin phenotype inactivates during in vitro passage (22).
Daily passage of strain 8325 and enumeration on rabbit blood agar was performed for a week (Fig. 5D). Statistically significant increases in the α-hemolysin-negative (αH−) phenotype were detected for ΔmutY, ΔmutS, ΔmutL, and ΔmutSL mutants at days 2 (ΔmutS only), 3, and 4 of passage. High percentages of αH− colonies were detected in 8325 from day 5 to day 7 (Fig. 5C). The total viable count for each strain at each day of passage is shown in Fig. 6. In each strain tested, once α-hemolysin inactivation in any strain was detected, the α-hemolysin inactivation rate was linear before a plateau of near 100% of the total viable count was reached. This result suggests that the αH− phenotype is selected for in TSB cultures regardless of the DNA repair ability of the strain; however, the absence of an effective DNA repair system increased the speed at which this phenotype became visible. We also need to consider that, based on the growth conditions utilized for this experiment, there may be selection for inactivation of agr. The rapid burst of metabolic activity induced by agr results in an intense metabolic burden, a consequence of which is increased concentrations of ROS and RNS, which are likely detrimental and responsible for frequent spontaneous agr mutations and inactivation in the laboratory (4, 22, 57, 58).
In strain HG003, the result was more complex (Fig. 5D and andE).E). The increased expression of sigB in HG003, due to repair of positive activator rsbU, dampened the α-hemolysin phenotype, as reported previously (31), reducing the halo diameter surrounding α-hemolysin-positive colonies. The importance of this phenotype became remarkable, as αH− and hyper-α-hemolysin-positive (αH++) colonies could be detected during passage. The HG003 experiment was performed as with 8325, except passage was carried out for 10 days because α-hemolysin-positive colonies could still be detected among the HG003 bacteria at day 7. Only the ΔmutS (day 2) and ΔmutSL (days 2 and 3) mutants demonstrated a significantly higher proportion of αH− colonies than HG003. Surprisingly, the ΔmutS (day 10), ΔmutL (day 10), and ΔmutM (day 2 to 8) mutants exhibited a significant reduction in αH− colonies compared to HG003. This was most striking in the ΔmutM strain, as the rate of αH inactivation was significantly reduced. The reason for this result in ΔmutM of HG003 but not 8325 is unknown, but the data may suggest that DNA repair is somehow modulated when sigB is expressed. In contrast to the ΔmutM strain, the reduction in the αH− phenotype exhibited by the ΔmutS and ΔmutL mutants at day 10 was a consequence of the increased prevalence of the αH++ phenotype (Fig. 5D). An example of the αH++ phenotype is seen in Fig. 5A.
This αH++ phenotype may be explained by mutations that enhance the expression of positive regulators (i.e., agr, saeRS, arlRS, mgrA, and sarZ) or reduce or inactivate the expression of repressors (i.e., rsbU, sigB, sarA, sarT, and rot) of α-hemolysin expression or of the hla promoter itself (32). The shift in the α-hemolysin phenotype observed over time in the ΔmutS and ΔmutL mutants is fascinating and demonstrates the potential for generating a diverse set of α-hemolysin-expressing clones that could be selected from the many divergent microenvironments that exist within a host. It is unclear whether the elevated mutation frequency or the shift in mutation type (i.e., AT→GC) observed with ΔmutS and ΔmutL mutants promoted this diversity in the α-hemolysin phenotypes. However, based on the lower detection of AT→GC transitions observed in the ΔmutSL mutant (19%) than in the ΔmutS (44%) or ΔmutL (42%) mutant and the fact that S. aureus exhibits an AT-rich genome, we suspect that the shift in mutation type may be most important for the production of αH++ phenotype in these mutant populations.
Alternatively, the accelerated loss of α-hemolysin and staphyloxanthin synthesis (see above) during in vitro passage in ΔmutY (α-hemolysin only), ΔmutS, ΔmutL, and ΔmutSL mutator strains suggests that these strains may have the potential to generate reduced virulence phenotypes during infection. This prediction is supported by the reduced fitness of the ΔmutL strain under conditions of coinoculation with its parent 4220 strain in a rat osteomyelitis infection assay (59). Besides the caution that should be exercised with the use of 4220 in infection assays (described above), a limitation of this osteomyelitis competition study is the lack of sustained antibiotic treatment during the course of infection, which appears to be important for reducing the cost of deleterious mutations associated with P. aeruginosa hypermutators during murine airway infection (60).
Considering that chronic treatment with the glycopeptide antibiotic vancomycin seems to be effective in selecting for MMR mutators and reduced-virulence vancomycin-intermediate S. aureus (VISA) isolates (2, 16, 17), it is worth speculating that mutators may hasten this reduction in virulence during chronic vancomycin therapy. Although the idea remains to be proven, it is suspected that the reduced virulence found in VISA isolates may mask immune detection (61). The potential for vancomycin to drive resistance evolution certainly seems conceivable based on prior reports showing that killing by bactericidal antibiotics is reliant on production of ROS (55). Although we failed to show that exposure to H2O2 can heighten mutation frequency in GO or MMR mutants (see Table S3 in the supplemental material), it remains possible that continued low-dose exposure to ROS due to vancomycin therapy might preferentially enhance resistance in mutators.
However, this is not to say that vancomycin is the only bactericidal drug capable of producing this antimicrobial resistance/immune system evasion phenomenon. For instance, frequent exposure to inhibitors of protein translation (i.e., gentamicin, linezolid) or folate synthesis (i.e., trimethoprim-sulfamethoxazole) selects for small-colony variants (SCVs) which carry mutations in alleles encoding targets for those antimicrobials (10, 62). The reduced metabolism displayed by SCVs causes a virulence factor alteration (i.e., increased adhesins, decreased toxins/proteases) that promotes intracellular survival in nonprofessional phagocytes (63). Interestingly, linkage of the mutator phenotype with thymidine-dependent SCVs suggests that a consequence of mutators may be the increased production of SCVs during infection. In support of this possibility, in vitro exposure of mutators to gentamicin was shown to increase the production of SCVs (64). It is also worth noting here that the intracellular environment of nonprofessional phagocytes or exposure to, e.g., pyocyanin or alkyl-hydroxyquinoline N-oxide secreted by P. aeruginosa is also capable of generating SCVs (65, 66). Based on the myriad of environment factors that may induce the production of SCVs, it is interesting that mutators may potentiate this known immune evasion mechanism to promote relapsing or persistent infection.
To determine the genetic cause of the αH− phenotype, we performed DNA sequence analysis of agr. Our search started here because agr (i) positively regulates α-hemolysin and (ii) was previously shown to be a hot spot for mutations that inactivate of α-hemolysin expression (22, 31). Single αH− colonies were picked from three replicates of 8325, GO, and MMR mutants following 7 days of passage and assessed for mutations in agr. Consistent with prior reports (5), all agr mutations detected were found in either agrC (histidine kinase) or agrA (response regulator) (Table 6). As expected, the mutation types found in agr were primarily substitutions that mimicked those found from sequence analysis of rpoB (Table 5), except for strain 8325, where two of the replicates encoded GC→TA transversions. In each case, these nucleotide substitutions led to either truncated proteins or amino changes that were not tolerated by the SIFT (Sorting Intolerant from Tolerant) algorithm (67). Interestingly, in one replicate from ΔmutM and ΔmutY strains, no mutation in agr could be detected, suggesting that in GO mutants, mutations to either hla or another regulator(s) of agr or hla might have led to this phenotype.
Although agr dysfunctional mutations are commonly reported from infections or serial passages (5), what benefit they provide to fitness is unclear. Some of the known consequences of agr dysfunction for virulence include (i) increased expression of staphylococcal superantigen-like secreted proteins (68), (ii) reduced colony spreading (69), (iii) increased potential for mammalian cell colonization resulting from increasing expression of the fibronectin binding protein A and B adhesins (70), and (iv) reduced susceptibility to thrombin-induced platelet microbicidal protein (71). Considering the adaptation to nutrient limitation provided by lasR mutants during the evolution of P. aeruginosa (72), it is conceivable that a reduced metabolic burden due to accelerated loss of agr function might provide a growth advantage for S. aureus mutators in TSB. However, the complexity of the infection environment is likely to select for phenotypes associated with agr dysfunction that are distinct from faster growth. As evidence of this possibility, it has been demonstrated that agr dysfunction reduces proinflammatory cytokine production from human endothelial cells, providing a condition that may favor enhanced intracellular survival (73).
We have shown here that inactivation of the MMR and GO systems in S. aureus produces a mutator phenotype, modifies the mutation spectrum, and accelerates the production of a heterogeneous bacterial population that may significantly alter the course of infection. Furthermore, we found that mutM and mutY are partially responsible for the clearance of promutagenic 8-oxo-dG-associated mutations in the DNA. These findings, combined with lack of sensitivity to ROS, suggest that S. aureus mutators may have a pathoadaptive value, leading to persistent infections. We also show that increased expression of sigB, as evidenced by strain HG003, reduces the mutation frequency and subsequently modifies the evolution of virulence factor production in some mutators. Whether the sum of the genomic changes produced by an increased spontaneous-mutation frequency overcomes the cost of deleterious mutations and favors the production of persistent infections remains to be proven. However, it is clear that in several cases (i.e., somatic hypermutation in B and T cells, tumor cell resistance to chemotherapy, HIV resistance to antiviral therapy, genetic drift in influenza virus), increased production of random mutations can be a powerful force positively shaping the evolution of organisms in general. It is expected that a more thorough and detailed understanding of S. aureus evolution will shed light on further opportunities for targeted therapeutic intervention during human infection.
This work was supported by funding from NIH R01 AI059111 and The J.R. Oishei Foundation (S.R.G.).
We thank Friedrich Götz (University of Tübingen, Tübingen, Germany) for kindly providing S. aureus strains, Taeok Bae (University of Indiana, Gary, Indiana) for kindly providing deletion vector pKOR1, the E. coli Genetic Stock Center at Yale for kindly providing E. coli BW25113 and the ΔmutT mutant, and Robert Quivey and Kaisha Gonzalez (University of Rochester Medical Center, Center for Oral Biology) for valuable advice and technical assistance with the DNA glycosylase assays.
Published ahead of print 30 November 2012
Supplemental material for this article may be found at http://dx.doi.org/10.1128/JB.00733-12.