Increased Expression of Efflux Pump norA Drives the Rapid Evolutionary Trajectory from Tolerance to Resistance against Ciprofloxacin in Staphylococcus aureus

ABSTRACT

The intensively intermittent use of antibiotics promotes the rapid evolution of tolerance, which may lead to resistance acquisition in the following evolutionary trajectory. In addition to directly exporting antibiotics as an instant resistance strategy, efflux pumps are overexpressed in tolerant strains. To investigate how efflux pumps participate in resistance development from tolerance to resistance, we performed in vitro evolutional experiments against the antibiotic ciprofloxacin in norA efflux pump mutants of Staphylococcus aureus. These experiments demonstrated that overexpression of norA rapidly facilitated the development of ciprofloxacin resistance from tolerance to resistance through elevated spontaneous mutations. The generated resistance mutations were further fixed in the population by increasing survival ability. The observed Ser80Phe and Glu84Lys mutations in the topoisomerase IV ParC (GrlA in S. aureus) may be responsible for tolerant strains to develop resistance to ciprofloxacin since it has been reported that such mutations disrupt the water-metal ion bridge between quinolones and ParC. MepA and Sav1866 are related to the same antibiotic (ciprofloxacin) susceptibility as NorA, and they also contributed to resistance development against ciprofloxacin. MgrA positively regulated NorA expression and the development of ciprofloxacin resistance. Importantly, blocking the evolutionary pathway by coadministering ciprofloxacin with the efflux pump inhibitor reserpine effectively delayed the resistance acquisition in an in vitro experiment. This study illustrated the role of efflux pumps in the evolutionary trajectory from tolerance to resistance. The delayed resistance development caused by the efflux pump inhibitor illuminates a possible strategy for postponing the resistance acquisition from tolerance to resistance by disrupting efflux pumps.

INTRODUCTION

Antimicrobial resistance has increasingly grown as a globally significant problem concerning every country, irrespective of income levels, health care system, and economic output. If the problem continues, around 10 million people are predicted to die annually due to a lack of effective antibiotics to treat bacterial infection by 2050, while 2% to 3.5% of the gross domestic product, up to $100 trillion, would be reduced (1). Among the most resistant bacteria, Staphylococcus aureus, an important opportunistic pathogen in humans, has caused high mortality in patients due to metastatic or complicated infections such as endocarditis, sepsis, and pyemia (2). The increasing emergence of S. aureus-resistant strains (mainly methicillin-resistant S. aureus [MRSA]) has hindered clinical treatment since the first case was published in 1961 (3). For example, MRSA accounts for up to 50% of S. aureus bloodstream infections in parts of Asia (4). Thus, solving this looming crisis requires the development of novel antimicrobials, local control practices, and innovative approaches for predicting and preventing the development of antibiotic resistance.

Well-identified resistance mechanisms include changing antibiotic targets by mutations, modifying (and protecting) targets, directly modifying antibiotics, and preventing access to targets both by reduced permeability and increased efflux (5). These resistance strategies assist bacteria to avoid antibiotic attacks by decreasing drug concentrations or increasing MICs (6). However, when bacteria are exposed to antibiotics at high levels of concentrations above the MIC, the decrease in the drug’s concentration does not help bacteria much. In this case, tolerance, which increases resilience to antibiotic treatment duration, enables microorganisms to survive transient exposure to high concentrations of antibiotics in a nonspecific way (7). For example, tolerance caused by dormancy protects bacteria from the lethality of many types of antibiotics whose action requires growth, such as quinolone (8). Unlike the awareness of increasing resistance, tolerance, despite being responsible for substantial morbidity and mortality, is commonly neglected due to a lack of standard clinical tests (9). Consequently, tolerance has become one of the major causes of persistent infections where a residual fraction of bacteria can resume growth after treatment is stopped, leading to infection relapse. Recent studies have shown that antibiotic tolerance facilitates the evolution of drug resistance both under laboratory conditions and in chronic infections (10–12). For example, when intermittently exposed to the high concentrations of antibiotics used in the clinic, the bacterium Escherichia coli first adaptively develops tolerance (13), with tolerance ending up evolving to resistance (10). However, we know little about the evolutional trajectory from the development of tolerance to resistance.

Moreillon and Tomasz have proposed that resistant mutants can emerge from tolerant cells. Recent experimental efforts support this hypothesis in various kinds of bacteria (14). During the long-term rifampin exposure in Mycobacterium tuberculosis, Sebastian et al. demonstrated that tolerance was a source of de novo resistant mutants (15). Another study found that the increased tolerance preceded the emergence and spread of ampicillin resistance, conferring mutations in E. coli populations (10). Nguyen et al. used a murine infection model to demonstrate that strongly decreased antibiotic tolerance abolished resistance development (16). All of these findings suggest a widespread link between tolerance and the development of resistance. However, how tolerance facilitates the development of resistance is still not clear.

Efflux pumps can transport a wide range of structurally dissimilar substrates, and they can be categorized into six families depending on structural characteristics, mechanisms of action, and energy sources required by transport (17). Efflux pumps are wildly characterized as a resistance mechanism (5). For example, the AcrAB pump, a member of the resistance-nodulation-cell division family, conferred resistance by transient overexpression in Gram-negative E. coli (18). The expression of several multidrug efflux genes was higher in tolerant cells, implying the potential role of efflux pumps in tolerance (8), but little is known about efflux pumps’ roles in resistance development, from tolerance to resistance. The elevated expression of acrAB promoted spontaneous mutations within single cells in E. coli, suggesting a positive association between resistance mutations and efflux pump overexpression (19). Additionally, a recent study demonstrated that elevated norA (major facilitator superfamily) expression increased the growth rate of ciprofloxacin-resistant mutants in S. aureus (20). However, how the efflux pumps participate in the resistance development from tolerance remains unclear.

By taking advantage of the availability of large numbers of tolerant strains in our laboratory, we determined whether efflux pumps in S. aureus would promote resistance development from tolerance. These tolerant S. aureus strains were previously obtained by intermittent exposure to different antibiotics in our laboratories. A ciprofloxacin (quinolone)-tolerant isolate was chosen for this study, as it was commonly used to treat S. aureus infections, and its mechanisms were already well characterized (21). This study aimed to determine how efflux pumps were involved in the resistance development of ciprofloxacin from tolerance to resistance in S. aureus.

RESULTS

Increasing expression of efflux pumps across tolerant strains.

To explore the role of efflux pumps in antibiotic tolerance, we first measured expressional levels of six selected efflux pump genes in nine tolerant strains of S. aureus by real-time PCR. These tolerant strains were previously acquired by intermittently exposing (3, 5, or 8 h) wild-type Newman to TSB medium supplemented with 20× MICs of seven different antibiotics (Fig. 1A). Additionally, mutations responsible for tolerance phenotype have been genetically and experimentally identified (see Table S1 in the supplemental material). Genes for efflux pumps in S. aureus are located in either the chromosome or plasmids. Here, considering plasmids’ mobility and heterogeneity, we exclusively focused on chromosomally encoded efflux pumps. We chose one or two representative efflux pumps from each family to measure their expressional levels for consideration of the broad scope of efflux pumps in terms of families and substrates (Fig. 2A). For example, only NorA was chosen in the MFS family since other similar pumps (NorB and NorC) in the family (MFS) have the same substrate, ciprofloxacin. Generally, nine tolerant strains expressed significantly higher levels of efflux pump genes than the wild-type strain Newman (Fig. 2B). Since the mutations were responsible for tolerance generation in the tolerance strains and none of these mutations were located in the genes for the efflux pumps, the variations in levels of pump expression were the results of tolerance phenotype. The tolerant strains which evolved from vancomycin (glycopeptide) and cefazolin (β-lactams) selection, except for VAN.C10.T5, showed relatively lower expression levels of efflux pumps. In contrast, those which evolved from ciprofloxacin (quinolones), oxacillin (β-lactams), flucloxacillin (β-lactams), meropenem (β-lactams), and imipenem (β-lactam) selection had the highest efflux pump expression (Fig. 2B). Among the six selected efflux pump genes, expressional levels remarkably varied across different strains, with the highest expressional levels of lmrS, mepA, and sav1866 overall, suggesting that these efflux pumps might play a more general role in resistance evolution across S. aureus (Fig. 2B).

FIG 1.

Tolerance facilitated the resistance development against ciprofloxacin in S. aureus. (A) Previous scheme for tolerant strains. (B) Experimental scheme for intermittent antibiotic challenge in this study. Tolerant strain CIP.C17.T8 and the wild-type Newman strain were used. (C) Tolerant phenotype of parent strain CIP.C17.T8 was consistently challenged with the lethal concentration of antibiotics for 36 h during which survival population was counted by plating on solid plates every 12 h. (D) MIC of evolved lines relative to the ancestral MIC of the first cycle. The change of relative MIC value for strain CIP.C17.T8 is with evolved Newman lines relative to its initial MIC as control.

FIG 2.

Expression levels of efflux pumps in different tolerant strains. (A) Families of multidrug efflux pumps in S. aureus. The ATP-binding cassette (ABC) family includes efflux pumps AbcA and Sav1866, the multidrug and toxin extrusion (MATE) family includes efflux pump MepA, the major facilitator superfamily (MFS) includes efflux pumps NorA and LmrS, and efflux pump SepA has not been characterized to any family. AbcA is responsible for oxacillin extrusion, and NorA, Sav1866, and MepA are responsible for ciprofloxacin extrusion. Ciprofloxacin targets two essential enzymes, DNA gyrase and topoisomerase IV. The efflux pump norA pumps ciprofloxacin out with help from electrochemical gradients before ciprofloxacin accesses its functional loci. (B) Heatmap showing relative expression levels of efflux pumps in wild-type and tolerant stains. The scale above the heatmap indicates log₂-normalized transcript abundance of tolerant strains relative to the mean expression level of wild-type strains.

Tolerance facilitated the resistance development against ciprofloxacin in S. aureus.

Since all the tolerant strains are genetically identical to Newman except for the mutations responsible for tolerance, the increased expression of efflux pumps in the above-described tolerant strains prompted us to question whether efflux pump-overexpressed tolerance could accelerate resistance development in S. aureus. To address this question, we performed in vitro adaptive evolution experiments in the CIP.C17.T8 strain, which has the typical tolerant phenotype (Fig. 1C) caused by the identified mutation in stbD (Table S1). The tolerant strain CIP.C17.T8 was chosen since it was evolved from ciprofloxacin, which was commonly used to treat S. aureus infections. This tolerant S. aureus strain and the isogenic wild-type (Newman) strain were intermittently exposed to lethal concentrations of ciprofloxacin (20× MIC), respectively, for 10 to 21 cycles (Fig. 1B). Strikingly, the tolerant strain CIP.C17.T8 became resistant to ciprofloxacin at the sixth cycle (MIC, 4 μg/mL), while Newman took eight cycles to reach the same resistance levels. Subsequently, the CIP.C17.T8 strain reached the highest MIC value (8 μg/mL) in the seventh cycle (Fig. 1D). Our results collectively demonstrated that the tolerance facilitated the rapid acquisition of resistance in S. aureus under in vitro selection conditions.

Overexpression and deficiency of norA influence resistance development.

Since the tolerant strain CIP.C17.T8 with enhanced activities of efflux pumps developed antibiotic resistance in S. aureus faster than the parent strain, we hypothesized that efflux pumps might contribute to resistance development from tolerance to resistance. Considering NorA’s well-known role in extruding ciprofloxacin, we focused on the efflux pump gene norA for later experiments. To directly determine the impact of the expressional level of efflux pumps on resistance development, we cloned norA genes into the wild-type strain Newman to create the norA overexpression mutant. Meanwhile, we constructed the norA knockout mutant in the tolerant strain CIP.C17.T8 and its corresponding complementary mutant. All above-constructed mutants were subjected to evolutionary experiments in 20× MIC of ciprofloxacin for 10 to 21 cycles using the protocol in Fig. 1B.

Overexpression of norA led to the acquisition of resistance, while the wild-type Newman strain remained susceptible to the antibiotics, with no changes in MIC values (Fig. 3A). On the other hand, knockout of norA delayed the development of resistance to ciprofloxacin compared to the tolerant strain by two cycles (from the 6th to the 8th cycle) (Fig. 3B). Additionally, our complements of efflux pumps restored the development of resistance to ciprofloxacin (Fig. 3B). Overall, results from overexpression and deleted mutants collectively demonstrated that increased expression of efflux pumps facilitated the rapid development from tolerance to resistance.

FIG 3.

Resistance development of efflux pump mutants. The MIC of evolved lines relative to the ancestral MIC of the first cycle, with Newman lines relative to its initial MIC as control. (A) Resistance development for the norA-overexpressed strain (red line) and the wild-type strain (gray line). (B) Resistance development of norA knockout strain (blue line), the tolerant strain (salmon lines), and norA complementary strain (tangerine line).

Elevated expression of norA increased the likelihood of occurrence of resistance mutations.

One possible explanation for the contribution of increased expression of efflux pumps to resistance development is the change in mutation rates. To determine whether overexpression of efflux pumps facilitated resistance evolution through increased mutations, the spontaneous mutation frequency in norA mutants was calculated. We performed these measurements in the media without antibiotics to rule out mutations induced by antibiotic stress. As shown in Fig. 4, overexpression of norA significantly increased random mutation rates compared to the control group. On the other hand, knockouts of norA significantly decreased mutation rates in comparison with the tolerant strains. Complementation in the knockout strains restored the mutation rates. These results highlight that the elevated expression of efflux pumps promoted the genetic mutations and provided an opportunity for the occurrence of specific benefit-oriented resistance mutations.

FIG 4.

Random mutation rates of either norA overexpression and knockout strains. Random mutation rates of the WT strain, the norA overexpression mutant, the CIP.C17.T8 strain, the norA knockout mutant, and the complementary mutant, respectively. n ≥ 3 biological replicates. ns, not significant; *, P < 0.05; **, P < 0.01. Error bars represent the standard deviation of more than three triplicates.

Elevated expression of norA increased bacterial survival ability.

Another strategy for increased resistance development could be increased survival capacities, which allow the bacteria to generate as many resistance mutations as possible before the population becomes extinct under antibiotic selection. To test this plausibility, we monitored the population dynamics of norA mutants over 36 h under 20× MIC of the ciprofloxacin treatment. The strain with norA overexpression showed higher survival ability than wild-type bacteria (Fig. 5A), while the knockout of norA dramatically reduced bacterial survival under antibiotic stress (Fig. 5B). These results collectively suggest that the high expression level of efflux pumps provides an extra opportunity for bacterial survival, allowing existing resistance mutations to fix in the population.

FIG 5.

Effect of norA expression on survival ability. (A) Survival ability of the norA overexpression mutant (red line). The survival rate is plotted over time. The gray lines represent the mean survival of the control lab strain Newman. Data were normalized to the number of CFU at time zero. (B) Survival ability of CIP.C17.T8 strain (salmon lines) and norA knockout mutant (blue line) under 20× MIC concentrate of ciprofloxacin. Data are presented as the means ± SD from at least three biological replicates.

The effect of norA on the resistance mutations of topoisomerase IV and DNA gyrase.

To determine the effect of efflux pumps on the outcome of adaptation, we isolated 30 evolved ciprofloxacin-resistant clones from the final cycle of the CIP.C17.T8 strain and 30 evolved ciprofloxacin-resistant clones from the final cycle of ΔnorA mutant. The canonical mechanism for S. aureus to evolve ciprofloxacin resistance is by point mutations that alter ciprofloxacin targets, including topoisomerase IV ParC and ParE (GrlA and GrlB in S. aureus), or DNA gyrases GyrA and GyrB. To test the genetic basis of resistance development from the CIP.C17.T8 strain and ΔnorA strain, we sequenced regions of parC, parE, gyrA, and gyrB in these selected colonies.

As shown in Fig. 6, each of the 60 clones contained at least one mutation, while clones 17 and 35 contained two mutations in comparison with the CIP.C17.T8 strain and ΔnorA mutant (Fig. 6A). All clones except clones 19 and 45 had a single mutation in the parC region (E84K). Instead, clone 19 contained a mutation at position 116 of parC, while clone 45 contained a mutation at position 80 of parC. However, we found no evidence of a difference in the genetic basis of resistance development between the CIP.C17.T8 strain and the ΔnorA mutant, as both of them possess the same frequency of mutations in each sequenced gene (Fig. 6A). In addition, clones 17 and 35 contained a mutation at position 517 of gyrB. All of the observed mutations in parC and gyrB of this study have been reported to be the mechanism leading to resistance against ciprofloxacin in S. aureus (21). Ciprofloxacin binds to topoisomerase IV by a water-metal ion bridge, which is mediated by Ser80 and acidic residue Glu84 (Fig. 6B). The most frequent mutations of Ser80 mutation and its four amino acids downstream of acidic residue Glu84 in parC disrupted ciprofloxacin-enzyme interactions by the disruption of the formation of the water-metal ion bridge (Fig. 6B).

FIG 6.

Frequency of ciprofloxacin resistance mutations across CIP.C17.T8 and ΔnorA stains. (A) The column panel shows the identified mutations in parC, parE, gyrA, and gyrB. The first 30 colonies from the CIP.C17.T8 and the last 30 colonies from the ΔnorA stain are ranked in the row panel. Colors on the heatmap indicate the frequency of each mutation. The right heatmap shows the total mutation frequency of 30 clones. (B) Diagram for how Ser80Phe and Glu84Lys mutations of topoisomerase IV break the water-metal ion bridge between quinolones and topoisomerase IV. Ciprofloxacin is colored black, and the Mg²⁺ and water molecules are colored green and black. The coordinating serine and acidic residues are depicted as green and red. Black dashed lines present the divalent metal ion interacting with four water molecules. The red dashed line indicates a hydrogen bond between the glutamic acid side-chain carboxyl group and one water molecule. The blue and green dashed lines indicate hydrogen bonds between the serine side-chain hydroxyl group and two water molecules. The middle picture shows the sequence alignment of the conserved serine and acidic residue (highlighting) coordinates of the water-metal ion bridge between CIP.C17.T8 and resistance mutants. Mutations of Ser80Phe (from green to blue) and Glu84Lys (from red to orange) fail to form the hydrogen bond (red, blue, and green dashed lines) between quinolones and topoisomerase IV.

Multiple efflux pumps that extrude ciprofloxacin accelerated ciprofloxacin resistance development.

Since NorA is not the only efflux pump that extrudes ciprofloxacin, other ciprofloxacin-targeted pumps might also participate in developing ciprofloxacin resistance. Here, to determine whether other pumps affect the resistance development against ciprofloxacin from tolerance to resistance, we chose two other pumps, MepA and Sav1866, from different families from NorA but that are also involved in ciprofloxacin transportation (22) to cover a wide range of families for revealing the general function of efflux pumps. We created additional ΔmepA and Δsav1866 mutants in the CIP.C17.T8 strain. Results showed that knockout of mepA and sav1866 kept the bacteria susceptible to ciprofloxacin (Fig. 7A), suggesting that mepA and sav1866 may affect the resistance development against ciprofloxacin from tolerance to resistance in a similar way as norA. Considering the high expression of these pumps in the tolerant strains (Fig. 2A), it can be inferred that the facilitation from tolerance to resistance was attributed by multiple efflux pumps. To determine how a set of efflux pumps contribute to the development from tolerance to the resistance of the same type of antibiotic, ciprofloxacin, we measured expressional levels of the other two pump genes in ΔnorA, ΔmepA, and Δsav1866 mutants, respectively. The deficiency of one efflux pump directly led to an increased complementary expression of the other two, especially in ΔnorA and Δsav1866 knockout strains, with up to 80-fold changes in the relative expression level, implying a less critical role of mepA than of norA (Fig. 7B to D). Multiple ciprofloxacin-targeted efflux pump genes (norA, mepA, and sav1866 from different families) may complementarily promote the development from tolerance to resistance against ciprofloxacin. This interaction between pumps in the tolerance context provides the possibility to figure out multiple targets and strategies for blocking tolerance evolution.

FIG 7.

Effect of other efflux pumps on resistance development against ciprofloxacin. (A) mepA knockout (blue line with the square box icon) and sav1866 knockout strains (blue line with the “x” icon) evolved resistance earlier than the control strain (salmon lines) in evolution experiments. (B to D) Relative expression levels of the other two efflux pumps in norA knockout (B), mepA knockout (C), and savl866 knockout (D) mutants.

Regulator MgrA positively regulates the resistance development of ciprofloxacin.

The strong association between norA and resistance development prompted us to look for upstream regulators, intending to provide a potential target to dampen resistance evolution. MgrA is a well-known regulator for NorA. Overexpression of mgrA resulted in increased expression of norA, which is responsible for the decrease in susceptibility to hydrophilic quinolones (norfloxacin and ciprofloxacin) (23). Here, we used mgrA as a representative to mechanistically clarify the impact of the regulator on antibiotic resistance development from an experimental perspective, so we created a ΔmgrA mutant. Results from reverse transcription-quantitative PCR (qRT-PCR) showed that the mgrA knockout mutant decreased the expression level of norA, providing evidence for the positive regulatory association between MgrA and NorA (Fig. 8A). During periodic exposure to ciprofloxacin, the knockout of mgrA kept the bacteria susceptible to ciprofloxacin (Fig. 8B), suggesting the possibility of delaying the development of resistance by inhibiting MgrA.

FIG 8.

Effect of the regulator MgrA on resistance evolution. (A) Expression levels of norA in mgrA knockout strain with β-actin as control by real-time PCR. (B) Evolutional experiments of ΔmgrA mutant and its corresponding complementary mutant.

Combatting resistance development of ciprofloxacin by the efflux pump inhibitor reserpine.

The facilitation of resistance development by increased expression of efflux pumps illuminates a possible strategy for dampening resistance development by blocking efflux pumps. To test this feasibility, reserpine, which can inhibit the activity of several efflux pumps, including NorA and MepA (24), was chosen to conduct the evolutionary experiment by periodically exposing CIP.C17.T8 strains to media containing 20 times the MIC value of ciprofloxacin either with or without an extra 33 μM reserpine.

As shown in Fig. 9A, reserpine effectively delayed the emergence of ciprofloxacin’s resistance by three cycles (from the 6th cycle to the 9th cycle). Additionally, the killing assay performed in the presence of ciprofloxacin with or without reserpine demonstrated that reserpine shortened an order of magnitude of killing time for CIP.C17.T8 (Fig. 9B). This delayed resistance development and decreased survival collectively suggested that the combination of ciprofloxacin and reserpine could win more time for clinical treatment without sacrificing the killing efficiency against bacteria before antibiotic resistance fails the treatment.

FIG 9.

Blocking the evolution of ciprofloxacin resistance by reserpine. (A) Evolution experiments with cultures started from a ciprofloxacin-tolerant strain (salmon line) that evolved resistance earlier than the one with reserpine (blue line). (B) Killing assay in liquid medium of clones isolated from ciprofloxacin-tolerant strain CIP.C17.T8 with or without reserpine under 20× MIC of ciprofloxacin.

DISCUSSION

Both tolerance and resistance are important reasons for the failure of antibiotic treatments. However, tolerance is still largely neglected clinically compared to resistance (7, 25). In this study, we focused on the process from tolerance to resistance because tolerance is easily classified as susceptibility due to the lack of measurement standards clinically, ignoring its potential role as a reservoir for resistance evolution (25). This neglection will cause relapsed infection and treatment failure because of the development of resistance against clinically available antibiotics. Our in vitro evolutional results that the tolerant strain CIP.C17.T8 acquired resistance more rapidly than the wild-type strain Newman highlighted the importance of tolerance in developing resistance (26). Resistance development from tolerance has been reported both in Gram-positive pathogens such as S. aureus (11) and Gram-negative pathogens such as Pseudomonas aeruginosa (12). Recognizing and understanding the process from tolerance to resistance will be the first step to finding new strategies to block resistance development.

The results from our current study clearly show that efflux pumps promoted the development of resistance from tolerance. This study is different from most previous studies in terms of concentrations of antibiotics to which bacteria are exposed. Until now, most studies focused on the evolutional process of resistance under concentrations near the antibiotic MIC (20, 27). This can be achieved by extruding antibiotics out of bacterial cells to directly cut off the access of antibiotics to their functional targets (28). Some mutations acquired in efflux pump genes or their regulators permanently result in overexpression of efflux pumps and consequently confer high resistance to antibiotics (29–31). Our study revealed that under a high concentration of antibiotics, the increased activity of efflux pumps facilitated resistance development from tolerance. Intermittent exposure of wild-type Newman to high doses of ciprofloxacin led to the rapid evolution of hypertolerance but not high-level resistance (Fig. 1A). Once established, tolerant isolates overexpressed efflux pumps and ultimately became resistant through genetic changes (Fig. 2B; Fig. 6). Efflux pumps were necessary for this process from tolerance to resistance because the knockout of norA delayed or disrupted the resistance acquisition.

The underlying mechanisms of how efflux pumps facilitated tolerant isolates to develop resistance were investigated in this study. Our previous genome sequencing of tolerant strain CIP.C17.T8 identified only one mutation in stbD across the whole genome (see Table S1 in the supplemental material). That means that the CIP.C17.T8 strain was genetically identical to the strain Newman except for the stbD mutation. The elevated expression of efflux pumps (Fig. 2B) in the tolerant strains might be involved in the development from tolerance to resistance. Several efflux pump mutants of norA were made in the CIP.C17.T8 strain to test this hypothesis. During the evolution from tolerance to resistance, overexpression of norA led to higher mutation rates, which increased opportunities for the emergence of resistance mutations. Similar results were also found by El Meouche and Dunlop that higher acrAB (components of multidrug pump AcrAB-TolC) expression of E. coli cells led to higher mutation frequencies (19). At the same time, the decreased mutation rates in the norA knockout mutant were observed, providing additional evidence for the crucial role of efflux pumps in mutations. Second, the generated resistance mutations were further fixed in the whole bacterial population by increasing the survival ability to expand to their descendants. This mechanism is theoretically supported by the model proposed by Windels et al. (32) that increased survival and mutation rates jointly affect the evolution of clinical resistance in E. coli. Our findings further corroborated the early S. aureus study’s conclusion that norA expression potentiated evolution by increasing the survival benefit of DNA topoisomerase mutations with increasing survival ability under ciprofloxacin treatment (20).

In addition to norA, we observed that knockouts of mepA and sav1866 enabled their strains to be susceptible to ciprofloxacin for 12 cycles, suggesting the involvement of many efflux pumps in resistance development instead of a specific efflux pump. The high expression of mepA and sav1866 in the norA knockout strain implied a complementary function of these three efflux pumps in S. aureus. This type of complementation was further observed in the ΔmepA and Δsav1866 strains. Nevertheless, knockout of one efflux pump was sufficient to delay resistance development. Efflux pumps NorA, MepA, and Sav1866 come from different families, and all target ciprofloxacin. Other efflux pumps in the same families with the same substrates possibly have similar effects. For example, NorB and NorC may also contribute to the resistance development against ciprofloxacin like NorA. It is plausible that the development from tolerance to resistance is facilitated by multiple efflux pumps.

Ciprofloxacin binds to topoisomerase IV (ParC, ParE) and DNA gyrase (GyrA, GyrB), inhibiting the repair of DNA breaks and leading to irreversible damage to the genome (33, 34). Ciprofloxacin resistance can result from specific mutations in these two enzymes, which weaken the binding between the antibiotic and these enzymes (35). By comparing the ciprofloxacin-resistant clones from the tolerant strain CIP.C17.T8 and ciprofloxacin-resistant clones from the ΔnorA mutant, we observed similar genetic mutations in the target genes, suggesting that the norA background did not change the resistance mutation preference of these known ciprofloxacin resistance genes.

To effectively avoid the possibly devastating consequences caused by tolerance, more novel strategies should be developed to combat antibiotic resistance evolved from tolerance (36). Since increased expression of efflux pumps positively participated in the process from tolerance to resistance and knockouts of efflux pumps effectively delayed resistance evolution with decreased survival ability, we studied the effect of using the efflux pump inhibitor, reserpine, of several efflux pumps, including NorA and MepA, on blocking resistance development, with good results. A decrease in survival ability under ciprofloxacin caused by the inhibitor added additional evidence to the feasibility of the proposed strategy that the inhibitor of efflux pumps promoted bacterial death at the same time. This blocking strategy was also confirmed in infected female mice: decreased expression of a drug efflux pump increased antibiotic-mediated clearance of gonococci from the genital tract (37). More importantly, the fact that efflux pump inhibitors significantly reduce biofilm formation in Salmonella enterica (38) and Pseudomonas aeruginosa (39, 40) highlights the potential of inhibitors to reduce further bacterial infection ability. Because of the formation of biofilms, the surface-attached communities of bacteria embedded in an extracellular matrix enhanced the persistence of chronic infections (41). However, we need to verify our conclusions in in vivo infection and clinically test the proposed strategy for dampening resistance evolution by the inhibitors of efflux pumps. The effect of efflux pump inhibitors in S. aureus needs to be confirmed at the in vivo level in future research.

As summarized in Fig. 10, elevated expression of efflux pumps facilitated resistance development through two strategies. On the one hand, elevated expression of efflux pumps directly pumps antibiotics out of bacterial cells as an instantly resistant tactic. On the other hand, when wild-type strains are evolved into the tolerant state under a lethal concentration of antibiotics, efflux pumps also indirectly promote resistance development from tolerance to resistance by contributing to the generation and fixation of antibiotic mutations under some regulators such as MgrA in S. aureus.

FIG 10.

Proposed model explains how the efflux pump accelerates the resistance evolution in S. aureus. Under the lethal concentration of antibiotics, bacteria first evolved for tolerance for survival. From tolerance to resistance, the elevated expression of efflux pumps facilitated the resistant evolution of antibiotics by increasing the opportunity for antibiotic mutation generation, which was further fixed in the population supported by increased survival ability. This process is regulated by regulators such as MgrA. This facilitation was attributed to multiple efflux pumps that share the same substrate.

MATERIALS AND METHODS

Bacterial strains and growth conditions.

All tolerant strains of S. aureus used in this study are listed in Table S1 in the supplemental material. In our previous experiments, tolerant strains were acquired by intermittently challenging wild-type strain Newman with an extremely high concentration (20× MIC) of 7 different kinds of antibiotics (ciprofloxacin, oxacillin, imipenem, flucloxacillin, meropenem, cefazolin, and vancomycin). Each cyclic evolution protocol consists of three steps. For the first killing step, an overnight culture (0.5 mL; ~2 × 10⁹ bacteria) was diluted 1:100 into 50 mL TSB supplemented with 20× MIC of antibiotics and incubated for 3, 5, or 8 h at 37°C with shaking at 180 rpm. For the second antibiotic washing step, the antibiotic-containing culture was removed by washing twice in phosphate-buffered saline (PBS) (20 min centrifugation at 1,500 × g). The antibiotic-free culture was resuspended in 1 mL fresh TSB medium with selective antibiotics for overnight growth at 37°C until the new round began. After 9 to 20 cycles, the MIC of the population culture from each cycle was measured. Clones were picked from the cycles before resistance development, and their tolerance was measured. The ones with typical tolerance phenotypes were picked out and purified, and their tolerance phenotype was remeasured. Then, we sequenced genomes of these selected tolerant strains, and all sequenced mutations are shown in Table S1. After restoring the mutated gene and measuring tolerance, genes shown in Table S1 were identified as being responsible for each strain’s tolerance phenotype (our unpublished data).

The genetically tractable model strain Newman is susceptible to ciprofloxacin and oxacillin. The E. coli IM08B strain was stored in our laboratory. S. aureus strains were cultured in TSB with orbital shaking or on TSB agar for incubation at 37°C. E. coli strains were cultured in Luria-Bertani (LB) medium under the same culture condition. For strain storage, bacterial culture was mixed with glycerol to a final volume of 50% (vol/vol) and stored at −80°C.

MIC determination for parental strains.

The MIC of each evolutional strain was determined by the broth microdilution method in 96-well plates. Briefly, three morphologically similar colonies of each strain were resuspended in TSB for overnight growth. Then, the culture was adjusted to the optical density at 600 nm (OD₆₀₀ ≈ 0.16 to 0.18) to ensure it would contain approximately 1 × 10⁸ cells/mL. This standardized inoculum was diluted a further 40-fold in Mueller-Hinton 2 broth (Sigma-Aldrich Co., St. Louis, MO, USA) containing antibiotics at a concentration gradient between 64 mg L⁻¹ and 0.03125 mg L⁻¹. After 24 h of incubation at 37°C, the lowest concentration which completely inhibited bacterial growth was considered the final MIC value. The Clinical and Laboratory Standards Institute (CLSI) defined MIC interpretive criteria of ciprofloxacin as ≤1 μg/mL for susceptibility and ≥4 μg/mL for resistance in S. aureus (42). The MIC value of ciprofloxacin in parental Newman and CIP.C17.T8 strains was 1 μg/mL.

Measuring expressional levels of efflux pumps by qRT-PCR.

Frozen stocks of tolerant strains and wild-type Newman were diluted 1:1,000 into fresh TSB medium for overnight culture. For each tolerant strain group, 5 μL of the tolerant overnight culture and Newman were inoculated into 5 mL TSB, which contained 20× MIC of corresponding antibiotics to match the conditions of the previous evolutional experiment. After 4.5 h of shaking at 180 rpm in a 37°C incubator, tolerant and wild-type bacterial cells were collected by centrifuging for 5 min at 5,000 × g. Total RNA was extracted using the manufacturer's protocol for Gram-positive bacteria of the E.Z.N.A. HP Total RNA kit (Omega, Norcross, GA, USA). After assessing the purity by measuring absorbance at 230, 260, and 280 nm and visualizing migration on a 0.7% agarose gel, extracted RNA was prepared for reverse transcription into cDNA with a cDNA TransScript One-Step genomic DNA (gDNA) removal and cDNA synthesis supermix (TransGen, Beijing, China).

The expressional levels of target genes were determined by quantitative PCR in a 25-μL TransStart Tip Green qPCR supermix (TransGen) using 2 μL of diluted cDNA as the template. The normalized expression was calculated using the 2^−ΔΔCT threshold cycle method using 16S rRNA as a housekeeping reference gene. The expression of the determined efflux pumps in tolerant strains was log₂ normalized relative to the mean expression level of the wild-type strain Newman. Mean fold change values were equivalent to the normalized expression ratio. All the primers for qPCR are listed in Table S2.

Construction of mutants.

Construction of norA was performed according to the well-established procedure as described by Li et al. (43). The expression of norA, mepA, and sav1866 in the wild-type strain Newman was used as control. The norA-overexpressed vector pSE1 was introduced into Newman to overexpress efflux pumps. Genes mepA, norA, sav1866, and mgrA were knocked out in CIP.C17.T8 strains using the pKZ2 vector to obtain ΔmepA, ΔnorA, Δsav1866, and ΔmgrA mutants, respectively. The shuttle plasmids pSC1 with the knockout gene were used to restore norA and mgrA in their corresponding knockout mutants. All primers used in this section are listed in Table S2.

All the original vectors were first linearized with EcoRI and KpnI restriction enzymes and purified using the universal DNA purification kit (Tiangen, Beijing, China). Then, the efflux pump genes were amplified from the genomic DNA of Newman by PCR using primers with 5′ overhangs designed for Gibson assembly (Table S2). These linearized vectors and amplified genes were assembled using pEASY-Basic seamless cloning and assembly kit (TransGen). Then, the assembled vectors were transformed into E. coli DH5a competent cells (TransGen).

Selected clones from LB plates with carbenicillin (100 mg L⁻¹) were confirmed by PCR and Sanger sequencing. Constructed expressional, knockout, and overexpressing vectors were first transformed into IM08B and then into targeted S. aureus tolerant strains by electroporation using the protocol described (44, 45). S. aureus transformants were selected from TSB agar plates with chloramphenicol (10 mg L⁻¹) and verified by PCR and Sanger sequencing.

In vitro adaptive evolution experiment.

Passaging experiments were performed following the slightly modified protocol from reference 13 and our previous treatment (Table S1). Overall, populations of mutant (overexpression, knockout, and complement strains) and tolerant strains were cyclically exposed to high concentrations of antibiotics (20× MIC) for a total of 7 to 21 iterations until antibiotic resistance was detected.

The detailed cyclic evolutional protocol consisted of three steps, antibiotic treatment, washing off the antibiotic, and overnight growth. First, an overnight culture (20 μL) was inoculated at a 1:100 ratio into 2 mL TSB supplemented with the antibiotic concentration of 20× MIC and incubated at 37°C with shaking for 4.5 h. Second, the antibiotic-containing medium was removed by washing twice in with centrifugation at 1,500 × g for 20 min. Finally, the washed bacterial cells were resuspended in 2 mL fresh TSB for overnight growth. Half of the regrown culture (1 mL) was prepared for the next passaging and another half after the confirmation of culture purity by observing the bacterial morphology after plating 100 μL overnight culture on the TSA plates frozen at −80°C for further analysis.

This assay was also performed to determine the effect of the efflux pump inhibitors on resistance evolvability by exposing CIP.C17.T8 to 20× MIC of ciprofloxacin, either with or without 33 μM reserpine.

Mutation rate estimation.

The strains were regrown from −80°C stock on the MH2 plate. Three colonies of each strain were inoculated in the TSB medium. After 24 h of incubation, cultures were subject to a bottleneck by diluting them 10⁷ times to minimize the presence of preexisting stationary-phase mutants. The diluted bacterial cultures were used to establish eight replicate cultures per colony. After overnight incubation at 37°C, a 5-μL culture of 8 replicates was diluted 10⁷ or 10⁸ times and plated on TSB agar plates to estimate population density.

Meanwhile, 100 μL of each culture was plated on TSB agar plates containing 100 mg L⁻¹ of the antibiotic rifampin. TSB plates were incubated for 24 h at 37°C; selective plates were incubated for 48 to 72 h at 37°C. The mutation frequency was calculated by dividing the CFU mL⁻¹ of selective antibiotic agar plates by that of the antibiotic-free agar plates.

Measurements of survival ability.

For measurements of survival ability under antibiotics, overnight cultures grown from a single colony in TSB were diluted at 1:100 in fresh medium supplemented with antibiotics at the concentration of 20× MIC. As for inhibitors, an extra 33 μM reserpine was added to the ciprofloxacin treatment. At indicated time points (12 h, 24 h, and 36 h), aliquots of the cultures (500 μL) were sampled, diluted to the appropriate dilutions, and plated on TSB plates. CFU were measured after over a 48-h period of incubation to make sure that all CFU had appeared. Care was taken to avoid the effects of the leftover antibiotic on plating efficiency by washing the culture before plating with PBS. All measurements were repeated in at least three independent experiments with more than three repetitions.

Statistical analysis.

Data were analyzed using GraphPad Prism version 9.0.0 for Mac. One-way analysis of variance (ANOVA) followed by least significant difference (LSD) or Dunnett’s T3 multiple-comparison test was performed. Significant differences were defined by two-tailed P values (*, P < 0.05; **, P < 0.01).