Salicylate Increases Fitness Cost Associated with MarA-Mediated Antibiotic Resistance

Abstract

Antibiotic resistance is generally associated with a fitness deficit resulting from the burden of producing and maintaining resistance machinery. This additional cost suggests that resistant bacteria will be outcompeted by susceptible bacteria in conditions without antibiotics. However, in practice, this process is slow in part because of regulation that minimizes expression of these genes in the absence of antibiotics. This suggests that if it were possible to turn on their expression, the cost would increase, thereby accelerating removal of resistant strains. Experimental and theoretical studies have shown that environmental chemicals can change the fitness cost associated with resistance and therefore have a significant impact on population dynamics. The multiple antibiotic resistance activator (MarA) is a clinically important regulator in Escherichia coli that activates downstream genes to increase resistance against multiple classes of antibiotics. Salicylate is an inducer of MarA that can be found in the environment and derepresses marA’s expression. In this study, we sought to unravel the interplay between salicylate and the fitness cost of MarA-mediated antibiotic resistance. Using salicylate as an inducer of MarA, we found that a wide spectrum of concentrations can increase burden in resistant strains compared to susceptible strains. Induction resulted in rapid exclusion of resistant bacteria from mixed populations of antibiotic-resistant and susceptible cells. A mathematical model captures the process and predicts its effect in various environmental conditions. Our work provides a quantitative understanding of salicylate exposure on the fitness of different MarA variants and suggests that salicylate can lead to selection against MarA-mediated resistant strains. More generally, our findings show that natural inducers may serve to bias population membership and could impact antibiotic resistance and other important phenotypes.

Significance

Antibiotic resistance is frequently associated with fitness costs, suggesting that bacterial populations should be biased away from drug resistance in the absence of antibiotic pressure. In practice, precise regulation of these resistance mechanisms largely eliminates this fitness deficit. However, environmental chemicals that are able to turn on expression can have a significant impact on the antibiotic resistance level of a given population. In this manuscript, we study the fitness cost of a clinically important resistance gene, marA, under exposure to its environmental inducer salicylate. We found that a wide spectrum of salicylate concentrations can turn on the resistance machinery, even in the absence of antibiotic stress. This work indicates the potential for important interaction effects between environmental chemicals and population composition.

Introduction

Antibiotic resistance is frequently associated with fitness deficits such as those resulting from burdensome expression of resistance proteins or from excessive energy consumption by resistance machinery (1, 2, 3, 4, 5). These observations suggest that in the absence of antibiotic pressure, a bacterial population should be biased away from drug resistance (6, 7, 8). In practice, compensatory mutations and precise regulation of burdensome protein expression reduce the effective cost, largely eliminating the fitness differences. For example, in an antibiotic-free environment, tetracycline-resistant bacteria that express the costly TetA efflux pump are expected to be outcompeted by susceptible strains (9). However, experiments have shown that the fitness cost of tetracycline resistance without antibiotic pressure is minimal (7, 10). Tight repression of tetA by TetR keeps costs low in the absence of inducers (7, 9). These regulatory approaches to reducing cost are common, and other examples include repression of the multiple antibiotic resistance activator marA by MarR (11) and the multidrug efflux pump acrAB by AcrR (12). However, certain environmental chemicals can serve as inducers that relieve repression and may change the fitness of antibiotic-resistant bacteria.

Environmental inducers include both the substrates that these mechanisms protect against (e.g., antibiotics) and other compounds. For example, the tetracycline decay product anhydrotetracycline (10), which is found widely in soil and wastewater (13, 14), significantly increases the fitness cost of tetracycline-resistant cells by releasing repression by TetR. The presence of anhydrotetracycline will select against resistant cells in the absence of antibiotic pressure because TetA is costly but provides no benefit (9, 10). Other environmental chemicals such as food preservatives (15) and pharmaceutical products (1) can impose fitness disadvantages by inducing higher levels of expression of resistance genes, which may lead to their eventual loss. A recent study demonstrated this effect by growing Escherichia coli for 2000 generations in the presence of benzoate, an inducer for multiple antibiotic resistance genes. The majority of benzoate-evolved strains acquired mutations in marA and its homolog rob, thereby lowering resistance against chloramphenicol and tetracycline (15).

MarA is a clinically important antibiotic resistance regulator conserved across enteric bacteria (16). It increases resistance levels by regulating over 40 downstream genes (17, 18, 19, 20, 21, 22, 23), including multidrug efflux pumps and porins (24, 25). Activation of these genes is taxing, resulting in decreased growth and a reduced fraction of cells with elevated MarA in a mixed community (1, 3). However, marA is under the control of its repressor MarR, which minimizes the cost in the absence of induction (21). MarA can also be modulated by its environmental inducer salicylate (1, 17), which binds directly with MarR and prevents repression of marA (16), significantly decreasing growth (1), which suggests an intriguing interplay between this environmental inducer and the cost of resistance machinery.

In this work, we focused on the cost of MarA-mediated resistance under salicylate exposure. We found that a wide range of concentrations of salicylate can induce MarA-mediated burden. AcrAB-TolC efflux pumps were found to be a major contributor to the salicylate-induced burden. The difference in cost under salicylate exposure leads to rapid exclusion of resistant strains in competitive settings. A mathematical model captures the process by which populations are biased away from resistance and predicts its effect in various salicylate conditions. This work suggests that the fitness cost of MarA-mediated antibiotic resistance can be amplified by salicylate, biasing populations toward susceptible strains.

Materials and Methods

Strains

We used four strains in this research: wild type (WT), MarA⁻, MarA⁺, and AcrB⁻. We used E. coli MG1655 as the WT strain. In the MarA⁻ strain, we deleted the marRAB operon, rob gene, and soxSR genes from E. coli MG1655 (26). In the MarA⁺ strain, we performed transversion mutations on MarR binding sites (23) in the chromosomal marRAB promoter to inhibit binding of MarR to the promoter region in E. coli MG1655 (26). Further details on the deletion and transversion strains are given in (26). The AcrB⁻ strain was derived from the Keio collection JW0451 (BW25113 ΔacrB::kan), and we removed the kanamycin resistance marker; details described in (27).

For strains in which we inserted the chromosomal fluorescent protein marker, we replaced the galK gene with sfgfp driven by a strong constitutive promoter as reported in (28). We generated a polymerase chain reaction fragment of the constitutive promoter-driven sfgfp with extensions homologous to the regions adjacent to the galK gene. We used the forward primer 5′-TTC ATA TTG TTC AGC GAC AGC TTG CTG TAC GGC AGG CAC CAG CTC TTC CGa gag gat cga gtt atc aaa aag a-3′ and the reverse primer 5′-TGC GCG CAG TCA GCG ATA TCC ATT TTC GCG AAT CCG GAG TGT AAG AAA TGg agt ttg gat ccc tat tat ttg ta-3′. Capitalized letters indicate the homologous recombination extensions. We then followed the recombination protocol in (29). After recombination, galK-deleted colonies were selected using the counterselection method reported in (30).

Minimal inhibitory concentration

Overnight cultures of WT, MarA⁻, MarA⁺, and AcrB⁻ were diluted 1:100 in Luria broth (LB) medium. Diluted cultures were incubated at 37°C with 200 revolutions per minute (rpm) shaking for 4 h. Cells were then incubated with carbenicillin (0–50 μg/mL in twofold dilutions) at 37°C for 24 h with 200 rpm shaking. The minimal inhibitory concentration was determined as the concentration of antibiotic at which no visible growth was observed.

Bacterial growth conditions

For all growth experiments, overnight cultures inoculated from a single colony were diluted 1:500 in LB medium. The diluted cultures were then precultured for 3 h at 37°C with shaking at 200 rpm. We refer to the end of this preculture as time point t = 0 h. Salicylate was added at t = 0 h when required; we then continued culturing at 37°C with 200 rpm shaking. Optical density (OD)_660nm readings were taken using a BioTek Synergy H1m plate reader (BioTek, Winooski, VT) each hour.

Competition experiments

For the competition experiments, each of two strains were cultured overnight separately, then diluted 1:1000 in fresh LB medium. Competition cocultures were created by mixing cultures of each of two strains together in equal proportion. Cocultures were then precultured for 3 h, and then at t = 0, salicylate was added where required, as in (1). OD_660nm readings were taken each hour using the plate reader. In addition, immediately after OD_660nm measurements, 5 μL of the sample was diluted into 200 μL sterile phosphate-buffered saline, and we measured fluorescent protein expression with a Guava easyCyte HT flow cytometer (Luminex, Austin, TX). Flow cytometry data were analyzed with custom MATLAB (The MathWorks, Natick, MA) scripts. Control experiments using strains without green fluorescent protein expression were carried out to determine the threshold for strain classification (Fig. S1).

Antibiotic killing experiments

Cultures for antibiotic killing experiments were created by diluting overnight cultures of each of the strains at the percentages shown in Fig. 4. Overnight cultures of each strain were first diluted 1:1000 in fresh LB medium and precultured for 3 h at 37°C with 200 rpm shaking, and then strains were mixed together at percentages as specified in the figure axis. 50 μg/mL of carbenicillin was added to the cocultures at t = 0. Cells were then centrifuged and resuspended in phosphate-buffered saline to wash out the antibiotics. Washed cultures were diluted and plated on LB agar for 24 h to determine the number of colony-forming units per milliliter (CFU/mL) after antibiotic exposure.

Figure 4.

Reducing fraction of MarA⁺ (and WT) cells increases susceptibility to antibiotic killing. Colony-forming units per milliliter (CFU/mL) of cocultures of MarA⁺ with (a) WT, (b) AcrB⁻, and (c) MarA⁻ and cocultures of WT with (d) AcrB⁻ and (e) MarA⁻ after applying 50 μg/mL carbenicillin for 4 h are shown. Data points show mean values and SD from n ≥ 3 biological replicates. NS, not significant; ^∗p < 0.05; ^∗∗p < 0.01; ^∗∗∗p < 0.001. Student’s t-test comparing against 100% MarA⁺ (a–c) or 100% WT (d–e). To see this figure in color, go online.

Dose-response experiments

Overnight cultures of the WT strain were diluted 1:50 in LB medium with 0, 0.6, 1.2, and 5 mM salicylate. Diluted cultures with various concentration of salicylate were then precultured for 3 h at 37°C with 200 rpm shaking. Cells were then incubated with carbenicillin (0–50 μg/mL in twofold dilutions) at 37°C for 4 h with 200 rpm shaking. OD_660nm readings were taken using the plate reader 4 h after carbenicillin exposure to determine cell density.

marA and acrAB transcriptional reporters

We used marA and acrAB transcriptional reporters in the WT strain to measure their transcriptional activity upon salicylate exposure. As described in (31), the promoter region of each gene was cloned upstream of the gene for cyan fluorescent protein on a low-copy (SC101 origin) plasmid. Salicylate was added at t = 0. The reporter fluorescence level was measured using microscopy after 3 h of salicylate exposure. Fluorescence distributions were calculated using kernel density.

Gompertz model

All growth curves were fitted to the modified Gompertz model (Eq. 1) described in (32). Carrying capacity (A), growth rate (μ), and lag time (λ) were determined by least-squares fitting to the mean of the experimental data using a custom MATLAB script.

$$\text{OD}660=\text{A}\times \exp \left\{-\exp \left[\frac{\mu \times e}{A}\times \left(\lambda -t\right)+1\right]\right\}$$ (1)

Selection coefficient

The selection coefficients were calculated using the regression model (Eq. 2) described in (33) after 4 h of competition with or without exposure to salicylate.

$$\text{S}=\left[ln\frac{R\left(t\right)}{R\left(0\right)}\right]/\left[\text{t}\right]$$ (2)

S is the selection coefficient; R(t) is the final competition ratio of mutants to the WT strain, and R(0) is the initial competition ratio of mutants to WT strain; t is number of generations in the competition experiment, corresponding to the number of doublings of OD_660nm observed (34).

Lotka-Volterra model

The modified Lotka-Volterra equations (Eqs. 3 and 4) from (35) were used to simulate competitive growth in the co-culture:

$$\frac{d{N}_1}{dt}={\mu }_1\times \frac{N_1}{A_1}\times \left[A_1-N_1-\left(\alpha \times N_2\right)\right],$$ (3)

$$\frac{d{N}_2}{dt}={\mu }_2\times \frac{N_2}{A_2}\times \left[A_2-N_2-\left(\alpha \times N_1\right)\right].$$ (4)

N_i is the cell density of strain i, μ_i is the maximal growth rate, A_i is the carrying capacity, and α is the interaction matrix. We assume α = 1. Both μ_i and A_i were determined using single-strain experimental growth curve data fit to the Gompertz model, as described above. Long-time simulations for Fig. 5 assume dilution every eight generations to the initial cell density, modeling serial transfers. We assume that the generation time is 30 min in the laboratory environment. All simulations were conducted using custom MATLAB scripts.

Figure 5.

Mathematical modeling predicts salicylate can accelerate population shift toward antibiotic-susceptible strains. Predicted population compositions of competitions between (a and b) susceptible strains (AcrB⁻ or MarA⁻) and MarA⁺ or (c and d) susceptible strains (AcrB⁻ or MarA⁻) and WT obtained by extending the Lotka-Volterra model over 60 generations are shown. Lines represent the final population fraction of susceptible cells for a particular salicylate concentration. To see this figure in color, go online.

Results and Discussion

MarA-mediated antibiotic resistance is costly

To investigate MarA-mediated resistance, we employed three strains of E. coli: one in which MarA and its homologs Rob and SoxS are knocked out (denoted MarA⁻) (36), WT, and a strain in which MarA is overexpressed via mutations in the MarR binding sites on its own promoter (MarA⁺). To verify that these strains exhibit differences in antibiotic resistance, we first measured the minimal inhibitory concentration (MIC) of carbenicillin. As expected, we found that the MIC increases as MarA expression levels increase (Fig. S2).

To quantify the cost of MarA-mediated resistance, we measured the growth of MarA⁻, WT, and MarA⁺ strains in LB medium without antibiotics and used fits to the Gompertz model (Eq. 1) to extract parameters associated with growth from the data: growth rate (μ), lag time (λ), and carrying capacity (A) (32) (Fig. 1 a). The exact growth rate values derived from the Gompertz model are likely to be sensitive to the temporal resolution of the measurements, which will impact their exact numerical values. In this work, we were primarily concerned with relative growth rates and used identical measurement resolutions for experiments in which we fitted the data. We found that of the three model parameters, only the growth rate was strongly dependent on MarA expression levels (Fig. 1 b), and effects on lag time and carrying capacity were minimal (Fig. S3).

Figure 1.

MarA-mediated resistance is costly. (a) MarA⁻ growth curve is shown with experimental and modeling data. Black solid line is a fit to the Gompertz model. Data points show experimental data from six biological replicates. Error bars show SD. Three parameters are extracted from the Gompertz model: growth rate (μ), lag time (λ), and carrying capacity (A). (b) Growth rate extracted from the Gompertz model fits for each strain is shown. Data points show mean values and SD from at least four biological replicates. Lag time and carrying capacity values are shown in Fig. S3. (c) The fraction of cells of each strain over time after competition between MarA⁻ and either WT or MarA⁺ is shown, with initially equal proportions of the competing strains in well-mixed liquid cultures. Relative proportions were obtained using counts of fluorescent cells from flow cytometry. Error bars show SD from three biological replicates. NS, not significant; ^∗p < 0.05; ^∗∗p < 0.01; ^∗∗∗p < 0.001, Student’s t-test. To see this figure in color, go online.

To evaluate whether these differences in growth rates could alter the proportion of bacteria with different MarA expression levels in a mixed population, we carried out competition assays between MarA⁻ and either WT or MarA⁺ strains under antibiotic-free conditions (Fig. 1 c).

To confirm and visualize the growth rate differences in the strains, we integrated a gene encoding green fluorescent protein (sfgfp) into the genome of MarA⁻ and quantified the compositions of the mixed populations by flow cytometry (Materials and Methods; Fig. S1). With initially equal representation between the strains, we found that the MarA⁻ strain outgrew MarA⁺, but not WT, in a short time window (4 h). In a longer competition (24 h), the MarA⁻ strain outcompeted both WT and MarA⁺, as expected because of the lack of burdensome MarA expression. The change in population composition is strongest between the antibiotic-susceptible (MarA⁻) and antibiotic-resistant (MarA⁺) strains (Fig. 1 c), suggesting fitness differentials are amplified if MarA is overexpressed. As a control, we also competed WT cells with and without genomically integrated sfgfp and found that the strains showed equal fitness, as expected (Fig. S4). The inverse relationship between MarA-mediated resistance and fitness supports the concept that differences in fitness cost could result in exclusion of resistant bacteria.

Salicylate increases fitness cost in resistant strains compared to susceptible strains

Because MarA expression is costly, we reasoned that inducing its expression with salicylate should accelerate the process of biasing the population toward strains that lack MarA expression. To test this, we first measured expression from the marA promoter induced with various concentrations of salicylate (Fig. S5 a). Using transcriptional reporters for marA in WT cells, we observed a clear trend of increased expression as the salicylate concentration increased. We then measured the cost of marA expression induced by salicylate. As expected, increasing salicylate dosages have a clear negative impact on the growth of WT cells (Fig. 2 a). However, the salicylate growth reduction effect can also be observed in the MarA⁻ strain (Fig. 2 b), which has marA and its homologs deleted. This suggests that salicylate can also reduce growth in a MarA-independent manner, possibly through global effects such as inhibiting expression of ATP synthase (37), decreasing membrane potential (38), or reducing metabolic activity (26).

Figure 2.

Salicylate increases fitness cost of resistant strains. (a) Growth curves of WT strain under increasing salicylate concentrations are shown. Solid lines show Gompertz model fits. For experimental data, error bars show SD from four biological replicates. (b) Growth rates extracted from the Gompertz model for different strains at a range of salicylate concentrations show strain-specific trends between growth rate and salicylate concentration. Data points and error bars show experimental data from at least four biological replicates. (c) Relative growth rates between strains at increasing salicylate concentrations are shown. Data points show ratio of mean growth rate between two strains. Dashed lines show linear least-squares fit. Slope and its 95% confidence interval (in parentheses) of the linear fit are 0.1196 (0.0958, 0.1433) for μ_AcrB⁻/μ_MarA⁺; 0.0760 (0.0500, 0.1021) for μ_AcrB⁻/μ_WT; 0.0300 (0.0201, 0.0398) for μ_MarA⁻/μ_MarA⁺; and 0.0051 (−0.0125, 0.0226) for μ_MarA⁻/μ_WT. To see this figure in color, go online.

Because salicylate can increase antibiotic resistance and decrease cell growth, to measure the costs and benefits of salicylate induction, we further measured the dose-response curve between carbenicillin and salicylate in the WT strain (Fig. S6). We observed a severe reduction in growth at high concentrations of salicylate (5 mM) in the absence of antibiotics, indicative of costs imposed by salicylate exposure. However, the cells have a growth benefit from salicylate induction when the carbenicillin concentration is higher than 10 μg/mL, suggesting a tradeoff between benefits and costs of salicylate induction.

We next asked whether salicylate could increase the cost in strains with MarA expression (WT and MarA⁺) relative to the MarA⁻ strain. To do this, we compared the relative growth rates between MarA⁻ with either WT or MarA⁺, measured by the ratio of the growth rates between the two strains in increasing concentration of salicylate (Fig. 2 c). Surprisingly, we found a weak positive correlation between relative growth rate and salicylate concentration in MarA⁻ versus MarA⁺ but not in MarA⁻ versus WT. Note that MarA⁺ cannot respond to salicylate via the native repressor MarR because of mutations in its MarR binding sites, indicating that this response to salicylate involves factors beyond MarA overexpression, potentially through induction of the MarA homolog Rob (36). We also observed similar effects using fits to a Hill function (1) to quantify the growth cost in different strains as a function of salicylate concentration (Fig. S7).

AcrAB-TolC efflux pump is a major contributor to MarA-mediated resistance burden

We next asked where the cost of salicylate-induced MarA burden was coming from. MarA regulates many downstream genes, but the AcrAB-TolC efflux pump is known to play a critical role in resistance and imposes a significant energy cost (3). A transcriptional reporter for acrAB also showed that acrAB transcription can be induced by salicylate (Fig. S5 b). To test its role in salicylate-induced burden, we constructed AcrB⁻, an acrB knockout strain that renders the efflux pump nonfunctional. We also observed lower MIC values in AcrB⁻ compared to WT or MarA⁺ (Fig. S2).

We measured growth rate of AcrB⁻ with MarA⁺ and WT and compared it under various concentrations of salicylate (Fig. 2 b). Interestingly, salicylate has only a modest effect on the growth rate of AcrB⁻ below 2.5 mM concentrations. We also observed a positive linear relationship between salicylate and the relative growth rate of AcrB⁻ with WT and MarA⁺ (Fig. 2 c). These results demonstrate that salicylate imposes less burden if the efflux pump is deleted. It is important to note that the AcrB⁻ strain has intact copies of MarA and its homologs. Thus, AcrB⁻ can still respond to salicylate exposure through MarA and its homologs, but the cost will not be increased in low salicylate concentrations if AcrAB-TolC cannot be produced. These results suggest that salicylate can induce much higher burden in the resistant strain (MarA⁺) than in the susceptible strain (MarA⁻), and this burden is mainly contributed by the AcrAB-TolC efflux pump.

Salicylate accelerates competitive exclusion of resistant bacteria

We next sought to test whether using salicylate to increase the relative cost in the resistant strain (MarA⁺) accelerates its exclusion in competitive environments. Because we found AcrAB-TolC to be largely responsible for the salicylate-induced fitness cost, we first measured competitive exclusion between AcrB⁻ and MarA⁺. Using an AcrB⁻ strain with genomically integrated sfgfp, we measured the fraction of AcrB⁻ and MarA⁺ cells by flow cytometry and the growth of the competition culture using a plate reader.

In the absence of salicylate, small differences in growth rates are enough to allow the susceptible strain AcrB⁻ to outcompete the resistant strain MarA⁺, although the exclusion effect is not large (Fig. 3 a). When the salicylate dosage was raised to 0.6 mM, the fraction of AcrB⁻ cells increased because salicylate reduces the growth of the resistant strain MarA⁺ without imposing a significant impact on the susceptible strain AcrB⁻ (Fig. 3 b). This difference is further magnified at higher levels of salicylate exposure (Fig. 3 c).

Figure 3.

Salicylate accelerates competitive exclusion of resistant bacteria. Competitions of AcrB⁻ (blue) with MarA⁺ (red) in (a) 0 mM, (b) 0.6 mM, and (c) 5 mM salicylate seeded with initially equal populations are shown. The red and blue shaded areas represent the predicted growth of each subpopulation using the Lotka-Volterra model, with the total area representing the combined population growth. Competitions of WT with (d) AcrB⁻, (e) MarA⁻, and (f) MarA⁺ in various concentrations of salicylate seeded with initially equal populations are shown. Relative competitive fitness was measured by dividing cell counts of each mutant (N_AcrB⁻, N_MarA⁻, or N_MarA⁺) by cell counts of WT (N_WT). Cell counts were obtained by counting fluorescent cells from flow cytometry. Strains with genomically integrated sfgfp are AcrB⁻ in (a)–(d), MarA⁻ in (e), and WT in (f). Data points show mean values and SD from three biological replicates. ^∗p < 0.05; ^∗∗p < 0.01; ^∗∗∗p < 0.001, Student’s t-test. To see this figure in color, go online.

To predict the competition dynamics between strains in different concentration of salicylate, we employed a mathematical model that we fitted to the single-species growth rate (μ) and carrying capacity (A) data and used it to predict two-species competition under various salicylate conditions. The growth rates and carrying capacities calculated from single-species growth curves fit using the Gompertz model were applied to a Lotka-Volterra model for competitive growth (Eqs. 3 and 4). The model shows good agreement with the experimental data in the dynamics of both growth and the population fraction for each strain (Fig. 3, a–c).

We observed similar effects with salicylate when competing WT with AcrB⁻ (Fig. 3 d) and MarA⁻ (Fig. 3 e). However, competition between WT and MarA⁺ in salicylate did not lead to accelerated competitive exclusion (Fig. 3 f). Because salicylate can significantly decrease bacterial growth (Fig. 2 b), we verified that our results were not an artifact of differences in growth rate by calculating the selection coefficients during competition between WT and all mutants under salicylate exposure (Fig. S8). The selection coefficient takes into account growth differences at higher salicylate concentrations (33). We observed similar trends in selection coefficients and results with cell counts, in which AcrB⁻ and MarA⁻ have significantly higher fitnesses compared with WT at high concentrations of salicylate.

Reducing population fraction of resistant cells increases susceptibility to antibiotic exposure

To study the impact of the fraction of resistant strains on antibiotic resistance of a population, we subjected mixed populations with different proportions of susceptible (AcrB⁻ and MarA⁻) and resistant (MarA⁺) strains to a lethal dose of carbenicillin (50 μg/mL) for 4 h and then measured survival by counting the CFU/mL. The results show that carbenicillin resistance is directly related to the fraction of MarA⁺ cells in the culture (Fig. 4, a–c). We found similar results in mixed populations with WT and its susceptible counterparts (AcrB⁻ and MarA⁻) (Fig. 4, d–e), although the scale of CFU/mL change is not as substantial as in mixed populations with susceptible and resistant strains. Decreasing the proportion of MarA⁺ can significantly reduce the antibiotic resistance of the population. This result supports our hypothesis that antibiotic resistance levels decrease when a susceptible strain outcompetes a resistant strain.

Modeling population shifts under salicylate treatment

In the real-world environment, susceptible mutants may only constitute a small proportion of cells. To study the population dynamics starting from a very small initial proportion of the susceptible strain, we next used the Lotka-Volterra model to make predictions about the effects of salicylate exposure on mixed populations of susceptible mutants with resistant and WT strains (Fig. 5).

We simulated competitive exclusion by starting with various initial proportions of the susceptible strains (AcrB⁻ or MarA⁻) in a mixed population with MarA⁺ or WT and applied the model over the course of 60 generations (∼1 week in real time (7)). The model predicts that resistant bacteria (MarA⁺ or WT) will be outcompeted by their susceptible counterparts much more rapidly with salicylate exposure than under conditions without salicylate. For example, if the susceptible strain AcrB⁻ initially comprises 0.01% of a mixed population with MarA⁺ (Fig. 5 a), after 60 generations, more than 30% and 99% of MarA⁺ is excluded in the presence of 0.3 and 5 mM salicylate, respectively. These results are in stark contrast to the conditions without salicylate, in which less than 5% of the MarA⁺ can be outcompeted by AcrB⁻ with the same initial composition. Similar trends in population shifts are also observed under competion with the WT strain (Fig. 5 c) or in a mixed population of MarA⁻ with MarA⁺ or WT (Fig. 5, b and d). The model predicts that even with extremely low initial proportions, susceptible strains (AcrB⁻ or MarA⁻) can outgrow MarA⁺ or WT and that salicylate accelerates this process, shifting the population toward antibiotic-susceptible strains.

Conclusions

Antibiotic resistance mechanisms frequently impose a significant fitness cost, suggesting the potential for competitive exclusion of resistant strains. However, this process is usually slow because of downregulation of the resistance machinery in the absence of inducers. One study combining mathematical modeling and experimental data predicted that it would take ∼1.5 years to reverse tetracycline resistance by replacing 99.9% of the cells carrying the tetA resistance gene (7). However, in the natural environment, bacteria commonly interact with other chemicals that could potentially induce expression of resistance genes. Costs imposed by induction could bias population composition.

Resistant mutants with MarA overexpression are commonly found in clinical E. coli antibiotic-resistant isolates (39, 40, 41). In addition, recent laboratory evolution studies have also found that MarR, a repressor of MarA, frequently mutates under antibiotic stress (42), suggesting that the MarA-mediated resistant mutant (MarA⁺) may be in abundance in environments with frequent antibiotic exposure.

In this study, we sought to unravel the impact of salicylate exposure on MarA-mediated resistance cost. We established a mathematical model to predict the population dynamics under different levels of salicylate exposure. We found that salicylate imposes a higher cost in MarA-mediated resistant strains (MarA⁺) than in susceptible (MarA⁻ and AcrB⁻) strains. Competition assays between MarA⁺ and AcrB⁻ showed a positive correlation between salicylate concentration and the relative fitness cost. This suggests that with higher salicylate induction, the relative fitness cost of resistant strains will increase, accelerating their competitive exclusion. Mathematical modeling also predicted mixed population dynamics from single-strain growth parameters and confirmed the effectiveness of salicylate exposure on accelerating the population bias toward susceptible strains. The resulting lower proportion of resistant cells in the population leads to greater lethality to carbenicillin. We further explored the origin of salicylate-mediated cost and found that the AcrAB-TolC efflux pump contributes to escalated fitness cost in MarA-mediated resistant strains.

Other factors that might affect competitive exclusion include the effective concentrations of salicylate and other environmental chemicals. The physiological concentrations of salicylate in human circulating plasma have been measured and vary between 0.1 μM and 0.5 mM (15), overlapping with values we tested here (Fig. 5). Further research may enable understanding of the interplay between salicylate and other chemicals and their combinatorial effects.

Our study indicates the potential for important interaction effects between environmental chemicals and population composition. Understanding the interplay between environmental chemicals and the cost of resistance may help to explain and predict the population dynamics of resistant bacteria in the environment.

Author Contributions

T.W. and M.J.D. conceived and designed the experiments. T.W. and C.K. performed the experiments and analyzed the data. T.W. and M.J.D. wrote the manuscript with input from C.K. All authors gave final approval for publication.

Editor: Nathalie Balaban.