Tobramycin Adaptation Enhances Policing of Social Cheaters in Pseudomonas aeruginosa

ABSTRACT

The Pseudomonas aeruginosa LasR-LasI (LasR-I) quorum sensing system regulates secreted proteases that can be exploited by cheaters, such as quorum sensing receptor-defective (lasR) mutants. lasR mutants emerge in populations growing on casein as a sole source of carbon and energy. These mutants are exploitative cheaters because they avoid the substantial cost of engaging in quorum sensing. Previous studies showed that quorum sensing increases resistance to some antibiotics, such as tobramycin. Here, we show that tobramycin suppressed the emergence of lasR mutants in casein-passaged populations. Several mutations accumulated in those populations, indicating evidence of antibiotic adaptation. We found that mutations in one gene, ptsP, increased antibiotic resistance and also pleiotropically increased production of a quorum sensing-controlled phenazine, pyocyanin. When passaged on casein, ptsP mutants suppressed cheaters in a manner that was tobramycin independent. We found that the mechanism of cheater suppression in ptsP mutants relied on pyocyanin, which acts as a policing toxin by selectively blocking growth of cheaters. Thus, tobramycin suppresses lasR mutants through two mechanisms: first, through direct effects on cheaters and, second, by selecting mutations in ptsP that suppressed cheating in a tobramycin-independent manner. This work demonstrates how adaptive mutations can alter the dynamics of cooperator-cheater relationships, which might be important for populations adapting to antibiotics during interspecies competition or infections.

IMPORTANCE The opportunistic pathogen Pseudomonas aeruginosa is a model for understanding quorum sensing, a type of cell-cell signaling important for cooperation. Quorum sensing controls production of cooperative goods, such as exoenzymes, which are vulnerable to cheating by quorum sensing-defective mutants. Because uncontrolled cheating can ultimately cause a population to collapse, much focus has been on understanding how P. aeruginosa can control cheaters. We show that an antibiotic, tobramycin, can suppress cheaters in cooperating P. aeruginosa populations. Tobramycin suppresses cheaters directly because the cheaters are more susceptible to tobramycin than cooperators. Tobramycin also selects for mutations in a gene, ptsP, that suppresses cheaters independent of tobramycin through pleiotropic regulation of a policing toxin, pyocyanin. This work supports the idea that adaptation to antibiotics can have unexpected effects on the evolution of quorum sensing and has implications for understanding how cooperation evolves in dynamic bacterial communities.

INTRODUCTION

Many Proteobacteria have quorum sensing systems that sense and respond to N-acyl-homoserine lactones (AHLs) to cause population density-dependent changes in gene expression (1–3). AHL systems involve a LuxR family signal receptor and a LuxI family signal synthase (1, 4, 5). LuxR-LuxI (LuxR-I) systems in many Proteobacteria control production of exoproducts such as proteases and toxins, which can be considered public goods (3, 6, 7). Public goods can be used by any member of the population; however, some population members do not contribute to public good production and are called freeloaders or cheaters (7–9). An example of a cheater is individuals with null mutations in the gene coding for the LuxR family signal receptor; these mutants do not produce quorum sensing-dependent public goods but can still exploit public goods produced by quorum sensing cooperators in the population (8–10).

Cheater proliferation presents a serious threat to the stability of cooperating populations. Because public goods are metabolically costly, cheaters can overrun cooperators (8, 9, 11). Under conditions where the public goods are required for growth, uncontrolled cheating can lead to population collapse (7, 12–14). Mechanisms must exist to control cheating in order for cooperative phenotypes to be maintained. One such mechanism is policing (15, 16). In bacteria, cooperators police cheaters by linking toxin and toxin resistance factor production, such as through coregulation of each of these by quorum sensing (16–18). Quorum sensing can also control cheating through coregulation of public goods with private goods (13) that benefit only producing cells. Previously, we showed that the soil bacterium Chromobacterium subtsugae (formerly Chromobacterium violaceum) uses quorum sensing to coregulate a publicly available secreted protease with a privately available cell-associated tetracycline-specific efflux pump (10). When C. subtsugae is grown on casein as a sole carbon and energy source, quorum sensing-dependent protease production can be exploited by quorum sensing-defective cheaters. However, these cheaters are suppressed when tetracycline is included in the casein medium, because they do not express the efflux pump conferring resistance (10). This type of cheater suppression likely requires some other selective force ensuring that public and private goods are maintained under coregulation (19).

In this study, we sought to determine if antibiotics can restrain the emergence of quorum sensing-defective cheaters in bacteria other than C. subtsugae. We were also interested in understanding how adaptation under antibiotic selection can alter the dynamics of cooperation and cheating. Previous reports show that AHL quorum sensing regulates antibiotic resistance in Pseudomonas aeruginosa (20–23). There are two AHL quorum-sensing systems in P. aeruginosa, the LasR-LasI (LasR-I) and RhlR-RhlI (RhlR-I) systems. These two systems produce and respond to the signals 3-oxododecanoyl-homoserine lactone (3OC₁₂-HSL) and butanoyl-homoserine lactone (C₄-HSL), respectively (3, 24, 25). The systems are hierarchical, with LasR-I controlling activation of RhlR-I (25, 26). Under biofilm conditions, deletions in LasR were previously shown to cause sensitivity to tobramycin antibiotic in at least one strain of P. aeruginosa, PAO1 (20–22).

Here, we show that the LasR-I system increases tobramycin resistance under planktonic conditions in the P. aeruginosa strain PA14. We also show that tobramycin can suppress the emergence of lasR mutant cheaters in cooperating PA14 populations grown on casein, similar to previous observations with C. subtsugae (10). We sequenced the genomes of isolates from tobramycin-evolved populations. All of the isolates had mutations in or upstream of ptsP, a gene coding for phosphoenolpyruvate-protein phosphotransferase (EI^Ntr). This enzyme is the first in a global regulatory system known as the nitrogen phosphotransferase system (PTS^Ntr) (27). Mutations in ptsP are known to increase tobramycin resistance (28, 29). Interestingly, we observed that ptsP mutations can also lead to suppression of cheaters, even in populations passaged with no antibiotic. We demonstrate that cheater suppression is due to increased production of the toxin pyocyanin in the ptsP mutants (30). Our results show that selection by tobramycin can lead to both direct and pleiotropic effects on cheating. These results provide new information on policing mechanisms in P. aeruginosa and demonstrate how antibiotic selection can lead to changes in cooperative activity.

RESULTS

LasR promotes tobramycin resistance in P. aeruginosa PA14 planktonic cultures.

As an initial test that LasR contributes to tobramycin resistance under planktonic conditions, we determined the MIC of tobramycin against the laboratory strain PA14 or against PA14 with a deletion of lasR or lasI. To increase the MIC detection sensitivity, we generated two tobramycin dilution series that were staggered by using two different starting concentrations (31). We observed a small but reproducible 1.7-fold decrease in the ΔlasR mutant MIC relative to that of PA14 (Fig. 1A). The difference was observed when tobramycin was used to treat cultures grown to an optical density at 600 nm (OD₆₀₀) of 4 but not cultures at lower cell densities (see Fig. S1 in the supplemental material). After 24 h of treatment with 1.1 μg ml⁻¹ tobramycin, about 200-fold fewer ΔlasR mutant cells were recovered than wild-type cells (Fig. 1B). We also observed a decrease in the MIC of the ΔlasI mutant similar to that of the ΔlasR mutant (Fig. 1A). We could restore resistance to the ΔlasI mutant by adding synthetic 3OC₁₂-HSL (Fig. 1A). There was no significant difference between the wild type and the ΔrhlR mutant, supporting the hypothesis that the RhlR-I system is not important for the resistance phenotype. Further, the MIC of the ΔlasR ΔrhlR double mutant was similar to that of the ΔlasR single mutant. Together, these results show that the LasR-I system, but not the RhlR-I system, contributes to tobramycin resistance in planktonically grown P. aeruginosa strain PA14.

FIG 1.

lasR contributes to tobramycin resistance in planktonically grown P. aeruginosa PA14. (A) The MIC of tobramycin was determined for each of the strains indicated as described in Materials and Methods. 3OC₁₂-HSL was added before inoculation where indicated (10 μM final concentration). Statistical analysis was by one-way analysis of variance (ANOVA) and Dunnett’s multiple-comparison test with the wild type: ***, P < 0.001; ****, P < 0.0001. (B) Cells recovered following tobramycin treatment at 1.1 μg ml⁻¹. Cells were treated as described for panel A, and the surviving cells were enumerated by serial dilution and plating following treatment. The difference in effective concentrations of tobramycin in panels A and B is likely due to the varying potencies of different tobramycin stocks used for each set of experiments. For both panels A and B, the values shown represent the average results of three independent experiments and the error bars represent the standard deviations.

Tobramycin suppresses cheating in P. aeruginosa.

Our results support the idea that tobramycin could limit the emergence of lasR-mutated cheaters in cooperating P. aeruginosa populations. P. aeruginosa requires a LasR-controlled protease to grow in minimal medium with casein as the sole source of carbon and energy (8, 9). We initially verified that our strain shows a similar dependence on LasR for growth on casein (Fig. S2A). Next, we serially passaged PA14 in casein broth and monitored the emergence of lasR mutant cheaters. We distinguished lasR mutants by colony phenotypes as described in Materials and Methods (32, 33), and for a subset of the identified lasR mutants, we confirmed the location of the mutation using Sanger sequencing of PCR amplicons (Table S1). In populations with no antibiotic, lasR mutants emerged between 5 and 8 days and increased to 98 to 99% of the population (Fig. 2). The rapid increase in frequency of lasR mutants in the population shows that mutating lasR provides a fitness advantage to these individuals and supports that they are acting as cheaters, as has been demonstrated by prior work in other P. aeruginosa strains (8, 9). We also observed that none of our populations showed evidence of a collapse, which is also consistent with prior work (8, 9). It has been proposed that cooperation is maintained so long as cheaters remain below a certain percentage in the population, such as by hydrogen cyanide-dependent policing (16). To test the role of tobramycin in cheater emergence, we carried out our passaging experiment in casein broth with tobramycin added at a sublethal concentration (0.6 μg ml⁻¹ tobramycin). In the populations with tobramycin, lasR mutant frequencies remained low or below the detection level throughout the experiment (Fig. 2). These results show that tobramycin can suppress the emergence of LasR-mutated cheaters in casein-grown P. aeruginosa populations.

FIG 2.

Tobramycin suppresses the emergence of lasR mutant cheaters. P. aeruginosa populations were transferred daily in 1% casein broth for 32 days, and cheaters were enumerated every 4 days by patching as described in Materials and Methods. For each experiment, one culture was initially started with no tobramycin, and after 72 h, it was split into three cultures propagated under one of three conditions: (i) with no tobramycin (circles), (ii) with tobramycin added at 0.6 μg ml⁻¹ (squares), or (iii) with tobramycin added at 0.6 μg ml⁻¹ initially and increased by 50% every 4 days to a final concentration of 7.1 μg ml⁻¹ (triangles). For conditions ii and iii, tobramycin was added every other day just after transfer to fresh medium. The detection level of cheaters was 1%. The values shown represent the average results of three independent experiments, and the error bars represent the standard deviations.

Variants from tobramycin-treated populations undergo genetic adaptation.

To assess how antibiotic adaptation influences cheater emergence and suppression, we carried out a second experiment using higher tobramycin concentrations for stronger selection. We used an initial concentration of 0.6 μg ml⁻¹ tobramycin and increased the concentration by 50% every 4 days to reach a final concentration of 7.1 μg ml⁻¹. In three independent cultures, the lasR mutant population remained at less than 5% of the total population (Fig. 2). We did not observe any significant growth inhibition at any stage of the experiment, even at the highest tobramycin concentration (Fig. S2B), suggesting that genetic adaptation occurred during passage. To test this hypothesis, we isolated one representative variant from each of the passaged populations (variants T1, T2, and T3 from the 0.6 to 7.1 μg ml⁻¹ tobramycin-passaged populations and variants T4, T5, and T6 from the 0.6 μg ml⁻¹ tobramycin-passaged populations) and determined the MICs. All six variants showed a higher tobramycin MIC than the ancestor strain (Fig. S3A) or variants from identically treated populations with no added antibiotic (clones N1 to N3) (Fig. S3B).

To identify the mutations that accumulated in tobramycin-evolved variants, we sequenced the genomes of our six tobramycin-evolved variants. For comparison, we also sequenced the parent PA14 strain and clone N2 described above, an isolate from a population passaged with no antibiotic. We identified 3 to 6 mutations in each of the tobramycin-evolved variants that were not in either the parent PA14 or isolate N2 (Table 1). Most of the tobramycin-evolved variants had mutations in two genes: ptsP, which codes for phosphoenolpyruvate protein phosphotransferase, and fusA1, which codes for translation elongation factor EF-G1A and is considered essential (35). To verify the role of ptsP and fusA1 mutations in tobramycin resistance, we introduced mutations of each into the PA14 genome. We used ΔptsP or the fusA1 G1643A mutation from isolate T5, because fusA1 deletions are thought to be nonviable (35). We also constructed a ΔptsP fusA1 G1634A double mutant. We compared the MICs of the mutated strains with that of the PA14 parent (Fig. S3C and D). Each of the individual mutations increased PA14 resistance by about 2-fold, and combining mutations increased the MIC by about 3-fold (Fig. S3E).

TABLE 1.

Mutations in variants from tobramycin-passaged populations

Varianta	Gene mutationb
	ptsP	fusA1	Other gene mutation(s)
T1	1547T	G61A	pmrB, mdpA
T2	841ΔC	NDc	mexZ, rrf2, psdR, PA14_RS15595
T3	1547T	G61A	dppA3, ostA
T4	G392228Ad	A1655G	psdR
T5	G392228Ad	G1634A	mdpA
T6	A392231Gd	ND	mdpA, pstS, ftsH, nuoM, Δ4064707–4092544

^aVariants isolated from populations passaged with tobramycin at 0.6 to 7.1 μg ml⁻¹ (T1 to T3) or 0.6 μg ml⁻¹ (T4 to T6).

^bBoldface indicates promoter mutations and large chromosomal deletions that are given by genomic location; all other gene mutations are given by nucleotide location.

^cND, not detected.

^dDesignation indicates DNA sequence in the predicted promoter of ygdP, which is immediately upstream of ptsP. ygdP and ptsP are predicted to be cotranscribed using the DOOR database for prokaryotic operons (75), accessed through the Pseudomonas Genome database website (76).

ptsP mutant strains have enhanced LasR activity and pyocyanin production.

We focused our attention on the ptsP mutation because disrupting ptsP was previously shown to increase LasR activity (30), and we were interested in understanding how these effects could alter the cooperator-cheating dynamic. First, we tested the hypothesis that the LasR-LasI system is elevated at least in the T1, T2, and T3 mutants. To do so, we used a PlasI-gfp plasmid reporter (34), which showed ∼2- to 3-fold higher fluorescence in the ΔptsP mutant than that of the wild-type PA14 strain (Fig. S4A), similar to previous results (30). The T1, T2, and T3 variants carrying this reporter also showed ∼3-fold higher fluorescence than that of PA14. The elevated fluorescence levels in these variants could be restored to wild-type levels by introducing ptsP to the neutral attB site in the genome (Fig. 3A). The reporter activities of isolates N1, N2, and N3 were similar to that of PA14 (Fig. S4B). The difference in reporter activity in the PA14 ΔptsP mutant or the T1 to T3 strains could not be explained by differences in growth (Table S2).

FIG 3.

Effects of ptsP inactivation on LasR activity and pyocyanin production. (A) Activity of LasR. P. aeruginosa strains were electroporated with a plasmid containing a LasR-responsive green fluorescent protein (GFP) reporter (pBS351). Reported values are fluorescence normalized to culture density at 18 h of growth. (B) Pyocyanin production. Cultures were inoculated into pyocyanin-producing medium, grown for 18 h, and extracted before quantification of pyocyanin as described in Materials and Methods. In all cases, reported values are micrograms of pyocyanin per milliliter normalized to culture density at the time of measurement. Strains carried either the empty CTX-1 cassette (CTX) or the CTX-1-ptsP cassette, inserted at the neutral attB site in the genome. The values represent the average results of three independent experiments, and the error bars represent the standard deviations. Statistical analysis was by one-way ANOVA and Dunnett’s multiple-comparison test with the wild type: **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

Disruptions in ptsP were previously shown to increase pyocyanin production by about 3-fold (30). To test whether pyocyanin production was similarly elevated in our tobramycin-evolved variants, we compared pyocyanin production by the T1, T2, and T3 variants with that of PA14. Our results showed that the variants have ∼3-fold-higher pyocyanin production than the PA14 parent and that pyocyanin production can be reduced to wild-type levels in the variants by introducing ptsP into the attB site in the genome (Fig. 3B). This increase was not observed in isolates N1 to N3 (Fig. S4C). Thus, ptsP mutations also increase pyocyanin production in strains passaged with tobramycin.

lasR mutant cheaters are suppressed in ptsP mutant populations.

Because ptsP mutations alter quorum sensing regulation, we predicted that these mutations might influence cheating dynamics even in the absence of tobramycin. Specifically, we predicted that lasR mutant cheaters might emerge at a higher rate in ptsP mutant populations due to the metabolic burden associated with higher LasR activity. To test this hypothesis, we passaged the T1, T2, and T3 variants in casein broth for 32 days in the absence of antibiotic and monitored the emergence of lasR mutant cheaters (Fig. 4A; Fig. S5). Surprisingly, cheaters emerged later in these populations than did those of the PA14 ancestor (23 to >32 days for T1 to T3 variants versus 6 to 8 days for PA14). For T2, cheaters were never observed even by 32 days. Further, the total cheater frequency in the T1 and T3 populations did not reach the same levels as that of PA14 (33 to 67% for T1 and T3 versus >98% for PA14) (Fig. S5A). There were no differences in cheating observed with variants N1 to N3 passaged with no tobramycin in comparison with PA14 (Fig. S5C). The total cell densities remained similar for all evolution experiments, suggesting that the differences in cheating were not due to differences in total population of the experiments (Fig. S5B and D).

FIG 4.

ptsP inactivation suppresses cheaters. P. aeruginosa populations were transferred daily in 1% casein broth for 32 days, and cheaters were enumerated every 4 days by patching as described in Materials and Methods. Values shown indicate the number of days until lasR mutant cheaters emerge in populations of wild-type PA14 (WT) and PA14 variants (T1, T2, and T3) from tobramycin-evolved populations (A) and the WT or the ΔptsP mutant carrying either the empty CTX-1 cassette (CTX), or the CTX-1-ptsP (CTX-ptsP) cassette, inserted at the neutral attB site in the genome (B). Cheater emergence is defined as when lasR mutants reach a frequency in the population of 8%, at which point it was generally observed that cheater frequencies continued increasing rather than decreasing or remaining constant. Full data sets of cheater frequency and total populations are shown in Fig. S5 in the supplemental material. Values represent the average results of three independent experiments, and the error bars represent the standard deviations. Statistical analysis was by one-way ANOVA and Dunnett’s multiple-comparison test with the wild type for panel A and by one-way ANOVA and Tukey’s multiple-comparison test for panel B: ****, P < 0.0001.

We hypothesized that the difference in cheating observed with the tobramycin-evolved variants was due to the disrupted ptsP gene in these strains. To test this hypothesis, we tested cheating in a ΔptsP mutant. We passaged the ΔptsP mutant carrying either CTX-1 or the CTX-1-ptsP cassette in 1% casein broth for 32 days and monitored cheater frequency over time. We also monitored cheating in PA14 and PA14 CTX-1. We observed no differences in cheating between PA14 CTX-1 and PA14, verifying that the CTX-1 cassette had no effect on cheating. However, cheaters emerged later and did not reach the same frequency in the ΔptsP-CTX-1 populations as they did with either PA14 strain (Fig. 4B; Fig. S5E). Cheating in PA14 was indistinguishable from that in the ΔptsP-CTX-1-ptsP populations (Fig. 4B; Fig. S5E). There were no notable differences in total growth of any of these populations (Fig. S5F). Together, our results show that cheating is suppressed in ptsP mutant populations compared with that of wild-type PA14. The observation that the lasR mutants are not completely suppressed in ptsP mutant populations suggests that there is still some fitness benefit associated with mutating lasR, although that benefit is reduced.

Pyocyanin produced by ΔptsP mutants is active against lasR mutant cheaters.

A potential explanation for delayed cheating in the ΔptsP mutant populations is that the ΔptsP mutant has a growth advantage over the PA14 ancestral strain. However, PA14 and the ΔptsP mutant showed identical growth rates (Table S2), suggesting that there is another explanation. We hypothesized that the ΔptsP mutant produces a toxin that inhibits growth of lasR mutant cheaters. To test this hypothesis, we filtered culture fluid from a ΔptsP mutant and tested its ability to inhibit growth of logarithmically growing PA14 or PA14 with deletions of ptsP, lasR, or both ptsP and lasR. We performed identical treatments with filtered fluid from PA14 or with unspent culture medium (untreated control) and measured cell growth after 24 h. Treatment with PA14 culture fluid had no effects on growth, and results were similar to that of the untreated control (Fig. 5). However, treatment with the ΔptsP mutant filtrate reduced growth of the ΔlasR mutant by 83-fold and that of ΔptsP ΔlasR mutants by 13,000-fold, compared with no treatment. Thus, the ΔptsP mutant secretes a substance that is growth inhibitive to lasR mutant strains and particularly to ΔptsP ΔlasR mutants.

FIG 5.

ΔptsP mutant cultures have antimicrobial activity against lasR mutants. Shown are final cell densities of P. aeruginosa strains treated for 24 h with filtered fluid from 1% casein-grown cultures of untreated (black), wild-type PA14 (light gray), or the ptsP mutant (dark gray). The initial cell density of the treated cultures was 5 × 10⁵ to 7 × 10⁵ cells ml⁻¹. The values represent the average results of three independent experiments, and the error bars represent the standard deviations. Statistical analysis was by two-way ANOVA and Dunnett’s multiple-comparison test with the untreated culture for each strain: *, P < 0.05; **, P < 0.01.

We hypothesized that the activity in the ΔptsP mutant culture fluid was due to pyocyanin. This hypothesis was based on our result that ptsP mutations increase pyocyanin production (Fig. 3) and prior work demonstrating that pyocyanin can be toxic to lasR mutants under certain conditions (17, 36). Pyocyanin is produced from the phzA1-G1 (phz1) and phzA2-G2 (phz2) gene clusters and the phzM and phzS genes (Fig. 6A) (37, 38). Several of the pyocyanin intermediates are toxic, such as 5-methylphenazine 1-carboxylic acid (5-Me-PCA) (39–41). We tested several gene mutations in this pathway to determine whether any of the products have activity against lasR mutants. We deleted phzM, which is needed to convert the precursor phenazine-1-carboxylic acid (PCA) to 5-Me-PCA, phzS, which is needed to convert 5-Me-PCA to pyocyanin (37, 42), and phzH, which is not needed for pyocyanin production but is involved in production of another product of this pathway, phenazine-1-carboxamide (PCN) (37).

FIG 6.

ΔptsP mutants police cheaters using pyocyanin. (A) Pyocyanin biosynthesis steps. The products of the phzA1-G1 (phz1) and phzA2-G2 (phz2) gene clusters synthesize phenazine-1-carboxylic acid (PCA) from chorismate. PhzM and PhzH convert PCA into 5-methylphenazine 1-carboxylic acid (5-Me-PCA) or phenazine-1-carboxamide (PCN), respectively. PhzS converts 5-Me-PCA into pyocyanin. (B) Final cell densities of ΔptsP ΔlasR cells treated with filtered fluid from 1% casein-grown cultures of a ΔptsP mutant or a ΔptsP mutant with the entire phz1 and phz2 operons deleted (Δphz1-2) or with phzM, phzS, or phzH deleted. The initial cell density of the treated cultures was 5 × 10⁵ to 7 × 10⁵ cells ml⁻¹. Data are the average results of three independent experiments with standard deviations. Statistical analysis was by one-way ANOVA and Tukey’s multiple-comparison test with the ΔptsP mutant: *, P < 0.05. (C) Final cell densities of wild-type PA14 (WT) or the ΔptsP ΔlasR mutant treated with 0 or 10 μg ml⁻¹ pyocyanin and filtered fluid from 1% casein-grown ΔptsP ΔphzS cultures. The initial cell density of treated cultures was 4 × 10⁵ to 6 × 10⁵ CFU ml⁻¹. Data are the average results of four independent experiments with standard deviations. Statistical analysis was by two-way ANOVA and multiple-comparison test with no-pyocyanin control: *, P < 0.05. (D) Cheater suppression during competition in coculture. Cocultures were inoculated with the lasR mutant and each cooperator strain at an initial ratio of 1:99 (cheater to cooperator) in 1% casein broth and transferred to fresh medium daily for 3 days. On day 3, cheaters were enumerated by patching as described in Materials and Methods. Each data point represents an independent experiment. The horizontal line represents the mean, and the vertical line represents the standard deviation of all the experiments in each set. Statistical analysis was by one-way ANOVA with Tukey’s multiple-comparison test of each set with the wild-type cooperator experiments: **, P < 0.01; ***, P < 0.001. Total population densities of each experiment are shown in Fig. S6 in the supplemental material.

We initially tested whether pyocyanin biosynthesis contributes to the antimicrobial activity of a ptsP mutant by making a ΔptsP Δphz1 Δphz2 triple mutant. We compared the antimicrobial activity in culture fluid of this strain with that of the single ΔptsP mutant and tested it against the ΔptsP ΔlasR mutant. Deleting both phz1 and phz2 abolished the antimicrobial activity observed with the ΔptsP single mutant (Fig. 6B). Mutations in phzM or phzS similarly abolished this activity (Fig. 6B). However, no effects were observed by deleting phzH (Fig. 6B), which is important for production of the pyocyanin biproduct PCN but not pyocyanin itself. These results support that the antimicrobial activity in ΔptsP mutant culture fluid is due to pyocyanin. We also tested whether commercially supplied pyocyanin could restore antimicrobial activity of the ΔptsP ΔphzS mutant by adding 10 μg ml⁻¹ pyocyanin to ΔptsP ΔphzS culture filtrates prior to treating ΔptsP ΔlasR cells. Adding pyocyanin reduced ΔptsP ΔlasR mutant growth by ∼10⁷-fold, whereas PA14 growth was only reduced by ∼10-fold (Fig. 6C). Together, these results support the conclusion that pyocyanin can selectively limit growth of ΔptsP ΔlasR mutants.

Pyocyanin production leads to enhanced policing by ΔptsP mutants.

We also tested the role of pyocyanin in policing cheaters using competition experiments. We mixed the ΔlasR mutant with a cooperator starting at a 1:99 (cheater-to-cooperator) ratio in 1% casein medium and transferred the population daily for 3 days. The results are shown in Fig. 6D. In cocultures with PA14 as the cooperator, ΔlasR mutants increased from an initial frequency of 1% to a final frequency of 56%. However, in identical competition experiments with the ΔptsP mutant, the ΔlasR mutants increased from 1% to only 16% final frequency, consistent with limited cheater proliferation observed with the ΔptsP mutant (Fig. 4B). Deleting phzM in PA14 did not alter the proliferation of ΔlasR mutants from that of PA14, demonstrating that pyocyanin is not important for policing ΔlasR mutant cheaters in PA14 under the conditions of our experiment. However, deleting phzM in the ΔptsP mutant caused the final frequency of ΔlasR mutants to increase to 88%. Thus, pyocyanin is responsible for cheater suppression in the ΔptsP mutant. We find it interesting that ΔlasR cheaters increase to a higher frequency in competition experiments with a ΔptsP ΔphzM mutant than with the ΔphzM mutant or the PA14 parent (88% versus ∼50% final ΔlasR cheater frequency). It is possible that deleting phzM in a ΔptsP mutant causes some deleterious effect on fitness that gives the ΔlasR mutants an additional advantage. Together, our results support that the cheater suppression observed in the ΔptsP mutant is dependent on pyocyanin.

DISCUSSION

Previous studies have demonstrated that P. aeruginosa quorum sensing plays a small but appreciable role in antibiotic resistance (20–23). Here, we confirm the role of quorum sensing in antibiotic resistance and also show that antibiotics can stabilize quorum sensing by directly suppressing cheaters and also by selecting mutations in ptsP that have pleiotropic effects on cheating. The ptsP mutations cause antibiotic-independent cheater suppression by enhancing production of the policing toxin, pyocyanin. Importantly, the results demonstrate how tobramycin adaptation can have pleiotropic effects on cheating. Together, these results support the idea that cheater control mechanisms can change under selection by environmental pressures such as antibiotics.

Mutations in ptsP were genetic adaptations to the antibiotic tobramycin. Adaptive mutations have been shown to influence cooperation and cheating in other studies (43–45). For example, adaptive mutations can improve the fitness of cooperators so they can outcompete cheaters (19, 44) or cause rewiring of cooperative traits and allow the cheaters to cooperate again (43, 52). In this study, adaptive mutations in ptsP were selected by tobramycin. The mutations in ptsP had pleiotropic effects on pyocyanin production that led to policing of cheaters. Our results show how adaptation to antibiotics could have important effects on cooperation and cheating. In natural settings, adaptation could be much more complex. For example, populations may encounter multiple antibiotics or other stressors or face competition from other species for nutrients. In addition, the adaptive mutations might not be homogeneous throughout the population, for example, if there is spatial structure created by the formation of biofilms.

In our study, cheater suppression was caused by increased production of the toxin pyocyanin. Pyocyanin was previously shown to be involved in policing lasR mutant cheaters (17, 36). Pyocyanin leads to the generation of highly toxic hydroxyl radicals (46–48), which can cause oxidative stress and cell death (49, 50). LasR mutant cheaters have been shown to be more susceptible to pyocyanin (51), possibly because they fail to upregulate enzymes involved in relieving pyocyanin-induced oxidative stress such as catalase and superoxide dismutase (SOD) (17, 47, 49, 51). Detoxifying enzymes produced by cooperators can also have some benefits to the cheaters (19), which may explain how lasR cheaters were still able to increase in frequency in ptsP mutant populations (Fig. 6). It will be interesting to determine whether pyocyanin production could have similar effects on stabilizing quorum sensing in natural communities, such as infections.

Mutations in ptsP have been previously shown to increase pyocyanin production (30). PtsP, along with two other enzymes, PtsN and PtsO, make up a poorly understood system called the nitrogen phosphotransferase system (PTS^Ntr). In Escherichia coli, PTS^Ntr is involved in regulating diverse physiological changes in response to nitrogen starvation (53, 54). It is unclear if PTS^Ntr plays a similar role in P. aeruginosa. In P. aeruginosa, PtsP has been shown to contribute to pathogenesis (55), biofilm formation (27), and quorum sensing regulation (30). Mutations in ptsP have also been shown to contribute to tobramycin resistance (28, 29), consistent with our results. The mechanistic pathway by which PtsP contributes to pathogenesis, biofilm formation, and tobramycin resistance is as yet unknown. In the case of quorum sensing, it is thought that PtsP somehow represses LasR through the antiactivator QscR (30), although it is not clear if PtsP acts entirely through QscR or modifies quorum sensing through other pathways. The important effects of mutating PtsP on P. aeruginosa virulence and virulence-associated behaviors suggest that PtsP and the PTS^Ntr system might have potential as a new target for therapeutic development.

Our results also show that quorum sensing makes a small but appreciable contribution to antibiotic resistance in P. aeruginosa strain PA14 under planktonic conditions (Fig. 1). Similar results have been reported for other strains and species, for example, in P. aeruginosa PAO1 in biofilms (20–23) and in C. subtsugae (10). In C. subtsugae, resistance is attributed to an efflux pump, CdeAB-OprA, which is transcriptionally activated by the CviR quorum sensing receptor in response to the cognate quorum sensing signal N-hexanoyl-homoserine lactone (10, 31). In P. aeruginosa, there may be multiple factors contributing to antibiotic resistance. There are at least three efflux pumps known to have overlapping specificity for aminoglycosides: MexAB (56, 57), MexXY (58, 59), and PA1874-1877 (60). There are also aminoglycoside-inactivating enzymes (56). Quorum control of antibiotic resistance may provide an important evolutionary benefit. For example, quorum sensing may increase resistance to protect against self-produced toxins or synchronize resistance factor expression across members of the population (61) to protect neighboring cells from exported antibiotic (62). Understanding how and why quorum sensing contributes to antibiotic resistance will provide important new information about the biology of quorum sensing and will be relevant to designing new therapies that function by blocking quorum sensing systems.

MATERIALS AND METHODS

Culture conditions and reagents.

Bacteria were routinely grown in Luria-Bertani broth (LB) or LB buffered to pH 7 with 50 mM 3-(morpholino)-propanesulfonic acid (MOPS) or on LB agar (LBA; 1.5% [wt/vol] Bacto agar). Growth media for specific experiments were M9-caseinate (casein broth; 6 g liter⁻¹ Na₂HPO₄, 3 g liter⁻¹ KH₂PO₄, 0.5 g liter⁻¹ NaCl, 1 g liter⁻¹ NH₄Cl, pH 7.4, 1% sodium caseinate), MOPS minimal medium [25 mM d-glucose, freshly prepared 5 μM FeSO₄, 15 mM NH₄Cl, and 2 mM K₂HPO₄ added to a 1× MOPS base buffer consisting of 50 mM MOPS, 4 mM Tricine, 50 mM NaCl, 1 mM K₂SO₄, 50 μM MgCl₂, 10 μM CaCl₂, 0.3 μM (NH₄)₆Mo₇O₂₄, 40 μM H₃BO₃, 3 μM cobalt(II) acetate, 1 μM CuSO₄, 8 μM MnSO₄, and 1 μM ZnSO₄] (63), a modified M9-Casamino Acids broth (200 ml liter⁻¹ 10× M9, 0.6 g liter⁻¹ thiamine hydrochloride, 1 ml liter⁻¹ 1 M MgSO₄, 1 ml liter⁻¹ 0.2 M CaCl₂, 10 g liter⁻¹ Casamino Acids; freshly prepared and filter sterilized), pyocyanin producing medium (PPM) (64), and 4% skim milk agar (SMA) (9). All P. aeruginosa broth cultures were grown at 37°C with shaking at 250 rpm, in 18-mm test tubes (for 2-ml cultures), 125-ml baffled flasks (10-ml cultures), or 250-ml baffled flasks (60-ml cultures), unless otherwise specified. For E. coli, 100 μg ml⁻¹ carbenicillin, 15 μg ml⁻¹ gentamicin, and 10 μg ml⁻¹ tetracycline were used. For P. aeruginosa, 300 μg ml⁻¹ carbenicillin, 50 to 200 μg ml⁻¹ gentamicin, and 200 μg ml⁻¹ tetracycline were used. AHLs were purchased from Cayman Chemicals (MI, USA) and handled as described previously (10). Genomic or plasmid DNA was extracted using a Qiagen Puregene core A kit (Hilden, Germany) or an IBI Scientific plasmid purification miniprep kit (IA, USA), while PCR products were purified using IBI Scientific PCR cleanup/gel extraction kits, according to the manufacturer’s protocol. All antibiotics were purchased from GoldBio (MO, USA) except for tetracycline, which is from Fisher Scientific (PA, USA). Pyocyanin and dimethyl sulfoxide (DMSO; solvent for pyocyanin) were purchased from Sigma-Aldrich (MO, USA) and Acros Organics (WI, USA), respectively.

Bacterial strains and strain construction.

All bacterial strains, plasmids, and primers used in this study are listed in Tables 2 to 5. P. aeruginosa strain UCBPP-PA14 (PA14) (65) and PA14 derivatives were used for these studies. Markerless deletions in specific loci of P. aeruginosa PA14 were generated using allelic exchange as described previously (66). To generate plasmids for allelic exchange, DNA fragments with the mutated or deleted gene allele plus 500 bp flanking DNA were generated by PCR using primer-incorporated restriction enzyme sites. The PCR product was digested and ligated to pEXG2 (fusA1 G1634A) or pEX18Ap (ΔlasR and ΔrhlR) and transformed into the appropriate P. aeruginosa strain. The plasmids for ΔlasI and ΔptsP mutants were described elsewhere (27, 67). Merodiploids were selected on pseudomonas isolation agar (PIA)-carbenicillin (300 μg ml⁻¹) for ΔlasR and ΔrhlR mutants, PIA-gentamicin (200 μg ml⁻¹) for ΔlasI and ΔptsP mutants, and LBA-gentamicin (50 μg ml⁻¹) for the fusA1 G1634A mutant. Deletion mutants were counterselected using NaCl-free 15% sucrose. Putative mutants were verified through antibiotic sensitivity tests and gene-targeted Sanger sequencing. Plasmid transformations were described previously (27, 34, 68, 69). Complementation strains were constructed by integrating the mini‐CTX‐1 vector at the neutral chromosomal attB locus (27, 69).

TABLE 2.

P. aeruginosa and E. coli strains used in this study

Strain	Relevant properties	Reference or source
P. aeruginosa
UCBPP-PA14 (PA14)	Ancestral wild type	65
PA14 ΔlasR	PA14 with a deletion of lasR	This study
PA14 ΔrhlR	PA14 with a deletion of rhlR	This study
PA14 ΔlasR ΔrhlR	PA14 ΔlasR with a deletion of rhlR	This study
PA14 ΔlasI	PA14 with a deletion of lasI	This study
PA14 ΔptsP	PA14 with a deletion of ptsP	This study
PA14 ΔptsP ΔlasR	PA14 ΔptsP with a deletion of lasR	This study
PA14 ΔphzM	PA14 with a deletion of phzM	39
PA14 ΔphzM ΔptsP	PA14 ΔphzM with a deletion of ptsP	This study
PA14 ΔphzS	PA14 with a deletion of phzS	39
PA14 ΔptsP ΔphzS	PA14 ΔphzS with a deletion of ptsP	This study
PA14 Δphz1 Δphz2	PA14 with a deletion of phz1 and phz2	82
PA14 Δphz1 Δphz2 ΔptsP	PA14 Δphz1 Δphz2 with a deletion of ptsP	This study
PA14 ΔphzH	PA14 with a deletion of phzH	83
PA14 ΔphzH ΔptsP	PA14 ΔphzH with a deletion of ptsP	This study
PA14 fusA1 G1634A	PA14 with the fusA1 G1634A mutation	This study
PA14 ΔptsP fusA1 G1634A	PA14 ΔptsP with the fusA1 G1634A mutation	This study
PA14 ΔptsP CTX-1-ptsP	PA14 ΔptsP with attB::CTX-1-ptsP insertion	This study
PA14 ΔptsP CTX-1	PA14 ΔptsP with attB::CTX-1 insertion	This study
PA14 CTX-1	PA14 with attB::CTX-1 insertion	This study
Escherichia coli
DH5α	F− ϕ80lacZΔM15 Δ(lacZYA-argF)U169 hsdR17(rK− mK+) recA1 endA1 phoA supE44 thi-1 gyrA96 relA1 λ−	Invitrogen
S17-1	recA pro hsdR RP4-2-Tc::Mu-km::Tn7	77
SM10	thi thr leu tonA lacY supE recA::RP4-2-Tc::Mu Km λpir	77
Rho3	thi-1 thr-1 leuB26 tonA21 lacY1 supE44 recA integrated RP4-2 Tcr::Mu (λpir +) Δasd::FRT ΔaphA::FRT	78

TABLE 3.

P. aeruginosa PA14 variants

Strain	Relevant properties	Reference or source
Variants isolated after daily transfer for 32 days in 1% casein
T1	Isolate from expt with tobramycin added at 0.6 to 7.1 μg ml−1	This study
T2	Isolate from expt with tobramycin added at 0.6 to 7.1 μg ml−1	This study
T3	Isolate from expt with tobramycin added at 0.6 to 7.1 μg ml−1	This study
T4	Isolate from expt with tobramycin added at 0.6 μg ml−1	This study
T5	Isolate from expt with tobramycin added at 0.6 μg ml−1	This study
T6	Isolate from expt with tobramycin added at 0.6 μg ml−1	This study
N1	Isolate from expt with no added tobramycin	This study
N2	Isolate from expt with no added tobramycin	This study
N3	Isolate from expt with no added tobramycin	This study
Modified variants
T1 CTX-1	T1 ΔptsP with attB::CTX-1 insertion	This study
T1 CTX-1-ptsP	T1 ΔptsP with attB::CTX-1-ptsP insertion	This study
T2 CTX-1	T2 ΔptsP with attB::CTX-1 insertion	This study
T2 CTX-1-ptsP	T2 ΔptsP with attB::CTX-1-ptsP insertion	This study
T3 CTX-1	T3 ΔptsP with attB::CTX-1 insertion	This study
T3 CTX-1-ptsP	T3 ΔptsP with attB::CTX-1-ptsP insertion	This study

TABLE 4.

Plasmids used in this study

Plasmid	Relevant properties	Reference or source
pEX18Ap	Suicide vector, Apr	79
pDA8	pEX18Ap containing ΔlasR with flanking sequences, Apr	D. An and M. Parsek, unpublished
pDA9	pEX18Ap containing ΔrhlR with flanking sequences, Apr	D. An and M. Parsek, unpublished
pEXG2	Suicide vector, Gmr	80
pEXG2-ΔlasI	pEXG2 containing ΔlasI with flanking sequences, Gmr	67
pEXG2-ptsP	pEXG2 containing ΔptsP with flanking sequences, Gmr	27
pEXG2-fusA1 G1634A	pEXG2 containing G1634A mutation in fusA1, Gmr	This study
pPROBE-GT	Broad-host-range pVS1/p15a GFP reporter, Gmr	81
pBS351	pPROBE-GT with −1 through −501 5′ region of lasI, Gmr	34
CTX-1	mini-CTX-1, P. aeruginosa integrative plasmid, Tetr	69
CTX-1-ptsP	CTX-1 containing the ptsP gene, Tetr	27

TABLE 5.

Primers used in this study

Primer	Sequencea
lasR-A	GCGCGCGAATTCAACATGGTCACCTCCAGCA
lasR-B	GCTGAGAGGCAAGATCAGAGCTGCAGCATAGCGCTACGTTCTTCTT
lasR-C	GCAAGAAGAACGTAGCGCTATG CTGCAGCTCTGATCTTGCCTCTCA
lasR-D	GCGCGCAAGCTTGTTACCGTCACCAGCGTCT
rhlR-A	GCGCGCGAATTCGCTGTTCGACGGCAGTAT
rhlR-B	GCGCGTCGAACTTCTTCTGGATCTGCAGCATTGCAGTAAGCCCTGA
rhlR-C	GCTCAGGGCTTACTGCAATGCTGCAGATCCAGAAGAAGTTCGACGC
rhlR-D	GCGCGCAAGCTTCGGACCGCAGAGAGACTA
rgaoligo74b	TAATAAAAGCTTCTCGGTGAAAGGCAAGAAAGA
rgaoligo75b	TAATAATCTAGAGCTTCGGCGTATTTGGAGAA

^aBold text indicates restriction sequences.

^brgaoligo74 and rgaoligo75 were used to incorporate the fusA1 G1634A mutation. Other oligonucleotides were used for lasR or rhlR deletions as indicated by their name.

Evolution experiments.

To prepare the inoculum for evolution experiments, overnight (18-h) pure cultures were grown in LB-MOPS, diluted 1:50 into 2 ml LB-MOPS, and grown to an OD₆₀₀ of ∼3.5. To start the experiment, 50 μl from this starter culture was transferred to 2 ml fresh casein broth in an 18-mm tube. At 24-h intervals, cultures were diluted 1:50 into fresh casein broth in a new tube. Tobramycin was added every other day where indicated, similar to previous experiments with C. subtsugae (10). The number of CFU ml⁻¹ for each lineage was determined by viable plate counts every 96 h. The percentage of lasR mutant cheaters (lasR cheater) was determined by patching 100 colonies, unless otherwise specified, on SMA. LasR cheaters form flattened colonies with an iridescent, metallic sheen surface and decreased skim milk proteolysis in SMA (9, 32, 33). Sixteen lasR mutants were sequence verified by Sanger sequencing (see Table S1 in the supplemental material).

Whole-genome sequencing.

Genomic DNA was extracted using the Qiagen Puregene yeast/bacteria kit, and a sequencing library was constructed with 350-bp inserts (strain T2) or 200-bp inserts (all other strains). Sequencing was performed using an Illumina HiSeq 4000 (for strain T2) or Illumina MiSeq system with ∼25× coverage (all other strains). The raw reads were aligned to the P. aeruginosa UCBPP-PA14 reference genome (UCBPP-PA14 GenBank accession no. NC_008463) by use of Strand NGS (Bangalore, India) software v3.1.1, using a pipeline described previously (70). Mutations of interest were verified by gene-targeted Sanger sequencing.

Antimicrobial susceptibility assays.

Tobramycin susceptibility was determined by MIC according to the 2020 guidelines of the Clinical and Laboratory Standards Institute (CLSI) (84), using a modified dilution method. Briefly, tobramycin was added to MOPS minimal medium and successively diluted 2-fold in a 200-μl volume in 2-ml tubes. For each experiment, two dilution series were staggered by starting them at different tobramycin concentrations to cover a broader range of concentrations. The MICs were determined as follows. We prepared P. aeruginosa starter cultures by growing them in LB to an OD₆₀₀ of 3.2 to 4 (Fig. 1) or as indicated in Fig. S1 in the supplemental material. The starter cultures were then diluted 1:40 into each tube containing tobramycin to start the MIC experiment. The inoculated tubes were incubated with shaking for 20 h. After incubation, turbidity was measured using a BioTek Synergy 2 plate reader. The MIC was defined as the lowest concentration of tobramycin (μg ml⁻¹) in which bacterial growth was not measurable. In some cases, the number of CFU ml⁻¹ was also determined by viable plate counts.

To determine susceptibility to P. aeruginosa culture fluid, we prepared culture fluid by inoculating overnight cultures to an OD₆₀₀ of 0.1 into 60 ml casein broth, grew these cultures for 20 h, and then passed the cultures through a 0.22-μm-pore-size filter to remove cells. The filtered fluid was mixed with 100 μl M9-Casamino Acids broth to a final volume of 2 ml, and this was inoculated at an initial OD₆₀₀ of 0.001 with either the wild type or ΔlasR, ΔptsP, or ΔptsP ΔlasR mutant from logarithmic-stage LB-MOPS cultures at an OD₆₀₀ of 0.2 to 0.6. The M9-Casamino Acids-filtrate mixture was incubated for 24 h, and the initial and final population counts (CFU ml⁻¹) were enumerated by colony counting on plates.

Measurements of LasR activity and pyocyanin production.

To measure LasR activity, we first introduced the LasR-responsive plasmid pBS351 into P. aeruginosa mutants or wild-type PA14 strains by electroporation (34). Electrocompetent cells were prepared from overnight cultures using 300 mM sucrose (71). Transformants were selected on LB agar using gentamicin at 50 to 200 μg ml⁻¹ and routinely grown with gentamicin (50 μg ml⁻¹ for agar and 15 μg ml⁻¹ for broth) for plasmid maintenance. Transformants were grown in LB-MOPS with 15 μg ml⁻¹ gentamicin for 18 h and washed with phosphate-buffered saline (PBS), and fluorescence was measured using a BioTek Synergy 2 plate reader. For pyocyanin measurements, we extracted pyocyanin as described previously (72–74). Briefly, cells were grown for 18 h in pyocyanin-producing medium (64), and 5 ml whole culture was extracted with 2 ml chloroform. The organic layer was separated and extracted a second time with 0.2 N HCl. The absorbance of the aqueous layer was measured at 520 nm and multiplied by 17.072 to calculate the pyocyanin concentration (μg ml⁻¹) (73). LasR activity and pyocyanin measurements were normalized to culture density (optical density at 600 nm) for reporting data.

Coculture assays.

Coculture experiments were conducted in 2 ml casein broth in 18-mm test tubes. To prepare the inoculum, overnight (18 h) pure cultures were grown in LB-MOPS, diluted to an OD₆₀₀ of 0.025 for cheaters or 0.05 for cooperators into LB-MOPS, and grown to an OD₆₀₀ of ∼3.5 before combining them at a 99:1 (cooperator-to-cheater) ratio. This mixture was then diluted 1:40 into casein broth to start the coculture. Cocultures were diluted 1:40 into fresh casein broth in a new tube every 24 h until the end of the experiment at 72 h. The initial and final total population counts (CFU ml⁻¹) were determined by viable plate counts. The percentage of lasR cheaters was determined by patching 200 colonies on SMA.

Data availability.

Sequence reads for the ancestral strain PA14 (SAMN16823471) and the tobramycin-evolved isolates (SAMN16823472 to SAMN16823478) can be found in the NCBI Sequence Read Archive under BioProject PRJNA678537.