Fig 1.
Antibiotic dose response curves for P. aeruginosa PAO1.
We exposed PAO1 to all four antibiotics in two experimental media: CAA+Tf (iron-limited casamino acids medium with transferrin) and CAS (casein medium). Except for meropenem, higher concentrations of antibiotics were required to inhibit PAO1 in CAS compared with CAA+Tf. Dots show means ± standard error across six replicates. All data are scaled relative to the drug-free treatment. Data stem from two independent experiments using different dilution series. The red dots indicate the highest concentration used for the respective experiments, from which 7 serial dilution steps were tested. Curves were fitted with either log-logistic functions (in CAA+Tf) or with three-parameter Weibull functions (in CAS). The underlying data for this figure can be found at https://doi.org/10.6084/m9.figshare.12515364.
Fig 2.
Antivirulence dose response curves for P. aeruginosa PAO1 (growth and virulence factor production).
We exposed PAO1 to the antivirulence compounds gallium (inhibiting pyoverdine-mediated iron uptake) and furanone C-30 (blocking QS response, including protease production) both in media in which the targeted virulence factors are expressed and required (iron-limited CAA+Tf medium for gallium and CAS medium for furanone) and in control media, in which the targeted virulence factors are not required (iron-supplemented CAA+Fe medium for gallium and protein digested CAA for furanone). (A) Dose-response curves for growth show that both antivirulence compounds reduced bacterial growth, but more so in media in which the targeted virulence factor is expressed. This demonstrates that there is a concentration window where the antivirulence compounds have no toxic effects on bacterial cells and just limit growth due to virulence factor quenching. (B) Dose-response curves for virulence factor production show that gallium and furanone C-30 effectively inhibit pyoverdine and protease production, respectively, in a concentration-dependent manner. Dots show means ± standard errors across six replicates. All data are scaled relative to the drug-free treatment. Data stem from two independent experiments using different dilution series. The red dots indicate the highest concentration used for the respective experiments, from which 7 serial dilution steps were tested. Curves were fitted with either log-logistic functions (in CAA+Tf) or with three-parameter Weibull functions (in CAS). The underlying data for this figure can be found at https://doi.org/10.6084/m9.figshare.12515364. CAA, casamino acid medium; CAS, casein medium; QS, quorum sensing; Tf, human apo-transferrin.
Fig 3.
Dose-response curves for P. aeruginosa PAO1 under antibiotic-antivirulence combination treatments.
Dose-response curves for growth and virulence factor production for PAO1 were assessed for 9×9 drug concentration matrixes involving the four antibiotics combined with either gallium (A) or furanone C-30 (B). Experiments were carried out in media in which the corresponding virulence factors are required for growth (pyoverdine: CAA+Tf; protease: CAS). Growth and virulence factor production were measured after 48 hours. All values are scaled relative to the untreated control, and data points show the mean across 12 replicates from two independent experiments. We used spline functions to fit the dose-response curves. The underlying data for this figure can be found at https://doi.org/10.6084/m9.figshare.12515364. CAA, casamino acid medium; CAS, casein medium; Tf, human apo-transferrin.
Fig 4.
Drug interaction heatmaps for antibiotic-antivirulence combination treatments.
We used the Bliss independence model to calculate the degree of synergy for every single drug combination with regard to growth suppression and virulence factor quenching shown in Fig 3. Heatmaps depicting variation in drug interactions ranging from antagonism (blue) to synergy (red) are shown for gallium-antibiotic combinations (A-D for growth; E-H for pyoverdine production) and furanone-antibiotic combinations (I-L for growth; M-P for protease production). All calculations are based on 12 replicates from two independent experiments. The underlying data for this figure can be found at https://doi.org/10.6084/m9.figshare.12515364.
Fig 5.
Effect of combination treatment on growth of AtbR clones.
Test of whether the addition of gallium (A) or furanone (B) can restore growth suppression in AtbR clones (in orange) relative to the susceptible wild-type (WT; in black). Under antibiotic treatment and in the absence of antivirulence compounds, all AtbR clones grew significantly better than the WT (two-sample t tests, −26.63 ≤ t21–40 ≤ −3.03, p < 0.01 for all treatment combinations; n.s. = nonsignificant; *p < 0.05; **p < 0.01; ***p < 0.001), a result that holds for both scaled (as shown above) and absolute growth. In the presence of antivirulence compounds (upper series of panels without antibiotics [−Atb]; lower series of panels with antibiotics [+Atb]), growth suppression was restored in six out of eight cases. The exceptions were the ciprofloxacin-furanone and meropenem-furanone combinations. The bottom series of panels shows the degree of drug synergy for the WT and the AtbR clones. All cell density values (measured with flow cytometry as number of events detected in 5 μL of culture, after 24 hours) are scaled relative to the untreated control. All data are shown as means ± standard errors across a minimum of 16 replicates from 4 to 6 independent experiments. The underlying data for this figure can be found at https://doi.org/10.6084/m9.figshare.12515364. AtbR clones, antibotic resistant clones.
Fig 6.
Effect of combination treatment on the relative fitness of AtbR clones.
Test of whether antivirulence compounds alone or in combination with antibiotics can abrogate or revert selection for antibiotic resistance. All AtbR clones were competed against the susceptible WT for 24 hours, starting at a 1:9 ratio. The dashed lines denote fitness parity, where none of the competing strains has a fitness advantage. In the absence of any treatment, all AtbR clones showed a fitness disadvantage (fitness values < 0) compared to the WT, demonstrating the cost of resistance. When treated with antivirulence compounds alone, the AtbR clones did not experience any selective advantage in 14 out of 16 cases (exception: colistin-gallium combinations). When treated with antibiotics alone, all AtbR clones experienced significant fitness advantages (fitness values > 0), as expected. When antivirulence compounds were added as adjuvants to antibiotics, the fitness advantage of AtbR clones was reduced, abrogated, or reversed for six out of eight drug combinations. All data are shown as means ± standard errors across a minimum of 16 replicates from 4 to 7 independent experiments. Significance levels are based on t tests or ANOVAs: n.s. = non-significant; *p < 0.05; **p < 0.01; ***p < 0.001. See S1 Table for full details on the statistical analyses. The underlying data for this figure can be found at https://doi.org/10.6084/m9.figshare.12515364. - Atb, treatments without antibiotics; + Atb, treatments with antibiotics; AtbR clones, antibiotic resistant clones; CAA, casamino acid medium; CAS, casein medium; Tf, human apo-transferrin; WT, wild-type.
Table 1.
List of mutations in the AtbR clones.