Alternative Evolutionary Paths to Bacterial Antibiotic Resistance Cause Distinct Collateral Effects

Abstract

When bacteria evolve resistance against a particular antibiotic, they may simultaneously gain increased sensitivity against a second one. Such collateral sensitivity may be exploited to develop novel, sustainable antibiotic treatment strategies aimed at containing the current, dramatic spread of drug resistance. To date, the presence and molecular basis of collateral sensitivity has only been studied in few bacterial species and is unknown for opportunistic human pathogens such as Pseudomonas aeruginosa. In the present study, we assessed patterns of collateral effects by experimentally evolving 160 independent populations of P. aeruginosa to high levels of resistance against eight commonly used antibiotics. The bacteria evolved resistance rapidly and expressed both collateral sensitivity and cross-resistance. The pattern of such collateral effects differed to those previously reported for other bacterial species, suggesting interspecific differences in the underlying evolutionary trade-offs. Intriguingly, we also identified contrasting patterns of collateral sensitivity and cross-resistance among the replicate populations adapted to the same drug. Whole-genome sequencing of 81 independently evolved populations revealed distinct evolutionary paths of resistance to the selective drug, which determined whether bacteria became cross-resistant or collaterally sensitive towards others. Based on genomic and functional genetic analysis, we demonstrate that collateral sensitivity can result from resistance mutations in regulatory genes such as nalC or mexZ, which mediate aminoglycoside sensitivity in β-lactam-adapted populations, or the two-component regulatory system gene pmrB, which enhances penicillin sensitivity in gentamicin-resistant populations. Our findings highlight substantial variation in the evolved collateral effects among replicates, which in turn determine their potential in antibiotic therapy.

Keywords: Pseudomonas aeruginosa; antibiotic resistance; collateral sensitivity; experimental evolution; trade-offs.

Fig. 1.

Directional selection of highly resistant P. aeruginosa. (A) Illustration of the experimental design used for the selection of resistant populations. Twenty replicate populations for each of the eight included antibiotics (table 1) and a control without antibiotic (a total of 180 populations) were serially transferred every 12 h into fresh medium and, for the drug treatments, increasing concentrations of each drug. Selection was initiated at 0.5 times the concentration inhibiting >90% of growth (IC90) and concluded at ∼40 times of the IC90. (B–I) Ten populations for each antibiotic were subsequently evaluated for their growth on different concentrations of the drug experienced during the experiment. Dose-response curves are shown in the left panels and IC90 fold changes in the right panels. The ten replicate populations are shown in different colors, whereas the black line represents the ancestral P. aeruginosa PA14.

Fig. 2.

Relative fitness in the absence of antibiotics. Shown, from top to bottom, is fitness relative to the average of the evolved control populations, calculated for growth rate, lag time, and maximum yield for all populations adapted to one of the eight antibiotics (X-axis) after 24 h of growth in antibiotic-free media. Colored points represent the replicate populations and the horizontal grey crossbars indicate the mean for each antibiotic. Black dashed lines highlight equality to the controls; values above indicate a fitness advantage whereas values below denote a cost. Plase note that for the lag phase, this is inverted: values larger than 1 indicate a longer time spent in lag phase and thus a fitness cost. Asterisks on top of each panel indicate significant difference from 1 (i.e., a significant change in fitness) using a Wilcoxon Rank test with probability adjustment based on the false discovery rate (FDR) to account for multiple testing.

Fig. 3.

Evolution of collateral sensitivity and resistance. (A) Illustration of the quantification of evolved collateral resistance or sensitivity. Bacterial growth (relative to a no-drug environment) of all evolved populations and the ancestral P. aeruginosa PA14 is first measured across concentrations of the various antibiotics. The area under the curve of the ancestor is subsequently subtracted from that of each population; resulting negative values indicate sensitivity (bottom panel), whereas positive values denote resistance (top panel). (B) The collateral profiles of all experimentally evolved populations (rows refer to the drugs used during experimental evolution), challenged against all other antibiotics (as indicated by columns). The vertical bars within each block represent the replicate populations. The different shades of purple or green highlight the extent of sensitivity or resistance, respectively. (C) We counted the total number of cases, for which adaptation to a particular antibiotic (listed in the middle) led to collateral sensitivity or resistance (direct adaptation; left panel), and also the total number of cases, for which sensitivity or resistance towards the focal antibiotic was observed upon adaptation to any of the other drugs (indirect adaptation; right panel).

Fig. 4.

Chemical similarity correlates with frequency of collateral resistance. Pairwise Jaccard’s similarity indexes were calculated based on the chemical fingerprints of each antibiotic. The frequency of collateral resistance (FCR) was then determined as $FCR\hspace{0.17em}=\hspace{0.17em}(R_{A\to B}\hspace{0.17em}+\hspace{0.17em}{R}_{B\to A})/L_{AB}$, where R_A→B is the number of populations resistant to drug A with cross-resistance to drug B (and vice versa for R_B→A), and L_AB is the total number of populations adapted to A and B. A significant correlation was then found between the chemical similarity and the FCR (Spearman’s rank correlation). Each point corresponds to a chemical comparison between any two given drugs. Labels are shown for some, but not all, of these pairwise comparisons.

Fig. 5.

Genomics of adaptation. Distribution of the number of genes with substitutions per (A) antibiotic used in the evolution experiment, (B) type of mutational change, and (C) average frequency class within the replicate populations. (D) Functional effect of mutations found in coding regions of the listed genes (vertical axis, left side) across evolution experiments with different antibiotics (horizontal axis). Functional information (right side) is inferred from a combined analysis using DAVID, the Pseudomonas database and publications. Different shades of red indicate the percentage of affected populations per evolution experiment with a particular antibiotic.

Fig. 6.

Genomic determinants of collateral sensitivity. We employed a hierarchical clustering analysis using the Ward’s criterion method and bootstrapping to identify the genomic determinants of variation in collateral profiles in four treatments. In two of these treatments, replicate populations that had adapted to either GEN (A, B) or STR (C, D) produced variation in collateral profile to PIT and CAR. In the other two treatments, replicate populations that had adapted to either PIT (E, F) or CAR (G, H) showed such variation towards GEN and STR. We first evaluated the clustering of populations adapted to GEN (A), STR (C), PIT (E), and CAR (G) based on the strength of collateral effects to the other two drugs (fig. 3B), highlighting clusters of those with collateral sensitivity (lines in different shades of purple, see legend to the right) versus those with collateral resistance (lines in different shades of green). The circled numbers always indicate the same replicate populations from a particular treatment across the related panels. Populations clustering together based on their collateral effects also often clustered together based on their genomic profile (B, D, F, H). The mutated genes present in the various clusters are given to the right of the dendrograms, followed by letters for their functional annotation in brackets (see legend to the right for the annotation categories). If a specific cluster mainly included populations associated with collateral sensitivity, then the gene names are given in purple. In cases, where clusters mainly included populations associated with collateral resistance, the gene name is given in green.

Fig. 7.

Functional analysis of different regulatory genes. We focused on four specific mutations identified in the evolved populations to associate with collateral sensitivity and introduced these into the ancestral PA14 strain. The resulting mutants were then tested against various concentrations of CAR, PIT, GEN, and STR (from left to right). In all cases, the ancestral PA14 (always in black) and the adapted population (always in darker colors), from which the particular mutation was extracted, were tested simultaneously with the corresponding constructed mutants. Points and error bars show the mean OD ± SD of five technical replicates per antibiotic concentration. For each set of bacterial populations challenged against a particular drug, we performed a GLM followed by Tukey’s honest significant difference (HSD) test. For a summary of the statistical results see supplementary table S6, Supplementary Material online.