Fig 1.
Collateral effects are pervasive and vary across parallel evolution experiments in E. faecalis.
(A) E. faecalis strain V583 was exposed to increasing concentrations of a single antibiotic over an 8-day serial-passage experiment with daily 200-fold dilutions (maximum of 60 generations total; see Materials and methods). The evolution was performed in quadruplicate for each drug and repeated for a total of 15 drugs (Table 1). After 8 days, a single mutant was isolated from each population. (B) The IC50 for each of 15 drugs was estimated for all 60 mutants by nonlinear fitting of a dose-response curve (relative OD) to a Hill-like function (Materials and methods). (C) Main panel: resistance (red) or sensitivity (blue) of each evolved mutant (horizontal axis; 15 drugs × 4 mutants per drug) to each drug (vertical axis) is quantified by the log2-transformed relative increase in the IC50 of the testing drug relative to that of WT (V583) cells. Although the color scale ranges from a 4× decrease to a 4× increase in IC50, it should be noted that both resistance to the selecting drug (diagonal blocks) and collateral effects can be significantly higher. Each column of the heat map represents a collateral sensitivity profile for one mutant. Right panel: enlarged first column from main panel. Mutants isolated from replicate populations evolved to DAP exhibit diverse sensitivity profiles. Although all mutants are resistant to the selecting drug (DAP), mutants may exhibit either sensitivity or resistance to other drugs. For example, the first and last replicates exhibit cross-resistance to CRO, whereas replicate 2 exhibits collateral sensitivity, and replicate 3 shows little effect. Data underlying this figure can be found in S1 Data. AMP, ampicillin; CHL, chloramphenicol; CIP, ciprofloxacin; CRO, ceftriaxone; DAP, daptomycin; DOX, doxycycline; FOF, fosfomycin; IC50, half-maximal inhibitory concentration; LVX, levofloxacin; LZD, linezolid; mut, mutant; NIT, nitrofurantoin; OD, optical density; OXA, oxacillin; RIF, rifampicin; SPT, spectinomycin; TET, tetracycline; TGC, tigecycline; WT, wild-type.
Table 1.
Table of antibiotics used in this study and their targets.
Fig 2.
Growth costs and lag times for isolates selected by different antibiotics.
(A) Example OD time series for single isolates selected by each of the 15 drugs. Blue or red circles correspond to the isolate, and black circles correspond to ancestral strains. Light green lines show fits to logistic growth function [60] given by g(t) = g0+Kc(1+exp(4μ(λ−t)/Kc+2))−1, where μ is the maximum specific growth rate, λ is the lag time, and Kc is the carrying capacity. To reduce the number of free parameters, we fix Kc = 0.5 to match that of the ancestral strain. (B) Maximum specific growth rate (μ, left) and lag time (λ, right) in drug-free media for isolates from each of the four populations selected by each drug. All values are normalized by the values measured in the ancestral strain. Error bars are standard errors of the mean estimated from three technical replicates for each isolate. (C) Left panel: variability in replicates for all 15 drugs versus the (log2-scaled) fold increase in IC50 to the selecting drug (Spearman ρ = 0.58, p = 0.03 including the SPT mutants; ρ = 0.82, p < 10−3, without the SPT mutants). Variability is defined as , where m = 4 is the number of replicates, and di is the Euclidean distance between mutant i and the centroid formed by all vectors corresponding to a given selecting drug (S2 Fig). Right panel: histogram of Euclidean distances between collateral profiles in pairs of isolates selected by the same (red) or different (blue) drugs. Distributions exhibit significantly different means (p<10−3, Welch t test). To emphasize collateral, rather than direct, effects, the component(s) of each collateral profile corresponding to the selecting drug(s) were removed prior to calculating variability and pairwise Euclidean distances. Data underlying this figure can be found in S1 Data. AMP, ampicillin; CHL, chloramphenicol; CIP, ciprofloxacin; CRO, ceftriaxone; DAP, daptomycin; DOX, doxycycline; FOF, fosfomycin; LVX, levofloxacin; LZD, linezolid; mut, mutant; NIT, nitrofurantoin; OD, optical density; OXA, oxacillin; RIF, rifampicin; SPT, spectinomycin; TET, tetracycline; TGC, tigecycline.
Fig 3.
Collateral effects can lead to frequent or high-level resistance to nonselecting drugs.
(A) Estimated dose-response curves (fit to Hill-like function) for all mutants tested against DAP. Strains evolved to DAP (blue) and all other drugs (red) frequently exhibit increased resistance to DAP relative to WT (black, individual replicates; dotted black line, mean IC50). Right inset: Approximately 64% of all drug-evolved mutants exhibit increased DAP resistance, while only 11% exhibit collateral sensitivity. (B) Fractional change in CHL IC50 for mutants evolved to LZD (blue). The width of the green line represents the confidence interval (± 3 standard errors of the mean measured over eight replicate measurements) for the (normalized) CHL IC50 in WT cells. For comparison, the red lines represent the final (day 8) CHL resistance achieved in populations evolved directly to CHL. Inset: schematic depicting two hypothetical paths to different CHL resistance maximums. The green circle represents the sensitive WT. Evolution can occur to CHL directly (red line) or to CHL collaterally through LZD resistance (blue line). The LZD evolution depicts early collateral sensitivity before ultimately achieving a higher total resistance. (C) CHL resistance (log2-scaled change in IC50 relative to ancestor) for LZD-selected isolates at day 5 (purple) and day 8 (blue) and for individual colony isolates (four) for each of the four CHL-selected populations (black). Arrows indicate 12 isolates chosen for Sanger sequencing. (D) Mutations observed in four different genes associated with LZD resistance in each of the 12 selected isolates from panel C. (E) CHL resistance and number of 23S mutations in LZD isolates on days 5 and 8. Data underlying this figure can be found in S1 Data. CHL, chloramphenicol; DAP, daptomycin; IC50, half-maximal inhibitory concentration; LZD, linezolid; MIC, maximum inhibitory concentration; Mut, mutant; OD, optical density; Sens, sensitivity; WT, wild-type.
Table 2.
Mutations identified in selected populations.
Fig 4.
Hierarchical clustering of collateral sensitivity profiles partitions mutants into groups selected by known drug classes.
Heat map with ordering of rows (testing drug) and columns (four replicate experiments with the same selecting drug) determined via hierarchical clustering. Colormap and scale are identical to those used in Fig 1. Collateral profiles (columns) for mutants selected by drugs from known drug classes (here labeled A–G) cluster together; if clusters are defined using the dashed line (top), there are seven distinct clusters, each corresponding to a particular drug class: (A) cell wall synthesis inhibitors (AMP, OXA, CRO, FOF), (B) tetracyclines (TET, DOX, TGC), (C) lipopeptides (DAP), (D) oxazolidinones (LZD), (E) fluoroquinolones (CIP, LVX), (F) aminocyclitols (SPT), and (G) antimycobacterials (RIF). When clustering the testing drugs (rows), drugs from the same class are frequently but not always clustered together. For example, cell wall drugs such as AMP, OXA, and CRO form a distinct cluster that does not include FOF (bottom four rows). See also S1 Table. AMP, ampicillin; CIP, ciprofloxacin; CRO, ceftriaxone; DAP, daptomycin; DOX, doxycycline; FOF, fosfomycin; LVX, levofloxacin; LZD, linezolid; OXA, oxacillin; RIF, rifampicin; SPT, spectinomycin; TET, tetracycline; TGC, tigecycline.
Fig 5.
Simulated optimal drug sequences constrain resistance on long timescales and outperform simple collateral sensitivity cycles.
(A) Discretized collateral sensitivity or resistance Cd∈{−2,−1,0,1,2} for a selection of six drugs. For each selecting drug, the heat map shows the level of cross-resistance or sensitivity (Cd) to each testing drug (the subscript d indicates the profiles are discretized) for nr = 4 independently evolved populations. See Fig 1 for original (nondiscretized) data. (B) Average level of resistance (〈R(t)〉) to the applied drug for policies with γ = 0 (red), γ = 0.7 (black), γ = 0.78 (magenta), and γ = 0.99 (blue). Resistance to each drug is characterized by 11 discrete levels arbitrarily labeled with integer values from −1 (least resistant) to 9 (most resistant). At time 0, the population starts in the second lowest resistance level (0) for all drugs. Symbols (circles, triangles, squares) are the means of 103 independent simulations of the MDP, with error bars ± SEM. Solid lines are numerical calculations using exact Markov chain calculations (see Materials and methods). Light red line, long-term optimal policy (γ = 0.99) calculated using the data in (A) but with collateral sensitivity values set to 0. Black shaded line, randomly cycled drugs (±SEM). (C) The time-dependent probability P(Drug) of choosing each of the six drugs when the optimal policy (γ = 0.99) is used. Inset, steady-state distribution Pss (Drug). (D) The probability P(Resist) of the population exhibiting a particular level of resistance to the applied drug when the optimal policy (γ = 0.99) is used. Inset, steady-state distribution Pss (Drug). (E) Steady-state joint probability distribution P(current drug, next drug) for consecutive time steps when the optimal policy (γ = 0.99) is used. (F) Average level of resistance (〈R(t)〉) to the applied drug for collateral sensitivity cycles of 2 (dark green, LZD-RIF), 3 (pink, AMP-RIF-LZD), or 4 (dark green, AMP-RIF-TGC-LZD) drugs are compared with MDP policies with γ = 0 (short-term, red) and γ = 0.99 (long-term, blue). For visualizing the results of the collateral sensitivity cycles, which give rise to periodic behavior with large amplitude, the curves show a moving time average (window size 10 steps), but the original curves are shown transparently in the background. Data underlying this figure can be found in S1 Data. AMP, ampicillin; DAP, daptomycin; FOF, fosfomycin; LZD, linezolid; MDP, Markov decision process; RIF, rifampicin; TGC, tigecycline.
Fig 6.
Optimized drug sequences reduce cumulative growth and adaptation rates in laboratory evolution experiments.
(A) Resistance (red) or sensitivity (blue) of each evolved mutant (horizontal axis; 4 drugs × 8 mutant per drug) to each drug (vertical axis) following 2 days of selection is quantified by the log2-transformed relative increase in the IC50 of the testing drug relative to that of wild-type (V583) cells. (B) Top: distribution of applied drug at time step 20 (approximate steady state) calculated over all realizations of the stochastic process using an optimal policy with γ = 0.9. Bottom: sequence of applied drug from one particular realization of the stochastic process with the optimal policy (γ = 0.9). (C–E) Cumulative population growth over time for populations exposed to single-drug sequences ([C], blue), two-drug sequences ([D], magenta), a four-drug sequence ([E], red), or the optimal sequence from panel B (black curves, all panels). Transparent lines represent individual replicate experiments and each thicker dark line corresponds to a mean over replicates. Dashed line, drug-free control (normalized to a growth of 1 at the end of the experiment). (F) Adaptation rate for single-drug (blue), two-drug (magenta), four-drug (red), and optimal sequences (black). Error bars are standard errors across replicates. Adaptation rate is defined as the slope of the best-fit linear regression describing time series of daily growth (see S15 Fig). Data underlying this figure can be found in S1 Data. AMP, ampicillin; FOF, fosfomycin; IC50, half-maximal inhibitory concentration; RIF, rifampicin; TGC, tigecycline.