Mechanisms of drug interactions between translation-inhibiting antibiotics

Abstract

Antibiotics that interfere with translation, when combined, interact in diverse and difficult-to-predict ways. Here, we explain these interactions by "translation bottlenecks": points in the translation cycle where antibiotics block ribosomal progression. To elucidate the underlying mechanisms of drug interactions between translation inhibitors, we generate translation bottlenecks genetically using inducible control of translation factors that regulate well-defined translation cycle steps. These perturbations accurately mimic antibiotic action and drug interactions, supporting that the interplay of different translation bottlenecks causes these interactions. We further show that growth laws, combined with drug uptake and binding kinetics, enable the direct prediction of a large fraction of observed interactions, yet fail to predict suppression. However, varying two translation bottlenecks simultaneously supports that dense traffic of ribosomes and competition for translation factors account for the previously unexplained suppression. These results highlight the importance of "continuous epistasis" in bacterial physiology.

Fig. 1. Antibiotics targeting different translation steps show diverse drug interactions.

a, b Schematic of the translation cycle and translation inhibitors. Translation factors are shown in dark gray boxes. The stability of the large subunit is mediated by Der and initiation by initiation factors (IFs). Elongation factors Tu and G (EF-Tu, EF-G) catalyze ribosome progression. The release of GDP from EF-Tu is facilitated by EF-Ts. Release factors (RFs) facilitate the ejection of the finished peptide from the ribosome, whose recycling is mediated by the factor for ribosome recycling (Frr). Translation inhibitors are shown in white boxes (abbreviations in Table 1). c Examples of response surfaces for different antibiotic combinations corresponding to different interaction types (left column) and examples of growth curves (right column). Left: dose–response surfaces for different drug combinations. Grayscale shows a normalized growth rate as a function of concentrations of two antibiotics. Drug interactions are determined based on the shape of the contour lines of equal growth (isoboles). If the addition of the second drug has the same effect as increasing the concentration of the first, the isoboles are straight lines. Deviations from this additive expectation reveal synergism (the combined effect is stronger and isoboles curve towards the origin), antagonism (the effect is weaker and isoboles curve away from the origin), or suppression (at least one of the drugs loses potency due to the other). Symbols show drug conditions in which growth curves shown in the right column were measured. Right: Growth curves (i.e., time courses of luminescence) for four conditions (no drug, individual drugs, and combination); thin gray line shows the additive expectation of the growth curve for the combined stress; luminescence (photon count-per-second) on y-axes is a proxy for the number of bacteria (Methods). Symbols on the growth curves indicate the condition used: no symbol, triangle, square and a circle correspond to no drug, CHL-only, the second drug only (see left plots), and the combination of both, respectively. The growth curves were shifted in time to originate from the same point at time zero. d Drug-interaction network of translation inhibitors. Color-code is as in (c); dashed gray lines denote additivity. Each drug interaction was measured twice.

Fig. 2. Mathematical model partially predicts drug interactions.

a Schematic of antibiotic binding and transport into the cell. Antibiotics (circles) bind to the unbound ribosomes (gray) in the first binding step (above dashed line); bound ribosomes can be bound by a second antibiotic (second binding step; below dashed line). b Schematic of antibiotics binding independently (top) or competing for the same binding site (bottom). c Growth laws link intracellular ribosome concentration to the growth rate. Solid line: ribosome concentration when the growth rate is varied by varying nutrient quality; dashed lines: ribosome concentration when the growth rate is lowered by perturbation of translation. Circles show data from ref. . d Data points are dose–response curves for ERM and KSG; lines show the best-fits of the mathematical model. The best-fit values of the response parameter α that encapsulates kinetic and physiological parameters (Supplementary Information) are shown. Both shallow (top panel, ERM) and steep (bottom panel, KSG) dose–response curves are observed. e Examples of predicted dose–response surfaces. The scatter plot depicts the correlation between predicted and measured growth rates. Means and error bars (standard deviation) of predicted growth rates are estimated from n = 100 bootstrap repetitions. The binding scheme assumed is indicated on the bottom right and Pearson’s ρ on the top left. Predicted and measured dose–response surfaces are shown below the scatter plot. Color of 20% isobole (bottom) denotes the type of predicted interaction. The model correctly predicts response surfaces for KSG-ERM, ERM-TET, and ERM-CHL, yet it fails to predict the interaction between STR and KSG.

Fig. 3. Artificial translation bottlenecks strongly affect antibiotic efficacy.

a Schematic of synthetic regulation introduced to control the expression of a translation factor x, which creates an artificial bottleneck in translation at a well-defined stage; lacI codes for the Lac repressor, which represses the P_LlacO-1-promoter (Methods,). b Constructs were made for six translation factors mediating 50S stability (der), initiation (infB), recycling (frr), translocation (fusA), tRNA delivery (tufAB) and GDP release (tsf), respectively. Higher expression alleviates the artificial bottleneck. Thicker lines or arrows indicate higher rates. c Translation factor induction curves (upper row) and response surfaces over the inducer-antibiotic grid for different antibiotics (KSG and FUS, middle and bottom row, respectively) in combination with different bottlenecks (50S stability, initiation, and translocation). Full induction of the translation factor rescues wild type growth; increasing bottleneck severity leads to a smooth decrease in growth rate to zero. Induction curves were measured in n = 8 technical replicates, and the median value of non-zero growth rates was calculated. Comparison of the response surfaces with independent expectation (dashed purple line) identify alleviation (orange line) or aggravation (dark blue line). d Columns show bottleneck dependency vectors in color-code; dependency vectors quantify the response of a given antibiotic to the translation bottlenecks (Supplementary Information). e Clustering of the bottleneck dependency vectors upon dimensionality-reduction by principal component analysis (PCA; Supplementary Information). Circles show dependency vectors projected onto the first two principal components (PC1, and PC2); colors indicate cluster identity. The extended cluster areas shown are convex hulls of bootstrapped projections (denoted by dots). Projections of the three additional antibiotics LAM, NIT, and TMP are denoted by a purple triangle, blue square, and green pentagon, respectively. We estimated the p-value by clustering n = 10⁴ reshuffled datasets with added noise and counting the fraction of instances that matched the shown clustering result. See Supplementary Equations (19–20) and Supplementary Fig. 3e; we did not use a standard statistical test.

Fig. 4. Translation factor deprivation mimics the action of equivalent antibiotics.

a Schematic of translation as a sequence of steps (white), catalyzed by translation factors (gray). In the absence of perturbations, ribosomes progress through the steps unimpeded, resulting in unperturbed growth. b Schematic of perturbed translation. Top: as the abundance of factor F1 is lowered (smaller factor symbol), the rate of step 1 decreases (thinner arrows) and ribosomes queue in front of the bottleneck. Bottom: the same rate is reduced by an antibiotic. The effects of factor deprivation and antibiotic action on growth are equivalent. c Schematic of conversion of inducer concentration b (here for the translocation factor) into the mimicked antibiotic concentration c (here: CRY). For each inducer concentration b, the growth rate from the induction curve g(b) is determined and the same growth rate on the antibiotic dose–response curve y(c) is identified (gray dashed line); the inverse function of the dose–response curve yields the equivalent antibiotic concentration as $c=y^{-1}\left(g\left(b\right)\right)$. d The resulting conversion of inducer concentration b into antibiotic concentration c for three different pairs of equivalent perturbations: CRY-translocation (gray), KSG-initiation (yellow), and TET-tRNA delivery (orange). e Inducer-antibiotic response surface (left) and mimicked antibiotic-antibiotic response surface (right) upon conversion of inducer concentration as in c, d. Purple dashed line shows isobole for multiplicative responses at relative growth rate 0.2. The remapped response surface is additive, corroborating the equivalence of CRY and translocation factor deprivation. f–i Comparison of response surfaces remapped to the additive expectations. The bottlenecks and antibiotics are shown on the bottom right, respectively. Errors in LI and in expected and remapped responses were evaluated by bootstrapping (Supplementary Methods, Supplementary Fig. 4). f Additive expectation from e and remapped response surface agree (ρ = 0.99). g As f, but for a recycling bottleneck. The large and statistically significant discrepancy in LI from 0 indicates that CRY and a recycling bottleneck are not equivalent. h As f, but for KSG and an initiation bottleneck (ρ = 0.98). i As f, but for TET and a tRNA delivery bottleneck (ρ = 0.99).

Fig. 5. Translation bottlenecks can predict antibiotic interactions.

a Example of drug-interaction prediction based on the equivalent translation bottlenecks. The drug interaction between CHL and an antibiotic that targets initiation can be predicted through mimicking the initiation inhibition by limiting the expression of initiation factor (infB). b The response surface of CHL combined with the inducer for the initiation (infB) bottleneck shows mild alleviation. This response surface contains information about the interaction between CHL and any antibiotic that interferes with initiation. The inducer axis is remapped into mimicked antibiotic concentration (Fig. 4c, d). c Left: resultant prediction of the response surface for the initiation-inhibiting antibiotic KSG and CHL. Right: measured KSG-CHL response surface for direct comparison; strong antagonism is observed as predicted. d A point-by-point comparison of predicted and measured response surfaces (Pearson’s ρ = 0.98). e Schematic showing antibiotics and their equivalent translation factor bottlenecks. Drug interactions with these antibiotics can be predicted for any antibiotic with a known response to the equivalent bottleneck. Color-code shows cluster identity from Fig. 3e. f Comparison of predicted and measured response surfaces for different antibiotics in combination with antibiotics that have a factor analog. Top row: scatter plots as in (d); bottom row: predicted and measured response surfaces, respectively. Remapping correctly predicts antagonism (KSG-LCY), suppression (FUS-CHL), strong antagonism (KSG-STR), and additivity (TET-CHL).

Fig. 6. Ribosome traffic jams underlie suppression between inhibition of translocation and initiation.

a Schematic of ribosomes progressing along a transcript—a stuck ribosome can cause a traffic jam. Ribosomes undergo factor-mediated initiation events with attempt rate ζ and translocation with attempt rate γ. Expression of initiation and elongation factors are controlled by the level of inducer (IPTG and aTc, respectively). b Results of all-or-nothing growth assay: bacteria grow only when both essential factors are induced. c Left: measured growth rate response surface for the dual inducible promoter strain from (a) as a function of both inducer concentrations; the red line shows the ridge of maximum growth. Right: cross-section of the response surface along the dashed purple line (gray circles) and at maximal aTc induction (white circles); solid lines are smoothed profiles. Black arrow denotes a decrease in translocation; if initiation is lowered simultaneously with translocation (orange arrow), growth reduction is smaller. d Schematic of the theoretical model: translation is described as an ensemble of transcripts competing for the limited and growth-rate-dependent pool of ribosomes. Ribosomes advance on transcripts as described by a generalized totally asymmetric simple exclusion process (TASEP) for particles of size L (see a and text). When γ < ζ(1 + L^1/2), ribosomes saturate and traffic jams develop, resulting in a drop in elongation and growth (black arrow, the transition happens at the black triangle) (Supplementary Methods^,,). When ζ < γ/(1 + L^1/2), a phase transition occurs (green triangle): traffic jams dissolve—elongation and growth increase (along the green arrow). e Left: the growth rate predicted by the generalized TASEP model recapitulates suppression of translocation inhibition by lowered initiation; note that, unlike in (c), axes show the concentrations of translation factors. States below and to the right of the green line are in the translocation limiting regime. Right: cross-sections of the response surface. As the initiation factor level is decreased, the critical point of the phase transition (green triangle) is reached; growth starts increasing after passing the critical point, and decreases again after passing the maximum (red square) as the number of translating ribosomes becomes limiting. f Bottleneck dependency (BD) score quantifies the deviation from independent expectation (BD = 0) for the response surfaces in (c, e); heights of bars corresponds to the medians and error bars are 90% bootstrap confidence intervals. Medians and confidence intervals were estimated from n = 100 bootstrap data points.