The spread of antimicrobial resistance has become a serious public health concern, making once-treatable diseases deadly again and undermining the achievements of modern medicine1,2. Drug combinations can help to fight multi-drug-resistant bacterial infections, yet they are largely unexplored and rarely used in clinics. Here we profile almost 3,000 dose-resolved combinations of antibiotics, human-targeted drugs and food additives in six strains from three Gram-negative pathogens—Escherichia coli, Salmonella enterica serovar Typhimurium and Pseudomonas aeruginosa—to identify general principles for antibacterial drug combinations and understand their potential. Despite the phylogenetic relatedness of the three species, more than 70% of the drug–drug interactions that we detected are species-specific and 20% display strain specificity, revealing a large potential for narrow-spectrum therapies. Overall, antagonisms are more common than synergies and occur almost exclusively between drugs that target different cellular processes, whereas synergies are more conserved and are enriched in drugs that target the same process. We provide mechanistic insights into this dichotomy and further dissect the interactions of the food additive vanillin. Finally, we demonstrate that several synergies are effective against multi-drug-resistant clinical isolates in vitro and during infections of the larvae of the greater wax moth Galleria mellonella, with one reverting resistance to the last-resort antibiotic colistin.
Access optionsAccess options
Rent or Buy article
Get time limited or full article access on ReadCube.
All prices are NET prices.
Subscribe to Journal
Get full journal access for 1 year
only $3.90 per issue
All prices are NET prices.
VAT will be added later in the checkout.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
We thank P. Beltrao (EBI) and T. Bollenbach (University of Cologne) for providing feedback on the manuscript; K. M. Pos (Goethe University) for the anti-AcrA antibody; D. Helm and the EMBL Proteomics Core Facility for help with mass spectrometry experiments; the EMBL GBCS and the Centre for Statistical Analysis for advice on data analysis; S. Riedel-Christ for help with G. mellonella experiments; and the members of the Typas laboratory for discussions. This work was partially supported by EMBL internal funding, the Sofja Kovalevskaja Award of the Alexander von Humboldt Foundation to A.Ty., the JPIAMR Combinatorials grant to F.B. (ANR) and A.Ty. (BMBF), and the DFG (FOR 2251) to S.G. A.M. and J.S. are supported by a fellowship from the EMBL Interdisciplinary Postdoc (EIPOD) program under Marie Curie Actions COFUND.Reviewer information
Nature thanks A. Koul, K. Lewis and the other anonymous reviewer(s) for their contribution to the peer review of this work.
Extended data figures and tables
Extended Data Fig. 1 High-throughput profiling of pairwise drug combinations in Gram-negative bacteria.
a, Drug and species selection for screen. The 79 drugs used in the combinatorial screen are grouped according to categories (Supplementary Table 1). Antibacterial agents are grouped by target, with the exception of antibiotic classes for which enough representatives were screened (>2) to form a separate category (β-lactams, macrolides, tetracyclines, fluoroquinolones and aminoglycosides). Classification of human-targeted drugs and food additives is not further refined, because for most of these the mode of action is unclear. A subset of 62 arrayed drugs was profiled against 79 drugs in all 6 strains (75 drugs were common to all strains and are depicted in the heat map). Strains are colour-coded according to species: yellow, E. coli; red, S. Typhimurium; green, P. aeruginosa. b, Quantification of drug–drug interactions. Growth was profiled by measuring optical density (OD595 nm) over time in the presence of no, one and both drugs. x and y correspond to particular concentrations of drugs X and Y. Interactions were defined according to Bliss independence. Significantly lower or higher fitness than the expectation (fx × fy) indicates synergy or antagonism, respectively. Synergy and antagonism were assessed by growth in 4 × 4 checkerboards (Methods).
a, Flowchart of the data analysis pipeline. b, Estimating single-drug fitness of arrayed drugs. As drug–drug interactions are rare, the slope of the line of best fit between gaq (growth with double drug) and gq (growth with query drug alone, which was deduced from the average of the top 5% growing wells across plates within a batch), across plates (np) of query drugs within a batch, corresponds to a proxy of the fitness of the arrayed drug alone, fa (see Methods). r denotes the Pearson correlation coefficient between gaq and gq across np. Well A9 from E. coli BW25113, containing 3 μg ml−1 spectinomycin, is shown as an example of arrayed drugs with several interactions. Several query drugs deviated from the expected fitness (light grey points), and therefore only half of the plates corresponding to the interquartile range of gag/gq were used to estimate fa. c, Density distributions of the first, second and third quartiles of Bliss-score (ε) distributions for E. coli. Q1, Q2 and Q3 denote the median of the first, second and third quartiles of ε distributions, respectively. n denotes the number of drug combinations used. Source Data
a, High replicate correlation for single- and double-drug treatments. Transparent box plots contain Pearson correlation coefficients between plates of the same batch that contain only arrayed drugs (for which LB was used instead of the second drug). n represents the total number of correlations. Full box plots contain Pearson correlation coefficients between double-drug replicate wells within the same plate, across all plates. n represents the number of wells used for correlation, nmax = (62 drugs + 1 LB) × 3 concentrations = 189. Only wells with median growth above 0.1 were taken into account for this correlation analysis (see b). For all box plots the centre line, limits, whiskers and points correspond to the median, upper and lower quartiles, 1.5 × interquartile range and outliers, respectively. b, Wells with lower median growth have lower replicate correlation. The double-drug correlation coefficients used to generate the box plot from a are plotted as a function of the median growth of all wells across all plates for E. coli IAI1. Wells with overall lower growth (due to the strong inhibition of the arrayed drug) are less reproducible owing to a combination of the lower spread of growth values and the sigmoidal nature of the drug–dose response curves. c, Drug–drug interactions are rare. Density distributions of all Bliss scores (ε) obtained per strain. d, The ability to detect synergies and antagonisms depends on the effects of single-drug treatments. Bliss scores (ε) are plotted as function of expected fitness (fx × fy) for all drug concentration ratios for all combinations in E. coli BW25113 (as an example). Box plots summarizing both variables are shown besides the axes (n = 99,907 Bliss scores; centre line, limits, whiskers and points correspond to the median, upper and lower quartiles, 1.5 × interquartile range and outliers, respectively). Blind spots for detecting antagonism and synergy are indicated; both of these are based on the expected fitness (see also Extended Data Fig. 4c, d), and are therefore dependent on the growth of the strain with the single drugs. The number of drug combinations falling in the blind spot for antagonism is larger, owing to the number of drugs used in the screen that do not inhibit E. coli on their own. e, Scatter plot of the number of interactions per drug versus the minimum fitness of the drug alone (as obtained in the screen, Supplementary Table 1). Strong and weak interactions are represented. n denotes the total number of interactions and r is the Pearson correlation coefficient. Strains are colour-coded as above. f, Density distributions of the number of interactions per drug for all strains. Source Data
a, The validation set is enriched in synergies and antagonisms to better assess true and false positives. Comparison of percentages of synergy and antagonism between the screen and validation set. Both strong and weak interactions (Fig. 2b) are accounted for in the screen tally. b, Number of benchmarked interactions per strain. c, d, Sensitivity analysis of the statistical thresholds for calling interactions. c, The total number of interactions as a function of the expected fitness (fx × fy) cutoff was used to restrict the ε distributions to relevant drug concentrations. Strong drug–drug interactions are classified according to the ε distribution in which they were significant: complete distribution only (that is, all expected fitness wells), relevant wells only (that is, all wells with fx × fy > cutoff for synergies and all wells with fx × fy < (1 − cutoff) for antagonisms), or in both. Weak drug–drug interactions are independently assigned and represented in white. We selected an expected fitness cutoff of 0.2, as this cutoff resulted in the largest number of total interactions detected, with the highest precision and recall (91 and 74%, respectively) after benchmarking against the validation dataset. d, Receiver operating characteristic curve for the screen across different P value thresholds (10,000 repetitions of a two-sided permutation test of Wilcoxon rank-sum test after correction for multiple testing, see Methods) as a unique criterion for assigning interactions. The selected P value (0.05) for the screen threshold is indicated by a grey cross. Sensitivity to additional parameters for calling hits is shown: allowing interactions to be either antagonisms or synergies but not both (one-sided); as well as strong and weak interaction thresholds. True- and false-positive rates were estimated based on the validation dataset. Precision and recall for the final and best-performing set of parameters are shown: one-sided interactions, P < 0.05, fx × fy cutoff = 0.2 and |ε| > 0.1 for strong interactions, |ε| > 0.06 for weak interactions. TP, true positive; TN, true negative; FP, false positive; FN, false negative. n indicates the total number of benchmarked drug combinations (Supplementary Table 3). e, Synergies between β-lactams according to the Loewe additivity interaction model. The results of 8 × 8 checkerboards for 3 combinations between β-lactams in 4 strains are shown. The grey line in each plot represents the null hypothesis in the Loewe additivity model and the black line corresponds to the IC50 isobole, which was estimated by fitting a logistic curve to the interpolated drug concentrations (coloured dots, Methods). Piperacillin did not reach 50% growth inhibition in E. coli, thus IC20 and IC40 isoboles were used for the amoxicillin + piperacillin combination in E. coli BW25113 and E. coli IAI1, respectively. Source Data
a, The bar plot illustrates the division of benchmarked drug combinations according to their degree of conservation within species. The pie chart shows the proportion of false positives (FP), true positives (TP), false negatives (FN) and true negatives (TN) within non-comparable drug–drug interactions. b, Combination of amoxicillin with cefotaxime in P. aeruginosa as an example of a non-comparable drug–drug interaction. Top box, the results of the screen. Left, Bliss scores as function of expected fitness for both strains. Right, a density distribution of the Bliss scores. n denotes the total number of Bliss scores, Q1 and Q3 indicate the Bliss score for the first and third quartiles, respectively. Antagonism was detected only for PAO1 (Q3 > 0.1). PA14 was resistant to both drugs at concentrations screened (top left panel), rendering the detection of antagonism impossible. Bottom box, benchmarking results indicate that the interaction is antagonistic in both strains, albeit weaker in PA14 and visible mostly at higher concentrations. The colour intensity on checkerboard reflects fitness and black dots correspond to drug ratios in which the Bliss score is above 0.1. Source Data
a, The bar plot illustrates the division of benchmarked drug combinations as in Extended Data Fig. 5a. The pie chart shows the proportion false positives and true positives within weak conserved interactions. b, Combination of doxycycline with amikacin in S. Typhimurium as an example of a weak conserved drug–drug interaction. Top box, the results of the screen. Left, Bliss scores as a function of expected fitness for both strains. Right, a density distribution of the Bliss scores. n denotes the total number of Bliss scores, Q1 and Q3 indicate the Bliss score for quartiles 1 and 3, respectively. A strong synergy was detected only for ST14028 (Q1 < −0.1), and a weak conserved synergy was assigned afterwards to ST LT2 (Q1 < −0.06). Bottom box, the benchmarking results confirm that the interaction is synergistic in both strains. The colour intensity on checkerboard reflects fitness and black dots correspond to drug ratios in which the Bliss score is below −0.1. Source Data
a, b, Drug category interaction networks. Nodes represent drug categories according to Extended Data Fig. 1a, and plotted as in Fig. 1b. Conserved interactions, including weak conserved interactions, are shown here. One of the most well-known and broadly used synergies is that of aminoglycosides and β-lactams45. Consistent with its use against P. aeruginosa in clinics, we detected multiple strong synergies between specific members of the two antibiotic classes in P. aeruginosa but fewer interactions in the other two species. c, d, Drug–drug interactions across cellular processes. Representation as in a, b but grouping drug categories targeting the same general cellular process. e, Quantification of synergy and antagonism in the networks from a, b and the corresponding χ2-test P value. As in E. coli (Fig. 1), antagonism occurs more frequently than synergy and almost exclusively between drugs belonging to different categories in S. Typhimurium and P. aeruginosa. In P. aeruginosa, there are very few interactions occurring between drugs of the same category (within the group). Source Data
Extended Data Fig. 8 Drug antagonisms are often due to a decrease in intracellular drug concentrations.
a, Cartoon of possible modes of action for drug–drug interactions that function via modulation of the intracellular drug concentration. A given drug (antagonist, blue) inhibits the uptake or promotes the efflux of another drug (black), and thus decreases its intracellular concentration. b, Different antagonists (see Methods for concentrations) of gentamicin (red, 5 μg ml−1) and ciprofloxacin (yellow, 2.5 μg ml−1) identified in our screen for E. coli BW25113 also rescue the killing effect of the two bactericidal drugs in the same strain, or its parental MG1655 (top right and top left panels, respectively). With the exception of clindamycin (for gentamicin) and curcumin (for ciprofloxacin), all other antagonists decrease the intracellular concentration of their interacting drug (bottom panels). Gentamicin was detected by using radiolabelled compound, and ciprofloxacin with LC–MS/MS (Methods). The degree of rescue (top panels) in many cases follows the decrease in intracellular concentration (bottom panels), which implies that most of these interactions depend at least partially on modulating the intracellular concentration of the antagonized drug. c, Antagonisms are resolved in E. coli BW25113 mutants that lack key components that control the intracellular concentration of the antagonized drug. Aminoglycosides depend on proton motive force-energized uptake, and thus on respiratory complexes7,46; ciprofloxacin is effluxed by AcrAB–TolC29,47. For gentamicin, most interactions are resolved when respiration is defected, even the interaction with clindamycin (which does not modulate intracellular gentamicin concentration, see b); this presumably occurs because the mode of action and import of aminoglycosides are linked by a positive feedback loop7,48. For ciprofloxacin, antagonisms with paraquat and caffeine are resolved in the ΔacrA mutant, which implies that both compounds induce the AcrAB–TolC pump (well-established for paraquat49). By contrast, interactions with curcumin, benzalkonium and doxycycline remain largely intact in the ΔacrA mutant. The first interaction is expected, as curcumin does not modulate intracellular ciprofloxacin concentration (see b). In the other two cases, other component(s) besides AcrAB–TolC may be responsible for the altered ciprofloxacin import and/or export; for example, ciprofloxacin uses OmpF to enter the cell50. Ciprofloxacin and gentamicin concentrations were adjusted in all strains according to MIC (70% and 100% MIC for ciprofloxacin and gentamicin, respectively; all drug concentrations are listed in Supplementary Table 6). Bliss interaction scores (ε) were calculated as in the screen. Bar plots and error bars in b, c represent the average and s.d., respectively, across n independent biological replicates. d, Gentamicin and ciprofloxacin antagonism networks for E. coli BW. Nodes represent drugs coloured according to targeted cellular process (as in Extended Data Fig. 1a). Full and dashed edges represent antagonistic drug–drug interactions for which intracellular antibiotic concentration was and was not measured, respectively. Drug interactions that result in decreased intracellular concentration of the antagonized drug are represented by black edges. e, Quantification of antagonistic drug–drug interactions from the networks in (d). The bars for fluoroquinolones and aminoglycosides account for an extrapolation of antagonistic interactions to all other members of the two classes, assuming that they behave in the same way as ciprofloxacin and gentamicin, respectively. Source Data
Extended Data Fig. 9 Drug–drug interactions are largely conserved within species and only partially driven by mode of action.
a, b, Drug–drug interactions are conserved in S. Typhimurium (a) and P. aeruginosa (b). Scatter plot of interaction scores in the two strains of each species; only strong interactions for at least one strain are shown. Colours and grouping as in Fig. 2a. r denotes the Pearson correlation and n denotes the total number interactions plotted. The lower correlation in P. aeruginosa is presumably due to fewer and weaker interactions. c, Drug interaction profiles are driven by phylogeny. Clustering of strains based on the Pearson correlation of their drug interaction profiles (taking into account all pairwise drug combinations; n = 2,759–2,883 depending on the species). Strains of the same species cluster together; the two enterobacterial species—E. coli and S. Typhimurium—behave more similarly to one another than either does to the phylogenetically more-distant P. aeruginosa. d, Conserved drug–drug interaction network. Nodes represent individual drugs grouped and coloured by targeted cellular process (as in Extended Data Fig. 1a). Drug names are represented by three-letter codes (given in Supplementary Table 1). Dashed and full edges correspond to conserved interactions between two or three species, respectively. Many of the human-targeted drugs, such as loperamide, verapamil and procaine, exhibit a general potentiating effect that is similar to that of membrane-targeting drugs. This suggests that these drugs may also facilitate drug uptake or impair efflux, consistent with previous reports on the role of loperamide in E. coli and verapamil in Mycobacterium tuberculosis4,51. e, Monochromaticity between all drug categories. The monochromaticity index (MI) reflects whether interactions between drugs of two categories are more synergistic (MI= −1) or antagonistic (MI = 1) than the background proportion of synergy and antagonism. The MI equals zero when interactions between two drug categories have the same proportion of synergy and antagonism as all interactions together (Methods). The MI was calculated using all interactions from the six strains for all category pairs that had at least two interactions. White cells in the heat map correspond to category pairs for which no (or an insufficient number of) interactions were observed. f, Human-targeted drugs, and LPS or proton motive force inhibitors, are strong and promiscuous adjuvants. Density distributions of the monochromaticity indices per drug category from e are shown. n denotes the number of drugs in each category and i the number of interactions in which they are involved. Source Data
Rows depict the 75 drugs common to all strains (coloured according to drug category, see Extended Data Fig. 1a), and columns depict their interactions with other drugs in all six strains tested. Clustering was done using the median of the ε distributions, uncentred correlation and average linkage. Source Data
Extended Data Fig. 11 Active synergies against Gram-negative MDR clinical isolates in vitro and in the G. mellonella infection model.
Both human-targeted drugs (which have recently been found to have an extended effect on bacteria52) and food additives can promote the action of antibiotics in MDR strains, indicating that their use as antibacterial adjuvants should be explored further. a, Drug combinations active against MDR E. coli and K. pneumoniae clinical isolates (see also Fig. 4). Interactions are shown as 8 × 8 checkerboards and synergies have a black bold border. Drug pairs are the same for each row of panel a, and are indicated at the first checkerboard in each row. The species in which the interaction was detected in the screen are indicated after the last checkerboard in each row. Concentrations increase in equal steps per drug (see legend); only minimal and maximal concentrations are shown for the first strain of each species. Apart from colistin, the same concentration ranges were used for all E. coli and K. pneumoniae MDR strains. One of two replicates is shown. b, Drug synergies against the same MDR strains in the G. mellonella infection model. Larvae were infected by E. coli and K. pneumoniae MDR isolates (106 and 104 CFUs, respectively) and left untreated, treated with single drugs or with the drug combination. The percentage of surviving larvae was monitored at indicated intervals after infection. n = 10 larvae per treatment. The averages of four biological replicates are plotted; error bars depict s.d. Source Data
a, The spectinomycin MIC decreases upon addition of 100 μg ml−1 vanillin in the wild-type E. coli BW25113, as well as in E. coli single-gene knockouts of members of the AcrAB–TolC efflux pump or its MarA regulator. Thus, the vanillin–spectinomycin synergy is independent of the effect of vanillin on AcrAB–TolC (Fig. 3). b, Synergy is specific to vanillin–spectinomycin, as spectinomycin is antagonized by 500 μg ml−1 of the vanillin-related compound aspirin, thereby increasing the MIC by approximately threefold. c, Profiling the vanillin–spectinomycin combination in the E. coli BW Keio collection26 to deconvolute its mode of action. Violin plots of the drug–drug interaction scores (ε) of all mutants (n = 9,216; Methods) are presented for the vanillin–spectinomycin combination (synergy) and, as control, for the combination of vanillin with another aminoglycoside amikacin (antagonism). The interaction scores of the two mdfA deletion clones present in the Keio library are indicated by red dots. The vanillin–spectinomycin synergy is lost in the absence of mdfA but the vanillin–amikacin antagonism remains unaffected, which indicates that the vanillin–spectinomycin synergy depends specifically on MdfA. d, Deletion of mdfA leads to an increased spectinomycin MIC and abolishes the synergy with vanillin, independent of the presence or absence of AcrAB–TolC. Mild overexpression of mdfA from a plasmid (pmdfA, Methods) further enhances the synergy with vanillin, decreasing the spectinomycin MIC by about twofold (compared to the MIC of the combination in the wild type). e, Overexpression of mdfA leads to increased spectinomycin sensitivity, even though the MIC does not change. The growth of E. coli BW25113 carrying a plasmid with mdfA cloned in it (pmdfA; no inducer, mild overexpression) or the empty vector (BW) was measured (OD595 nm after 8 h) over twofold serial dilutions of spectinomycin and normalized to the no-drug growth of the corresponding strain (white and black dots represent the average of n = 3 independent biological replicates, error bars represent s.d.). The spectinomycin dose response was computed using a logistic fit of the averaged data points (MICs are calculated by fitting individual replicates, and then averaging). Fitted curves are represented by full and dashed lines for pmdfA and E. coli BW25113, respectively. f, Vanillin leads to accumulation of spectinomycin in the cell in an mdfA-dependent manner. Intracellular spectinomycin is measured with the tritiated compound (Methods). Bar plots and error bars in a, b, d, f represent the average and s.d., respectively, across n independent biological replicates. Source Data
This file contains the Supplementary Discussion and Supplementary Fig. 1 (the uncropped immunoblot scans related to Fig. 3b).
This file contains Supplementary Tables 1–7.
This zipped file contains Supplementary Data 1–11 and a guide to the Supplementary Data.