Genetic architecture of intrinsic antibiotic susceptibility

Abstract

Background: Antibiotic exposure rapidly selects for more resistant bacterial strains, and both a drug's chemical structure and a bacterium's cellular network affect the types of mutations acquired.

Methodology/principal findings: To better characterize the genetic determinants of antibiotic susceptibility, we exposed a transposon-mutagenized library of Escherichia coli to each of 17 antibiotics that encompass a wide range of drug classes and mechanisms of action. Propagating the library for multiple generations with drug concentrations that moderately inhibited the growth of the isogenic parental strain caused the abundance of strains with even minor fitness advantages or disadvantages to change measurably and reproducibly. Using a microarray-based genetic footprinting strategy, we then determined the quantitative contribution of each gene to E. coli's intrinsic antibiotic susceptibility. We found both loci whose removal increased general antibiotic tolerance as well as pathways whose down-regulation increased tolerance to specific drugs and drug classes. The beneficial mutations identified span multiple pathways, and we identified pairs of mutations that individually provide only minor decreases in antibiotic susceptibility but that combine to provide higher tolerance.

Conclusions/significance: Our results illustrate that a wide-range of mutations can modulate the activity of many cellular resistance processes and demonstrate that E. coli has a large mutational target size for increasing antibiotic tolerance. Furthermore, the work suggests that clinical levels of antibiotic resistance might develop through the sequential accumulation of chromosomal mutations of small individual effect.

Figure 1. Overview of experimental protocol.

(A) An aliquot of a library containing ∼5×10⁵ mutants each with a single transposon insertion was taken from frozen stock, grown overnight in LB, pelleted, washed, and resuspended at 2% inoculum in fresh M9-media containing an antibiotic at the chosen concentration (Table 1). Each day, an aliquot was frozen, and 2% of the culture was transferred to fresh media to continue the selection. Genetic footprinting was performed on frozen samples to amplify the region of genomic DNA adjacent to the transposon in each of the mutants . DNA was subsequently labeled and hybridized with a reference of labeled genomic DNA to spotted microarrays . (B) Dose response curves were used to select drug concentrations. For each antibiotic, fresh media containing various drug concentrations was inoculated with overnight culture of the wild-type strain. Growth was monitored using OD₆₀₀ readings. Shown are the curves for amikacin; curves for all other antibiotics are in Figure S1. Typically, we selected moderately inhibitory drug concentrations that reduced the growth after 14 hours by 30–50%. (C) Separation of DNA on an agarose gel provided a qualitative depiction of the population diversity after each day of selection. Shown are the amplified Tn-adjacent DNA from all seven days of one of the ampicillin selections. Selections performed without antibiotic showed no discernable banding pattern (Figure S2). Gel images for all selections with antibiotics are in Figure S3.

Figure 2. Selection rates during direct competitions.

Selection rates (generations/day) were calculated as (log₂(A(t₁)/A(t₀))−log₂(B(t₁)/B(t₀)))/(t₁−t₀) . A(t₀) and B(t₀) are, respectively, the mutant and the wild-type population sizes at t₀, the beginning of the competition, and A(t₁) and B(t₁) are the mutant and the wild-type population sizes at t₁, the end of the competition. Shown are the average and standard deviation of three repetitions. The selection rate for the trpA::kan mutant in amikacin was calculated after two days of enrichment to correspond with the samples hybridized. The trpA::kan strain's reliance on tryptophan from lysed wild-type cells prevents the mutant from taking over the culture, and during additional transfers, the wild-type strain showed a competitive advantage. Selection rates for other strains were insensitive to the competition duration.

Figure 3. Overlap between genes influencing fitness in partially inhibitory concentrations of different antibiotics.

Squares on the main diagonals indicate the number of genes whose disruption caused a significant fitness effect (See Materials and Methods). Genes causing general changes in antibiotic susceptibility (Figure 7) were excluded. The lower left (upper right) triangle reports on genes whose disruption was beneficial (deleterious) to E. coli in the presence of the indicated antibiotic. Off-diagonal squares indicate how many genes caused significant fitness changes in both antibiotics when disrupted. The shading shows the likelihood of an overlap of the indicated size or larger occurring by chance and was calculated using the hypergeometric distribution. P-values were corrected for multiple testing. Erythromycin and fusidic acid are not shown as the only genes whose disruption affected fitness caused general changes in susceptibility.

Figure 4. Disruption of electron transport and oxidative respiration reduces susceptibility to aminoglycosides.

(A) The heatmaps, in which hierarchical clustering was used to order both the genes and the drugs , show loci whose disruption changed susceptibility to all four aminoglycosides tested (See Materials and Methods). Table S1 lists the genes with annotations. (B) Of the 73 transposon insertions regions identified as beneficial in all four aminoglycosides, the 48 shown are expected to reduce Fenton reaction-based oxidative damage. Following exposure to lethal concentrations of bactericidal antibiotics, the oxidative electron transport chain depletes the NADH pool, generating high levels of superoxide, which removes iron from iron-sulfur clusters . The free iron subsequently generates hydroxyl radicals through the Fenton reaction . Removal of key catabolic enzymes should shrink the NADH pool and reduce the flux through the electron transport chain. The media used lacks cysteine, the sulfur donor for iron-sulfur center synthesis , so disruption of cysteine biosynthesis should reduce the availability of sulfur for iron-sulfur centers. The iron-sulfur center synthesis genes shown are not specific for NADH dehydrogenase I, and their disruption should reduce the number of iron-sulfur clusters throughout the cell. Q: ubiquinone; FMN: flavin mononucleotide; FAD: flavin adenine dinucleotide

Figure 5. Reduced flagella synthesis is advantageous in β-lactams.

(A) The heatmaps shows loci whose disruption changed susceptibility to all three β-lactams tested (See Materials and Methods). Hierarchical clustering was used to order both genes and drugs . Table S2 lists the genes with annotations. (B) Both transposon insertions that disrupt genes that encode flagella components as well as insertions that indirectly reduce flagella synthesis by activating the Rcs system are beneficial. The core components of the Rcs system are RcsC, a hybrid sensor kinase, RcsD, a histidine phosphotransferase, and RcsB, a DNA-binding response regulator . Other components are RcsF, a lipoprotein that activates RcsC , , and RcsA, a transcription factor that forms a heterodimer with RcsB . Together, RcsA and RcsB repress transcription of flhDC, the master regulator of flagella synthesis . RcsA is a target of the Lon protease , and insertions in lon, which stabilize RcsA, are beneficial. RcsC and RcsD both transfer phosphate to as well as remove phosphate from RcsB, resulting in higher activation of the Rcs system in rcsC or rcsD mutants than in wildtype , . Insertions in mdoG and mdoH, which encode proteins that synthesize osmoregulated periplasmic glucans (OPGs), reduce motility by activating the Rcs system . The beneficial effects of mdoG, modH, and rcsC disruptions are not limited to β-lactams (Figure 7).

Figure 6. Genetic and chemical perturbations of the folate biosynthesis pathway.

Sulfamonomethoxine inhibits FolP, a dihydropteroate synthase ; trimethoprim inhibits FolA, the cell's main dihydrofolate reductase (DHFR) . FolM, which also acts as a DHFR, is not inhibited by trimethoprim . Mutants lacking folM or folX are less sensitive to both to trimethoprim and sulfamonomethoxine.

Figure 7. Genes altering susceptibility to three or more classes of antibiotics.

Yellow (blue) indicates that transposon insertions in or near a gene were beneficial (deleterious). Black indicates no significant effect; gray indicates missing data. Antibiotics with the same target are written in the same color. Sulfamonomethoxine and trimethoprim inhibit different enzymes in the folic acid biosynthesis pathway; placing them in separate classes did not alter the results. Z-scores were calculated as described in Materials and Methods.