The evolution of antimicrobial peptide resistance in Pseudomonas aeruginosa is shaped by strong epistatic interactions

Abstract

Colistin is an antimicrobial peptide that has become the only remaining alternative for the treatment of multidrug-resistant Gram-negative bacterial infections, but little is known of how clinical levels of colistin resistance evolve. We use in vitro experimental evolution and whole-genome sequencing of colistin-resistant Pseudomonas aeruginosa isolates from cystic fibrosis patients to reconstruct the molecular evolutionary pathways open for high-level colistin resistance. We show that the evolution of resistance is a complex, multistep process that requires mutation in at least five independent loci that synergistically create the phenotype. Strong intergenic epistasis limits the number of possible evolutionary pathways to resistance. Mutations in transcriptional regulators are essential for resistance evolution and function as nodes that potentiate further evolution towards higher resistance by functionalizing and increasing the effect of the other mutations. These results add to our understanding of clinical antimicrobial peptide resistance and the prediction of resistance evolution.

Figure 1. High-level colistin resistance in clinical isolates is multi-factorial.

(a) Maximum-parsimonious phylogenetic tree representing the evolutionary relationship of P. aeruginosa isolates from patient B3. The tree shows the relationship between isolates based on 334 SNPs identified from genome sequencing with an equal number of mutational events (that is, perfect consistency). The P. aeruginosa reference isolate PAO1 sequence was used as outgroup to determine the root of the tree. Branch lengths are proportional to number of SNPs between isolates. Date of isolation is in parentheses after the strain designation. (b) MICs (μg ml⁻¹) from microbroth dilution of P. aeruginosa PAO1 (PAO1), the B3 clinical isolates, B3-CFI pmrAB deletion (B3-CFI Δ pmrAB) and P. aeruginosa PAO1 allelic replacement mutant with pmrB^{634G>A, 742G>A} allele from B3-CFI (PAO1 pmrB^B3−CFI). Horizontal lines represent median MIC values of three or five replicates. Data underlying the figure is also presented in Supplementary Tables 2 and 3.

Figure 2. Experimental evolution of high-level colistin resistance in P. aeruginosa.

(a) Colistin MIC values (μg ml⁻¹) for Wildtype (WT) ancestral P. aeruginosa PAO1; SNRB through SNRP are colistin-resistant clones from surviving selection lineages. CON1 through CON5 are clones from control lineages grown in LB medium in the absence of colistin. CON1 is a mutator strain with mutation in mutS. Colistin MIC was determined by microbroth dilution assays. Horizontal lines indicate the median and symbols indicate biological replicates for MIC determination (at least three replicates were used per strain except for CON5). (b) Parallel evolution of colistin resistance in vitro. Shown are operons or genes with mutations in more than three colistin-resistant strains. The symbols contain the number of mutations at the locus. Strain SNRO did not have a mutation in ORF PA5194 but had a mutation immediately upstream and a mutation downstream of the gene. Circles, genes affecting arnB-operon regulation; squares, genes affecting LPS biosynthesis or outer membrane assembly; hexagons, genes affecting other functions; dotted line, in vitro selected strains and clinical isolates.

Figure 3. Evolutionary pathways to colistin resistance.

(a) Phenotypes of the 32 (2⁵) constructed genotypes in PAO1 based on mutations linked to resistance in the colistin-resistant strain SNRC with colistin MIC (μg ml⁻¹) for each genotype. (b) Phenotypes of the 8 (2³) constructed genotypes in PAO1 based on the alleles of pmrAB^{634G>A,742G>A} (pmrAB^B3−CFI), opr86^1525A>G (opr86^B3−CFI) and lpxC^551A>G (lpxC^B3−CFI) linked to resistance in colistin-resistant strains from patient B3 (B3-20M and B3-CFI) with colistin MIC (μg ml⁻¹) for each genotype. For both figures, node labels denote genes affected by the mutation. Blue, low colistin MIC; red, high colistin MIC. Line thickness indicates relative MIC increase. Paths to high-level resistance deemed inaccessible to selection because they had no increase in MIC at any step were omitted for clarity. Wildtype (WT), ancestral P. aeruginosa PAO1. Colistin MIC was determined by microbroth dilution assays and values are the median of at least three biological replicates.

Figure 4. Epistasis in colistin resistance evolution.

Graphs show proportional MIC change in 16 allelic backgrounds with increasing MIC upon addition of mutations in pmrB, opr86, lpxC, PA5194 or PA5005 from the colistin-resistant strain SNRC. Each point represents a single strain. In some cases, points are on top of each other as indicated by the numbers in parentheses. Colistin MIC was determined by microbroth dilution assay and presented as the median of at least three biological replicates.

Figure 5. Mutations in regulators potentiate evolution of high-level colistin resistance.

Relative estimates for frequency of resistance to 16 μg ml⁻¹ colistin (unless otherwise indicated) was determined for different strains of P. aeruginosa PAO1 using 96-well microtiter plate assays. Plates were scored for growth after 48 hrs by measuring optical density (OD) at 600 nm. Resistance as growth was scored for wells with OD>0.06. Dotted horizontal line, number of wells (96) investigated; wildtype (WT) P. aeruginosa PAO1; pmrB, phoQ, opr86, lpxC, PA5194 and PA5005, P. aeruginosa allelic replacement mutants with mutations from SNRC or SNRD (phoQ^779T>G). ΔpmrABΔphoPQ, genes encoding the two-component systems were deleted; ΔarnB arnB gene of the LPS modification operon was deleted. Each point represents a single experiment. In some cases more than one symbol occupies the same space. Horizontal lines show the median for each strain. Results are based on at least three biological replicates.