Sequential Acquisition of Virulence and Fluoroquinolone Resistance Has Shaped the Evolution of Escherichia coli ST131

Abstract

Escherichia coli ST131 is the most frequently isolated fluoroquinolone-resistant (FQR) E. coli clone worldwide and a major cause of urinary tract and bloodstream infections. Although originally identified through its association with the CTX-M-15 extended-spectrum β-lactamase resistance gene, global genomic epidemiology studies have failed to resolve the geographical and temporal origin of the ST131 ancestor. Here, we developed a framework for the reanalysis of publically available genomes from different countries and used this data set to reconstruct the evolutionary steps that led to the emergence of FQR ST131. Using Bayesian estimation, we show that point mutations in chromosomal genes that confer FQR coincide with the first clinical use of fluoroquinolone in 1986 and illustrate the impact of this pivotal event on the rapid population expansion of ST131 worldwide from an apparent origin in North America. Furthermore, we identify virulence factor acquisition events that predate the development of FQR, suggesting that the gain of virulence-associated genes followed by the tandem development of antibiotic resistance primed the successful global dissemination of ST131.

Importance: Escherichia coli sequence type 131 (ST131) is a recently emerged and globally disseminated multidrug-resistant clone frequently associated with human urinary tract and bloodstream infections. In this study, we have used two large publically available genomic data sets to define a number of critical steps in the evolution of this important pathogen. We show that resistance to fluoroquinolones, a class of broad-spectrum antibiotic used extensively in human medicine and veterinary practice, developed in ST131 soon after the introduction of these antibiotics in the United States, most likely in North America. We also mapped the acquisition of several fitness and virulence determinants by ST131 and demonstrate these events occurred prior to the development of fluoroquinolone resistance. Thus, ST131 has emerged by stealth, first acquiring genes associated with an increased capacity to cause human infection, and then gaining a resistance armory that has driven its massive population expansion across the globe.

FIG 1

Geographical diversity of the combined data set across clades and time. (A) Stacked histogram showing the number of strains in clades A, B, C1, and C2 according to their country of origin. The color scheme is shown in the legend on the left along with abbreviated country names. (B) Box-and-whisker plot showing the distribution according to year of isolation for all strains based on their country of origin. Country abbreviations: KOR, South Korea; IND, India; NZ, New Zealand; AUS, Australia; POR, Portugal; UK, United Kingdom; ESP, Spain; CAN, Canada; USA, United States of America.

FIG 2

Maximum likelihood phylogenetic tree of ST131 strains. The phylogram was built from 5,471 nonrecombinant SNPs using maximum likelihood (ML). Branch support was performed by 1,000 bootstrap replicates (see Fig. 2B in the supplemental material). The scale bar indicates the number of substitution SNPs. Taxon labels for clades A, B, and C are colored red, orange, and green, respectively. Seven strains sharing intermediate characteristics between clades B and C are colored pink. Of note, clade A strains were collapsed and the clade A-specific branches shortened for display. Metadata are represented as circles as follows: year of isolation in gray and gradient and geographical region in assorted colors as depicted in the legend. Allelic profiling information is shown as colored strips surrounding the phylogram (from inner to outer) for the fimH, parC, gyrA, and CTX-M genes. Two additional distinctions were made for some fimH variants: “Untypeable” corresponds to a strain with a truncated or missing fimH gene, and “Pseudogene” corresponds to a strain in which fimH is disrupted by an insertion sequence. Clades B0 to B5 and C0 subclades are shown as arcs in the outermost ring, with arrows and dotted lines denoting the division between subclades C1 and C2.

FIG 3

Prevalence of antibiotic resistance-associated genes by (A) clade alone and (B) clade and country. The box-and-whisker plot shows the number of resistance-associated genes per (A) clade and (B) per clade and country. Colors correspond to clades as follows: clade A, red; clade B, orange; clade C1, light green; and clade C2, dark green. Screening was done using Srst2 (42) against ARGannot, with a minimum of depth of 15× read coverage. No., number. Country abbreviations: AUS, Australia; CAN, Canada; NZ, New Zealand; ESP, Spain; USA, United States of America; UK, United Kingdom. P values are indicated as follows: ***, P < 0.001; **, P < 0.01; and *, P < 0.05.

FIG 4

Evolutionary scenario of the emergence of ST131 clades (B and C) A time-calibrated phylogeny was reconstructed using BEAST 2.0 based on 3,779-bp nonrecombinant SNPs for the 172 clade B and C strains. Of all combinations tested (see the summary of BEAST analysis results in Data Set S1 in the supplemental material), the one combining the GTR substitution model, a constant relaxed clock model, and the Bayesian skyline population tree model was preferred. (A) Maximum clade credibility tree colored according to clade origin as shown on the right with B in orange, intermediate B0 and C0 in pink, C1 in light green, and C2 in dark green. The x axis indicates emergence time estimates of the corresponding strains. Major evolutionary events are also indicated by an arrow pointing at the branch onto which they are predicted to have occurred (the position along the branch is arbitrary). Three categories of major events are displayed, namely, MGE and genomic island (GI) insertion events in black, allelic change acquired through recombination events in blue, and allelic change acquired through point mutation events in purple. Of note, the two point mutations indicated by purple arrowheads and pointing at the branch from which clades C1 and C2 originate confer resistance to fluoroquinolone, for which the first introduction is indicated in the bottom timeline by a red arrowhead. (B) Unrooted phylogenetic tree built on the same 3,779-bp nonrecombinant set of SNPs using maximum likelihood (ML). Branch support was performed by 1,000 bootstrap replicates. Intermediate strain names and predicted acquisition of fimH30 are indicated on the tree. (C) The Bayesian skyline plot illustrates the predicted demographic changes of the ST131 clade B and C population since the mid-1940s. The black curve indicates the effective population size (Ne), with 95% confidence intervals shown in blue.

FIG 5

Geographic location of the most recent common ancestor (MRCA) of ST131 major clades. Individual probabilities were predicted from 10 independent BEAST analyses of randomly subsampled data to limit bias related to overrepresentation of some locations. The mean probabilities of the geographic locations of the MRCA for (A) clades B and C and (B) clade C are shown as a box-and-whiskers plot colored according to country using the scheme as described in Fig. 1. Countries are labeled on the x axis by abbreviation: USA, United States of America; CAN, Canada; ESP, Spain; UK, United Kingdom; AUS, Australia; NZ, New Zealand; IND, India.