Selective pressure of antibiotic pollution on bacteria of importance to public health

Abstract

Background: Many bacteria of clinical importance survive and may grow in different environments. Antibiotic pollution may exert on them a selective pressure leading to an increase in the prevalence of resistance.

Objectives: In this study we sought to determine whether environmental concentrations of antibiotics and concentrations representing action limits used in environmental risk assessment may exert a selective pressure on clinically relevant bacteria in the environment.

Methods: We used bacterial inhibition as an assessment end point to link antibiotic selective pressures to the prevalence of resistance in bacterial populations. Species sensitivity distributions were derived for three antibiotics by fitting log-logistic models to end points calculated from minimum inhibitory concentration (MIC) distributions based on worldwide data collated by the European Committee on Antimicrobial Susceptibility Testing (EUCAST). To place bacteria represented in these distributions in a broader context, we performed a brief phylogenetic analysis. The potentially affected fraction of bacterial genera at measured environmental concentrations of antibiotics and environmental risk assessment action limits was used as a proxy for antibiotic selective pressure. Measured environmental concentrations and environmental risk assessment action limits were also directly compared to wild-type cut-off values.

Results: The potentially affected fraction of bacterial genera estimated based on antibiotic concentrations measured in water environments is ≤ 7%. We estimated that measured environmental concentrations in river sediments, swine feces lagoons, liquid manure, and farmed soil inhibit wild-type populations in up to 60%, 92%, 100%, and 30% of bacterial genera, respectively. At concentrations used as action limits in environmental risk assessment, erythromycin and ciprofloxacin were estimated to inhibit wild-type populations in up to 25% and 76% of bacterial genera.

Conclusions: Measured environmental concentrations of antibiotics, as well as concentrations representing environmental risk assessment action limits, are high enough to exert a selective pressure on clinically relevant bacteria that may lead to an increase in the prevalence of resistance.

Figure 1

Conceptual link between the MIC₅₀ and antibiotic concentrations > CO_WT with resistance prevalence in a universe (Ω) including resistant (black) and wild-type (gray) populations, and its relation to the MIC distribution. Antibiotic concentrations > CO_WT completely inhibit wild-types, and resistance prevalence in the active (i.e., growing) population will be 100%. Concentrations of antibiotics ≤ CO_WT, such as the MIC₅₀, will inhibit a fraction of the wild-type population (e.g., 50%). If we assume equal growth rates of wild-type and resistant populations, the prevalence of antibiotic resistance will increase in the active population.

Figure 2

Unrooted neighbor-joining dendrogram of species (by LTP name) represented in the pooled 16S rRNA alignment of ciprofloxacin, erythromycin, and tetracycline. Species highlighted in red were not included in the species sensitivity distributions due to lack of evidence of growth in the environment. Bar units indicate the number of nucleotide substitutions per site. Black nodes indicate ≥ 70% bootstrap support; blue nodes indicate < 70% bootstrap support. Full species names are available in Supplemental Material, Table S1 (http://dx.doi.org/10.1289/ehp.1104650).

Figure 3

Histograms showing the range of pairwise evolutionary distances covered by (A) the pooled 16S rRNA alignment of species represented in the MIC distributions of all three antibiotics, and (B) the entire LTP 16S rRNA alignment for the domain Bacteria. Pairwise evolutionary distances in Awere calculated from the same 16S rRNA alignment used to construct the dendrogram in Figure 2.

Figure 4

Species sensitivity distributions derived for the NOEC (black) and MIC₅₀ (red) curves with overlayed empirical cumulative distributions (dots). Fitted curves represent the bootstrap estimate and 95% bootstrap CIs for the log-logistic model. Dashed and solid blue vertical lines represent the VICH phase I aquatic and soil action limits, respectively. (A) Ciprofloxacin: NOEC (R² = 0.90; p < 0.0001), MIC₅₀ (R² = 0.91; p < 0.0001); bootstrap estimate of model parameters: NOEC species sensitivity distribution (α = –2.1, β = 0.37), MIC₅₀ species sensitivity distribution (α = –1.0, β = 0.54). (B) Erythromycin: NOEC (R² = 0.96; p <0.001), MIC₅₀ (R² = 0.97; p < 0.0001); bootstrap estimate of model parameters: NOEC species sensitivity distribution (α = –1.13, β = 0.29), MIC₅₀ species sensitivity distribution (α = –0.014, β = 0.48). (C) Tetracycline: NOEC (R² = 0.97; p <0.0001), MIC₅₀ (R² = 0.98; p < 0.0001); bootstrap estimate of model parameters: NOEC species sensitivity distribution (α = –0.84, β = 0.27), MIC₅₀ species sensitivity distribution (α = 0.2, β = 0.33). The potentially affected fraction of bacterial genera at a given antibiotic concentration is read from the y-axis at the point in which the antibiotic concentration intersects with the species sensitivity distribution. For example, a concentration of 100 ppb (i.e., log₁₀ concentration = 2) of ciprofloxacin inhibits approximately one-half of the wild-type population (i.e., red MIC₅₀ curve) in 54% of the bacterial genera and at least some individuals (i.e., black NOEC curve) in 95% of the bacterial genera.

Figure 5

MICs ≥ CO_WT for bacterial taxa in the MIC distributions of ciprofloxacin (A), erythromycin (B), and tetracycline (C). Symbols represent the CO_WT for different genera; dashed vertical lines extend up to the maximum MIC beyond the CO_WT; and horizontal lines represent antibiotic concentrations. The x-axis represents the number of bacterial taxa with a defined CO_WT for each antibiotic, and the y-axis indicates antibiotic concentrations.