Responses of the Soil Bacterial Community, Resistome, and Mobilome to a Decade of Annual Exposure to Macrolide Antibiotics

Abstract

Biosolids that are applied to agricultural soil as an organic fertilizer are frequently contaminated with pharmaceutical residues that have persisted during wastewater treatment and partitioned into the organic phase. Macrolide antibiotics, which serve as a critically important human medicine, have been detected within biosolids. To determine the impacts of macrolide antibiotics on soil bacteria, every year for a decade, a series of replicated field plots received an application of a mixture of erythromycin, clarithromycin, and azithromycin at a realistic (0.1 mg kg soil^-1) or an unrealistically high (10 mg kg soil^-1) dose or were left untreated. The effects of repeated antibiotic exposure on the soil bacterial community, resistome, mobilome, and integron gene cassette content were evaluated by 16S rRNA and integron gene cassette amplicon sequencing, as well as whole-metagenome sequencing. At the unrealistically high dose, the overall diversity of the resistome and mobilome was altered, as 21 clinically important antibiotic resistance genes predicted to encode resistance to 10 different antibiotic drug classes were increased and 20 mobile genetic element variants (tnpA, intI1, tnpAN, and IS91) were increased. In contrast, at the realistic dose, no effect was observed on the overall diversity of the soil bacterial community, resistome, mobilome, or integron gene cassette-carrying genes. Overall, these results suggest that macrolide antibiotics entrained into soil at concentrations anticipated with biosolid applications would not result in major changes to these endpoints. IMPORTANCE Biosolids, produced from the treatment of sewage sludge, are rich in plant nutrients and are a valuable alternative to inorganic fertilizer when applied to agricultural soil. However, the use of biosolids in agriculture, which are frequently contaminated with pharmaceuticals, such as macrolide antibiotics, may pose a risk to human health by selecting for antibiotic resistance genes that could be transferred to plant-based food destined for human consumption. The consequences of long-term, repeated macrolide antibiotic exposure on the diversity of the soil bacterial community, resistome, and mobilome were evaluated. At unrealistically high concentrations, macrolide antibiotics alter the overall diversity of the resistome and mobilome, enriching for antibiotic resistance genes and mobile genetic elements of concern to human health. However, at realistic antibiotic concentrations, no effect on these endpoints was observed, suggesting that current biosolids land management practices are unlikely to pose a risk to human health due to macrolide antibiotic contamination alone.

Keywords: antimicrobial resistance; biosolids; macrolides; soil microbiology.

FIG 1

Chao1 richness (a) and composition (b) of bacterial taxa in antibiotic-exposed (0.1 and 10 mg kg⁻¹) and untreated control soil. (a) Number of observed bacterial taxa (richness) from the whole-metagenome (n = 3) and 16S rRNA amplicon (n = 4) analyses within each treatment group. Horizontal lines connect samples for visual clarity. Statistically significant differences between treatments groups were assessed with a one-way ANOVA. ns, indicates that differences in richness were not significant. (b) Principal-component analysis (PCA) ordination plots of the CLR-transformed Aitchison distances of soil bacterial taxa. Percentages of total variance explained by each principal component (PC1 and PC2) are displayed in the axis titles. Shaded ellipses correspond to 95% confidence intervals of treatment groups. P and F values are from a PERMANOVA with 999 permutations.

FIG 2

Chao1 richness (a to c) and principal-component analysis (PCA) of the compositions (d to f) of antibiotic resistance genes, mobile genetic element variants, and integron gene cassette open reading frames in antibiotic-exposed (0.1 to 10 mg kg⁻¹) and untreated control soil. Richness was measured as the number of unique antibiotic resistance genes (n = 3) (a), mobile genetic element variants (n = 3) (b), and integron gene cassette open reading frames (n = 4) (c) detected within each treatment group. Horizontal lines connect samples for visual clarity. Statistically significant differences between treatment groups were assessed with a one-way ANOVA and Tukey’s all-pairs test post hoc. ns, indicates that differences in richness were not significant. PCA ordination plots of the CLR-transformed Aitchison distances of antibiotic resistance genes (d), mobile genetic element variants (e), and integron gene cassette open reading frames (f). Percentages of total variance explained by each principal component (PC1 and PC2) are displayed in the axis titles. Shaded ellipses correspond to 95% confidence intervals of treatment groups. P and F values are from a PERMANOVA with 999 permutations.

FIG 3

(a and b) Mean fold changes (effect sizes) of antibiotic resistance gene relative abundances in antibiotic-exposed soil (0.1, 10 mg kg⁻¹) relative to untreated control soil (a) and (b) the antibiotic drug classes (b) to which resistance is predicted (right). The abbreviated name of the predicted antibiotic resistance gene (ARG) is listed in the middle. (a) Fold changes for both treatment groups (blue, low dose; pink, high dose) are shown only for the genes that were significantly differentially abundant (P < 0.05) within at least one treatment group relative to the control. The black vertical line at zero represents no difference in relative abundance compared with the control group. Horizontal lines intersecting with points are error bars, indicating the extent of Bonferroni-adjusted 95% confidence intervals of effect sizes. Points without error bars indicate that the confidence interval is smaller than the diameter of the point. Filled points represent genes whose abundances were significantly different from the unexposed control soil, and open points represent abundances that were not significantly different. (b) Antibiotic drug classes (and triclosan) listed on the x axis are abbreviated as follows: AMINO, aminoglycoside; CEPH, cephalosporin; PENAM, penam; TRICL, triclosan; PENEM, penem; MONO, monobactam; FLUORO, fluoroquinolone; SLFNM, sulfonamide; SLFON, sulfone; DIAMIN, diaminopyrimidine; ACRID, acridine dye; PHENI, phenicol; CARBA, carbapenem; MACRO, macrolide; TETRA, tetracycline.

FIG 4

Principal-component analysis (PCA) ordination plots of the CLR-transformed Aitchison distances of antibiotic resistance genes, grouped by their target drug classes. Only the ordination plots with statistically significant differences in composition between treatment groups are shown (using a PERMANOVA with 999 permutations). Percentages of total variance explained by each principal component (PC1 and PC2) are displayed in the axis titles. Shaded ellipses correspond to 95% confidence intervals of treatment groups.

FIG 5

Mean fold changes (effect sizes) of mobile genetic element (MGE) variant relative abundances in antibiotic-exposed soil (0.1 and 10 mg kg⁻¹) relative to unexposed control soil, and the GenBank accession numbers of the host genomes of the variants. Fold changes for both treatment groups (blue, low dose; pink, high dose) are shown only for the mobile genetic elements that were significantly differentially abundant (P < 0.05) within at least one treatment group relative to the control. The black vertical line at zero represents no difference in relative abundance compared with the control group. Horizontal lines intersecting with points are error bars, indicating the extent of Bonferroni-adjusted 95% confidence intervals of effect sizes. Points without error bars indicate that the confidence interval is smaller than the diameter of the point. Filled points represent variants whose abundances were significantly different from the unexposed control soil, and open points represent abundances that were not significantly different.