The antibiotic resistome of swine manure is significantly altered by association with the Musca domestica larvae gut microbiome

Abstract

The overuse of antibiotics as veterinary feed additives is potentially contributing to a significant reservoir of antibiotic resistance in agricultural farmlands via the application of antibiotic-contaminated manure. Vermicomposting of swine manure using housefly larvae is a promising biotechnology for waste reduction and control of antibiotic pollution. To determine how vermicomposting influences antibiotic resistance traits in swine manure, we explored the resistome and associated bacterial community dynamics during larvae gut transit over 6 days of treatment. In total, 94 out of 158 antibiotic resistance genes (ARGs) were significantly attenuated (by 85%), while 23 were significantly enriched (3.9-fold) following vermicomposting. The manure-borne bacterial community showed a decrease in the relative abundance of Bacteroidetes, and an increase in Proteobacteria, specifically Ignatzschineria, following gut transit. ARG attenuation was significantly correlated with changes in microbial community succession, especially reduction in Clostridiales and Bacteroidales. Six genomes were assembled from the manure, vermicompost (final product) and gut samples, including Pseudomonas, Providencia, Enterococcus, Bacteroides and Alcanivorax. Transposon-linked ARGs were more abundant in gut-associated bacteria compared with those from manure and vermicompost. Further, ARG-transposon gene cassettes had a high degree of synteny between metagenomic assemblies from gut and vermicompost samples, highlighting the significant contribution of gut microbiota through horizontal gene transfer to the resistome of vermicompost. In conclusion, the larvae gut microbiome significantly influences manure-borne community succession and the antibiotic resistome during animal manure processing.

Figure 1

Attenuation or enrichment of ARGs. (a) Profiling of ARG abundance for all observed samples (with 12 replicates for C0d and S6d and 3 for all other samples; 6 S0d samples were marked as C0d here, because they are biologically the same). Each row is the relative gene abundance from a single primer set, with only those that showed significant (P<0.05) changes among samples determined by analysis of variance plus Duncan tests represented. Red indicates greater correlation, blue indicates less correlation, and green represents median values. ARGs highlighted in yellow are specified in Supplementary Table S5; and all datasets are outlined in Supplementary Table S3. (b) The descriptive statistics for attenuated resistance genes (the relative abundance, expressed as S6d/C0d when <1). The box plots denote 25th to 75th percentile; horizontal line, median; and whiskers, 10th and 90th percentile. Symbols along x axis indicate genes conferring resistance to aminoglycoside (Am), beta_lactamase (Be), (flor)/(chlor)/(am)phenicol (Ph), (flouro)quinilone (Qu), MLSB (Ml), Multidrug (Mu), Tetracylines (Te), Vancomycin (Va) and Sulfonamide (Su). (c) The descriptive statistics for enriched resistance genes (S6d/C0d >1). Fold change values are considered significant only if the range created by standard deviations of the average are entirely <1 (attenuated) or entirely>1 (enriched) (see Supplementary Table S4 for the datasets).

Figure 2

Bacterial community structure from 16S rRNA analysis, and functional predictions using PICRUST. (a) PCoA plots based on weighted UniFrac distances of samples. Red triangle=manure samples under vermicomposting (red triangle); green circle=manure samples under traditional composting without larvae input; blue square=larvae gut samples. The average distances for each category by connecting all points with minimal total length divided by the total numbers of these involved samples are further shown. (b) Relative abundance (%) of taxa at the genus level, clustered using UPGMA on the weighted UniFrac distances. Taxa with <1% of reads were merged together as ‘others' while ‘unknown' represents unclassified taxa at the genus level (see Supplementary Table S6). (c) PICRUST predictions using KEGG subsystem level II. Raw manure (C0d+S0d), larvae gut (L1d to L6d) and vermicompost (S6d) are compared using non-parametric Kruskal–Wallis test with Dunn's multiple comparison. Only functions with >1% abundance are shown. Red on y axis indicates increased (P<0.05) functional abundance in vermicompost compared with raw manure (see Supplementary Table S13).

Figure 3

Local similarity analyses of the potential linkages between the dynamics of ARGs and bacterial taxa/OTUs during vermicomposting. Blue octagons represent taxon at phylum, class, order, family or genus level; orange circles represent OTUs (97% similarity threshold); and green diamonds represent ARGs. Node size represents average relative abundance across replicates. Edge color represents local similarity score, and weight represents P-value, where a lower P-value will have a wider edge. (a) Subnetwork organized between significantly attenuated ARGs and taxa/OTUs. (b) Subnetwork organized between significantly enriched ARGs and taxa/OTUs. (c) Subnetwork organized between significantly attenuated ARGs and family members belonging to the order Clostridiales. (d) Subnetwork organized between significantly attenuated ARGs and family members belonging to the order Bacteroidales. (e) Subnetwork organized between significantly attenuated ARGs and OTUs/taxa belonging to the family Bacteroidaceae. To reduce the network complexity, pairs illustrated here were filtered with a significant P-value<0.005. The basic network parameters for original network diagrams (before data filtration) are shown in an inserted table. Also see Supplementary Table S7 for taxonomic information for each OTU in network.

Figure 4

Larvae gut reshapes bacterial community compositions in manure samples. (a) Source Tracking of microbial communities in manure under vermicomposting from day 1 (S1d) to day 6 (S6d). ‘Unknown' are taxa that could not be significantly soured to either raw manure or larvae gut. (b) Exclusive and shared OTUs (non-singleton OTUs, based on 97% reads similarity) between raw manure (C0d and S0d), larvae gut (L1d to L6d) and vermicompost (S6d), with number representing OTUs found in each segment (see Supplementary Table S8). (c) Oligotyping of Ignatzschineria and Pseudomonas. The dynamics for each identified oligotype are shown from 1 day (S1d) to 6 day (S6d) under vermicomposting (see Supplementary Table S6).