Microbial Ecology

, Volume 59, Issue 3, pp 487–498

Molecular Ecology Of Macrolide–Lincosamide–Streptogramin B Methylases in Waste Lagoons and Subsurface Waters Associated with Swine Production

Authors

  • Satoshi Koike
    • Department of Animal SciencesUniversity of Illinois at Urbana-Champaign
    • Research Faculty of AgricultureHokkaido University
  • Rustam I. Aminov
    • Rowett Institute of Nutrition and HealthUniversity of Aberdeen
  • A. C. Yannarell
    • Department of Animal SciencesUniversity of Illinois at Urbana-Champaign
    • Institute for Genomic BiologyUniversity of Illinois at Urbana-Champaign
  • Holly D. Gans
    • Department of Animal SciencesUniversity of Illinois at Urbana-Champaign
  • Ivan G. Krapac
    • Illinois State Geological Survey
  • Joanne C. Chee-Sanford
    • USDA Agricultural Research Service
    • Department of Animal SciencesUniversity of Illinois at Urbana-Champaign
    • Institute for Genomic BiologyUniversity of Illinois at Urbana-Champaign
Environmental Microbiology

DOI: 10.1007/s00248-009-9610-0

Cite this article as:
Koike, S., Aminov, R.I., Yannarell, A.C. et al. Microb Ecol (2010) 59: 487. doi:10.1007/s00248-009-9610-0

Abstract

RNA methylase genes are common antibiotic resistance determinants for multiple drugs of the macrolide, lincosamide, and streptogramin B (MLSB) families. We used molecular methods to investigate the diversity, distribution, and abundance of MLSB methylases in waste lagoons and groundwater wells at two swine farms with a history of tylosin (a macrolide antibiotic structurally related to erythromycin) and tetracycline usage. Phylogenetic analysis guided primer design for quantification of MLSB resistance genes found in tylosin-producing Streptomyces (tlr(B), tlr(D)) and commensal/pathogenic bacteria (erm(A), erm(B), erm(C), erm(F), erm(G), erm(Q)). The near absence of tlr genes at these sites suggested a lack of native antibiotic-producing organisms. The gene combination erm(ABCF) was found in all lagoon samples analyzed. These four genes were also detected with high frequency in wells previously found to be contaminated by lagoon leakage. A weak correlation was found between the distribution of erm genes and previously reported patterns of tetracycline resistance determinants, suggesting that dissemination of these genes into the environment is not necessarily linked. Considerations of gene origins in history (i.e., phylogeny) and gene distributions in the landscape provide a useful “molecular ecology” framework for studying environmental spread of antibiotic resistance.

Introduction

The widespread use of antimicrobial agents has contributed to an increase in antibiotic resistance in pathogenic and commensal microorganisms of humans and animals. In the USA, antibiotics are commonly used in animal agriculture for disease treatment, prophylaxis, and growth promotion, and these practices can select for antibiotic resistance in animal-associated bacteria [43]. Recent studies suggest that animal waste can contribute antibiotic-resistant bacteria and/or their genes to the environment [1, 2, 8, 11, 17, 19, 44, 45]. Thus, animal production facilities are increasingly viewed as potential reservoirs for antibiotic resistance [29, 38].

The antibiotic tylosin is commonly used in poultry, swine, and cattle feedlots in the USA [9]. Tylosin belongs to the macrolide family of antibiotics along with erythromycin, which is an important alternative therapeutic in treating infections in humans [31]. Tylosin and erythromycin inhibit protein synthesis in bacteria by binding to the 50S ribosomal subunit, and one of the most widespread mechanisms of macrolide resistance in clinical isolates involves the methylation of a specific adenine residue (A2058) on the 23S rRNA [35, 36]. This methylation disrupts the binding of macrolides, as well as the structurally unrelated lincosamide and streptogramin B antibiotics [35, 36]. The RNA methylase genes responsible for this are known as erm genes [36] due to their initial characterization in erythromycin-resistant bacteria. However, erm genes can confer resistance to a broad range of antibiotics in the macrolide, lincosamide, and streptogramin B (MLSB) families [34]. For instance, erm(B) has been shown to confer resistance to multiple MLSB drugs [14, 20]. Exposure to any of the MLSB drugs may provide positive selection pressure for erm genes [34], and tylosin use in animal production has been shown to lead to increased levels of erythromycin resistance in bacterial isolates [5, 24]. Administration of tylosin to animals can also increase the incidence of bacterial resistance to unrelated antibiotics such as tetracycline [10].

Because tylosin usage in animal production has the potential to select for erm genes that confer multidrug resistance, it is important to understand the factors that contribute to the dissemination and persistence of these genes in environments associated with animal agriculture. These include factors that relate to dispersal and persistence of antibiotic-resistant bacteria to nearby environments, those that can increase the frequency of antibiotic resistance genes in host populations, and those that lead to the spread of these genes, via horizontal gene transfer (HGT), to new host species. Molecular methods, such as quantitative polymerase chain reaction (qPCR), can be used to detect patterns in gene distribution without the need to cultivate or identify host organisms. Thus, they are useful tools to characterize the emergent spatial and temporal patterns of antibiotic resistance that result from: (1) gene selection, (2) population dynamics of host organisms, (3) movement of host organisms to new environments, (4) transmission of mobile genetic elements, or (5) some combination of these phenomena. An understanding of these patterns can help to identify factors that contribute the spread of antibiotic resistance, especially when these patterns can be related to environmental drivers or potential sources of resistance genes in the landscape.

Our objective in this work is to characterize spatial and temporal patterns of RNA methylase gene distributions in agriculture-associated environments. We report on a 3-year, farm-scale study investigating the distribution of eight different RNA methylase genes in waste treatment lagoons and groundwater at two swine production facilities with different hydrogeology (Fig. 1). Phylogenetic analyses were used to guide primer design for culture-independent detection and quantification of these genes. By comparing genes found in lagoons and wells with those of on-site control wells, we evaluate the hypothesis that hog waste lagoons can be sources of antibiotic resistance for agriculture-associated environments. In addition, we draw upon long-term observations at these facilities [8, 19, 22, 23, 25, 28] to relate the patterns in erm gene distributions to previously described patterns of chemical contamination [22, 23, 25, 28] and the occurrence of genes conferring resistance to tetracycline antibiotics [8, 21].
https://static-content.springer.com/image/art%3A10.1007%2Fs00248-009-9610-0/MediaObjects/248_2009_9610_Fig1_HTML.gif
Figure 1

Location of monitoring wells at sites A and C. Stratigraphic columns indicate the locations and characteristics of sand layers. Large open arrows indicate the direction of groundwater flow, and small arrows show direction of stream flow. The locations of monitoring wells are indicated by circles. Numbers in parentheses indicate well depths in meters. Wells A6, A8, A9, A11, and A12 are heavily impacted by lagoon leakage. Wells A1, A2, A7, and C1 are background control wells

Methods

Strains and Plasmids

The following positive control strains and plasmids were used for RNA methylase (erm and tlr) genes: erm(A), Tn554 in chromosome of Staphylococcus aureus RN1389 [40]; erm(B), pAC1 in Streptococcus pyogenes AC1 and pJH1 in Enterococcus faecalis JH2-2 [40]; erm(C), pE194 in S. aureus RN4220 [40]; erm(F), pBF4 in Bacteroides fragilis V-479-1 [33]; erm(G), Tn7853 in chromosome of Bacteroides thetaiotaomicron [12]; erm(Q), plasmid JIR432 in Clostridium perfringens [6]; tlr(B), plasmid pSD14c in Streptomyces fradiae [27]; tlr(D), on plasmid pSD57c in S. fradiae [7]. These strains were kindly provided by A. A. Salyers (Department of Microbiology, University of Illinois at Urbana-Champaign, Urbana, USA) and S. Douthwaite (Department of Biochemistry and Molecular Biology, University of Southern Denmark, Odense M, Denmark). All strains were grown and maintained as described in the original reference.

Site Description

Samples were collected from groundwater and waste lagoons at two privately operated swine confinement facilities, referred to hereafter as sites A and C (Fig. 1). At the time of this study, the antibiotic regimens at both sites included chlortetracycline and tylosin (personal communication with facilities operators; specific dosages and the timings of administration were not revealed). For treatment of swine manure, both sites used unlined lagoons that were periodically pumped for application to crop fields. Site A was underlain by a fine-grained silt loam (low permeability), but its hydrogeology also included two shallow, highly permeable sand layers (Fig. 1). Site C was underlain by low-permeability silt loam (Fig. 1). Thus, the two sites differed significantly in terms of groundwater permeability in the vicinity of waste lagoons. Sixteen wells at site A and six at site C were previously installed in locations around the lagoons [25]. The locations of these wells (Fig. 1) were meant to capture a range of suspected waste contamination from lagoons, based on electromagnetic terrain conductivity surveys [25]. On-site environmental control wells (three wells at site A and one at site C) were installed above the waste lagoons in the hydrological flow path (Fig. 1); these wells would not receive any groundwater flowing from lagoons but could potentially receive antibiotic resistance genes originating from other sources in the landscape. More detailed descriptions of the operations and hydrogeology of these sites have been described elsewhere [8, 25, 28].

Chemical measurements taken from these wells since 1996 [22, 23, 25, 28] show that the following wells (Fig. 1) are significantly impacted by lagoon leakage: A6, A8, A9, A11, and A12, while wells A13, C2, C3, and C4 showed moderate or occasional leakage. In addition, wells A6, A8, A9, A11, and A12 have been found to frequently contain tetracycline resistance genes (tet) also found in the waste lagoons [21]. This led Koike and colleagues to classify wells A6, A8, A9, A11, and A12 as “impacted” by tet genes from the lagoon.

Microbial Community Samples and DNA Extraction

To compare the distribution of RNA methylase genes (erm and tlr) and tet genes, we use the same samples that were previously collected to investigate tetracycline resistance at these facilities [21, 22]. Briefly, groundwater samples from all wells and lagoons were collected in May 2001 (both sites), April (site A) or May (site C) 2002, and March (site A) or February (site C) 2003. To avoid sample cross-contamination, polyethylene bailers dedicated to each well were alcohol-sterilized and rinsed prior to sample collection, and wells were purged by removing 1.5 to three well volumes of groundwater before sample collection. Lagoon samples consisted of eight pooled 2-L subsamples collected from arbitrary locations around the lagoon [1]. All samples were stored in clean, sterilized bottles, kept on ice in the field, and refrigerated (4°C) until DNA extraction.

Sample processing and DNA extraction were described previously [21]. In summary, groundwater (1 L) and lagoon (100 mL) samples were centrifuged at 17,700×g for 20 min at 4°C. The pellets were retained and washed with a phosphate-buffered saline. Total DNA was extracted from the pellets using a freeze–thaw method [42]. The crude DNA was purified with polyvinylpolypyrrolidone and Sepharose 2B (Sigma Chemical, St. Louis, MO, USA) as described by Zhou [46] and Miller [30]. A DU 7500 spectrophotometer (Beckman, Fullerton, CA, USA) was used to measure DNA concentration, which was then adjusted to 10 ng/μL for each sample.

Primer Design and PCR

The nucleotide sequences of all methylase genes available at the time of primer design and conferring resistance to MLSB antibiotics were downloaded from the GenBank database [4]. Sequences were aligned using the multiple sequence alignment software ClustalX ver. 1.83 [41]. Based on phylogenetic analyses (see below), the corresponding clusters in the alignment file were inspected by eye to design PCR primers that are specific and cover all diversity within the sequences belonging to the same gene cluster (Table 1). These oligonucleotides were initially tested in silico using the blastn algorithm and nr nucleotide sequence database (GenBank) to ensure a 100% match with the genes of interest and to rule out false priming (i.e., no matches with other genes). For each gene, laboratory validation was performed with the corresponding positive control plasmid and strain, and all other methylase gene-bearing strains and plasmids were used as templates for negative controls. PCR was conducted using Amplitaq Gold DNA polymerase (Applied Biosystems, Foster City, CA, USA). The reaction mixture contained 0.5 μM of each primer, 1.5 mM MgCl2, 0.2 mM of each deoxyribonucleotide triphosphate, PCR buffer II, 1.25 U of Amplitaq Gold DNA polymerase and 10 ng of template DNA in a total volume of 25 μL. Amplification of all genes began with denaturation at 94°C for 10 min. With the exception of erm(Q), denaturation was followed by 40 cycles of 30 s at 94°C, 30 s at the annealing temperature (Table 1), and 30 s of extension at 72°C. A final extension of 72°C for 10 min was used for all samples. For erm(Q), a step-down PCR was employed to eliminate primer dimer formation: 35 cycles were performed using an annealing temperature of 68°C, followed by 15 cycles using an annealing temperature of 60°C; because of this potential primer dimer difficulty, erm(Q) primers were not used for qPCR. PCR product aliquots were analyzed by electrophoresis on 2.0% (wt/vol) agarose gel followed by staining with ethidium bromide.
Table 1

PCR primers targeting methylase genes

Primer

Class targeted

Sequence (5′–3′)

PCR annealing temp (°C)

Amplicon size (bp)

Reference

Erm A-FW

erm(A)

AGTCAGGCTAAATATAGCTATC

63

157

This study

Erm A -RV

CAAGAACAATCAATACAGAGTCTAC

Erm B-FW

erm(B)

GGTTGCTCTTGCACACTCAAG

65

191

This study

Erm B-RV

CAGTTGACGATATTCTCGATTG

Erm B-1a

erm(B)

GAAAARGTACTCAACCAAATAATAA

57

639

Sutcliffe et al. [40]

Erm B-2

AGTAAYGGTACTTAAATYGTTTAC

Erm C-FW

erm(C)

AATCGTGGAATACGGGTTTGC

63

293

This study

Erm C-RV

CGTCAATTCCTGCATGTTTTAAGG

Erm F-FW

erm(F)

TCTGGGAGGTTCCATTGTCC

65

424

This study

Erm F-RV

TTCAGGGACAACTTCCAGC

Erm G-FW

erm(G)

GTGAGGTAACTCGTAATAAGCTG

63

255

This study

Erm G-RV

CCTCTGCCATTAACAGCAATG

Erm Q-FW

erm(Q)

CACCAACTGATATGTGGCTAG

68, 60

154

This study

Erm Q-RV

CTAGGCATGGGATGGAAGTC

Tlr B-FW

tlr(B)b

GTGTCCTGGAGGAGTTCGAG

63

111

This study

Tlr B-RV

AGCGGAAGTGTGTCCCATAC

Tlr D-FW

tlr(D)b

GTCAACGACGACTTCACGAC

65

186

This study

Tlr D-RV

ACTGGGCGTTGAAGAGATTG

aUsed in sequence analysis

bTlr(B) is also known as erm(32); tlr(D) is also known as erm(N)

PCR was performed on all groundwater and lagoon samples to detect the presence of the six erm and two tlr genes reported in Table 1. For any environmental sample that did not show amplification of the target gene, PCR was repeated using serially diluted positive control plasmid (see below) spiked into the sample to rule out the possibility of PCR inhibition. However, we found no evidence of PCR inhibition in any sample.

Quantitative PCR Assays

All groundwater and lagoon samples were used as templates for quantification of erm(A), erm(B), erm(C), erm(F), and bacterial 16S rDNA genes. We did not perform qPCR on erm(G) or the tlr genes due their low frequency of occurrence (see “Results”), and qPCR was not performed for erm(Q) due to the aforementioned potential for primer dimer formation. qPCR conditions were identical to those described for PCR (Table 1), and 16S rDNA PCR conditions followed Koike et al. [21]. Positive control plasmids for qPCR standards were prepared by cloning PCR product from positive control strains using the primers and conditions of Table 1. Briefly, PCR products were excised and purified from agarose gel with the QIAquick gel extraction kit (QIAGEN, Valencia, CA, USA), and then ligated into the pGEM-T EASY vector (Promega, Madison, WI, USA) and transformed into competent Escherichia coli JM109. After confirming the proper-size insert with PCR, the recombinant plasmids were purified with the QIAprep spin miniprep kit (QIAGEN). The concentration of each plasmid was determined by spectroscopy, and the plasmid copy number was calculated based on the length (base pair) and the molecular weight of DNA. Tenfold serial dilution of each plasmid was then used to create qPCR standards. qPCR of well and lagoon samples was done using a GeneAmp 9600 thermal cycler in conjunction with the GeneAmp 5700 sequence detection system (Applied Biosystems, Foster City, CA, USA). The SYBR GREEN PCR Master Mix (Applied Biosystems) was used with 10 ng of template DNA for all reactions, and the plasmid standards from tenfold serial dilution was run for each set of reactions for all genes. Duplicate reactions were run for each sample. Data analysis was performed using the GeneAmp 5700 SDS software (Applied Biosystems). For each sample, the number of copies of erm was normalized against the copy number of 16S rDNA.

Sequencing of Environmental erm(B)

Comparative analysis of erm(B) sequences from sites A and C was conducted using the primer set of Sutcliffe and colleagues [40], which amplifies 639 bp of erm(B). Samples from 2003 of wells A9, A11, and the lagoon at site A and wells C1, C4, and the lagoon at site C were used to construct the erm(B) clone libraries using the pGEM-T EASY system as described above. Ten clones from each well/lagoon were randomly selected for sequencing, which was performed by the W.M. Keck Center for Comparative and Functional Genomics at the University of Illinois. Sequences were compared with available sequences of erm(B) in the GenBank database using BLAST, and phylogenetic analysis by Bayesian inference was performed (see below). All sequences generated in this study have been deposited in GenBank under accession numbers (EU168273–EU168331).

Phylogenetic and Numerical Analyses

Maximum likelihood (ML) and Bayesian inference (BI) were used to infer phylogenetic trees and estimate the clade support. ML analysis included the run of 100 resampled trees using DNAML in the PHYLIP package, v.3.6 (http://evolution.genetics.washington.edu/phylip.html). For BI analysis, posterior probabilities were calculated using a Bayesian Markov chain Monte Carlo method with MrBayes v.3.1.1 program [18, 37], using a 4 by 4 nucleotide substitution model with gamma-distributed site variation and a proportion of invariable sites. The priors were left at the defaults (i.e., uninformative priors) in MrBayes. Ten or 20 million generations were run, with sampling every 100 generations.

Patterns in gene occurrence in wells and lagoons were investigated in a multivariate framework based on the results of PCR screening. In each sample, all erm genes were scored as “present” or “absent,” and the resulting arrays for all sample pairs were compared using the “Jaccard” (dis-)similarity coefficient, which is commonly used for ecological comparisons of community composition for species presence/absence data [26]. For any given pair of samples, the \( {\text{Jaccard coefficient}} = 1 - \left[ {a/\left( {a + b + c} \right)} \right] \), where a is the number of erm genes found in both samples and b and c are the number of genes found in only one sample. The resulting dissimilarity matrix was used to cluster samples using average-linkage agglomerative clustering. A Jaccard dissimilarity matrix was also computed for tet genes in these samples using the data of Koike and colleagues [21], and the erm and tet data were compared with the Mantel test [26]. To determine whether certain genes had a tendency to co-occur in wells, the Jaccard coefficient was used as a measure of association (i.e., “R-mode” sensu Legendre and Legendre [26]) for erm and tet genes across all groundwater samples, and the resulting matrix was clustered by average linkage. All numerical analyses were carried out using the R environment for statistical computing [32] with functions provided by packages “vegan” and “labdsv.”

Results

Phylogenetic Analysis

Maximum likelihood and Bayesian inference produced erm gene trees with similar topology and high clade support; only the Bayesian tree is presented here (Fig. 2). With a 100% clade support (posterior probability), this analysis shows that RNA methylases are organized into two major clusters, one of which is represented by high-G + C bacteria, including antibiotic-producing Streptomyces, and another by low-G + C commensal, pathogenic, and environmental bacteria (Fig. 2). PCR primers (Table 1) were designed to amplify six MLSB methylases found in cluster 1 (erm(A), erm(B), erm(C), erm(F), erm(G), and erm(Q)) and two genes from cluster 2 (tlr(B) and tlr(D)). In laboratory validation tests of PCR primers (Table 1), amplicons of the expected sizes were produced with positive controls, and product was not detected using incongruent primer–template combinations.
https://static-content.springer.com/image/art%3A10.1007%2Fs00248-009-9610-0/MediaObjects/248_2009_9610_Fig2a_HTML.gifhttps://static-content.springer.com/image/art%3A10.1007%2Fs00248-009-9610-0/MediaObjects/248_2009_9610_Fig2b_HTML.gif
Figure 2

Phylogenetic tree (Bayesian inference) of MSLB resistance genes encoding methylases. The posterior probabilities of clades are shown as interior node labels, with values less than 1.00 omitted in the figure. The scale bar shows expected changes per site; however, branch lengths of cluster 2 and two S. fradiae strains are not in scale. GenBank accession numbers of sequences used in this analysis are given in parenthesis. Tlr(B) is also known as erm(32); tlr(D) is also known as erm(N)

Distribution of erm Genes on Swine Farms

The lagoons were always found to contain erm(A), erm(B), erm(C), and erm(F) (Table 2). Erm(Q) was found in all lagoon samples except one (site C in 2003; Table 2). In lagoons, erm(G) was only found at site C in 2002 and site A in 2003. The cluster 2 genes (tlr) were never found in lagoons. Erm(A), erm(B), erm(C), and erm(F) were found in multiple site A and C wells (Table 2), and these four genes were significantly more likely to occur in wells previously reported to be impacted by lagoon leakage [2123, 25, 28] as compared with other wells (Fisher’s exact test p values >0.001 for erm(A), erm(B), and erm(C); 0.007 for erm(F)). With only two exceptions, all samples from impacted wells contained erm(B) and erm(C) (Fig. 3). Neither erm(A) nor erm(F) was found outside of waste lagoons or impacted wells (Table 2; Fig. 3). The precise combination of genes in wells A6, A9, and A11 was identical in any given year (Fig. 3). At site C, where the hydrological setting limits lagoon leakage (Fig. 1), the detection frequencies of all genes in wells were not statistically distinguishable from those of background control wells.
Table 2

Proportion of water samples testing positive for cluster 1 MLSB genes

Samplesa

erm(A)

erm(B)

erm(C)

erm(F)

erm(G)

erm(Q)

Site A lagoons

 

2001

+

+

+

+

+

2002

+

+

+

+

+

2003

+

+

+

+

+

+

Site C lagoons

 

2001

+

+

+

+

+

2002

+

+

+

+

+

+

2003

+

+

+

+

Site A wells

Impactedb

2001 (n = 5)

0.60

0.80

0.60

2002 (n = 5)

1.00

1.00

1.00

-

-

0.20

2003 (n = 5)

0.80

1.00

0.80

0.60

-

-

Nonimpacted

2001 (n = 9)

2002 (n = 11)

0.18

0.18

0.09

2003 (n = 11)

0.18

0.27

Site C wells

 

2001 (n = 5)

0.2

0.2

2002 (n = 6)

0.17

0.50

0.17

 

2003 (n = 6)

0.50

0.67

Other numbers are [positive detections/n]

no detection in lagoon or any wells, + detection in lagoon

aNumber in parenthesis indicates total groundwater wells sampled; n = 1 for lagoon

b“Impacted” wells were previously determined to receive chemical and biological contamination from lagoon leakage

https://static-content.springer.com/image/art%3A10.1007%2Fs00248-009-9610-0/MediaObjects/248_2009_9610_Fig3_HTML.gif
Figure 3

Clustering of the gene complement of samples based on Jaccard’s coefficient. Clustering method was agglomerative, with nodes assigned by average linkage. All samples with at least one erm gene present are included. Well numbers are identical to those in Fig. 1, and numbers in parenthesis indicate the year in which samples were collected. The combination of genes present in each cluster is indicated at the nodes. Wells heavily impacted by lagoon leakage are indicated with asterisks. Lagoon samples (“lag”) are in boldface

Quantification of erm Genes

In laboratory validation tests, all primer sets detected single copies of their targets with the exception of erm(Q), which had a detection limit of 100 copies (data not shown). We restricted our quantification of environmental genes to the empirically determined linear amplification range of our qPCR standards. For each of the four primer sets used for qPCR, the linear range of quantification was as follows: 59 to 5.9 × 108 copies for Erm A-FW/RV (regression coefficient 0.9913, 99.7% amplification efficiency); 40 to 4.0 × 108 copies for Erm B-FW/RV (regression coefficient 0.9936, 100.6% amplification efficiency); 34 to 3.4 × 108 copies for Erm C-FW/RV (regression coefficient 0.995, 99.3% amplification efficiency); 36 to 3.6 × 108 copies for Erm F-FW/RV (regression coefficient 0.9935, 102.3% amplification efficiency).

Despite the presence of erm genes in multiple wells, the copy numbers of genes in water samples was nearly always below the empirically determined quantification limit (e.g., <59 copies for erm(A)). Erm(A), erm(B), or erm(C) copy numbers were within the linear quantification range in only nine samples (Fig. 4), and all erm(F) detections were below the limit of quantification. In general, erm(A), erm(B), and erm(C) were quantifiable in lagoon samples (Fig. 4), where erm(B) was the most abundant.
https://static-content.springer.com/image/art%3A10.1007%2Fs00248-009-9610-0/MediaObjects/248_2009_9610_Fig4_HTML.gif
Figure 4

Relative copy number of MSLB resistance genes encoding methylases from selected environmental samples. Only samples with quantifiable gene copy numbers are shown. “A-lag” and “C-lag” refer to waste lagoons from sites A and C, respectively, and well IDs are as indicated in Fig. 1. Error bars represent standard deviations for qPCR estimation. Asterisk, bars reaching the dashed line indicate samples that were PCR positive for the respective gene, but relative copies numbers were below the lower limit of quantification (Table 1)

Co-occurrence of erm and Tetracycline Resistance Determinants

Using samples previously screened for tet genes [21], we found a significant but low correlation (Mantel’s R = 0.39, p < 0.001) between tet and erm genes in groundwater wells. Exclusion of impacted wells from the analysis eliminates this correlation (R = −0.0035, p = 0.487). Considered at the farm scale, the specific combination of tet genes found in any well was not a good predictor of which erm genes may be present, as demonstrated by the lack of strong associations of erm and tet genes across well samples (Fig. 5).
https://static-content.springer.com/image/art%3A10.1007%2Fs00248-009-9610-0/MediaObjects/248_2009_9610_Fig5_HTML.gif
Figure 5

Association of methylase genes for MSLB resistance and tetracycline resistance genes. The association statistic is Jaccard’s coefficient based on gene presence/absence across all well samples. Clustering method was agglomerative, with nodes assigned by average linkage. Tet(M), tet(O), tet(Q), and tet(W) encode tetracycline resistance by ribosomal protection; tet(C), tet(H), and tet(Z) encode tetracycline resistance by efflux pump

Discussion

This study utilized phylogenetic reconstruction of RNA methylase genes to aid in primer design and to identify gene targets of both evolutionary and environmental relevance. Based on this phylogeny, we hypothesized that the presence of different cluster 1 and cluster 2 methylases may be indicative of different sources of antibiotic resistance. Cluster 1 genes like erm(B), erm(C), erm(G), and erm(F), and erm(Q) have been recovered from both soil and commensal/pathogenic bacteria, while erm(A) has generally been recovered from clinically relevant strains [35, 36]. Furthermore, cluster 1 genes show a high degree of sequence similarity within the principal gene subclusters; for example, nearly identical erm(B) sequences have been recovered from pathogens (e.g., Streptococcus pyogenes), commensals (e.g., Lactobacillus), and soil bacteria (e.g., Bacillus cereus; Fig. 2). This is indicative of widespread horizontal gene transfer, as has been previously suggested to occur within the GI tract ecosystem [38], where the close association of these organisms forms a “genetic exchange community.” The second cluster (cluster 2; Fig. 2) is deeply separated from the first and is composed of two subclusters, one of which contains genes found primarily in MLSB antibiotic-producing strains. Aminov and Mackie [3] have recently observed that tetracycline and vancomycin resistance genes show this same phylogenetic separation of genes from pathogens/commensals versus those found in producers of the respective antibiotics. The deep-branching structure of these gene trees (e.g., Fig. 2) suggests that diversification within these gene families occurred prior to modern human antibiotic use, with little evidence of recent transfer from antibiotic-producing bacteria to clinically relevant organisms. Thus, a prevalence of cluster 2 genes is to be expected in the presence of large native populations of antibiotic-producing organisms, while a prevalence of cluster 1 genes is suggestive of other sources, such as antibiotic use on site or elsewhere in the landscape [39].

In this study, the cluster 1 methylases were predominant among MLSB-resistant bacteria associated with animal waste. This is consistent with previous observations [3, 11, 34]. In a similar study that included Ohio swine farms, Chen and colleagues [11] found erm(A), erm(B), and erm(F) in all swine manure (n = 16) and swine lagoon (n = 6) samples examined, but erm(C) was found with reduced frequency in both manure and lagoon samples. The increased prevalence of erm(C) in waste lagoons in the present study suggests that there are differences in the autecology of erm(C)-hosting organisms in Ohio and Illinois herds. However, methodological differences between these two studies may also have been partly responsible. For example, different erm(C) PCR primers and protocols were by Chen and colleagues [11], and this may have led to differential specificity, sensitivity, or amplification efficiency as compared to the current study. Additionally, Chen and colleagues [11] used sample-derived, linear qPCR standards as opposed to the plasmid-based standards used here, and this may have caused differences in qPCR calibration. Thus, proper caution is needed when comparing the results of any qPCR-based environmental studies.

Of the cluster 2 genes, tlr(D) was found only once in a 2003 groundwater sample (well C3), and tlr(B) was never detected (data not shown). This suggests that these farms may lack significant populations of tylosin-producing S. fradiae; however, a more thorough investigation of cluster 2 genes is necessary to quantify the potential influence of native MSLB-producing bacteria in the soil and groundwater at these sites. Additionally, it must be noted that the genes targeted in the present study represent a subset of the possible macrolide resistance mechanisms [13].

The cluster 1 genes erm(A), erm(B), erm(C), and erm(F) were also the most commonly detected genes in groundwater wells. In particular, these genes were frequently detected in a spatially explicit fashion, occurring in shallow wells close to the site A lagoon (Figs. 1 and 3) and that have been previously shown to be impacted by lagoon leakage [2123, 25, 28]. In addition, the impacted wells A6, A9, and A11 had the same combination of erm genes in any given year (Fig. 3). These spatial and temporal correlations implicate lagoon leakage as the principle source of these four genes in groundwater wells.

On the other hand, erm(G) and erm(Q) were infrequently detected in wells (Table 2) and did not show any notable spatial or temporal distribution patterns (Fig. 3). A likely explanation for this infrequent detection outside of lagoon water is that the anaerobic organisms that host these genes (Fig. 2) have not been able to effectively disperse to the groundwater environment. Alternately, the sporadic appearance of these genes and their lack of spatial association with waste lagoons suggest that they may originate from a nonlagoon source (or multiple sources). On three occasions, erm(C) was detected in background control wells A1, A2, and A7 (Fig. 3), which strongly points to an “upstream” source for this gene. Singer and colleagues [39] argue that antibiotic resistance may be more properly viewed as a landscape-level problem, and multiple potential sources of antibiotic-resistant organisms and their genes should be considered for a full accounting of their occurrence. Our data suggest that there is another source of RNA methylase genes in the landscape, or there may be a low-level “background” presence of these genes in native bacteria in the environment. The frequent occurrence of erm(A), erm(B), erm(C), and erm(F) in impacted groundwater wells is a spatially defined signal imposed upon this background.

The most frequent and abundant gene at these sites was erm(B) (Figs. 3 and 4), which was present in lagoon water from 4 × 102 to 104 copies per million copies of bacterial 16S rDNA. This is roughly an order of magnitude lower than the abundance of erm(B) in Ohio swine lagoons as determined by Chen and colleagues [11]. As it was below the quantification range of our qPCR, erm(F) abundance in sites A and C lagoons was at least three orders of magnitude lower than that determined by Chen and colleagues [11], whereas lagoon erm(C) was higher in the present study, and erm(A) was comparable. As before, these differences may be methodological (primers and/or qPCR standards), or they may represent regional/ecological differences in the microbial assemblages under consideration.

With the exception of three samples from wells A8, A9, and A11, erm gene copy numbers in all well samples were below the qPCR range of quantification. For erm(A), erm(B), and erm(C), this corresponds to a reduction in numbers by at least three orders of magnitude in groundwater as compared to lagoons. A likely explanation of this reduction is due to limited dispersal of erm-hosting bacteria to groundwater. Previous investigation of tetracycline resistance genes (tet) at these sites [21] found an average tenfold reduction of tet in groundwater as compared to lagoons. This may reflect differential movement or persistence of erm-harboring bacteria in groundwater as compared with tet-harboring bacteria. Alternately, as macrolide antibiotics tend to have shorter half-lives in manure than tetracyclines [9], this may reflect a lower selection pressure for the maintenance of erm genes than tet genes, assuming that manure or lagoon water is the primary source of antibiotics at these sites. Using liquid chromatography with tandem mass spectroscopy, we were unable to detect macrolide antibiotics in any of the groundwater samples analyzed in this study, and macrolides were detected in two lagoon samples at each site, at concentrations of <3 mg/L (data not shown).

Resistance genes for different antibiotics have been observed together in the same organism and even on the same genetic element [3, 16, 34, 38], and this raises that possibility that selection due to one antibiotic can coselect for resistance genes to another [39]. For example, an increase in tetracycline resistance gene abundance was recently reported in fecal bacteria of cattle fed subtherapeutic doses of tylosin but not tetracycline [10]. In the present study, erm and tet genes were mildly correlated only if impacted wells were considered (Fig. 5). Therefore, it is unlikely that co-occurrence of these genes in the same host organism or on the same mobile genetic element was principally responsible for the patterns identified; instead, passive leakage from the lagoon, leading to a high diversity of both tet and erm genes in impacted wells, is the likely cause of this correlation. The strongest patterns of co-occurrence, which have been previously noted [21], were among the four ribosomal protection genes tet(M), tet(O), tet(Q), and tet(W) and also among the three efflux pump genes tet(C), tet(H), and tet(Z) (Fig. 5). There was a tendency for erm(B) to co-occur with the efflux tet genes and for erm(C) to co-occur with tet genes as a whole (Fig. 5). Erm(A), erm(F), erm(G), and erm(Q) did not show strong associations with any other genes. The overall lack of strong associations between erm and tet genes suggests that coselection is not a likely explanation for the patterns reported here. This also suggests that the distribution of genes in these systems reflects the individual ecological limitations of host species and differences in transmissibility and host spectrum of resistance-conferring mobile genetic elements. However, a more explicit focus on physical linkages between resistance genes (e.g., accumulation on mobile genetic elements) should be the focus of future environmental research regarding the spread of antibiotic resistance.

Ecological Perspectives on Gene Distribution

The distribution of antibiotic resistance genes at these farms suggests that a strong environmental signal (the lagoon) is superimposed upon weaker, more ephemeral background signal. The former has a clear spatial relationship with the lagoon (Table 2; Fig. 3), while genes involved in the latter exhibit more individualistic spatial and temporal patterns. The distributions of these genes at these sites are emergent features of factors that influence gene frequencies in host populations, movement of resistant organisms to new environments, and transmission of genes via HGT to new organisms. Previous authors [3, 38, 39] have suggested that studies of antibiotic resistance in the environment should focus on gene distributions in much the same way that ecologists study the distributions of organisms. In the present work, we also considered the evolutionary history of these gene families, and we used gene phylogeny to help interpret patterns of gene occurrence. For example, the lack of cluster 2 tlr genes suggests that the presence of native tylosin-producing bacteria is not a major selective force for MLSB resistance at these sites.

Erm(B) and erm(C) were the most frequently detected and most abundant genes in this study (Table 2; Fig. 4). Erm(B) has been frequently found in MLSB-resistant bacteria associated with animals and animal waste [3, 14, 15, 20], and it was the most common and abundant erm gene found in swine manure samples in a previous study [11]. However, erm(B) has also been found in soil bacteria such as Bacillus (Fig. 2; [34, 35]). We obtained DNA sequences of erm(B) from lagoon, background well, and impacted well samples from sites A and C, but all erm(B) sequences were over 99% similar by nucleotide identity (data not shown). Thus, it was not possible to identify distinct clusters of “native” erm(B) from “swine” erm(B) as was previously seen for tet(W) at these sites [21]. The widespread distribution of nearly identical erm(B) sequences in a variety of species (Fig. 2) means that we cannot definitely conclude that erm(B) in our groundwater samples came from a single source population.

Erm(C) has also been found in both animal-associated and soil bacteria (Fig. 2; [34, 35]). While erm(C) tended to co-occur with erm(B) in impacted wells, it was also observed more frequently in other well samples, including background wells A1, A2, A7, and C1 (Fig. 4). This may indicate the presence of alternative sources of erm(C) at these sites, for instance, native populations of erm(C) soil or groundwater bacteria. If true, this could explain the higher detection frequency and abundance of erm(C) in this study in comparison with previous work [11]. These types of gene ecological considerations should help inform future studies of antibiotic resistance in the environment and the impact of concentrated animal feeding operations.

Acknowledgements

This research was supported by funding from the USDA NRI Competitive Grants Program 26.0 (award nos. 2001-35102-10774 and 2005-35102-16424). Support was also provided by the Scottish Government Rural and Environmental Research and Analysis Directorate and by Hatch funding to the Agricultural Experimental Station at the University of Illinois at Urbana-Champaign.

Copyright information

© Springer Science+Business Media, LLC 2009