1 Introduction

Pollution by human and animal wastes is a common threat to water resources. Untreated human sewage frequently enters waterways as point source discharges (intentional or accidental) or via combined sewer overflows (CSO) from wastewater treatment plants (WWTPs; Stoner 2005; Whitlock et al. 2002; Wiggins et al. 1999). For example, from January 1, 2002, to December 31, 2005, WWTPs across Michigan reported having 2,542 CSO events, totaling 270 billion liters of discharged sewage (Michigan Department of Environmental Quality 2004a, b, 2005). Animal waste seeps into surface water as non-point source pollution as a result of housing large numbers of animals in confined areas, or as runoff from liquid manure applications onto croplands (a primary method of disposing of the waste generated by confined animal feeding operations (CAFOs); Carpenter et al. 1998; Esiobu et al. 2002).

Both human and animal wastes add nutrients (e.g., nitrogen, phosphorus; Carpenter et al. 1998; Tabbara 2003) as well as fecal coliform bacteria (Esiobu et al. 2002) to receiving waters. Excessive nutrients promote eutrophication, which decreases water transparency, creates foul odor and taste, depletes oxygen, causes fish kills, reduces biodiversity, and decreases esthetic, recreational, and property values along waterways (Boesch et al. 2001; Carpenter et al. 1998; Wetzel 2001). To reduce eutrophication, nutrient levels are often controlled via regulation and mitigation treatments (e.g., Carpenter et al. 1998; Cooke et al. 2005; Wetzel 2001). Human and animal wastes may also contaminate waterways with pathogenic bacteria (Burton and Engelkirk 2004; Guan and Holley 2003; Pell 1997). To avoid the difficulty and expense of direct testing for all possible types of pathogenic bacteria transmittable by feces, standard microbiological examination of water centers on determining the abundance of indicator bacteria, such as the fecal coliform bacteria (American Public Health Association 1998; Griffin et al. 2001). These bacteria are commonly found in the gut and feces of warm-blooded animals, and their presence in water environments can indicate contamination with human or animal feces (and by extension, contamination of water by pathogenic bacteria possibly present in feces; American Public Health Association 1998; Ash et al. 2002; Chao et al. 2003; Griffin et al. 2001). The greater the fecal coliform count, the greater the probability there is of contracting diseases from waterborne pathogenic bacteria (Mitchell and Stapp 1997). In Michigan, the Michigan Department of Natural Resources and Environment (MDNRE, formerly Michigan Department of Environmental Quality) monitors surface waters for a specific fecal coliform bacterium (Escherichia coli), rather than the entire fecal coliform group (Michigan Department of Environmental Quality 1999). In contrast, some other states in this region monitor fecal coliforms in surface waters. Both Ohio and Illinois require that surface waters used for extensive contact (e.g., swimming) have no more than 10% of samples exceeding a maximum fecal coliform count of 400 colony-forming units per 100 ml (Illinois Pollution Control Board 2002).

Antimicrobial agent resistance is an emerging global concern to both public and veterinary health. Antibiotics are heavily used to treat disease in both humans and animals, and there is a pattern of antibiotic resistance and transfer emerging among bacterial populations in proportion to the use of antibiotics, especially in agriculture (Levy 1997; Oppegaard et al. 2001). Antibiotics have been added to animal feed as growth promoters for some time, and the animal production industry has been identified as a potential reservoir for resistant Enterobacteriaceae (Witte 1997). The use of antibacterial drugs for prophylactic or therapeutic purposes in humans and for veterinary and agricultural purposes has provided selective pressure favoring the survival and spread of resistant organisms (Hooper 2002; Mascaretti 2003; Taber 2002). These resistant bacteria may transfer their resistance to previously non-resistant pathogenic bacteria or directly infect humans with bacterial diseases that cannot be treated by conventional antimicrobial therapies (Khachatourians 1998). The potential for antibiotic exposure and resistance development in human and animal gastrointestinal tracts, coupled with relatively great abundance in waters contaminated with human and animal waste, makes the fecal coliform bacteria a logical focal group for studies of antibiotic resistance and transfer in aquatic environments.

Despite active research investigating how antibiotic resistance enters and is maintained in the environment (D’Costa et al. 2006; Esiobu et al. 2002; Graves et al. 2002; Iwane et al. 2001; Rice et al. 1995; Sayah et al. 2005; Schwartz et al. 2003; Whitlock et al. 2002; Witte 1997), determining the mechanisms by which bacteria can transfer resistant genes within and across species (Arana et al. 1997, 2001; Bell et al. 1980; Oppegaard et al. 2001; Salyers et al. 2002), and identifying and characterizing new antibiotic resistance genes (Aminov et al. 2001, 2002; Chopra and Roberts 2001; Davies 1997; Gevers et al. 2003; Roberts 2005) and integrons (France et al. 2005; Hall 1997; Hall et al. 2003), the global ecological impact of antibiotic resistance and risks to human and veterinary health are yet to be determined. In the USA, there are no national or state water quality regulations requiring testing or reporting on antibiotic-resistant organisms (American Public Health Association 2005; Michigan Department of Environmental Quality 1999; US Environmental Protection Agency 2002).

In order to understand how CAFOs and WWTPs affect bacterial antibiotic resistance and transfer, we simultaneously conducted traditional water quality measurements (pH, temperature, turbidity, total phosphorus, and fecal coliform density) and microbiological and molecular techniques for determining patterns of antibiotic resistance and potential transfer of resistance genes at CAFO-impacted and unimpacted sites and at sites upstream and downstream of WWTPs. We also examined the relationships between traditional water quality parameters and conducted a preliminary analysis of the genetic location and mode of transfer of antibiotic resistance elements.

2 Materials and Methods

2.1 Study Design and Site Descriptions

Ten sites were selected for this study, based on the presence of previously documented point source and non-point source contamination (see following section) and proximity to public access (Table 1). Six sites were located in an agricultural region near Hudson, MI. Three of the agricultural sites were waterways approximately 1–2.5 km from different CAFO farms and were classified as agriculturally impaired (AI) based on the following criteria: (1) the waterways received direct runoff from fields sprayed with liquid manure from CAFO farms; (2) previous observations of fecal coliform densities greater than 1,000 colony-forming unit (CFU)/100 ml (Khachatourians 1998; J. Pernicano personal communication); (3) ratings of “poor” on biological (macroinvertebrate) assessments of these waterways (Michigan Department of Environmental Quality 2003a); and (4) the MDNRE listing these waterways on their non-attainment and total maximum daily load lists (Michigan Department of Environmental Quality 2004b). Three additional agricultural sites located in the same farming region were classified as agriculturally unimpaired (AUI) since they did not receive direct runoff from manure-sprayed fields and were not located near a CAFO farm. Two of these three sites (sites SJT8 and SJC7) have been classified by MDNRE as “least impacted” streams in this location and have been previously identified as reference streams (Michigan Department of Environmental Quality 2003a). Four additional sites were selected approximately 1–1.5 km both upstream and downstream of the Ann Arbor, MI, and Chelsea, MI, WWTPs in the Huron River watershed. Like all WWTPs, these facilities occasionally release partially treated and untreated sewage. The most recent such episode prior to this study was the release of ∼50 million liters of sewage by the Ann Arbor WWTP in August 2003 (Michigan Department of Environmental Quality 2003b).

Table 1 Site descriptions: designations, locations, and average discharge
Full size table

2.2 Sampling and Physical and Chemical Water Quality Procedures

The initial and final sampling dates (June 3, 2004, and August 9, 2004, respectively) were chosen during distinct dry periods (zero precipitation measured within the proceeding 72 h), and the middle three sampling periods (June 16, 2004, July 7, 2004, and July 18, 2004) occurred after rainstorms where >1 cm of precipitation was recorded within 24 h (Michigan State Climatology, http://climate.geo.msu.edu/Semcog/SEMdaily/2004/, SEMCOG Daily Precipitation Summary). Site VH25 was completely dry by the end of the study, so this site was sampled on only four out of five possible dates.

At each site, a handheld probe was used to measure pH and temperature (YSI 63, Yellow Springs Instrument Co., Yellow Springs, OH). Turbidity was measured in the field using a Hach 2100P portable turbidimeter. For total phosphorus measurements, 100 ml of water was collected into an acid-washed Nalgene bottle and stored at −20°C for later analysis using the protocol of Wetzel and Likens (2000). For fecal coliform isolation, 100 ml of water was collected into two sterile 50-ml Falcon conical tubes and kept on ice (<8 h) until it was processed in the laboratory (Harwood et al. 2000). All water samples were obtained by submerging the collection container approximately 1 cm below the water surface before opening the lid to collect the sample.

Fecal coliform density was determined as follows: water samples of various volumes (100, 1.0, 0.1, and 0.01 ml) were brought to a final volume of 100 ml with 0.01 M phosphate-buffered saline (0.138 M NaCl and 0.0027 M KCl, pH 7.4 at 25°C) and filtered onto a 0.45-μm filter (Pall Life Sciences, East Hills, NY) using a vacuum system as described by Mitchell and Stapp (1997) and the Standard Methods for the Examination of Water and Wastewater (American Public Health Association 1998). The filter membranes were transferred to Petri plates (BD Falcon, BD Biosciences, Bedford, MA) containing fecal coliform selective agar (modified fecal coliform, mFC; BBL, BD Biosciences, Franklin Lakes, NJ) without antibiotics and were incubated at 44.5°C in a humidified incubator for 24 h. For the first sampling date (June 3, 2004), 100-, 0.1-, and 0.01-ml volume water samples were used; filtration of 100 ml yielded colonies with confluent growth, while the low-volume samples (0.1 and 0.01 ml) had no growth. As a result, fecal coliform densities from June 3, 2004, were not considered in any further analyses, and volumes of 100 and 1 ml were analyzed for the remaining four sampling dates (June 17, 2004, July 7, 2004, July 18, 2004, and August 9, 2004). After incubation, bacterial density (CFU/per 100 ml) was determined for each site on each sampling using the colony counts from either the 100- or 1-ml filtered plate, depending on which plate produced countable isolated colonies.

2.3 Bacterial Isolation and Antibiotic Resistance Susceptibility

Individual colonies were picked from distinctly isolated typical colonies as recommended by the Standard Methods for the Examination of Water and Wastewater (American Public Health Association 1998). Approximately 30 isolates per site per sampling date were analyzed. To test for multiple antibiotic resistance, individual bacterial isolates were replica-plated onto tryptic soy agar (TSA; BD Biosciences, Franklin Lakes, NJ) supplemented with one of five antibiotics: ampicillin (A), kanamycin (K), streptomycin (S), chlorotetracycline (C), and oxytetracycline (O) (Sigma-Aldrich, St. Louis, MO). These antibiotics are commonly used in both human and veterinary medicine and demonstrate different mechanisms and pathways (inhibiting cell wall synthesis [ampicillin] vs. protein synthesis [streptomycin, kanamycin, and the tetracyclines]; Esiobu et al. 2002). The antibiotic concentration for all drugs was 20 μg/ml for samples processed from the initial measurement date (June 3, 2004). The antibiotic concentrations were increased to 30 μg/ml for the remainder of the sampling dates (June 17, 2004, July 7, 2004, July 18, 2004, and August 9, 2004) to ensure complete inhibition of the control bacterial strain in the conjugation assay. Some authors consider these to be high antibiotic concentrations for screening environmental bacteria (D’Costa et al. 2006), but they are similar to concentrations used in other screening studies (i.e., 5–80 μg/ml; Harwood et al. 2000; Wiggins et al. 1999). The plates were incubated at 37°C for 24 h. Bacterial isolates were scored as positive for resistance if there was visible, tangible growth. No growth or trace colony development was recorded as antibiotic susceptible. All resistant isolates were preserved in tryptic soy broth (TSB, BD Biosciences, Franklin Lakes, NJ) supplemented with the appropriate antibiotic and 10% glycerol and frozen at −80°C.

2.4 Conjugation Testing

An S-resistant Salmonella typhimurium strain, EM1000, was used as the recipient in all conjugation testing. EM1000 was originally obtained as SGSC452 from the Salmonella Genetic Stock Center (Calgary, Alberta, Canada; Bullas and Ryu 1983). All environmental isolates not resistant to S were tested as donor strains in individual conjugation assays. Isolates demonstrating multiple-drug resistance were processed and tested separately for the ability to transfer each type of drug resistance gene. For example, after the replica-plating assay, if a particular isolate grew on an A plate and also on a separate K plate, the isolate was subjected to two conjugation assays. In the first assay, the transconjugants would be selected on a dual antibiotic plate containing A and S, and in the second assay, the selective plate would include K and S.

An A- and K-resistant E. coli strain, SM10-TnphoA, was used as a positive control donor for testing the efficiency of conjugation. SM10-TnphoA was obtained originally from John Mekalanos (Taylor et al. 1989). Prior to conjugation, individual bacterial cultures were grown overnight at 37°C in TSB containing 30 μg/ml of the appropriate antibiotic (S for EM1000; A and K for SM10-TnphoA; and either A, K, C, or O for the environmental isolates). The bacterial cells were pelleted by centrifugation and resuspended in fresh TSB lacking antibiotic. Conjugation experiments were conducted as described by Bell et al. (1983). Briefly, equal volumes (0.1 ml) of stationary-phase cells from the donor strain (an individual environmental isolate or the SM10-TnphoA positive control donor) and recipient strain (EM1000) were mixed in a 1.5-ml tube with 0.8 ml of antibiotic-free TSB and incubated at 37°C for 2–4 h. After the incubation, mating mixtures were plated onto TSA containing dual antibiotics (S and either A, K, C, or O all at 30 μg/ml) and incubated for 24 h at 37°C to select for transconjugants. Transconjugants of interest were preserved in TSB supplemented with the appropriate antibiotic and 10% glycerol and frozen at −80°C.

Negative conjugation control experiments lacking either the donor or recipient bacteria were included in all conjugation experiments to check for mutation of either strain to antibiotic resistance. No mutation to specific antibiotic resistance was seen with any donor strains or the universal S. typhimurium EM1000 recipient strain.

2.5 PCR Screening for Tetracycline Resistance Genes

Environmental isolates positive for tetracycline resistance and their respective transconjugants were screened for specific tetracycline resistance genes using the polymerase chain reaction (PCR). The tetracycline resistance genes tested in this study included tet (B), tet (C), tet (E), tet (H), tet (Y), and tet (Z) based on primer sets from Aminov et al. (2002); tet (M) and tet (W) from Aminov et al. (2001); and tet (K) and tet (L) from Gevers et al. (2003). The presence of class I and II integrons was screened using primer sets from France et al. (2005). A typical PCR reaction was performed with a 50-μl mixture of the following reagents: 1-μl DNA template (50 ng) or bacterial cell suspension (one colony suspended in 10 μl H2O), 0.2-μM primers, and 45-μl Platinum PCR Supermix (Invitrogen Corporation, Carlsbad, CA) supplemented with MgCl2 to a final concentration of 2.0 mM. PCR amplification was performed as follows: 95°C for 5 min (one cycle), 94°C for 30 s, 50°C for 30 s, 72°C for 30 s (25 cycles), and 72°C for 7 min (one cycle) using an MJ Research PTC-200 thermocycler (Bio-Rad Laboratories, Hercules, CA). If amplification products were not detected in the first assay, the samples were subjected to another round of amplification using the same cycling protocol with the exception of reducing the annealing temperature to 45–48°C. PCR amplification products were analyzed by gel electrophoresis on a 2.5% (wt/vol) agarose gel (NuSieve; FMC Bioproducts, Rockland, MD) and stained with ethidium bromide. The Promega 1 KB Plus Ladder (Promega Co., Madison, WI) was used for verifying DNA fragment sizes. Amplification products were purified by the use of a Qiagen QIAquick PCR purification kit (Qiagen Inc., Valencia, CA) and sequenced at the University of Michigan DNA Sequencing Core (Ann Arbor, MI). The DNA sequences generated from environmental isolates were compared to known sequences using the Basic Local Alignment Search Tool family of programs from NCBI (http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=pubmed&list_uids=15215342&dopt=Citation; McGinnis and Madden 2004).

Plasmid DNA was purified from environmental isolates and their transconjugants using the Qiagen QIAprep plasmid purification system. The following modifications in the procedure were performed to optimize purification of low-copy plasmids and cosmids: (1) volumes of buffers P1, P2, and N3 were doubled, (2) the optional PB wash step was included, and (3) H2O heated to 70°C was used to elute DNA from the QIAprep membrane.

2.6 Identification of Environmental Bacteria

Each bacterial isolate that screened positive for a particular tetracycline resistance gene was identified using the RapiD 20 E system (bioMerieux, Inc., Durham, NC). Individual isolates were inoculated on media containing trypticase soy agar with 5% sheep blood BBL (Becton Dickinson, Franklin Lakes, NJ) and incubated overnight at 37°C. A cell suspension was prepared using one to four well-isolated colonies diluted in 2 ml 0.85% NaCl to a turbidity equivalent to 0.5 McFarland. The cell suspension was immediately used to inoculate a RapiD 20 E test strip and incubated for 4 h at 37°C. The test results were entered into the RapiD 20 E Analytical Profile Index, and the species identification was provided at the bioMerieux apiweb site (https://apiweb.biomerieux.com/servlet/Authenticate?action=prepareLogin).

2.7 Statistical Analysis

Differences between the AI and AUI sites in (1) the proportion of samples which exceeded established water quality thresholds for total phosphorus (0.1 mg/l, Lind 1985), turbidity (40.1 NTU, Mitchell and Stapp 1997), and fecal coliform density (400 CFU/100 ml, Michigan Department of Environmental Quality 1999), (2) the proportions of single- vs multi-drug resistant bacterial isolates, and (3) the proportions of bacterial isolates able to transfer their resistance during conjugation assays were tested using Fisher’s exact test (Zar 1999). Identical comparisons were examined for sites upstream and downstream of WWTPs. Spearman’s rank correlation was used to examine the relationships among turbidity, temperature, total phosphorus, pH, and fecal coliform density across all samples and sites. For correlation analyses, fecal coliform densities plated as too numerous to count (TNTC) were coded as an arbitrarily large number greater than 7,900 CFU/100 ml (the largest recorded value). The SYSTAT computer package (v 10.2, SYSTAT Inc., Chicago, Illinois) was used to perform all statistical analyses for this study.

3 Results

3.1 Physical, Chemical, and Biological Water Quality

Temperature, pH, and turbidity were similar among the agricultural streams and within the WWTP sites (Table 2). Turbidity tended to be increased by rain events, especially in the agricultural streams. AI sites were one to eight times more phosphorus rich than all other sites. Fecal coliform densities ranged from 70 to >7,900 CFU/100 ml. For some measurements, particularly those taken during rain events and at AI sites, quantitative fecal coliform densities could not be determined as a result of confluent colony growth regardless of the water dilution (Table 3). The fecal coliform densities at agricultural sites ranged from 700 to TNTC, while fecal coliform densities in waterways near WWTPs ranged from 70 to 2,300 CFU/100 ml. The highest fecal coliform densities were measured after rain events and occurred most frequently at the AI sites (Table 3). Turbidity, total phosphorus concentration, and fecal coliform density were highly correlated (turbidity–phosphorus n = 49, r s = 0.790, p < 0.001; turbidity–fecal coliform n = 39, r s = 0.688, p < 0.001; phosphorus–fecal coliform n = 39, r s = 0.684, p < 0.001), with AI sites tending to have the greatest values for these three parameters. Turbidity, total phosphorus concentration, and fecal coliform density were all negatively correlated with pH (turbidity–pH n = 49, r s = −0.371, p < 0.01; phosphorus–pH n = 49, r s = −0.606, p < 0.001; fecal coliform–pH n = 39, r s = −0.471, p < 0.01). AI streams were significantly more likely than AUI streams to exceed a total phosphorus threshold of 0.1 mg/l (p < 0.001) and marginally more likely to exceed a turbidity threshold of 40.1 NTU (p = 0.08, Table 4). There was no difference in the total phosphorus or turbidity exceedance frequencies between sites upstream and downstream of WWTPs (p always >0.05). Although the AI sites tended to have greater fecal coliform densities than the AUI sites, there was no difference in the frequency of fecal coliform densities >400 CFU/100 ml (p > 0.05), nor was there any difference observed in fecal coliform exceedance frequencies between sites upstream and downstream of WWTPs (p > 0.05, Table 4).

Table 2 Water chemistry ranges across sampling dates for agriculturally impaired and agriculturally unimpaired sites
Full size table
Table 3 Fecal coliform densities for each grouping of sites across sampling dates from filtered plates (CFU/100 ml)
Full size table
Table 4 Patterns of water quality and multi-drug resistance at study sites
Full size table

3.2 Antibiotic Resistance Patterns

A total of 830 fecal coliform isolates were collected from mFC agar plates and replica-plated on TSA supplemented with one of five antibiotics. Overall, 98.3% of the fecal coliform isolates selected were resistant to at least one antibiotic (Table 5). Fourteen of the isolates (1.7%) did not grow on any of the antibiotic plates, 640 isolates (77.1%) were resistant to only A, and the remaining 176 resistant isolates (21.2%) were resistant to two or more antibiotics. The only single-drug resistance pattern observed was to A, and all the multi-drug isolates included A resistance with the exception of one isolate that was KCO resistant (Table 5).

Table 5 Percentage of antibiotic-resistant fecal coliform isolates from individual study sites
Full size table

The AI sites had a significantly greater proportion of isolates that were resistant to multiple antibiotics (41.6% vs. 16.5%) and a lower proportion of isolates resistant to only ampicillin (58.4% vs. 83.5%) than the AUI sites (p < 0.001, Tables 4 and 5). The sites upstream and downstream of WWTPs had similar proportions of isolates resistant to only ampicillin (81.0% vs. 89.1%) and multi-drug resistant (19.0% vs. 10.9%) (p > 0.05, Tables 4 and 5).

3.3 Conjugal Transfer of Antibiotic Resistance Genes

Conjugation assays were conducted on 735 non-S-resistant isolates. Of the 640 single-drug resistant isolates (all were A-only resistant), 456 (71.3%) produced viable resistant transconjugants. For the multi-drug-resistant isolates (n = 95), 27.4% were able to transfer all resistance determinants (AK, AO, and ACO), and 81.1% were able to transfer at least one multi-drug resistance pattern (AK, AO, AC, ACO, and AKCO) with A being the most prevalent resistance determinant transferred (Table 6). The proportion of isolates from each site category able to transfer resistance in laboratory conjugation assays ranged from 60.0% to 85.6% (Table 4) and did not differ significantly between either the AI and AUI sites (p > 0.05) or the sites upstream and downstream of WWTPs (p > 0.05).

Table 6 Resistance pattern observed after gene transfer for non-streptomycin isolates
Full size table

A subset of AO, AC, and ACO multi-drug-resistant isolates (n = 22) and their transconjugants were further screened for the presence of specific tetracycline resistance genes and class I and II integrons. In all, 13 isolates were tet positive: six isolates were identified carrying the tet (B) gene, five with tet (C), and two with both tet (B) and tet (C). No class I or II integrons or tet (E, H, K, L, M, Y, W, or Z) genes were detected. The 13 tet-positive isolates were individually identified as E. coli by the RapiD 20 E system (Table 7). The identity of all tet (B) and tet (C) PCR products was confirmed by DNA sequence analysis.

Table 7 Identification of tetracycline (tet) resistance genes in the environmental isolates and their transconjugants
Full size table

To characterize the potential location and mode of horizontal transfer of tet (B) and tet (C) genes between the environmental isolates and the conjugative recipient, plasmid DNA was purified and PCR-amplified using primer sets for tet (B) and tet (C) (see Table 7). All eight isolates positive for the tet (B) gene showed the presence of tet (B) on plasmid DNA, as did their respective transconjugants, with the exception of two isolates (VH19-7R1 and VH25-16R1). For these two isolates, their transconjugants could not be subcultured and maintained on tetracycline-selective media (at 30 μg/ml C or O). Repeated conjugation experiments for isolates VH19-7R1 and VH25-16R1 failed to produce any viable colonies selected on tetracycline plates despite the evidence that the donor isolates contain plasmid DNA with the tet (B) gene. In this study, chromosomal DNA was not definitively screened in the absence of plasmid DNA.

For the tet (C) determinant, three different patterns of gene transfer were observed (Table 7): (1) In two out of eight isolates (SJ7-5-I and HC-1-R2), the tet (C) gene was identified with plasmid DNA in both the donor isolates and their transconjugants; (2) tet (C) was identified in total DNA from isolates VH19-8R3, VH19-11F, and MC2-9R1 and their transconjugants, but not associated with any plasmid DNA; and (3) for isolates MC1-22R4 and MC1-1F, the tet (C) gene was present in total DNA from the isolate but was unidentifiable in any plasmid DNA or total DNA from the transconjugant even though viable tetracycline-resistant colonies appeared after conjugation.

For the two isolates carrying both tet (B) and tet (C) genes (VH19-11F and MC2-9R1), both tet genes were identified in total DNA (in the isolates and transconjugants). The tet (B) gene was also amplified from plasmid DNA isolated from both isolates and their transconjugants, but this did not occur with the tet (C) gene (Table 7). Figure 1 shows gel PCR amplification products produced using tet (B) and tet (C) primer pairs with DNA from bacterial cells and isolated plasmid from isolate MC2-9R1 and its transconjugant, MC2-9R1T2-1.

Fig. 1
figure 1

PCR screening analysis of tet (B), left-side amplicons (206 bp), and tet (C), right-side amplicons (207 bp), from a representative tetracycline-resistant bacterial isolate MC2-9R1 and its transconjugant, MC2-9R1T2-1. Lanes 1 and 12, 1 KB Plus ladder (Promega); lanes 2 and 8, isolate MC2-9R1 cells; lanes 3 and 9, isolate MC2-9R1 plasmid DNA; lanes 4 and 10, transconjugant MC2-9R1T2-1 cells; lane 5, transconjugant MC2-9R1T2-1 plasmid DNA; lanes 6 and 11, Salmonella strain EM1000 cells (conjugal recipient, negative control); lane 7, pBR322 plasmid DNA (positive control for the tet (C) gene). The lower bands in lanes 2, 6, 9, and 11 are DNA primers

Full size image

4 Discussion

Chemical water quality at the WWTP sites was good; all WWTP samples had total phosphorus concentrations <0.1 mg/l, which is considered acceptable or unpolluted (Lind 1985; Mitchell and Stapp 1997), and turbidities of <40.1 NTU, which is considered excellent to good (Mitchell and Stapp 1997). Fecal coliform densities at the WWTP sites were marginal, as 56% of samples surpassed the 10% exceedance threshold (<400 CFU/100 ml). No statistically significant differences were detected in water chemistry measurements and fecal coliform densities between the upstream and downstream sites. These results suggest that little, if any, WWTP point source pollution occurred during the study period. MDNRE (Michigan Department of Environmental Quality 2004a) confirmed that no untreated sewage was discharged from either the Chelsea or the Ann Arbor WWTPs during or immediately preceding the study period.

In contrast, the AI sites located near CAFO farms had indicators of poor water quality compared to reference sites in the same area. Total phosphorus concentrations were much more likely (100% vs. 26.7%) to exceed established thresholds in the AI sites than AUI sites, and turbidity was also somewhat more likely (35.0% vs. 6.7%) to exceed threshold values. Although the frequency of fecal coliform densities exceeding 400 CFU/100 ml did not differ between AI and AUI sites, only the AI sites had fecal coliform densities TNTC, indicating that maximal densities of fecal coliforms occurred at AI sites.

As expected, water quality decreased following heavy precipitation. At all study sites, turbidity levels were elevated after rain events compared to the initial and final measurements taken during dry periods. Increased phosphorus levels were also detected after precipitation in the agriculturally impacted areas, and fecal coliform densities were much higher after precipitation. The strong correlation of turbidity, total phosphorus, and fecal coliform densities suggests a common source for these parameters. Elevated total phosphorus, turbidity, and fecal coliform densities are presumed to be the direct result of runoff from nearby tiled fields sprayed with liquid manure as reported by MDNRE in numerous previous waste discharge infractions by the CAFO farms in close proximity to our AI sites (Michigan Department of Environmental Quality 2003a, 2004b). Our data are consistent with recent biological surveys of the same waterways, which classified the AI sites as poor habitat, supporting only macroinvertebrate and fish communities that could survive in waters with low dissolved oxygen for sustained periods, whereas several of the AUI sites were listed as “reference” or “least impacted” systems in this region (Michigan Department of Environmental Quality 2003a).

Elevated nutrient levels can also improve the efficiency of horizontal gene transfer between species by the bacterial processes of transformation (Davison 1999; Quintiliani et al. 1999; Thomas and Nielsen 2005) and conjugation (Arana et al. 1997, 2001). High fecal coliform densities, as seen in this study, increase the likelihood of humans and animals coming in contact with pathogenic bacteria. Implementing a quantitative risk assessment analysis, as modeled by Hurd (2006), could help identify the “chain of causal events that connect on-farm antibiotic use to additional days of human illness caused by infections with resistant bacteria.”

Antibiotic-resistant fecal coliform bacteria were found in all water samples collected during the course of this study, with A resistance the most prevalent (77.1% overall, Table 4). This proportion of A resistance in fecal coliform bacteria parallels other river and waterway reports (Ash et al. 2002; McKeon et al. 1995; Ogan and Nwiika 1993). Some studies have reported lower frequencies of A resistance yet still found that resistance to A was the most common form of resistance (Iwane et al. 2001; Niemi et al. 1983). In contrast, other studies of soil, water, and manure samples from farm locations have found tetracycline and other drug resistance patterns in higher proportions than A resistance (Esiobu et al. 2002; Sayah et al. 2005). One plausible explanation for the differences seen in antibiotic resistance patterns is the source of the environmental sample, specifically human waste vs. animal waste. Harwood et al. (2000) determined the antibiotic resistance pattern from fecal coliforms isolated from domestic wastewater and various animal feces using four different concentrations of antibiotics; they showed a significantly greater percentage of A resistance in fecal coliform isolates from human sources (62% at 10 μg/ml A) than animal feces (15%) and a significantly greater percentage of C resistance in cattle feces (58% at 20 μg/ml C) compared to the human sources (35%). This could explain why we observed a significantly lower proportion of A resistance (58.4%) at the AI sites (sites near animal waste contamination) than at the AUI sites (83.5%) and sites upstream (81.0%) and downstream (89.1%) of the WWTPs.

The proportion of multi-drug resistance observed at the AI sites near CAFO farms (41.6%) was almost three times greater than at the AUI sites (16.5%). Certain multiple-resistance combinations were more common at some AI sites than others, perhaps reflecting site-specific antibiotic use patterns. In any case, the high proportion of multi-drug resistance at AI sites suggests that the fecal bacterial populations in these locations were subjected to conditions that fostered the acquisition of multiple-resistance determinants. In addition to likely increases in antibiotic-resistant bacteria from animals fed antibiotic-laden feed (Davies 1997; Lu et al. 2004; Wegener et al. 1999), soil-dwelling bacteria are thought to be a significant reservoir of resistance determinants (D’Costa et al. 2006), and studies on crop soils fertilized with animal manure show that horizontal transfer between fecal and soil bacteria is facilitated by the high nutrient availability of manure (Cooke 1976; Schmitt et al. 2006). We observed that fecal coliform levels increased after rain events at the AI sites, where periodic spraying of liquid manure was reported in nearby fields (Kauffman and Melmoth 2003; Michigan Department of Environmental Quality 2003a). Natural water environments may allow fecal coliforms a better selective advantage in becoming antibiotic resistant over other habitats such as soil, sand, or sewage effluent (Cooke 1976). Survival rates tend to be higher in bacteria containing plasmids (Arana et al. 1997, 2001; Ash et al. 2002; Schwartz et al. 2003), and the conjugative transfer of plasmids from one bacterium to another tends to occur more readily in aquatic habitats (Lebaron et al. 1993). The high rate of success for gene transfer as seen in our conjugation assays (83.3% overall) suggests that the environmental isolates carried conjugative plasmids or transposons. Of the multi-drug-resistant isolates, most exhibited resistance to combinations of antimicrobial drugs which included A. This is an indication that multiple-resistance genes may coexist on one plasmid (Davison 1999; Quintiliani et al. 1999; Sayah et al. 2005), a single conjugative transposon (Pembroke et al. 2002; Waters 1999), or an integron (Mazel 2004; White et al. 2001). This condition is particularly disconcerting given that exposure to one antibiotic agent may result in resistance to others without previous exposure (Sayah et al. 2005) or cost to bacterial fitness (Aminov et al. 2001).

The conjugation assay in combination with PCR was used to identify genetic patterns of transfer in bacterial populations. Using 10 primer pairs for the tetracycline resistance genes most commonly found in E. coli and other Gram-negative bacteria, we detected the presence of the tet (B) and tet (C) determinants in 13 of 22 ACO-resistant isolates. In strains isolated at all study sites, tet (B) was associated with plasmid DNA in isolates and their transconjugants, while tet (C) showed three different patterns of gene transfer. The first pattern of gene transfer showed that the tet (C) gene was associated with plasmid DNA in isolates and transconjugants from the AUI sites. The second pattern showed that the tet (C) gene from the UPWWTP isolates associated exclusively with total DNA in both the original isolates and the transconjugants and not with plasmid DNA; this suggests that the tet (C) gene was located on chromosomal DNA or on a large plasmid or cosmid that could not be purified by our methods. The final pattern of gene transfer found the tet (C) gene in total DNA in the isolate, but the tet (C) gene was undetectable in the resistant transconjugant. These results suggest that the transconjugant in the third pattern could have received a different tetracycline resistance gene from the donor isolate that was not detected by our PCR screening method (e.g., isolates MCI-22R4 and MCI-1F; Table 7). Two of 13 isolates (Table 7; isolates VH19-11F and MC2-9R1) showed the presence of both tet (B) and tet (C) genes, with tet (B), but not tet (C), associated with plasmid DNA, suggesting that the tet (B) and tet (C) genes are not genetically linked. Furthermore, both the differential distribution of the tet (B) and tet (C) genes in the VH19-11F and MC2-9R1 isolates (Table 7) between total and plasmid DNA suggested that the plasmid DNA preparations were not contaminated with chromosomal DNA. Additional studies with a larger population of tetracycline-resistant isolates would be needed to better characterize these genetic patterns.

Based on current chemical and biological water quality standards (turbidity, total phosphorus concentrations, and fecal coliform densities), study sites near the WWTPs were considered environmentally healthy, yet had high levels of single- and multi-drug-resistant fecal coliform bacteria (>81% for A alone and >10.9% for multi-drug resistance, Tables 4 and 5). Agricultural sites, especially the AI sites near CAFOs, had much lower measures of traditional water quality and also had high levels of single and multi-drug-resistant bacteria, with multi-drug resistance greatest (41.6%) at the AI sites near CAFOs. The risk to human and animal health posed by the high incidence of antibiotic resistance and gene transfer is unknown. Traditional measures of chemical and biological water quality do not appear to be direct surrogates for detection of the prevalence of antibiotic resistance. Those parameters most elevated in the AI sites (e.g., total phosphorus concentrations, turbidity), however, may have some predictive ability for the prevalence of multiple-drug resistance. We echo previous suggestions (Esiobu et al. 2002; Sayah et al. 2005) that testing for antibiotic resistance genes in bacterial strains become part of the standard methods for examining and regulating water quality and wastewater discharge in areas at high risk for pollution from human and animal waste.