Introduction

Over the past several decades, there has been an increased emergence of antibiotic-resistant bacteria and associated genes within surface waters, apparently due to increases in the usage of antibiotics for human treatments and for animal production (Lebkowska 2009). The presence of antibiotic-resistant bacteria in many types of environmental waters, including coastal waters, creeks, rivers, lakes, wells, agricultural runoff, and groundwater, has been well-documented (Baquero et al. 2008; Czekalski et al. 2012; Junco et al. 2001; Li et al. 2013; Rice et al. 1995). Irrigation waters have been suggested as a potential pathway for human exposure to wildlife strains of antibiotic-resistant bacteria (Allen et al. 2010). Surface waters and bottom sediments harbor a diverse community of human-, animal-, and environmental-indigenous bacteria (Pachepsky et al. 2011), some of which may have an inherent resistance to certain antimicrobial agents (Baquero et al. 2008). Surface waters used for irrigation have been implicated in the transmission of water-borne pathogens to fresh produce, which has resulted in the increased regulation of their microbiological quality (Gerba 2009; Pachepsky et al. 2011). Various types of water distribution systems (i.e., pipelines dispensing drinking water, municipal wastewater, hospital discharge water, and treated wastewater) have been reported to disseminate antibiotic-resistant bacteria (Armstrong et al. 1981; Czekalski et al. 2012; Gomez-Alvarez et al. 2012; Junco et al. 2001; Lebkowska 2009; Yomoda et al. 2003), but the role of irrigation systems in doing so as well has not yet been explored.

Bacterial antibiotic resistance genes, which may be encoded in the chromosome or in extrachromosomal plasmids, elicit biochemical defense mechanisms that allow for tolerance of the corresponding antibiotic compound, and those encoded in the latter can be transferred among or acquired by suitable bacteria via conjugation, transformation, or transduction. Yomoda et al. (2003) reported proliferation of antibiotic resistance genes in several wards throughout a hospital in Japan and suggested that this was caused by gene transfer via conjugation between Pseudomonas strains in the hospital water pipelines. Junco et al. (2001) suggested that antibiotic resistance genes in a municipal wastewater system in Spain contributed to the presence of antibiotic-resistant strains of fecal enterococci in proximate coastal water used for human recreation. In Switzerland, multi-drug-resistant strains of heterotrophic bacteria were enumerated from water and sediment near a drinking water pipeline intake in Lake Geneva and were considered to be related to nearby incoming discharge from a wastewater treatment plant that had contained bacteria with similar antibiotic resistances in its pipelines (Czekalski et al. 2012). According to Barkay et al. (1995), conjugation is common in nature when the density of metabolically active bacterial cells is high and two cells suitable for the exchange are situated near each other, such as within biofilms. Hausner and Wuertz (1999) performed an in situ study and described biofilm structure to impact the rate of plasmid transfer from donor Escherichia coli to recipient Alcaligenes eutrophus. Biofilms are considered a preferable habitat for many microorganisms (De Lancy Pulcini 2001; Stanley and Lazazzera 2004; Wimpenny 1996), and those that form within wastewater, drinking water, and surface water systems have been shown to serve as transient, as well as long-term habitats for antibiotic-resistant bacteria (Gomez-Alvarez et al. 2012; Schwartz et al. 2003). Biofilms that form within irrigation systems have been suggested to incorporate pathogens and nonpathogenic bacteria (Armstrong et al. 1981; Pachepsky et al. 2012; Sadovski et al. 1978; Shelton et al. 2012, 2013); however, they have not yet been studied as a reservoir for antibiotic-resistant bacteria.

The objectives of this work were to determine: (a) if antibiotic-resistant bacteria grow within a routinely used irrigation system, in the residual pipe water that remains in the system between irrigation events, and in the biofilms that develop on the substrata and (b) if the percentages of antibiotic-resistant bacteria within the general population of a bacterial group living in the biofilm (i.e., fecal coliforms, total coliforms, and total heterotrophs) differ among bacterial groups and/or change over time.

Materials and methods

Experimental procedure

Experiments were conducted at the USDA-ARS BARC South Farm Research Site (Beltsville, MD, USA). The design of the irrigation system assembled at the site is shown in Fig. 1. The two parallel main lines were constructed with new PVC pipes that had an internal diameter of 2.5 in. (6.35 cm). Pipeline coupons (6 in.; 15.34 cm) were cut from the main lines, and four-coupon sets were inserted into the middle of the pipe in front of the first and third sprinkler on each line. Coupons in each set were connected to each other and to the pipe with stock rubber couplings and stainless steel clamps (1001-44, Fernco Inc., Davison, MI) that covered the coupon and pipe contact areas. Water from the nearby perennial Paint Branch Creek was used for all irrigation events. A sprinkler pump (NSPHE Series, Monarch Industries, Winnipeg, Canada) in the edge-of-field pump house was set to draw water from the creek and direct flow to a PVC pipe splitter that connected to the two irrigation lines. The irrigation system was placed on fairly level ground with a moderate grass cover (estimated slope of 0–2 %).

Fig. 1
figure 1

Schematic of the irrigation system. Pipelines 1 (bottom) and 2 (top) have the same dimensions from the ends of the PVC splitter

Full size image

The initial irrigation event occurred on September 9, 2013 (day 0). Subsequent events took place on September 16, 2013 (day 7), September 23, 2013 (day 14), and September 30, 2013 (day 21). For each event, the pump was turned on and approximately 100 L (i.e., >1 irrigation system volume) of creek water was “flushed” through the system with the end valves open to eliminate any debris and discard remaining residual water. After closing the end valves, samples of creek water at the intake hose were taken in triplicate with a 70 % ethanol disinfected 1-L grab sampler and transferred to sterile Whirl-Pak® (B01065, Nasco, Fort Atkinson, WI) bags at that time (hour 0) and 2 h later, just before turning off the irrigation pump; six creek water samples were collected during each event. Water remained in the closed system between irrigation events. Prior to running the system on each of the latter three events, clamps around one of the coupons at locations 1a, 1c, 2a, and 2c (Fig. 1) were loosened and a portion of the residual water remaining from the previous irrigation event was collected in a sterile Whirl-Pak® bag; one coupon was extracted from each location and wrapped with plastic wrap, pipes were moved to close the gaps from the removed coupons, and clamps were tightened to reseal the pipeline. The creek water samples, residual water samples, and coupons were stored and transported to the lab on ice. All irrigation events took place within the time window of 8:00 a.m. and 10:30 a.m.

Discharge rates for sprinklers 1a, 1b, and 1c were 12.2 ± 0.4, 8.0 ± 0.3, and 8.0 ± 0.2 L min−1, respectively. Those for sprinklers 2a, 2b, and 2c were 8.4 ± 0.3, 12.2 ± 0.1, and 7.7 ± 0.5 L min−1, respectively. Temperatures inside the pipe were monitored every 5 min with a HOBO U20 stage logger (Onset Computer Corporation, Bourne, MA) that was housed within the PVC splitter, and air temperature was measured at a weather station in the pump house.

Microbiological analyses

Creek water and residual water samples were analyzed for fecal coliforms, total coliforms, and total heterotrophic bacteria (THB). Fecal coliforms were enumerated on MacConkey agar (R453802, Remel Inc., Lenexa, KS) spread plates that were incubated at 44.5 °C for 24 h; total coliforms were enumerated on CHROMagar™ ECC (EF322, DRG International Inc., Mountainside, NJ) spread plates that were incubated at 37 °C for 24 h; THB were enumerated on L agar (R452322, Remel Inc., Lenexa, KS) spread plates that were incubated at 37 °C for 24 h. Ampicillin (Ap)- and tetracycline (Tc)-resistant fecal coliforms, total coliforms, and THB were enumerated via spread plating on their respective agars that had been supplemented with standard breakpoint concentrations of Ap or Tc (i.e., 32 and 16 mg mL−1, respectively [USDA-ARS 2014]) during agar preparation. All bacterial concentrations in creek water samples were expressed as colony-forming units (CFU) per milliliter.

Surface-associated cells and potential biofilms were obtained from the coupons by using methods described in Shelton et al. (2013). The biofilm samples were analyzed for concentrations of fecal coliforms, total coliforms, and THB, as well as for these bacteria that were resistant to Ap and Tc by using the agars described above, and the concentrations were expressed as CFU per square centimeters of pipe interior. Biofilm proteins were measured with the Micro BCA™ Protein Assay (23235, Thermo Fisher Scientific Inc.) and expressed as microgram protein per square centimeters.

Isolates of THB, total coliforms, and fecal coliforms from the biofilm samples were analyzed for resistance to a panel of antibiotics that included Ap, Tc, cefoxitin (Fx), chloramphenicol (Cm), ciprofloxacin (Cp), kanymycin (Km), and streptomycin (Sm). Each week, 100 THB isolates, 50 total coliform isolates, and 50 fecal coliform isolates were randomly selected from their respective agar plates for the panel test. The isolates were replicated onto L agar plates and incubated at 37 °C for 24 h. The biomass from these L agar plates was further replicated onto a panel of eight Mueller Hinton agar plates (225250, Becton Dickinson Microbiology Systems, Sparks, MD), including seven plates supplemented with standard breakpoint concentrations of antibiotics Ap, Fx, Cm, Cp, Km, Sm, or Tc (32, 30, 32, 2, 64, 64, and 16 mg mL−1, respectively [USDA-ARS 2014]) and one control plate, and then were incubated at 37 °C for 24 h. Growth on Mueller Hinton-supplemented agars indicated resistance to the antibiotic amendment. The total percentages of fecal coliform, total coliform, and THB isolates that were resistant to Ap, Fx, Cm, Cp, Km, Sm, or Tc were determined by dividing the number of positives for antibiotic resistance (in cases where the control was also positive, which was >95 % of the time) by the number of total control positives. Ap and Fx are both β-lactams (β-L) and Sm and Km are both aminoglycosides (AG), with each of the latter being a newer generation in the respective class. The mean percentages of biofilm-associated fecal coliforms, total coliforms, and THB that were resistant to Ap and Fx and those which were resistant to Sm and Km were calculated to represent general resistance to the β-L class and the AG class, respectively. Grouping resistance for antibiotic classes was important for multiple-drug resistance assessments, since resistance to two compounds in the same class is generally conferred by the same gene/mechanism. In total, resistance to five antibiotic classes (i.e., Cm, Cp, Tc, β-Ls, and AGs) was determined. The percentages of isolates of each bacterial group that were resistant to 0, 1, 2, 3, 4, or all 5 of the antibiotic classes on the panel were organized into histograms to show the relative amounts of multiple-drug resistance.

Statistical analyses

Most of the statistical distributions of measured values were not normal in this work. Therefore, we used the nonparametric statistical tests. The statistical software PAST (Hammer et al. 2001) was used to: (a) apply the Kruskal-Wallis test to estimate the probabilities of sameness in median concentrations in the creek water from week to week and (b) apply the Kolmogrov-Smirnoff test to determine the probabilities of sameness in the histograms of multiple-drug resistance among bacterial groups on each irrigation event date and for each bacterial group on consecutive irrigation event dates.

Results

Temperature

The average daily air temperature was 17.8 ± 4.2 °C with the daily min and max of 11.8 ± 5.0 °C and 24.5 ± 4.9 °C, respectively. Time series of the air temperatures and the temperatures measured inside the irrigation pipe are shown in Fig. 2. The trend of slight cooling over time can be observed and the pipe min temperatures were often cooler than that of the air (Fig. 2).

Fig. 2
figure 2

Air temperatures and temperatures inside the irrigation pipes throughout the experiment

Full size image

Bacterial growth in residual water and in biofilm

The concentrations of each bacterial group that passed into the irrigation system were often specific to the sampling date (Table 1). There were significant differences in median values of the weekly intake water concentrations of THB (p = 0.005), total coliforms (p = 0.003), fecal coliforms (p = 0.002), Ap-resistant THB (p = 0.002), and Tc-resistant total coliforms (p = 0.033 [representative of Tc-resistant total coliform concentrations on day 7 and day 14; other days were BDL (Table 1)]). There were no significant differences in the intake water concentrations of Tc-resistant THB from week to week (p = 0.343) or Ap-resistant total coliforms (p = 0.173) from week to week. The remaining bacteria that were monitored at the creek water intake (i.e., Ap- and Tc- resistant fecal coliforms) could not be statistically analyzed for temporal differences due to more than half of the collected samples being below the detection limit (i.e., <4 CFU mL−1).

Table 1 Concentrations of total heterotrophic bacteria (THB), total coliforms, fecal coliforms, and the Ap- and Tc-resistant bacteria in each bacterial group in water in pipes
Full size table

Bacteria grew substantially in the residual water that remained in the pipes during each 1-week sitting time between irrigation events. The concentrations of THB, total coliforms, fecal coliforms, and Ap- and Tc-resistant bacteria in each bacterial group that were measured in the residual water were often equal to or greater than their concentrations in the creek intake water from 1 week earlier (Table 1). The concentrations of Ap-resistant bacteria in the intake water and in the residual water were always greater than those of the Tc-resistant bacteria (Table 1).

Protein contents of the material that was scraped from the irrigation pipeline coupons indicated substantial and steady growth of biofilm throughout the experiment, and accordingly, bacterial concentrations in the biofilm displayed a trend of increasing over time (Fig. 3).

Fig. 3
figure 3

Biofilm growth throughout experiment. Protein content in the four pipeline locations (ref. Fig. 1) (a). Average ± standard error of concentrations of THB (b), total coliforms (c), and fecal coliforms (d) in the biofilm (filled symbols represent all bacteria in the bacterial group, and open symbols represent only Ap-resistant bacteria in the bacterial group)

Full size image

The concentrations of biofilm-associated Ap-resistant THB, total coliforms, and fecal coliforms were, in most cases, proportional to the general concentrations of each respective bacterial group (Fig. 3, right); that is, when the concentration of a particular bacterial group increased, the Ap-resistant concentration of the same bacterial group appeared to increase at a similar proportion. Tc-resistant THB, total coliforms, and fecal coliforms were also detected in several biofilm samples; however, most samples were below the detection limit (i.e., <0.4 CFU cm−2). In the biofilm, concentrations of Ap-resistant bacteria exceeded those of Tc-resistant bacteria.

Antibiotic resistance panel and multiple-drug-resistant bacteria in biofilm

The percentages of THB, total coliform, and fecal coliform isolates that were resistant to each of the antibiotics on the panel varied throughout the study. In general, all biofilm-associated bacteria demonstrated relatively high levels of resistance to the β-Ls, while resistance to Cm, Tc, and the AGs appeared to be more limited, and resistance to Cp was very low or nonexistent (Fig. 4). Compared with total coliforms and THB, fecal coliforms demonstrated higher levels of resistance to Cm throughout the study. It is noticeable that for any bacterial group, when the percentage of isolates that were resistant to a specific antibiotic/antibiotic class was initially relatively high on the first sampling date (i.e., >50 % on day 7), the relatively high levels of resistance persisted on subsequent irrigation events, and when the percentage of resistant isolates were initially relatively low to begin (i.e., <10 % on day 7), the percentages often tended to increase from week to week (Fig. 4).

Fig. 4
figure 4

Ratio of antibiotic-resistant isolates from the biofilm to total isolates from the biofilm on a log scale throughout the study

Full size image

Multiple-drug resistance was observed in many of the biofilm isolates. In fact, greater than 65 % of the biofilm-associated fecal coliforms appeared to be resistant to two or more antibiotic classes (Fig. 5). The distributions of multiple-drug-resistant THB, total coliforms, and fecal coliforms (i.e., the histograms in Fig. 5) were compared on each irrigation event date. The distributions of multiple-drug-resistant THB and total coliforms were not significantly different on any event date (p > 0.05), but those for THB and fecal coliforms were significantly different on day 7 and day 21, and that for total coliforms and that for fecal coliforms were significantly different on all irrigation event dates (Supplementary Information, Table S-1). Different groups of indicator bacteria experienced different numbers of multiple-antibiotic resistances.

Fig. 5
figure 5

Histograms showing the relative amounts of antibiotic classes that the biofilm-associated isolates were resistant against, out of the five classes that were tested on the panel

Full size image

The multiple-drug resistance distributions of each bacterial group were compared on consecutive event dates to see if the patterns changed over time. The distributions of multiple-drug-resistant THB did not significantly differ on day 7 and day 14 (p = 0.142) or on day 14 and day 21 (p = 0.243). However, there were substantial differences in the distributions of multiple-drug-resistant total coliforms on day 7 and day 14 (p < 0.001) and on day 14 and day 21 (p < 0.059). There were also substantial differences in the distributions of multiple-drug-resistant fecal coliforms on day 7 and day 14 (p = 0.007) and on day 14 and day 21 (p < 0.080). While the multiple-drug resistance trends of the broad group of biofilm-associated THB were not different during consecutive irrigation events, there were high probabilities of difference in the multiple-drug resistance of the more specific groups, total coliforms and fecal coliforms, during consecutive events.

Discussion

The excessive usage of antibiotics in agriculture and aquaculture and for treatment of human infection has presumably lead to inputs that have caused the emergence of natural environments as reservoirs for antibiotic-resistant bacteria and antibiotic resistance genes (Cabello 2006; Martinez 2008). Surface water resources harbor antibiotic-resistant bacteria and antibiotic resistance genes, and bottom sediments are a hot spot for horizontal gene transfer and allow antibiotic resistance to spread within the bacterial community, especially under stress conditions (e.g., in the presence of heavy metals) (Devarajan et al. 2015). Antibiotic resistance genes have been found in human pathogens in areas without a history of antibiotic contamination (Pallecchi et al. 2008), suggesting that they may even be well-maintained and proliferate in natural ecosystems (Martinez 2009). The environmental dissemination of antibiotics, antibiotic-resistant bacteria, and antibiotic resistance genes becomes a serious issue because (1) antibiotics within the environment challenge microbial populations and can induce restructuring of native communities, having adverse effects on microbial functional diversity and ecosystem nutrient cycling (Martinez 2009), (2) antibiotic resistance genes can eventually spread to human and animal populations via environmental contact or entering the food supply (Smith et al. 2005), and (3) human pathogens can potentially acquire antibiotic resistance genes through horizontal gene transfer (Pallecchi et al. 2008). While antibiotics are considered as the most successful family of drugs so far developed for the benefit of human health, thwarting the spread of bacterial-related illnesses may be challenged in the future if antibiotic prescription options become limited due to the widespread presence of genes that render them ineffective (Martinez 2009). The persistence of antibiotic-resistant bacteria and antibiotic resistance genes in aquatic environments is important with regard to potential human and environmental risks.

The surface water that was used as the irrigation source was part of a suburban watershed. The creek ran through a wooded riparian strip dividing two agricultural fields at the USDA-Beltsville Agricultural Research Center. A bike trail, baseball fields, and a camping site were several kilometers upstream from the irrigation site, and the area further upstream was forested, with suburban housing at the outskirts. The antibiotic-resistant bacteria that were enumerated from the creek water and the irrigation system may have originated from agricultural, anthropogenic, or wildlife (e.g., deer) inputs. Genomic elements associated with resistance may have entered the water body via these sources as well. It is worth noting that the fecal coliforms enumerated in this work were not often E. coli; we attempted to enumerate E. coli in the water and biofilm samples, but the majority of samples were below the detection limit. The fecal coliforms enumerated could have been another coliform with inherent resistance to certain antibiotics, e.g., Pseudomonas being resistant to β-Ls, hence the relatively high levels of bacterial resistance to Ap. Regardless, having measured fecal coliform concentrations in the creek water at relatively high concentrations (Table 1), it suggests a substantial level of pollution and nutrient availability for bacteria in this irrigation source.

We observed significant differences in the concentrations of some of the measured bacteria, including antibiotic-resistant bacteria, in the creek intake water from week to week. This may be attributable to temporal changes in environmental factors that affect the concentrations of bacteria in the aquatic environment. For example, temperature, which impacts growth and survival of E. coli as well as other indicator microorganisms and pathogens in surface waters (Blaustein et al. 2013; Flint 1987; Pachepsky et al. 2014), changed at a slow and steady rate throughout the month-long study (Fig. 2). Temperature may also impact concentrations of bacterial predators that are ubiquitous in freshwater environments, such as protozoans and zooplankton (Barker and Brown 1994; Williamson et al. 2002). Even slight shifts in dominant predator groups, which could have occurred over the course of our study in response to temperature changes, have been reported to strongly influence coliform population survival rates in aquatic environments (Brettar and Hofle 1992; McCambridge and McMeekin 1980). Furthermore, streambed resuspension could have had a major influence on temporal variations in bacterial concentrations in the creek water. Freshwater sediments have been suggested as an important reservoir for antibiotic resistance genes and as hot spot for horizontal gene transfer (Devarajan et al. 2015). Bacterial resuspension, which may affect bacterial survival in waters (Harmel et al. 2010), can occur during high flow conditions in the event of rain. That occurred, there were three rainfall events that occurred at the USDA-ARS South Farm Research Site during the timeframe of the experiment. The rainfall depths that were measured at the nearby weather station were 15.5, 0.5, and 28.7 mm on September 12, 2013, September 16, 2013, and September 22, 2013, respectively. The heaviest rainfall event occurred 1 day prior to the third irrigation event and a very light misting of rainfall occurred on the day of the second irrigation event (began during the irrigation event). Thus, rainfall may have played a role in the temporal variation of bacterial concentrations. Impacts of environmental factors on aquatic communities may at least partially explain why some of the concentrations of (antibiotic-resistant) bacteria were different from week to week. Other physicochemical parameters, such as organic matter content, metal pollutant concentrations, sediment depth, and grain size, have been reported to be significantly correlated with the persistence of bacterial load, genetic markers of indicator bacteria, and antibiotic resistance genes in the aquatic environment (Devarajan et al. 2015; Thevenon et al. 2012).

The internal environment of the irrigation pipeline appeared to be conducive to bacterial survival and growth. Protein steadily accumulated on the pipe walls as the study progressed, and all tested bacteria grew, including antibiotic-resistant bacteria. Concentrations of bacteria within the residual water were generally greater than those in the intake water (Table 1). The anomaly to this statement, which is for total coliforms and fecal coliforms on day 21, can be partially explained by the daily min temperature within the irrigation pipes for that date and even a few days prior being at or below 5 °C (Fig. 2). Coliform bacteria metabolism is drastically slowed and their affinity for substrate becomes lower in cold, low-temperature environments (Blaustein et al. 2013). If temperature drops to about 5 °C or below, E. coli can be driven into a viable but not culturable (VBNC) state, which allows continuous survival without the ability to divide (Garcia-Armizen and Servais 2004; Na et al. 2006). It is possible that some of the coliform bacteria in the populations in residual water and/or in biofilm entered a VBNC state around the time of the last sampling event and were not available for enumeration on the growth media used.

Our previous work at the same experimental site also indicated bacterial growth in irrigation pipe water and in the biofilms on the substrata (Pachepsky et al. 2012; Shelton et al. 2012, 2013). High probabilities of differences in bacterial concentrations were observed at the system intake and outputs, and biofilms were identified as a source and sink for microbiota passing through the system (Pachepsky et al. 2012; Shelton et al. 2012; Shelton et al. 2013). The potential for biofilms in irrigation pipelines to impact the microbial quality of delivered water was originally suggested by Sadovski et al. (1978). The effluent from drip irrigation lines was monitored for traceable microorganisms that were added to treated wastewater, and small peaks were observed in bacterial concentrations on days 2 and 8 and in viral counts on day 8. They hypothesized that the elution of microorganisms from the irrigation system was affected by adsorption of viruses and bacteria to the organic matter which lined the inside walls of the irrigation system; repeated irrigations may have resulted in the release of the organisms. Conceptually, release and retention of antibiotic-resistant bacteria from biofilms may also occur, and the concentrations of antibiotic-resistant bacteria that exit the system may be quite different from the concentrations at the intake.

In this study, we observed variable amounts of resistance of the biofilm-associated bacteria to different antibiotics, which has previously been described in water distribution systems and environmental water sources (Armstrong et al. 1981; Czekalski et al. 2012; Esioubu et al. 2002; Li et al. 2013; Schwartz et al. 2003; Wellington et al. 2013). The percentages of bacterial resistance to a given antibiotic class were subject to change over time as the biofilm in the pipe developed. With the assumption that major changes in environmental selective pressure within the irrigation pipes were unlikely to have occurred throughout the month-long study, this finding may have simply been a factor of stochastic bacterial incorporation into the biofilm on a weekly basis. Any temporal changes in measured concentrations of antibiotic-resistant bacteria in the biofilm may be attributable to the concentrations being different at the intake water on a weekly basis. While horizontal gene transfer is an important factor in proliferation of antibiotic-resistant bacteria in the natural environment, since we did not enumerate genomic material in this study, it cannot be concluded that this mechanism was a cause for the observed changes in antibiotic resistance in the biofilm. Rather, the flux was probably a result of shifts in the microbial community where opportunistic bacteria from the aquatic environment that happened to have an acquired form of resistance, whether that be a resistance gene encoded on a plasmid, a resistance gene that resulted from mutation(s), or innate resistance based on phenotype, were able to grow.

In agreement with previous studies, antibiotic resistance within the biofilm varied among bacterial groups (Esioubu et al. 2002; Thevenon et al. 2012; Devarajan et al. 2015). Compared to total coliforms and fecal coliforms, THB displayed a slightly higher resistance to Cp (Fig. 4). Since all fecal coliforms are total coliforms and all total coliforms are THB, it appears that there were other Cp-resistant THB in the biofilm. Alternatively, compared to THB and total coliforms, fecal coliforms appeared to have higher frequencies of resistance to Cm throughout the study. The biofilm-associated total coliforms and THB that were not fecal coliforms, which could have occurred in higher concentrations than the fecal coliforms, must have had relatively low frequencies of resistance to Cm. The significant differences between the distributions of multiple-drug-resistant isolates of fecal coliforms and isolates of THB and the significant differences between the distributions of multiple-drug-resistant isolates of fecal coliforms and isolates of total coliforms (ref. Supplementary Information, Table S-1) were probably related to the relatively higher frequency of Cm-resistant fecal coliforms. Given the number or isolates of each group that were tested, there is a high likelihood that the significant differences were not random. The observed differences in multiple-antibiotic resistance of different bacterial groups from week to week may be partially attributable to fluctuations of the general populations of these bacteria within the biofilm community. Resistance trends are likely to correlate with the abundance and type of bacterial strains that are present in the habitat (Esioubu et al. 2002).

This is the first work to show that antibiotic-resistant bacteria grow in the biofilms that form in irrigation systems, along with the generic bacterial populations. The notion that natural aquatic environments are an important reservoir for antibiotic-resistant bacteria is underscored by our conclusion that irrigation systems utilizing surface waters contribute to the dissemination of antibiotic-resistant bacteria, potentially into the food supply. The biofilms that form within irrigation systems are a complex environment, and the concentrations of antibiotic-resistant and multiple-drug-resistant bacteria in these dynamic communities can change over time and differ among bacterial groups.