Environ Sci Pollut Res DOI 10.1007/s11356-014-2911-y

RESEARCH ARTICLE

Antimicrobial potential of the ionophore monensin on freshwater biofilm bacteria Cynthia L Winkworth & Gavin Lear

Received: 16 December 2013 / Accepted: 14 April 2014 # Springer-Verlag Berlin Heidelberg 2014

Abstract Microorganisms play key roles in stream ecosystems, but comparatively little is known about the resilience of freshwater bacterial communities and their susceptibility to the chemical by-products of agricultural land use. Antibiotics used in the agricultural sector are of particular concern and have been detected in waterways associated with agricultural land. Despite widespread agricultural intensification globally and the sector's high antibiotic use, the effects of agricultural antibiotic by-products on stream microbial communities have yet to be characterised. We investigated the impacts of the antibiotic monensin on microbial biofilm communities in a simulated contamination event using streamside-replicated channels. A 24-h pulse experiment in flow channels precolonised by stream biofilm microbial communities contrasted the effects of monensin concentrations ranging from realistic to extreme toxicity levels (1–550 ug L−1). Biofilm community composition was characterised immediately before and after the pulse for several weeks using automated ribosomal intergenic spacer analysis. Despite applying acutely toxic levels of monensin, only limited effects to biofilm community composition were detected immediately after antibiotic application, and these disappeared within 4 days. Rather, temporal factors drove biofilm differences, highlighting the overriding importance of

Responsible editor: Robert Duran Electronic supplementary material The online version of this article (doi:10.1007/s11356-014-2911-y) contains supplementary material, which is available to authorized users. C. L. Winkworth (*) Department of Zoology, University of Otago, PO Box 56, Dunedin 9054, New Zealand e-mail: [email protected] G. Lear School of Biological Sciences, The University of Auckland, Auckland, New Zealand

wider, catchment-level, physiochemical hydrological influences on structuring freshwater biofilm communities, as opposed to localised and sporadic agricultural surface runoff contamination events containing antibiotics. Keywords Monensin . Effluent . ARISA . Biofilm . Freshwater . Bacterial community

Introduction Global agricultural land use intensification is typically characterised by the input of a wide range of contaminants to streams, with the potential to affect resident organisms (Winkworth 2013). While macro-organism responses have been well studied (Hillis et al. 2007; McGregor et al. 2007; Choung et al. 2013), knowledge of the resilience or susceptibility of microorganisms to such contaminants is limited (Pesce et al. 2010), despite being vital components of stream ecosystem functioning. Performing a range of tasks crucial for effective ecosystem functioning, like leaf decay (Das et al. 2007), nutrient processing (Battin et al. 2003), carbon cycling (Finlay et al. 1993) and oxygen metabolism (Young et al. 2008), as well as being a basal food source (Costanzo et al. 2005), changes to microbial biofilm communities could result in key ecological processes being disrupted. Antibiotics are recognised as emerging chemicals of concern in aquatic environments (Engemann et al. 2008), since they are specifically designed to cause biological effects (Halling-Sørenson et al. 1998), yet their effect on freshwater microbial processes or community structure remain unclear. While multiple uses exist for antibiotics, the focus here is in dairy-based agricultural settings given the sector’s high antibiotic use for disease treatment, feed efficiency and growth promotion (Sarmah et al. 2006). Globally, agricultural antibiotic use is widespread (Sarmah et al. 2006), with the

Environ Sci Pollut Res

subsequent excretion by agricultural animals of up to 50–70 % of administered antibiotics as unmetabolised or degraded yet still active compounds reported (Hillis et al. 2007; Sarmah et al. 2006; Watanabe et al. 2008). As chemical and biological contamination of waterways in agricultural settings from surface runoff is well documented (Dossky 2001; Mawdsley et al. 1995), it is not surprising antibiotic compounds have now also been detected in waters traversing agricultural land (Costanzo et al. 2005; Kemper 2008; Watkinson et al. 2009). It should be noted however, that detection likely reflects recent analysis improvements (Castiglioni et al. 2005; Song et al. 2007), as opposed to being truly ‘emergent’ (Lapworth et al. 2012), given the long history of antibiotic use in agricultural settings (Khachatourians 1998). Agricultural antibiotic use for disease treatment, growth promotion and improvements to feed efficiency in animals represent a considerable proportion of total antibiotic use in many countries (Kemper 2008; Sarmah et al. 2006). While some overlap exists with antibiotics employed for both human and animal disease treatment, the majority of antibiotics used in the agricultural sector is for nontherapeutic purposes (Sarmah et al. 2006). One particular group of antibiotics, collectively known as ionophores, is used extensively in agricultural settings to improve feed efficiency and reduce bloat (Hansen et al. 2009). By actively altering microbial rumen flora, the commonly used ionophore monensin works to promote feed breakdown to propionic acid and reduce methane gas production (Russell and Houlihan 2003). However, significant levels of unmetabolised monensin and degraded yet still active compounds are excreted into the environment (Donoho 1984). Displaying a low affinity to bind to soil (Dolliver et al. 2008), slow photolysis (Watanabe et al. 2008) and the ability to persist in surface waters (Lissemore et al. 2006), it is not surprising that monensin has been detected in waterways (Watanabe et al. 2008). While some knowledge about monensin ecotoxicity to aquatic organisms exists, it is largely restricted to organisms at a higher trophic level. Some algal species have been reported to be negatively affected (Salis 2010) and zooplankton indirectly affected as a result of algal food source modifications in mesocosm studies employing 50 ug L−1 (Hillis et al. 2007). Any negative impacts of antibiotics on bacterial biofilm nutrient cycling and decomposition remain largely unknown, as do negative effects on biofilm quantity and/or quality as a food source for higher trophic levels organisms that consume it. Over the past two decades, the New Zealand landscape has undergone significant land-use change, primarily the conversion from low-density beef and sheep farming to intensive dairy farming on the South Island (PCE 2004). A dairy herd population of over 4.6 million cows now exists, up from 2.4 million 20 years ago (LIC 2007). Of particular concern is the high use of antibiotics in the New Zealand agricultural sector, 57 % of total antibiotic use in 2000 (Sarmah et al. 2006),

indicating an inevitable increase of antibiotics on the landscape given the substantial dairy herd population increase. Related with these land use changes has been the increased input of a wide range of chemical and biological contaminants to rivers and streams (PCE 2004), often via surface runoff as a result of saturation excess overland flow during the winter months (Muirhead et al 2006). Addressing knowledge gaps as to the effect of agricultural antibiotic use on freshwater microbial biofilms is paramount. Of particular importance is identification of any alterations to key ecosystem services like the decomposition of organic matter and biochemical pathways (Sigee 2005). Being comprised of longer-term microbial residents as compared with microbes present transiently in the water column, stream biofilms provide an appropriate focus for the study of community-level responses to anthropogenic pressures. Here, a manipulative, streamside channel experiment was performed to quantify freshwater bacterial community-level responses to agricultural contaminants. The contaminants chosen for investigation in this study were dairy cow effluent and the routinely used ionophore antibiotic monensin over a range of toxicity levels. Simulating a saturation excess surface runoff event typical of New Zealand conditions (Winkworth et al. 2008), established stream channel biofilms were assessed immediately prior to and following a 24-h contamination pulse, as well as over a 17-day recovery period. To overcome the difficulties of culturing microbes from the environment (Schmeisser et al. 2007; Pusch et al. 1998), biofilm communities were screened using a DNA-fingerprinting approach (Automated Ribosomal Integenic Spacer Analyses, or ARISA). We predicted that biofilm bacterial communities would change in response to increasing antibiotic toxicity, with the least similarity comparing communities subject to no antibiotic versus the highest concentrations. We further predicted that, following a pulse of contamination, similarity between the bacterial community compositions across the contaminant treatments would first decrease but later increase as hydrological physiochemical parameters outweighed pulse effects. Channel shading, simulating waterway riparian plantings, was predicted to increase the degree of toxicity observed due to slower monensin UV photolysis, thereby impacting bacterial community structure over a longer period of time. Effluent was predicted to provide both a source of additional microorganisms (altering community structure) and nutrients (increasing leaf decomposition rates).

Methods Experimental setup A manipulative field-based experiment was performed in the austral summer between 22 December 2009 and 31 January 2010 in 80 circular streamside channels supplied continuously

Environ Sci Pollut Res

with water from the Kauru River, a third-order stream in North Otago, New Zealand (170°44.6′ East, 45°6.5′ South; 98 m above sea level). The Kauru River catchment (124 km2) received a mean annual rainfall of 817 mm in 2010, resulting in a mean annual river discharge of 1.29 m3/s adjacent to our field setup, as measured by the Otago Regional Council. The catchment is characterised by low-density sheep and beef stock grazing (0.1–3 animals per hectare) on grasslands comprising either native tussock grass species or low-producing exotic grasses (Matthaei et al. 2010) that had no significant pre-exposure to monensin. The river water is nutrient-poor (Wagenhoff et al. 2012) but contains diverse and abundant algal (Liess et al. 2008) and invertebrate communities (Herrmann 2009). The general experimental field setup has been described previously in detail by Liess et al. (2008). Briefly, it consisted of a two-level scaffold, the top level of which supported five polythene header tanks (135 L) supplied entirely and continuously with river water by a centrifugal pump (average rate of 166 L m−1) that, in turn, fed each of 16 individual circular stream channels via gravity, supported on the lower scaffold level (Fig. 1; modified from Liess et al. 2008). Header tank water entered the nylon channels (Microwave Ring Moulds, Interworld, Auckland, NZ; 25 cm external diameter, 6 cm inner diameter) via a plastic jet at an acute angle to create circular flow, before discharging over the lowered inner channel edge. Tap regulators between the jets and the header tank permitted adjustment of flow, with current velocity checked every second day to maintain approximately 7.0 cm s−1. Experimental design A full factorial, repeated-measures design was used to assess the individual and combined effects of the agricultural antibiotic monensin, dairy farm effluent and light on biofilm communities. The experiment comprised a 23-day colonisation period to facilitate natural colonisation of the channels by

stream biofilm communities and invertebrates followed by a 17-day manipulation. The study design included five antibiotic levels (0, 1, 55, 300 or 550 μg monensin L−1), two dairy effluent levels (effluent added or no effluent added), two light levels (minimally shaded and light reduced) and four replicates of each treatment combination, resulting in 80 experimental units per collection day. Contaminants Antibiotic and effluent treatments were both applied during a 24-h pulse on the first day of the experimental phase, after which the stream channels were allowed to recover for 17 days (Fig. 2). In contrast, light levels were manipulated for the entire 23-day colonisation and 17-day experimental periods (41 days total). Dairy farm effluent and the agricultural antibiotic monensin (sold commercially as Rumensin® Trough Treatment; Div. Eli Lilly and Co NZ Ltd, Auckland, NZ) were applied as continuous flows for 24 h to approximate a realworld situation of dairy farm runoff following a heavy rainfall event, while shading simulated the presence of riparian stream-side vegetation. During the experimental phase, monensin was continuously dripped into four header tanks from four 300-L groundlevel barrels (each containing nominally 1, 55, 300, 550 μg monensin L in Kauru River water) using battery-driven fluidmetering pumps (FMI CERAMPUMP® Lab Pump QBG, Fluid Metering, Inc. USA) at 1.9 L per hour over the 24-h pulse period. A fifth treatment level of 0 μg antibiotic delivered Kauru River water from a fifth ground-level barrel to the fifth header tank over the 24-h pulse period in the same manner as the antibiotic treatments. The four nominal monensin concentration levels equated to those previously detected in agricultural streams (1 μg L−1; Lissemore et al. 2006), the recommended concentration of monensin administered to each dairy cow per day via reticulation drinking water systems (300 μg L−1) and levels identified as toxicity

0 -1 µg L

1 µ g L-1

55 µ g L -1

300 µ g L -1

Channels

Header tanks

Water enters the channels via header tanks from the river

550 µ g L-1

Channels: Minimally shaded No effluent Effluent added Shaded No effluent Effluent added *note only one of the four replicate channels shown per header tank

Outflow water exits from the channels Fig. 1 Schematic diagram of experimental field setup showing header tank conditions, flow of water and subsequent treatment combinations of circular stream channels

Environ Sci Pollut Res Antibiotic & Effluent pulse applied Light manipulation applied throughout Colonisation period

Recovery period

Experiment duration in days (* indicates sample collection days) -23

-0*+1*

+5*

+17*

Fig. 2 Overview and timeline of the experiment. Shaded area indicates the timing of antibiotic and effluent pulse; dotted area indicated the recovery post-pulse period; asterisks indicate sampling dates

thresholds for aquatic organisms (55 μg L−1 chronic exposure level, 550 μg L−1 acute exposure level; as reported by the Lilly Aquatic Exposure Guideline; LAEG; ELANCO Rumensin Trough Treatment MSDS 2007). Light assay Light treatments consisted of either minimally shaded or lightreduced levels, with light reduced in half the channels using hoods made of garden shade cloth (Nylex, Melbourne, Australia) that covered the channel entirely and as utilised by previous stream channel experiments (Lange et al. 2011; Liess et al. 2008). Shade cloth was applied to the selected channels at the beginning of the colonisation period. All 80 channels were individually covered entirely with fabric hoods (DuPontTM Tyvek®, Wilmington, DE, USA) to standardise air circulation above the channel water surface, as well as reduce the introduction of aerial invertebrates and debris. For the light-reduced treatments, shade cloth was placed on top of the fabric hood. Channel light intensity and temperature were measured continuously during the experimental period using HOBO® Temperature/Light Pendant Data Loggers (Onset Computer Corp., Pocasset, MA, USA) in four channels each for minimally shaded and light-reduced treatment conditions. Shade cloth reduced light intensity by an average of 73 % (17,066 lux versus 64,479 lux) and temperature by an average of 6.6 % (20.2 °C versus 21.6 °C), as measured at 12 pm, in the shaded versus minimally shaded channels, respectively. Treatment conditions were assigned to the five header tanks and 80 channels prior to the colonisation period. The five different antibiotic treatment levels were randomly assigned to one each of the five header tanks, with four of the 16 channels per header tank each assigned to the different light treatments (shade/effluent, shade/no effluent, no shade/ effluent, no shade/no effluent). Biofilm assay Eight unglazed terracotta tiles (100×100×14 mm) pre-soaked in distilled water for 24 h to remove any manufacturing surface contaminants were placed rough-side down around the bottom of each channel. Space constraints within the channels prevented vertical tile placement. As minimal

channel sedimentation had been observed in previous channel-based experiments (Piggott et al. 2012), largely owing to how river water was delivered via the header-tank pump setup, sedimentation on the horizontally placed tiles was considered minimal. Tiles were exposed to river water flowing continuously through the channels without further disturbance for the 23-day colonisation period to facilitate biofilm development. Two days prior to the antibiotic and effluent pulses, tiles were moved to shade-treatment appropriate holding tanks and fed continuously by flowing river water for a few hours while channel walls were wiped clear of biofilm and any other debris. This insured multiple tiles were not connected by the same biofilm, which otherwise would have become disruptive when attempting to sample individual tiles during the experimental timeframe and also ensured a degree of standardisation of initial biofilm colonisation of the leaf packs. The colonised tiles were randomly redistributed back into either minimally shaded or light-reduced channels, ensuring the original colonisation period light treatment was maintained. Leaf packs As a bioassay to measure organic matter decay, leaf packs comprising 10 g wet-weight of the fast decomposing native shrub mahoe Melicytus ramiflorus Forst and Forst (10–15 fresh leaves stacked and riveted together) were added to the side of each newly wiped-clean channel opposite the inflow jet to minimise turbulence 2 days prior to the antibiotic and effluent pulses. Leaf packs were left undisturbed for the duration of the experimental period until sampled 16 days postpulse (Day 17) to determine decomposition rates between treatments. Sample collection Biofilm on the terracotta tiles was sampled from all treatment combinations on four collection days: immediately pre- (day 0) and post-pulse (day 1), then post-pulse on day 5 and concluding with a final sampling occasion on day 17 (2 light levels×2 effluent levels×5 antibiotic levels×4 replicate tiles× 4 collection days=320 biofilm samples). Sampled tiles were replaced with cleaned terracotta tiles but not resampled during

Environ Sci Pollut Res

the experiment. On each collection date, one toothbrush previously sterilised in a 10 % bleach solution for 1 h and rinsed thrice using Milli-Q water was used to scrub the entire surface area of each sampled terracotta tile (one clean toothbrush per tile). Loosened biofilm was collected in a 50-ml conical tube (BD Falcon™, Franklin Lakes NJ, USA) using laboratorysourced distilled water. Any visible invertebrates were removed prior to scrubbing to minimise interference with subsequent molecular analysis. Tubes were immediately stored on ice and frozen at the end of each collection day without centrifugation for long-term storage at −20 °C in the dark until processing. Laboratory procedures Leaf packs Leaf packs collected on Day 17 (16 days post-pulse) were rinsed clean of any associated invertebrates and transported on ice to the laboratory for processing. Five individual leaves per pack from every treatment combination were measured for Leaf Strength, defined as the weight required to force a blunt metal pin through the leaf using a penetrometer (avoiding leaf veins; Young et al. 2008) and compared with the averaged strength of 25 fresh leaves measured at the time leaf packs were prepared pre-experiment. Leaf biomass was calculated for every leaf pack as ash-free dry mass using standard methods (APHA 1998) and compared with the ash-free dry mass average of ten packs prepared and measured at the same time as those prepared for placement in the channels. DNA extraction Nucleic acids were isolated from the biofilm samples using PowerBiofilm™ DNA Isolation Kits (MO BIO Laboratories, Inc., Carlsbard, CA, USA). The manufacturers protocol was followed except for the initial homogenisation step, where samples were instead homogenised using a Geno/Grinder® 2010 (SPEX CertiPrep Group L.L.C., Metuchen, NJ, USA) for 2 min at 1,500 strokes m−1, due to equipment availability. ARISA Automated ribosomal intergenic spacer analysis (ARISA) was performed on all 320 samples using the primer pair SDBact (5’-TGCGGCTGGATCCCCTCCTT-3’; Life Technologies New Zealand Ltd, Auckland, NZ) and LDBact (5’-CCGG GTTTCCCCATTCGG; Life Technologies New Zealand Ltd, Auckland, NZ; Ranjard et al. 2001), with the following amplification conditions: (1) 95 °C for 5 min; (2) 30 cycles of 95 °C for 30 s, 61.5 °C for 30 s, 72 °C for 90 s and (3) 72 °C for 10 min. The SDBact primer was labeled at the 5’-end with HEX (6-carboxyhexafluorescein) fluorochrome (Life

Technologies New Zealand Ltd, Auckland, NZ). The PCR mix comprised 0.5× Sahara Master Mix (Bioline, London, UK), 200–300 nM each of the forward and reverse primers, purified DNA and Milli-Q water to a total volume of 25 μL. Amplified products were visualised on Tris-buffered 1 % agarose gels (VWR International Ltd, Poole, UK; Pure Science Ltd, Wellington, NZ) containing 3 % SYBR Safe DNA gel stain (Invitrogen™ New Zealand Ltd, Auckland, NZ). Products were purified (Zymo DNA clean and Concentrator kit; Ngaio Diagnostics Ltd., Nelson, NZ) and 1 μL combined with 10 μL Hi Di formamide and an internal LIZ1200 standard (ABI Ltd., Melbourne, Australia) before being heat-treated (95 °C, 5 min) and subsequently cooled on ice. To generate ARISA profiles of bacterial community structure, samples were run on a 3130XL Capillary Genetic Analyser (Life Technologies New Zealand Ltd, Auckland, NZ) using a 50-cm capillary and standard genemapper protocol, but with an increased run time (15 kV,18 h). Data analysis Leaf packs Decomposition rates estimated from Leaf Decay and Leaf Strength were calculated in terms of an exponential decay coefficient (k per day; Young et al. 2008) and refer to the proportion of leaf material lost per day. ARISA GENEMAPPER software (v. 3.7; Life Technologies NZ Ltd, Auckland, NZ) was used to assign a fragment length (in nucleotide base pairs) to ARISA peaks, via comparison with the standard ladder (LIZ1200; Life Technologies NZ Ltd, Auckland, NZ). To include the maximum number of peaks whilst excluding background fluorescence, only peaks with a fluorescence value of 50 U or greater were analysed. As the 16S–23S region is thought to range between approximately 140 and 1,530 base pairs (Fisher and Triplett 1999), fragments 1,000 bp. The total area under the curve was normalised to 100 to remove differences in profiles caused by different initial DNA template quantities and peak size rounded to the nearest whole number. Using this approach, the data generated for every sample were the relative abundance of each of 850 bacterial taxa (which represent the length of the intergenic region of constituent bacteria, from 150 to 1,000 bp). Data analysis The similarity in bacterial community composition among samples was documented using a Bray-Curtis similarity

Environ Sci Pollut Res

measure. To visualise multivariate patterns in bacterial community data, multidimensional scaling (MDS) of the ARISA data was done in PRIMER v.6 (Primer-E Ltd., Plymouth, UK). To test for statistically significant differences among ARISA datasets, permutational analyses of variance (PERMANOVA) were completed by comparing Bray Curtis distances among sample profiles using the ‘ a d d - o n ’ P E R M A N O VA + p a c k a g e f o r P R I M E R (Anderson et al. 2008). PERMANOVA + allowed the multivariate information to be partitioned according to the full experimental design (including all interaction terms) and made no assumptions regarding the distributions of the original variables as all P values were obtained by permutation. The overall experimental design consisted of four factors: Day (fixed: with four levels, 0, 1, 5 and 17 days), Shade (fixed: with two levels, yes or no), Effluent (fixed: with two levels: yes and no) and antibiotic (fixed: with five levels, 0, 1, 55, 300 and 550 μg L–1). In addition to the PERMANOVA tests, contrasts were used to compare differences among each treatment combination. Direct multivariate analogues to the usual ANOVA estimates for variance components (Searle et al. 1992) were used to quantify the variability associated with each source of variation in the model. In order to be on the same measurement scale as the original Bray Curtis measure used for this analysis, these were expressed in terms of their square root. Additionally, ARISA datasets were compared using multivariate dispersion index values in PRIMER (MVDISP; based on Bray Curtis distance among samples), which provide a relative ‘score’ of the multivariate variability within each of the groups in a single ordination.

Table 1 PERMANOVA of leaf-litter decay rates obtained from flow chamber mesocosms on the basis of the Bray-Curtis similarity measure, showing the partitioning of multivariate variation and tests for the factors of Shade (Sh), Effluent (Ef) and Antibiotic (An), and their interactions Source of variation

df

SS

F

P

Sh Ef An An × Sh An × Ef Sh × Ef An × Sh × Ef Residuals

1 1 4 4 4 1 4 60

2 10 191 20 4 2 55 756

0.16 0.76 3.80 0.39 0.08 0.16 1.10

0.696 0.376 0.009 0.813 0.988 0.697 0.367

d.f. degrees of freedom, SS sum of squares, F F-ratio of the explained variance to the unexplained variance, P significance values obtained using 9,999 permutations of residuals under a reduced model, Sq. root square root of the component of variation attributable to that factor in the model, in units of Bray-Curtis measure

Results Leaf decay Of the three different treatment conditions of shading, effluent and antibiotic, only the antibiotic was found to have significantly affected leaf litter decay rates overall (PERMANOVA, P