Dissipation and Transport of Veterinary Sulfonamide Antibiotics after

Sep 29, 2007 - Dissipation and Transport of Veterinary Sulfonamide Antibiotics after Manure Application to Grassland in a Small Catchment. Krispin Sto...
1 downloads 15 Views 389KB Size
Environ. Sci. Technol. 2007, 41, 7349-7355

Dissipation and Transport of Veterinary Sulfonamide Antibiotics after Manure Application to Grassland in a Small Catchment K R I S P I N S T O O B , †,‡,| H E I N Z P . S I N G E R , † STEPHAN R. MUELLER,§ R E N EÄ P . S C H W A R Z E N B A C H , ‡ A N D C H R I S T I A N H . S T A M M * ,† Swiss Federal Institute of Aquatic Science and Technology, Eawag, 8600 Du ¨ bendorf, Switzerland, Institute of Biogeochemistry and Pollution Dynamics, ETH Zurich, 8092 Zurich, Switzerland, and Federal Office for the Environment (FOEN), Papiermu ¨ hlestrasse 172, CH-3003, Bern; Switzerland

The heavy use of veterinary antibiotics in modern animal production causes concern about risks of spreading antibiotic resistance after manure applications to agricultural fields. We report on a field study aiming at elucidating the fate of sulfonamide (SA) antibiotics in grassland soils and their transport to surface water. Two controlled manure applications were carried out under different weather conditions. After both applications, the SA concentrations in pore water and the total soil content declined rapidly. This stage of fast decline was followed by a second one during which the SA were rather persistent. More than 15% of the SAs applied were still present in the soil 3 months after application, always exceeding 100 µg/ kg topsoil. The apparent SA sorption increased strongly with time. Accordingly, the risk for SA losses to water bodies decreased within 2 weeks to very low values. In contrast to SA concentrations in the soil, losses to the brook were strongly influenced by the weather conditions after the two manure applications. The overall losses were 15 times larger (about 0.5% of applied SA) during the wet conditions of May 2003 compared to the dry conditions following the first application (March 2003).

Introduction Antibiotics are an integral part of modern livestock production. In the U.S., the consumption for veterinary purposes amounts to about 9000 tonnes per year (1), and about 5000 t are administered annually in the European Union, which is about one-third of the entire consumption (2). The administration of antibiotics in livestock involves the risk of introducing active substances into the environment because high percentages of some antibiotics are excreted in their active form. To assess the risk caused by such residues it has to be known how they persist and distribute in the environment after their release. Among the different classes of veterinary antibiotics, the sulfonamides (SA) are of particular interest. On the one hand, * Corresponding author phone: +41-1-823 55 65; fax: +41-1-823 58 26; e-mail: [email protected]. † Eawag. ‡ ETH Zurich. § Federal Office for the Environment (FOEN). | Present address: RCC, Ltd., Zelgliweg 1, 4452 Itigen, Switzerland. 10.1021/es070840e CCC: $37.00 Published on Web 09/29/2007

 2007 American Chemical Society

they are important because of their frequent use, e.g., within the EU (3) and in Switzerland (4). Because of their high excretion rates (5), and their persistence (6) they are expected to be present in manure. Indeed, up to 20 mg SA/kg of liquid manure have been found in samples from different farms in Switzerland (7). The application of manure containing such concentrations results in loads of several hundred grams of sulfonamides per ha and application. Such rates are in the same order of magnitude as herbicide applications in crop production, which are well-known to cause surfaces water pollution (8). Furthermore, SAs have been considered as very mobile in soils due to their supposed weak sorption (9, 10). Indeed, several plot and field studies support this view. A comparative study for example, investigating losses of different antibiotics from arable soils plots, revealed very small apparent Kd values for sulfamethazine (11). We obtained similar values for the same compound on grassland plots (12). These data suggest a significant potential for SA losses to waters bodies. Leaching of SA to groundwater has actually been demonstrated on manured fields (13) and overland flow was identified as an important loss mechanism (14-16). The general view of SAs as being very mobile has been challenged more recently by new findings from field and laboratory experiments. In a controlled field study on an arable clay soil, SA recoveries from the soil after 1 day did not exceed 15% and a rapid disappearance in the soil was observed (17). In batch experiments, we showed that SA sorption got much more pronounced the longer the contact time with the sorbents. This held for organic material like compost or humic acids as well as for inorganic sorbents such as iron hydroxide or clay minerals (18, 19). Additionally, it has been demonstrated that SAs may get transformed either by microbiological, enzymatic (outside of living cells), or purely chemical processes (20, 21). Part of the discrepancies between studies on SA mobility may be attributed to different time scales of contact time of SAs with the soils. Weak sorption was mainly found during short-term experiments. Additionally, factors like pH, presence of manure, soil properties, or moisture conditions differing between the experiments may also have contributed to the partially contradicting results. Laboratory experiments for example suggest increasing dissipation with higher soil moisture contents (22). How the different controlling factors interact in the field, is still insufficiently understood due to the limited number of field studies investigating the behavior of SA under real-world conditions. Accordingly, the aim of this study was to quantify the fate of selected veterinary SA in the topsoil of a grassland site and to link the temporal development of soil content with the SA losses to surface water under variable weather conditions. This was achieved with two controlled manure applications followed by highfrequency monitoring of the SA concentration in the topsoil as well as in the brook over 3 months.

Experimental Section Compounds. Five different SA antibiotics were chosen for this study (Table 1, Table S4). The selection was based on their use as veterinary medicines in Switzerland. Additionally, sulfamethoxazole was included as the most important sulfonamide in human medicine. All five SA have two pKavalues and speciation changes from anionic to neutral and even cationic when decreasing the pH within the environmentally relevant range. All unlabeled compounds were purchased from Sigma-Aldrich (Buchs, Switzerland) with a purity of at least 98%. VOL. 41, NO. 21, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

7349

TABLE 1. Structures and Substance Properties of the Sulfonamides Investigated. References for the Properties Are Provided in Table S4 (Supporting Information)

a

Calculated values using Advanced Chemistry Development (ACD/Labs) Software Solaris V4.67.

Rainfall data were collected by an automatic meteorological station (Manifold, equipped with Micrologger 21X, Campell Scientific, Inc.) in 10 min intervals from March to July 2003. Soil moisture status was determined in 60 min intervals in a soil profile on Field 2 at three different depths (2.5, 7.5, and 15 cm). One thermistor was installed per depth; TDR probes and tensiometers were installed in duplicates. The water table was measured manually every few days with five piezometers forming a transect across the two fields (see Figure S5, Supporting Information). A water sampling station was installed at the outlet of the catchment about 500 meters downstream of the two fields. It consisted of a wooden channel of 5 m length with a defined cross section. More details can be found in ref 23. The catchment discharge was recorded automatically every 5 min using ultrasonic Doppler probes (Isco 750 area velocity flow module, Isco, Inc.).

FIGURE 1. Aerial view of the investigated catchment with the water sampling station at the outlet of the catchment and the two manured fields. Site Description and Instrumentation. The study site (Figure 1) was located in the catchment of Lake Greifensee, southeast of Zu ¨ rich, Switzerland. The two fields investigated are slightly sloped (average slope Field 1: 4-6°, Field 2: 1.53.5°) toward the brook separating them. The size of each field was about 0.4 ha. The soil was classified as loamy Eutric Cambisol (FAO) with 33.4% sand and 23% clay, an organic carbon content of 3.6%, and a pH of 5.7 (0.01 M CaCl2; all values hold for the top 15 cm). Further soil parameters are given in ref 14. 7350

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 21, 2007

Controlled Manure Applications. Due to the lack of pig farms in the catchment and the neighboring areas, no manure containing significant amounts of sulfonamides are used in the study catchment. To the best of our knowledge the site chosen had not received any manure containing sulfonamides in the past decade preceding our experiment. Accordingly, the manure used in this experimentsimported from a large pig farm in a different regionswas the only source for these compounds. The data confirmed the assumption that no measurable background contamination was present in the soil or brook water. The manure was applied according to good agricultural practice using a band spreader yielding distinct manure tracks of about 5 cm width separated by about 20 cm of untreated surface (see p. 78 in ref 24). We carried out two controlled applications (I and II, Table S5): at the beginning of the growing season on March 24th and after the first mowing on May 8th, 2003. Ten-meter wide buffer strips were maintained between the brook and the manured area. The manure of the same origin and quality as described in ref 14, which was

applied, contained sulfamethazine due to regular treatment of the pigs with medical feed. The manure used for a given field and a given application was spiked with one additional sulfonamide (sulfadiazine, sulfadimethoxine, sulfamethoxazole, or sulfathiazole) used as marker (Table S5, Supporting Information). The marker compounds were prepared as buffered sulfonamide solutions using 3 M tris buffer and ultrasonication. The spiking of the manure was accomplished by first filling the tank with one-third of the liquid manure, then adding the marker solution and finally filling in the remaining two-thirds of the manure. Spiking occurred several hours before the spreading. Sampling Strategies and Analytical Procedures. The manure was sampled from the tank at the beginning, during, and at the end of the application procedure. In addition, the manure reaching the ground during the application was collected on plates containing 1 L of distilled water to control the homogeneity of the mixing in the tank. All liquid manure samples were homogenized with a kitchen blender a few hours after the application and stored in 1 L plastic containers at -20 °C until pressurized liquid extraction (PLE). For PLE, 3 mL of liquid manure were dripped into an extraction cell prefilled with diatomaceous earth and extracted like the soil samples. Details of the PLE method are given in ref 25. Briefly, a mixture of buffered water (pH 9) and acetonitrile (85:15 (v/v)) served as solvent for the extraction at 200 °C and 10 MPa conducted during 14 min with an ASE 200 (Dionex, Sunnyvale, CA) accelerated solvent extractor. After extraction, the cell was flushed with an additional pore volume of the extraction solvent and purged with N2. One mL of the extract was spiked with internal standard solution containing isotope labeled derivatives of the five investigated sulfonamides, buffered to pH 4.5 with acetate buffer and filtered prior to analysis with reverse phase liquid chromatography-tandem mass spectrometry (LC-MS/MS). Absolute extraction recoveries determined in spiked manure (spiking concentration: 5 mg/L, contact time 90 min) were 89 ( 13% indicating almost quantitative extraction of the sulfonamides. Soil samples (0-5 cm depth) were taken with a 5 cm diameter split tube core sampler. Results from a previous plot study on the same field indicated that this depth would be a reasonable choice (12) because the SAs are strongly retained in that soil layer. Due to the band spreader application, the spatial pattern on the field was very heterogeneous at the scale of the diameter of the core sampler. This required an adapted spatial sampling strategy to determine average field concentrations. The manure tracks caused by the band spreader were marked right after the applications. Five cores were taken in a row perpendicular to the manure tracks (see Figure 4-1 in ref 24) covering the width of a manure track and the space between. The same track was sampled at five different locations distributed over the field. These “five times five” cores were pooled to one composite sample. Three replicates of three different manure lines were taken on every sampling date except after mowing (day 45 after application I). Additionally, one random sample composed of 24 cores was taken on each field on every sampling day. All samples were weighed, stored at -20 °C, milled under frozen CO2, and homogenized with a sample splitter (Rentsch, Haan, Germany) before analysis. The soil was sampled 17 times: two weeks prior to application and 1, 4, 8, 11, 18, 24, 31, 36, 45 days after the first and 1, 4, 7, 11, 18, 28, 41 days after the second application. Samples prior to application and from days 1, 4, 8, 11, and 45 after application I and from days 1, 4, 11, and 41 after application II were selected for analysis. Two different methods of analysis were applied: (i) pressurized liquid extraction (PLE) to extract SA as exhaustively as possible from the soil (see above) and (ii) analysis

FIGURE 2. Time-series of rainfall (A), discharge (B), concentration (C), and cumulative load of sulfamethazine (D) in the brook after application. The second manure application is marked with the dashed line. Details for the indicated events (e.g., timing) are reported in Table S6.

FIGURE 3. Temporal development of the sulfonamide content in the topsoil (extracted by PLE). Average of triplicates with standard deviation, Bars: applied rates by manuring (gray: sulfamethazine, white: marker compound). The error bars indicate the standard deviations. of the pore water to assess the fraction available for transport. The pore water was gained by ultra centrifugation of 20 g soil VOL. 41, NO. 21, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

7351

FIGURE 4. Temporal development of the sulfonamide content in pore water of the topsoil. The two manure applications are marked with the dashed lines. The error bars indicate the standard deviations. in 38.5 mL thick-wall polycarbonate tubes at 160 000g at 20 °C for 30 min (Centrikon T-2000 with fixed-angle rotor TFT 70.38, Kontron Instruments, Zurich, Switzerland). One mL of the supernatant was buffered, spiked with internal standards, and filtered prior to LC-MS/MS analysis. At least two different samples per sampling date were taken for analysis, and at least one of them was extracted in duplicate. Measuring the sulfonamide concentration in the pore water as well as the “total” soil content by PLE allowed calculating a distribution ratio between pore water and soil matrix (see the Supporting Information). Losses of the sulfonamides to the brook were assessed by taking flow proportional composite water samples. A subsample of 90 mL was collected every 5-25 m3 of discharge, depending on the flow conditions. Ten sub-samples were automatically pooled to one composite sample by the sampler (Isco 6700, Isco, Inc.). Samples were transferred into 1 L glass bottles and stored in the dark at 4 °C. A total of 638 samples was taken in the period from March to July 2003, of which 313 were measured with fully automated online solid-phase extraction LC-MS/MS, using column switching techniques (details given in ref 26). Briefly, 20 mL of the filtered samples were spiked with internal standard solution, buffered to pH 4.5 and enriched automatically on an Oasis HLB extraction cartridge. The sorbed analytes were directly eluted to LCMS/MS with a mixture of water (pH 2.7) and methanol (50: 50 (v/v)). The mobile phase for LC gradient elution was composed of the SPE-eluate and 10 mM ammonia acetate at different flow ratios.

Results and Discussion Rainfall and Hydrological Conditions. The periods after the applications differed substantially with regard to rainfall timing. Application I was followed by a dry week and the first rainfall occurred after 7 days (Figure 2; event I-1). It caused little discharge in the brook (Table S6), which was completely dry before. In contrast, the period following application II was wet, starting with an intensive storm in the night right after application (event II-1). The subsequent week was dominated by several rain events rendering the soil close to saturation throughout the profile (Figure S6, Supporting Information). Brook discharge reacted quickly to rainfall but 7352

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 21, 2007

dropped to low values within few days. The wettest period of the study period was about two weeks after application II with the subsequent discharge event II-5 lasting over a whole week (Figure 2). Due to these wet conditions and more runoff after application II, higher losses to the brook could be expected during this period. Dissipation in Top Soil. In spite of the significantly warmer and wetter soil conditions after the second applications (Figure S6), the SA fate in soil was very similar during both periods (Figure 3). One day after application, the total amount recovered by PLE was 30-60% of the amount applied on both fields and after both applications except for sulfamethoxazole. Initially, dissipation rate was fast but slowed down considerably (Figure 3, Figure S1, S6, and S7) and 3 months after application, residuals of the markers of application I (sulfadiazine and sulfathiazole) amounted still to more than 15% of their applied mass. In absolute values, we detected the two compounds always at levels exceeding 100 µg/kg soil. For a new pharmaceutical in the process of registration, this value would trigger a detailed risk analysis (27). These data suggest a higher stability of sulfonamides in the fields than reported by others (17). One possible methodological reason is the extraction procedure we used. Without the harsh extraction conditions, the apparent dissipation would have been much more pronounced (25). The only exception was sulfamethoxazole (marker for application II, Field 1): The recovered amount was very small (8% on day 1) and concentrations fell below detection limits within a few days. Smaller recoveries of sulfamethoxazole compared to the other sulfonamides were also observed in spike experiments with soil (25). An alternative explanation for the stability of the SA was proposed by Wang et al. (28) with their availability-adjusted first order (AAFO) model. This model assumes that the dissipation process follows first order but the availability of the compounds decreases exponentially with time (SI-2, Supporting Information). It predicts a decreasing dissipation rate with time, which is in agreement with a two-stage sorption process that we observed in batch experiments (19). This approach describes the observed field data very well. The fitted model explains between 91 and 99% of the observed variance (Table S1, Figure S1). It yields half-lives for the PLE extractable amounts ranging from 10 d (SDMO), 17 d (STA) to 48 d for SDA. Sulfamethazine ranked in between with 32 d determined on both fields independently. However, it should be noted that this model predicts an accumulation of SA because for t f ∞ the predicted concentration generally does not approach zero (eq 1b, SI-2, Supporting Information). So far no data have been reported indicating such an accumulation process (e.g., ref 29). This suggests that the model does not properly describe the behavior of SAs on long timescales. Dissipation in Pore Water. The mass detected in pore water 1 day after application was about 15-30 times less than the extractable mass determined with PLE and accounted for a few percentages of the applied amount only (Figure 4). As for the PLE extractable soil content, dissipation could not be properly described by first-order kinetics over the entire period. The deviations were even more pronounced than for the PLE extracts (Figure S8, Supporting Information). Interestingly, sulfamethoxazole (marker Field 2 application II), which showed low PLE concentrations, exhibited similar pore water concentrations as the other SAs. Accordingly, the pore water content made up about 50% of the total content. The reasons for this exceptional behavior could not be clarified based on the available field data. Distribution Between Soil Matrix and Pore Water. As a measure for SA sorption, the distribution ratio between soil matrix and pore water (Dsw) was calculated (see the Sup-

TABLE 2. Cumulative Load and Normalized Losses of Sulfamethazine and the Marker Compounds for Field 1 (Sulfamethoxazole) and Field 2 (Sulfadimethoxine) during the Different Rain Eventsa cumulative load (g)

normalized lossesb

sulfamarker marker sulfamarker marker marker event methazine field 1 field 2 methazine field 1 field 2 average

FIGURE 5. Temporal changes of the distribution ratios (Dsw) between pore water and soil matrix of the different sulfonamides on the two fields. The details of the calculations are described in the Supporting Information. porting Information). For all substances, this ratio strongly increased with time (Figure 5) as assumed by the AAFO model: Dsw ranged from about 10-15 L/kg 1 day after application and reached values of about 100 (sulfadiazine, sulfadimethoxine and sulfamethazine) to 250 (sulfathiazole) 7-8 weeks later. The exception was sulfamethoxazole, which showed a very small distribution ratio (Dsw ) 0.4 L/kg). In contrast to the expectations of the AAFO model, the SA availability did not reveal the predicted exponential decline. A power law function (see SI-2, Supporting Information) described the data much better explaining 99% of the variance for all compounds, while the exponential model only accounted for 78-87% (Table S2). Combined with the first-order dissipation for the available fraction, the power law could also successfully describe the PLE dissipation (Table S3). Overall, these results indicate that the temporal increase of sorption was a main factor controlling dissipation but different models for the sorption kinetics may result in similar dissipation rates. The Dsw from the first days after application were larger than apparent sorption coefficients (Kd) of SA reported from short-term laboratory experiments. Thiele-Bruhn (10), for example, reported Kd values between 0.1 and 10 L/kg in different soils with comparable pHs. The observed distribution ratios in this field experiment were also large compared to apparent Koc values we have measured in sterile batch experiments (19). After 1 d, the field data corresponded to apparent Koc values of about 300-450 L/kg, which is about 2-4 times larger than under lab conditions. After two weeks, these values were above 1500 L/kg exceeding the batch values also 2-4 times. Interestingly, the field data reported here are in agreement with apparent Kd values of 10-20 L/kg estimated from plot studies on Field 1 with a contact time of 1 day (14). The differences compared to sterile batch experiments could possibly be explained by faster biological degradation in the pore water without instantaneous equilibration (see below). In any case, our data demonstrate that sulfonamides are not as mobile as sometimes suggested in the literature (9, 10). The sorption of sulfonamides varies with pH (18, 30). Therefore, one might argue that the sulfonamides were more mobile right after the manure application due to its basic pH. The increase of Dsw could then be explained by pH reequilibration with time. However, soil pH returned to its previous value within few days (data not shown), whereas Dsw still increased strongly with time. The increase of Dsw over time could also be due to nonlinear sorption. With the total concentration decreasing due to dissipation, the relative SA amount that is sorbed would increase. Indeed, the relationship between sulfona-

I-1 I-2 I-3 sum I II-1 II-2 II-3 II-4 II-5 II-6 II-7 sum II

0.07 0.09 0.01 0.18 0.96 0.25 0.55 0.32 0.58 0.03 0.06 2.77

na na na

na na na

1.05 0.26 0.25 0.12 0.18 nd nd 1.87

0.14 0.06 0.26 0.12 0.13 nd nd 0.72

1.8% 3.2% 0.7% 8.7% 3.5% 12.2% 16.2% 50.9% 4.5% 10.1%

na na na

na na na

13.4% 4.8% 7.1% 7.5% 26.7% nd nd

4.0% 8.7% 2.4% 3.6% 16.6% 11.9% 18.8% 13.2% 50.5% 38.6% nd nd

a na, not applied; nd, not detected. b Normalized losses were determined by division of the cumulative load (per event) by the amount present in pore water before the particular rain event.

mide concentration in the pore water and the soil matrix could be described with a nonlinear Freundlich isotherm (Freundlich n e 0.31; Table S7). However, long-term desorption experiments have revealed a pronounced sorptiondesorption hysteresis (31) suggesting that the increasing Dsw values observed in our field experiment were mainly due to nonequilibrium condition rather than pronounced nonlinearity of sorption. Transport to Surface Water. Based on the transport behavior of herbicides (8) or phosphate (32) in areas where preferential flow and surface runoff largely contribute to transport, we expected SA losses to surface waters to be driven mainly by rainfall events. This expectation was fully met as can be seen from Figure 2 depicting the pronounced concentration dynamics in the brook. The high SA concentration peaks were only observed for about 2 weeks after manure applications (Figure 2). This short period of SA losses agrees with the fast decline of SA available in the pore water. Desorption from the soil matrix was shown to be too slow for replenishing the mobile sulfonamide pool within a few hours or days (31). Due to the shorter time span between manuring and the rainfall, maximal sulfamethazine concentrations were about 20 times higher after application II compared to application I. Up to 3300 ng/L were determined in event II-1, whereas the peak concentration reached 150 ng/L during the first event after application I. Accordingly, the total input into the brook after application II (2.77 g) was about 15 times larger than following the first application (0.18 g). For application II, not only the first rain event after the application, but also event II-5 taking place two weeks later led to a substantial input into the brook (Table 2). To account for the dissipation on the fields, the cumulative loads were normalized to the amount present in pore water right before the particular discharge event (normalized losses; Table 2). In general, the normalized losses after application II (3.5-16% of the amount in the pore water) were much larger than after application I (0.7-3.2%). For event II-5, an exceptional normalized loss of more than 50% was calculated. A closer comparison of application I and II was possible for event I-2 and event II-4 due to almost identical soil moisture conditions, rain intensities, and runoff ratios. Despite these similarities, the normalized SMA loss was about 5 times larger for event II-4 than for I-2. This observation points to the influence of the weather conditions after application for the subsequent transport to the brook. Possibly, SAs had migrated into aggregates to a larger extent VOL. 41, NO. 21, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

7353

during the dry conditions after the first application rendering them less available for mobilization into fast flow components like runoff (33). Environmental Implications. The fate of the sulfonamide antibiotics after manure application in the soil revealed a two-stage dissipation process. Initially, SA concentration decreased rapidly in the topsoil. Therefore, significant losses to surface waters were limited to a rather short period of about 2 weeks after the manure application. Given the timing of manure application as well as the soil conditions (8) in the area, one can consider the losses of 0.5% of the applied amount under wet conditions as a worst case for this type of land use and climatic conditions. Accordingly, one can expect lower concentrations and smaller relative loads in other areas. This was confirmed by monitoring data in several tributaries of Lake Sempach in Switzerland draining small agricultural catchments. This region is characterized by intensive pig production such that a heavy use of SA has to be expected representing a worst case situation in Switzerland. None of the monitoring data revealed event-averaged SA concentration exceeding 100 ng/L (24). Together with the very small values found in the lake water (