This work was approved by animal care and use committees of the University of South Florida (W3228) and the University of Florida (023-08WEC). All animals used were treated humanely and with regard for alleviation of suffering.
The mesocosm experiment was conducted at the University of Florida’s Gulf Coast Research and Education Center during July and August 2008 (35 days total). Mesocosms consisted of cattle water tanks (1.8 m diameter, 60 cm deep, ~ 1,100 L) containing 800 L water, 300 g leaf litter, and local zooplankton, phytoplankton, periphyton, insect, gastropod, and crayfish species [see Supplemental Material, Table S2 (doi:10.1289/ehp.1002956)]. Mesocosms were covered with 60% shade cloth to prevent overheating and entry or escape of animals. Each tank received 10 Rana sphenocephala
(southern leopard frog) tadpoles from eight clutches (collected at N 28°06.759´, W 082°23.014´) and 25 Osteopilus septentrionalis
(Cuban treefrog) tadpoles (all at Gosner stages 25–28; Gosner 1960
) from five clutches (collected at N 28°03.537´, W 082°25.410´).
Tanks were arranged in a randomized block design with four replicates of each treatment (a total of 16 tanks). There were two control treatments, receiving either 50 mL of water or 50 mL acetone solvent (used to dissolve chlorothalonil). Tanks for the remaining two treatments received chlorothalonil (technical grade, purity > 98%; Chemservice, West Chester, PA) dissolved in 50 mL acetone so that nominal concentrations in the tanks were either one time the EEC (1×; 164 μg/L) or two times the EEC (2×; 328 μg/L). Tanks were dosed the same day as the amphibians were added, and targeted nominal concentrations closely matched the actual concentrations (1×, 172 μg/L; 2×, 351 μg/L; spiked recovery efficiencies, 95%). Thus, for simplicity and consistency across the experiments in this article, we refer to the nominal concentrations. Several water quality and chemistry variables were quantified at various times during the experiment [see Supplemental Material, “Mesocosm Experimental Methods” and Tables S3 andS4 (doi:10.1289/ehp.1002956)]. Standardized dip net sampling of each tank was conducted the third day of the experiment to quantify any rapid mortality associated with chlorothalonil exposure. The number of metamorphosed frogs was noted daily, and tadpole survival was determined 5 weeks after dosing.
Laboratory experiment I.
We obtained Hyla squirella
and O. septentrionalis
from multiple, thoroughly mixed clutches collected from two adjacent retention ponds in Tampa, Florida, in July 2008 (N 28°0.322´, W 82°19.532´). We employed a completely randomized design with 21 32-L glass aquaria, each filled with 10 L artificial spring water (Cohen et al. 1980
), with water hardness of 62.7 ppm (5B Hardness Test Kit; HACH Co., Loveland, CO) and pH ~ 7.0). Aquaria were maintained in a laboratory at the University of South Florida at 27°C and on a 14:10-hr light:dark cycle. Each aquarium received five H. squirella
and 15 O. septentrionalis
tadpoles (Gosner stages 25–28), and tadpoles were fed boiled organic spinach daily. We used five treatments of technical grade chlorothalonil (purity > 98%; Chemservice; actual concentrations, 0.176, 1.76, 17.6, 176, and 1,760 μg/L) and two control treatments [water and solvent (500 ng/L acetone)], with three replicates per treatment. The targeted nominal concentration for the chlorothalonil stock was 1,640 μg/L, and the actual concentration was 1,760 μg/L (spiked recovery efficiencies, 95%). All of the other concentrations were attained through serial dilutions of this stock solution. Again, for simplicity and consistency across the experiments, we refer to the nominal concentrations. We quantified frog survival and preserved dead tadpoles 12 hr after the start of the experiment and then every 24 hr for 4 days (there were no water changes). Surviving tadpoles were euthanized and preserved (70% ethanol) at the end of the experiment.
Laboratory experiment II. The same protocols used in laboratory experiment I were used in this experiment, conducted in October 2008, with the following exceptions. We tested three tadpole species: R. sphenocephala, O. septentrionalis, and H. cinerea (all starting at Gosner stage 25). We employed a completely randomized design with 144 500-mL mason jars, each filled with 300 mL artificial spring water and each receiving three tadpoles of a single species. Species were isolated in this experiment because O. septentrionalis was occasionally observed depredating H. squirella in laboratory experiment I. The jars received one of six chlorothalonil treatments (0.0164, 0.164, 1.64, 16.4, 82.0, or 164 μg/L) or water or solvent. We used the same stock solution as in laboratory experiment I. A single water change occurred on day 7, and each jar was redosed at that time. There were six replicates per species per treatment. The number of surviving tadpoles was noted after 4 hr, 24 hr, and then every 24 hr, for 10 days, and all dead tadpoles were removed and preserved in formalin at those times.
To quantify the effects of chlorothalonil on tadpole livers and immune cells, at the end of the experiment one arbitrarily selected O. septentrionalis
from each replicate was euthanized, embedded in paraffin, sectioned, and stained with hematoxylin and eosin. We used O. septentrionalis
for liver, immune, and corticosterone quantification because it had the lowest mortality of the three species and thus offered us the most survivors per tissue. To test whether chlorothalonil exposure affected liver tissue integrity, we used ImageJ64 software (Rasband 2010
) to calculate liver tissue density, following ImageJ’s Quantifying Stained Liver Tissue
(Burger and Burge 2009
), which reports the density of stained tissue within a designated area. To test whether chlorothalonil exposure affected density of liver immune cells, we counted the number of melanomacrophages and granulocytes per field of view at 400× magnification. Melanomacrophages and granulocytes are leukocytes that help defend against a variety of parasites (Rohr et al. 2008b
). Because of the morphological similarity among granulocytes, we conservatively categorized all granule-
containing immune cells as granulocytes, but most were likely eosinophils.
We used O. septentrionalis
tadpoles (Gosner stages 25–28; the same population as used in laboratory experiment II) to quantify the effect of chlorothalonil exposure on frog corticosterone levels, a steroid hormone elevated in response to natural and anthropogenic stressors, including pesticides (Martin et al. 2010
). We used the same general protocols as described in laboratory experiment II and the following treatments: 0.164, 16.4, 82, and 164 μg/L chlorothalonil, and water and solvent controls. These treatments were crossed with one of three chlorothalonil exposure durations: 4, 28, or 100 hr (n
= 3, 2, and 3, respectively). The exception, however, was that tadpoles exposed to 164 μg/L chlorothalonil were only exposed for 4 hr because they did not survive for 28 or 100 hr of exposure. This design resulted in 43 independent replicates. After the appropriate exposure duration, tadpoles were euthanized, and individual tadpoles were weighed (to 0.0001 g) and homogenized in ultrapure water. Tritiated corticosterone (2,000 cpm) was then added to each sample to quantify recoveries postextraction. We used a corticosterone enzyme immunoassay (EIA) kit (catalog no. 900-097; Assay Designs, Ann Arbor, MI) to quantify hormone levels in each sample. Individual recoveries (mean, 55.3%) and tadpole mass measurements were used to estimate corticosterone per gram of tadpole tissue. Detailed methods for this EIA kit and a discussion of its potential limitations are provided in Supplemental Material (doi:10.1289/ehp.1002956).
Statistical analyses. For all experiments and responses, we compared the water and solvent controls. Because we found no difference between these treatments (p > 0.328), we pooled the two treatments into one “control” group for all subsequent analyses.
For the mesocosm experiment, all analyses were conducted on the arcsine-square-root–transformed proportion of R. sphenocephala and O. septentrionalis surviving to the end of the experiment, controlling for the four spatial blocks. We tested whether chlorothalonil was associated with mortality relative to the control treatments by conducting a permutation-based multivariate regression analysis. For the laboratory experiment, we arcsine-square-root transformed the proportion of tadpoles surviving until the end of the experiment and log transformed hours to death, mass of survivors, amount of liver damage, and melanomacrophage and granulocyte counts to meet parametric assumptions. For the liver and immune analyses, we log transformed chlorothalonil concentration and weighted the time to death analyses by the number of animals that died per replicate. If a dose response appeared linear, chlorothalonil concentration was treated as a continuous predictor in a regression model (liver density). If a dose response was nonlinear but relatively simple (one inflection point), chlorothalonil concentration was treated as a continuous predictor, and we used polynomial regression with type II sums of squares to fit the data (immune responses). If a response was nonlinear and relatively complex (more than one apparent inflection point), chlorothalonil concentrations were treated as levels of a categorical predictor followed by Fisher’s least significant difference (LSD) multiple comparison test to determine which levels were different from one another (proportion of tadpoles that survived and time to death). As an additional test for nonmonotonicity (hump-shaped dose response), we eliminated the highest concentrations, which typically caused considerable mortality, and used polynomial regression to test for a quadratic dose–response relationship with the remaining concentrations. For the immune responses, we conducted a multivariate polynomial regression model with melanomacrophages and granulocytes as responses and followed it by univariate analyses on each response variable. We log-log transformed these relationships to improve fit and meet the assumption of the polynomial regression.
For the corticosterone experiment, we conducted polynomial regression (using least trimmed squares) with chlorothalonil concentration as a continuous predictor and log-
transformed corticosterone as the response variable. All statistical analyses were conducted with Statistica (version 8.0; Statsoft, Tulsa, OK). We did not calculate LC50 (concentration that results in death of 50% of individuals by a given time) values for any responses because all three dose–response experiments showed evidence of nonmonotonicity, which would violate the assumptions of LC50 calculations.