Arch Environ Contam Toxicol (2011) 60:281–289 DOI 10.1007/s00244-010-9623-1
Acute Toxicity and Effects Analysis of Endosulfan Sulfate to Freshwater Fish Species John F. Carriger • Tham C. Hoang • Gary M. Rand • Piero R. Gardinali • Joffre Castro
Received: 25 June 2010 / Accepted: 25 October 2010 / Published online: 3 December 2010 Ó Springer Science+Business Media, LLC 2010
Abstract Endosulfan sulfate is a persistent environmental metabolite of endosulfan, an organochlorine insecticide–acaricide presently registered by the United States Environmental Protection Agency. There is, however, limited acute fish toxicity data for endosulfan sulfate. This study determines the acute toxicity (LC50s and LC10s) of endosulfan sulfate to three inland Florida native fish species (mosquitofish [Gambusia affinis]; least killifish [Heterandria formosa]; and sailfin mollies [Poecilia latipinna]) as well as fathead minnows (Pimephales promelas). Ninety-six-h acute toxicity tests were conducted with each fish species under flow-through conditions. For all of the above-mentioned fish species, 96-h LC50 estimates ranged from 2.1 to 3.5 lg/L endosulfan sulfate. The 96-h LC10 estimates ranged from 0.8 to 2.1 lg/L endosulfan sulfate. Of all of the fish tested, the least killifish appeared to be the most sensitive to endosulfan sulfate exposure. The abovementioned data were combined with previous acute toxicity data for endosulfan sulfate and freshwater fish for an effects analysis. The effects analysis estimated hazardous concentrations expected to exceed 5, 10, and 50% of the
J. F. Carriger T. C. Hoang G. M. Rand (&) Department of Earth and Environment, Ecotoxicology and Risk Assessment Laboratory, Southeast Environmental Research Center, Florida International University, Biscayne Bay Campus, 3000 NE 151st Street, North Miami, FL 33181, USA e-mail:
[email protected] P. R. Gardinali Department of Chemistry & Biochemistry, Southeast Environmental Research Center, Florida International University, Biscayne Bay Campus, 3000 NE 151st Street, North Miami, FL 33181, USA J. Castro Everglades National Park, Homestead, FL, USA
fish species’ acute LC50 or LC10 values (HC5, HC10, and HC50). The endosulfan sulfate freshwater-fish acute tests were also compared with the available freshwater-fish acute toxicity data for technical endosulfan. Technical endosulfan is a mixture of a- and b-endosulfan. The LC50s had a wider range for technical endosulfan, and their distribution produced a lower HC10 than for endosulfan sulfate. The number of freshwater-fish LC50s for endosulfan sulfate is much smaller than the number available for technical endosulfan, reflecting priorities in examining the toxicity of the parent compounds of pesticides. The toxicity test results and effects analyses provided acute effect values for endosulfan sulfate and freshwater fish that might be applied in future screening level ecologic risk assessments. The effects analyses also discussed several deficiencies in conventional methods for setting water-quality criteria and determining ecologic effects from acute toxicity tests.
Endosulfan is a neurotoxic organochlorine insecticide– acaricide of the cyclodiene family of pesticides. It is one of the remaining organochlorine pesticides registered under the Federal Insecticide Fungicide and Rodenticide Act by the United States Environmental Protection Agency (USEPA). Endosulfan formulations are used for insect pests and mite control in agriculture, such as row crops, fruit trees, greenhouse plants, and vegetables. Technicalgrade endosulfan is a mixture of stereoisomers, designated ‘‘a’’ and ‘‘b,’’ in a 7:3 ratio, respectively (Wan et al. 2005). Several transformation products of endosulfan have been identified in the environment¯such as endosulfan diol, endosulfan lactone, endosulfan ether, endosulfan alcohol, and endosulfan a-hydroxy ether¯but endosulfan sulfate is one of the main products that has residual activity from application to application (Stewart and Cairns 1974;
123
282
Martens 1976; Van Dyk and Van der Linde 1976; Martens 1977; Miles and Moy 1979; Rao and Murty 1980; Cotham and Bidleman 1989; Awasthi et al. 2000). Among these products, endosulfan sulfate is also the most frequently detected in sediment, soil, and water (Leonard et al. 2000). In South Florida, endosulfan is one of the major insecticides used in vegetable production. According to the South Florida Water Management District, endosulfan isomers and endosulfan sulfate were detected in surface waters and benthic sediments at several locations in the south Miami-Dade County farming areas at concentrations exceeding the chronic surface water-quality standard of 0.056 lg/L (Pfeuffer and Matson 1998 to 2007). A recent probabilistic aquatic ecologic risk assessment in south Florida freshwater systems indicates that based on 15 years of surface water data, there are potential risks of total endosulfan concentrations (i.e., a ? b ? endosulfan sulfate) in certain locations (Rand et al. 2010). The acute toxicity of endosulfan (technical) to aquatic organisms has been studied (Wan et al. 2005; Berrill et al. 1998; Sunderam et al. 1992; Nebeker et al. 1983; Haider and Imbaray 1986; USEPA 2002b, 2007). Information on the toxicity of endosulfan sulfate is limited, but the available data do indicate that the toxicity of endosulfan and endosulfan sulfate are similar (German Federal Environment Agency 2007; USEPA 2002b, 2007; Wan et al. 2005; You et al. 2004). For example, the 96-h LC50 values for bluegill sunfish were 1.4 lg/L for endosulfan sulfate and 1.8–3.3 lg/ L for technical endosulfan (Wan et al. 2005; German Federal Environment Agency 2007; Nebeker et al. 1983). This study, therefore, examined the acute toxicity (96-h LC50) of endosulfan sulfate to three native fish species in south Florida (mosquitofish [Gambusia affinis]; least killifish [Heterandria Formosa]; and sailfin mollies [Poecilia latipinna]) as well as fathead minnows (Pimephales promelas), a standard freshwater fish toxicity test species. Small freshwater demersal fish were selected for toxicity testing due to the higher sensitivity of this group to mortality from endosulfan sulfate acute exposure (Carriger and Rand 2008) as well as their importance in the Everglades food web. Endosulfan sulfate freshwater fish LC50 data were fit to cumulative distribution functions (CDFs) for estimating concentration-effect benchmarks. The CDFs were created for the 96-h fish LC50s generated by our laboratory and with freshwater fish LC50s from previous studies by the USEPA (2007) and Wan et al. (2005). The species sensitivity distribution (SSD) approach was chosen to construct distributions of LC50s. From the SSDs, lower centile effect concentrations (e.g., the 5th and 10th centile hazardous concentrations or HC5 and HC10 values) were estimated. When toxicity data from laboratory studies are properly fit to distributions using the SSD approach, one can predict what proportion of species toxicity data
123
Arch Environ Contam Toxicol (2011) 60:281–289
(e.g., HC10 or 10th centile) is likely to be exceeded by any environmental concentration(s). Likewise, in an inverse fashion, one can estimate concentrations for effect levels. For example, if we were to determine the 10th centile concentration from a regression (or HC10), we could state that 10% of the species in a certain trophic group, such as fish, are likely to have toxicity values below this concentration. Lower centiles on a SSD are generally used as effects benchmarks for ecologic risk assessment under the assumption that protecting a high proportion of the species toxicity values is a conservative approach for protecting ecosystem function (Carriger and Rand 2008).
Materials and Methods Organisms The four fish species used in this study were fathead minnow (P. promelas), mosquitofish (G. affinis), least killifish (H. formosa), and sailfin mollies (Poecilia latipinna). Fish were obtained from commercial suppliers. Fish were acclimated in carbon-filtered, ultraviolet (UV)-sterilized city water, under standard laboratory conditions (25 ± 1°C, 16:8-h light-to-dark photoperiod) and fed daily a commercial brine shrimp flake food during this period. Toxicity Tests All 96-h acute toxicity tests were based on USEPA (2002a) standard methods. The four toxicity tests were conducted in 18-L glass chambers. Tests were conducted under flowthrough conditions using a modified proportional flowthrough dilutor, as originally described in Mount and Brungs (1967). Test water was carbon filtered and UV-sterilized city water. Water volume flowed through test chambers at a rate of five volume replacements per day. Temperature and photoperiod were kept at 25°C ± 1°C and 16:8 h light to dark, respectively. Five treatment concentrations of endosulfan sulfate (nominal 0.625, 1.25, 2.5, 5, and 10 lg/L), with two replicates each, and solvent (dimethylformamide) controls were used for each test. Ten fish were used for each replicate. Fish were fed 48 h before test initiation and were not fed during the tests. Mortality measurements at 24, 48, 72, and 96 h were used to estimate lethal effect concentrations expected to produce 50 and 10% fish mortality (LC50s and LC10s). Chemical Analyses During the tests, water temperature, dissolved oxygen concentration (DO), and pH were measured daily for each replicate with a YSI 55 m (YSI, Yellow Springs, OH) and an
Arch Environ Contam Toxicol (2011) 60:281–289
Accumet meter AP63 (Fisher Scientific, Pittsburgh, PA), respectively. Total hardness and alkalinity were measured at test initiation and termination (96 h) by titrating with 0.01 M ethylene diamine tetra acetate and 0.02 H2SO4 solutions, respectively. Endosulfan sulfate concentrations were measured in water samples taken from each replicate at test initiation and termination by gas chromatography with electron capture detection (GC/ECD) (Agilent Technologies, Palo Alto, CA) using modifications of EPA method 8081. Briefly, water samples (50–250 mL) were extracted by liquid–liquid extraction against pesticide-grade-methylene chloride (3 9 10 mL). The combined organic extract was dried over combusted sodium sulfate and, after evaporation and solvent exchange, analyzed by GC/ECD. The extracts were analyzed only for endosulfan sulfate according to the method developed by Sericano et al. (1998) using a Hewlett Packard 5980II gas chromatograph (HP 5880II GC) (Agilent Technologies, Palo Alto, CA) with electron capture detector using a 30-m, 0.25-mm i.d., 25-lm DB-5 fused silica capillary column from Agilent Technologies. The HP 5880II gas chromatograph was operated in splitless mode using ultra high purity (UHP)-grade helium as a carrier gas (25 psi) and UHP-grade nitrogen for make-up gas (40 psi). PCB103 was used a surrogate standard and added to the water samples before extractions. Recovery of PCB103 (77–105%) was assessed by using a GC standard (TCMX) added to sample extracts just before the GC analysis. Batch quality assurance and quality control included analysis of blanks, fortified blanks, and duplicates. A fivepoint calibration curve was used for quantitative determination. The method detection limit for endosulfan sulfate was statistically determined (n = 7) to be 0.005 lg/L. Recovery of endosulfan sulfate from fortified samples ranged from 76 to 115%. Toxicity Test Data Analysis LC50s were determined for each 24-h interval. All calculations were based on measured concentrations of endosulfan sulfate. The concentration–response data were fit with three models (log-normal, log-logistic, and Gompertz) using maximum likelihood methods described in Newman (1995). The maximum likelihood fitting of the models was conducted using PROC PROBIT in SAS software (version 9.2) of the SAS system for Windows 2002–2008 (SAS Cary, NC). Concentration data were log-transformed before fitting the models. The model that best fit the data for each time interval and test was chosen. The fit of each of the models was compared using the results of v2 tests (the Pearson goodness-of-fit v2). Models were rejected if significant heterogeneity was found between observed and expected values (a = 0.05). The slopes and slope SEs of
283
the best-fit maximum likelihood models are reported for the 24-h concentration–response curves. When maximum likelihood methods failed to fit the data, an LC50 was estimated using trimmed SpearmanKarber methods, and a 96-h LC10 was estimated using nonlinear interpolation with bootstrapping. For the latter, observations were resampled 200 times to obtain the nonlinear interpolated LC10 and fiducial limits. Bootstrapped fiducial limits were expanded with a multiplier based on the low number of replicates according to Norberg-King (1993). Before applying the nonlinear interpolation, concentration data were log-transformed and response data were logit-transformed. Trimmed Spearman-Karber and nonlinear interpolation calculations were performed in ToxCalc version 5.0.23 (Tidepool Scientific Software 2004). Several mortalities occurred in the fathead-minnow control treatments after 24 h. Therefore, Abbott’s formula for LC50 estimates or Abbott’s corrected isotonized mean responses for LC10 estimates was used to adjust the mortalities in exposed treatments based on the amount of control mortalities. Control mortalities for fathead minnows never exceeded 10% in any of the four replicate fathead-minnow control tanks. SSDs A Microsoft Excel-based program (SSD Master v2.0; Intrinsik Environmental Sciences) was used for creating SSDs and deriving associated statistics (Rodney and Moore 2008). The SSDs were created for the fish species data generated by our laboratory and for the same data with additional endosulfan sulfate fish LC50s from other laboratories. Using SSD Master, log-normal, log-logistic, Gompertz, Weibull, and Fisher-Tippett sigmoidal distributions were fit to the acute toxicity data (i.e., LC50s, LC10s). Plotting positions were derived using Hazen’s method (pi = (i - 0.5)/n). Distribution parameters and error estimates were generated using maximum likelihood methods. Anderson–Darling statistics (A2) were produced in SSD Master (Rodney and Moore 2008) and used to compare relative fits, with lower values of the statistic indicating a better fit, and for rejecting distributions that might not fit the data well (a = 0.10; critical value = 1.933). As indicated in Rodney and Moore (2008), the recommended minimum n value for Anderson–Darling is five plotting points. However, only LC50 data for four fish species were available from our laboratory and only LC50 data for two additional fish were available from the literature, so this requirement was not met when only data from the current study were considered. Therefore, the mean square error (MSE) was used as a comparison tool for distributions with fewer than five plotting points. From the SSDs, we
123
284
Arch Environ Contam Toxicol (2011) 60:281–289
estimated the hazardous endosulfan sulfate water concentrations (HCx) expected to exceed certain percentages (x = 5%, 10%, or 50%) of the freshwater fish acute-toxicity values.
Results For all of the tests, hardness and alkalinity ranged from 56–69 and 63–69 mg/L as CaCO3, respectively. Dissolved oxygen concentrations ranged from 6.3 to 7.7 mg/L. Temperature and pH ranged from 24 to 25°C and 7.5–7.9, respectively. The measured concentration averages for each treatment (and ranges) are listed in Table 1. The 24- to 96-h LC50s and LC10s from the acute toxicity of endosulfan sulfate to the four fish species are listed in Table 2. The 24-h LC50s ranged from 4.2 lg/L (H. formosa) to 8.4 lg/L (P. latipinna), and the 96-h LC50s ranged from 2.1 lg/L (H. formosa) to 3.5 lg/L (P. latipinna). An LC10 was not calculated for fathead minnows beyond 24 h due to poor fit of the maximum likelihood models to the data. However, the 24-h LC10 using a log-normal model was 2.7 lg/L. A 96-h LC10 calculated using the nonlinear interpolation method with bootstrapping was 0.8 lg/L. Table 1 Summary of the measured concentrations in treatments for the acute toxicity tests with endosulfan sulfate and freshwater fisha Species
Least killifish
Mosquitofish
Sailfin molly
Fathead minnow
a
Nominal concentration (lg/L)
Measured concentration (range) (lg/L)
0.625
0.394 (0.383–0.404)
1.25
0.929 (0.823–1.021)
2.5
1.767 (1.483–1.940)
5
3.254 (3.109–3.464)
10
6.843 (6.675–7.092)
0.625
0.466 (0.447–0.477)
1.25
0.937 (0.884–1.036)
2.5
1.819 (1.544–2.012)
5
3.510 (3.322–3.637)
10
9.195 (8.209–10.099)
0.625
0.514 (0.472–0.569)
1.25
1.312 (1.212–1.527)
2.5
2.720 (2.548–2.842)
5 10
5.762 (5.039–6.401) 10.168 (9.317–11.389)
0.625
Discussion
0.408 (0.383–0.439)
1.25
0.900 (0.792–1.076)
2.5
1.625 (1.493–1.707)
5
3.586 (3.180–4.033)
10
9.362 (7.741–10.668)
Concentrations were taken from both replicates in each treatment on test initiation (day 0) and at test termination (day 4)
123
Data for fathead minnows at 96 h did not increase monotonically, so the interpolation method applied a smoothing to the response means. However, the 95% fiducial limits were wide for this estimate, ranging from 0.2 to 6.5 lg/L. For purposes of creating the SSDs, the 96-h LC10 of 0.8 lg/L was used. The statistics from the best-fit SSD CDF models are shown in Table 3. Figure 1(a–c) shows the data points for log-transformed concentration variables with the fit of each of the candidate distributions. The 5th, 10th, and 50th centile hazard concentration estimates (i.e., HC5, HC10, and HC50 values) for the LC50s from the four freshwater species tested were 1.65, 1.83, and 2.67 lg/L, respectively (Table 3). The Anderson–Darling fit statistics could not be used for data strictly from the current study due to the low number of plotting points (n = 4 fish species). The MSE values were used for a rough comparison and indicate that the Weibull distribution displayed in Fig. 1a fits the logtransformed concentration data the best (MSE = 0.0077). When the candidate SSD models were fit to the LC10 estimates, the Gompertz distribution fit the data the best (MSE = 0.0008) (Table 3). The Weibull distribution had the second best fit (MSE = 0.0010). In Fig. 1b, the Gompertz sigmoidal curve follows the data points better than any of the other distributions. The Gompertz distribution also had the lowest range between the upper and lower fiducial intervals for all of the hazardous concentration estimates. The log-normal and log-logistic models had similar MSE values, and the Fisher-Tippet model had the poorest fit. The 5, 10, and 50th centile hazard concentration estimates for the LC10 estimates for freshwater fish species were 0.64, 0.80, and 1.47 lg/L, respectively (Table 3). Table 3 also lists the statistics from the best-fit SSD created for all available freshwater fish acute toxicity LC50 data. The 5, 10, and 50th centile hazard concentration estimates were 1.10, 1.41, and 2.68 lg/L for the candidate distributions (Table 3). The tighter fiducial intervals for the HC50 values in all of the distributions is expected due to its central location on the SSDs. The Anderson–Darling statistics did not indicate that the data did not come from each of the specified distributions in Fig. 1c (the null hypothesis that the data came from the specified distribution was not rejected for each distribution at a = 0.10).
Comparisons of Aquatic Toxicity Between Endosulfan Sulfate and Technical Endosulfan The LC50s for endosulfan sulfate in this study were similar from species to species and were approximately two times higher than the LC50 value reported by Wan et al. (2005)
Arch Environ Contam Toxicol (2011) 60:281–289
285
Table 2 Results from the acute toxicity tests with measured endosulfan sulfate concentrations and freshwater fish speciesa Species
Exposure time (h)
LC50 (95% FL) (lg/L)
LC10 (95% FL) (lg/L)
Model/method used
P. promelas
24
4.575 (3.696–5.778)
2.681 (1.731–3.371)
Log-normal
48 72 G. affinis
H. formosa
P. latipinna
4.096 (3.382–4.960)
–
3.818 (3.107–4.692)
–
Slope (SE) 5.5209 (1.182)
Pearson goodness of fit v2 0.1977
TSK
–
–
TSK
–
–
–
–
96
3.047 (2.385–3.893)
0.811 (0.225–6.506)
TSK/NonlinInt
24
6.033 (4.857–7.387)
3.960 (2.676–4.909)
Log-normal
7.0085 (1.439)
0.0027
48
3.647 (3.004–5.550)
1.988 (0.833– 2.517)
Gompertz
3.1054 (1.021)
0.3872
72 96
2.617 (2.238–3.061) 2.273 (1.963–2.661)
1.821 (1.314–2.146) 1.623 (1.200–1.892)
Log-normal Log-normal
8.1364 (1.775) 8.7514 (1.953)
0.0030 0.0080
24
4.156 (3.439–5.107)
2.389 (1.589–2.963)
Log-normal
5.3321 (1.085)
1.0245
48
2.566 (2.343–2.810)
72
2.183 (1.856–2.557)
1.479 (1.031–1.760)
Gompertz
4.8344 (1.073)
3.5213
96
2.058 (1.747–2.443)
1.339 (0.894–1.608)
Gompertz
4.3832 (1.025)
1.1729
24
8.429 (6.828–11.786)
4.244 (2.210–5.465)
Log-normal
4.3015 (1.154)
0.8896
48
5.035 (4.023–5.849)
3.473 (1.956–4.254)
Log-logistic
13.6241 (3.714)
2.4528
72
4.257 (3.457–5.144)
2.631 (1.705–3.278)
Log-logistic
10.5103 (2.199)
1.5562
96
3.506 (2.874–4.304)
2.104 (1.385–2.614)
Log-logistic
9.9086 (2.079)
4.1167
TSK
FL fiducial limit, TSK trimmed Spearman-Karber method for calculating the LC50, NonlinInt nonlinear interpolation method with bootstrapping for determining the LC10 a
Results were calculated using measured concentrations
Table 3 SSD statistics for acute log-transformed LC50s of endosulfan sulfate and four freshwater fish species tested for the current study, log-transformed LC10s of endosulfan sulfate from tests in the current
study, and LC50s from the current study plus two additional fish species from the literature (all data)
Distribution
SSD for LC50 values from the current study (Weibull)
SSD for LC10s from the current study (Gompertz)
SSD for LC50 values for all available data (Gompertz)
HC5 (lg/L) (95% FI)
1.65 (1.01–2.71)
0.64 (0.43–0.95)
1.10 (0.84–1.45)
HC10 (lg/L) (95% FI)
1.83 (1.16–2.89)
0.80 (0.59–1.09)
1.41 (1.14–1.74)
HC50 (lg/L) (95% FI)
2.67 (2.03–3.50)
1.47 (1.25–1.74)
2.68 (2.41–2.98)
Scale
0.4677
0.1399
0.1483
3.8986
0.2198
0.4825
0.0154
0.0015
0.0131
Mean square error (MSE)
0.0077
0.0008
0.0033
Anderson–Darling statistic (A2)
–
–
0.192
Location/shape Sum of squared error terms (
P 2 ei )
FI Fiducial interval
for Oncorhychus mykiss, a coldwater fish. However, the LC50 values for Daphnia magna (48 h) and Hyalella azteca (96 h) were approximately 1000–2000 times and 2 times higher than the LC50 values reported in this study, respectively (Wan et al. 2005). Knauf and Schulze (1973) also found that the LC50 values of endosulfan sulfate ranged from 1 to 10 lg/L for Poecilia reticulata. The differences in sensitivity of vertebrate and invertebrate species to endosulfan sulfate may be due to differences in mechanisms of uptake, accumulation, and mode of action. With technical-grade endosulfan, there was also a difference in
sensitivity of vertebrate and invertebrate species (Nebeker et al. 1983; Sunderam et al. 1992; Wan et al. 2005; Hose et al. 2003). The acute LC50 estimate for endosulfan sulfate and G. affinis in this study was similar to the toxicity of technical-grade endosulfan to G. affinis found by Sunderam et al. (1992). The 96-h LC50 values in this study and in the study of Sunderam et al. (1992) were 2.3 and 2.8 lg/L, respectively. Wan et al. (2005) also found that the toxicity of endosulfan sulfate and technical-grade endosulfan to O. mykiss and H. azteca was similar. However, the 96-h LC50 values for
123
286
Arch Environ Contam Toxicol (2011) 60:281–289
(b) 1.0 Proportion of LC10s exceeded
Proportion of LC50s exceeded
(a) 1.0 0.8
0.6
0.4 Normal Logistic Gompertz Weibull Fisher-Tippett
0.2
0.0 0.1
1
10
0.8
0.6
0.4 Normal Logistic Gompertz Weibull Fisher-Tippett
0.2
0.0 0.1
100
1
10
100
Concentration ( µg/L)
Concentration ( µg/L)
Proportion of LC50s exceeded
(c) 1.0 0.8
0.6
0.4 Normal Logistic Gompertz Weibull Fisher-Tippett
0.2
0.0 0.1
1
10
100
Concentration ( µg/L)
Fig. 1 SSDs for a acute log-transformed LC50 values of endosulfan sulfate and four freshwater fish species tested for the current study, b acute log-transformed LC10s of endosulfan sulfate and four freshwater fish species tested for the current study, and c acute log-
transformed LC50s of endosulfan sulfate and four freshwater fish species tested for the current study plus two additional fish species from the literature
technical-grade endosulfan for P. promelas reported by Nebeker et al. (1983) (1–1.7 lg/L) and by Lemke (1980) (0.29–1.67 lg/L) were less than the 96-h LC50 value (3.0 lg/L) found in this study for endosulfan sulfate. From the SSDs, for endosulfan sulfate LC10s and LC50s, the HCx estimates were approximately two times lower for the distributions of LC10s versus the distributions of LC50s when data were considered from the current studies. The closeness of the LC50 and LC10 concentrations are indicative of the steep slope of the concentration-effect curves that can arise when species have a specific receptor site for a chemical (Aquatic Effects Dialogue Group 1994; ECOFRAM 1999). The steep slopes of the SSDs further indicate the tight range of acute sensitivity that fish species have for endosulfan sulfate. This was also observed with acute toxicity SSDs developed for freshwater and saltwater fish and technical endosulfan by Carriger and Rand (2008). Overall, the data from past and present studies indicate that endosulfan sulfate does have a similar toxicity for freshwater fish compared with the parent compound (technical endosulfan). The HC5 value for all fish species
and technical endosulfan was 0.31 lg/L in Hose and van den Brink (2004). This value is lower than the HC5 values in the current study for endosulfan sulfate (the lowest HC5 value was 1.10 lg/L). The HC10 for the acute freshwater fish technical endosulfan SSD from Carriger and Rand (2008) was 0.38 lg/L. This value was also lower than the HC10s calculated in the current study (1.41 lg/L). The HC50 for all fish was found to be 2.64 lg/L by Hose and van den Brink (2004). This value was approximately equivalent to the HC50 values for endosulfan sulfate in the current study (highest HC50 value was 2.68 lg/L). The HC50 from endosulfan sulfate LC10 values was approximately 1 lg/L lower than the HC50 for LC50s and technical endosulfan in Hose and van den Brink (2004). Interestingly, the HC5 from Hose and van den Brink (2004) and the HC10 from Carriger and Rand (2008) from LC50 SSDs and technical endosulfan were approximately twice as low as the same estimates for LC10 SSDs and endosulfan sulfate. However, the SSDs for technical endosulfan were based on a larger database of acute toxicity values and were therefore generally considered more reliable.
123
Arch Environ Contam Toxicol (2011) 60:281–289
Fig. 2 Log-normally cumulatively distributed species sensitivity distributions for fish and technical endosulfan (black squares) and endosulfan sulfate (gray circles)
Figure 2 shows a comparison of the SSDs for technical endosulfan and endosulfan sulfate using acute toxicity values for fish. The two SSDs were constructed using leastsquared log-normal regressions. From Fig. 2, it is apparent that the HC10s (tenth centile intercepts) are approximately one order of magnitude different and that the toxicity data for technical endosulfan have greater variability than for endosulfan sulfate. The LC50s for endosulfan sulfate were within the range of LC50s for technical endosulfan, and the HC50 values were similar. However, the range of LC50s is much greater for technical endosulfan, and thus the resulting slope of the SSD is less steep. As discussed in de Zwart (2002), SSDs for chemicals with similar modes of action generally have similar slopes if the n value is sufficiently high and the species found in each distribution are similar. The dissimilarity in slopes between technical endosulfan and endosulfan sulfate could be due to the lower number of species tested for endosulfan sulfate compared with the parent compound (de Zwart 2002). Other factors that could produce different slopes for SSDs for chemicals with similar modes of actions are the inclusion of studies with different methods, varying test conditions, and the wider range of fish species LC50 data for technical endosulfan. When the SSD for technical endosulfan and freshwater and saltwater arthropods were compared with the SSDs for freshwater and saltwater fish in Carriger and Rand (2008), the slope was found to be much steeper for fish than arthropods in both cases. Importance of Lethal Dose End Points Regulatory thresholds are often applied for effects characterization based on the LC50 for compounds. For example, the USEPA (2002c) advocates multiplying acute LC50s by 0.5 to establish a level of concern for nonendangered
287
species potentially exposed to pesticides. However, the LC50 is not fully representative of the entire concentration– response curve, and a lower effects benchmark (e.g., LC5 or LC10) might be a more appropriate safety factor for a level of concern due to the slope of the toxicity curve (Aquatic Effects Dialogue Group 1994). The Aquatic Effects Dialogue Group (1994) recommended using LC5s from acute toxicity data. After analyzing approximately 200 data sets, Moore and Caux (1997) claimed that estimates lower than the LC10, such as the LC5, were generally too dependent on model selection and too unreliable based on the large fiducial intervals. Therefore, the LC10 might be a better general choice than the LC5 for a lower-level effects benchmark. For analyzing toxicity data, extrapolation from concentration effect curves has been advocated for deriving protective concentrations in lieu of no-effect concentrations (NOECs) estimated using hypothesis-testing techniques (Landis and Yu 2004). Concentrations estimated from testing the null hypothesis of an NOEC are driven by experimental design issues, including the typical lack of a power analysis or poorly considered b value, as well as issues with replication (Newman 2008). A previous study observed that the majority of hypothesis-tested NOECs estimated in a data set of toxicity tests were higher than the 10% effect concentration for both quantal and continuous data sets (Moore and Caux 1997), despite the customary 0.05 a value. In one data set, Crane and Grosso (2002) noted an extreme case where the hypothesis-tested NOEC was equivalent to the lower fiducial limit of the LC50. The use of LC10s is therefore a generally conservative approach in lieu of deriving hypothesis-tested NOECs from standard acute-toxicity test experimental designs (e.g., two replicates and five concentration treatments), such as the ones in the current studies. One drawback of relying on probit-type analyses for obtaining an LC10 are the wide fiducial intervals often found from estimates at lower ends of a concentrationeffect curve (Landis and Yu 2004). This drawback prompted us to represent sensitive acute effects at 24-h durations with an LC10 instead of an LC5. In this study, the LC10 fiducial intervals were reasonable at 96 h for all species, except fathead minnows. However, in addition to statistical issues with estimating an NOEC from regression or hypothesis-testing work, the ecologic value or relevance to a true NOEC level for a receiving aquatic system is something in need of more consideration. Another drawback is the lack of understanding of what an LCx or ECx value means in terms of a true NOEC. Fox (2008, 2009) discussed some important issues in applying effective concentrations (i.e., LC/ECx) from experiments as proxies for NOECs. These issues include how to interpret an SSD HCx derived from ECx data and the lack of a
123
288
correspondence between what an ECx is and what an NOEC should be. Better dose–response characterizations and the development of integrated models was argued to be a way forward in estimating safety values for chemicals by Fox (2009). Building integrated risk models through end points with greater ecologic relevance would be one way of attempting to answer what an acute or chronic NOEC might be. Uncertainties in LCx and HCx End Points for WaterQuality Criteria Applications The acute SSD approach can consider a variety of types of experimental conditions and species in its implementations and thus is a useful supplementary tool for screening-level ecologic risk assessments that examine exceedences of LC50s. However, the input data used and typically available for the SSD approach does not address ecologic functions, time-varying exposures, species interactions, and loss of services from risks to aquatic communities, among other issues. The application of LC50s to hazard and risk assessment is common due to precedence and the wide availability of toxicity data that use these measures. Timeto-event toxicity end points, as developed in the works of Michael C. Newman (e.g., Newman 1995 et al.), are more appropriate than ECx end points for building models that can assess the acute toxicity and ecologic risks of chemicals. A better understanding of the risks of endosulfan sulfate exposures to fish populations as a function of time and concentration, such as that described in Unger et al. (2007), to a subset of these test species has been described in an additional article (Carriger et al. 2010). Acknowledgments This study was funded by the Everglades National Park through cooperative agreement number H5297-040133. We thank Dwayne Moore of Intrinsik Environmental Sciences, Inc. for providing us with a copy of SSD Master. This is SERC contribution number 499.
References Aquatic Effects Dialogue Group (1994) Aquatic effects dialogue group: pesticide risk assessment and mitigation. Society of Environmental Toxicology and Chemistry and SETAC Foundation for Environmental Education, Pensacola, FL Awasthi N, Ahuja R, Kumar A (2000) Factors influencing the degradation of soil-applied endosulfan isomers. Soil Biol Biochem 32:1697–1705 Berrill M, Coulson D, McGillivray Lise, Pauli B (1998) Toxicity of endosulfan to aquatic stages of anuran amphibians. Environ Toxicol Chem 17:1738–1744 Carriger JF, Rand GM (2008) Aquatic risk assessment of pesticides in surface waters in and adjacent to the Everglades and Biscayne National Parks: II. Probabilistic analyses. Ecotoxicology 17: 680–696 Carriger JF, Hoang TC, Rand GM (2010) Survival analysis of mosquitofish (Gambusia affinis) and least killifish (Heterandria
123
Arch Environ Contam Toxicol (2011) 60:281–289 formosa) exposed to endosulfan sulfate. Arch Environ Contam Toxicol 58:1015–1022 Cotham WE Jr, Bidleman TF (1989) Degradation of malathion, endosulfan, and fenvalerate in seawater and seawater/sediment microcosms. J Agric Food Chem 37:824–828 Crane M, Grosso A (2002) Time to event analysis of standard ecotoxicity data. In: Crane M, Newman MC, Chapman PF, Fenlon J (eds) Risk assessment with time to event models. Lewis, CRC Press, Boca Raton, FL De Zwart D (2002) Observed regularities in SSDs for aquatic species. In: Posthuma L, Suter GW II, Traas TP (eds) Species sensitivity distributions in ecotoxicology. Lewis, CRC Press, Boca Raton, FL ECOFRAM (1999) ECOFRAM aquatic report. Ecological committee on FIFRA risk assessment methods: report of the aquatic workgroup. United States Environmental Protection Agency, Office of Pesticide Programs, Washington, DC Fox D (2008) NECS, NOECS and the EC. Australas J Ecotoxicol 14:7–9 Fox D (2009) Is the ECX a legitimate surrogate for a NOEC. Integr Environ Assess Manage 5:351–353 German Federal Environment Agency (2007) Endosulfan: draft dossier prepared in support of a proposal of endosulfan to be considered as a candidate for inclusion in the Annexes to the Stockholm Convention. Umweltbundesamt, Dessau Haider S, Imbaray RM (1986) Relative toxicity of technical material and commercial formulation of malathion and endosulfan to a fresh water fish, Channa punctata. Ecotoxicol Environ Safe 11:347–351 Hose GC, Van den Brink PJ (2004) Confirming the species-sensitivity distribution concept for endosulfan using laboratory, mesocosm, and field data. Arch Environ Contam Toxicol 47:511–520 Hose GC, Hyne RV, Lim RP (2003) Toxicity of endosulfan to Atalophlebia spp. (Ephemeroptera) in the laboratory, mesocosm, and field. Environ Toxicol Chem 22:3062–3068 Knauf W, Schulze EF (1973) New findings on the toxicity of endosulfan and its metabolites to aquatic organisms. Meded Fac Landbouwwetensch. Rijksuniv Gent 38:717–732 Landis WG, Yu MH (2004) Introduction to environmental toxicology: impacts of chemicals upon ecological systems, 3rd edn. Lewis, CRC Press, Boca Raton, FL Lemke AE (1980) Interlaboratory comparison acute testing set. Comprehensive report. In: Ambient water quality criteria. EPA 440/5-80-046. United States Environmental Protection Agency, Washington, DC, pp B-1–B-35 Leonard AW, Hyne RV, Lim RP, Pablo F, Van den Brink PJ (2000) Riverine endosulfan concentrations in the Namoi River, Australia: link to cotton field runoff and macroinvertebrate population densities. Environ Toxicol Chem 19(6):1540–1551 Martens R (1976) Degradation of [8,9-14C] endosulfan by soil microorganisms. Appl Environ Microbiol 31(6):853–858 Martens R (1977) Degradation of endosulfan-8, 9-14C in soil under different conditions. Bull Environ Contam Toxicol 17(4): 438–446 Miles JRW, Moy P (1979) Degradation of endosulfan and its metabolites by a mixed culture of soil microorganisms. Bull Environ Contam Toxicol 23:13–19 Moore DRJ, Caux PY (1997) Estimating low toxic effects. Environ Toxicol Chem 16:794–801 Mount DI, Brungs WA (1967) A simplified dosing apparatus for fish toxicology studies. Wat Res 1:21–29 Nebeker AV, McCrady JK, Mshar R, McAuliffe CK (1983) Relative sensitivity of Daphnia magna, rainbow trout and fathead minnows to endosulfan. Environ Toxicol Chem 1:69–72 Newman MC (1995) Quantitative methods in aquatic ecotoxicology. Lewis, Boca Raton, FL
Arch Environ Contam Toxicol (2011) 60:281–289 Newman MC (2008) What exactly are you inferring?¯A closer look at hypothesis testing. Environ Toxicol Chem 27:1013–1019 Norberg-King T (1993) A linear interpolation method for sublethal toxicity: the inhibition concentration (ICp) approach (version 2.0). National Effluent Toxicity Assessment Center Technical Report 03-93, United States Environmental Protection Agency, Environmental Research Laboratory, Duluth, MN Pfeuffer R, Matson F (1998–2007) Pesticide surface water report. Quarterly sampling event. South Florida Water Management District, West Palm Beach, FL Rand GM, Carriger JF, Gardinali PR, Castro J (2010) Endosulfan and its metabolite, endosulfan sulfate, in freshwater ecosystems of south Florida: a probabilistic aquatic ecological risk assessment. Ecotoxicology 19:879–900 Rao DMR, Murty AS (1980) Persistence of endosulfan in soils. J Agric Food Chem 28:1099–1101 Rodney S, Moore D (2008) Development of an excel-based tool for fitting and evaluating species sensitivity distributions: final report. Intrinsik Science, Ottawa, Ontario Sericano JL, Gardinali PR, Wade TL (1998) Quantitative determination of chlorinated hydrocarbons. In: Lauenstein GG, Cantillo AY (eds) Sampling and analytical methods of the national status and Trends Program Mussel Watch Project: 1993–1996 Update. NOAA Technical Memorandum NOS ORCA 130. National Oceanic and Atmospheric Administration (NOAA), National Ocean Service, Silver Spring, MD, pp 160–167 Software Tidepool Scientific (2004) ToxCalc: toxicity data analysis software v5.0.26. Tideppol Scientific Software, McKinleyville, CA Stewart DKR, Cairns KG (1974) Endosulfan persistence in soil and uptake by potato tubers. J Agric Food Chem 22(6):984–986 Sunderam RIM, Cheng DMH, Thompson GB (1992) Toxicity of endosulfan to native and introduced fish in Australia. Environ Toxicol Chem 11:1469–1476 Unger MA, Newman MC, Vadas GG (2007) Predicting survival of grass shrimp (Palaemnetes pugio) during ethylnaphthalene,
289 dimethylnaphthaalene, and phenanthrene exposures differing in concentration and duration. Environ Toxicol Chem 26:528–534 United States Environmental Protection Agency (2002a) Methods for measuring the acute toxicity of effluents and receiving waters to freshwater and marine organisms. EPA-821-R-02-012. United States Environmental Protection Agency, Washington, DC United States Environmental Protection Agency (2002b) Reregistration eligibility decision for endosulfan. EPA 738-R-02-013, United States Environmental Protection Agency, Prevention, Pesticides and Toxic Substances, Washington, DC United States Environmental Protection Agency (2002c) Reregistration eligibility science chapter for atrazine: environmental fate and effects chapter. Appendix XV. Comparison of OW and OPP methods for aquatic assessments. United States Environmental Protection Agency, Office of Prevention, Pesticides and Toxic Substances, Washington, DC United States Environmental Protection Agency (2007) Appendix 1 to 2007 addendum: environmental fate and ecological risk assessment of endosulfan. United States Environmental Protection Agency, Office of Prevention, Pesticides, and Toxic Substances, Washington, DC Van Dyk LP, Van der Linde A (1976) Persistence of endosulfan in soils of the Loskop Dam irrigation area. Agrochemophysica 8:31–34 Wan MT, Kuo J, Buday C, Schroeder G, Van Aggelen G, Pasternak J (2005) Toxicity of a-, b-, (a ? b)-endosulfan and their formulated and degradation products to Daphnia magna, Hyalella azteca, Oncorhynchus mykiss, Oncorhynchuc kisutch, and biological implications in streams. Environ Toxicol Chem 24:1146–1154 You J, Schuler LJ, Lydy MJ (2004) Acute toxicity of sediment-sorbed endrin, methoxychlo and endosulfan to Hyalella azteca and Chironomus tentans. Bull Environ Contam Toxicol 73:457–464
123