From Individual to Population Level Effects of

Apr 9, 2010 - 6150 mg/kg dry weight (d.w.) of sediments, depending on the studied life-cycle trait. Actual zinc ... spiked sediments (87 mg d.w.) and ...
1 downloads 9 Views 374KB Size
Download PDF
Environ. Sci. Technol. 2010, 44, 3566–3571

From Individual to Population Level Effects of Toxicants in the Tubicifid Branchiura sowerbyi Using Threshold Effect Models in a Bayesian Framework V I R G I N I E D U C R O T , * ,† E L I S E B I L L O I R , ‡ ´ RY,§ ALEXANDRE R. R. PE JEANNE GARRIC,| AND SANDRINE CHARLES‡ INRA (Institut National de la Recherche Agronomique), ` UMR985 Ecologie et Sante´ des Ecosystemes, E´quipe E´cotoxicologie et Qualite´ des Milieux Aquatiques, Agrocampus Ouest, 65 rue de Saint Brieuc, F-35042, Rennes, France, Universite´ de Lyon, F-69000, Lyon, Universite´ Lyon 1, CNRS, UMR5558, Laboratoire de Biome´trie et Biologie E´volutive, F-69622, Villeurbanne, France, INERIS, Unite´ METO, Parc Alata - BP 2, F-60550 Verneuil-en-Halatte, France, and Laboratoire d’Ecotoxicologie, Cemagref, 3bis quai Chauveau, CP 220, F-69009 Lyon, France

Received January 4, 2010. Revised manuscript received March 23, 2010. Accepted March 25, 2010.

Effects of zinc were studied in the freshwater worm Branchiura sowerbyi using partial and full life-cycle tests. Only newborn and juveniles were sensitive to zinc, displaying effects on survival, growth, and age at first brood at environmentally relevant concentrations. Threshold effect models were proposed to assess toxic effects on individuals. They were fitted to lifecycle test data using Bayesian inference and adequately described life-history trait data in exposed organisms. The daily asymptotic growth rate of theoretical populations was then simulated with a matrix population model, based upon individual-level outputs. Population-level outputs were in accordance with existing literature for controls. Working in a Bayesian framework allowed incorporating parameter uncertainty in the simulation of the population-level response to zinc exposure, thus increasing the relevance of test results in the context of ecological risk assessment.

Introduction Endobenthic tubificids are widely used to assess toxicity of sediment-associated contaminants because they exhibit intimate contact with the substrate during their whole lifecycle, and because they are exposed through various routes such as tegument, gills, and digestive tract (1, 2). Among tubificids, Branchiura sowerbyi has been recognized as a valuable alternative to the standard species Tubifex tubifex from both the ecological and the toxicological points of view (3). Indeed, this worldwide-distributed species (4) is one of the most pollutant-sensitive freshwater oligochaete species * Corresponding author phone: +33 223 485 625; fax: +33 223 485 440; e-mail: [email protected]. † INRA. ‡ Universite´ de Lyon. § INERIS. | d’Ecotoxicologie, Cemagref. 3566

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 9, 2010

(5). Effects of toxicants on B. sowerbyi have mostly been characterized in 96-h mortality tests (see for instance refs 6 and 7). Only one study referring to chronic toxicity tests with this species has been published (8). Probably due to the scarcity of data from chronic tests, population-level responses of worms exposed to contaminated sediments have not been studied so far. However, this species is often dominant in lakes and is very productive in organic matter rich sediments (6, 9), thus representing a major food source for fish (10). Therefore, effects of toxicants on population dynamics may lead to a decline in fish stock. Assessment of population sustainability and productivity based on appropriate toxicity tests implemented on B. sowerbyi would thus provide relevant data to support fish management and the implementation of environmental policies (9). In this context, our study aimed at assessing population level effects of zinc, chosen as a representative of heavy metals (11), on B. sowerbyi. Sensitivity windows along the life-cycle of the worm were first investigated using a suite of five partial life-cycle (PLC) tests on survival, growth, and reproduction in juveniles and adults. A full life-cycle (FLC) test was also implemented to assess interactions between life-cycle traits influencing the response of exposed individuals. Data on individual performances were analyzed using threshold effect models, which included stress functions inspired from DEBtox models (12). This latter framework was chosen because it provided a generic and ecotoxicologically sound approach to link survival, growth, and reproduction to exposure concentration/duration. It also allowed encompassing some of the statistical and ecotoxicological issues of the classical effect criteria (e.g. No Observed Effect Concentration), which are questionable indices (12, 13). Point estimates and probability distributions of model parameters were estimated using Bayesian inference. Such a method has already been successfully used to estimate parameters of interest in ecotoxicology and their uncertainty simultaneously from growth and reproduction data (14). Individuallevel effect models outputs were included into a matrix model (15) based on the partial life-cycle graph of the worm (16). The population-level response was finally assessed through the daily asymptotic population growth rate depending on the exposure concentration.

Material and Methods Partial and Full Life-Cycle Tests Design. Natural fine sediments were sampled, conditioned, and spiked with zinc (chemical purity ) 99.9%) according to the rolling jars method, as detailed in previous studies (17), after sediment moisture and background zinc level has been determined. Nominal concentrations varied in the range 0 (control) 6150 mg/kg dry weight (d.w.) of sediments, depending on the studied life-cycle trait. Actual zinc concentrations were measured at the end of the toxicity tests in natural and spiked sediments, pore and overlying water, using microwave mineralization followed by inductively coupled plasma mass spectrometry (quantification limit ) 1 mg/kg). Test chambers consisted of 600 mL glass beakers filled with 200 mL of zincspiked sediments (87 mg d.w.) and 300 mL of reconstituted water, which was a mixture of a natural drilled ground water and deionized water with the following properties at the beginning of the test: pH ) 7.75, conductivity ) 300 ( 30 µS/cm, and hardness ) 100 mg/L CaCO3. Tests began three days after the preparation of the test chambers in order to allow the settlement of sediments and the re-establishment of redox gradients (18). 10.1021/es903860w

 2010 American Chemical Society

Published on Web 04/09/2010

Five independent partial life-cycle (PLC) tests and a full life-cycle (FLC) test were implemented (see Table S1, Supporting Information, for a detailed comparison of test designs). Survival and growth were monitored in 28 days PLC tests with juveniles (mean age of individuals: 14 ( 7 days), subadults (40 ( 7 day-old worms), and adults (60 ( 7 day-old worms). Fecundity and hatching rate were monitored in two independent 42 days PLC tests implemented with adults of homogenous fresh weight (91 ( 12 mg) and cocoons retrieved from our cultures, respectively. The FLC test consisted of the continuous exposure of individuals from the juvenile to the adult stage and ended when the mean survival of the controls dropped under 70%, which corresponds to the threshold fixed by regulation for the validity of control data in toxicity tests with benthic invertebrates (19). Test organisms were sampled in cultures (see ref 20 for culture conditions) by gentle sieving of the substrate (mesh size ) 315 µm) and randomly introduced in the test chambers. Three and nine replicates of ten worms per tested concentration were used in the PLC and FLC tests, respectively. Organisms were kept at 21 °C under gentle aeration, exposed to 8:16 h light:dark photoperiod and fed ad libitum with grinded fish flakes (1.6 and 3.2 mg of Tetramin-TetraWerke, Melle, Germany- per worm per working day for juveniles and adults, respectively). Temperature, pH, conductivity, dissolved oxygen, nitrite and ammonia concentrations were monitored weekly in the overlaying water. Life-Cycle Trait Measurements. Individual performances were weekly monitored in the PLC tests: from days 14 to 28 in the survival/growth tests and from days 21 to 42 in the reproduction/hatching tests. Therefore, worms were gently sieved out of the sediment. Survivors were counted, blotted, weighted to the nearest 0.01 mg, and placed in renewed test chambers. The number of viable cocoons (i.e., neither opaque nor black) and the number of eggs per cocoon were counted using a light table. Fecundity was defined as the mean number of newborn produced per individual. The mean hatching rate was estimated in independent experiments, where cocoons from our culture were exposed to zinc-spiked sediments during 42 days. In the FLC test, individual performances were measured as described above, in independent replicates, at days 123, 151, and 179 (end of the test). Based on these data, we implemented a three-step modeling approach: • The “estimation” step consisted of identifying all parameters of individual effect models by fitting on PLC experimental data that showed significant effect of zinc (i.e., juveniles survival and growth - see the Results section). • The “calibration” step allowed identifying some parameters of the matrix population model based on information stemming from sparse FLC data (age at first brood, adult survival, and fecundity) and literature (20). • The “extrapolation” step was based on simulations with the matrix population model, in which inputs came from steps 1 (juvenile survival) and 2 (other inputs). Individual-Level Effect Modeling: The “Estimation” Step. Individual-level effects were modeled using threshold effect models, which included stress functions on parameters that described individual performances. Those stress functions were inspired from DEBtox models under the assumption of rapid kinetics (12). Hence, impairment of individual performances was supposed to occur only above an exposure concentration threshold called the No Effect Concentration (NEC), which was specific for every life-cycle trait. Effects were then supposed to be proportional to the difference between the exposure concentration c and the NEC. The proportionality constant indicated the intensity of toxic effects. Under these hypotheses, we formulated effect models

describing juvenile survival and growth in B. sowerbyi exposed to zinc. Adult survival and fecundity were calculated in control conditions only, because no significant effect occurred in exposed organisms for these parameters when compared to controls (see the Results section). The juvenile survival probability until time t at exposure concentration c was modeled using a classical survival function as proposed in DEBtox models (12). The hypothesis of a rapid kinetics for zinc led to the following exponential decay model with a stress function on the shape parameter p(t, c) ) e-b×(1+σs(c))×t with σs(c) ) ks × max(0, c - NECs) (1) with b representing the natural mortality rate (1/d), σs(c) representing the survival stress function (with NECs the no effect concentration for survival, in mg/kg), and ks representing the intensity effect coefficient on juvenile survival (kg/mg). According to experimental data and previous growth studies in control conditions (20), juvenile growth was simply described by a linear model including a stress function on the worm growth rate W(t, c) ) W0 + γ × (1 - σg(c)) × t with σg(c) ) min(1, kg × max(0, c - NECg))

(2)

with W0 representing the weight at the beginning of the experiment (mg), γ representing the daily worm growth rate (mg/d), σg(c) representing the growth stress function (with NECg the no effect concentration for growth, in mg/kg), and kg representing the juvenile growth effect intensity coefficient (kg/mg). In this first “estimation” step, threshold effect models were fitted to growth and survival data using Bayesian inference. In such an approach, all parameters are considered as unknown random variables with probability distributions. Bayesian inference consists of an update from chosen prior probability distributions to posterior probability distributions given the data. Parameter estimates are thus given under the guise of posterior probability distributions, from which point estimates and credible intervals can easily be deduced, as described in ref 14. In our paper, 2.5% and 97.5% quantiles are provided in square brackets for all parameter estimates. Prior probability distributions and units are provided for each model parameter in Table S2 (Supporting Information). Data were related to model outputs assuming a normal error model for growth and a binomial error model for survival. Empirical posterior probability distributions of all the parameters were obtained using Monte Carlo Markov Chains (21). Probability distributions of growth and survival predicted data were also simulated and then compared to the observed data for a posterior predictive checking (21). In contrast to a data/model fitting representation, posterior predictive checking also takes into account the chosen error model. All calculations were implemented in WinBugs (Imperial College London 22, 23). The model specification files we implemented are provided in the Supporting Information. Results were graphically represented using the R software (24). Population-Level Effect Modeling: The “Calibration” and “Extrapolation” Steps. As proposed for polychaetes (25), population-level effects of zinc in B. sowerbyi were extrapolated using a two stage-classified matrix model, where only juvenile and adult stages were distinguished. As hypothesized in partial life-cycle models (16), juveniles were assumed to survive with an average probability until puberty, and adults were supposed to contribute to the population increase with a constant average fertility rate from their first to their last breeding. Based on these hypotheses, the asymptotic population growth rate at each time step of the model (day) and VOL. 44, NO. 9, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

3567

at exposure concentration c was given by the following characteristic equation (25) λ(c)z(c)+1 - Sa × λ(c)z(c) - Sj(c) × F(c) ) 0

(3)

where λ(c) is the daily asymptotic population growth rate (1/d), z(c) is the age at first brood (d), Sj(c) is the probability to survive until the age at first brood, Sa is the daily adult survival rate, and F is the daily adult fecundity. Sj(c) was derived from the juvenile survival model (eq 1) as follows Sj(c) ) p(z(c), c) S Sj(c) ) e-b×(1+σs(c))×z(c)

(5)

with definitions and parameter values of b and σs(c) identical to those in eq 1. The age at first brood z(c) was modeled versus c with a stress function, as follows z(c) ) z0 × (1 + σz(c)) with σz(c) ) kz × max(0, c - NECz) (4) where z0 is the control age at first brood (d), σz(c) is the maturation stress function (with NECz the no effect concentration for maturation, in mg/kg), and kz is the maturation effect intensity coefficient (kg/mg). Based on PLC test results (see the Results section), effects on the age at first brood were supposed to be directly linked to effects on growth, leading to the hypothesis NECz ) NECg. Next, both literature (20) and our FLC test results for controls were used to calibrate parameter kz. We used a uniform probability distribution for z0 between bounds 57 and 62 days, and based on FLC tests, adults may reproduce from day 116 ((7 days) when exposed to the highest tested concentration. The maximum value of z(c), zmax, was then drawn in a uniform probability distribution between bounds 109 and 125 days. Finally, combining probability distributions on z0, zmax, and NECz into eq 4, Monte Carlo simulations led to a calibrated probability distribution for kz. Biological meaning and units of parameters involved in the “calibration” step are summarized in Table S3 (Supporting Information). The daily adult survival rate (Sa) and the daily adult fecundity (F) did not depend on the exposure concentration (see the Results section). The probability distribution of Sa was calibrated from the mean adult survival rate distribution at day 179 over the whole range of exposure concentrations (data from FLC tests) and then expressed per day. The probability distribution of F was estimated from the distribution of the mean number of eggs per survivor after 28, 35, and 42 days of reproduction over the whole range of exposure concentrations (data from PLC tests). We assumed that all eggs were viable and that the embryo survival rate varied in the fixed range given in ref 20. Once all parameter distributions were estimated or calibrated, the last “extrapolation” modeling step provided Monte Carlo simulations of the daily asymptotic population growth rate, implemented under the R software for exposure concentrations ranging from 0 to 3350 mg/kg d.w.

Results Exposure Conditions. Physicochemical parameters remained stable during the tests, maximal temperature variation being (1°C, pH being comprised between 7.6 and 8.0, conductivity ranging from 270 to 330 µS/cm, and dissolved oxygen remaining above 80%. Nitrate and ammonia concentrations remained below 0.2 and 5.0 mg/L, respectively, which were not toxic values for the worms. Background zinc concentration in control sediments (i.e. spiked with 0 mg/kg d.w. of zinc) ranged from 55 to 72 mg/kg d.w. Actual tested zinc concentrations were in the range 409-3769 mg/kg d.w. 3568

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 9, 2010

(see Table S1, Supporting Information). Zinc concentrations in overlaying and pore water remained below 5 mg/L. They were not accounted for in models, exposure being assumed to occur mainly due to contact and ingestion of sediment in conveyer-belt feeders like B. sowerbyi. Short-Term Effects of Zinc on Adults. Based on 28 days PLC test results in the range 1087-3769 mg/kg d.w., subadult and adult survivals (Figure S1, Supporting Information) were not significantly impaired until exposure concentration reached 3317 mg/kg d.w. This concentration was 25 times higher than the average concentration calculated based upon measurements in 455 sediments samples, which were collected in various locations in France (130 mg/kg d.w. (26)). Therefore, we assumed that adult worms would not experience adverse effects on survival under average field concentration. Furthermore, no significant effect on the growth of subadult and adult worms occurred during the PLC tests. Adult survival and growth thus remained independent of c in our population model. PLC tests on reproduction (42 days) in the range 419-1651 mg/kg d.w. indicated that the onset of reproduction in mature worms was not delayed. Moreover, the number of breeding worms, the number of cocoons laid per breeding individual, and the number of eggs per cocoon were similar in controls and in exposed worms. No significant effect of zinc on the hatching rate was evidenced in this concentration range. We thus assumed that zinc would not induce deleterious effects on reproduction in mature adults exposed to average field concentration and considered that fecundity did not depend on c in our population model. Yet, 67% (s.d.) 34%) of worms born in sediments where zinc concentrations exceeded 551 mg/kg exhibited morphological abnormalities in gills (i.e., lack or malformation of respiratory filaments, or spare gills). The increase in abnormal worm proportion could not be significantly related to zinc concentration, due to the lack of replicates in our test design concerning the variability of this endpoint. It was thus not accounted for in the population model. Short-Term Effects of Zinc on Juveniles. Based on the 28 days PLC survival test results in the range 1307-2961 mg/kg d.w., a significant decrease in juvenile survival (Kruskall-Wallis test, p3000.0]). These results confirmed that populationlevel effects may not be reliably simulated based only on the impairment of the most sensitive life-cycle trait (31). Relevance of the Proposed Modeling Strategy. The present approach provided simple test methods and mathematical models to assess long-term effects of toxicants in B. sowerbyi populations. It allowed the use of short-term test data (PLC tests) to estimate individual-level effect parameter values that were biologically meaningful and did not depend on exposure durations. It also allowed to properly model individual performances in control and exposed worms, as shown in Figures 1 and 2 and in the Supporting Information. A major advantage of this approach consisted of its ability to reliably describe the time-course of toxic effect with a daily time step. Such information was directly useful in population-level models and was obtained at a limited experimental cost when compared to dedicated life-table response experiments. Using information stemming from both PLC (“estimation” step) and FLC tests (“calibration” step) allowed ultimately extrapolating reliable population-level outputs. Indeed, estimated values for adult fecundity (F), maximal age at first brood (zmax), and adult survival rate (Sa) were in accordance with values measured at the laboratory in experimental populations of B. sowerbyi reared at 17 °C (32). The daily asymptotic population growth rate we simulated for controls was in accordance with results from other studies, using either a classical exponential growth model (30) or a system of differential equations based on exponential growth to describe the dynamics of various size classes (32). Moreover, our approach was more relevant than existing ones (30, 32) in attempting to extrapolate population-level effects of toxicants in B. sowerbyi. Indeed, it allowed extrapolating effects on individual growth, survival, and maturity in population-level outputs, notably by including the potential increase in age at first brood (eq 4) what was generally ignored in population risk assessment. Yet, life span, growth, and reproduction are highly plastic in natural populations of B. sowerbyi, being mainly determined by temperature and food availability (4). The influence of those modulating factors was not accounted for in our models. Therefore, including them as covariables in individual-level effect models, using e.g. process-based models derived from the Dynamic Energy Budget theory (33), would probably improve the relevance of further approaches towards field populations. Interpreting test data using Bayesian inference was especially useful to deal with parameter uncertainty because (i) parameter estimates were provided as probability distributions reflecting parameter uncertainty at the individual level, which could then be extrapolated at the population level, and (ii) simulated data were provided with their distribution and then compared to the observations for posterior predictive checking. This method ensured improving the relevance of projection matrix outputs by accounting for the uncertainty between model parameters while simu-

lating population response to the toxicant. This is a major benefit of our approach, which is thus particularly worth using in the context of ecological risk assessment.

Acknowledgments The authors thank the CEMAGREF, which funded this study and enabled V. Ducrot to obtain the presented data during her Ph.D., helped by Raphae¨l Mons & Herve´ Que´au who provided efficient technical assistance during the experiments. The authors express their thanks to Victoria Suntov (University Lyon 1) who provided statistical test results during her master training course and to Thierry Caquet (INRA) for his great help in improving this manuscript. Finally, we thank three anonymous referees for their very relevant and helpful comments.

Supporting Information Available It presents full data obtained from PLC and FLC tests and describes the model specification files in WinBugs, items that we used in the estimation and calibration steps of the modeling. It also provides prior and posterior distributions and point estimates of all model parameters. It finally presents the calibrated relationship between age at first brood and exposure concentration. This material is available free of charge via the Internet at http://pubs.acs.org.

Literature Cited (1) Ingersoll, C. G.; Ankley, G. T.; Benoit, D. A.; Brunson, E. L.; Burton, G. A.; Dwyer, F. J.; Hoke, R. A.; Landrum, P. F.; NorbergKing, T. J.; Winger, P. V. Toxicity and bioaccumulation of sediment-associated contaminants using freshwater invertebrates: a review of methods and applications. Environ. Toxicol. Chem. 1995, 14, 1885–1894. (2) Reynoldson, T. B. Interactions between sediment contaminants and benthic organisms. Hydrobiologia 1987, 146, 53–66. (3) Marchese, M. R.; Brinkhurst, R. O. A comparison of two tubificid oligochaete species as candidates for sublethal bioassay tests relevant to subtropical and tropical regions. Hydrobiologia 1996, 334, 163–168. (4) Carroll, J. H.; Dorris, T. C. The life history of Branchiura sowerbyi. Am. Midl. Nat. 1972, 87, 413–422. (5) Chapman, P. M.; Farrell, M. A.; Brinkhurst, R. O. Relative tolerances of selected aquatic oligochaetes to individual pollutants and environmental factors. Aquat. Toxicol. 1982, 2, 47– 67. (6) Naqvi, S. M. Z. Toxicity of twenty-three insecticides to a tubificid worm Branchiura sowerbyi from the Mississippi delta. J. Econ. Entomol. 1973, 66, 70–74. (7) Ghosal, T. K.; Kaviraj, A. Influence of poultry litter on the toxicity of cadmium to aquatic organisms. Bull. Environ. Contam. Toxicol. 1996, 57, 1009–1015. (8) Casellato, S.; Aiello, R.; Negrisolo, P. A.; Seno, M. Long-term experiment on Branchiura sowerbyi (Beddard) (Oligochaeta, Tubificidae) using sediment treated with LAS (Linear Alkylbenzene Sulfonate). Hydrobiologia 1992, 232, 169–173. (9) Raburu, P.; Mavuti, K. M.; Harper, D. M.; Clark, F. L. Population structure and secondary productivity of Limnodrilus hoffmeisteri (Claparede) and Branchiura sowerbyi (Beddard) in the profundal zone of Lake Naivasha, Kenya. Hydrobiologia 2002, 488, 153– 161. (10) Yanling, L. Annual production of Branchiura sowerbyi (Oligochaeta, Tubificidae) in the Donghu Lake, Wuhan, China. Chin. J. Oceanol. Limnol. 1984, 2, 102–108. (11) Guidance document on control of toxicity test precision using reference toxicant; Report number SPE 1/RM/12; Environment Canada: Ottawa, Ontario, 1990. (12) Kooijman, S. A. L. M.; Bedaux, J. J. M. The analysis of aquatic toxicity data; VU University Press: Amsterdam, The Netherlands, 1996.

(13) Isnard, P. F. P.; Roman, G.; Babut, M.; Bastien, P.; Bintein, S.; Esserme´ant, L.; Fe´rard, J. F.; Gallotti-Schmitt, S.; Saouter, E.; Saroli, M.; Thie´baud, H.; Tomassone, R.; Vindimian, E. Statistical analysis of regulatory ecotoxicology tests. Chemosphere 2001, 45, 659–669. (14) Billoir, E.; Delignette-Muller, M.-L.; Pery, A. R. R.; Charles, S. A Bayesian approach to analyzing ecotoxicological data. Environ. Sci. Technol. 2008, 42, 8978–8984. (15) Caswell, H. Matrix Population Models, construction, analysis, and interpretation, 2nd ed.; Sinauer Assiociates: Sunderland, MA, U.S.A., 2001. (16) Oli, M. K. Partial life-cycle analysis, a simplified model for postbreeding census data. Ecol. Model. 2003, 168, 101–108. (17) Ducrot, V.; Pe´ry, A. R. R.; Mons, R.; Queau, H.; Charles, S.; Garric, J. Dynamic energy budgets as a basis to model population-level effects of zinc-spiked sediments in the Gastropod Valvata piscinalis. Environ. Toxicol. Chem. 2007, 26, 1774–1783. (18) Simpson, S. L.; Angel, B. M.; Jolley, D. F. Metal equilibration in laboratory-contaminated (spiked) sediments used for the development of whole-sediment toxicity tests. Chemosphere 2004, 54, 597–609. (19) Test method for measuring the toxicity of sediment-associated contaminants with freshwater invertebrates; Report number E 1706 95-b; American Society for Testing Materials: Philadelphia, PA, U.S.A., 2000. (20) Ducrot, V.; Pe´ry, A. R. R.; Que´au, H.; Mons, R.; Lafont, M.; Garric, J. Rearing and estimation of life-cycle parameters of the tubificid worm Branchiura sowerbyi: application to ecotoxicity testing. Sci. Tot. Environ. 2007, 384, 252–263. (21) Gilks, W. R.; Richardson, S.; Spiegelhalter, D. J. Markov Chain Monte Carlo in practice; Chapman and Hall/CRC: Boca Raton, FL, U.S.A., 1996. (22) Lunn, D. J.; Thomas, A.; Best, N.; Spiegelhalter, D. Winbugs a Bayesian modelling framework: concepts, structure, and extensibility. Stat. Comput. 2000, 10, 325–337. (23) Spiegelhalter, D.; Thomas, A.; Best, N.; Lunn, D. WinBUGS Version 1.4 User Manual; MRC Biostatistics Unit: Cambridge, U.K., 2003. (24) R Development Core Team. R: A Language and Environment for Statistical Computing; R Foundation for Statistical Computing: Vienna, Austria, 2007. (25) Levin, L. A.; Caswell, H.; Bridges, T. S.; DiBacco, C.; Cabrera, D.; Plaia, G. Demographic response of estuarine polychaetes to pollutants: life table response experiments. Ecol. Appl. 1996, 6, 1295–1313. (26) Garric, J. Status report on aquatic micropollution in French inland waters. Rev. Med. Vet. 2002, 153, 337–344. (27) Tachet, H.; Richoux, P.; Bournaud, M.; Usseglio-Polatera, P. Inverte´bre´s d’eau douce, syste´matique, biologie, e´cologie; Centre National de la Recherche Scientifique: Paris, France, 2000. (28) Naitoh, T.; Takabatake, I. Observations on the tail movements and the contractions of the dorsal blood vessel and intestine in an anomalously double-tailed worm of Oligochaeta, Branchiura sowerbyi (Beddard): a case report with implications for physiological mechanisms of contractility. Mem. Fac. Sci. Shimane Univ. 1995, 29, 45–50. (29) Aston, R. J. The effect of temperature on the life cycle, growth and fecundity of. Branchiura sowerbyi (Oligochaeta, Tubifcidae). J. Zool. London 1968, 154, 29–40. (30) Aston, R. J.; Sadler, K.; Milner, A. G. P. The effects of temperature on the culture of Branchiura sowerbyi (Oligochaeta, Tubificidae) on activated sludge. Aquaculture 1982, 29, 137–145. (31) Kammenga, J. E.; Busschers, M.; Van Straalen, N. M.; Jepson, P. C.; Bakker, J. Stress-induced fitness reduction is not determined by the most sensitive life-cycle trait. Funct. Ecol. 1996, 10, 106–111. (32) Bonacina, C.; Pasteris, A.; Bonomi, G.; Marzuoli, D. Quantitative observations on the population ecology of Branchiura sowerbyi (Oligochaeta, Tubificidae). Hydrobiologia 1994, 278, 267–274. (33) Kooijman, S. A. L. M. Dynamic energy and mass budgets in biological systems, 2nd; University Press: Cambridge, 2000.

ES903860W

VOL. 44, NO. 9, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

3571