Environ. Sci. Technol. 2006, 40, 6026-6031
Mechanistic Models to Perform Population Risk Assessment with the Midge Chironomus Riparius: Application to Heavy Metals A L E X A N D R E R . R . P EÄ R Y , * , † ALAIN GEFFARD,‡ AND JEANNE GARRIC† Laboratoire d’e´cotoxicologie, Cemagref, 3bis quai Chauveau, CP 220, 69336 Lyon, Cedex 9, France, and Laboratoire d’Eco-Toxicologie, Universite´ de Reims EA 2069 URVVC, Moulin de la Housse 51687 Reims Cedex 2, France
Mechanistic models can substantially contribute to population risk assessment to assess effects on population and to increase the relevance of the toxicity parameters estimated at an individual level. We use four mechanistic models to change the scale from concentration to effects on individuals and from individuals to population with the midge Chironomus riparius: a kinetics model; an energy-based effects model, linking effects on the life cycle and compound body residues; a matrix approach to derive population growth rate; and an energy-based population model to derive carrying capacity. The whole “model battery” was applied to cadmium and copper. The data came from growth, survival, and reproduction tests. We also incorporated information about compounds physiological mode of action and kinetics. Thresholds at population level were derived through comparisons with our control database. We showed that our two population endpoints (carrying capacity and population growth rate) provide complementary information about toxicity risks, even if, in our study, population growth rate appeared to be slightly more sensitive than carrying capacity. We found population no effect concentration of, respectively, 0.42 and 9.3 mg/kg for cadmium and copper. We also showed that information about physiological mode of action was relevant, whereas a kinetics test was unnecessary.
Introduction The environmental risk assessment of toxicants still needs more relevant studies at the population level. The change of scale obtained through population dynamics modeling permits us to integrate in one parameter all the effects observed at the individual level and to assess effects at a biological level much closer to the ecosystems than a simple bioassay data analysis. We consider here population endpoints which we believe to be complementary when assessing perturbations on the populations. First, population growth rate is the most currently used population endpoint by ecotoxicologists when studying populations (1). In most cases, it is calculated in optimal conditions, and a significant effect on this parameter accounts for an increased risk of dis* Corresponding author phone: 0033 4 72 20 87 88; fax: 0033 4 78 47 78 75; e-mail:
[email protected]. † Laboratoire d’e ´ cotoxicologie, Cemagref. ‡ Universite ´ de Reims. 6026
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 19, 2006
appearance, a decrease of the ability to colonize a new environment, or a decrease of the recovery rate. We recently proposed (2) another population endpoint for C. riparius, the carrying capacity (the mean number of individuals that a system can handle in a long-term perspective). This endpoint accounts for the mean number of individuals present per time unity. Any perturbation would have consequences for all the species linked through food chains to the species we studied. Moreover, our study of the carrying capacity is analogous to the threshold food concentration concept (3), which calculates the energetic requirements at population steady state and provides a basis to study the outcome of food competition between species from the same family. Population risk assessment is relevant only if the parameters used to perform the change of scale from individual to population are relevant. To achieve this, we chose energybased models derived from DEBtox models (4). These models comport two parts. An effect module links effects to the body residues of the tested compound. Indeed, in most of the cases, the compound is only active when it has entered into the test organisms. That is why some authors suggest expressing toxicity results in terms of internal concentration (5-6). In the energy-based models, effects on one given parameter of the life cycle are proportional to the difference between body residues and a threshold value. A kinetics module, assuming a one-compartment kinetics, links body residues and external concentration, taking time into account. These models present many advantages when compared with purely statistical approaches. First, the estimated parameters are time-independent, in contrast to the NOEC or the LC50 which values depend on the duration of the test (4). Second, the expected dose-response profiles are realistic when compared to actual data (7). Moreover, the models presented here can contribute to the estimation of the physiological mode of action of contaminants, which correspond to the energy process (assimilation, growth, maintenance) targeted by the compound. For instance, we showed that an increase of energy costs was more likely to be the actual copper physiological mode of action on Chironomus growth (8). In the present paper, we first measure effects of cadmiumspiked artificial sediments on the kinetics of accumulation and on the life cycle parameters of C. riparius. The data are analyzed with our models, and projections at population level are derived from population growth rate in ad libitum feeding conditions and on the carrying capacity for cadmium and also for copper, as we have obtained the data from previous experiments (7-9). The parameters estimation at an individual level are performed with and without the incorporation of kinetics-measured parameters and the change of scale from individual to population are performed for the two possible physiological modes of action (effects on food assimilation or on growth) to investigate the importance of acquiring this information on parameter estimation and on the change of scale from individual to population. This is important because any additional information about kinetics or physiological mode of action demands one supplementary test, which increases the global cost of the risk assessment. To assess population thresholds, results are compared to the distribution of the control population endpoints we obtained using our population models and our control database in standard conditions.
Experimental Section As the experimental data for copper and their analysis according to energy-based models has been extracted from other publications (7-9), we focus on cadmium toxicity tests 10.1021/es0607234 CCC: $33.50
2006 American Chemical Society Published on Web 09/01/2006
in the first parts of this section (experiments and data analysis at individual level). Sediment Spiking. We used an artificial sediment: silica sand with the following particle size distribution: 90% between 50 and 200 µm and 10% under 50 µm. We chose this sediment because we previously showed that, among all the sediments we spiked in our laboratory, it is the one leading to the lowest thresholds. These thresholds are consequently protective for a large range of sediments. Three weeks prior to the spiking, 30 kg of sediments were maintained with 25 liters of water and a small amount of food (0.5 g Tetramin fish food) to allow bacterial development. To spike cadmium into sediment, we placed 1.98 kg wet sediment into 2 liter jars together with cadmium (CdCl2) dissolved in 0.8 liters of water. Jars were then rolled for 6 h, maintained at test temperature, and swirled manually each day for 10 days. The concentration of the compounds was measured in the spiked sediment and in the overlying water at the last day of the fourth instar growth test. The measurements were performed by the Laboratoire d’analyze des eaux in the Cemagref, Lyon. Growth and Survival Tests. The test beakers were prepared 3 days before the beginning of each test. They were filled with 0.1 liter sediment and 0.4 liter water (half from demineralised water and half from an uncontaminated spring near our laboratory; the pH is 8.1 and the conductivity is 400 µS/cm). Using the results of previous tests, we used exposure concentrations of 0, 0.33, 1, 3, and 9 mg/kg. Tests organisms (10 per beaker) from our laboratory culture were introduced into each glass-beaker ( nine per concentration). Three experiments were performed with the second (2 days after hatching), third (4 days after hatching) and fourth instars (6 days after hatching). Instar was checked with head capsule width measurements. The beakers were set in a water bath at 21 °C with a 16:8 h light:dark photoperiod, and water was aerated. Midges were fed each day with 0.6 mg Tetramin fish food (Tetrawerke, Melle, Germany) per individual, which are ad libitum conditions (8). Growth effects were assessed by measuring the length daily during 3 days. Each day, we randomly chose three beakers per concentration for measurements. Additionaly, twenty organisms were measured at the start of the tests. Length was measured using a binocular microscope fitted with a calibrated eyepiece micrometer. Prior to measurements, organisms were killed using a solution of 20% formaldehyde and 80% water. They were kept less than 10 s to avoid distortion of the shape. An additional fourth instar growth test comparing ad libitum and food-limited conditions (0.1 mg food per larva per day) was performed to assess the physiological mode of action of cadmium (decrease of feeding rate or increase of growth costs). Individuals were exposed at the beginning of the fourth instar (6 days after hatching), during 4 days in ad libitum conditions, and 5 days in food-limited conditions. In these conditions, if the physiological mode of action is decrease of feeding rate, high effect in ad libitum conditions would result in much lower effect in food-limited conditions because, despite a reduced assimilation rate, all the food available within a 1 day period will be assimilated and transformed into biomass. We used the same concentrations as in the growth tests and five replicates per concentration and feeding level. Reproduction Tests. The assays were initiated with 4 days old larvae (end of the second larval instar). At day 0 of the test, 10 organisms were randomly introduced into each glassbeaker. There were six replicates per cadmium concentration. We used the same exposure concentrations as for the growth tests and the beakers were set in the same conditions as previously used. Feeding level was 0.6 mg per larva per day. The beakers were covered with a net trap to prevent imagoes from escaping. From the very start of emergences,
imagoes were withdrawn daily from the test beakers by aspiration and counted. Females emerging from each replicate corresponding to a defined concentration were gathered into a bottle (Pyrex, one liter) containing a small quantity of water in order to receive the egg laying. Bottles were immediately stored at 23 °C after addition of two males per female taken in the laboratory culture to ensure that enough males were available from the test beakers to fertilize the emerging females. The egg-masses were removed daily out of each bottle. The number of eggs per egg-mass was then counted. Each egg mass was removed and put into a 5 mL tube with 2 mL H2SO4, 2 N overnight. The following day, the tubes were agitated to dissociate the eggs and then were counted using a binocular microscope. For each egg mass, measurements were made three times to reduce experimental errors. Kinetics Tests. Kinetics tests can seem uninformative in a DEBtox context: in the energy-based models, kinetics parameters, and in particular elimination rate, are directly deduced from the effects data. Here, we plan to investigate this point. For each sediment, we followed cadmium kinetics in fourth instar larvae for nominal concentrations 0, 1, and 3 mg/kg. Test conditions were the same as previously used. Three groups of 10 organisms at day 0 and three beakers per concentration at day 1, 2, and 3, were randomly taken, then the larvae were left for 10 min at room temperature in a 100 mL solution of 3 × 10-3 M EDTA to remove cadmium from the surface of the organisms, as we performed with copper in previous experiments (10). Afterward, organisms were frozen at -80 °C, freeze-dried, and weighed. Before metal analysis, a hot acid mineralization was realized by the addition of nitric acid (1 mL by 0.5 g wet weight). After 12 h at 80 °C, the sample was adjusted at a known volume (1 mL) with deionised water and analyzed by flameless atomic absorption spectrophotometry (AAS). Internal calibration was equally applied with lobster hepatopancreas tissue (TORT2, NRC-CNRC) as reference material (our value 26.2 ( 1.4 mg/kg; certified value 26.7 ( 0.6 mg/kg). Data Analysis for the Tests. The kinetics is assumed to correspond to a one-compartment model:
dci (t) ) ke × (ce(t) - ci(t)) dt
(1)
with ke the elimination rate, ce(t) the exposure concentration, and ci(t) the concentration in the tissue scaled by the bioconcentration factor, ku/ke, where ku is the uptake rate of the compound. The scaled concentration ci(t) is proportional to the concentration in the tissue, but has the dimension of an external concentration. Parameters (ke and bioconcentration factor) are estimated through the leastsquare method. The analysis of growth data is detailed in a previous paper (8). In ad libitum conditions, growth was linear for each instar, with a growth rate depending on the instar. The equations describing length growth from 1 day to the following one are then as follows:
ln+1 ) ln + a, γ‚(ln+13 - ln3) )
if (ln + a)3 - ln3
