Bioamplification as a Bioaccumulation Mechanism for Persistent Organic Pollutants (POPs) in Wildlife

Abstract

Bioaccumulation describes the process by which anthropogenic chemicals are taken up by organisms from their environment and diet and are subsequently assimilated and distributed into tissues (Arnot and Gobas 2003; Borgå et al. 2004; Mackay and Fraser 2000). Thus, bioaccumulation is a central framework within ecotoxicology, because it helps define the maximum concentration that can be achieved by an organism in its tissues, relative to the exposure media, and helps determine the potential chemical dose/toxicity experienced by an individual. Therefore, understanding the dynamic processes that regulate chemical bioaccumulation in animals is essential for protecting species, ecosystems, and ultimately human health (Arnot and Gobas 2004; Kelly et al. 2004).

Appendices

10 Appendix 1: Description of the Yellow Perch Model and Associated Simulations

Simulations were performed by finite difference, using a daily time step. The model was set up using a Microsoft Excel spreadsheet, and the calculation algorithms and/or constants for each variable of the model are summarized in Table 4. The model was initialized by assuming that day 1 corresponded to a 1-year-old fish (365 days old) hatched on May 20 of the previous year. The model was initialized at this life stage, because of a lack of data on growth, proximate composition changes and chemical toxicokinetics in larvae fish during the first year. The model was run for 2,920 days (i.e., 8 years from the simulation initialization). In the simulations, fish growth was allowed to change as a function of age, following the precedent of a von Bertalanffy model fitted to 22 North American populations of yellow perch (Jackson et al. 2008). The von Bertalanffy model was used to predict the length (cm) of fish at a given age (in days) for the constant temperature scenario. Body length (cm) was converted to a body weight (g) value, by using a linear regression equation fitted to the log₁₀ of body length versus log₁₀ of body weight (this approach relied on unpublished data that was generated for aquaculture-reared yellow perch; see Table 4).

Table 4 Model parameters and variables used in yellow perch non-steady state bioaccumulation simulations

For the dynamic temperature scenario, growth rates from the baseline simulations were further modified to make growth rates temperature dependent, whereas overall growth rates generally matched those of the constant temperature simulation. The modifications to the basic growth model are described as follows. On May 20 of each year (i.e., reflecting the assumed birthday of the fish), the body weight of the fish was set to be equal to the body weight estimated for the constant temperature simulation. Body weights for other days in the year were calculated by a finite difference method as: BW_(t) = BW_(t−1) + GC_YI × T; where GC_YI is the temperature-dependent growth constant for a given year interval and T is the water temperature (°C). The GC_YI was iteratively fit for each year class so that the predicted body weight of fish at the next May 20th date corresponded to the body weight predicted under the constant temperature scenario. The fitted coefficients for each year class are described in the footnotes of Table 4.

Whole body lipid contents (g lipid per g body weight) were predicted from the temperature-dependent algorithm used for medium and large yellow perch described in Drouillard et al. (2009). For the constant temperature simulations, all fish had a constant fractional lipid content of 0.086 (by weight). For the dynamic temperature scenario, lipid content ratios varied between 0.043 and 0.093, depending on the temperature. Water content of the fish was predicted from the negative correlation between %moisture and %lipid as described in Drouillard et al. (2009). The lean dry weight of fish was subsequently estimated by subtracting the water and lipid weights from the fish body weight. Daily changes in whole body lipid and lean dry weights were used to estimate growth increments, growth energy costs and lipid equivalent contents for each time step (see Table 4 for details). Daily estimates of food consumption rates (g food consumed per g body weight per day) and gill ventilation rates (mL water ventilated per g body weight per day) for yellow perch were calculated from temperature and size dependent algorithms as described in Drouillard et al. (2009) and detailed in Table 4.

The uptake portion of the toxicokinetics sub-model used the calculation methods of Arnot and Gobas (2004). Uptake flux of chemical into fish from water (ng/d) was estimated as the product of gill ventilation rate (mL per g body weight per day), chemical extraction efficiency across gills (unitless), chemical concentration in water (ng per mL), and body weight (g) of fish. In both scenarios, the recommended value of 0.54 was used for the chemical extraction efficiency across gills. The uptake flux of chemical into fish from food (ng/d) was estimated as the product of the feeding rate, dietary assimilation efficiency of the chemical, chemical concentration in food and body weight of fish. A constant value of 0.6 was used as the dietary assimilation efficiency value of the chemical. Chemical loss from fish (ng/d) via gills and feces were estimated as detailed in Drouillard et al. (2009), and described in Table 4. For simplicity of the simulations, the same chemical extraction efficiency across gills was used as the uptake algorithm. Similarly, the organism/feces chemical exchange efficiency was set to be equal to the dietary chemical assimilation efficiency described above.

Yellow perch model simulations were performed to predict bioaccumulation and bioamplification of a negligibly metabolized POP compound having a log K _OW of 6.5. Two different simulation scenarios were established to compare bioamplification under an artificial baseline condition of constant temperatures and a seasonally dynamic scenario that is consistent with temperate aquatic ecosystems. The baseline scenario involved a simulation, in which temperature was kept constant (21 °C; at the species optimum) across seasons and years. The seasonal scenario involved a simulation, in which an annual temperature profile that is consistent with measurements made in aquaculture ponds of Southern Ontario, was applied to predict the seasonal variation in lipid content and its impact on chemical bioamplification. The same annual temperature profile was cycled across simulation years. In both simulations, the model was initialized by assuming Age 1 fish were in chemical equilibrium with water. Fish were assumed to feed on a stable food source (constant in its proximate composition, energy density and chemical concentrations) over the duration of the simulation. The concentration of chemical in food was set so that the food was in equilibrium with the concentration of chemical in water.

The yellow perch non-steady state bioaccumulation model was developed using a Microsoft Excel spreadsheet. The model is a finite difference model run at a daily time steps for 2,920 days. The model was initialized with a temperature of 15.9 °C and fish aged 365 d (May 20, year 1), body weight of 4.84 g, lipid content of 7.5%, moisture content of 71.8%. The initial fish concentration was set to be in equilibrium with water. Food concentration was set to be in equilibrium with water. The food concentration we held constant throughout the experimental duration.

11 Appendix 2: Description of the Herring Gull Model and Associated Simulations

The herring gull bioenergetic and toxicokinetics model is described in several papers (Norstrom et al. 1986a, b, 2007; Clark et al. 1987, 1988; Drouillard et al. 2003). Most commonly, the model has been used to describe non-steady bioaccumulation of POPs in adult female life stages over multi-year periods. However, the basic algorithms for chick growth and bioenergetics of early and late life stages for both sexes are detailed in Norstrom et al. (1986b). For simplicity, the herring gull model simulations were formulated for male birds to circumvent the need to consider chemical depuration by egg laying, and to maximize predictions of bioamplification in the species. The model was expanded to include three linked life-stages: a chick stage (post-pipping to fledging), immature adult (post fledging to 3.8 years), and a reproductive adult (3.8 years to 8 years) stage. The chick stage immediately experiences bioamplification from maternally deposited residues (Drouillard et al. 2003), followed by growth dilution until fledging. The subadult male experiences seasonal temperature variation and proximate composition, but does not participate in reproductive activities (such as courtship, or the feeding and foraging costs associated with rearing a clutch of chicks). Adult males experience additional foraging costs associated with the later activities. Although herring gulls have much longer life spans than the 8-year period of the model simulation, an 8-year duration was selected for consistency with yellow perch simulations.

For each life stage, the bioenergetic sub-model predicts growth, proximate composition, and food consumption as outlined in Norstrom et al. (1986a). The toxicokinetics model only considers chemical uptake from food, since air uptake by birds is negligible (Drouillard et al. 2012). Similar to the yellow perch model, the uptake flux of chemical into the bird from food (ng/d) was estimated as the product of the feeding rate, dietary assimilation efficiency of chemical, chemical concentration in food, and body weight of fish. A constant value of 0.9 was used as the dietary assimilation efficiency value for the chemical and was derived from data collected for ring doves (Drouillard and Norstrom 2000). The toxicokinetic parameters necessary to describe chemical elimination included the plasma/fat partition coefficient (K _PF) and plasma clearance coefficient (k′_pc). For model simulations, the values of K _PF and k′_pc for mirex, measured in juvenile herring gulls (Clark et al. 1987), were used and were assumed to be constant across the different life stages. Mirex was chosen to represent a highly hydrophobic POP that is negligibly biotransformed in birds. A modification to the herring gull model not applied in previous publications of the model was that the chemical outflux was measured by multiplying the k′_pc by the body weight and lipid equivalent concentration of chemical in the animal tissues. Past iterations of the module used the lipid normalized concentration. This change was made to make the model more consistent with the yellow perch model. However, it should be noted that the fish and bird models differ fundamentally in how elimination flux is treated. In the herring gull model, k′_PC is a constant, and elimination flux varies over the year only as a result of changes in the proximate composition (lipid equivalent content) of the animal. In the fish model, elimination flux of the chemical depends on variation in proximate composition, as well as variation in gill ventilation and feeding rates. These differences result in a de-coupling of feeding rates and chemical elimination in birds that causes lags in return to steady state, following sudden shifts in feeding rates. This is exemplified in Fig. 2c for the constant temperature, post-adult male simulation. No attempts were made to harmonize the two model organisms into a single toxicokinetic model, because we preferred to preserve the characteristics and attributes of the models that had been addressed in their original publications.

As for the yellow perch models, two simulation scenarios were established for herring gulls. The baseline simulation kept temperature and photoperiod constant at 21 °C and 12 h/d across seasons and years. The seasonal scenario used temperature and photoperiod data from Lake Ontario that had been collected during 1997. The model used monthly mean temperature and photoperiod data and interpolated temperature and photoperiods for each day of the simulation. The model cycled the same annual temperature profile across all years in each simulation. The model was initialized using a fresh egg concentration predicted from a 10-year adult female model simulation (Clark et al. 1988), wherein the female bird was fed a constant diet of the same concentration and energy density as that used for male simulations. The female simulation used a constant temperature and photoperiod to initialize the constant temperature simulation, and a variable temperature and photoperiod equivalent to the Lake Ontario profile to initialize the dynamic temperature simulation. The simulated fresh egg concentration (μg/kg wet weight) from adult female simulations was multiplied by the egg weight (85 g), and was divided by the fresh egg lipid content (7.2 g) as reported by Drouillard et al. (2003) to estimate the lipid normalized egg concentration. Bioamplification of fresh egg residues in the pipping embryo was accounted for by multiplying the fresh egg lipid normalized concentration by a factor of 3.1 (Drouillard et al. 2003), and multiplied by the lipid equivalent content of the newly pipped chick to determine the total mass of chemical in the chick. The bird was grown out and was assumed to feed on a constant food source of proximate composition, energy density, and chemical concentration that was similar for the duration of the study. A full description of model parameters and algorithms employed is presented in Table 5.

Table 5 Model parameters and variables used in herring gull non-steady state bioaccumulation simulations

The herring gull toxicokinetic model output was copied onto a Microsoft Excel spreadsheet. The model is a finite difference model and was run at a daily time steps for 3,137 days. The model was initialized with a 1-day-old pipping male chick hatched on May 29, 1997. The chick sub-model was used to calculate growth, proximate composition, bioenergetics, and chemical toxicokinetics between days 1 and 88. The output from the chick sub-model was linked to a subadult male model between days 89 and 1,398, i.e., up to 4 years. In the subadult model, immature birds were assumed to not participate in reproductive activities and therefore had no courtship feeding or chick provisioning costs. The output from the subadult male model was linked to a reproductive adult male model which covered simulation days 1,399–3,136.