Aquatic Toxicology 86 (2008) 470–484 Modeling larval fish behavior: Scaling the sublethal effects of methylmercury to population-relevant endpointsଝ Cheryl A. Murphy a,∗, Kenneth A. Rose a, Mar´ıa del Carmen Alvarez b,1, Lee A. Fuiman b a Louisiana State University, Department of Oceanography & Coastal Sciences, Energy, Coast and Environment Building, Baton Rouge, LA 70803, USA b The University of Texas Marine Science Institute, 750 Channel View Drive, Port Aransas, TX 78373, USA Received 8 October 2007; received in revised form 14 December 2007; accepted 20 December 2007 Abstract Expressing the sublethal effects of contaminants measured on individual fish as cohort and population responses would greatly help in their inter- pretation. Our approach combines laboratory studies with coupled statistical and individual-based models to simulate the effects of methylmercury (MeHg) on Atlantic croaker larval survival and growth. We used results of video-taped laboratory experiments on the effects of MeHg on larval behavioral responses to artificial predatory stimuli. Laboratory results were analyzed with a regression tree to obtain the probability of control and MeHg-exposed larvae escaping a real predatory fish attack. Measured changes in swimming speeds and regression tree-predicted escape abilities induced by MeHg exposure were then inputted into an individual-based larval fish cohort model. The individual-based model predicted larval-stage growth and survival under baseline (control) conditions, and low- and high-dose MeHg exposure under two alternative predator composition scenarios (medusa-dominated and predatory fish-dominated). Under MeHg exposure, stage survival was 7–19% of baseline (control) survival, and the roughly 33-day stage duration was extended by about 1–4 days. MeHg effects on larval growth dominated the response under the medusa- dominated predator composition, while predation played a more important role under the fish-dominated predator composition. Simulation results suggest that MeHg exposures near extreme maximum values observed in field studies can have a significant impact on larval cohort dynamics, and that the characteristics of the predator–prey interactions can greatly influence the underlying causes of the predicted responses. © 2007 Elsevier B.V. All rights reserved. Keywords: Methylmercury; Predator–prey interactions; Individual-based model; Regression tree; Larval fish; Behavior; Population-level effects 1. Introduction ology, metabolism, homeostasis, and behavior of an organism (Damstra et al., 2002). Despite intensive study on how endocrine The past few decades of toxicological research have demon- disruptors operate mechanistically (Rotchell and Ostrander, strated that it is difficult to establish a link between sublethal 2003), it often remains unclear as to how the subtle changes effects of contaminants measured on individuals and indices of caused by endocrine disruptors affect an individual’s perfor- population health (Mills and Chichester, 2005). Yet, endocrine mance in a natural setting, or how these changes relate to disruptors, which interrupt normal signaling processes within population-level responses. Ecological death is the phenomenon and between organisms, generally manifest themselves as subtle where animals are not overtly harmed by a contaminant, but changes in the reproductive system, molecular biology, physi- exposure causes alterations in their behavior or physiology such that the organism is unable to function ecologically as expected (Mesa et al., 1994; Scott and Sloman, 2004). ଝ Although the research described in this article has been funded wholly or in Behavior is often used as an endpoint in toxicological studies part by the United States Environmental Protection Agency through cooperative agreement R 829458 to CEER-GOM, it has not been subjected to the Agency’s because an animal’s behavior represents an integrated expres- required peer and policy review and therefore does not necessarily reflect the sion of its physiological response to its environment (Clotfelter views of the Agency and no official endorsement should be inferred. et al., 2004). Usually, toxicological studies that monitor behavior ∗ Corresponding author. Current address: Department of Fisheries and include behaviors that show disruption of certain physiological Wildlife, Natural Resources Building, Michigan State University, East Lansing, mechanisms. For example, in striped mullet (Mugil cephalus), MI 48824-1222, USA. Tel.: +1 517 432 7771; fax: +1 517 432 1699. E-mail address: [email protected] (C.A. Murphy). methylmercury (MeHg) was shown to decrease levels of sero- 1 Current address: Institute of Estuarine and Coastal Studies (IECS), Univer- tonin, which were then associated with the progressive loss of sity of Hull, Hull HU6 7RX, UK. motor control (Thomas et al., 1981). Increasingly, toxicologi- 0166-445X/$ – see front matter © 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.aquatox.2007.12.009 C.A. Murphy et al. / Aquatic Toxicology 86 (2008) 470–484 471 cal experiments have been focused on behavioral endpoints that late behavioral effects associated with contaminant exposure. reflect how a contaminant will affect ecological health or fit- Atlantic croaker spawn offshore, and the egg, yolk-sac larval, ness of the exposed individual (Scott and Sloman, 2004). The and early planktonic feeding larval stages experience high mor- behaviors selected for such ecological health experiments usu- tality rates while in the ocean on their way to their juvenile ally relate to an animal’s foraging ability (Weis et al., 2001), nursery grounds in estuaries (Diamond et al., 1999). Early life predator-evasion skills (Weis and Weis, 1995a,b; Zhou and Weis, stages that exhibit high and variable mortality rates can be impor- 1998; Faulk et al., 1999; McCarthy et al., 2003; Alvarez and tant regulators of year class strength and recruitment dynamics Fuiman, 2006), or reproduction (Bjerselius et al., 2001). In only in fish (Houde, 1989; Leggett and Deblois, 1994). The ocean a few instances, the effects on predator and prey interactions larval stage of croaker may be especially sensitive to behavioral were related to cohort-level responses (e.g., Alvarez and Fuiman, effects because early ocean larvae have to undergo sufficient 2005) or to population-level responses (e.g., Fleeger et al., 2003; physiological and behavioral development to exhibit successful Alvarez et al., 2006). foraging (exogenous feeding) and predator evasion. Laboratory Individual-based models (IBMs) are well suited to link labo- experiments have demonstrated how contaminants affect the for- ratory behavioral results to population-relevant indices (Rose et aging and predator-evasion behaviors of Atlantic croaker larvae al., 1999, 2003). Individual-based models track each individual; (Faulk et al., 1999; McCarthy et al., 2003; Alvarez et al., 2006). the sum over all individuals at any given time is the cohort or the population. Laboratory toxicological effects that are reported 2.2. Laboratory studies for the individual can be easily imposed on processes repre- sented in an individual-based model, either directly or first via We used the results of a laboratory study of the effects of some conversion from a statistical model. If the IBM simulates MeHg on larval croaker swimming speed and predator-evasion the appropriate processes, sublethal effects, such as changes in skills (see Alvarez et al., 2006). Briefly, uncontaminated and behavior, can be imposed on individuals and the IBM used to contaminated food was fed to adult female croaker, who were predict the population-level response. then spawned and their eggs were collected and analyzed for In this paper, we used an IBM to scale the laboratory- MeHg. Eggs from individual adults were grouped together based measured effects of maternally transferred MeHg on larval on the MeHg concentrations measured in their eggs: control Atlantic croaker (Micropogonias undulatus) behavior to larval- (<0.05 ␮gg−1 egg), low-dose exposure (0.01–3.5 ␮gg−1), and stage survival and duration. MeHg has been shown to reduce high-dose exposure (>3.5 ␮gg−1). Eggs from the control, low, serotonin levels in the brain of fish, inhibit normal development and high MeHg dose treatments were then hatched in the labo- of the hypothalamic serotonergic system (Tsai et al., 1995), and ratory and resulting larvae were evaluated for their growth rates to affect locomotor activity and impair prey capture abilities and survival skills. At days 1, 3, 6, 11, and 17 after hatching, (Smith and Weis, 1997). In this analysis, we use the experimental larvae were removed from their rearing tanks and their lengths results of Alvarez et al. (2006), who documented MeHg effects were measured. Day 3 larvae had just completed yolk absorp- on the locomotor activity and predator-evasion skills of larval tion and were labeled as “yolk” larvae. The oil globule had been croaker. Expressing these behavioral effects as changes in stage completely absorbed by day 6, and these larvae were labeled as duration and stage survival helps to place these sublethal effects “oil” larvae. Days 11 and 17 larvae marked 4 and 11 days after into an ecological context. Stage duration and survival are fun- oil absorption, subsequently these larvae were labeled “oil+4” damental parameters in life tables and matrix projection models and “oil+11.” of population dynamics. We simulate the effects of a low and Larval fish from each of the age classes (yolk, oil, oil+4 and high dose of MeHg on larval-stage