Grantee Research Project Results
Final Report: Integrated Modeling Approaches to Support Systems-Based Ecological Risk Assessment
EPA Grant Number: R835795Title: Integrated Modeling Approaches to Support Systems-Based Ecological Risk Assessment
Investigators: von Stackelberg, Katherine Ellen , Stark, John D , Landis, Wayne G. , Sunderland, Elsie M.
Institution: President and Fellow of Harvard College , Western Washington University , Washington State University
Current Institution: President and Fellow of Harvard College , Washington State University , Western Washington University
EPA Project Officer: Aja, Hayley
Project Period: June 1, 2015 through May 31, 2017 (Extended to December 31, 2018)
Project Amount: $651,708
RFA: Systems-Based Research for Evaluating Ecological Impacts of Manufactured Chemicals (2014) RFA Text | Recipients Lists
Research Category: Chemical Safety for Sustainability
Objective:
In a complex and changing environment and with an increasing emphasis on sustainability of coupled human-environment systems, reductionist approaches to environmental management that fail to consider feedback loops, multiple stressors, and spatial and temporal characteristics of exposures and populations no longer suffice. We demonstrate how existing data and models can be integrated using a Bayesian Network – Relative Risk Modeling (BN-RRM) framework that explicitly links molecular initiating events to regulatory outcomes of interest. The flexible approach allows multiple stressors linked to multiple outcomes. The objectives of the research are to: A. Develop quantitative adverse outcome pathways (qAOPs) for at least one legacy contaminant (e.g., organophosphate pesticides) and one emerging contaminant (e.g., per- and polyfluorinated compounds) and develop exposure-response profiles for use in a larger BN-RRM; B. Develop several ecological risk case studies demonstrating the integration and application of one or more underlying process models to synthesize and integrate available data across levels of biological organization, including exposure estimates, quantitative AOPs, non-chemical stressors, and population models; and, C. Apply the BN-RRM framework to demonstrate how the approach generates quantitative predictions of potential ecological risk impacts at scales relevant to policy development and regulatory decision making.
Summary/Accomplishments (Outputs/Outcomes):
The research team developed several BN-RRMs within an ecological risk case study focused on Chinook salmon populations in the Pacific Northwest and exposures to a single organophosphate pesticide (OP), two environmental stressors as proxy for habitat and water quality (e.g., temperature and dissolved oxygen), the combination of these, and finally, exposure to multiple OPs (e.g., chemical mixtures). The BN-RRMs were developed in Netica, publicly available software that can read any Netica files and is free for models with a limited number of individual nodes. Exposures to OPs were modeled using an established AOP for acetylcholinesterase (AChE) inhibition, the primary documented pathway by which OPs exert their toxic action. Neurotoxicity induced by OPs can reduce predator avoidance, homing fidelity, swimming speed and foraging ability in exposed fish, which translates to increases in mortality in adult and juvenile fish, an outcome directly relevant to the population model. The population model was conducted based on a single salmon population, and extended to include a consideration of metapopulations with four watershed areas in the Pacific Northwest.Models were developed for a baseline case of no stressors, exposures to single OPs, mixtures of OPs (e.g., chlorpyrifos, malathion, and diazinon), environmental stressors (e.g., dissolved oxygen and temperature), and a 20% known population decline (irrespective of cause and used as a benchmark against which to evaluate the OP and environmental stressor exposures). Risk was defined as the probability of falling below the published management benchmark of a Chinook population of at least 500,000 fish. Baseline risk (assuming no stressor exposures) was 2%, in other words, there is a 2% predicted probability of fish populations falling below 500,000. Results show that all scenarios save for baseline predict declines in Chinook salmon population size distributions that fall well below established management goals for salmon populations in Puget Sound. Environmental stressors are associated with the highest predicted population declines (anywhere between 51% and 90%, depending on the watershed and season), and exposure to OPs (in the absence of environmental stressors) is associated with probabilities of falling below 500,000 ranging from 3% to 27%, depending on season and watershed. Mixture effects were also evident from these models. A BN-RRM based on the binary combination of exposures to malathion and diazinon resulted in a 67% probability of falling below the benchmark population size of 500,000 assuming an additive model. A synergistic model of exposures to diazinon and malathion (as opposed to additive) resulted in a predicted risk of 75%, and 8% increase from the additive model, suggesting a synergistic combination. Chlorpyrifos and malathion synergistically resulted in a predicted risk of 69%. No other synergisms were noted, suggesting that malathion exposures, in conjunction with other OP exposures, result in greater than additive risks to salmon populations. In principle, developing a quantitative AOP (qAOP) requires an understanding of the mathematical relationships between each molecular initiating event (MIE) and key event (KE) in the adverse outcome pathway, and then an understanding of how exposure to one or more environmental stressors alters those relationships. In practice, AOPs are virtually never quantitative, and there are no examples of use of the AOP framework to estimate actual ecological risks that the research team could find in the literature. Even the AOPWiki (www.aopwiki.org), which represents a formal online community of practice for AOP development, has no examples of qAOPs. Some of the key challenges that occur when quantifying AOPs includes the role of adaptation, variability in responses across individuals, variability in responses within individuals, the role of compensating mechanisms and interactions with other endogenous and exogenous constituents, feedback mechanisms, and pathway saturation. Some of these responses require a more complex conceptual model of the relationship between MIEs and KEs, and others require a better understanding of uncertainty and variability, for example, in a probabilistic framework. The research team found no examples of a quantitative treatment of any of these potential modifiers to AOPs – even in the absence of a particular stressor exposure – in the literature. The Pacific Northwest Chinook salmon database was selected as a representative case study for a legacy contaminant for which data were available with which to quantify relationshipsbetween MIEs and KEs in an AOP context, as well as quantitative exposure-response data for phenotypic behavioral outcomes in fish relevant for population sustainability. In addition, many years worth of exposure data are available for OPs in the Puget Sound system, and Chinook populations are closely monitored. Several existing population models for these salmon exist, reducing the uncertainty in the baseline parameters of fecundity and mortality, necessary inputs for the population models.
Per- and poly-fluorinated compound (PFAS) exposures and outcomes proved more circumspect. This was a case study for which the goal was to evaluate the feasibility of using a BN-RRM for a non-legacy or emerging contaminant for which data availability was likely to be far less than for a legacy contaminant. However, the complexity of exposures to PFAS make this approach very challenging. Immunotoxicity (specifically reductions in antibody production and hypersensitivity) is a well-established outcome for exposure to one of the most common PFASs -- perfluorooctane sulfonate (PFOS). PFOS is also shown to be an endocrine disruptor in the US EPA ToxCast program. The other most common PFAS, perfluorooctanoic acid (PFOA) has not been shown to be an endocrine disruptor. The most sensitive assays and associated AC50s (lowest concentration leading to activation of the pathway) for PFOA are related to CYP – activation of cytochrome P450, as well as phosphatase proteins implicated in autoimmune disorders. PFOS activated these assays at concentrations nearly two orders of magnitude lower than PFOA.
Conclusions:
In principle, if an AOP can be established, and it is understood how exposure to the contaminant(s) initiates key events, then it should be possible to use the BN framework to link key events where each key event is represented by a node, conditioned on the previous node. The challenge is that the suite of in vitro tests are conducted in isolation, so the induction of any key event is not systemically linked to any previous event. Ideally, case learning could be used to establish conditional probabilities across nodes, but because the underlying data are not linked in any way, there is little justification for that in a quantitative sense.
Journal Articles on this Report : 2 Displayed | Download in RIS Format
Other project views: | All 13 publications | 2 publications in selected types | All 2 journal articles |
---|
Type | Citation | ||
---|---|---|---|
|
Landis W, Chu V, Graham S, Jarris M, Markiewicz A, Mitchell C, von Stackelberg K, Stark J. Integration of Chlorpyrifos Acetylcholinesterase Inhibition, Water Temperature, and Dissolved Oxygen Concentration into a Regional Scale Multiple Stressor Risk Assessment Estimating Risk to Chinook Salmon. INTEGRATED ENVIORNMENTAL ASSESSMENT AND MANAGEMENT 2019;16(1):28-42. |
R835795 (Final) |
Exit |
|
Mitchell CJ, Lawrence E, Chu VR, Harris MJ, Landis WG, von Stacekberg KE, Stark JD. Integrating Metapopulation Dynamics into a Bayesian Network Relative Risk Model:Assessing Risk of Pesticides to Chinook Salmon (Oncorhynchus tshawytscha) in an Ecological Context. INTEGRATED ENVIORNMENTAL ASSESSMENT AND MANAGEMENT2020;17(1):95-109. |
R835795 (Final) |
Exit |
Supplemental Keywords:
uncertainty, systematic review, influence diagrams, Bayes network model, probabilistic, decision analysis, integrated assessment modelsProgress and Final Reports:
Original AbstractThe perspectives, information and conclusions conveyed in research project abstracts, progress reports, final reports, journal abstracts and journal publications convey the viewpoints of the principal investigator and may not represent the views and policies of ORD and EPA. Conclusions drawn by the principal investigators have not been reviewed by the Agency.