Michael S. Breen

Endocrine disrupting chemicals in fish: Developing exposure indicators and predictive models of effects based on mechanism of action

Knowledge of possible toxic mechanisms (or modes) of action (MOA) of chemicals can provide valuable insights as to appropriate methods for assessing exposure and effects, thereby reducing uncertainties related to extrapolation across species, endpoints and chemical structure. However, MOA-based testing seldom has been used for assessing the ecological risk of chemicals. This is in part because past regulatory mandates have focused more on adverse effects of chemicals (reductions in survival, growth or reproduction) than the pathways through which these effects are elicited. A recent departure from this involves endocrine-disrupting chemicals (EDCs), where there is a need to understand both MOA and adverse outcomes. To achieve this understanding, advances in predictive approaches are required whereby mechanistic changes caused by chemicals at the molecular level can be translated into apical responses meaningful to ecological risk assessment. In this paper we provide an overview and illustrative results from a large, integrated project that assesses the effects of EDCs on two small fish models, the fathead minnow (Pimephales promelas) and zebrafish (Danio rerio). For this work a systems-based approach is being used to delineate toxicity pathways for 12 model EDCs with different known or hypothesized toxic MOA. The studies employ a combination of state-of-the-art genomic (transcriptomic, proteomic, metabolomic), bioinformatic and modeling approaches, in conjunction with whole animal testing, to develop response linkages across biological levels of organization. This understanding forms the basis for predictive approaches for species, endpoint and chemical extrapolation. Although our project is focused specifically on EDCs in fish, we believe that the basic conceptual approach has utility for systematically assessing exposure and effects of chemicals with other MOA across a variety of biological systems.

The near-road exposures and effects of urban air pollutants study (NEXUS): Study design and methods

The Near-Road Exposures and Effects of Urban Air Pollutants Study (NEXUS) was designed to examine the relationship between near-roadway exposures to air pollutants and respiratory outcomes in a cohort of asthmatic children who live close to major roadways in Detroit, Michigan USA. From September 2010 to December 2012 a total of 139 children with asthma, ages 6-14, were enrolled in the study on the basis of the proximity of their home to major roadways that carried different amounts of diesel traffic. The goal of the study was to investigate the effects of traffic-associated exposures on adverse respiratory outcomes, biomolecular markers of inflammatory and oxidative stress, and how these exposures affect the frequency and severity of respiratory viral infections in a cohort of children with asthma. An integrated measurement and modeling approach was used to quantitatively estimate the contribution of traffic sources to near-roadway air pollution and evaluate predictive models for assessing the impact of near-roadway pollution on childrens exposures. Two intensive field campaigns were conducted in Fall 2010 and Spring 2011 to measure a suite of air pollutants including PM2.5 mass and composition, oxides of nitrogen (NO and NO2), carbon monoxide, and black carbon indoors and outdoors of 25 participants homes, at two area schools, and along a spatial transect adjacent to I-96, a major highway in Detroit. These data were used to evaluate and refine models to estimate air quality and exposures for each child on a daily basis for the health analyses. The study design and methods are described, and selected measurement results from the Fall 2010 field intensive are presented to illustrate the design and successful implementation of the study. These data provide evidence of roadway impacts and exposure variability between study participants that will be further explored for associations with the health measures.

Mechanistic Computational Model of Ovarian Steroidogenesis to Predict Biochemical Responses to Endocrine Active Compounds

Sex steroids, which have an important role in a wide range of physiological and pathological processes, are synthesized primarily in the gonads and adrenal glands through a series of enzyme-mediated reactions. The activity of steroidogenic enzymes can be altered by a variety of endocrine active compounds (EAC), some of which are therapeutics and others that are environmental contaminants. A steady-state computational model of the intraovarian metabolic network was developed to predict the synthesis and secretion of testosterone (T) and estradiol (E2), and their responses to EAC. Model predictions were compared to data from an in vitro steroidogenesis assay with ovary explants from a small fish model, the fathead minnow. Model parameters were estimated using an iterative optimization algorithm. Model-predicted concentrations of T and E2 closely correspond to the time–course data from baseline (control) experiments, and dose–response data from experiments with the EAC, fadrozole (FAD). A sensitivity analysis of the model parameters identified specific transport and metabolic processes that most influence the concentrations of T and E2, which included uptake of cholesterol into the ovary, secretion of androstenedione (AD) from the ovary, and conversions of AD to T, and AD to estrone (E1). The sensitivity analysis also indicated the E1 pathway as the preferred pathway for E2 synthesis, as compared to the T pathway. Our study demonstrates the feasibility of using the steroidogenesis model to predict T and E2 concentrations, in vitro, while reducing model complexity with a steady-state assumption. This capability could be useful for pharmaceutical development and environmental health assessments with EAC.

Air Change Rates and Interzonal Flows in Residences, and the Need for Multi-Zone Models for Exposure and Health Analyses

Air change rates (ACRs) and interzonal flows are key determinants of indoor air quality (IAQ) and building energy use. This paper characterizes ACRs and interzonal flows in 126 houses, and evaluates effects of these parameters on IAQ. ACRs measured using weeklong tracer measurements in several seasons averaged 0.73 ± 0.76 h−1 (median = 0.57 h−1, n = 263) in the general living area, and much higher, 1.66 ± 1.50 h−1 (median = 1.23 h−1, n = 253) in bedrooms. Living area ACRs were highest in winter and lowest in spring; bedroom ACRs were highest in summer and lowest in spring. Bedrooms received an average of 55 ± 18% of air from elsewhere in the house; the living area received only 26 ± 20% from the bedroom. Interzonal flows did not depend on season, indoor smoking or the presence of air conditioners. A two-zone IAQ model calibrated for the field study showed large differences in pollutant levels between the living area and bedroom, and the key parameters affecting IAQ were emission rates, emission source locations, air filter use, ACRs, interzonal flows, outdoor concentrations, and PM penetration factors. The single-zone models that are commonly used for residences have substantial limitations and may inadequately represent pollutant concentrations and exposures in bedrooms and potentially other environments other where people spend a substantial fraction of time.

GPS-based microenvironment tracker (MicroTrac) model to estimate time–location of individuals for air pollution exposure assessments: Model evaluation in central North Carolina

A critical aspect of air pollution exposure assessment is the estimation of the time spent by individuals in various microenvironments (ME). Accounting for the time spent in different ME with different pollutant concentrations can reduce exposure misclassifications, while failure to do so can add uncertainty and bias to risk estimates. In this study, a classification model, called MicroTrac, was developed to estimate time of day and duration spent in eight ME (indoors and outdoors at home, work, school; inside vehicles; other locations) from global positioning system (GPS) data and geocoded building boundaries. Based on a panel study, MicroTrac estimates were compared with 24-h diary data from nine participants, with corresponding GPS data and building boundaries of home, school, and work. MicroTrac correctly classified the ME for 99.5% of the daily time spent by the participants. The capability of MicroTrac could help to reduce the time–location uncertainty in air pollution exposure models and exposure metrics for individuals in health studies.

Developing Predictive Approaches to Characterize Adaptive Responses of the Reproductive Endocrine Axis to Aromatase Inhibition: II. Computational Modeling

Endocrine-disrupting chemicals can affect reproduction and development in humans and wildlife. We developed a computational model of the hypothalamic-pituitary-gonadal (HPG) axis in female fathead minnows to predict dose-response and time-course (DRTC) behaviors for endocrine effects of the aromatase inhibitor, fadrozole (FAD). The model describes adaptive responses to endocrine stress involving regulated secretion of a generic gonadotropin (LH/FSH) from the hypothalamic-pituitary complex. For model development, we used plasma 17β-estradiol (E2) concentrations and ovarian cytochrome P450 (CYP) 19A aromatase mRNA data from two time-course experiments, each of which included both an exposure and a depuration phase, and plasma E2 data from a third 4-day study. Model parameters were estimated using E2 concentrations for 0, 0.5, and 3 µg/l FAD exposure concentrations, and good fits to these data were obtained. The model accurately predicted CYP19A mRNA fold changes for controls and three FAD doses (0, 0.5, and 3 µg/l) and plasma E2 dose response from the 4-day study. Comparing the model-predicted DRTC with experimental data provided insight into how the feedback control mechanisms in the HPG axis mediate these changes: specifically, adaptive changes in plasma E2 levels occurring during exposure and overshoot occurring postexposure. This study demonstrates the value of mechanistic modeling to examine and predict dynamic behaviors in perturbed systems. As this work progresses, we will obtain a refined understanding of how adaptive responses within the vertebrate HPG axis affect DRTC behaviors for aromatase inhibitors and other types of endocrine-active chemicals and apply that knowledge in support of risk assessments.

A modeling framework for characterizing near-road air pollutant concentration at community scales

In this study, we combine information from transportation network, traffic emissions, and dispersion model to develop a framework to inform exposure estimates for traffic-related air pollutants (TRAPs) with a high spatial resolution. A Research LINE source dispersion model (R-LINE) is used to model multiple TRAPs from roadways at Census-block level for two U.S. regions. We used a novel Space/Time Ordinary Kriging (STOK) approach that uses data from monitoring networks to provide urban background concentrations. To reduce the computational burden, we developed and applied the METeorologically-weighted Averaging for Risk and Exposure (METARE) approach with R-LINE, where a set of selected meteorological data and annual average daily traffic (AADT) are used to obtain annual averages. Compared with explicit modeling, using METARE reduces CPU-time by 88-fold (46.8h versus 32min), while still retaining accuracy of exposure estimates. We show two examples in the Piedmont region in North Carolina (~105,000 receptors) and Portland, Maine (~7000 receptors) to characterize near-road air quality. Concentrations for NOx, PM2.5, and benzene in Portland drop by over 40% within 200m away from the roadway. The concentration drop in North Carolina is less than that in Portland, as previously shown in an observation-based study, showing the robustness of our approach. Heavy-duty diesel vehicles (HDDV) contribute over 55% of NOx and PM2.5 near interstate highways, while light-duty gasoline vehicles (LDGV) contribute over 50% of benzene to urban areas where multiple roadways intersect. Normalized mean error (NME) between explicit modeling and METARE in Portland ranges from 12.6 to 14.5% and normalized mean bias (NMB) ranges from -12.9 to -11.2%. When considering a static emission rate (i.e. the emission does not have temporal variability), both NME and NMB improved (10.5% and -9.5%). Modeled concentrations in Detroit, Michigan at an array of near-road monitors are within a factor of 2 of observed values for CO but not NOx.

Air Pollution Exposure Model for Individuals (EMI) in Health Studies: Evaluation for Ambient PM2.5 in Central North Carolina

Air pollution health studies of fine particulate matter (diameter ≤2.5 μm, PM2.5) often use outdoor concentrations as exposure surrogates. Failure to account for variability of indoor infiltration of ambient PM2.5 and time indoors can induce exposure errors. We developed and evaluated an exposure model for individuals (EMI), which predicts five tiers of individual-level exposure metrics for ambient PM2.5 using outdoor concentrations, questionnaires, weather, and time-location information. We linked a mechanistic air exchange rate (AER) model to a mass-balance PM2.5 infiltration model to predict residential AER (Tier 1), infiltration factors (Tier 2), indoor concentrations (Tier 3), personal exposure factors (Tier 4), and personal exposures (Tier 5) for ambient PM2.5. Using cross-validation, individual predictions were compared to 591 daily measurements from 31 homes (Tiers 1-3) and participants (Tiers 4-5) in central North Carolina. Median absolute differences were 39% (0.17 h(-1)) for Tier 1, 18% (0.10) for Tier 2, 20% (2.0 μg/m(3)) for Tier 3, 18% (0.10) for Tier 4, and 20% (1.8 μg/m(3)) for Tier 5. The capability of EMI could help reduce the uncertainty of ambient PM2.5 exposure metrics used in health studies.

Modeling Spatial and Temporal Variability of Residential Air Exchange Rates for the Near-Road Exposures and Effects of Urban Air Pollutants Study (NEXUS)

Air pollution health studies often use outdoor concentrations as exposure surrogates. Failure to account for variability of residential infiltration of outdoor pollutants can induce exposure errors and lead to bias and incorrect confidence intervals in health effect estimates. The residential air exchange rate (AER), which is the rate of exchange of indoor air with outdoor air, is an important determinant for house-to-house (spatial) and temporal variations of air pollution infiltration. Our goal was to evaluate and apply mechanistic models to predict AERs for 213 homes in the Near-Road Exposures and Effects of Urban Air Pollutants Study (NEXUS), a cohort study of traffic-related air pollution exposures and respiratory effects in asthmatic children living near major roads in Detroit, Michigan. We used a previously developed model (LBL), which predicts AER from meteorology and questionnaire data on building characteristics related to air leakage, and an extended version of this model (LBLX) that includes natural ventilation from open windows. As a critical and novel aspect of our AER modeling approach, we performed a cross validation, which included both parameter estimation (i.e., model calibration) and model evaluation, based on daily AER measurements from a subset of 24 study homes on five consecutive days during two seasons. The measured AER varied between 0.09 and 3.48 h−1 with a median of 0.64 h−1. For the individual model-predicted and measured AER, the median absolute difference was 29% (0.19 h‑1) for both the LBL and LBLX models. The LBL and LBLX models predicted 59% and 61% of the variance in the AER, respectively. Daily AER predictions for all 213 homes during the three year study (2010–2012) showed considerable house-to-house variations from building leakage differences, and temporal variations from outdoor temperature and wind speed fluctuations. Using this novel approach, NEXUS will be one of the first epidemiology studies to apply calibrated and home-specific AER models, and to include the spatial and temporal variations of AER for over 200 individual homes across multiple years into an exposure assessment in support of improving risk estimates.

Mechanistic Computational Model of Steroidogenesis in H295R Cells: Role of Oxysterols and Cell Proliferation to Improve Predictability of Biochemical Response to Endocrine Active Chemical—Metyrapone

The human adrenocortical carcinoma cell line H295R is being used as an in vitro steroidogenesis screening assay to assess the impact of endocrine active chemicals (EACs) capable of altering steroid biosynthesis. To enhance the interpretation and quantitative application of measurement data in risk assessments, we are developing a mechanistic computational model of adrenal steroidogenesis in H295R cells to predict the synthesis of steroids from cholesterol (CHOL) and their biochemical response to EACs. We previously developed a deterministic model that describes the biosynthetic pathways for the conversion of CHOL to steroids and the kinetics for enzyme inhibition by the EAC, metyrapone (MET). In this study, we extended our dynamic model by (1) including a cell proliferation model supported by additional experiments and (2) adding a pathway for the biosynthesis of oxysterols (OXY), which are endogenous products of CHOL not linked to steroidogenesis. The cell proliferation model predictions closely matched the time-course measurements of the number of viable H295R cells. The extended steroidogenesis model estimates closely correspond to the measured time-course concentrations of CHOL and 14 adrenal steroids both in the cells and in the medium and the calculated time-course concentrations of OXY from control and MET-exposed cells. Our study demonstrates the improvement of the extended, more biologically realistic model to predict CHOL and steroid concentrations in H295R cells and medium and their dynamic biochemical response to the EAC, MET. This mechanistic modeling capability could help define mechanisms of action for poorly characterized chemicals for predictive risk assessments.