A quantitative source-to-outcome case study to demonstrate the

This work presents a case study of a hypothetical contaminated site to demonstrate a. 18 quantitative approach for implementing the AEP framework and ...
0 downloads 0 Views 846KB Size
Subscriber access provided by LUNDS UNIV

Ecotoxicology and Human Environmental Health

A quantitative source-to-outcome case study to demonstrate the integration of human health and ecological endpoints using the Aggregate Exposure Pathway and Adverse Outcome Pathway frameworks David E Hines, Rory B. Conolly, and Annie M Jarabek Environ. Sci. Technol., Just Accepted Manuscript • DOI: 10.1021/acs.est.9b04639 • Publication Date (Web): 22 Aug 2019 Downloaded from pubs.acs.org on August 24, 2019

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 33

Environmental Science & Technology

1 2 3

A quantitative source-to-outcome case study to demonstrate the integration of human health and ecological endpoints using the Aggregate Exposure Pathway and Adverse Outcome Pathway frameworks

4

David E. Hines1, Rory B. Conolly*1, and Annie M. Jarabek2

5 6 7 8 9 10 11 12

Affiliations: 1U.S. Environmental Protection Agency, Office of Research and Development, National Health and Environmental Effects Research Laboratory, Integrated Systems Toxicology Division, Research Triangle Park, North Carolina 27709, United States 2U.S. Environmental Protection Agency, Office of Research and Development, National Center for Environmental Assessment, Research Triangle Park, North Carolina 27709, United States *Corresponding author; phone: 919-541-3350, email: [email protected] Abstract:

13

Exposure to environmental contaminants can lead to adverse outcomes in both human and non-human

14

receptors. The Aggregate Exposure Pathway (AEP) and Adverse Outcome Pathway (AOP) frameworks

15

can mechanistically inform cumulative risk assessment for human health and ecological endpoints by

16

linking together environmental transport and transformation, external exposure, toxicokinetics, and

17

toxicodynamics. This work presents a case study of a hypothetical contaminated site to demonstrate a

18

quantitative approach for implementing the AEP framework and linking this framework to AOPs. We

19

construct an AEP transport and transformation model, then quantify external exposure pathways for

20

humans, fishes, and small herbivorous mammals at the hypothetical site. A Monte Carlo approach was

21

used to address parameter variability. Source apportionment was quantified for each species and

22

published pharmacokinetic models were used to estimate internal target site exposure from external

23

exposures. Published dose-response data for a multi-species AOP network were used to interpret AEP

24

results in the context of species-specific effects. This work demonstrates 1) the construction, analysis,

25

and application of a quantitative AEP model, 2) the utility of AEPs for organizing mechanistic exposure

26

data and highlighting data gaps, and 3) the advantages provided by a source-to-outcome construct for

27

leveraging exposure data and to aid transparency regarding assumptions.

28 29

Key Words: Cumulative Risk Assessment, Ecological Network Analysis, Source apportionment, PBPK models, Target Site Exposure, Aggregate Exposure Pathway, Adverse Outcome Pathway 1 ACS Paragon Plus Environment

Environmental Science & Technology

30

Acronyms:

31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50

Absorption, Distribution, Metabolism and Elimination (ADME) Adverse Outcome (AO) Adverse Outcome Pathway (AOP) Aggregate Exposure Pathway (AEP) Cumulative Risk Assessment (CRA) European Union System for the Evaluation of Substances (EUSES) In Vitro-In Vivo Extrapolation (IVIVE) Key Event (KE) Key Exposure State (KES) Modeling ENvironment for TOtal Risk (MENTOR) Molecular Initiating Event (MIE) National Research Council (NRC) Perchlorate anion (ClO4-) Physiologically Based Pharmacokinetic (PBPK) Reference dose (RfD) Sodium-Iodide Symporter (NIS) Target site exposure (TSE) Triiodothyronine (T3) Thyroxin (T4) Thyroid Stimulating Hormone (TSH)

2 ACS Paragon Plus Environment

Page 2 of 33

Page 3 of 33

51

Environmental Science & Technology

1. Introduction

52

Risk assessors integrate knowledge about exposure and toxicity pathways to evaluate the

53

potential risk of adverse outcomes (AOs) from environmental contaminants and other stressors1.

54

Mechanistic approaches provide advantages for risk assessors because they describe the causal

55

pathways from sources of contamination to AOs and therefore can facilitate a science-based evaluation

56

of the current knowledge, data gaps, and uncertainties in risk assessment results2. While human health

57

outcomes are often of primary concern, non-human species are also exposed to contaminants and may

58

be adversely affected at different exposure levels than humans. The National Research Council (NRC)1

59

emphasized the importance of considering both human health and ecological endpoints in cumulative

60

risk assessments (CRAs), which evaluate the combined effects of multiple chemicals and non-chemical

61

stressors on organisms for the purposes of identifying AO risk3,4. However, the mechanisms resulting in

62

AOs vary across organisms and contaminants. Thus, integrating mechanistic data from human health

63

and ecological endpoints in support of site-based risk assessments and CRAs is challenging5.

64

Exposure science is a critical part of risk assessments that emphasizes the evaluation of

65

interactions between relevant physical, chemical or biologic stressors and their receptors6 and is

66

evolving rapidly given the emerging technologies that allow measurement in various biological samples

67

and mechanistic characterization of the exposome7. Over the past several decades, researchers have

68

developed approaches for quantifying and analyzing the mechanisms behind these interactions8,9. For

69

example, Georgopoulos and Lioy (1994)10 presented a framework and twelve-step process to guide the

70

collection and management of data characterizing exposure and dose in humans. This theoretical

71

framework was later adapted for use in computational studies and guided the development of

72

implementations such as the Modeling ENvironment for TOtal Risk (MENTOR), a mechanistic source-to-

73

dose toolbox for exposure assessment11. Environmental risk assessment case studies have incorporated

74

probabilistic approaches for exposure prediction based on distributions of parameters for mechanistic 3 ACS Paragon Plus Environment

Environmental Science & Technology

75

data, but these approaches are rarely implemented in regulatory contexts with multi-media assessment

76

tools such as the European Union System for the Evaluation of Substances (EUSES)9. Nevertheless, the

77

insight provided by mechanistic exposure assessment tools is critically important for predicting relevant

78

AOs from toxicity data12, as the exposure pathways leading to human health and ecological AOs may

79

vary across species. Additionally, absorption, distribution, metabolism and elimination (ADME)

80

processes differ across toxicants and organisms, leading to variation in the internal concentration, or

81

target site exposure (TSE), of a contaminant that results from external exposures13,14,15. Tools such as

82

physiologically based pharmacokinetic (PBPK) models can provide quantitative, mechanistic descriptions

83

of ADME processes to predict TSEs, and therefore play a key role in associating external exposures with

84

internal dose estimates for multiple species in CRAs16,17. Further developing tools to understand and

85

analyze the mechanisms behind external exposure pathways and ADME processes is essential if risk

86

assessors are to integrate human health and ecological endpoints into CRAs for regulatory

87

applications7,18.

88

Toxicologists face similar challenges to those faced by exposure scientists regarding

89

physiological differences among organisms when integrating human health and ecological endpoints.

90

Understanding the mechanistic similarities governing ADME and tissue responses in organisms can,

91

however, make quantitative extrapolation and data comparison across species possible17,19,20. The

92

Adverse Outcome Pathway (AOP) framework21 organizes knowledge about these mechanisms by linking

93

a biological perturbation at the molecular level, termed a Molecular Initiating Event (MIE), to an AO at

94

the organism or population level. A TSE describes the concentration of toxicant at the MIE, and the

95

mechanisms leading from an MIE to an AO are represented by a series of causal Key Events (KEs) that

96

are empirically measurable steps along the AOP22. Quantitative AOP (qAOP) models, which provide

97

mathematical descriptions of the relationships among KEs, may aid in reducing uncertainties in cross-

98

species extrapolations and facilitate data integration in risk assessments23,24. Additionally, multiple AOPs 4 ACS Paragon Plus Environment

Page 4 of 33

Page 5 of 33

99 100 101

Environmental Science & Technology

with shared KEs can be combined into AOP networks of interacting pathways25, facilitating the organization of mechanistic toxicity data across organisms. While AOPs have been proposed for use in toxicity screening and risk assessment26,27,28, they do

102

not incorporate exposure data describing the events leading to activation of an MIE. Teeguarden et al.

103

(2016)16 introduced the Aggregate Exposure Pathway (AEP) framework, which utilizes existing concepts

104

in exposure science to complement the AOP framework by providing an analogous structure for

105

organizing exposure data. In the AEP framework, sources of contaminants are linked to TSEs through a

106

series of empirically measurable Key Exposure States (KES) that describe the environmental transport

107

and transformation of chemicals, as well as ADME interactions within organisms29,30. A variety of model

108

types, including transport and transformation models, could be used within this framework to

109

quantitatively evaluate the external exposure pathways within AEPs, and PBPK models can be applied to

110

estimate TSEs from these aggregate external exposures31. Thus, AEP and AOP models together

111

comprehensively cover conceptual site models16,17,32 and AEPs can facilitate quantitative analyses to

112

compliment AOPs for site-based or regional CRAs33,34.

113

Recently, Hines et al. (2018)35 presented a case study to demonstrate the utility of a joint AEP-

114

AOP construct for integrating mechanistic human health and ecological endpoints to inform community-

115

based CRAs. That case study used the perchlorate anion (ClO4-) along with the AOP for the inhibition of

116

iodide uptake into the thyroid by the sodium-iodide symporter (NIS) as a data-rich example because this

117

AOP is highly conserved and acts across vertebrate species36,37. Specifically, Hines et al. (2018)35 showed

118

how mechanistic toxicity data for NIS inhibition and subsequent effects could be organized across an

119

AOP network to identify similarities and differences in dose response relationships for a diverse set of

120

organisms and used a qualitative AEP description to show how this approach could be used to inform a

121

site-specific CRA. The current work expands on Hines et al. (2018)35 by developing a transport and

122

transformation AEP model for this ClO4- case study; to our knowledge this is the first such quantitative 5 ACS Paragon Plus Environment

Environmental Science & Technology

123

AEP model in the published literature. Drawing on available environmental data for ClO4-, we evaluate

124

species-specific exposures at a hypothetical contaminated site through transport and transformation

125

network modeling, then link external exposures to MIEs through TSEs using published pharmacokinetic

126

models. Network analysis is used to identify the source apportionment of exposure to three

127

representative groups of organisms: human, fish, and small herbivorous mammals to provide examples

128

of how this source-to-outcome analysis can help identify at-risk species or groups of organisms. Our

129

results 1) demonstrate the type of data needed to construct a quantitative AEP model, 2) provide an

130

example of how AEPs can organize knowledge to highlight data gaps, 3) illustrate how various exposure

131

pathways intersect at the MIE as the critical input to an AOP, 4) show how quantitative AEPs could be

132

applied in a risk assessment or regulatory setting, and 5) highlight the benefits of using this approach for

133

integrating of human health and ecological endpoints.

134

2. Methods

135

We construct a quantitative AEP model to demonstrate how this approach enables evaluation of

136

species-specific external exposures and source apportionment, then connect this AEP to AOP data from

137

Hines et al. (2018) to complete a source-to-outcome analysis. We us a mass balance transport and

138

transformation model of ClO4- at a hypothetical site to provide a complete example of the construction,

139

data requirements, and applications of this type of model.

140

2.1 Hypothetical contaminated site

141

We created a hypothetical contaminated site by piecing together data from multiple locations to

142

demonstrate the construction and application of the quantitative AEP model with the available data

143

(Tables S1 and S2 of the Supporting Information). This hypothetical site receives ClO4- contamination

144

from three sources: atmospheric deposition, surface water runoff, and groundwater input, and

145

considers six interacting environmental compartments that represent different components of the 6 ACS Paragon Plus Environment

Page 6 of 33

Page 7 of 33

Environmental Science & Technology

146

ecosystem (Figure 1). ClO4- is highly mobile and environmentally stable36, therefore we assumed that it

147

can be transported abiotically between these compartments, removed from the system by flowing

148

surface water or groundwater, or be absorbed and accumulate in aquatic plants, terrestrial grass, or

149

terrestrial shrubs36,38,39. The volumes of biotic and abiotic compartments in the model were held

150

constant at 10,000 m3 for surface water, 5,000 m3 for groundwater, 5,000 m3 for soil, 900 m3 for aquatic

151

vegetation, 500 m3 for terrestrial grass, and 1,000 m3 for terrestrial shrubs to provide physical

152

constraints that represented a contaminated site with a relatively large volume of surface water and

153

active biotic components.

154

Three types of organisms: humans, fishes, and small herbivorous mammals, were considered to

155

interact with the environmental compartments at the hypothetical contaminated site to provide a

156

diverse set of example organisms for demonstrating the integration of mechanistic data from multiple

157

species into CRAs (Figure 1). The species used were selected based on availability of published ClO4-

158

PBPK models and toxicity data for the NIS inhibition AOP. Fishes were represented by zebrafish (Danio

159

rerio) and small herbivorous mammals by rats (Rattus sp.) and meadow voles (Microtus sp.). These

160

species may or may not be relevant for actual sites but are used in this work as illustrative examples.

161

2.2 Quantitative AEP model

162

The AEP model was constructed from a set of six differential equations representing the

163

environmental compartments and potential exposure routes for the hypothetical contaminated site.

164

Transport of ClO4- between compartments was modeled as the product of a transport (or uptake) rate (𝑟

165

) and the amount of ClO4- in the donor compartment; equations for each compartment were then

166

constructed by combining the appropriate input and output flows. For example, the differential

167

equation for the surface water compartment (EQ.1) was constructed as

168

𝑑𝑆𝑊 𝑑𝑡

= 𝑆𝑊𝑖𝑛 + 𝐴𝑇𝑀𝑖𝑛 + 𝑟𝐺𝑊,𝑆𝑊𝐺𝑊 + 𝑟𝐴𝑃,𝑆𝑊𝐴𝑃 ― 𝑟𝑆𝑊,𝐺𝑊𝑆𝑊 ― 𝑟𝑆𝑊,𝐴𝑃𝑆𝑊 ― 𝑟𝑆𝑊,𝐵𝑆𝑊 EQ.1 7 ACS Paragon Plus Environment

Environmental Science & Technology

169

where 𝑆𝑊, 𝐺𝑊, and 𝐴𝑃 are the amounts (g) of ClO4- in the surface water, groundwater, and

170

aquatic plant compartments that are used to calculate ClO4- concentrations, 𝑟𝐺𝑊,𝑆𝑊 and 𝑟𝐴𝑃,𝑆𝑊 are the

171

daily transfer rates (d-1) from groundwater and aquatic plants to surface water, and 𝑟𝑆𝑊,𝐺𝑊, 𝑟𝑆𝑊,𝐴𝑃, and

172

𝑟𝑆𝑊,𝐵 are daily transfer rates (d-1) from surface water to groundwater, aquatic plants, and outside of the

173

system boundary (𝐵), respectively. ClO4- inputs (g/d) to the surface water compartment from exterior

174

surface water contamination (𝑆𝑊𝑖𝑛) and atmospheric deposition (𝐴𝑇𝑀𝑖𝑛) were held constant for this

175

hypothetical model. A full list of equations is available in the Supporting Information (Equations, Tables

176

S1 and S2). The model was constructed in R Version 3.2.2 using the enaR package Version 3.040.

177

The ClO4- uptake rate into the terrestrial grass compartment was taken from experimental

178

measurements reported by Susarla et el. (2000)38, while loss from the terrestrial grass compartment was

179

estimated based on bioaccumulation levels discussed in EPA (2002)36 and Hatzinger et al. (2015)41. The

180

AEP model tracks ClO4- flows among compartments as g ClO4-/d, and the amount in each compartment is

181

reported in g ClO4-. Due to the uncertainty inherent in parameterizing model components using

182

disparate literature sources, we assigned each parameter a range of plausible values based on the range

183

of reported values. A complete list of the parameters and their corresponding ranges is given in Tables

184

S1 and S2 of the Supporting Information.

185

Variability in model predictions caused by parameter variability was assessed using a Monte

186

Carlo approach. In this analysis, parameter values were determined by sampling uniformly from ranges

187

given in Tables S1 and S2 to generate 10,000 sets of initial conditions. For each initial condition, the

188

model was run to steady state and mass balance was verified for all model components. While we used

189

a uniform sampling distribution because we obtained parameters from unrelated studies to define each

190

model flow, it is important to note that other sampling distributions (e.g. truncated log-normal) may be

191

appropriate for actual sites if data are available. Each set of parameters represented a single ecosystem

8 ACS Paragon Plus Environment

Page 8 of 33

Page 9 of 33

Environmental Science & Technology

192

network model42,43 and the ranges in the steady state mass of ClO4- in the compartments are measures

193

of the impact of the variability of parameter values for the model predictions.

194

2.3 Exposure Scenarios

195

We applied three scenarios to the quantitative AEP model at the hypothetical site: mild,

196

moderate, and high ClO4- contamination. In the mild scenario, ClO4- was assumed to be present in all

197

compartments, but at concentrations below the ranges reported for documented sites36,41 and at least

198

an order of magnitude lower than the other hypothetical scenarios. The moderate scenario maintained

199

the same transfer rate parameter restrictions as the mild scenario, but increased ClO4- input through all

200

contamination sources (surface water, atmospheric deposition, and groundwater) by ten-fold, resulting

201

in an approximate increase to environmental concentrations of an order of magnitude. The high

202

contamination scenario maintained the conditions of the moderate scenario but included an additional

203

ten-fold increase over the moderate scenario in groundwater input (100 times greater than the mild

204

scenario) to simulate a ClO4- spill into this compartment. Environmental ClO4- concentrations in the high

205

contamination scenario were well above typical environmental concentrations36,41. Monte Carlo analysis

206

was conducted for each of the scenarios resulting in a total of 30,000 network model parameterizations

207

(10,000 for each scenario).

208

2.4 Evaluation of external exposure

209

Concentrations of ClO4- were calculated based on compartment volumes and predicted mass for

210

each parameterization and are shown in the Supporting Information (Table S3). The resulting

211

concentration distributions for each compartment represented the environmental concentrations in

212

each scenario. However, the external exposure pathways that are relevant to organisms depend not

213

only on environmental concentrations, but also on species-specific behavior44. For this exercise, in lieu

214

of species-specific exposure data, we assumed different behavioral regimens for each type of organism 9 ACS Paragon Plus Environment

Environmental Science & Technology

215

examined and used these regimens in conjunction with distributions of environmental concentrations to

216

demonstrate how external exposure to each organism can be predicted in each scenario. The drinking

217

of well water (groundwater) in pregnant women was considered as the only direct exposure pathway for

218

humans in this hypothetical case study to provide a simplified example, but other exposure pathways

219

such as food consumption may be relevant for a ClO4- risk assessment. Pregnant women were selected

220

as the target population because the fetus is a vulnerable life stage to developmental neurotoxicity

221

through NIS inhibition45,46,47,48. In this scenario, we assumed a body mass for a full-term pregnant female

222

of 72.3kg49 and a daily water intake of 2L, although other lifestages (such as women of childbearing age)

223

could also be considered50. For fishes, the ClO4- exposure was equivalent to the concentration of the

224

surface water because ClO4- is highly mobile, stable, and has poor complexing properties35,51. For small

225

mammals, we assumed herbivorous behavior, and that ClO4- exposure came from both consumption of

226

terrestrial grass as well as drinking of surface water. Additionally, we assumed a total daily grass and

227

water intake of 0.005 g/kg, and a body mass for the small mammal of 0.044 kg, which approximates that

228

of voles36,52. Voles (Microtus sp.) can rely heavily on moisture content in food for water53, therefore we

229

assumed a daily intake that consisted of 0.00475 g terrestrial grass (95%) and 0.00025 g surface water

230

(5%).

231

After combining distributions of environmental concentrations and exposure-related behaviors

232

to calculate external exposures (intake) for each species, we calculated the apportionment of each ClO4-

233

source (surface water, atmosphere, and groundwater) to each organism. Ecological Network Analysis, a

234

type of network modeling commonly used to track the transport and transformation of matter through

235

ecosystem networks54,55,56 facilitated this task. Specifically, we applied a subset of Ecological Network

236

Analysis called Environ Analysis, which can partition the flows in a network model to show the

237

contribution of each boundary flow (surface water, atmosphere, and groundwater) to each network

238

component through both direct and indirect pathways. For example, ClO4- in contaminated 10 ACS Paragon Plus Environment

Page 10 of 33

Page 11 of 33

Environmental Science & Technology

239

groundwater consumed by a human may have entered the ecosystem through groundwater inputs

240

(direct exposure pathway) or may have entered through surface water and was later transported to

241

groundwater (indirect exposure pathway); Environ Analysis quantifies the contribution of each exposure

242

pathway. This type of analysis differs from other network analyses, such as analysis of directed acyclic

243

graphs, that have recently been proposed to draw causal inferences for CRA57 because Ecological

244

Network Analysis techniques focus specifically on mass-conservative movement of material, including

245

cyclic and acyclic flows. Thus, Ecological Network Analysis is amenable to support quantitative analysis

246

of exposure pathways54. We analyzed each model parameterization, producing a distribution of source

247

apportionment predictions for each organism. A detailed description of the calculations of Environ

248

Analysis can be found in Patten (1978)58, Patten and Matis (1982)59, and Fath and Patten (1999)54.

249

2.5 Linking external exposure to TSE and MIE

250

The external exposures calculated from the AEP network and behavioral assumptions were

251

linked to TSEs to evaluate potential activation of MIEs in humans and small herbivorous mammals using

252

published PBPK models. For humans, we used a published model for NIS inhibition in a pregnant mother

253

and fetus from Lumen et al. (2013)49, which allowed for simultaneous adjustment of ClO4- and iodide

254

doses. We selected an iodide intake rate of 75 µg/iodide/d to represent an iodide deficient woman60, as

255

iodide deficiency results in more vulnerability to adverse effects from NIS inhibition36,61. PBPK Model

256

simulations were run using ClO4- doses corresponding to the 1st percentile, median, and 99th percentile

257

of external exposures predicted by the AEP model for humans in each scenario (Table 1), as well as using

258

a dose of zero to observe baseline model behavior. The PBPK model calculated the concentration of

259

free ClO4- in the serum (as the TSE), as well as the effects on iodide uptake into the thyroid at the NIS

260

(the MIE). Percent NIS inhibition was calculated by comparing the iodide uptake rate into the thyroid in

261

each simulation to the zero-toxicant simulation. For small herbivorous mammals, we used a published

262

rat PBPK model for NIS inhibition from ClO4- from Merrill et al (2003)62 in a similar manner to the model 11 ACS Paragon Plus Environment

Environmental Science & Technology

Page 12 of 33

263

used for humans. Parameters for the rat model were not altered from those used in Merrill et al.

264

(2003)62, except to use an iodide dose of 0.033 mg/kg/d, a body weight of 0.044 kg, and a consumption

265

rate of 0.005 g/kg/d to represent a small herbivorous mammal based on the meadow vole36,52. As with

266

humans, ClO4- doses corresponding to the 1st percentile, median, and 99th percentile exposures

267

predicted by the AEP exposure network were tested and compared with a zero-dose simulation to

268

estimate percent NIS inhibition. For fishes, which can equilibrate with their aqueous environment51, no

269

published dosimetry models were available. Therefore, we assumed that the concentration of ClO4- in

270

the surface water was equivalent to the TSE (Dr. John Nichols, personal communication, 14 October

271

2016). Percent NIS inhibition was not estimated in fishes due to the lack of a relevant published

272

dosimetry model.

273

2.6 Linking TSE and MIE to AOP toxicity data

274

Hines et al. (2018)35 used the AOP framework to integrate dose-response data for mechanistic

275

endpoints leading from NIS inhibition to AOs in multiple species, including those considered in this case

276

study. In the 2018 case study, predictions from the Lumen et al. (2013)49 model were used to inform

277

dose-response relationships for humans, while data for endpoints characterizing each KE were

278

assembled from literature sources to describe dose-response relationships in other species35. These data

279

were organized along the NIS inhibition AOP to make comparisons across taxa, and detailed descriptions

280

of the KEs, endpoints, and literature sources for this dose-response data are available in Hines et al.

281

(2018)35. In the present study, we used the AEP exposure modeling results as inputs to these dose-

282

response data to evaluate species-specific effects of NIS inhibitor contamination. We calculated a

283

hazard index (𝐻𝐼) for each species according to EPA guidance using the formula

284

𝑛

𝐸𝑖

EQ. 2

𝐻𝐼 = ∑𝑖 = 1𝐴𝐿𝑖

12 ACS Paragon Plus Environment

Page 13 of 33

285

Environmental Science & Technology

where 𝑖 is each exposure source, 𝐸 is the exposure level, and 𝐴𝐿 is the acceptable limit of

286

exposure63. 𝐴𝐿 was set to the lowest reported activation of a KE in a species35 to provide the most

287

conservative estimate of 𝐻𝐼, and we used the 1% and 99% dose for the predicted distribution of each

288

species in each scenario to provide the range of possible results. An 𝐻𝐼 of 1 or below indicates that

289

exposure levels are below the 𝐴𝐿, while a value greater than 1 indicates that exposure levels are above

290

the 𝐴𝐿 and may be cause for concern.

291

3. Results and discussion

292

The ClO4- case study provides an example application of techniques for site-specific evaluation of

293

exposure, toxicity, and risk in both human and non-human targets. We present the results of our

294

analysis for the hypothetical contaminated site and discuss the utility of a combined AEP-AOP

295

framework in future CRA efforts.

296

3.1 Species-specific external exposures

297

The sets of 10,000 AEP network models provided distributions of predictions for ClO4-

298

concentration in each environmental compartment (Table S3) that reflect the variability in the model

299

parameters. In each scenario, the ClO4- concentrations in terrestrial grass and shrubs were 50 to 100-

300

fold higher than the concentration in groundwater due to bioaccumulation, which is consistent with

301

environmental predictions for this contaminant36. Little to no bioaccumulation was predicted for aquatic

302

plants, which was also consistent with literature64. The toxicant concentrations of all compartments

303

increased across the mild, moderate, and high contamination scenarios, with large increases in

304

groundwater and grass concentrations in the high contamination scenario due to the simulated

305

groundwater spill (Table S3). When combined with the behavioral assumptions for each species, these

306

environmental concentrations provided species-specific distributions of external exposure predictions in

307

each scenario (Table 1). Due to the modeled accumulation of ClO4- in terrestrial grass and the high 13 ACS Paragon Plus Environment

Environmental Science & Technology

308

contribution of grass to their diet, doses were two orders of magnitude higher in the small herbivorous

309

mammals than in humans or fish. This result is consistent with the fact that ClO4- has been detected in

310

the milk of grazing animals such as cows65,66, and supports the notion that contaminated grass can be an

311

important pathway for ClO4- exposure in some terrestrial mammals.

312

3.2 Source apportionment for exposure pathways

313

There was a dominant source contributor to exposure for each species in the mild scenario

314

(Figure 2). For humans and small herbivorous mammals, ClO4- inputs from groundwater contamination

315

were responsible for 95% to 99% and 91% to 96% of exposure across the set of 10,000 networks,

316

respectively, while surface water input contributions ranged from 69% to 98% of exposure to fishes

317

(Figure 2A). In the moderate scenario, the contributions from all sources were multiplied by ten.

318

Therefore, ClO4- concentrations were higher than those for the mild scenario, but no change was

319

observed in the source apportionment due to the linear assumptions of the hypothetical exposure

320

model (Figure 2B). This finding highlights that source apportionment and environmental concentration

321

vary independently in this analysis. Although there is a linear increase in environmental concentration

322

with no change to source apportionment between the mild and moderate contamination scenarios,

323

non-linear relationships in the ADME and dose-response properties of the AOP imply that linear

324

increases in external exposure do not necessarily result in linear increases in AO risk for each species;

325

and can be further modified by animal behavior (below). The high contamination scenario, which

326

simulated a groundwater spill, resulted in changes in both source apportionment and environmental

327

concentration; groundwater contribution ranged from 99% to 100% in humans, 95% to 99% in small

328

herbivorous mammals, and from 17% to 80% in fishes (Figure 2C). This large range of predicted source

329

apportionment values in fishes was a result of a tradeoff between surface water and groundwater

330

inputs. Surface water contributions ranged from 19% to 82% across the set of 10,000 networks for this

331

scenario and showed a tight negative correlation with groundwater inputs (Spearman’s rho = -1, 14 ACS Paragon Plus Environment

Page 14 of 33

Page 15 of 33

Environmental Science & Technology

332

p