Toxic Mixtures in TimeâThe Sequence Makes ... - ACS Publications

Subscriber access provided by UNIVERSITY OF LEEDS

Article

Toxic mixtures in time – the sequence makes the poison Roman Ashauer, Isabel O'Connor, and Beate I. Escher Environ. Sci. Technol., Just Accepted Manuscript • DOI: 10.1021/acs.est.6b06163 • Publication Date (Web): 08 Feb 2017 Downloaded from http://pubs.acs.org on February 12, 2017

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 free 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 accessible to all readers and 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.

Environmental Science & Technology 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 22

Environmental Science & Technology

TOC art. For Table of Contents only. TOC art. For Table of Contents 275x190mm (96 x 96 DPI)

ACS Paragon Plus Environment


Toxic mixtures in time – the sequence makes the poison

1 2

Roman Ashauer1,2*, Isabel O’Connor1, Beate I. Escher1,3,4

3 4 5 6 7 8 9 10 11 12

1

Department of Environmental Toxicology, Eawag - Swiss Federal Institute of Aquatic

Science and Technology, 8600 Dübendorf, Switzerland 2

Environment Department, University of York, Heslington, York YO10 5DD, United

Kingdom 3

Cell Toxicology, Helmholtz Centre for Environmental Research – UFZ, Leipzig, Germany

4

Environmental Toxicology, Center for Applied Geosciences, Eberhard Karls University

Tübingen, Germany

13 14 15 16

*Corresponding author:

17

Phone: +44-(0)-1904-324052

18

Email: [email protected].

19 20 21

ACS Paragon1Plus Environment

Page 2 of 22

Page 3 of 22


22

Abstract: “The dose makes the poison.” This principle assumes that once a chemical is

23

cleared out of the organism (toxicokinetic recovery), it no longer has any effect. However, it

24

overlooks the other process of re-establishing homeostasis, toxicodynamic recovery, which

25

can be fast or slow depending on the chemical. Therefore, when organisms are exposed to

26

two toxicants in sequence, the toxicity can differ if their order is reversed. We test this

27

hypothesis with the freshwater crustacean Gammarus pulex and four toxicants that act on

28

different targets (diazinon, propiconazole, 4,6-dinitro-o-cresol, 4-nitrobenzyl chloride). We

29

found clearly different toxicity when the exposure order of two toxicants was reversed, while

30

maintaining the same dose. Slow toxicodynamic recovery caused carry-over toxicity in

31

subsequent exposures, thereby resulting in a sequence effect – but only when toxicodynamic

32

recovery was slow relative to the interval between exposures. This suggests that carry-over

33

toxicity is a useful proxy for organism fitness and that risk assessment methods should be

34

revised as they currently could underestimate risk. We provide the first evidence that carry-

35

over toxicity occurs amongst chemicals acting on different targets and when exposure is

36

several days apart. It is therefore not only the dose that makes the poison but also the

37

exposure sequence.

38 39 40

Keywords: Toxicology, dose, mixture toxicity, chemical safety assessment

41



42

Introduction

43

The paradigm “the dose makes the poison”, that is, that chemicals can be toxic or non-toxic

44

depending on their dose, stems back to Paracelsus. Given that exposure to toxicants in the

45

environment is often episodic and repeated 1, 2 and organisms are exposed to several toxicants

46

at the same time or after each other, it is important to also take into consideration the

47

temporal aspects of toxicity 3-5. The time required by an organism to recover from damage

48

due to previous exposure events is defined as the organism recovery time 6, 7. Organism

49

recovery results from toxicokinetic recovery, i.e. biotransformation, distribution and

50

excretion of the toxicant, and toxicodynamic recovery, i.e. re-establishment of homeostasis.

51

In previous studies 7-9, we developed toxicokinetic-toxicodynamic models using toxicants

52

from different chemical classes and found that certain toxicants exhibit slow toxicodynamic

53

recovery in Gammarus pulex. The modelling suggested that even after the substance was

54

eliminated from the organism, more time was necessary to re-establish homeostasis. The

55

conventional wisdom, as reflected by most current chemical risk assessment practices10, is

56

that when a chemical is eliminated from an organism (toxicokinetic recovery), it no longer

57

exerts an effect. This overlooks a subtler process of toxicodynamic recovery, which is

58

determined by the mode of toxic action. Both aspects of recovery are related to the

59

characteristics of the substance 3, 9, where toxicokinetic parameters are related to

60

hydrophobicity 11 and toxicodynamic parameters to mechanism of toxicity 9 and hence

61

molecular structure 12.

62

What if subsequent exposure events occur in intervals that are sufficient for

63

toxicokinetic recovery but not for toxicodynamic recovery? We already know that

64

toxicodynamic build-up of damage can lead to carry-over toxicity from subsequent pulses of

65

the same toxicant 7 and from subsequent pulses of different toxicants that act on the same

66

target 6. Therefore toxicokinetic-toxicodynamic models were developed recently, that extend

67

the classical dose paradigm by complementing internal concentrations with scaled damage as

68

the dose metric 9, 13. These early insights into toxicokinetics and toxicodynamics led to the

69

sequence effect hypothesis: when organisms are exposed to two toxicants in sequence, the

70

toxic effects may differ if that sequence is reversed 6.

71

To date this hypothesis has only been tested once, with carbaryl and chlorpyrifos in

72

G.pulex, two toxicants that both inhibit the same enzyme. The empirical evidence supported

73

it but it could be explained by a build-up of inhibited enzyme 6 and therefore did not

74

challenge the current paradigm. More recently we gained a much deeper understanding of

75

toxicokinetics 14 and toxicodynamics 9 of a diverse range of toxicants in G. pulex, in ACS Paragon3Plus Environment

Page 4 of 22

Page 5 of 22


76

particular how toxicokinetic and toxicodynamic recovery times differ among chemicals.

77

Based on this new knowledge we now propose a significant broadening of the sequence

78

effect hypothesis: it also applies to toxicants that act on different targets and via different

79

mechanisms. Would carry-over toxicity also occur after subsequent pulses of different

80

toxicants that act on different targets with different toxic modes of action? This is an

81

important question because sequential exposure to toxicants is the norm, but the current risk

82

assessment paradigm pays little attention to possible interactions of different toxicants over

83

time. We think that substances acting on different targets can cause carry-over toxicity, which

84

we view as a proxy for reduced organism fitness. If the sequence effect occurs for a wide

85

range of toxicants with different mechanisms of toxicity, then we are potentially severely

86

underestimating risk from chemicals.

87

The sequence effect hypothesis can be tested experimentally because it predicts that

88

the toxicity of two subsequent pulsed exposure events is more pronounced if we first expose

89

organisms to the substance with a slower recovery time, followed by exposure to the

90

substance with a faster recovery time. Conversely, exposure to the substance with a shorter

91

recovery time followed by the substance with a longer recovery time will result in less carry-

92

over toxicity and have a weaker impact on the organism.

93

We tested the sequence effect hypothesis in the amphipod crustacean Gammarus

94

pulex using four model toxicants that differ in their toxic mode of action. G.pulex are

95

particularly suitable for this type of study because of their relatively long life-span and

96

sufficiently large size to measure toxicant body residues. Diazinon [DIAZ] is an insecticide

97

that inhibits acetylcholinesterase 15, propiconazole [PCZ] is a fungicide that likely acts as a

98

baseline toxicant in G. pulex 16, 4,6-dinitro-o-cresol [DNOC] is an uncoupler of oxidative

99

phosphorylation 17, and 4-nitrobenzyl chloride [NBCl] is a reactive toxicant that forms

100

covalent bonds with proteins and other biomolecules 18. Therefore, each toxicant acts

101

according to a different mode of action, and their mixture toxicity should be explained by the

102

model of independent action 19, 20. Previously, we determined the times required for 95%

103

elimination of the toxicologically relevant metabolites and for 95% recovery of

104

toxicodynamic damage in G. pulex 7, 8, 14, 21 for each compound individually (Table 1). All

105

four substances have toxicokinetic recovery times of less than three days, whereas the

106

toxicodynamic recovery times are short for PCZ (< 3d) and DNOC (29000d). Based on this knowledge, we designed four experiments,

108

each involving two toxic experimental groups and a control group. The two toxic

109

experimental groups employed two subsequent pulsed exposures to two different toxicants ACS Paragon4Plus Environment


110

(Fig. 1). The two toxic experimental groups in each experiment differed only in the sequence

111

of the toxicants applied.

112

We chose a pulse interval that was long enough to allow toxicokinetic recovery (such

113

that no residual chemical remained in the body at the onset of the second pulse (Supporting

114

Information Fig. S1–S4)), but too short to allow complete toxicodynamic recovery after

115

DIAZ and NBCl exposure. Hence, we expected carry-over toxicity for those experimental

116

groups in which the first applied pulse was a toxicant with slow toxicodynamic recovery

117

(DIAZ or NBCl) but not in the groups in which the first applied pulse was a toxicant with fast

118

toxicodynamic recovery (PCZ and DNOC). Consequently, we expected to observe a

119

sequence effect, i.e. differences in survival at the end of the experiment, in the first, second

120

and third experiments (Fig. 1). In the fourth experiment, we expected both experimental

121

groups to exhibit carry-over toxicity due to slow toxicodynamic recovery of both toxicants,

122

and therefore did not expect the sequence effect.

123

We also carried out toxicokinetic-toxicodynamic mixture toxicity modelling. This

124

dynamic modelling expands the applicability domain of mixture toxicity models to pulsed or

125

time-variable exposures. The model structure was derived by assuming that the toxicants act

126

on different targets and therefore independently. The model parameters were taken from

127

previous studies. This enables us to compare model forecasts with new experimental data

128

from this study to assess how good the model predictions are.

129 130

Materials and methods

131

Experiments

132

Experimental procedures closely followed established protocols 6, 7, 9. Adult G. pulex were

133

collected in a headwater stream in the Itziker Ried, Switzerland near Zurich (E 702150, N

134

2360850) and acclimatized to 13°C (Table S2). G. pulex were fed horse-chestnut leaf discs

135

conditioned with Cladiosporum herbarum22 fungi. Experiments were performed in pyrex

136

beakers filled with 500 mL pre-aerated artificial pond water22 and at 13°C under a 12:12 h

137

light:dark cycle. Each beaker initially contained 10 G. pulex and the total number of test

138

organisms in each toxic experimental group is given in Tables S6, S 8, S10 and S12. Test

139

organisms were a natural mixture ratio of adult males and females and the numbers used in

140

each experiment varied depending on the availability of organisms at the Itziker Ried. Leaf

141

discs were added for ad libitum feeding, ranging from three to five leaf discs per beaker

142

depending on the number of alive G.pulex. Live organisms were counted by gently prodding

143

with a spatula and observing their resulting movement (Tables S6, S8, S10, and S12). Each ACS Paragon5Plus Environment

Page 6 of 22

Page 7 of 22


144

experiment included also a control group which was not treated with toxicants. No blinding

145

was employed. Any neonates that were released by females during the experiments were not

146

counted. Dead individuals were removed.

147

The test concentrations were chosen based on previous toxicity studies7-9 to achieve partial

148

mortality after the first pulse. The test chemicals were dosed as a mixture of 14C-labelled and

149

unlabelled substances to allow quantification of exposure concentrations. Radiolabelled

150

DIAZ and PCZ were supplied by the Institute of Isotopes (Budapest, Hungary). Radiolabelled

151

DNOC and NBCl were sourced from American Radiolabeled Chemicals (St. Louis, MO,

152

USA). Unlabelled chemicals were of analytical grade and supplied by Sigma-Aldrich (Buchs,

153

Switzerland). Dosing stocks of labelled and unlabelled test compounds were prepared in

154

acetone as solvent. Each beaker was dosed by pipetting the required volume of stock into the

155

medium, followed by gentle stirring with a glass rod. Concentrations of the solvent acetone in

156

the test medium were 0.1% or lower (Table S4), i.e. three orders of magnitude below

157

concentrations toxic to crustaceans23. Actual exposure concentrations were measured by

158

liquid scintillation counting (Tri-Carb 2200CA, Packard, USA) using 1 mL aliquots of the

159

medium (Tables S5, S7, S9, and S11), correcting for efficiency using internal standards,

160

subtracting background activity using control medium samples and converting radioactivity

161

to molar concentrations by using the known mixture ratio of labelled and unlabelled

162

substance. Immediately before and at the end of an exposure pulse, organisms were separated

163

from the medium by using a tea strainer, the test medium was discarded, and the organisms

164

were rinsed with clean water and then placed in beakers with fresh artificial pond water and

165

fresh leaf discs. The pH, conductivity and dissolved oxygen concentration of the medium

166

were monitored (Table S4).

167

Our experimental design aimed to maximise the number of test organisms that we could

168

handle so that we would maximise experimental power. However no formal power analysis

169

was carried out and the experimental design was constrained by the practicalities of the

170

experiment.

171

Statistical Analysis

172

Statistical testing was performed using GraphPad Prism (v6.03, GraphPad Software Inc., La

173

Jolla, CA, USA). To detect the sequence effect, we compared the numbers of dead and live

174

individuals at the end of the experimental groups using contingency table analysis, more

175

specifically, Fisher’s exact test (Fig. 1B, D, F, H, J, L, N, P). For evaluation of carry-over

176

toxicity, we compared the survival curves after the first and second pulses for each



Page 8 of 22

177

experimental group using the Gehan-Breslow-Wilcoxon test (Fig. 2). These survival curves

178

were constructed by recording the absolute number of alive G.pulex at the start of a pulse

179

(this could be the first or the second pulse in an experimental group) as the initial number of

180

organisms and then following each group until either the end of the experiment or the onset

181

of the second pulse. In other words: we compared survival curves during and following the

182

pulsed exposure to the same toxicant in different groups of the same experiment, once as the

183

first pulse and again as the second pulse. If there was a difference, then we concluded that

184

carry-over toxicity had occurred because we used the same dose for a single toxicant in both

185

of its exposure pulses within each experiment. The only difference was that the order was

186

reversed in the second experimental group. All four experiments were analysed following the

187

same method.

188

Toxicokinetic-toxicodynamic modelling

189

Toxicokinetics: modelling the internal dose

190

The toxicokinetic model was calibrated previously using time series data of measured internal

191

concentrations in pulsed exposure experiments7-9,

192

concentrations in G. pulex was simulated using the previously established toxicokinetic

193

models and parameters from those studies:

194

195

, ( ) ,$ ( )

14

. Here, the time-course of internal

= () × , − , () × , − ∑" , () × ! ," #

(1)

= , () × ! ," − ," () × ,"

(2),

196

where Cinternal,p(t) is the concentration of the parent compound in the organism [nmol/kgw.w.],

197

Cw(t) is the concentration of the parent compound in water [nmol/L], Cinternal,j(t) is the

198

concentration of biotransformation product j in the organism [nmol/kgw.w.], kin,p is the uptake

199

clearance coefficient [L/(kgw.w.×d)], kout,p is the elimination rate constant of the parent

200

compound [1/d], kmet,j is the first-order biotransformation rate constant for the formation of

201

metabolite j [1/d], and kout,j is the elimination rate constant of the biotransformation product j.

202

Internal concentrations of propiconazole (PCZ) and 4,6-dinitro-o-cresol (DNOC) were

203

directly used as inputs in the toxicodynamic model; by contrast, for diazinon (Diaz) and 4-

204

nitrobenzyl chloride (NBCl), biotransformation products were the driving variables for the

205

toxicodynamic model (see Tables 1 & S1, also for model parameter values). In the case of 4-

206

nitrobenzyl chloride, the metabolites are unidentified and considered to act as baseline

207

toxicants9. For diazinon, metabolite 1 is diazoxon, which is the toxicologically relevant


Page 9 of 22


208

molecule (Table 1). For propiconazole and 4,6-dinitro-o-cresol, we included only the parent

209

compounds in the model, because biotransformation in G.pulex is negligible8, 14.

210

Toxicodynamics: modelling survival

211

We used the two limit cases stochastic death (SD) and individual tolerance (IT) of the

212

General Unified Threshold model of Survival (GUTS)

213

concentrations and scaled damage

9, 13

13

with explicitly modelled internal

as the dose metric. These models have been described

9, 13

214

previously

215

parameters are given in Table S2. Scaled damage is a proxy for the toxicodynamic state of

216

the organism, is generally inferred by model fitting 9, 13 and it is calculated as follows:

217

dDscaled (t ) = k r × (Csum tox, organism (t ) − Dscaled (t ) ) dt

218

where Dscaled(t) is the time course of the scaled damage [nmol/kgw.w.], t is time [d], Csum tox,

219

organism(t)

220

active chemicals in the organism [nmol/kgw.w.], and kr is the damage recovery rate constant

221

[1/d]. The damage recovery rate constant captures the time course of toxicodynamics, and we

222

use two different parameters, kr SD and kr IT, for the two limit cases of GUTS.

223 224

Effects of scaled damage on survival in GUTS-SD

225

The hazard rate is the probability of an organism dying at a given point in time and is

226

calculated as follows8, 13, 24:

227

%( )

, but we repeat their description here for clarity. The toxicodynamic model

(3),

is the time course of the sums of the internal concentrations of the toxicologically

= & × '() (*+, () − -, 0) + ℎ_2345367

(4),

228

where dH(t)/dt is the hazard rate [1/d], kk is the killing rate constant [kgw.w./(nmol×d)],

229

Dscaled(t) is the time course of the scaled damage [nmol/kgw.w.], t is time [d], z is the threshold

230

[nmol/kgw.w.], and h_controls is the background hazard rate (control mortality rate, assumed

231

to be constant during the experiment, see Table S3) [1/d]. The survival probability, i.e. the

232

probability of an individual surviving until time t, is given as follows:

233

S (t ) = e − H ( t )

234

where S(t) is the survival probability [unitless].

235

Effects of scaled damage on survival in GUTS-IT

(5),



Page 10 of 22

236

We assume a log-logistic distribution of the threshold in the study population (individual

237

tolerance)8, 13. Then, the cumulative log-logistic distribution of the tolerance threshold in the

238

study population, which changes over time as individuals die, is calculated as follows:

239

F (t ) =

240

where F(t) is the cumulative log-logistic distribution of the tolerance threshold over time

241

[unitless], Dscaled(t) is the time course of the scaled damage [nmol/kgw.w.], t is time [d], τ is

242

time [d], α is the median of the distribution [nmol/kgw.w.], and β is the shape parameter of the

243

distribution [unitless]. Under the assumption of individual tolerance, the survival probability

244

is then given as follows 8, 13:

245

S (t ) = (1 − F (t ) ) × e − h _ controls×t

246

where S(t) is the survival probability [unitless], and h_controls is the background hazard rate,

247

i.e. the control mortality rate, which is assumed to be constant over time [1/d].

248 249

Mixture toxicity model

250

Mixture toxicity model assumptions

251

Mixture toxicity models generally assume either that the toxicants act on the same target or

252

on different targets 25, 26. Independent action is the model for the second assumption, which

253

applies to our combinations of toxicants. Traditional mixture toxicity models are designed for

254

exposure to multiple toxicants at the same time and for constant concentrations. We expand

255

the applicability domain of the mixture toxicity modelling to fluctuating, pulsed or sequential

256

exposures using a dynamic model, GUTS 9, 13. As we assume independent action for

257

compounds with different modes of action, we sum the hazards from the different toxicants 9,

258

1  max Dscaled (τ )   1 +  0< τ < t   α  

(6),

−β

(7),

27

, which is equivalent to multiplying the survival probabilities.

259

Mixture toxicity model equation

260

The same mixture toxicity model is used for GUTS-SD and GUTS-IT:

261

8!9 () = : ; ∑ %

262

which is equivalent to

263

8!9 () = ∏ 8 ()

(8),

(9)


Page 11 of 22


264

Smix (t) is the survival probability under toxic stress from i toxicants. In equation (8), we sum

265

the hazards, which is equivalent to multiplying the survival probabilities (eq. 9) and

266

corresponds to the assumption that toxicants act on different target sites or via different

267

toxicity pathways.

268

Treatment of NBCl metabolites in the toxicity model

269

In the case of NBCl, the internal concentration of the parent compound was modelled as a

270

reactive toxicant, whereas the sum of the internal concentrations of the metabolites was

271

modelled as a baseline toxicant, as in a previous study 9. We used the toxicodynamic

272

parameters of 1,2,3-trichlorobenzene, a well-known prototype baseline toxicant, to simulate

273

the effects of the metabolites assumed to act via baseline toxicity 9. As we assumed the two

274

modes of toxic action are independent, we modelled separate scaled damages for each using

275

equations (3) to (7) and then multiplied the resulting survival probabilities:

276

8,!= () = 8=+ 9, ? () × 8+,@, 9, ? () × : ;A_, +×

277 278

Results and discussion

279

Mixture toxicity model prediction

280

(10)

We compared the survival prediction determined by the mixture toxicity models with

281

the observed survival (Fig. S1 to S4). The agreement or disagreement between the predicted

282

and observed survival curves after the first pulses is a measure of the inter-experimental

283

variability. In half of the cases, the predicted survival after the first pulse did not match the

284

observed survival very well (Fig. S1G, S3G, S4G, H). Only in experiment two was survival

285

after the first pulses predicted accurately in both groups (Fig. S2G, H); in this experiment,

286

survival after the second pulses was also predicted accurately (Fig. S2G, H). The

287

disagreement between the predicted and observed survival after the second pulse in

288

experiment one (Diaz, Fig. S1H) or three (NBCl, Fig. S3H) is consistent with the

289

disagreement seen after the first pulse and is not necessarily an indicator that the mixture

290

toxicity models fail to predict full recovery after PCZ (experiment one) or DNOC

291

(experiment three). In experiment four the model predicted that all organisms would die

292

almost immediately following exposure to NBCl, thus assessment of the predicted mortality

293

after DIAZ in this experiment is difficult.

294 295

We also compared the mixture toxicity model prediction with the observed survival curve in the five experimental halves where we found evidence of carry-over toxicity



296

(survival after PCZ, DNOC, DNOC, DIAZ and NBCl in experiments one, two, three, four

297

and four, respectively; Fig. 2 A, C, E, G, H) and observed in four out of five cases that the

298

observed mortality following the respective second pulses exceeded the predicted mortality

299

(Fig. S6 to S10). From that observation we conclude that the toxicants interact, even if

300

exposure is days apart, and that the observed mortality following the second pulses is more

301

than just continued mortality from the other toxicant of the respective first pulses in all the

302

‘sequence’ treatments and the ‘inverse sequence’ treatment of experiment four (i.e. those with

303

carry-over toxicity).

304

We need more studies of this type to better understand how the model performs for

305

different substances and classes of chemicals and where the limits of its predictive

306

capabilities are28. Uncertainty arising from calibration experiments themselves can be

307

included in the model predictions28 to better understand inter-experimental variability.

308

Interesting questions arise also from the temporal mismatch in the onset of predicted and

309

observed mortality (particularly NBCl, see Fig. S3 & S4) as well as our choice of toxic

310

molecules (Table 1).

311

In summary, we found no evidence that the mixture models fail to predict carry-over

312

toxicity or the sequence effect. However, we did find large inter-experimental variability

313

evidenced by a disagreement between the predicted and observed mortalities. Since the

314

models were calibrated based on previously published experiments, this disagreement only

315

indicates that the variability in survival from one pulsed exposure experiment to the next is

316

too large to assess the performance of the model prediction. However, the experimental

317

evidence depicted in Figures S1 to S4 fully supports the sequence effect hypothesis.

318

Conclusions drawn from comparisons within an experiment, such as in this study, are not

319

subject to the limitations of inter-experimental variability. Variability between experiments

320

could possibly be reduced by using laboratory-cultured organisms.

321

Furthermore we identified a knowledge gap in the suite of toxicokinetic-

322

toxicodynamic mixture toxicity models available. We provide equations for summing up

323

hazard rates for toxicants which are assumed to act independently. However it is also

324

conceivable to model the combined effects of toxicants that act on different targets by

325

summing up the state variable scaled damage. However there are two challenges. First,

326

summing up damage has been previously used to model combined effects of toxicants that act

327

on the same target6, 29. This raises a question: How can we reconcile that previous work with

328

any future use of the same approach for toxicants acting on different targets? The task of


Page 12 of 22

Page 13 of 22


329

reconciling this previous work with newer studies is further complicated by subtle issues and

330

inconsistencies in the models used in the older studies13. Second, the model equations for

331

summing up scaled damage in GUTS-SD are straightforward to derive, however the model

332

equations for summing up scaled damage in GUTS-IT are unknown and not trivial to derive.

333

The sequence effect

334

We observed the sequence effect in experiments one, two and three (Fig. 1). Contingency

335

table analysis of the number of dead and alive individuals at the end of each experimental

336

group indicated a sequence effect for the combinations of DIAZ and PCZ (p

Toxic Mixtures in TimeâThe Sequence Makes ... - ACS Publications

Recommend Documents

Toxic Mixtures in TimeâThe Sequence Makes ... - ACS Publications