Modeling the Sensitivity of Aquatic Macroinvertebrates to Chemicals

Subscriber access provided by UNIV OF LOUISIANA

Ecotoxicology and Human Environmental Health

Modeling the sensitivity of aquatic macroinvertebrates to chemicals using traits Sanne van den Berg, Hans Baveco, Emma Butler, Frederik De Laender, Andreas Focks, Antonio Franco, Cecilie Rendal, and Paul J. van den Brink Environ. Sci. Technol., Just Accepted Manuscript • DOI: 10.1021/acs.est.9b00893 • Publication Date (Web): 22 Apr 2019 Downloaded from http://pubs.acs.org on April 25, 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

Modeling the sensitivity of aquatic

2

macroinvertebrates to chemicals using traits

3

Sanne J. P. Van den Berg*,1,2, Hans Baveco3, Emma Butler4, Frederik De Laender2, Andreas

4

Focks3, Antonio Franco4, Cecilie Rendal4, Paul J. Van den Brink1,3

5

1Aquatic

6

6700 AA Wageningen, the Netherlands

7

2Department

8

3Wageningen

9

4Safety

Ecology and Water Quality Management Group, Wageningen University, P.O. Box 47,

of Biology, University of Namur, Rue de Bruxelles 61, 5000 Namur, Belgium Environmental Research, P.O. Box 47, 6700 AA Wageningen, Netherlands

and Environmental Assurance Centre, Unilever, Colworth Science Park, Sharnbrook

10

MK441LQ, United Kingdom

11

*Corresponding

12

KEYWORDS

13

Predictive ecotoxicology, ecotoxicological modeling, database approach, mode of action,

14

ecological scenarios, macroinvertebrate sensitivity

15

ABSTRACT

author ([email protected])

1

ACS Paragon Plus Environment


16

In this study, a trait-based macroinvertebrate sensitivity modeling tool is presented that provides

17

two main outcomes: (1) it constructs a macroinvertebrate sensitivity ranking and, subsequently, a

18

predictive trait model for each one of a diverse set of pre-defined Modes of Action (MOAs) and

19

(2) it reveals data gaps and restrictions, helping with the direction of future research. Besides

20

revealing taxonomic patterns of species sensitivity, we find that there was not one genus, family

21

or class which was most sensitive to all MOAs, and that common test taxa were often not the

22

most sensitive at all. Traits like life cycle duration and feeding mode were identified as important

23

in explaining species sensitivity. For 71% of the species, no or incomplete traits data were

24

available, making the lack of trait data the main obstacle in model construction. Research focus

25

should therefore be on completing trait databases, and enhancing them with finer morphological

26

traits, focusing on the toxicodynamics of the chemical (e.g. target site distribution). Further

27

improved sensitivity models can help with the creation of ecological scenarios by predicting the

28

sensitivity of untested species. Through this development, our approach can help reduce animal

29

testing and contribute towards a new predictive ecotoxicology framework.

30

TOC ART

2


Page 2 of 33

Page 3 of 33


31 32

1.

Introduction

33

In the environmental risk assessment (ERA) of chemicals it is essential to determine the

34

environmental threshold concentration below which ecosystem structure and functioning

35

experience no adverse impacts. In order to set this threshold, a key challenge in ERA remains the

36

extrapolation of effects of toxicants found for a limited number of standard test species to many

37

additional species. Ecosystems are generally populated by hundreds to thousands of species, and

38

each species has the potential to show a different sensitivity towards one of the hundreds or even

39

thousands of different chemical compounds that can be present in our ecosystems.1-2

40

Experimental testing of this innumerable amount of species-chemical combinations, plus any

41

possible environmentally realistic mixture of those chemicals, is impossible. We therefore need 3



42

to improve current modeling approaches and make them flexible for application to any

43

geographic region and any set of abiotic conditions.

44

Traditional approaches trying to incorporate species diversity into risk assessment include the

45

application of uncertainty factors3 and the fitting of species sensitivity distributions (SSDs)4 to

46

available toxicity data. While both methods are extensively used (and frequently combined), they

47

are aimed at being protective rather than predictive and as such, still maintain large uncertainty

48

due to both a limited knowledge of the mechanisms underlying species sensitivity and a lack of

49

taxonomic diversity. This lack of taxonomic diversity is especially true for uncertainty factors,

50

but also holds for SSDs. Regulatory frameworks require SSDs to contain 10 to 15 species in

51

total,5 but divided over the different organism groups (e.g. fish, crustaceans, algae), this results in

52

only 1 or 2 organisms from each organism group, which are often comprised out of the same set

53

of standard test species. For the extrapolation across chemicals, Quantitative Structure-Activity

54

Relationships (QSARs) are commonly used.6 QSARs use chemical characteristics to predict the

55

toxicity of many chemicals for a certain species. Due to their large demand for experimental

56

toxicity data, however, QSAR-models are often built only for specific standard test species like

57

Daphnia magna and Oncorhynchus mykiss and, therefore, fail to account for the large species

58

diversity of real ecosystems. In the last decade, trait-based approaches have been introduced to

59

overcome this lack of realism.7-8 These approaches incorporate more ecological realism into ERA

60

by considering traits to provide a clear mechanistic link between exposure and effects, making it

61

possible to extrapolate species sensitivities over chemicals acting by the same Mode Of Action

62

(MOA). 4


Page 4 of 33

Page 5 of 33


63

After the introduction of trait-based approaches as a potential tool in ERA around a decade ago,8-

64

11

65

biological traits to predict species sensitivity. They performed a Principal Components Analysis

66

(PCA)19 on a species-by-substance matrix (12 macroinvertebrate species, 15 chemicals covering

67

several MOAs), where they introduced a species-by-traits matrix as a set of nominal, passive

68

explanatory variables. They found that up to 71% of the variability in the sensitivity could be

69

explained with only four species traits. In a later study, Rubach, Baird and Van den Brink12

70

developed the approach further, now using single and multiple linear regression instead of PCA,

71

and dividing the chemicals into groups according to their MOA. MOA has proven to be a strong

72

determinant of species sensitivity and is, therefore, seen as a promising alternative to chemical

73

class-based predictive toxicity modeling.20-21 Rubach , Baird and Van den Brink12 defined the

74

Mode Specific Sensitivity (MSS) value of each species as the average relative sensitivity of the

75

species to a group of chemicals with the same MOA. Single and multiple linear regressions

76

between the MSS values and species trait data at the family level explained up to 70% of the

77

variation in invertebrate sensitivity to 3 groups of insecticides. Recently, the approach by

78

Rubach, Baird and Van den Brink12 has been extended by incorporating relationships between

79

species traits and sensitivity into predictive models that could potentially be used in risk

80

assessment.14

81

So far, these predictive models have only been built for insecticides, whilst it remains open

82

whether the same traits are important in explaining invertebrate sensitivity to other groups of

83

chemicals. Therefore, in this study, we develop predictive models of macroinvertebrate

they have been rapidly evolving.7, 12-18 Baird and Van den Brink8 were among the first to use

5



84

sensitivity to a larger and more diverse set of MOAs. We first optimize the sensitivity prediction

85

method of Rubach et al.12 and Rico and Van den Brink14 using both literature research and

86

comparative analysis. The methods, results and discussion of this technical optimization are

87

given in the Supplementary Information (S1). Next, and as the main focus of this article, we

88

study the influence of MOA on species sensitivity rankings and, subsequently, on the resulting

89

predictive trait models. By studying which traits are included in the trait-sensitivity models, we

90

aim to get a better mechanistic understanding of how species sensitivity is determined by MOA.

91

In a final evaluation, we test the power of species sensitivity predictions and assess if model

92

performance depends on data availability. As a result of our research, we deliver a trait-based

93

macroinvertebrate sensitivity modeling tool that builds predictive models for potential use in risk

94

assessment for a diverse set of pre-defined MOAs.

95

2.

Methods

96

2.1.

Overall methodology

97

The prediction tool was developed in the R environment, and automates sensitivity rankings of

98

aquatic macroinvertebrates and the construction of a corresponding set of trait-based models from

99

a MOA-based chemical grouping (Figure 1). A hands-on explanation of the R tool and all R

100

scripts and databases required to run the tool are available in the Supporting Information (S1 and

101

S2 respectively). The tool consists of four different parts: database collection, preliminary input

102

data processing, data processing, and production of output. Each part in turn, consists of multiple

103

steps, the essentials of which will be presented briefly. 6


Page 6 of 33

Page 7 of 33


104

2.2.

Database collection and preliminary input data processing

105

Several public databases were utilized to obtain data on MOA, toxicity, chemical properties,

106

traits and taxonomy. We accept the limitations and errors of the databases exploited in this study,

107

as it is not within the scope of this study to perform a quality assessment on the databases used.

108

The extensive (1213 chemicals) MOA database developed by Barron et al.20 was used to extract a

109

list of chemicals with their CAS numbers and their division into 6 groups of broad MOA

110

(narcosis, acetylcholinesterase (AChE) inhibition, ion/osmoregulatory/circulatory (IOC)

111

impairment, neurotoxicity, reactivity, and electron transport inhibition), and their subdivision into

112

31 groups of specific MOA (see Table 1 in ref 20). The database was cleaned (e.g. chemicals for

113

which no specific MOA was defined were removed) and prepared for further processing (e.g.

114

specific MOAs with long names were replaced by a letter to ensure readability). For more details

115

on pre-processing, see the corresponding R script in the SI (S2).

116 117

Figure 1. Structure of the developed R

118

tool, divided into the four different

119

parts: database collection (blue),

120

preliminary input data processing

121

(light grey), data processing (dark

122

grey) and production of output (red).

123

MSS refers to Mode Specific 7



124

Sensitivity; LOOCV refers to Leave-One-Out-Cross-Validation.

125 126 127 128

The US Environmental Protection Agency (EPA) ECOTOX database22 was selected as the source

129

of toxicity data. See instructions in S1 on how to download the ECOTOX database. Only the

130

ECOTOX tables tests, results and species are incorporated into the R tool.

131

A chemical properties database was acquired by batch-running all CAS numbers available in the

132

MOA database in the EPI (Estimation Programs Interface) Suite programs KOWWIN, MPBPVP

133

and WSKOWWIN, respectively obtaining data on logKow (octanol-water partitioning

134

coefficient); melting point, boiling point and vapor pressure; and water solubility.23 Modelled

135

values were only used when experimental data were lacking. Eventually, only data on water

136

solubility are used in the tool (as a check for realistic concentration values). All data are kept in

137

however, to enable future flexibility of the tool. Data on molecular weight were obtained and

138

added to the chemical properties database by extracting SMILES (Simplified Molecular-Input

139

Line-Entry System) of all MOA CAS numbers from the SMILECAS database (also available

140

through EPI Suite), and calculating the molecular weight based on these SMILES using the rcdk

141

package in R (version 3.4.5).24

8


Page 8 of 33

Page 9 of 33


142

The Tachet database provides a coding of 22 biological and ecological traits describing 472

143

species known to live in French freshwaters.25-26 The database is based on ‘a very large and

144

scattered published expert knowledge and diverse literature sources ... We also included

145

unpublished observations of ourselves and colleagues’.26 From this, we deduced that the database

146

represents the average trait state of European species and is, therefore, suitable for our study.

147

Since our interest lies in predicting species sensitivity and we want to avoid overfitting, only the

148

following traits for which we could hypothesize a mechanistic relation with sensitivity were

149

extracted:12 maximum potential size (i.e. ≤ 0.25 cm, > 0.25 – 0.5 cm), life cycle duration (i.e. ≤ 1

150

year, > 1 year), potential number of cycles per year (i.e. < 1, > 1), dispersal mode (i.e. aquatic

151

passive, aerial active), respiration mode (i.e. tegument, gill), feeding mode (i.e. shredder,

152

scraper), current velocity preferendum (i.e. slow, fast), salinity preferendum (i.e. freshwater,

153

brackish water), temperature preferendum (i.e. < 15°C, > 15°C) and pH preferendum (i.e. ≤ 4, >

154

5.5-6) (see Table S1 for an overview of the traits and the corresponding trait categories).

155

To facilitate the cross-linking of information among the different databases, the Taxonomy

156

database of the NCBI (National Centre for Biotechnology Information)27-28 was used to extract

157

the scientific names, along with the taxonomic rank and unique id, of all the species present in

158

both the ECOTOX and the Tachet database. For this we used the taxize package in R (version

159

0.9.0).29 A copy of the ECOTOX species table and the Tachet database, both with updated

160

taxonomy, are provided in the SI (S2).

161

2.3.

Data processing

9



162

Data processing has been executed once for each broad and once for each specific MOA (36

163

times in total). The data processing part of the tool consists out of eight consecutive steps (Figure

164

1): i) first Mode Specific Sensitivity (MSS) calculation, ii) removal of species without traits

165

information, iii) second MSS calculation, iv) species-traits matching, v) preparation of trait data,

166

vi) optimization, vii) removal of redundant traits, and, viii) Leave-One-Out-Cross-Validation

167

(LOOCV).

168

The MSS values are calculated twice; once including all species for which sufficient toxicity data

169

are available (step i), and once only including species for which we have ascertained that also

170

trait data are available (step iii). In between the two MSS calculations, species for which no trait

171

data are available are removed (step ii). Performing the MSS calculation twice is necessary,

172

because the MSS values depend on relative sensitivities, and therefore are influenced by the in-

173

or exclusion of species. The MSS calculations are implemented as described by Rubach, Baird

174

and Van den Brink.12 In short, only toxicity tests lasting between 1 and 4 days and studying the

175

effect of chemical stress on mortality are included in the analysis. Within each chemical, first the

176

log transformed LC50 or EC50 values are normalized using the mean and standard deviation of

177

all sensitivity values found for that chemical, resulting in a relative sensitivity of each species

178

towards that chemical. Subsequently, these relative sensitivities are averaged over all chemicals

179

belonging to the same MOA, resulting in an MSS value. Importantly, there were three

180

fundamental differences in our method compared to the method of Rubach, Baird and Van den

181

Brink:12 the unit of the toxicity data (we use mol/L instead of µg/L), an extra check for realism of

182

concentration values (should not exceed solubility), and the selection of tests with the longest 10


Page 10 of 33

Page 11 of 33


183

exposure duration to enlarge the chance of reaching equilibrium concentration (see discussion for

184

more extensive explanation).

185

After the second MSS calculation, species-traits matching is executed at the lowest taxonomic

186

level possible (step iv). Since species level trait data are scarce, and traits measured at genus and

187

species level are strongly correlated,30 species-traits matching is done at genus level. Prior to

188

species-traits matching, MSS values are converted to genus level by averaging the MSS values of

189

all species belonging to the same genus. Traits data are also converted to genus level by taking

190

the median of the original fuzzy codes of all taxa belonging to that genus. This fuzzy coding

191

scheme is specifically developed for describing species traits data, and is ideal for differentiating

192

species by their affinities to different trait modalities (i.e. categories) belonging to several traits

193

(for further explanation of fuzzy codes, see ref 31).

194

Next, the trait data are prepared for linear regression by expressing continuous traits (e.g. size) as

195

weighted averages of the different trait modalities,12, 14, 25 and factorial traits (e.g. mode of

196

respiration) as fixed within species (step v).12, 14 The latter means that for each of the factorial

197

traits, the modality for which the taxon has the highest affinity is selected as the modality this

198

taxon carries and uses throughout its entire lifespan. In case of equal affinity to more than one

199

trait modality (e.g. 40% gills, 40% skin respiration), a missing value is inserted. These missing

200

values are problematic for (multiple) linear regression, because regression requires closed

201

datasets with no missing values. To solve this, an optimization step is performed (step vi), which

202

removes all gaps from the dataset by deleting any species with missing trait values. 11



Page 12 of 33

203

Next, redundant traits are removed to avoid overfitting (step vii). Traits are considered redundant

204

i) when they are clearly aliased with other traits (collinearity), and ii) when they do not show

205

enough variation in their different trait modalities. The tool tackles these two issues through i) a

206

collinearity maximum of 0.7 (see ref 32 for a review of different methods to deal with

207

collinearity), removing all traits that exceed this maximum, starting with those that correlate with

208

the largest number of other traits or, in case multiple traits match this criterion, with the largest

209

exceedance of the collinearity maximum, and through ii) a minimum on the trait modality

210

diversity index (derived from the Shannon diversity index33), removing all traits with a trait

211

modality diversity below this minimum.

212

Finally, a Leave-One-Out-Cross-Validation (LOOCV) is executed to quantify the predictive

213

power of the model (step viii). LOOCV is done by successively leaving out one species from the

214

training dataset and building the MSS model based on the remaining species (for explanation of

215

model building, see section 2.4). LOOCV was preferred above k-fold cross-validation, because

216

LOOCV is better in giving a reliable picture of the true R2 than k-fold cross-validation.34 With

217

the produced model, the MSS value of the left out species is predicted, and afterwards compared

218

to the known value by calculating the squared error. From the LOOCV, the mean squared

219

prediction error (MSPE) and a prediction coefficient (P2) were calculated as follows:

220

𝑀𝑃𝑆𝐸 = ∑𝑖 = 1

221

𝑃2 = 1 ―

𝑛

(𝑦𝑖 ― ŷ𝑖)2

eqn. 1

𝑛

𝑀𝑆𝑃𝐸

eqn. 2

𝑠2𝑦

12


Page 13 of 33


222

where ŷ𝑖 is the estimated MSS value, 𝑦𝑖 is the calculated MSS value, and 𝑠2𝑦 is the variance of all

223

MSS values in the dataset. When all predictions perfectly match observations, P2 equals 1. A

224

negative P2 value indicates that prediction errors exceed the total variance of the sensitivity

225

values, and therefore the model has poor accuracy.

226

2.4.

227

Each MOA output consists of the MSS ranking and the best MSS model. The MSS ranking is

228

obtained by sorting the MSS values found in the first MSS calculation from low (most sensitive)

229

to high (least sensitive). To visualize the taxonomic patterns in species sensitivity, we made heat

230

maps of the MSS rankings, dividing the MSS values into four bins, ranging from sensitive to

231

tolerant (MSS ≤ -1; -1 – 0; 0 – 1; ≥ 1). The best MSS model is selected in two steps. First, all

232

possible linear models are constructed using the regsubset function of the R package leaps

233

(version 3.0).35 Next, the best model is selected from all constructed models based on the small

234

sample unbiased Akaike’s Information Criterion (AICc), which takes both model fit and model

235

complexity into consideration and which is additionally extended with a bias correction term for

236

small sample size.36

237

3.

238

Figures of all MSS rankings resulting from the first MSS calculation and tables comparing within

239

and among broad MOA relative sensitivities can be found in S1 and S3 respectively. For the

240

interpretation of the results, it is important to realize that the MSS values represent the relative

241

sensitivity of a species to a group of chemicals with the same MOA. The results of the MSS 13

Production of output

Results and discussion



242

rankings (section 3.1) and the MSS models (section 3.2) are not described together, because the

243

MSS models result from a further reduced and transformed MSS dataset (Figure 1).

244

3.1.

245

We found large differences in the MSS rankings of species depending on MOA, both within the

246

broad MOAs (between the specific MOAs belonging to one broad MOA), and between the

247

different broad MOAs. Differences between broad MOAs were, however, larger than differences

248

within broad MOAs. This can be seen when comparing the standard deviation (SD) of the MSS

249

values across the broad MOAs (0.68), with the SD of the MSS values within the broad MOAs

250

(0.26, 0.38, 0.4, 0.52, 0.59, and 0.63 for within respectively the broad MOAs AChE inhibition,

251

narcosis, IOC impairment, electron transport inhibition, neurotoxicity, and reactivity) (S3).

252

Having a higher SD between the broad MOAs compared to within the broad MOAs confirms that

253

grouping into specific MOAs is helpful for modeling species sensitivity, as we expected.20-21, 37

MSS rankings

14


Page 14 of 33

Page 15 of 33


254 255 256 257 258 259 260 261 262 263 264

Figure 2. Heat map displaying taxonomic

265

distribution of species sensitivity towards the

266

different MOAs including a) all species with an

267

MSS value, and b) only species which have been

268

tested on 5 or 6 MOAs. MSS values are divided into

269

four bins, ranging from sensitive to tolerant. White

270

indicates absence of data. Upper cladogram shows

271

similarities between MOAs. 15



272

A heat map of the MSS rankings shows the general taxonomic pattern of species sensitivity and

273

tolerance to the different MOAs (Figure 2). For AChE inhibition, for instance, arthropods are

274

most sensitive and mollusks are most tolerant. Whether crustaceans or insects are the most

275

sensitive arthropods to AChE inhibition remains unclear, since both groups contain comparable

276

numbers of sensitive and tolerant genera. We see similar results in two closely related studies12, 14

277

as well as in a review of semi-field experiments.38 The sensitivity pattern for narcotic chemicals

278

is closely related to the sensitivity pattern of AChE inhibition, with crustacean and insect genera

279

containing the largest number of sensitive genera. We compared this result with two outdoor

280

microcosm studies performed with the fungicide azoxystrobin, classified as a narcotic chemical.20

281

In the study of Zafar et al.,39 significant effects were found for only one insect species

282

(Chaoborus obscuripes) and for none of the crustacean species tested. In the study from Cole et

283

al.40 (summarized in ref 41), the mollusk Sphaeriidae was the only species showing negative

284

effects on occurrence after a constant exposure of 10 µg/L, followed by negative effects on

285

Gammaridae, Oligochaeta and Planorbidae at 30 µg/L, and on Asellidae at 100 µg/L. No negative

286

effects on any of the insect groups tested (Hemiptera, Chaoboridae, Chironomidae) were found.

287

The discrepancy between our results and the two microcosm studies might be due to the wrong

288

assignment of narcosis as the MOA of azoxystrobin. Indeed, although two studies20, 42 classified

289

azoxystrobin as a narcotic chemical, two other classification schemes (the QSAR Toolbox

290

developed by the OECD, and Toxtree) assigned a non-narcotic MOA to azoxystrobin.21 In the

291

last paragraph of section 3.2.1 we discuss the causes and effects of MOA misclassification in

292

more detail. 16


Page 16 of 33

Page 17 of 33


293

Besides general patterns in species sensitivity, the heat map also shows that there is not one

294

genus, family or class which is sensitive to all MOAs (Figure 2a). When looking over all MOAs,

295

the arthropod phylum contained the largest fraction of species classified as being more sensitive

296

than average, followed by nematodes, mollusks and annelids (Figure 2a). Genera belonging to the

297

phyla Bryozoa, Cnidaria, Platyhelminthes and Rotifera were never more sensitive than average.

298

At class level, Hexanauplia contained the largest fraction of genera more sensitive than average,

299

which makes sense, because of their relatively small size and, herewith, large surface to volume

300

ratio. Another result is that common test taxa are often not the most sensitive at all (Figure 2b).

301

We find that Daphnia never belongs to the most sensitive group, and other commonly tested taxa

302

(e.g. Asellus, Procambarus, and Chironomus) are for the majority of the MOAs found to be

303

medium tolerant. Only the commonly tested taxon Ceriodaphnia shows high sensitivity to two

304

MOAs: AChE inhibition and reactivity.

305

The information presented in Figure 2a can also be used to identify accurately for which MOA-

306

taxon combinations data is lacking. Ciliophora have, for instance, been studied more frequently

307

for reactive chemicals, whilst Nematodes have been studied more frequently for the MOAs IOC

308

impairment and electron transport inhibition (Fig. 2). At a more general level, the MOAs IOC

309

impairment, electron transport inhibition and reactivity have been studied to a lesser extent than

310

the others. Especially data on insect species are missing for these MOAs, which might explain the

311

counterintuitive result of seemingly reduced sensitivity of insects to these MOAs. Indeed, a field

312

study performed in the 1970s shows that two mayfly (Ephemeroptera) species significantly

313

reduced in occurrence after application of the electron transport inhibitor Antimycin A, whilst all 17



314

other invertebrate taxonomic groups present at the study site remained unaffected.43 This

315

indicates that certain insect species are indeed sensitive to electron transport inhibition, but these

316

species are not included in the ECOTOX database. That the taxonomic focus of some MOAs

317

differs can lead, besides to a misinterpretation of results as just has been demonstrated, to a bias

318

in the sensitivity ranking, and subsequently, to a bias in the sensitivity models. That the

319

taxonomic composition of the species assemblage used to construct models is important is well

320

known. Maltby et al.44 found, for example, that hazardous concentrations (HC5) derived from

321

arthropod SSDs were significantly lower than those derived from non-arthropod invertebrates. In

322

order to avoid any bias in predictive modeling, we should, therefore, not only ensure high

323

taxonomic coverage, but also evenness across the different taxonomic groups.

324

3.2.

325

The final modeling effort resulted in 12 significant (p < 0.05) models explaining 31 to 90 percent

326

of the variation in MSS values, covering 5 broad MOAs and 7 specific MOAs (Table 1). Some of

327

the MOAs show overlap in the traits that explained sensitivity best. For example life cycle

328

duration (life) and feeding mode (feeding) are included in the models for 3 of the 6 broad MOAs.

329

This is a good indicator that these traits are in general important in explaining species sensitivity.

330

Results from several studies12, 14, 17 confirm this by including the same or closely related traits in

331

their models. Additionally, an extensive review of potential sensitivity related traits classify

332

feeding mode and life cycle duration as traits known to have an established link with sensitivity

333

for several taxa.7 Other traits selected by our modeling effect (e.g. temperature and salinity

Sensitivity models

18


Page 18 of 33

Page 19 of 33


334

preferendum) are only included in explaining one broad MOA, which might indicate that these

335

traits are less important for determining species sensitivity. For temperature preferendum, this is

336

indeed confirmed by the relatively low prediction coefficient of the models including this trait

337

(Table 1). One of the models containing salinity preference (alicyclic GABA antagonism),

338

however, has one of the highest prediction coefficients, and should, therefore, be considered

339

important. Indeed, a relationship between salinity preference and the toxicity of a GABA

340

antagonist makes sense, because GABA is one of the most common neurotransmitters,45 and is

341

therefore potentially influenced by the presence or absence of strategies to deal with salt stress

342

(see ref 46 for strategies to deal with salt stress).

343

For some traits, the correlation to sensitivity is positive for some MOAs and negative for other

344

MOAs (e.g. dispersal, Table 1). Species were more sensitive to narcosis when they preferred

345

more neutral waters (pH preferendum), were capable to disperse actively (dispersal) and breathed

346

through their tegument (see Table S1 for the included the trait modalities). For reactivity,

347

however, we find that the relationship between dispersal and sensitivity is the exact opposite, and

348

the more passive organisms were found to be more sensitive. This can be explained by the large

349

gap in insect genera for the MOA reactivity (Figure 2). When comparing the data included in the

350

multiple linear regression, the dataset for reactivity only contains some insensitive Odonata and

351

Diptera genera that are classified to disperse actively, whilst the dataset for narcosis contains both

352

sensitive Ephemeroptera and sensitive Trichoptera genera. Because of this unbalance, it remains

353

unclear whether the direction of the trait-sensitivity relationship of the two MOAs is indeed

354

mechanistic, or it is simply an artefact of data availability. We think that the latter is true, because 19



355

the trait profile of species sensitive to narcotic chemicals matches perfectly with the trait profiles

356

of insects belonging to the orders Ephemeroptera, Plecoptera and Trichoptera (EPT), which are

357

generally known as sensitive species.47 It can be hypothesized that our trait selection does not

358

contain the best traits, and the patterns we see only arise because traits are phylogenetic

359

correlated, i.e. belong to trait syndromes.47-48 Further analyses including phylogenetic

360

information can probably further elucidate this.

20


Page 20 of 33

Page 21 of 33


361

Table 1. Model coefficients of the best models (smallest AICc) that were found significant (p ≤ 0.05) for the different MOAs using

362

exhaustive linear regression analysis. Model fit is shown as the adjusted R2 (R2), and predictive power is shown as the prediction

363

coefficient (P2). See table S1 for an explanation of the traits and trait modalities used in this analysis.

Broad MOA

Specific MOA

Narcosis

Life

Life

Max.

pH

Dispersal

Respiration

cycle

cycles

Feeding

Temp.

potential

Velocity

Salinity

pref.

mode

mode

dur.

yr-1

mode

pref.

size

pref.

pref.

-0.32

-0.42

0.51

Nonpolar -0.44

Neurotoxicity

1.26

0.16

Alicyclic GABA antagonism AChE inhibition

-1.11

Organophosphate Carbamate Reactivity

-0.75

364

a)

-0.711

0.36

0.189

2.44

0.31

-0.29

6.13

0.44

0.193

-0.66 1.16

-0.74

-0.92

0.41

-0.027

-0.15

-0.83

0.33

-0.33

0.62

-0.125

0.67

0.255

0.9

-1.611

0.41

-0.446

0.48

0.326

-1.48

0.46 -0.91

Uncoupling oxidative phosphorylation

IOC impairment

0.42

-0.17

Chromate ETIa

0.013

0.53 -0.35

0.25 -1.6

Electron Transport Inhibition

21


P2

0.33 -0.77

Polar

R2

-1.26

-0.64 1.04


365

3.3.

Data gaps and potential for improvement

366

Low availability of traits data has in earlier studies been described as one of the major threats to

367

the successful implementation of trait-based approaches in ecotoxicology.7, 13 We confirm this by

368

showing that the lack of trait data was the main obstacle in model construction. Over all the

369

MOAs, an average of only 12% (SD ±5%) of the species for which we had sufficient data to

370

calculate the first MSS value were included in the construction of the MSS models. For an

371

average of 56% (±10%) of the species, no match at all could be found in the trait database. For

372

15% (±5%) of the species, a match could be found in the trait database, but values on one or

373

multiple traits of interest were lacking, and the species were therefore removed during the

374

optimization process. It is generally known that traits databases hold a high number of gaps, and

375

15% of the species with missing values matches closely to 18% of the species having an

376

incomplete trait description in a comparable trait database for Europe.49 Due to the large loss of

377

taxa in the analysis because of missing traits information, also only 20% (±7%) of the toxicity

378

tests used to calculate the first MSS value ended up as an underlying data point in the multiple

379

linear regression analysis. This indicates that a large part of the available toxicity data remains

380

unused, merely because of insufficient trait data. Future research should therefore focus on

381

completing, and simultaneously extending the taxonomic extent of existing trait databases. This

382

can be achieved by obtaining new data, but also by combining existing trait databases, although

383

the latter could become complicated due to existing differences in methodologies and categories

384

between the different trait databases. Baird et al.50 review the technological challenges in creating

22


Page 22 of 33

Page 23 of 33


385

and sharing traits data; and Culp et al.51 show examples of traits databases currently available for

386

different habitats and taxonomic groups (cf. Table 2).

387

In our analysis 16% of the traits we deemed important regarding species sensitivity did not enter

388

the multiple linear regression process, because they were highly correlated to another trait. This

389

could indicate the permanent presence of collinearities between some of the sensitivity-related

390

traits we selected (also defined as trait-syndromes48). However, there was a clear relationship

391

between the number of traits and the number of species going into the multiple linear regression

392

process, indicating that collinearity was primarily found in small datasets (Figure S4). Therefore,

393

traits were in most cases only correlating with each other due to chance, originating from the

394

small size of the remaining datasets after all data processing steps. Besides the sensitivity-related

395

traits selected and evaluated in this study, it can be helpful to enhance trait databases with finer

396

morphological characteristics which are more directly related to bioaccumulation (e.g. lipid

397

content), or to the internal distribution of the chemicals (e.g. target site distribution). Buchwalter

398

and Luoma52 found, for example, that although well-studied traits like body size or gill size did

399

not explain macroinvertebrate sensitivity to heavy metals, the relative number of ionoregulatory

400

cells did relate to dissolved metal uptake rates. Rubach et al.53 also tried to improve trait-models

401

by using finer morphological traits. Additionally, they measured the majority of these traits on

402

individuals captured in the same season (although a different year) and in the same biogeographic

403

region (sometimes even the exact same location) as they were collected for toxicity testing, which

404

reduces the presence of trait variability due to phenotypic or seasonal intraspecific variability.

405

However, their results show only a minor increase in model fit by measuring traits, even with the 23



406

lowered intraspecific variability, both for the single linear regression and for the multiple linear

407

regression (compare Table 2 from ref 12 with Table 3 and Figure 2 from ref 53).

408

There are other aspects that could improve trait-based models, three of which we will discuss in

409

more detail. First, measurement of internal tissue concentrations (i.e. the tissue residue) in

410

addition to external exposure concentrations would improve toxicity predictions, because toxic

411

effects are more closely related to the concentration inside the organism than the concentration in

412

the water.54-55 Although this has already been proven in the eighties, research on the (partial)

413

incorporation of tissue residues into ERA has only started the previous decade (e.g. in the biotic

414

ligand model56), and has still not been accepted as a standard in chemical regulation. We hope

415

that the recent development and increasing importance of mechanistic effect models like

416

toxicokinetics-toxicodynamics models (e.g. the General Unified Threshold Model of Survival57,

417

or the Dynamic Energy Budget58) will lead to a more frequent measurement and publication of

418

internal concentrations. Especially since these process-based models explain the link between

419

internal concentrations and dynamic processes underlying toxic responses, and additionally

420

enable extrapolation from standard test conditions, e.g. to different exposure scenarios.59-61

421

Rubach et al.53 additionally demonstrated that quantitative links between the parameters of these

422

mechanistic effect models fitted on internal concentration data and traits are substantially

423

stronger than quantitative links between classical sensitivity endpoints (e.g. EC50, LC50) and

424

traits. Measurements of internal concentrations could therefore improve our models in two ways:

425

one, by using internal instead of water concentrations, and two, by trying to explain TKTD

24


Page 24 of 33

Page 25 of 33


426

parameters, instead of LC50 values, which could subsequently be used to model EC50 values at

427

any exposure profile of interest,

428

Second, and especially important in the absence of internal concentrations, the exposure duration

429

of toxicity tests should be long enough to ensure approximate equilibrium between external and

430

internal concentrations. Without going too deep into the discussion of acute versus chronic

431

exposure, biotransformation, and elimination, we do want to emphasize that failure to obtain

432

equilibrium concentration can result in an underestimation of effects, simply because under the

433

same exposure time, a smaller organism reaches a higher internal concentration faster than a

434

larger organism.55 This means that for an acute toxicity test, different organisms may require a

435

different minimum exposure time. We tried to account for this by selecting the toxicity test with

436

longest exposure duration where possible. However, it can still be that some of the variability we

437

see in Figure 2 may be associated with differences in exposure duration (e.g. in one toxicity test,

438

a species was exposed for 48 hours to chemical A, whilst in another toxicity test, the same

439

species was exposed for 96 hours to chemical B).

440

Finally, improvements can be made by integrating Bayesian methods into the current approach.

441

Approximate Bayesian Computation can distinguish which mechanisms contribute most to

442

patterns observed in the data,62 and can thereby help optimizing model complexity. Bayesian

443

model averaging can additionally incorporate model-selection uncertainty into statistically

444

derived predictive models, and can therefore provide a more realistic estimation of model

445

uncertainty compared to the approach used in our study.63-64 25



446

Interestingly, data availability alone could not explain differences in model performance. An

447

increase in the number of toxicity tests, chemicals or genera did not improve model performance

448

substantially (Figures S1, S2, and S3). Actually, model fit reduces slightly with an increase in the

449

number of tests, chemicals or genera. However, cross-validation results show an exact opposite

450

trend, with a slight reduction in error with an increase in data availability. We think that model

451

performance does not increase with data availability due to one (or a combination) of the

452

following three reasons. First, the chemical groups (in our case, the MOAs) may insufficiently

453

differentiate the chemicals according to the effects that they cause in invertebrates. As in any

454

kind of grouping, mistakes or missing information could result in a wrong grouping in MOA for

455

the chemicals (see ref 21, 37 for studies on MOA classification errors). Additionally, MOA is not a

456

constant property of a compound, but may vary between species or life stage, depending, for

457

instance, on the availability of target sites, exposure duration or frequency, or the endpoint of

458

interest.65 Photosynthetic inhibitors, for example, are specific toxicants towards primary

459

producers, but are often baseline toxicants towards invertebrates.65 The chance that a chemical

460

expresses multiple MOAs in different species becomes smaller, however, when analysis is

461

restricted to species belonging to the same organism group (e.g. invertebrates, as in this study),

462

because the basic biochemical systems and molecular targets affected by each MOA may be

463

generally conserved across many species.66 A second aspect that could have prevented increased

464

model performance with increased data availability can be that we missed the right traits to

465

mechanistically explain the relationship between MOA and sensitivity. As mentioned before,

466

obtaining additional and finer morphological traits can help improve trait-based models, 26


Page 26 of 33

Page 27 of 33


467

especially if these traits explain the toxicodynamics of the chemical, since processes related to

468

toxicodynamics are currently not covered by the traits in our selection. Finally we argue that the

469

models could not achieve a better model performance because, regardless of data availability,

470

trait data alone may be insufficient in explaining species sensitivity. Several studies show that

471

complementing trait data with data on phylogenetics greatly enhances model performance, and

472

that both traits and phylogenetic indicators explain a distinct part of species sensitivity.67-68 Other

473

attempts combining phylogenetics and physiochemical properties into predictive models have

474

also proven successful,1, 69 although phylogenetics is still impossible to include in an inclusive

475

approach as we performed in this study. Nevertheless, we suggest to further explore the potential

476

of species traits for sensitivity predictions, as they enable large scale applicability and increased

477

mechanistic understanding.

478

3.4.

479

Our prediction tool is an addition to the fast growing and evolving science behind environmental

480

risk assessment, which can be easily amended in the future when more data or better data

481

processing and analysis procedures are identified. It is very flexible, and facilitates i) testing the

482

effectiveness of different kinds of chemical grouping, e.g. based on (physical) mechanism of

483

action,70 ii) testing the predictive value of ‘new’ traits, hypothesized to have a relationship with

484

species sensitivity, iii) the addition of new predictors, e.g. based on phylogenetics or

485

physiochemical properties, and iv) repeating the modeling exercise for different groups of

486

organisms (e.g. fish or algae), as long as trait data are available. Analyses comparable to those

Future direction

27



487

reported herein would take days to weeks of preparation time to collect all input data, followed

488

by more time to conduct and compile all analyses, requiring multiple types of commercial

489

software. Using the software and algorithms developed for this work, these intensive efforts can

490

now be compiled in only a few hours, already has all main open-source databases incorporated

491

and uses only R, a free software environment.

492

Both the sensitivity rankings and the sensitivity models produced in this study are an important

493

step forward on the challenging road towards predictive ecotoxicology with reduced animal

494

testing.71 The sensitivity rankings help reduce animal testing by guiding future taxonomic focus

495

of toxicity tests depending on the estimated MOA of the new chemicals. The sensitivity models

496

can predict the sensitivity of species or even entire communities never tested before, avoiding

497

additional animal tests. This allows us to rank the sensitivity of species occurring in any

498

community composition anywhere, or in other words, to determine the worst-case ecological

499

scenario for any aquatic ecosystem. Developing such ecological scenarios will fill one of the

500

remaining gaps in the construction of environmental scenarios72-74 deemed necessary for future

501

ecological risk assessment frameworks. Together with exposure scenarios, ecological scenarios

502

will form the basis for developing spatial-temporal explicit population-, community- and

503

ecosystem-level effect models for use in prospective environmental risk assessment for

504

chemicals.

505

SUPPORTING INFORMATION

28


Page 28 of 33

Page 29 of 33


506

(S1) PDF file: i) Hands-on explanation of the R tool, ii) Downloading instructions for the

507

ECOTOX database, iii) Table S1. Description of traits and trait categories, iv) Figures S1 – S4.

508

Relationships between model input and model performance, v) Comparison of modeling methods

509

and decisions, vi) Figures S12 – S42. MSS ranking for each Broad and Specific MOA. (S2) Zip

510

folder: i) All required R scripts, ii) Copy of required databases. (S3) Excel file: i) Tables

511

comparing within and among broad MOA rankings, ii) Modeling sets resulting from running the

512

R tool using the different practices.

513

ACKNOWLEDGEMENTS

514

This project has been funded by Unilever under the title ‘Ecological Scenarios and Models for

515

Use in chemical risk assessment’ (ESMU). Thanks to Pablo Rodriguez for assisting in

516

programming issues, to Merel Kooi with help in ggplot, and to Lara Schuijt for commenting on

517

schematic figures. The authors would also like to express their gratitude to the editors involved

518

and the anonymous reviewers for their insightful comments and suggestions.

519

AUTHOR CONTRIBUTIONS

520

The manuscript was written through contributions of all authors. All authors have given approval

521

to the final version of the manuscript.

522

REFERENCES

523 524 525 526 527

1. Guénard, G.; von der Ohe, P. C.; Walker, S. C.; Lek, S.; Legendre, P., Using phylogenetic information and chemical properties to predict species tolerances to pesticides. Proceedings of the Royal Society of London B: Biological Sciences 2014, 281 (1789), 20133239. 2. Guénard, G.; Ohe, P. C.; de Zwart, D.; Legendre, P.; Lek, S., Using phylogenetic information to predict species tolerances to toxic chemicals. Ecological Applications 2011, 21 (8), 3178-3190.

29



528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586

3. Chapman, P. M.; Fairbrother, A.; Brown, D., A critical evaluation of safety (uncertainty) factors for ecological risk assessment. Environmental Toxicology and Chemistry 1998, 17 (1), 99-108. 4. Posthuma, L.; Suter II, G. W.; Traas, T. P., Species sensitivity distributions in ecotoxicology. CRC press: 2001. 5. Belanger, S.; Barron, M.; Craig, P.; Dyer, S.; Galay-Burgos, M.; Hamer, M.; Marshall, S.; Posthuma, L.; Raimondo, S.; Whitehouse, P., Future needs and recommendations in the development of species sensitivity distributions: Estimating toxicity thresholds for aquatic ecological communities and assessing impacts of chemical exposures. Integrated Environmental Assessment and Management 2017, 13 (4), 664-674. 6. Donkin, P., Quantitative Structure-Activity Relationships. In Handbook of Ecotoxicology, Blackwell Publishing Ltd.: 2009; pp 785-811. 7. Rubach, M. N.; Ashauer, R.; Buchwalter, D. B.; De Lange, H. J.; Hamer, M.; Preuss, T. G.; Töpke, K.; Maund, S. J., Framework for traits‐based assessment in ecotoxicology. Integrated environmental assessment and management 2011, 7 (2), 172-186. 8. Baird, D. J.; Van den Brink, P. J., Using biological traits to predict species sensitivity to toxic substances. Ecotoxicol Environ Saf 2007, 67 (2), 296-301. 9. Liess, M.; Ohe, P. C. V. D., Analyzing effects of pesticides on invertebrate communities in streams. Environmental Toxicology and Chemistry 2005, 24 (4), 954-965. 10. Baird, D. J.; Rubach, M. N.; Van den Brink, P. J., Trait‐based ecological risk assessment (TERA): The new frontier? Integrated Environmental Assessment and Management 2008, 4 (1), 2-3. 11. Buchwalter, D. B.; Cain, D. J.; Clements, W. H.; Luoma, S. N., Using Biodynamic Models to Reconcile Differences Between Laboratory Toxicity Tests and Field Biomonitoring with Aquatic Insects. Environmental Science & Technology 2007, 41 (13), 4821-4828. 12. Rubach, M. N.; Baird, D. J.; Van den Brink, P. J., A new method for ranking mode-specific sensitivity of freshwater arthropods to insecticides and its relationship to biological traits. Environmental Toxicology and Chemistry 2010, 29 (2), 476-487. 13. Van den Brink, P. J.; Alexander, A. C.; Desrosiers, M.; Goedkoop, W.; Goethals, P. L.; Liess, M.; Dyer, S. D., Traits‐based approaches in bioassessment and ecological risk assessment: Strengths, weaknesses, opportunities and threats. Integrated environmental assessment and management 2011, 7 (2), 198-208. 14. Rico, A.; Van den Brink, P. J., Evaluating aquatic invertebrate vulnerability to insecticides based on intrinsic sensitivity, biological traits, and toxic mode of action. Environ Toxicol Chem 2015, 34 (8), 1907-17. 15. Liess, M.; Beketov, M., Traits and stress: keys to identify community effects of low levels of toxicants in test systems. Ecotoxicology 2011, 20 (6), 1328-40. 16. Segner, H., Moving beyond a descriptive aquatic toxicology: the value of biological process and trait information. Aquat Toxicol 2011, 105 (3-4 Suppl), 50-5. 17. Ippolito, A.; Todeschini, R.; Vighi, M., Sensitivity assessment of freshwater macroinvertebrates to pesticides using biological traits. Ecotoxicology 2012, 21 (2), 336-52. 18. Baert, J. M.; De Laender, F.; Janssen, C. R., The Consequences of Nonrandomness in Species-Sensitivity in Relation to Functional Traits for Ecosystem-Level Effects of Chemicals. Environ Sci Technol 2017, 51 (12), 72287235. 19. Braak, C. J. F. t., Ordination. In Data analysis in community and landscape ecology, Cambridge University Press: Cambridge, 1995; pp 9-173. 20. Barron, M. G.; Lilavois, C. R.; Martin, T. M., MOAtox: A comprehensive mode of action and acute aquatic toxicity database for predictive model development. Aquat Toxicol 2015, 161, 102-7. 21. Kienzler, A.; Barron, M. G.; Belanger, S. E.; Beasley, A.; Embry, M. R., Mode of Action (MOA) Assignment Classifications for Ecotoxicology: An Evaluation of Approaches. Environ Sci Technol 2017, 51 (17), 10203-10211. 22. USEPA ECOTOX User Guide: ECOTOXicology Knowledgebase System. . (accessed 03-30-2017). 23. USEPA Estimation Programs Interface Suite for Microsoft Windows, 4.11; Washington, DC, USA, 2018. 24. Guha, R., Chemical Informatics Functionality in R. Journal of Statistical Software 2007, 6 (18). 25. Tachet, H.; Richoux, P.; Bournaud, M.; Usseglio-Polatera, P., Invertébrés d'eau douce: systématique, biologie, écologie. CNRS éditions Paris: 2000; Vol. 15. 26. Usseglio‐Polatera, P.; Bournaud, M.; Richoux, P.; Tachet, H., Biological and ecological traits of benthic freshwater macroinvertebrates: relationships and definition of groups with similar traits. Freshwater Biology 2000, 43 (2), 175-205. 27. Sayers, E. W.; Barrett, T.; Benson, D. A.; Bryant, S. H.; Canese, K.; Chetvernin, V.; Church, D. M.; DiCuccio, M.; Edgar, R.; Federhen, S.; Feolo, M.; Geer, L. Y.; Helmberg, W.; Kapustin, Y.; Landsman, D.; Lipman, D. J.; Madden, T. L.; Maglott, D. R.; Miller, V.; Mizrachi, I.; Ostell, J.; Pruitt, K. D.; Schuler, G. D.; Sequeira, E.; Sherry, S. T.; Shumway, M.; Sirotkin, K.; Souvorov, A.; Starchenko, G.; Tatusova, T. A.; Wagner, L.; Yaschenko, E.; Ye, J., Database resources of the National Center for Biotechnology Information. Nucleic Acids Res 2009, 37 (Database issue), D5-15.

30


Page 30 of 33

Page 31 of 33

587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643


28. Benson, D. A.; Karsch-Mizrachi, I.; Lipman, D. J.; Ostell, J.; Sayers, E. W., GenBank. Nucleic Acids Res 2009, 37 (Database issue), D26-31. 29. Chamberlain, S.; Szöcs, E. taxize - taxonomic search and retrieval in R., F1000Research: 2013. 30. Dolédec, S.; Olivier, J.; Statzner, B., Accurate description of the abundance of taxa and their biological traits in stream invertebrate communities: effects of taxonomic and spatial resolution. Archiv für Hydrobiologie 2000, 148 (1), 25-43. 31. Chevenet, F.; Dolédec, S.; Chessel, D., A fuzzy coding approach for the analysis of long‐term ecological data. Freshwater biology 1994, 31 (3), 295-309. 32. Dormann, C. F.; Elith, J.; Bacher, S.; Buchmann, C.; Carl, G.; Carré, G.; Marquéz, J. R. G.; Gruber, B.; Lafourcade, B.; Leitão, P. J.; Münkemüller, T.; McClean, C.; Osborne, P. E.; Reineking, B.; Schröder, B.; Skidmore, A. K.; Zurell, D.; Lautenbach, S., Collinearity: a review of methods to deal with it and a simulation study evaluating their performance. Ecography 2013, 36 (1), 27-46. 33. Spellerberg, I. F.; Fedor, P. J., A tribute to Claude Shannon (1916–2001) and a plea for more rigorous use of species richness, species diversity and the ‘Shannon–Wiener’ Index. Global Ecology and Biogeography 2003, 12 (3), 177-179. 34. Hawkins, D. M.; Basak, S. C.; Mills, D., Assessing Model Fit by Cross-Validation. Journal of Chemical Information and Computer Sciences 2003, 43 (2), 579-586. 35. Lumley, T.; Miller, A. Leaps: Regression Subset Selection. https://CRAN.R-project.org/package=leaps (accessed 22-1-2018). 36. Johnson, J. B.; Omland, K. S., Model selection in ecology and evolution. Trends Ecol Evol 2004, 19 (2), 101-8. 37. Martin, T.; Young, D.; Lilavois, C.; Barron, M., Comparison of global and mode of action-based models for aquatic toxicity. SAR and QSAR in Environmental Research 2015, 26 (3), 245-262. 38. Wijngaarden, R. P. V.; Brock, T. C.; Van Den Brink, P. J., Threshold levels for effects of insecticides in freshwater ecosystems: a review. Ecotoxicology 2005, 14 (3), 355. 39. Zafar, M. I.; Belgers, J. D. M.; Van Wijngaarden, R. P. A.; Matser, A.; Van den Brink, P. J., Ecological impacts of time-variable exposure regimes to the fungicide azoxystrobin on freshwater communities in outdoor microcosms. Ecotoxicology 2012, 21 (4), 1024-1038. 40. Cole, J. F. H.; Everett, C. J.; Gentle, W.; Ashwell, J. A.; Goggin, U. Azoxystrobin: an outdoor pond microcosm study (summarised and cited in EFSA 2009). ZENECA Agrochemicals. Jealott’s Hill International Research Centre. : 2000. 41. EFSA, Council directive 91/414/EEC. Azoxystrobin. Report and proposed decision of the United Kingdom made to the European Commission under commission regulation 737/2007. Accessed via http://dar.efsa.europa.eu/dar-web/provision. 2009. 42. Bradbury, S. P., Predicting modes of toxic action from chemical structure: an overview. SAR and QSAR in Environmental Research 1994, 2 (1-2), 89-104. 43. Morrison, B., An investigation into the effects of the piscicide antimycin A on the fish and invertebrates of a Scottish stream. Aquaculture Research 1979, 10 (3), 111-122. 44. Maltby, L.; Blake, N.; Brock, T. C. M.; Van den Brink, P. J., Insecticide species sensitivity distributions: Importance of test species selection and relevance to aquatic ecosystems. Environmental Toxicology and Chemistry 2005, 24 (2), 379-388. 45. Nelson, D. L.; Cox, M. M., Lehninger principles of biochemistry. 5th ed. ed.; W.H. Freeman and Company: New York - Basingstoke, 2008; p 371-410. 46. Rivera-Ingraham, G. A.; Lignot, J.-H., Osmoregulation, bioenergetics and oxidative stress in coastal marine invertebrates: raising the questions for future research. Journal of Experimental Biology 2017, 220 (10), 1749-1760. 47. Verberk, W.; Van Noordwijk, C.; Hildrew, A., Delivering on a promise: integrating species traits to transform descriptive community ecology into a predictive science. 2013. 48. Poff, N. L.; Olden, J. D.; Vieira, N. K.; Finn, D. S.; Simmons, M. P.; Kondratieff, B. C., Functional trait niches of North American lotic insects: traits-based ecological applications in light of phylogenetic relationships. Journal of the North American Benthological Society 2006, 25 (4), 730-755. 49. Gayraud, S.; Statzner, B.; Bady, P.; Haybachp, A.; Schöll, F.; Usseglio‐Polatera, P.; Bacchi, M., Invertebrate traits for the biomonitoring of large European rivers: an initial assessment of alternative metrics. Freshwater Biology 2003, 48 (11), 2045-2064. 50. Baird, D. J.; Baker, C. J.; Brua, R. B.; Hajibabaei, M.; McNicol, K.; Pascoe, T. J.; de Zwart, D., Toward a knowledge infrastructure for traits‐based ecological risk assessment. Integrated environmental assessment and management 2011, 7 (2), 209-215.

31



644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701

51. Culp, J. M.; Armanini, D. G.; Dunbar, M. J.; Orlofske, J. M.; Poff, N. L.; Pollard, A. I.; Yates, A. G.; Hose, G. C., Incorporating traits in aquatic biomonitoring to enhance causal diagnosis and prediction. Integr Environ Assess Manag 2011, 7 (2), 187-97. 52. Buchwalter, D. B.; Luoma, S. N., Differences in Dissolved Cadmium and Zinc Uptake among Stream Insects: Mechanistic Explanations. Environmental Science & Technology 2005, 39 (2), 498-504. 53. Rubach, M. N.; Baird, D. J.; Boerwinkel, M. C.; Maund, S. J.; Roessink, I.; Van den Brink, P. J., Species traits as predictors for intrinsic sensitivity of aquatic invertebrates to the insecticide chlorpyrifos. Ecotoxicology 2012, 21 (7), 2088-101. 54. Friant, S. L.; Henry, L., Relationship between toxicity of certain organic compounds and their concentrations in tissues of aquatic organisms: A perspective. Chemosphere 1985, 14 (11), 1897-1907. 55. McCarty, L.; Landrum, P.; Luoma, S.; Meador, J.; Merten, A.; Shephard, B.; van Wezel, A., Advancing environmental toxicology through chemical dosimetry: External exposures versus tissue residues. Integrated Environmental Assessment and Management 2011, 7 (1), 7-27. 56. Paquin, P. R.; Gorsuch, J. W.; Apte, S.; Batley, G. E.; Bowles, K. C.; Campbell, P. G. C.; Delos, C. G.; Di Toro, D. M.; Dwyer, R. L.; Galvez, F.; Gensemer, R. W.; Goss, G. G.; Hogstrand, C.; Janssen, C. R.; McGeer, J. C.; Naddy, R. B.; Playle, R. C.; Santore, R. C.; Schneider, U.; Stubblefield, W. A.; Wood, C. M.; Wu, K. B., The biotic ligand model: a historical overview. Comparative Biochemistry and Physiology Part C: Toxicology & Pharmacology 2002, 133 (1), 3-35. 57. Jager, T.; Albert, C.; Preuss, T. G.; Ashauer, R., General unified threshold model of survival--a toxicokinetic-toxicodynamic framework for ecotoxicology. Environ Sci Technol 2011, 45 (7), 2529-40. 58. Jager, T.; Martin, B. T.; Zimmer, E. I., DEBkiss or the quest for the simplest generic model of animal life history. J Theor Biol 2013, 328, 9-18. 59. Ashauer, R.; Boxall, A. B.; Brown, C. D., New ecotoxicological model to simulate survival of aquatic invertebrates after exposure to fluctuating and sequential pulses of pesticides. Environmental science & technology 2007, 41 (4), 1480-1486. 60. Ashauer, R.; Hintermeister, A.; Caravatti, I.; Kretschmann, A.; Escher, B. I., Toxicokinetic and toxicodynamic modeling explains carry-over toxicity from exposure to diazinon by slow organism recovery. Environmental Science & Technology 2010, 44 (10), 3963-3971. 61. Ashauer, R.; Boxall, A. B.; Brown, C. D., Modeling combined effects of pulsed exposure to carbaryl and chlorpyrifos on Gammarus pulex. Environmental science & technology 2007, 41 (15), 5535-5541. 62. Hartig, F.; Calabrese, J. M.; Reineking, B.; Wiegand, T.; Huth, A., Statistical inference for stochastic simulation models – theory and application. Ecology Letters 2011, 14 (8), 816-827. 63. Clark, J. S., Why environmental scientists are becoming Bayesians. Ecology letters 2005, 8 (1), 2-14. 64. Wintle, B. A.; McCarthy, M. A.; Volinsky, C. T.; Kavanagh, R. P., The use of Bayesian model averaging to better represent uncertainty in ecological models. Conservation Biology 2003, 17 (6), 1579-1590. 65. Nendza, M.; Muller, M. T., Discriminating toxicant classes by MOA: 2. Physico-Chemical Descriptors. Quantitative Structure-Activity Relationships 2000, 19, 581-598. 66. LaLone, C. A.; Villeneuve, D. L.; Lyons, D.; Helgen, H. W.; Robinson, S. L.; Swintek, J. A.; Saari, T. W.; Ankley, G. T., Editor's Highlight: Sequence Alignment to Predict Across Species Susceptibility (SeqAPASS): A WebBased Tool for Addressing the Challenges of Cross-Species Extrapolation of Chemical Toxicity. Toxicol Sci 2016, 153 (2), 228-45. 67. Pilière, A. F. H.; Verberk, W. C. E. P.; Gräwe, M.; Breure, A. M.; Dyer, S. D.; Posthuma, L.; de Zwart, D.; Huijbregts, M. A. J.; Schipper, A. M., On the importance of trait interrelationships for understanding environmental responses of stream macroinvertebrates. Freshwater Biology 2016, 61 (2), 181-194. 68. Poteat, M. D.; Jacobus, L. M.; Buchwalter, D. B., The importance of retaining a phylogenetic perspective in traits-based community analyses. Freshwater Biology 2015, 60 (7), 1330-1339. 69. Malaj, E.; Guénard, G.; Schäfer, R. B.; Ohe, P. C., Evolutionary patterns and physicochemical properties explain macroinvertebrate sensitivity to heavy metals. Ecological applications 2016, 26 (4), 1249-1259. 70. Ashauer, R.; Jager, T., Physiological modes of action across species and toxicants: the key to predictive ecotoxicology. Environ Sci Process Impacts 2018, 20 (1), 48-57. 71. Villeneuve, D. L.; Garcia‐Reyero, N., Vision & strategy: Predictive ecotoxicology in the 21st century. Environmental toxicology and chemistry 2011, 30 (1), 1-8. 72. Rico, A.; Van den Brink, P. J.; Gylstra, R.; Focks, A.; Brock, T., Developing ecological scenarios for the prospective aquatic risk assessment of pesticides. Integrated environmental assessment and management 2016, 510-521. 73. Franco, A.; Price, O. R.; Marshall, S.; Jolliet, O.; Van den Brink, P. J.; Rico, A.; Focks, A.; De Laender, F.; Ashauer, R., Towards refined environmental scenarios for ecological risk assessment of down-the-drain chemicals in freshwater environments. Integrated Environmental Assessment and Management 2016, 233-248.

32


Page 32 of 33

Page 33 of 33

702 703 704


74. De Laender, F.; Morselli, M.; Baveco, H.; Van den Brink, P. J.; Di Guardo, A., Theoretically exploring direct and indirect chemical effects across ecological and exposure scenarios using mechanistic fate and effects modelling. Environ Int 2015, 74, 181-90.

705

33


Modeling the Sensitivity of Aquatic Macroinvertebrates to Chemicals

Recommend Documents