OPEN ACCESS DOCUMENT

• Environmental Pollution 2017 220, 1231-1243

• DOI: 10.1016/j.envpol.2016.11.010

1

1

2

3

4

5

6

7

8

9

10

12

Assessment of chlorpyrifos toxic effects in zebrafish (Danio rerio) metabolism

Cristian Gómez-Canelaa*, Eva Pratsb, Benjamí Piñaa, Romà Taulera

aDepartment of Environmental Chemistry, IDAEA-CSIC, Jordi Girona 18-26, 08034 Barcelona, Catalonia, Spain.

bCentre d’Investigació I Desenvolupament, CID-CSIC, Jordi Girona 18-26, 08034 Barcelona, Catalonia, Spain.

*Corresponding authorEmail: cristian.gomez@cid.csic.es (C. Gómez-Canela)Tel: +34 93 400 61 00Fax: +34 93 204 59 04

Abstract

In this work the effect of chlorpyrifos exposure on metabolic profiles of zebrafish

muscle was evaluated by liquid chromatography coupled to high resolution mass

spectrometry. Different chemometric tools based on the selection of Regions of Interest

and on Multivariate Curve-Resolution-Alternating Least Squares are proposed for the

analysis of the complex data sets generated in the different exposure experiments.

Analysis of Variance Simultaneous Component Analysis of changes on metabolite

peak profile areas showed significant chlorpyrifos concentration and exposure time-

dependent changes, clearly differentiating between exposed and non-exposed samples

and between short (2h) and long exposure times (6h or 24h). The changes observed in

the concentrations of 50 muscle metabolites are indicative of induction of oxidative

stress, of a general disruption of neurotransmitter metabolism, and of muscle

exhaustion. These three effects are intimately related to the toxicity of chlorpyrifos.

Capsule Abstract

The combination of LC-HRMS and chemometric tools allowed the postulation of

affected metabolic pathways by the action of chlorpyrifos on the zebrafish muscle.

Keywords

Zebrafish; environmental pollution; LC-HRMS; metabolomics; ROI-MCR-ALS.

2

11

12

13

1415

1617

1819202122

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

34

1. Introduction

Pesticides have been widely used in most sectors of the agricultural

production during the last decades. They play a key role in providing reliable

supplies of agricultural products. Nevertheless, farmers are not always aware of

their exposure or lack proper knowledge and training (Oerke and Dehne, 2004).

China, India and other developing countries have substantially increased

production of pesticides in recent years (Jin et al., 2015). Organophosphates

are among the most widely used pesticides, intended to selectively inhibit insect

acetylcholinesterases. However, their affinity for vertebrate esterases results in

severe toxic effects for non-target organisms, including aquatic vertebrates

(Koyama, 1996; Li et al., 2014; Senthilkumar et al., 2001), which makes them to

be considered as very damaging for the aquatic ecosystems (Aktar et al., 2009).

A well-known organophosphate pesticide, chlorpyrifos (CPF) is extensively

used for controlling agriculture and household pests all over the world, although

its use is currently restricted for applications in inhabited areas, due to its

proved moderately toxicity in humans (Nolan et al., 1984; Sumon et al., 2016;

U.S. Environmental Protection Agency (USEPA), 2002). Toxic effects of

organophosphate pesticides (including CPF) have been linked to nausea,

dizziness, confusion, increased heart rate, respiratory failure, and even death. A

number of previous studies have indicated that CPF exposures in aquatic

organisms are associated with a wide range of toxic effects including

nephrotoxicity, oxidative stress, genotoxic and mutagenic effects, alterations in

swimming performance, as well as effects on development (Ali et al., 2009;

Kavitha and Rao, 2008; Sandahl et al., 2005).

In the last years, zebrafish (Danio rerio), a tropical freshwater cyprinid has

become an excellent animal model for studies on molecular biology, vertebrate

development, and toxicology. Zebrafish is now widely used in many research

areas due to several advantages like easy availability, low maintenance cost

and breeding in laboratory conditions (Lawrence, 2007). Analysis of its

complete genome and the study of its gene expression patterns in a variety of

conditions revealed its high degree of likeness to humans in fundamental

genetic, developmental, and physiological processes. A 70% of protein-coding

human genes are related to genes found in zebrafish, and 84% of genes known

3

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

61

62

63

64

65

66

67

68

69

70

71

72

73

74

75

76

56

to be associated with human diseases have a zebrafish counterpart (Howe et

al., 2013). Besides, it is one of the few fish species with a well annotated

metabolome (Okuda et al., 2008). All these characteristics configured zebrafish

as the ideal vertebrate model in risk assessment for both environmental and

human health (Scholtz et al, 2008). Metabolomics has been revealed itself as

an emerging technique to evaluate physiological responses associated to drug

action, toxic effects, and metabolic disorders (Peng et al., 2015). Many different

analytical methodologies are currently used for metabolomics analysis,

including direct infusion mass spectrometry (DI-MS) (Højer-Pedersen et al.,

2008), liquid chromatography coupled mass spectrometry (LC-MS) (Bajad et al.,

2006), gas chromatography coupled to MS (GC-MS) (Lu et al., 2008), two

dimensional GC coupled to MS (GCxGC-MS) and proton nuclear magnetic

resonance (1H-NMR) (Asiago et al., 2010). In the last years, coinciding with the

emergence of the LC coupled to high resolution mass spectrometry (HRMS),

different works have used this analytical methodology because of the better

sensitivity and enhanced identification capabilities regarding to other techniques

(Dervilly-Pinel et al., 2014; Gertsman et al., 2013; Le Boucher et al., 2015). In

addition, chemometrics is presently a well-established field in chemical data

analysis and has recently been proven to be valuable in the analysis of -omic

data (Boccard and Rutledge, 2013; Chadeau-Hyam et al., 2013; Farrés et al.,

2014; Gorrochategui et al., 2016).

Thus, the aim of this study was to evaluate the changes observed in the

metabolic profiles of zebrafish muscle as a consequence of chlorpyrifos

exposure using liquid chromatography coupled to high-resolution mass

spectrometry (LC-HRMS). The choice of muscle (instead of brain or liver) was

made attending that it is a relatively large tissue, easy to dissect, and its

functioning is affected by the inhibition of acetylcholinesterase and other

putative organophosphate-targeted esterases (Tilton et al., 2011; Lopes et al.,

2011). Moreover, the use of LC-HRMS with the Orbitrap allowed better

sensitivity and enhanced identification capabilities than other techniques such

as LC-MS/MS (Gómez-Canela et al., 2013). Regions of Interest (ROI)

(Gorrochategui et al., 2015) and Multivariate Curve-Resolution-Alternating Least

Squares (MCR-ALS) (Pere-Trepat et al., 2005; Tauler, 1995) were employed as

4

77

78

79

80

81

82

83

84

85

86

87

88

89

90

91

92

93

94

95

96

97

98

99

100

101

102

103

104

105

106

107

108

109

78

general approaches for proper investigation and resolution of complex and

extensive LC-MS data sets (in full spectral scan mode). In this case, the use of

ROI allowed an easy and reliable compression of the information-rich massive

amounts of data generated by MS Orbitrap instruments. Moreover, multiple

chromatographic coelutions and difficulties in the identification of a large

number of metabolites were more easily solved by the proposed MCR-ALS

method. In addition, Analysis of Variance Simultaneous Component Analysis

(ASCA) (Barker and Rayens, 2003) was applied to MCR-ALS resolved peak

profile areas to investigate what metabolites were more influenced by CPF

levels and exposure time. To our knowledge, this is the first time that zebrafish

is used as a model to perform a study of changes in muscle metabolites

induced by the exposure to the organophosphate pesticide chlorpyrifos.

2. Experimental2.1.Chemicals and materials

Pure standard metabolites used for the analytical method development

(amino acids, sugars, nucleotides, nucleosides and others) as well as

chlorpyrifos (CPF) HPLC grade standard used for the zebrafish exposures were

supplied from Sigma-Aldrich (St. Louis, USA). Stock individual standard

solutions (1000 ng µL-1) were prepared dissolving accurate amounts of pure

standards in ultra-pure water (HPLC grade). A standard mixture sample of

these compounds was prepared at 10 ng µL-1 concentration level also in HPLC

water. Acetonitrile, dimethilsulfoxide (DMSO) and HPLC grade water were

obtained from Merck (Darmstadt, Germany). Wild-type zebrafish were obtained

from Piscicultura Superior (Barcelona, Spain) and maintained in fish water

(reverse-osmosis purified water containing 90 µg mL-1 of Instant Ocean®, 0.58

mM CaSO4.2H2O) at 28±1°C under a 12:12 light:dark photoperiod in our

facilities under standard conditions. Only adult male specimens were selected

for the tests. Animals were starved for 24 h prior to the exposure period. All

procedures were approved by the Institutional Animal Care and Use Committee

at the Research and Development Centre of the Spanish Research Council

(CID-CSIC) and conducted in accordance with institutional guidelines under a

license from the local government (DAAM 7964).

5

110

111

112

113

114

115

116

117

118

119

120

121

122

123

124

125

126

127

128

129

130

131

132

133

134

135

136

137

138

139

140

141

142

910

2.2.Acute toxicity test and exposure experiment of chlorpyrifos

First of all, the LC50 value of chlorpyrifos in zebrafish was estimated to

perform the exposure experiment. Acute toxicity assessment was performed

following the Organization for Economic Co-operation and Development

(OECD) guideline (Organization for Economic Cooperation and Development

(OECD), 1992). Preliminary exposure experiments were carried out in Pyrex®

beakers, each containing 1 L of fish water at 28±1ºC and three adult male.

Live/dead animals were counted after 24-h by gently prodding and observing

movement of appendages. According to OECD 203, fish are considered dead if

there is no visible movement (e.g. gill movements) and if touching of the caudal

peduncle produces no reaction. Initially, different CPF concentrations were

tested starting with 0.71, 1.43, 2.86, 5.71, 6, 6.5 and 7 µM. Control (fish water

only) and solvent controls showed no measurable toxicity. Following this, the

test procedure was optimized further by adjusting concentrations to enable a

better estimation of LC50. Concentrations where 100, 50 and 0% of the animals

died were repeated two or three times using 10 animals for each chosen

concentration in tanks containing 4 L of fish water to ensure reliable results.

Lethal median concentration effects and its 95% confidence interval (CI) were

estimated by fitting immobility concentration responses to the Hill regression

model (Eq. 1).

Eq. 1

where, I(Ci) is the proportion of immobile animals at concentration Ci; Ci is the

concentration of the respective compound (i); LC50 is the median lethal

concentration to the 50% of population and Hill is the shape constant. Once

stated the aquatic toxicity of CPF in zebrafish (LC50= 6 µM), the main

experiment consisted of zebrafish expositions at 2 and 5 µM, both of which lay

below the LC50 value to limit the number of dead organisms. Specifically, 2 µM

6

Hill

i

i

LCC

CI

50

1

1)(

143

144

145

146

147

148

149

150

151

152

153

154

155

156

157

158

159

160

161

162

163

164

165

166

167

168

169

170

171

172

173

174

175

1112

was chosen as a concentration below LC10 and 5 µM was chosen as a

concentration between LC10 and LC50.

2.3.Sample preparation and extraction

Sexually mature zebrafish were placed in 4 L aerated glass tanks with fish

water at 28±1°C at a rate of 4 animals L-1. Two concentrations of CPF (2µM and

5 µM) dissolved in fish water containing 0.05% DMSO were tested. Vehicle

controls were carried out in parallel under the same conditions. Five fishes were

recovered from each tank at 2, 6 and 24 h (with a total of 15 specimens in each

tank). Animals from each experimental group were anesthetized in ice. Skeletal

muscle was obtained from the caudal region, introduced in microcentrifuge

tubes, immediately frozen in liquid nitrogen and stored at -80ºC until extraction.

Figure SI1 shows the experimental design applied in the present study.

Samples were spiked with 5 ng μL-1 of IS (methionine sulfone) that was used as

extraction and analytical control. Three hundred microliters of a MeOH:H2O

(90:10) mixture were added to each 100 mg of dissected zebrafish muscle.

Then, samples were homogenized in an ultrasonic homogenizer (BRANSON

Sonifier® 150) during 5 min. After this step, samples were shaken for 20 min in

a vibrating plate and then centrifuged for 10 min at 15000 g at 4ºC. During all

the experimental procedure the samples were kept in ice. The supernatant was

filtered with 0.20 µM PTFE filters (DISMIC®-13JP, ADVANTEC®) and then kept

in amber chromatographic vials at -80 ºC (to avoid any possible degradation)

until LC-HRMS analysis.

2.4.LC-HRMS

An Orbitrap/Exactive mass spectrometer from Thermo Fischer Scientific

(Bremen, Germany) equipped with a heated electrospray ionization (H-ESI)

source was used. The system was equipped with a HTC PAL autosampler and

a Surveyor MS Plus pump. A TSKgel-Amide 80 column (2 x 250 mm, 5 µm)

from Tosoh Bioscience GmbH (Stuttgart, Germany) was chosen to carry out the

experiments being the best option in comparison to Luna NH2 100A (2 x 250

mm, 5 µm) column from Phenomenex (Torrance, California, USA) that offered

7

176

177

178

179

180

181

182

183

184

185

186

187

188

189

190

191

192

193

194

195

196

197

198

199

200

201

202

203

204

205

206

207

1314

worse results in terms of resolution. The mobile phase composition consisted of

binary mixtures with acetonitrile (A) and 5 mM ammonium acetate (pH 5) in

HPLC water (B). Gradient elution started at 75% A and 25% B, decreased to

70% A in 8 min and decreased to 40% A in 4 min, held for 5 min. Initial

conditions were attained in 3 min and the system was stabilized for 10 min. The

flow rate was set at 150 µL min−1 and 5 µL were injected. Some metabolites

were detected under positive ESI mode, some under negative ESI mode and

others by ESI+/ESI- mode. In the case of the metabolites detected by

positive/negative ionization, compounds with better resolution were used. Full

scan acquisition over a mass range of 80-800 Da was performed at 70,000

FWHM and spray voltage at 3.00 kV, capillary voltage at 30 V, skimmer voltage

at 28 V and tube lens voltage at 130 V were used. Sheath gas flow rate at 30

arbitrary units (au), auxiliary gas flow rate at 10 au and capillary temperature at

300 ºC were chosen. All mass spectrometer conditions used in this work were

following those of a previous study about the metabolome of Gammarus pulex

using LC-HRMS (Gómez-Canela et al., 2016). Solvent blanks did not contain

any of the investigated metabolites, indicating no carry-over effect in any of the

LC-HRMS runs.

3. Chemometric data arrangement, preprocessing and analysis

Chemometric data analysis included different multivariate data analysis

methods like Multivariate Curve Resolution-Alternating Least Squares (MCR-

ALS) and Analysis of Variance Simultaneous Component Analysis (ASCA).

Matlab 2013b (Mathworks Inc. Natick, MA, USA), MCR-ALS Toolbox (Jaumot et

al., 2015; Jaumot et al., 2005) and PLS Toolbox 7.3.1 (Eigenvector Research

Inc., Wenatchee, WA, USA) were used as computer programming

environments for all chemometric analyses.

8

208

209

210

211

212

213

214

215

216

217

218

219

220

221

222

223

224

225

226

227

228

229

230

231

232

233

234

235

236

237

1516

3.1.Data preprocessing (Regions of interest, ROI)

Each LC-HRMS chromatographic run recorded for every sample resulted in a

data file which was converted to NetCDF format through Thermo Xcalibur 2.1

(Thermo Scientific, San Jose, CA) software and further imported into MATLAB

environment. Using mzcdfread and mzcdf2peaks functions from the MATLAB

Bioinformatics ToolboxTM, data were loaded into MATLAB workspace and

transformed into data matrices. A reduced size data matrix containing only

relevant LC-HRMS was constructed using an in-house written MATLAB routine

which search the regions of interest (ROI) (Gorrochategui et al., 2015). ROI are

defined from mass traces with significant MS intensity values, higher than a

predetermined signal-to-noise ratio threshold. ROI should also contain a

minimum number of consecutive high density data points, compressed within a

particular mass deviation, typically set to a generous multiple of the mass

accuracy of the mass spectrometer. These ROI parameters allow ionic signals

or noise to be filtered out (Gorrochategui et al., 2015). Every ROI identifies a

possible metabolite elution profile. A ROI threshold value 105 was selected in

our data sets (the maximum MS measured intensity). In addition, the m/z

deviation, expressed in mass units, was set to 0.001 Da/e as optimal value.

Finally, the minimum number of retention times chosen to be considered in a

ROI was selected to be 5. Once these three parameters were chosen, MSroi

individual data matrices (from each chromatographic run) were column-wise

augmented to include several chromatographic runs in the same analysis. More

details about the ROI procedure can be encountered in Gorrochategui et al.

study (Gorrochategui et al., 2015). Three different augmented data matrices

(one for each exposure time 2, 6 and 24h) containing information about the 45

muscle zebrafish samples were obtained in positive and negative ionization

mode. Figure 1 shows the column-wise augmented data matrices and its

respective decomposition.

9

238

239

240

241

242

243

244

245

246

247

248

249

250

251

252

253

254

255

256

257

258

259

260

261

262

263

264

265

266

267

268

1718

3.2.MCR-ALS analysis of MSroi data matrices

Once MSroi data matrices were appropriately prepared and augmented for

their simultaneous study they were submitted to MCR-ALS analysis (Tauler,

1995). Initialization of the MCR-ALS procedure when applied to MSroi

augmented data matrices was performed using estimates of spectra profiles

found at the purest elution times (Windig, 2010). Constraints used during ALS

optimization were non-negativity for both elution and mass spectra profiles, and

spectra equal length to reduce intensity ambiguities. The number of MCR-ALS

selected components was selected according the following criteria: a) a large

part of the experimental data variance was explained; b) no further

improvement was observed when this number was increased; and c) the

shapes of the resolved profiles were physically meaningfull. This number can

also include possible contributions of the background signal and of solvent

gradient changes During the MCR-ALS optimization procedure non-negativity

constraints were applied to the component profiles (elution and spectra profiles)

and the pure spectra were normalized to equal maximum height.

Results of MCR-ALS produced a set of elution/concentration profiles

resolved for every individual component in the different chromatographic runs

simultaneously analyzed, and a set of MS pure spectra, one for every

component, common for all chromatographic runs (see Figure 1). Peak areas

and heights of the MCR-ALS resolved concentration profiles in control and

exposed samples were estimated and compared. This information was then

used for the evaluation of CPF effects and possible metabolite identification

from their MCR-ALS MS resolved spectra. Accurate mass assignations were

possible, up to four decimal places, allowing the mass (positively and negatively

ionized) identification of possible metabolites candidates. Chromatographic

experimental data were first corrected for MS sensitivity and sample size

changes using the probabilistic quotient normalization (PQN) method (Dieterle

et al., 2006). This method scales all the peak intensities of a particular spectrum

using the median of the quotients of the spectrum amplitudes in relation to a

reference spectrum that in our case was a control sample. PQN normalization is

highly recommendable for cases were size effects are high, and internal

normalization is not suitable because it would destroy relative peak information

10

269

270

271

272

273

274

275

276

277

278

279

280

281

282

283

284

285

286

287

288

289

290

291

292

293

294

295

296

297

298

299

300

301

1920

within the chromatogram (different detector sensitivities). More details about the

PQN procedure can be found in Dieterle et al. study (Dieterle et al., 2006). In

the present study, PQN was especially useful to normalize the peak areas and

to help in the identification of sample clusters (see section 3.5).

3.3. Identification and statistical analysis of zebrafish metabolites

Identification of the metabolites corresponding to the 50 resolved

chromatographic peaks was performed using Human Metaboloma Database

(HMDB) to determine metabolite compounds (Wishart et al., 2013). Different

confirmation criteria were established to ensure unequivocal identification of

metabolites:

1. By using HRMS (Orbitrap, Exactive), precursor ions were identified with

four decimal digits, restricting significantly the number of possible

candidates for each component.

2. Accurate mass measurements of the molecular and product ions should

had an error <5 ppm, with a high resolving power of 50,000 FWHM.

3. For the metabolites where standards have been used, the retention time

shift between the standards and the samples was lower than 2%.

4. Final metabolites identified were checked online in databases such as

KEGG (Kyoto Encyclopedia of Genes and Genomes), where global

maps show an overall picture of metabolism (Okuda et al., 2008)

In those cases where several metabolites were identified with the same exact

mass (m/z), the next steps were followed to choose the correct compound:

a. the assigned compound corresponded to the metabolite with the minimum

mass error value of the measured m/z;

b. protonated [M+H]+ and deprotonated [M-H]- molecules were chosen as a

priority option;

c. the candidates must exist in the zebrafish metabolome database (HMDB

in our case).

11

302

303

304

305

306

307

308

309

310

311

312

313

314

315

316

317

318

319

320

321

322

323

324

325

326

327

328

329

330

2122

Finally, the relative area fold change (FC) between controls and exposed

samples for all metabolites identified were calculated according to equation 2:

Fold change (FC) = (Peak areaexposed sample)/(Peak areacontrol sample) eq. 2

3.4. Analysis of Variance Simultaneous Component Analysis (ASCA)

ASCA is a multivariate data analysis approach that combines the statistical

advantages of ANOVA to separate the variance sources, and the advantages of

Principal Component Analysis (PCA) for eliminating covariation among

variables and explain maximum variance. In the ASCA methodology, each PCA

model is fitted to each factor matrix individually (Jansen et al., 2005). Assuming

the general factor ANOVA model for balanced data, a permutation test can be

set to check the statistical significance of the effects of all factors and their

interactions (Vis et al., 2007). The null hypothesis (H0) assumes that there is no

effect of the factor. A permutation test is performed by randomly permuting the

original data matrix (i.e. 10,000 permutations) and recalculating the sum of

squares of the factors. More details about the ASCA method and its application

in metabolomics studies are given in several works (Gorrochategui et al., 2016;

Timmerman et al., 2015).

3.5. Cluster analysis

The autoscaled peak areas studied by ASCA and previously resolved by

MCR-ALS were represented in a heat map with dendrograms showing

hierarchical clustering (clustergram). Clustergram performs a hierarchical

clustering analysis of values in the input data matrix and displays a heat map

with row and column dendrograms of the clustering. In our case, the rows in the

input matrix were the metabolites and columns, the samples. The heat map of

the metabolite peak areas was calculated from clustergram function in PLS

Toolbox 7.3.1 (Eigenvector Research Inc., Wenatchee, WA, USA) that is based

on the DeRisi et al. (DeRisi et al., 1997) and Eisen et al. (Eisen et al., 1998)

studies.

12

331

332

333

334

335

336

337

338

339

340

341

342

343

344

345

346

347

348

349

350

351

352

353

354

355

356

357

358

359

360

2324

4. Results and discussion4.1. MCR-ALS of full scan LC-HRMS chromatograms

MCR-ALS was applied to the augmented data matrices obtained from control

and exposed samples at 2h, 6h and 24h and using positive and negative MS

ionization. Between 40 and 100 components were initially postulated. However,

taking into account that the contributions related to background and solvents

(gradient), 50 components were finally resolved by MCR-ALS, explaining 99.2%

and 99.4% of data variance, respectively for positive and negative ionization

matrices. As an example, Figure 2 shows the MCR-ALS results obtained in the

analysis of the MSroi augmented data matrix at 6h of exposure time and

positive ionization. Fifteen samples divided in five controls (C1, C2, C3, C4 and

C5), five exposed at 2 µM of CPF (T1, T2, T3, T4 and T5) and five exposed at 5

µM of CPF (Z1, Z2, Z3, Z4 and Z5) were grouped. Elution profiles resolved for

these fifteen samples are represented in Figure 2A and the respective elution

profiles for the component 1 (Figure 2B). Concentrations of this metabolite

changed significantly for samples exposed to 2 µM (T1-T5) and 5 µM CPF (Z1-

Z5) compared to control samples (C1-C5). The mass spectrum resolved by

MCR-ALS for this component is shown in Figure 2C, with its most abundant ion

at m/z of 124.0074 (Taurine). Finally, in Figure 2D, the bar plot of the relative

peak areas of this first component for the control and treated samples are given.

The effect of CPF on this component is clearly shown from the drastically

increase of the peak area values of the five replicates of the MCR-ALS resolved

elution profiles of this component in control samples compared to those of the

same component in the ten treated samples.

Table 1 shows the list of metabolites finally identified, including retention

times, KEGG numbers, molecular formula and exact mass for the parental

compounds. Both positive and negative ionization modes were tested; when a

same metabolite could be detected under both methods, the data shown

correspond to the mode with better resolution. A wide range of biomolecules

were identified, comprising amino acids, sugars, nucleotides, nucleosides,

13

361

362

363

364

365

366

367

368

369

370

371

372

373

374

375

376

377

378

379

380

381

382

383

384

385

386

387

388

389

390

391

2526

organic acids, etc. In total, fifty metabolites were univocally identified, except for

L-leucine and L-isoleucine, with identical exact mass and elution times.

4.2. ASCA results: Simultaneous evaluation of CPF dose and exposure time

effects on zebrafish muscle tissues

A first overview of the amount of variance related to the design factors can be

obtained by separating this variation into contributions from the different factors.

In this study, statistical significances of the two categorical factors (i.e., dose of

CPF and exposure time) and of their interaction were evaluated separately.

ASCA results are summarized in Table 2. Results of this evaluation indicated

that a dominant part of variation is coming from residual (non-factor) variability

(residuals= 46.9%) and another part is coming from the applied factors

(≤25.2%). Strong effects were observed for the dose and exposure time factor

which were statistically assessed from the results of the permutation test (see

Table 2, p= 0.001) On the other hand, results of the permutation test confirmed

that the interaction of these two factors was not significant (p> 0.05).

PCA scores of the first component for the “dose” factor are shown in Figure

3A. Samples exposed to 5 µM CPF were significantly different to control

samples. In addition, the scores of the second component also differentiated

between control and exposed samples at 2 µM CPF, and between samples

exposed to 5 µM and to 2 µM CPF. PC1 and PC2 explained respectively

72.97% and 27.03% of the variation observed for this factor. Metabolites with

the largest loading in PC1, and therefore showing the largest changes upon

CPF exposure, were tyramine, L-valine, L-acetylcarnitine, dethiobiotin, L-

carnitine, 3-dehydroxicarnitine, L-glutamic acid and L-amminobutyric acid

(Figure 3B).

Scores for the “time” factor data matrix are shown in Figure 4A. PC1 (65.40%

of the total data variance) scores showed time dependent variations in muscle's

metabolome from 2h to 24h of exposure, whereas PC2 (34.60% of the total

data variance) showed specific effects only observed after 6h of exposure

(Figure 4A). PC1 loadings (Figure 3B) show that β-alanine, adenosine,

14

392

393

394

395

396

397

398

399

400

401

402

403

404

405

406

407

408

409

410

411

412

413

414

415

416

417

418

419

420

421

422

2728

adenosine diphosphate ribose (ADP-ribose), and adenosine monophosphate

(AMP) could be potential markers of time effects. Metabolites significant at 6h of

exposed time in PC2 loadings were oleic acid, D-maltose, β-alanine, ADP, and

ADP-ribose. Finally, PC1 scores of the interaction matrix did not show any

specific pattern, as there was no systematic increasing or decreasing trend in

the sample scores at the different doses of exposure respect to exposed times

(Table 2).

4.3. Biological interpretation of metabolic changes

Metabolites with the highest ASCA loadings (values changing up/down more

than ten times) were used for the biological interpretation and were identified as

possible biomarkers (Figures 3B and 4B). Table SI1 shows the fold changes,

FC values, of all identified metabolites calculated following equation 2

(described in the previous section 3.4.). However, of all of them, only FC values

of the 27 metabolites with higher ASCA loadings have been represented. Figure

5 shows FC values for the identified metabolites postulated as possible

biomarkers after 2, 6 and 24h of CPF exposure (plot A, B and C respectively) as

well as the significance level between exposed and control samples calculated

through a t-test. At 2h of exposition time (plot A) and 2 µM of CPF exposure,

dethiobiotin, niacinamide, myristoleic acid, L-acetylcarnitine, L-carnitine, 3-

dehydroxicarnitine, 5-oxoliproline, tyramine, taurine, L-valine, L-isoleucine/L-

leucine and inosinic acid increased their concentrations (FC values > 1)

whereas the acids docosahexanoic, linoleic, oleic, L-glutamic, GABA as well as

glutathione, hypoxanthine, adenine and D-maltose showed a decrease in their

concentrations (FC values < 1) regarding control samples. However, at 5 µM of

CPF exposure, 17 metabolites increased their concentrations whereas only 8

metabolites showed a decrease in their concentrations (see Figure 5A). When

zebrafish were exposed for 6h at 2 µM CPF, 10 muscle metabolites increased

their concentrations regarding control samples and 13 had the opposite effect.

The rest of the metabolites showed no big changes in their concentrations.

Otherwise, at 5 µM CPF, 15 metabolites increased their concentrations

regarding control samples being L-leucine/L-isoleucine and adenine the

metabolites with the highest FC values. In this case, 9 metabolites decreased

15

423

424

425

426

427

428

429

430

431

432

433

434

435

436

437

438

439

440

441

442

443

444

445

446

447

448

449

450

451

452

453

454

455

2930

their concentrations during the 6h of exposure time (Figure 5B). Finally, during

24h of exposure at 2 µM CPF, 15 metabolites increased their concentrations

regarding control samples (L-acetylcarnitine and tyramine with highest values),

and 9 metabolites decreased their concentrations. Instead, at 5 µM CPF, 18

metabolites increased their concentrations regarding control samples and only 7

compounds decreased their concentrations (Figure 5C).

Hierarchical clustering of the auto-scaled metabolite peak areas distribute

metabolites in three major clusters, termed as A, B and C in Figure 6. Cluster A

includes compounds that became depleted during CPF treatment, while

metabolites in clusters B and C increased their concentrations. This

accumulation was transient for metabolites in cluster C, as variation were only

observed after 2h of treatment, and only at the highest CPF dose tested. In

contrast, changes observed in the metabolites included in cluster B were

apparently stable, at least up to 24h of treatment; some of them appeared also

dose-dependent. Some of these metabolic changes can be related to known

toxic effects of CPF. For example, sub chronic or chronic exposure to CPF (and

other organophosphorates) showed possible induction of oxidative stress,

allegedly a main mechanism of organophosphorous toxicity (Čolović et al.,

2013) (Table 3). This is consistent with the observed glutathione depletion and

with the consequent increase of concentration of its precursor 5-oxoliproline (Jin

et al., 2015). The known effect of CPF as inhibitor of acetylcholinesterase

(AChE), and the subsequent increase of physiological levels of acetylcholine

(ACh), are also consistent with the changes observed in the levels of several

neuroactive substances such as GABA, L-glutamate, tyramine and taurine

(Table 3). In addition, other variations were observed which can be related with

changes in the energy metabolism of muscle. For example, there is a general

concentration decrease of at least four long-chain fatty acids

(docosahexanoate, linoleate, oleate and myristoleate), followed by an increase

of concentrations of different carnitine derivatives and of some of its short

organic acid conjugates. Since these conjugates are related to fatty acid

degradation, these results suggest a possible over-consumption of the reserve

lipids. A similar conclusion can be drawn from the depletion of maltose levels (a

likely metabolite from the digestion of muscular glycogen) and from the increase

16

456

457

458

459

460

461

462

463

464

465

466

467

468

469

470

471

472

473

474

475

476

477

478

479

480

481

482

483

484

485

486

487

488

3132

of lactic acid concentrations (sugar anoxic metabolism) as well as from the

glycolysis of the intermediate glycerol-3-phosphate (Table 1/Table 3). Similarly,

the increase of L-valine, L-leucine/isoleucine and L-glutamine may also be

related to protein degradation, a common last resource for restoring energy

levels. All these concentration changes are consistent with an exhaustion of the

muscular tissue, likely to occur with the hyper contracture of axial muscle fibers

that was also observed in zebrafish embryos treated with chlorpyrifos-oxon, the

recognized neurotoxic derivative of CPF (Faria et al., 2015). Other observed

changes, like those observed for the four metabolites related to purine

metabolism and cofactor synthesis, are not directly related to known

physiological effects of CPF, although the decrease in ascorbate levels could be

related to the oxidative stress (Table 3).

Conclusions

The analysis of metabolite changes in zebrafish muscle upon exposure to

chlorpyrifos (CPF) reflected different adverse effects related to

organophosphate toxicity, such as oxidative stress, disruption of

neurotransmitter metabolism, and energy exhaustion. These effects occurred at

concentrations much higher than those found in water bodies or soils, below the

10 µg/L or 100 µg/Kg marks, respectively (Ensminger et al., 2011; Zhang et al.,

2012), but comparable to those used for human health risk assessment in fish

or rodents (10 µM and 0.5-5 mg/Kg b.w., respectively) (Eaton et al., 2008;

Koenig et al., 2016). LC-HRMS data were performed using a chemometric

strategy based on the selection of the LC-MS regions of interest (ROI) and

multivariate curve resolution (MCR-ALS) simultaneous analysis of control and

CPF exposed samples. ASCA of MCR-ALS resolved LC-MS peak profile areas

revealed that both CPF dose and exposure times produced significant changes

in metabolite concentrations compared to zebrafish control samples. Moreover,

samples exposed at 5 µM and 2 µM CPF showed dose-dependent effects in

comparison with control samples. Significant changes in metabolite

concentrations were observed, which were discussed in terms of their possible

involvement in several different biochemical pathways, with reflected metabolite

17

489

490

491

492

493

494

495

496

497

498

499

500

501

502

503

504

505

506

507

508

509

510

511

512

513

514

515

516

517

518

519

520

3334

accumulation and depletion processes. Thus, a general fatty acid degradation

effect was observed in agreement with an increase of carnitine and of some of

its short organic acid conjugates. Observed changes in muscle metabolome of

zebrafish were consistent with the induction of oxidative stress, a general

disruption of neurotransmitter metabolism and the exhaustion of the muscle

itself probably related to the excess of ACh function in motoneurons, intimately

related to the toxic effect of chlorpyrifos and other organophosphate pesticides.

In summary, the combination of LC-HRMS metabolomics and chemometric data

analysis allowed the postulation of possible affected metabolic pathways by the

action of the CPF pesticide on the zebrafish muscle.

Acknowledgements

The research leading to these results has received funding from the European

Research Council under European Union’s Seven Framework Programme

(FP/2007-2013)/ERC Grant Agreement n.320737.

18

521

522

523

524

525

526

527

528

529

530

531

532

533

534

535

3536

19

Metabolites KEGG numberExact mass

(m/z)Molecular formula

Docosahexanoic acid C06429 327.2334 C22H32O2

Linoleic acid C01595 279.2335 C18H32O2

β-cyprinol sulphate C05468 531.2999 C27H48O8S

Oleic acid C00712 281.2477 C18H34O2

Palmitic acid C00249 255.2331 C16H32O2

Myristoleic acid C08322 227.2002 C14H26O2

Niacinamide C00153 123.0550 C6H6NO2

L-Lactic acid C00186 89.0243 C3H6O3

Adenine C00147 134.0471 C5H5N5

Adenosine C00212 268.1037 C10H13N5O4

Hypoxanthine C00262 137.0455 C5H4N4O

2-Ketobutyric acid C00109 203.0560 C4H6O3

N2-succinyl-L-ornithine C03415 267.0752 C9H16N2O5

Inosine C00294 267.0734 C10H12N4O5

Threonic acid C01620 135.0303 C4H8O5

Tyramine C00483 120.0805 C8H11NO [M+H-H

L-Phenylalanine C00079 164.0717 C9H11NO2

AMP C00020 346.0567 C10H14N5O7P

Glycerol-3-phosphate C00093 171.0070 C3H9O6P

Piperidine C01746 86.0961 C5H11N

L-Isoleucine C00407130.0873 C6H13NO2

L-Leucine C00123

Inosinic acid C00130 347.0390 C10H13N4O8P

Phosphoric acid C00009 96.9693 H3PO4

Taurine C00245 124.0074 C2H7NO3S

Glutathione C00051 306.0763 C10H17N3O6S

ADP C00008 426.0234 C10H15N5O10P2

ɣ-Aminobutyric acid (GABA) C00334 102.0558 C4H9NO2

L-Glutamic acid C00025 146.0456 C5H9NO4

L-Valine C00183 118.0857 C5H11NO2

Propionylcarnitine C03017 218.1384 C10H19NO4

L-Proline C00148 114.0559 C5H9NO2

N-Acetylhistidine C02997 198.0871 C8H11N3O3

Glucose-6-phosphate C00668 259.0225 C6H13O9P

Phosphocreatine C02305 210.0282 C4H10N3O5P

ADP-ribose C00301 540.0534 C15H23N5O14P2 [M-H-H

L-Acetylcarnitine C02571 204.1229 C9H18NO4

β-Alanine C00099 90.0547 C3H7NO2

Creatine C00300 130.0621 C4H9N3O2

N-Acetylornithine C00437 173.0928 C7H14N2O3

ATP C00002 505.9898 C10H16N5O13P3

L-Threonine C00188 118.0507 C4H9NO3

5-oxoproline C01879 130.0496 C5H7NO3

L-Glutamine C00064 145.0616 C5H10N2O3

Dethiobiotin C01909 215.1385 C10H18N2O3

Creatinine C00791 112.0519 C4H7N3O2

L-Carnitine C00318 162.1124 C7H15NO3

D-Maltose C00208 325.1122 C12H22O11 [M+H-H

536

3738

Table 1. Tentative identification of zebrafish metabolites associated to MCR-ALS

resolved elution profiles with significant peak area changes between the different

experimental conditions tested.

20

537

538

539

540

541

542

543

544

545

546

547

548

549

550

551

3940

Table 2. ASCA results: Significance and partitioning of the total variance into the

individual terms corresponding to factors and interaction.

Factor Cum EigenVala Percentage of variationb

Significance (p-value)

Cc 8.52 17.1 <0.01

Td 12.59 25.2 <0.01

C x T 5.39 10.8 0.998

Residuals 46.9

aCumulative EigenVal. bPercentage of variation expressed as sums of squared

deviations from the overall mean. c C= Concentration factor (controls, 5 µM and 2 µM

levels). dT= exposed time factor (2h, 6h and 24h levels).

21

553

554

555

556

557

558

559

560

561

562

563

564

565

566

567

568

4142

Table 3. Most important metabolites from ASCA analysis and their possible metabolic

pathway changed upon CPF exposure.

Metabolite KEGG

number Cluster Pathway (I) Pathway (II) Possible metabolic change

Threonic acid C01620 A Ascorbate

metabolism Cofactor

metabolism

Disruption of cofactor

metabolism Dethiobiotin C01909 B Biotin metabolism

Niacinamide C00153 C Nicotinate

metabolism

Docosahexanoic acid C06429 A

Fatty acid

metabolism

Fatty acid

metabolism Depletion of fatty acids

Linoleic acid C01595 A

Oleic acid C00712 A

Myristoleic acid C08322 A

L-Acetylcarnitine C02571 B

Fatty acid

metabolism

Fatty acid

metabolism Fatty acid degradation

L-Carnitine C00318 B

Propionylcarnitine C03017 C

3-Dehydroxicarnitine C05543 C

Glutathione C00051 A Glutathione

metabolism

Glutathione

metabolism Glutathione depletion

5-Oxoliproline C01879 B

GABA C00334 A

Neuroactive ligand Neuroactive

ligand

GABA/Glutamate depletion,

disruption of other

neuroactive ligands

L-Glutamic acid C00025 A

Tyramine C00483 B

Taurine C00245 C

L-Valine C00183 B

Protein degradation Protein

degradation

Protein degradation

(starvation) L-Isoleucine C00407 B

L-Glutamine C00064 C

Hypoxanthine C00262 A

Purine metabolism Purine

metabolism Purine metabolism disruption

Inosinic acid C00130 B

Adenine C00147 B

Inosine C00294 C

D-Maltose C00208 A

Sugar metabolism Sugar

metabolism

Starvation (glycogen to

lactate metabolism) Glycerol-3-phosphate C00093 B

L-Lactic acid C00186 C

22

569

570

571

4344

Figure 1. Column-wise augmented data matrices containing 45 zebrafish muscle samples in each exposure time. Also, its respective decomposition after MCR-ALS is showed.

23

572

573574

4546

Figure 2. Example of MCR-ALS results in the analysis of MSroi augmented data matrix including fifteen samples: five control samples (C1, C2, C3, C4 and C5), five 2 µM CPF exposed samples (T1, T2, T3, T4 and T5) and five 5 µM CPF exposed samples (Z1, Z2, Z3, Z4 and Z5), at 6h of exposure time. (A) Elution profiles of the 50 components resolved by MCR-ALS, where each component (metabolite) is shown with a different color. (B) MCR-ALS resolved elution profiles of component 1 in these fifteen (five control and ten exposed) samples (C) Mass spectrum of the first component resolved by MCR-ALS with its most abundant ion at m/z of 124.0074, and (D) Profile peak areas of the component 1 in these fifteen samples.

24

575

576

577578579580581582583584585

586

587

4748

Figure 3. ASCA results: (A) SCA scores plot of the “dose” factor matrix. Color symbols indicate the different samples studied: yellow diamonds are controls, green squares are the zebrafish samples at 2 µM of CPF and red triangles are the zebrafish samples at 5 µM of CPF. (B) SCA PC1 loadings plot of dose sample factor matrix with their metabolite identification in the x-axis. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article).

25

588

589

590591592593594595

596

597

4950

Figure 4. ASCA results: (A) SCA scores plot of the “time” factor matrix. Symbols indicate the different samples studied. Red diamonds, green squares, and blue triangles correspond to replicate zebrafish samples collected after 2h, 6h, and 24h of exposure time, respectively. (B) SCA PC1 loadings plot of the replicate time samples factor data matrix with their metabolite identification in the x-axis. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article).

26

598599600601602603604605

5152

Figure 5. Fold change (FC) of the selected metabolites at 2h (A), 6h (B) and 24h (C) of exposition time. The fold change has been calculated from equation 2 in section 3.4. Identification of the selected metabolites is given in the x-axis of the plots. *: p value < 0.05; **: p value < 0.01; ***: p value < 0.005

27

606

607608609610

5354

Figure 6. Heat map of CPF effects on zebrafish metabolite concentration changes obtained from the auto-scaled peak areas of the selected

28

611612

5556

metabolites at each experimental condition (five replicates). Metabolites are represented by their KEGG number. A, B and C show three major

clusters defined by hyerarchical clustering (after PQN normalization (Dieterle et al., 2006)).

29

613

614

5758

REFERENCES

Aktar, W., Sengupta, D., Chowdhury, A., 2009. Impact of pesticides use in agriculture: their benefits and hazards. Interdisciplinary toxicology 2, 1-12.

Ali, D., Nagpure, N.S., Kumar, S., Kumar, R., Kushwaha, B., Lakra, W.S., 2009. Assessment of genotoxic and mutagenic effects of chlorpyrifos in freshwater fish Channa punctatus (Bloch) using micronucleus assay and alkaline single-cell gel electrophoresis. Food Chem. Toxicol. 47, 650-656.

Asiago, V.M., Alvarado, L.Z., Shanaiah, N., Gowda, G.A.N., Owusu-Sarfo, K., Ballas, R.A., Raftery, D., 2010. Early detection of recurrent breast cancer using metabolite profiling. Cancer Res. 70, 8309-8318.

Bajad, S.U., Lu, W., Kimball, E.H., Yuan, J., Peterson, C., Rabinowitz, J.D., 2006. Separation and quantitation of water soluble cellular metabolites by hydrophilic interaction chromatography-tandem mass spectrometry. J. Chromatogr. A 1125, 76-88.

Barker, M., Rayens, W., 2003. Partial least squares for discrimination. J. Chemom. 17, 166-173.Boccard, J., Rutledge, D.N., 2013. A consensus orthogonal partial least squares discriminant

analysis (OPLS-DA) strategy for multiblock Omics data fusion. Analytica Chimica Acta 769, 30-39.

Chadeau-Hyam, M., Campanella, G., Jombart, T., Bottolo, L., Portengen, L., Vineis, P., Liquet, B., Vermeulen, R.C.H., 2013. Deciphering the complex: Methodological overview of statistical models to derive OMICS-based biomarkers. Environmental and Molecular Mutagenesis 54, 542-557.

Čolović, M.B., Krstić, D.Z., Lazarević-Pašti, T.D., Bondžić, A.M., Vasić, V.M., 2013. Acetylcholinesterase inhibitors: Pharmacology and toxicology. Curr. Neuropharmacol. 11, 315-335.

DeRisi, J.L., Iyer, V.R., Brown, P.O., 1997. Exploring the metabolic and genetic control of gene expression on a genomic scale. Science 278, 680-686.

Dervilly-Pinel, G., Chereau, S., Cesbron, N., Monteau, F., Le Bizec, B., 2014. LC-HRMS based metabolomics screening model to detect various β-agonists treatments in bovines. Metabolomics 11, 403-411.

Dieterle, F., Ross, A., Schlotterbeck, G., Senn, H., 2006. Probabilistic quotient normalization as robust method to account for dilution of complex biological mixtures. Application in1H NMR metabonomics. Anal. Chem. 78, 4281-4290.

Eaton, D.L., Daroff, R.B., Autrup, H., Bridges, J., Buffler, P., Costa, L.G., Coyle, J., McKhann, G., Mobley, W.C., Nadel, L., Neubert, D., Schulte-Hermann, R., Spencer, P.S., 2008. Review of the toxicology of chlorpyrifos with an emphasis on human exposure and neurodevelopment. Critical Reviews in Toxicology 38, 1-125.

Eisen, M.B., Spellman, P.T., Brown, P.O., Botstein, D., 1998. Cluster analysis and display of genome-wide expression patterns. Proceedings of the National Academy of Sciences of the United States of America 95, 14863-14868.

Ensminger, M., Bergin, R., Spurlock, F., Goh, K.S., 2011. Pesticide concentrations in water and sediment and associated invertebrate toxicity in Del Puerto and Orestimba Creeks, California, 2007-2008. Environmental Monitoring and Assessment 175, 573-587.

Faria, M., Garcia-Reyero, N., Padrós, F., Babin, P.J., Sebastián, D., Cachot, J., Prats, E., Arick Ii, M., Rial, E., Knoll-Gellida, A., Mathieu, G., Le Bihanic, F., Escalon, B.L., Zorzano, A., Soares, A.M.V.M., Ralduá, D., 2015. Zebrafish Models for Human Acute Organophosphorus Poisoning. Sci. Rep. 5. Article num. 15591.

Farrés, M., Piña, B., Tauler, R., 2014. Chemometric evaluation of Saccharomyces cerevisiae metabolic profiles using LC–MS. Metabolomics 11, 210-224.

Gertsman, I., Gangoiti, J.A., Barshop, B.A., 2013. Validation of a dual LC-HRMS platform for clinical metabolic diagnosis in serum, bridging quantitative analysis and untargeted metabolomics. Metabolomics, 1-12.

Gómez-Canela, C., Cortés-Francisco, N., Ventura, F., Caixach, J., Lacorte, S., 2013. Liquid chromatography coupled to tandem mass spectrometry and high resolution mass spectrometry as analytical tools to characterize multi-class cytostatic compounds. Journal of Chromatography A 1276, 78-94.

30

615

616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650

651652653654655656

657658659660661662663664665666667668669

5960

Gómez-Canela, C., Miller, T.H., Bury, N.R., Tauler, R., Barron, L.P., 2016. Targeted metabolomics of Gammarus pulex following controlled exposures to selected pharmaceuticals in water. Sci. Total Environ. 562, 777-788.

Gorrochategui, E., Jaumot, J., Tauler, R., 2015. A protocol for LC-MS metabolomic data processing using chemometric tools. Protocol Exchange. doi:10.1038/protex.2015.102.

Gorrochategui, E., Lacorte, S., Tauler, R., Martin, F.L., 2016. Perfluoroalkylated Substance Effects in Xenopus laevis A6 Kidney Epithelial Cells Determined by ATR-FTIR Spectroscopy and Chemometric Analysis. Chem. Res. Toxicol. 29, 924-932.

Gorrochategui, E.; Jaumot, J.; Lacorte, S.; Tauler, R., Data analysis strategies for targeted and untargeted LC-MS metabolomic studies: Overview and workflow. TrAC Trends Anal. Chem. 2016, 82, 425-442.

Højer-Pedersen, J., Smedsgaard, J., Nielsen, J., 2008. The yeast metabolome addressed by electrospray ionization mass spectrometry: Initiation of a mass spectral library and its applications for metabolic footprinting by direct infusion mass spectrometry. Metabolomics 4, 393-405.

Howe, K., Clark, M.D., Torroja, C.F., Torrance, J. et al. 2013. The zebrafish reference genome sequence and its relationship to the human genome. Nature 496, 498-503.

Jansen, J.J., Hoefsloot, H.C.J., Van Der Greef, J., Timmerman, M.E., Westerhuis, J.A., Smilde, A.K., 2005. ASCA: Analysis of multivariate data obtained from an experimental design. J. Chemom. 19, 469-481.

Jaumot, J., de Juan, A., Tauler, R., 2015. MCR-ALS GUI 2.0: New features and applications. Chemom. Intell. Lab. Syst. 140, 1-12.

Jaumot, J., Gargallo, R., De Juan, A., Tauler, R., 2005. A graphical user-friendly interface for MCR-ALS: A new tool for multivariate curve resolution in MATLAB. Chemom. Intell. Lab. Syst. 76, 101-110.

Jin, Y., Liu, Z., Peng, T., Fu, Z., 2015. The toxicity of chlorpyrifos on the early life stage of zebrafish: Asurvey on the endpoints at development, locomotor behavior, oxidative stress and immunotoxicity. Fish Shellfish Immunol. 43, 405-414.

Kavitha, P., Rao, J.V., 2008. Toxic effects of chlorpyrifos on antioxidant enzymes and target enzyme acetylcholinesterase interaction in mosquito fish, Gambusia affinis. Environ. Toxicol. Pharmacol. 26, 192-198.

Koenig, J.A., Dao, T.L., Kan, R.K., Shih, T.M., 2016. Zebrafish as a model for acetylcholinesterase-inhibiting organophosphorus agent exposure and oxime reactivation, in: Laskin, J.D. (Ed.), Countermeasures against Chemical Threats. Blackwell Science Publ, Oxford, pp. 68-77.

Koyama, J., 1996. Vertebral deformity susceptibilities of marine fishes exposed to herbicide. Bulletin of Environmental Contamination and Toxicology 56, 655-662.

Lawrence, C., 2007. The husbandry of zebrafish (Danio rerio): A review. Aquaculture 269, 1-20.Le Boucher, C., Courant, F., Royer, A.L., Jeanson, S., Lortal, S., Dervilly-Pinel, G., Thierry, A.,

Le Bizec, B., 2015. LC–HRMS fingerprinting as an efficient approach to highlight fine differences in cheese metabolome during ripening. Metabolomics 11, 1117-1130.

Lopes, R.M., Silva, M.V., de Salles, J.B., Bastos, V., Bastos, J.C., 2014. Cholinesterase activity of muscle tissue from freshwater fishes: characterization and sensitivity analysis to the organophosphate methyl-paraoxon. Environmental Toxicology and Chemistry 33, 1331-1336.

Li, W., Tai, L., Liu, J., Gai, Z., Ding, G., 2014. Monitoring of pesticide residues levels in fresh vegetable form Heibei Province, North China. Environ. Monit. Assess. 186, 6341-6349.

Lu, H., Liang, Y., Dunn, W.B., Shen, H., Kell, D.B., 2008. Comparative evaluation of software for deconvolution of metabolomics data based on GC-TOF-MS. TrAC, Trends Anal. Chem. 27, 215-227.

Nolan, R.J., Rick, D.L., Freshour, N.L., Saunders, J.H., 1984. Chlorpyrifos: Pharmacokinetics in human volunteers. Toxicology and Applied Pharmacology 73, 8-15.Organization for Economic Cooperation and Development (OECD), 1992. Test No. 203: Fish, Acute Toxicity Test. OECD Guidelines for the Testing of Chemicals. OECD Publishing, Section 2, OECD Publishing, Paris.Oerke, E.C., Dehne, H.W., 2004. Safeguarding production—losses in major crops and the role

of crop protection. Crop Protection 23, 275-285.Okuda, S., Yamada, T., Hamajima, M., Itoh, M., Katayama, T., Bork, P., Goto, S., Kanehisa, M.,

2008. KEGG Atlas mapping for global analysis of metabolic pathways. Nucleic Acids Research 36, W423-W426.

31

670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728

6162

Peng, B., Li, H., Peng, X.-X., 2015. Functional metabolomics: from biomarker discovery to metabolome reprogramming. Protein & Cell 6, 628-637.

Pere-Trepat, E., Lacorte, S., Tauler, R., 2005. Solving liquid chromatography mass spectrometry coelution problems in the analysis of environmental samples by multivariate curve resolution. J. Chromatogr. A 1096, 111-122.

Sandahl, J.F., Baldwin, D.H., Jenkins, J.J., Scholz, N.L., 2005. Comparative thresholds for acetylcholinesterase inhibition and behavioral impairment in coho salmon exposed to chlorpyrifos. Environ. Toxicol. Chem. 24, 136-145.

Senthilkumar, K., Kannan, K., Subramanian, A., Tanabe, S., 2001. Accumulation of organochlorine pesticides and polychlorinated biphenyls in sediments, aquatic organisms, birds, bird eggs and bat collected from south India. Environmental Science and Pollution Research 8, 35-47.

Scholz, S., Fischer, S., Gundel, U., Kuster, E., Luckenbach, T., Voelker, D., 2008. The zebrafish embryo model in environmental risk assessment--applications beyond acute toxicity testing. Environ Sci Pollut Res Int 15, 394-404.

Sumon, K.A., Rico, A., Ter Horst, M.M.S., Van den Brink, P.J., Haque, M.M., Rashid, H., 2016. Risk assessment of pesticides used in rice-prawn concurrent systems in Bangladesh. Sci. Total Environ. 568, 498-506.

Tauler, R., 1995. Multivariate curve resolution applied to second order data. Chemom. Intell. Lab. Syst. 30, 133-146.

Tilton, F.A., Bammler, T.K., Gallagher, E.P., 2011. Swimming impairment and acetylcholinesterase inhibition in zebrafish exposed to copper or chlorpyrifos separately, or as mixtures. Comparative Biochemistry and Physiology C-Toxicology & Pharmacology 153, 9-16.

Timmerman, M.E., Hoefsloot, H.C.J., Smilde, A.K., Ceulemans, E., 2015. Scaling in ANOVA-simultaneous component analysis. Metabolomics 11, 1265-1276.

U.S. Environmental Protection Agency (USEPA), 2002. Interim Reregistration Eligibility Decision for Chlorpyrifos.Vis, D.J., Westerhuis, J.A., Smilde, A.K., van der Greef, J., 2007. Statistical validation of megavariate effects in ASCA. BMC Bioinformatics 8.

Windig, W., 2010. Two-Way Data Analysis: Detection of Purest Variables. Compr. Chemom. Elsevier, pp. 275-307.

Wishart, D.S., Jewison, T., Guo, A.C., Wilson, M., Knox, C., Liu, Y., Djoumbou, Y., Mandal, R., Aziat, F., Dong, E., Bouatra, S., Sinelnikov, I., Arndt, D., Xia, J., Liu, P., Yallou, F., Bjorndahl, T., Perez-Pineiro, R., Eisner, R., Allen, F., Neveu, V., Greiner, R., Scalbert, A., 2013. HMDB 3.0-The Human Metabolome Database in 2013. Nucleic Acids Res. 41, D801-D807.

Zhang, X., Starner, K., Spurlock, F., 2012. Analysis of Chlorpyrifos Agricultural Use in Regions of Frequent Surface Water Detections in California, USA. Bulletin of Environmental Contamination and Toxicology 89, 978-984.

32

729730

731732733734735736737738739740741742743744745746747748749750751752753754755756757

758759760761762763764

765766767

768

769

6364