Ecotoxicology and Environmental Safety 145 (2017) 583–590

Contents lists available at ScienceDirect

Ecotoxicology and Environmental Safety

journal homepage: www.elsevier.com/locate/ecoenv

Analysis of neurobehavioural data by chemometric methods in MARK

ecotoxicological studies
⁎
Cristian Gómez-Canelaa, , Eva Pratsb, Romà Taulera, Demetrio Raldúaa
a
Department of Environmental Chemistry, IDAEA-CSIC, Jordi Girona 18-26, 08034 Barcelona, Catalonia, Spain
b
Centre d′Investigació i Desenvolupament, CID-CSIC, Jordi Girona 18-26, 08034 Barcelona, Catalonia, Spain

A R T I C L E I N F O A B S T R A C T

Keywords: Incorporation of chemometric tools in behavioural data management workflows allows for the early identifi-
Zebrafish cation of most relevant endpoints complementarily to statistical confirmatory approaches. In this work, the
Open field test effects of two model neurotoxicants, chlorpyrifos (CPF) and nicotine, exposures on behavioural profiles of adult
Behaviour zebrafish at three different times (2, 6 and 24 h) were evaluated using open field test (OFT) paradigm experi-
Chemometric tools
ments. Two chemometric methods like Principal Component Analysis (PCA) and Analysis of Variance-
Simultaneous Component Analysis (ASCA) have been used to interpret the changes observed in the obtained
behavioural data. A decreased of the locomotor activity, an anxiolytic effect and an altered exploratory beha-
viour were the most affected behavioural endpoints in the CPF exposures. However, an increase of the locomotor
activity and an anxiogenic effect were observed in the nicotine exposures. Finally, an excellent correlation
between the ASCA results and the results obtained using traditional statistical procedures for both compounds
were encountered.

1. Introduction available for using in laboratory animals, including zebrafish, for

studying specific neurotoxic effects of environmental pollutants and
Humans are routinely exposed to a wide range of environmental drugs (Levin and Chen, 2004; Mora-Zamorano et al., 2016; Oliveri
pollutants potentially inducing neurotoxicity (De Duffard and Duffard, et al., 2015). The recent development of different video-tracking soft-
1996; Jones and Miller, 2008). Different animal models have been used ware in neuroscience research has enabled standardize and automate
in attempt to understand the neurobiological basis of the adverse effects behavioural endpoints, promoting reproducibility and allowing for
induced by these neurotoxicants on the human central and peripheral multiple endpoints to be recorded at once (Cachat et al., 2011). Specific
nervous system. Zebrafish (Danio rerio) is a vertebrate model species video-tracking systems to automate behavioural studies in zebrafish are
increasingly used in biomedical research, drug discovery and safety commercially available from companies such as Noldus Information
pharmacology (Brittijn et al., 2009; Lieschke and Currie, 2007; Raldúa Technology (Ethovision® XT) and Viewpoint (ZebraLab). These systems
and Piña, 2014). Low cost, well-established molecular genetic tools and are able to analyse simultaneously dozens of dependent variables, in-
high conservation of the main physiological processes involved in cluding those related with movement, location, path or direction, for
nervous system morphogenesis and maintenance all combine to make each sampling time (commonly 30 samples per second) and experi-
zebrafish a promising animal model for neuroscience research, in- mental condition. Once obtained all these data from the video-tracking
cluding neurobehavioural toxicology (Babin et al., 2014; Faria et al., systems, behavioural endpoints are usually compared between control
2015; Raldúa and Piña, 2014; Stewart et al., 2015). Until recently, the and treated animals by using standard statistical analysis. Thus, after
main criterion for determining the neurotoxicity of one chemical was assessing the normality of the different distributions and the homo-
the presence of neuropathological adverse effects. However, beha- scedasticity of the groups, the most suitable parametric or non-para-
vioural endpoints have been recently incorporated to the neurotox- metric test are usually selected. This process, however, can become
icology screening protocols, and these functional endpoints are now extremely time-consuming, as often hundreds of distributions need to
used routinely to detect and characterize potential neurotoxicity of be tested individually. Therefore, methodologies allowing the prior-
chemicals (De Duffard and Duffard, 1996). itization of those endpoints really relevant for a further statistical
A wide variety of behavioural methods and paradigms are currently analysis and confirmation are urgently needed in the neurobehavioural

⁎
Corresponding author.
E-mail address: cristian.gomez@cid.csic.es (C. Gómez-Canela).

http://dx.doi.org/10.1016/j.ecoenv.2017.08.013
Received 18 April 2017; Received in revised form 2 August 2017; Accepted 3 August 2017
0147-6513/ © 2017 Elsevier Inc. All rights reserved.
C. Gómez-Canela et al. Ecotoxicology and Environmental Safety 145 (2017) 583–590

toxicology field. Although multivariate data analysis chemometric in order to avoid the inter-sessions habituation response in the beha-
techniques, including Principal Component Analysis, PCA (Farrés et al., vioural tests.
2015) and the Analysis of Variance and Simultaneous Component
Analysis (ASCA) (Jansen et al., 2005; Smilde et al., 2005), could be 2.3. Behavioural test
extremely useful for this role, their use in zebrafish neurobehavioural
research is still scarce. The zebrafish open field test was performed according to Grossman
Thus, our hypothesis in this work is that chemometric tools are et al. (2010). An experimental setup for monitoring and recording 4 fish
useful to predict the most relevant behavioural endpoints altered by simultaneously was used (Supporting information, Fig. S1). OFT was
neurotoxicants. As a proof of principle of this new approach, adult performed in four circular trunked conical white plastic tanks (testing
zebrafish have been exposed to chlorpyrifos (CPF) and nicotine, two tanks; 22.5 cm lower diameter × 25.0 cm upper diameter × 26.0 cm
well-known neurotoxic compounds modulating anxiety-like behaviour height) containing 5 L of fish water at 28 °C. Following the exposure to
in mammalian models (Lopez-Crespo et al., 2007). Adults were selected the carrier and 5 μM CPF or 50 μM nicotine for the selected times, fish
instead of embryos or larvae, as they exhibit a more complex behaviour were individually transferred to the testing tanks. The first 6 min of the
(Cachat et al., 2011). The open field test (OFT) and the video tracking trial were video-recorded (MPEG1 format, 30 fps at 640 × 480) with a
system Ethovision XT 11.5 (Noldus Technologies) was selected for the video surveillance software (Security Monitor Pro; DeskShare Inc,
behavioural analyses, as this experimental paradigm evaluates the Plainview, NY, USA) linked to USB 2.0 web-cameras (Microsoft LifeCam
natural neophobic response, providing information on both locomotor Studio; Microsoft Corporation, Redmon, WA, USA) placed on the top of
and anxiety-related behaviour (Stewart et al., 2010). The obtained data the testing tanks. Two anti-flicker LED tubes (TUT8-ST28-NFL; AS de
were first analysed with the most time-consuming statistical procedures LED®, Valencia, Spain) mounted on both sides of the test tanks provided
and then, the behavioural profiles obtained with the two chemometric uniform illumination for the video-recording. After recording, videos
methods (PCA and ASCA) were compared. Results of this study strongly were analysed by Ethovision XT 11.5 and the distance travelled (cm),
supported that chemometric analysis of video-tracking data not only average velocity (cm/s) and the duration and frequency of immobile
offers new possibilities to extract information, but also illustrates new (changed area ≤ 3%), mobile (changed area between 3.1% and 69.9%),
effects obtained in behavioural assays, allowing for a save of time and and highly mobile (changed area ≥ 70%) states were determined. In
for a reduction of the efforts in the statistical confirmatory analysis of order to assess any effect of the CPF and nicotine treatments on the
the most relevant endpoints. thigmotaxism, OFT arena was virtually divided into periphery (area
within 5 cm from the walls) and center. Moreover, as zebrafish exhibit a
2. Material and methods robust preferred loci (homebases) in the OFT (Cachat et al., 2013;
Stewart et al., 2010), arena was divided in four virtual zones and
2.1. Animals and housing identification of the homebase (HB) and non-homebase (NH) quadrants
was based in the criteria described by Stewart et al. (2010). Behavioural
Adult wild-type zebrafish, obtained from a local commercial dis- endpoints analysed in both HB and NH included the distance travelled
tributor (Piscicultura Superior, Barcelona, Spain), were maintained in and the cumulative time spent.
fish water [reverse-osmosis purified water containing 90 μg/mL of
Instant Ocean (Aquarium Systems, Sarrebourg, France) and 0.58 mM 2.4. Statistical analysis
CaSO4·2H2O] at 28 ± 1 °C and a 12 L:12D photoperiod in the Research
and Development Centre of the Spanish Research Council (CID-CSIC) Statistical analyses were performed using SPSS software v22 (IBM,
facilities for at least 5 weeks before the experiments. All procedures Chicago, IL, USA). The data were presented as the mean ± standard
were conducted in accordance with the institutional guidelines under a error of the mean (SEM) unless stated otherwise. Normal distribution
license from the local government (DAMM 7964) and were approved by (Kolmogorov-Smirnov test) and homoscedasticity (Levene's test) of the
the Institutional Animal Care and Use Committee at the Spanish data were analysed. Pairwise statistical significance was determined
Research Council. with Student's t-test or Mann-Whitney U-test as appropriate. Differences
were considered statistically significant at p < 0.05. For a simple ex-
2.2. Experimental procedure perimental design like the one used here, 216 normality tests and 108
homoscedasticity tests were needed before any inference could be done
Chlorpyrifos (CPF; CAS♯2921-88-2, purity ≥ 99.5%) and (-)-nico- about the most suitable statistical test. Student t-test was used to
tine (nicotine; CAS♯54-11-5, purity ≥ 99%) were purchased from compare the different behavioural endpoints between control and
Sigma-Aldrich (St. Louis, MO, USA). Stock solution 10 mM CPF was treated fish when distributions fitted normality and homoscedasticity
prepared in dimethyl sulfoxide (DMSO). A range finding test was con- criteria. On the other hand, Mann-Whitney U-test was used when dis-
ducted for each compound in order to find the maximum tolerated tributions were not normal and/or variances were unequal, even con-
concentration (MTC), defined as the concentration for which no leth- sidering that some reports suggest that homoscedasticity is also needed
ality was observed above that seen in vehicle treated siblings (Raldúa for using this non-parametric test (Kasuya, 2001). In addition, the
and Babin, 2009). The MTC for each compound, 5 μM CPF and 50 μM multiple comparison Tukey-Kramer test (Tukey, 1977) was applied for
nicotine, was used in the exposure solutions for behavioural analyses. multiple testing correction in the univariate case. As stated recently
On the day of the experiment, fresh exposure solutions were pre- (Saccenti et al., 2014), different situations can be encountered in
pared. Whereas CPF exposure solution was prepared by dilution of the practice depending on the correlation among variables, and it is better
stock solution in fish water, with a final concentration of 0.05% DMSO, to confirm the results using both type of approaches, univariate (with
nicotine exposure solution was prepared by direct dilution of product in the multicomparison test) and multivariate approaches.
the fish water. Locomotor behaviour was tested in zebrafish simulta-
neously in control (0.05% DMSO in fish water for CPF and fish water 2.5. Chemometric analyses
only for nicotine samples) and treated (5 µM CPF and 50 µM nicotine)
groups at 2, 6 or 24 h after CPF and nicotine exposures. Between 7 and Chemometric data analysis included two different multivariate data
12 replicates of 2–3 independent experiments were analysed depending analysis methods, Principal Component Analysis (PCA) and ANOVA-
on each condition. Exposures were performed at 28 ± 1 °C in aerated Simultaneous Component Analysis (ASCA). MATLAB R2016a
glass beakers containing 4000 mL of exposure solution and four fish (Mathworks Inc. Natick, MA, USA) and PLS Toolbox 8.2 (Eigenvector
were conducted. All fishes used in this study were experimentally naïve, Research Inc., Wenatchee, WA, USA) were used as computer

584
C. Gómez-Canela et al. Ecotoxicology and Environmental Safety 145 (2017) 583–590

programming environments for all chemometric analyses. (Fig. 3A). Thus, a significant increase in both the distance travelled and
velocity was found 6 and 24 h after exposure to nicotine (Fig. 3A).
2.5.1. Principal Component Analysis (PCA) Mobility states of zebrafish were also significantly altered by nicotine.
PCA is a commonly used method for reducing the dimensionality of On the other hand, high mobility and mobility frequencies significantly
complex data sets and it is frequently used for unsupervised pattern decreased 6 and 24 h after exposure. The duration of high mobility
recognition of the dominant effects and clusters on a set of samples. state, however, significantly increased 24 h after nicotine exposure. At
PCA decomposes the experimental data matrix, D, into the product of this time point the frequency of the immobility state was significantly
two factor matrices using a bilinear orthogonal decomposition reduced by nicotine.
(Esbensen and Geladi, 2010; Wold et al., 1987). D = XYT+E, where CPF and nicotine exhibited different effects on the zebrafish anxiety
principal component scores, X, and loadings, YT explain most of the behaviour. Thus, a significant increase of the time spent in the center
data variance (information) of the original dataset, leaving the un- with a concomitant decrease of the time spent in the periphery was
explained residual variance in E. Plots of scores and loadings of the first found at the three selected time points after CPF exposure (Fig. 2B).
principal components display the main sources of variance in the ori- This decrease in the thigmotactic swimming found in the open field
ginal data and the interrelationships among samples and variables. In paradigm indicates a specific anxiolytic-like effect of CPF in adult
this work, PCA was performed on the autoscalled OFT matrix of both 54 zebrafish. Moreover, the relationship between CPF-exposure and the
and 55 zebrafish group of samples treated by the two studied neuro- increase observed in the time spent in the immobility state found in our
toxicants (CPF and nicotine) at three different exposure times (2 h, 6 h study is consistent with the reported increase in the number of animals
and 24 h). The number of principal components was selected on the undergoing freeze responses observed in adult zebrafish 24 h after CPF
basis of scree plots (Jackson, 2004). exposure (Tilton et al., 2011). Nicotine, on the contrary, exhibited an
anxiogenic effect, increasing significantly the time spent in the per-
2.5.2. Analysis of Variance-Simultaneous Component Analysis (ASCA) iphery and reducing the time spent in the center of the arena 24 h after
ASCA is a multivariate data analysis approach that combines the exposure (Fig. 3B). While homebase formation represents an important
statistical advantages of ANOVA to separate the variance sources, and aspect of animal exploration, data on the effect of environmental pol-
the advantages of PCA for eliminating the covariation among variables lutants and drugs on homebase behaviour are still scarce (Cachat et al.,
and to explain maximum variance. ASCA can be understood as a direct 2013; Stewart et al., 2010). When the effect of CPF on the homebases
generalization of the analysis of variance (ANOVA) for univariate data formation was examined (Fig. 2C), a mild but significant effect was
to the multivariate case (Smilde et al., 2005). In this method, the in- found only 2 h after exposure. At this time CPF-treated zebrafish moved
formation of the experimental design structure of datasets (i.e., un- longer distances and spent more time in the HBs than control zebrafish.
derlying factors such as exposure time, dose or combinations thereof) is These results indicate that CPF modify also the zebrafish exploratory
incorporated, enabling a better understanding of the underlying bio- behaviour. Nicotine, in contrast, did not modify the exploratory beha-
logical information of them. In the ASCA methodology, each PCA model viour of zebrafish at any selected time point (Fig. 3C).
is fitted to each ANOVA separated factor data matrix individually All these results were confirmed when the multiple comparison
(Jansen et al., 2005). Assuming the general factor ANOVA model for Tukey-Kramer test was applied (Tukey, 1977). In all cases, the results
balanced data, a permutation test is used to check the statistical sig- were similar than the student's t-test and Mann-Whitney U-test applied
nificance of the effects of all factors and their interactions (Vis et al., to every one of the behaviour endpoint considered separately.
2007). This permutation test, assess whether there is no significant ef-
fect of the considered factor (H0, null hypothesis), and it is performed 3.2. PCA results
by randomly permuting the original data matrix (i.e. 10,000 permuta-
tions) (Vis et al., 2007). PCA was applied to the video-tracking data results for exploratory
In this work, ASCA was applied to the OFT experimental data sets in analysis. Both CPF and nicotine treated samples were compared with
the presence of CPF (Fig. 1) and nicotine (Fig. S2), to study the effects of their respective control samples. In the case of CPF, three principal
two factors, treatment (controls and 5 µM CPF or 50 µM nicotine, re- components explained 67.1% of the data variance. PC1 explains
spectively) and exposure time (2, 6 and 24 h), with a number of levels 41.43% of the data variance and separates the samples in relation to the
differing among experiments and their interaction. Statistical sig- treatment factor (Fig. S3A). Samples grouped in the negative side of PC1
nificances of the two factors and of their possible interaction were axis were those exposed to CPF whereas samples grouped on the posi-
evaluated by the permutation test, using 10,000 permutations (Vis tive side of PC1 axis were the control samples. Variances explained by
et al., 2007). PC2 and PC3 were related to other variability sources non-dependent of
treatment factor. In the case of nicotine, PCA with three principal
3. Results and discussion components explained a 76.02% of the data variance. PC1 explains
36.47% of the data variance, but it does not describe the differences
3.1. Statistical univariate analysis results between control and nicotine-treated animals (Fig. S3B). When data
from both treatments were simultaneously analysed, no improvement
As commented before, Student's t-test or Mann-Whitney U-test were in the interpretation was gained and the separation between control
used to calculate statistically significant differences between control and treated samples was worse. To further investigate the possible ef-
and exposed samples, considering p values < 0.05. fects of exposure time factor, the ASCA method was used to show the
When the OFT was used to analyse the effect of CPF on zebrafish level of the significance of both factors, treatment and time, and their
locomotor activity, a significant decrease in both the distance travelled possible interaction.
and velocity was found at all selected time points (Fig. 2A). Mobility
states were also significantly altered by CPF exposure calculated 3.3. ASCA results: evaluation of CPF and nicotine treatment and exposure
through t-test (Fig. 2A). Thus, in spite that the high mobility frequency time effects on behaviour of zebrafish
increased after 2 and 24 h exposure to CPF, duration of this mobility
state decreased at all the selected time points in CPF-treated fish. Statistical significances of the two categorical factors (i.e., treatment
Moreover, both frequency and duration of mobile state significantly and exposure time) and of their interaction were evaluated by the ASCA
increased from 6 h CPF exposure onwards. Finally, both frequency and method. Table 1 shows the ASCA results indicating the significance and
duration of the immobility state increased after 2 and 24 h of CPF ex- partitioning of the total variance. For CPF treatment, natural variability
posure. Nicotine exhibited the opposite effect on locomotor activity was the dominant part (residuals ≥ 67%), and only a minor part could

585
C. Gómez-Canela et al. Ecotoxicology and Environmental Safety 145 (2017) 583–590

Fig. 1. Structure of the experimental data sets arranged for PCA and ASCA analyses in the CPF experiment (similar data arrangements for nicotine). Each rectangle represents a zebrafish
sample. Data sets i at each exposure time are shown in different colours (green: 2 h; blue: 6 h and red: 24 h). Data sets from control and treated zebrafish samples are further arranged into
an augmented data matrix, as indicated in the right-hand side of the figure. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of
this article.)

be assigned to factors and interaction (~33%), with a higher effect of periphery, time in periphery, entrance to center and highly mobile
the treatment (different CPF concentrations) factor (27.8%). Results of duration are the variables more affected by exposure to nicotine.
the permutation test confirmed the larger significance (p < 0.01) of the Conversely, the rest of the variables studied (latency in periphery, time
treatment factor and in contrast, no significant p-values for the exposure in center, latency in center, highly mobile frequency, mobile frequency
time factor, neither for the interaction between factors were reported and duration, immobile frequency and duration, homebase distance
(Table 1). SCA PC1 scores (of the first component) of factor the treat- and duration, and non-homebase distance and duration) had higher
ment factor shown in Fig. 4A indicate that control samples were dif- loadings for control untreated samples. On the other hand, there was no
ferent to samples exposed to CPF. SCA PC1 loadings (Fig. 4B) indicated clear interpretation of the effects of exposure time and for this reason;
that the more affected variables by CPF exposure were related to highly the plot of SCA PC1 scores of exposure time factor data matrix is not
mobile state frequency, mobile and immobile states frequency, dura- shown. Finally, there was no systematic increasing or decreasing in-
tion, homebase distance and duration. In contrast, the rest of the teraction trend in the sample scores between treatment factor regard to
variables studied (distance moved, velocity, entrance and velocity to exposure time factor (Table 1).
periphery, time in periphery, entrance to center, highly mobile duration PCA and ASCA results reported in this manuscript for the effects
and non-homebase distances) did not show differences between CPF caused by the treatment with both compounds, CPF and nicotine, have
treated and control samples. On the other hand, SCA PC1 scores of the shown a good agreement with the results obtained simultaneously with
interaction data matrix did not show any specific pattern. traditional statistical tests. In fact, SCA PC1 loadings presented in
Table 1 also shows the ASCA results for nicotine. Natural variability Figs. 4B and 5B summarize perfectly the statistical results presented in
was even more dominant in this case (residuals ≥ 83%) than for CPF Figs. 2 and 3. Thus, SCA PC1 loadings identified the decrease in loco-
(residuals ≥ 67%), and the factors and interaction accounted only for motor activity observed in animals exposed to CPF, with the exposed
~17% of the total observed variance. Permutation test showed statis- fish swimming less distance, at lower velocity, spending less time at
tical significance for the treatment factor (p < 0.001) and also for the highly mobility state and more time at mobile and immobile states.
exposed time factor (p = 0.0084). In contrast, the interaction between These results were also consistent with the reduction in locomotion
these two factors was not significant. SCA PC1 scores of the first com- observed in rats 2 days after a single high dose of CPF (Lopez-Crespo
ponent for the treatment factor separated control samples from samples et al., 2007) and with the decrease in the swimming rates found in adult
exposed to nicotine (Fig. 5A). Therefore, SCA PC1 loadings (Fig. 5B) zebrafish exposed to 0.6 µM CPF for 24 h (Tilton et al., 2011). SCA PC1
show that the endpoints distance moved, velocity, entrance to loadings identified also a specific anxiolytic-like effect of CPF in adult

586
C. Gómez-Canela et al. Ecotoxicology and Environmental Safety 145 (2017) 583–590

Fig. 2. Time-course of behavioural effects induced by the acute exposure to 5 µM chlorpyrifos (CPF) in adult zebrafish. The 6-min open field test (OFT) was selected and different
endpoints related with locomotion (A), anxiety (B) and exploratory (C) behaviours were analysed. Representative 2D traces of control- and CPF-treated fish 24 h after exposure generated
by Ethovision XT11.5 are also shown. Data are reported as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001.

587
C. Gómez-Canela et al. Ecotoxicology and Environmental Safety 145 (2017) 583–590

Fig. 3. Time-course of behavioural effects induced by the acute exposure to 50 µM nicotine in adult zebrafish. The 6-min open field test (OFT) was selected and different endpoints related
with locomotion (A), anxiety (B) and exploratory (C) behaviours were analysed. Representative 2D traces of control- and nicotine-treated fish 24 h after exposure generated by Ethovision
XT11.5 are also shown. Data are reported as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001.

588
C. Gómez-Canela et al. Ecotoxicology and Environmental Safety 145 (2017) 583–590

Table 1
ASCA results: Statistical significances and partitioning of the total variance into the in-
dividual terms corresponding to factors and their interaction.

Neurotoxicant Factor Cum Percentage of Significance (p-

EigenVala variationb value)

Chlorpyrifos Treatment 5.01 27.8 0.001

Exposed time 0.41 2.30 0.676
Treatment x 0.50 2.81 0.999
Exposed time
Residuals 67.4
Nicotine Treatment 1.34 7.44 0.0001
Exposed time 1.08 6.02 0.0084
Treatment x 0.72 3.98 0.9879
Exposed time
Residuals 82.56

a
Cumulative EigenVal.
b
Percentage of variation expressed as sums of squared deviations from the overall
mean. Treatment factor (control and 5 µM in the case of CPF; and control and 50 µM in the
case of nicotine). Exposed time factor (2 h, 6 h and 24 h levels). In bold, p-values ≤ 0.001.

Fig. 5. ASCA results of the OFT data from s in control and nicotine-treated adult zebrafish
samples. (A) SCA scores plot for the “dose” factor matrix. Color symbols indicate the
different samples studied: green diamonds are controls and red squares are the zebrafish
samples at 50 µM of nicotine and (B) SCA PC1 loadings plot of dose sample factor matrix
with their variables in the x-axis. (For interpretation of the references to color in this
figure legend, the reader is referred to the web version of this article.)

reports on the behavioural effects of CPF on vertebrates (Lopez-Crespo

et al., 2007; Richendrfer et al., 2012).
The behavioural effects of nicotine identified by the SCA PC1
loading were also encountered in the statistical analysis of the loco-
motor activity data and with the concomitant anxiogenic-like effect.
These behavioural effects of nicotine seem to be dependent on the route
of administration, the behavioural state of the experimental subjects,
the behavioural paradigm explored, the animal species, as well as on
the strain (Picciotto et al., 2002). Although an anxiolytic-like effect of
acute nicotine treatment has been described in different species, in-
cluding adult zebrafish (Cachat et al., 2011; Levin et al., 2007;
Sackerman et al., 2010), the anxiogenic-like effect nicotine found in the
present study is consistent with a similar effect found in different stu-
dies on mice and rat using different paradigms (Elliott et al., 2004;
Fig. 4. ASCA results of OFT data from control and CPF-treated adult zebrafish samples.
Kenny et al., 2000; Picciotto et al., 2002; Trigo et al., 2009).
(A) SCA scores plot for the “dose” factor matrix. Color symbols indicate the different
samples studied: green diamonds are controls and red squares are the zebrafish samples at
5 µM CPF and (B) SCA PC1 loadings plot for the dose sample factor matrix with their 4. Conclusions
variables in the x-axis. (For interpretation of the references to color in this figure legend,
the reader is referred to the web version of this article.) This study explores the possibility of using chemometric methods to
reduce the heavy workload commonly associated to regular statistical
zebrafish, characterized by a decrease in the thigmotactic swimming: analysis of data generated during the neurobehavioural protocols using
CPF-exposed zebrafish spent significantly more time swimming in the video-tracking technologies. The hypothesis tested is that the applica-
center of the arena than the control fish. This effect was not only en- tion of principal component analysis (PCA) and analysis of variance-
countered in the statistical analysis but also is consistent with different simultaneous component analysis (ASCA) as initial chemometric mul-
tivariate data analysis steps in behavioural data management workflow

589
C. Gómez-Canela et al. Ecotoxicology and Environmental Safety 145 (2017) 583–590

allows for an early fast identification of the most relevant endpoints Elegante, M., Elkhayat, S., Tan, J., 2010. Characterization of behavioral and endo-
crine effects of LSD on zebrafish. Behav. brain Res. 214, 277–284.
altered by the different treatments. Then, these endpoints should be Jackson, J.E., 2004. Frontmatter. A User's Guide to Principal Components. John
prioritized for further confirmatory statistical analyses. Adult zebrafish Wiley & Sons, Inc (pp. i-3).
were exposed to chlorpyrifos (CPF) and nicotine and changes in beha- Jansen, J.J., Hoefsloot, H.C., van der Greef, J., Timmerman, M.E., Westerhuis, J.A.,
Smilde, A.K., 2005. ASCA: analysis of multivariate data obtained from an experi-
viour have been analysed at three different times using the OFT para- mental design. J. Chemom. 19, 469–481.
digm. The most relevant behavioural effects induced by CPF (decreased Jones, D.C., Miller, G.W., 2008. The effects of environmental neurotoxicants on the do-
locomotor activity, anxiolytic effect, altered exploratory behaviour) paminergic system: a possible role in drug addiction. Biochem. Pharmacol. 76,
569–581.
and nicotine (increased locomotor activity and anxiogenic effect) Kasuya, E., 2001. Mann-Whitney U test when variances are unequal. Anim. Behav. 61,
identified by a time-consuming statistical analyses were also success- 1247–1249.
fully identified by using ASCA. The results presented in this manuscript Kenny, P.J., Cheeta, S., E. File, S., 2000. Anxiogenic effects of nicotine in the dorsal
hippocampus are mediated by 5-HT(1A) and not by muscarinic M1 receptors.
support that the incorporation of the proposed chemometric methods in
Neuropharmacology 39, 300–307.
neurobehavioural assessment studies of neurotoxic effects of chemicals Levin, E.D., Bencan, Z., Cerutti, D.T., 2007. Anxiolytic effects of nicotine in zebrafish.
and drugs can be very useful as a preliminary fast screening step. Physiol. Behav. 90, 54–58.
Levin, E.D., Chen, E., 2004. Nicotinic involvement in memory function in zebrafish.
Neurotoxicol. Teratol. 26, 731–735.
Acknowledgements Lieschke, G.J., Currie, P.D., 2007. Animal models of human disease: zebrafish swim into
view. Nat. Rev. Genet. 8, 353–367.
This study was funded by the European Research Council under Lopez-Crespo, G., Carvajal, F., Flores, P., Sánchez-Santed, F., Sánchez-Amate, M., 2007.
Time course of biochemical and behavioural effects of a single high dose of chlor-
European Union's Seven Framework Programme (FP/2007–2013)/ERC pyrifos. Neurotoxicology 28, 541–547.
Grant Agreement n. 320737, the NATO SfP project MD.SFPP 984777 Mora-Zamorano, F.X., Svoboda, K.R., Carvan III, M.J., 2016. The nicotine-evoked loco-
(D.R.), and the Spanish Government (CTM2014-51985-R; D.R.). The motor response: a behavioral paradigm for toxicity screening in zebrafish (Danio
rerio) embryos and eleutheroembryos exposed to methylmercury. PLoS One 11,
authors thank Mr. Marc Mañas for his valuable assistance building the e0154570.
Open Field Test setup. Marc F. Nuñez, Michele Celentano and Aurora Oliveri, A., Bailey, J., Levin, E., 2015. Developmental exposure to organophosphate flame
Costa are gratefully thanked for their help in the OFT experiments. retardants causes behavioral effects in larval and adult zebrafish. Neurotoxicol.
Teratol. 52, 220–227.
Picciotto, M.R., Brunzell, D.H., Caldarone, B.J., 2002. Effect of nicotine and nicotinic
Appendix A. Supplementary material receptors on anxiety and depression. NeuroReport 13, 1097–1106.
Raldúa, D., Babin, P.J., 2009. Simple, rapid zebrafish larva bioassay for assessing the
potential of chemical pollutants and drugs to disrupt thyroid gland function. Environ.
Supplementary data associated with this article can be found in the
Sci. Technol. 43, 6844–6850.
online version at http://dx.doi.org/10.1016/j.ecoenv.2017.08.013. Raldúa, D., Piña, B., 2014. In vivo zebrafish assays for analyzing drug toxicity. Expert
Opin. Drug Metab. Toxicol. 10, 685–697.
References Richendrfer, H., Pelkowski, S.D., Colwill, R.M., Créton, R., 2012. Developmental sub-
chronic exposure to chlorpyrifos reduces anxiety-related behavior in zebrafish larvae.
Neurotoxicol. Teratol. 34, 458–465.
Babin, P.J., Goizet, C., Raldua, D., 2014. Zebrafish models of human motor neuron dis- Saccenti, E., Hoefsloot, H.C.J., Smilde, A.K., Westerhuis, J.A., Hendriks, M.M.W.B., 2014.
eases: advantages and limitations. Progress Neurobiol. 118, 36–58. Reflections on univariate and multivariate analysis of metabolomics data.
Brittijn, S.A., Duivesteijn, S.J., Belmamoune, M., Bertens, L.F.M., Bitter, W., De Bruijn, Metabolomics 10, 361–374.
J.D., Champagne, D.L., Cuppen, E., Flik, G., Vandenbroucke-Grauls, C.M., Janssen, Sackerman, J., Donegan, J.J., Cunningham, C.S., Nguyen, N.N., Lawless, K., Long, A.,
R.A.J., De Jong, I.M.L., De Kloet, E.R., Kros, A., Meijer, A.H., Metz, J.R., Van Der Sar, Benno, R.H., Gould, G.G., 2010. Zebrafish Behavior in Novel Environments: effects of
A.M., Schaaf, M.J.M., Schulte-Merker, S., Spaink, H.P., Tak, P.P., Verbeek, F.J., Acute Exposure to Anxiolytic Compounds and Choice of Danio rerio Line. Int. J.
Vervoordeldonk, M.J., Vonk, F.J., Witte, F., Yuan, H., Richardson, M.K., 2009. Comp. Psychol. / ISCP; Spons. Int. Soc. Comp. Psychol. Univ. Calabr. 23, 43–61.
Zebrafish development and regeneration: new tools for biomedical research. Int. J. Smilde, A.K., Jansen, J.J., Hoefsloot, H.C., Lamers, R.-J.A., Van Der Greef, J.,
Dev. Biol. 53, 835–850. Timmerman, M.E., 2005. ANOVA-simultaneous component analysis (ASCA): a new
Cachat, J., Kyzar, E.J., Collins, C., Gaikwad, S., Green, J., Roth, A., El-Ounsi, M., Davis, A., tool for analyzing designed metabolomics data. Bioinformatics 21, 3043–3048.
Pham, M., Landsman, S., 2013. Unique and potent effects of acute ibogaine on zeb- Stewart, A., Cachat, J., Wong, K., Gaikwad, S., Gilder, T., DiLeo, J., Chang, K., Utterback,
rafish: the developing utility of novel aquatic models for hallucinogenic drug re- E., Kalueff, A.V., 2010. Homebase behavior of zebrafish in novelty-based paradigms.
search. Behav. Brain Res. 236, 258–269. Behav. Process. 85, 198–203.
Cachat, J.M., Stewart, A., Utterback, E., Kyzar, E., Hart, P.C., Carlos, D., Gaikwad, S., Stewart, A.M., Grieco, F., Tegelenbosch, R.A., Kyzar, E.J., Nguyen, M., Kaluyeva, A.,
Hook, M., Rhymes, K., Kalueff, A.V., 2011. Deconstructing adult zebrafish behavior Song, C., Noldus, L.P., Kalueff, A.V., 2015. A novel 3D method of locomotor analysis
with swim trace visualizations. Zebrafish Neurobehav. Protoc. 191–201. in adult zebrafish: implications for automated detection of CNS drug-evoked phe-
De Duffard, A.E., Duffard, R., 1996. Behavioral toxicology, risk assessment, and chlori- notypes. J. Neurosci. Methods 255, 66–74.
nated hydrocarbons. Environ. Health Perspect. 104, 353. Tilton, F.A., Bammler, T.K., Gallagher, E.P., 2011. Swimming impairment and acet-
Elliott, B.M., Faraday, M.M., Phillips, J.M., Grunberg, N.E., 2004. Effects of nicotine on ylcholinesterase inhibition in zebrafish exposed to copper or chlorpyrifos separately,
elevated plus maze and locomotor activity in male and female adolescent and adult or as mixtures. Comp. Biochem. Physiol. Part C: Toxicol. Pharmacol. 153, 9–16.
rats. Pharmacol. Biochem. Behav. 77, 21–28. Trigo, J.M., Zimmer, A., Maldonado, R., 2009. Nicotine anxiogenic and rewarding effects
Esbensen, K.H., Geladi, P., 2010. Principal Component Analysis: Concept, Geometrical are decreased in mice lacking β-endorphin. Neuropharmacology 56, 1147–1153.
Interpretation, Mathematical Background, Algorithms, History, Practice. Tukey, J.W., 1977. Some thoughts on clinical trials, especially problems of multiplicity.
Comprehensive Chemometrics. Elsevierpp. 211–226. Science 198, 679–684.
Faria, M., Garcia-Reyero, N., Padrós, F., Babin, P.J., Sebastián, D., Cachot, J., Prats, E., Vis, D.J., Westerhuis, J.A., Smilde, A.K., van der Greef, J., 2007. Statistical validation of
Arick II, M., Rial, E., Knoll-Gellida, A., 2015. Zebrafish models for human acute or- megavariate effects in ASCA. BMC Bioinforma. 8.
ganophosphorus poisoning. Sci. Rep. 5. Wold, S., Esbensen, K., Geladi, P., 1987. Principal component analysis. Chemom. Intell.
Farrés, M., Piña, B., Tauler, R., 2015. Chemometric evaluation of Saccharomyces cere- Lab. Syst. 2, 37–52.
visiae metabolic profiles using LC–MS. Metabolomics 11, 210–224.
Grossman, L., Utterback, E., Stewart, A., Gaikwad, S., Chung, K.M., Suciu, C., Wong, K.,

590