Introduction

Xenobiotics are increasingly being released into the environment as a result of human activities. The compounds of one particular group of xenobiotics, persistent organic pollutants (POPs) persist in the environment, have lipophilic properties, and are very stable chemically, making the bioaccumulation of these molecules highly likely. Aquatic compartments, including sediment, often constitute the ultimate reservoir of these compounds, so fish may be exposed to POPs at various stages in their lives.

Polycyclic aromatic hydrocarbons (PAHs) constitute a large and diverse family of POPs formed by the association of several aromatic rings. This family of molecules also includes PAH derivatives with alkyl, nitrogen, sulfur, or oxygen substitutions. PAHs and their derivatives are produced during the combustion of organic matter (pyrolytic PAHs) and are present in oils (petrogenic PAHs). Pyrolytic PAHs are released into the atmosphere, leading to their deposition on water and soil. PAHs deposited on soil can enter water compartments through runoff. Petrogenic PAHs may be discharged directly into water as a result of oil spills or naval or offshore oil drilling activities (Latimer and Zheng 2003). PAHs are also introduced into the environment through natural processes, such as oil leaks or diagenesis. In all cases, PAHs are present in the form of complex mixtures containing large amounts of individual PAHs and derivatives. The composition of mixtures of PAHs depends on their origin. Pyrolytic mixtures contain a high proportion of heavy PAHs, such as benzo[a]pyrene (BaP), and few alkylated PAHs, whereas petrogenic mixtures contain a smaller proportion of heavy PAHs, with a higher proportion of light PAHs and larger amounts of alkylated PAHs (Benlahcen et al. 1997; Budzinski et al. 1997; Latimer and Zheng 2003; Neff 1979; Yunker et al. 2002). The release of PAHs into the environment has increased in recent decades (Eisler 1987; Shen et al. 2013; Zhang and Tao 2009). Indeed, monitoring networks have documented PAH concentrations of up to 50 μg g−1 dry weight (dw) in sediment from various affected aquatic ecosystems, although the concentrations in highly contaminated areas were mostly found to be in the 10 μg g−1 range (Baumard et al. 1998; Benlahcen et al. 1997; Cachot et al. 2006; Johnson et al. 2007; Varanasi et al. 1993; Yanagida et al. 2012). The concentration of PAHs in the biota depends on the concentration of these molecules in the immediate surroundings and trophic level. For example, the total concentration of the 16 PAHs used as indicators by the US Environmental Protection Agency (US-EPA) in the copepod Eurytemora affinis in the Seine Estuary has been reported to be in the 165–3,866 ng g−1 dw range (Cailleaud et al. 2007) and, in mussels, this concentration may reach 1,600 ng g−1 dw (Le Goff et al. 2006). In cases of acute accidental exposure, the concentration in mussels may be even higher, with values reaching 3,000 ng g−1 dw reported after the Erika oil spill (Jeanneret et al. 2002) and of 14,350 ng g −1 dw recorded after the Exxon Valdez oil spill (Payne et al. 2008). PAHs accumulate in animals located at lower levels of the food chain because they are poorly metabolized in these species. Food is thus an important source of exposure, as demonstrated by the high concentrations of PAHs found in the stomachs of fish caught in contaminated areas (Johnson et al. 2007; Varanasi et al. 1993; Yanagida et al. 2012).

There is growing evidence to suggest that POPs can act as neurotoxicants, particularly in cases of early exposure (Ottinger et al. 2013; Scott and Sloman 2004). Only a few studies have described the behavioral disruptions following PAH exposure in fish. In most cases, this was achieved by waterborne exposure of juveniles to single compound (Almeida et al. 2012; Correia et al. 2007; Goncalves et al. 2008; Gravato and Guilhermino 2009; Oliveira et al. 2012). One study also reported additive effects for a mixture of three PAHs (Goncalves et al. 2008). These studies have generally concluded that PAHs cause lethargy, resulting in a decrease in the swimming activities of exposed fish.

Early stages are considered to be more sensitive than later stages, but the number of studies focusing on early stages (embryo-larval stage) is even smaller. For example, brief exposure of medaka embryos to sediment spiked with benzo[a]anthracene has been shown to reduce larval activity after a photomotor challenge in medaka (Le Bihanic, personal communication). Finally, the exposure of Pacific herring eggs to weathered oil has several adverse effects, including a decrease in the ability of the larvae to swim (Carls et al. 1999). We have shown that the brief exposure (from 4 to 96 h postfertilization) of zebrafish embryos to a mixture of benzo[a]pyrene, pyrene, and phenanthrene leads to lower levels of locomotor activity and/or the induction of anxiety-like behavior in adults (Vignet et al. 2014a). These findings are consistent with observations showing that exposure to PAHs is associated with several types of neurobehavioral disruption, including anxiety and attention disorders, in humans and rodents (Bouayed et al. 2009; Edwards et al. 2010; Jedrychowski et al. 2003; Perera et al. 2012; Saunders et al. 2001, 2003; Tang et al. 2008; Xia et al. 2011). It has also recently been shown that perinatal exposure to a mixture of 16 PAHs induces anxiety-related behavior in adult rats (Crépeaux et al. 2012, 2013).

Quantitative behavior analysis is now considered to be a good indicator in studies of pharmacology, toxicology, ecotoxicology, and in cognition or neurobiology (Ali et al. 2011a, b; Champagne et al. 2010). The use of model fish species, such as medaka and zebrafish, has developed exponentially over the last decade, and the use of these species in this context is now widely accepted (Champagne et al. 2010; Creton 2009; Norton and Bally-Cuif 2010; Steenbergen et al. 2011). Studies of fish behavior as a means of investigating the neurotoxic effects of compounds and the underlying mechanisms of toxicity are becoming increasingly common. Fish have been used as models of human disorders in studies on compounds such as alcohol (Gerlai et al. 2000; MacPhail et al. 2009) and for the testing of drugs to treat anxiety (Champagne et al. 2010; Egan et al. 2009; Maximino et al. 2010). In the particular context of ecotoxicology, individual responses correspond to the interface between ecological factors and toxicity mechanisms, making individual behavior a particularly relevant and integrative indicator of the effects of pollutants (Bailey et al. 2013; Hinton et al. 2005; MacPhail et al. 2009).

We report here the effects of long-term dietary exposure to three environmental mixtures of PAHs in zebrafish, in a dose-response study with three concentrations. Fish were fed diets spiked with mixtures approximating relevant environmental concentrations, from their first meal at 5 days postfertilization (dpf). We analyzed several behavioral traits, including locomotion, exploration, and anxiety, successively in 2-month-old juveniles and 6-month-old adults. In addition to measuring behavioral toxicity and evaluating the neurotoxicity of PAHs in a model species, we also evaluated the potential ecological consequences for fish populations.

Materials and methods

Preparation of PAH-contaminated diets

Three aromatic fractions were used for exposure experiments: (1) a pyrolytic fraction (PY) extracted from sediments collected from a polluted site of the Seine Estuary (Oissel, France) and (2) two petrogenic fractions obtained from Erika fuel (heavy oil; HO) and Arabian Light crude oil (LO). PAHs were extracted as previously described (Cachot et al. 2006). Zebrafish were exposed from their first meal (at 5 dpf) onward, through spiked food pellets. The food pellets obtained from the supplier were 0.5 mm in diameter (INICIO Plus 0.5, Biomar, France); these pellets were ground and used to prepare age-appropriate food pellets of four sizes: ≤125, 125–315, 315–500, and ≥500 μm (Vignet et al. 2014b). Three concentrations, 0.3X, 1X, and 3X, were used, with the 1X concentration corresponding to the Σ[16 EPA PAH] at 5 μg g−1 dw food, representative of the concentrations found in mollusks in the Seine Estuary. The names of the diets indicate the origin of the fraction and its concentration. For example, the 1X pyrolytic fraction diet was named PY-1X. For each exposure, a control treatment was included, corresponding to the plain food treated as for the spiked food with dichloromethane, the carrier solvent used for PAH spiking.

Fish exposure

This study was conducted with the approval of the French Animal Care Committee under the official license held by M.-L. Bégout (17-010).

We used the zebrafish wild-type TU strain (ZFIN ID: 76 ZDB-GENO-990623-3), which was established at our platform (PEP—http://wwz.ifremer.fr/pep) 6 years ago from large batches of individuals originating from the Amagen platform (Gif/Yvette, France) and the Pasteur Institute fish facility (Paris, France). The exposure procedures are described in detail in the article by Vignet et al. (2014b). Briefly, eggs were obtained by the random pairwise mating of zebrafish placed together in spawning boxes (AquaSchwarz, Germany) overnight. Eggs were collected the next morning, and the rate of fertilization was assessed within 2 h of collection: only spawns with a fertilization rate greater than 80 % were retained. We ensured that incubation conditions were as homogeneous as possible, by mixing at least five spawns and then randomly sampling 50 embryos and transferring them to as many Petri dishes as required for the planned number of treatments. Exposure experiments with PY, HO, and LO were carried out successively (PY in year 1 of the project, HO and LO in year 2, see details in Vignet et al. (2014b)) and all exposures were performed in triplicate. Embryos and larvae were maintained at 28 °C, in clean medium, in Petri dishes in an incubator subjected to the same photoperiod as for adults. From 5 dpf onwards, each replicate were reared in separate 1-l tanks. After 12 days, they were transferred to 10-l tanks in a flow through system with water renewal, the daily rate of which was increased with age, reaching 40 % after 1 month. Standard water quality and rearing methods were used (Lawrence 2007; Vignet et al. 2014b). From 5 dpf onwards, the fish were fed twice daily with size-adapted spiked pellets and once daily with plain Artemia nauplii. The fish were allowed ad libitum access to the food pellets from 5 dpf until the first biometric measurements were taken: at 2 months postfertilization (mpf) for HO and LO and at 3 mpf for PY. Thereafter, the quantity of food provided was equivalent to 2 % of the biomass in each tank, with the exception of the PY-spiked food, for which we provided ~5 % of the biomass in the tank until 3 mpf when we decreased to 2 %. Rations were adapted on a monthly basis, following biometric measurements. This led to differences in growth between the control fish, which were heavier and longer for PY (at 6 mpf, body mass (mg) was 707 ± 22; body length (mm) was 31.1 ± 0.3) than for HO and LO (body mass (mg): HO = 265 ± 7, LO = 246 ± 6; body length (mm): HO = 25.2 ± 0.2, LO = 24.3 ± 0.2) as described in detail in Vignet et al. (2014b). These differences in body size may partly account for the differences in behavioral responses observed between control PY and control HO/LO fish.

Behavioral experiments

Behavioral experiments were performed with 2-month-old undifferentiated juveniles and 6-month-old adult males and female (in equal proportions, except for HO-3X, for which it was not possible to distinguish between the sexes on the basis of phenotype). Different sets of animals were used for each behavioral test, except for the photomotor response in adults, which was coupled with the 24-h swimming activity test, and the maze test, which was coupled with the novel tank test, these tests being performed consecutively on the same day (see below for the detailed protocol). HO-3X fish were not tested at the age of 6 months in the Z-maze and novel tank tests, because of their low survival (Larcher et al. In revision; Vignet et al. 2014b). Tests were carried out in a dedicated room maintained at 27 °C ± 1 °C, with a 14:10 photoperiod synchronized with that of the rearing room so as to minimize unwanted correlated effects. The light was switched on at 08:30 and there were no twilight transition periods.

24-h swimming activity in adults

At each session and for each fraction, 6-month-old fish exposed to each concentration were randomly placed in 3-l tanks (24.5 × 15 × 13.5 cm, AquaBox® 3, AquaSchwarz) filled with 1.5 l of system water, to avoid tank position and session bias. The 12 tanks (3 fish per concentration) were isolated from neighboring tanks by opaque walls. They were placed on top of an infrared apparatus (IR floor 1 × 1 m, Noldus, The Netherlands). During the day, the room was lit with two LED spotlights (LED Technology Master LEDSPOT 16 W 2700 K, illumination of 30 lx near the tanks). During the night or dark period, the spotlights were turned off and infrared light from the floor was used to record fish movements (illumination <1 lx near the tanks). Fish were placed in the tanks at 17:00 on the day before the experiment, to allow them to acclimate overnight. Video recordings began the next day at 12:30 and lasted 24 h. We tested 12 fish for each concentration. We ran four sessions for each fraction (PY, HO, and LO) and the water was changed after each session.

Photomotor response in juveniles and adults

We used the same setup to monitor photomotor responses, as an indicator of complex phenotypes of stress and anxiety (Vignet et al. 2013). We recorded swimming activity for 1 h (light on-1) and then challenged the fish by subjecting them to a sudden dark period (15-min, light off), continuing the video recordings for an additional 1.5 h (light on-2). The photomotor response was monitored after one night of acclimation, beginning at 12:30 for 2-month-old fish or at the end of the 24-h recording for 6-month-old fish. For both tests, the dependent variables measured were swimming path length (distance traveled; cm) and time spent immobile (time; s). We tested 12 fish for each concentration. We ran four sessions for each fraction (PY, HO, and LO) and the water was changed after each session.

Exploration of a new environment—maze challenges

In addition to providing information about the ability of fish to explore their environment, the swimming characteristics recorded during exploration provide information about the ways in which fish cope with this novel environment. Two experimental setups were used: maze challenges (a T-maze or a Z-maze) and a novel tank challenge.

The T-maze was used for PY fish (in year 1 of the project). This maze was adapted slightly from that described by Ninkovic and Bally-Cuif (2006) and comprised a shallow area, with a water depth of 5 cm, consisting of the stem of the T (46 cm long), and its two arms (one leading to a deep area and one in the opposite direction). The total length of the arms was 66 cm. The deep area was 15 cm deep (water depth of 10 cm), 23 cm wide, and 23 cm long, and it contained marbles and plastic grass, and was therefore considered to be a favorable area (Ninkovic and Bally-Cuif 2006). Fish were individually transferred into a 1-l aquarium at 16:00 on day 1 and were challenged on day 2, between 9:00 and 11:00. Fish were placed in the start area of the maze and their swimming characteristics were recorded for 5 min. We tested 12 fish per concentration and the dependent variables measured were time to the first exit from start area (in s), the time spent in each area (start, shallow, deep; in s), individual success or failure to reach the deep area, and time to first entry into the deep area (s).

For the HO and LO mixtures, in year 2 of the project, a Z-maze (70 × 45 × 15 cm) adapted from that described by Chapman et al. (2010) was preferred over the T-maze. We decided to use the Z-maze for these experiments because it is more suitable than the T-maze for the evaluation of exploration (Chapman et al. 2010), the T-maze being more appropriate for assessments of learning, which was not the main focus of this experiment. The Z-maze consisted of a refuge zone and four arms (see Fig. 1 in Chapman et al. 2010). Each arm was divided virtually into four quarters of equal size, referred to as quadrants. The maze thus consisted of 16 quadrants. Fish were individually placed in a 1-l aquarium at 16:00 on day 1 and challenged on day 2, between 9:00 and 11:00. Fish were introduced into the refuge zone (equivalent to three quadrants, covered with a black slab and ending in a sliding door). The start area thus consisted of these three covered quadrants and the adjacent one. The next two arms formed the close area (close to the refuge zone) and the last two arms formed the distant area. The fish were left for 2 min in the refuge zone and the door was then opened. Fish activity was recorded in the 5 min following the first excursion of the fish from the refuge zone (detection in the first quadrant of the close area). The variables measured were the time taken to leave the refuge (time to first exit from the refuge zone; in s), the time spent in each area (start, close, distant; s) during the trial, and individual success or failure to reach quadrant 16 (the farthest from the refuge zone). We also determined the number of quadrants into which the fish ventured within the maze (16 indicating that the fish reached the end of the maze), as an index of individual tendency to explore (Chapman et al. 2010). We converted these data into proportions for statistical analysis. We tested 12 fish per fraction and per concentration.

Fig. 1
figure 1

Swimming activity of 6-month-old fish as a function of photoperiod, for each fraction: PY, HO, and LO. a, b, c Mean distance traveled (cm/30 min) averaged per period (day vs. night). d, e, f Mean time spent immobile (s/30 min) averaged per period (day vs. night). Values are means ± SEM; n = 12 fish per fraction and concentration; * indicates a significant difference at p < 0.05

Full size image

Exploration of a new environment—the novel tank challenge

The novel tank challenge was performed from 14:00 to 16:00 on the same day as the maze exploration test (which was performed from 9:00 to 11:00). Between tests, the fish were individually kept in the 1-l aquarium in the same room. The fish were tested in the same chronological order in both tests. Fish were transferred to a novel tank (trapezoid 1.5-l tank; Aquatic Habitats, Apopka, FL; size in cm: height 15.2 × width 7.1 × length 27.9 at the top and 22.5 at bottom) and filmed for 6 min from the side. We tested 12 fish per fraction and concentration. For space occupancy analysis, tanks were separated into two areas as previously described (Egan et al. 2009): a “top” area corresponding to one half of the volume and a “bottom” area corresponding to the other half. The dependent variables measured were time spent in each area (top, bottom; in s) per minute, total distance traveled (cm), and time spent immobile (s).

Data recording and analysis

For swimming activity, maze exploration, and novel tank experiments, videos were recorded with an analog ICD-48E camera (Ikegami) and a 2.7–13.5-mm lens (Fujinon), linked to a PC with an acquisition card and Ethovision XT software (Noldus). In all experiments, EthoVision XT software was used for track extraction and analysis. Data were acquired by EthoVision XT at a rate of 25 frames per second, and variables were nested for further analysis, every 30 min for the 24-h swimming tests on 6-month-old fish and every minute during the photomotor response test (distance traveled in cm and time spent immobile in s). In photomotor response tests, the distance traveled was also summed over 15-min periods (light on-1, light off, light on-2). Nesting was carried out every 5 min (duration of the test) in maze experiments and every 1 or 6 min in novel tank experiments (distance traveled in cm, time spent in top area in s, and time spent immobile in s). For assessments of the time spent immobile, we used the mobility detection function of EthoVision XT, which has a lower threshold for separating immobility from mobility, and an upper threshold for separating mobility from highly mobile. The lower threshold was set at 20 % for assessments of immobility, indicating that a change of no more than 20 % in the pixels of a detected object between two consecutive samples would be considered to indicate immobility (see EthoVision XT Reference Manual).

Statistical analysis

Statistical analyses were carried out with Statistica 9.0 (Statsoft, Tulsa, OK, USA) software. Generalized linear models (GLM) were applied in each case, for statistical assessment of the effect of PAH concentration on measurements of locomotor activity, photomotor response, the tendency to explore, and the anxiety level. For each fraction (PY, HO, or LO), we tested concentration (control, 0.3X, 1X, and 3X) and period (day vs. night in 24-h locomotion; light on-1, light off, light on-2 during the photomotor response test; min-1 to 6 in the novel tank test), or area (start, shallow, and deep in the T-maze; start, close, or distant in Z-maze; top in the novel tank test) as fixed factors. Fish and session were treated as random factors in this repeated measures analysis, to overcome the problem of pseudoreplication. Sex was also treated as a fixed factor in analyses of 6-month-old fish. Body mass was tested as a covariable but was never found to have a significant effect. We also analyzed the interaction between concentration and period. Throughout the results section, we report only significant fixed factor or interaction effects. Individual successes or failures to reach the deep area (T-maze) or quadrant 16 (Z-maze) were compared between concentrations in chi2 tests followed by two-tailed paired Fisher’s exact tests (Control vs. each concentration). The proportion of the maze explored was subjected to arcsine transformation before GLM analysis. All statistical tests were carried out at the 5 % significance level. Newman-Keuls post hoc tests were carried out in each case. The results are reported in the text and figures as means ± SEM.

Results

This work is part of a larger program including analyses of the changes in several physiological variables following exposure to the diets described here. Chemical analyses of the fractions, diets, and metabolites are presented elsewhere (Vignet et al. 2014b). Briefly, the PY fraction contained a high proportion of heavy PAHs and almost no methylated derivatives. The HO fraction contained moderate levels of heavy PAHs and methylated derivatives. The LO fraction contained low levels of heavy PAHs and a high level of alkylated derivatives. The concentrations of the 16 US-EPA PAHs in 1X diets were PY-1X: 4,505 ± 1,527, LO-1X: 2,739 ± 231, and HO-1X: 1,887 ± 130 ng g−1dw. The total concentration of PAHs (including methylated derivatives) measured in the 1X diets were: 5,816 ± 1,433, 4,663 ± 360, and 6,726 ± 278 ng g−1dw for PY, HO, and LO, respectively. The quantification of hydroxylated metabolites in 15 dpf larvae confirmed that the fish had effectively been exposed, and the results obtained were consistent with diet composition (Vignet et al. 2014b).

24-h locomotion in adults

In 6-month-old adults, we assessed 24-h locomotion and compared the distance traveled between the day and night periods (Fig. 1a–c). No differences were observed between the day and night periods for PY and LO, whereas fish exposed to HO displayed significantly higher levels of activity during the day than during the night (F = 10.23, p < 0.001). Concentration and sex had no significant effect and there was no significant interaction between fixed factors. During both day and night periods, the fish remained immobile for about 50–60 % of the total time (Fig. 1d–f). For fish on PY diets, the time spent immobile was greater during the day than during the night (F = 44.84; p < 0.001), but neither concentration nor sex had any significant effect on this variable and no significant interactions were detected. A similar pattern, with longer periods of immobility during the day than during the night, was observed for fish on HO diets (F = 46.43; p < 0.001) but, in this case, a difference was also observed between concentrations (F = 13.24; p < 0.001) and there was a significant interaction between period and concentration (F = 3.09; p < 0.01). Indeed, fish on HO diets spent less time immobile during the day than control fish, for HO-1X and HO-3X (p < 0.05 and p < 0.001 respectively), whereas fish on the HO-3X diet spent less time immobile than the control fish during the night (p < 0.001). Sex had no effect. Fish on LO diets were less immobile during the day than control fish (F = 119.63; p < 0.001) and significant differences were observed between concentrations (F = 5.25; p < 0.01), but there was no interaction and no effect of sex. In this case, all fish fed LO diets spent a significantly shorter time immobile during the day than the control fish.

Photomotor response in juveniles and adults

GLM analyses revealed significant differences in the photomotor response between periods, for 2-month-old PY fish (F = 16.95; p < 0.001; Fig. 2a) with higher levels of activity during and after the dark challenge. There was a trend for PY-3X juveniles to display a smaller increase in activity during the dark period, but there was no significant difference between concentrations, whatever the period (Fig. 2a, focus graph). In 6-month-old fish, similar significant differences between periods (F = 42.33; p < 0.001) were observed, but with no effect of concentration or sex (data not shown).

Fig. 2
figure 2

Photomotor response of 2-month-old fish, for each fraction: PY, HO, and LO. a, c, e Distance traveled (cm/min) over 1 h, before a 15-min lights-off period (gray shadowing) followed by an additional 1.5 h of monitoring. The superimposed view shows the distance traveled summed per 15-min period before, during and after the lights-off period. b, d, f Time spent immobile (s/15 min) per 15-min period before, during, and after the lights-off period. Values are means ± SEM; n = 12 fish per fraction and concentration; * indicates a significant difference at p < 0.05

Full size image

Two-month-old fish spent more time immobile before the challenge than after the challenge (Fig. 2b; period effect F = 64.63, p < 0.001). A concentration effect (F = 3.36, p < 0.05) was observed (PY-1X fish spent more time immobile during light on-1, p < 0.05) but no significant interaction was detected. For 6-month-old fish, we observed only a similar period effect (F = 214.82, p < 0.001, data not shown).

In 2-month-old fish exposed to the HO fraction, there were significant differences between periods (F = 7.26; p < 0.001; Fig. 2c), with greater locomotion during and after the challenge, but no significant concentration effect or interaction was detected. Similar results were obtained for 6-month-old fish (data not shown), with significant differences between periods (F = 55.94; p < 0.001) and between concentrations (F = 3.73; p < 0.05), and there was a significant interaction between period and concentration (F = 2.23; p < 0.05), with H0-3X fish displaying lower levels of locomotor activity during the dark period (p < 0.001) and the following light on-2 period (p < 0.05).

A period effect on the time spent immobile was observed in 2-month-old fish (Fig. 2d; F = 6.40, p < 0.01), with a small decrease during and after the challenge. There was no concentration effect but a significant interaction between period and concentration was detected (F = 3.14, p < 0.001): HO-3X fish spent more time immobile during the period in which the light was switched off (p < 0.05). For 6-month-old fish, a period effect was observed, with the same pattern as for 2-month-old fish (F = 104.04, p < 0.001, data not shown), together with a concentration effect (F = 7.36, p < 0.001) and a significant interaction (p < 0.001): HO-3X fish spent less time immobile than control fish during the light on-1 and light on-2 periods (p < 0.01). A sex effect (F = 6.85, p < 0.01) was also detected, with males spending less time immobile (262.51 ± 28.88 s) than females (322.44 ± 35.19 s). No other significant interaction was detected.

The locomotor activity of 2-month-old LO-3X fish was lower than that of fish fed other concentrations of LO (concentration effect, F = 12.68; p < 0.001) throughout the photomotor response challenge, with no difference in the distance traveled whatever the light period for LO-3X (Fig. 2e). A period effect was seen for other fish (F = 3.42; p < 0.05), but there was no significant interaction. Similar responses were observed for 6-month-old LO-3X fish, with lower levels of activity (data not shown) and a significant difference between periods (F = 31.54; p < 0.001), together with a sex effect (F = 6.19, p < 0.05): female fish were less active (1,611 ± 98.58 cm / 15 min) than male fish (1,843 ± 106.95 cm / 15 min) with no other significant interactions detected.

Two-month-old fish (Fig. 2f) spent more time immobile before the challenge (period effect F = 18.10, p < 0.001) but there was no concentration effect and no significant interaction. Six-month-old fish also spent more time immobile before the challenge (period effect F = 67.85, p < 0.001, data not shown) and there was a concentration effect for fish of this age (F = 4.70, p < 0.01): LO-1X and LO-3X fish spent less time immobile during light on-2 (p < 0.001 and p < 0.05 respectively), but no significant interaction was detected. A sex effect was found (F = 15.54, p < 0.001) with males spending less time immobile (217.51 ± 23.85 s) than females (276.21 ± 26.09 s). No other significant interactions were observed.

Exploration of a novel environment

We investigated the exploration of the T-maze (PY diets) and the Z-maze (HO and LO diets) in 2- and 6-month-old fish (data not shown for the 6-month-old fish).

For all three diets, there was a trend towards a dose-dependent increase in the time taken to leave the start area, but a significant difference between concentrations was observed only for fish fed HO diets (F = 5.27; p < 0.01) with fish on the HO-3X diet taking almost twice as long to exit the start area for the first time as the other fish fed HO (p < 0.01; Fig. 3). A significant sex effect was observed for LO fish (F = 7.73, p < 0.01), with males leaving the start area earlier (36.12 ± 38.34 s) than females (58.42 ± 40.23 s). No other significant interactions were measured.

Fig. 3
figure 3

Maze exploration by 2-month-old fish for each fraction: PY, HO, and LO. a Time (s) to first exit from the start area in the T-maze (PY fraction). b, c Time (s) to first exit from the start area in the Z-maze (HO and LO fractions, respectively). Values are means ± SEM; n = 12 fish per fraction and concentration; different letters indicate significant differences at p < 0.05

Full size image

In the T-maze challenge, whatever the concentration, 2-month-old fish spent more time in the shallow area than in the other two areas (F = 21.36; p < 0.001; Fig. 4a). No other interactions were measured. Similar results were obtained for 6-month-old fish (F = 10.20, p < 0.001; data not shown).

Fig. 4
figure 4

Maze exploration by 2-month-old fish for each fraction: PY, HO, and LO. a Total time spent (s) in the start, shallow and deep areas of the T-maze (PY fraction). b, c Total time spent (s) in the start, close and distant areas of the Z-maze (HO and LO fractions, respectively). Values are means ± SEM; n = 12 fish per fraction and concentration; different letters indicate significant differences at p < 0.05

Full size image

For fish on the HO diet, differences were observed in the time spent in the various areas (F = 14.70; p < 0.001; Fig. 4b) and there was a significant interaction between area and concentration (F = 2.83; p = 0.013): a dose-dependent decrease in the time spent in the distant area was observed, with HO-3X fish spending only a quarter the time spent by control fish in this area (p < 0.05). A significant difference in the time spent in the various areas was observed in 6-month-old fish (F = 8.13; p < 0.001; HO-3X fish not tested), and post hoc tests revealed that this difference was due to less time being spent in the start area. There was no effect of sex and no other interaction.

Two-month-old fish on LO diets spent more time in the close area than control fish (F = 10.65; p < 0.001; Fig. 4c) but no difference between concentrations and no interactions were observed. No differences were observed with 6-month-old adults.

We evaluated the exploratory tendencies of the tested fish further, by analyzing individual success or failure to reach the deep zone (T-maze) or quadrant 16 (the farthest away from the start area in the Z-maze; Fig. 5a). For 2-month-old fish on PY diets, success in reaching the deep zone depended on concentration (chi2 = 8.0, p < 0.05), with 0.3X and 3X diets giving similar results to those obtained for control fish and lower scores obtained for the 1X diet (only half the fish reached the deep area). The success of HO fish in reaching the farthest area of the maze also depended on concentration (chi2 = 16.6, p < 0.001): higher success rates were obtained for control and 0.3X fish than for 1X and 3X fish. Success rates were similar for all fish fed with LO diets (6 to 9), regardless of concentration.

Fig. 5
figure 5

Maze exploration by 2-month-old fish for each fraction: PY, HO, and LO. a Proportion of individuals successfully reaching the deep zone of the T-maze or the farthest quadrant (16) of the Z-maze. b Time (s) to first entry into the deep area of the T-maze (PY) or the farthest quadrant of the Z-maze (HO, LO). c Proportion of Z-maze explored in HO and LO fish. Values are means ± SEM; the numbers of fish reaching the deep area or quadrant 16 per fraction and concentration are indicated on the bars; different letters indicate significant differences at p < 0.05

Full size image

For each fraction, we focused our analyses on fish reaching the deep zone or quadrant 16. Whatever the fraction, the time taken by 2-month-old fish to reach the deep zone or quadrant 16 was not affected by the exposure, although fish fed with PY or HO tended to take longer to reach these zones than control fish (Fig. 5b). In 6-month-old fish, a significant concentration effect was shown only for HO (F = 3.75, p < 0.05): control fish took significantly less time to reach quadrant 16 than fish fed with the 0.3X and 1X diets. A sex effect was observed for LO fish (F = 10.95, p < 0.01), with males taking half the time to reach quadrant 16 (77.46 ± 14.60 s) than females (150.55 ± 17.71 s). No other interactions were observed.

For the Z-maze, the tendency to explore, analyzed by determining the most distant quadrant reached, decreased significantly with increasing dose in HO fish (F = 4.50; p < 0.01, Fig. 5c), with HO-1X fish reaching a significantly lower numbered quadrant than 0.3X fish (p < 0.05), with a further halving of the quadrant number reached for HO-3X fish (p < 0.001). A similar tendency for a decrease in exploratory tendency with increasing dose was observed in 6-month-old fish (F = 5.17; p < 0.05; data not shown). No significant effect was detected for LO fish.

Novel tank test

In all cases, regardless of the fraction and concentration considered, the proportion of time spent in the top area increased with time during the 6-min challenge (p < 0.001; Fig. 6). In 2-month-old fish fed PY, the time spent in the top area increased significantly over time (F = 19.49, p < 0.001) and concentration effects were observed (F = 5.15; p < 0.001; Fig. 6a), with PY-3X fish spending less time in the top area than control fish (p < 0.05), but no interactions were observed. The distance traveled and the time spent immobile were similar for all concentrations (Fig. 6b-c). In 6-month-old fish, the same increase in the time spent in the top area over time was observed (F = 19.50; p < 0.001), together with similar differences between concentrations (F = 3.14; p < 0.05), with 6-month-old PY-1X fish spending less time than control fish in the top area (p < 0.05). No sex effects or other interactions were detected.

Fig. 6
figure 6

Novel tank challenge in 2-month-old fish for each fraction: PY, HO, and LO. a, d, g Mean time spent (s) per minute (Min-1 to Min-6) in the top area of the novel tank. b, e, h Total distance traveled (cm) over the 6-min challenge. c, f, i Total time spent immobile (s) over the 6-min challenge. Values are means ± SEM; n = 12 fish per fraction and concentration; different letters and * indicate significant differences at p < 0.05

Full size image

In 2-month-old HO fish, significant differences over time (F = 10.21; p < 0.001) and between concentrations (F = 21.93, p < 0.001) were observed during the 6-min challenge, with HO-1X and HO-3X fish spending less time in the top area than control fish (p < 0.001, for both concentrations). For 6-month-old HO fish, only time was a significant factor (F = 3.90; p < 0.01), with fish showing the same pattern as 2-month-old fish. The distance traveled was also significantly lower for 2-month-old HO-3X fish than for fish on other HO diets (F = 7.41; p < 0.001), and these fish tended to spend longer periods of time immobile (p = 0.078). In tests on 6-month-old fish, significant differences were observed only for the time spent in the top area at different times (F = 3.90; p < 0.01), with fish displaying the same pattern as 2-month-old fish. No effect of sex and no interactions were detected.

In 2-month-old LO fish, we observed an effect of time (F = 7.44, p < 0.001) and of concentration (F = 2.88; p < 0.05), with LO-3X fish spending less time in the top area than control fish. LO-3X also traveled shorter distances than control fish (F = 4.44, p < 0.01; post hoc p < 0.01) and there was no significant difference in time spent immobile between concentrations. In 6-month-old fish, we detected effects of time (F = 12.32, p < 0.001) and concentration on the time spent in the top area (F = 2.64; p < 0.05): contaminated fish spent less time in the top area than control fish. There was also a sex effect (F = 6.35, p < 0.05), with males spending more time in the top area (22.71 ± 17.64 s) than females (18.79 ± 16.15 s). No other significant interactions were detected.

Discussion

In this study, we exposed fish to three PAH mixtures through diet, beginning at the first meal (5 dpf), and continuously thereafter (Vignet et al. 2014b). This exposure route was chosen because it is a major contamination pathway for fish in natural conditions (Johnson et al. 2007; Varanasi et al. 1993; Yanagida et al. 2012). The three mixtures used were extracted aromatic fractions with different compositions, representatives of different mixtures encountered in the environment. PY was characterized by a high proportion of heavy PAHs and a very low proportion of alkylated PAHs. LO contained a high proportion of alkylated PAHs and a low proportion of heavy PAHs, whereas HO had intermediate proportions of both groups of PAHs. The total concentration of PAHs was in the 4.6–6.7 μg g−1 dw range for the intermediate 1X concentration. This concentration was close to the targeted concentration of 5 μg g−1 dw, reported in copepods and bivalves in chronically contaminated areas (Cailleaud et al. 2007; Jeanneret et al. 2002; Le Goff et al. 2006; Payne et al. 2008), but lower than concentrations occasionally found after accidental pollution incidents, such as the 14.4 μg g−1 measured in mussels after the Exxon Valdez oil spill (Payne et al. 2008). The levels of exposure used in this study were, therefore, representative of different situations occurring in the environment. Heavy metals and other organic pollutants (e.g., PCBs) have been shown to modify behavior. Aromatic fractions were prepared with methods excluding the copurification of heavy metals. PCBs are copurified with PAHs, but there are no PCBs in petrogenic fractions. For the PY fraction extracted from Oissel sediment, previous reports have indicated the presence of PCBs at very low concentrations (∑PCBs = 70 ng g−1and dioxin-like congener CB118 = 8.4 ng g−1 (Cachot et al. 2006)). We therefore conclude that the behavioral defects reported here are essentially due to the PAHs present in the different fractions.

Our results for the set of variables studied here indicate that all three fractions caused behavioral disruptions. The pattern of disruption was similar, but there were clear differences between fractions in terms of the severity of the disruptions observed. The HO fraction was the most deleterious, followed by the LO and PY fractions. These results suggest that alkylated PAHs play an important role in triggering behavioral disruption, as the LO fraction (mostly alkylated PAHs) was more toxic than the PY fraction (heavy PAHs). The higher toxicity of the HO fraction than of the other two fractions suggests a possible additive effect of heavy and alkylated PAHs.

Major advances have been made in the identification and characterization of behavioral disruptions in zebrafish larva, juveniles, and adults, in terms of basic traits, such as stress or anxiety levels, and more complex traits, such as social interactions and learning abilities (Arthur and Levin 2001; Bailey et al. 2013; Cachat et al. 2010a; Champagne et al. 2010; Egan et al. 2009; Norton and Bally-Cuif 2010; Steenbergen et al. 2011; Stewart et al. 2012). Tests and variables for evaluating the disruption of these traits have also been defined. These tests have also been used to evaluate the effects of exposure to environmental pollutants on fish behavioral responses per se or as an integrated indicator of physiopathological disruptions (Baraban et al. 2005; He et al. 2011; Kusch et al. 2008; Levin et al. 2003, 2011; Nakayama et al. 2005; Péan et al. 2013; Saili et al. 2012; Schmidt et al. 2005; Timme-Laragy et al. 2006).

There have been few studies of the consequences of PAH exposure for fish behavior. Most of the studies published to date were performed after brief waterborne exposure to a single PAH or a simple mixture of individual PAHs. In these studies, the variables monitored were swimming ability (Almeida et al. 2012; Gravato and Guilhermino 2009; Oliveira et al. 2012) or behavioral traits evaluated in groups (Correia et al. 2007; Goncalves et al. 2008). Exposure to single compounds has been shown to lead to a decrease in swimming ability, evaluated in a device similar to a swim tunnel, and the variables considered in these studies were resistance to a counter current flow and/or swimming velocity (Almeida et al. 2012; Gravato and Guilhermino 2009; Oliveira et al. 2012). These effects are of the utmost importance in an ecological context, because the disruption of swimming ability may greatly impair the ability of the fish to find and capture preys or to avoid predators. However, these effects may have a number of causes not related to behavior per se; such as an impaired energy budget or respiration. In the studies focusing on behavioral traits, both swimming ability and social interactions were monitored (Correia et al. 2007; Goncalves et al. 2008). In these cases, monitoring was performed on groups of fish and the authors concluded that locomotor activity decreased after exposure to fluorene, pyrene, or phenanthrene, or to a mixture of these three PAHs and that social interactions decreased after exposure to phenanthrene (Correia et al. 2007; Goncalves et al. 2008). In these group studies, it is difficult to identify the behavioral disruptions affecting individual fish, due to potential interactions between fish reflecting group structure or hierarchy. Nevertheless, all these articles concluded that exposure led to a decrease in swimming activity, which was presented as a lethargy-mediated effect of the tested PAHs.

The protocols used here monitored individual behavior and were chosen for the evaluation of several behavioral traits: circadian spontaneous swimming activity, response to a light challenge, exploration ability, and behavioral response to a novel environment. The tests used and the variables extracted could also be used for the evaluation of integrative traits, such as anxiety level. Hence, our results can be analyzed in terms of two groups of broad effects. Firstly, control fish spent more than half the time [57–72 %] immobile, during both the day and night periods, whereas the proportion of time spent immobile was significantly lower for HO-3X fish (~20 % of the time, during both periods) and for HO-1X and LO-0.3X, LO-1X and LO-3X fish during the day period (<40 % of the time). Thus, the contaminated fish had shorter resting times and therefore, probably, a greater demand for energy. They also traveled a similar distance to control fish, suggesting a more continuous low swimming speed. Alternatively, these fish may have been displaying bouts of high activity due to “erratic” movements, indicative of anxiety (Cachat et al. 2010b; Egan et al. 2009).

This second interpretation would be consistent with the second group of effects observed. Behavioral responses in mazes or the novel tank were characterized by a delay in exit from the refuge zone, lower levels of exploration (in mazes; HO-1X and HO-3X), and less time spent in the top area (in the novel tank; PY-3X, HO-1X, HO-3X, and LO-3X). All these findings indicate an increase in anxiety (Cachat et al. 2010b; Egan et al. 2009; Stewart et al. 2010, 2011). High cortisol concentrations are often considered an indicator of anxiety/stress (Alsop and Vijayan 2008; Cachat et al. 2010b). An increase in cortisol concentration has been observed in trout after exposure to BaP and a prototypic AhR agonist, β-naphthoflavone (Gesto et al. 2008; Tintos et al. 2008), and in three-spined stickleback after exposure to produced water (Knag and Taugbol 2013). These findings are consistent with exposure to PAHs triggering anxiety. Further studies, with the inclusion of anxiolytic and anxiogenic drugs with the different fractions and concentrations, together with cortisol determinations in the water, would be useful for testing this hypothesis (Felix et al. 2013).

In addition to these principal findings, we also found that some responses were weaker in adult fish than in juveniles, suggesting that the sensitivity of these tests is greater in juveniles, which should be preferred for such studies. Sex effects were also occasionally observed; suggesting that the sex of the fish tested should also be taken into account whenever possible.

A strong induction of tumorigenesis has been reported in fish exposed to all three fractions tested here (Larcher et al. In revision), and some of the behavioral disruptions identified here could be a consequence of this. However, no tumors were found in the nervous system of the fish, ruling out a direct effect of tumors on behavior. It remains possible that a global failure of metabolism due to tumor development hinders the expression of some types of behavior. However, respiratory metabolism was evaluated for both PY- and HO-exposed fish and no effect of exposure was detected (Julie Lucas, personal communication). This suggests that no such mechanism is involved in the onset of behavioral disruptions described here.

In the framework of this project, we obtained several lines of evidence for endocrine disruption, such as a shift in the timing of hormone expression or spawning disruption (unpublished results). Steroid hormones regulate reproductive and nonreproductive behavior and, as a consequence, endocrine disruption modifies individual or social behavior in fish (Belanger et al. 2010; Filby et al. 2012; Oliveira et al. 2009; Saaristo et al. 2010; Salierno and Kane 2009; Soffker et al. 2012; Stacey 2003) and other species (Ottinger et al. 2013; Panzica et al. 2007). It is therefore tempting to suggest that the endocrine disruption activity of the fractions tested contributed to the behavioral modifications described here.

However, the mechanisms underlying these behavioral disruptions remain largely unknown. Monoamines, including serotonin and dopamine, are involved in the control of a wide number of biological processes (see reviews (Høglund et al. 2005; Øverli et al. 2005)). BaP and β-naphthoflavone disrupt dopaminergic and serotonergic systems in the brain of rainbow trout, as revealed by an increase in the turnover of these neurotransmitters (Gesto et al. 2008, 2009). In addition, as experimental exposure began early in this study (5 dpf), at a time at which neurogenesis was still underway and the nervous system was maturing (Kastenhuber et al. 2010; McGraw et al. 2012; Souza and Tropepe 2011), the phenotypes observed may reflect this early disruption. Similar effects have been demonstrated following the embryonic disruption of monoamine systems (Dennis et al. 2013; Shabanov et al. 2005; Silva et al. 2013) or early exposure to silver or cadmium (Kusch et al. 2008; Powers et al. 2011). Such a delayed effect has also been demonstrated in zebrafish, following the exposure of embryos or larvae to PAH mixtures (Vignet et al. 2014a), in rat, after perinatal exposure (Crépeaux et al. 2012) and after the exposure of Pacific herring embryos to weathered crude oil (Carls et al. 1999).

These results establish parallels between the effects of chronic exposure in fish and the neurobehavioral disruptions observed in humans and rodents after exposure to PAHs (Bouayed et al. 2009; Crépeaux et al. 2012, 2013; Edwards et al. 2010; Jedrychowski et al. 2003; Perera et al. 2012; Saunders et al. 2001, 2003; Tang et al. 2008; Xia et al. 2011). In addition to confirming findings for humans, studies of this kind provide an opportunity to decipher the underlying mechanisms. Finally, this study provides support for the use of behavioral endpoints in the framework of environmental quality assessment.

In conclusion, behavioral disruptions have been observed in zebrafish, after chronic, long-term dietary exposure to PAHs. These disruptions, indicative of an increase of anxiety, may have detrimental consequences in terms of the performance of the exposed fish, their ability to survive and to explore their environment and, hence, their ability to contribute to next generation. Furthermore, the severity of the disruption depended on the mixture used, with the HO and LO fractions having a more severe effect than PY. More behavioral analyses, associated with molecular research, are required to identify the mechanisms underlying the observed disruption and the particular toxicity of alkylated PAHs.