royalsocietypublishing.org/journal/rspb Research Cite this article: Sadoul B, Alfonso S, Goold C, Pratlong M, Rialle S, Geffroy B, Bégout M-L. 2022 Transcriptomic profiles of consistent risk-taking behaviour across time and contexts in European sea bass. Proc. R. Soc. B 289: 20220399. https://doi.org/10.1098/rspb.2022.0399 Received: 1 March 2022 Accepted: 14 April 2022 Subject Category: Behaviour Subject Areas: behaviour, molecular biology, physiology Keywords: coping style, behavioural syndrome, fish, personality, anti-predator behaviour, intra-individual variability Author for correspondence: Bastien Sadoul e-mail: bastien.sadoul@agrocampus-ouest.fr † Co-last. Electronic supplementary material is available online at https://doi.org/10.6084/m9.figshare. c.5958649. Transcriptomic profiles of consistent risk-taking behaviour across time and contexts in European sea bass Bastien Sadoul1,2, Sébastien Alfonso1,3, Conor Goold4, Marine Pratlong5, Stéphanie Rialle5, Benjamin Geffroy1,† and Marie-Laure Bégout1,† 1 MARBEC, University of Montpellier, CNRS, Ifremer, IRD, Palavas-Les-Flots, France DECOD, Ecosystem Dynamics and Sustainability, Institut Agro, Ifremer, INRAE, Rennes, France 3 COISPA Technology and Research, Experimental Station for the Study of Sea Resources, Bari, Italy 4 Faculty of Biological Sciences, University of Leeds, LS2 9JT, UK 5 MGX-Montpellier GenomiX, University of Montpellier, CNRS, INSERM, Montpellier, France 2 BS, 0000-0001-5968-3983; SA, 0000-0002-2471-2876; CG, 0000-0002-9198-0889; MP, 0000-0002-3406-9829; BG, 0000-0001-6120-1103; M-LB, 0000-0003-1416-3479 Bolder individuals have greater access to food sources and reproductive partners but are also at increased risk of predation. Boldness is believed to be consistent across time and contexts, but few studies have investigated the stability of this trait across variable environments, such as varying stress loads or long periods of time. Moreover, the underlying molecular components of boldness are poorly studied. Here, we report that boldness of 1154 European sea bass, evaluated using group risk-taking tests, is consistent over seven months and for individuals subjected to multiple environments, including a chronically stressful environment. Differences in risk-taking behaviour were further supported by differences observed in the responses to a novel environment test: shy individuals displayed more group dispersion, more thigmotaxic behaviour and lower activity levels. Transcriptomic analyses performed on extreme phenotypes revealed that bold individuals display greater expression for genes involved in social and exploration behaviours, and memory in the pituitary, and genes involved in immunity and responses to stimuli in the head kidney. This study demonstrates that personality traits come with an underpinning molecular signature, especially in organs involved in the endocrine and immune systems. As such, our results help to depict state–behaviour feedback mechanisms, previously proposed as key in shaping animal personality. 1. Introduction For most animals, boldness could be viewed as a risky strategy as it may increase the probability of being predated. At the same time, risk-taking behaviour of bold individuals also increases their likelihood of finding more suitable environments (e.g. for food or mating). This trade-off is known to favour bolder individuals when predation is low and shyer individuals when predation is high [1]. Risk-taking behaviour is therefore tightly linked to the eco-evolutionary dynamic of species in a given environment. Animals maximize their fitness through plasticity (i.e. variation that depends on contexts within individuals) and selection of adaptive behaviours (i.e. variation over generations within species). Indeed, it has previously been demonstrated in numerous species that the environment drives the average behaviour of a population [2,3]. The most evident examples relate to the effects of differences in predation risks between two environments, including as a consequence of human presence [4]. Nevertheless, within a given population, individuals differ in their behavioural phenotypes. Some are more prone to take risks than others, and strong heterogeneity in risk-taking has been related to global population sizes and larger spatial ranges [5]. Interestingly, these inter-individual differences in © 2022 The Author(s) Published by the Royal Society. All rights reserved. 2. Material and methods 2 (a) Consistency of group risk-taking behaviour over time and different environmental conditions royalsocietypublishing.org/journal/rspb (i) Experimental animals All fish used in this study were part of a larger experiment which led to a first publication [29], providing all details on the rearing protocol. Briefly, 2053 European sea bass were individually tagged at 175 ± 3 days post-fertilization (dpf ), using an ISO PIT tag (8 × 1.4 mm), to ensure individual identification. They were dispatched in three 1.5 m3 tanks and supplied with water at a constant temperature of 21°C. At 255 ± 2 dpf and 358 ± 2 dpf, a total of 288 and 492 fish respectively were randomly extracted from the tank for two other studies, one being published [29]. From tagging to the end of the experiment (at 462 dpf ), 119 (out of 2053, 5.8%) fish died of unknown reasons. (ii) Group risk-taking tests Over the whole experiment (described below) fish were evaluated in four group risk-taking tests (GRTs), performed following a previously described protocol [26]. All GRTs were conducted in a tank of the same size and shape as the rearing tanks, and supplied with water from the same system. The tank was vertically divided in two parts of equal volume with an opaque screen. In the middle of the screen, a 12 cm hole was surrounded by a circular RFID-reading antenna (DORSET, The Netherlands) connected to a computer. The hole was blocked with a door that could be opened without being seen by the fish. Fish were fasted for 24 h prior to the start of the experiment. They were then lightly anaesthetized (100 ppm of Benzocaine, 150 g l−1, E1501, Sigma, Saint Louis, MO, USA) and transferred between 10.00 and 11.00 on one side of the tank that we covered with a black tarp in order to get complete darkness. The other side was not covered in order to create a riskier zone. After 2 h of habituation, the door was opened to allow the fish to move freely to the other side of the tank; PIT tags were read by the antenna and the time of first passage (Latency to exit (s)) was stored. In order to avoid a too long starvation period, we censored the test to 20 h. After the test was completed, fish were anaesthetized (300 ppm of Benzocaine) and were measured for length (cm) and weight (g) and transferred back into their original rearing tank. (iii) Experimental protocol At 255 ± 4 dpf, the first risk-taking test (GRT1) was conducted for each rearing tank, providing a risk-taking phenotype for the 1748 fish available (figure 1). At 309 dpf, fish from all three tanks were transferred in six tanks of 1 m3 with equal number of fish from each initial rearing tank (figure 1). At 336 dpf, and for a period of three weeks, three tanks were stressed as previously detailed [29], while the three other tanks were kept under normal rearing conditions (figure 1). Briefly, stressors consisted in random light flashes, chasings and confinements randomly programmed over the three weeks. At 358 ± 1 dpf, a subsample from each tank was extracted for the analyses published in another paper [29]. The remaining fish (n = 1187) were transferred to three 1 m3 tank making sure that previously stressed and control fish were equally represented. Two weeks later, at 373 ± 1 dpf, fish from each tank were evaluated in a new risk-taking test (GRT2) performed for each tank (figure 1). A total of 1187 individuals were phenotyped at GRT2. The third GRT (GRT3) was performed six weeks later at 423 ± 3 dpf, on each tank for the 1181 remaining fish. At 443 dpf, all fish were then transferred in a unique 5 m3 tank, and at 463 dpf, the last risk-taking test was performed at once in a 5 m3 tank (GRT4) on the 1154 remaining fish. All fish were then euthanized and a random subsample (n = 606) was sexed Proc. R. Soc. B 289: 20220399 risk-taking behaviour are relatively consistent over time and contexts [6], and being bold or shy is thus considered as a personality trait in animals [7]. Personality traits have been observed for a wide range of terrestrial and aquatic species and linked to fitness in a given environment [8]. The emergence of personalities were previously proposed to be linked to key life-history traits, such as metabolism and growth, through a state–behaviour feedback [9,10]. In this framework, individuals that are more prone to take risks also exhibit lower hypothalamo–pituitary–interrenal/adrenal axis responses, but stronger hypothalamo–sympathetic activity [11]. These interconnections between behaviour and stress physiology are part of the concept of coping styles [11] and might lead individuals to consistently differ in behaviour. Yet, it remains unclear to what extent personality traits are governed by the genetic architecture of an individual and/or depend on the environment it will experience [12–14]. In addition, we still lack information regarding the molecular mechanisms underlying these risk-taking-related behaviours, although some efforts have been made to elucidate the genetic component of traits, e.g. boldness: [15]. Most studies of animal personality focus on short time periods or one life stage. This focus provides a limited view of how personality interacts with physiology, life stage, experience and the environment [16]. Understanding molecular mechanisms behind key behavioural phenotypes, such as the boldness/shyness continuum, is of primary importance in behavioural ecology. In this study, we characterize the boldness of European sea bass (Dicentrarchus labrax) by assessing the consistency of risk-taking-related behaviour across multiple environments, challenges and time. Then, we highlight molecular pathways in the brain, the pituitary and the head kidney associated with these divergent behavioural phenotypes. These three organs were chosen because (i) the brain centralizes perceptions of the environment and controls behaviour and physiology [17], (ii) the pituitary secretes hormones regulating multiple processes including growth, sexual maturity, stress response and energetic balances [18] and (iii) the head kidney, analogous to the adrenal gland in mammals, plays a central role in the immune system and stress regulation by secreting cortisol and catecholamines, the major stress hormones in fish [19,20]. The European sea bass was used as a model species, because of its ecological and economic importance [21]. Previous work demonstrated strong interindividual variation in metabolism [22], stress response [23], feeding behaviour [24] and coping style [25] of the species. Consistency in risk-taking behaviour over three months has also been demonstrated in European sea bass [26] but has never been challenged by environmental variation, especially by extreme events such as a chronic stress period. Also, within a population, correlations exist between molecular phenotypes related to behaviour and coping capacities [27], but no conserved transcriptomic signature for boldness has been shown in the sea bass whole brain [28]. The present study aims (i) to investigate the longterm consistency in risk-taking behaviour of individuals from a large sea bass population (n = 1154), (ii) to demonstrate that extreme individuals (i.e. very shy or very bold) contrastingly respond to a different behavioural test and then (iii) to depict molecular differences in gene expression across specific organs (brain, pituitary and head kidney) in those extreme individuals. G RT 3 n /h ove yp l e ox nv ia iro te nm st ta en nk t tra ns fe r G RT 4 se sa xin m g pl an in d g br ai n tra ns fe G r RT 2 ta nk tra ns fe r ta nk 423 429 463 chronic stress Time (dpf) 175 255 309 336 358 N=3 N=3 N=3 N=3 N=3 N=3 N=3 N=3 N=3 1748 1187 443 478 N=1 1181 96 1154 Figure 1. Experimental protocol assessing consistency of risk-taking behaviour over time and environments. Fish were reared in replicated tanks (N = number of rearing tanks). A total of four GRTs were performed on each rearing tank. Colour of the tank illustrates a different rearing condition. Until 309 dpf, juveniles were reared in 1.5 m3 tanks (dark green tank), they were then transferred in 1 m3 tanks (light green tanks) and half of the tanks were stressed (blue tanks) from 336 to 358 dpf. At the end of the stress period, fish were mixed and transferred to other 1 m3 tanks (yellow tanks) until 443 dpf when they were all transferred in a unique 5 m3 tank (orange tank). (Online version in colour.) (figure 1) indicating that the population was composed of 54% of females (329 females versus 277 males). (iv) Statistical analyses of group risk-taking test data We analysed the GRT data with hierarchical Bayesian regression with a lognormal likelihood function in Stan [30] and cmdstanr [31] in R v. 4.0.5 [32]. We included data from 1748 fish who participated in GRT1 and who were assigned a chronic stress group. Missing data due to fish extracted from the experiment (see previous section) were imputed in the Bayesian model rather than being discarded. Censored values were accounted for via imputation with a lower bound of 20 h. Fish were categorized as bold or shy at the first GRT (at 255 dpf ) based on whether they were above (shy) or below (bold) the median latency in their tank. We included GRT as an observation-level predictor and bold/shy (bold = 875), chronic stress (yes = 874) and sex (female = 598, male = 494) as individual-level predictors. Six-hundred-andfifty-five fish were missing sex assignment, which was accounted for by marginalizing out of the sex variable for those fish [33]. Four models were run and compared using the expected log predictive density (ELPD) estimated via Pareto-smoothed importance sampling leave-one-out cross-validation (ELPD-LOO [34]) in the loo package [35]. Model 1 included random intercepts (i.e. personality) for individuals, model 2 extended model 1 with random slopes by GRT time (i.e. behavioural plasticity) for individuals, model 3 extended model 2 by including a random intercept of initial tank and model 4 extended model 3 by modelling heterogeneous residual variances by individual (i.e. behavioural predictability). From these models, we calculated behavioural repeatability (intraclass correlation coefficients; ICCs) at each GRT time point (i.e. conditional repeatability [36]) as well as the coefficient of variation for predictability as a measure of behavioural predictability effect size [37]. We summarize the parameters and results using the posterior mean and 95% highest density interval (HDI; the most probable values). (b) Group behavioural response to novel environment and hypoxia challenge exits or no exit at all) or bold behavioural type (three early exits) were selected and challenged in novel environment and a hypoxia challenge at 429 ± 2 dpf (mean ± s.e., 91.0 ± 6.6 g and 194 ± 4.8 cm). This test was performed in groups of eight fish with similar behaviour (i.e. n = six trials*eight individuals per behavioural type, i.e. bold versus shy). The trials to monitor the group behavioural response to a novel environment and hypoxia were carried out according to the protocol described in [29] and detailed in the electronic supplementary material, SMM1. (ii) Statistical analyses of behavioural data in the novel environment test Measured variables (time spent in the periphery, velocity and inter-individual distances) during the novel environment and hypoxia challenge were analysed independently before and after the hypoxia challenge. A linear mixed model was fitted with risk-taking behavioural phenotype (bold or shy), time (categorical) and their interaction as fixed effects and the day of experiment and trial as random effects. A Tukey post hoc test was completed with the glht function from the multcomp package [38] to test significant differences between the levels of a significant fixed effect. (c) Transcriptomic signatures between bold and shy individuals (i) Samplings At 478 ± 2 dpf, i.e. two weeks after GRT4 (figure 1), fish were caught from their rearing tank and euthanized using 1500 ppm of Benzocaine. Sampling of the whole brain, pituitary and head kidney was performed on five individuals per behavioural phenotype chosen among the individuals also tested in the novel environment test (i.e. 49 days after the novel environment test). The 30 samples collected (five individuals × two behavioural phenotypes × three tissues) were flash-frozen in liquid nitrogen and kept at −80°C until further analyses. Individuals were chosen to be equally represented for each sex (two shy males, three shy females, three bold males and two bold females). (i) Behavioural tests in a novel environment (ii) RNA sequencing and analyses Among fish screened during GRT1, GRT2 and GRT3, two subsamples of 48 individuals displaying consistent shy (three late Details on RNA extraction, sequencing and analyses are provided in the electronic supplementary material, SMM2. Briefly, Proc. R. Soc. B 289: 20220399 no. fish in the test 373 royalsocietypublishing.org/journal/rspb Events in di vi du al ta G gg RT in 1 g 3 3. Results The average ICC for between-individual differences was 0.38 (95% HDI: [0.30, 0.47]) and was highest at GRT1 (GRT1: 0.44, HDI: [0.38, 0.50]; GRT2: 0.33, HDI: [0.29, 0.37]; GRT3: 0.41, HDI: [0.36, 0.46]; GRT4: 0.36, HDI: [0.31, 0.42]). The ICC for between-tank differences was very low (0.003, HDI: [0.000, 0.012]). The model with the highest ELPD was model 3 (ELPD-LOO: −9832.8; model 1 = −10118.0, model 2 = −10071.8, model 4 = −9835.5) including behavioural predictability. The coefficient of variation for behavioural predictability was relatively high at 0.57 (HDI: [0.51, 0.64]). Fish had generally lower latencies to leave the sheltered areas at GRT1 and GRT2 compared to GRTs 3 and 4, but there were interactions between GRT and fish types (figure 2). Fish categorized as bold (see estimates in figure 2) at GRT1 had significantly lower latencies (between 3 and 5 h) at the following GRTs than shy fish. This was repeatable across the rearing tanks (electronic supplementary material, figure S1). Female fish had a tendency for lower latencies than male fish across the GRTs, but the difference was only significant at GRT3 and GRT4 (approximately 2–3 h quicker). Chronic stress had no credible influence on latency to leave the shelter. (b) Group behavioural response to novel environment and hypoxia challenge After introduction in the novel environment, both behavioural phenotypes (bold and shy) showed first a low swimming velocity which gradually increased to reach a plateau after 30 min. Bold individuals showed a tendency for higher swimming velocity after 10 min compared to shy individuals ( p = 0.057; figure 3a). At introduction in the novel environment, shy individuals showed increased dispersion in the tank (figure 3b; p < 0.05) and spent more time in the periphery (figure 3c; p < 0.05) than bold individuals. After 20 min, both behavioural phenotypes showed a similar group behaviour pattern which became constant after 30 min (figure 3c). Both behavioural phenotypes displayed a same response to the hypoxia challenge for the three behavioural measures (figure 3). (c) Transcriptomic signatures between bold and shy individuals In the pituitary and the head kidney, 556 and 141 genes were DE between bold and shy individuals respectively. In the brain, only six genes were DE and four of them were annotated (nr2e3, glipr2, dsg2 and pla2g1b). In the 4. Discussion (a) Consistency of group risk-taking behaviour over time and different environmental conditions Mean latency to exit the covered side of the GRT tank greatly varied over time, indicating that environment and/or age impact risk-taking behaviour. Although we are unable to identify causes of this variation, it supports that risk-taking behaviour is relatively plastic and can be affected by biotic or abiotic changes [25,40]. Nevertheless, within our studied population, individuals showed heterogeneous risk-taking behaviour which was moderately consistent over time and environmental conditions. This is particularly interesting given that repeatability was assessed over almost seven months and was challenged by varying the rearing conditions and stress loads. Consequently, risk-taking behaviour can be considered plastic over time and environmental conditions for the whole population but, at the same time, environmental conditions are not sufficient to trigger long lasting changes in individual risk-taking within a population. In this study, a GRT was performed two weeks after the end of the chronic stress protocol. The absence of differences in risk-taking behaviour between the chronically stressed individuals and controls suggest that the measured behaviour is resilient (i.e. significant change but quick recovery) or resistant (little change) to different stress loads. Stress is known to affect behaviour, including risktaking [41,42]. Behavioural measures are also generally considered as reliable markers of welfare [29,43,44]. Nevertheless, to our knowledge, no previous studies have investigated long-term post-stress differences in risk-taking behaviour in fishes; behavioural differences being observed only during the stress protocol or right after. Yet, some events occurring during fish life history impacted boldness [45,46], which was not the case in our study, suggesting that this measured behaviour, performed in group, is resilient toward stressors. 4 Proc. R. Soc. B 289: 20220399 (a) Consistency of group risk-taking behaviour over time and different environmental conditions pituitary, the biological process ‘Behaviour’ (GO:0007610) was, out of level 2 biological processes, the most differentially represented within the genes DE between bold and shy individuals (electronic supplementary material, table S1). Most of these behaviour-related genes were upregulated in bold individuals (figure 4a). All DE genes related to social behaviour, exploration behaviour and memory were upregulated in bold individuals (figure 4b–d). Only casp3, trpv1 and arrdc3 were found downregulated in shy individuals (figure 4a). In the head kidney, nine biological processes out of all level two biological processes were overrepresented among DE genes, with ‘immune system process’ being the most significant (GO:0002376). Most DE genes related to immune system processes were upregulated in bold individuals (figure 5a). All DE genes related to immune system development and leucocyte activation were found to be upregulated in bold individuals except for pawr (figure 5b,c). After adjustment of the p-values to lower false discovery rates, six biological processes were still significant in the head kidney, confirming the true biological signal underpinning transcriptomic differences (electronic supplementary material, table S1 and figure S2). royalsocietypublishing.org/journal/rspb after libraries were constructed, validated and quantified, they were sequenced in equimolar amounts using a HiSeq 2500 (Illumina, San Diego, CA, USA). Reads were then aligned to the Dicentrarchus labrax genome (NCBI, reference GCA_000689215.1) with a set of gene model annotations [39]. Differentially expressed (DE) genes between bold and shy were identified within each organ, while the effect of sex was accounted for. Gene ontology (GO) analysis was then performed to categorize each gene within a biological process. (a) GRT1 (255 dpf) GRT3 (423 dpf) GRT4 (463 dpf) –4.89 (HDI: (–6.31, –3.34)) –4.74 (HDI: (–5.65, –3.79)) –3.25 (HDI: (–4.71, –1.92)) latency (h) 30 20 10 0 bold GRT1 (255 dpf) bold shy bold shy bold shy GRT2 (373 dpf) GRT3 (423 dpf) GRT4 (463 dpf) –1.02 (HDI: (–0.45, 2.46)) –0.14 (HDI: (–1.04, 0.74)) 0.3 (HDI: (–1.1, 1.66)) 25 latency (h) 20 15 10 5 0.11 (HDI: (–0.56, 0.77)) 0 no (c) yes GRT1 (255 dpf) no yes no yes no yes GRT2 (373 dpf) GRT3 (423 dpf) GRT4 (463 dpf) –1.56 (HDI: (–3.58, 0.45)) –2.41 (HDI: (–3.8, –1.06)) –2.28 (HDI: (–4.42, –0.31)) 25 latency (h) 20 15 10 5 0.02 (HDI: (–0.76, 0.82)) 0 female male female male female male female male Figure 2. Latencies (in hours) at the four GRTs by at 255 dpf by (a) (estimates not shown for GRT1 because boldness was a direct function of latencies), (b) chronically stressed between GRT1 and GRT2, and (c) sex. Points show the posterior mean and vertical line segments represent the 95% HDI. Text annotations illustrate the posterior differences between each estimate and text in bold font highlights significant differences (i.e. the 95% HDI do not include zero). (Online version in colour.) We observed moderate consistency in the latency to leave the initial area over a relatively long period of time (ICC = 0.38), and for whether fish left the initial area at all (ICC = 0.35). Previous studies have also demonstrated consistency in fish behaviour overtime within a population [15,40,47,48], but only few exceeded 45 days [26,47,49]. Longterm consistency in personality traits have, nevertheless, been studied in multiple mammal species. Multiple of these studies observed changes once puberty or sexual maturation occurs [50,51]. To our knowledge, only one study investigated boldness consistency in fishes over maturation and sex change and observed a strong loss of consistency once sea bream (Sparus aurata, protandric species) became adults [47]. Our study investigated behaviour from 255 to 462 dpf, while sea bass experienced puberty (at least for males), suggesting that repeatable risk-taking behaviour occurs in European sea bass over important ontogenetic transitions. Altogether, our results advocate for resilient risk-taking behaviour over time and environmental conditions and further consolidate the bold-shy continuum as a personality trait in European sea bass. (b) Group behavioural response to novel environment and hypoxia Both bold and shy fish showed similar expected behavioural responses (i.e. increasing velocity and cohesion and reducing time in the periphery) during the acclimation period in the novel environment test [29,43]. Differences between bold and shy groups were observed at the very beginning of the acclimation period. We identified higher thigmotaxic behaviour for shy individuals compared to bold, classically indicative of anxiety-like behaviour [52]. Shy individuals also showed less cohesion than bold individuals at the beginning of the acclimation period. Based on the high cohesion values observed at the end of the acclimation period for both groups, we suspect that values at the start of the Proc. R. Soc. B 289: 20220399 (b) shy 5 royalsocietypublishing.org/journal/rspb GRT2 (373 dpf) (b) velocity (cm s–1) mean interindividual distance (cm) 8 p = 0.057 6 behaviour 4 bold shy 0 25 100 bold shy 35 30 * 25 125 (c) 0 25 50 75 time (min) 100 125 25 50 75 time (min) 100 125 (d) 200 behaviour oxygen saturation (%) time spent in the periphery (%) behaviour Proc. R. Soc. B 289: 20220399 50 75 time (min) 6 bold shy 75 50 25 150 100 50 * 0 25 50 75 time (min) 100 125 0 Figure 3. Group behaviour in a novel environment before and during a hypoxia challenge depending on fish behaviour in the GRTs (bold in orange, n = 48; shy in grey, n = 48). (a) Mean velocity (cm s−1), (b) inter-individual distance (cm) and (c) time spent in the periphery (%) are illustrated during a 60 min acclimation period to the novel environment and during a hypoxia challenge (starting at the red line) obtained by reducing the oxygen saturation down to 20% (d). Data are represented as mean ± s.e.m. The difference in behaviour between the two fish groups at a single time point is illustrated with an asterisk ( p < 0.05). (Online version in colour.) experiment are a marker of fear in shy groups. The tendency for higher activity in bold animals during the acclimation period could also be characteristic of fearless individuals [53,54]. Results from the two behavioural tests are therefore consistent. In response to the hypoxia challenge, both bold and shy fish decreased their swimming activity and enhanced the distance between individuals. These behaviours allow to, respectively, reduce energetic costs and maximize oxygen uptake [55]. These results suggest that both phenotypes respond similarly, in their behaviour, to a life-threatening challenge. (c) Transcriptomic signatures between bold and shy individuals Bold and shy individuals showed only limited differences in the expression of genes in the whole brain. This contrasts with a previous study highlighting 246 DEGs between behaviours in the whole brain [28], while we only observed 6. Nevertheless, in Rey et al. [28], the deployed risk-taking test was different and involved an hypoxia condition, forcing individuals to leave the sheltered area, and this was demonstrated to phenotype for a different behaviour in sea bass [15]. Gene expressions in the pituitary were the most significantly different between bold and shy individuals. The pituitary gland of bony fishes is composed of seven royalsocietypublishing.org/journal/rspb (a) endocrine-involved cell types [18] under the control of the hypothalamus in the brain. They release in the circulatory system hormones playing key roles in many biological processes, such as stress response, behaviour or growth [18]. Interestingly, multiple genes related to the ‘behaviour’ were significantly different between individuals categorized as bold or shy. Individuals categorized as bold overall showed an increased expression of genes related to social, exploration behaviours and memory. For instance, the gene dlg4 was upregulated in bold individuals. The knock out of this gene leads to reduced social behaviour and increased anxiety behaviour in mice [56]. This correlates well with the significant behaviour observed during the novel environment test, with bold animals showing increased grouping (lower inter-individual distance) and lower anxiety (lower time spent in the periphery) than shy individuals. Shy individuals had also reduced expression of jph3, encoding a protein of the junctophilin family, that was previously demonstrated to have an active role in exploratory behaviour in mice [57]. This higher exploration tendency for bold individual is consistent with behavioural responses both in GRT and novel environment test, and consistent with classic scheme of features underlying divergent coping style, i.e. proactive versus reactive individuals [10]. Genes related to memory and learning capacities were generally downregulated in shy individuals (e.g. egr1, npas4). In some cases, bolder and faster explorers also are fast learners (a) (b) behaviour 7 social behaviour no. copies 0 behaviour bold shy 1.0 (c) nr xn 2 np as 4 –1 nl gn 3 dl g4 0.1 gr id 1 exploration behaviour no. copies 100.0 behaviour bold shy 10.0 1.0 (d) slc 4a 10 dl g4 0.1 jp h3 1 10.0 memory 1000.0 no. copies 100.0 behaviour bold shy 10.0 1.0 cy p7 b1 eg r1 jp h3 np as se 4 rp in slc f1 8a sn 2 ap 25 th ad cy 8 ap p 0.1 gene Figure 4. Differences between bold and shy individuals in the expression of genes related to the biological process ‘behaviour’ in the pituitary of European sea bass. (a) Heatmap of all genes DE in the GO ‘behaviour’. (b–d ) Raw expressions on a log-scale for the two behavioural phenotypes of genes involved in social behaviour, exploration behaviour and memory (transparent triangles or rectangles for bold or shy individuals, respectively). Mean and standard error are also illustrated. (Online version in colour.) [58]. At the opposite, it can also be argued that shy individuals are more efficient in reverse learning and display more elaborated memory processes [59]. Links between cognition and animal personality, including boldness, could be influenced by many factors and are still debated [59,60]. In European sea bass, previous study observed increased expression of egr1, a gene involved in neurogenesis, of individuals categorized as shy in an individual novel environment test [25], while the opposite has been observed in the present study. Differences may be explained by the fact that the GRT and individual novel environment are measuring different behavioural responses, potentially due the context, isolation versus group testing [25]. These differences could also be due to difference in sampling time (after the experiment versus control) and thus differences in fish stress state, or due to the difference in brain parts sampled. This example illustrates that multiple sampling in various context are a key to understand the inter-individual variability. In the head kidney, many genes were also DE between the two groups. The head kidney is central in the stress axis regulation, by releasing cortisol and catecholamines into the blood. The analogue adrenal gland in mammals was previously demonstrated to show differences in the expression of genes related to glucocorticoid receptor signalling between two clear behavioural phenotypes in pigs [61]. While in our study a significant enrichment for ‘response to stimulus’ was observed, none of the DE genes were part of the GO ‘glucocorticoid receptor signalling pathway’ (GO:0042921). In teleosts, the head kidney is also involved in the immune system, by producing and maturing white blood cells and hosting antibody-producing cells [19,20]. The genes that are part of the ‘immune system processes’ were the most different between the two groups, with individuals categorized as bold showing overall increased gene expression. Although the link between behaviour and immune system is well documented in mammals, this link Proc. R. Soc. B 289: 20220399 th lypd1 slc4a10 nlgn3 snap25 slc8a2 cyp7b1 cntnap4 pum1 ctnnd2 astn1 app lamb1 sobp adcy8 serpinf1 apba1 npas4 jun nr1d2 mbd5 nrxn2 jph3 egr1 egr2 grid1 dlg4 strbp spg11 apba1 2 royalsocietypublishing.org/journal/rspb 100.0 casp3 trpv1 arrdc3 (a) immune system process 8 (b) irf4 2 3000 prkcb klhl6 1 2 no. copies pik3cd 1000 behaviour bold shy 300 tnrc6c no tc h2 pi k3 r1 tn rc 6c gene Ipin1 krf13 (c) leukocyte activation traf2 pik3r1 1000 tsc22d3 tagap no. copies ddit4 behaviour bold shy 100 dhx58 rpl30 pawr irf 4 itg av no tc h2 pa wr pi k3 cd pi k3 r1 pr kc b 10 trim25 gene Figure 5. Differences between bold and shy individuals in the expression of genes related to the biological process ‘immune system process’ in the head kidney of European sea bass. (a) Heatmap of all genes DE in the GO ‘immune system process’. (b,c) Raw expressions on a log-scale for the two behavioural phenotypes of genes involved in immune system development and leucocyte activation (transparent triangles or rectangles for bold or shy individuals, respectively). Mean and standard error are illustrated. (Online version in colour.) has been investigated in only a few fish studies reporting distinct immune capacities between behavioural phenotypes [53,62]. Whether the higher expression of immune genes observed for bold European sea bass in our study translates to increased immune functions, later affecting responses to pathogens, still needs to be demonstrated. Ethics. Experiments were authorized by ethics committee agreement APAFIS#10745 and all procedures involving animals were in accordance with the ethical standards of the institution and followed the recommendations of Directive 2010/63/EU. Data accessibility. All data and models used are available in the electronic supplementary material [66]. Raw sequencing data are available under the GEO accession no. GSE195636. Authors’ contributions. B.S.: conceptualization, data curation, formal 5. Conclusion Our study demonstrates that inter-individual differences in risk-taking behaviour are consistent over time and environmental conditions. Differences observed in the GRT also translated into differences in a novel environment test in smaller groups, where bold individuals adapted faster to a novel environment than shy fish. Finally, inter-individual differences also correlated with transcriptomic changes in the pituitary and head kidney, mostly through differences in behaviour- and immune-related genes. These results highlight the importance of investigating inter-individual differences in behaviour and physiology for the ecology of species as they probably explain dispersal and/or migration at the population scale [63,64] and are as such at the forefront of the response to global changes [65]. analysis, methodology, visualization, writing—original draft; S.A.: data curation, methodology, software and writing—review and editing; C.G.: data curation, methodology and writing—review and editing; M.P.: data curation and formal analysis; S.R.: data curation, formal analysis and writing—review and editing; B.G.: data curation, formal analysis, investigation and writing—review and editing; M.-L.B.: conceptualization, funding acquisition, project administration, supervision and writing—review and editing. All authors gave final approval for publication and agreed to be held accountable for the work performed therein. Conflict of interest declaration. We declare we have no competing interests. Funding. This work was funded by the ERANET COFASP project SUSHIFISH (ANR-15-COFA-0002-01) and ERANET ANIHWA WINFISH (grant no. ANR-14- ANWA-0008). MGX acknowledges financial support from France Genomique National infrastructure funded as part of ‘Investissement d’Avenir’ program managed by Agence Nationale pour la Recherche (contract ANR-10-INBS-09). Acknowledgements. We also thank Peter Biro for stimulating discussions regarding statistics. Finally, authors are also grateful to Emilie Levavasseur for the European sea bass drawing in figure 1. Proc. R. Soc. B 289: 20220399 notch2 kl f1 3 ets1 irf 4 100 –1 et s1 itgav royalsocietypublishing.org/journal/rspb immune system development References 2. 3. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. Ferrari S, Horri K, Allal F, Vergnet A, Benhaim D, Vandeputte M, Chatain B, Bégout ML. 2016 Heritability of boldness and hypoxia avoidance in European seabass, Dicentrarchus labrax. PLoS ONE 11, e0168506. (doi:10.1371/journal.pone.0168506) 16. Stamps J, Groothuis TGG. 2010 The development of animal personality: relevance, concepts and perspectives. Biol. Rev. 85, 301–325. (doi:10.1111/j. 1469-185X.2009.00103.x) 17. Butler AB. 2011 Functional morphology of the brains of ray-finned fishes. In Encyclopedia of fish physiology: from genome to environment (ed. AP Farrell), pp. 37–45. London, UK: Elsevier. 18. McMillan DB. 2011 The pituitary gland or hypophysis. In Encyclopedia of fish physiology: from genome to environment (ed. AP Farrell), pp. 1457–1465. London, UK: Elsevier. 19. Geven EJW, Klaren PHM. 2017 The teleost head kidney: integrating thyroid and immune signalling. Dev. Comp. Immunol. 66, 73–83. (doi:10.1016/j.dci. 2016.06.025) 20. Rauta PR, Nayak B, Das S. 2012 Immune system and immune responses in fish and their role in comparative immunity study: a model for higher organisms. Immunol. Lett. 148, 23–33. (doi:10. 1016/j.imlet.2012.08.003) 21. Vandeputte M, Gagnaire PA, Allal F. 2019 The European sea bass: a key marine fish model in the wild and in aquaculture. Anim. Genet. 50, 195–206. (doi:10.1111/age.12779) 22. Killen SS, Marras S, Ryan MR, Domenici P, McKenzie DJ. 2012 A relationship between metabolic rate and risk-taking behaviour is revealed during hypoxia in juvenile European sea bass. Funct. Ecol. 26, 134–143. (doi:10.1111/j.1365-2435.2011.01920.x) 23. Ferrari S, Rey S, Høglund E, Øverli Ø, Chatain B, MacKenzie S, Bégout ML. 2020 Physiological responses during acute stress recovery depend on stress coping style in European sea bass, Dicentrarchus labrax. Physiol. Behav. 216, 112801. (doi:10.1016/j.physbeh.2020.112801) 24. Ferrari S, Benhaïm D, Colchen T, Chatain B, Bégout ML. 2014 First links between self-feeding behaviour and personality traits in European seabass, Dicentrarchus labrax. Appl. Anim. Behav. Sci. 161, 131–141. (doi:10.1016/j.applanim.2014.09.019) 25. Alfonso S, Sadoul B, Gesto M, Joassard L, Chatain B, Geffroy B, Bégout ML. 2019 Coping styles in European sea bass: the link between boldness, stress response and neurogenesis. Physiol. Behav. 207, 76–85. (doi:10.1016/j.physbeh.2019. 04.020) 26. Millot S, Bégout ML, Chatain B. 2009 Risk-taking behaviour variation over time in sea bass Dicentrarchus labrax: effects of day-night alternation, fish phenotypic characteristics and selection for growth. J. Fish Biol. 75, 1733–1749. (doi:10.1111/j.1095-8649.2009.02425.x) 27. Sadoul B, Alfonso S, Bessa E, Bouchareb A, Blondeau-Bidet E, Clair P, Chatain B, Bégout ML, 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. Geffroy B. 2018 Enhanced brain expression of genes related to cell proliferation and neural differentiation is associated with cortisol receptor expression in fishes. Gen. Comp. Endocrinol. 267, 76–81. (doi:10.1016/j.ygcen.2018.06.001) Rey S, Jin X, Damsgård B, Bégout ML, Mackenzie S. 2021 Analysis across diverse fish species highlights no conserved transcriptome signature for proactive behaviour. BMC Genomics 22, 33. (doi:10.1186/ s12864-020-07317-z) Sadoul B, Alfonso S, Cousin X, Prunet P, Bégout ML, Leguen I. 2021 Global assessment of the response to chronic stress in European sea bass. Aquaculture 544, 737072. (doi:10.1016/j.aquaculture.2021. 737072) Carpenter B et al. 2017 Stan: a probabilistic programming language. J. Stat. Softw. 76(1), 1–32. (doi:10.18637/jss.v076.i01) Gabry J, Cešnovar R. 2020 CmdStanR: the R interface to CmdStan. See mc-stan.org/cmdstanr/ reference/cmdstanr-package.html. R Core Team. 2021 R: a language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. McElreath R. 2020 Statistical rethinking: a Bayesian course with examples in R and stan. Boca Raton, FL: CRC Press. Vehtari A, Gelman A, Gabry J. 2017 Practical Bayesian model evaluation using leave-one-out cross-validation and WAIC. Stat. Comput. 27, 1413–1432. (doi:10.1007/s11222-016-9696-4) Vehtari A, Gabry J, Magnusson M, Yao Y, Bürkner PC, Paananen T, Gelman A. 2020 loo: efficient leaveone-out cross-validation and WAIC for Bayesian models. See https://mc-stan.org/loo/ Nakagawa S, Schielzeth H. 2010 Repeatability for Gaussian and non-Gaussian data: a practical guide for biologists. Biol. Rev. Camb. Phil. Soc. 85, 935–956. (doi:10.1111/j.1469-185X.2010. 00141.x) Cleasby IR, Nakagawa S, Schielzeth H. 2015 Quantifying the predictability of behaviour: statistical approaches for the study of betweenindividual variation in the within-individual variance. Methods Ecol. Evol. 6, 27–37. (doi:10. 1111/2041-210X.12281) Hothorn T, Bretz F, Westfall P. 2008 Simultaneous inference in general parametric models. Biom. J. 50, 346–363. (doi:10.1002/bimj.200810425) Tine M et al. 2014 European sea bass genome and its variation provide insights into adaptation to euryhalinity and speciation. Nat. Commun. 5, 1–10. (doi:10.1038/ncomms6770) Alfonso S, Peyrafort M, Cousin X, Bégout ML. 2020 Zebrafish Danio rerio shows behavioural crosscontext consistency at larval and juvenile stages but no consistency between stages. J. Fish Biol. 96, 1411–1421. (doi:10.1111/jfb.14310) Biro PA, Beckmann C, Stamps JA. 2010 Small within-day increases in temperature affects boldness Proc. R. Soc. B 289: 20220399 4. Quinn JL, Cole EF, Bates J, Payne RW, Cresswell W. 2012 Personality predicts individual responsiveness to the risks of starvation and predation. Proc. R. Soc. B 279, 1919–1926. (doi:10.1098/rspb.2011.2227) Creel S. 2018 The control of risk hypothesis: reactive vs. proactive antipredator responses and stressmediated vs. food-mediated costs of response. Ecol. Lett. 21, 947–956. (doi:10.1111/ele.12975) Møller AP, Kwiecinski Z, Tryjanowski P. 2017 Prey reduce risk-taking and abundance in the proximity of predators. Curr. Zool. 63, 591–598. (doi:10.1093/ cz/zow114) Geffroy B, Sadoul B, Putman BJ, Berger-Tal O, Garamszegi LZ, Møller AP, Blumstein DT. 2020 Evolutionary dynamics in the Anthropocene: life history and intensity of human contact shape antipredator responses. PLoS Biol. 18, e3000818. (doi:10.1371/journal.pbio.3000818) Møller AP, Garamszegi LZ. 2012 Between individual variation in risk-taking behavior and its life history consequences. Behav. Ecol. 23, 843–853. (doi:10. 1093/beheco/ars040) Sih A, Bell A, Johnson JC. 2004 Behavioral syndromes: an ecological and evolutionary overview. Trends Ecol. Evol. 19, 372–378. (doi:10.1016/j.tree. 2004.04.009) Gosling SD, John OP. 1999 Personality dimensions in nonhuman animals: a cross-species review. Curr. Dir. Psychol. Sci. 8, 69–75. (doi:10.1111/1467-8721. 00017) Smith BR, Blumstein DT. 2008 Fitness consequences of personality: a meta-analysis. Behav. Ecol. 19, 448–455. (doi:10.1093/beheco/arm144) Biro PA, Stamps JA. 2008 Are animal personality traits linked to life-history productivity? Trends Ecol. Evol. 23, 361–368. (doi:10.1016/j.tree.2008. 04.003) Réale D, Garant D, Humphries MM, Bergeron P, Careau V, Montiglio PO. 2010 Personality and the emergence of the pace-of-life syndrome concept at the population level. Phil. Trans. R. Soc. B 365, 4051–4063. (doi:10.1098/rstb.2010.0208) Koolhaas JM, Korte SM, De Boer SF, Van Der Vegt BJ, Van Reenen CG, Hopster H, De Jong IC, Ruis MAW, Blokhuis HJ. 1999 Coping styles in animals: current status in behavior and stress-physiology. Neurosci. Biobehav. Rev. 23, 925–935. (doi:10.1016/ S0149-7634(99)00026-3) Bergmüller R, Taborsky M. 2010 Animal personality due to social niche specialisation. Trends Ecol. Evol. 25, 504–511. (doi:10.1016/j.tree.2010.06.012) Bierbach D, Laskowski KL, Wolf M. 2017 Behavioural individuality in clonal fish arises despite nearidentical rearing conditions. Nat. Commun. 8, 15361. (doi:10.1038/ncomms15361) Christensen KA, Brunelli JP, Wheeler PA, Thorgaard GH. 2014 Antipredator behavior QTL: differences in rainbow trout clonal lines derived from wild and hatchery populations. Behav. Genet. 44, 535–546. (doi:10.1007/s10519-014-9663-9) royalsocietypublishing.org/journal/rspb 1. 9 43. 44. 46. 47. 48. 49. 59. 60. 61. 62. 63. 64. 65. 66. Trans. R. Soc. B 373, 20170282. (doi:10.1098/rstb. 2017.0282) Sih A, Del Giudice M. 2012 Linking behavioural syndromes and cognition: a behavioural ecology perspective. Phil. Trans. R. Soc. B 367, 2762–2772. (doi:10.1098/rstb.2012.0216) Baker MR, Wong RY. 2019 Contextual fear learning and memory differ between stress coping styles in zebrafish. Sci. Rep. 9, 9935. (doi:10.1038/s41598019-46319-0) Gley K, Murani E, Trakooljul N, Zebunke M, Puppe B, Wimmers K, Ponsuksili S. 2019 Transcriptome profiles of hypothalamus and adrenal gland linked to haplotype related to coping behavior in pigs. Sci. Rep. 9, 13038. (doi:10.1038/s41598-01949521-2) Kirsten K, Fior D, Kreutz LC, Barcellos LJG. 2018 First description of behavior and immune system relationship in fish. Sci. Rep. 8, 846. (doi:10.1038/ s41598-018-19276-3) Cote J, Clobert J, Brodin T, Fogarty S, Sih A. 2010 Personality-dependent dispersal: characterization, ontogeny and consequences for spatially structured populations. Phil. Trans. R. Soc. B 365, 4065–4076. (doi:10.1098/rstb.2010.0176) Cote J, Fogarty S, Weinersmith K, Brodin T, Sih A. 2010 Personality traits and dispersal tendency in the invasive mosquitofish (Gambusia affinis). Proc. R. Soc. B 277, 1571–1579. (doi:10.1098/rspb.2009. 2128) Geffroy B, Alfonso S, Sadoul B, Blumstein DT. 2020 A world for reactive phenotypes. Front. Conserv. Sci. 1, 5. (doi:10.3389/fcosc.2020.611919) Sadoul B, Alfonso S, Goold C, Pratlong M, Rialle S, Geffroy B, Bégout ML. 2022 Transcriptomic profiles of consistent risk-taking behaviour across time and contexts in European sea bass. FigShare. (https://doi. org/10.6084/m9.figshare.c.5958649) 10 Proc. R. Soc. B 289: 20220399 45. 50. Kanda LL, Louon L, Straley K. 2012 Stability in activity and boldness across time and context in captive Siberian dwarf hamsters. Ethology 118, 518–533. (doi:10.1111/j.1439-0310.2012.02038.x) 51. Neave HW, Costa JHC, Weary DM, von Keyserlingk MAG. 2020 Long-term consistency of personality traits of cattle. R. Soc. Open Sci. 7, 191849. (doi:10. 1098/rsos.191849) 52. Maximino C, de Brito TM, da Silva Batista AW, Herculano AM, Morato S, Gouveia A. 2010 Measuring anxiety in zebrafish: a critical review. Behav. Brain Res. 214, 157–171. (doi:10.1016/j.bbr. 2010.05.031) 53. Castanheira MF et al. 2017 Coping styles in farmed fish: consequences for aquaculture. Rev. Aquac. 9, 23–41. (doi:10.1111/raq.12100) 54. Sneddon LU. 2003 The bold and the shy: individual differences in rainbow trout. J. Fish Biol. 62, 971–975. (doi:10.1046/j.1095-8649.2003.00084.x) 55. Gesto M, Zupa W, Alfonso S, Spedicato MT, Lembo G, Carbonara P. 2020 Using acoustic telemetry to assess behavioral responses to acute hypoxia and ammonia exposure in farmed rainbow trout of different competitive ability. Appl. Anim. Behav. Sci. 230, 105084. (doi:10.1016/j.applanim.2020.105084) 56. Feyder M et al. 2010 Association of mouse Dlg4 (PSD-95) gene deletion and human DLG4 gene variation with phenotypes relevant to autism spectrum disorders and Williams’ syndrome. Am. J. Psychiatry 167, 1508–1517. (doi:10.1176/ appi.ajp.2010.10040484) 57. Moriguchi S et al. 2006 Functional uncoupling between Ca2+ release and afterhyperpolarization in mutant hippocampal neurons lacking junctophilins. Proc. Natl Acad. Sci. USA 103, 10 811–10 816. (doi:10.1073/pnas.0509863103) 58. Dougherty LR, Guillette LM. 2018 Linking personality and cognition: a meta-analysis. Phil. royalsocietypublishing.org/journal/rspb 42. and alters personality in coral reef fish. Proc. R. Soc. B 277, 71–77. (doi:10.1098/rspb.2009.1346) Culumber ZW. 2020 Thermal stress increases activity and risk-taking behavior but not anxiety in a livebearing fish. Environ. Biol. Fishes 103, 313–317. (doi:10.1007/s10641-020-00966-9) Alfonso S, Sadoul B, Cousin X, Bégout ML. 2020 Spatial distribution and activity patterns as welfare indicators in response to water quality changes in European sea bass, Dicentrarchus labrax. Appl. Anim. Behav. Sci. 226, 104974. (doi:10.1016/j.applanim.2020.104974) Martins CIM et al. 2012 Behavioural indicators of welfare in farmed fish. Fish Physiol. Biochem. 38, 17–41. (doi:10.1007/s10695-011-9518-8) Chapman BB, Morrell LJ, Krause J. 2010 Unpredictability in food supply during early life influences boldness in fish. Behav. Ecol. 21, 501–506. (doi:10.1093/beheco/arq003) Ferrari S, Leguay D, Vergnet A, Vidal MO, Chatain B, Bégout ML. 2016 Unpredictability in food supply during early life influences growth and boldness in European seabass, Dicentrarchus labrax. Appl. Anim. Behav. Sci. 180, 147–156. (doi:10.1016/j.applanim.2016.04.017) Castanheira MF, Martínez Páramo S, Figueiredo F, Cerqueira M, Millot S, Oliveira CCV, Martins CIM, Conceição LEC. 2016 Are coping styles consistent in the teleost fish Sparus aurata through sexual maturation and sex reversal? Fish Physiol. Biochem. 42, 1441–1452. (doi:10.1007/s10695-016-0231-5) Castanheira MF, Herrera M, Costas B, Conceição LEC, Martins CIM. 2013 Can we predict personality in fish? Searching for consistency over time and across contexts. PLoS ONE 8, e62037. (doi:10.1371/journal. pone.0062037) Geffroy B, Sadoul B, Bardonnet A. 2015 Behavioural syndrome in juvenile eels and its ecological implications. Behaviour 152, 147–166. (doi:10. 1163/1568539X-00003236)