Altered photoperiod affects behaviour in a key fish grazer, the thinlip grey mullet, Chelon ramada

Abstract

The aim of this study is to address experimentally the immediate impacts of an alteration of the photoperiod, simulating an extreme artificial light pollution, on the behaviour and stress of a key grazer fish common in temperate coastal ecosystems in Europe, the thinlip grey mullet, Chelon ramada. Fish behaviour, assessed by total distance travelled, swimming speed, distance between individuals in fish groups and angular velocity analysed over 6 periods of 5‐min slots during 24 h, was shown to vary depending on the photoperiod conditions and time. During the 5‐min sequences analysed during the day, fish in the control photoperiod condition (alternate photoperiod, AP) swam slower, travelled less distance and were closer to their conspecifics than fish in the altered conditions [direct or indirect continuous photoperiod (iCP), dCP and iCP, respectively]. For the 5‐min sequences analysed at night, fish exposed to direct continuous lighting (dCP) swam significantly faster, hence travelled a longer distance and further apart than fish exposed to iCP or alternate normal photoperiod (AP). Angular velocity was similar for all conditions. Exploratory cortisol measures conducted in the water revealed lower quantities in the iCP condition than in the AP and dCP conditions. Although based upon an experimental design exaggerating the natural photoperiod alteration as compared with most real‐life situations, results obtained highlight the importance of addressing behavioural responses of organisms at the individual level to tackle the global and specific impacts of diurnal cycle alteration by artificial light in urbanised aquatic ecosystems. This study is also the first to report on the behavioural response of a common wild species playing a key role in the functioning of aquatic ecosystems to a widespread stress in coastal environments.

1. INTRODUCTION

Artificial light at night (ALAN) affects more than 80% of the planet, with over 99% of the North American and European human populations living under a light‐polluted sky (Falchi et al., 2016; Liu et al., 2024). ALAN also affects biodiversity, including aquatic species (Gaston et al., 2015). Yet, in 2022, from a total of 524 scientific publications dealing with the effect of light pollution on organisms or ecosystems, only 20% explicitly addressed aquatic species (compared with 59% addressing terrestrial organisms and 21% of general approaches; Bassi et al., 2022). Meanwhile, a growing body of scientific studies reported diverse effects of this form of physical pollution on various attributes of aquatic communities, including spatial and temporal distributions of organisms, species reproduction, individual mobility and behaviour, or trophic interactions (Ganguly & Candolin, 2023; He et al., 2024; Hölker et al., 2023). Considering ALAN can penetrate water bodies up to 20 m depth depending on water clarity (Smyth et al., 2021), this form of stress is thus likely to affect many features of aquatic ecosystem functioning. Such effects are particularly important in urbanised regions, where aquatic ecosystems may be subjected to sky glow or intense (although often indirect) illumination at night, especially in estuaries (Zapata et al., 2019) and coastal waters (Garratt et al., 2019; Maggi et al., 2019). Despite increasing efforts from public stakeholders to limit the impact of ALAN on some urban areas, light‐polluted skies remain strikingly visible and dominate the nocturnal coastal environments throughout the planet (Falchi et al., 2016; Liu et al., 2024).

One of the major impacts of ALAN on biodiversity is its interaction with natural photoperiod (day length) and associated diurnal cycles, partly modifying organisms' biology and life history (Bradshaw & Holzapfel, 2007). Globally, ALAN has been shown to disrupt or mask natural signals regulating nyctemeral cycles (Tidau et al., 2021), leading to the alteration of diel or seasonal patterns in species distributions (Riley et al., 2012; Vowles & Kemp, 2021), individual activity patterns and energy expenses (Pulgar et al., 2019; Quintanilla‐Ahumada et al., 2024), reproduction (Brüning et al., 2011), risk taking related to predation (Perkin, 2013; Tałanda et al., 2018), or the rhythmicity of feeding activities (Lynn et al., 2024; Xu et al., 2023). Most studies dealing with the impacts of photoperiod alteration by ALAN in aquatic organisms rely upon experimental approaches and have targeted various biological models including periphyton (Roux et al., 2024), invertebrates (gastropods, Mondy et al., 2021; zooplankton, Tałanda et al., 2018; decapods, Steell et al., 2020; amphipods, He et al., 2024; echinoderms, Xu et al., 2023; bivalves, Botté et al., 2023), and some vertebrates including marine mammals (Stanton & Cowart, 2024), reptiles (Taylor et al., 2022) and fishes. For the latter group, which represents the highest diversity of Vertebrates on Earth, ALAN impacts were studied on some freshwater species (common perch Perca fluviatilis L., Franke et al., 2013; roach Rutilus rutilus (L.), Brüning et al., 2015; common bleak Alburnus alburnus (L.), Brüning, Hölker, et al., 2018; Brüning, Kloas, et al., 2018; chub Squalius cephalus (L.), Brüning et al., 2011; smallmouth bass Micropterus dolomieu Lacepède 1802, Foster et al., 2016; common rudd Scardinius erythrophthalmus (L.), Tałanda et al., 2018; mosquitofish Gambusia affinis (Baird & Girard 1853), Miner et al., 2021; guppy Poecilia reticulata (Peters 1859), Kurvers et al., 2018) but also on a few migratory diadromous species (Atlantic salmon Salmo salar L., Riley et al., 2012; European eel Anguilla anguilla (L.), Vowles & Kemp, 2021) and some marine species (rockfish Girella laevifrons (Tschudi 1846), Pulgar et al., 2019; clownfish Amphiprion ocellaris Cuvier 1830, Fobert et al., 2019; bonefish Albula vulpes (L.), Szekeres et al., 2017; surgeonfish Acanthurus triostegus (L.), O'Connor et al., 2019).

The thinlip grey mullet, Chelon ramada (Risso 1827), is one of the most common species in West Atlantic coasts (Nelson et al., 2016). Although poorly studied, its conservation status is evaluated as ‘least concerned’ by the IUCN, but recent data suggest an increasing fishing effort on the species in Europe (Dindart et al., 2024). It occurs in a wide range of habitats, including coastal, brackish and freshwater environments within estuarine systems (Laffaille et al., 2002; Lebreton et al., 2011), often in urbanised conditions (Velasque et al., 2023), and its distribution area is expected to expand with climate change (Lassalle & Rochard, 2009). Additionally, C. ramada has a varying diet over its life cycle, with juveniles being primary consumers, that is, grazing microphytobenthos (or more generally biofilms) on mudflats or other intertidal habitats (Lebreton et al., 2011, 2013), while adults are considered limnivorous (Carassou, Vildier, et al., 2025; Laffaille et al., 2002). The species therefore has an important contribution to linkages between trophic levels and between benthic and pelagic components in aquatic ecosystems. The migratory behaviour of C. ramada, whose catadromous life cycle is supported by pronounced osmoregulatory capacities, also emphasises the role of this fish species as an important vector of energy transfers between adjacent marine, estuarine and freshwater environments, on which important ecosystem services rely upon (Lassalle & Rochard, 2009; Truchy et al., 2025; Whitfield, 2016). C. ramada is also considered a pertinent indicator species in pollution biomonitoring programs (Chine et al., 2021; Zaoui et al., 2023). Finally, it is also known to be sensible to light/dark alternations (Trancart et al., 2011) and to detect colours (El‐Sayed & El‐Ghobashy, 2011). Based on its ecological characteristics and role in ecosystem functioning, C. ramada thus represents an interesting biological model to address fish responses to light pollution.

In order to address the impacts of physical, chemical or any other stress factor on Vertebrates, behavioural responses are often earlier indicators of altered conditions (Wong & Candolin, 2015). Behavioural studies were indeed shown to be sensitive, powerful and ecologically meaningful approaches to assess, for instance, ecotoxicological effects (Bertram et al., 2022), linked to survival, growth and reproduction (Pyle & Ford, 2017; Robinson, 2009). The main behavioural responses studied in fish ethology are general activity, thigmotaxis and group cohesion (Alfonso et al., 2020; Sadoul et al., 2017). More precisely, variations in fish swimming behaviour were shown to efficiently reveal important responses to ecotoxicological stress, in particular fish body orientation and movement, including posture, frequency and duration and speed of movements (Porras‐Rivera et al., 2024). Swimming behaviour can also efficiently demonstrate responses of aquatic vertebrates to toxic stress by pesticides (Shuman‐Goodier & Propper, 2016) or fungicides (Andrade et al., 2016). In addition to the aforementioned variables, angular velocity has been considered an indicator of swimming complexity for the sea bass (Dicentrarchus labrax (L.), Benhaïm et al., 2012, 2013) or juvenile empire gudgeons [Hypseleotris compressa (Krefft 1864), Melvin, 2016]. In aquaculture, angular velocity is also being used to control the well‐being of captive animals, as angular velocity relates in nature to fish ability to accelerate and display fast manoeuvres essential to their survival, especially in terms of predator encounters (Alfonso et al., 2020; An et al., 2021; Paniccia et al., 2022). For example, the effect of an increased water temperature was shown to decrease fish angular velocity in the delta smelt (Hypomesus transpacificus McAllister 1963, Davis et al., 2019). Swimming speed and angular velocity were also successfully combined to address the behavioural response of juvenile H. compressa to water treatment modalities in aquaculture (Melvin, 2016).

The objective of this study is to compare the short‐term behavioural responses of juvenile C. ramada exposed to an altered or a normal photoperiod under experimental conditions. The behavioural metrics analysed include total distance travelled and velocity, distance between individuals and angular velocity, relating to fish mobility, sociality and stress response, respectively. Additionally, some exploratory measures of cortisol were performed in the water and fish plasma samples in an attempt to test for other physiological responses of the fish to the aforementioned exposure conditions. As cortisol basal values remain unknown for the model wild species C. ramada, those exploratory measures mostly aimed at experiencing a protocol allowing for stress assessment in this species for which data on cortisol expression are lacking. The photoperiod conditions were designed to represent a worst‐case scenario of altered urban conditions, in which ALAN induces continuous illumination of aquatic environments with daytime‐like intensity, both day and night. These stressful conditions were compared with typical ‘natural’ conditions, which are unaffected by urban light pollution and characterised by high‐intensity illumination of aquatic environments during the daytime only. Based on knowledge on the model species ecology and on the available literature described above, our main hypotheses are that (i) C. ramada displays different swimming patterns depending on photoperiod exposure conditions, especially under continuous direct light exposure; (ii) cortisol levels measured in the water increase over time, with a higher concentration at night under a continuous photoperiod (CP) with direct and indirect light exposures; and (iii) cortisol levels in the plasma are correlated with cortisol levels in the water at the end of the exposure period.

2. METHODS

2.1. Ethics statement

The care and use of experimental animals complied with guidelines and policies from the Comité National de Réflexion Ethique sur l'Expérimentation Animale under the authority of the French Ministry of Higher Education and Research, under the agreement number APAFIS#38009‐2022102818104872 v2. Experiments were performed at the INRAE experimentation station in Saint‐Seurin‐sur‐l'Isle (under structure agreement for experiments on animals B33‐478‐001). The necessary sampling authorisations were obtained from the Direction Interrégionale de la Mer Sud‐Atlantique, unité Réglementation des Pêches Poitou/Charente/Aquitaine, France.

2.2. Fish collection and acclimatisation

Juvenile fish were captured on 5 April 2023 in a nearby estuarine natural environment from the vessel L'Esturial using surface channel nets of section 4.0 × 1.0 m, at low‐speed (1–1.5 knots). Salinity at the time of capture ranged from 8.2 to 11.4, with surface temperatures averaging 13.3°C ± 0.5°C. A total of 100 fish were transported in plastic bags pre‐filled with estuarine water salted to 3 salinity units above the ambient environment (to reduce stress; Tacchi et al., 2015; Favero et al., 2019; Luz & Favero, 2024) and inflated with pure oxygen. Temperature and salinity were measured in the ambient water and in the bags using a portable conductimeter (FDO 700IG, TETRACON 700IG, WTW), then upon arrival at the experimental site where the fish were placed in quarantine for sanitary observations. A Vibrio anguillarum infection occurred from April 26 to April 28, 2023. Antibiotic treatment thus had to be carried out with the use of OxyTetraCycline 75% by 30 min of balneation at 0.15 g L⁻¹, the treatment being conducted twice at 48 h interval (as per formal veterinary prescription). Mortality from this disease reached 45% of the stock. The remaining fish, the number of which (N = 45) was sufficient to conduct the experiment, were then acclimatised for experimental conditions (visual and sound context, water medium from a mixture of the adjacent river water and seawater from a dedicated stock, feeding conditions) over 1 month.

As C. ramada juveniles are known to abundantly occur in brackish and freshwater reaches of estuaries (where they were heading upon their capture in the oligohaline section of the Gironde estuary), the experiment was conducted in freshwater, in order to facilitate water quality management over the course of the experiment, and knowing C. ramada can tolerate large salinity variations (Dindart et al., 2024; Rochard et al., 2025). The fish were thus conditioned to river water over slow and progressive thresholds during a 3‐week acclimatisation period (5 April 2023 to 23 May 2023). Salinity decreases were limited to thresholds of 5 units per week. Meanwhile, the fish were fed with mixtures prepared at the laboratory (including commercial pellets and sticks, frozen mussels, chironomid larvae). The water temperature, dissolved oxygen and conductivity were monitored continuously with automatic probes throughout the whole acclimatisation period. Light conditions varied between the quarantine and acclimatisation periods to ensure maximum comfort for the animals, starting with obscure conditions at fishing retrieval and progressively achieving light conditions adjusted to future experimental conditions (see below). A ‘normal’ photoperiod, including alternate illumination and extinction periods as expected in an environment unaffected by ALAN, was maintained throughout the whole acclimatisation process, until fish were transferred to experimental tanks. Extinction and illumination period durations were parameterised over a 24 h‐cycle as a function of the ephemeris at the time of the acclimatisation procedure.

2.3. Experimental design

The fish were exposed to three light conditions during 72 h (3 consecutive day/night periods). Day and night durations were fitted to the natural ephemeris of the experimental period for the conditions of alternate photoperiod (AP). AP conditions indeed consisted in 14:30 h of day and 9:30 h of night (light on from 06:00 to 21:30; off from 21:30 to 06:00), with 30‐min gradual night‐day transitions, while two other fish groups were exposed to a CP designed to induce an intense continuous day‐like light stress. The later condition included two contrasted orientations of the light source: direct (dCP: light projected in the direction of the water surface of the aquarium) and indirect (iCP: light projected towards the ceiling in the opposite direction of the water surface of the aquarium at a similar distance; Figure 1). The dCP modality was designed to mimic a very drastic nocturnal lighting, sometimes, although rarely, occurring in the field close to, for example, sports, transport or industrial infrastructures in heavily urbanised areas. The iCP modality intends at illustrating a more usual situation where lamps are more gently reaching the aquatic environment although their lighting adds up to the ambient skyglow in the area.

FIGURE 1.

(a) Experimental design used to address behavioural and stress responses of C. ramada juveniles to photoperiod modification by intense light pollution. The design relies upon exposure to a natural, alternate, photoperiod (day‐night: AP) or a modified, continuous, photoperiod (day‐day: CP) with direct (dCP) or indirect (iCP) light exposures. (b) Workflow of the processes of dataset selection for cortisol and behavioural main results. On the left, the process for analysis of full results presented in the supplements; on the right, process for analysis of the curated behavioural and selected cortisol data presented in the main results.

The lamps used were 1047 mm long JBL LED SOLAR NATUR 57W, with a module for LED SOLAR CONTROL. The distance between the water surface and lamps, and dimensions of the aquaria (length, width, depth), were identical between the three exposure conditions. The illumination conditions were set up to 4000 K during the day for AP and continuously for CP (colour temperature chosen of 4000 K corresponding to neutral white, equivalent to daylight). Each aquarium (three in total) included five fish (n = 15) randomly assigned to each modality from the acclimatised stock on the day before the start of the experiment. Light intensity was kept constant throughout the experiment, which is a drastic situation as compared with most real‐life situations (where intensity produced by urban lighting at night is largely lower than the light intensity produced by solar radiations during the day; Gaston et al., 2013). However, this helped ensure that measured effects would result from a photoperiod alteration, responding to our objective, and not from light intensity variations during the relatively short exposure periods considered.

The experiment was repeated three times, on three different fish lots from the same 45 fish stock (23–26 May 2023, 30 May to 2 June 2023, 6–9 June 2023), during three successive weeks with three photoperiod conditions identical in all trials. In between experiments, the unexposed fish were maintained in their acclimatisation tank under a natural AP similar to that in AP during the experiments. Each week, 15 fish were thus introduced into three separate aquariums (five fish in each experimental condition: AP, dCP and iCP). The tanks consisted of 30 L aquaria associated with individual 30 L reserves of water allowing for water re‐circulation and sampling (Figure 1). A cardboard opaque container avoiding any visual disturbance to the fish during the experiments surrounded the aquaria. Automatic probes allowed for continuous measurements of temperature, dissolved oxygen and conductivity. The nitrogen products (NH₄, NO₂) were monitored (using ammonia tank tests LCK 304 HACH®; nitrite tank tests LCK341 HACH®, with the HACH® lange DR3900 spectrophotometer) every 6 h during the 72 h of each experimental trial, as well as before fish transfers (Table 1).

TABLE 1.

Summary of average exposure conditions (values are mean ± standard errors) over the experiment (second trial; see Table S1 for values observed for the other experimental trials).

Variables	N	Conditions	Values	QL
Temperature (°C)	13	AP	22.1 ± 0.1a
	13	dCP	20.9 ± 0.1b
	13	iCP	21.7 ± 0.1c
Dissolved oxygen (mg L−1)	13	AP	8.3 ± 0.0a
	13	dCP	8.5 ± 0.0b
	13	iCP	8.2 ± 0.0c
Ammonium concentration (mg L−1)	39	All	0.1 ± 0.0	0.015
Nitrite concentration (mg L−1)	39	All	0.1 ± 0.0	0.015
Light intensity in the water (μmol m−2 s−1)	20	AP and dCP	36.8 ± 3.1a
	10	iCP	0.8 ± 0.0b

Note: Different superscript letters indicate statistical differences between individual conditions (‘All’) or specific effects of condition exposure (AP, dCP, iCP) as per Dunn's post‐hoc tests.

Abbreviation: QL, quantification limit.

Individual cameras (BASLER) were set up above each aquarium with COMPUTAR VARI FOCAL objectives and an Infrared light filter (Infrarot 850). Infrared floors (in‐house construction with BLANKO IR lights) were installed underneath each aquarium to continuously record the activities of the animals throughout the experiment during both day‐ and night‐times (the use of infrared also allowing to avoid issues related to the reflection of light on the water surface which could impede the acquisition of behavioural observations). The fish behaviour was recorded using tracking software (Ethovision XT Noldus, The Netherlands).

At the end of the experiment, a quantum sensor LI‐193 with light meter LI‐250A (Biosciences GmbH) was used to measure light intensity (in μmol m⁻² s⁻¹) in the water at different longitudinal and vertical positions within the tanks in order to characterise actual light exposure conditions for each modality (Table 1). All the fish were genotyped to verify the species as the Mugilidae species can sometimes be confused at the juvenile stage. At the end of the experiments, each fish was genotyped from a clip from the pectoral fin. All individuals were confirmed as C. ramada (analysis entrusted to GenoScreen, 1 rue du Prof. Calmette, 59,000 Lille, France; sequences reading by J‐D Durand, UMR MARBEC, IRD, Montpellier, France). The fish size ranged from 12.0 to 15.3 cm total length (with an average size of 13.7 ± 0.2 cm) for a weight range of 13.3–31.1 g (with an average weight of 20.1 ± 1.2 g), which confirmed their juvenile stage (C. ramada having a length at maturity averaging 25.9 cm, with a range of 25–32 cm, https://fishbase.se). There was no significant difference in fish size between exposure conditions and among the three trials (H = 2.23 for condition, H = 1.19 for among trial, p‐value >0.05).

2.4. Behavioural data acquisition and sampling

We sampled six periods for video interpretation within the 72 h of each experiment. The selected sequences were positioned along the 72 h course of the experiment so as to obtain behaviour observations during successive day and night sequences as follows: 06:30–07:30 (day), 12:00–13:00 (day), 17:00–18:00 (day), 21:30–22:30 (night), 23:00–00:00 (night) and 03:00–04:00 (night), the 5‐min sequences selected being positioned in the middle of the time periods (see Figure 1b). The videos were acquired at 25 frames per second using Multiple Arena and Social Interaction Modules. The 'subject size’ (dimensional parameter of the object being tracked, set up in pixels) was kept constant throughout all data acquisition periods, with only tracked objects and background colours and contrasts varying to adjust to specific exposure conditions and optimise fish detection (see Table S1).

To address the quality of positioning data extracted from the videos, we used the occurrence of undetected subjects. The index software ‘subject not found’ provided by Ethovision XT (Noldus) was expressed in percentage (%). The value of this index corresponds to the proportion of time during which the fish could not be correctly detected and monitored by Ethovision XT Noldus within a given sequence. It averaged 19% for AP and dCP conditions, and 12% for iCP condition for the whole dataset, with notable differences depending on exposure conditions (due to targeted contrasted light conditions and variations in fish activities—see Table S2). Despite rigorous parameterisation of the cameras and the acquisition system, resulting errors were thus high during the tracking process, due to loss of tracked individuals (happening when individuals do not move at all or very little, or on the contrary when they move erratically, but also when several individuals cross each other, or when reflection of fish images on the aquarium glass walls impedes their detection by the software). We accordingly manually corrected the individual trajectories of fish for a sub‐sample of the recordings, in order to obtain a curated dataset, the results from which were confronted to results obtained from similar analyses of the whole, uncurated, dataset. The total duration of the analysed curated data set (6 × 5 min = 30 min) is comparable to other previously studies addressing fish behaviour under experimental conditions (Alfonso et al., 2020; Handy et al., 1999). The main results of the study presented herein therefore rely upon the analysis of the curated sub‐set of behavioural data, while results obtained from similar analysis of the full, uncurated dataset (and including aforementioned detection biases), are presented in the Supporting Information and discussed. The curated dataset covers a total of six 5‐min video tracks (in the middle of the time periods) from the second experimental trial (30 May to 2 June 2023), and includes three daytime and three night‐time tracks for each of the three exposure modalities. The behavioural variables analysed include the total distance travelled over the sequence (in cm), mean velocity (in cm s⁻¹), mean distance between fish (in cm) and angular velocity (in deg. s⁻¹). For each variable, the results were differentiated between ‘day’ and ‘night’ periods as defined by the AP condition. The mean distance between fish provided by the software is measured at the scale of individuals (e.g., subject 1 is x cm apart from subject 2), and was thus averaged among individuals (as a proxy of shoaling) within a given condition.

2.5. Exploratory protocol for stress measurement using cortisol concentrations

Blood samples were collected from every exposed fish at the end of the experiment using a syringe and after anaesthesia with benzocaine (30 mg L⁻¹). The blood was recovered by puncture in the caudal vein. Blood was immediately placed in Eppendorf tubes in contact with 20 μL of heparin at 1667 IU mL⁻¹, centrifuged at 3500 rpm for 10 min to separate plasma from red cells, leucocytes and platelets (Buffy Coat), and stored at −20°C until further processing. Cortisol ELISA 402710 NEOGEN kits, a Multiskan FC microplate photometer with incubator (Thermofisher®) and a Skanlt Reader Software (Thermofisher®) were then used to estimate cortisol quantities in the samples, each sample being analysed using duplicate Elisa plates according to the manufacturer's instructions, comprising a blank and a standard. Each plate used was produced with its own standard curve, based on the standards supplied in the kit. Variation coefficients were calculated following duplicates per plate, with an average of 6.5% (always below 40%). The data were expressed in ng of cortisol per mL of plasma per g for each fish. The cortisol plasma results presented in the following sections rely upon a sub‐set of 14 samples from the 15 fish successfully sampled during the second experimental trial (selected for video data interpretation, see earlier section) for clarity. Data from all blood samples (including fish from the three successive experimental trials) are presented in Supporting Information (Figure S3) and discussed.

Cortisol was also measured in the water of the experimental systems over the course of the experiment, following protocols developed by Sadoul and Geffroy (2019). Briefly, 2 L of water were sampled in the 30 L reserves associated to each aquarium every 6 h over the course of the exposure periods (for 117 samples in total for the three successive trials, including 39 samples for the second trial). The water was first filtered on Whatman® GF/F filters (Glass Microfiber filters CatNo.1825‐047) with a porosity of 0.45 μm and a diameter of 47 mm, before being gently (25 mL min⁻¹ maximum) sieved through SEP‐PAK cartridges (C18 Plus Short 360 mg Sorbent/55–105‐μm particle size, reversed‐Phase Ph 2–8 × 50 WATERS®). The cartridges were activated with 10 mL of absolute methanol and 10 mL of MilliQ water beforehand, and placed in the refrigerator before being sent to the analytical laboratory (SELARL Vet'eau, 1289 rue des Pyrénées, 31330 Grenade, France). Four 96‐well plates (ELISA Cortisol Saliva from TECAN, France) were used for the three successive trials, from which four wells were blanks, and 6 × 2 were standards that ranged from 0 to 30 ng mL⁻¹. Each plate used was produced with its own standard curve, based on the standards supplied in the kit. Variation coefficients were calculated following duplicates per plate, with an average of 4.4% (always below 119%). The cortisol concentrations were then related to fish biomass, thus expressed in ng of cortisol per mL of water and per g of fish (this allowed to take into account possible variations in fish size and weights in our samples). As for plasma cortisol, the cortisol concentrations in the water presented in the following Results section correspond to samples obtained during the second experimental trial (selected for behavioural data analysis), for clarity. The whole dataset on cortisol concentrations in water (and for the three successive experimental trials) is presented in Supporting Information (Figure S2), and discussed.

2.6. Statistical analyses

Statistical analyses were performed using R (4.4.0) and R studio (2023.12.1 Build 402). First, all data were presented as average values ± standard errors (with package plotrix). Given the low sample size and resulting non‐normality and heteroscedasticity of the data, non‐parametric statistical methods were used. Kruskall‐Wallis tests were used to compare the experimental conditions (temperature, dissolved oxygen, light intensity, nitrogen products), behavioural variables (total distance travelled, velocity, distance between fish and angular velocity) and cortisol concentrations (in blood and water) between exposure conditions (AP, dCP, iCP, i.e., 2 degrees of freedom) with differences considered significant if p‐value <0.05. When significant differences were detected, Dunn post‐hoc tests were used to test differences between conditions by pairs (AP vs. dCP, AP vs. iCP, dCP vs. iCP) with differences considered significant if p‐value <0.05. Wilcoxon‐Mann–Whitney tests were performed to address differences between exposure periods (day vs. night) in water cortisol concentrations and behavioural variables (differences were considered significant when p‐value <0.05).

3. RESULTS

3.1. Experimental conditions

The temperature and dissolved oxygen varied between conditions, with the lowest values observed in dCP and highest in AP during the second experimental trial (Table 1). The difference was, however, less than 1.2°C between conditions. Ammonium and nitrite levels were identical in the three conditions over the experiment and remained low. The light intensity measured in the water also varied between conditions, with the highest intensities in AP and dCP, and lowest in iCP (Table 1).

3.2. Fish behaviour

3.2.1. Total distance travelled

The total distance travelled by the fish was similar between day and night periods within each condition throughout the experiment (H = 1019, p‐value > 0.05; Figure 2). The fish in dCP were, however, significantly more active (swam more distance; H = 16.9, p‐value < 0.05) over the three 5‐min sequences analysed than in AP and iCP. During the day, the fish travelled a lower total distance in AP than in the other conditions over the three 5‐min sequences analysed (H = 13.5, p‐value < 0.05; Figure 2), with an average distance covered of 1425 ± 205 cm in AP, 2147 ± 90 cm in iCP and dCP. At night, the distance travelled by the fish also varied between conditions (H = 9.72, p‐value < 0.05; Figure 2) and was significantly greater in dCP (average of 2488 ± 190 cm) than in AP and iCP (1678 ± 124 cm).

FIGURE 2.

Total distance travelled (in cm) by C. ramada juveniles over three 5‐min periods during the day (left) and three 5‐min periods during the night (right) throughout the experiment (see Methods for details on the selection procedure of analysed sequences; N = 90). AP, alternate photoperiod (black diamonds); dCP, direct continuous photoperiod (grey dots); iCP, indirect continuous photoperiod (light grey triangles). See Methods for experimental design. See Figure S1a for similar analysis of the whole, uncurated, dataset. Labels (diamonds, circles and triangles) are individual values. Thick horizontal lines are medians; boxes illustrate the 25th to 75th quantiles, and upper and lower limits of whiskers at plus or less than 1.5 × IQR, where IQR is the interquartile range or distance between the first and third quartiles.

When the whole (uncurated) dataset was considered, the patterns observed in terms of differences among conditions were affected by variations in fish detection performances (see Methods). However, the results also indicated a lower total distance travelled by the fish in AP conditions (see Figure S1a, H = 13.9, p‐value < 0.05).

3.2.2. Mean velocity

Overall, the mean velocity of fish was similar during day and night periods within each condition throughout the experiment (H = 1019, p‐value >0.05). Over the six 5‐min sequences analysed, the fish in dCP were significantly faster (with an average swimming speed of 7.9 ± 0.4 cm s⁻¹) than in AP and iCP (H = 16.9, p‐value <0.05; average 5.7 ± 0.3 cm s⁻¹; Figure 3). During the day, the fish moved slower during the three 5‐min sequences analysed in AP than in other conditions (H = 13.5, p‐value <0.05; Figure 3), with an average swimming speed of 4.8 ± 0.7 cm s⁻¹ in AP, vs. 7.1 ± 0.3 cm s⁻¹ in iCP and dCP (Figure 3). At night, the mean velocity of fish also varied between conditions (H = 9.72, p‐value <0.05; Figure 3) and was significantly faster in dCP (average of 8.3 ± 0.6 cm s⁻¹) than in AP and iCP (5.6 ± 0.4 cm⁻¹) over the three 5‐min sequences analysed (Figure 3).

FIGURE 3.

Mean velocity of C. ramada juveniles (in cm s⁻¹) over three 5‐min periods during the day (left) and three 5‐min periods during the night (right) throughout the experiment (N = 90; see Methods for details on the selection procedure of analysed sequences). AP, alternate photoperiod (black diamonds); dCP, direct continuous photoperiod (grey dots); iCP, indirect continuous photoperiod (light grey triangles). See Methods for experimental design. See Figure S1b for similar analysis of the whole, uncurated, dataset. Labels (diamonds, circles and triangles) are individual values. Thick horizontal lines are medians; boxes illustrate the 25th to 75th quantiles, and upper and lower limits of whiskers include values at plus or less than 1.5 × IQR, where IQR is the interquartile range or distance between the first and third quartiles.

When the whole (uncurated) dataset was considered, the patterns observed in terms of differences among conditions were affected by variations in fish detection performances (see Methods). However, the results also indicated that overall fish moved significantly faster in dCP conditions than in AP and iCP (H = 13.9, see Figure S1b).

3.2.3. Mean distance between fish (group cohesion)

The average distance between individuals was similar during day and night periods within each condition throughout the experiment (H = 962, p‐value >0.05). However, the distance between individuals varied between all conditions of exposure (H = 23.5, p‐value <0.05; AP < iCP < dCP with averages of 26.8 cm ± 0.5 cm, 28.7 cm ± 0.6 cm and 30.6 cm ± 0.3 cm, respectively). Additionally, during the day, the individuals were closer to one another in AP than in any other conditions (H = 25.3, p‐value <0.05; Figure 4), with an average distance of 25.2 ± 0.5 cm between individuals in AP, 30.4 ± 0.4 cm in iCP and dCP. At night, the distance between individuals also varied between conditions (H = 9.98, p‐value <0.05; Figure 4) and was significantly greater in dCP (average of 30.5 ± 0.4 cm) than in AP and iCP (27.9 ± 0.4 cm).

FIGURE 4.

Mean distance between individuals (in cm) of C. ramada juveniles over three 5‐min periods during the day (left) and three 5‐min periods during the night (right) throughout the experiment (N = 90; see Methods for details on the selection procedure of analysed sequences). AP, alternate photoperiod (black diamonds); dCP, direct continuous photoperiod (grey circles); iCP, indirect continuous photoperiod (light grey triangles). See Methods for experimental design. See Figure S1c for similar analysis of the whole, uncurated, dataset. Labels (diamonds, circles and triangles) are individual values. Thick horizontal lines are medians; boxes illustrate the 25th to 75th quantiles, and upper and lower limits of whiskers include values at plus or less than 1.5 × IQR, where IQR is the interquartile range or distance between the first and third quartiles.

When the whole (uncurated) dataset was used, the patterns observed in terms of differences between conditions were affected by variations in fish detection (see Methods). However, the results also indicated an overall lower distance between individuals in AP (H = 24.9, see Figure S1c).

3.2.4. Angular velocity

The angular velocity displayed by fish was similar whatever the period and condition throughout the experiment (H = 1085 for day and night comparison, H = 1.18 for conditions, p‐values >0.05; Figure 5). The average angular velocity was 0.5 ± 0.9 deg. s⁻¹. When the whole (uncurated) dataset was used, the patterns observed in terms of differences between conditions were affected by variations in fish detection performances (see Methods) but the results also indicated no significant difference between conditions (see Figure S1d).

FIGURE 5.

Angular velocity (deg s⁻¹) by C. ramada juveniles over three 5‐min periods during the day (left) and three 5‐min periods during the night (right) throughout the experiment (N = 90; see Methods for details on the selection procedure of analysed sequences) AP, alternate photoperiod (black diamonds); dCP, direct continuous photoperiod (grey circles); iCP, indirect continuous photoperiod (light grey triangles). See Methods for experimental design. See Figure S1d for similar analysis of the whole, uncurated, dataset. Labels (diamonds, circles and triangles) are individual values. Thick horizontal lines are medians; boxes illustrate the 25th to 75th quantiles, and upper and lower limits of whiskers include values at plus or less than 1.5 × IQR, where IQR is the interquartile range or distance between the first and third quartiles.

3.3. Exploration of cortisol production as a measure of physiological stress

3.3.1. Waterborne cortisol

The cortisol concentration excreted in the water by the fish averaged across conditions did not vary between day and night periods (H = 117.5, p‐value >0.05). The cortisol concentrations were significantly lower in iCP (0.27 ± 0.03 ng mL⁻¹ g⁻¹; H = 20.9, p‐value <0.05) as compared with AP and dCP (0.61 ± 0.05 ng mL⁻¹ g⁻¹; Figure 6). The fish from AP conditions excreted significantly more cortisol than fish in iCP for day and night data together (AP: 0.73 ± 0.07 ng mL⁻¹ g⁻¹, iCP: 0.27 ± 0.03 ng mL⁻¹ g⁻¹; H = 12.8 for day data, H = 8.47 for night data, p‐value <0.05). dCP was statistically similar to other conditions during both day and night periods (H = 12.8 for day data, H = 8.47 for night data, p‐value >0.05; dCP: 0.49 ± 0.05 ng mL⁻¹ g⁻¹). When the whole dataset was used, the concentrations of cortisol excreted by fish in the water were not statistically different between conditions and throughout the three experimental trials (H = 1804.5 comparison day and night data, H = 5.22 for condition, H = 3.38 for day data, H = 1.85 for night data, p‐values >0.05; see Figure S2).

FIGURE 6.

Cortisol in water excreted in ng mL⁻¹ g⁻¹ as a function of time for the three conditions (N = 39). Sun and moon symbols indicate values obtained during day and night periods, respectively. AP, Alternate photoperiod (top panel); dCP, Direct continuous photoperiod (middle panel); iCP, Indirect continuous photoperiod (bottom panel). Labels (dots and triangles) are individual values on tanks. See Methods for experimental design. ‘D0’, ‘D1’, ‘D2’ and ‘D3’ correspond to the days of the experiment (days 1 to 3). See Figure S2 for similar analysis of the whole, uncurated dataset.

3.3.2. Plasma cortisol levels at the end of the exposure period

The estimated sampled plasma volumes ranged from 20 to 150 μL. The cortisol concentrations quantified at the end of the exposure period were similar in all conditions (H = 4.5, p‐value > 0.05), with mean concentrations of 57.24 ± 8.51 ng g⁻¹ (Figure 7). When the whole dataset was used, cortisol concentrations were not statistically different among conditions either (H = 1.51, p‐value > 0.05; see Figure S3).

FIGURE 7.

Cortisol concentration quantified in the plasma of C. ramada juveniles (in ng g⁻¹) throughout the experiment (N = 14). AP, Alternate photoperiod (black dots, N = 4 due to one failed puncture); dCP, Direct continuous photoperiod (grey triangles, N = 5); iCP, Indirect continuous photoperiod (light grey squares, N = 5). Labels (circles, triangles and squares) are individual values. See Methods for experimental design. See Figure S3 for similar analysis of the whole, uncurated, dataset. Thick horizontal lines are median values. Boxes include all values included within the 25th to 75th quantiles, and upper and lower limits of whiskers include values at plus or less than 1.5 × IQR, where IQR is the interquartile range or distance between the first and third quartiles.

There was no correlation between plasma cortisol and water cortisol at the end of the experiment (df = 7, t = −1.1, p‐value = 0.3 and R ² = −0.4; N = 6), neither for the whole data set (df = 7, t = −1.1, p‐value = 0.3 and R ² = −0.4; N = 24).

4. DISCUSSION

4.1. Experimental conditions

During our experiment, the temperature was shown to significantly vary among conditions, although the difference did not exceed 1.2°C between AP and dCP (globally, the temperature varied from 19.3 to 22.1°C between conditions over the successive trials, this variation being related to seasonal fluctuations in river temperature over the course of the project). Such small temperature variations are unlikely to affect the behaviour of C. ramada juveniles, which are known to migrate between marine, brackish and sometimes freshwater habitats following hydrodynamic cues related to mixed effects of currents and salinity (Trancart et al., 2011, 2012). Hence, those fish can easily cope with small variations in temperature or salinity in natural environments (Trancart et al., 2016), which also facilitated their gentle acclimatisation from oligohaline to freshwater conditions over the experimental process. Additionally, dissolved oxygen and nitrogen‐associated excretion products were scrupulously monitored during the experiment, which ensured optimal conditions for fish survival and well‐being throughout the project. During the second experimental trial, the average dissolved oxygen and nitrogen product (ammonium and nitrite) concentrations in the aquaria were 8.3 mg L⁻¹ and 0.1 mg L⁻¹, respectively, both low values ensuring optimal survival conditions for the fish (Ciji & Akhtar, 2020) (over the three successive trials, they were at 8.7 to 8.2 mg L⁻¹ and 0.1 mg L⁻¹, respectively). In AP aquaria, in which lights were set up for day/night successions over 72 h identical to the photoperiod used during acclimatisation period, the LED lamps used, specifically designed for aquarium use, limited warming of the water. Furthermore, dawn and dusk transitions were gradual (30 min) in terms of light intensity [Péan et al., 2013 having shown that brutal variations in light intensity without twilight during day/night transitions induce rapid changes in swimming and stress in zebrafish Danio rerio (Hamilton 1822)]. The light exposure tested during this experiment therefore only differed in terms of photoperiod conditions (presence or absence of a diurnal cycle), but not in terms of variations in light intensity occurring at dawn or dusk, nor to any other behavioural indirect effects arising from other experimental conditions, including animal habituation.

4.2. Behaviour responses of C. ramada juveniles to ALAN

Our results indicate that the interruption of diurnal cycles resulting from the constant use of 4000 K LED lamps influences the swimming behaviour and stress of juvenile fish exposed during 72 h. Our experimental design was however conceived to test for a drastic absence of photoperiod (also using a neutral white light colour of 4000 K, although most lamps in managed urban environments are 3000 K, warm white, LEDs). However, studies conducted using more realistic environmental conditions (particularly lower light intensities at night) also showed that ALAN affects the behaviour of marine animals in relation to their biological rhythms, such as valve opening and associated trophic activity in oysters (Botté et al., 2023). In another experimental study using lamps set up at 5000 K (cool white), 12 h day‐day or day‐night modalities during 60 days, ALAN was observed to affect clownfish Amphiprion ocellaris reproduction, with eggs incubated in the presence of ALAN simply not hatching (Fobert et al., 2019). Therefore, ALAN can affect many aspects of fish and other aquatic animals biology and ecology, including fish reproduction (Brüning et al., 2011; Brüning, Kloas, et al., 2018; Fobert et al., 2019), migration (Vowles & Kemp, 2021), but also diel behaviour and activities influenced by a circadian clock (Franke et al., 2013; Kurvers et al., 2018; Miner et al., 2021; this study).

Fish behavioural response patterns varied between day and night for some variables. During the day (or 5 min between 06:30 and 07:30, 12:00 and 13:00, 17:00 and 18:00), the fish exposed to AP (or our ‘control’ condition with a normal, alternate, photoperiod) travelled less distance, swam slower and kept closer together, therefore had a lower activity level, while the fish exposed to continuous lighting (dCP and iCP) travelled longer distance, swam faster and were further away from one another, therefore displaying a higher activity level. At night (5 min between 21:30–22:30, 23:00–00:00 and 03:00–04:00), the fish exposed to direct continuous lighting (dCP) also swam significantly faster, longer distance and showed higher inter‐individual distances than the fish exposed to an indirect continuous photoperiod (iCP) or to AP. When the whole (uncurated) behavioural dataset was used, the fish in AP ‘control’ condition still were less active and displayed tighter shoal than in exposed CP conditions (including both direct and indirect exposure modalities). Meanwhile, the fish in dCP conditions were more active than in AP and iCP. These results suggest that continuous and intense lighting at night modifies fish movements and activity, and alter cohesion among individuals in a group, with exposed fish maintaining an intense swimming activity at night although C. ramada juveniles are known to use diurnal selective tidal stream transport in nature, as do other catadromous species (Trancart et al., 2012). It also has already been shown that C. ramada young of the year initiate migratory movements in response to water flow direction and light–dark cycle (Trancart et al., 2011). Their feeding activity is also known to increase at sunrise and to remain then stable throughout the day (Gisbert et al., 1997). Therefore, an altered photoperiod can modify the diel rhythm of their movement behaviour, as well as other related activities including foraging, which involves a somewhat stationary position and happens mostly at day for many marine‐estuarine species with developed vision cues such as C. ramada (Almeida, 2003; Carassou, Benhaïm, et al., 2025). Note that studies conducted in aquaculture contexts also suggest that C. ramada larvae can distinguish colours (El‐Sayed & El‐Ghobashy, 2011). Our results, although obtained in necessarily simplified and exaggerated conditions (involving notable stress for wild species kept captive, see below), suggest ALAN can affect mullet activities in natural habitats, including swimming behaviour and associated essential functions regulated by nyctemeral cycles, such as resting, feeding or migrating. To broaden these conclusions, these first results need to be confronted against more realistic nocturnal lighting conditions, that is, using lower intensity at night for example, a situation more plausible in real‐life contexts (Botté et al., 2023). However, our results are in line with other previous studies reporting diverse impacts of ALAN on fish behaviour, either in experimental or in natural conditions and based on diverse species models.

For example, ALAN was shown to induce a swimming behaviour likely to increase predation risks in P. reticulate (Kurvers et al., 2018), to affect the natural rhythmicity of activities in G. laevifrons (Pulgar et al., 2019) and M. dolomieu (Foster et al., 2016), or the swimming behaviour of the model species D. rerio (Shafiei Sabet et al., 2016). Migratory behaviours can also be affected, as observed for S. salar (Riley et al., 2012) or A. anguilla (Vowles & Kemp, 2021). The effects of ALAN on larval recruitment have also been reported in A. triostegus (O'Connor et al., 2019). Fish behavioural responses to light pollution therefore stands for diverse fish life stages, multiple species and a diversity of ecological guilds.

4.3. Exploratory measures of physiological stress using cortisol concentration

In our study, in addition to behavioural response variables, we tried to quantify fish stress response using cortisol concentrations in the water every 6 h, in an attempt to catch up the dynamics of cortisol excretion over the 72‐h exposure period. However, we observed no clear temporal patterns in cortisol quantities, which also remained similar between conditions although concentrations in iCP were unexpectedly lower than in other conditions. We did not observe any difference in plasma cortisol concentrations between exposure conditions either. Finally, cortisol estimations from plasma samples appeared poorly correlated to the waterborne cortisol levels measured at the end of the experiment, contrary to other studies on other species (Gabor & Contreras, 2012; Wong et al., 2008). Based on this absence of significant results, and considering no data on C. ramada cortisol production are available so far (including basal cortisol estimations in water, blood, or any other substrate), it remains difficult to properly interpret those exploratory results further. On other species, other studies also reported that cortisol levels were not affected by ALAN for several species (e.g., P. fluviatilis, Brüning et al., 2015 ; S. salar, Newman et al., 2015; A. vulpes, Szekeres et al., 2017 or G. affinis, Miner et al., 2021).

Besides the absence of cortisol baseline values, we did not observe cortisol accumulation in the water over time, contrary to our expectations. This may indicate an inadequate water sampling frequency (every 6 h in this instance). Plasma cortisol concentrations were indeed shown to peak within 5–30 min of handling stress in fry and juveniles of jundiá [Rhamdia quelen (Quoy & Gaimard, 1824), Koakoski et al., 2012], whereas in D. labrax, peak levels occurred about 1 h after a challenge such as chasing or emersion (Fanouraki et al., 2008). In fact, cortisol is a relatively stable molecule that, once released in the water, has a half‐time of 16 h at 12°C (Ellis et al., 2004, 2007). In any future research using similar experimental conditions, it would thus be beneficial to sample the water at times closer to the stress initiation, that is, 15 min after switching on and 15 min after switching off the lamps for example. Studies have for example already demonstrated very rapid cortisol production after a punctual stress phenomenon (on A. anguilla under the effect of anxiolytics, Sebihi et al., 2025; after fish handling effect in R. quelen, Koakoski et al., 2012). The relative increase in waterborne cortisol measured in AP over time may be explained by other ‘perceived factors’ (Barton, 2002) occurring prior experiment, such as handling, capture, confinement, or transport which can be stressful especially for wild animals.

Cortisol excretion in response to light relies upon direct impact of the hypothalamic‐pituitary‐interrenal (HPI) axis in fish. The absence of direct effect of the exposure conditions on the HPI axis of C. ramada juveniles may explain the abscence of clear cortisol excretion signals. However, this study is the first to report cortisol values for this species in the literature, so that the basal levels of the hormone in fish blood or its dynamics over time in response to a stress remain largely unknown. Therefore, our cortisol measurements represent an exploratory avenue of research that will require further improvements in order to address how C. ramada produces cortisol in basal and stressful contexts in the future. Finally, the unexpected poor correlation between cortisol measured in the water and in fish blood at the end of the exposure period may arise from other methodological avenues including different protocols to sample each matrix. For plasma cortisol, the literature reports strong impacts of the semi‐invasive sampling procedure (whatever the method used for blood extraction, here realised through caudal vein punctures, Sadoul & Geffroy, 2019). The fish anaesthesia method, and time between fish anaesthesia or euthanasia and blood sampling, can also affect the resulting cortisol measures in the plasma. In D. labrax for example, the plasma cortisol concentration increased from 50 to 250 ng L⁻¹ within 12 min of fish death (Sadoul & Geffroy, 2019). On the other hand, water cortisol measures, as representative of the total amount excreted by five fish together in individual aquaria, capture a high inter‐individual variability (as expressed by the relatively high coefficient of variation related to these ELISA estimations, which could reach 119% at the maximum). Finally, the high cortisol values obtained from both methods may be specific to the species metabolism itself, as cortisol levels are highly species‐dependent, but could also arise from bias related to challenges in acclimatising wild animals to captivity, consequences of the antibiotic treatment (although concentration and duration of the treatment were applied similarly for the whole fish stock), or fish handling for blood sampling procedure. The reuptake of waterborne cortisol by fish is also possible (Lemos et al., 2023; Scott & Ellis, 2007). The hydrophobic properties of the steroid molecule also increase its adsorption on plastic materials (pipes) from the experimental apparatus (Sadoul & Geffroy, 2019), which may also have played a role in explaining the poor correlations observed between values from the two procedures. All these factors, related to methodological levers, may play a role in explaining our results.

5. CONCLUSION

Although our experimentation was designed to test for an extreme alteration of diurnal cycles, results obtained indicate that behavioural performances of C. ramada juveniles were affected by the nocturnal light alteration, and those behavioural responses are shown to directly relate to survival, growth and reproduction of the species in nature (Pyle & Ford, 2017). These results emphasise that behaviour is a sensitive integrative endpoint to study the sub‐lethal effects of specific stressors in controlled conditions, while these effects are often underestimated in field‐based studies (Robinson, 2009; Wong & Candolin, 2015). While this study enabled us to report some behavioural responses of C. ramada to an emergent urban stress factor for the first time, we however observed no effect of continuous illumination on cortisol measurements, the interpretation of which is constrained by methodological levers. Observed responses on the swimming activity and group cohesion behaviour still suggest possible impacts of extreme light pollution on foraging, resting and migration activities of this tolerant species in nature. These impacts could potentially affect the multiple provisioning and regulatory ecological services associated to this species in many regions of the world, including its important role in terms of energy transfers to upper trophic levels and between adjacent components of aquatic meta‐ecosystems. Further studies are nevertheless needed to better understand ALAN impacts in a more realistic context, such as, for example, addressing the impact of lower intensity lights with or without the additional effect of ambient skyglows on other wild fish species. Furthermore, addressing possible differential responses depending on previous exposure conditions in fish native environments (examining possible habituation effects on behavioural or stress responses for example) would be an interesting research avenue. Finally, we also suggest that cortisol‐based methodologies need development when applied to wild species representative of realistic conditions in nature.

AUTHOR CONTRIBUTIONS

CR contributed to the design of the experimental set‐up, led the experiment, including data sampling, analysis and interpretation, and drafted the manuscript. AB designed the experimental set‐up, analysed cortisol samples and revised the manuscript. PC contributed to the experimental design, to data sampling and substantively revised the manuscript. FV led and managed the experimental design set‐up, contributed to data sampling and revised the manuscript. PJ helped in designing the experimental set‐up, advised on animal welfare–related issues and substantively revised the manuscript. MLB advised on cortisol measurement protocols, helped in behavioural data interpretation and substantively revised the manuscript. SM supervised the PhD student, contributed to data interpretation and substantively revised the manuscript. LC supervised the PhD student, secured the funding, coordinated the research project, contributed to the experimental design set‐up, advised on protocols and data interpretation and substantively revised the manuscript. All authors read and approved the final manuscript.

FUNDING INFORMATION

This work was supported by the Institut National de Recherche pour l'Agriculture, l'Alimentation et l'Environnement (AQUA department and the Nature Division of Bordeaux Métropole Agence de l'Eau Adour Garonne, Département de la Gironde, and Région Nouvelle‐Aquitaine; France. as part of the shared research and development agreement relating to the Biodiver'Cité action plan including grants from Bordeaux Métropole, Agence de l'Eau Adour Garonne, Département de la Gironde, and Région Nouvelle‐Aquitaine; France). This work was carried out with the support (equipment and staff) provided by the XPO scientific infrastructure in Saint‐Seurin‐sur‐l'Isle (https://doi.org/10.17180/brey-mr38; part of LIFE [Living in Freshwater and Estuaries] research infrastructure of INRAE and AnaEE‐France).

Supporting information

Table S1. Parameters of detection settings on Ethovision XT Noldus for each condition AP (alternate photoperiod), dCP (direct continuous photoperiod) and iCP (indirect continuous photoperiod).

Table S2. Number of videos and average ‘subject not found’ values given by Ethovision XT Noldus for each condition of the experimentation (AP, alternate photoperiod; dCP, direct continuous photoperiod; iCP, indirect continuous photoperiod).

Table S3. Summary of average exposure conditions (values are mean ± standard errors) over the experiments. Different superscript letters indicate statistical differences between individual conditions (‘All’) or specific effect of condition exposure (Dunn tests). T corresponds to the experimental trials, as follows: Trial 1, T1, was conducted from 23/05/23 to 26/05/23; trial 2, T2, from 30/05/23 to 2/06/23; and trial 3, T3, from 6/06/23 to 9/06/23. AP, alternate photoperiod; dCP, direct continuous photoperiod; iCP, indirect continuous photoperiod. QL stands for quantification limits.

Figure S1. Behavioural variables on uncurated dataset: (a) Total distance travelled in cm during 1 h during day and night periods for the three light conditions. (b) Mean velocity in cm s⁻¹ during 1 h during day and night periods for the three light conditions. (c) Mean distance between fish in cm during 1 h during day and night periods for the three light conditions. (d) Angular velocity in deg. s⁻¹ during 1 h during day and night periods for the three light conditions. AP, alternate photoperiod (N = 190); dCP, direct continuous photoperiod (N = 190); iCP, indirect continuous photoperiod (N = 179, so 559 in total). The parameter ‘Subject not found in percent’ given by the Ethovision XT (Noldus, The Netherlands) corresponds to the percent of time when fish were not followed, and is classified by three groups: Subject not found <15% (light grey dots), Subject not found [15–30] % (grey triangles), Subject not found >30% (black squares). Thick horizontal lines are median values. Boxes include all values included within the 25th to 75th quantiles, and upper and lower limits of whiskers include values at plus or less than 1.5 × IQR, where IQR is the interquartile range or distance between the first and third quartiles.

Figure S2. Cortisol in water excreted in ng mL⁻¹ g⁻¹ as a function of time for the three conditions for each week experiments (N = 117). Sun and moon symbols indicate values obtained during day and night periods, respectively (T1: week 1, grey dots; T2: week 2, black triangles; T3: week 3, grey squares). AP, alternate photoperiod (top panel); dCP, direct continuous photoperiod (middle panel); iCP, indirect continuous photoperiod (bottom panel). See Methods for experimental design.

Figure S3. Cortisol concentration in juvenile thinlip mullet plasma (in ng g⁻¹) throughout the experiment (N = 37). AP, alternate photoperiod (left); dCP, direct continuous photoperiod (middle); iCP, indirect continuous photoperiod (right); T1, week 1, grey dots; T2, week 2, black triangles; T3, week 3, grey squares. See Methods for experimental design. Thick horizontal lines are median values. Boxes include all values included within the 25th to 75th quantiles, and upper and lower limits of whiskers include values at plus or less than 1.5 × IQR, where IQR is the interquartile range or distance between the first and third quartiles.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.