Alteration of cleaner wrasse cognition and brain morphology under marine heatwaves

Abstract

en

1. Heatwaves, exacerbated by global warming, are progressively affecting various ecosystems, with coral reefs among the most susceptible. Within these ecosystems, cleaner wrasses (Labroides dimidiatus) engage in cooperative interactions with client fish by removing ectoparasites and play an essential role in sustaining client abundance and diversity.
2. In 2016, the northern section of the Great Barrier Reef experienced widespread and intense bleaching due to unparalleled ocean temperatures associated with a marine heatwave. While prior studies have connected changes in fish densities following this heatwave to modifications in cleaner fish cognitive performance, the immediate impact of heatwave exposure on cleaner fish cognition and brain structure has yet to be investigated.
3. Here, we exposed cleaner wrasses to a laboratory-simulated Category 1 marine heatwave for 55 days, mirroring the 2016 Great Barrier Reef event. Cleaners' cognitive performance was evaluated through a visual discrimination task during the heatwave and after a 14-day recovery phase. This was followed by an analysis of brain development 30 days after the cessation of the marine heatwave.
4. Our results demonstrate that although heatwave exposure temporarily hindered cognitive performance, these deficits were recoverable. Interestingly, cleaner fish brain morphology, measured after recovery, underwent significant changes. Specifically, despite cleaners exposed to heatwaves having notably larger brains, their telencephalon was substantially smaller, while their brainstem was enlarged.
5. These findings indicate that while some cognitive effects may be reversible, marine heatwave exposure leads to lasting alterations in brain morphology, particularly in regions associated with higher cognitive functions and social behaviour. This raises questions about the potential impact on more complex tasks that rely on these brain regions. We argue that the significant disruptions in cleaners' cognitive performance observed months after the 2016 due to neurological impairments linked to brain morphological changes. If so, a mere recovery of fish densities may not necessarily lead to a restoration of cognitive performance, as experiencing marine heatwaves might induce life-long morphological alterations in fish. Our results underscore marine heatwaves' intricate and enduring impact on cleaner fish, emphasizing the need for comprehensive strategies to safeguard these vital components of coral reef ecosystems.

Read the free Plain Language Summary for this article on the Journal blog.

Resumo

pt

1. As ondas de calor, intensificadas pelo aquecimento global, estão a afetar progressivamente vários ecossistemas, e os recifes de coral estão entre os mais susceptíveis. Dentro destes ecossistemas, os bodiões limpadores (Labroides dimidiatus) desempenham um papel crucial ao remover ectoparasitas dos peixes clientes, contribuindo para a manutenção da sua abundância e diversidade.
2. Em 2016, a secção norte da Grande Barreira de Coral sofreu um branqueamento massivo devido a uma onda de calor marinha sem precedentes. Embora um estudo anterior tenha relacionado alterações na densidade de peixes a modificações no desempenho cognitivo dos bodiões-limpadores após este evento, o impacto imediato da exposição a ondas de calor na cognição e na estrutura cerebral destes peixes ainda não foi explorado.
3. Neste estudo, bodiões-limpadores foram expostos a uma onda de calor marinha de Categoria 1 simulada em laboratório durante 55 dias, que replicou o evento de 2016 na Grande Barreira de Coral. O desempenho cognitivo dos limpadores foi avaliado através de uma tarefa de discriminação visual durante a onda de calor (dias 50 a 55 de exposição) e após uma fase de recuperação de 14 dias. Posteriormente, foi realizada uma análise da densidade cerebral, 30 dias após a interrupção da onda de calor.
4. Os resultados revelaram que, embora a exposição à onda de calor tenha causado um défice cognitivo temporário, este foi recuperável. Curiosamente, observou-se uma alteração significativa na morfologia cerebral dos peixes limpadores após a recuperação: os limpadores expostos a ondas de calor apresentaram cérebros proporcionalmente maiores (relação cérebro-corpo), com um telencéfalo substancialmente menor (relação telencéfalo-cérebro), e um tronco encefálico aumentado (relação tronco encefálico-cérebro).
5. Estas descobertas sugerem que, embora os efeitos cognitivos imediatos sejam reversíveis, a exposição a ondas de calor marinhas provoca alterações duradouras na morfologia cerebral, particularmente em regiões associadas a funções cognitivas e ao comportamento social. Argumentamos que as perturbações cognitivas observadas nos bodiões-limpadores meses após os eventos de 2016 podem não se dever apenas a mudanças na densidade de peixes, mas também a alterações neurológicas associadas a modificações morfológicas cerebrais. Se assim for, a simples recuperação da densidade de peixes poderá não ser suficiente para restaurar o desempenho cognitivo, uma vez que a exposição a ondas de calor marinhas pode induzir alterações morfológicas permanentes nos peixes. Estes resultados destacam o impacto complexo e duradouro das ondas de calor marinhas sobre os bodiões-limpadores, sublinhando a necessidade de estratégias abrangentes para proteger estes componentes vitais dos ecossistemas dos recifes de coral.

1 INTRODUCTION

Marine ecosystems, particularly coral reefs, are facing unprecedented challenges in the face of anthropogenic climate change. The escalating trend of gradual warming is accompanied by the rise of marine heatwaves (MHWs)—temperature anomalies that exceed the 90th percentile of the local climatology for 5 days or more (Hobday et al., 2016). These extreme climatic events are increasing in frequency, intensity, duration and spatial extent (Oliver et al., 2018), emerging as a significant threat to coral reef fish. For tropical ectotherms living near their thermal maximum, heightened temperatures can lead to a decrease in metabolic scope—the energy available for activities beyond essential physiological functions—limiting the efficiency of other important ecologic processes such as development, growth and reproductive capacity (Brown et al., 2004; Farrell et al., 2008; Rummer et al., 2014; Tewksbury et al., 2008). In response, fish can either migrate to habitats with more favourable temperatures or adapt to the new environmental conditions through genetic changes or environmentally induced phenotypic plasticity (Nati et al., 2016; Sandblom et al., 2014; Schulte et al., 2011; Seebacher et al., 2015). However, the extent, pace, and cost of acclimation to warming remain largely unknown for fish and other vertebrates (Bradshaw & Holzapfel, 2006; Morgan et al., 2020; Sandblom et al., 2016). Another vital process that can become impaired at high temperatures is cognition (Danner et al., 2021; Toni et al., 2019). Cognitive abilities in fish encompass a range of mental processes crucial for fitness-related behaviours, such as foraging, predator avoidance, and engagement in social interactions (Morand-Ferron et al., 2016). While warming-induced effects on cognitive performance of ectotherms are poorly studied, research suggests it may underlie behavioural changes through different physiological and neurobiological mechanisms (Ramírez-Calero et al., 2023; Silveira et al., 2023; Toni et al., 2019; Závorka et al., 2020). For example, Závorka et al. (2020) found that in warmer environments, maintaining high aerobic scope in fish comes at the cost of altered brain morphology and reduced capacity to explore new environments.

In the coral reefs of the Indo-Pacific Ocean, the bluestreak cleaner wrasse, Labroides dimidiatus (hereafter referred to as ‘cleaner’), plays a crucial role in enhancing the fitness of other reef fish (so called ‘clients’) by removing parasites and dead tissue during cooperative cleaning interactions (Bshary, 2003; Grutter, 1999). However, conflict arises due to cleaners' preference for feeding on client mucus—an energetically costly resource for clients to produce and thus constitutes cheating (Grutter & Bshary, 2003). Notably, this deceptive behaviour is limited to non-predatory clients, as predators carry the threat of reciprocity (Bshary & Bronstein, 2004). To mitigate cheating and maintain cooperation, non-predatory clients employ several partner control mechanisms (Bshary & Grutter, 2005). Resident clients with access to a single cleaning station aggressively chase cleaners as punishment (Bshary & Grutter, 2002). Visitors with large home ranges and access to multiple cleaning stations respond to cheating with the termination of interaction and switch of cleaner (partner switching; Bshary & Schaffer, 2002). Additionally, clients eavesdrop on occurring interactions and choose cleaners who provide a good service and avoid those they witness cheating (Bshary & Grutter, 2002). To navigate the balance of these interactions and optimize food intake, strategically sophisticated cleaners resort to a set of social techniques, adjusting their behaviour flexibly given the specifics of a situation. In line with biological market theory (Noë et al., 1991), cleaners adjust their conduct based on client type, prioritizing visitor (and ephemeral) clients over resident (and permanent) ones (Bshary & Noë, 2003). Furthermore, in the presence of observers (‘bystanders’) and potential future clients, cleaners enhance their cooperative behaviour to increase their image score (Bshary & Grutter, 2006; Pinto et al., 2011). Adding complexity is the cleaners' ability to manipulate client decisions through tactile stimulation (i.e. stroking their pelvic and pectoral fins on the clients' dorsal area). This massage-type behaviour reduces stress levels, allowing cleaners to prolong interactions and reconcile with clients after cheating (Grutter, 2004; Soares et al., 2011).

Over the past two decades, cleaner wrasses have emerged as a model species for investigating fish social cognition (Binning et al., 2018; Bshary & Grutter, 2002). Their remarkable cognitive abilities, encompassing learning and memory, are crucial for maintaining mutualism and its positive impact on the ecosystem, extending beyond individual interactions to influence population and community health (Clague et al., 2011; Grutter et al., 2003; Wagner et al., 2015; Waldie et al., 2011). Therefore, alterations to the interaction behaviour of L. dimidiatus due to environmental conditions could have cascading consequences for tropical fish communities and the ecosystems they inhabit.

Previous research showed that elevated temperatures can adversely affect cleaning behaviour (Paula et al., 2019; Triki et al., 2018). Specifically, under warming conditions, L. dimidiatus exhibited reduced interaction frequencies with its surgeonfish (Paula et al., 2019). After the 2016 El Niño event, during which the northern section of the Great Barrier Reef underwent mass bleaching due to a marine heatwave, Triki et al. (2018) reported a disproportionate decline in cleaner density and cleaner-to-client ratios and a reduced cognitive strategic sophistication among adult cleaners. The authors suggested that shifts in biological market conditions, characterized by high demand and low supply, possibly explain the observed decline in the cognitive performance of cleaners. Yet, they also postulate an alternative hypothesis—the direct effects of marine heatwave conditions on cleaners' physiological status.

In this study, we investigate whether the response to environmental stress in a laboratory setting would parallel the observations of diminished strategic sophistication in cleaners, as reported by Triki et al. (2018) in the field, and if it could be related to a physiological impact. To explore this, individuals were subjected to a simulated Category 1 marine heatwave (i.e. moderate magnitude of scale, reaching 1°C–2°C above the mean climatology; Hobday et al., 2018) lasting 55 days, replicating the conditions of the 2016 El Niño event. The potential alignment of our laboratory findings with the field observations by Triki et al. (2018) would suggest that the impairment of cleaners' cognitive abilities could be linked to the direct consequences of a temperature affecting cleaners' physiological mechanisms. We measured cognitive performance through a visual discrimination task during the heatwave and after a 14-day recovery phase. Finally, brain density analysis was conducted 30 days following the conclusion of the exposure period.

2 METHODOLOGY

2.1 Fish subjects

Bluestreak cleaner wrasse, Labroides dimidiatus (n = 20), measuring ~5.7 cm in length (±0.5), were wild-caught by local fishermen in the Maldives and transported to the aquatic facilities of Laboratório Marítimo da Guia, Cascais, Portugal. Each cleaner was individually housed in glass tanks (50 × 20 × 25 cm) within flow-through aquatic systems, provided with shelter (i.e. PVC tube; 2 cm diameter, 10 cm length), and fed ad libitum with Mysis shrimp (Gamma Slice Mysis Shrimp—Neomysis japonica, TMC) twice a day.

2.2 Housing and experimental setup

The flow-through aquatic systems (Figure S1) utilized natural seawater (NSW) pumped from the sea, filtered (0.35 μm), UV-irradiated (Vecton 300, TMC) and delivered to mixing tanks through a dripping system, ensuring constant water renewal. Each mixing tank featured biological (bio-balls), mechanical (100 μm filter, TMC), and physical filtration (ReefSkimPro 400, TMC). To maintain appropriate oxygen levels, filtered air was injected directly into the mixing reservoirs using a compressor (Medo Blower LA-120A, Nitto Kohki, Japan, sourced from UK branch, Derbyshire) and soda lime (Sofnolime, soda lime, Molecular Products Ltd). The water was then supplied to three experimental aquaria via a water pump (TMC, V2 Power Pump, 2150 L h⁻¹), undergoing a second stage of UV-irradiation (Vecton 300, TMC) before delivery. The photoperiod was set at 12 h of light and 12 h of darkness. Daily assessments of ammonia and nitrite levels, using colorimetric test kits (Aquamerk, Merck Millipore), ensured levels remained below detectable thresholds.

After arrival, cleaners underwent a 5-day temperature-stable acclimation period before exposure to two experimental treatments: control (n = 10, 28.8°C) and a simulated scenario of a category I marine heatwave (n = 10, max. 29.8°C) (Table 1). Temperature control was automated, monitored every 5 min, and adjusted with an error margin of 0.15°C (Profilux 3.1N, GLH). Cooling was regulated by chillers (TK1000, Teco), and heating was managed by submerged heaters (300 W, TMC). Manual daily monitoring of seawater temperature and oxygen (oximeter VWR DO220), salinity (V2 refractometer, TMC) and pH (826 pH mobile, Metrohm) complemented the automatic systems. Seawater parameters of experimental setups are summarized in Table S1.

TABLE 1. Replication statement.

Scale of inference	Scale at which the factor of interest is applied	Number of replicates at the appropriate scale
Individuals	Tanks	10 × Control treatment 10 × MHW treatment

2.3 Heatwave simulation

A 35-year dataset (11/11/1983 to 11/11/2018) for sea surface temperature (SST) of Lizard Island (LI), in the Great Barrier Reef, Australia (14°40′04.2″ S, 145°27′53.0″ E) was obtained from the National Oceanic and Atmospheric Administration (NOAA). Following the methods outlined by Hobday et al. (2016), we estimated the frequency and intensity of the MHWs in LI; 75% of the duration corresponded to moderate (Category I) MHW, 15% to strong (Category II) MHW, and 10% of the time showed no heatwaves (see Figure S2).

In this study, we used Category I MHW conditions, as these represent the more typical and prevalent intensity level of MHWs in the region. The onset rate (0.14°C day⁻¹) and decline rate (0.8°C day⁻¹) were determined based on this categorization. The duration (55 days) was selected based on a heatwave occurrence in the Great Barrier Reef between 16 February and 3 April 2016. Modulation analyses were conducted using the statistical software R version 3.4 (R Core Team, 2017).

MHWs are defined as periods of at least 5 days during which SST exceeds the climatological 90th percentile threshold based on a long baseline period (Hobday et al., 2016). The estimated SST on Lizard Island varied between 24.2°C in winter and 29°C in summer. The 90th percentile threshold ranged from 24.8°C to 29.8°C in winter and summer, respectively.

The MHW treatment was initiated after the 5-day acclimation period. The target temperatures of the treatments were 28.8°C (control; n = 10) and 29.8°C (MHWI; n = 10). The MHW was initiated by gradually increasing the temperature over a period of 6 days (ramp period) to reach the initial temperature of the treatment. After 55 days of exposure, the temperature was lowered to control temperatures, and cleaners started a recovery period of 30 days, after which they were euthanised for brain removal (see Figure S3).

2.4 Behavioural trials: Learning task

All cleaners were subjected to a learning task during the last week of the MHW exposure period and again after a 14-day recovery period. The experimental framework employed in our study involved a visual association task, requiring fish to learn to locate a food reward based on colour cues. The experimental design used was a ‘simple market’ version adapted from the biological market theory experiment first introduced by Bshary and Grutter (2002).

In our setup (Figure 1), cleaners were tested in their individual holding aquariums partitioned into holding (approx. 1/3 of tank length) and experimental compartments (approx. 2/3 of tank length) using an opaque barrier. Fish were initially directed to the holding compartment (Figure 1a). Within the experimental compartment, two Plexiglas plates (8 cm × 4 cm) were positioned, differing in colour of decorative stripes. The colours of the plates varied between the last week of exposure (orange and grey) and after the 14-day recovery period (pink and brown). One colour plate featured an accessible food reward (correct plate), while the other colour plate offered an inaccessible food reward (covered with tape; incorrect plate). The colours of the plates also varied between individuals within the same treatment. For example, for half the fish in the control treatment, the correct colour plate was orange, while for the other half it was grey. Needle-punctured holes in the tape ensured that the smell cue was identical between plates, and food was smeared on the back of the plates to remain unseen by cleaners during the selection process. The role of each plate (correct or incorrect) was pre-defined and counterbalanced between individuals within each treatment, and plate positions (left or right) were systematically counterbalanced over 10 successive trials. A vertical Plexiglas partition between the plates ensured that fish could access only one plate, allowing the experimenter to determine when a definitive choice had been made.

FIGURE 1

Schematic representation of the colour discrimination task setup. (a) Fish were initially directed into a holding compartment using an opaque barrier. (b) The barrier was lifted, allowing access to the experimental compartment and two different plates. (c) A correct choice allowed both plates to remain in the aquarium, enabling access to a food reward. (d) An incorrect choice resulted in the immediate removal of the correct plate, preventing access to the reward. Illustration by B. P. Pereira.

At the onset of each experimental trial, the partition was lifted, granting the cleaner the freedom to forage and access the plates (Figure 1b). The moment the tip of the fish's snout passed the threshold of the vertical Plexiglas partition separating the plates, the plate choice was recorded. A correct choice allowed both plates to stay inside the experimental tank (Figure 2c), while an incorrect choice led to the immediate removal of the correct plate, thereby preventing cleaners from accessing the food item (Figure 1d). The incorrect plate remained in the tank for 2 min to allow exploration. After each trial, the cleaner was returned to the holding compartment, and the setup was prepared for the next trial. The inter-trial interval averaged 1 h and never exceeded 2 h. Cleaners underwent testing 20 times per day, divided into two sessions of 10 trials, with a maximum of 100 trials (10 sessions). The task was solved when an individual chose the correct plate at least 9/10 times within a single session, 8/10 times in two consecutive sessions, or 7/10 times in three consecutive sessions. Failure to meet this criterion in a total of 100 trials categorized individuals as unsuccessful in the task.

FIGURE 2

Cleaner wrasse cognitive performance under marine heatwave exposure (a, b) and after a 14-day recovery period (c, d). Graphs (a) and (c) depict the percentage of successful individuals solving the task through the number of trials. Dot plots (b) and (d) illustrate the number of trials required by each fish (individual dots) to complete the task per treatment; dots above the dashed line represent fish that failed to complete the task in 100 trials.

2.5 Brain development

Following the 30-day recovery period, all cleaners underwent euthanasia using a lethal dose of MS-222 (250 mg/L), followed by spinal cord sectioning. Subsequently, standard length (SL) and weight were recorded for each fish, and their brains were dissected into five macro-regions (telencephalon, diencephalon, midbrain, cerebellum and brainstem; Figure S4). The whole brain as well as each individual region was then weighed with a precision scale (accuracy of 0.00001 g).

2.6 Statistical analysis

All analysis was performed using R, version 3.4.3 (RStudio Team, 2022). Data exploration and model validation used the HighstatLibV10 R library from Highland Statistics (Zuur et al., 2009, 2010). Specifically, this process involved checking for collinearity, outliers and heteroscedasticity.

To examine the learning abilities of cleaners during and after exposure to heatwaves, we conducted survival analyses, given the time-to-event nature of our data. To estimate the survival function, Kaplan–Meier estimates were used (‘surfit’), allowing for the understanding of the probability of completing the cognitive task over time under different conditions. A log-rank test (‘survdiff’) was performed on these estimates to compare the survival curves from the two treatments. This test helped ascertain whether there were significant differences in cognitive task completion rates between the two treatments. Survival curves were plotted using the ‘ggsurvplot’ function from the R package ‘survminer’ (Kassambra & Kosinski, 2018). To assess the impact of heatwave exposure on learning abilities and post-recovery cognition, including the potential confounding effect of colour perception, we employed Cox proportional hazards models (Figure S5). Model fit and assumptions, namely, the proportional hazards assumption were done using the ‘coxph’ function and ‘ggcoxdiagnostics’, respectively, from the R package ‘survival’ (Therneau & Grambsch, 2000).

To assess the impact of heatwave exposure on brain development, we used weighted linear models (to include the effect of body weight or brain weight) using the ‘glm’ function from the ‘stats’ package (Fox & Weisberg, 2011). These models analysed differences in brain–body mass ratio and ratios of specific brain regions (telencephalon, optic tectum, diencephalon, cerebellum and brain stem) relative to total brain mass between control and heatwave conditions. Each ratio served as our response variable, while treatment groups (control vs. heatwave) were considered predictor variables. Two individuals from the control group were excluded from the analysis of the brain–body weight ratio because their body weight was considerably lower than that of the other participants (1.9 and 1.63 g), which could have skewed the overall results. The average body mass for the control group, excluding these two fish, was 2.99 ± 0.40 g. However, when these underweight individuals were included, the average body mass dropped to 2.75 ± 0.63 g, indicating greater variability within the group. This discrepancy suggests that these fish might have experienced feeding issues prior to the experiment, leading to disproportionately low body weight. Analysis of variance was conducted using the function ‘Anova’ from the package ‘car’ (Fox & Weisberg, 2011). Assumptions of homogeneity of variances and normality of residuals were verified for each model (Figures S7–S11) using the function ‘check_model’ from the package ‘performance’ (Lüdecke et al., 2021).

2.7 Ethical note

This study was conducted under the approval of Faculdade de Ciências da Universidade de Lisboa, Animal Welfare Body (ORBEA—Statement 01/2017) and Direção-Geral de Alimentação e Veterinária (DGAV Permit 2018-05-23-010275) in accordance with the requirements imposed by Directive 2010/63/EU of the European Parliament and of the Council of 22 September 2010 on the protection of animals used for scientific purposes.

3 RESULTS

3.1 Cognitive test

Both heatwave-exposed and control fish actively participated in the visual discrimination task, demonstrating a strong interest in obtaining a food reward throughout the trials.

During the exposure period, we observed a significant impact of the MHW treatment on the cleaners' performance in the visual discrimination task (Figure 2a,b; χ² = 14.9; df = 1; p < 0.001). Under control conditions, all cleaners successfully completed the task, with nine out of 10 fish meeting the criteria within 40 trials. In contrast, in the MHW treatment, only six out of 10 cleaners completed the task within the 100-trial limit, taking an extended duration to meet the criteria, ranging from 50 to 80 trials.

However, following the cessation of heatwave exposure and a 14-day recovery period, task performance of the MHW treatment group was identical to that observed for control individuals (Figure 2c,d; χ² = 0.1; df = 1; p = 0.8).

3.2 Brain development

Brain/body ratio was higher in individuals subjected to the MHW treatment group than in the control group (Figure 3a; χ² = 21.2, df = 1, p < 0.001).

FIGURE 3

Brain development of cleaner wrasses Labroides dimidiatus in two treatments. Plots illustrate variations in (a) brain-to-body mass ratio and (b) specific brain regions-to-brain mass ratios (telencephalon-brain ratio; optic tetum-brain ratio; diencephalon-brain ratio; cerebellum-brain ratio; brain stem-brain ratio), among cleaners exposed to control and heatwave conditions. Each data point represents an individual cleaner's ratio.

Analysis of the brain's macro-regions revealed further differences: individuals exposed to the MHW treatment exhibited a significantly reduced telencephalon/brain ratio (Figure 3b; χ² = 10.2, df = 1, p = 0.001) and a significantly higher brain stem/body ratio (Figure 3b; χ² = 4.2, df = 1, p = 0.04). No significant differences in the optic tectum/body ratio (Figure 3b; χ² = 0.8, df = 1, p = 0.36), diencephalon/body ratio (Figure 3b; χ² = 0.338, df = 1, p = 0.561), or cerebellum/body ratio (Figure 3b; χ² = 1.5, df = 1, p = 0.22) were observed between the MHW treatment and control groups.

4 DISCUSSION

In this study, we examined the effects of a simulated Category 1 marine heatwave on the cognitive performance of the cleaner wrasse L. dimidiatus. After subjecting the fish to elevated temperatures for 55 days, we observed a significant impact on their ability to solve a visual discrimination task. Concurrently, we noted changes in brain density among the fish acclimated to warmer conditions. Although cognitive impairments were found to be temporary—cleaner wrasse were able to recover their cognitive abilities within 14 days—a 30-day recovery period was insufficient to fully reverse the observed morphological alterations, suggesting a complex interplay between thermal stress, cognitive function and neurobiological adaptations.

Cognition encompasses a series of neural mechanisms through which animals make behavioural decisions (Shettleworth, 2010). After acquiring information from the environment through sensory systems, individuals employ attentional mechanisms to select pertinent stimuli, such as identifying different client types in the case of cleaner wrasses. Subsequently, they utilize learning mechanisms to form associations and predictions, like recognizing which clients offer the best opportunities for food. Memory mechanisms then store this acquired information, allowing individuals to remember past interactions with clients and adjust their behaviour accordingly. Finally, decision-making mechanisms come into play, where individuals weigh various factors to determine appropriate responses, such as whether to prioritize the cleaning of a particular client. These decisions may require inhibitory control; for example, refraining from engaging in risky behaviours like cheating with certain clients or in certain contexts. While evidence suggests that warming may impact the cognition of ectotherms, the underlying mechanisms remain poorly understood.

Our study unveiled a significant impact of moderate warming on the associative learning capacity of cleaners, as assessed through the visual discrimination task (Figure 2). Fish exposed to the MHW exhibited a lower success rate and prolonged task completion time compared with cleaners in the control treatment. This lower cognitive performance was in line with the effects observed in the wild, using a more complex associative learning task (biological market task—BM) suggested to be linked to conditions of low cleaner density and low cleaner-to-client ratios (Triki et al., 2018). Given that all cleaners in this study originated from the same location and generally performed well under control conditions, our results cannot be attributed to a lack of social complexity.

Similar to the findings of Paula et al. (2019), where cleaners failed to solve the BM task under high CO₂ levels, the observed discrepancies in this study likely stem from the effects of environmental stress through different pathways, either interconnected or occurring independently. Increased temperature may have directly impaired the neural function and neurotransmitter activity of cleaners, potentially leading to alterations in cognitive performance (Van Hook, 2020). A study examining the neuromolecular mechanisms underlying the behavioural disruption of cleaning interactions in response to projected warming conditions (+3°C) revealed transcriptional reprogramming, primarily linked to stress, heat shock proteins, hypoxia and behaviour (Ramírez-Calero et al., 2023). Similarly, Toni et al. (2019) demonstrated that zebrafish exposed to elevated temperatures (+2°C) for 21 days exhibited reduced synaptic protein production and neurotransmitter function, ultimately resulting in impaired learning.

Additionally, the observed prolonged task completion in warm-acclimated cleaners suggests a heightened effort to integrate new information and extract relevant patterns. Processing of thermal information by the brain may disrupt attention and exploration of other competing stimuli, leading to a decrease in overall cognitive functions such as learning, memory, or decision-making. Consistent with our findings, Silveira et al. (2023) demonstrated that damselfish exposed to moderate and high temperatures exhibited reduced learning efficiency, memory and accuracy in cognitive tasks.

Finally, the disruption of cognitive processes under increased temperature could be associated with trade-offs in energy allocation. It is known that fish energetic demands (i.e. metabolic rate) predictably increase with water temperatures (Abram et al., 2017; Alfonso et al., 2021). To sustain their basic metabolic needs under warmer conditions, fish may limit the availability of energy for non-essential functions such as cognition. While enhanced cognitive abilities could potentially benefit cleaners in acquiring food, the metabolic demands associated with elevated temperatures likely outweigh the benefits gained from increased foraging efficiency.

In contrast to prior research indicating a positive correlation between brain size and cognitive ability (Kotrschal et al., 2013a; Marhounová et al., 2019), our investigation revealed a paradoxical finding: warm-acclimated fish that exhibited diminished cognitive performance had a higher brain-to-body ratio. This discrepancy suggests that the increase in brain volume associated with thermal compensation did not translate into an expansion of functional brain tissue, as further evidenced by the disproportionately smaller telencephalon – the brain region that controls key cognitive functions, such as memory, spatial memory, fear conditioning and decision-making (López et al., 2000; O'Connell & Hofmann, 2011; Portavella et al., 2002; Salas, Broglio, et al., 1996; Salas, Rodríguez, et al., 1996; Triki et al., 2021). Since the brain is the most energetically costly organ to develop and sustain (Aiello & Wheeler, 1995; Kotrschal et al., 2013b), energetic limitations might lead to greater development of brain regions that are crucial for a specific context (Gonda et al., 2013; Pike et al., 2018; Závorka et al., 2020). Therefore, this reduction in telencephalon size may reflect a reallocation of resources away from higher cognitive functions towards maintaining essential physiological processes in response to environmental stressors. However, the specific anatomical factors responsible for the differences in telencephalon size among the sampled fish remain unclear. The decrease in telencephalon size could result from a reduction in the number of neurons (Marhounová et al., 2019), alterations in connectivity (Triki et al., 2020) or a combination of both. We also observed an enlarged brainstem in warm-acclimated fish, which may be linked to the vital role of this region in regulating essential physiological functions, such as respiratory control, osmoregulation, sensory integration, motor coordination and autonomic functions, all essential for thriving in warmer environments, which may be particularly crucial under conditions of environmental warming (Helfman et al., 2009). For example, the enlargement of the brainstem could represent an adaptive response aimed at optimizing oxygen uptake, thereby supporting increased metabolic demands in warm-acclimated individuals (Portner & Knust, 2007). Further research elucidating the functional significance of these morphological alterations is warranted to unravel the complex interplay between thermal acclimation, brain morphology, and cognitive performance in ectothermic organisms.

Interestingly, despite the rapid restoration of cleaners' cognitive functions only 14 days after the cessation of heatwave conditions, the morphological alterations in the brain endured for at least 30 days. It underscores the remarkable dynamic capacity for neural plasticity in fish brains throughout adulthood, a feature distinct from higher vertebrates (Gould et al., 1999; Hastings et al., 2000). This adaptive response probably enabled cleaners to swiftly regain their cognitive abilities after the cessation of the stressor, despite the sustained changes in brain structures. This persistence of morphological alteration could indicate a continued response to prolonged thermal stress on neurobiological structures or a slower process of morphological reorganization or adaptation. Noteworthy, there were some variations in task performance between the two time points in the control group, where the number of successful individuals and trials required differed. These variations can be attributed to the fact that during the recovery period, individuals had to unlearn the initial colour association and learn a new one. This transition likely made the task more challenging, leading to a slight increase in the number of unsuccessful individuals and the number of trials required to complete the task. The introduction of new colours added complexity to the task, accounting for these differences in performance.

In conclusion, this study sheds light on the interaction between thermal stress, cognitive function, and neurobiological adaptations in cleaner wrasse. Our findings suggest that when exposed to a moderate yet prolonged marine heatwave, cleaners may undergo thermal regulation processes that impair cognitive functions and lead to changes in brain morphology. While acknowledging the potential influence of various ecological factors, such as market conditions and habitat destruction, we confirm our initial hypothesis that the loss of strategic sophistication following environmental perturbations extends beyond social dynamics. If that is the case, a mere recovery of fish densities may not necessarily lead to a restoration of cognitive performance.

Although in our study, we observed transient cognitive impairments, it is also important to consider that cognitive function is multifaceted and recovery in one aspect does not necessarily imply complete restoration of all cognitive processes, especially considering the observed alterations in brain structures. If colour association is a relatively simple task for cleaners, it prompts the question of what the consequences for more complex ones are. For instance, it is known that cleaner fish can recognize the faces of familiar and unfamiliar individuals in social groups, adjusting their behaviour accordingly (Kohda et al., 2023). This ability likely relies on the telencephalon, which, as aforementioned, is crucial for processing complex cognitive information like long-term memory and decision-making. If thermal stress impaired the telencephalon's functionality, it could disrupt these and other sophisticated social techniques, potentially affecting the fish's ability to maintain social hierarchies, engage effectively in mutualistic interactions, or adapt to changes in social dynamics. For example, the reduced ability to identify clients could result in cleaning inefficiency and decreased benefits for clients. It could also lead to increased predation risk if cleaners fail to distinguish or properly manage predatory clients and engage in opportunistic behaviours.

Supporting this idea but, at the same time, controversial to our results, an artificial selection study by Triki et al. (2023) found that an enlarged telencephalon in guppies was positively associated with enhanced performance in key executive functions—cognitive flexibility, inhibitory control and working memory—yet did not significantly impact simpler associative learning tasks, such as colour discrimination. These findings are consistent with earlier lesion and ablation studies showing that fish without a functioning telencephalon can still form basic associations but struggle with more complex tasks (López et al., 2000; Overmeir & Hollis, 1983; Overmeir & Papini, 1986; Salas, Broglio, et al., 1996; Salas, Rodríguez, et al., 1996; Savage, 1980). In our study, even basic cognitive abilities were impaired under warming conditions, suggesting that thermal stress may have broader impacts on cognition, possibly affecting multiple neural pathways and cognitive processes simultaneously.

If the magnitude of heat-induced cognitive impairment primarily depends on task complexity (Hancock & Vasmatzidis, 2003) and the time it takes for cognitive abilities to return to baseline is influenced by the severity of the initial impairment, it is possible that the relatively simple task employed in our study underestimated the long-term consequences of MHW on cleaner wrasse cognition. More complex cognitive tasks, such as those involving natural social interactions, spatial orientation, or reversal learning, may reveal more persistent cognitive deficits.

Additionally, this study also prompts consideration of the wider implications of more frequent and intense marine heatwaves, as well as potential multistressor scenarios. Further research is necessary to understand how increased water temperatures, combined with other environmental stressors, might induce permanent alterations in brain function and structure, especially considering different life stages. The developmental environment experienced by juvenile fishes has been shown to significantly influence adult cognition and behaviour (e.g. Brockmark & Johnsson, 2010; Chapman et al., 2008; Peele et al., 2023; Vila Pouca et al., 2018). Furthermore, the pronounced neuronal plasticity during developmental stages, owing to the rapid growth of the brain (Knickmeyer et al., 2008), suggests that juvenile cleaners may be particularly vulnerable to temperature effects on brain structure and function.

Nevertheless, even the short-term effects on cognition observed in the current study could disrupt cleaning behaviour, with potential fitness consequences for both cleaners and clients, and cascading effects on ecosystem dynamics. We highlight the urgency for holistic approaches to protect these crucial components of coral reef ecosystems.

AUTHOR CONTRIBUTIONS

Beatriz Pereira led the writing of the manuscript with contributions from José Ricardo Paula and Rui Rosa; José Ricardo Paula and Rui Rosa conceived the idea and designed the methodology; Beatriz Pereira, Lígia Cascalheira and José Ricardo Paula collected the data; Beatriz Pereira, Lígia Cascalheira and José Ricardo Paula analysed the data. All authors contributed critically to the drafts and gave final approval for publication.

ACKNOWLEDGEMENTS

We would like to acknowledge all members of Laboratório Marítimo da Guia who contributed with comments and help throughout this work. This work was supported by FCT—Fundação para a Ciência e Tecnologia, I.P., within the project grant PTDC/BIA-BMA/0080/2021—ChangingMoods to JRP, the scientific employment stimulus programme 2021.01030.CEECIND to JRP, the strategic project UIDB/04292/2020 granted to MARE, project LA/P/0069/2020 granted to the Associate Laboratory ARNET, and a PhD fellowship 2021.06590.BD to BPP.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

Open Research

DATA AVAILABILITY STATEMENT

Data deposited in the Dryad Digital Repository: https://doi.org/10.5061/dryad.9zw3r22rn (Pereira et al., 2025).