Social experience influences thermal sensitivity: lessons from an amphibious mangrove fish

Warming of the oceans can cause species migrations (García-Molinos et al., 2016), decline in marine biomass (Lotze et al., 2019), higher extinction rates (Wabnitz et al., 2018) as well as changes in community structure (MacNeil et al., 2010) and ecosystem functions (Petchey et al., 1999). Therefore, predicted ocean warming will have critical biotic impacts (Madeira et al., 2018), especially on ectotherms, which rely largely on their environment to thermoregulate. The physiological response of ectotherms to warming can be illustrated with the help of thermal performance curves (TPCs) (Schulte et al., 2011; Sinclair et al., 2016). TPCs model how an ectotherm's body temperature affects its instantaneous performance and define the temperature interval in which a species can survive (Krenek et al., 2011). In these theoretical curves, the animal's pejus temperature lies outside the optimal thermal range and is the temperature at which physiological processes cease to be optimal and ‘get worse’ (Pörtner, 2010). Organisms can survive temperatures beyond their pejus temperature thanks to physiological mechanisms such as the heat shock response, metabolic depression and antioxidative defense but they can only do so for a limited time (Pörtner et al., 2017). In general, the pejus temperatures of tropical marine ectotherms are closer to their critical thermal maximum (CT_max) than their temperate counterparts, making them particularly vulnerable to warming temperatures (Sunday et al., 2014; Huey et al., 2009; Krenek et al., 2012). Thus, it is crucial that we understand the factors that can affect the pejus temperature of tropical marine species to understand their thermal limits.

The impact of temperature depends not only on the degree of warming but also on the susceptibility of species to warming (Deutsch et al., 2008; Tewksbury et al., 2008). This susceptibility, defined by the width of the TPC, can be influenced by a variety of factors, including genetics (Manis and Claussen, 1986), age (Bowler and Terblanche, 2008), sex (Sornom et al., 2010), body size (Cheung et al., 2013; Peck et al., 2009), recent thermal history (Galbraith et al., 2012) and, important for this study, sociality (LeBlanc et al., 2011). For example, social signals disrupt the adaptive thermal response of wild tropical mangrove rivulus (Kryptolebias marmoratus), causing these amphibious fish to emerge from water – a marker of their pejus temperature – at higher temperatures than their isolated counterparts (Currie and Tattersall, 2018). In gregarious juvenile lake sturgeon (Acipenser fulvescens), the presence of conspecifics during acute thermal exposure causes a decrease in both endocrine and cellular stress responses (Yusishen et al., 2020). The presence of conspecifics has also been shown to modify the preferred and threshold temperatures of the black-axil chromis (Chromis atripectoralis) (Nay et al., 2021). Not only can acute social stimulation affect a fish's response to temperature but social experience from prior social stimulation may also be influential. Indeed, social experience can affect an individual's basal cortisol concentrations and neurophysiological systems and, in turn, affect behavioural traits (as seen in Neolamprologus pulcher; Antunes et al., 2021a,b). Despite the growing evidence that social stimulation affects responses to temperature in fishes, the nature of the social interactions and the mechanism underpinning these responses remain unclear.

For both behavioural and physiological reactions to changing temperature, animals first need to be able to perceive or sense temperature. We know that thermosensation depends on the activation of gated ion channels in the nervous system and the skin (Caterina et al., 1997). Most of these thermal receptors belong to the transient receptor potential (TRP) ion channel family (Castillo et al., 2018) and are highly conserved across taxa (Saito and Shingai, 2006; Saito et al., 2011). Six thermosensitive TRP channels have been identified in teleosts and two (TRPV1 and TRPV4) help to perceive ambient environmental temperature (Saito and Shingai, 2006; Saito et al., 2011). All of these thermoTRP channels have distinct temperature intervals at which they are activated (Islas, 2017). Although TRPV1 is the best studied thermoTRP channel, we have limited knowledge of its mechanism of activation in any organism. In addition, its function varies across species, which means we must generalize cautiously (Caterina et al., 1997; Saito and Shingai, 2006; García-Ávila and Islas, 2019). It is thought that TRPV1 is activated at temperatures exceeding 42°C (Caterina et al., 2000; Caterina, 2007) but its activation temperature in fish seems to be lower, at ∼32–33°C [zebrafish (Danio rerio): Gau et al., 2013; rainbow trout (Oncorhynchus mykiss): Ashley et al., 2007), although there are limited data available. Given that the social environment affects how several fish species behaviourally and physiologically respond to temperature, we hypothesized that an individual's ability to sense temperature could be influenced by its social environment. If socially modulated thermal risk could be explained by variation in thermosensation, then changes in the social environment would result in differences in thermal sensitivity.

We used isogenic lineages of the amphibious, tropical mangrove rivulus in different social situations to determine the possible effects of sociality on thermal biology. The mangrove rivulus is a simultaneous hermaphrodite fish and is capable of producing highly homozygous fertilized eggs (Taylor et al., 2001). Since genetic profiles and relatedness can influence how individuals interact (Nakamura et al., 2016), studying multiple isogenic lineages allowed us to investigate the effect of genetics on sociality (Hsu et al., 2008). In the wild, these fish appear to exhibit environmentally dependent social behaviour. During the wet season, these fish occupy different microhabitats including mangrove crab burrows, where they may be found alone or in small groups of ∼2–5 individuals (Kristensen, 1970; Taylor et al., 2008). Mangrove rivulus are also found dispersed in small ephemeral ponds or standing water (Taylor et al., 2008). During the dry season, these fish may densely pack together in moist logs in larger groups of up to 100 fish (Taylor et al., 2008).

Amphibious fishes emerge because of both abiotic (e.g. habitat drying, poor water quality) and biotic (e.g. feeding, reproduction, predation) factors (reviewed in Turko and Wright, 2015). Here, we capitalized on the observation that both lab (Gibson et al., 2015) and wild (Currie and Tattersall, 2018) mangrove rivulus will emerge with increasing water temperature. Given this propensity to emerge coupled with the heterogeneous nature of rivulus' social environment, we tested the hypothesis that the social environment modulates physiological and behavioural thermal responses within and between isogenic lineages of fish. We raised fish in isolation and they received either a brief 24 h period of social experience or no experience. We then exposed the fish to acute warming while socially stimulating some fish with their mirror reflection or conspecifics. Different behaviours were also measured in order to assess whether any differences in temperature response were due to aggression or affiliation towards the social stimulus. We first predicted that socially stimulated fish (fish with a mirror or conspecifics) and experienced fish would emerge from water at a higher temperature than isolated and socially naive fish, respectively. In an attempt to understand the mechanism underpinning socially influenced thermal responses, we further hypothesized that the social environment would result in a desensitization of thermoreceptors. To this end, we predicted that socially experienced fish treated with the well-known TRPV1 agonist, capsaicin, would require higher concentrations to elicit a thermal escape response (i.e. emersion) compared with socially naive fish, suggesting a desensitization of these thermal receptors.

We performed all experiments with adult mangrove rivulus hermaphrodites [Kryptolebias marmoratus (Poey 1880)] housed in a breeding colony at the Acadia University Animal Care Facility, Wolfville, NS, Canada. The three isogenic lineages in this colony, Belize 1 (50.91), Belize 2 (Dan06) and Honduras (Hon11), were originally caught in Twin Cayes, Belize in 1991, in Dangriga, Belize in 2006 and in the Bay Islands, Honduras in 1996, respectively (Tatarenkov et al., 2010). The three lineages are genetically divergent; each strain has been bred in the laboratory for more than 45 generations and has multiple loci differences (Tatarenkov et al., 2010). All fish were adults older than 3 months and reared individually in 120 ml cups of synthetic seawater made with Instant Ocean (Pets and Pond Canada) and reverse osmosis water. The cups were held in constant conditions (80 ml of 15 ppt water, 12 h:12 h light:dark cycle, 30–75% humidity). We performed water changes biweekly and maintained the colony on a thermal diel cycle of 25–28°C (Ellison et al., 2012). The minimum and maximum temperatures were reached at approximately 04:00 h and 18:00 h, respectively. We fed fish three times per week with live Artemia sp. nauplii; frozen bloodworms supplemented the fish's diet once per week. The Acadia University Animal Care Committee approved all protocols in this study (#05-19 and #02-22).

We provided both mirrors and live conspecifics to study whether or not the source of social stimulation had an effect. We use ‘social stimulation’ to describe an event in which fish respond to a perceived (mirror) or live conspecific and ‘social experience’ to describe a fish that was previously socially stimulated for 24 h.

We performed three experimental series where we tested whether: (1) social stimulation affects physiological and behavioural thermal responses of isogenic lineages of fish; (2) social experience results in distinct physiological and behavioural responses; and (3) a desensitization of thermal receptors is responsible for socially modulated thermal responses.

All fish were raised in isolation and had never interacted with conspecifics prior to the experiments. For socially experienced fish, we placed focal fish in a cylindrical container with a diameter of 11.2 cm filled with 300 ml of 15 ppt salt water. The container was divided into three equal chambers with fiberglass mesh that allowed the fish to interact visually and chemically with two other fish while preventing physical interaction. Because the container was cylindrical, all three fish could equally interact with each other regardless of which chamber we placed them in. Therefore, we randomly assigned chambers to the fish. We habituated the fish to the presence of their size-matched conspecifics for 24±1 h before transferring them to the experimental container for the start of the experiment. Socially naive fish were not manipulated prior to their transfer to the experimental container.

Here, we tested the hypothesis that social stimulation alters behavioural and physiological thermal responses in three distinct laboratory lineages of mangrove rivulus (Fig. 1A). We measured the emersion temperature (T_em: temperature at which the gills of the fish emerged out of water) and the critical thermal maximum (CT_max: temperature at which the fish could not maintain dorsal-ventral equilibrium; Cox, 1974; Beitinger et al., 2000; Gibson et al., 2015)_. We interpret T_em as the pejus temperature or T_pej, the water temperature beyond the optimal temperature (T_opt) when performance ‘gets worse’ and the fish escapes (Pörtner, 2010).

We measured T_em and CT_max of socially naive fish in isolation (control) and with a mirror from all three lineages (Fig. 1; Belize 1: naive control, n=16 and naive mirror, n=20; Belize 2: naive control, n=13 and naive mirror n=12; Honduras: naive control, n=20 and naive mirror, n=20). We analysed the behaviour of the fish post hoc using video recordings (see ‘Analysis’).

Unlike wild mangrove rivulus (Currie and Tattersall, 2018), laboratory lineages did not differ in emergence behaviour in the presence or absence of a mirror (see series 1 results). Although not a schooling fish, wild mangrove rivulus would not necessarily be socially naive; thus, we hypothesized that even brief previous social experience would affect the response to temperature. To address this second hypothesis, we used one lineage (Honduras) and compared socially naive and socially experienced fish both isolated during the acute thermal stress (naive n=20; experienced n=19; series 2a). We then compared socially naive and socially experienced fish that were stimulated by a mirror during the acute thermal stress (Fig. 1B; naive n=20; experienced n=22; series 2b). The naive fish used in series 2 are the same as in series 1. We conducted the series 2 experiments 2 months after series 1. Based on our past studies and given that all fish were adults (i.e. over 3 months old) and of similar size and age, we are confident that the age difference is not a confounding variable. In a third experiment, we replaced the mirror with two live size-matched conspecifics (series 2c). We performed this experiment in a different chamber from series 2a and series 2b owing to the addition of the two conspecifics (Fig. 1). We measured T_em in all three experiments and measured CT_max in series 2a and 2b. Since CT_max did not differ between socially naive and socially experienced fish in series 2a and 2b, we did not measure it in series 2c.

Here, we measured the emersion of socially naive and experienced fish but instead of increasing the temperature, we used increasing concentrations of the specific TRPV1 agonist, capsaicin (Caterina et al., 1997) to elicit chemical warming. We used fish from the Honduras lineage for these experiments. We measured the emersion concentration (C_em: the concentration of capsaicin at which the gills of the fish emerged out of water), the number of times the fish attempted to emerge from the water as well as the critical concentration maximum (CC_max: the concentration of capsaicin at which the fish loses equilibrium), if and when it occurred. We made these measurements on both socially naive and experienced fish (Fig. 1C) and used control groups that were treated with water or 0.1% DMSO (capsaicin vehicle) (naive water, n=10; naive DMSO, n=10; naive capsaicin, n=13; experienced water, n=10; experienced DMSO, n=10; experienced capsaicin, n=13). Capsaicin, water and DMSO were gradually added to one of two corners of the chamber depending on which was furthest from the fish at that moment.

We performed all experiments between 12:00 h and 17:00 h and recorded them using a camcorder (Canon VIXIA HF200) for post hoc behavioural analyses. Whenever we removed a fish from the experimental container after determination of T_em or CT_max, we returned it to its individual cup which held 80 ml of 15 ppt water at 32–34°C and then immediately placed in a water bath at 27°C until their return to the colony (within 2 h). We measured the length (to the nearest mm) and mass (to the nearest mg) of fish at least 24 h after they were subjected to a test. When we used focal fish in more than one experiment, the time interval between experiments was at least 4 months.

The T_em experiments started with a 1 h familiarization in the new chamber. During this time, we kept the chamber in a 25–28°C water bath and then heated the chamber at a rate of 1°C min⁻¹ in a water bath (VWR, model: WBE05) containing 2 l water. The experiment ended when the gills of the fish emerged from the water (T_em) (Gibson et al., 2015; Currie and Tattersall, 2018). In all T_em measurements, we used a steel temperature probe connected to a Lab Pro (Vernier Software & Technology) to measure the water temperature to the tenth of a degree every second.

For T_em measurements in which the fish was isolated or had access to a mirror, we placed the focal fish in one of four identical and visually isolated experimental chambers (L×W×H: 8 cm× 4 cm×7 cm) filled with 150 ml of 15 ppt water. If the experiment required social stimulation, we removed an opaque barrier against the long side of the chamber to reveal either a mirror (series 1 and series 2b) or two live conspecifics (series 2c). We revealed the mirror at the beginning of heating and revealed the two stimulus fish 10 min before the start of the heating; this ensured that the focal fish perceived them and was socially stimulated throughout the acute warming. A fiberglass mesh barrier allowed focal and stimulus fish to interact visually and chemically but not physically. The focal fish were unfamiliar to the two stimulus fish and all fish were of the same lineage. To closely observe the focal fish, we carefully removed both stimulus fish from the experimental container when the water reached 40°C (close to CT_max) and returned them to their individual cups. When we allowed the fish to interact with two live conspecifics, the focal fish chamber (L×W×H: 8 cm× 4 cm×7 cm) held 150 ml of 15 ppt salt water while the two stimulus fish chambers (each 4 cm× 4 cm×7 cm) on the other side of the fiberglass mesh held 75 ml each of 15 ppt salt water.

We measured CT_max in a way similar to T_em, but with a few modifications. We filled the chamber used in series 1, series 2a and series 2b with 190 ml water and covered it with cellophane to prevent the fish from emerging as the water warmed. We ended the experiment when the fish could not maintain dorsal–ventral equilibrium (Currie and Tattersall, 2018). We measured CT_max on the same fish as we measured T_em, but performed the second measurement 48 h after the first. To evaluate whether the order of testing had an impact on the results, we alternated the order in which each test was performed.

At the beginning of each experiment, we placed the fish in one of two identical and visually isolated experimental chambers (L×W×H: 8 cm× 4 cm×7 cm) filled with 150 ml of 15 ppt water and surrounded by a 2 cm wide ledge and a 10 cm high wall. We placed a stir bar at the bottom of each chamber, which was separated from the fish by a 1.3 cm tall fiberglass mesh insert. We set the stirring speed so that the water in the chamber was mixed homogeneously in approximately 20 s. We kept the chamber at 26.5±2.5°C with one of two stirring hot plates (Thermolyne, model: SP18425; Fisher Scientific, model: SPN105082). After a 1 h familiarization period with the chamber, we chemically heated the water at a rate of 20 μmol l⁻¹ min⁻¹ by gently pipetting 480 μl of 3.2 mmol l⁻¹ capsaicin at one of the two corners of the experimental chamber, depending on the location of the fish. The capsaicin solution was added every 30 s to a maximum of 200 μmol l⁻¹ capsaicin (similarly to Endo et al., 2020 in zebrafish). Since few studies have exposed fish to waterborne capsaicin, we performed preliminary experiments to determine the capsaicin concentration that would cause emersion (200 μmol l⁻¹), but not loss of equilibrium or mortality (Fig. S1). The experiment ended when the fish reached CC_max or 15 min after the start of the experiment, whichever came first. We prepared a stock solution of 195 mmol l⁻¹ capsaicin in DMSO and the working solution of 3.2 mmol l⁻¹ capsaicin was prepared daily by diluting the stock solution in 15 ppt water. We repeated the same protocol with 0.1% DMSO in 15 ppt water instead of a capsaicin solution in the control experiments.

We tracked behaviour through post hoc video analysis of 89 experiments in series 1 and 2 (control naive, n=13; control experienced, n=15; mirror naive, n=17; mirror experienced, n=15; fish naive, n=14; fish experienced, n=14) using EthoVision XT v. 15.0 (Noldus Information Technology). The experiments were analysed from start to finish (∼15 min). We recorded six behaviours: (i) the latency to resume normal swimming activity after transfer to experimental chamber, (ii) latency to first approach the interaction zone (defined as one body length of the individual, starting at the barrier between the social stimulus and the focal fish), (iii) total time spent in the interaction zone (association time), (iv) time spent swimming, (v) time in contact with the social stimulus barrier (Li et al., 2018; Edenbrow and Croft, 2013) and (vi) the frequency of aggressive approaches towards the social stimulus (fish or mirror). To measure behaviour elicited by the social stimulus alone, without warming, we visually analysed 5 min at the start of each experiment starting after 1 min to ensure that the fish had resumed normal activity. We manually measured aggressive approaches and defined aggression as attempted bites and mouth wrestling against the social stimulus barrier, swimming in bursts along or towards the social stimulus barrier and lateral displays performed parallel to the social stimulus barrier (control naive, n=13; control experienced, n=15; mirror naive, n=16; mirror experienced, n=15; fish naive, n=14; fish experienced, n=13). Frequency of aggressive approaches is included in association time given that we measured time spent in the interaction zone (i.e. association time) independently of behaviour. We analysed these six behavioural markers of aggression and activity to possibly explain differences between socially stimulated fish and isolated fish as well as between socially experienced and socially naive fish. We also compared behavioural markers of fish with different sources of social stimulation (i.e. mirror versus live conspecifics).

We performed all statistical analyses using R (https://www.r-project.org/) and GraphPad Prism 9.2.0 for series 3, all with a critical alpha value of 0.05. We used Shapiro–Wilk and Bartlett's tests to test the assumptions of normality and homoscedasticity, respectively.

In series 1, the effects of lineage and treatment (control or mirror) and their interaction on T_em and CT_max were evaluated using linear mixed-effects models with the nlme package (v. 3.1-162; https://CRAN.R-project.org/package=nlme). Group (experienced or naive) and treatment (control or mirror) and their interaction term were defined as fixed effects and condition index, mass or length as random effects.

In series 2, T_em of naive and experienced fish from series 2a and 2b were compared using linear mixed effects models with or without mirror and prior social experience or not as fixed effects, and mass, length or condition index as random effects. The effect of conspecifics on emersion temperature (series 2c) was tested similarly, with social experience as fixed effect and mass, length or condition index as random effects. We performed two-way ANOVAs to compare the CT_max of naive and experienced fish from series 2a and 2b. The independent variables of the two-way ANOVAs were social experience and treatment (control or mirror). The residuals of series 2 two-way ANOVA respected the assumption of normality but deviated slightly from the assumption of homoscedasticity. We judged the test to be acceptable despite the modest deviation from homoscedasticity. Finally, to analyse the effects of group and treatment on the aggressive approach frequency, we used a zero-inflated model to account for excessive zeros in the count data. The model was fitted using the ‘zeroinfl’ function in the pscl package (https://CRAN.R-project.org/package=pscl) in R. We used t-tests to compare association time between socially naive and socially experienced fish. We computed linear best fit to study the relationship between association time and T_em, and to analyse the relationship between aggressive approach frequency and T_em of socially naive and socially experienced fish.

To compare the capsaicin response of socially naive and socially experienced fish in series 3, we used dose-response curves and compared the half maximal effective concentration (EC₅₀) of capsaicin of socially naive and socially experienced fish using a modified extra-sum-of-squares F-test.

In socially naive fish from three isogenic lineages, the social stimulus provided by the mirror did not significantly affect T_em when compared with values in fish with no mirror (Fig. S2). The social stimulus provided by the mirror also had no significant effect on CT_max (Table 1).

Given that each lineage is isogenic, we were interested in the intraspecific variation in each of our conditions as a gauge of trait plasticity. When comparing across lineages, we found no significant differences in T_em or CT_max (Table S1, S2). Mass and length of fish did not influence T_em or CT_max. However, Fulton's condition factor (k) did affect CT_max; its effect varied between social stimuli (control or mirror) and among lineages. In Belize 1 and Belize 2 fish, CT_max was positively correlated to the condition factor while this was not the case in the Honduras lineage. In addition, the variability of T_em or CT_max did not differ among lineages. Within lineages, only Honduras fish showed higher variability when isolated compared with fish faced with a mirror.

As with the socially naive mangrove rivulus, the T_em and CT_max of socially experienced mangrove rivulus did not differ whether they had access to a mirror or not (Table 1). However, T_em was significantly higher in socially experienced mangrove rivulus compared with socially naive fish. This was only true during control experiments when fish were not socially stimulated by mirrors (Fig. 2; P=0.021). Even though mirror reflections masked the effect of social experience on T_em, we did not observe a significant interaction between social experience and social stimulation (P=0.145). Furthermore, when we provided a social stimulus in the form of live conspecifics during the thermal stress, the effect of social experience on the T_em observed with control fish also disappeared (Fig. 3; P=0.268).

Despite differences in T_em, there was no significant difference between the CT_max of socially naive fish and socially experienced fish (Table 1). We did not detect any effects of previous social experience (P=0.736) or the presence of the mirror (P=0.743) on CT_max (not shown). Variability of T_em and CT_max did not differ between socially naive and experienced fish except for T_em of fish with mirror reflections. It is noteworthy that some of the T_em measurements were beyond the CT_max of the fish, whether they were socially naive, socially experienced, isolated or faced with a mirror. This was the case for all three lineages.

We compared six behavioural markers of aggression and association between control, mirror and fish treatment groups of the Honduras strain in series 2. Since aggression and sociality are not always correlated (Lacasse and Aubin-Horth, 2014), it was important not to exclusively record markers of aggression. The association time for focal fish did not differ between socially naive and socially experienced fish when they were isolated during the experiment (Fig. 4A) or when they had access to a mirror (Fig. 4B). However, when the fish were with live conspecifics during the experiment, socially naive fish spent significantly more time interacting with them than socially experienced fish (Fig. 4C; P=0.0285). The association time was not significantly correlated with T_em in any experimental group except for socially experienced fish faced with a mirror (Fig. S3). Aggressive approach frequency differed between isolated fish (control) and fish with mirrors (mirror) showing that mirrors and blank walls were perceived differently. The frequency of aggressive approaches was lower in socially naive fish (Fig. 5; P<0.001) and the presence of a mirror amplified the influence of prior social experience (Fig. 5; P=0.012). However, this effect disappeared when fish had the opportunity to interact with conspecifics (Fig. S4; P=0.634). Aggression was not correlated with T_em, whether the fish were socially naive, experienced or had social stimulation. While we observed differences in aggression in socially experienced fish with the mirror compared with the control fish (Fig. S5), the mirror caused the difference in T_em between socially naive and socially experienced fish to disappear (Fig. 2).

We then tested if the desensitization of TRPV1 could provide a mechanism to explain how social experience influences the thermal response of mangrove rivulus. By specifically activating TRPV1 receptors with capsaicin, we were able to test whether or not the difference we observed in T_em in different social contexts was because of different thermal sensing capacities. Because we observed a significant correlation between fish mass and capsaicin concentration at first emersion for socially naive fish, we normalized the capsaicin concentration per mg of fish (Fig. 6; Fig. S5). Socially experienced fish emerged at significantly higher capsaicin concentrations (Fig. 6; naive, EC₅₀=2.921 μmol l⁻¹ mg⁻¹, 95% CI=2.632–3.257 μmol l⁻¹ mg⁻¹; experienced, EC₅₀=5.286 μmol l⁻¹ mg⁻¹, 95% CI=4.101–6.808 μmol l⁻¹ mg⁻¹; P<0.0001) compared with naive fish. This difference between socially naive and experienced fish was true whether or not capsaicin concentration was normalized to the mass of the fish.

Whether fish were socially naive or experienced, fish exposed only to water or DMSO emerged significantly less than fish exposed to capsaicin and never reached CC_max (Fig. S1). In fact, only 1 of the 40 controls emerged and it did so only once (Fig. S1). Despite emerging at significantly higher capsaicin concentrations, socially experienced fish and socially naive fish emerged at the same frequency (Fig. S1).

We tested the hypothesis that social stimulation and brief periods of social experience affect physiological and behavioural thermal responses within and between isogenic lineages of fish owing to varying thermosensing capacities. Our results indicate that acute social stimulation with a mirror reflection or with live conspecifics does not affect the thermal response in lab lineages, unlike in wild fish (Currie and Tattersall, 2018). However, fish raised in isolation and then briefly exposed to other fish (i.e. socially experienced) had an increased pejus temperature (T_em), an effect masked by acute social stimulation (mirror or live fish). We present evidence that socially experienced fish have different responses to a TRPV1 agonist compared with naive fish, suggesting that the mechanism underlying socially influenced thermal responses is linked to thermosensation. Specifically, socially experienced fish emerged at capsaicin concentrations almost twice as high as levels in socially naive fish.

We first stimulated a naive focal fish by exposing them to a mirror or two live conspecifics to the experimental chamber. We did this with three distinct isogenic lineages in an attempt to determine whether or not there is a genetic component to the response. Because the presence of a mirror delayed the emersion temperature in wild K. marmoratus (Currie and Tattersall, 2018), we predicted that our laboratory isogenic fish would exhibit the same delayed emersion with social stimuli from a mirror or with conspecifics. Surprisingly, the presence of a mirror had no effect on the T_em or CT_max in our three lineages of laboratory K. marmoratus nor did the presence of two conspecifics. The absence of significant differences in thermal response among our three isogenic lineages suggests that the socially influenced thermal responses we were studying are not driven by genetics. Because of this, we performed all subsequent experiments with a single lineage (Honduras). It is worth noting that, unlike our study, Currie and Tattersall (2018) used a paired design where each fish was tested with and without a mirror. However, the most likely reason for the differences between wild and laboratory fish is the source of the fish. Unlike our laboratory lineages, wild fish are not socially naive.

It is well established that experiments on wild and laboratory animals do not always yield similar results, especially in behavioural and endocrine studies (Calisi and Bentley, 2009). More specifically, laboratory fish have been shown to have different thermal and social responses than their wild counterparts. For example, laboratory-held zebrafish (Danio rerio) have reduced thermal plasticity (Morgan et al., 2022) and have significantly higher thermal tolerances (CT_max) than wild zebrafish (Morgan et al., 2019) while the opposite was true in trout (Salmo trutta, Salvelinus fontinalis and Oncorhynchus mykiss: Carline and Machung, 2001). Furthermore, domesticated fish generally have modified aggressive behaviour (Lucas et al., 2004; Jonsson and Jonsson, 2006; Campbell et al., 2015) and courtship behaviour (Fleming et al., 1996). In particular, our laboratory fish are homozygous, whereas wild fish are not (Ellison et al., 2012; Tatarenkov et al., 2010). Moreover, our K. marmoratus had also been raised in isolation and had significantly different social histories than the wild mangrove rivulus, which might also explain why social cues had different effects. Why wild mangrove rivulus responded more strongly to mirrors (Currie and Tattersall, 2018) than our laboratory fish is unknown, but it may be due to genetic differences between our laboratory lineages and the wild populations, different environmental holding conditions and/or different social histories (Mayer et al., 2011; Price, 1999).

We know that social experience can affect various behaviours, including how fish respond to temperature. For example, the social status and social stimulation of rainbow trout (O. mykiss) influenced their heat shock response and thermal tolerance (LeBlanc et al., 2011; Bard et al., 2021, respectively). Convict cichlids (Amatitlania nigrofasciata) that had previously socialized with mixed-sex groups were able to create pair bonds with new fish more frequently than fish that only had same-sex social experience (Little et al., 2017). In the same species, dyadic contests between individuals housed in isolation lead to more violent interactions (e.g. mouthwrestling and attack-bite sequences) than contests between individuals that had been housed in communal tanks (Earley et al., 2006). Given the influence of past experiences on physiology and behaviour, we thought it important to determine if socially naive and experienced fish would have different thermal responses. Interestingly, socially experienced fish had a significantly larger T_em than socially naive fish; however, there was no effect on CT_max. The latter finding may not be surprising since CT_max is a physiological marker and, although this trait is plastic (i.e. Morgan et al., 2019; Hirakawa and Salinas, 2020), physiology is often less variable than behaviour (Llewelyn et al., 2016; Telemeco et al., 2009; Huey et al., 2003). Our results are also consistent with Currie and Tattersall (2018), where K. marmoratus emersion temperatures, but not CT_max, were influenced by the presence of a mirror. Since socially experienced fish were able to withstand higher temperatures while maintaining their CT_max, they may be using physiological mechanisms that allow them to better cope with temperature elevations. If that is the case, social experience may be particularly important for species that live near their thermal maxima.

The influence of social experience on pejus temperature did not appear to be a result of aggressiveness or association time with the social stimulus. The difference in pejus temperature between socially experienced and naive fish disappeared when fish could interact with their mirror reflections or see and smell live conspecifics during thermal stress. Through video analysis, we confirmed that this finding was not due to socially experienced fish displaying increased interest or aggression towards the mirror. As with mirrors, fish did not express more aggression towards live conspecifics whether they were socially experienced or not. However, socially naive fish spent more time with conspecifics than socially experienced fish did. There are two possible competing hypotheses explaining why the effect of social experience was lost with an acute social interaction: (1) the presence of perceived conspecifics (mirror reflection or live fish) diminishes the stress response of socially experienced fish allowing them to accurately perceive temperature and emerge earlier, or (2) the presence of perceived conspecifics is stressful for the socially naive fish and the resulting stress response is preventing them from perceiving temperature accurately, thus delaying their emersion. If the second hypothesis is supported, we would expect both socially naive and experienced fish faced with conspecifics (real or perceived) to have a higher pejus temperature compared with control fish. However, this is not what we observed. Therefore, it seems more likely that the presence of perceived conspecifics is lowering the pejus temperature of socially experienced fish, supporting the first hypothesis. It is tempting to speculate that we are observing social buffering, where the presence of social partners reduces the stress response of an individual (Hennessy et al., 2009; Kiyokawa et al., 2007; Hostinar et al., 2013). Social buffering has been demonstrated in fishes (Culbert et al., 2019; Faustino et al., 2017; Yusishen et al., 2020); however, confirmation of this will require future experiments.

It should be noted that we observed similar responses in socially experienced fish regardless of whether they were exposed to their mirror reflections or conspecifics. Mirrors are commonly used as a social stimulus with vertebrates that are not capable of self-recognition, including fish, to study behaviour towards conspecifics (Li et al., 2018; Earley et al., 2000). In some fish species, including K. marmoratus, an individual's reflection can evoke greater aggression responses than live conspecifics (Baenninger, 1966; Dore et al., 1978; Earley et al., 2000). However, mangrove rivulus do not interact with mirrors in exactly the same way as they do with live conspecifics (Li et al., 2018): they spent less time interacting with mirrors, launched fewer attacks and switched orientation more frequently (Li et al., 2018). These muted social responses towards mirror reflections have been linked to differences in immediate-early gene expression revealing that mirror reflections activate the neural social decision-making network differently than live conspecifics do (Clayton et al., 2019; Weitekamp and Hofmann, 2017; Li et al., 2018). Therefore, if instead of providing a purely social stimulation as we intended, mirror reflections were perceived as being unfamiliar and frightening (Li et al., 2018), it is surprising that they masked the effect of social experience in the same way as live conspecifics. Despite mirrors eliciting different behaviour compared with live conspecifics, we did find that socially naive fish with mirrors displayed more aggression than socially naive control fish, confirming that mirrors did elicit responses from our laboratory fish.

Our overarching hypothesis was that the delayed emersion of socially experienced fish was the result of desensitized TRPV1 thermoreceptors. In support of this, socially experienced fish emerged at almost twice the capsaicin concentration as socially naive fish did, suggesting that there is likely a relationship between social experience and thermosensation through the TRPV1 channel. We know that individuals can preferentially choose being social over responding to temperature (i.e. Currie and Tattersall, 2018; Nay et al., 2021). Perhaps this trade-off is driven by modified thermosensation of TRPV1. Alternatively, it is also possible that there is no change in temperature perception but the social context leads to different responses because of ecological trade-offs with surfacing (e.g. Pineda et al., 2020). For example, an experienced fish may be more territorial and willing to risk a higher temperature and stay in water to protect this resource, whereas a naive fish may not appreciate the potential ecological risks of emersion (e.g. predation, habitat defence) and emerge at a lower temperature. It will be interesting to evaluate potential changes in risk perception with social experience in future work. That said, capsaicin is a specific TRPV1 agonist and our observation that experienced fish require higher concentrations of capsaicin to emerge points to some change in thermal perception.

If social experience impacts how fish perceive temperature, the ‘master abiotic factor’ for ectotherms (Brett, 1971), and influences their habitat use (e.g. emerging versus staying submerged), sociality could be considerably more important to survival than previously thought. To our knowledge, there are no studies linking social behaviour to TRP channel activation in any animal. Given that TRP channels are highly conserved among vertebrates (Saito et al., 2011), a relationship between sociality and TRP channels could potentially influence behaviour not only in fish, but also in other vertebrates. Notably, insects also share TRP channels with vertebrates (Wei et al., 2015) and might also exhibit social behaviour modulated by TRP channels. However, TRP channels can differ biophysically between species (see varying TRPA1 activation temperatures in rodents, reptiles and humans; Story et al., 2003; Gracheva et al., 2010; Chen et al., 2013) and so caution is required when generalizing across species (García-Ávila and Islas, 2019). It is also important to note that TRPV1 is not the only thermoreceptor in fish. TRPV4 is also sensitive to temperature (Boltana et al., 2018) and the number and types of thermoreceptors differs between classes of vertebrates (Nisembaum et al., 2015; Boltana et al., 2018). Studies specifically looking at TRP channel expression and activation in different species, including fishes, would provide important information to directly establish a link between sociality and thermosensation and help determine if this phenomenon is conserved across taxa.

In conclusion, we show that social context affects the thermal response of mangrove rivulus to high temperatures. Notably, the effect of social context depended on the source of the social stimulation and the social experience of the individual. We demonstrated that social experience influenced how fish perceived temperature causing them to expose themselves longer to critically warm water. Compared with socially naive fish, socially experienced fish emerged at significantly higher capsaicin concentrations, suggesting that the TRPV1 channels of socially experienced fish are less responsive. At this point, we do not know if the decreased thermosensation was through a desensitization of TRPV1 channel itself and/or if socially experienced fish have fewer TRPV1 channels. TRPV1 seems to be involved, but thermal sensation may be more complex given fish have multiple thermosensitive TRP channels. Since emersion behaviour of mangrove rivulus revealed that sociality affects thermal responses and that social experience may impact how fish perceive temperature, past and present social interactions should be carefully considered when studying thermal responses of aquatic ectotherms. This is especially relevant given that in the next few decades, all species – particularly those living in already warm and polar areas – will need to cope with temperatures beyond their historical experience (IPCC, 2022). Understanding how sociality impacts thermal responses in fishes is crucial to accurately predicting how aquatic animals will cope with climate warming.

We thank Dawn Miner and the Acadia University Animal Care Facility as well as undergraduate volunteers for assistance with animal care.