Cannabinoid (CB) receptors are widespread in the nervous system and influence a variety of behaviors. Weakly electric fish have been a useful model system in the study of the neural basis of behavior, but we know nothing of the role played by the CB system. Here, we determine the overall behavioral effect of a nonselective CB receptor agonist, namely Δ9-tetrahydrocannabinol (THC), in the weakly electric fish Apte-ronotus leptorhynchus. Using various behavioral paradigms involving social stimuli, we show that THC decreases locomotor behavior, as in many species, and influences communication and social behavior. Across the different experiments, we found that the propensity to emit communication signals (chirps) and seek social interactions was affected in a context-dependent manner. We explicitly tested this hypothesis by comparing the behavioral effects of THC injection in fish placed in a novel versus a familiar social and physical environment. THC-injected fish were less likely to chirp than control fish in familiar situations but not in novel ones. The tendency to be in close proximity to other fish was affected only in novel environments, with control fish clustering more than THC-injected ones. By identifying behaviors affected by CB agonists, our study can guide further comparative and neurophysiological studies of the role of the CB system using a weakly electric fish as a model.
Endocannabinoids serve as one of the main retrograde neurotransmitter in the brain [Kreitzer, 2002]. Cannabinoid (CB) receptors are prevalent throughout the nervous system [Svizenska et al., 2008] and are found in vertebrates [Elphick, 2002, 2012] including nonmammalian species (e.g., lampreys [Kettunen et al., 2005], teleost fish [Cottone et al., 2005; Lam et al., 2006], reptiles [Newman et al., 2007], and birds [Alonso-Ferrero et al., 2006; Stincic and Hyson, 2008]). The role of the CB system in neural function and its influence on behavior present many similarities across vertebrate species [Dalton et al., 2009; Elphick, 2012]. Agonism of the CB receptor 1 subtype (CB1) decreases locomotor activity in several vertebrates [Soderstrom et al., 2000; Soderstrom and Johnson, 2001; Valenti et al., 2005] in a pattern similar to that observed in mammals [Sañudo-Peña et al., 2000]. Effects on appetite [Soderstrom et al., 2004; Valenti et al., 2005], learning, and memory [Soderstrom and Johnson, 2003] and a propensity to produce communication signals [Soderstrom and Johnson, 2001] are affected in nonmammals as in mammals.
Particularly relevant to our study, social interactions in several species (e.g., zebrafish [Barba-Escobedo and Gould, 2012] and mice [Umathe et al., 2009]) are enhanced by CB agonists. The influence of CB on anxiety is well documented [Moreira and Wotjak, 2010] and CB agonists can alter the reaction to a novel environment (e.g., zebrafish [Connors et al., 2014] and mice [Haller et al., 2004]). In mice, the effect on stress/anxiety and the effects on social interaction seem to be opposite in direction, depending on the exact experimental and environmental conditions [Navarro et al., 1997; Akinshola et al., 1999; Haller et al., 2004; Krug and Clark, 2015].
These studies are among the many studies on mammals detailing the complexity of the behavioral effect of CB, but researchers are still determining which of these effects are generalizable across vertebrates and which are species-specific. Without data, we can only speculate about which effects will be observed in any given species. In this paper, we aim to detail some of the main effects of a CB agonist on the behavior of weakly electric fish, a model system that has led to important advances in our understanding of the neural basis of behavior.
Weakly electric fish have historically been a very useful model system for advancing our understanding of both sensory processing [Zakon, 2003; Maler, 2007; Allen and Marsat, 2018] and sensory-motor integration [Sawtell, 2017], and can be credited with providing some of the most detailed understanding of how specific behavioral responses can be generated by sensory-to-motor neural pathways [Heiligenberg, 1991; Rose, 2004]. This deep understanding of the neural basis of complex behavior facilitates linking specific behavioral effects of CB with their underling neural mechanisms. Furthermore, there are over 200 species of closely related Gymnotiformes [Albert and Crampton, 2005], each displaying differences in their behavior. Gymnotiformes are therefore well-suited for comparative studies that could reveal how the CB system has evolved in relation to behavior.
Gymnotiformes vary in social grouping, from gregarious to solitary [Stamper et al., 2010]. Apteronotus leptorhynchus tends to be found in groups consisting of a few individuals, in creeks and banks populated by large numbers of conspecifics. These fish will readily interact when exposed to another individual, and the nature of this interaction depends on dominance status, familiarity, physiological state, etc. As a nocturnal species, A. leptorhynchus tends to hide during the day, and is territorial regarding hiding places [Moller, 1995]. They also prefer confined spaces, although they do explore their environment.
Communication signals and behavior have been particularly well-studied behaviorally and physiologically [Walz et al., 2013]. Social interactions in weakly electric fish are largely mediated by their electrosense and the electric field that they generate. A. leptorhynchus continuously generates a weak (mV) electric field that oscillates in polarity at a frequency of 600–1,000 Hz. The base frequency of this electric organ discharge (EOD) can signal sex, maturity, and individual identification, because EOD frequency is stable in an individual [Zakon and Smith, 2009; Harvey-Girard et al., 2010].
When the electric fields of 2 nearby fish interact, the resulting perceived electrical signal has an ongoing modulation in amplitude, called a beat. The frequency of this beat allows the fish to estimate the EOD frequency of a conspecific relative to its own EOD, and remains as an ongoing signal in the background of any conspecific interactions [Hopkins and Popper, 2005]. These fish produce transient EOD modulations [Turner et al., 2007] during social interactions.
One such modulation is the well-characterized jamming avoidance response (JAR). This is elicited when 2 fish with relatively similar frequencies come into close proximity. Both fish change their baseline frequency by a few hertz, thereby avoiding a slow beat modulation that may jam their ability to electrolocate [Heiligenberg, 1991]. The main communication signals, known as chirps, are very short, and are characterized as transient increases in EOD frequency which usually last for 10–100 ms [Turner et al., 2007]. Transient EOD modulations like chirps and JAR are key components of conspecific interactions including aggressive and agnostic interactions or courtship [Henninger, 2015].
There are 2 cloned CB receptor subtypes: CB1 and CB2 [Matsuda et al., 1990; Munro et al., 1993]. Both receptors are expressed heterogeneously throughout the body, with CB1 expressed on neurons. Generally, the behavioral effects of CB are mediated by CB1, whereas the anti-inflammatory effects are mediated by CB2 [Donvito et al., 2018]. The presence of CB1 in the brain of A. leptorhynchus has been confirmed [Harvey-Girard et al., 2013], and its distribution in the nervous system presents many similarities to other teleost fish as well as mammals. Nothing is known about the role of the CB system in the physiology and behavior of Gymnotiformes. The overarching goal of this study was to make a first step towards using this powerful model, to better understand the role of the CB system in shaping the neural substrate of behavior. In doing so, our behavioral analysis reveals an intricate effect of CB on social interaction. We first characterized the overall behavioral effect of the prototypical CB receptor agonist Δ9-tetrahydrocannabinol (THC) on A. leptorhynchus behavior. The initial results allowed us to hypothesize that THC-induced behavioral alterations are dose- and context-dependent.
Fish Care and Use
Wild-caught A. leptorhynchus adult males and females were acquired from commercial suppliers. Typically, the fish were housed in 100-L tanks containing up to 10 individuals. PVC tubes were placed in the tank to provide hiding spots and environmental enrichment. The condition of the water was monitored daily and maintained in the normal range (pH 6.5–7.5; conductivity 200–300 µS; temperature 26.5–27.5°C). Water conditions in experimental tanks were identical to home tanks. Fish were acclimated in the laboratory at least 3 weeks before being used for experiments. Those from a given tank were kept together to ensure familiarity with the other individuals. Housing facilities were on an inverted 12: 12 light cycle; experiments were conducted during the active nocturnal phase of the fish. The fish to be used for the experiments were identified with a label on the tank and, when needed, individually marked with small fluorescent elastomer implants under the skin (VIE tags; Northwest Marine Technologies Inc., Shaw Island, WA, USA). We thereby ensured that a fish was not used for further experiments until it had recovered for at least 7 days, and that no behavioral, electromotor, or physical symptoms remained from the previous experiment. Fish care procedures and the experimental procedures described below were approved by WVU’s Animal Care and Use Committee.
Studies on the behavioral effects of systemic THC injection in other species helped us to delineate the appropriate dose range and testing condition to reveal the most likely effects. Concentrations up to 100 mg/kg were used, a concentration known in mice to be high but nonlethal [Varvel et al., 2005].
The fish were individually weighed prior to injection. Injection volume was 0.01 mL/g in all cases. THC was generously provided by the NIDA Drug Supply Program (Bethesda, MD, USA) and dissolved in a vehicle consisting of 5% ethanol, 5% Kolliphor EL (Sigma-Aldrich, St. Louis, MO, USA), and 90% saline [Kinsey and Cole, 2013]. THC concentration in the solution was 0 mg/mL (control), 3.3, 6.7, or 10 mg/mL, leading to injections of 0, 33, 67, or 100 mg/kg of body mass. Single-use 0.3-mL syringes with 30-gauge needles were used for intramuscular injection in the epaxial muscle. These injections were very quick with minimal disturbance to the fish. Long-term visible injuries to the skin or muscle did not occur.
Chirp Chamber Experiments
A. leptorhynchus spends a large portion of time in small burrows and root masses where there is protection from predators. In laboratory aquaria, the fish are given PVC tubes for use as hiding places. We modified one of these PVC tubes by opening the sides and replacing them with plastic mesh, in order to be able to stimulate and record from the fish while they were in the tube. A. leptorhynchus individuals readily respond to conspecific communication signals when hiding in the tube, and this paradigm is widely used to quantify the responsiveness of the fish to various stimuli [Harvey-Girard et al., 2010].
The experimental setup consisted of a 75-L aquarium containing a chirp chamber (i.e., the modified PVC tube). A stimulation dipole was placed 10 cm apart on 1 side, 20 cm away from the chirp chamber, and a recording dipole was attached to the ends of the tube. The stimulation and recording dipoles were permanently installed in a perpendicular orientation so that the recording dipole picked up the fish’s signal strongly, but not the stimulus.
Individual fish were introduced into the chirp chamber, and the opening was closed to prevent them from leaving the tube. After 5 min of acclimatization, the first set of stimuli were played and electromotor behavior was recorded. The fish was then taken out of the tank, injected, placed back in the tank and after 15–20 min, guided into the chirp chamber. After an additional 5 min, the set of stimuli was repeated. Stimuli consisted of sinusoidal signals mimicking a conspecific EOD. The frequency was adjusted relative to the fish’s own frequency to create beat amplitude modulations of specific frequencies ranging from –120 to +120 Hz. To do so, the fish’s EOD frequency was measured before each stimulation using the frequency estimate of the fish’s signal displayed by the oscilloscope (GwInstek GDS-2074A, Good Will Instrument Co., Taiwan). Some stimuli were simply constant EOD frequencies whereas others contained transient chirp modulations. These modulations were Gaussian increases in frequency of 120 Hz that lasted 14 ms, and were presented twice per second, which corresponds to the type 2 chirps used in numerous previous studies [Marsat et al., 2009]. Each stimulus type was played for 1 min with 2-min pauses in between. Stimuli were created and played with custom MATLAB (MathWorks, Natick, MA, USA) scripts controlling the computer’s generic sound card. Signals were produced at 44,100 Hz and isolated before being sent to the tank (model 2200 Analog stimulus isolator, AM Systems, Sequim, WA, USA). Stimulus intensity was calibrated to cause EOD intensities similar to those produced by an average conspecific. To achieve this calibration, we placed a small recording dipole in the chirp chamber perpendicular to the chamber’s main axis. We recorded and averaged the intensity of the signal emitted by 10 fish placed directly above the chirp chamber (perpendicular to its axis). The stimuli were adjusted to cause the same intensity. Signals were amplified (model 1700 amplifier, AM Systems) and recorded (at a sampling rate of 44,100 Hz) with the same sound card and custom MATLAB script used for stimulation. Analysis of the recordings was also done using MATLAB scripts by visualizing the recordings as spectrograms (Fig. 1, 2). Changes in baseline EOD frequency of the fish (JAR response) and the occurrence of chirps were visually identified by the experimenter (blinded to experimental conditions), and then recorded through the graphical interface.
Aggressive Encounter Experiments
Fish were injected and placed in a small 30 × 30 cm tank with a water depth of 10 cm, in order to constrain the movements of the fish to mostly 2-dimensional. The tank was further divided in half with a diagonal, plastic-mesh panel. The half used for the experiments contained only the stimulus and recording dipoles, whereas the opposite half contained the inlets/outlets for warm water and an air bubbler. The recording dipole was placed at opposite corners of the tank, and the stimulus dipole (5 cm between the poles) was placed in the middle of the compartment, perpendicular to the recording dipole. After a variable acclimatization time (0.5, 1, 2, or 4 h depending on the test group; see Results), ten 1-min-long stimuli were presented with 2-min pauses in between. The stimuli consisted of an EOD with a frequency that was 10 Hz below the fish’s own, and small chirps identical to the ones presented in the chirp chamber experiment. The stimulation and recording equipment were also identical to the chirp chamber rig. The tank was enclosed in an opaque box to block all external light. Infrared LED illumination was provided from below, and video recordings were captured using a webcam (model C920, Logitech, Newark, CA, USA) with its IR filter removed. In a first set of trials (a waiting time of 20 min), video recordings were scored by visually identifying and counting lunges. Lunges were operationally defined as rapid forward movements. These stereotyped motor patterns are easily identifiable. In a small subset of recordings, a second experimenter rescored the trial. In every case, the lunges identified by the first and second experimenter were identical (r = 1). For the second set of experiments, lunges were also identified [Hupé and Lewis., 2008] and, in addition, EOD recordings were processed by quantifying chirping and JAR frequency changes. Furthermore, the videos were processed with ANY-maze video analysis software (Stoelting Co., Wood Dale, IL, USA) to track the position of each fish. The software identified the fish based on contrast and used the tip of the frontal end to determine the x and y coordinates for each frame. We verified visually that the program correctly identified the fish and its frontal edge. Using the coordinates, we quantified several aspects of the fish’s locomotor behavior, including time near the stimulus dipole (i.e., ≤2.5 cm from either pole), distance traveled, and mean swimming speed. These different measures gave similar results, with swimming speed showing less variability, and we therefore show swimming speeds in the Results section.
Social Setting Experiments
This set of experiments used a large 1 × 1 m tank (20 cm water depth) divided into 8 compartments; a ninth central compartment was not used. The compartments were connected to each other by 3-cm-wide openings, starting at 5 cm from the bottom of the tank and extending up to the surface of the water. Each compartment contained a hiding tube, carbon-rod recording dipoles located in opposite corners, and inlets and outlets connected to a sump for the circulation of clean, warmed, and oxygenated water. The tank was covered with an opaque box and kept in the dark during the recordings. An IR video camera was used to monitor the fish without disrupting the experiment. After being injected with THC or vehicle, 4 fish were placed in different compartments on opposite corners of the tank. The fish in the “familiar environment” group were left to acclimate to the tank for 24 h before injection. EOD signals from each compartment were recorded for 3 h after injection. Signals were amplified (model 1700 amplifier, AM Systems), and data were acquired at a sampling rate of 20 kHz for each of the 8 channels, using an AD/DA board (PCIe-6343, National Instruments, Austin, TX, USA). Analysis relied on a custom MATLAB script to visualize the recordings as spectrograms, as described for the chirp chamber experiments, but extended to accommodate 8 separate recording channels. For each 1 s of recording, an experimenter determined where each fish was by comparing the EOD frequencies and amplitude present in each compartment. Because the opening between compartments was small and the recording dipole orientation was kept constant in each compartment, the strength of an EOD signal of a given fish was always stronger in the compartment that contained the fish compared to the recordings of this signal from adjacent compartments. Simple comparison of the spectrograms and power spectra of the recordings from each compartment allowed determination of where the signal was strongest for a given fish’s EOD frequency. In a separate set of pilot experiments, the results of 120 single estimates of fish position per this method were compared with the position of fish in a video recording of the tank to ensure the reliability of the method. Rather than tracking the position of individual fish, the number of fish present in each compartment at the beginning of each 1-s portion of recordings was quantified. Movements of fish from one tank to the next were reflected as a change in the number of fish in the corresponding compartment. The variability of the number of fish per compartment (i.e., “position variability”) was used as a measure of fish movement between compartments. In very rare cases, if 2 fish traded respective compartments within 1 s, the fish count in these 2 compartments would not change. In such rare cases, our estimated fish measure may slightly underestimate movement across compartments. Given that such an underestimation, however small in magnitude, would occur in equal proportion across chambers and test conditions, the risk of type 2 error is minimal. Prior to statistical analysis, we ascertained that our data were normally distributed, using a Kolmogorov-Smirnov test. Normality was confirmed in all cases except for the results displayed in Figure 3. Parametric or nonparametric statistical tests were applied accordingly.
Five different behavioral paradigms were used, as detailed in the coming sections. The first was the simple observation and electrical monitoring of fish that were injected, released into a 75-L tank, and free to move. The fish were monitored for 20 min as an acclimatization step before the chirp chamber recordings described in the next section. At high doses, the behavioral effects were obvious and rapid, starting within minutes of injection. At the highest dose of 100 mg/kg, nearly every fish displayed the same behavior. Once released in the tank after injection, the fish would swim rapidly around the tank ≥1 times, then drop to the bottom of the tank and become nearly motionless. They would occasionally initiate a short bout of energetic swimming that would last a few seconds. The period of immobility lasted up to several hours. Lower doses had more moderate effects on the swimming in a dose-dependent manner, and higher doses had more severe effects (see next sections).
The most surprising effect of the high-dose THC injection (100 mg/kg) was on the generation of the electric field. It is important to point out that these fish generate an electric field continuously throughout their adult lives. Drastic procedures, such as spinal transection, must be performed to prevent the electric field being generated. However, high doses of THC disrupted and often eliminated the generation of the EOD. The typical initial effect was for the EOD to be erratic, decrease in frequency with time, and then suddenly jump back up, with bouts where it had multiple frequency peaks (Fig. 1a). In the most affected fish, EOD generation stopped and started abruptly, with longer and longer periods of silence, so that the EOD was completely silenced within 10–20 min. We would like to point out as a casual observation that, although all other observable behavioral effects remitted within hours, the fish that had their EOD disrupted and silenced did not recover it until days later. This recovery seemed to be slow, with the EOD gaining amplitude with time. We did not quantify this long-term phenomenon, which should be the topic of a thorough investigation of the effects of CB agonists on EOD-generating mechanisms. We did, however, quantify the propensity for this initial disruption in EOD generation. At a given time, the disruption was either clearly present or not, with no graded transition between states. We quantified the percentage of fish that displayed these erratic EOD disruptions and the percentage of time for which the disruptions were observed (Fig. 1b). The effect was observed in all the fish for injections of 100 mg/kg THC. Lower doses did not cause this effect, except for a single trial where a fish injected with 67 mg/kg THC displayed a brief interruption.
Electromotor Response to Communication Signals in a Hiding Chamber
The first set of stimuli tested consisted of simple beat modulations that replicate the sinusoidal amplitude modulation experienced by the fish when in close proximity to a conspecific. Fish were tested before (Fig. 3a) and after (Fig. 3b) THC injection. The typical JAR was not systematically affected by THC, and there was no effect of dose on the JAR (Kruskal-Wallis test, p = 0.4). This lack of effect of THC on the JAR was useful in confirming that the effects described in the following sections do not simply reflect a generalized lack of responsiveness by the fish, but rather the selective effects of THC on some behaviors but not others.
Chirp stimuli were also presented, and consisted of constant beats (frequencies of –120 to 120 Hz) interrupted twice per second by a small chirp (type 2). The fish tended to respond to these signals with chirps, often produced as an echo response, and it is known that low-frequency beats elicit more response chirps than high-frequency beats (Fig. 3c; Hupé and Lewis ). After THC treatment (Fig. 3d; online suppl. Fig. S1; see www.karger.com/doi/10.1159/000490171 for all online suppl. material), the chirping response was significantly affected in a dose-dependent manner (Kruskal-Wallis test, p = 0.016). Chirping was almost completely suppressed at the 2 higher doses, reduced at the weaker dose (33 mg/kg), and unaffected by the vehicle injections. Chirp chambers constitute a safe, low-stress environment, and our results showed that under such conditions the fish’s tendency to chirp in response to a conspecific signal is reduced by a CB agonist.
Aggressive Encounter Scenario
One of the most frequent and obvious social behaviors of these fish is agonistic interactions. Agonistic behavior can be easily triggered by putting 2 fish into a confined space [Hupé and Lewis, 2008]. Similarly, it can be artificially elicited when 1 fish is presented with a mimic of another fish’s signal. The probability of observing aggressive interactions can be increased by (1) removing any hiding place, and (2) using a stimulus EOD that is close in frequency to the fish (i.e., a low-frequency beat) and that contains chirps. We thus constrained a fish to one-half of a 30 × 30 cm tank, in the middle of which a stimulation dipole was placed (Fig. 4a). After a 20-min acclimatization, the stimulus was presented for 1 min, followed by 2 min of rest without the stimulus. This stimulus-rest cycle was repeated 10 times.
As expected, the fish responded to the stimulus with characteristic aggressive swimming movements, consisting of high-speed movement towards and around the stimulation electrodes and stereotypical lunges [Hupe et al., 2008]. A. leptorhynchus individuals fight by lunging and biting at each other (online suppl. information: videos illustrating the behavior). Similarly, when presented with an electrical mimic, the target fish lunged toward the stimulus dipole and made biting motions in the target areas near the probe that corresponded with the body position of a real fish. These lunges were most frequent in the first few trials with sequences of 10 stimulations (Fig. 4b), indicating that the fish habituated to the artificial fish stimuli. Injections of weak and moderate doses of THC significantly decreased the propensity of the fish to lunge (ANOVA, p = 0.02). Note that this suppression of lunging was not due to the physical injection procedure, because the responses of vehicle control fish did not differ from noninjected controls (ANOVA, p = 0.3).
We used this small-tank agonistic stimulation paradigm to evaluate the duration and time course of the behavioral effects of THC. A 67-mg/kg THC dose was used in the following experiments based on the data presented above that this dose has clear behavioral effects without the risk of the loss of a fish’s EOD and ability to move that was seen at 100 mg/kg. We repeated the aggressive encounter experiment described above, but varied the latency between the injection and the stimulation (from 0.5 to 4 h). In addition to lunging, we quantified each fish’s JAR response, average swimming speed, and chirping response. JAR was unaffected by THC, confirming, once again, that the effects observed are selective (Fig. 5a; ANOVA, p = 0.18). The swimming speed was decreased by THC in the first 2 h following injection (Fig. 5b; ANOVA, p = 4 × 10–8). This decrease had diminished after 2 h and vanished by 4 h (Tukey HSD, p < 0.01 for 0.5–2 h and p = 0.32 for 4 h). We suggest that this time course represents that of the overall effect of THC injections and thus argue that, for these type of injections, THC will affect the fish’s behavior for 2–3 h.
As shown in Figure 4b, THC-injected fish lunged significantly less frequently than controls when tested within 1 h of injection (Fig. 5c; ANOVA, p = 7 × 10–6; Tukey HSD, p < 0.05 for 0.5 and 1 h and p > 0.05 for 2 and 4 h). When control fish were newly placed in a tank and presented with chirp stimuli, they responded with a relatively high number of attack movements; this propensity decreased if they had been acclimated to the tank for several hours. THC injections reduced the frequency of these initial aggressive movements to the same level as in control fish that had been acclimated. These data show that THC decreased aggressive lunges in fish placed in an aggressive encounter scenario.
Aside from decreasing lunging, THC significantly increased chirping responses in this context, but only when the fish was given a short time to acclimate to the new environment (Fig. 5d; ANOVA, p = 0.003; Tukey HSD, p > 0.05 for 1, 2, and 4 h). Interestingly, the THC-induced increase in chirping observed here is opposite to the effect observed in the chirp chamber. We will discuss the interpretation of this result in the Discussion section, but it is noteworthy that previous research indicates that chirping can either be a sign of aggression, or a signal to deescalate aggression when produced as an echo response between the 2 fish [Hupé, 2012]. We therefore calculated the probability that the chirps produced were in response to 1 of the stimulus chirps (i.e., following it closely in time). Focusing on the experiments where the fish was tested 0.5 h after injection, we calculated that 65% of the chirps produced by the fish were indeed echo responses. The likelihood that the chirps produced were echo chirps (i.e., chirps preceded by stimulus chirps) was similar for THC-injected vs. control fish (0.61 ± 0.21 vs. 0.7 ± 0.18 respectively; t test, p = 0.37). Considering that echo chirps are thought to be signals to deescalate aggressiveness, the increase in chirping was consistent with the decrease in lunging and aggressiveness.
Social Environment in Novel versus Familiar Conditions
The results of our experiments so far indicate that THC affected thesocial behavior of A. leptorhynchus in a context-specific manner. For example, in the chirp chamber, communication behavior decreased, whereas in the free-swimming aggressive scenario, communication behavior increased, potentially to deescalate aggression. We next aimed to test if these different effects on social interaction were indeed the consequence of the 2 different contexts. To do so, we used a large tank that was separated into small, interconnected compartments, each containing a single hiding tube for the fish (Fig. 6a). This arrangement was designed to provide a comfortable living space where several fish can reside without stress caused by limited space or hiding tubes. The experiment involved 4 fish being placed in the tank. The “familiar” scenario used 4 fish from the same housing tank, with an established dominance hierarchy. The fish were placed in the tank 24 h before the start of the experiment. Thus, the sociophysical environment was also familiar at the beginning of the experiment. The “novel” scenario used 4 fish from 4 different home tanks. In addition, there was no acclimation to the test tank before testing began. These fish were therefore subject to a novel sociophysical environment. Data were collected over the 3 h following injection (see Methods) of either 67 mg/kg THC or vehicle. By visualizing a spectrogram of the EOD signals recorded from each compartment, fish location chirping rates were easily recorded (Fig. 6b).
Consistent with the decreased motor activity observed in the other tests, the THC-injected fish changed compartments less often than their control counterparts (Fig. 7a; ANOVA followed by Tukey HSD, p = 6 × 10–7 for the effect of THC). This effect was most pronounced shortly after injection (Fig. 7b) and dissipated 2.5 h after injection. The novel versus familiar treatment groups showed no differences in position variability (ANOVA followed by Tukey HSD, p = 0.38 for the effect of sociophysical environment).
One of the most striking effects of THC injection was observed for the novel condition, by comparing the dispersion of fish throughout the compartments (ANOVA, p = 5 × 10–6 comparing only the % time found in “clusters of 1”). Control fish were very likely to be found in groups of 2–3 individuals, often even sharing the same tube. This huddling behavior is observed in a variety of fish species in response to stressful situations such as novel environments [Miller, 1963; Johnston and Glasgow, 2015]. Although no publication has documented this behavior in Gymnotiformes, we routinely observe this clustering when a group of fish is placed in a novel tank. It was therefore not surprising to see a higher level of clustering in the novel-control test groups compared to the familiar-control test groups (Fig. 7c). In this study, clustering in the novel-THC group was as low as clustering for the fish in a familiar environment. Clustering was also low in the familiar environment, and THC injection did not further lower clustering.
THC injections did influence chirping propensity; the familiar-THC groups chirped less than their control counterparts (Fig. 7d; t test for familiar control vs. THC, p = 0.013) throughout the duration of the recording (Fig. 7e). Both groups exposed to the novel environment chirped at similarly high rates (t test, p = 0.69). THC injection thus affected distinct aspects of the behavior in novel versus familiar environments. In novel environments, they clustered less than the control groups, but continued communicating with each other at the same rate, whereas THC injections decreased chirping in familiar environments.
The behavioral effects of CB agonists described here are consistent with observations of other species, with one of the most marked effects being a decrease in locomotor activity [Rodriguez de Fonseca et al., 1998] (Fig. 5b, 7a). The influence of THC on locomotor activity in A. leptorhynchus led to a decrease in swimming, but only with THC of very high doses. THC injections also resulted in a disruption of electromotor activity. Our study focused on communication behavior and social interactions. Our data indicate that THC injection did not merely induce a general decrease in all behavior, as would be expected if the animals were simply sedated. Rather, we see a specific effect on some behaviors and a lack of effect on others (JAR). Chirping in THC-injected fish was decreased in the chirp chamber experiments when presented with simple beat stimuli. It also decreased when animals were placed in a familiar context with several other fish, but was not affected by THC in an unfamiliar context. In contrast, THC caused an increase in chirping in isolated, freely moving fish presented with chirp stimuli. Our experiment also showed that THC injections could reduce certain social behaviors. It decreased aggressive lunges in fish placed in a new tank and presented with chirp stimuli, and decreased the likelihood of fish to cluster in groups when placed in a new social and physical environment. Our results thus support our hypothesis that THC behavioral effects are context-dependent.
CB1 is expressed in a wide range of regions in the brain [Harvey-Girard et al., 2013], including motor and sensory areas as well as a high concentration in higher brain areas (e.g., the forebrain). Our study clearly shows that electromotor activity is disrupted at high THC doses. Harvey-Girard et. al.  showed that the pre-pacemaker nucleus, a region modulating the pacemaker that drives the generation of the electric field, is rich in CB receptors. We cannot rule out the possibility that EOD generation is disrupted at a site other than the pre-pacemaker nucleus. Nevertheless, lesions in the pre-pacemaker nucleus can cause changes in EOD frequency and regularity [Moortgat et al., 1998]. Thus, it seems plausible that CB could affect the pre-pacemaker dynamic leading to observed changes in EOD output.
It could be suggested that most of the effects observed in this study result from the influence of the CB system on the motor system. For example, it could be suggested that a generalized depression of the motor system in THC-injected fish caused the slower swimming speed, the decrease in aggressive lunging, or the decreased clustering. We argue, however, that an effect on the motor system is unlikely to explain all our observations. The data indicate that THC sometimes increases chirping and sometimes decreases chirping. Also, clustering was affected in a context-dependent manner (Fig. 7c) whereas the effects on the amount of movement were not context-dependent (Fig. 7a). We therefore suggest that although an effect on the motor system can explain some of our observations, the data indicate that multiple systems are affected by THC dose and environmental context.
The lower levels of the sensory system involved in electrocommunication are also a major focus of research in weakly electric fish. Specifically, the electrosensory lateral line lobe and the feedback pathway through parallel fibers have a well-understood role in processing electrosensory signals (review [Hopkins and Popper, 2005; Maler, 2007; Marsat et al., 2012]). This network has several features specialized in processing communication signals. The main output cells of the electrosensory lateral line lobe, the pyramidal cells, receive massive feedback from higher levels to enhance the processing of communication signals. One of these sources of feedback, the granule cells of the caudal cerebellum, express CB receptors and exhibit plasticity [Bol et al., 2011]. The CB system may thus influence the processing of communication signals as early as the primary electrosensory area of the central nervous system. CB1 is also present in higher areas of the sensory system such as the torus semicircularis [Harvey-Girard et al., 2013]. Once again however, it seems unlikely that a change in sensory processing plays a large role in explaining the main behavioral effect of THC observed in this study. For example, chirping decreased in the chirp chamber experiment even though the JAR was unaffected. If THC causes a change in responses to beat stimuli in this context, you would expect both behaviors to be affected. Also, chirping did not always decrease in response to THC injection, whereas a change in perceptual abilities would be expected to affect responsiveness in all situations. Nevertheless, we do not rule out the possibility that the sensory effects of THC may have affected the behavior of the fish in our experiments.
It is likely that whole-brain dynamics, including those of higher brain regions, are affected by the CB system. A possible explanation for the context dependence of some of the behavioral effects of THC injection is that different contexts cause state changes in the animal and influence levels of stress, anxiety, or aggressiveness. We detail how our results support this suggestion in the following 2 paragraphs.
THC caused a decrease in clustering in a novel social and physical environment but not in a familiar one. This change in clustering can be understood based on the typical reaction of these fish to novel environments. Serval species of fish display huddling behavior when placed in a novel environment or are subjected to stress [Miller, 1963; Johnston and Glasgow, 2015]. It is frequently observed in A. leptorhynchus when they are introduced to a new environment. The THC-induced decrease in clustering could be the consequence of a decrease in the anxiety-like response in a novel environment.
It has been shown that chirping correlates with increased or decreased aggressiveness depending on the pattern of chirp exchanges [Hupé, 2012]. Specifically, when emitted as an echo response to a chirping conspecific, it mediates a deescalation of aggression. Conversely, if chirps are not produced in response to the other fish’s chirping, this correlates with an escalation of aggression. In our experiments, the fish injected with THC increased chirping as an echo-response to a stimulus containing chirps. This experiment was designed to replicate aggressive interactions, and so the increase in echo-response chirping could be interpreted as a deescalation signal. This interpretation is also supported by the parallel decrease in aggressive lunges observed.
Our results therefore suggest that THC causes a decrease in aggressiveness and, potentially, a decrease in stress/anxiety, and that these effects could explain the context-dependency of some of the behavioral effects observed. This hypothesis is strengthened by a rich literature on the link between the CB system and anxiety-like behaviors, including several studies on teleost species such as the zebrafish. Activation of the CB system has a clear anxiolytic/anxiogenic effect in zebrafish where low or medium doses of agonist lead to decreased anxiety-like behavior but increase anxiety-like behavior at high doses [Krug and Clark, 2015]. For example, Stewart and Kalueff  tested freely swimming fish and found that, in addition to decreased mobility, high doses of THC caused a decrease in time spent in the upper portion of the tank, an anxiogenic-like response. In contrast, Barba-Escobedo and Gould  described an anxiolytic-like effect of low doses of the synthetic CB WIN55,212-2 leading to more shoaling and increased exploration of portions of the maze that were lit up. As noted by Krug and Clark , the effect of CB agonists in zebrafish in various experiments depends on many variables, including dose but also context.
The literature on mice shows similar context-dependent effects of CB agonists. Some studies show selective decreases in aggressive behaviors while others show the induction of fearfulness and the inhibition of social behaviors [Miczek, 1978; Cutler and Mackintosh, 1984; van Ree et al., 1984]. Similarly, in mice in an elevated plus-maze which is a common screen for anxiolytic drugs, CB antagonists exerted both anxiogenic- and anxiolytic-like effects, depending on the dose and context [Navarro et al., 1997; Akinshola et al., 1999].
The role of context in the influence of the CB system on anxiety and social behavior was explicitly tested in a study using CB1 knockout mice [Haller et al., 2004]. In the home cages, an intruder elicited more aggressive interactions in CB1-deficient mice than in wild-type controls that expressed normal levels of CB1. In a novel environment, introduction of a conspecific induced fewer interactions and lower aggressiveness in the knockout mice. In addition, CB1 deletion decreased the amount of time spent in the open arms of an elevated plus-maze, but only when the maze was brightly lit, indicating that CB1 effects on exploration are context-dependent, possibly due to interactions between CB signaling and the hypothalamic-pituitary-adrenal axis.
Considering the widespread expression of CB receptors throughout the brain [Patel et al., 2017], it is not surprising that their systemic activation/deactivation has complex effects that are task-dependent. It is noteworthy that many of the behavioral effects described here are comparable to the effects observed in other vertebrates. Therefore, in conclusion, we argue that our study lays the basis for future studies using the advantages of weakly electric fish as a model system to understand the dynamics of the CB system in a generalizable way.
There were no conflicts of interest.
B.N. and T.O. contributed equally to this work.