Acute Citalopram administration modulates anxiety in response to the context associated with a robotic stimulus in zebrafish

Abstract

Background:

Anxiety represents one of the most urgent health challenges in Western Countries, whereby it is associated with major medical and societal costs. A common therapeutic approach is the use of selective serotonin reuptake inhibitors, such as Citalopram. However, this treatment of choice is characterized by incomplete efficacy and potential side effects. Preclinical research is needed to detail the mechanisms underlying therapeutic efficacy of available treatments.

Methods:

Zebrafish, a rapidly emerging model species, constitutes an excellent candidate for high-throughput studies in behavioral pharmacology. Here, we present a robotics-based experimental paradigm to investigate the effects of acute Citalopram administration on conditioned place aversion. We trained adult subjects in a three-partitioned tank, consisting of one central and two lateral compartments: the latter were associated either with a fear eliciting robotic stimulus or with an empty environment. Following training, we implemented an automated three-dimensional tracking system to assess the spatial association and detail individual phenotype in a stimulus-free test session.

Results:

We observed a linear dose-response profile with respect to geotaxis, with increasing Citalopram concentrations reducing the tendency to swim near the bottom of the tank. Although control subjects failed to exhibit the predicted conditioned aversion, we found preliminary evidence that Citalopram may affect sexes differentially, with male subjects showing increased conditioned aversion at low Citalopram concentration.

Conclusions:

Experimental paradigms based on robotics and three-dimensional tracking can contribute methodological advancements in zebrafish behavioral psychopharmacology.

INTRODUCTION

The global burden of mental disorders is constantly increasing and is expected to account for more than 6% of the total disability-adjusted life years by 2030 [1]. Among them, anxiety represents one of the most widespread mood disorders [2, 3]. Its multifactorial etiology includes genetic and environmental factors, and its pathophysiology is highly heterogeneous [4]. Alterations in monoaminergic neurotransmission have been shown to represent a common endophenotype in anxious subjects [5–9]. Compounds enhancing monoaminergic neurotransmission, through inhibition of its reuptake, have been found to yield controversial results, with several studies reporting anxiety reduction [10–12] and others reporting anxiety increase [13].

Citalopram, a selective serotonin reuptake inhibitor (SSRI) considered the treatment of choice in depression, is often used for the treatment of anxiety [10, 11]. However, several limitations have been recognized for Citalopram and other SSRIs; for example, they usually exert anxiogenic effects after acute administration, and they may cause a wide range of mild to severe side effects [13, 14]. The earliest discoveries concerning SSRIs and their effects on mental disorders were primarily driven by serendipity [15]; only recently has hypothesis-driven research focused on these compounds [14, 16]. Further clinical and preclinical studies are required to expand the knowledge on SSRIs and serotonin as mood regulators.

Although laboratory rodents are traditionally the species of choice for neuropharmacological studies [17, 18], zebrafish (Danio rerio) is gaining popularity as a neuropharmacological animal model. Zebrafish has been widely used in preclinical studies, due to its many advantages, including low-cost maintenance, high reproduction rate, rapid development, fully sequenced genome, and translucent, easy-to-manipulate embryos [19–21]. Zebrafish physiology and neuroanatomy are highly homologous to humans [22–25]. They exhibit a rich behavioral repertoire that is sensitive to pharmacological manipulations, thereby affording elevated predictive validity [26–28]. While these characteristics denote zebrafish as an excellent tool for behavioral pharmacology [29–33], further research is required to improve protocols and methodologies in zebrafish studies.

One of the main challenges of using zebrafish in pharmacological research is translating experimental paradigms traditionally designed for rodents to zebrafish [27, 28, 34, 35]. Several conditioning paradigms have been used to study the effects of drug administration on animal behavior, especially on spatial learning and memory. In particular, Pavlovian fear conditioning constitutes an established experimental paradigm that dates back to the origins of ethological experiments. This paradigm rests upon the association of a given environmental stimulus (conditioned stimulus) with the anticipation of an aversive event (unconditioned stimulus), through single or multiple training sessions [36]. Once the animal learns the association between the conditioned and the unconditioned stimulus, the former is sufficient to evoke a fear response [36]. The stimulus location can represent a conditioned stimulus itself, and it can be used to create an aversion towards a given section of the experimental area. This mechanism constitutes the basis of the conditioned place avoidance (CPA) paradigm, which was first developed for rodents and later adapted to zebrafish [37, 38].

Ethorobotics constitutes a promising tool toward the improvement of experimental protocols to study zebrafish behavior [39–42]. Biologically-inspired robots that imitate morphophysiological features and behavioral patterns of live animals can be used to replace live stimuli. Such a replacement has been associated with a reduced variability in experimental data, which increases statistical power, thereby reducing the number animals required for hypothesis testing [43, 44]. With specific reference to CPA paradigms, harmless robots can replace traditional noxious stimuli that are generally used to condition experimental subjects [37, 38, 45]. Ultimately, ethorobotics may promote advancements toward the 3R’s principle (replacement, reduction, and refinement) for animal welfare [46].

Towards advancing classical experimental paradigms through ethorobotics, we recently developed several biologically-inspired robots that can elicit fear and anxiety in zebrafish [43, 47–51]. In our most recent study [49], we exposed live zebrafish to a robotic stimulus that mimicked a tightly packed school of conspecifics. In live groups, highly coordinated and polarized swimming patterns are triggered by an imminent threat and signal to other fish the need of an escape response [52, 53]. Through repeated exposure to this robotic stimulus, we conditioned zebrafish to avoid a specific section of the experimental arena [49]. Additionally, to assess the predictive validity of our CPA paradigm, we tested whether the anxiolytic properties of ethanol mitigated the conditioned aversion towards the robotic stimulus. In accordance with our predictions, ethanol-treated subjects did not show the same aversion exhibited by vehicle-treated, control subjects.

Herein, we aimed to extend this paradigm to the study of acute Citalopram administration. We repeatedly exposed zebrafish to a robotic school, composed of three silicone replicas of a male adult zebrafish and maneuvered along three-dimensional trajectories that were reminiscent of highly coordinated and polarized swimming patterns. The robotic platform was placed in a predetermined side of the tank to form a spatial association between the stimulus and its location. We tested the conditioned fish in a stimulus-free environment, in response to vehicle (0 mg/L) or Citalopram (30 and 100 mg/L concentrations) administration. We hypothesized that zebrafish would learn to associate the robotic-paired section of the apparatus as a potential threat, thus showing avoidance and anxiety-related behavior. We anticipated a linear dose-response curve to Citalopram administration [48, 54, 55]. The behavioral phenotype of zebrafish was scored in three-dimensions by fusing information from two independent camera views, using an in-house developed automated tracking software. This aspect represents a further advantage in terms of the 3R’s principle, whereby three-dimensional observations can reduce the number of subjects up to 50% [56] and refine behavioral phenotyping of anxiety [55, 57] .

MATERIALS AND METHODS

ETHICAL STATEMENTS

Experiments were performed following the relevant guidelines and regulations set forth by the University Animal Welfare Committee (UAWC) of New York University under protocol number 13–1424.

ANIMAL HOUSING

A total of 78 wild-type zebrafish with an equal sex-ratio were purchased from Carolina Biological Supply Co. (Burlington, NC, USA). Upon arrival, the fish were kept in 10 L vivarium tanks (Pentair Aquatic Eco-Systems Locations, Cary, NC, USA) with a density of no more than 12 fish per tank for an acclimation period of at least 14 days prior to the beginning of the experiment. The housing tanks were kept under a photoperiod of 12 h light/12 h dark. Water temperature and pH were maintained at 26 °C and 7.2, respectively.

Male and female fish were kept in separate tanks to facilitate the sex identification and to avoid mating behavior. A day before the start of the experiments, each fish was relocated to an individual 1.5 L housing tank to keep track of individual identities. Following the recommendations of [58] and the study by Lawrence, et al. [59], fish were fed once a day, approximatively at 6.30 PM, with commercial flake food (O.S.I. Marine lab Inc., Burlingame, CA, USA); minor changes to the feeding schedule were needed during the COVID-19 pandemic.

ROBOTIC PLATFORM

The robotic platform consisted of two main parts: a three-dimensional manipulator, and a school of conspecific replicas. Building on the work of Kim, et al. [60], we developed a three-dimensional manipulator capable of creating an accurate ethogram of a school of conspecifics using four degrees-of-freedom. The school of conspecific replicas could move in three dimensions and rotate around its center about a vertical axis (yaw) in a coordinated manner.

The robotic manipulator had two servo motors (MG995R, Tower Pro, Taiwan), one DC motor with an encoder (52 RPM planetary gear motor, Actobotics, Winfield, KS), and two stepper motors (NEMA 17, Adafruit, New York, NY). Two servo motors were actuated simultaneously to maneuver the school along X-axis (Fig. 1) with a 10-cm range. The DC motor enabled the movement along Y-axis in a 25-cm range. One stepper motor was used to adjust the height of the shoal from 5 cm above ground up to the top of the water column. The other stepper motor aligned the school along its swimming direction, providing a biologically meaningful representation.

Figure 1:

Computer-assisted design of the experimental setup showing the robotic platform, the school of replicas, live fish, doors, and cameras. Red, green, and yellow arrows represent the coordinate system: X-, Y-, and Z-axis, respectively.

We employed two motor shields (Adafruit, New York, NY) connected to two microcontrollers (Arduino, Ivrea, Italy) to drive all the motors. We designated the microcontrollers in master-slave configuration and paired the master microcontroller with an ethernet shield (Adafruit, New York, NY). The ethernet shield allowed for communication between the microcontrollers and the computer through a user datagram protocol (UDP). The time traces of position and speed were loaded in MATLAB R2019b (MathWorks, Natick, MA), and communicated to the microcontrollers at 2Hz.

The robotic platform operated in an open-loop configuration. Specifically, the school was maneuvered along a predetermined trajectory, which duplicated previous motion data of a single fish swimming in an equivalent tank (16 cm × 30 cm × 30 cm; length × width × height) [49, 61].

SCHOOL OF REPLICAS

We fabricated three conspecific replicas of zebrafish and arranged them in a tightly packed school. The rationale for choosing a school of three replicas stems from the need to compromise between a salient stimulus and a practically-realizable one with our hardware. From a theoretical point of view, the examination of inter-individual distance for different group sizes from the literature suggests that inter-individual distance of fish plateaus for groups of three fish [62–64]. As a result, three was the smallest group size that preserved the saliency of a real school. From a practical point of view, the limited range of motion of the robotic platform prompted us to minimize the number of replicas in order to instrument it with a more natural motion pattern.

Based on the images of a male adult zebrafish, we modeled a replica using a computer-assisted drawing software [60]. We then formed a mold around the replica and 3D-printed the mold. The mold was filled with silicone (Dragon Skin 10, Smooth-On, Macungie, PA), whose compliance allowed the replicas to passively undulate their body while swimming. Replicas were hand-painted, reproducing the natural patterns of a zebrafish, and a pair of glass eyes were glued on each of them. They were arranged in a tightly packed school formation and connected to the acrylic rod actuated by the robotic platform (Fig. 2). The nearest neighbor distance between the replicas was 3 cm, similar to the density of a school that is under a high state of agitation [52].

Figure 2:

The tightly packed school of conspecific replicas of zebrafish.

The school of replicas performed a highly coordinated movement throughout the experiments. The motion patterns of the replicas can be found in the Supplementary Data files.

EXPERIMENTAL APPARATUS

The experimental apparatus was composed of three transparent Plexiglas tanks: the main rectangular tank (experimental tank) and two smaller rectangular tanks placed next to its shorter sides. The experimental tank was 42 cm × 30 cm × 30 cm in length, width, and height, respectively (Fig. 1). Two transparent acrylic partitions divided this tank into three compartments, with the central one of 12 cm in length, and the lateral ones of 15 cm in length. The small tanks, placed along the length of the experimental tank, measured 16 cm × 30 cm × 30 cm in length, width, and height, respectively. All the tanks were filled up to 15 cm of water level with room temperature (26°C) tap water. The bottom of all the tanks and the sides of the smaller tanks were covered with white opaque contact paper to prevent any external visual interference and provide a homogeneous background. For the same reasons, a white panel was placed along the length of the experimental tanks. The front side of the tanks was uncovered to allow for the camera recording.

Two cameras (C920, Logitech, Lausanne, Switzerland) recorded each experimental session, at 30 frames per second, with a resolution of 640 × 480 pixels. One webcam was positioned 72 cm above the water surface (top view). The other one was positioned 82 cm from the front of the tanks, at the height of 12 cm from the bottom of the tank (front view). The experimental setup was illuminated by two 250 lumen LED lamps (Target, Minneapolis, MN), which projected the light towards a white panel placed behind the front camera (ambient lighting was measured at 200 lux). The area was surrounded with black curtains to isolate the experimental arena from the external environment.

EXPERIMENTAL PROCEDURE

Conditioned place aversion experiments were performed in January and February 2020. Three experimental groups, each composed of five males and five females, were trained and tested. Each subject was trained and tested over four consecutive days; the training trials were performed during the first three days, one in the morning (9.30 – 12.30 AM) and one in the afternoon (2.30 – 5.30 PM), while the test trials were performed on the fourth day (11 AM – 4 PM). The experiments were conducted in five experimental batches comprising six subjects (one male and one female fish for each of the three experimental groups). Each fish was trained twice per day for a total of six trainings.

In the training trials (conditioning phase), the fish swam in the main tank, while the robotic school was placed in one of the two lateral smaller tanks. The location of the robotic school was counterbalanced across subjects: we maintained the robotic school in the same lateral tank for an experimental batch, and we then switched its location for the subsequent batch. The robotic apparatus was removed before testing. During training, a fish was gently hand-netted from its housing tank to the middle compartment of the main experimental tank, with both sliding doors closed. After one minute of acclimation period, we lifted one of the panels to let the fish swim into either the stimulus-paired compartment or the empty one. If we lifted the stimulus-paired panel in the morning, we would lift the empty-paired panel in the afternoon, and vice versa.

Simultaneous to the door opening, the robotic platform started its movement. We lowered the panel immediately after the fish entered the lateral compartment. In case the fish did not leave the middle compartment within a minute after the panel was lifted, we gently pushed the fish using a transparent acrylic bar. The fish swam in the side compartment for 15 minutes, while the robotic school was in motion. At the end of the training, the robotic school stopped, and we lifted the panel and gently pushed the fish with the transparent bar into the middle compartment. We then hand-netted the fish and transferred it back to its housing tank.

Each test consisted of a fish swimming in a stimulus-free environment after drug administration. The subject was gently hand-netted from its housing tank into a 0.5 L beaker, containing either the Citalopram solution (30 mg/L or 100 mg/L) or water, and remained there for five minutes (the water in the beaker was always taken from the housing tanks). Then, the fish was hand-netted to the middle compartment of the main tank, with both panels lowered. After one minute of acclimation period, we simultaneously lifted both panels to let the fish swim in any of the three compartments freely for 10 minutes. When the experiment was concluded, the fish was relocated to its housing tank.

An independent study was conducted to detail the avoidance response to the stimulus in October 2020. Specifically, we performed binary choice tests in which a naïve subject was allowed to swim in the same tank used for the conditioned placed preference tests (albeit without panels) in the presence of the following stimuli: the same moving stimulus as the conditioned place aversion tests; a stationary version of the stimulus; and both the moving and stationary stimuli. A total of 48 experimentally naïve animals with 1:1 sex ratio were tested (16 per condition). Sex and conditions were counterbalanced across experimental sessions.

TRACKING SOFTWARE

To obtain a complete ethogram of the experimental fish during conditioned place aversion tests, we used the tracking software presented in [49]. We acquired images from two views of the tank: the front view, from which we captured the X–Z plane, and the top view, from which we recorded the X–Y plane. For each video, we created a binary mask to highlight a region of interest. Within that region, we estimated a background image by taking the most repeated intensity values of pixels over 1,000 random images. We subtracted the background from all the frames and applied a threshold on the images. As a result, we obtained binary images where the potential fish targets were marked with a true value; upon examining connected regions within these images and contrasting them with the typical fish size, we identified the centroid of the fish and extracted trajectories for both views.

We calibrated the cameras using a checkerboard pattern [65]. We identified points on the frames and their 3D correspondences to estimate the rotation and translation matrices of the cameras. Finally, we triangulated the three-dimensional trajectory of the fish using the information obtained from both cameras. Representative trajectories of six experimental subjects are displayed in Fig. 3. For the independent tests detailing avoidance response, we focused on the X–Z plane.

Figure 3:

Representative trajectories of six fish during testing. Zebrafish swimming in the experimental arena for 10 minutes after an habituation phase of one minute with (a,b) 0 mg/L, (c,d) 30 mg/L, and (e,f) 100 mg/L Citalopram administration. Males are in (a,c,e), and females in (b,d,f).

BEHAVIORAL SCORING

Three-dimensional position data obtained over 10 bins of one minute were used to calculate the following parameters during testing [48, 49].

Distance from the bottom of the tank:

Distance from the bottom of the tank: the perpendicular distance between the centroid of the fish and the bottom of the tank, measured in cm, ranging from 0 to 15. Low values indicate geotaxis, that is, a preference for the lower part of the water column, an anxiety-related behavior common in zebrafish [26];

Avoidance index:

Avoidance index: T_E/(T_E+T_S) × 100, wherein T_E is the time spent in the stimulus free compartment (empty), and T_S is the time spent in the stimulus-associated compartment. Values significantly greater than 50% indicate aversion, while values significantly smaller than 50% indicate attraction;

Distance from the stimulus:

Distance from the stimulus: the perpendicular distance between the centroid of the fish and the lateral wall adjacent to the stimulus tank, measured in cm, ranging from 0 to 42. High values indicate aversion, while low values indicate attraction;

Freezing time:

Freezing time: percent time in which the fish was exhibiting freezing behavior, quantified as the time in which the fish was confined within a 2-cm sphere for more than two seconds. Similar to geotaxis, freezing is an anxiety-related measure [26];

Spatial entropy:

Spatial entropy: Shannon entropy from binning the tank with a uniform grid whose cubic cell dimension was approximately two body lengths. It is used as a measure of the exploration of the tank (high values indicate a higher tendency to explore the tank) [48, 49];

Average speed:

Average speed: measured in cm/s as the time-average of the magnitude of the first-order numerical derivative of the position time series;

Average absolute acceleration:

Average absolute acceleration: measured in cm/s² as the time-average of the magnitude of the second-order numerical derivative of the position time series;

Average turn rate:

Average turn rate: measured in rad/s as the time-average of the first-order numerical derivative of the heading angle, calculated from the swimming direction.

The last four parameters can be used to measure fish locomotory activity, including erratic movements of the fish that tend to increase in fish exposed to stressors or exhibiting anxiety state [26].

During training, we measured the distance from the lateral wall adjacent to the stimulus (the empty compartment or the compartment containing the robotic school) tank to understand whether the fish, constrained in the training compartment, displayed an avoidance response after the first two minutes of habituation. Likewise, in the independent experiment, we focused on the distance from the wall separating the moving or stationary robotic school.

STATISTICAL ANALYSIS

Before conducting the analysis of the conditioned place aversion, we inspected videos and raw data to identify and exclude the outliers. We defined an outlier as a subject who exhibited a freezing behavior for more the 80% of each time bin. Hence, two experimental animals were excluded from the analysis: one female belonging to the control group and one male in the Citalopram 30 mg/L group. We used the Shapiro-Wilk test to verify the normality of the data organized according to treatment and sex.

The avoidance index for each group was compared to the 50% hypothesized value using a one sample, one-tailed t-test. Since we conducted multiple comparisons, we controlled for type-1 errors (false positive findings). Therefore, the p-values obtained from one sample, one-tailed t-test were corrected using Bonferroni correction; the same procedure was utilized in [43]. The same procedure was used to assess avoidance in the independent binary choice tests.

The comparison among treatments and between sexes, for each parameter, was performed with repeated measures analysis of variance (ANOVA), calculated over the 10 minutes of the test trial, with each minute representing a time bin. Time bins were considered as a within-subject factor, while the treatment (Control, Citalopram 30mg/L, Citalopram 100 mg/L) and the sex (Female, Male) were between-subject factors. The general model for the analysis of the conditioning phase data (average distance from the stimulus pooled across the training sessions) was an ANOVA with two factors: sex and side (empty, containing the robotic school); due to accidents incurred during data storage, three trials, pertaining to different experimental subjects, were excluded from the analysis. We then computed Bonferroni post-hoc pairwise comparisons between the treatments and between sexes within each treatment for the testing sessions.

Statistical analysis was performed using Statview II (Abacus Concepts, CA, USA) or SPSS 26 (IBM, NY, USA). Results are expressed as mean ± standard error of the mean (SEM). The significance level was set at p<0.05.

RESULTS

MODULATORY EFFECT ON CITALOPRAM

Distance from the bottom:

In accordance with predictions, the distance from the bottom increased with the Citalopram concentration in a dose-dependent manner (Condition: F_(2,16) = 16.40, p < 0.01). Specifically, Bonferroni comparisons showed significant differences between Control and 30 mg/L group (mean diff. = 3.58, crit. diff. = 2.14, p < 0.01), Control and 100 mg/L group (mean diff. = 6.28, crit. diff. = 2.09, p < 0.01) and between 30mg/L and 100 mg/L groups (mean diff. = 2.66, crit. diff. = 2.09, p = 0.01), but not between sexes within each group. This profile was neither affected by time nor sex (Time bins: F_(9,144) = 1.36. p = 0.21; Sex: F_(2,16) = 1.26, p = 0.28), although a significant interaction between these three factors was noted (Time bins × Condition × Sex: F_(18,144) = 2.15, p < 0.01) (Fig. 4a).

Figure 4:

Comparisons among treatments and between sexes for the following parameters: (a) Distance from the bottom of the tank. (b) Conditioned aversion with 95% confidence intervals (whiskers) measured as the time spent in the empty compartment divided by the sum of time spent in the empty and the stimulus-paired compartment during the test session. The dashed line represents the chance level (50%). The asterisk indicates a significant difference from the chance, i.e., avoidance or attraction towards the stimulus associated compartment. (c) Distance from the stimulus tank. (d) Spatial entropy. (e) Average acceleration. (f) Average turn rate. In (a)—(f), pink denotes female subjects, blue indicates male subjects, and different patterns refer to the different conditions. Also, histograms represent means of the groups with standard error; different letters on top of the histograms indicate a significant difference among the means. (g) Average speed with standard error (whiskers) for female subjects. (h) Average speed with standard error (whiskers) for male subjects. In (g) and (h), whiskers indicate standard errors, cyan circles represent the Control group, green triangles represent 30 mg/L group, and red squares represent 100 mg/L group. Pound indicates significant time variation over the 10 minutes of the test session.

Avoidance index:

During the test session, experimental subjects did not show a preference for any of the lateral compartments (t₂₇ = 0.50, p = 0.62). Separate analyses conducted in the three experimental groups revealed that only the Citalopram 30 mg/l subjects displayed significant avoidance of the stimulus-associated compartment (t₈ = 3.12, p = 0.02), while neither control (t₈ = −0.55, p = 0.90) nor Citalopram 100 mg/L (t₉ = −0.22, p = 1.00) showed any preference. When analyzing the sex effect, we did not register a preference in control subjects (female: t₄ = −0.91, p = 1.00, male: t₃ = 0.70, p = 1.00), but we noted a significant conditioned avoidance for Citalopram 30 mg/L males (t₄ = 4.40, p = 0.04) and not in Citalopram 30 mg/L females (t₃ = 0.95, p = 1.00). Neither males nor females in the Citalopram 100mg/L group displayed avoidance (female: t₄ = 0.97, p = 1.00, male t₄ = −1.16, p = 0.93) (Fig. 4b).

Distance from the stimulus:

The distance from the stimulus tank did not differ among the treatments (Condition: F_(2,16) = 0.78, p = 0.47). No difference was detected even when considering the combined effect of treatment and sex (Condition × Sex: F_(2,16) = 2.06, p = 0.16). Likewise, this profile did not vary over time (Time bins: F_(9,144) = 0.79, p = 0.62) (Fig. 4c).

Spatial entropy:

Spatial entropy was undistinguishable across treatments (Condition: F_(2,16) = 1.23, p = 0.31), and remained constant in time (Time bins: F_(9,144) = 1.11, p = 0.36). The effect of Citalopram administration was not influenced by sex (Condition × Sex: F _(2,16) = 1.97 p = 0.17) (Fig. 4d).

Average acceleration:

The average acceleration did not differ among the three experimental groups (Condition: F_(2,16) = 0.01, p = 0.99), neither did it vary throughout the test session (Time bins: F_(9,144) = 1.43, p = 0.18). Finally, male and female subjects exhibited an undistinguishable response to Citalopram administration (Condition × Sex: F _(2,16) = 0.69, p = 0.51) (Fig. 4e).

Average turn rate:

Similar to other locomotion-related metrics, Citalopram administration did not elicit a differential response in experimental subjects (Condition: F_(2,16) = 0.24, p = 0.79). Average turn rate did not vary in time (Time bins: F_(9,144) = 1.58 p = 0.12) and was not affected by the interaction between sex and treatment (Condition × Sex: F _(2,16) = 1.49, p = 0.25) (Fig. 4f).

Average speed:

The average levels of speed were indistinguishable among the experimental groups (Condition: F_(2,16) = 0.00, p = 0.99). No significant interaction between condition and sex was detected (Condition × Sex: F _(2,16) = 1.41, p = 0.27). Yet, average speed varied in time (Time bins: F_(9,144) = 3.39, p < 0.001), whereby post-hoc comparison indicate a significant decrease from the second to the last time bin (mean diff. = 1.088, crit. diff. = 0.897, p <0.01) (Fig. 4g,h).

AVERSIVE NATURE OF THE ROBOTIC STIMULUS

During training, we observed that fish exhibited aversion towards the robotic school. Specifically, when allowed to swim in the tank associated with one or the other stimulus, they swam closer to the wall associated with the empty compartment than the one associated with the robotic school (Side: F_(1,56) = 17.31, p < 0.001). Male and female subjects did not differ in their avoidance response (Sex: F_(1,56) = 0.06, p = 0.81), and such finding was not influenced by the stimulus side (Sex × Side: F_(1,56) = 1.94, p = 0.17) (Fig. 5)..

Figure 5:

Distance from the stimulus-associated compartment during training. Pink bars represent the mean distances from the stimulus of female subjects, and blue bars indicate the mean distances of male subjects. Whiskers denote the standard error of the mean, whereas different letters show a significant difference among the means.

In the binary choice tests, we did not observe a consistent avoidance of any robotic stimulus, be it moving or stationary. When fish were given a choice between the compartment with the moving stimulus and the empty compartment, they did not exhibit avoidance (Avoidance index = 45.28 ± 3.38%; t₁₅ = −1.40, p = 0.28). Similarly, fish did not exhibit a spatial preference when given a choice between the stationary stimulus and the empty compartment (Avoidance index = 56.99 ± 5.99%; t₁₅ = 1.17, p = 0.39). Finally, when the animals were exposed to both the robotic stimuli located, they did not exhibit a preference for either of them (Avoidance index for the moving stimulus = 55.32 ± 5.31%; t₁₅ = 1.00, p = 0.50). These profiles were not affected by sex (p > 0.68).

DISCUSSION

The main aims of the present study were to confirm that the presentation of a fear-evoking robotic stimulus resulted in conditioned avoidance and whether this response was modulated by acute Citalopram administration. The robotic stimulus was designed to mimic a tightly packed school of conspecifics moving in a highly coordinated manner [49]. We proposed that the visual appearance and motion pattern of the replicas signaled to the test subject an imminent threat, thereby evoking an anxiety-related state. Accordingly, we observed that such a stimulus induces conditioned avoidance in a stimulus-free test session [49]. Here, subjects were exposed to this robotic stimulus for six training sessions over three days, and on the fourth day, they were tested in the absence of the stimulus in response to Citalopram administration (0, 30, and 100 mg/L).

The three-dimensional ethogram of the fish was obtained using an in-house tracking software, described and validated in our previous work [55, 56]. Through this software, we examined the geotactic response, aversion from the stimulus, and locomotory measures, such as speed, acceleration, and turn rate to infer a relationship between acute Citalopram administration and zebrafish behavior. Experimental results support the prediction that zebrafish behavior was modulated by Citalopram administration, in the form of weaker geotactic behavior for increasing Citalopram concentrations. The control subjects showed a tendency to stay near the middle of the water column, while the two groups of treated subjects displayed a marked tendency to swim closer to the water surface. This tendency was more pronounced in the group exposed to the higher Citalopram concentration, showing that increasing Citalopram concentration is related to a decrease in geotaxis.

Geotaxis is traditionally associated with anxiety [26, 66] and its predictive validity has been confirmed in numerous studies using different anxiolytic compounds, including nicotine [34, 54], ethanol [55, 67], benzodiazepines (chlordiazepoxide and Diazepam [68]), 5-HT receptor agonists (Buspirone [68, 69]), and SSRI’s (Fluoxetine [31] and Citalopram [48, 54, 55]). Our study contributes to this bulk of knowledge by confirming that acute Citalopram administration reduces the geotactic response of adult zebrafish. Analogous results on geotaxis have been documented in [48, 54, 55]. Sackerman and his collaborators [54] used a novel tank diving test to demonstrate that acute Citalopram administration (100 mg/L) reduces geotaxis. In [55], a similar test was conducted using different Citalopram concentrations (30, 50, and 100 mg/L). In [48], the effect of Citalopram on geotaxis was tested in a conditioned place aversion test using an aerial robotic predator as a fear-evoking stimulus. These studies suggest that Citalopram concentration is linearly associated with a reduction in anxiety, be the latter an unfamiliar environment or a conditioned place aversion test.

Herein, we used a robotic school of conspecifics as a conditioning stimulus and tested the effect of different Citalopram concentrations. In agreement with our predictions, we registered a linear dose-response profile in the geotactic behavior of the experimental subjects, demonstrating the anxiolytic effect of acute Citalopram administration in this fear conditioning paradigm. We acknowledge that subjects treated with the 100 mg/L concentration often raised to the surface of the water, exhibiting the so-called “surfacing” behavior [26]. This behavior is associated with buoyancy dysregulation and it is principally evoked by neuroactive compounds [26]. In particular, in SSRI-treated fish, surfacing has been associated with serotonin syndrome - the adverse reaction to an acute overwhelming increase of serotonin levels [70–72]. Although this finding may suggest that geotaxis could have been caused by serotonin syndrome, the absence of surfacing in the 30 mg/L group that swam in the top third of the water column would not support this tenet. In the light of the linearity of the dose-response curve, a more parsimonious model would favor the presence of a single mechanism underlying the effect of Citalopram on geotaxis.

The analysis of locomotory patterns was also in agreement with previous observations, whereby we observed a significant reduction in the activity throughout the test session. Locomotory activity was assessed by studying four different three-dimensional metrics: spatial entropy, average speed, average acceleration, and average turn rate. We detected a significant reduction of the average speed throughout the test session, which was mirrored by a similar trend in the average acceleration. This temporal pattern reflects the typical habituation profile that has been observed for zebrafish [73]. In agreement with previous data on acute Citalopram administration [48], we did not register differences in average speed and acceleration among experimental conditions.

While we anticipated avoidance of the robotic stimulus, based on our earlier results in [49], this claim was not confirmed by the present data. Neither control nor treated subjects showed significant aversion towards the side paired with the robotic stimulus during testing. However, our results suggest that males and females responded differently to the robotic stimulus during testing. While females failed to show any side bias regardless of Citalopram administration, males exposed to the lower Citalopram concentration exhibited a significant aversion towards the stimulus-paired compartment. In humans, men and women usually show a differential response to many antidepressants, including SSRIs [74–76]. One explanation for this differential effect is the influence of estrogen on the serotoninergic system [75]. Studies in non-human primates have highlighted that estrogen has a strong influence on serotonin synthesis, transport, and uptake, thereby modulating females’ responsiveness to Citalopram administration [77]. Although an increasing number of rodent-based studies on SSRIs reported sex-specific behavioral responses [78, 79], the effect of sex on zebrafish behavior is still understudied [80, 81].

While we did not aim at disentangling the effects of sex, we decided to test male and female subjects to increase the heterogeneity of our study population, thereby strengthening the generalizability of our findings. As a result, the statistical analysis of the sex factor was inherently underpowered and only allowed the identification of large effect sizes. Under this premise, our findings suggest that male and female zebrafish responded differently to the robotic stimulus and Citalopram administration. Low Citalopram concentration elicited avoidance in male subjects, but not in females. Moreover, although not significant, males had a higher locomotory activity than females, and this trend was abolished by Citalopram administration. Similarly, vehicle-treated females showed a stronger geotactic tendency than males, and the reduction in geotactic response caused by Citalopram was stronger for females.

Albeit characterized by limited statistical power, these preliminary results may highlight sex-specific patterns in the serotoninergic pathways of zebrafish, similar to mammals. Particularly important will be to precisely evaluate zebrafish appraisal of the robotic stimulus, designed to mimic a school of male subjects, by experimental individuals. Sex has an important role on schooling [82], which may reverberate in a differential tendency in our study. Further studies using a modified robotic apparatus, mimicking female or mixed schools, will be necessary to verify this hypothesis.

The response of subjects during training as well as independent binary choice tests points at richer picture of the role of repeated exposure to fear-eliciting robotic stimuli that should be the object of further investigation. While results from the training sessions confirmed our prediction that experimental subjects should exhibit a robust aversion towards the robotic school, binary choice tests did not reveal any aversion for a robotic stimulus, be it stationary or moving. When given a choice between a robotic stimulus and an empty compartment, fish displayed a preference that approximated chance; likewise, when given the choice between a stationary and a moving robotic stimulus, they were not repelled by either. This further strengthens the importance of repeated exposure while devising and conducting conditioned place aversion tests.

We cannot exclude that behavioral sensitization may contribute to explain the difference between the results we observed on aversion in binary choice tests and in the training phase of CPA experiments. A single prolonged exposure to a live predator has been shown to increase geotaxis over time in novel tank diving tests, in contrast with control subjects that decrease geotaxis in time [83]. Stewart et al., 2014 proposed that this behavior could be reminiscent of symptoms of post-traumatic stress disorder wherein a prolonged exposure to an aversive stimulus results into a persistent trace culminating in aberrant responses to the re-exposure to the same stimulus, albeit of a smaller intensity. Similar results were reported in [84], using conspecific alarm substance as a traumatic stressor in a conditioned place aversion paradigm. The relevance of repeated exposures to a biologically-salient fearful stimulus in the induction of behavioral sensitization has been epitomized by Gibson et al. [85] with a study conducted in fruit flies. The authors observed that experimental Drosophila exhibited a fear response following the presentation of “moving overhead translational stimuli (shadows)” and that such a response was stronger if, instead of a single time, the stimuli were presented repeatedly. While future research is needed to clarify this aspect, it is tenable that a combination of repeated exposure and time-dependent sensitization contributed to the aversion observed in our study.

In summary, we provided further evidence for the anxiolytic effect of acute Citalopram administration on zebrafish behavior. Additionally, although we did not observe the predicted avoidance of the robotic stimulus, we pinpointed some factors that can affect the appraisal of its potentially fearful nature. Our study adds to literature aiming at improving the suitability of zebrafish as a model species in psychopharmacology. This claim rests upon the refinements of the proposed experimental paradigm that integrates robotic stimuli, automated three-dimensional scoring of behavior, and a random block statistical design accounting for heterogeneous study populations.

• Robots can replace harmful stimuli in zebrafish fear-conditioning test

• Three-dimensional scoring of zebrafish behavior is important to study anxiety

• Acute Citalopram administration modulates anxiety in zebrafish

• Zebrafish geotactic response decreases with Citalopram concentration

• Sex could have an effect on zebrafish response to Citalopram administration