Coordination and persistence of aggressive visual communication in Siamese fighting fish

Abstract

Outside acoustic communication, little is known about how animals coordinate social turn taking and how the brain drives engagement in these social interactions. Using Siamese fighting fish (Betta splendens), we discover dynamic visual features of an opponent and behavioral sequences that drive visually driven turn-taking aggressive behavior. Lesions of the telencephalon show that it is unnecessary for coordinating turn taking but is required for persistent participation in aggressive interactions. Circumscribed lesions of the caudal dorsomedial telencephalon (cDm; the fish pallial amygdala) recapitulated the telencephalic lesions. Furthermore, ventral telencephalic regions and the thalamic preglomerular complex, all of which project to cDm, show increased activity during aggressive interactions. Our work highlights how dynamic visual cues shape the rhythm of social interactions at multiple timescales. The results point to the vertebrate pallial amygdala as a region with an evolutionarily conserved role in regulating the persistence of emotional states, including those that promote engagement in social interactions.

Keywords: Betta splendens; CP: Neuroscience; aggression; amygdala; animal communication; behavioral neuroscience; emotional states; neuroethology; social behavior; turn taking; visual neuroscience.

Figure 1.. Betta take turns during aggressive displays and scale their aggressive response to match that of their opponent.

(A) Opponents are placed in the same tank and multiple aggressive behaviors (flaring, tail beating, and biting) are scored. (B) Top: first flare and bite start times observed in fighting arenas in Thailand. Dots denote latency to first flare or bite in competitions between independent dyads, and vertical lines denote the mean. Bottom: ethogram raster plot of a single fight in a laboratory setting. (C) Behavioral paradigm consisting of a dyad in neighboring tanks. Videos were recorded from the top and side. (D) Fish key points (red dots) and contour (green outline) are tracked to quantify behavior. (E) Correlation of the proportion of time flaring between randomly chosen opponents 1 and 2. Each point denotes a dyad. (F) Correlation of the proportion of time flaring of either opponent 1 or 2 with the head-to-tail length of their respective opponent (blue) and the difference in head-to-tail length between opponents (red).

Figure 2.. Naturalistic animations evoke as strong a flaring response as a conspecific.

(A) Conspecific paradigm: two opponents in neighboring tanks. (B) Naturalistic animation paradigm: one opponent faces a naturalistic animation modeled after a male displaying aggression (see STAR Methods). (C) Heatmaps showing tank occupancy (from a top view) by fish during isolation and exposure. (D) Average distance to the stimulus. (E) Polar plots showing distributions of the head orientation during isolation and exposure. (F) Proportion of time spent facing within 180° of the stimulus. Dotted line denotes chance level. (G) Average persistence of the proportion of time flaring over the course of the exposure period (dashed lines) with exponential curve fit (solid lines). Arrowheads point to the time when the proportion of time flaring decreases to 0.75× of the max. (H) Proportion of time flaring during isolation and exposure. (D, F, and H) Thin lines denote individuals, and horizontal thick lines denote the median. p values were determined by mixed-model ANOVA.

Figure 3.. Betta coordinate their flare response with dynamic visual cues.

(A) Selected frames from one loop of a naturalistic betta animation. (B) Raster plot (top) and histogram (bottom) showing flaring during each time point in the animation. (C) Z-scored flaring frequency (mean ± SEM). (D) Correlation between flaring of opponent 1 and behaviors of the animation. (E) Correlation between flaring of opponent 1 or 2 and behaviors of their opponent. (F and G) Peri-event time histograms (mean ± SEM) of changes in flare probability, lateral orientation probability, and normalized elevation (see STAR Methods) of the virtual fish (F) or real opponent (G) aligned to either the onset (left) or offset (right) of flaring bout of the opponent.

Figure 4.. Flaring of a stimulus is not necessary for synchronizing flaring but promotes more persistent flare responses.

(A) Behavior setup for presenting two phasic animation types (no flare or flare) to fish in a balanced order. (B) Animations were composed of two alternating phases with distinct combination of flare, orientation, and speed state. (C) Frequency of flaring bout onsets (mean ± SEM) aligned to a loop of the animation. (D) Proportion of time flaring against each animation. (E) Average flaring onset frequency for each phase (1 or 2) in each animation. (D and E) Thin gray lines denote individuals, and horizontal thick lines denote the median. p values were determined by a two-tailed paired t test (D) or mixed-methods ANOVA (E). (F) Average persistence of the proportion of time flaring over the course of the exposure period (dashed lines) with exponential curve fit (solid lines). Arrowheads point to the time when the proportion of time flaring decreases to 0.75× of the max.

Figure 5.. Flaring increases when the opponent is close to the surface.

(A) Left: phasic animation in which the stimulus maintains its elevation. Right: animation was shown at different distances from the surface and bottom of water by varying the elevation of the animation and the height of the water. (B) Proportion of time flaring against stimuli with varying distances from surface and bottom. (C) Proportion of time flaring against the naturalistic animation versus the phasic animation at elevation closest to the surface. (B and C) Boxes denote interquartile range with whiskers at 1.5× the interquartile range and lines at the median. (D) Heatmaps of exemplar individual betta against each elevation condition.

Figure 6.. Caudal Dm promotes persistent engagement in an aggressive interaction.

(A) Left: betta were exposed to either a conspecific or empty tank for a brain-wide activity screen. Right: example sections of rDm and cDm with pS6+ cells. (B) Quantification of pS6+ cells in rDm and cDm following conspecific or empty tank exposure. p values were determined by Welch’s t test. For other brain regions, see Figure S9. (C) Custom survival surgery set up for Dm lesioning and experimental timeline. (D) Example sections showing targeted lesions of rDm and cDm. Arrows point to the brain region removed by the lesion. (E) Example telencephalon removal with intact optic lobes and optic nerves following removal. (F) Correlations between flaring of fish with rDm (red), cDm (blue), or telencephalon (purple) lesions pre- and post-surgery with behaviors of the naturalistic animation. (G) Change in the proportion of time flaring toward animation (left) and conspecific (right) following lesions of rDm, cDm, or telencephalon, normalized by change in the proportion of time spent flaring of sham-operated individuals. p values were determined by mixed-methods ANOVA. (H–J) Normalized average persistence of time flaring over the course of the exposure (left: animation; right: conspecific) period (dashed lines) with exponential curve fit (solid lines), normalized to the maximum flaring of each individual. Arrowheads point to the time when the proportion of time flaring decreases to 0.5× of the max following sham or rDm lesions (H), sham or cDm lesions (I), and sham or telencephalon lesions (J).

Figure 7.. Neuronal projections to rDm and cDm.

(A) ctb-555 was injected into either rDm (n = 3) or cDm (n = 3). (B) Diagram of afferents to rostral and caudal Dm. All ctb+ regions were consistently found in all three fish injected in each region. (C–I) Coronal sections of fish injected into rDm (left) or cDm (right) ordered along rostro-caudal axis with areas of interest highlighted. All highlighted areas were found in 3/3 rDm- or cDm-injected fish. Cp, central posterior thalamic nucleus; Dc3/4, central part of the dorsal telencephalon, subdivision 3/4; Dp, dorsal posterior thalamic nucleus; Hc, caudal zone of periventricular hypothalamus; Hd, dorsal zone of periventricular hypothalamus; ICL, internal cellular layer of the olfactory bulb; lHb, lateral habenula; NLTd, lateral tuberal nucleus, dorsal part; PGc, caudal preglomerular nucleus; PGl, lateral preglomerular nucleus; PGm, medial preglomerular nucleus; PPa, parvocellular preoptic nucleus, anterior part; PPp, parvocellular preoptic nucleus, posterior part; Ppd, dorsal periventricular pretectal nucleus; TL, torus longitudinalis; TPp, periventricular nucleus of the posterior tuberculum; Val, lateral division of valvula cerebelli; Vp, postcommissural nucleus of the ventral telencephalon; Vs, supracommissural nucleus of the ventral telencephalon. Nomenclature was informed by betta, cichlid, and zebrafish atlases. Scale bars: 400 μm.