Development of social behaviour in young zebrafish

Adult zebrafish are robustly social animals whereas larva is not. We designed an assay to determine at what stage of development zebrafish begin to interact with and prefer other fish. One week old zebrafish do not show significant social preference whereas most 3 weeks old zebrafish strongly prefer to remain in a compartment where they can view conspecifics. However, for some individuals, the presence of conspecifics drives avoidance instead of attraction. Social preference is dependent on vision and requires viewing fish of a similar age/size. In addition, over the same 1–3 weeks period larval zebrafish increasingly tend to coordinate their movements, a simple form of social interaction. Finally, social preference and coupled interactions are differentially modified by an NMDAR antagonist and acute exposure to ethanol, both of which are known to alter social behavior in adult zebrafish.


Introduction
Human infants exhibit social behaviours from birth (Xiao et al., 2014). Throughout 3 life these innate social drives provide the substrate for learning more complex forms 4 of human interaction. Disruptions to early social behaviour may impair the 5 development of normal adult sociality, and may contribute to disorders such as autism 6 (Banerjee et al., 2014). Since the neural circuitry that underlies human innate social 7 behaviour is established in utero, very little is understood about its normal and 8 pathological development, anatomy, and function. 9 Early developing social behaviours, such as the preference to observe and however, do not exhibit the overt shoaling and schooling behaviours that are readily 23 apparent in adults. In order to shoal fish must prefer to approach and remain near 24 conspecifics. There is some evidence that this preference might develop as early as however, requires the group to move in a polarized and coordinated manner and has 28 thus far only been described in adults(Miller and Gerlai, 2012). 29 Social behaviour encompasses more than simply preferring to be near 30 members of the same species. For instance, individuals may coordinate their 31 behaviour with other members of the same social group. Such coordination is obvious 32 in the case of schooling fish, where individuals align their body orientation and 33 synchronize their movements, but it is also present in social mammals. For example, 34 humans will unconsciously coordinate a diverse range of behaviours, such as yawning, 35 eye blinks and posture (Richardson et al., 2007;Sebanz et al., 2006), and this is 36 thought to provide a foundation for more elaborate forms of social communication 37 and cooperation. 38 Here we set out to investigate early social interactions in zebrafish and to 39 determine if the establishment of preference for the presence of conspecifics is 40 contemporaneous with individuals beginning to coordinate their behaviour. We have 41 designed a novel social preference/interaction assay for zebrafish larvae that 42 continuously monitors, with high temporal resolution, the detailed behaviour of 43 individuals freely choosing to observe or avoid a counterpart. This assay demonstrates 44 that social behaviours develop gradually and are robust by three weeks post-45 fertilisation. We have also used the assay to characterize the effects of substances 46 known to influence the social behaviour of adults. 47 48

50
Zebrafish husbandry. AB strain zebrafish (Danio rerio) to be tested were bred, raised 1 and housed in the same environment. All fish were obtained from natural spawning 2 and housed in groups of roughly 50 fish, and kept at 14h light/10h dark cycle. Fish 3 were fed two times per day from 4 dpf with dry food diet from SAFE Diets (particle 4 size 50-100) and twice with salt water rotifer (Brachionus Plicatilis) until 10 dpf; then 5 twice a day with dry food diet (particle size 100-200), and with a combination of salt 6 water rotifer and brine shrimp (Artemia salina) until 15 days; finally twice a day with 7 dry food diet (particle size 200-300) and with brine shrimp until used in the 8 experiments. All the fish run in the behavioural assay were fed in the morning. All 9 test fish were paired with age-matched siblings as the social cue. The experiments 10 described were approved by local ethical committee and the UK Home Office. 11 12 Behavioural assay. Experiments were performed in a custom-built behavioural setup 13 (Fig. 1A) that was assembled from structural framing (Misumi, DE) and 14 optomechanics (Thorlabs, USA). The videography system comprised a high-speed 15 camera (Flea3, PointGrey, CA), infrared LED backlight (Advanced Illumination, 16 USA, 880 nm), infrared filter (R70, Hoya, JP), and a vari-focal lens (Fujinon, JP).

17
Fish were imaged in a custom-built behavioural arena that was fabricated with a laser-18 cutter from 5 mm thick opaque white acrylic, sealed with silicone, and with glass 19 window partitions; the multi-chamber design is shown in Fig. 1A. The dimensions of 20 a single behavioural arena was 4 cm x 3.2 cm. The viewing chamber that contained a 21 single or multiple SC fish was 1.5 cm square, and the width of the passage between 22 the two arms of the arena was 6 mm. The water level height was 5 mm, the 23 temperature of the water was around 25'C, pH was 7.0, and the conductivity was 445 24 µS.The arena was supported on a transparent base covered on one side with diffusive 25 gel paper (Rosco Cinegel, USA  Fig. 1B) from the number of frames spent 43 in the equivalent region on the opposite side of the chamber (blue tracking in Fig. 1B).

44
This difference was then divided by the total number of frames recorded in the two 45 analysis compartments during the experiment, resulting in a value varying between -1 46 and 1.
["SPI = (SC frames -Non SC frames)/ Total frames"]. The SPI during the 47 acclimation period, for which there is no SC, was computed with reference to the 48 randomized side of the chamber on which the SC would be added in the subsequent 49 experimental phase. 50 The compressed video from each experiment could be repeatedly re-analysed using 1 custom written bulk-processing routines in Python (https://www.dreo-2 sci.com/resources/). A motion trajectory for each fish was computed by first 3 segmenting the binary particle for each fish from the background and then measuring 4 the change in pixel values, for that particle, from one frame to the next. This resultant 5 frame-by-frame segmented motion value provided a very stable time series for 6 identifying the peaks of individual bouts and then testing for interaction between the 7 observer and SC fish. 8 9 Statistical analysis: 10 The SPI distributions for many Ac vs SC conditions (one week old, MK-801, EtOH, 11 etc.) are normally distributed, however some conditions (in which the value 12 approaches the SPI bound of +/-1), are clearly non-normal. We therefore used the 13 same non-parametric statistic test for all comparisons in the study: a Wilcoxon 14 signed-rank test of paired samples. solution in fish water in order to obtain a working concentration of 100 µM. Zebrafish 20 were washed with fish water before placing them in the chamber for recordings.

21
For ethanol experiments, low (0.125%) or a high ethanol (0.5%) concentrations were 22 obtained by diluting ethanol in fish water. Fish were exposed with one of the two 23 ethanol concentrations for 1 h prior to, and during experiments. 24 25 26

28
Fish develop strong social preference and interactions by three weeks of age 29 We designed a behavioural chamber in which zebrafish fish could swim freely 30 between two arms, but in only one they could view conspecific siblings through a 31 glass partition. Six such chambers were simultaneously monitored with an infrared 32 high-speed camera and automated tracking software recorded the behaviour (position 33 and orientation) of the observer fish ( Fig. 1A, Movie 1). Following 15 minutes in the 34 chamber without conspecifics (acclimation (Ac) period), a single or three conspecifics 35 were added to one of the adjacent compartments, randomly selected, and the 36 behaviour of the observer fish was monitored for an additional 15 minutes (social cue 37 (SC) period). There was no bias between compartment arms in the acclimation phase 38 for fish at any age, nor if the fish were monitored for a further 15 minutes following 39 the acclimation phase without adding the SC (Supp. Figure 1E-F). 40 Three week old zebrafish consistently showed a very strong bias to remain in 41 the arm of the chamber adjacent to the SC (Fig. 1B, compare Supp. Movie 2 (one 42 week) to Supp. Movie 3 (three week)). To quantify the tendency for each tested fish 43 to spend time in one or the other arm of the chamber, we defined a social preference 44 index (SPI) (see Materials and Methods). A positive SPI indicates a preference for the 45 chamber arm with the SC and a negative SPI indicates an aversion for the SC. The 46 SPI was computed for all tested one, two and three week old fish with and without the 47 presence of multiple (Fig. 1D) or a single conspecific ( Fig 1E). One week old larvae 48 exhibited a very weak, but significant, preference for an SC arm (Fig. 1D, top; Ac vs 49 SC p=0.006) containing multiple conspecifics, however, this preference bias was 1 absent when only a single fish was placed in the SC arm ( Fig. 1E, top; Ac vs SC 2 p=0.9). In contrast, the SPI of two week old larvae was strongly shifted towards 3 positive values when viewing both multiple (Fig. 1D, middle; Ac vs SC, p= 4.4*10 -10 ) 4 and single conspecifics (Fig. 1E, middle; Ac vs SC, p=1.4*10 -11 , Supp. Fig. 1D). By 5 three weeks, SC preference strengthened further, with many values close to 1, 6 reflecting the strong bias of some observer fish to remain almost entirely on the side 7 of the conspecifics (multiple: Fig. 1D, bottom; Ac vs SC, p=2.0*10 -13 , single: Fig. 1E,  8 bottom; Ac vs SC,p=1.4*10 -15 ). 9 A small minority of three week old fish had strong negative SPIs (Fig 1D,  10 bottom). These fish exhibited an aversive response to the conspecifics, preferring to 11 stay in the chamber away from the SC (Supp. Fig. 1B and Supp. Movie 4). Such 12 aversive behaviour was rarely observed in younger fish suggesting that, as for positive 13 social interaction, social aversion also increases throughout development. either the left or right eye was found in this assay (Sovrano and Andrew, 2006).

22
We next set out to investigate what sensory cues contribute to the displayed 23 social preference.

25
Social preference requires visual observation of conspecifics with a similar age 26 Although visual stimulation seemed the most likely source of the preference for the 27 SC, it was possible that some olfactory or tactile cues may pass between the chambers 28 of the observer and SC fish. Consequently, we compared preference behaviour for 29 three week old fish tested in the dark to those tested in light ( Fig. 2A).

30
Removal of background illumination completely abolished the tendency of 31 observer fish to orient towards the conspecific viewing chamber (Fig. 2B) and social 32 preference was abolished, as evidenced by the distribution of SPIs (Fig. 2C, and Supp. 33 Fig. 2A; Ac vs SC, p=0.57). Furthermore, with normal background illumination, 34 replacing the transparent window with an opaque barrier also eliminated preference 35 for the SC (Supp. Fig. 2B). These experiments provide strong evidence that the social 36 preference behaviour of three week old zebrafish depends on vision. 37 The data above indicates that during the first three weeks of their life, 38 zebrafish develop a robust social preference to view age-matched conspecifics. 39 However, during this time they also change significantly in size, doubling their head 40 to tail length (Supp. Fig. 1A). To assay whether the size/age of the SC fish influences 41 social preference, we monitored the behaviour of one and three week old zebrafish 42 presented with larger/older or smaller/younger conspecifics as the SC (Fig. 2D and  43 Supp. Fig 2C). 44 One week old fish not only showed no preference for three week old fish, but 45 also a slight aversion to viewing the larger fish, supporting the conclusion that the 46 development of social preference reflects maturation of the observer and is not simply 47 dependent on the age/size of the stimulus (Fig. 2D left; Ac vs SC, p=0.002; SPI 48 difference (Ac-SC) = -0.111). Three week old fish also did not exhibit a strong social 49 preference when presented with one week old fish as the SC (Fig. 2D, right; Ac vs SC, 50 p=0.02; SPI difference (Ac-SC) = 0.106). However, the broader distribution of SPIs 1 suggests that the smaller/younger fish may still influence the behaviour of the 2 larger/older observing fish, which could be due to fish becoming progressively more 3 responsive to any moving objects within their environment.

5
Zebrafish coordinate their movement 6 Three week old fish display robust visually-driven social preference; high-speed 7 videography additionally allowed us to investigate the extent to which the behaviour 8 of the SC fish influenced the behaviour of the observer (Supp. Movie 5). 9 Young zebrafish tend to move in small bouts of activity consisting of discrete 10 swims or turns (Orger et al., 2008). Individual bouts were detected by identifying a 11 peak in the motion tracking signal (Fig. 3A, top trace). Averaging all of these bout 12 time-courses (Fig. 3B) revealed a pre-and post-bout quiescent period, the timings of 13 which reflected the periodicity of movement. These quiescent periods shortened from 14 one to three weeks of age as the mean bout frequency increased ~50%, from 0.79 Hz 15 at one week to 1.22 Hz at three weeks. As observed in other behavioural contexts, 16 these movement bouts were composed of a mixture of forward swims and orienting 17 turns ( Fig. 3C)(Orger et al., 2008). 18 We next asked whether a motion bout produced by the SC fish influenced the 19 movement of the observer. Short time windows of the motion trajectories from the 20 observer fish, normalized by each individual's average motion peak, were extracted 21 and aligned to the bouts of the SC fish (Fig. 3A, middle trace) and were then averaged 22 over all bouts (Fig. 3A, bottom trace). This generated a 'bout triggered average' 23 (BTA), which is an estimate of how the motion bout of the SC influences movement 24 of the observer.

25
A clear interaction between the movement of the SC fish and the observer was 26 present at all stages of development. Notably, a bout of motion by a SC fish was, on 27 average, coincident with a synchronous increase in motion by the observer fish. The 28 strength of this motion coupling increased substantially over development (Fig. 3d), 29 correlating with the enhancement in positive social preference. This visual coupling 30 behaviour was, unsurprisingly, absent for fish in the dark (Fig. 3E). These results 31 indicate that not only do three week old fish prefer to be with conspecifics, but that 32 their behaviour is more strongly coupled with that of their social partners. 33 34 Social preference and interaction are differentially impaired by drug exposure 35 We next assayed whether pharmacological manipulations that affect sociality in adult 36 animals similarly influence the manifestation of social behaviour in young zebrafish. the overall amount of swimming. However, we found that three week old fish 45 exhibited no social preference ( Fig. 4A and Supp. Fig. 2D; Ac versus SC period, 46 p=0.51). This result suggests that blocking NMDA receptors interferes with circuitry 47 required for social interactions, both in larvae and in adult fish. In addition, video-48 tracking revealed a significant alteration of movement dynamics in MK-801 treated 49 larvae. The treated fish produced swim bouts lacking the pre-and post-bout quiescent 50 periods (Fig. 4B), and a near total loss of conventional forward swims (Fig. 4C). 1 Every bout involved a change in orientation (i.e. turning), which is consistent with the 2 "erratic movements" observed in MK-801 treated adult fish(Sison and Gerlai, 2011) 3 and reveals a substantial disruption to the movement control system after drug 4 treatment. These altered bout dynamics also produced an asymmetry in the pattern of 5 social interaction (Fig. 4D); the motion of the observer fish strongly influenced the 6 movement of the untreated SC fish (dip prior to 0 ms), but the drug-treated observer 7 was much less influenced by the movement of the SC fish (smaller dip after 0 ms; Fig.  8 4D). Furthermore, the synchronicity peak at 0 milliseconds lag was abolished. 9 Acute exposure to high concentrations of ethanol are also known to influence Consequently, we exposed three week old fish to low (0.125%) and high (0.5%) 12 levels of ethanol 1 hour prior to and during testing in the social assay ( Fig. 4E-L). The 13 influence of ethanol exposure was concentration dependent. Fish exposed to low 14 ethanol retained a strong SPI (Fig. 4E, and Supp. Fig. 2D; Ac vs SC period p= 5.8*10 -15 8 ) and their bout dynamics (Fig. 4F) and composition (swims vs turns) (Fig. 4G) were 16 unaffected. Furthermore, the strength of their BTA interaction was similar to age-17 equivalent untreated fish. In contrast, upon exposure to a higher concentration of 18 ethanol, social preference was greatly reduced and the SPI distribution was not 19 significantly different from the acclimation period ( Fig. 4L and Supp. Fig. 2D, Ac vs 20 SC p=0.07). Remarkably, despite this loss of social preference by zebrafish exposed 21 to high ethanol concentrations, movement dynamics (Fig. 4J), distributions of swim 22 turns (Fig. 4K), and the strength of movement coupling with other fish (Fig. 4L) were 23 not substantially affected. These intriguing results suggest that social preference and 24 interactions with other individuals, each a fundamental component of social behaviour, 25 can be decoupled by pharmacological, and likely other, manipulations. 26 27 28

30
The development of social preference 31 We have shown that zebrafish gradually develop a "social" preference, which we 32 define as the tendency to remain in a chamber that provides visual access to 33 conspecifics; a behaviour that is absent in one week old fish, begins to emerge by two 34 weeks, and is very robust at three weeks. This preference is visually driven and does 35 not solely depend on the age/size of the conspecific partners as one week old larvae 36 show no interest in larger three week old fish. These results suggest that social 37 preference arises with the development of neural systems that appear or mature during 38 the second and third weeks of life. For instance, it is known that some brain areas, 39 such as the pallium (Dirian et al., 2014), undergo extensive growth during the 40 establishing period of social preference. 41 Whether the preference to observe conspecifics reflects a drive to shoal/school 42 or aggression (Gerlai et al., 2000) is difficult to distinguish with this assay. Both of 43 these behaviours involve multi-modal stimuli (olfactory and tactile), which are 44 prevented in our assay by the transparent barrier dividing the observer and social cue. 45 Nevertheless, our assay demonstrates that visual cues alone can drive social 46 behaviours and is thus easily translated to experimental setups that require restrained 47 fish, but allow for detailed investigation of the underlying neural activity, e.g. two-48 photon microscopy. 49 We have not yet fully characterized the specific visual features that drive 1 social preference. However, a simple preference for moving stimuli is unlikely to 2 explain the response since three week old fish show little interest in viewing moving 3 younger/smaller conspecifics. Our assay has demonstrated that visual stimulation is 4 sufficient to drive social behaviour at three weeks of age. The presentation of visual 5 "social" stimuli to restrained fish is much more straightforward than attempting to 6 recapitulate the complex tactile and olfactory stimuli that are also involved in 7 schooling/shoaling interactions (Cornelia H et al. 2012). Therefore, this social 8 behaviour assay will considerably facilitate future studies to characterize the changes 9 in neural circuitry that correlate with this fundamental behavioural transition. 10 It is intriguing that not all fish develop a positive response to conspecifics as 11 some individuals exhibit avoidance behaviour when other fish come into view. This 12 result warrants further investigation. For instance, our assay could be used to 13 determine whether fish exhibiting aversive behaviour retain this negative social bias 14 after multiple presentations of the SC or whether different environmental, 15 pharmacological and genetic manipulations predispose developing zebrafish to 16 express more aversive or attractive social behaviours.

18
Social interaction as a coupled dynamic system 19 We found that when a fish observes the movement of a conspecific, its own 20 swimming is affected. This visually-mediated coupling of movement is already 21 present in one week old larvae, but strengthens considerably over the following weeks. 22 Coupling the motion of one fish to that of another is an important prerequisite for the 23 coordinated behaviour that predominates in groups of schooling fish (Miller and 24 Gerlai, 2012). The temporal profile of this movement coupling, notably its remarkable 25 synchronicity, is reminiscent of the coordinated movements apparent in other social 26 organisms, including humans (Richardson et al., 2007;Sebanz et al., 2006). 27 However, synchronization of behaviour is observed in many physical systems with 28 coupled dynamics, such as two metronomes on a shared surface (Pantaleone, 2002) as 29 well as for many biological rhythms (Winfree, 1967). If any two periodic movement 30 generators are sensitive to the motion of one another, then they will act as coupled 31 oscillators and will have a natural tendency to synchronize. We have found such a 32 coupling of observation of movement to movement generation in young zebrafish, but 33 whether such coupling dynamics are important for the shoaling/schooling interactions 34 of adult fish, or any other species demonstrating coordinated synchronous movements, 35 warrants further investigation. 36 Disruptions to the coordination of behaviours, such as the loss of synchronized eye-37 blinking in autistic subjects (Sears et al., 1994;Senju et al., 2007), are now being 38 identified as potentially important biomarkers of disease that may facilitate early 39 diagnosis and intervention.

41
Pharmacological manipulation of early social behaviour 42 The NMDA receptor antagonist MK-801 disrupted both social preference and 43 interaction, whereas alcohol exposure disrupted only preference, leaving intact the 44 ability of fish to couple their movements. This suggests that these two aspects of the 45 social behaviour can be at least partially disassociated. 46 In addition to confirming that acute treatment with 100 µM MK-801 disrupts 47 social preference in larvae, as was previously shown in adults(Sison and Gerlai, 2011), 48 we also found that it greatly alters underlying movement dynamics. MK-801 treated 49 larvae show strongly reduced bout periodicity and do not produce conventional 50 forward swims, which could explain the observed deficit in coordinated behaviour. 1 Such movement impairments will also affect the ability of treated fish to shoal, and 2 might explain why adult zebrafish exposed to a lower concentration (5 µM) of MK-3 801 exhibited disrupted shoal cohesion(Maaswinkel et al., 2013). 4 In contrast to NMDA receptor blockade, fish exposed to high concentrations 5 of ethanol exhibited no disruption of intrinsic movement dynamics and show wild-6 type levels of movement interaction, but social preference was severely disrupted. 7 These results highlight our assay's sensitivity to distinguish components of social 8 behaviour, preference and interaction, which could be separately impaired by 9 different pathologies. Consequently this assay should be well suited for analysis of a manipulations that have been linked to developmental disorders affecting social 12 behaviour.

14
Future Directions 15 If we are to identify and characterize the neural circuits that underlie the development 16 of social behaviour in zebrafish, it will be necessary to adapt our assay to enable the 17 monitoring and manipulation of neural activity in vivo during social interactions. 18 Fortunately, the brains of zebrafish are still small and relatively transparent during the 19 developmental stages at which we observe the onset of social preference and coupled 20 interactions, and are thus amenable to optical and optogenetic techniques for 21 anatomical and functional investigation of whole-brain circuitry(Ahrens and Engert,                         shows the test chamber indicating in red the side in which the SC fish is visible and in 47 blue the side in which it is not. The plots on the right indicate how movement bouts 48 were analysed. Top plot shows movements bouts of the SC fish. Peaks in movement 49 trajectories were identified with a dual-threshold algorithm (upper threshold dotted 50 line is 3*standard deviation (3*SD) and lower threshold dotted line is 2*SD from 1 baseline). The middle plot shows the movement bouts of the observer, test fish. The 2 movement peaks of the SC fish were used to extract short time windows of the 3 movement trajectories of the observer fish trajectory (2 seconds either side of the SC 4 fish movement peak). The bottom plot shows the 'bout-triggered-average' (BTA) 5 movement for the observer fish which was computed by averaging movements across 6 all of the four second time windows aligned to the SC peak movement. BTAs were 7 computed separately depending on whether the observer fish could view the SC or not 8 (left schematic). B, The average bout motion time-course for all SC fish, normalized 9 to the peak movement of each fish, at different developmental stages. The average 10 bouts are overlaid to highlight changes in the kinetics between one and three weeks of 11 age. C, Scatter plot presentation of all bouts, where each bout is represented by a 12 single point that indicates the change (Δ) in position and orientation that occurred 13 during that bout (n = 247779 bouts). D, BTAs of one-to three week old observer fish 14 motion aligned to movement bouts of single SC fish (red) or plotted when the SC was 15 not visible (blue) (one week: n = 106, two week: n = 136, three week: n = 163). E, 16 BTAs for fish monitored in darkness when on the same (red) or opposite side (blue) 17 of the SC (n = 90). Histogram of SPIs revealing no apparent preference for the SC and (inset) body 24 orientations showed little or no direction towards the SC chamber (zero position). 25 SPIs during the acclimation periods are shown in Supp. Figure 2D.