The aim of this study was to model in zebrafish the effects of shank3 genetic mutations on the social contagion of an affective state, to identify underlying attentional and affective-sate recognition mechanisms, and to quantify associated changes in neuroplasticity at the gene expression level.

Animals and husbandry

We used naïve adult zebrafish, Danio rerio, aged 6–12 months of the genetically modified shank3a mutant line (n = 12) and its wild-type background siblings as controls (n = 11). Fish were raised under laboratory conditions in the fish facility of the Gulbenkian Institute of Science, housed in groups of 11 fish in 1.5 L aquaria of a recirculating system (ZebraTec, 93 Tecniplast) kept at 27–28OC, 7.5 ± 0.2 pH, 1000 μSm conductivity and 14 L: 10 D photoperiod. Animals were fed a combination of live (Artemia salina) and dry food (Gemma). Welfare and health-maintenance protocols included a previously described approach where animals were kept with minimal external stress, full social and environmental enrichment, regular observational body condition and health checks, and free from known pathogens via sentinel testing.

Genotypic characterization of shank3 mutants

To generate mutants for shank3a, the following Crispr target sequence for exon2 was identified using the ChopChop CRISPR design tool (chopchop.cbu.uib.no), GGCTCTGGTTGAGTGTGCAG. The generation of the sgRNA guide was obtained by using the technique described by Gagnon et al. [44]. The following DNA sequence was ordered from Invitrogen, TAATACGACTCACTATAGGCTCTGGTTGAGTGTG CAGGTTTTAGAGCTAGAAATAGCAAGTTAAAATAAGGC TAGTCCGT TATCAACTTG AAAAAGTGGCACCGAGTCGGTGCTTTTAAA, and annealed to the complementary oligo as described in the paper by Gagnon et al. The gRNA was generated using the HiScribe T7 High Yield RNA Synthesis Kit (NEB) followed by DNase I (NEB) digestion and purification with RNeasy MiniKit (Qiagen). The Cas9 from plasmid pT3TS-nCas9n (Addgene) was linearised with XbaI (NEB) and capped mRNA generated with the mMessage mMachine T3 Transcription Kit (Life Technologies) followed by polyadenylation with the Poly(A) Tailing Kit (Life Technologies). The synthesised mRNA was purified using the RNeasy MiniKit. Cas9 mRNA and shank3a gRNA were co-injected into one-cell stage embryos at a concentration of 200 ng and 75 ng per embryo, respectively.

Genomic DNA was extracted from single injected embryos by incubation in 50 μl base solution (1.25 M KOH and 10 mM EDTA) at 95 °C for 30 min followed by addition of 50 μl neutralization solution (2 M Tris HCL). To identify mutants, a region of 291 bp around the CRISPR target site was PCR-amplified with Taq DNA polymerase (Invitrogen) using primers Fwd 5′- GCT CTG GTG ACT TTG GTT GA -3′ and Rev 5′- CCT TCA CAC AGG TCA GAG AAG -3′. A frameshift mutation with a deletion of 5 bp in exon2 was identified (sequence CGAATCCTCTGCACACTCAACCA; deletion shown in bold), leading to compromised downstream functional domains according to gene structure (Ensembl ID: ENSDARG00000063332; NCBI ID: 557,701) and resulting to severe changes to amino-acid sequence and consequent alterations in protein structure compared to the wild-type protein (Fig. 1a–c). To confirm the expected non-sense mediated decay effects of this frameshift mutation, we ran whole-brain qPCR analyses in a sub-sample of our experimental animals and calculated relative expression levels of shank3 RNA, which were found significantly reduced in mutant animals compared to wild types (t 16 = 2.66, P = 0.0172; Fig. 1d).

Experimental set-up and procedure

To test social fear contagion, we used a previously validated two-alternative video-playback approach [19]. The custom-built set-up includes a long corridor (arena: 14.5L, 29.5 × 14.5 × 11 cm) placed on top of an infra-red lightbox inside a dark cabinet, with two monitors on either side (Asus VG248, 1080 HD, 144 Hz rapid refresh rate) presenting focal fish with concurrent pre-recorded videos of a conspecific demonstrator, and with an overhanging camera for recording (Fig. 2a). The corridor is separated in three equal parts constructed by removable transparent dividers. This enables the use of a two-part protocol where focal animals are first acclimatised in the central compartment and allowed to observe and encode conflicting behavioural states presented between the two monitors and then given access to either monitor to test discrimination via local preferences (Fig. 2b).

Fig. 2

Experimental assessment of shank3a mutation effects on social contagion and underlying recognition and attention components. (a) The set-up included a corridor separated in three equal parts by removable transparent dividers, a camera with a birds-eye-view of the arena and monitors on either side of the tank for displaying demonstrator videos. (b) The experiment included three phases. Acclimation to the arena and background videos for 10 min, where baseline mobility during the last 5 min could be assessed. Observation of two contrasting videos from the central compartment for 5 min, where the demonstrator exhibited either neutral behaviour (control) or periodic distress (stimulus: erratic and freezing). Test of local preferences for 10 min, following removal of dividers and access to the whole tank, while both videos displayed the demonstrator in a neutral state. During the observation phase, fish mobility, orientation towards the distress behaviour (heading: 0–180°) and repetition of the observed erratic and freezing behaviour were measured. During the test phase, discrimination based on the local preference for either video was used to assess the recall of recognised differences between distress and neutral state. (c) Mobility in terms of total distance travelled, was lower in Shank3a mutants only during observation phase, suggesting no motor deficits are present. (d) Temporal changes in the directional changes exceeded baseline thresholds, and together with the proportion time erratic response was exhibited, following analogous behaviour in the stimulus video, were markedly lower in mutants, compared to wild-type animals. (e) Immobility, used to measure freezing, was greater in mutants, but this related to temporal differences in velocity that revealed an overall low activity in mutants compared to the freezing bouts in wild types. (f) Attention towards the distressed stimulus, compared to the neutral control, was greater in the wild-type animals but not in the Shank3a mutants. (g) Local preference scores revealed recognition deficits in the Shank3a mutants compared to wild types. Heat maps are representative examples with the least deviation from the mean. [*P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001]

During observation, the two playbacks exhibit the same demonstrator in two discrete states: a consistent neutral state (baseline swimming; control) on one monitor and an intermittent distress state (three bouts of erratic and freezing acts; affective stimulus) on the other. This allows the quantification of contagion in observers (erratic and freezing behaviour) and shifts in attention between observed neutral and distress states, while the use of the same demonstrator controlled for effects from individual differences that may be independent from behavioural state. During the observation phase animals are also able to collect information on the state presented on either monitor, which is required for recognition and quantified in the test phase. During the test phase, both monitors present the demonstrator in a neutral state, and thus cumulative local preferences can indicate discrimination dependent on the recall of previously recognised differences in observed behaviour (distress from neutral) rather than any effects from current differences in stimuli, e.g. differences in movement or in signals of imminent local danger.

Demonstrators and video-playbacks

Playback videos were constructed by recording two naïve wild-type zebrafish unfamiliar and unrelated to the focal fish. Recordings were done in 1.5-L tanks, which were first video captured before the introduction of the demonstrators for use during the acclimatization phase of experiments. The tanks included an overhanging PVC tubing (diameter: 0.8 mm internal, 2.4 mm external) and obscured by their surroundings by opaque covers, but in view of a camera (goPro hero3 + , 60 fps, 1080 pixel resolution) kept in place outside one of the tanks’ glass walls, and behind an opaque acrylic sheet with a customised cut-out for the camera lens, that could be remotely operated to record behaviour. Demonstrators were individually kept in these tanks overnight and recorded the following day. During recording, we extracted a 200 s capture of the demonstrator in undisturbed baseline behaviour (neutral swimming) followed by the introduction of 0.75 ml of alarm substance from the overhanging tube and the recording of the behavioural response. The alarm substance used to elicit demonstrator response was extracted from live fish and appropriately stored using a previously reported protocol [19]. The response consisted of a stereotyped repertoire of erratic behaviour, i.e. fast zig-zagging movements, and freezing, characterised by immobility on the bottom of the tank with fast opercular movement.

Videos were then edited (VSDC© software; v. 6.3.6.18; Flash-Integro LLC, 2019) and included a 10-min playback of the demonstrator housing tank, used as background during acclimation; a 5-min video with the demonstrator exhibiting baseline swimming behaviour (neutral control), which included 3 repetitions of a 100 s recorded period; a 5-min video of the demonstrator periodically exhibiting distress, including 3 repetitions of a 60 s swimming period followed by a 40 s bout of the distress behaviour, i.e. erratic and freezing. During tests, videos were scaled on monitors to live-size proportions and focal fish were exposed to one of two playback replicates, one of a male and one of a female demonstrator, counterbalanced across individuals of either genotype (shank3a mutants or wild types). This controlled for demonstrator effects being either sex-specific or due to any individual phenotypic property.

Behavioural test and recording protocols

In order to limit effects from prior social interactions, on the eve of experimental testing animals were moved to overnight individual housing in 1.5 L tanks kept at the same conditions as their original housing. On the day of testing animals were individually placed in the central division of the experimental arena and allowed to acclimatise for 10 min to playbacks of the empty demonstrator tank. Then, animals were kept in the central compartment and allowed to observe the two videos of the conflicting behavioural states (neutral and distress) for 5 min, with the side on which either state was presented being counterbalanced across experimental animals to control for side biases. After this period, videos shifted to the 10-min presentation of the neutral behaviour playback on both monitors, and the dividers delimiting the central compartment were lifted, allowing access to the whole arena (Fig. 2b). This enabled the quantification of cumulative local preferences towards either playback. Both the observation and test phase were recorded from above using via an infra-red sensitive camera (Henelec 300B; acquisition at 30 fps) for enhancing contrast against the infra-red light-box on the bottom of the tank. Videos were fed to a remote computer and recorded via the Pinnacle Studio software (v. 12, http://www.pinnaclesys.com). Individual recordings were then analyzed using the commercially video tracking software Ethovision XT© 11.0 (Noldus Inc., The Netherlands).

Behavioural data extraction

From the recordings of each phase of the experiment, the movement of animals was automatically tracked and movement data (x, y coordinates) analysed to measure behaviour (Additional file 3: Data S1). Across experimental phases, we extracted overall mobility measures using the total distance travelled by animals in each period, including baseline mobility during the last 5 min of the acclimation period, as well as mobility during the observation and test phases. From movement data during observation, within the central compartment in which animals were restricted, attention to videos was measured by the absolute compass heading (x direction relative to the distress-stimulus video, ranging from 0° to 180°) and contagion was measured by the proportion time spent in erratic movement [acceleration > 8 cm/s2 and > 5 changes in direction/sec (> 90°)] and freezing (velocity < 0.2 cm/s), and directional change frequency (no. clockwise + no. anti-clockwise, > 90O) and velocity (cm/s) were additionally used to better qualify behavioural outputs. From the recordings of the test phase, during which access to the full tank was allowed, the third of the tank next to each video was set as regions of interest (ROIs) representing interaction zones and the cumulative time spent within each ROI was measured.

RNA extraction and cDNA synthesis

For RNA extraction, a subsample of 9 fish per genotype were euthanised with an overdose of tricaine solution (MS222, Pharmaq; 500–1000 mg/L), the whole brain was collected directly into 500 µl lysis buffer (RNeasy Lipid Tissue Mini Kit-Qiagen) and stored at − 80◦C. Total RNA was extracted with the RNeasy Lipid Tissue Mini Kit (Qiagen) according to the manufacturer’s instructions, and the concentration, as well as the purity ratios (260 nm/280 nm and 260 nm/230 nm) estimated in the NanoDrop 2000 (Thermo Scientific). The same concentration of RNA for each sample was then reverse transcribed to cDNA (iScript cDNA Synthesis Kit, Biorad), following the manufacturer’s instructions. Briefly, a mix of nuclease-free water, 5 × iScript reaction mix (4 µl), iScript reverse transcriptase (1 µl), and RNA template were prepared in a 1.5 µl sterile tube in a final volume of 20 µl, and incubated in a PCR thermocycler in the following conditions: 5-min priming at 25 °C, 60-min reverse transcription at 42 °C, 5-min reverse transcription inactivation at 85 °C, and then and kept at 4 °C until tube collection. The samples were subsequently stored at -20 °C until further use.

Gene expression

Quantitative real-time PCR (qPCR) was performed for the target genes (bdnf, npas4, nlgn1, nlgn2, wnt3, neurod), and the eukaryotic translation elongation factor 1 alpha 1, like 1 (eef1a1l1) was used as a reference gene [43]. Primers for the gene Shank3a were designed on Primer 3 (primer3.ut.ee), tested for quality in the FastPCR 5.4., and the PCR product was sent to sequence to confirm the amplicon size and the alignment with the target sequence. The qRT-PCR’s were performed in the Applied Biosystems quantstudio 7 thermocycler (7900 HT, Thermofisher) in 8 μl reactions, with SYBR Green PCR Master Mix (Applied Biosystems, Life Technologies), the primers at a concentration of 50 μM and the cDNA samples diluted 1:10. Thermocycling conditions were 5 min at 95° C, followed by 40 cycles: 95 °C for 30 s, specific annealing temperature for each primer for 30 s (Additional file 5: Table S1), and 72 °C. A melting curve was also included, with a program from 55 to 95 °C with 0.5° C increase changes and the presence of a single reaction product in each well was confirmed. All reactions were performed in triplicate, and the technical replicates were run on the same plate. To calculate the relative expression, the ΔΔCt method ( 2–∆∆Ct) was used (Additional file 4: Data S2), where Ct is the qPCR-generated cycle threshold and ΔCt is calculated by:

$$\Delta Ct = Ct_}\,}} \right)}} - Ct_}\,}} \right)}}$$

and ΔΔCt by:

$$\Delta \Delta Ct = \Delta Ct_}\,}} \right)}} - \Delta Ct_}\,}} \right)}}$$

Analysis

Statistical analyses, calculations and graphical representations were carried out using the softwares Minitab® (v.17; Minitab Inc., State College, PA) and GraphPad© Prism (v.8.4.2; GraphPad Software LLC, San Diego, CA). The visualization of the genetic mutation was performed by the ApE© software (v2.0.5) and of the 3D protein structure by the open-source software ColabFold [45] that utilises MMseqs2 with AlphaFold2. Figures were edited and completed with illustrations using the software Adobe® Illustrator® (CS6, v.16.0.0; Adobe Systems Inc.) and Inkscape© (v. 1.2.2; Free Software Foundation Inc.).

From the behavioral test, the cumulative distance travelled by each animal (in cm) during the latter 5 min of the habituation period and during both the observation and test phase was compared between mutants and wildtypes using Welch’s 2-sample t-tests (due to unequal sample sizes). For the observation phase, we evaluated the social contagion of distress by measuring the percentage time exhibiting the behaviors observed in the distressed demonstrator, i.e. erratic and freezing (as per the defined kinematic thresholds), as well as average velocity for better screening of freezing bouts. These measures were compared between wild-type animals and shank3a mutants using a general linear model that included time bin, to test for temporal variations, and sex as added factors. Further, the number of directional turns exhibited by focal animals was compared to the maximum exhibited by demonstrators during their neutral swimming pattern to assess any motor deficits, using 1-sample Poisson rate tests, and compared between genotypes using 2-sample Poisson rate tests. Genotypic and temporal variations in average velocity were assessed via a general linear model for differentiating freezing bouts from irresponsiveness.

To characterise attention, we first tested whether mean absolute heading towards distress (0°–180°) significantly differed between genotype and sex using a general linear model. We then tested if it differed from divided orientation between distress and control playbacks (µ ≠ 90°) via 1-sample t-tests, both in wild-type animals and shank3a mutants. Finally, we quantified attentional differences by comparing mean absolute heading between wild-type animals and shank3a mutants using Welch’s 2-sample t-tests (due to unequal sample sizes).

To quantify recognition of the distress-state, we calculated individual local preference scores (PS) based on the time individuals spend in the ROI near the distressed-state stimulus (TS) compared to the control video (TC) using:

$$PS = \frac - T_ } \right)}} + T_ } \right)}}$$

where values range between -1 (full preference for control) and 1 (full preference for stimulus). To identify sex and genotype effects on PS, we used a general linear model. Then, discrimination ability in both wild types and mutants was ascertained by testing whether the mean PS for each group was significantly different form 0, using 1-sample t-tests. Comparisons between wild-type and mutant fish were performed using Welch’s 2-sample t-tests (due to unequal sample sizes).

For the characterization of shank3 mutation effects on neuroplasticity, the relative expression of target genes was compared between a subsample of mutants and wild types by unpaired t-tests, with Welch’s approach for cases with unequal samples sizes (due to quality-control sample exclusions), and Mann–Whitney U tests when measures did not conform to normality. We then measured inter-correlations in expression levels between genetic markers, and performed correlation-based cluster analysis (complete linkage based on absolute correlation coefficients, |r|) followed by correlation-based principal components analysis (PCA), with varimax rotation (best for smaller sample sizes [46]), to identify groups of associated genes and extract separate neuroplasticity component scores relating to differentiated gene groups. Finally, multiple linear regression analyses were used to examine the effect of different neuroplasticity components to attentional control and state recognition measures.