Brain-wide visual habituation networks in wild type and fmr1 zebrafish

Habituation is a form of learning during which animals stop responding to repetitive stimuli, and deficits in habituation are characteristic of several psychiatric disorders. Due to technical challenges, the brain-wide networks mediating habituation are poorly understood. Here we report brain-wide calcium imaging during larval zebrafish habituation to repeated visual looming stimuli. We show that different functional categories of loom-sensitive neurons are located in characteristic locations throughout the brain, and that both the functional properties of their networks and the resulting behavior can be modulated by stimulus saliency and timing. Using graph theory, we identify a visual circuit that habituates minimally, a moderately habituating midbrain population proposed to mediate the sensorimotor transformation, and downstream circuit elements responsible for higher order representations and the delivery of behavior. Zebrafish larvae carrying a mutation in the fmr1 gene have a systematic shift toward sustained premotor activity in this network, and show slower behavioral habituation.


Supplementary Figure 5. Anatomical distributions of five habituating clusters.
For each functional cluster shown in Figure 2, a dorsal view, lateral view, and four coronal virtual sections are shown. The ranges included in the coronal sections are indicated by the height of the corresponding boxes in the top left panel. Figure 6: Correlation of habituation dynamics between free-swimming and brain responses. a. Raster plots of ROIs that responded to the auditory stimuli (left) and to both loom and sounds (right). Color bar represents the SD. b. Top view of the anatomical locations of the ROIs in a. Left for the sound only responsive ROIs (in purple) and right for the loom and sound responsive ROIs (in cyan). c. Correlation analysis of habituation dynamics for the freeswimming behaviour and the habituating responses in eight brain regions. d. The average response profiles for strongly habituating (left) moderately habituating (center), and weakly habituating (right) tectal ROIs in each of the habituation paradigms. e. Fitted curves to the Moderately habituating tectal normalized responses during the first block of habituation. f. Recovery responses after first block of stimuli of the Moderately habituating tectal normalized responses. Black horizontal bar represents the median. g. Pearson correlations between the normalized responses of the tectum and their group-matched free-swimming probability of responses. Error bars show mean and 95% CI. Figure 7. Validation of graph theory approach: Number of nodes and single ROI graph analysis. a. Matrices generated with a scalable number of nodes (99 nodes, first row; 197 nodes, second row; and 368 nodes, third row) and all loom responsive ROIs of a sample fish (fourth row) for the pre-loom moment and the 1 st , 2 nd , 3 rd , 10 th , and 11 th looms. b. Density indices of pair-wise node connections from panel a (colored lines) and the all-ROI matrices of eleven individual fish from the f20 group (grey lines). c. Participation coefficients of the same graphs as in panel b (color labelling is the same). Figure 8. Validation of graph theory approach: Null models and leaveone-out cross validation. a. Raster plots of the nodes' activity for each of the generated models. Left: average activity of f20 fish nodes. Middle: adjusted-amplitude Fourier transform surrogate time series from the average model. Right: adjusted-amplitude Fourier transform surrogate time series from the average model where the AAFT was applied only to the loom presentation windows. b. Matrices generated with the models (first to third row) and the f20 dataset (fourth row) for the pre-loom moment, 1 st -5 th , 10 th , and 11 th looms. c. Density index of the connections of the matrices in panel b (colored lines) and the four stimulus train graphs (f20, f60, s20 and s60). d. Participation coefficient of the same graphs as in panel c. e. Sample sets each composed of five matrices generated with the leave-one-out cross-validation method for the pre-loom moment, 1 st -3 rd , 10 th , and 11 th looms (the fish excluded from each row is indicated at the left). f. Density index of the connections of the leave-one-out generated group-average matrices (grey lines, n=11) and the f20 whole group average matrix (black line) for the 1 st -3 rd , 10 th , and 11 th looms. Figure 9. Spatially sorted brain-wide graphs for WT and fmr1 -/larvae. Edges with strengths above 0.75 are shown between all pairs of nodes for trials 1, 2, 3, 10, and 11; nodes are arranged by brain region. The nodes' functional clusters are identified by color. Empty nodes (black) are added to spatially match the right side (ipsilateral to the visual stimulus) to the left side, despite the latter having fewer nodes. Abbreviations are the same as in Figure 6: Pallium, Pal; subpallium, Sp; thalamus, Th; habenula, Hb; pretectum, Pt; tectum, Tec; tegmentum, Tg; cerebellum, Cb; and hindbrain, HB.

Supplementary Figure 10. Dynamic community measures and optimization of γ and ω parameters.
a. Range of the average flexibility, cohesion, and promiscuity results based on combinations of the γ and ω parameters for the multilayer graphs of the WT, Hets, and fmr1 mutants. The median is given on the top of each subfigure. b. Steps for selecting the γ and ω values. Top left, difference in Q (maximized modularity quality, see Methods) of the WT graph and the temporal null model of the WT graph at various γ and ω values. Top middle, relative variance of the maximum Q of the WT graph. Higher values represent the combinations with lowest variance. Top right, optimization of Q by combining the Q-Qt difference and the relative variance of the WT graphs. Higher values indicate combinations of parameters that have a larger difference from the temporal null model and also show low variance. Bottom left, average of the optimized Q of the 3 genotype datasets; optimized Q is defined as the mean value of the maximized Q that have the highest difference from the null model and the lowest variance. Bottom middle, same as bottom left but applying the limiting rules for community detection results (See Methods). Bottom right, combination of values that are above the mean in the bottom middle panel. This set of parameters was used to estimate the community measures reported in Figure 6.
Supplementary Figure 11. Persistent homology analysis and motor related strongly habituating ROIs. a-e. Barcode graphs and lifetime sums of the persistent homology analysis performed in leave-one out group-averaged matrices (see Methods, WT=10 and fmr1 -/-=11). a. Dimension 1 barcode graphs for fmr1 mutants at the 1 st loom. b. Dimension 1 barcode graphs for fmr1 mutants and WT at the 5 th loom. c. Dimension 1 barcode graphs for fmr1 mutants and WT at the 11 th loom. d. Dimension 0 lifetime sums of group-averaged matrices of fmr1 mutants (n=11) and WT (n=10) at pre-loom and 20 loom time points. e. Dimension 2 lifetime sums of group-averaged matrices of fmr1 mutants (n=11) and WT (n=10) at pre-loom and 20 loom time points. Centre indicate means and error bars indicate 95% CIs in e and d. f-h. Prominence of motor-correlated strongly habituating ROIs in the hindbrain. f. The distribution of all strongly habituating neurons across the brain for the f20 stimulus train. g. The subset of the ROIs from panel (f) that show >1 s.d. correlation with motor responses during loom stimuli on a trial-by-trial basis. h. The proportion of strongly habituating ROIs that shows this motor correlation, by brain region.

Supplementary Table 1. Analysis of proportion of ROIs in each region
Two-way ANOVA test (one-sided) performed to compare the differences in cluster proportion of ROIs across brain regions in the f20, f60, s20 and s60 datasets.

Supplementary Table 2. Analysis of cluster identity of loom-responsive ROIs
Two-way ANOVA test (one-sided) performed to compare the differences in cluster identity within brain regions across the f20, f60, s20 and s60 datasets.