IO neurons can be divided into distinct morpho-anatomical types. A, Dorsal and lateral views of the average expression of the hspGFFDMC28C line used in this study (green, N = 39 fish; line from Takeuchi et al., 2015) registered to a common reference larval zebrafish brain (gray), showing strong signal in the IO and in the CFs in the cerebellum. In this and subsequent panels: ro, rostral direction; l, left; r, right; c, caudal; d, dorsal; v, ventral; scale bars, 100 µm; vertical dashed lines indicate the midline of the brain. Teal rectangle outlines the area shown in B, C, D, and F. B, Example of a single labeled IO neuron (magenta). C, Axon reconstruction of that neuron. The inset shows its dendritic morphology, and the asterisk indicates its axon. D, Axon reconstruction of all labeled IO neurons (N = 53 neurons from 39 larvae) color coded by soma location (red, left IO, teal, right IO), showing that IO neurons project contralaterally. E, Examples of IO neurons that were divided into two morphological classes: unipolar neurons (green) that have a single dendritic tree arborized along the midline, and multipolar neurons (magenta) that have bi- or tri-polar dendritic trees. Asterisks indicate axons. For the complete dataset (N = 16 unipolar, 19 multipolar and 18 ambiguous neurons) see Figure 1-1. F, Axon reconstruction of all unipolar (green) and multipolar neurons (magenta), showing that the morphological type of a neuron is predictive of its projection pattern and its location within the IO.

The majority of IO neurons are sensitive to translational and rotational motion, direction-selective and spatially organized. A, Left, two-photon laser-scanning microscope (2P LSM) used for calcium imaging. Middle, same as Figure 1A; ro, rostral direction; l, left; r, right; c, caudal; scale bars, 100 µm. Teal rectangle outlines the imaged area. Right, maximum intensity z-projection of the anatomy stack from an example fish, showing a typical imaging field of view with individual IO neurons. B, Stimulation protocol. Fish were presented with translational and rotational motion, with each stimulus lasting 21.4 s (6 s stationary, 10 s moving, 5.4 s stationary). Translational gratings moved at 10 mm/s and rotational windmill at 22.5 °/s. For each imaging plane, we presented translational gratings in 8 different directions in a randomized order, followed by clockwise (CW) and counter-clockwise (CCW) rotational motion. C, Average responses of four forward-selective example neurons to translational and rotational motion. DS, direction-selective neurons responding exclusively to translation motion, DS + CW or DS + CCW, direction-selective neurons that also responded to CW or CCW rotation, DS + CW + CCW, direction-selective neurons that also responded to rotation in both directions. Shadows represent SEM across repetitions. Horizontal dotted lines represent each neuron baseline. In C and D, vertical dotted lines separate stationary and moving periods of the stimulus, vertical solid lines separate visual stimuli. D, Left, raster plot for all direction-selective neurons (N = 608 neurons from 12 fish), grouped by response types and sorted by PD (coded by the circular color wheel). Each row represents a neuron’s average response to the ten stimuli shown on top. Right, average response of all neurons in each group to CW and CCW stimuli. Shadows represent SEM across neurons. E, Left, distribution of PDs of all direction-selective neurons and of the 4 different groups of neurons independently. 0° represents forward, 90° rightward, 180° backward and −90° leftward directions. Right, probability distribution of the cosine of PD for each of the four groups. F, Spatial distribution of PDs within the IO (N = 967 active neurons, including 608 direction-selective ones, from 12 fish). Each dot represents a neuron color-coded for PD as represented in the color wheel, or gray for not direction-selective neurons. Small italic roman numbers indicate the location of neurons shown in C. In F and G: ro, rostral direction; l, left; r, right; c, caudal; scale bars, 25 µm. Green and magenta curves show the rostral–caudal distribution of forward- and backward-preferring neurons respectively, separated according to the sign of the PD cosine. Blue and orange curves show the left-right distribution of left- and right-preferring neurons respectively, separated into forward-preferring (top) and backward-preferring (bottom) groups. G, Spatial distribution of rotation sensitivity within the IO. Each dot represents a neuron color-coded for its response to CW (red) and CCW (blue) stimuli. White lines on the color bars denote 99^th and 95^th percentiles, used for presenting the spatial distributions of top 1%, top 5%, and remaining 95% active neurons.

The majority of IO neurons receive input from both eyes with a contralateral bias. A, Monocular visual stimuli. Fish were presented with translational and rotational motion, with each stimulus lasting 21.4 s (6 s stationary, 10 s moving, 5.4 s stationary). Translational gratings moved at 10 mm/s and rotational windmill at 22.5°/s. Each stimulus was presented three times per imaging plane: to the left eye, to the right eye, and binocularly. To avoid contralateral contamination during monocular stimulation, the two visual fields were separated by a vertical 0.5 mm black patch that was positioned below the fish body. In addition, we had a 55° cut-off in front of the fish (27.5° in each eye) to prevent stimulation of the eyes’ binocular zone. B, The trial structure was the same as in the binocular stimulation experiment (translational gratings in 8 different directions in a randomized order, followed by clockwise (CW) and counter-clockwise (CCW) rotational motion). To minimize possible contribution of light onset/offset to the responses, monocular and binocular stimulation were performed in groups: left, followed by right and finally binocular stimulation. Binocular stimulation block also included converging (conv) and diverging (div) rotational motion. C, Examples of neurons’ responses to monocular stimulation of the left (cyan) and right (yellow) visual fields and to binocular stimulation (magenta). Vertical dotted lines separate stationary and moving periods of the stimulus. Horizontal dotted lines represent each neuron baseline. Vertical solid lines separate different visual stimuli. For comparison purposes, all example neurons monocular CW and CCW responses are repeated in binocular convergence and divergence stimuli (dashed lines). Shadows represent SEM across repetitions. D, Distribution of neurons’ monocular index. Monocular bias is color coded with a red (contra) through gray (binocular) to blue (ipsi) gradient. Dashed line indicates unbiased binocular neurons. Note that the distribution is shifted to the right, indicating that IO neurons are in general more sensitive to contralateral stimulation. E, Spatial distribution of monocular bias within the IO, color coded as histogram in D; ro, rostral direction; l, left; r, right; c, caudal; scale bar, 25 µm. Small italic roman numbers indicate the location of neurons shown in C. N = 518 neurons (of which 511 were active) from six fish.

Functional organization of the IO maps onto its morpho-anatomical organization. A, Left, light-sheet microscope used for fast volumetric calcium imaging. Right, same as Figure 1A. Teal rectangle outlines the imaged area. B, Stimulation protocol. The trial structure was the same as in the binocular stimulation experiment [translational gratings in 8 different directions in a randomized order, followed by clockwise (CW) and counter-clockwise (CCW) rotational motion]. Additionally, we included converging (conv) and diverging (div) rotational motion. Each stimulus lasted 21 s (6 s stationary, 10 s moving, 5 s stationary). Translational gratings moved at 10 mm/s and rotational windmill at 22.5 °/s. For each fish, we presented this stimulus set five times. C, Max z-projection of the distribution of active voxels categorized as forward selective (green), backward selective (blue) or rotation selective (red), averaged across fish. D i, Max z-projections of the distribution of active voxels categorized as left selective (green) or right selective (magenta) in the entire imaging field of view. D ii, Max lateral projection of the distribution of active voxels selective for ipsiversive motion (green) and for contraversive motion (magenta) within the IO. E i, Max z-projection patterns of unipolar and multipolar neurons from Figure 1F, color-coded depending on their morphological type and left-right location within the IO. E ii, Lateral projection of the location of unipolar (green) and multipolar neurons (magenta) within the IO. F, Overlay of D and E. Note that the distribution of active voxels in the IO in the light-sheet imaging data includes signals not only from cell somata but also from surrounding neuropil, which accounts for the more lateral spread compared to the soma distribution. See Table 1 for quantification of the overlap. N = 28 fish; ro, rostral direction; l, left; r, right; c, caudal; v, ventral; d, dorsal; scale bars, 100 µm.