FIGURE SUMMARY
Title

Retinotectal circuitry of larval zebrafish is adapted to detection and pursuit of prey

Authors
Förster, D., Helmbrecht, T.O., Mearns, D.S., Jordan, L., Mokayes, N., Baier, H.
Source
Full text @ Elife

Experimental paradigm for studying location-specific processing in the tectum.

(A) In a typical hunting sequence, the fish detects prey in its peripheral visual field (1), ultimately turns and approaches to bring the prey image into its central binocular field (2). Hypothetically, the retinotectal map might be adapted to this location- and size-specific representation of the prey object. (B) Sketch of the tectum showing previously described cell types and neuropil layers. (C) Schematic for functional imaging setup. (D) On the left: Region of interest (ROI) for imaging tectal cell responses and exemplary expression of nuclear-localized GCaMP6s. On the right: ROI for RGC imaging and expression of GCaMP6s in RGC axons under control of ath5:Gal4. (E) Stimulus protocol. Arrows below stimulus representation indicate object movement, first in nasal, then in temporal direction. See Materials and methods for details. Scale bar in (D): 20 µm.

Functional clustering of tectal cells.

(A) Normalized calcium transients of all exemplars (gray) and average transients of all cells (colored) for the 14 tectal clusters. (B) Validation of cluster number by the silhouette coefficient. A minimal number of 14 clusters was chosen (dashed blue line). (C) 3D representation of the three main principal components (PCs) for all tectal cells. Data points are colored by their corresponding cluster. (D) The chosen number of three PCs explains 74.9% of the variance in the measured data (black curve). Shuffled data (blue curve) resulted in a lower average explained variance (57.6% for the three main PCs). (E) Functional cluster distribution of superficial interneurons (SINs) and neuropil interneurons (NINs), imaged in the elavl3:nls-GCaMP6s line (n = number of cells from three fish). Cluster numbers are indicated.

Functional clustering of tectal cells.

(A) Normalized calcium transients of all exemplars (gray) and average transients of all cells (colored) for the 14 tectal clusters. (B) Validation of cluster number by the silhouette coefficient. A minimal number of 14 clusters was chosen (dashed blue line). (C) 3D representation of the three main principal components (PCs) for all tectal cells. Data points are colored by their corresponding cluster. (D) The chosen number of three PCs explains 74.9% of the variance in the measured data (black curve). Shuffled data (blue curve) resulted in a lower average explained variance (57.6% for the three main PCs). (E) Functional cluster distribution of superficial interneurons (SINs) and neuropil interneurons (NINs), imaged in the elavl3:nls-GCaMP6s line (n = number of cells from three fish). Cluster numbers are indicated.

Functional clustering of RGC types.

(A) Normalized calcium transients of all exemplars (gray) and average transients of all pixels (colored) for the 10 RGC clusters. (B) Silhouette plot for cluster validation. A number of 10 RGC clusters was chosen (dashed blue line) as it significantly improves the modeling correlation compared to four clusters (see Figure 4—figure supplement 1A). (C) Comparison of RGC cluster distribution for three independent larvae. Data from the first larva (first bars) are shown in Figure 2 and were used for modeling. (D) Quantification of 30° dot- (light blue), 5° dot- (red), and looming+dark ramp/flash-responsive (dark blue) pixels in the segmented tectal layers, throughout the whole image stack shown in Figure 3C.

Functional clustering of RGC types.

(A) Normalized calcium transients of all exemplars (gray) and average transients of all pixels (colored) for the 10 RGC clusters. (B) Silhouette plot for cluster validation. A number of 10 RGC clusters was chosen (dashed blue line) as it significantly improves the modeling correlation compared to four clusters (see Figure 4—figure supplement 1A). (C) Comparison of RGC cluster distribution for three independent larvae. Data from the first larva (first bars) are shown in Figure 2 and were used for modeling. (D) Quantification of 30° dot- (light blue), 5° dot- (red), and looming+dark ramp/flash-responsive (dark blue) pixels in the segmented tectal layers, throughout the whole image stack shown in Figure 3C.

Figure panels showing the active RGC pixels of three imaging planes from <xref rid='fig3' ref-type='fig'>Figure 3C</xref> separately for the three relevant clusters (30° dot, 5° dot, looming+dark ramp/flash).

Skin is outlined by white-dashed line. Scale bar: 50 µm.

Linear modeling parameters.

(A) Modeling prediction quality shown as correlation values R2score and Rtrace, as a function of RGC cluster number. (B) Mean squared error (MSE) of modeling prediction as a function of the regularization parameter λ. See Materials and methods for details. (C) Correlation (R) between predicted and measured scores for modeling data (green) compared to a randomized model (gray). See Materials and methods for details.

Linear modeling parameters.

(A) Modeling prediction quality shown as correlation values R2score and Rtrace, as a function of RGC cluster number. (B) Mean squared error (MSE) of modeling prediction as a function of the regularization parameter λ. See Materials and methods for details. (C) Correlation (R) between predicted and measured scores for modeling data (green) compared to a randomized model (gray). See Materials and methods for details.

Comparison and quantification of tectal cell morphologies.

(A) Workflow for combining single-cell data. A common reference marker (elavl3:lyn-tagRFP) allowed co-registration of randomly labeled single cells (BGUG method), functionally selected single cells (FuGIMA method), and RGC expression patterns as landmarks for tectal laminae. (B) Dorsal overview of collected tectal interneurons from BGUG dataset (blue) and FuGIMA dataset (orange). (C) Tectal sideview. Dorsal is up and anterior is left. (D) Average proportional branch length of tectal cell neurites in the respective layers, for randomly labeled (blue) and functionally selected cells (orange). Error bars are SEM. ***: p < 0.001, and *: p < 0.05.

Comparison and quantification of tectal cell morphologies.

(A) Workflow for combining single-cell data. A common reference marker (elavl3:lyn-tagRFP) allowed co-registration of randomly labeled single cells (BGUG method), functionally selected single cells (FuGIMA method), and RGC expression patterns as landmarks for tectal laminae. (B) Dorsal overview of collected tectal interneurons from BGUG dataset (blue) and FuGIMA dataset (orange). (C) Tectal sideview. Dorsal is up and anterior is left. (D) Average proportional branch length of tectal cell neurites in the respective layers, for randomly labeled (blue) and functionally selected cells (orange). Error bars are SEM. ***: p < 0.001, and *: p < 0.05.

Tectal interneuron catalog.

Schematic representation of all identified morphotypes for tectal interneurons. The number of collected BGUG cells for each type is presented below. Red morphotypes have only been identified by the FuGIMA method. Identified functional types are indicated below. L: looming, D: dark ramp, B: bright ramp, 5°: small dot, 30°: big dot, BF: bright flash. PVIN: periventricular interneuron, NIN: neuropil interneuron, SIN: superficial interneuron, ms: monostratified, bs: bistratified, ts: tristratified, 4 s: tetrastratified, ps: partially stratified, ns: non-stratified/diffuse.

Quantification of tectal cell arbor size.

(A) Illustration of arbor quantification for a tristratified tectal cell. (B) Average arbor area in different tectal laminae. Error bars are SEM. (C) Ratio of deep vs. superficial arbor area within the same cells, for 5° dot-, Looming-, and Luminance-responsive cell types. Error bars are SEM. **: p < 0.01, and *: p < 0.05. (D) Illustration of columnar-shaped, 5° dot-responsive cells (red) and cone-shaped, looming-responsive cells (purple).

Functional compartmentalization of the tectum along the anterior-posterior axis.

(A) Distribution of tectal cell bodies from 30°-dot (blue), 5°-dot (red) and 5°-dot-forward (yellow) response clusters. Anterior (A), medial (M) and posterior (P) positions of the tectum are indicated. Graph shows probability density function for cell body distribution. Integrals are colored according to their functional cluster with p-values characterizing the difference from the distribution of all sampled cells (gray integral). (B) Tectal sideview of registered FuGIMA neurons showing the distribution of 30°-dot (blue), 5°-dot non-DS (red) and 5°-dot-DS (yellow) cells. (C) Quantification of proportional neurite branch length of tectal cells in the anterior tectum. N equals number of cells. **: p = 0.006,; *:p = 0.014.

Small size-tuned RGC inputs are essential for small-object processing in the tectum.

(A) Experimental setup for RGC axon ablations. Larvae are expressing mCherry in RGCs and nulear GCaMP6s panneuronally. The eye contralateral to the ablation site is visually stimulated and the ipsilateral tectal cells are functionally imaged before and after the ablations. As a control, the eye ipsilateral to the ablation site is stimulated and the contralateral tectal cells are imaged in the same fish. (B) Sideview of mCherry expression in RGCs at 6 dpf shows the most lateral axon bundle, which leaves AF7 for the SO layer (arrow). (C) Dorsal view of single image planes showing the axon fibers of interest in the contralateral (control, upper panel) and ipsilateral (ablated, lower panel) pretectum of the same fish. (D) Single functional image planes, projected over time, showing nuclear GCaMP6s expression in the ipsilateral tectum, before (6 dpf, left) and after (7 dpf, right) ablation. Pixels are color-coded by preference for 5° dot (magenta) or looming (cyan) stimuli. (E) Number of cells per image plane (out of two fish), which are responsive to a 5° dot stimulus, before and after ablations in the ipsilateral and the contralateral tectum. (F) Same as (E), showing the number of cells responsive to a looming stimulus. (G) Fraction of 5°-dot- and looming-responsive cells after ablations in the ipsilateral and contralateral tectum. Error bars are SEM. ***: p = 0.0006; n.s.: p = 0.46. N equals number of cells from two independent fish. Scale bars in (B): 30 µm, (C): 20 µm, and (D): 50 µm.

Tectal cell ablation controls and enucleation experiments.

(A–C) In our free-swimming prey capture assay, 'sham' ablated larvae (gray), in which non-responsive tectal cells have been removed, did not behave differently from untreated ('ctrl', green), or agarose-embedded and released ('embed', purple) larvae, in terms of time spent in prey capture (A), initial J-turn kinematics (B), or direction selectivity index (C). (D) Cluster color-coded functional cell responses in both anterior tecta for the experiment shown in Figure 8A (arrow heads highlighting cells selected for ablations). Midline is indicated by a dashed white line. Note the tectal cell responses in the left, non-stimulated tectum. (E) Enucleated larvae expressing elavl3:nls-GCaMP6s were functionally imaged as described, and the functional cluster distribution of tectal cells was compared to untreated fish. Error bars are SEM. Functional cluster colors are described on the right. Scale bar in (D): 20 µm.

Tectal cell ablation controls and enucleation experiments.

(A–C) In our free-swimming prey capture assay, 'sham' ablated larvae (gray), in which non-responsive tectal cells have been removed, did not behave differently from untreated ('ctrl', green), or agarose-embedded and released ('embed', purple) larvae, in terms of time spent in prey capture (A), initial J-turn kinematics (B), or direction selectivity index (C). (D) Cluster color-coded functional cell responses in both anterior tecta for the experiment shown in Figure 8A (arrow heads highlighting cells selected for ablations). Midline is indicated by a dashed white line. Note the tectal cell responses in the left, non-stimulated tectum. (E) Enucleated larvae expressing elavl3:nls-GCaMP6s were functionally imaged as described, and the functional cluster distribution of tectal cells was compared to untreated fish. Error bars are SEM. Functional cluster colors are described on the right. Scale bar in (D): 20 µm.

Acknowledgments
This image is the copyrighted work of the attributed author or publisher, and ZFIN has permission only to display this image to its users. Additional permissions should be obtained from the applicable author or publisher of the image. Full text @ Elife