Direction selectivity in the larval zebrafish tectum is mediated by asymmetric inhibition

Abstract

The extraction of the direction of motion is an important computation performed by many sensory systems and in particular, the mechanism by which direction-selective retinal ganglion cells (DS-RGCs) in the retina acquire their selective properties, has been studied extensively. However, whether DS-RGCs simply relay this information to downstream areas or whether additional and potentially de novo processing occurs in these recipient structures is a matter of great interest. Neurons in the larval zebrafish tectum, the largest retino-recipent area in this animal, show direction-selective (DS) responses to moving visual stimuli but how these properties are acquired is still unknown. In order to study this, we first used two-photon calcium imaging to classify the population responses of tectal cells to bars moving at different speeds and in different directions. Subsequently, we performed in vivo whole cell electrophysiology on these DS tectal neurons and we found that their inhibitory inputs were strongly biased toward the null direction of motion, whereas the excitatory inputs showed little selectivity. In addition, we found that excitatory currents evoked by a stimulus moving in the preferred direction occurred before the inhibitory currents whereas a stimulus moving in the null direction evoked currents in the reverse temporal order. The membrane potential modulations resulting from these currents were enhanced by the spike generation mechanism to generate amplified direction selectivity in the spike output. Thus, our results implicate a local inhibitory circuit in generating direction selectivity in tectal neurons.

Keywords: asymmetric inhibition; direction selectivity; tectum; vision; zebrafish.

Figure 1

Two-photon calcium imaging of tectal responses to moving bars. (A) Schematic of the experimental setup. Larval zebrafish were placed in a custom chamber (chamber not shown) and presented bars moving caudo-rostrally (CR) or rostro-caudally (RC) to the left eye. Calcium imaging and electrophysiological recordings were performed on the right tectum. (B) (Left) Average image of a two-photon stack with three neurons highlighted as ROIs. Scale bar represents 20 μm (r, rostral; c, caudal; m, medial; l, lateral; 85% of active cells in this example fish were sensitive to motion). (Right) Delta f/f (df/f) traces of the highlighted neurons. Red lines indicate stimulus moving in the CR direction and blue lines indicate RC direction at 60°/s. The cell in magenta was direction selective (DS) for the CR direction, the cell in cyan was non-DS and the cell in green was insensitive to motion. (C,D) In (C), Average df/f responses (n = 3 trials) of two cells to different speeds are shown. Red indicates motion in the CR direction and blue motion in the RC direction. Darker colors imply slower speeds and lighter colors faster speeds. The responses to slowly moving bars are delayed with respect to the onset of the stimulus as the time taken for the bar to enter the receptive fields of the neurons is longer. This is seen in the second cell's (bottom) response to slower stimuli. In (D), the single trial peak df/f responses (filled circles) of the same two cells to bars moving in the CR (red) or RC (blue) directions at different speeds are shown. The responses were fit with a smoothing spline (light blue or red lines) and the preferred speed was estimated. The cell on top shows band-pass tuning, while the cell on the bottom shows low pass tuning.

Figure 2

Population responses to bars moving at different speeds and directions. (A) The percentage of active cells that are motion sensitive, that are direction selective at a speed of 60°/s, and that show responses modulated by speed (mean ± SD). (B) Direction selectivity indices (DSIs) of individual tectal cells at different speeds are shown as raster ticks. Cells to the right of the red dashed line are CR selective and cells to the left of the blue line are RC selective. The green lines show the dependence of DSI on the speed of the stimulus for three cells. The cells shown have low (SD = 0.04), medium (SD = 0.12), and high (SD = 0.44) variability of DSIs at different stimulus speeds. (C) The histogram of preferred speeds for all speed tuned cells in the CR direction (red, n = 61 cells) and RC direction (blue, n = 73 cells).

Figure 3

Inhibitory currents are biased toward the null direction of motion. (A) Two-photon fluorescence image of a tectal neuron filled with dye (Alexa 594) from the recording electrode. Tectal neurons project dendritic arbors into a neuropil where they receive retinal ganglion cell axonal input from the contralateral eye. Scale bar represents 10 μm. (B) Voltage clamp (Exci and Inhi) and current clamp (Spike) recordings of a tectal cell's response (for the duration the stimulus was presented) to bars moving in the CR direction (red) and the RC direction (blue). For the currents, the mean trace (thicker red/blue line) is shown superimposed over five trials (lighter red/blue lines). This cell is RC selective and has strong inhibition from the null CR direction. (C and D) Spike-DSI for all cells (n = 17) plotted vs Exci-DSI and Inhi-DSI, respectively.

Figure 4

Inhibition precedes excitation in the null direction and follows it in the preferred direction. (A) Average excitatory (black) and inhibitory (gray) current profiles from a DS cell to bars in the preferred and the null directions are shown after normalization to illustrate the temporal relationship between them. (B) The latency between excitatory and inhibitory currents for preferred and null directions for all the DS cells (n = 6 preferred, n = 11 null, see text). Negative values mean inhibition precedes excitation.

Figure 5

Sharpening of membrane potential DS by the spike generation mechanism. (A) Membrane potential recordings in current clamp mode of the same cell as Figure 3B to bars moving in CR (red) and RC (blue) directions. The spikes were filtered out by using a median filter. The mean trace (dark) is superimposed over traces from n = 5 individual trials (light). (B) Spike-DSI plotted vs Memb-DSI for all the recorded cells (n = 17).

Figure 6

The proposed model for the direction-selective circuit in the larval zebrafish tectum. The DS tectal neuron receives excitatory inputs from RGCs (open circles). This neuron also receives an inhibitory input (filled circle) from an interneuron that is topographically shifted toward the null direction of the DS tectal neuron. The interneuron is itself direction selective but for the opposite direction to that of the DS tectal neuron being considered. Thus, if a bar were to move in the preferred direction of the DS tectal neuron (orange arrow), the excitatory currents from the RGC terminals would arrive before the input from the Inhibitory DS interneuron. Also, the inhibition from the interneuron would be low. In the null direction for the DS tectal neuron, higher inhibition from the Inhibitory DS interneuron would arrive before the excitatory inputs from the RGC terminals.