Gap Junctions Contribute to Differential Light Adaptation across Direction-Selective Retinal Ganglion Cells

Abstract

Direction-selective ganglion cells (DSGCs) deliver signals from the retina to multiple brain areas to indicate the presence and direction of motion. Delivering reliable signals in response to motion is critical across light levels. Here we determine how populations of DSGCs adapt to changes in light level, from moonlight to daylight. Using large-scale measurements of neural activity, we demonstrate that the population of DSGCs switches encoding strategies across light levels. Specifically, the direction tuning of superior (upward)-preferring ON-OFF DSGCs becomes broader at low light levels, whereas other DSGCs exhibit stable tuning. Using a conditional knockout of gap junctions, we show that this differential adaptation among superior-preferring ON-OFF DSGCs is caused by connexin36-mediated electrical coupling and differences in effective GABAergic inhibition. Furthermore, this adaptation strategy is beneficial for balancing motion detection and direction estimation at the lower signal-to-noise ratio encountered at night. These results provide insights into how light adaptation impacts motion encoding in the retina.

Keywords: Connexin-36; cell types; classification; detection sensitivity; direction discrimination; multielectrode arrays; neural coding; population codes.

Figure 1.. Functional classification of DSGCs from multielectrode array recordings.

A-B. Spike rasters of an example DSGC (A) and a non-DSGC (B) to square-wave gratings moving in 8 directions at a speed of 480 μm/s (spatial period, 960 μm; temporal period, 2 s), 50% Michelson contrast, and background light intensity was 10000 R*/rod/s. Each grating was represented pseudo-randomly 3 times. Polar plots summarize the average response in each direction normalized by the direction producing the highest firing rate. C. Vector sum magnitudes for drifting gratings presented at two different speeds (480 μm/s and 240 μm/s) from all RGCs identified in one MEA recording. Cells were clustered into DSGCs (magenta, n = 77 cells from 1 retina) and non-DSGCs (black, n = 328 cells from 1 retina) (see Methods). D. Histogram of vector magnitudes for RGCs in (C) in response to the 240 μm/s gratings. E-F. Speed tuning curves of ON DSGCs (E) and ON-OFF DSGCs (F); gray shows all cells of each type and blue and red curves highlight an example ON and ON-OFF DSGC, respectively. Normalized response is computed based on vector sum magnitudes (see Methods). G. The scatter plot shows the weights from principal component analysis of the speed tuning curves of all DSGCs in one recording: weights for first and third principal components are plotted. Cells were clustered into ON DSGCs (blue, n = 16 cells from 1 retina) and ON-OFF DSGCs (red, n = 61 cells from 1 retina) with a two-Gaussian mixture model. Polar plots indicate 3 ON DSGCs (blue) aligned with three cardinal axes (left, blue) and ON-OFF DSGCs aligned with 4 cardinal axes (right, red) Note: coordinates are in visual space such that 'superior' refers to upward motion in the environment (ventral on the retina). S: Superior; A: Anterior; I: Inferior; P: Posterior. H. ON (blue) and ON-OFF (red) DSGC locations for each preferred direction estimated from electrical images (see Methods). Grey hexagons indicate MEA border. Each circle has a diameter of 150 μm to approximate average dendritic field size of mouse DSGCs. I. Micrograph of whole-mount Hb9∷eGFP retina on MEA. Bright dots are eGFP expressing cells, which are ON-OFF superior DSGCs. Black dots are electrodes. Red circles indicate soma locations of ON-OFF superior DSGCs estimated from electrical images (Els, see Methods and Figure S1).

Figure 2.. Direction tuning of superior DSGC is dependent on ambient light level.

A-D. Response of example ON-OFF DSGCs to gratings moving in 8 directions at a speed of 480 μm/s (spatial period, 960 μm; temporal period, 2 s) and 50% Weber contrast across 5 light levels. Polar plot summarize direction tuning across light levels normalized to the preferred direction response at each light level. E-H. Normalized average spike count as a function of stimulus direction across cells for each type in one retina. (mean ± SEM; data from 1 retina; superior: n = 18 cells, anterior: n = 22 cells, inferior: n = 7 cells, posterior: n = 11 cells). Note: spontaneous (baseline) spiking was subtracted from each cell prior to normalizing and averaging their tuning curves. See also Figure S2 for direction tuning curves of ON DSGCs. I. Direction selectivity index (See Methods) as a function of background light intensity. Cells from two retinas were combined. (mean ± SEM; Retina 1: same as E-H. Retina 2: superior: n = 20 cells, anterior: n = 14 cells, inferior: n = 6 cells, posterior: n = 6 cells) J. Tuning width (See Methods) as a function of background light intensity. Cells from two retinas were combined (mean ± SEM; same as I). K. Average direction tuning curves for all 4 types of DSGCs at two different light levels (top: 10 R*/rod/s, bottom: 10,000 R*/rod/s, mean ± SEM; cells from one retina, same as E-H).

Figure 3.. Electrical coupling among superior ON-OFF DSGCs is reduced in FACx mice.

A. Schematic of the floxed Cx36 and FACx transgenic mouse lines. B. Circuit diagram showing the site of Cx36 knock-out in FACx mice. C. Neurobiotin labeling of superior ON-OFF DSGCs. Left: Hb9∷eGFP retina. Somata of superior ON-OFF DSGCs are eGFP⁺ (green). Injection of the gap junction-permeable tracer Neurobiotin into a single eGFP⁺ cell revealed dye coupling to several neighboring somata (red). Right: FACx retina. Somata of CreER expressing cells are tdTomato⁺ (green). Injection of Neurobiotin into a single tdTomato⁺ revealed some residual coupling with nearby cells (red). Injections were performed under mesopic conditions (50-500 Rh*/rod/s). See also Figure S3 for details. D-F. Example cross-correlograms computed for spike trains evoked by fine-scale checkerboard white noise (see Methods) at 10,000 R*/rod/s for 3 example pairs of superior DSGCs in C57/bl6 (D), superior DSGCs in FACx (E), and anterior DSGCs in C57/bl6 (F). See also Figure S4 for change of correlated activity across light levels. G-H. Electrical coupling networks for 2 C57/bl6 (G) and 2 FACx retinas (H). Hexagons indicate outline of MEA. Each circle indicates estimated locations of superior ON-OFF DSGC from their electrical images. Black line connecting circles indicate electrically coupling cell pairs (see Methods). Red circles and lines identify numbered example cell pairs in D and E.

Figure 4.. Direction tuning of superior DSGCs at low light levels is sharpened in FACx mice.

A. Average direction tuning curves at 1 R*/rod/s of superior DSGCs in C57 mice (solid black, n = 18 cells from 1 retina), FACx mice (dash black, n = 37 cells from 2 retinas), and the other three types of DSGCs in C57 (solid gray, n = 28 cells from 1 retina) and FACx (dash gray, n = 47 cells from 2 retinas). Note, spontaneous (baseline) spiking was subtracted from responses prior to normalizing and averaging the tuning curves. Data are represented as mean ± SEM. B. Average DSI of superior and other DSGCs in WT (2 retinas, superior: n = 38 cells, others: n = 63 cells) and FACx (2 retinas, superior: n = 37 cells, others: n = 47 cells) mice as a function of background light level. Data are represented as mean ± SEM. See also Figure S5 for details.

Figure 5.. Gabazine equates the null direction responses among all types of ON-OFF DSGCs at low light level.

A. Top: spike rasters of a superior DSGC in control (left) and 15 μM SR (right) in response to 10 repeats of bar (80% positive contrast) moving in the null direction at a speed of 960 μm/s, background light intensity of 4 R*/rod/s. Bottom: spike rasters of a posterior DSGC. Middle: overlaying PSTHs of the superior and posterior DSGCs in control and SR conditions. B-C. Average number of spikes in a 1 second time bin in response to a bar moving in the null direction, as a function of bar contrast, in C57 (B, 2 retinas, superior n = 20 cells, others n = 50 cells) and FACx (C, 2 retinas, superior n = 11 cells, others n = 23 cells). Data are represented as mean ± SEM. D-E. Summary of spike counts of superior and other DSGCs in control and SR conditions at low (5%) and high (80%) contrasts in C57 (D) and FACx (E) (Same population of cells as in B-C). Data are represented as mean ± SEM. See also Figure S5 for a LIF model simulation.

Figure 6.. Impact of direction tuning width on motion detection and discrimination.

A. Schematic of simulated motion detection (left) and direction discrimination (right) tasks (see Methods). B. Black circles indicate SNR of individual DSGC (all direction preferences) at 20% bar contrast at two different light levels (low light level: 1 R*/rod/s, high light level: 10,000 R*/rod/s). Red circle indicates the average SNR across cells. (mean ± SEM; 1 retina, n = 64 cells). Data from another retina have same result (not shown). C. Simulated discrimination error (red) and detection rate (blue) as a function of direction tuning width (α in equation (2), β fixed to 1) at rod (1 R*/rod/s, R_max = 1.5) and cone (10,000 R*/rod/s, R_max = 7.7) light levels. Red stars mark the lowest points on the discrimination error curves. Dashed lines mark the tuning widths measured from MEA recording using drifting gratings. D-E. Normalized performance changes (see Methods) in discrimination (red) and detection (blue) tasks as a function of the number of broad tuning curves in the DSGC population, at rod (1 R*/rod/s; D) and cone (10,000 R*/rod/s; E) light levels. See also Figure S6 for detection and discrimination as a function of stimulus direction.