Versatile functional roles of horizontal cells in the retinal circuit

Abstract

In the retinal circuit, environmental light signals are converted into electrical signals that can be decoded properly by the brain. At the first synapse of the visual system, information flow from photoreceptors to bipolar cells is modulated by horizontal cells (HCs), however, their functional contribution to retinal output and individual visual function is not fully understood. In the current study, we investigated functional roles for HCs in retinal ganglion cell (RGC) response properties and optokinetic responses by establishing a HC-depleted mouse line. We observed that HC depletion impairs the antagonistic center-surround receptive field formation of RGCs, supporting a previously reported HC function revealed by pharmacological approaches. In addition, we found that HC loss reduces both the ON and OFF response diversities of RGCs, impairs adjustment of the sensitivity to ambient light at the retinal output level, and alters spatial frequency tuning at an individual level. Taken together, our current study suggests multiple functional aspects of HCs crucial for visual processing.

Figure 1

Histological characterization of the dHC retina. (a) Schematic diagram of the BAC-Cx57-CreERT2 construct, in which the CreERT2-pA cassette is integrated into the translation start site of the mouse Connexin57 (Cx57) gene, and the NSE-DTA construct, in which the expression of Cre-mediated diphtheria toxin A (DTA) is induced under the control of neuron-specific enolase (NSE) promoter. (b) Schematic diagram of schedule for tamoxifen administration and harvest of retinas. (c) HCs were immunostained with an anti-Calbindin antibody (arrowheads) in control and dHC retinas. (d) Toluidine blue staining of control and dHC retinas. (e–h) Immunohistochemical analysis of synaptic connections between PRs and BCs in control and dHC retinas using marker antibodies as follows: PKCα (rod BC), Ctbp2 (synaptic ribbon), Trpm1 (ON BC), Pikachurin (PR terminal), mGluR6 (BC dendritic terminal), and PKARIIβ (type 3b OFF cone BC). Asterisks indicate cone terminals. (i) Ultrastructural analysis of PR ribbon synapses in control and dHC retinas by transmission electron microscopy. B, BC dendrite; H, HC process; R, synaptic ribbon. Asterisks indicate space due to lack of HC dendrites. (j) Quantification of PR ribbon synapses containing bipolar cell dendrites. The numbers of ribbon synapses quantified are 134 (rods, 130; cones, 4) for control and 105 (rods, 103; cones, 2) for dHC retinas (n = 3 retinas from three animals for each genotype). Nuclei were stained with DAPI (blue). ONL, outer nuclear layer; OPL, outer plexiform layer; INL, inner nuclear layer; IPL, inner plexiform layer; GCL, ganglion cell layer. Data are presented as mean ± SD. p = 0.12. n.s., not significant.

Figure 2

Changes in firing properties of RGCs by the depletion of horizontal cells. (a) Light-evoked responses recorded from control and dHC retinas. Examples of transient (upper) and sustained (lower) responses evoked by light stimulation (1,600 × 1,600 μm square, 2 s, yellow). Raw data for one trial (top), spike raster plots (middle), and the PSTH (30 ms bin, 10 trials, bottom). Inset, sorted spikes. (b) Left, Pearson’s r between trials. Right, CV of inter-spike interval. These measures were calculated from light onset (0 s) to 4 s. Gray, 186 RGCs in 6 control retinas; cyan, 136 RGCs in 4 dHC retinas. Bars, mean ± SD; n.s., not significant. (c) Left, histograms of response duration (upper, control; lower, dHC). Duration is defined as a half width at half maximum of PSTH. Histograms are fitted by Gaussian functions (control, red line, sum of two Gaussians, each of which correspond to transient [dotted black] and sustained [dotted gray]; dHC, blue line, single Gaussian). Right, cumulative probability of the histograms. gray, control. blue, dHC. p = 5.05 × 10⁻⁴. (d) Population analysis of light-evoked responses. Abscissa, time to the PSTH peak from the light onset or from the light offset. Ordinate, median firing rate for each RGC. Distribution of light-evoked RGC response property in the dHC retinas was different from that in control retinas (114 ON response and 108 OFF response in control retinas, 78 ON response and 79 OFF response in dHC retinas; ON latency, Kolmogorov-Smirnov test, p = 0.0044, OFF latency, Kolmogorov-Smirnov test, p = 6.59 × 10⁻⁶; ON firing rate, Kolmogorov-Smirnov test, p = 6.58 × 10⁻⁶, OFF firing rate, Kolmogorov-Smirnov test, p = 7.77 × 10⁻¹¹).

Figure 3

Horizontal cells mediate the surround inhibition of RGCs. (a,b) Principal component analysis for the aggregate of computed ON STAs (a) and OFF STAs (b). gray, 148 RGCs in 3 control retinas. blue, 134 RGCs in 3 dHC retinas. Histograms were calculated by 0.004-bin in each axis. insets, examples of STAs. Areas denoted by dotted ines showed aggregation of STAs (gray, control; blue, dHC). In ON STAs, RGCs within the first quadrant, which show the fast time course, were 93% of total 62 RGCs in the dHC retinas (blue, dHC). In OFF STAs, RGCs within the third quadrant, which show the fast time course, were 78% of the total 72 RGCs in the dHC retinas (blue, dHC). (c) Examples of the estimated RFs of the control (left) and dHC (right) retinas. (d) Relationship between the distance from the RF center and the mean pixel intensity (black, control; blue, dHC) of the sample RFs (c). Values were normalized by the mean intensity at the RF center. bars, SD. Inset values, surround indexes denoted minimum mean intensity. (e) Histograms of surround indexes calculated by 0.03-bin. inset numbers, median of each distribution (gray, control; blue, dHC). (f) Area of the RF estimated for each RGC in control and dHC retinas. bars, mean ± SD, unpaired t-test, p = 2.21 × 10⁻²², ***p < 0.001.

Figure 4

Responses to the drifting gratings. (a) Response examples of three types of frequency tuning. Upper, control; lower, dHC. Mean spike numbers were calculated from spikes evoked during drifting phase. (b) Left, matrix of observed spike numbers. We quantified the frequency preference of each cell as a frequency condition (red triangles) causing the max response (asterisk). Right, histograms of the frequency preference (black, control; blue, dHC). Upper, SF; lower, TF. Multiple comparisons were based on the adjusted standardized residual (**p < 0.01, *p < 0.05). 159 RGCs in 4 control retinas. 161 RGCs in 4 dHC retinas. p = 0.0067 (SF, 0.0075), p = 0.0334 (SF, 0.15), p = 0.042 (SF, 0.3), p = 0.075 (SF, 0.6), p = 0.195 (SF, 1.2), p = 0.337 (SF, 2.4), p = 0.009 (TF, 1), p = 0.207 (TF, 3), p = 0.015 (TF, 6).

Figure 5

Global light adaptation mediated by horizontal cells. (a) Stimulus schema. Intensity of the test flash (1,200 × 1,200 μm square, 2 s) was changed (3.85, 4.82, 8.64, 10.07, 13.27 cd/m²) on various background intensities (2,800 × 2,800 μm square, 0, 1.8, 4.6 cd/m²). (b) Responses to the test flash (3.85, 8.64, 13.27 cd/m², PSTH of 0.01 s-bin) on the background with low (black; 0 cd/m²) or medium intensity (red; 1.8 cd/m²). yellow, test flash onset. Left, control. Right, dHC. (c) Spike number for 0.4 s from the test flash onset on various background intensities. Values were normalized to the spike number during the maximum test flash. Bars, mean ± SD. (d) Time-to-peak latency of the PSTH (denoted in c) among various background intensities. Values were normalized to the latency during the maximum test flash. Bars, mean ± SD. (e) Left, histograms of the intensity causing 50% of the max spike numbers (R _50%) with the low (gray, +0) or medium (red, +1.8) background intensity. Upper, control; lower, dHC. Right, cumulative distribution of R _50%. 156 RGCs in 4 control retinas. 148 RGCs in 4 dHC retinas. Kolmorov-Smirnov test, p = 7.79 × 10⁻¹⁰ in control retina, p = 0.75 in dHC.

Figure 6

Optokinetic responses (OKRs) in dHC mice. (a) The visual stimulus presented on three LCD monitors set around the mouse (19-inch LCD; refresh rate, 75 Hz; size, 270° × 65.7°). (b) The visual stimuli were moving sinusoidal gratings with one of five spatial frequencies (SF) selected randomly from lookup table: 0.0313, 0.0625, 0.125, 0.25, or 0.5 cycles/deg in a given trial. The temporal frequency (TF) was selected from 0.1875, 0.375, 0.75, 1.5, 3, 6, 12, or 24 Hz. (c–f) Eye velocities of the initial OKR responses averaged over the mice are shown by diameter of the circle. Filled circles: statistically significant (t-test, p < 0.05) responses (c,d). Heat map plots of the best-fit Gaussian functions (e,f). (c,e: control mice (n = 5), d,f: dHC mice (n = 5)). (g–j) Mean amplitudes of the slow phase eye velocity of the late OKR response represented by the diameter of the circle. Filled symbols: statistically significant (t-test, p < 0.05) responses (g,h). Heat map plots of the best-fit Gaussian functions (i,j). (g,i: control mice (n = 8), h,j: dHC mice (n = 7)). (k,l) Comparison of properties of the spatial frequency in initial and late OKRs. The gain (eye velocity/ stimulus velocity) of the eye movement was calculated. The sum of gains at all temporal frequencies at each spatial frequency is shown. (k: initial OKR, control mice (n = 5), dHC mice (n = 5), p = 0.00322 (SF 0.03), p = 0.3673 (SF 0.06), p = 0.0497 (SF 0.125), p = 0.475 (SF 0.25), p = 0.943 (SF 0.5), l: late OKR, control mice (n = 8), dHC mice (n = 7), p = 0.185 (SF 0.03), p = 0.448 (SF 0.06), p = 0.0450 (SF 0.125), p = 0.0088 (SF 0.25), p = 0.657 (SF 0.5)). (m,n) The difference in the optimal spatial frequency (sfo) (m: initial OKR, n: late OKR). p = 0.091 (m), p = 0.0028 (n). Data are presented as mean ± SD.