Experimental framework for distinguishing the effect of reduced cone input, circuit changes, and functional compensation following partial cone loss.

(A) Illustration of known cone pathways to AOFF-S and AOFF-T ganglion cells. AOFF-S receives approximately ≤4% inputs from type 1b and 50% inputs from type 2 (T2) cone bipolar cells (CBCs) (Della Santina et al. 2016 and Yu et al. 2018). AOFF-T receives 40% inputs from T3a and 18% of inputs from T4 CBCs (Yu et al. 2018). Each CBC type receives different degrees of inhibitory inputs from AII and other amacrine cell types (top pie charts). The schematic depicts T2 and T3a CBCs as primary inputs to AOFF-S and AOFF-T ganglion cells, respectively. Each ganglion cell type receives glycinergic and GABAergic inhibitory inputs and equivalent contributions from AII amacrine cells (bottom pie charts). If cone ablation affects circuits before divergence of the OFF pathways, then the following sites are implicated (white): cones, horizontal cells, common amacrine cells. If cone ablation affects circuits after divergence of the OFF channels, then the following locations are implicated (blue): bipolar cells (T2 or T3a), different amacrine cells that send their inputs to sustained or transient pathways. (B-E) Comparison of three conditions and their state of cones and the retinal circuit: (1) control condition has a full complement of cones and a control circuit (control); (2) DTR condition has half the cones and a DTR circuit (cone-DTR); (3) half stimulation condition has stimulation of half the cones and a control circuit (partial). Triangles represent comparisons among three conditions that would be interpreted as (B) no change, mechanisms explained by (C) half cone stimulation, (D) circuit changes, or (E) compensation. See also Figure S2.

Partial cone loss causes differential temporal changes to AOFF ganglion cell types.

(A) Example spatio-temporal filters in response to a bar noise stimulus under three conditions for either AOFF-S (rows 1, 3) or AOFF-T (rows 2, 4) under voltage clamp for measuring excitation (rows 1-2) or inhibition (rows 3-4). (B) Average temporal filters of (odd rows) AOFF-S and (even rows) AOFF-T ganglion cells measuring excitation (rows 1-2) and inhibition (rows 3-4) for control (black), cone-DTR (magenta) and partial stimulation (green). (C-D) Box plots of first principal components (PC1) for each pairwise comparison of conditions for AOFF-S and AOFF-T ganglion cell excitation. Interpretation of mechanisms in the triangle center. (E) Difference of deltas results for the null hypothesis for equivalent changes in the temporal filters between AOFF-S and AOFF-T ganglion cells in each comparison of conditions (bootstrapped distribution of 10,000 iterations with noted standard deviations) and the actual difference of deltas (circles with error bars). Significant difference from the null hypothesis displayed on the right. (F) Features extracted from temporal filters: time to peak (TTP), time to trough (TTT), peak amplitude (PA), trough amplitude (TA), and time to zero crossing (TTZ). (G-H) Average temporal filters under control and cone-DTR conditions for (G) AOFF-S and (H) AOFF-T ganglion cells, which show a significant difference between control vs. cone-DTR, and a version of the control temporal filter shifted by multiple features to match the cone-DTR temporal filter. (I) Normalized distance between the temporal filter in control (ordinate = 0) and cone-DTR (ordinate = 1) when the control temporal filter is changed by each feature for AOFF-S (tan) and AOFF-T (blue) ganglion cells. (J) Average excitatory current normalized to the flash strength of AOFF-S (left) and AOFF-T (right) to a 30ms decrement in light. P-values in box plots indicate rank sum comparison between each pair of conditions after correcting for multiple comparisons with the Holm method (C-D). P-values in difference of deltas plot (E) indicate significant differences between AOFF-S and AOFF-T ganglion cells from a permutation test with correction for multiple comparisons. The following asterisks indicate p values: * ≤ 0.05, ** ≤ 0.01, *** ≤ 0.005. See also Figure S2 and Dataset S1.

Partial cone loss causes presynaptic glycine and direct GABA inhibition to mediate changes in the cone-DTR temporal filters of AOFF-T ganglion cells.

(A-B, D-E) Schematic of retinal pathways leading to the AOFF-S (left) and AOFF-T (right) ganglion cells and the circuit motifs for direct excitation (white), presynaptic inhibition from (A) GABA (yellow) or (B) glycine (green), and direct inhibition from (D) GABA (yellow) or (E) glycine (green) that would be blocked by (A, D) picrotoxin, an antagonist of GABAA and GABAC receptors, or (B, E) strychnine, an antagonist of glycine receptors. (C) Illustration of the difference filter computed by subtracting the temporal filters from after and before pharmacological application for each cell. (F, H, K, M) Average difference between Ames and (F, K) picrotoxin or (H, M) strychnine. (G, I, L, N) Box plots of first principal components (PC1) across cell types and conditions for the differences between (G, L) Ames and picrotoxin or (I, N) Ames and strychnine from (G, I) excitation temporal filters and (L, N) inhibition temporal filters. (J, O) Results of the comparison of PC1 and the interpretation of mechanisms for (J) excitatory temporal filters and (O) inhibitory temporal filters. P-values in box plots indicate rank sum comparison between each pair of conditions; p-values are corrected for multiple comparisons with the Holm method (G, I, L, N). See also Figure S3 and Dataset S1.

Partial cone loss causes differential changes in spatial filters of AOFF ganglion cell types.

(A-B, G-H) Box plots of first principal components (PC1) for (A, G) AOFF-S and (B, H) AOFF-T ganglion cells between conditions from (A-B) excitation and (G-H) voltages, which showed significant differences across conditions. Interpretation of mechanisms in the triangle centers. (C, I) Difference of deltas results for the null hypothesis for equivalent changes in the spatial filters between AOFF-S and AOFF-T ganglion cells in each comparison of conditions (bootstrapped distribution of 10,000 iterations with noted standard deviations) and the actual difference of deltas (circles with error bars) for (C) excitation and (I) voltages. Significant difference from the null hypothesis displayed on the right. (D, J) Features extracted from spatial filters: center height, surround height, center width, and surround width for (D) excitation and (J) voltages. (E, K-L) Average spatial filters under control and cone-DTR conditions for (E) AOFF-T ganglion cell excitation and (K) AOFF-S and (L) AOFF-T ganglion cell voltages, which show a significant difference between control vs. cone-DTR, and a version of the control spatial filter shifted by multiple parameters to match the cone-DTR spatial filter. (F, M) Normalized distance between the spatial filter in control (ordinate = 0) and cone-DTR (ordinate = 1) when the control spatial filter is changed by each parameter for (F) AOFF-T ganglion cell excitation and (M) AOFF-S and AOFF-T ganglion cell voltages. P-values in box plots indicate rank sum comparison between each pair of conditions; p-values are corrected for multiple comparisons with the Holm method (A-B, G-H). P-values in the difference of deltas plots indicate significant differences between AOFF-S and AOFF-T ganglion cells from a permutation test with correction for multiple comparisons (C, I). See also Figure S4 and Dataset S1.

Partial cone loss causes excitation and direct GABAergic and glycinergic inhibition to mediate changes in the cone-DTR spatial filters of AOFF-T ganglion cells.

(A-B, D-E) Schematic of retinal pathways leading to the (left) AOFF-S and (right) AOFF-T ganglion cells and the circuit motifs for direct excitation (white), presynaptic inhibition from (A) GABA (yellow) or (B) glycine (green), and direct inhibition from (D) GABA (yellow) or (E) glycine (green) that would be blocked by (A, D) picrotoxin, an antagonist of GABAA and GABAC receptors, or (B, E) strychnine, an antagonist of glycine receptors. (C) Illustration of the difference filter computed from subtracting the spatial filters from after and before pharmacological application for each cell. (F, H, K, M) Average difference between Ames and (F, K) picrotoxin or (H, M) strychnine. Variance captured by each principal component noted on the corresponding axis. (G, I, L, N) Box plots of first principal components (PC1) across cell types and conditions for the differences between (G, L) picrotoxin and Ames or (I, N) strychnine and Ames from (G, I) excitation spatial filters and (L, N) inhibition spatial filters. (J, O) Results of the comparison of PC1 and the interpretation of mechanisms for (J) excitatory spatial filters and (O) inhibitory spatial filters. P-values in box plots indicate rank sum comparison between each pair of conditions; p-values are corrected for multiple comparisons with the Holm method (G, I, L, N). See also Figure S3 and Dataset S1.

Partial cone loss affects nonlinearities at multiple levels leading to partial recovery of AOFF ganglion cell outputs.

(A) Three exemplar nonlinearities from control (black), cone-DTR retina (magenta), and partial stimulation (green) for excitation, inhibition, subthreshold voltages, and spikes from individual (top) AOFF-S and (bottom) AOFF-T ganglion cells. (B-C, E-F, H-I, K-L). Box plots of first principal components (PC1) for AOFF-S and AOFF-T ganglion cells between conditions from (B-C) excitation, (E-F) inhibition, (H-I) voltages, (K-L) spikes. Interpretation of mechanisms in the triangle centers. (D, G, J, M) Difference of deltas results for the null hypothesis for equivalent changes in the nonlinearity between AOFF-S and AOFF-T ganglion cells in each comparison of conditions (bootstrapped distribution of 10,000 iterations with noted standard deviations) and the actual difference of deltas (circles with error bars) for (D) excitation, (G) inhibition, (J) voltages, and (M) spikes. Significant difference from the null hypothesis displayed on the right. P-values in box plots indicate rank sum comparison between each pair of conditions; p-values are corrected for multiple comparisons with the Holm method (B-C, E-F, H-I, K-L). P-values in the difference of deltas plots from a permutation test with correction for multiple comparisons (D, G, J, M). See also Figures 7, S5 and Dataset S1.

Partial cone loss has a differential impact on nonlinearities of input currents vs. output voltages and spikes.

(A, C, E, G) Principal component analysis was performed on the nonlinearities for (A) excitation and subthreshold voltages, (D) excitation and spikes, (E) inhibition and subthreshold voltages, (G) inhibition and spikes. (B, D, F, H) Using the projection onto the first principal component of nonlinearities across conditions, we analyzed the difference of deltas to determine differences between inputs vs. outputs within each cell type. P-values in the difference of deltas plots from a permutation test with correction for multiple comparisons (B, D, F, H). See also Dataset S1.

Partial cone loss causes excitation, direct GABA and glycine inhibition to mediate changes in the cone-DTR nonlinearities of AOFF ganglion cells.

(A-B, D-E) Schematic of retinal pathways leading to the (left) AOFF-S and (right) AOFF-T ganglion cells and the circuit motifs for direct excitation (white), presynaptic inhibition from (A) GABA (yellow) or (B) glycine (green), and direct inhibition from (D) GABA (yellow) or (E) glycine (green) that would be blocked by (A, D) picrotoxin, an antagonist of GABAA and GABAC receptors, or (B, E) strychnine, an antagonist of glycine receptors. (C) Illustration of the difference nonlinearity computed from subtracting the nonlinearities from after and before pharmacological application for each cell. (F, H, K, M) Average difference between Ames and (F, K) picrotoxin or (H, M) strychnine. (G, I, L, N) Box plots of first principal components (PC1) across cell types and conditions for the differences between (G, L) picrotoxin and Ames or (I, N) strychnine and Ames from (G, I) excitation nonlinearities and (L, N) inhibition nonlinearities. (J, O) Results of the comparison of PC1 and the interpretation of mechanisms for (J) excitatory and (O) inhibitory nonlinearities. P-values in box plots indicate rank sum comparison between each pair of conditions; p-values are corrected for multiple comparisons with the Holm method (G, I, L, N). See also Figure S3 and Dataset S1.

Excitatory and inhibitory synapses change in OFF pathways following partial cone loss.

(A) En face confocal images of T2 (rows 1-2) and T3a (rows 3-4) OFF CBC axons as labeled by synaptotagmin 2 (Syt2) and HCN4 (magenta) with CtBP2 (yellow) for excitatory output synapses, GlyRɑ1 for glycine receptor alpha-1, and GABAAβ2,3 (cyan) for GABAA receptor subunit beta 2-3 in control (rows 1, 3) and cone-DTR (rows 2, 4) retina. (B-C) The volume of cone bipolar cell axon terminals in cone-DTR retina remained constant. (D-G) While the puncta density of ribbon synapses was maintained in the axons of both cone bipolar cell populations (D-E), we found an increased volume of CtBP2 in both bipolar cell types (F-G). In contrast to the enlarged volume of excitatory synapses, that of inhibitory synapses remained unchanged. Interestingly, we found differential changes in the density of GlyRɑ1 and GABAARβ2,3 in T3a cone bipolar cell (D-E). These results demonstrate common adjustments of excitatory synaptic ribbons in response to partial cone loss whereas inhibitory synapses potentially undergo differential changes between T2 and T3a cone bipolar cells. (H) En face confocal images of control AOFF-S (rows 1-2) and AOFF-T (rows 3-4) ganglion cells labeled with PSD95, GlyRɑ1, gephyrin, and GABAAβ2,3 and in cone-DTR retina. Insets show a stretch of dendrites with puncta for rectangles in main images. (I-J) Average puncta volume of PSD95, gephyrin, GlyRɑ1, and GABAAβ2,3 within the dendrites of (I) AOFF-S and (J) AOFF-T ganglion cells. While the volume and density of PSD95 remained constant in AOFF-S ganglion cells, AOFF-T ganglion cells showed increased PSD95 puncta size (I-J). We also found increased volume of GlyRɑ1 and GABAARβ2,3 clusters in both ganglion cell types (I-J). (K-L) Average linear density of PSD95, gephyrin, GlyRɑ1, and GABAAβ2,3 within the dendrites of (K) AOFF-S and (L) AOFF-T ganglion cells. Despite a common increase in the volume of inhibitory receptors, AOFF ganglion cells exhibit differential changes to synaptic density (K-L). Density and volume of gephyrin were maintained in both cell types in cone-DTR retina. Box plots show median with IQR and whiskers from 10% to 90% of the data. P-values indicate rank sum comparison between cell types. See also Dataset S1.