Divergence of visual channels in the inner retina

Abstract

Bipolar cells form parallel channels that carry visual signals from the outer to the inner retina. Each type of bipolar cell is thought to carry a distinct visual message to select types of amacrine cells and ganglion cells. However, the number of ganglion cell types exceeds that of the bipolar cells providing their input, suggesting that bipolar cell signals diversify on transmission to ganglion cells. We explored in the salamander retina how signals from individual bipolar cells feed into multiple ganglion cells and found that each bipolar cell was able to evoke distinct responses among ganglion cells, differing in kinetics, adaptation and rectification properties. This signal divergence resulted primarily from interactions with amacrine cells that allowed each bipolar cell to send distinct signals to its target ganglion cells. Our findings indicate that individual bipolar cell-ganglion cell connections have distinct transfer functions. This expands the number of visual channels in the inner retina and enhances the computational power and feature selectivity of early visual processing.

Figure 1. Many ganglion cells respond to input from a single bipolar cell

(a) Schematic diagram of the experiment. A single bipolar cell (BC) is impaled with a sharp electrode and intracellularly stimulated by current injection (see d for example), while a population of ganglion cells (GCs) is simultaneously recorded with a multielectrode array. A, amacrine cell (AC); B, BC; G, GC; H, horizontal cell; P, photoreceptor; filled circle, excitatory synapse; open circle, inhibitory synapse. (b) Horizontal view of a BC, focusing on the axon arbors in the inner plexiform layer (top), and the vertical view across the soma (bottom). The arrows indicate locations of the image slices shown in the other panels, respectively. (c) The receptive field centers of an intracellularly recorded BC (green) and of 39 GCs on the electrode array (gray, unconnected; black, cyan, purple, and blue, connected; see d for connectivity analysis). Each outline represents a two-dimensional Gaussian fit to the receptive field profile (contour at one standard deviation; see Supplementary Fig. 1 for details). (d) Raster graph of GC spikes in response to inputs from a single BC (from c). Each row represents the spiking activity of a single GC, arranged in order of increasing distance from the BC (top to bottom). Either depolarizing (pink-shaded periods) or hyperpolarizing (blue-shaded periods) current pulses were delivered to the BC intracellularly (top trace; only the first three trials are shown here; see Supplementary Fig. 1 for more). The three representative GCs from c are shown in the respective colors.

Figure 2. Individual pairs of bipolar and ganglion cells have distinct transmission properties

(a) Responses of two GCs (top, raster graphs; bottom, peri-stimulus time histogram; yellow bins, significant deviation from spontaneous firing rate) to current stimulation of a single BC. Here and in subsequent figures the current stimulus is color coded (pink/blue) as in Fig 1d. Note sustained firing in one GC (left) but very transient firing in another (middle; magnified in the right panel). (b) Responses of a single GC to current stimulation of two different BCs (left and right; displayed as in a). (c) Population data for synaptic connections divergent from the same BC. Each dot represents a BC-GC connection (one column for each BC). Top: Peak latency evoked by BC depolarization (0.28±0.15 s, mean ± standard deviation from 633 GC responses total). Significant variation was found in GC responses to 38 out of 53 BCs (red). The variation among the connections of individual BCs explains 62% of the total variation, whereas the variation across different BCs explains only 38% (see Methods). Bottom: For the peak firing rate (6.6±11.0 spikes s⁻¹), significantly different GC responses were found in 43 BCs. The variation within connections from the same BC explains 67% of the total variation. (d) Population data for synaptic connections convergent on the same GC (displayed as in c). Inputs from different BCs can drive the same GC differently (5 out of 15 GCs for peak latency; 6 GCs for peak rate), and distinct dynamics can arise even with the same evoked firing rate (as in b, indicated by blue and green circles).

Figure 3. Dynamics of bipolar cell signals are diversified by amacrine circuits

(a, b) Spiking response of two GCs to current stimulation of a BC, with (top) and without (bottom) inputs from ACs. After blocking AC signals by 100 μM picrotoxin (PTX) and 1.0 μM strychnine (STR), the transient burst of spikes in one GC became considerably stronger and peaked later (a), whereas the sustained response in the other GC became stronger but peaked earlier (b). (c, d) Summary of the effects of blocking inhibitory synaptic transmission on the peak firing rate (c) and the peak latency (d) evoked by single BC depolarization. Scatter plots compare the GC responses with (abscissa) and without (ordinate) inhibitory transmission (66 GCs in total from 6 BCs indicated by different colors; blue and green circles indicate those in a and b, respectively). With the inhibitory circuits active, the peak firing rate was lower (c; p<0.001, sign-test; control, 12.4±13.7 spikes s⁻¹; drug, 32.7±29.1 spikes s⁻¹; mean ± standard deviation) but there was a greater range in the peak latency (d; p<0.001, Levene’s test; control, 0.27±0.13 s; drug, 0.26±0.09 s). In all six experiments, blocking AC signals made sustained responses more transient and transient responses more sustained. Insets in d correspond to curve fits for the examples in a (dark and light blue) and b (dark and light green).

Figure 4. Interactions with amacrine cells can control the kinetics of connections between bipolar and ganglion cells

We simulated how a step change in the input current to a BC is transduced into an evoked GC firing rate in the presence of four distinct types of AC inputs (see top circuit diagram and Methods): tonic presynaptic inhibition of the BC terminal (b), tonic postsynaptic inhibition of the GC (d), feedback presynaptic inhibition (a), and feedforward postsynaptic inhibition (c). For all four types of inhibitory interactions, the evoked firing rate decreases as the inhibitory effects become stronger (trace color from white to black; see Supplementary Fig. 3 for the effects of BC baseline potential). However, the effects on the response kinetics vary (compare to experiments in Fig. 3). Tonic presynaptic inhibition prevents synaptic depletion and thus extends the GC response in time (b). Tonic postsynaptic inhibition affects the spiking threshold but not the release dynamics of BC terminals, and thus the peak latency remains unchanged (d). Both feedback presynaptic inhibition (a) and feedforward postsynaptic inhibition (c) shorten the GC responses by truncating the later component of the excitation. They differ, however, in that presynaptic inhibition slows vesicle release and thus prevents rapid synaptic depression, producing weaker but prolonged postsynaptic responses over an intermediate regime.

Figure 5. Adaptation of bipolar cell signals depends on interaction with amacrine cells

(a–d) Responses of four simultaneously recorded GCs to depolarization of a single BC, and their evolution over trials. a, left: Raster graph showing spikes during the first 300 ms of depolarizing current, delivered in many successive 6-s-long trials (Fig. 1d); gray line is a linear fit to the peak latency over trials (gray, nonsignificant change; dark gray, significant increase or desensitization; light gray, significant decrease or sensitization). a, right: Variation of the peak firing rate over trials with a linear fit. b–d: Responses of three additional GCs with different characteristics (displayed as in a). (e) Population data for the slow changes in the peak latency (left) and peak rate (right). Each data point represents the adapting behavior of one BC-GC connection, estimated by the slopes of the linear fits as in a–d (cross, significant change; dot, nonsignificant change; colored circles indicate those from a–d and f). Each column shows the connections of one BC (sorted in order of increasing mean latency changes). The stacked histograms are obtained from 129 BC-GC connections in total (gray, nonsignificant change; dark gray, significant desensitization; light gray, significant sensitization). For latency adaptation, 20 out of 24 BCs showed significant variation among their BC-GC connections, and this variation originating from individual BCs explains 85% of the total variation. For peak rate adaptation, 14 BCs showed significant variation and that accounts for 59% of the total. (f) Spiking responses of a GC to BC depolarization before (left) and after (right) pharmacological block of AC signals by 100 μM picrotoxin (PTX) and 1.0 μM strychnine (STR). Displayed as in a–d. (g) Population data (57 GCs total) for the adapting changes in the peak latency and peak rate over trials in absence of AC transmission (displayed as in e). After the block of AC signals, GCs showed desensitization more frequently for both the latency (p<0.002; χ²-test) and the peak rate (p<0.002; χ²-test).

Figure 6. Adaptation is specific to individual pairs of bipolar and ganglion cells

(a) Responses of a single GC to depolarization of two different BCs (displayed as in Fig. 5a). The response to one BC showed strong desensitization over time (left), whereas that to another BC did not, despite a higher peak firing rate (right). (b) Population data for the slow changes in peak firing rate and latency evoked by depolarization of two different BCs (displayed as in Fig. 5e; colored circles indicate those in a). Inputs from different BCs can lead to different adapting behaviors in the same GCs (4 out of 7 GCs for the peak latency change; 3 GCs for the peak rate change). (c) Responses of a GC to two different inputs: current injected into a BC (top, black trace) and visual stimulation in an annulus that does not drive the injected BC (cyan; contrast-reversal grating). The GC fired on BC depolarization (middle, raster graph). This response declined over subsequent current stimulations (bottom, PSTH; gray, 95% confidence interval; green, single exponential fit). The GC also fired on the visual stimulation, both before (r_before) and after (r_after) the current injection. (d, e) Results from many such experiments (6 BCs and 44 GCs; diamonds indicate the example in c). Most GCs showed desensitization in response to consecutive BC depolarizations (d, histogram of decay constants for an exponential fit as in c). This, however, did not affect the GC’s responses to other BCs driven by the visual stimulus (e; p>0.06, sign-test).

Figure 7. Rectifying and nonrectifying transmission from bipolar cells

(a) Response of two simultaneously recorded GCs to current injection of a single BC. One GC responded to both BC depolarization and hyperpolarization with opposite sign (top; “nonrectifying” transmission, black) and showed a rebound response at the end of the hyperpolarization (Supplementary Fig. 3). In contrast, the other GC responded only to the depolarization (bottom; “rectifying” transmission, red) and did not show the rebound response. (b) Population results of rectification index from many such paired recordings (0.50±0.48, mean ± standard deviation from 127 GCs in total; see Methods for details). Bottom: Each dot represents one BC-GC connection (black and red circles from a), and each row corresponds to one BC. Top: Stacked histogram across all BC-GC connections (black, “nonrectifying” connections; red and cyan, “rectifying” connections transmitting primarily on BC depolarization or on hyperpolarization, respectively). Both types of connections were found in GC responses to 17 out of 27 BCs (such as in a). The variation of the index within individual BCs accounted for 41% of the total variation. (c) Response of two GCs to a single BC under pharmacological block of inhibitory transmission (displayed as in a but a different pair from a). (d) Population results of rectification tested without AC signaling (25 GCs in total). Despite an increase in the evoked firing rates (Fig. 3c), the inhibitory transmission blockers did not significantly change the rectification index (p>0.06; rank-sum) or the observed frequencies of rectifying and nonrectifying connections across the populations (p>0.6; χ²-test).

Figure 8. Amacrine cells can gate individual bipolar cell signals

(a) Response of two GCs to BC stimulation alone (top) or in conjunction with visual stimulation in a distant annulus (bottom). The visual stimulus served to drive lateral AC circuits (see top circuit diagram and Methods). Under these conditions the response of one GC to the central BC was suppressed (left), whereas that of the other GC was enhanced (right). (b) Changes in effective strength of BC-GC connections elicited by distant visual stimulation (−0.47±0.87; mean ± standard deviation from 221 GCs in total). Left: Each dot represents a BC-GC connection (colored circles from a), and each column is one BC. 15 out of 20 BCs showed distinct modulations among their connections. This variation from individual BC signals accounted for 59% of the total variation. Right: stacked histogram across all connections. Background stimulation weakened 65 connections (cyan; see left side of the circuit diagram at top) but strengthened 15 connections (red; right side). (c) The effects of background stimulation on convergent connections from two different BCs. Display as in b, but columns correspond to individual GCs. In 3 of 9 cases, the two BC-GC connections experienced significantly different gating (see Supplementary Fig. 4 for an example). (d) The effects of background visual stimulation on the transmission from central BCs to GCs before (left) and after (right) applying inhibitory transmission blockers. The drug application eliminated both the suppressive and facilitatory gating effects (p<0.007, χ²-test; p<0.02, Levene’s test; see Supplementary Fig. 5 for an example).