Measuring responses and transmission in an amacrine cell pathway.

(A) Diagram of experimental setup for simultaneous intracellular and multielectrode array recording: a uniform field randomly flickering visual stimulus with a white noise Gaussian distribution is projected through a video monitor onto an intact isolated salamander retina. Simultaneously, Gaussian white noise current (orange trace) was injected intracellularly into an amacrine cell (current condition) or the amacrine cell membrane voltage (green trace) was recorded intracellularly using a sharp microelectrode (control condition). A multielectrode array of 60 extracellular electrodes recorded spiking activities of multiple ganglion cells simultaneously. Multiple traces on the left represent spiking response of multiple ganglion cells to a single stimulus trial during the control condition (orange spikes) and current condition (green spikes). (B) The membrane potential response of a sustained Off-type amacrine cell (top row), the response of a transient On–Off type amacrine cell (middle row), to a uniform field flashing stimulus (bottom row). (C) (Top) Receptive fields of a single amacrine cell and multiple ganglion cells. Each oval indicates one standard deviation of a two- dimensional Gaussian fit to the receptive field mapped using a white noise checkerboard stimulus. Red oval indicates an amacrine cell receptive field; black ovals indicate ganglion cells affected by current injection into the amacrine cell. Grey ovals are unaffected ganglion cells. (Bottom) Temporal filters computed for recorded cells, including: amacrine visual filter (red), computed by correlating a uniform-field visual stimulus and the amacrine cell membrane potential; ganglion cell visual filters, computed by correlating the visual stimulus and ganglion spikes for the affected ganglion cells (black) and all other recorded ganglion cells (grey); and an amacrine transmission filter, between the amacrine cell and one affected ganglion cell (blue) computed by correlating a white noise current stimulus injected into the amacrine cell and the ganglion cell’s spikes.

Contributions of amacrine cells to the average ganglion cell visual feature.

(A) Illustration of a model containing two main pathways, an amacrine pathway whose feature is shown by the red linear filter, and another pathway representing other ganglion cell features whose average is shown by the blue linear filter. The outputs of these two pathways are combined by a two-dimensional nonlinear function to generate the ganglion cell’s firing rate. (B) (Left column, top) Sample amacrine–ganglion cell pair for which amacrine pathway feature (red curve) contributed to the ganglion cell’s STA (black curve) as indicated by the observation that the STA and the oSTA (blue curve) are different. (Left column, bottom) The raw stimulus distribution (light grey) and the spike-triggered stimulus ensemble distribution (black), as well as the amacrine pathway nonlinear response function (computed as a quotient of the spike-triggered and raw stimulus distributions) (green), when the stimulus is projected onto the amacrine pathway feature (red). (Right column) Same as left column for an amacrine cell for which the oSTA was similar to the STA. (C) Histogram of center-surround weighting for STA (black) and oSTA (blue) across 39 amacrine–ganglion cell pairs, expressed as the signed area curve of their STA and oSTA filters. (D) Histogram of the observed difference between the STA and oSTA for 39 amacrine–ganglion cell pairs, expressed as the angle difference in degrees. Colored symbols indicate the angle differences corresponding to the sample cell pairs shown in (B). (E) Relationship between difference between the time to peak of an amacrine pathway filter and the target ganglion cell STA (y-axis) versus the difference between STA and oSTA across 39 amacrine–ganglion cell pairs.

Differential modulation of multiple distinct ganglion cell visual features by an amacrine cell.

(A) The multi-pathway model: The space of visual features encoded by a ganglion cell was decomposed into its principal components (light blue traces) using STC analysis on the stimulus space when the amacrine pathway feature is projected out. Then for each of the ganglion cell’s feature dimensions and the amacrine pathway’s feature dimension, a two-dimensional nonlinear firing rate function is computed. (B) (Left) Illustrates a nonlinear interaction between the amacrine pathway and one of the orthogonal STC features in (A) for a pair of an amacrine cell and a ganglion cell. Red trace represents the amacrine pathway feature; black trace represents the orthogonal STC feature; and green traces show one-dimensional nonlinear response functions computed from the two-dimensional instantaneous firing rate nonlinearity for four bins of amacrine pathway output values specified by the colorbar. (C) Amacrine–ganglion cell’s features nonlinearity characterization: (Top) Orthogonal STC significant dimensions representing the ganglion cell’s other features, excluding the amacrine pathway feature, for a sample amacrine–ganglion cell pair. (Middle) Two-dimensional firing rate nonlinearity as a function of the amacrine pathway output (y- axis) and the projection of the stimulus on each orthogonal feature (x-axis). The y-axis is the projection of stimuli on the amacrine pathway feature, and the x-axis is the projection of the stimuli on the corresponding ganglion cell’s visual feature. The two-dimensional instantaneous firing rate was computed as a quotient of the spike-conditional and raw stimulus distributions. Lighter regions correspond to higher ganglion cell firing rate (FR) in the two-dimensional stimulus subspace. (Bottom) One-dimensional slices of each two-dimensional firing rate as a function of the projected stimulus onto the corresponding ganglion cell’s feature (x-axis) for different levels of the amacrine cell’s feature output. Trace color indicates different levels of the amacrine feature output, indicated by green bars in the middle row. Because the amacrine cell is inhibitory (its transmission filter has a negative peak), a high level of the amacrine pathway’s preferred feature roughly corresponds to the amacrine cell being more hyperpolarized (darker green colors).

Two types of modulatory effects of amacrine cells on the ganglion cell population.

(A) Firing rate response nonlinearities from example amacrine–ganglion cell’s feature pairs, representing five different types of nonlinear modulatory effects observed across 321 amacrine– ganglion cell’s feature pairs. The amacrine cell’s nonlinear effects include combinations of additive and multiplicative transformations of the ganglion cell’s nonlinear response function as the output of the amacrine pathway varied from weaker to stronger inhibition. (B) Illustrations of how changes in threshold, sensitivity, gain, or offset and polarity reversal as shown for sample amacrine– ganglion cell’s feature pairs in (A) can be obtained by simple additive or multiplicative operations on the ganglion cell’s nonlinear response function. (C) Distribution of different types of nonlinear modulatory effects identified by a multi-pathway model framework across 321 amacrine–ganglion cell’s feature pairs, which include changes in the response gain, sensitivity, threshold, and gain- sensitivity modulation (GSM) index, estimated via fitting a piecewise linear approximation of a sigmoidal function to the ganglion cell’s nonlinear response function. Dashed lines indicate the median of each histogram. (D) Center heatmap shows the joint distribution of response threshold modulation and gain-sensitivity modulation index, showing how these response variables change together when inhibition from the amacrine pathway gets stronger. Each data point (n = 321) represents the Pearson correlation coefficient between different levels of amacrine pathway polarization and the corresponding values of each response variable. Inset nonlinear functions illustrate changes in the nonlinear properties of a ganglion cell’s response to distinct visual features, associated with the two types of functional effects of amacrine cells, sensitivity modulation (top left), and gain modulation (bottom right). (E) The correlation coefficient of the change in the response nonlinearity parameters across visual features of each ganglion cell for each type of modulatory effect in (D) (median±SEM), showing variation of the effect across different modulated features for the same amacrine–ganglion cell pair.

Amacrine cells create two types of context-dependent modulation of ganglion cells.

Schematic diagram of how visual features of different amacrine cells represent a context that drives two types of modulation. The first is the control of input sensitivity of visual features prior to summation within the ganglion cell. The second is control of the output gain of the ganglion cell, which acts similarly on all visual features of the ganglion cell. Colors used for differentiating different connections match those used in figure 4D-F.