Stimulus dependence of barrel cortex directional selectivity

Abstract

Neurons throughout the rat vibrissa somatosensory pathway are sensitive to the angular direction of whisker movement. Could this sensitivity help rats discriminate stimuli? Here we use a simple computational model of cortical neurons to analyze the robustness of directional selectivity. In the model, directional preference emerges from tuning of synaptic conductance amplitude and latency, as in recent experimental findings. We find that directional selectivity during stimulation with random deflection sequences is strongly dependent on the mean deflection frequency: Selectivity is weakened at high frequencies even when each individual deflection evokes strong directional tuning. This variability of directional selectivity is due to generic properties of synaptic integration by the neuronal membrane, and is therefore likely to hold under very general physiological conditions. Our results suggest that directional selectivity depends on stimulus context. It may participate in tasks involving brief whisker contact, such as detection of object position, but is likely to be weakened in tasks involving sustained whisker exploration (e.g., texture discrimination).

Figure 1. Definition of directionally selective model neuron.

A: Latency tuning: directional dependence of excitatory and inhibitory synaptic conductance peaks (left) and latencies (right). Symbols: Gray squares, excitatory conductance; white squares, inhibitory conductance; black circles, total synaptic conductance. B: Polar plot of directional selectivity of model neuron tested with discrete single deflections in one of eight possible directions (symbolically represented at bottom: solid line is 0° deflection). Responses are on a radial scale normalized to the preferred direction (PD, set to 0°) response (outer circle = 1). Symbols: Circles, gray line: synaptic responses (mean peak evoked synaptic potential per deflection). Squares, black line: spiking responses (mean evoked number of spikes per deflection, counted over a 20 ms window).

Figure 2. Directional selectivity is sensitive to stimulus frequency.

A: Plot of directional selectivity tested with 20 Hz random deflection sequence. Deflections occurred at exponentially distributed intervals (symbolically represented at bottom left: each vertical line is a deflection) and were across neighboring positions on a diamond spatial grid (bottom right). Thus, there was a “whisker displacement” from one position to a neighbor every 50 ms on average. Response scales and symbols as in Fig. 1B. B: Plot of directional selectivity tested with 200 Hz random deflection sequence. Except for frequency, all other parameters were as for A (symbols at bottom represent the faster sequence and the unchanged position grid). Directional selectivity was substantially weakened.

Figure 3. Robustness to changes in key model parameters.

All panels represent the “Selectivity Index” (see Materials and Methods) computed on spiking responses. A: Dependence on time window for response estimation. Selectivity index as a function of stimulus frequency assessed using two different integration time windows after each deflection (10 and 20 ms). B: Dependence on short-term synaptic plasticity. The selectivity index was evaluated for conditions differing in how synaptic depression properties were matched across excitatory and inhibitory conductances (ND: no depression, same model as in other figures; SBD, WBD and NBD defined in Results). C: Dependence on choice of grid geometry on which whisker deflections were defined (square: deflections on square grid; diamond: on diamond grid; random: random walk deflections; see Materials and Methods for full explanation and Fig. 2, bottom, for schematic of diamond grid). In all cases, response selectivity was weakened at high frequencies. Error bars: standard error of mean across trials.