Control of prestimulus activity related to improved sensory coding within a discrimination task

Abstract

Network state influences the processing of incoming stimuli. It is reasonable to expect, therefore, that animals might adjust cortical activity to improve sensory coding of behaviorally relevant stimuli. We tested this hypothesis, recording single-neuron activity from gustatory cortex (GC) in rats engaged in a two-alternative forced-choice taste discrimination task, and assaying the responses of these same neurons when the rats received the stimuli passively. We found that the task context affected the GC network state (reducing beta- and gamma-band field potential activity) and changed prestimulus and taste-induced single-neuron activity: before the stimulus, the activity of already low-firing neurons was further reduced, a change that was followed by comparable reductions of taste responses themselves. These changes served to sharpen taste selectivity, mainly by reducing responses to suboptimal stimuli. This sharpening of taste selectivity was specifically attributable to neurons with decreased prestimulus activities. Our results suggest the importance of prestimulus activity control for improving sensory coding within the task context.

Figure 1.

Behavioral contexts and task performance. A, Schematic diagrams of the taste discrimination task (task; left) and the passive taste delivery (passive; right). In an experimental box, three ports for nose poking (triangles) were set. In task, rats obtained pseudorandomly selected taste solutions by nose poking the center port (N, NaCl; S, sucrose; C, citric acid; Q, quinine), and then were required to poke the side port associated with the delivered taste stimulus to obtain water reward. In passive, the ports were covered and the taste solutions were presented pseudorandomly, without nose poking on the part of the rat. In both contexts, the taste solutions and water were delivered directly to the rat's oral cavity through a manifold of thin tubes inserted into the implanted cannula (see main text for details). B, C, Task performance (B) and reaction time (C) are independent of the taste type (n = 28 sessions from 5 rats; p > 0.1 by one-way ANOVA). All, Average of across tastes. Error bars indicate SEM. D, E, Recording areas. D, Nissl-stained image of a recording site (arrow). An electrolytic lesion was made by passing current through electrode wires after the conclusion of the experiment. E, The black boxes indicate all recording areas, reconstructed by electrolytic lesions and tracks of electrodes, taking into account the numbers of recording sessions (electrodes were moved ventrally ∼80 μm after each recording). Tips of electrode bundles were spread over several hundred micrometers (<500 μm). Activity was recorded mainly from dysgranular insular cortex (DG), but also from granular (GI) and agranular insular (AI) cortex. For simplicity of presentation, all recording areas are projected on one hemisphere, but recordings were performed bilaterally in three rats. Schematic diagram was reprinted with publisher's permission from Paxinos and Watson (1998).

Figure 2.

Behavioral context modifies LFP. A, E, Averaged power spectra of all recording sites (n = 214 recording sites) before (0.5 s period from −0.5 to 0.0 s) (A) and after (0.5 s period from +0.3 to +0.8 s) (E) NaCl delivery. Power was lower in the beta and gamma frequency ranges in both prestimulus and poststimulus epochs of task (black lines). The three lines indicate mean and mean ± SE. B–D, F–H, Averaged power in beta (15–30 Hz) (B, F), low-gamma (40–50 Hz) (C, G), and high-gamma (70–85 Hz) (D, H) frequency ranges before (B–D) and after stimulus (F–H). In task, LFP power was significantly lower than in passive at most frequencies. B, G, p < 0.05 for main effects of context and taste by 2W-RM-ANOVA. C, H, p < 0.05 for simple main effect of context in all tastes after significant context–taste interaction. D, p < 0.05 for simple main effect of context in N (but not other tastes) after significant context–taste interaction. F, No significant difference between contexts. N, NaCl; S, sucrose; C, citric acid; Q, quinine. I, Averaged GEP in NaCl (N) trials (n = 133 recording sites) in task (black lines) and passive (gray, mean and mean ± SE). The GEP amplitude was smaller in task. J, Between-context comparison of the peak amplitudes of the GEP (filled bars, task; open bars, passive). N = 133, 127, 104, and 76 for N, S, C, and Q, respectively. Only the data containing significant peaks or valleys at least in one context (>3 times the SD of 500 ms before stimulus epoch) (also see Materials and Methods) were used. Amplitudes in task were significantly lower than those in passive for all tastes (all values of p < 0.05 by Wilcoxon's signed rank test with Bonferroni's correction). The difference between contexts was not affected by the block order in most conditions: for 8 comparisons—4 per taste × 2 possible orders—the task versus passive difference failed to reach significance in only one (Q in passive-first sessions). K, Between-context comparison of peak times of the GEP (filled bar, task; open bar, passive). N = 133, 127, 104, and 76 in N, S, C, and Q. There was no difference between contexts for any taste, nor any effect of block order. Error bars indicate SEM.

Figure 3.

Behavioral context modifies single-unit activities. Examples of taste-responsive neural firing during task (black) and passive (red). Raster graphs (top panels) indicate action potential timings (vertical hash marks) in individual trials (each row) relative to the taste delivery (at 0 s); the colored bars represent event timing during task [magenta before stimulus, center port entry; magenta after stimulus, movement start; green and yellow, right and left port entry; light blue, reward delivery (located outside the figures in many trials)]. Peristimulus time histograms (bottom panels) were computed with 0.2 s sliding windows (0.05 s steps). The neurons shown in A and B produced weak taste responses that were suppressed in task (i.e., to N, S, and C) and strong taste responses that were comparable between contexts (to Q). The neuron shown in C produced taste responses that increased in task. Prestimulus activity was modified in the neurons shown in B and D.

Figure 4.

Prestimulus and poststimulus activity is modified by task in a firing rate-dependent manner. A, A between-context comparison of prestimulus firing. Each dot indicates the mean prestimulus firing rate (FR) (0.5 s period before stimulus delivery) for each taste-responsive neuron. Means were computed for all trials across tastes in each context. Neurons with 0 firing rates were excluded for purposes of log scale plotting, but included in the statistical firing rate comparison (n = 118). B, The MIs of low firing frequency neurons were significantly less than 0, indicating the reduction of prestimulus activity in task. Asterisks indicate p < 0.05 by Wilcoxon's signed rank test with the Bonferroni's correction (n = 45, 38, 24, and 10 from left to right; total N = 117 neuron—one neuron was removed for lack of firing in either context). MI in prestimulus epoch was computed using the average firing rates for all tastes from each neuron. C, Comparison of firing rate of taste responses between contexts. All significant taste responses pooled across four tastes (n = 319 taste responses; n = 81, 84, 81, and 73 responses in N, S, C, and Q, respectively) are plotted, save for 0 firing rate cases. D, The modulation indices of low-amplitude taste responses were significantly less than 0. The asterisk indicates p < 0.05 by Wilcoxon's signed rank test with the Bonferroni's correction (n = 64, 94, 105, and 53 from left to right). Taste responses were pooled across tastes (total n = 316 responses; 3 taste responses were removed for lack of firing). Error bars indicate SEM.

Figure 5.

Trial-to-trial variability of neural activity is modified by task. A, The Fano factor (trial-to-trial firing rate variability) decreases at the time of port entry in task (black lines in left); a similar reduction occurs at the time of stimulus delivery in passive (gray lines in right). The Fano factor was computed with 0.2 s moving window (0.05 s step). The three lines indicate mean and mean ± SE. B, In task, the Fano factor during the postentry epoch is significantly lower than that during the pre-entry epoch (n = 86; asterisk: p < 0.05 by simple main effect of the epoch in task after significant context–epoch interaction). C, The Fano factor in task is significantly lower than that in passive during the prestimulus but not poststimulus epoch (n = 82; asterisk: p < 0.05 by simple main effect of the context in pretaste delivery epoch after significant context–epoch interaction). Significant reductions between epochs were only observed in passive (p < 0.05 by simple main effect of epoch in passive).

Figure 6.

Taste selectivity increases in task. A, The number of tastes which induce significant taste responses in each neuron is lower in task (n = 118). The sizes of circles indicate the number of neurons in a particular group. The biggest circle ([x, y] = [4, 4]) indicates 15 neurons, and the smallest circles (e.g., [4, 1]) indicate 1 neuron. B, C, By both sharpness (B) and strength (C) indices, taste selectivity is higher in task (n = 117; 1 neuron was removed because it failed to fire in either context). The asterisks indicate p < 0.05 by Wilcoxon's singed rank test. Error bars indicate SEM.

Figure 7.

Taste response modulation is related to prestimulus activity changes. A, The MI of taste responses is positively correlated with that in the prestimulus epoch. Each dot indicates a significant taste response (n = 306 responses pooled across tastes; n = 78, 78, 78, and 72 responses to N, S, C, and Q). Thirteen responses were removed because of 0 firings either in prestimulus or poststimulus epoch. Here, prestimulus MI was computed for each response, using the mean firing rate in each taste (compare B). B, A between-context comparison of firing rates (FR) of taste responses in pre-high (filled circles; n = 135 responses pooled across tastes from 47 pre-high neurons) and pre-low neurons (open circles; n = 180 responses pooled across tastes from 70 pre-low neurons). Each neuron was classified as either pre-high or pre-low neurons via analysis of prestimulus MI (MI-pre; pre-high: MI-pre > 0 and pre-low: MI-pre < 0; 1 neuron was not classified as either type because the neuron failed to fire in prestimulus epoch). Significant taste responses in both types of neurons were plotted together, save for neurons that fired no action potentials. Here, MI-pre was computed using the averaged firing rates across tastes for each neuron (compare A). Many responses in pre-low neurons fall below the diagonal (Y = X) line, indicating that task tends to decrease taste responses in pre-low neurons. Four responses were removed because the neurons failed to fire action potentials in either prestimulus or taste epochs. C, The mean modulation indices of taste responses (MI-post) in pre-high neuron (n = 135 responses pooled across tastes) and pre-low neurons (n = 180 responses pooled across tastes) are significantly higher and lower than 0, respectively. The asterisk indicates p < 0.05 by Wilcoxon's signed rank test with Bonferroni's correction. Four responses were removed because the neurons failed to fire action potentials in either prestimulus or taste epochs. D–F, A between-context comparison of the number of tastes inducing significant responses (D), selectivity sharpness index (E), and selectivity strength index (F) in pre-high (filled circle) and pre-low neurons (open circle). D, The context effect, but not the context–neuron type interaction is significant (p < 0.05 by 2W-RM-ANOVA; pre-high, n = 47; pre-low, n = 70 neurons). E, F, Asterisks indicate p < 0.05 by simple main effects of context in pre-low neurons after significant context–neuron type interactions (pre-high, n = 47; pre-low, n = 69; 1 neuron was removed from pre-low neurons because it failed to fire action potentials in response to any tastes). Error bars indicate SEM.

Figure 8.

Simultaneously recorded ensembles carry more taste information in task. A, An example of simultaneously recorded ensemble activity in task (left) and passive (right). Each row indicates one trial of activity sorted by taste type, and each column indicates the responses of one single neuron. Color code indicates normalized firing rate (FR) in each neuron. Neurons were sorted by best stimulus in task (N, NaCl; S, sucrose; C, citric acid; Q, quinine from left to right in the left panel). Neurons shown here were all recorded from distinct electrodes in the same session. B, Classification of GC neural activity was more successful in task (n = 14 ensembles containing 4–11 neurons; p < 0.05 by Wilcoxon's signed rank test). The classification analysis compared “similarity” between one trial activity (each row in A) and the averaged activity of each taste computed from the rest of trials (for detail, see Materials and Methods). Classification was performed in each context separately.