Electrophysiological Data Of Striatum And Cortex During Timing

We have recently begun testing the hypotheses generated by the SBF model by recording from ensembles of neurons in both the striatum and cingulate cortex of rats while they perform an interval timing task (Matell et al., in press; for general methods on ensemble electrophysiology, see Nicolelis et al., 1997). As with all electrophysiological investigations in behaving animals, it is necessary to design the behavioral task so that the neural activity related to the cognitive variables of interest (in this case, the perception of time) can be dissociated from the motor activity resulting from task performance. While never an easy feat in freely behaving animals, this dissociation is all the more difficult in the study of timing, as the behaviors of the rats evolve as a function of time (Fetterman et al., 1998).

15.4.1 Methods

To overcome this difficulty, we utilized a matched behavior analysis, in which we compared the same behavior across different periods in time, so that time itself was the primary variable that differed. To this end, we trained the rats on a multiple-duration, fixed-interval procedure. Specifically, rats were trained that food reinforcement could be earned for the first lever press either 10 or 40 sec after stimulus onset. Responses prior to the criterion time had no consequence. Only one response lever was available, and a barrier was constructed around this lever so that the rat could only respond with its right forepaw. In this manner, all responses on the lever would be largely identical in terms of motor activity. The trial ended after the delivery of reinforcement at either the early (10 sec) or late (40 sec) duration. No information was provided to the rat as to which duration would be reinforced on each trial.

This reinforcement schedule produced two bursts of lever pressing on each trial, one press burst occurring around the short duration and, if reinforcement was not delivered (i.e., if the late duration was primed), a second burst of pressing occurred around the late duration. Over trials, the distribution of lever pressing on late trials had a peak around 10 sec and a scallop up to 40 sec. By choosing reinforcement probabilities for the two criterion durations that were inversely proportional to the reinforcement densities, the rats' peak rates of lever pressing at these two durations were nearly identical. The distribution of lever pressing on late trials from an individual rat in this task is shown in Figure 15.2.

15.4.2 Neural Firing Patterns

Given the similarity of response rates for pressing at the two durations, the moment-by-moment behavior of the rat was approximately the same when the rat was pressing for the early-duration reward and when it was pressing for the late-duration reward. Therefore, the neural activity corresponding to these periods of pressing should be identical in terms of overt motor aspects, and as such, differences in neural activity can be interpreted as resulting from the processing of different durations. The firing pattern of a striatal neuron while the rat behaved on the late trials is shown in Figure

10 second Peak-type Striatal Unit

I—I Striatal Unit Lever Presses

FIGURE 15.2 Distribution of lever presses and striatal activity as a function of trial time. Rats were trained that they could earn food at either 10 or 40 sec from stimulus onset (0 sec). Data are from trials in which food was available at 40 sec. Rates of lever pressing peaked at both 10 and 40 sec, demonstrating the rat's temporally specific expectation of food reward. In contrast, the striatal neuron fired maximally at 10 sec, but showed significantly less activity at 40 sec, thereby demonstrating its temporal specificity. The shape of the neural activity is classified as a peak because the magnitude of firing rate change from the baseline at 10 sec is greater than twice the change at 40 sec. Data have been smoothed by a 3-sec running mean, and the y-axis is partially shown to emphasize the match between striatal activity and behavior.

I—I Striatal Unit Lever Presses

FIGURE 15.2 Distribution of lever presses and striatal activity as a function of trial time. Rats were trained that they could earn food at either 10 or 40 sec from stimulus onset (0 sec). Data are from trials in which food was available at 40 sec. Rates of lever pressing peaked at both 10 and 40 sec, demonstrating the rat's temporally specific expectation of food reward. In contrast, the striatal neuron fired maximally at 10 sec, but showed significantly less activity at 40 sec, thereby demonstrating its temporal specificity. The shape of the neural activity is classified as a peak because the magnitude of firing rate change from the baseline at 10 sec is greater than twice the change at 40 sec. Data have been smoothed by a 3-sec running mean, and the y-axis is partially shown to emphasize the match between striatal activity and behavior.

15.2. As can be seen, there was a phasic increase in neural activity around 10 sec without a robust increase as time approached 40 sec.

In order to quantitatively evaluate whether this neural activity at 10 sec was significantly different from that at 40 sec, we compared the number of spikes that occurred during an early vs. late press period. The lengths of these press periods were matched so that the number of presses, width of the press window, and therefore mean rate of pressing were all identical across the two durations. We found significantly different firing rates in 28% of the striatal neurons recorded and 20% of the cortical neurons recorded. Of these differentially active striatal neurons, 93% had activity levels during one or both of the press periods that crossed a 99% confidence interval, indicating that they were firing at near-maximal levels. Similarly, 91% of the differentially active cortical neurons crossed this confidence interval. Having maximal (or minimal) activity at the criterion durations and having differences in activity across durations suggest that these neurons were primarily involved in the encoding of specific signal durations.

Of those striatal neurons showing both differences across duration and maximal activity at these signal durations, 71% showed activity patterns similar to that illustrated in Figure 15.2, which we classified as peak shaped because the magnitude of change in firing rate during the early-duration press period was more than twice the magnitude of change during the late-duration press period. In addition to this peak firing pattern, 14% of the primarily timing striatal neurons showed a modu-latory firing pattern, which we defined as showing peaks in firing associated with one or both of the behavioral peaks, but where the magnitude of change across durations was less than twofold. Also, 14% of the primarily timing striatal neurons showed ramp-like activity patterns, in which the firing rate showed monotonic changes as a function of time. In contrast, 40% of the primarily timing cortical neurons had peak-type changes, 50% were classified as modulatory, and 20% as ramp (see Figure 15.3).

Further inspection of the rats' behavior during these press periods using a finegrained video analysis showed that their behavior alternated between bursts of pressing and checking the food magazine for reward. Because the levels of magazine checking often varied as a function of time in the trial, and could potentially account for the neural activity differences found across signal durations in the mean functions, the data were reanalyzed by comparing the neural firing rates during only those periods of time in which the rat was engaged in a press burst. Any differences in press topography (e.g., press duration, interpress interval, release-press interval) that may have changed as a function of signal duration were controlled for by using these values as covariates in the analysis.

The results of this motor-controlled analysis showed that a total of 12 of 54 (22%) striatal and 8 of 54 (15%) cortical neurons had differences in neural activity as a function of signal duration that could not be explained by any measured motor variables. Of these neurons, eight striatal and four cortical neurons were retained from the earlier mean function analysis, and four striatal and four cortical neurons were added (i.e., they did not show significant difference when testing without the press covariates). Of those striatal neurons retained, 75% were peak, 12% were modulatory, and 12% were ramp. Of the cortical neurons retained, 0% were peak, 50% were modulatory, and 50% were ramp. Those neurons that were added were not given shape classifications because their mean functions were not indicative of the duration-related differences found in the latter analysis.

15.4.3 Discussion

Finding a different magnitude of neural activity at one vs. the other criterion duration provides compelling evidence that striatal and cortical neurons are intimately involved in interval timing. However, the shapes of their neural activity patterns tells us considerably more than general involvement in timing. As can be seen in Figure 15.2, the firing rate modulation of this striatal neuron was peak shaped, firing maximally around the first criterion time of 10 sec, which matches the decision stage output function hypothesized by the SBF model (Matell and Meck, 2000, in press) and other interval timing models (Gibbon and Church, 1984; Staddon and Higa, 1999). Because the majority of striatal neurons showing a difference in firing rate

40-second Modulatory-type Cortical Unit

Time

Cortical Unit Lever Presses

40-second Ramp-type Cortical Unit

0 0

Post a comment