Behavioral performance - Neural activity in the posterior thalamus and hippocampus during time

Activity of single neurons was recorded from the hippocampus and thalamus (Fig. 4) of five rats performing a time interval discrimination task (Fig. 1). In this task, the animals had to discriminate six different durations of time interval that were presented in random order into long (3979 ms) or short (3629 ms) ones and navigate to the corresponding goal locations to obtain water reward (Fig. 1). The probability for the animal to choose the long target ( ) increased as a function

of the sample interval duration, which was well accounted for by a logistic regression model (R²= 0.81±0.005; Fig. 4). The animals chose the correct target in 82.6 0.07% of trials.

B. Example of Neuronal responses

A total of 411 and 459 well isolated single units were recorded in this task and the majority were putative pyramidal cells (n = 330, 80.3%; Fig. 3) in hippocampus. Since, it is unclear to discriminate putative pyramidal cells from putative inhibitory neurons in thalamic single units, all 459 well isolated single units were used for all analyses. Of these, only those units with mean firing rate

≥ 0.1 Hz during sample interval presentation were subject to analysis. Diverse

types of neuronal activity profiles were observed during sample interval

presentation in hippocampus and thalamus (Fig. 4-5, respectively). Of these, a number of single units were a monotonically changing activity profile (“ramping activity”) (Bodner et al., 1997; Durstewitz and Seamans, 2006); many neurons gradually increased or decreased their activity over time.

Fig. 4. Choice behavior The graphs show the fraction of long-target choices ( ) as a function of sample interval duration. The error bars denote SEM. The solid lines were determined by logistic regression and the shading indicates 95%

confidence interval.

Fig. 5. Examples of individual activity of hippocampus Spike raster and spike density functions (σ = 100 ms) are shown for example hippocampus. The

two rows (upper row is raster plot; lower row is spike density functions) represent single unit. Each column means six different interval duration. Trials were grouped according to the length of sample interval and the abscissa denotes time since the onset of each sample interval (The left-most column means the shortest one). Gray vertical lines denote the onset of each time interval. As shown, diverse types of neuronal activity were observed during sample interval presentation.

Fig. 6. Examples of individual activity of thalamus Spike raster and spike density functions (σ = 100 ms) are shown for example thalamus. The two rows

(upper row is raster plot; lower row is spike density functions) represent single unit. Each column means six different interval duration. Trials were grouped according to the length of sample interval and the abscissa denotes time since the onset of each sample interval (The left-most column means the shortest one). Gray vertical lines denote the onset of each time interval. As shown, diverse types of neuronal activity were observed during sample interval presentation.

C. Temporal Information

Hippocampal and thalamic neurons with mean firing rate ≥ 0.1 Hz during

sample interval presentation (n = 147, 413; respectively) were subject to analysis, and only correct trials were analyzed unless noted otherwise. I examined whether hippocampal and thalamic neurons transmitted information on the elapse of time.

This was done in two different ways. First, Sample intervals were classified into short or long ones based on neuronal ensemble activity during the last 500 ms of each sample interval using a leave one out cross validation procedure (length classification). Length classificationbased on all recorded units across sessions (n

= 147, hippocampus; n = 413, thalamus) assuming independence among neurons was significantly above chance level (80.0% correct; binomial test, p <<

0.01,hippocampus; 95.1% correct; binomial test, p << 0.01, thalamus ; Fig. 7A).

Second, I examined how well the hippocampal and thalamic neuronal ensemble kept track of the elapse of time (decoding elapsed time; each interval was divided into 10 equal-duration bins). All recorded units (n = 147, hippocampus; n = 413, thalamus) were used for decoding elapsed time assuming independence among neurons, the mean error (the distance between the actual and predicted bins) was 1.4±0.08 bin for the longest sample interval (4784 ms) in hippocampal neuronal ensemble, 1.1±0.08 bin in thalamic neuronal ensemble, which were well below chance level in both neuronal ensemble (paired t test, p

<0.01, hippocampus; p < 0.01, thalamus; Fig. 8A).

Neuron dropping analyses revealed that > 150 neurons were needed for asymptotic performance of neural decoding (Fig. 7B, 8B). These results suggest that many more neurons are required to accurately keep track of elapsed time in the range of a few seconds compared with representing the animal’s choice in a simple binary-choice task.

Fig. 7. Neuronal Ensemble Decoding of Choice Hippocampal and thalamic neuronal ensembles convey temporal information. Sample intervals were classified into short or long ones based on neuronal ensemble activity during the last 500 ms of each sample interval using a leave -one-out cross-validation procedure. Only correct trials were included in the analysis. Gray, trial -by-trial decoding results (0, short-target choice; 1 long-target choice); black, their mean and SEM (A). Ensemble size, hippocampus n = 147; thalamus n = 413 neurons. (B) Results of a neurondropping analysis for A.

Fig. 8. Neuronal Ensemble Decoding of the elapse of time The longest sample interval (4784 ms) was divided into ten equal-duration bins, and the order of the middle eight bins was determined based on neuronal ensemble activity within each bin. Only correct trials were included in the analysis. Gray, trial -by-trial decoding results (0, short-target choice; 1 long-target choice); black, their mean and SEM. Thalamus-matched, decoding results based on randomly selected 147 neurons (A). (B) Results of a neurondropping analysis for A.

D. Individual Neuronal Responses

A reliable pattern of sequential activation of neuronal activity was observed in the CA1 region of hippocampus during the delay period of a memory task (Pastalkova et al., 2008; Gill et al., 2011; Itskov et al., 2011; MacDonald et al., 2011; Kraus et al., 2013). The pattern of sequential activation of simultaneously recorded neurons for the longest sample interval duration (Fig.9) was reliable across trials and lasted 4784 ms without repeating itself in hippocampus but not in thalamus. Therefore, I hypothesized that the population spiking activity of only hippocampal neurons at any point in time during a trial could be used to infer elapsed time.

To investigate the difference activity profiles between hippocampal and thalamic neurons, activity profiles of all recorded hippocampal and thalamic neurons during the longest sample duration were shown together (minimum firing rate, 0.1 Hz; Fig. 10). Hippocampal neurons tended to be active briefly, whereas thalamic neurons tended to be active largely for the entire interval duration.

Fig. 9. Example of simultaneously recorded hippocampal and thalamic neurons during the longest sample duration. Each row of all graphs represents normalized mean firing rate (50 ms bins) of the longest interval duration trials for one neuron (minimum firing rate, 0.1 Hz). The neurons were sorted by the latency of peak firing rate. First column indicate average peak firing rate of one session in hippocampus (A) and thalamus (B). Second and third column indicate examples of single trial for the first column. Scale bar’s range means normalized cell firing rate.

Fig. 10. Neuronal responses of all recorded neurons. Activity profiles of all recorded hippocampal and thalamic neurons during the longest sample duration are shown together (minimum firing rate, 0.1 Hz). Hippocampal neurons tended to be active briefly, whereas thalamic neurons tended to be active largely for the entire interval duration. Scale bar’s range means normalized cell firing rate.

E. Response Duration

To investigate individual neuronal response of hippocampal neurons is different from individual neuronal response of thalamic neurons during a given sample duration, the duration width between the maximum and half -maximum of each neuron’s spike density function and frequency histograms for the duration was quantified. Frequency of hippocampal neurons which have the most narrow width during the longest sample duration is larger than thalamic neurons. As the width is wider, the number of thalamic neurons are larger than that of hippocampal neurons (Fig. 11A). I also examined the duration width between the maximum and half-maximum for the each six different sample duration. In all of six sample interval, width of thalamic neurons is much wider than that of hippocampal neurons. These results suggest that hippocampal neurons convey temporal information sequentially, whereas thalamic neurons transport temporal information generally during the delay duration.

Fig. 11 Durations of neuronal response (A) Frequency histograms for the duration between the maximum and half-maximum of each neuron’s spike density function (σ = 100 ms) during the longest sample interval (4784 ms; orange,

hippocampus; green, thalamus). When the maximal activity was located in the middle of the interval so that two estimates of maximal -half maximal duration were available, the longer one was taken. (B) The mean duration between the maximum and half-maximum for all six sample interval durations.

F. Multiple linear regression

To explore whether there were cells that fired reliably during a particular period within the longest sample duration, a generalized linear model of firing rate during the delay period using a normal function was constructed (see Materials and Methods). The full model included dependence of neural activity on elapsed time (i.e., bin number, T) considering other confounding factors of the animal’s behavioral variations (X-position (X), Y-position (Y) and displacement (D)) during each bin and previous goal choice (PC). Significant fractions conveyed neural signals for the elapse of time in both brain region (T, hippocampus n = 55, 36.4%, binomial test p << 0.01; thalamus n = 206, 46.7%, p

<< 0.01). I also examined whether the amount of temporal information is different between hippocampal and thalamic neurons by χ² test. Number of cells between hippocampal and thalamic neurons modulated by elapse of time is significantly different. (p = 0.049, Table 1). These results indicate that thalamic neurons convey temporal information with a climbing activity than hippocampal neurons.

Table 1. Neural signals for the elapsed time

G. Time Field

Firing rate of profile of each neuron triggered at the beginning of sample intervals was calculated from action potentials generated during animal ’s behavior.

In order to distinguish place field of hippocampal neurons, “time fields” were defined as the areas of an increased firing rate in a given sample interval. To identify a time field, a spike density function (σ = 100 ms) for the longest time

interval for each neuron was generated, and divided it into 50ms bins. A time field was defined as the minimum 5 adjacent bins with mean firing rate in each bin >

1.5 SD above the mean. The width of individual fields was determined by the duration between the maximum and half-maximum of each neuron’s spike density function during the longest sample interval (4784 ms). When the maximal activity was located in the middle of the interval so that two estimates of maximal -half maximal duration were available, the longer one was taken.

To investigated whether the difference proportion of neurons between hippocampus and thalamus, χ²test was performed. Significantly large proportion of hippocampal neurons have a time field compare to thalamic neurons (p = 0.0061, Table 2).

49 Table 2. Number of cells with a time field

IV. DISCUSSION

I examined neuronal activity in the hippocampus and thalamus while rats were performing a temporal discrimination task, and obtained two major proposals.

First, both hippocampal and thalamic neuronal populations convey information about the elapse of time. Second, neural processes underlying transmission of temporal information are different between the two areas. Thalamic neurons tended to transmit temporal information based on monotonically changing activity profiles, while hippocampal neurons tended to convey temporal information based on sequential activation of multiple neurons. Second, according to the different way of transportation of temporal information in both areas, different timing mechanisms might be adopted during duration’s encoding.

A. Role of hippocampus in interval timing

Conflicting results have been reported regarding the role of hippocampus in interval timing. Damage to the hippocampus in birds and rats had no effect on their timing ability or on their ability to discriminate visual stimuli on the basis of either orientation or size (Dietrich et al., 1997; Colombo et al., 2001). Also, following hippocampal damage, rats often respond haphazardly, that is, their response rate is not influenced by the time since the last response (Meck et al., 1984).

On the other hand, a large and growing body of literatures indicates a key role for the hippocampus in encoding and retrieving time in the range of seconds to minutes, in humans and animals and across a broad range of behavioral paradigms.

Several studies suggested that the hippocampus is involved in maintaining and retrieving information about interval timing. Such evidence comes from individuals with selective hippocampal damage (Mayes and Montaldi, 2001;

Spiers et al., 2001) and fMRI studies (Kumaran and Maguire, 2006; Staresina and Davachi, 2006; Ekstrom and Bookheimer, 2007; Lehn et al., 2009; Tubridy and Davachi, 2011).

A particularly striking example (MacDonald et al., 2011; Macdonald et al., 2013) showed that very large proportion of hippocampal neurons in rats encodes sequential events and hippocampal neuronal activity bridges and disambiguates the identical empty delay between the object and odor that compose each sequence. Hippocampal neurons fired at particular times during key events that occur reliably at particular moments (the objects and odors), and “time cells”

encoded sequential moments during an extended discontiguity between those identifiable event. The evidence that neurons that fire at particular moments in the delay period are “time cells” parallels the evidence that hippocampal neurons that fire at particular locations in space are : “place cells”. Previous work on hippocampal neuronal activity in rats performing T-maze alternation tasks has shown that hippocampal neuronal ensembles similarly disambiguate overlapping

spatial routes (Frank et al., 2000; Wood et al., 2000; Shapiro et al., 2006). In an extension of these studies, Pastalkova et al. (2008) revealed the existence of hippocampal neurons that fire at specific moments as rats walk on a running wheel between trials, and some of these cells distinguished subsequent left and right turn trials. The present observation indicates that hippocampal neurons also encode specific times between non-spatial events and disambiguate non-spatial sequences.

Our examination of changes in hippocampal neuron’s firing patterns following discrimination of the sample duration reveals that hippocampal neurons respond in sequential ways. A previous study’s task for rats is object-delay odor sequences, in which animals are required to memorize the association between object and odor and recall it during the delay duration. Therefore, they named hippocampal neurons that fired at specific time as “time cells”, and the information of time cells can be mixed with memory for the association between object and odor. The present findings reveal that a large proportion of hippocampal neurons conveys temporal information based on sequential discharges in a purely temporal discrimination task.

B. Role of thalamus in interval timing

Animals can predict the time of occurrence of a forthcoming event relative to a preceding stimulus, i.e. the interval time between those two, given previous

learning experience with the temporal contingency between them. Accumulating evidence suggests that a particular pattern of neural activity observed during tasks involving fixed temporal intervals might carry interval time information; the activity of some cortical and subcortical neurons ramps up slowly and linearly during time intervals, like a temporal integrator, and peaks around the time at which an event is due to occur. The slope of this climbing activity, and hence also the peak time, adjust to the length of a temporal interval during repetitive experience with it (Durstewitz and Seamans, 2006).

Komura et al. (2004) found linearly climbing acti vity in thalamic neurons in a trace conditioning protocol where a predictive stimulus precede d a reward by a fixed interval. Climbing activity was triggered by the predictive (conditioned) stimulus and peaked around the time at which the reward was due. As the delay between the offset of the predictor and the delivery of reward was either increased or decreased, the slope of climbing activity adjusted within a few trials to the new temporal interval.

In present study, I found diverse types of neural activity during the sample duration including ramping activity for each of six different sample duration (Fig.

6). Significant fractions conveyed information on the elapse of time based on monotonically changing activity profiles (e.g., n = 206, 46.7% during the longest sample duration, Table 1). These results suggest that the posterior thalamic region might be involved in interval timing and that is conveys temporal information

54 largely based on ramping activity.

C. Relationship with the other brain regions

Timing-related neural activity has been found in many different areas of the brain, which is consistent with the view that the brain is equipped with multiple intrinsic clocks rather than a central dedicated clock (Mauk and Buonomano, 2004;

Ivry and Schlerf, 2008). Numerous brain structures such as cerebellum (Ivry et al., 2002; Spencer et al., 2005), orbito frontal cortex (Tsujimoto et al., 2009), parietal cortex (Leon and Shadlen, 2003; Bueti and Walsh, 2009), prefrontal cortex (Kim et al., 2013), and supplementary motor area (Mita et al., 2009; Kotz and Schwartze, 2011; Shinomoto et al., 2011) transmit temporal information even though underlying mechanisms might differ (Matell and Meck, 2004; Buhusi and Meck, 2005). My results showed that hippocampal and thalamic neural signals related to interval timing were not particularly weak compared to those in the other brain regions. Considering these results and anatomical connection patterns among these brain subregions, I suggest that hippocampus and posterior thalamus provide separate temporal information signals to the cortico -striatal circuits.

Cortical projections to the striatum are topographically ordered in a series of parallel anatomical ‘loops’ running from neocortex to the striatum, pallidum, thalamus, and back to neocortex (Pennartz et al., 2009). This loop has been proposed as the major site for interval timing (Matell and Meck, 2004; Mathai and

55 Smith, 2011).

Nevertheless, I cannot rule out the possibility that hippocampus and thalamus or other regions convey temporal information independently because there are direct connections between subregions apart from cortico -striatal loop:

hippocampus and thalamus also projects to prefrontal cortex. To examine this possibility, it is needed to record multiple regions simultaneously across hippocampus, striatum, thalamus and frontal cortex in an interval timing paradigm.

D. Dedicated vs. distributed time models

Dedicated models have been proposed for temporal processing. In this view, timing in the nervous system would be analogous to that in the computer, in which a central clock sends out information to many other component s of the computer.

In the psychological literature on timing, by far the most influential model has been the centralized internal clock model (Treisman, 1963; Buonomano and Karmarkar, 2002). Centralized internal clocks are hypothetical mechanisms in which a neural pacemaker generates pulses; the number of pulses relating to a physical time interval is recorded by a counter. Centralized internal clock models are generally dedicated: one clock is used for all timing tasks (Buonomano and Karmarkar, 2002).

If timing relies on a dedicated mechanism, significant signals about durations would not be expected in hippocampus and thalamus, since prefrontal cortex is

already known to convey a large amount of temporal information (Kim et al., 2013). However, in the present study, hippocampus and thalamus have a large proportion of significant cells according to the elapsed time (Table 1), and have also conveyed temporal information on the elapse of time (Fig.8). In addition, hippocampal neural populations are shown sequential activity during the longest interval duration. A large proportion of thalamic neurons are also shown climbing activity during the delay duration (Table 1). In the present study, timing-related neural activity was not strongly confounded with motor responses, because the animals initiated navigation only after a time-interval offset. By carefully analyzing potential influence of uncontrolled behavioral variations on neural activity, both hippocampal and thalamic neurons conveyed temporal information on the elapse of time that cannot be explained by behavioral variations during sample intervals. Therefore, these results are consistent with distributed model that there is no specialized brain system for representing temporal information, asserting that time is inherent in neural dynamics.

An essential feature of this model is that temporal representation is context dependent. This property not only implies modality specificity but also that, even within a modality, the representation of a particular interval will be state dependent. One study examined performance variability of a group of l3 subjects in eight different tasks that involved the processing of temporal intervals in the subsecond range (Merchant et al., 2008). These tasks differed in their

sensorimotor processing (perception vs. production), the modality of the stimuli used to define the intervals (auditory vs. visual), and the number of intervals (one or four). The performance variability was larger not only in perceptual tasks than that in motor-timing tasks, but also when using visual rather than auditory stimuli,

문서에서 Neural activity in the posterior thalamus and hippocampus during time interval discrimination in rats (페이지 41-0)