Synchronous processing of temporal information across the hippocampus, striatum, and orbitofrontal cortex

Reviewer #1 (Public review):

Summary:

It is known that neuronal activity in several brain regions encodes interval time. However, how interval time is encoded across distributed brain regions remains unclear. By simultaneously recording neuronal activity from the hippocampal CA1, dorsal striatum, and orbitofrontal cortex during a temporal bisection task, the authors showed that elapsed time during the interval period is encoded similarly across these regions and that the neuronal activity of time cells across these regions tends to be synchronized within 100 ms. Using Bayesian decoding, they demonstrated that the interval time decoded from the firing activity of time cells in these regions correlated with the rats' decisions and that the times decoded from the neuronal activity of different brain regions were correlated. The sound experiments and analyses support most of the main conclusions of this paper.

Strengths:

They used a temporal bisection task in which the effects of time and distance can be dissociated. The test trials successfully revealed the relationship between the interval time estimated by Bayesian decoding and the animal's judgment of long versus short interval times. Simultaneous recording of neuronal activity from the hippocampal CA1, dorsal striatum, and orbitofrontal cortex, which is technically challenging, allowed comparison of interval time encoding across brain regions and the degree of synchrony between neurons from different brain regions.

Weaknesses:

Some analyses were not explained in detail, making it difficult to assess whether their results support the authors' conclusions.

Reviewer #2 (Public review):

Summary:

In this work, the authors examined how neural activity related to temporal information is distributed and coordinated throughout the hippocampus, dorsal striatum, and orbitofrontal cortex. Rats were forced to run for fixed time intervals on a treadmill and make a decision based on whether the interval was long (10s) or short (5s). Under these conditions time cells were observed across all examined brain regions. The primary finding of the authors is that synchronized activity between time cells across brain regions is entrained into the theta cycle. This observation is used to support the central claim that the sharing of temporal information is mediated by the theta oscillation.

Strengths:

By simultaneously recording several brain regions in an interval discrimination task, the authors provide a valuable dataset for understanding how temporal information is processed and distributed throughout relevant networks.

Weaknesses:

Several methodological concerns should be addressed and a more focused analysis should be performed to strengthen the central claims of this work.

Major Concerns

(1) The restriction to only use time cells to understand temporal information processing. Other mechanisms of encoding time, like population clocks and ramping, have been characterized in the striatum and frontal cortex, and these dynamics might contain more temporal information than the subset of cells that meet the statistical criteria for being a time cell. Furthermore, time cells in the OFC, and DS in particular, appear to be heavily biased towards the beginning of treadmill running. This raises the question of whether temporal information can be encoded by neurons other than time cells in these two regions.

(2) The results of the Bayesian decoding analysis should be expanded on. In particular, the performance of each decoder above the chance level is not quantified. Comparing the performance of decoders trained on all cells to the performance of decoders trained on time cells alone would partially address the question of whether or not time cells are the only cells that can encode temporal information in the DS and OFC.

(3) The decoding results for the test trials appear different from the results in the authors' previous publication (Shimbo et. al., 2021). There, differences in decoded time between the selected-long and selected-short trials emerged after 5s, the duration of the short trials. This was to be expected given the following two reasons. First, from the task design, it is unclear that the animal can distinguish trial types (long, short, or test) until after the first 5 seconds of treadmill running, making it logical for differences in decoded time to emerge only after this point. Second, time cell activity was identical in the first 5s of the long and short trials as shown in Figure 2A. Here, however, the differences in decoded time during the selected-long and selected-short test trials emerge within the first 2s of treadmill running. Could the authors explain this discrepancy?

Furthermore, in Figure 6B, at 3 seconds of running time, the decoded time for selected-long and selected-short trials shows a difference of nearly 2 seconds, with no further increase as running time progresses. In contrast, at 2 seconds of running time, there is no significant difference in decoded time for DS and OFC, while CA1 shows a slight increase in the decoded time for selected-long trials. This pattern suggests a sudden jump in the encoded time for selected-long trials between 2 and 3 seconds. However, without explicitly showing the raw data, it is difficult to interpret this result and other results from the decoding analysis.

Minor Concerns

(1) It is not clear how the Bayes decoder was trained. Does the training data come entirely from the long trials?

(2) For Figure 5D, even if only one of two neurons in a pair has its spike rate modulated by theta, wouldn't the expectation be that synchronous spike events between these two neurons would be modulated by theta as well? This analysis might benefit from shuffling methods to determine if the mean resultant length of synchronous spike events is larger than the chance level.

(3) In Figure 5A, the authors suggest that 'the synchronization of time cells was modulated by theta oscillation.' However, it is unclear whether the population exhibits a preferred theta phase or the phase preference only occurs at the individual cell level. If there is no preference on the population level, how would the authors interpret this result?

Reviewer #3 (Public review):

Summary:

This study examines neural activity recorded simultaneously in the hippocampus, dorsal striatum, and orbitofrontal cortex as rats performed an interval timing task. The analyses primarily focus on the activity of "time cells" which are neurons that fire at specific moments during the intervals. In this experiment, the intervals consist of periods when animals are running on a treadmill before selecting the arm associated with the interval duration. The results show that the theta oscillations induced by this running behavior were observed across the three regions and that this strong oscillation modulated the activity of neurons across regions. While these findings are correlative in nature, they provide an important characterization of activity patterns across regions during complex behavior. However, more research is needed to determine whether these activity patterns specifically contribute to temporal coding.

Strengths:

(1) Overall, the paper is very well written. Although I have specific concerns about the review of the relevant literature and the interpretation of the results (see below), I do want to commend the authors for their efforts toward presenting this complex work in an accessible manner.

(2) The study is well designed and the quality of the electrophysiological data collected from multiple brain regions in such a challenging behavioral experiment is impressive. This work is a technical tour de force.

(3) The analyses are very thorough, statistically rigorous, and clearly explained and visualized. The authors provide a thoughtful mixture of example data (at the level of individual cells or animals) and aggregated data (at the group or session level) to properly explain and quantify the activity patterns of interest.