Sub-second dynamics of theta-gamma coupling in hippocampal CA1

Lu Zhang, John Lee, Christopher J. Rozell, Annabelle C. SingereLife 2019 · DOI

TL;DR

This paper introduces a novel method to analyze the rapid, sub-second dynamics of theta-gamma oscillations in the hippocampus, revealing four distinct functional states that the brain switches between frequently. This work provides a new lens for understanding how the hippocampus dynamically processes information.

What Problem Does This Paper Address?

The hippocampus, a key brain region for memory and navigation, exhibits complex oscillatory activity, particularly the nesting of faster gamma oscillations within slower theta oscillations. These theta-gamma coupling patterns are thought to reflect different internal brain states and communication patterns between brain regions. However, existing methods typically average neural signals over long periods, obscuring the rapid, cycle-to-cycle changes that might be crucial for flexible cognitive functions. This paper aims to address this limitation by developing a method to precisely track these dynamic theta-gamma coupling states at a single-cycle resolution.

Methods

The authors developed a two-step approach:

  1. Cycle-by-cycle classification: They used wavelet analysis to extract frequency and phase information for each individual theta cycle. This data was then fed into a k-means clustering algorithm, guided by community clustering to determine the optimal number of states (four in this case). This method focuses on both the frequency content of gamma oscillations and their phase relationship with theta.
  2. State transition analysis: Once theta cycles were classified into distinct states, the authors modeled the transitions between these states using Markov chain theory. This allowed them to quantify how often the brain switches between states and the probabilities of these transitions.

Datasets

The study utilized electrophysiological recordings from two previously published datasets:

  • Hc-11 dataset: Recordings from four male Long-Evans rats during novel spatial learning tasks. This dataset was used for clustering, state occurrence/transitions, spike-field analysis, and place cell analysis.
  • Hc-3 dataset: Recordings from eleven male Long-Evans rats in CA1, CA3, or entorhinal cortex (EC) during sleep or linear maze sessions. This dataset was primarily used for CA3-CA1 and EC-CA1 LFP pairwise phase consistency analysis.

Key Insights and Ideas

  • The hippocampus is a rapid state-shifter: The most striking finding is the demonstration that the hippocampus doesn't just maintain a state for extended periods; it frequently switches between distinct theta-gamma coupling states on a cycle-to-cycle basis (within milliseconds).
  • Four distinct functional states: The clustering revealed four stable theta-gamma coupling states, which the authors named S-gamma, M-gamma, EF-gamma (early fast gamma), and LF-gamma (late fast gamma). These states are not just frequency differences; they also differ in their phase relationships and functional roles.
  • States have distinct connectivity profiles: The identified states show differential coupling with other brain regions. S-gamma states are strongly coupled with CA3 (involved in memory retrieval), while M- and EF-gamma states show stronger coupling with the entorhinal cortex (involved in sensory processing). This suggests states are tailored for different computational tasks.
  • Neural coding varies by state: Place cell activity, including spatial information and phase precession, differs significantly across these states. S-gamma states are associated with lower spatial information and less phase precession, suggesting a role in slower, less spatially precise processing.

Key Results

  • Four identified theta-gamma coupling states: S-gamma (30-50 Hz), M-gamma (60-120 Hz), EF-gamma (early phase, >120 Hz), and LF-gamma (late phase, >120 Hz).
  • Rapid state transitions: Transition probabilities between states ranged from 0.15 to 0.36, with the probability of remaining in the same state less than 0.5, indicating frequent switching.
  • Behavior-dependent state dynamics: State occurrences and transitions changed significantly during active exploration (track sessions) compared to rest periods, with S-gamma decreasing and M/EF-gamma increasing during navigation.
  • Differential connectivity: Stronger CA3-CA1 coupling was observed during S-gamma states, while stronger EC-CA1 coupling was seen in M-, EF-, and LF-gamma states.
  • Neural coding differences: Place cells exhibited lower peak firing rates and spatial information during S-gamma states. M- and LF-gamma states showed a greater contribution to theta phase precession.

Limitations and Future Work

  • Methodological robustness: While the method showed robustness across recording depths and animals, cross-validation revealed variability, suggesting that applying the identified states directly as a standard might require caution. The authors recommend others apply their method to their own data.
  • Tetrode data variability: Clustering results from tetrode recordings showed some variation, with recommendations to select channels with slow gamma at the peak of theta for analysis.
  • Three vs. four states: The authors note that further investigation is needed to definitively establish whether EF- and LF-gamma are truly distinct or if there are only three main fast gamma states.
  • Future directions: The authors suggest that their method provides a new avenue for investigating oscillatory brain dynamics broadly, enabling more precise studies of how these dynamic states relate to different brain computations and functions. They also highlight the importance of studying cycle-by-cycle transitions to fully understand hippocampal function.

Summary updated

Abstract

Oscillatory brain activity reflects different internal brain states including neurons' excitatory state and synchrony among neurons. However, characterizing these states is complicated by the fact that different oscillations are often coupled, such as gamma oscillations nested in theta in the hippocampus, and changes in coupling are thought to reflect distinct states. Here, we describe a new method to separate single oscillatory cycles into distinct states based on frequency and phase coupling. Using this method, we identified four theta-gamma coupling states in rat hippocampal CA1. These states differed in abundance across behaviors, phase synchrony with other hippocampal subregions, and neural coding properties suggesting that these states are functionally distinct. We captured cycle-to-cycle changes in oscillatory coupling states and found frequent switching between theta-gamma states showing that the hippocampus rapidly shifts between different functional states. This method provides a new approach to investigate oscillatory brain dynamics broadly.
BibTeX
@inproceedings{zhang2019sub,
  author = {Lu Zhang and John Lee and Christopher J. Rozell and Annabelle C. Singer},
  title = {{Sub-second dynamics of theta-gamma coupling in hippocampal CA1}},
  booktitle = {eLife},
  year = {2019},
  doi = {10.7554/elife.44320},
}
Figure