Hippocampal neural fluctuation between memory encoding and retrieval states during a working memory task in humans

Yusuke Watanabe, Yuji Ikegaya, Takufumi YanagisawabioRxiv 2023 · DOI

TL;DR

This study reveals that the hippocampus exhibits distinct neural activity patterns during working memory encoding and retrieval, with these patterns becoming more pronounced with higher memory loads. Crucially, sharp-wave ripples (SWRs), a key hippocampal oscillation, appear to play a role in shifting these neural states towards retrieval.

What Problem Does This Paper Address?

Working memory (WM) is fundamental to our daily lives, yet its underlying neural mechanisms, particularly the role of the hippocampus, remain incompletely understood. While the hippocampus is known for memory consolidation and retrieval, its specific contribution to dynamic WM processes like encoding and retrieval is less clear. Previous research has shown conflicting results, with some studies suggesting the hippocampus isn't critical for basic WM, while others point to its involvement in specific phases. This paper aims to clarify the hippocampus's dynamic role in WM by investigating how neural activity within it changes across different task phases and how these changes relate to memory load and specific neural events like sharp-wave ripples (SWRs).

Methods

The researchers employed Gaussian-process factor analysis (GPFA), a dimensionality reduction technique, to model the complex multi-unit activity recorded from the medial temporal lobe (MTL) of epilepsy patients. GPFA allows for the visualization of neural activity as trajectories in a low-dimensional space, making it easier to track changes in neural states over time. They also developed methods to detect and analyze sharp-wave ripples (SWRs), transient, high-frequency oscillations in the hippocampus, which are thought to be involved in memory processes. The study focused on analyzing these neural trajectories and SWRs during different phases of a working memory task.

Datasets

The study utilized an open dataset of intracranial electroencephalogram (iEEG) data and multi-unit activity from nine patients with epilepsy. These patients performed a modified Sternberg task, a standard paradigm for studying working memory.

Key Insights and Ideas

  • Hippocampal Trajectories Reflect WM States: The hippocampus doesn't just passively store information; its neural activity dynamically shifts its "state" as it moves between encoding and retrieval phases of a working memory task. These shifts are more pronounced than in other MTL regions like the entorhinal cortex and amygdala.
  • Memory Load Matters: The degree to which the hippocampus transitions between encoding and retrieval states is directly influenced by the amount of information being held in working memory. More information leads to a more pronounced shift.
  • SWRs as State Shifters: Sharp-wave ripples (SWRs), a prominent hippocampal oscillation, appear to be more than just markers of memory replay; they seem to actively facilitate the transition of hippocampal neural states towards retrieval.
  • Hippocampus as a Dynamic Integrator: The findings suggest the hippocampus plays a more active and dynamic role in working memory than previously thought, acting as a crucial integrator of information across different cognitive states.

Key Results

  • The trajectory distance between different phases of the Sternberg task was significantly larger in the hippocampus compared to the entorhinal cortex and amygdala.
  • This trajectory distance between encoding and retrieval phases was dependent on memory load.
  • A transient increase in neural trajectory was observed during SWRs.
  • The direction of hippocampal neural trajectories fluctuated between encoding and retrieval states, with a bias towards the retrieval state during SWR periods.
  • Specifically, the distance from the origin (representing neural state) during mid-SWR periods was greater than during pre-SWR periods, suggesting an active shift.

Limitations and Future Work

  • Limited to Putative CA1 Regions: The analysis was restricted to putative CA1 regions of the hippocampus, which might not capture the full picture of hippocampal involvement.
  • Potential Bias in SWR Detection: The authors acknowledge that the selection of channels might have introduced a bias, potentially exaggerating the increase in trajectory distance from the origin during SWRs.
  • Future Directions: The study provides a foundation for further investigation into the precise mechanisms by which SWRs influence hippocampal states during working memory and how these dynamics contribute to memory recall. Future work could explore these dynamics in more detail across different types of memory tasks and in a broader range of hippocampal subregions.

Summary updated

Abstract

Abstract Background Working memory (WM) is essential for everyday life, yet its neural mechanism remains unclear. Although the hippocampus plays a critical role in memory consolidation and retrieval, its role in WM tasks has yet to be fully elucidated. We hypothesized that multiunit activities in the hippocampus alter their representations depending on the memory load and phase of a WM task, particularly in relation to sharp-wave ripple complexes (SWRs), which are known as various cognitive biomarkers. Methods We used an open dataset of intracranial electroencephalogram (iEEG) data and multiunit activity recorded from the medial temporal lobe (MTL) of nine patients with epilepsy. The MTL includes the hippocampus, entorhinal cortex, and amygdala. During the recording, all subjects performed an eight-second Sternberg test, in which they memorized sets of four, six, or eight letters (encoding phase), waited for three seconds (maintenance phase), and recalled whether a probe letter was included (Match IN task) or not (Mismatch OUT task) (retrieval phase). We used Gaussian-process factor analysis to visualize the neural trajectories of multiunit activity in MTL regions during the task. We also detected SWRs from the iEEG data in the hippocampus. Findings We found that the trajectory distance between phases of the Sternberg task was larger in the hippocampus compared to the entorhinal cortex and amygdala. Additionally, the trajectory distance between the encoding and retrieval phases was memory load dependent. Moreover, a transient trajectory increase was detected during SWRs. Finally, the trajectory direction of the hippocampus fluctuated between the encoding and retrieval states, and the balance of the fluctuation was shifted to the retrieval state during SWR periods. Interpretation Our results demonstrate the involvement of the hippocampus during a WM task. Furthermore, it is suggested that SWR in the retrieval phase plays a role in memory retrieval for a WM task. Our results provide new insight into the two-stage model of memory formation.
BibTeX
@misc{watanabe2023hippocampal,
  author = {Yusuke Watanabe and Yuji Ikegaya and Takufumi Yanagisawa},
  title = {{Hippocampal neural fluctuation between memory encoding and retrieval states during a working memory task in humans}},
  year = {2023},
  doi = {10.1101/2023.04.28.538785},
  howpublished = {bioRxiv},
}
Figure