中央研究院 Logo

Integrating recent information is critical for guiding behavior and predictions. This involves tracking variables such as distance traveled or time elapsed, all of which require a starting point. By establishing the starting point, the brain can segment the continuous flow of experiences into discrete units and integrate information within each unit. Yet, the neuronal mechanisms that underlie both the initiation and execution of integration remain largely unknown. Using virtual reality setups, we developed behavioral paradigms where a mouse is required to integrate distance or time. Our study revealed two functionally distinct subpopulations of hippocampal CA1 pyramidal neurons -- PyrDown and PyrUp neurons. These subpopulations exhibited contrasting responses at the start of integration: PyrDown neurons shut down their activity at the start and then slowly ramped up towards the reward, while PyrUp neurons increased their activity first and then gradually declined. We also found that these two subpopulations are modulated by different types of inhibitory interneurons. Parvalbumin inhibitory interneurons preferentially mediate the initial shutdown of PyrDown neurons and are necessary for initiating information integration, while somatostatin interneurons primarily regulate the activity of PyrUp neurons and are required for accurate integration. In summary, we uncover potential circuit motifs that support different stages of the information integration process.