The mammalian hippocampus, a brain structure important for the initial storage and subsequent recall of memories for specific experiences, spontaneously replays traces of experience during sleep and wakeful rest. Disrupting replay causes memory deficits, but it remains unclear why. Early theories of replay function suggested that replay should recapitulate previous experience as accurately as possible, but recent work has shown that replay content is much more diverse -- including, for instance, spatial paths to a goal, paths not taken, and never-experienced paths. The proposed work investigates what factors determine the content of replay, and what function(s) replay serves. The content of replay is measured under various conditions, and the consequences of disrupting its effects on a target brain region tested. This work will provide new insights into the mystery of how the brain organizes experience to build knowledge structures and then draws upon that knowledge to support adaptive behavior. Accessing replay content, and many other questions in systems neuroscience, rely crucially on statistics and computational models, which are barriers to entry for many otherwise talented young neuroscientists. This project provides a multi-pronged training environment for quantitative skills in neuroscience, with activities ranging from short, visually driven introductory workshops aimed at demystifying terminology, to fully immersive research experiences at summer schools and in the laboratory. The project plans include making all materials, including videos and hands-on tutorials working with real neural data, publicly available online.
Prominent theories of hippocampal replay posit that specific factors such as experience, surprise/uncertainty, and economic value determine what is replayed. However, the effects of manipulating these factors have rarely been tested with controlled, parametric experimental designs. Similarly, the suspicion that replay content should differ depending on whether a task actually requires the hippocampus or not remains untested. The proposed work addresses these issues through the recording and decoding of hippocampal replay content on tightly controlled tasks that manipulate these factors while holding the others constant. Complementing these classic proposals about replay are new ideas derived from reinforcement learning models, which suggest that replay serves to update state and/or action values following errors in the prediction of reward. This idea is tested by optogenetically inhibiting neural activity in the ventral striatum, a structure important for reward learning, specifically during hippocampal replay events, and observing the effects on behavior and on replay content. Supporting these experimental aims are two methodological aims that seek to improve rigor and reproducibility of replay analysis: (1) construction of open data analysis pipelines that follow best practices in software engineering such as unit testing, and (2) development of models that generate synthetic replay events so that the outcomes of different analysis approaches can be tested against a known ground truth. Together, this work provides strong tests of the leading proposals regarding the content and function of hippocampal replay. This project is jointly funded by the Modulation Program of Integrative Organismal Systems in the Directorate for Biological Sciences, the Established Program to Stimulate Competitive Research (EPSCoR), and the Mathematical Biology Program of the Division of Mathematical Sciences in the Directorate for Mathematical and Physical Sciences.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
