2022 Theses Doctoral
Dynamic and compressed memory coding in the hippocampus
A longstanding goal in neuroscience is to provide a biological understanding of episodic memory, our conscious recollection of prior experience. While the hippocampus is thought to be a critical locus for episodic learning in the mammalian brain, the nature of its involvement is unsettled. This thesis details several investigations that attempt to probe the neural mechanisms that support the encoding and organization of new experiences into memory.
Throughout the included works, we utilize in vivo two-photon fluorescence microscopy and calcium imaging to study the functional dynamics of hippocampal networks in mice during memory-guided behavior. To begin, Chapter 2 examines how neural coding in hippocampal area CA1 is modified during trace fear conditioning, a common model of episodic learning in rodents that requires linking events separated in time. We longitudinally tracked network activity throughout the entire learning process, analyzing how changes in hippocampal representations paralleled behavioral expression of conditioned fear. Our data indicated that, contrary to contemporary theories, the hippocampus does not generate sequences of persistent activity to learn the temporal association. Instead, neural firing rates were reorganized by learning to encode the relevant stimuli in a temporally variable manner, which could reflect a fundamentally different mode of information transmission and learning across longer time intervals.
The remaining chapters concern place cells---neurons identified in the hippocampus that are active only in specific locations of an animals' environment. In Chapter 3, we use mouse virtual reality to explore how the hippocampus constructs representations of novel environments. Through multiple lines of analysis, we identify signatures of place cells that acquire spatial tuning through a particularly rapid form of synaptic plasticity. These motifs were enriched specifically during novel exploration, suggesting that the hippocampus can dynamical tune its learning rate to rapidly encode memories of new experiences. Finally, Chapter 4 expands a model of hippocampal computation that explains the emergence of place cells through a more general mechanism of efficient memory coding. In theory and experiment, we identified properties of place cells that systematically varied with the compressibility of sensory information in the environment. Our preliminary data suggests that hippocampal coding adapts to the statistics of experience to compress correlated patterns, a computation generically useful for memory and which naturally extends to representation of variables beyond physical space.
This item is currently under embargo. It will be available starting 2024-08-04.
More About This Work
- Academic Units
- Neurobiology and Behavior
- Thesis Advisors
- Losonczy, Attila
- Fusi, Stefano
- Ph.D., Columbia University
- Published Here
- August 10, 2022