UC Irvine

The ability to temporally organize our memories is critical to daily life function and impaired in several cognitive disorders. Considerable research shows that this fundamental capacity is conserved across modalities and species, and that it critically depends on the hippocampus. However, it remains unclear how the hippocampus supports this capacity. Accumulating evidence from rodent electrophysiology points to two main coding properties as potential neural mechanisms. First, ensemble activity exhibits sequential firing fields (or “time cell” activity) during individual stimuli or intervals. Although this form of temporal coding has been shown to provide significant temporal information during such task events, a direct link with the ability to remember sequences of events has yet to be established. Second, hippocampus ensemble activity has also been shown to represent sequences of spatial locations, which can be reactivated during “offline” moments such as during sharp-wave ripple events. However, it remains to be determined whether this form of sequence coding extends to nonspatial information and thus reflects a fundamental process supporting the memory for event sequences. To address these critical issues, I analyzed existing datasets in which CA1 neural activity was recorded in rats performing a nonspatial sequence memory task, a task previously shown to have strong behavioral parallels in rats and humans and to engage comparable circuits across species. The central aim of this thesis was to investigate temporal and sequence coding properties in hippocampus and determine their respective contributions to order memory judgments. To do so, I took advantage of this rich behavioral paradigm and applied advanced statistical tools designed to quantify these coding properties during specific task events and intervals in between them. The background and significance of my research are described in Chapter 1, my specific research aims (outlined below) are described in Chapters 2-4, and a general discussion of our findings and their implications is included in Chapter 5. Aim 1: Determine if CA1 ensemble activity provides a temporal signal that carries event specific information and bridges across event sequences (Chapter 2). Specifically, I examined whether the temporal coding is stimulus-specific and important for correct order decisions, captures sequential relationships among stimuli through a lag effect, and bridges across sequences of stimuli. To do so, I applied a Bayesian model to quantify the temporal information contained in the hippocampal ensemble activity during discrete task events as well as across sequences of events. Aim 2: Build statistical framework to quantify sequence replay of discrete nonspatial states during offline periods of our sequence memory task (Chapter 3). I developed a statistical framework to overcome the shortcomings of earlier models which are primarily designed to quantify the replay of continuous spatial states. Specifically, my objective was to build a framework which can detect the replay of odor sequences and determine if their content and direction conform to our predictions. To do so, I adapted the "sequenceness metric" from a human MEG study and expanded it to identify the precise order in which odors are replayed within replay moments and make it suitable to the electrophysiological data from our sequence memory experiment. Aim 3: Determine if nonspatial event sequences are reactivated during hippocampal replay moments and if the direction and content of the replay reflects trial-specific demands and behavior (Chapter 4). I investigated whether replay mainly reflects sequences of discrete events reflecting task critical associations and if its content and direction vary based on the interval type (temporal location in the sequence). Specifically, I tested whether replay is primarily in forward direction, involving upcoming odor sequences during rest within sequence presentation, and in reverse direction, involving odor sequences leading to sequence completion during rest at the sequence's end. To do so, I utilized the framework from chapter 3.