Behavior-dependent spatial maps enable efficient theta phase coding

Abstract

Navigation through space involves learning and representing relationships between past, current and future locations. In mammals, this might rely on the hippocampal theta phase code, where in each cycle of the theta oscillation, spatial representations start behind the animals location and then sweep forward. However, the exact relationship between phase and represented and true positions remains unclear. Developing a quantitative framework for the theta phase code, we formalize two previous notions: in spatial sweeps, different phases of theta encode positions at fixed distances behind or ahead of the animal, whereas in temporal sweeps, they encode positions reached at fixed time intervals into the past or future. These two schemes predict very different position representations during theta depending on the animals running speed. Paradoxically, in some studies sweep length has been shown to increase proportionally with running speed, consistent with temporal sweeps, whereas in other studies individual place field parameters such as field size and phase precession slope were shown to remain constant with speed, consistent with spatial sweeps. Here, we introduce a third option: behavior-dependent sweeps, according to which sweep length and place field properties vary across the environment depending on the running speed characteristic of each location. Analyzing single-cell and population variables in parallel in recordings from rat CA1 place cells and comparing them to model simulations, we show that behavior-dependent sweeps uniquely account for all relevant variables. This coding scheme combines features and advantages of both spatial and temporal sweeps, revealing an efficient hippocampal code. SignificanceTo learn the structure of the world and the consequences of our actions, information about the past must be carried through to the present and linked to what is currently happening. To plan, desired future states and the predicted outcomes of actions must be represented. In mammals, including humans, hippocampal neurons are thought to encode such representations of past, present and future states at different phases of the theta oscillation. However, the precise hippocampal phase code remains unknown. We show that two previous ideas are incompatible with each other and with rat experimental data. So, we propose a new coding scheme that synthesizes features from both ideas and accounts for all relevant observations.