How does hippocampal activity represent the external world in a manner that reflects changing behavior goals?

While most often studied for their reliable spatial tuning, neurons within the hippocampus respond to other non-spatial features, especially when these cues are required to solve an ongoing task. However, it is unclear whether spatial and task-relevant cognitive tuning co-exist, compete, or cooperate within the hippocampus. Moreover, if hippocampal firing provides the framework for the formation of episodic memories, such a balance between spatial and cognitive tuning likely continuously fluctuates as a result of task demand.

I designed a task where the need for representing continuous change in space or other task variables (acoustic cue) was purposefully juxtaposed.

Acoustic Cue-guided Navigation Task

Mice were trained to run on a linear track with 7 equally spaced ports. During non-auditory trials, mice received a water reward at either end of the track. When auditory cues were relevant, a tone ascended at random speeds. The changes in frequency were driven by the spatial position of the mouse (closed loop). Frequencies were thus changing with movement but were decoupled from space from trial to trial. Mice used auditory tones to identify when the target frequency (22 kHz) was reached and licked at the closest water port.

Example trials where the mouse uses the ascending auditory frequency to guide it to the correct reward ports. Notice head sweeps adjacent to ports as the mouse is running down the track, deliberating on whether the target frequency has been reached.

Hippocampal cells represented spatial coordinates as well as task-relevant actions (such as approach to ports, or path length), but not purely sensory tuning

Trajectories (gray) segregated by trial type (choice of the mouse). Spikes from a single cell overlaid in red with average rates for each trial type shown in shades of blue. Top, rate aligned to position, Bottom, rate aligned to the choice port (dashed line).

When auditory cues were relevant, many hippocampal neurons fired in a non-spatial manner. Rather than responding to the tones, these cells reflected action plans (action cell). These were not trivial motor signals because the firing rates of these cells varied according to the uncertainty of outcomes (notice rate modulation by trial type in the rates aligned to the choice) and reflected moments of deliberation at earlier ports. Despite spatial cues being non-informative, spatial tuning persisted (spatial cell). Using a statistical model, I found that action and spatial tuning fell on a continuum – several cells were modulated by both variables.

Hippocampal spatial and action tuning combined to generate distinct subsets of active cells for each trial type. At the population level, this resulted in strikingly unique representations for distinct trial types (which port the mouse chose).

Hippocampal activity on the auditory task visualized as a manifold. Dots are 100 ms time bins colored by trial type. The manifold branches out (gray arrows) before the end of trials, distinguishing between trials.

I propose that the hippocampus combines “where I am” and “what I am doing” signals to distinguish between individual trials – providing a framework for hippocampal activity representing flexible behavior goals.

My future research work seeks to identify progressive stages of this sensory cue to cognitive variable transformation.

This project is being performed in collaboration with Athina Apostolelli (Master's student, Buzsaki lab), Winnie Yang (Graduate student, Buzsaki lab), Edoardo Balzani (Flatiron Institute) and Cristina Savin (Center for Neural Science, NYU)