Boundary Vector Cells Encode a Future-Biased Spectrum of Positions in the Rat

Abstract

Spatial tuning is a hallmark property of neural firing in the hippocampal formation. Yet, that tuning is often less well correlated with the instantaneous current position of an animal than it is with an integrated version of the past or future state of the animal. Whether that encoding is biased towards past or future states and the extent to which it shows fixed versus multi-scale encoding varies across circuits and cell types. The temporal encoding properties of boundary vector cells of the subiculum are not well established. To address this here, we re-analyzed recordings of BVCs described previously by Lever et al. (2009) with multiple approaches. In the first, we asked if adding a temporal offset between the rat position and the spiking of a BVC increased the apparent spatial tuning in the firing rate map. We found that aligning BVC spiking with future states maximized the rate map spatial tuning. These results were mirrored in a second analysis that, instead of optimizing rate map spatial tuning, optimized how well the firing rate map predicted the BVC spiking. The second analysis also allowed us to ask whether that encoding is focused on a particular temporal horizon or whether the encoding captures behavior at multiple scales. To this end, for a given recording, we asked "How much time-integration of the behavioral state is the observed spiking most consistent with?" We observed a wide spectrum of time-constants of integration across cells, indicating that BVCs form a multiscale encoding of future states. The distribution of both offsets and integration rates observed across BVCs did not differ significantly from other, non-BVC, subiculum neurons. Taken together, these findings indicate that BVCs, along with other subiculum neurons, form a multi-scale encoding of future states.