TRANSIENT CA1 PLACE CELL REMAPPING BETWEEN TWO VISUALLY IDENTICAL ENVIRONMENTS

M.C. Fuhs ¹*; S.R. VanRhoads ²; A.E. Casale ²; B.L. McNaughton ²; D.S. Touretzky ¹

1. Computer Science Dept., Carnegie Mellon Univ, Pittsburgh, PA, USA
2. NSMA, Univ Arizona, Tucson, AZ, USA

Rats foraged in two identical, connected boxes with 180^o opposite orientations (VanRhoads et al., SFN Abs., 2004). In each session, boxes were visited in an A-B-A-B sequence; each visit lasted 4.5 min. Box A place fields were stable across visits. Rat 1 remapped in box B in the first visit. Rat 2 remapped in the second visit. Rat 3 remapped in the third visit (second session). A Poisson hidden Markov model (P-HMM) reconstruction gives a finer grain view. For each rat’s first session in which place field correlations indicated remapping, the first box A and B visits were divided into 250 ms bins and a vector of spike counts constructed for each. The second visits of the session served as a “reference” for their respective maps. The sequence of roughly 2000 vectors from the first visits were used to infer the most likely path through a three-state P-HMM. States “Box A” and “Box B” were expected to emit Poison-distributed spike count vectors whose means were the location-specific mean firing rates measured during the reference visits. A third state reflected place cell inactivity. The reconstruction confirmed that rats 1 and 3 remapped in box B during their first and third visits, respectively. Though cells were less active early in the visit, the reconstruction is consistent with the interpretation that remapping occurred upon entry into box B. Rat 2 did not appear to remap until the second box B visit, but reconstruction of the first visit uncovered activity more similar to the box B map during the first 27 sec; then the rat switched back to the box A map for the remainder of the visit. This suggests that remapping may initially be transient and unstable but will stabilize via learning mechanisms over time. The box B map may actually be an extension of the box A map into the box B space; self-localization then reset the hippocampus back onto the box A portion of the map, perhaps after adjusting the head direction estimate to be consistent with the rotated cues in box. B.

Support Contributed By: MH059932

Key words: