In dominant views, a neuron becomes a functional hub because of its special position within a circuit. In our recently accepted paper on Science Advances, we find experimental evidence supporting a much more democratic view in which almost a majority of recorded single units could serve as hub at least for some time and for a special function.
We analyze hippocampal, enthorinal and prefrontal cortex rat recordings during anesthesia and natural sleep and compute, beyond firing rate, also active information storage and information sharing, i.e. the capacity of a firing patterns to maintain and buffer information through time or to copy the carried information content into the firing pattern of other neurons. We find that there are discrete computing states in which neurons have different involvement in these different primitive computations. (e.g. some neuron serve as a storage hub and some other as a sharing hub). Furthermore, a same neuron may be (or not) a hub in a given computing state but stop (or start) being one in another one. In this view, computational roles of distinct units become flexible and emergent rather than hardwired.
Computing states do not form regular sequences, but stochastic-like switching occur in a way only partially explained by brain state transitons (e.g. REM vs non-REM sleep). However, the specific brain state controls the statistical complexity of computing states sequences, suggesting that state-dependent information processing may occur within different brain states and that different brain states could be associated to (unknown) computations of different combinatorial sophistication.
To know more, check:
- W. Clawson, A.F. Vicente, M. Ferraris, C. Bernard, D. Battaglia* & P. Quilichini* (2019) Computing hubs in the hippocampus and cortex. Science Advances5, eaa4843. [* Shared last authorship].