Traveling UP states in the post-subiculum reveal an anatomical gradient of intrinsic properties

Cortical activity is characterized by state-specific dynamics arising from the interplay between connectivity, cellular diversity, and intrinsic properties. During non-Rapid Eye Movement (NREM) sleep, cortical population activity alternates between periods of neuronal firing (“UP” states) and neuronal silence (“DOWN” states). Patterns of neuronal activity at DOWN-to-UP (DU) transitions have functional relevance beyond sleep: they are related to sensory coding during wakefulness and support homeostatic processes and memory consolidation. Despite this functional importance, the factors that organize these spiking patterns remain unknown but mechanisms that rely on network connectivity or intrinsic excitability have been proposed.

In order to elucidate the mechanisms that organize spontaneous activity, we recorded populations of neurons in the head-direction cortex (HDC, i.e., post-subiculum), where the behavioral correlates of most neurons are well accounted for. Neuronal tuning to HD was independent of anatomical position. However, while UP-DOWN (UD) transitions were synchronous along the dorsoventral (DV) axis, we observed sequential activation of neurons at DU transitions. To understand the mechanisms underlying these traveling waves at UP state onset, we built a computational model with a linear array of recurrently connected adapting units and compared the effects of different biophysical gradients. We found that, unlike gradients in local connectivity, excitability/input, and adaptive current, a gradient in rectifying current (Ih) was able to uniquely reproduce the experimental observations, and predict a yet-unobserved relationship between UP onset and post-DOWN rebound activity. Subsequent ex vivo intracellular recordings confirmed the predicted DV gradient of Ih in HDC.

In conclusion, precisely organized spontaneous population activity patterns may be independent of circuit features and sensory coding but instead may only reflect intrinsic neuronal properties. Yet, the resulting traveling waves have the potential to anatomically segment computation in output structures like the medial entorhinal cortex (MEC) and indirectly, the hippocampus.