Unexpected Rule-Changes in a Working Memory Task Shape the Firing of Histologically Identified Delay-Tuned Neurons in the Prefrontal Cortex.

Working memory-guided behaviors require memory retention during delay periods, when subsets of prefrontal neurons have been reported to exhibit persistently elevated firing. What happens to delay activity when information stored in working memory is no longer relevant for guiding behavior? In this study, we perform juxtacellular recording and labeling of delay-tuned (-elevated or -suppressed) neurons in the prelimbic cortex of freely moving rats, performing a familiar delayed cue-matching-to-place task. Unexpectedly, novel task-rules are introduced, rendering information held in working memory irrelevant. Following successful strategy switching within one session, delay-tuned neurons are filled with neurobiotin for histological analysis. Delay-elevated neurons include pyramidal cells with large heterogeneity of soma-dendritic distribution, molecular expression profiles, and task-relevant activity. Rule change induces heterogenous adjustments on individual neurons and ensembles' activity but cumulates in balanced firing rate reorganizations across cortical layers. Our results demonstrate divergent cellular and network dynamics when an abrupt change in task rules interferes with working memory.

The Timing of Sensory-Guided Behavioral Response is Represented in the Mouse Primary Somatosensory Cortex.

Whisker-guided decision making in mice is thought to critically depend on information processing occurring in the primary somatosensory cortex. However, it is not clear if neuronal activity in this "early" sensory region contains information about the timing and speed of motor response. To address this question we designed a new task in which freely moving mice learned to associate a whisker stimulus to reward delivery. The task was tailored in such a way that a wide range of delays between whisker stimulation and reward collection were observed due to differences of motivation and perception. After training, mice were anesthetized and neuronal responses evoked by stimulating trained and untrained whiskers were recorded across several cortical columns of barrel cortex. We found a strong correlation between the delay of the mouse behavioral response and the timing of multiunit activity evoked by the trained whisker, outside its principal cortical column, in layers 4 and 5A but not in layer 2/3. Circuit mapping ex vivo revealed this effect was associated with a weakening of layer 4 to layer 2/3 projection. We conclude that the processes controlling the propagation of key sensory inputs to naive cortical columns and the timing of sensory-guided action are linked.

Divisions of Identified Parvalbumin-Expressing Basket Cells during Working Memory-Guided Decision Making.

Parvalbumin-expressing basket cells tightly control cortical networks and fire remarkably stereotyped during network oscillations and simple behaviors. How can these cells support multifaceted situations with different behavioral options and complex temporal sequences? We recorded from identified parvalbumin-expressing basket cells in prefrontal cortex of freely moving rats performing a multidimensional delayed cue-matching-to-place task, juxtacellularly filled recorded neurons for unequivocal histological identification, and determined their activity during temporally structured task episodes, associative working-memory, and stimulus-guided choice behavior. We show that parvalbumin-expressing basket cells do not fire homogenously, but individual cells were recruited or inhibited during different task episodes. Firing of individual basket cells was correlated with amount of presynaptic VIP (vasoactive intestinal polypeptide)-expressing GABAergic input. Together with subsets of pyramidal neurons, activity of basket cells differentiated for sequential actions and stimulus-guided choice behavior. Thus, interneurons of the same cell type can be recruited into different neuronal ensembles with distinct firing patterns to support multi-layered cognitive computations.