Changes in patterns of activity within the medial prefrontal cortex enable rodents, non-human primates, and humans to update their behavior to adapt to changes in the environment, e.g., during cognitive tasks. Within medial prefrontal cortex, inhibitory neurons expressing parvalbumin are important for updating strategies in a rule shift task. Nevertheless, causal mechanisms through which parvalbumin neurons recruit specific circuits to produce prefrontal network dynamics that switch from maintaining to updating task-related patterns of activity remain unknown. Here we identify a new long-range inhibitory projection from prefrontal parvalbumin neurons to the contralateral cortex that mediates this process. Whereas nonspecifically inhibiting all callosal projections does not prevent mice from learning rule shifts, selectively inhibiting these callosal parvalbumin projections disrupts rule shift learning, desynchronizes gamma-frequency activity that is necessary for learning, and suppresses changes in prefrontal activity patterns that normally accompany rule shifts. Thus, callosal parvalbumin projections switch prefrontal circuits from maintaining to updating behavioral strategies by synchronizing callosal communication and preventing it from inappropriately maintaining outdated neural representations. These findings may explain how deficits in prefrontal parvalbumin neurons and gamma synchrony cause impaired behavioral flexibility in schizophrenia, and identify long-range projections from prefrontal parvalbumin neurons as a key circuit locus for understanding and treating these deficits.

Changes in patterns of activity within the medial prefrontal cortex enable rodents, non-human primates and humans to update their behaviour to adapt to changes in the environment—for example, during cognitive tasks1–5. Parvalbumin-expressing inhibitory neurons in the medial prefrontal cortex are important for learning new strategies during a rule-shift task6–8, but the circuit interactions that switch prefrontal network dynamics from maintaining to updating task-related patterns of activity remain unknown. Here we describe a mechanism that links parvalbumin-expressing neurons, a new callosal inhibitory connection, and changes in task representations. Whereas nonspecifically inhibiting all callosal projections does not prevent mice from learning rule shifts or disrupt the evolution of activity patterns, selectively inhibiting only callosal projections of parvalbumin-expressing neurons impairs rule-shift learning, desynchronizes the gamma-frequency activity that is necessary for learning8 and suppresses the reorganization of prefrontal activity patterns that normally accompanies rule-shift learning. This dissociation reveals how callosal parvalbumin-expressing projections switch the operating mode of prefrontal circuits from maintenance to updating by transmitting gamma synchrony and gating the ability of other callosal inputs to maintain previously established neural representations. Thus, callosal projections originating from parvalbumin-expressing neurons represent a key circuit locus for understanding and correcting the deficits in behavioural flexibility and gamma synchrony that have been implicated in schizophrenia and related conditions9,10.

Neural synchronization at gamma (∼40 Hz) frequencies is believed to contribute to brain function and be deficient in disorders including Alzheimer’s disease and schizophrenia. Gamma-frequency sensory stimulation has been proposed as a non-invasive treatment for deficient gamma synchrony and associated cognitive deficits, and has been shown to be effective in mouse models of Alzheimer’s disease. However, both the mechanism and applicability of this approach remain unclear. Here we tested this approach using mutant (Dlx5/6+/-) mice which have deficits in gamma synchrony and the ability to learn to shift between rules which use different types of cues to indicate reward locations. 40 Hz auditory stimulation rescues rule shifting in Dlx5/6+/- mice. However, this improvement does not outlast the period of stimulation, and is not associated with normalized gamma synchrony, measured using genetically encoded voltage indicators. These results show how gamma-frequency sensory stimulation may improve behavior without fully restoring normal circuit function.

No abstract