Mechanisms of cognitive flexibility in primate fronto-striatal circuits

Abstract

Adapting behavior to changing rewards entails momentary adjustment of single neuron firing and synchronization patterns across the primate fronto-striatal network. But how changes in neural activity realize behavioral adaptation has remained elusive. This thesis outlines a series of studies that aim to elucidate which task variables trigger behavioral changes, which brain areas in the fronto-striatal network are essential for adapting behavior, and how separable neuron types contribute to flexible learning. The thesis shows that putative inhibitory interneurons, particularly fast spiking interneurons (FSIs), play important roles in encoding learning variables. In comparison to other neuron types, FSI activity correlated stronger with (i) in the striatum how well behavior is adjusted during learning, (ii) in the anterior cingulate cortex (ACC) required demands for behavioral adjustment based on mismatching reward expectations and experienced outcomes, and (iii) in the lateral prefrontal cortex (LPFC) how uncertain a subject is about the most rewarding choices between different choice options. In the ACC and LPFC, the learning-specific activity of interneurons was linked to gamma-synchronous network activity. The thesis uses these findings of interneuron-specific learning correlates and proposes three area-specific circuit motifs. These circuit motifs explain three testable mechanistic roles of interneurons, in the anterior cingulate cortex to reduce outcome uncertainty, in the lateral prefrontal cortex to reduce choice uncertainty, and in the striatum to gate learning relevant information. The thesis then extends existing behavioral paradigms for studying how subjects adapt to behavioral change and illustrates that the expectation of gains versus losses differentially modulates learning efficiency. This finding suggests a major role of motivation to bias adaptive behavior. The thesis shows that the motivational component is causally supported by the ACC, revealed by showing that flexible learning from losses is disrupted by transcranial ultrasound stimulation (TUS) in the ACC. Together, these findings demonstrate how flexible learning is dependent on unique neuron types (FSIs in the fronto-striatal network), network nodes (anterior cingulate cortex), and behavioral task demands (dependence of motivational demands). These results advance our neurophysiological and conceptual understanding of flexible learning mechanisms in the fronto-striatal network and illustrate the versatility of a cross-level approach in neuroscience.