Role of Feed-Forward Inhibition in Neocortical Information Processing: Implications for Neurological Disorders

Abstract

A major well-documented feature of cortical functional organization is the presence of prominent broadly tuned feed-forward inhibition in the input layer 4, in which local layer 4 inhibitory cells receive direct thalamocortical input and in turn suppress responses of neighboring layer 4 excitatory cells to their thalamocortical drive, thereby sharpening their receptive field properties. Here we review the evidence that the presence of broadly tuned feed-forward inhibition in layer 4 turns local layer 4 domains into functional analogs of Radial Basis Function networks, enabling layer 4 to contribute importantly to sensory information processing as a pluripotent function linearizer: i.e., it performs such a transform of afferent inputs to a cortical column that makes possible for neurons in the upper layers of the column to learn and perform their complex functions using primarily linear operations. Feed-forward inhibition is subserved by fast-acting basket cells and slow-acting neurogliaform cells, which rely on GABAA and GABAB receptor-mediated inhibition, respectively. Their respective contributions can be observed by measuring tactile stimulus detection threshold using step vs. ramp vibrotactile stimuli. The static (step) threshold reflects basket-mediated inhibition, whereas the difference between the static and dynamic (ramp) thresholds reflects neurogliaform-mediated inhibition. Our feed-forward inhibition metric, which is based on the static and dynamic detection thresholds, can provide significant insight about the neurological health of the cortical circuitry, given that neurogliaform cells are engaged in on-demand energy homeostasis of cortical networks through their local release of insulin. For example, we have found this metric to be below normal in adolescents with autism spectrum disorder, but highly elevated in type 2 diabetes. Maladaptive feed-forward inhibition can have significant downstream implications for cortical information processing, and our metric can potentially be an effective means for evaluating a number of cortical abnormalities.