The metabolic cost of maintaining the cortical GABAergic interneuron network György Buzsáki Center for Molecular and Behavioral Neuroscience, Rutgers, The State University of New Jersey, Newark, NJ 07102

Blood oxygen level-dependent (BOLD) functional magnetic resonance imaging (fMRI) is a reliable tool to infer neuronal activity in the intact brain. However , two major issues remain to be clarified. The first is the exact nature of neuronal activity responsible for BOLD variations. The second issue is the relative energy consumption by principal cells vs. interneurons. Inhibitory interneurons are the most economic solution for increasing computational diversity, yet keeping axonal wiring length and energy costs at minimum. Although smaller in number, GABAergic interneurons have several-fold higher firing rates than principal cells. Because the greatest single demand on energy is for pumping Na⁺ out of neurons, the cost of maintaining the interneuron net is considerable. Comparison of firing rates, axonal length, dendritic arbor and the IPSCs they produce, we calculate that maintaining the metric (oscillations) for principal cell computation by the interneuron network is expensive and may require more than a quarter of the metabolic costs of principal cells. Furthermore, a subgroup of interneurons contains the synthesizing enzyme of NO and their axon terminals are often found in the immediate proximity of microvessels. Thus, interneurons not only integrate the activity of a large number of principal cells but may also signal metabolic needs to the local blood vessels. Because interneuronal activity varies with behavior, their context-dependent activity should affect imaging-based measurements (fMRI, PET) of brain signals.