Parvalbumin-Expressing Interneurons Implement Dynamic Gain Control in Cortical Networks as Assessed through All-Optical Interrogation
Inhibition in cortical networks is administered by a diverse array of specialized cell types with distinct intrinsic properties and connectivity patterns. This stark heterogeneity in cellular form and function suggests that different inhibitory subtypes may actually implement distinct computations that rely on their specific morphologies and functional positions within the network. For example, inhibition has been suggested to regulate the global level of input responsiveness (sensitivity/gain), sharpen responses to specific inputs (selectivity), and carry out suppressive interactions between those responses (competition), all of which might be mediated by distinct cellular subclasses. To causally explore this possibility, we optogenetically targeted parvalbumin-expressing interneurons in mouse visual cortex, and combined high speed (50 Hz) two-photon imaging, in vivo electrophysiology, and a novel form of single cell optical stimulation in order to isolate the role of Pv inhibition during the well established cortical computations of the visual system. Specifically, we injected the adenoassociated viral vector double-floxed inverted open reading frame ChR2-mCherry (AAV DIO ChR2-mCherry), with Cre-dependent expression of ChR2, into PV-Cre knock-in mice. We observed robust light-activated responses in mCherry-labeled fast-spiking neurons (>100 Hz) in both brain slices and visually-engaged cortex, and found that optically driving these neurons effected functional shutdown in neighboring pyramidal cells in vitro and in vivo. By targeting this shutdown to specifically timed intervals within the receipt of visual stimuli by the cortical network, we could explore the functional impact of Pv inhibition on the resolution of orientation and contrast responses. For orientation responses measured through cell-attached recordings in vivo, we observed a Pv-driven suppression of visual responses in proportion to the magnitude of those responses, suggesting that Pv neurons implement a "division" of responsiveness equivalent to a network shunting effect. This result was confirmed using high-speed two-photon imaging in conjunction with the optical stimulation. Similarly, when activating visual cortical neurons with increasing levels of network drive, by presenting optimal stimuli of increasing contrast, we found that Pv activation had much stronger attenuation of the higher contrasts (stronger network drive) compared to the lower ones (weaker drive), again suggesting a normalization of network state via divisive gain control. Finally, we present a novel methodology for optically activating localized neurons in vivo at many locations on a rapid timescale, and combine this methodology with simultaneous high-speed two-photon imaging to enable single cells to be toggled during concurrent visualization of distributed network responses. Taken together our results strongly implicate parvalbumin-containing inhibitory interneurons as arbiters of dynamic gain control in the cortex.
Society for Neuroscience Abstract, 2011.