Methods and tool development for robust biological voltage imaging

Abstract

Electrical potential differences across membranes play important roles throughout biology, particularly in the brain, where propagating electrical waves called action potentials and smaller subthreshold perturbations carry information within neurons. These electrical signals can be measured at tissue-scale resolution by external electrodes or at a higher resolution by electrodes inserted into the brain. For the highest resolution measurements that avoid many of the problems associated with electrode insertion into the brain, the use of dyes or proteins that transduce the electrical signal into an optical readout is a rapidly maturing technique. While voltage imaging has already enabled significant scientific advances, I address in this dissertation several barriers that have stood in the way of broader use of voltage imaging. First, I worked to overcome the loss of voltage sensitivity under two-photon imaging conditions hitherto seen in microbial rhodopsin voltage indicators. This class of genetically encoded voltage indicators (GEVIs) provides superior response kinetics under one-photon conditions but loses sensitivity under two-photon conditions. I experimentally characterized the photocycle of the Förster resonance energy transfer (FRET)-opsin GEVIs, Voltron1 and Voltron2, and used the results of these experiments to rationally design two-photon imaging conditions that restored voltage sensitivity. I demonstrated this technique for two-photon voltage imaging with Voltron2 in barrel cortex of a live mouse. These results open the door to high-speed two-photon voltage imaging of FRET-opsin GEVIs in vivo and provide insight into the reporters’ photocycle that is useful both for robust one-photon imaging and for future development of two-photon-optimized rhodopsin-based GEVIs. Second, I bring together experimental and theoretical work to provide a set of well-characterized upper bounds to two-photon voltage imaging performance in vivo. While aspects of this have previously been addressed, there has not been a single work addressing these various limits with a specific focus on two-photon voltage imaging. As voltage imaging operates under different constraints than other more familiar types of functional imaging, this treatment is necessary to set realistic expectations, delineate the most productive avenues for optimization, and provide a common theoretical groundwork for comparing voltage imaging performance. Among other conclusions, we found that current technologies are limited to high quality imaging of neurons under standard conditions at depths greater than 300μm in vivo. Third, I present Luminos, an open-source MATLAB-based software package for highly synchronized control of high-speed microscopes. The distinct constraints of voltage imaging place distinct constraints on the required instrumentation. Existing general-purpose control libraries are inadequate for the speed and synchronization required for voltage imaging experiments involving complex electrical and optical stimulation and recording. Rather than produce control code narrowly-tailored to a specific microscope or experiment, our lab embarked on a project to develop a modular customizable control suite that is now used for data acquisition on all of the seven custom microscopes in our lab and has been publicly released with the aim of making voltage imaging instrumentation more accessible to the broader field. Fourth, I performed a theoretical and experimental treatment of the calibration of intensitybased voltage indicators to an absolute voltage scale. Because of unknown expression levels, background, optical efficiency, and other factors, intensity-based imaging of a single reporter provides only relative signals that cannot be calibrated either to an external scale or across significant spatial or temporal extent. I mathematically and experimentally analyzed the feasibility of calibrating voltage responses with a second independently expressed voltage indicator and found that a combination of a linear indicator with a nonlinear indicator tailored to the voltage range of interest can provide a resilient calibration to an absolute scale. Likely due to a publication bias towards reporting more linear indicators, high-performance nonlinear indicators are not currently available, but this theoretical treatment motivates their development and dissemination. Finally, I discuss the technical outlook for voltage imaging in light of this work. By characterizing the performance limits of two-photon voltage imaging, uncovering the mechanism of two-photon voltage insensitivity in opsin-GEVIs, and proposing a method for dual indicator absolute voltage imaging, I provide insight into the most productive directions for future voltage indicator development. Both the demonstration of two-photon voltage imaging with Voltron2 and the release of the Luminos software are important steps towards making voltage imaging more broadly useful and accessible. Important work remains in further optimizing opsin-GEVIs for two-photon imaging and in making voltage imaging hardware more accessible.