BPS2026 – Engineering far-red genetically encoded voltage indicators for in vivo imaging

Recording electrical activity from ensembles of neurons is a fundamental challenge to decode brain function. Voltage imaging using genetically encoded fluorescent voltage indicators (GEVIs) has emerged as a powerful approach for detecting neuronal activity with high spatial and temporal resolution. Existing GEVIs predominantly utilize green/yellow light excitation, which compromises performance in tissue due to high light scattering, autofluorescence, and absorption in this spectral range. Here, we introduce a GEVI employing a novel voltage-sensing mechanism that modulates fluorescence of bright far-red dyes like the rhodamine dye JF669, enabling deep-tissue imaging with enhanced signal-to-noise ratio. To evolve this GEVI, we developed an assay to measure fluorescence response to elicited action potentials of newly designed proteins in primary neuron cultures using simultaneous field stimulation and rapid fluorescence imaging. We then improved their sensitivity and brightness focusing on their performance with the far-red dye JF669. To do this, we integrated rational protein design approach with machine learning-guided protein engineering. We trained predictive models to map protein sequence to functional response and then applied these predictors in iterative cycles of machine-learning prediction and wet-lab testing. This strategy yielded GEVI variants with enhanced voltage sensitivity and brightness. Experimental validation in cultured neurons demonstrates that the GEVI is capable of reliably detecting action potentials as well as subthreshold fluctuations, achieving a high signal-to-noise ratio. In addition, we show that our new GEVI can report membrane potential with a pallet of different color rhodamine dyes: JF525, JF552, JF585, JF608, JF646, and JF669. This work establishes a new class of GEVIs that enable voltage imaging at tissue depths previously inaccessible with existing indicators, offering powerful new opportunities for dissecting neural activity in vivo.