Engineering Novel Rhodopsins for Neuroscience

Citation

Bedbrook, Claire Nicole (2018) Engineering Novel Rhodopsins for Neuroscience. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/7RA0-BC29. https://resolver.caltech.edu/CaltechTHESIS:05302018-095950094

Abstract

The overarching goal of my PhD research has been engineering proteins capable of controlling and reading out neural activity to advance neuroscience research. I engineered light-gated microbial rhodopsins, primarily focusing on the algal derived, light-gated channel, channelrhodopsin (ChR), which can be used to modulate neuronal activity with light. This work has required overcoming three major challenges. First, rhodopsins are trans-membrane proteins, which are inherently difficult to engineer because the sequence and structural determinants of membrane protein expression and plasma membrane localization are highly constrained and poorly understood (Chapter 3-5). Second, protein properties of interest for neuroscience applications are assayed using very low throughput patch-clamp electrophysiology preventing the use of high-throughput assays required for directed evolution experiments (Chapter 2, 5-6). And third, in vivo application of these improved tools require either retention or optimization of multiple protein properties in a single protein tool; for example, we must optimize expression and localization of these algal membrane proteins in mammalian cells while at the same time optimizing kinetic and functional properties (Chapter 5-6). These challenges restricted the field to low-throughput, conservative methods for discovery of improved ChRs, e.g., structure-guided mutagenesis and testing of natural ChR variants. I used an alternative approach: data-driven machine learning to model the fitness landscape of ChRs for different properties of interest and applying these models to select ChR sequences with optimal combinations of properties (Chapters 5-6). ChR variants identified from this work have unprecedented conductance properties and light sensitivity that could enable non-invasive activation of populations of cells throughout the nervous system. These ChRs have the potential to change how optogenetics experiments are done. This work is a convincing demonstration of the power of machine learning guided protein engineering for a class of proteins that present multiple engineering challenges. A component of the novel application of these new ChR tools relies on recent advances in gene delivery throughout the nervous system facilitated by engineered AAVs (Chapter 7). And finally, I developed a behavioral tracking system to monitor behavior and demonstrate sleep behavior in the jellyfish Cassiopea, the most primitive organism to have this behavior formally characterized (Chapter 8).

Thesis Files