Our protein engineering team has synthesized the world’s first channelrhodopsins that respond robustly to ambient indoor lighting. This breakthrough resulted from an AI/machine learning directed campaign of the vast number of possible protein sequences. We’ve released a comprehensive paper and a video explainer detailing its discovery and characterization.
Optogenetics, first introduced twenty years ago, has changed how neuroscience is done: it allows scientists to ask what happens when you make specific sets of neurons fire, rather than just observing what their activity is correlated with. In the field of optogenetics, opsins are the proteins responsible for converting light into chemical signals (or, in our case, electrochemical signals). Opsins are found across the animal and microbial kingdoms, from human retinas to single-cell algae. In neuroscience applications, different types of opsins are genetically expressed in specific types of neurons or other excitable cells, where they function as a sort of switch, allowing the cells to respond to light. Different types of opsins can either excite or inhibit a cell in response to light.
Early optogenetic actuators were relatively insensitive to light, requiring bright lasers or LEDs to work. The sensitivity of opsins has been steadily improving (PDF) for years, but researchers haven’t yet seen reliable and robust action potentials to ambient light levels found indoors.
Our family of opsins, called WAChRs, allows neurons to be reliably activated with indoor office lights. To test this, we devised the simplest experiment we could think of—we patched a cell in the dark, opened the curtain, and measured the response to the room lights in what became known as the “curtain test”. When the curtain was opened, the neuron activated, and when the curtain was closed, it stopped.
We’re excited for the new types of experiments that low-light optogenetics can enable. WAChRs represent not only a possible pathway to treatments for vision restoration and other conditions, but for neuroscience applications in the field of optogenetics more broadly, where researchers use these tools to study the neural circuits underpinning Parkinson’s disease, epilepsy, depression, and other psychiatric disorders.
To make it, we first converted an inhibitory channel into an excitatory channel. We then identified mutations that made it more powerful, and changed the speed of its response. Response time and sensitivity are typically inversely related such that an improvement in one requires a tradeoff from the other: we identified a family of WAChR variants that offers improvements across the speed-sensitivity frontier.
Historically, it has been laborious and time-consuming to measure the functional properties of channelrhodopsins, but if you want to succeed in a protein engineering campaign you need to be able to make a lot of measurements. We developed a new method that lets us screen at a larger scale than people have been able to do before. We also made heavy use of machine learning models to help us come up with designs for new channelrhodopsins, and to pick the most valuable ones to test in the lab. Our filtering pipeline specifically used preliminary filters based on sequences and then used subsequent filters based on structure. The models improved as we fed them more data, so we continually got better as we went. We’ve screened over 1,750 opsins so far, and more every week. We believe these opsins represent a significant advance to the channelrhodopsin toolkit, and we’re excited to continue developing new and better proteins for patients and scientists.
