Jennifer Doudna's group used AI to create compact genome editors that outperformed the natural TnpB protein in human cells
Jennifer Doudna's group used AI to create compact genome editors that outperformed the natural TnpB protein in human cells
On 16 July, Science published a study on SynTnpB, new variants of a small enzyme that guide RNA directs to a specific DNA site. The authors designed these variants using the protein's three-dimensional structure and evolutionary constraints, then tested them in living cells.
A genome editor works like a molecular tool with a specific address: guide RNA recognizes a DNA site, and the enzyme cuts or modifies it. For many applications, this tool must be delivered into the cell. The size of the enzyme determines whether it will fit into the delivery vehicle together with the required RNA. TnpB belongs to a group of compact ancestors of some CRISPR-Cas12 systems, making it suitable for delivery systems that must accommodate both the enzyme and the guide RNA.
Jennifer Doudna's group used TnpB as a starting point and separately designed two protein surfaces. One binds DNA, while the other holds the guide RNA. Evolutionary data indicated which amino acids could not be changed without impairing function. A protein inverse-folding model selected new sequences compatible with the specified three-dimensional structure of the remaining regions. The researchers then assembled combinations of these parts, screened out those that did not work in bacteria, and tested the best candidates in human and plant cells.
The best SynTnpB variants maintained or exceeded the activity of natural TnpB in human cells. One of the most active variants shared only 77% of its amino acids with the original protein. Cryo-electron microscopy showed how the designed enzyme holds RNA and DNA in two distinct positions. The laboratory produced a functional molecular machine with a new sequence and experimentally verified function.
The study tested the editors in bacteria, plants, and human cells. Delivery to specific tissues, unintended DNA cuts, and safety in humans will require separate evaluation. The model generated substantially different designs, and experimental testing identified those that actually cut DNA. This approach combines the design and testing of new proteins in the search for genome editors.
Source: Science; independent analysis: Chemical & Engineering News.