EXTRAORDINARY IBMB SEMINAR

Amyloid protein aggregates are pathological hallmarks of more than fifty human diseases including the most common neurodegenerative disorders. While Cryo-EM has revealed the structure of many different fibrils, the process by which a soluble protein nucleates the formation of amyloid fibrils is still poorly characterised. This is the key step to understand to be able to prevent or slow down the formation of amyloids, but the rules controlling the kinetics of the process are still unclear. This is partially due to how difficult it is to characterise amyloid formation in vitro and to do this at scale. We have developed a kinetic selection assay to report on amyloid nucleation and employed it with combinatorial mutagenesis and machine learning to reveal the transition state of the nucleation reaction of amyloid beta, the protein that aggregates in Alzheimer’s disease. By quantifying the nucleation rates of over 140,000 proteins we obtained an estimate of the changes in activation energy for every possible mutation in amyloid beta, as well as a quantification of the energetic couplings between over 600 pairs of mutations. This dataset allowed us to build a “portrait” of the amyloid beta nucleation transition state, revealing that amyloid fibrils nucleate starting from a structured C-terminal hydrophobic region where specific interactions are established.  While identifying the amyloid beta conformation to target to stop or slow down amyloid formation in Alzheimer’s disease, our work provides a general strategy based on combinatorial mutagenesis and selection that can be used to target the transition states of other biological reactions. What is more, these datasets can also be used to train new models that accurately predict amyloid formation from sequence.