Abstract
Specific insoluble protein aggregates are the hallmarks of many neurodegenerative diseases1–5. For example, cytoplasmic aggregates of the RNA-binding protein TDP-43 are observed in 97% of cases of Amyotrophic Lateral Sclerosis (ALS)6,7. However, it is still unclear for ALS and other diseases whether it is the insoluble aggregates or other forms of the mutated proteins that cause these diseases that are actually toxic to cells8–13. Here we address this question for TDP-43 by systematically mutating14 the protein and quantifying the effects on cellular toxicity. We generated >50,000 mutations in the intrinsically disordered prion-like domain (PRD) and observed that changes in hydrophobicity and aggregation potential are highly predictive of changes in toxicity. Surprisingly, however, increased hydrophobicity and cytoplasmic aggregation actually reduce cellular toxicity. Mutations have their strongest effects in a central region of the PRD, with variants that increase toxicity promoting the formation of more dynamic liquid-like condensates. The genetic interactions in double mutants reveal that specific structures exist in this ‘unstructured’ region in vivo. Our results demonstrate that deep mutagenesis is a powerful approach for probing the sequence-function relationships of intrinsically disordered proteins as well as their in vivo structural conformations. Moreover, we show that aggregation of TDP-43 is not harmful but actually protects cells, most likely by titrating the protein away from a toxic liquid-like phase.