To evaluate the artificial intelligence (AI)-guided AlphaFold algorithm for studying the binding interactions of human huntingtin and the aggregation of huntingtin peptides. Variants of huntingtin protein implicated in Huntington's disease were used as a model system to evaluate AlphaFold. Variants of huntingtin and huntingtin peptides with polyglutamine tracts (PQT) containing 21, 31, 51, or 78 glutamines were studied. The 3-dimensional structures of huntingtin variants and their interactions with huntingtin-associated protein-40 (HAP40) were obtained. Aggregation experiments were conducted with peptide sequences corresponding to variants of PQT, amino terminal sequence (NTS) plus PQT, NTS plus PQT plus proline rich region (PRR), and the 300 amino acid sequence from the NTS through HEAT3 of huntingtin. Oligomerization experiments with 1, 3, 6, or 12 peptide sequences were used to assess the quaternary structures of aggregates. The PQT and PQT plus NTS peptides formed a helical secondary structure that formed a central core in the quaternary structure of the aggregates The PRR formed an extended type II polyproline helix that did not participate in central core the aggregates. The distance between the amino and carboxyl termini of disease-linked 31Q, 51Q, and 78Q variants of full-length huntingtin was prominently decreased compared to the 21Q huntingtin. The interaction of HAP40 with the 78Q variant increased the distance between the amino and carboxyl termini. AlphaFold identified key tertiary structure changes in human huntingtin that have been independently corroborated in experimental models. The results highlight the utility of AlphaFold for hypothesis generation in pharmaceutical research.
Read full abstract