Lessons from Tau RT–QuIC: Structure, Kinetics, and Drug Discovery

Abstract

The aggregation of the microtubule–associated protein tau into filamentous inclusions is a hallmark of various neurodegenerative diseases collectively known as tauopathies. This group encompasses over 20 clinicopathological conditions, including Alzheimer’s disease (AD), which accounts for about 70% of dementia cases worldwide. Cryo–EM studies of amyloid filaments isolated from human brains have revealed that distinct tau conformers are associated with different diseases. These findings suggest a hierarchical classification of tauopathies based on filament structure. Disease–specific aggregate conformers are categorised according to the extent of their ordered cores, which align with the isoform compositions of tau inclusions in corresponding disorders. Understanding these structural variations provides a foundation for developing novel diagnostic methods and targeted therapeutic strategies. However, two translational gaps remain: (i) current seeding assays rely on heparin, which biases filament conformations and limits their potential for future diagnostic applications, and (ii) there is no structure–validated pipeline that links such assays to polymorph–selective inhibitor discovery. This thesis closes both gaps through an integrated biochemical, structural and computational programme. In my PhD research, various aspects of tau aggregation were explored using the Real–Time Quaking–Induced Conversion (RT–QuIC) assay for the selective amplification of tau filaments from human brain homogenates. By harnessing the self-propagating nature of amyloid aggregates, this approach enables the sensitive amplification of ex vivo seeds through the strategic selection of recombinant tau proteins and the fine-tuning of aggregation conditions in the presence of the amyloid–sensitive fluorescent dye thioflavin T (ThT). First, a heparin-free, salt-modulated RT–QuIC platform was engineered using two cysteine-free tau substrates, K12 and K11, encompassing the full repeat region. Under optimised conditions, specifically at the appropriate concentrations of trisodium citrate or sulfate, spontaneous nucleation is delayed, whereas seeded aggregation proceeds rapidly within a day. Using post–mortem brain homogenates, the workflow differentiates eight tauopathies and resolves 4R strains into four kinetically distinct fluorescence-defined clusters. ATR–FTIR confirms that amplified products retain conformer–specific β–sheet fingerprints. Second, the assay was coupled to structure-guided inhibitor discovery. A two stage in silico screen of about 2 million CNS–filtered small molecules, narrowed to 1000 candidates and 102 synthesised compounds, was iteratively refined with machine learning models trained on RT–QuIC kinetics. Three chemotypes emerged that lower the secondary nucleation rate constant (k₂) of AD–derived seeds by 94–97% at 20 µM while sparing reactions seeded with Pick’s disease seeds, establishing the first proof of principle for polymorph–selective anti–tau agents. Third, the potential to direct tau toward disease-relevant conformations was investigated. AD brain homogenate seeded the assembly of C322A 0N3R tau mutant to form paired helical filament–like structures solved at 1.9 Å resolution; the C322S 0N3R tau mutant yielded an ordered C–terminal filament structure, whereas wild–type 0N3R tau formed mostly non–native aggregates. Both mutant filaments retained seeding competence through serial amplification, showing that a single side chain substitution can steer the conformational landscape. Finally, the application of the platform was illustrated in various translational settings. It resolved an unclassified human tauopathy, distinguished seed–positive P301S from seed–negative S305N mutation and knock–in mice and revealed that human iPSC–derived microglia internalise fibrillar tau and re–secrete RT–QuIC–positive aggregates, identifying microglia as active modulators of extracellular proteostasis.