Direct observation and classification of heterogeneous protein aggregation in real-time through super-resolution and aggregational fingerprinting.

Abstract

Protein aggregation and misfolding are highly diminishing components of protein function and expression and are known to be a major part of multiple pathologies, including amyloidosis and neurologic disorders. Current understanding of protein aggregation mainly relies on ensemble measurements and diffraction-limited imaging, effectively averaging heterogeneous aggregation behavior and their kinetics. We introduced real-time photobleaching localization microscopy (REPLOM), a novel approach to super-resolution imaging to observe directly and in real-time heterogeneous aggregation growth. We generated a generic unsupervised machine learning (ML) framework for agnostic, automated protein aggregation analysis concluding what we term an aggregational fingerprint akin to our recent work in Pinholt et al. The aggregational fingerprint contains multiple spatial and kinetic descriptive features for each individual protein aggregate, enabling rapid and precise characterization of aggregation. By combining REPLOM and ML-driven aggregational fingerprinting we were able to track and characterize single aggregates and fibril formation in real-time revealing new heterogeneous growth. We expect our proven combined framework to generalize as a universal, automated protein aggregation analysis platform to drive further mechanistic insights.