A novel method to accurately locate and count large numbers of steps by photobleaching

• Published in: Molecular biology of the cell 2016-11, Vol.27 (22), p.3601-3615
• Main Authors: Tsekouras, Konstantinos, Custer, Thomas C, Jashnsaz, Hossein, Walter, Nils G, Pressé, Steve
• Format: Article
• Language: English
• Subjects: Algorithms; Bayes Theorem; Fluorescence; Fluorescent Dyes; Microscopy, Fluorescence - methods; Photobleaching; Signal-To-Noise Ratio; Statistics as Topic - methods
• ISSN: 1059-1524; 1939-4586
• DOI: 10.1091/mbc.e16-06-0404

Photobleaching event counting is a single-molecule fluorescence technique that is increasingly being used to determine the stoichiometry of protein and RNA complexes composed of many subunits in vivo as well as in vitro. By tagging protein or RNA subunits with fluorophores, activating them, and subsequently observing as the fluorophores photobleach, one obtains information on the number of subunits in a complex. The noise properties in a photobleaching time trace depend on the number of active fluorescent subunits. Thus, as fluorophores stochastically photobleach, noise properties of the time trace change stochastically, and these varying noise properties have created a challenge in identifying photobleaching steps in a time trace. Although photobleaching steps are often detected by eye, this method only works for high individual fluorophore emission signal-to-noise ratios and small numbers of fluorophores. With filtering methods or currently available algorithms, it is possible to reliably identify photobleaching steps for up to 20-30 fluorophores and signal-to-noise ratios down to ∼1. Here we present a new Bayesian method of counting steps in photobleaching time traces that takes into account stochastic noise variation in addition to complications such as overlapping photobleaching events that may arise from fluorophore interactions, as well as on-off blinking. Our method is capable of detecting ≥50 photobleaching steps even for signal-to-noise ratios as low as 0.1, can find up to ≥500 steps for more favorable noise profiles, and is computationally inexpensive.