Dynamic Mode Decomposition of Fluorescence Loss in Photobleaching Microscopy Data for Model-Free Analysis of Protein Transport and Aggregation in Living Cells

Abstract

The phase separation and aggregation of proteins are hallmarks of many neurodegenerative diseases. These processes can be studied in living cells using fluorescent protein constructs and quantitative live-cell imaging techniques, such as fluorescence recovery after photobleaching (FRAP) or the related fluorescence loss in photobleaching (FLIP). While the acquisition of FLIP images is straightforward on most commercial confocal microscope systems, the analysis and computational modeling of such data is challenging. Here, a novel model-free method is presented, which resolves complex spatiotemporal fluorescence-loss kinetics based on dynamic-mode decomposition (DMD) of FLIP live-cell image sequences. It is shown that the DMD of synthetic and experimental FLIP image series (DMD-FLIP) allows for the unequivocal discrimination of subcellular compartments, such as nuclei, cytoplasm, and protein condensates based on their differing transport and therefore fluorescence loss kinetics. By decomposing fluorescence-loss kinetics into distinct dynamic modes, DMD-FLIP will enable researchers to study protein dynamics at each time scale individually. Furthermore, it is shown that DMD-FLIP is very efficient in denoising confocal time series data. Thus, DMD-FLIP is an easy-to-use method for the model-free detection of barriers to protein diffusion, of phase-separated protein assemblies, and of insoluble protein aggregates. It should, therefore, find wide application in the analysis of protein transport and aggregation, in particular in relation to neurodegenerative diseases and the formation of protein condensates in living cells.

Dynamic Mode Decomposition of Multiphoton and Stimulated Emission Depletion Microscopy Data for Analysis of Fluorescent Probes in Cellular Membranes

“…Before using DMD, each of these datasets must be reshaped into a large matrix by concatenating the xand y-coordinate into a column vector of n = x + y entries for each measurement. This gives one column at each given instance of time, z-position or polarization angle, i.e., for the independent variables, t, z, or θ (see Figure 1 for an illustration of this reshaping for the variable z and for t, as applied to bleaching-based image segmentation and FLIP microscopy, in [41,42]). Thus, for each instance of the independent variable in the original image sequence, one obtains a column in a large matrix of the form [32,36]:…”

“…In live-cell imaging, matrix decomposition methods such as principal component analysis (PCA), non-negative matrix decomposition, or independent component analysis are often used to dissect and analyze dynamic processes [39,40]. In contrast, the use of DMD in microscopy is in its infancy and, to our knowledge, limited to our recent studies in which we showed the potential of this method for bleaching-based image segmentation and for the analysis of protein dynamics and aggregation in living cells upon fluorescence loss in photobleaching (FLIP) microscopy [41,42].…”

Per-pixel unmixing of spectrally overlapping fluorophores using intra-exposure excitation modulation

“…Fluorescence microscopy is a key technology in the life sciences, though it is limited in the number of components that can be visualized at once. Several solutions to address this limitation have been developed, including computational approaches based on spectral unmixing or the use of fluorescence properties other than the emission color, such as fluorescence anisotropy [1], fluorescence lifetime [2], or light-induced processes resulting in characteristic fluorescence dynamics in the emission [3,4,5,6].…”

Special Issue “Feature Papers in Biosensors Section 2022”

“…Nanostructures based on framework DNA hold excellent promise for molecular biology studies and versatile tools for biosensor applications as reviewed in [ 3 ]. Dynamic mode decomposition of fluorescence loss was also used for monitoring protein diffusion, protein assemblies and protein aggregates in living cells [ 8 ]. Biosensors based on immobilized bacteria or yeast can be used for toxicological monitoring in natural waters [ 4 ] or for determining biochemical oxygen demand [ 5 ].…”