Bayesian Fluorescence Correlation Microscopy to Resolve Heterogeneity in Membrane Protein Dynamics

Abstract

Fluorescence Correlation Spectroscopy (FCS) is a powerful tool to measure molecular dynamics with single molecule sensitivity, including local concentrations, aggregation states, and transport mechanisms. Due to its high spatial-temporal resolution and non-invasive nature, FCS has been widely used to probe molecular dynamics in living systems. In contrast to single point detection in traditional confocal FCS, advances in Electron-Multiplying Charge-Coupled Device (EM-CCD) camera technology, Total Internal Reflection Fluorescence Microscopy (TIRFM), and Single-Plane Illumination Microscopy (SPIM) now enable parallel measurements at hundreds to thousands of spatial locations either on cell surface membranes or within tissues. However, objective, automated Multiple Hypothesis Testing (MHT) for such highly parallel FCS measurements is challenging due to sampling limitations, sample heterogeneity, and the lack of automated procedures to accurately estimate the highly correlated noise that is present in Temporal Autocorrelation Functions (TACFs). Thus, automated, unbiased MHT procedures are critical for the proper analysis and interpretation of highly parallel FCS measurements as obtained from TIRFM and SPIM. Here, we apply a Bayesian inference procedure for MHT of competing models for FCS datasets. Simulated FCS measurements demonstrate that the Bayesian procedure selects the simplest model that describes the observed data, thereby capturing heterogeneity in the sample while avoiding over-fitting. Further, model probabilities provide a reliability test for the downstream interpretation of measured parameter values. We apply the procedure to TIRFM FCS data of phase-separated supported lipid bilayers and also Epidermal Growth Factor Receptor (EGFR), demonstrating that membrane and receptor heterogeneity and dynamics are captured when noise is sufficiently low. Our results demonstrate that the Bayesian approach provides an automated, unbiased procedure for FCS data analysis and interpretation with broad applicability to resolving heterogeneous molecular dynamics in biological processes interrogated using TIRFM and SPIM.