Improving Accuracy and Precision in Biological Applications of Fluorescence Lifetime Imaging Microscopy.

Abstract

The quantitative understanding of cellular and molecular responses in living cells is important for many reasons, including identifying potential molecular targets for treatments of diseases like cancer. Fluorescence lifetime imaging microscopy (FLIM) can quantitatively measure these responses in living cells by producing spatially resolved images of fluorophore lifetime, and has advantages over intensity-based measurements. However, in live-cell microscopy applications using high-intensity light sources such as lasers, maintaining biological viability remains critical. Although high-speed, time-gated FLIM significantly reduces light delivered to live cells, making measurements at low light levels remains a challenge affecting quantitative FLIM results. We can significantly improve both accuracy and precision in gated FLIM applications. We use fluorescence resonance energy transfer (FRET) with fluorescent proteins to detect molecular interactions in living cells: the use of FLIM, better fluorophores, and temperature / CO2 controls can improve live-cell FRET results with higher consistency, better statistics, and less non-specific FRET (for negative control comparisons, p-value = 0.93 (physiological) vs. 9.43E-05 (non-physiological)). Several lifetime determination methods are investigated to optimize gating schemes. We demonstrate a reduction in relative standard deviation (RSD) from 52.57% to 18.93% with optimized gating in an example under typical experimental conditions. We develop two novel total variation (TV) image denoising algorithms, FWTV (f-weighted TV) and UWTV (u-weighted TV), that can achieve significant improvements for real imaging systems. With live-cell images, they improve the precision of local lifetime determination without significantly altering the global mean lifetime values (<5% lifetime changes). Finally, by combining optimal gating and TV denoising, even low-light excitation can achieve precision better than that obtained in high-light cases (RSD = 12.76% at total photon counts (TC) = 100 vs. RSD = 23.03% at TC = 400). Therefore, high-intensity excitation of living cells can be avoided. Notable five-fold improvements in precision (RSD from 49.90% to 11.94%) are easily observed in our extreme low-light example. This study overcomes several challenges associated with making quantitative measurements of cellular responses, by enabling novel fluorescence lifetime map construction for better quantitation of molecular interactions and sub-cellular environmental changes in live cells.