Rapid global fitting of large fluorescence lifetime imaging microscopy datasets

Abstract

Fluorescence lifetime imaging (FLIM) is widely applied to obtain quantitative information from fluorescence signals, particularly using Förster Resonant Energy Transfer (FRET) measurements to map, for example, protein-protein interactions. Extracting FRET efficiencies or population fractions typically entails fitting data to complex fluorescence decay models but such experiments are frequently photon constrained, particularly for live cell or in vivo imaging, and this leads to unacceptable errors when analysing data on a pixel-wise basis. Lifetimes and population fractions may, however, be more robustly extracted using global analysis to simultaneously fit the fluorescence decay data of all pixels in an image or dataset to a multi-exponential model under the assumption that the lifetime components are invariant across the image (dataset). This approach is often considered to be prohibitively slow and/or computationally expensive but we present here a computationally efficient global analysis algorithm for the analysis of time-correlated single photon counting (TCSPC) or time-gated FLIM data based on variable projection. It makes efficient use of both computer processor and memory resources, requiring less than a minute to analyse time series and multiwell plate datasets with hundreds of FLIM images on standard personal computers. This lifetime analysis takes account of repetitive excitation, including fluorescence photons excited by earlier pulses contributing to the fit, and is able to accommodate time-varying backgrounds and instrument response functions. We demonstrate that this global approach allows us to readily fit time-resolved fluorescence data to complex models including a four-exponential model of a FRET system, for which the FRET efficiencies of the two species of a bi-exponential donor are linked, and polarisation-resolved lifetime data, where a fluorescence intensity and bi-exponential anisotropy decay model is applied to the analysis of live cell homo-FRET data. A software package implementing this algorithm, FLIMfit, is available under an open source licence through the Open Microscopy Environment.

Figure 1. Profiling of the CPU and memory requirements of the algorithm.

(A) Fractional core activity of the four cores while performing the fit in Experiment 1, colour coded by algorithm stage as shown in the flowchart in (B) which shows the main stages of the algorithm. (C) CPU time spent on the different algorithm stages in Experiment 1. (E) Memory requirements for a global fit against image number for (blue) a five frame time gated FLIM dataset, (red) a TCSPC dataset and (green) a two channel polarisation resolved dataset. Numbers exclude the memory required for the MATLAB runtime engine (300 MB).

Figure 2. Global analysis of a multiwell plate with varying concentrations of fluorescent dyes.

Global analysis was applied to a multiwell plate with varying concentrations of the fluorescent dyes Rhodamine B and Rhodamine 6G using a bi-exponential model. The relative concentration of Rhodamine 6G reduces across pairs of columns as described in the text. The dataset contains four fields FOV per well. A) plate map showing the measured fractional contribution of Rhodamine 6G for a representative FOV in each well. B) plot of the actual Rhodamine 6G contribution against measured contribution (crosses). C) plot of measured lifetime using a single exponential fit against actual Rhodamine 6G concentration. This dataset was collected as part of a previous study .

Figure 3. Global analysis of an IPA-3 dose-response dataset modulating the interaction between Rac1 and Pak1.

Global analysis was applied to a multiwell plate dose-response dataset showing the effect of the inhibitor IPA-3 on interaction between Rac1 and Pak1 using an mTurquoise variant of the FLAIR biosensor in COS-7 cells stimulated with EGF. A) representative images from each inhibitor concentration showing distribution of fraction undergoing FRET. B) examples of automatic image segmentation with (left) donor intensity and (right) acceptor images shown in grey-scale with coloured segmented cell regions overlaid. C) plot of fraction of donor molecules undergoing FRET against IPA-3 concentration, averaged across segmented cells with fitted dose-response curve. Error bars indicate 95% confidence intervals on average FRET fraction over segmented cells at each dose. White scale bar represents 100 µm.

Figure 4. Global analysis of an NMT inhibitor dose-response dataset modulating Gag aggregation.

Global analysis was performed across a multiwell plate dataset with HeLa cells expressing ECFP-Gag and EYFP-Gag with increasing levels of a NMT inhibitor using a bi-exponential donor FRET model. A) representative images from each inhibitor dose showing distribution of fraction Gag-CFP undergoing FRET. B) plot of fraction of Gag-CFP undergoing FRET against inhibitor concentration, averaged across wells with fitted dose-response curve. Error bars indicate 95% confidence intervals across wells. This dataset was collected as part of a previous study . White scale bar represents 100 µm.

Figure 5. Global and pixel-wise analysis of simulated TCSPC polarisation resolved image data.

Simulated polarisation resolved TCSPC data was generated with fluorescence lifetimes 3.0 and 1.2 ns and rotational correlation times of 30 ns and 1.0 ns. The simulated data was generated with the total initial anisotropy set to 0.4 across the image with the initial anisotropy contribution of the short component equal to 0.1, 0.2 and 0.3 in three bands from top to bottom across the image. (A) False colour images of the recovered initial anisotropy contribution for the long (left) and short (right) correlation time components analysed pixel-wise (top) and with global fitting (bottom); Histograms of estimates of the initial anisotropy contribution of the (B) long and (C) short correlation time components analysed pixel-wise (dashed lines) and with global fitting (solid lines).

Figure 6. Global analysis of a polarisation resolved homo-FRET TCSPC dataset reading out PtdIns(3,4,5)P₃ accumulation at the membrane.

A MEF transfected with EGFP-AKT-PH was imaged at two minute intervals and stimulated with 50 ng/ml PDGF after 6 minutes (indicated by black triangles). (A, top row) False colour map of the initial anisotropy contribution r ₂ associated with homo-FRET over the time course. (A, bottom row) Integrated fluorescence intensity images over the time course. (B, C) Initial anisotropy contributions spatially averaged over the cell: r₁ associated with the rotational correlation (B) and r₂ associated with homo-FRET (C). Error bars represent the standard deviation across the image. (D,E) Exemplar fluorescence decays from the region indicated by a white triangle in the first (D) and last (E) frame with fit (top) and normalised residuals (bottom). The thin, fainter lines represent the experimental data while the thick, bolder lines represent the fitted model. Black lines represent the parallel component while grey lines represent the perpendicular component. Data are representative of three experiments. White scale bar represents 20 µm.