Global and pixel kinetic data analysis for FRET detection by multi-photon time-domain FLIM

Abstract

FLIM/FRET is an extremely powerful technique that can microscopically locate nanometre-scale protein-protein interactions within live or fixed cells, both in vitro and in vivo. The key to performing sensitive FRET, via FLIM, besides the use of appropriate fluorophores, is the analysis of the time-resolved data present at each image pixel. The fluorescent transient will, in general, exhibit multi-exponential kinetics: at least two exponential components arise from both the interacting and non-interacting protein. We shall describe a novel method and computer program for the global analysis of time resolved data, either at the single level or through global analysis of grouped pixel data. Kinetic models are fitted using the Marquardt algorithm and iterative convolution of the excitation signal, in a computationally-efficient manner. The fitting accuracy and sensitivity of the algorithm has been tested using modelled data, including the addition of simulated Poisson noise and repetitive excitation pulses which are typical of a TCSPC system. We found that the increased signal to noise ratio offered by both global and invariance fitting is highly desirable. When fitting mono-exponential data, the effects of a ca.12.5 ns (ca. 80 MHz) repetitive excitation do not preclude the accurate extraction of populations with lifetimes in the range 0.1 to 10 ns, even when these effects are not represented in the fitting algorithm. Indeed, with global or invariance fitting of a 32x32 pixel area, the error in extracted lifetime can be lower than 0.4 % for signals with a peak of 500 photon counts or more. In FRET simulations, modelling GFP with a non-interacting lifetime of 2.15 ns, it was possible to accurately detect a 10 % interacting population with a lifetime of 0.8 ns.

Bayesian analysis of fluorescence lifetime imaging data

“…Automated selection of the valid data window is discussed in Ref. 6. In contrast, the Bayesian analysis routines operate on all of the available data, except for small portions at the start and end of the transient, potentially corrupted by consequences of dithering associated with the time-amplitude converter in the TCSPC electronics.…”

“…1 When significant counts are available, where Gaussian statistics closely approximate the Poissonian statistics associated with photon count data, data fitting is relatively straightforward and both least squares (LS) and maximum likelihood estimation (MLE) have been shown to perform well. [2][3][4][5][6][7] The collection of significant counts is however inevitably linked to longer experimental acquisition times. When only a low number of counts is available, resulting from either low fluorophore concentrations, or their propensity to fade or when dynamic processes need to be studied, difficulties in determining the correct lifetime, or establishing the error associated with lifetime estimation inevitably increase.…”

Effect of ethanol perturbation on viscosity and permeability of an inner membrane in Bacillus subtilis spores

“…TCSPC lifetime recording was performed over 200 temporal channels (final resolution of 128 ps/channel). FLIM images were analyzed using TRI2 software version 2.4.4.1 (Gray Institute, Oxford [36]). In order to obtain about 1000 counts in the pixel peak maxima, circular binning of 4 was used for analysis of the different wild type spore samples, whereas a circular binning of 2 or 3 was used for cotE gerE spores.…”

Towards high-throughput FLIM for protein-protein interaction screening of live cells and tissue microarrays

“…Figure 2 presents the time resolved analysis panel and shows how fits and data can be compared. Fitting is performed by iterative reconvolution with the instrument response function to minimize the Chi-square parameter, for which the Levenberg-Marquardt algorithm is used as described elsewhere [9].…”

“…We therefore develop in-house hardware as well as control and analysis software [8]. This includes novel software for global analysis of time-resolved data over multiple pixels and multiple images [9]. …”