New theory tremendously improves data evaluation of fluorescence correlation spectroscopy

Since its invention by Magde, Elson, and Webb (1) in the seventies of the last century, fluorescence correlation spectroscopy (FCS) has become a widely used and powerful technique in biophysical research (2). It is most often used for studying the mobility of molecules (diffusion, directed transport) (3) but also for looking at the kinetics of photophysical transitions (4,5), the temporal dynamics of conformation fluctuations (6), or the speed of rotational diffusion (7,8). The core idea is to record long time traces of the fluorescence intensity that is excited in and detected from the picoliter-sized detection volume of a confocal microscope and then to calculate the temporal correlation function of the observed intensity fluctuations. The dependence of this correlation function on lag time then reports about, e.g., how fast fluorescent molecules diffuse in and out of the detection volume (giving the diffusion coefficient of the observed molecular species), how fast molecules switch between different photophysical states (intersystem crossing, triple-state lifetime, etc.), or how fast the fluctuations are between different structural conformations.

However, for many years, it was extremely difficult to correctly estimate the accuracy of FCS measurements or how many fluorescence photons are required to obtain an unbiased and reliable result (9). The reason is that FCS is an intrinsically statistical method, which cumulatively gets better with increasing measurement time, but with a far-from-trivial behavior concerning the scattering and bias of correlation values from their infinite-time limit. Conventionally, the ideal correlation curves that are required and used for fitting experimental FCS correlation data are always calculated in the limit of infinite measurement time, but how much signal is needed such that the actually measured correlation curve is correctly fitted, in an unbiased way, remained an open question. The usual and pragmatic approach was to measure as long as the result does not change any further with increasing measurement time.

In their new paper (10), Kohler et al. now address this problem in a rigorous way. They develop a new, analytical, and accurate model of FCS measurements with finite measurement time. One surprising result is that short measurement times lead to considerable bias in the fit results (i.e., do not evenly scatter around the infinite-time limit). Even more and better, the authors derive exact fit curves that do take into account a finite measurement time, and they show that their new model allows for obtaining unbiased, correct fit results also for short measurement times. This has a huge impact on the application of FCS for diagnostics and screening, where one limitation has always been the prohibitive long measurement times required for obtaining reliable unbiased results. Also, for FCS applications in live-cell biology, the good news is that with the new model and data analysis, one can obtain reliable results on short timescales comparable to or shorter than the timescale of the global cellular dynamics (cellular motion, spatial reorganization of subcellular structures and the cytoskeleton, etc.). Thus, the new paper represents a quantum leap in FCS analysis, making applications of FCS feasible that were not realistic before due to finite measurement times and resulting inaccuracies of data evaluation.

Editor: Dylan Myers Owen.