The study of the dynamics of macromolecular complexes becomes essential to investigate molecular mechanisms in live cells. Different approaches are used to investigate protein–protein interactions in live cells. To this end, we are here interested in dual-color fluorescence correlation spectroscopy (FCS) [1]. The technique consists on the cross-correlation analysis of temporal fluorescence intensity fluctuations of two color fluorophores to follow the co-diffusion of two proteins labeled with these fluorescent tags. Dual-color FCS is performed at a single point and is capable of detecting interactions between proteins in which FRET does not necessarily occur, therefore provides complementary information relative to FRET in imaging mode. This approach overcomes the main limitation of FRET that requires the close proximity of the two fluorescent tags. The only requirement in order to detect interactions using dual-color FLCS is that both fluorescent proteins should diffuse and/or transiently interact with high molecular weight complexes or mobile subcellular structures. Recently, this approach was used in live cells giving novel insights on the study of macromolecular complexes, such as the MAP kinase complexes in yeast pheromone signaling [2], the dimerization of EGF receptors [3], the T cell receptor complexes [4], and the virus RNA polymerase assembly [5]. But couples of protein tags commonly used in live cells (ECFP/EYFP and EGFP/mCherry) present spectral overlap in their emission giving rise to spectral bleed-through that corrupts the cross-correlation data [5]. The cross-talk contribution from the shorter wavelengths emission species in the longer wavelengths channel is usually minimized by choosing red-shifted band-pass filter for the longer wavelengths channel and/or by decreasing the intensity of the laser exciting the shorter wavelengths emission species. But these tricks decrease the brightness of both fluorescent species, a parameter usually critical in dual-color FCS in live cells. The cross-talk can also be calibrated and its amount retrieved during data analysis [6]. But nonetheless it impedes the study of very low amounts of protein–protein interactions since at least 10 % of cross talk is always observed when working with green and red fluorescent proteins. The use of novel fluorescent proteins exhibiting a large Stokes shift in dual-color FCS experiments decreases the contribution of the spectral cross talk to the cross-correlation [7, 8], but the weak photo stability of some of them introduces other difficulties.

To eliminate spectral bleed-through in dual-color FCS, the switching method or pulsed interleaved excitation was introduced [9–11]. In this method, two alternating laser beams are controlled by an acousto-optic modulator and the synchronized correlation signals in both detectors are calculated as cross-correlation functions with no cross-talk. These approaches necessitate homemade developed confocal setups and cannot be easily implemented to commercial microscopes. A similar method was recently implemented in dual-color scanning FCS with a conventional confocal microscope to quantify ligand-receptor interactions in cell membranes of living zebrafish embryos [12]. In membranes, the diffusion of protein complexes is slow, which can be monitored by using the sequential dual-color scanning method. However, the speed of acquisition of line-scanning FCS in sequential dual color is too slow to investigate intracellular co-diffusion.

Here, we propose a FCCS-based method to overcome the EGFP and mCherry spectral cross talk in quantitative studies of protein–protein interactions in live cells by using a commercial confocal system (Microtime from PicoQuant, SP8 SMD from Leica or any conventional confocal TCSPC module) (e.g., PicoQuant or Becker & Hickl). Our approach is based on Fluorescence Lifetime Correlation Spectroscopy (FLCS). The use of a pulsed laser to perform FCS takes advantage of Time Correlated Single Photon Counting (TCSPC) to differentiate the contribution of two different fluorescence lifetime species to FCS file (FLCS, [13]). Indeed, each single photon is recorded with its global arrival time and the delay time with respect to the corresponding laser pulse. The FLCS method relies on the calculation of lifetime filters for each species having different fluorescence decay signatures in the TCSPC histogram. Each fluorescent component, in the fluorescent time trace, having different lifetime properties can be resolved on the basis of its fluorescence decay signature. FLCS was already used in spectroscopy to suppress scattered light and detector afterpulses [14], to study dynamic transition between two fluorescence lifetime states [15], to resolve different lifetime species [16], and to measure precise concentration values as low as picomolar of fluorophore [17]. FLCS was also employed in live cells to quantify the interaction of EGFR-neutralizing antibody labeled with Alexa488 and EGFR-GFP by fluorescence lifetime cross-correlation measurements [18]. Recently, we have shown that FLCS applied to dual-color EGFP and mCherry cross-correlation allows the determination of protein–protein interactions in living cells without any spectral bleed-through artifact [19].

By using two lasers to excite simultaneously two spectrally different fluorophores, a pulsed laser (470 nm which can excite a green fluorescent species, i.e., EGFP) and a continuous laser (561 nm to excite a red fluorescent protein, i.e., mCherry), dual-color FLCS provides a way to remove the spectral bleed-through of the green species into the red detection channel. The photons coming from the green species detected in the red channel are time correlated to the pulsed laser (see the fluorescence decay histogram in Fig. 1a, top) in contrast to the photons coming from the red species which do not harbor any fluorescence decay signature (see the constant emission histogram in Fig. 1a, middle). When both fluorophores are excited together, the fluorescence histogram contains the two signals (fluorescence decay for the spectral bleed-through of the green species and constant histogram for the red species, Fig. 1a, bottom). Lifetime filter functions are calculated from the two-component fluorescence decay analysis acquired in the red channel. For each time-resolved channel, the lifetime filter value gives the statistical weight of the counted photon for its corresponding species. In Fig. 1b are shown the filter corresponding to the fluorescence decay species (in blue) and this corresponding to the steady-state fluorescence (in pink). These different filters can be selected during auto- and cross-correlation process of the absolute time traces to take into account the different weighting of each individual photon. By using the lifetime filter of the fluorescence decay of EGFP in the green channel and the filter of the steady-state fluorescence of mCherry in the red channel during cross-correlation calculation of the green and red time trace, the spectral bleed-through artifact is completely removed. We stress the potentiality of the technique when applied to quantify protein–protein interactions in vivo where protein A is fused to EGFP and protein B is fused to mCherry. The cross-correlation amplitude is directly dependent on the quantity of protein complexes and is not related to both the relative concentration of green and red species and the relative laser intensities used during the experiment. The dual-color FLCS experimental procedure where the different laser intensities do not have to be controlled during the experiment is extremely powerful to quantitatively study protein interactions in live samples.