Splitting the Difference: Sorting Photons to Improve Quantitative Measurements in Correlation Spectroscopy

Main Text

Fluorescence cross-correlation spectroscopy (FCCS) is a widely used tool to measure interactions. It is based on the fact that two fluorescence signals are correlated if two labeled particles are interacting and thus possess similar spatial distribution and/or temporal dynamics. This information can be reliably extracted and quantified by the cross-correlation function between the signals. Because FCCS does not require direct interaction of the fluorophores, as, e.g., Förster resonance energy transfer (FRET) does, but only of the labeled particles, it directly reports on the interaction of interest and is less prone to false negatives. Importantly, FCCS can report on concentrations and dynamics quantitatively and was used to determine biomolecular dissociation constants (1) even in small organisms (2,3).

However, the accuracy of FCCS measurements has been hampered by the problem of having to align multiple lasers for the different labels to the same spot with high accuracy. Failure to do so can lead to strong errors and artifacts in FCCS (4,5). This led to the development of FCCS approaches that used a single laser for excitation (6) and the development of fluorescent proteins with long Stokes shifts to allow better distinction between proteins that have the same excitation but were emitting at different wavelengths (7,8). Although an improvement by removing the necessity for the alignment of multiple lasers, this did not eliminate the problem of differences in the wavelength-dependent observation volume and of cross talk between the detection channels. In their article, Štefl et al. ((9) in this issue of the Biophysical Journal) take this final step and not only excite the samples with the same laser but also detect the signal only in one wavelength channel. But how do they then differentiate between labels? For that purpose, they take advantage of a clever filtering approach first applied in fluorescence lifetime correlation spectroscopy that allowed differentiating fluorophores by lifetimes instead of wavelength (10). Fluorescence lifetime correlation spectroscopy excites the sample with a short laser pulse and then records the arrival times of the emitted photons. It then calculates statistical filters that determine the contribution of photons to one or the other label depending on the lifetimes of the fluorophores and the arrival time of the photons, which allows calculating correlation functions when sufficient photons were collected for the statistical filers to work accurately (11). This has multiple advantages. First, it allows the differentiation of two molecules that emit at the same wavelength only depending on their lifetimes. But it also permits the filtering out of the background signal because the background signal has, in general, a very different lifetime spectrum and can be removed. This allows a more accurate and unbiased determination of concentrations, an important advantage for quantitatively determining dissociation constants. It should be noted that this statistical filtering approach is not unique to lifetimes but can be used for other fluorophore properties. This concept, for instance, was further used in fluorescence spectral correlation spectroscopy, which also allowed better distinction of fluorophores even in the case of highly overlapping spectra (12).

For the application of lifetime statistical filters, it is, however, necessary to possess fluorescent proteins with sufficiently different lifetime histograms so they can be separated, as demonstrated by the authors by simulations. For this purpose, Štefl et al. have screened and engineered green fluorescent protein (GFP) variants to obtain a construct with a lifetime of only 1.8 ns, which they call short lifetime monomeric GFP (slmGFP). As a good partner, they identified Envy, a GFP-like protein with a lifetime of 3.2 ns and good brightness and photostability (9). To prove the applicability in vivo, the authors determine the interaction of proteins of the proteasome complex in yeast. The proteasome is a large complex consisting of multiple proteins and is an ideal test sample because proteins can be selected that belong to the same complex but have a sufficient distance to avoid FRET because FRET would require additional corrections (13). In a first proof-of-principle experiment, they used the two labels to determine the interaction of proteins belonging to the proteasome, Rpn7-Envy, the lid protein, and Pre6-slmGFP, a core protein using a single excitation wavelength and a single detection channel, a technique called single-color fluorescence lifetime cross-correlation spectroscopy (FLCCS). They validated the results by using spectrally different channels in dual-color FLCCS of Rpn7-3mCherry and Pre6-slmGFP.

Although this alone is an important improvement for the applicability of FLCCS in vivo, it is still restricted to two fluorophores so far. A similar issue exists in dual-color FCCS approaches in which the number of fluorophores that can be used simultaneously has always been limited because in biological samples, the main fluorophores used are still genetically encoded fluorescence proteins that have important advantages in labeling but have relatively wide emission spectra, limiting the number of fluorophores that can be detected simultaneously.

Therefore, in a final step, Štefl et al. extend their approach to three labels by simultaneously performing single-color FLCCS and dual-color FLCCS between Rad23-3mCherry, Pre6-slmGFP, and Rpn7-Envy, providing information on the three bimolecular interactions from a single measurement and being able to determine the three dissociation constants from a single yeast strain.

This work is a further step in the quantitative analysis of biomolecular interactions that removes potential artifacts, relies on fluorescent proteins, facilitates applications in vivo, excites the labels with one wavelength, and detects fluorescence in a single channel and if combined with multicolor FCCS, can be used to study multiple interactions simultaneously (9). The authors also predict that in the longer wavelength range, mScarlet (3.9 ns) and mCherry (1.5 ns) are possible partners for single-color FLCCS. This would make the extension to four labels for simultaneous measurements possible with a simple two-color system. This not only simplifies the required systems but also reduces the amounts of measurements to be performed.

Editor: Samrat Mukhopadhyay.