Toward identification and characterization of molecular interactions in living cells: quantifying molecular colocalization in live cell fluorescence microscopy.
Interactions between molecules are essential for living organisms. Protein-protein, protein-DNA, and protein-RNA interactions are among the most important regulatory mechanisms in cell biology that determine the fate of cells and are therefore the primary target for pharmaceutical interventions. Accordingly, the quantitative identification and characterization of molecular interactions is of great importance; however, it is also one of the most challenging tasks in microscopy.
In living cells, the assessment of interactions between molecular species is typically performed by fluorescent labeling of the interaction partners with spectrally distinct fluorophores and imaging in different color channels. Yet current methods for determining colocalization of molecules result in outcomes that can vary greatly depending on signal-to-noise ratios, threshold and background levels, or differences in intensity between channels.
A German team led by Thomas Huser (University of Bielefeld) has now developed a novel and quantitative method for determining the degree of colocalization in live-cell fluorescence microscopy images for two and even more data channels. Moreover, the new method enables the construction of images that directly classify areas of high colocalization. It is based on close investigation of the Euclidian norm of a vector extracted from a symmetrical correlation matrix.
Their starting point was the recently published correlation-matrix method, which was initially developed to quantitatively and quickly judge the quality of coincidence events in single-molecule fluorescence experiments. The algorithm analyzes the joint diffusion of two (or more) fluorescent molecules through a confocal excitation volume. This method is based on evaluating the Euclidian length Gamma of a vector derived from a 4 X 4 Hermitian correlation-matrix that contains temporal correlation coefficients calculated from single-molecule diffusion time traces. The essential step in adopting this technique to the analysis of images was the conversion of temporal correlation coefficients to spatial coefficients. An image or a region of interest had to be transcribed into a linear trace of the pixel-to-pixel intensity variation. The Euclidian length Gamma of the vector can then be calculated from the corredponding correlation-matrix.
A novel parameter derived from these calculations and called the Gamma-norm was introduced as a measure for colocalization. For two-color fluorescence experiments this norm adopts values between 0 (i.e. no interactions) and 2 (fully interacting binding partners). Based on this value a direct calculation of the fraction of pixels that do not colocalize or colocalize only randomly is possible.
Compared to other methods, the new correlation-matrix method shows a superior robustness against the influence of varying background or random noise contributions. Even at high noise levels, the Gamma-norm acts as a filter against these factors. An extension to the analysis of three- or multicolor images is straightforward. The technique can readily be applied to 3D and super-resolution microscopy data, as well as data obtained by other contrast methods, e.g. Raman scattering or second harmonic generation (SHG). It could even be expanded to visualize dynamic colocalization effects, i.e. in live-cell movie data.
In order to enable the wider research community to easily test and adopt the Gamma-norm analysis, the team developed a plugin for the widely used open-source image analysis software Fiji, which is available for download here.