Quantitative Colocalization Analysis in Fluorescence Microscopy

Background

The intricate spatial organization of macromolecules, such as proteins and nucleic acids, plays a central role in living systems. Macromolecules are organized through various mechanisms at different levels. First, for multicellular organisms, the production of many molecules is usually cell- or tissue-specific. Second, functional molecules often localize to specific organelles or subcellular locations in individual cells. Third, multiple molecules also form complexes through physical or biochemical interactions. Thus, decoding the localization of macromolecules, especially their colocalization with cellular structures and other macromolecules, provides critical information on their biological functions. Fluorescence microscopy has become an invaluable tool in cell biology because of its label specificity, sensitivity, and non-invasive nature. Using fluorescence microscopy, different molecular or cellular components can be labeled with distinct fluorophores, their localization imaged in separate channels, and their colocalization quantified by the overlap of their fluorescence signals [1].

Colocalization consists of two distinct components: co-occurrence and correlation, which reflect the extent of spatial and intensity co-distribution of two fluorophores, respectively [2,3]. To display colocalization qualitatively, the prevalent method is to show the superposition of individual fluorescent images obtained in different channels and to reflect the overlap of different fluorophores using the colors of the merged image. Scatterplots provide a semi-quantitative method for visualizing colocalization, in which the intensities of two fluorophores are plotted against each other pixel-by-pixel. Under proportional co-distribution, the points in a scatterplot would cluster around a straight line. In contrast, points would show a spread distribution in the case of a lack of co-distribution [2–4]. However, superposition and scatterplot do not provide quantitative measurements and cannot be used for comparing the extent of colocalization across different genetic backgrounds and experimental conditions.

Manders’ colocalization coefficients (M1 and M2)

M1, the fraction of color R overlapping with color G

$M1=\frac{\sum _i{R}_{i,colocal}}{\sum _i{R}_i}$(1)

Where R_i, colocal = R_i, if G_i > 0, and R_i, colocal = 0 if G_i = 0.

M2, the fraction of color G overlapping with color R

$M2=\frac{\sum _i{G}_{i,colocal}}{\sum _i{G}_i}$(2)

Where G_i, colocal = G_i, if R_i > 0, and G_i, colocal = 0 if R_i = 0.

Pearson’s correlation coefficient (PCC)

$PCC=\frac{\sum _{}(R_i-\overline{R})\times (G_i-\overline{G})}{\sqrt{\sum _{}{(R_i-\overline{R})}^2\times \sum _{}{(G_i-\overline{G})}^2}}$(3)

Where R_iand G_i represent the intensity values of two individual channels of pixel i, and and represent the mean intensities of these channels across the entire image.

Manders’ overlap coefficient (MOC)

$MOC=\frac{\sum _{}(R_i\times G_i)}{\sqrt{\sum _{}{R_i}^2\times \sum _{}{G_i}^2}}$(4)

To quantify the co-occurrence, M1 and M2 are the most popular metrics. M1 and M2 measure the co-occurrence fraction of color 1 with color 2 (M1) and the co-occurrence fraction of color 2 with color 1 (M2) [5]. The calculation of M1 and M2 requires the sum of the intensity for each fluorophore and the sum of the intensity for each fluorophore within the area where these two fluorophores overlap [5]. Thus, M1 and M2 provide intuitive and direct measurements of colocalization, which are independent of signal proportionality.

Pearson correlation coefficient (PCC) and Mander’s overlap coefficient (MOC) are widely used in the quantification of correlation. PCC quantifies to what extent the signal intensity variation in two color images can be explained by a simple, linear relationship [3]. PCC measures the pixel-by-pixel covariance between the two-color images, which is sensitive to both signal co-occurrence and correlation. The meaning of extremely high or low PCC is generally clear, but intermediate values are difficult to interpret. Values close to 1 or -1 indicate that the two-color images have a perfect positive or negative correlation, and values close to 0 indicate a lack of correlation [3]. Mander’s overlap coefficient (MOC) is mathematically similar to PCC. Regarding the analysis of colocalization from fluorescence images, the major difference between these two metrics is that MOC uses absolute fluorescence intensities, while PCC is calculated from the intensities after the subtraction of the mean [2]. Compared to PCC, MOC is almost independent of signal proportionality and sometimes measures co-occurrence indirectly and unpredictably [2,3]. Therefore, we choose co-occurrence fractions (M1 and M2) and PCC as the major indicators for colocalization.

Various software packages perform colocalization analysis using these metrics. However, many of these tools operate as a “black box,” providing little or no visualization of intermediate processing steps. As a result, users often determine signal thresholds through trial and error, leading to subjective, arbitrary, and potentially irreproducible results. Here, we present a protocol for analyzing colocalization in fluorescence microscopy images using coAnalyzer, a MATLAB-based software package. The graphical user interface (GUI) enables multiple ROI selection, one-step image merging, line scanning, data integration, and stepwise visualization of the analysis process. Additionally, coAnalyzer can be applied to image folders, facilitating efficient analysis of large datasets.