The following article requires a subscription:



(Format: HTML, PDF)

Summary: Fluorescence imaging of two independently labelled proteins is commonly used to determine their co-localization in cells. Antibody-mediated crosslinking can mediate the patching of such proteins at the cell surface, and their co-localization can serve to determine complex formation among them. However, manual analysis of such studies is both tedious and subjective. Here we present a digital co-localization analysis that is independent of the fluorescence intensity, is highly consistent and reproducible between observers, and dramatically reduces the analysis time. The approach presented is based on a segmentation procedure that creates binary objects, and then determines whether objects belonging to two different groups (e.g. green- and red-labelled) are co-localized. Two methods are used to determine co-localization. The 'overlap' analysis defines two objects as co-localized if the centre of mass of one falls within the area of the other. The 'nearest-neighbour distance' analysis considers two objects as co-localized if their centres are within a threshold distance determined by the imaging modality. To test the significance of the results, the analysis of the actual images is tested against randomized images generated by a method that creates images with uncorrelated distributions of objects from the two groups. The applicability of the algorithms presented to study protein interactions in live cells is demonstrated by co-patching studies on influenza haemagglutinin mutants that do or do not associate into mutual oligomers at the cell surface via binding to AP-2 adaptor complexes. The approach presented is potentially applicable to studies of co-localization by other methods (e.g. electron microscopy), and the nearest-neighbour distance method can also be adapted to study phenomena of correlated placement.

(C) 2003 Royal Microscopical Society