Dependence of descriptors of co‐localization on microscope spatiotemporal resolution and the choice of regions of interest

The advent and increasing availability of super‐resolution microscopies has prompted re‐searchers to re‐investigate questions of co‐localization and co‐clustering in the hope of providing more precise and relevant data. Here, we focus on the problem of studying inter‐organelle interfaces, a topic of growing interest in cell biology. We sought to identify mitochondria‐associated membrane (MAM) candidate sites from dual‐colour large‐field super‐resolution images. MAMs are specialized lipid microdomains of the endoplasmic reticulum (ER) in close apposition with mitochondria. Using total internal reflection fluorescence structured‐illumination microscopy (TIRF‐SIM, Brunstein et al., Optics Express, 2013), we achieved a three‐dimensional spatial resolution down to ∼100 nm. Based on experimental and simulated data, we studied how the spatio‐temporal resolution affects common descriptors of co‐localization. The apparent overlap scaled inversely with spatial resolution (as expected for objects that are in close apposition and do not merge), independently of the precise metrics used. Important for live‐cell imaging, organelle motility made measurements more uncertain, rendering statements of how physiological stimulations or pharmacologic manipulations affect co‐localization less robust. Organelle density, or equivalent, the choice of different subcellular regions of interest (ROIs) had a marked effect on the amount of co‐localization, as had the size of the ROI chosen. Our study calls for prudence when interpreting co‐localization data and suggests that cell and organelle motility, the choice of the ROI analysed, the effective spatiotemporal resolution all impact on the result and hence should systematically be stated, particularly when co‐localization arguments are used to assess the effect of drug application on cellular signalling pathways.