Differentiating between Membrane Topography and Molecular Clustering

Abstract

According to single-molecule localisation microscopy (SMLM) almost every plasma membrane protein is organised into clusters. However, an appreciable fraction of this clustering could be due to variations in membrane topography rather than genuine clustering since an increase in the local membrane density can make molecules that are randomly distributed appear to be clustered. Cells are neither flat nor smooth and membrane topography variations can be caused by membrane folding, undulations, protrusions like microvilli and lamellipodia as well as invaginations like caveolae and clathrin coated pits. To differentiate between topography-induced and genuine clustering variations in the membrane density needs to be factored out. Live cell dual colour SMLM, with the fluorophore DiI showing the membrane density together with the proteins the transferrin receptor (TfR) or the GPI-anchored CD59 was performed to demonstrate that apparent clustering due to topography variations can be identified in SMLM datasets. Pair correlation analysis reported both the two proteins and the membrane dye as being clustered. Converting the localisations to images and performing signal distribution analysis, on the other hand, revealed that the DiI and CD59 clusters are simply caused by variations in topography, i.e. the CD59 clusters are not genuine clusters whereas the TfR clusters are legitimate. Performing simulations on localisation distributions varying the partition ratio to clusters, blurring, cluster number, cluster size and image size we demonstrate that signal distribution analysis is a powerful tool to differentiate clustered from random distributions and to detect very low levels of clustering. We conclude that topography must be considered when deciding whether clusters are genuine or artefactual. Our findings imply that clustering may be less ubiquitous than the current literature suggests.