VE-cadherin junctional activity

VE-cadherin junctional activity of retinas was analysed by hand using the patching algorithm⁹. The algorithm processed each optical plane of the z-stack of the confocal retinal images (× 63 magnification) into random patches of 100 × 100 pixels based on a user-defined threshold for the anti-VE-cadherin signal. The resulting patches were subsequently blindly presented to the user who hand-classified the VE-cadherin morphology in each patch according to the key shown in Fig. 8h. Next, the classified patches were used to reconstruct images as three-dimensional-summed z-projections of the patch class colour overlaid on the VE-cadherin staining. The resulting colour(s) give insight into the extent of junctional heterogeneity of the vessels.