LS-3-O-2560 Intracellular distribution of phosphatidylinositol-3,5-bisphosphate revealed by quick-freezing and freeze-fracture replica labeling

Our knowledge of lipid distribution is limited because of the difficulty in visualizing lipids. The difficulty is a result of their fast lateral movement and unresponsiveness to chemical fixatives. To overcome this difficulty, we have developed a unique, electron-microscopic method, called quick-freezing and freeze-fracture replica labeling (QF-FRL). In this method, membrane lipids are immobilized by quick-freezing, which is followed by freeze-fracture and vacuum evaporation of carbon and platinum onto the hydrophobic side of the lipid monolayer, fixing the lipid distribution physically. The preparation, called a replica, is then treated with SDS to remove extramembranous materials and used for immunolabeling (Figure 1). Membranes in the replica are observed as a quadric surface, enabling the analysis of the two-dimensional distribution of lipids at the nanometer scale.
Phosphatidylinositol-3,5-bisphosphate (PI(3,5)P₂) is a low-abundance phosphoinositide, which is up-regulated in response to various cellular stresses. Mutations in its metabolizing enzymes affect endosomal functions and cause neuronal defects in higher animals. Despite of the functional importance, detailed localization of PI(3,5)P₂ remains unknown. Therefore, we developed a method to reveal the distribution of PI(3,5)P₂ using QF-FRL.
Several PI(3,5)P₂-binding proteins prepared as recombinant proteins were examined for binding specificity using replicas of different phosphoinositide-containing liposomes. Among them, S. cerevisiae ATG18p showed the most intense labeling for PI(3,5)P₂. ATG18p also reacted with PI(3)P, but this PI(3)P labeling was abolished by incubating with an excess amount of the p40 PX domain, which only binds to PI(3)P and not to PI(3,5)P₂ (Figure 2).
PI(3,5)P₂ labeling was observed in the vacuole of S. cerevisiae that was subjected to hyperosmotic stress to stimulate PI(3,5)P₂ generation. Interestingly, vacuolar membrane labeling was concentrated in a domain where the intramembrane particles (IMP) are excluded, whereas PI(3)P was labeled in the entire vacuolar membrane. The IMP-poor domains frequently coincided with the site of membrane contacts either between neighboring vacuoles or between the vacuole and the nucleus. The domain was often invaginated toward the lumen, which is reminiscent of microautophagy. These results suggest that PI(3,5)P₂ is generated and functions in restricted areas of the vacuolar membrane. This is the first nanoscale demonstration of the PI(3,5)P₂ distribution. Our approach should help analyze the function of PI(3,5)P₂ in both physiological and pathological contexts.