Lipid droplets (LDs) are central hubs for lipid storage and exchange, interfacing with the ER, mitochondria, peroxisomes, and lysosomes, and governing energy balance, signaling, and stress responses. Despite rich descriptive mapping, causally linking LD surface interactions to lipid fluxes remains challenging. In this PhD thesis, we developed a perturbative framework called ControLD: an engineered, specific, programmable system that encapsulates LDs within a liquidliquid phase-separated (LLPS) protein meshwork. By combining phase-separating scaffolds with LD-targeting modules, ControLD builds a tunable, reversible steric barrier at the LD surface, blocking interactions without chemical inhibitors. In mammalian cells, live imaging shows rapid, specific recruitment to LDs and controlled meshwork formation with minimal off-target assembly; LLPS enables on-demand trapping and release. We quantify how encapsulation remodels organelle interfaces and metabolism: limiting surface availability reduces neutral lipid influx under nutrient abundance, while trapped LDs resist consumption during starvation. Pharmacological and nutrient challenges highlight state-dependent roles of the LD surface in buffering lipid availability. To test portability across lineages, we translated ControLD to the diatom Phaeodactylum tricornutum (Pt), adapting targeting domains (including endogenous LD-binding proteins) and expression strategies. In Pt, ControLD forms robustly under multiple conditions; we compare encapsulation behaviors between models and infer potential physiological causes. Methodologically, this thesis delivers a modular toolkit for surface-localized LLPS meshworks with controlled affinity, coupling physical hindrance to metabolic readouts. By constraining LD accessibility rather than deleting components, ControLD isolates physical proximity from signaling and enzymatic partners. This establishes organelle trapping by engineered condensates as a general approach to probe organelle interfaces and lipid homeostasis. ​​​​​
​

​​Supervision: Juliette SALVAING (INRAE)
Co-supervision: Zoher GUEROU (CNRS)