Extracellular vesicles (EVs) are biological nanoparticles with important roles in intercellular communication and pathophysiology. Their capacity to transfer biomolecules between cells has sparked efforts to bioengineer EVs as drug delivery vehicles. However, a better understanding of EV biogenesis mechanisms and function is required to unleash their considerable therapeutic potential. Here we demonstrate a novel role for GAPDH, a glycolytic enzyme, in EV assembly and secretion, and we exploit these findings to develop a GAPDHbased methodology to load therapeutic siRNAs onto EVs for targeted drug delivery to the brain. In a series of experiments, we observe high levels of GAPDH binding to the outer surface of EVs via a phosphatidylserine binding motif, designated as G58, and discover that the tetrameric nature of GAPDH promotes extensive EV aggregation. Studies in a Drosophila EV biogenesis model demonstrate that GAPDH is absolutely required for normal generation of intraluminal vesicles in endosomal compartments and promotes vesicle clustering both inside and outside the cell. Fusing a GAPDH-derived G58 peptide to dsRNA-binding motifs permits highly efficient loading of RNA-based drugs such as siRNA onto the surface of EVs. Such vesicles efficiently deliver siRNA to target cells in vitro and into the brain of a Huntington's disease mouse model after systemic injection, resulting in silencing of the huntingtin gene in multiple anatomical regions of the brain and modulation of phenotypic features of disease.Taken together, our study demonstrates a novel role for GAPDH in EV biogenesis, and that the presence of free GAPDH binding sites on EVs can be effectively exploited to substantially enhance the therapeutic potential of EV-mediated drug delivery to the brain. Extended Figure 1 GAPDH present on outer surface of EVs binds to cleaved lactoferrin: (a) Protease digestion assay of EVs under native conditions and following detergent treatment (denatured conditions).Top left represents schematic drawing of protease digestion assay. CD81-GFP was used as positive control for protein extending into the lumen of EVs. After protease treatment, protein lysates were analysed with lactoferrin, GFP, GAPDH and Alix antibodies on western blots. Lane 1, standard protein marker; 2, nontreated EVs; 3, 5, 7 and 9, EVs incubated with pronase (mixture of proteases) for indicated times under native conditions; 4, 6, 8 and 10, EVs were treated with detergent and heated at 90 ˚C for 5 min to disrupt membranes followed by treatment with proteases. Assay was carried out at 37˚C. Digestion of lactoferrin