Intracellular membrane fusion has been mimicked in vitro using a mix of 17 purified proteins and lipid bilayers. This technical tour de force allows the study of how cells orchestrate and perform such fusion events.
Molecules in living organisms are constantly being replaced, yet cellular structures can maintain their identity for a lifetime1. About half of all biological processes involve membrane proteins, which must be delivered, and eventually removed, with great accuracy to regulate the constancy of structural identity. This delivery and removal is mediated by the membrane-bound organelles of eukaryotic cells, which communicate with each other by budding off vesicles and other transport packages, which travel to, and fuse with, target membranes. The organization and regulation of membrane-fusion reactions, then, are crucial for virtually every membrane-bound biological process.
But how is fusion of internal membranes achieved? To understand these mechanisms, researchers have strived to devise in vitro systems that mimic fusion events in vivo. In this issue (page 1091), Ohya et al.2 describe an enormous technical accomplishment — the self-assembly of 17 individual purified proteins into a fusion 'machine' whose activity and regulation recapitulate those of membrane fusion in an intact cell.
Two decades of investigation into membrane fusion have given us a set of key players. These include the SNARE family of membrane proteins and the Rab proteins, a subfamily of the Ras superfamily of GTPase enzymes. Mutations or pharmacological treatment of any of these components block fusion in vivo3. The membrane-bound SNARE proteins link the two membranes destined for fusion. They do this by interacting with partner SNARE proteins on the opposing membrane to form a stable, four-coiled bundle consisting of helices from several individual SNARE proteins4 (Fig. 1).
Rab GTPases, such as Rab5, are found both on membranes and in the cytosol. Rab proteins alternate between GDP-bound and GTP-bound forms, a switch controlled by proteins called guanine-nucleotide exchange factors. Like that of other members of the Ras superfamily, the nucleotide-dependent switch of Rab proteins is thought to control a downstream catalytic process. Rab proteins are involved in such processes as the regulation of organelle transport, the tethering of membranes before fusion, and the control of SNARE function5.
Each of these protein families is served by a constellation of accessory factors; these include NSF, an ATPase enzyme that helps to recycle the SNARE protein complex; the Sec1/Munc18 (SM) proteins, which together with SNAREs are essential for fusion; and various effector proteins that interact with GTP-bound, active Rab4,5.
SNARE complexes are structurally similar to viral envelope proteins known to catalyse membrane fusion6, so a paradigm emerged casting these rod-shaped helical bundles into the central role of 'minimal fusion machine', with all other proteins assigned to supporting roles such as regulation of the SNARE complex7. However, direct tests comparing the contributions of SNAREs alone with SNAREs plus all the other proteins essential for fusion have not been performed because of the complexities of assembling microgram quantities of membrane proteins in defined lipid environments together with other fragile purified proteins — requirements that stretch the limits of current biochemical technologies.
Ohya et al.2 overcome these obstacles, and reconstitute membrane fusion in vitro using physiological concentrations of 17 proteins derived from human membrane-bound organelles called endosomes, together with endosomal lipids. They quantify fusion by using a novel content-mixing assay in which light is emitted when the contents of the fusing endosomes interact. This ensures direct detection of complete membrane fusion, unlike lipid-mixing assays that confuse fusion with hemifusion — fusion of only one monolayer of each of the two lipid bilayer membranes. The result is that the reconstituted fusion is comparable in both extent and rate to the fusion of intact biological endosomes purified from cell extracts.
The reconstituted fusion is robust enough for Ohya and colleagues to increase the diameter of the resultant vesicles by up to tenfold over their initial size (the equivalent of a 100-fold increase in surface area, or fusion of 100 vesicles into one fusosome). Fusion using this new system therefore differs from that observed in reconstitutions involving only SNARE proteins, in which small changes in the diameter of fusion products are hard won. Rab5 GTPase seems to cooperate with SNARE proteins to direct and drive the specificity and extent of membrane fusion, in addition to the traditional roles ascribed to Ras-related GTPases as master regulators5. The exclusion of Rab5 from the mix eliminates fusion, and the effect of Rab5 addition can be detected only when the other Rab5 effectors are also present. This in vitro experiment recapitulates genetic experiments8 by showing the dependence of the fusion reaction on each participating protein, including the Rab proteins and their effectors. The study also reveals novel insights into the role of SNAREs and SNARE-protein accessory factors in stabilizing the recruitment of Rab5 effectors.
The view that emerges from Ohya and colleagues' work2 is that membrane fusion at physiological rates involves the synergistic interactions of an ensemble of proteins, and is not simply the sum of the actions of individual proteins. One way in which interacting clusters of SNAREs, Rabs, Rab effectors and SM proteins may cooperate to effect fusion is by changing local membrane morphology and energetics. Membrane-associated proteins can combine with lipids to form fluid proteolipid domains that drive vesicle budding9. This new work2 suggests that the endosomal-fusion machine may itself function as such a specialized domain, and be composed of the following: membrane proteins that self-associate (SNAREs)10; linker proteins such as PRA1/HuYIP3 that bind to both Rabs and SNAREs11; proteins such as Rab5 that are prenylated — a modification that favours clustering with unsaturated lipids in the membrane; and specific lipids with defined head groups, such as PI(3)P. It may help to have many different proteins all focusing their lipid-interaction energies on a hydrophobic patch of membrane to drive hemifusion12, much as rings of proteins can drive hemifission13. Perhaps one of the recruited proteins inserts amphipathic peptides — those with both hydrophobic and hydrophilic regions — into the lipid bilayer. This could buckle the membrane, and lower the energy needed for fusion. The calcium sensor synaptotagmin is thought to promote SNARE-mediated fusion of synaptic vesicles in neurons by a similar mechanism14.
Organelle fusion requires two nucleotide cycles: the GTP hydrolysis of the Rab cycle, and ATP hydrolysis to reset SNARE conformations after fusion is complete. We now have an excellent system in which to elucidate the functions of these nucleotide cycles in membrane fusion, their stoichiometries, and how they are thermodynamically coupled to fusion. We can also investigate how the reactions discriminate between different donor and acceptor membranes, and how fusion is linked to the translocation of Rabs between membranes and the cytosol. Solving these challenging problems will require a combination of approaches from fearless young scientists who are as well versed in the energetics and chemistry of membrane lipids as they are in the reconstitution of membrane proteins.
This study2 changes our view of the defined and separate roles of SNAREs and Rab proteins in membrane fusion, with Rab GTPases as conductors and SNARE proteins as responsive musicians. Rather, it seems that the players collaborate with each other to coordinate their activities, as in a chamber orchestra. The properties of fusion may arise from complex systems interactions among many components (Fig. 1). The stage is now set to explore how these collaborations take place in cellular membranes, how the score is adapted for different ensembles, and how fusion and fission are coordinated with sorting and with the dispatch of components to specific destinations.