p97, an abundant hexameric ATPase of the AAA family, is involved in homotypic membrane fusion. It is thought to disassemble SNARE complexes formed during the process of membrane fusion. Here, we report two structures: a crystal structure of the N-terminal and D1 ATPase domains of murine p97 at 2.9 A resolution, and a cryoelectron microscopy structure of full-length rat p97 at 18 A resolution. Together, these structures show that the D1 and D2 hexamers pack in a tail-to-tail arrangement, and that the N domain is flexible. A comparison with NSF D2 (ATP complex) reveals possible conformational changes induced by ATP hydrolysis. Given the D1 and D2 packing arrangement, we propose a ratchet mechanism for p97 during its ATP hydrolysis cycle.

At least two distinct ATPases, NSF and p97, are known to be involved in the heterotypic fusion of transport vesicles with their target membranes and the homotypic fusion of membrane compartments. The NSF-mediated fusion pathway is the best characterized, many of the components having been identified and their functions analysed. In contrast, none of the accessory proteins for the p97-mediated fusion pathway has been identified. Now we have identified the first such component, a protein of relative molecular mass 47,000 (p47), which forms a tight, stoichiometric complex with cytosolic p97 (one trimer of p47 per hexamer of p97). It is essential for the p97-mediated regrowth of Golgi cisternae from mitotic Golgi fragments, a process restricted to animal cells. As a homologue of p47 exists in budding yeast, this indicates that it might also be involved in other membrane fusion reactions catalysed by p97, such as karyogamy.

A cell-free system that mimics the reassembly of Golgi stacks at the end of mitosis requires two ATPases, NSF and p97, to rebuild Golgi cisternae. Morphological studies now show that alpha-SNAP, a component of the NSF pathway, can inhibit the p97 pathway, whereas p47, a component of the p97 pathway, can inhibit the NSF pathway. Anti-syntaxin 5 antibodies and a soluble, recombinant syntaxin 5 inhibited both pathways, suggesting that this t-SNARE is a common component. Biochemical studies confirmed this, showing that p47 binds directly to syntaxin 5 and competes for binding with alpha-SNAP. p47 also mediates the binding of p97 to syntaxin 5 and so plays an analogous role to alpha-SNAP, which mediates the binding of NSF.

NSF and p97 are ATPases required for the heterotypic fusion of transport vesicles with their target membranes and the homotypic fusion of organelles. NSF uses ATP hydrolysis to dissociate NSF/SNAPs/SNAREs complexes, separating the v- and t-SNAREs, which are then primed for subsequent rounds of fusion. In contrast, p97 does not dissociate the p97/p47/SNARE complex even in the presence of ATP. Now we have identified a novel essential factor for p97/p47-mediated membrane fusion, named VCIP135 (valosin-containing protein [VCP][p97]/p47 complex-interacting protein, p135), and show that it binds to the p97/p47/syntaxin5 complex and dissociates it via p97 catalyzed ATP hydrolysis. In living cells, VCIP135 and p47 are shown to function in Golgi and ER assembly.