The question concerning whether all membranes fuse according to the same mechanism has yet to be answered satisfactorily. During fusion of model membranes or viruses, membranes dock, the outer membrane leaflets mix (termed hemifusion), and finally the fusion pore opens and the contents mix. Viral fusion proteins consist of a membrane-disturbing 'fusion peptide' and a helical bundle that pin the membranes together. Although SNARE (soluble N-ethylmaleimide-sensitive factor attachment protein receptor) complexes form helical bundles with similar topology, it is unknown whether SNARE-dependent fusion events on intracellular membranes proceed through a hemifusion state. Here we identify the first hemifusion state for SNARE-dependent fusion of native membranes, and place it into a sequence of molecular events: formation of helical bundles by SNAREs precedes hemifusion; further progression to pore opening requires additional peptides. Thus, SNARE-dependent fusion may proceed along the same pathway as viral fusion: both use a docking mechanism via helical bundles and additional peptides to destabilize the membrane and efficiently induce lipid mixing. Our results suggest that a common lipidic intermediate may underlie all fusion reactions of lipid bilayers.

Fusion pore opening and expansion are considered the most energy-demanding steps in viral fusion. Whether this also applies to soluble N-ethyl-maleimide sensitive fusion protein attachment protein receptor (SNARE)– and Rab-dependent fusion events has been unknown. We have addressed the problem by characterizing the effects of lysophosphatidylcholine (LPC) and other late-stage inhibitors on lipid mixing and pore opening during vacuole fusion. LPC inhibits fusion by inducing positive curvature in the bilayer and changing its biophysical properties. The LPC block reversibly prevented formation of the hemifusion intermediate that allows lipid, but not content, mixing. Transition from hemifusion to pore opening was sensitive to guanosine-5′-(γ-thio)triphosphate. It required the vacuolar adenosine triphosphatase V0 sector and coincided with its transformation. Pore opening was rate limiting for the reaction. As with viral fusion, opening the fusion pore may be the most energy-demanding step for intracellular, SNARE-dependent fusion reactions, suggesting that fundamental aspects of lipid mixing and pore opening are related for both systems.

Pore models of membrane fusion postulate that cylinders of integral membrane proteins can initiate a fusion pore after conformational rearrangement of pore subunits. In the fusion of yeast vacuoles, V-ATPase V0 sectors, which contain a central cylinder of membrane integral proteolipid subunits, associate to form a transcomplex that might resemble an intermediate postulated in some pore models. We tested the role of V0 sectors in vacuole fusion. V0 functions in fusion and proton translocation could be experimentally separated via the differential effects of mutations and inhibitory antibodies. Inactivation of the V0 subunit Vph1p blocked fusion in the terminal reaction stage that is independent of a proton gradient. Δvph1 mutants were capable of docking and trans-SNARE pairing and of subsequent release of lumenal Ca2+, but they did not fuse. The Ca2+-releasing channel appears to be tightly coupled to V0 because inactivation of Vph1p by antibodies blocked Ca2+ release. Vph1 deletion on only one fusion partner sufficed to severely reduce fusion activity. The functional requirement for Vph1p correlates to V0 transcomplex formation in that both occur after docking and Ca2+ release. These observations establish V0 as a crucial factor in vacuole fusion acting downstream of trans-SNARE pairing.

Poxviruses, such as vaccinia virus (VV), replicate their DNA in endoplasmic-reticulum-enclosed cytoplasmic sites. Here, we compare the dynamics of the VV replication sites with those of the attenuated strain, modified VV Ankara (MVA). By live-cell imaging, small, early replication sites of both viruses undergo motility typical of microtubule (MT)-motor-mediated movement. Over time, growing replication sites of VV collect around the nucleus in a MT-dependent fashion, whereas those of MVA remain mostly scattered in the cytoplasm. Surprisingly, blocking the dynein function does not impair the perinuclear accumulation of large VV replication sites. Live-cell imaging demonstrates that in contrast to small replication sites, large sites do not display MT-motormediated motility. Instead, VV infection induces cellular contractility that facilitates the collection of growing replication sites around the nucleus. In a subset of cells (30-40%), this VV-induced contractility is alternated by phases of directed cell migration, suggesting that the two processes may be linked. The MVA-infected cells do not display contractility or cell migration, supporting the idea that these cellular activities facilitate the efficient accumulation of the VV replication sites around the nucleus. We propose that the recently described cytoskeletal rearrangements induced by VV are a prerequisite for the observed cell contractility and migration activities that apparently contribute to the organization of the complex cytoplasmic life cycle of VV.