Lipid bilayers repel each other through a so-called “hydration force” when they approach each other more closely than about 15–20 Å (4). To overcome this barrier, fusion requires a counteracting force, either from a bulk effect, such as pressure by polymer exclusion, or from a localized fusion facilitator, such as the presynaptic fusion complexes Gipson et al.(3) describe. Computational studies have suggested that formation of a localized contact would minimize the hydration force by minimizing the area of closely apposed membranes, and hence reduce the kinetic barrier, leading to a “hemifusion stalk” as a likely fusion intermediate (5). The cryo-ET images of Gipson et al.(3) are in accord with those proposals. The fusogens for synaptic-vesicle fusion are the “neuronal SNAREs”: synaptobrevin-2 (also known as VAMP2), anchored in the membrane of the synaptic vesicle; syntaxin-1, anchored in the plasma membrane; and SNAP-25, associated with the plasma membrane by palmitoylated cysteine residues. The Ca2+ sensor, synaptotagmin-1, anchored in the vesicle, completes a minimal, Ca2+-triggered fusion system, but addition of two other regulatory factors (complexin and

Munc13) change both the speed and yield of the response and, as shown in by Gipson et al.(3), the range of morphologies present. Like medical tomography, cryo-ET reconstitutes a 3D image of a single entity from a large number of projected views, related by tilting about a fixed axis (6).(In medical tomography, the X-ray beam rotates about the subject; in EM tomography, the stage bearing the sample tilts in a fixed beam.) Because electron damage severely limits the exposure, each tilt image is very noisy, and the achievable resolution is substantially more modest than can be reached with very large numbers of identical molecular objects. One way to enhance the signal-to-noise in cryo-ET is to use phasecontrast imaging, now made relatively practical by the Volta phase plates that Danev et al.[coauthors of the Gipson et al.(3) paper] have shown to be effective (7, 8). Biological samples are essentially pure phase objects; at perfect focus there is no detectable image. In the absence of a phase plate—the electron-beam equivalent of a phase plate for an optical phase-contrast microscope—the standard method for obtaining an interpretable image is careful underfocusing, with subsequent computational correction to restore uniform contrast. The phase plate retains important information lost by the underfocusing approach, while avoiding potential computational problems that result from variation in defocus across a titled sample. Gipson et al.(3) describe the contacts they see as “point contacts”—with a thin bridge of density (interpreted as the assembled protein fusion machinery) between the liposomes bearing vesicle proteins (synaptobrevin-2 and synaptotagmin) and those bearing plasma-membrane proteins (syntaxin-1 and SNAP-25)—and “long contacts,” with a laterally more extended region of density between the two liposomes. Addition of both regulatory proteins restricted the contacts largely to the former class and a spacing of∼ 100 Å between the centers of the two bilayers (ie,> 50 Å between their apposed surfaces). Moreover, after addition to Ca2+, which induces fusion on a millisecond timescale when these proteins are present, nearly all of the point contacts had disappeared, leaving only a few long contacts and