Membrane fusion occurs when SNAREpins fold up between lipid bilayers. How much energy is generated during SNAREpin folding and how this energy is coupled to the fusion of apposing membranes is unknown. We have used a surface forces apparatus to determine the energetics and dynamics of SNAREpin formation and characterize the different intermediate structures sampled by cognate SNAREs in the course of their assembly. The interaction energy-versus-distance profiles of assembling SNAREpins reveal that SNARE motifs begin to interact when the membranes are 8 nm apart. Even after very close approach of the bilayers (B2-4 nm), the SNAREpins remain partly unstructured in their membrane-proximal region. The energy stabilizing a single SNAREpin in this configuration (35 k B T) corresponds closely with the energy needed to fuse outer but not inner leaflets (hemifusion) of pure lipid bilayers (40-50 k B T). Intercellular communication and intracellular protein transport rely upon the fusion of cargo-containing vesicles with target membranes. As lipid bilayers are inherently stable, such fusion events are energetically costly and require specialized fusion proteins that harvest the energy made available during their own binding and folding to drive membrane disruption and merging 1-5. In neuronal synapses, the core of the fusion machinery consists of three proteins from the SNARE family: the synaptic vesicle (v)-SNARE protein VAMP-2 and the two target plasma membrane (t)-SNARE proteins syntaxin-1A and SNAP-25 (refs. 6-8). When separately reconstituted into synthetic liposomes or ectopically expressed on the surfaces of cells, neuronal v-and t-SNARE proteins are sufficient to drive membrane fusion through their assembly in the form of SNAREpins 2,9. The interacting domains of SNARE proteins (SNARE motifs) contain 60-70 amino acid residues; they are mostly unstruc-tured as monomers 10-12 and assemble in solution into a highly stable heterotrimer consisting of four a-helices aligned in parallel, with VAMP-2 and syntaxin-1A each contributing one helix and SNAP-25 contributing two helices 13,14. In the context of lipid bilayers, the assembly of SNAREs starts at their membrane-distal N termini and proceeds toward their membrane-proximal C termini (zipper model), a process that also includes passage through a stable intermediate binding state 15-21. This zipper-like assembly progressively brings the membranes into close apposition and creates a tight bridge between them that triggers lipid bilayer fusion. Progressive assembly of SNAREs may culminate in a release of energy sufficient to drive membrane merging. Alternatively, the assembling SNAREs may pass through a series of intermediates, each of which contributes enough energy for advancement through the successive stages of membrane fusion. Characterization of these inter-mediates requires the capacity to measure the interactions between membrane-associated proteins at nanometer distance resolutions. Thermodynamic and atomic force microscopy (AFM) measurements have successfully described the kinetics of SNARE assembly and disassembly in solution 22 and the rupture forces of SNARE complexes affixed to solid supports 23,24. However, none of these studies has been able to offer information about the dynamics and energetics of SNAREpin folding, including conformational changes and distance-energy correlations during SNARE assembly. Furthermore, the previous experiments were not performed in the context of lipid bilayers, preventing any investigation of the interplay of lipids and SNARE proteins in membrane interaction and fusion. Here, we have investigated these questions using the surface forces apparatus (SFA), which directly measures the interaction energy between two facing functionalized membranes as a function of their separation distance and makes it possible to identify molecular rearrangements of interacting species during their association 25. Direct measurements of force versus distance between membrane-embedded neuronal SNARE proteins (derived from mouse and rat) allow us to explore in real time the molecular details of SNAREpin formation across two lipid bilayers, including conformational changes, kinetics of association, binding energy and extent of assembly. RESULTS Interactions between SNAREs in apposing bilayers Force was measured between two mica-supported lipid bilayers reconstituted with the neuronal cognate t-and v-SNARE proteins (Fig. 1). The surfaces were approached toward each other and then