To study the mechanisms involved in membrane fusion, we visualized the fusion process of giant liposomes in real time by optical dark-field microscopy. To induce membrane fusion, we used (i) influenza hemagglutinin peptide (HA), a 20-aa peptide derived from the N-terminal fusion peptide region of the HA2 subunit, and (ii) two synthetic analogue peptides of HA, a negatively (E5) and positively (K5) charged analogue. We were able to visualize membrane fusion caused by E5 or by K5 alone, as well as by the mixture of these two peptides. The HA peptide however, did not induce membrane fusion, even at an acidic pH, which has been described as the optimal condition for the fusion of large unilamellar vesicles. Surprisingly, before membrane fusion, the shrinkage of liposomes was always observed. Our results suggest that a perturbation of lipid bilayers, which probably resulted from alterations in the bending folds of membranes, is a critical factor in fusion efficiency.giant liposome ͉ influenza hemagglutinin ͉ optical microscopy ͉ direct observation M embrane fusion plays an essential role in cellular activities such as exocytosis, endocytosis, and vesicle transport of various cellular organelles. Also, it is an essential step of infection by enveloped viruses. Because the fusion of biological membranes is so important, studies on the mechanism of membrane fusion have been performed. Many of those studies focused on membrane fusion driven by fusogenic peptides, for example, those derived from influenza hemagglutinin. Those studies demonstrated that there are several critical steps required for membrane fusion (1-4). To characterize membrane fusion, resonance-energy transfer and fluorescence-quenching methods have been commonly used (5, 6). The resonance-energy transfer method monitors the mixing of membrane lipids that occurs during fusion, whereas the f luorescence-quenching method monitors the mixing of vesicular contents. Light scattering to detect fusion kinetics, or electron microscopy to visualize the fusion process, has also been commonly used. Electron microscopy can provide snapshots of the changing membrane structures in detail (7), but it is not suitable for studying the time-dependent transition of the three-dimensional membrane morphology.For understanding the fusion mechanism, the real-time visualization of individual fusion processes is indispensable (6, 8). We used giant (1-20-m diameter) liposomes as a model for biological membrane vesicles, and we monitored their fusion process in real time by using optical high-intensity dark-field microscopy (8). Giant liposomes can be visualized by several optical microscopic methods (9-11). However, we have used dark-field microscopy because it gives the best high-contrast images of giant liposomes compared with other methods (8,(12)(13)(14)(15).To induce membrane fusion, we used three different peptides: HA, E5, and K5. HA is a 20-aa peptide derived from the N-terminal fusion peptide region of the influenza hemagglutinin HA2 subunit, whereas E5 and K5 are negativ...