It has been proposed that membrane fusion events such as virus-cell fu
sion proceed through a hemifusion intermediate, a state where lipids b
ut not contents of the fusing compartments mix. We engineered the infl
uenza hemagglutinin (HA) such that it would be anchored in membranesvi
a a glycosylphosphatidylinositol (GPI) tail. GPI-anchored HA forms a t
rimer that can bind red blood cells (RBCs) and change conformation und
er fusion-inducing conditions. Using RBCs labeled with fluorescent lip
id or fluorescent soluble content probes, we found that GPI-anchored H
A mediated lipid mixing with similar time course and efficiency as wt-
HA, yet did not mediate transfer of soluble contents. Hence, GPI-ancho
red HA appears to initiate, but not complete, a fusion reaction. We in
terpret our results as evidence for uncoupling a physiological fusion
reaction, for trapping a hemifusion intermediate, and for assigning a
role to a transmembrane domain in a fusion event.