Architecture of the influenza hemagglutinin membrane fusion site

The mechanism of influenza hemagglutinin (HA) mediated membrane fusion has been intensively studied for over 20 years after the bromelain-released ectodomain of HA at neutral pH was first crystallized. Nearly 10 years ago, the low-pH-induced "spring coiled" conformational change of HA was predicted from peptide chemistry and confirmed by crystallography. Other work has yielded a wealth of knowledge on the observed changes in HA fusion/hemifusion phenotypes as a function of site-specific mutations of HA, or added amphipathic molecules or particular IgGs. It is becoming clear that the conformational changes predicted by the crystallography are necessary to cause fusion and that interfering with these changes can block fusion or reduce it to hemifusion. What is not known is how the conformational changes cause fusion. In particular, while it is generally agreed that fusion requires an aggregate of HAs, how the aggregate may act to transduce the energy of the HA conformational changes to creating the initial fusion defect is not known. We have used a comprehensive mass action kinetic model of HA-mediated fusion to carry out a "meta-analysis" of several key data sets, using HA-expressing cells and using virions. The consensus result of these detailed kinetic studies was that the fusion site of influenza hemagglutinin (HA) is an aggregate with at least eight HAs. The high-energy conformational change of only two of these HAs within the aggregate permits the formation of the first fusion pore. This "8 and 2" result was required to best fit all the data. We review these studies and how this kinetic result can guide and constrain HA fusion models. The kinetic analysis suggests that the sequence of fusion intermediates starts with protein control and ends with lipid control, which makes sense. While curvature intermediates, e.g. the lipid stalk, are almost certainly within the fusion sequence, the "8 and 2" result does not suggest that they are the first step after HA aggregation. The stabilized hydrophobic defect model we have proposed as a precursor to the lipid stalk can form and is consistent with the "8 and 2" result.