**4. Discussion**

Our simulations indicate the highly dynamic nature of HAfp, which includes interconversions between open and closed surface conformations, as well as transient excursions into the membrane core. The lack of notable differences in the level of membrane perturbation by surface configurations of active and inactive mutants combined with apparently greater ability of the active ones to adopt transmembrane placement with highly membrane disruptive potential indicate that peptides' ability to dive into the membrane core may be the key element of their fusogenic function.

Indeed, experimentally observed activity of the three considered HAfp versions appears to correlate with the stability and insertion depth of their membrane-spanning configurations. According to our results, the actual HAfp fusogenic mechanism is related to partial burial of positively charged N-terminal amino group whose interaction with membrane phosphates promotes lipids tilting and, eventually, acyl tails protrusions. While such role of the N-terminal charge has already been postulated, its burial was ascribed to oblique peptide insertion whose existence is not supported by our study. Instead, we advocate that the experimental measurements consistent with such oblique insertion reflected an average over two, distinct states corresponding to in-plane and transmembrane orientations. While at first sight the relatively short length of HAfp helical hairpin arms

appears incompatible with transmembrane configurations, in fact, it is well suited to provide for efficient burial of the N-terminal amino group within bilayer core. An extension of the N-helix would lead to surfacing of the N-terminus and a stable, ye<sup>t</sup> benign membranespanning structure, typical for integral membrane proteins. Instead, the necessity to maintain a metastable deep configuration by keeping balanced interaction with aqueous compartment on both membrane sides imposes strict requirements for the kink region that likely explain the conservation of residues that support tight packing of both helices, such as the glycine ridge, as well as the existence of the essential W14, whose role in membrane penetration and water attraction was highlighted by our simulations.

The free energy difference between surface and deep configurations resulting from our simulations is large and suggests that the fraction of deeply inserted peptides may be rather small. We note, however, that our free energy calculations likely provide an upper limit to the actual value due to relatively small size of the membrane patch considered in our model. In reality, deep peptide insertions may be happening opportunistically supported by local membrane fluctuations. However, even rare such events ye<sup>t</sup> of relatively high fusogenic potential are enough to explain the course of experimentally observed vesicle fusion, whose time scale extends to minutes.
