In a previous work, we observed that different basic peptides were able to induce membrane invaginations in giant vesicles, a process suggesting “physical endocytosis” mechanism for the basic peptide uptake 
. However, the role of lipids in this phenomenon was not studied. In the present paper we investigated the influence of the membrane lipid composition in the changes induced by the homeodomain-derived peptide Penetratin.
Several studies suggest that increase in negatively charged phospholipid concentrations favours Penetratin and other basic peptide sequences binding to membranes by electrostatic adsorption. According to these authors, neutral phospholipids (PC) are unable to bind the peptide or to aggregate membranes 
. This discrepancy with our data showing that pure zwitterionic interfaces bind Penetratin can be explained taking into account that in some of the previous studies, the authors used Small Unilamellar Vesicles (SUV) with a highly positive curvature of membranes. We assume that the spreading of neutral phospholipid headgroups could result in the reduction of peptide binding. We have presently used weakly curved LUV and GUV because they have a radius of curvature close to those of cell membranes. In agreement, other studies have demonstrated that basic peptides 
and Penetratin are able to bind PC monolayers 
and flat bicelles 
. Our data indicate that Penetratin binding to membrane bilayers does not depend on the liquid ordered or liquid disordered physical state of the membrane because PC as well as SM/Chol (1/1) vesicles bind Penetratin. The addition of PG (10%) to the membranes increases to four fold the binding. Altogether the data are consistent with the ability of Penetratin to bind lipid membrane possibly by electrostatic interaction of basic residues with anionic headgroups or by forming hydrogen bonds. Guanidinium groups of arginine and the phosphate group of phospholipids are susceptible to act as donor and acceptor in the formation of hydrogen bond at neutral pH 
. Interestingly, in the case of poly-L-Lysine, a molecule without the guanidinium groups of arginine, the formation of tubes is abolished in GUV lacking anionic phospholipids 
Penetratin ability to induce vesicles aggregation by bridging adjacent membranes was correlated with binding 
. Bridging increases after surface membrane saturation by the charged peptide. Modelisation studies suggest that the amphipathic fusion peptide E5 of influenza
virus binds strongly and at a deeper distance from the aqueous interface in fluid dimirystoylphosphatidylcholine membranes compared to a rigid dipalmytoylphosphatidylcholine membrane 
. Here, we observe that membrane fluidity also favours bridging. This may also indicate that mobility of the peptide on the fluid surface favours the arrangement required to form inter-membrane bridges.
Membrane fluidity was critical in the occurrence of membrane deformations after CPP binding. In a membrane comprised of fluid disordered phase (Ld) such as expected for unsaturated PC around physiological temperatures, Penetratin was able to induce invaginations in giant vesicles (PC, DOPC and PC/PG (9/1)). On the contrary, the peptide showed no effect on a raft-like membrane (SM/Chol (1/1)) indicating that the rigidity of the membrane comprised with the ordered phase has restrained considerably tubulation. Roux et al. 
showed that the force required to form tubes from a liquid ordered membrane is 1.7 times stronger than for a liquid disordered membrane. Penetratin interaction energy is probably not strong enough to deform the ordered rigid domains.
It is noticeable that peptide binding is increased after the addition of PG in the ordered membranes. This induces peptide aggregation and an accumulation of small dimension vesicles (grapes). We have interpreted the occurrence of these grape-like vesicular structures by the peptide inefficiency to support the “normal” elongation of the highly curved thin tubes. Interruption of the tube elongation by membrane rigidity surrounding the tubulation starting point may eventually result in grape vesiculation. The resistance of the raft-like membranes to deformation by the peptide was confirmed by cryo-electron microscopy of LUV in which only the liquid disordered membranes undergo deformations in the peptide-membrane bridges.
The structure of membranes in the presence of Penetratin was also studied by 31
P-NMR and X-ray diffraction. Both methods revealed that the liquid ordered phases comprised of SM and Chol are resistant to membrane deformation by the peptide. On the contrary, both methods revealed perturbation of the lipid arrangement in the disordered phase. 31
P-NMR spectra showed a strong isotropic peak consistent with highly curved membrane structures. A possible explanation for the isotropic signal is the fast tumbling of small vesicles detached from the large MLV. Other possible explanations such as the transition of lamellar to cubic arrangement have been turned down by X-ray examination. Vesicle formation is also suggested by the X-ray diffractograms. The X-ray data showed a decrease in the Bragg's reflections corresponding to the lamellar arrangement. Due to vesiculation, the Bragg-peaks diminish and the typical form-factor contribution arising from uncorrelated bilayers becomes visible as the structure factor contribution of constructive interference Bragg's diffraction peaks decreases. It is also shown that the peptide separates partially the phospholipids into two lamellar arrangements with wide distance repeats as regard to the records in the absence of peptide. We hypothesize that the arrangement with the longer spacing corresponds to the increased inter-bilayer distance consistent with the measurement given by cryo-electron microscopy. Indeed, the distance repeat of the lamellar membrane perturbed by Penetratin is 7.05 nm as measured by X-ray diffraction, a value in agreement with the thickness of the lipid bilayer plus the electron dense peptide layer observed by cryo-electron microscopy (~7 nm). However, the Bragg peaks with the repeat distance of 8.54 nm would represent another lamellar arrangement which is probably related to the vesiculation intermediate arrangement. It might be interesting to speculate whether this thicker structure is related to the rod-like structures observed in membranes incubated with Tat peptide 
. However, neither a cubic nor an inverted hexagonal phase has been revealed by the present X-ray examination. The enlarged inter-chain distance which was detected by wide angle X-ray diffraction in the presence of Penetratin may be a clue to explain the mechanism of tube or vesicle formation. The enlarged inter-acyl chain distance is in agreement with the hypothesis of a negative curvature induced by the peptide as suggested previously 
, but it might also results from phase separation induced by the formation of peptide-lipid clusters.
In conclusion, we show that the homeodomain-derived basic peptide Penetratin is able to bind several types of membranes (ordered or disordered) but can only induce tubulation (“physical endocytosis”) in liquid disordered membranes. This can be seen in the absence of negatively charged phospholipids. This is relevant to eukaryotic cells where the external layer of the plasma membrane does not contain significant amounts of anionic phospholipids in the resting state. Interaction of Penetratin with the phosphate group of PC or SM of cell plasma membrane would be efficient to ensure the proper binding. Second, the tubulation effect of Penetratin on phospholipid membranes is only possible on membranes in the liquid disordered phase. We assume that this effect is due to the capability of the peptide to induce negative curvature in membranes. No tubule was observed in raft-like liquid ordered membranes exposed to Penetratin. The data suggest that the formation of lipid-peptide complexes, which requires fluidity, is critical and that the compactness of the raft-like domains is a barrier for cell penetration. Fretz has recently shown that perturbation of cell membrane domains by cholesterol depletion with methyl-beta-cyclodextrin increases polyarginine (R8) uptake independently of endocytosis 
possibly after the transition of the liquid ordered to disordered arrangement. Our data suggest that for the biological processes involving messenger proteins containing protein transduction domains (i.e. TAT and homeoproteins) as well as for therapeutic molecular vectors, the preferential cellular membrane target for penetration would be the non-raft fluid plasma membrane domains. In this case, the formation of invaginations: tubes in liquid disordered domains and vesiculation in mixed ordered/disordered domains could explain the metabolic energy independent mechanism of internalization. Experiments with messenger proteins and peptides on cell membranes are the perspectives for the future research.