Mode of Membrane Interaction and Fusogenic Properties of a de Novo Transmembrane Model Peptide Depend on the Length of the Hydrophobic Core*

Model peptides composed of alanine and leucine residues are often used to mimic single helical transmembrane domains. Many studies have been carried out to determine how they interact with membranes. However, few studies have investigated their lipid-destabilizing effect. We designed three peptides designated KALRs containing a hydrophobic stretch of 14, 18, or 22 alanines/leucines surrounded by charged amino acids. Molecular modeling simulations in an implicit membrane model as well as attenuated total reflection-Fourier transform infrared analyses show that KALR is a good model of a transmembrane helix. However, tryptophan fluorescence and attenuated total reflection-Fourier transform infrared spectroscopy indicate that the extent of binding and insertion into lipids increases with the length of the peptide hydrophobic core. Although binding can be directly correlated to peptide hydrophobicity, we show that insertion of peptides into a membrane is determined by the length of the peptide hydrophobic core. Functional studies were performed by measuring the ability of peptides to induce lipid mixing and leakage of liposomes. The data reveal that whereas KALR14 does not destabilize liposomal membranes, KALR18 and KALR22 induce 40 and 50% of lipid-mixing, and 65 and 80% of leakage, respectively. These results indicate that a transmembrane model peptide can induce liposome fusion in vitro if it is long enough. The reasons for the link between length and fusogenicity are discussed in relation to studies of transmembrane domains of viral fusion proteins. We propose that fusogenicity depends not only on peptide insertion but also on the ability of peptides to destabilize the two leaflets of the liposome membrane.

Biological membranes are a complex mixture of lipids that contain proteins, hydrocarbons, and other constituents (1). In addition, expression, folding, and insolubility problems can make the study of protein-membrane interactions very complicated and limit the results of interpretation (2,3). Therefore, peptide-membrane interactions are often studied with liposomes of strictly controlled composition (4 -8). Use of such model membrane systems also allows one to control the surrounding medium (4). The same principle has been used by several groups with simplified de novo model peptides composed of typical residues (9 -17). These approaches are useful for analyzing the impact of specific features of peptides and/or membranes on peptide-membrane interactions (12,18).
The effect of hydrophobicity of TM 6 model peptides on their interaction with membranes has been studied by several groups (16,19). A threshold of hydrophobicity is required for peptides to be able to insert into membranes and to adopt a transmembrane orientation (19). For example, peptides with a hydrophobic core composed of 24 alanine residues do not adopt a stable transmembrane orientation in phosphatidylcholine-hydrated membranes (15). Lewis et al. concluded that this is due to insufficient hydrophobicity rather than poor helicity (15). As a result, TM model peptides are usually composed of a mixture of alanine and leucine residues (16, 20 -29). This composition better mimics the mean hydrophobicity of natural TM domains than peptides composed of solely leucine residues (16,30). Almost all de novo TM peptides have aromatic or positively charged residues on either side of the hydrophobic stretch. Lysine and arginine residues are used, since they help to solubilize the peptide, to promote the monomer form, to favor a helical conformation, and to ensure the TM orientation (16,22). Moreover, it was shown that natural transmembrane ␣-helices domains are often flanked by lysine and arginine residues (31,32).
When the length of the hydrophobic stretch of a helical peptide fits the thickness of the membrane hydrophobic core, the peptide inserts into the membrane and adopts a TM orienta-tion (26,33). When it does not fit, either polar residues are exposed to the apolar medium, or hydrophobic residues are accessible to the water phase. This unfavorable phenomenon has been called hydrophobic mismatch (6,16,18,34). Positive and negative hydrophobic mismatches occur when the hydrophobic stretch of a peptide is too long or too short with respect to the membrane thickness, respectively (16). The peptide can adapt its structure or its orientation to overcome mismatch (23,25,35). It can also oligomerize to reduce the mismatch (for a review, see Ref. 16). If the mismatch is too great, insertion of the peptide into the membrane is reduced (16, 34 -37). It has been shown that membrane adaptation can also result as a response to a hydrophobic mismatch. 2 H NMR measurements with lipids containing a perdeuterated acyl chain, x-ray diffraction, and differential scanning calorimetry measurements revealed that a positive mismatch increases the lipid chain order, whereas a negative mismatch increases membrane disorder (38 -41). However, other studies using 15 N as well as 31 P NMR, differential scanning calorimetry, and ESR showed that peptides long enough or too long to span the membrane perturb the membrane order whereas peptides that are too short have less effect (23,24).
Few studies have focused on the ability of TM model peptides to induce membrane fusion. Hofmann et al. (42) showed that the fusogenicity of TM model peptides varies with the ratio of helix-promoting leucine and sheet-promoting valine residues and is enhanced if helix-destabilizing residues, such as glycine and proline, are present within their hydrophobic core. They further showed that fusogenicity of these peptides correlates with structural flexibility (42). The more flexible the peptide is, the more fusion is induced.
In this study, we analyzed the interaction of de novo TM peptides with membranes and in particular their ability to induce fusion. We designed three TM model peptides; all have a hydrophobic core made of alanines and leucines (14,18, and 22 residues), and both the N and C termini were extended with three positively charged residues. Interactions and fusogenic properties of each KALR with a membrane were studied by modeling and experimental approaches. , the hydrated lipid dispersion was exposed to five freeze-thaw cycles and passed 10 times through a polycarbonate membrane (0.1 m) under 20 bars pressure with an extruder (Lipex Biomembranes, Vancouver, Canada). To obtain DOPC/CHOL/DOPE/DOPG/SM (34/33/ 16/10/7 mol/mol/mol/mol/mol) small unilamellar vesicles (SUV), the hydrated film was sonicated (high intensity ultrasonic processor; Sigma) for 5 min at 50 W. A 10,000 ϫ g centrifugation for 2 min (Biofuge Pico, Van der Heyden, Heraeus, Germany) discarded titanium deposit and residual multilamellar vesicles. The phospholipid concentration was determined by phosphorus analysis (43).

EXPERIMENTAL PROCEDURES
Lipid-binding Experiments-In a polar environment, the tryptophan present in KALRs has a fluorescence spectrum with a maximum around 350 nm. Upon the addition of SUVs, the fluorescence maximum shifts to the blue, around 335 nm. The affinity of the peptide for the SUV was determined by adding increasing amounts of vesicles to 0.67 mM peptides dissolved in Me 2 SO, as previously described (44 -46). Fluorescence was recorded at room temperature ( exc , 280 nm; em , 335 nm) after each addition on an LS-50B PerkinElmer Life Sciences fluorimeter. The fluorescence values were then corrected by taking into account the dilution factor corresponding to the addition of the liposomes and by subtracting the corresponding blank (Me 2 SO with the same amount of SUVs). The ratio (R) of bound to total peptides and the dissociation constant (K d ) were calculated as described by Lear and DeGrado (46). Experiments were carried out in a buffer composed of 5 mM HEPES and 100 mM NaCl at pH 7.4.
Lipid Mixing Experiments-Mixing of liposome membranes was followed by measuring the fluorescence increase of octadecylrhodamine chloride B, a lipid soluble probe, after the fusion of labeled and unlabeled liposomes. Labeled liposomes were obtained by incorporating octadecylrhodamine chloride B in the dry lipid film at a 5% concentration of the total lipid weight. Labeled and unlabeled liposomes were mixed at a weight ratio of 1:4, respectively, and at a final concentration of 12.5 M in 5 mM HEPES and 100 mM of NaCl buffer at pH 7.4. 100% of fusion was determined by adding Triton X-100 at 2% to labeled/unlabeled (1:4) LUVs. Fluorescence was recorded at room temperature ( exc , 560 nm; em , 590 nm) on an LS-50B PerkinElmer Life Sciences fluorimeter. The tests were performed with LUVs and were repeated three times with different batches of peptide.
Leakage of Liposome Vesicle Contents-Vesicle leakage was monitored using an assay based on the quenching of HPTS by DPX (47). HPTS and DPX are both encapsulated in the aqueous phase of the same liposomes. Leakage of vesicles was followed by measuring the dequenching of HPTS released into the medium. Fluorescence was recorded at room temperature ( exc , 450 nm; em , 512 nm) on an LS-50B PerkinElmer fluorimeter. Liposomes (LUVs) were prepared as described above in 12.5 mM HPTS (45 mM NaCl), 45 mM DPX (20 mM NaCl), and 10 mM HEPES buffer at pH 7.4. Vesicles containing encapsu-lated HPTS and DPX were eluted in the void volume of a Sephadex G-75 column, with 5 mM HEPES and 100 mM NaCl buffer (pH 7.4). Assays were repeated three times with different batches of peptide.
Electron Microscopy-The effect of the peptides on SUV was examined by negative transmission staining electron microscopy. Prior to staining and fixing, suspensions of SUV at 4.5 mM (with or without lyophilized peptide) were incubated for 5 min at room temperature. A drop containing SUV alone or with a peptide was deposited onto a carbon-coated grid and negatively stained with 2% phosphotungstic acid (pH 6.8). The grids were observed by using a JEOL JEM 100B electron microscope (Japan Electron Optics Laboratory Co., Tokyo, Japan).
IR Spectroscopy Measurements-Spectra were recorded at room temperature on a Brüker Equinox 55 equipped with a liquid nitrogen-cooled mercury-cadmium-telluride detector at a resolution of 2 cm Ϫ1 by averaging 512 scans. Reference spectra of the germanium plate were automatically recorded after purging for 15 min with dry air and subtracted from the recently run sample spectrum. The plate was sealed in a universal sample holder and rehydrated by flushing the holder with N 2 saturated with D 2 O for 3 h at room temperature. Free peptide samples (50 g of peptide in TFE) and the lipid-bound peptides were spread out on a germanium ATR plate (50 ϫ 20 ϫ 2 mm; Aldrich Chimica) with an aperture of 45°, yielding 25 internal reflections. The lipid-membrane sample was prepared by incubation of 1 ml of peptides at 60 M with 6 ml of CHOL/DOPC/ DOPE/DOPG (40/34/16/10, mol/mol/mol/mol/mol) SUV at 500 M (1:50 peptide/lipid (mol/mol) ratio) in 5 mM HEPES and 100 mM NaCl buffer at pH 7.4 for 1 h at room temperature. After incubation, the lipid/peptide mixture was filtered through an anisotropic hydrophilic YM 30 membrane (cut-off 30 kDa) of a Centrifree micropartition system (Amicon) with a 13,000 ϫ g centrifugation for 45 min to separate lipid-associated from free peptides. Distilled water was used to recuperate the retained sample containing the peptide-lipid complex. To determine the peptide insertion rate in membranes by infrared spectra, the area of the amide I peak was divided by the area of the lipid CϭO peak and the number of amide groups in the peptide, as previously shown (35).
To evaluate the orientation of peptides in lipids, spectra were recorded at two orthogonal linear polarizations (90 and 0°) of the incident light. The dichroic spectrum was obtained by subtracting the spectrum recorded with polarized light at 0°from that at 90°. The angle between the germanium crystal and the dipole was calculated from the dichroic ratio R ATR /R ATR ϭ A(90°)/A(0°), where A(90°) is the absorbance of the selected dipole from a spectrum recorded with polarized light at 90°, and A(0°) is the absorbance of the same dipole from a 0°p olarized spectrum. The bands chosen to characterize the protein and phospholipid orientation are the amide I and the lipid s (CH 2 ), respectively (48,49).
Design of KALRs-Three peptides with a hydrophobic core of 14, 18, and 22 residues were designed ( Table 1). The hydrophobic core is an AA(LAAA) n motif with n ϭ 3, 4, and 5 for KALR 14 , KALR 18 , and KALR 22 , respectively. The hydrophobic core lengths were chosen to have variations of approximately one helix turn, corresponding to 6 Å (50). Each peptide is surrounded by three lysine residues at the N-terminal extremity and three arginine residues at the C-terminal extremity. At the center of the sequence of each peptide, a tryptophan residue replaces an alanine residue in order to have a fluorescent sensor.
Analysis by Molecular Modeling-The integral membrane protein and lipid association (IMPALA) method was used to analyze interaction of KALRs with membranes (51). The thickness of the implicit bilayer was extrapolated from experimental data on DOPC bilayers (16,52,53) and the contribution of SM/cholesterol to membrane thickness (54 -57). The thickness of the model membrane was 40 Å with a hydrophobic core of 31 Å to mimic membranes used in the experimental assays.
The IMPALA method simulates the insertion of peptides into the bilayer by adding energy restraint functions to the usual energy function of peptides. The lipid bilayer is defined by C(z), which represents an empirical function describing membrane properties as follows, where z is perpendicular to the membrane and has its origin at the center of the bilayer. The values ␣ and z0 are parameters fixed in such a way that C(͉z Ն 20 Å͉) ϭ 1 and C(͉z Յ 15.5 Å͉) ϭ 0. The value of the function is constant from Ϫ∞ to Ϫ20 Å (hydrophilic phase), from Ϫ15.5 to 15.5 Å (hydrophobic core), and from 20 Å to ∞ (hydrophilic phase).
Two restraints simulate the interaction between the peptide and the bilayer. The first one accounts for the effect that pushes hydrophobic atoms into the membrane (hydrophobic restraint) and hydrophilic atoms outside of it as follows, where N is the total number of atoms, S(i) is the solvent-accessible surface of atom i, Etr(i) is its transfer energy in units of accessible surface area, and C(zi) is the value of C(z) at the position zi of atom i.
The second restraint simulates the perturbation of the bilayer due to the insertion of the molecule (E lip ), where a lip is an empirical factor fixed to 0.018 (51).
Peptides were constructed as ␣-helices using Hyperchem 6.0 (Hypercube, Inc.) assigning values of and angles of Ϫ58 and Ϫ47°, respectively (50). Conformation of backbone and side chains was optimized with Hyperchem by a conjugated gradient procedure using the AMBER 3 force field with termination conditions fixed to a root mean square gradient of 0.01 k cal / (Å⅐mol). A systematic procedure was performed to predict the position of the peptide into the membrane. During this procedure, the peptide systematically moves from 40 to Ϫ40 Å with respect to the membrane center in steps of 1 Å. For each position along the z axis, 5000 random orientations are tested. At the end of the systematic analysis, the position and the orientation with minimal energy among the 405,000 positions tested are selected. The peptide structure was optimized by an angular dynamics procedure taking into account the force field of the membrane as described by Lins et al. (58). Random rotations of 2°maximum and random translations of 0.1 Å maximum of the peptide are allowed. Calculations were performed on an Intel Pentium 4, CPU 3.80 GHz, 4.00 gigabytes of RAM. Graphs were drawn using WinMGM (59).

Study of KALRs in an Implicit
Membrane Model-The IMPALA algorithm, which successfully predicted the interaction of many peptides with membranes (7,8,51,60), was used to predict the interaction of the KALRs with a model membrane mimicking plasma membranes (Fig. 1). Each KALR was folded as a standard helix, and its interaction at all levels of the membrane was analyzed by a systematic procedure. Then peptide was set at its best level of insertion and was left to adapt in the membrane by an angular dynamics procedure.
After angular dynamics, the three peptides remain mainly helical, although some structure adaptations occur to minimize the sum of intramolecular energy plus the IMPALA restraints (data not shown). The root mean square deviation values between the peptide backbone before and after angular dynamics are 1.1, 1.3, and 0.9 Å for KALR 22 , KALR 18 , and KALR 14 , respectively. Structural changes of KALR 22 have a minor effect on the length of the peptide (elongation of 0.5 Å). KALR 18 length increases from 35.9 to 40.8 Å due to the transition from ␣-helix to more extended structures, especially at the N-and C-terminal extremities. Like KALR 22 , KALR 14 undergoes relatively low backbone deformation and no elongation of structure (variation of 0.1 Å).
KALR 22 , KALR 18 , and KALR 14 angles, with respect to the membrane normal, are 10, 15, and 15°before angular dynamics, and 20, 10, and 10°afterward ( Figs. 1 and 2). Thus, the three KALRs adopt a perpendicular orientation. However, the transmembranous status depends on the peptide. KALR 22 spans the entire membrane, similarly folded before and after structure optimization. Its mass center is 2 Å away from the membrane center (Figs. 1 and 2). Its polar extremities are outside the hydrophobic core of the membrane, and its hydrophobic core is inside. Before the angular dynamics, KALR 18 does not traverse the membrane and has its charged N-terminal extremity in the hydrophobic core of the membrane (Fig. 1). Its mass center is 6 Å away from the membrane center (Fig. 2). After the angular dynamics, it traverses the membrane due to peptide elongation. Its two polar extremities are located outside the hydrophobic core of the membrane (Fig. 1), and its mass center is closer to the membrane center (1 Å). Compared with KALR 18 , the position of KALR 14 in the membrane does not  change after the angular dynamics. The peptide mass center is at 10 Å away from the membrane center (Fig. 2). In this position, the N-terminal extremity is exposed to the hydrophobic core of the membrane (Fig. 1). Hence, if all peptides are perpendicular to the membrane and two cross that membrane, one is restricted to one monolayer.
The energy profiles of KALR 22 in the membrane are similar before and after angular dynamics (Fig. 2). The peptide is well inserted into the membrane and does not need structural optimization to locate hydrophobic residues in the membrane core and hydrophilic residues at the interface. The energy profiles also show that the TM position is much more energetically favorable than non-TM positions. The energy profiles of KALR 18 are different before and after angular dynamics (Fig. 2). After optimization, the interaction of KALR 18 with the membrane is energetically more favorable than before, especially for TM positions with the mass centers of the peptide around the center of the membrane (Z ϭ 0 Å). This is due to elongation of the peptide backbone and the consequent migration of the polar N-terminal extremity from the apolar membrane core to polar lipid head groups. The energy profiles of KALR 14 in the membrane are similar before and after angular dynamics (Fig.  2). Since the peptide can only traverse one monolayer, the interaction energy is much less favorable than for longer peptides, and interfacial positions are more stable than positions at the center of the membrane. The ability of KALR 14 to adopt nonperpendicular positions is greater than for the other peptides.
Binding of KALRs to Membranes-The affinity of the three peptides for DOPC/CHOL/DOPE/DOPG/SM (34/33/16/10/7, mol/mol/mol/mol/mol) SUVs was monitored by following the fluorescence of the central tryptophan residue (Table 1). When the peptide interacts with lipids, the tryptophan emission wavelength shifts from 350 to 335 nm (Fig. 3A). The binding of KALRs to membranes was measured at 335 nm upon titration with SUVs. An increase in the fluorescence intensity corresponds to an increase in the amount of membrane-bound peptides. For the three peptides, fluorescence increases when SUVs are added (Fig. 3B). All three peptides bind strongly to membranes, although with different affinities. The partition coefficients calculated from the curves are 7 ϫ 10 5 , 2 ϫ 10 6 , and 4.5 ϫ 10 6 for KALR 14 , KALR 18 , and KALR 22 ,respectively. The data reveal a direct correlation between the peptide length and its binding capability; an increase in the peptide length by 4 residues results in about a 2.5-fold higher partition coefficient.
Structural Properties of KALRs-Fourier transform infrared (FTIR) spectroscopy was used to determine the structure of each KALR in TFE solvent and within membranes (Fig. 4, A and  B). The spectral position of the amide I vibration, between 1615.5 and 1694.5 cm Ϫ1 , was used to characterize the conformation of the peptide backbone (48). No significant difference was observed between the three peptides in the different media (Fig. 4). Indeed, the bands are centered around 1655 cm Ϫ1 , indicating that they are mainly helical in TFE as well as in phospholipids (Figs. 4, A and B).
As shown in Fig. 4B, the intensity of the amide I absorption band is sensitive to the length of the peptide when inserted into lipids. A decrease in the band intensity is observed when the length of the peptide decreases, with a considerable decrease for the shortest peptide. Since the buffer gives no significant FTIR signal (data not shown), these data indicate that the longest peptide is the most inserted into the bilayer. From the FTIR spectra, we evaluated the relative insertion of each peptide into membranes (35). If we consider that the KALR 22 peptide has an insertion rate of 100, KALR 18 and KALR 14 have a relative rate of 90 and 40, respectively. In summary, the FTIR analysis reveals that KALR inserts into membranes as an ␣-helix conformation and that the amount of inserted peptides depends on its length. Orientation of KALRs in the Bilayer-The orientation of peptides in lipid bilayers was calculated by recording the spectra obtained with polarized IR lights parallel and perpendicular with respect to the incidence plane. The amide I transition moment, around 1655 cm Ϫ1 , is parallel to the helix axis, whereas the transition moment for s(CH 2 ), around 2853 cm Ϫ1 , is perpendicular to the lipid chain axis. These bands were used to evaluate the orientation of the peptide helix axis and the lipid chain axis with respect to the germanium plate by calculating their dichroic ratios (48,49). The R ATR of the lipid s(CH 2 ) is 1.43 for all peptides. This corresponds to an angle of 41 Ϯ 5°b etween the acyl chain carbons and the normal to the germanium crystal (  Table 2). ATR-FTIR spectra thus indicate that KALRs are perpendicular helices in membranes.
Studying the Fusogenic Properties of KALRs-We first studied the fusogenic capacity of peptides by measuring their ability to induce lipid mixing of liposomes (Fig. 5, A and B). KALR 14 induces no significant liposome lipid mixing, even after 10 min of incubation. The longer peptides, KALR 18 and KALR 22 , induce lipid mixing of liposomes. The lipid mixing increases rapidly following the addition of KALR 22 and reaches a maximum after only 60 s. For KALR 18 , the maximum is reached after 240 s. Analysis of lipid mixing at different peptide/lipid ratios confirmed that the smallest peptide has no significant effect as compared with the longer peptides and that KALR 22 induces more lipid mixing than KALR 18 irrespective of the peptide/lipid ratio (Fig. 5B). The lipid mixing induced by KALR 22 and KALR 18 increases with the peptide/lipid ratio and is maximal at a ratio of 0.04. Above this ratio, lipid mixing decreases. KALR 22 induces a maximal lipid mixing of about 50% of the liposomes.
To further investigate the membrane-destabilizing effect of KALRs, we carried out leakage assays. The addition of KALR 14 to membrane liposomes does not increase leakage, whereas the addition of KALR 18 or KALR 22 does (Fig. 5C). Similar to the lipid mixing assays, the process is dose-dependent with a maximum response for a peptide/lipid ratio of 0.04. Above this ratio, leakage decreases. KALR 22 also induces more leakage than KALR 18 irrespective of the peptide/lipid ratio and the incubation time (Fig. 5, C and D). KALR 22 induces a maximal leakage of about 80% of the liposomes.
To analyze the effect of KALRs on membranes, we observed LUVs by negative stain electron microscopy before and after the addition of peptides (peptide/lipid ratio ϭ 0.04). Micrographs show that KALR 14 has no effect on LUVs (Fig. 6). The increase in vesicle size in the presence of KALR 18 demonstrates fusion. KALR 22 also increases vesicle size but to a larger extent.

DISCUSSION
In this paper, we examined the ability of hydrophobic TM model peptides with three hydrophobic core lengths to induce in vitro membrane fusion and its correlation with membrane binding, membrane insertion, and the ability to induce membrane destabilization. These peptides, called KALRs, are composed of hydrophobic residues (Ala and Leu residues) sur-  rounded by charged residues (Lys and Arg residues). The hydrophobic part is a repeat of 3, 4, and 5 AAAL motifs that result in a difference of approximately one turn of helix between peptides. Effect of the Hydrophobic Core Length of KALR on Insertion into Membranes-Lipid-binding assays and infrared spectroscopy show that the ability of KALRs to bind to lipids and to be incorporated into membranes increases with the length of the hydrophobic core of the peptides. These results could simply be explained by the hydrophobicity of peptides. Indeed, the addition of one AAAL motif increases the peptide and the helix structure's hydrophobicity (as seen in Fig. 2), and as a result, affinity for lipids is multiplied by 2.5, as measured in the lipidbinding assays. However, the insertion rate of KALRs into membranes calculated from FTIR spectra is not linearly related to the addition of AAAL motifs. Indeed, the insertion rate of KALR 14 is 40% that of KALR 22 , whereas the insertion rate of KALR 18 is 90%. This indicates that insertion of KALR peptides into lipids is not simply governed by hydrophobicity.
The hydrocarbon thickness of a fully hydrated DOPC bilayer is 26.8 -28.6 Å (16, 52, 53, 61-63). The addition of SM and cholesterol increases the thickness of the hydrophobic core of monounsaturated membranes by 2-4.1 Å (54 -57). It was also shown that in some cases, SM and cholesterol can induce the formation of more rigid lipidic domains 3-10 Å thicker than the rest of the membrane (56,64,65). We can conclude from these data that at least a large part of our DOPC membrane has a hydrophobic thickness between 28.8 and 32.7 Å. Thicker rigid domains could be present, but this cannot be assumed with certainty. That is why the membrane hydrophobic core is 31 Å thick in the molecular modeling simulation.
The first result of molecular modeling simulations is that all KALR helical peptides adopt a perpendicular orientation in membranes. This prediction is in accordance with the FTIR data of this study and with previous studies showing that Ala/ Leu peptides of different lengths are transmembrane helices in membranes (11,16,17,30,35).
The second result of molecular modeling simulations is that the compatibility between membranes and peptides depends on the peptide length. KALR 22 matches the membrane thickness; it inserts the hydrophobic residues inside the membrane hydrophobic core and allows the hydrophilic residues to protrude from the membrane. In contrast, KALR 14 is too short to traverse the membrane even with structural adaptations; the N terminus is located in the membrane core, which is energetically unfavorable. This situation corresponds to a negative hydrophobic mismatch. In between, KALR 18 is too short to span the membrane as a standard ␣-helix, but structural optimization elongates the helix and suppresses this incompatibility. These results are consistent with previous studies of peptide-membrane hydrophobic mismatch. Small distortions of ␣-helix structure can occur at the extremities of a TM model peptide too short to span a membrane (16,66). To reduce the mismatch, other adaptations, such as peptide oligomerization or lipid length reduction, can occur (for a review, see Ref. 16). However, when the mismatch is too great, the peptide is excluded from the membrane (23,34). This would explain why the KALR 14 is much less inserted into the membrane than KALR 18 and KALR 22 .  In conclusion, peptide-membrane interactions depend on peptide hydrophobicity and on the matching of peptide length with membrane thickness. Peptide hydrophobicity should mainly control its ability to bind to lipids. In addition, the peptide's length should affect the peptide's ability to insert into a membrane. KALR 18 should have reduced binding and insertion with respect to KALR 22 because of a lower hydrophobicity and of a small hydrophobic mismatch. The lower hydrophobicity and the large hydrophobic mismatch of KALR 14 should decrease its insertion into membranes.
Furthermore, this work serves as another piece of evidence that IMPALA can be used to predict how a peptide will interact with a membrane. IMPALA can predict the insertion and orientation, the impact on the peptide structure, and the degree of hydrophobic mismatch.
Effect of the Hydrophobic Core Length of KALR on Fusogenicity-Previous studies showed that peptides corresponding to the TM domain of the vesicular stomatitis virus G protein and of SNARE induce in vitro membrane fusion (67)(68)(69). It was proposed that the fusogenicity of TM domains depends on structural flexibility (67)(68)(69). The relationship between TM peptide fusogenicity and structural flexibility was further shown by using a TM model peptide with variable proportions of ␣-helix/␤-sheet-promoting residues (42). Here we showed that a TM model peptide can induce membrane fusion in vitro even if it contains only alanine and leucine ␣-helixpromoting residues (70 -72) as long as the length of the hydrophobic core is sufficient.
Our study goes in the same direction as the latter studies; the longer the peptide, the more it induces fusion in vitro. This could first be related to the number of membrane-inserted peptides. Destabilization of a membrane can be greater if more peptides insert into it (23,24). Membrane destabilization would then increase with KALR length, because the quantity of inserted peptides increases with length. However, KALR 14 does not induce membrane fusion whenever its insertion rate is 40% of the KALR 22 rate. Hence, fusogenicity must be controlled by additional parameters whose insertion rate is also supported by the fact that many TM peptides are not fusogenic.
Molecular modeling results suggest that one of the parameters regulating fusogenicity might be peptide length. Indeed, KALR 14 spans only one monolayer of the membrane, whereas KALR 18 and KALR 22 can traverse the bilayer. It was previously shown that hemagglutinin of the influenza A virus needs to have a TM domain of at least 17 residues to efficiently promote full fusion (81). When the TM domain has only 16 or fewer residues, the hemagglutinin cannot promote full fusion (81).
Interestingly, the minimal length for fusogenicity is in the same range for different peptides: 17 residues for HA2 and between 14 and 18 residues for the KALR peptide. An effect of the length of the TM domain on the fusion process was also described previously for other viruses, such as human immunodeficiency virus (82), Foamy virus (83), murine coronavirus (84), and murine leukemia virus (85), and for SNARE protein (86,87). The capacity to span the membrane might be an important parameter for inducing fusion.