Epsin N-terminal Homology Domain (ENTH) Activity as a Function of Membrane Tension*

The epsin N-terminal homology domain (ENTH) is a major player in clathrin-mediated endocytosis. To investigate the influence of initial membrane tension on ENTH binding and activity, we established a bilayer system based on adhered giant unilamellar vesicles (GUVs) to be able to control and adjust the membrane tension σ covering a broad regime. The shape of each individual adhered GUV as well as its adhesion area was monitored by spinning disc confocal laser microscopy. Control of σ in a range of 0.08-1.02 mN/m was achieved by altering the Mg2+ concentration in solution, which changes the surface adhesion energy per unit area of the GUVs. Specific binding of ENTH to phosphatidylinositol 4,5-bisphosphate leads to a substantial increase in adhesion area of the sessile GUV. At low tension (<0.1 mN/m) binding of ENTH can induce tubular structures, whereas at higher membrane tension the ENTH interaction deflates the sessile GUV and thereby increases the adhesion area. The increase in adhesion area is mainly attributed to a decrease in the area compressibility modulus KA. We propose that the insertion of the ENTH helix-0 into the membrane is largely responsible for the observed decrease in KA, which is supported by the observation that the mutant ENTH L6E shows a reduced increase in adhesion area. These results demonstrate that even in the absence of tubule formation, the area compressibility modulus and, as such, the bending rigidity of the membrane is considerably reduced upon ENTH binding. This renders membrane bending and tubule formation energetically less costly.

The epsin N-terminal homology domain (ENTH) is a major player in clathrin-mediated endocytosis. To investigate the influence of initial membrane tension on ENTH binding and activity, we established a bilayer system based on adhered giant unilamellar vesicles (GUVs) to be able to control and adjust the membrane tension covering a broad regime. The shape of each individual adhered GUV as well as its adhesion area was monitored by spinning disc confocal laser microscopy. Control of in a range of 0.08-1.02 mN/m was achieved by altering the Mg 2؉ concentration in solution, which changes the surface adhesion energy per unit area of the GUVs. Specific binding of ENTH to phosphatidylinositol 4,5-bisphosphate leads to a substantial increase in adhesion area of the sessile GUV. At low tension (<0.1 mN/m) binding of ENTH can induce tubular structures, whereas at higher membrane tension the ENTH interaction deflates the sessile GUV and thereby increases the adhesion area. The increase in adhesion area is mainly attributed to a decrease in the area compressibility modulus K A . We propose that the insertion of the ENTH helix-0 into the membrane is largely responsible for the observed decrease in K A , which is supported by the observation that the mutant ENTH L6E shows a reduced increase in adhesion area. These results demonstrate that even in the absence of tubule formation, the area compressibility modulus and, as such, the bending rigidity of the membrane is considerably reduced upon ENTH binding. This renders membrane bending and tubule formation energetically less costly.
Clathrin-mediated endocytosis is one of the key metabolic pathways for the uptake of macromolecules into eukaryotic cells (1)(2)(3)(4). Driven by a chain of remodeling events and an elaborate set of proteins acting in an orchestrated manner, an almost flat patch of plasma membrane is transformed into a closed, cargo-containing vesicle. As plasma membrane shape transformation is associated with significant local bending of the membrane, the process is highly sensitive to lateral membrane tension. Plasma membrane tension originates from two primary sources; that is, hydrostatic pressure across the lipid bilayer and cytoskeleton-membrane adhesion (5). Depending on the cell type, plasma membrane tensions span a range of roughly 0.003-0.45 mN/m (5)(6)(7)(8). Even though it had become clear already in the late 1990s that tension plays a role in exoand endocytosis (9,10), only in recent years has significant evidence been accumulated that membrane tension is of utmost importance for processes that rely on membrane remodeling (11)(12)(13)(14). Cells actively maintain and regulate their membrane tension and use it to control exo-and endocytosis (15). In K562 cells it has been reported that endocytosis is completely suppressed under hypoosmotic conditions (16). Generally, high lateral membrane tension suppresses membrane deformation as both stretching the lipid bilayer and opening bonds between the cytoskeleton and the plasma membrane requires a large amount of energy (17).
To control lateral membrane tension in an artificial well defined membrane system, we made use of giant unilamellar vesicles (GUVs) that adhere to an avidin-coated glass substrate. By adjusting the Mg 2ϩ concentration in solution, we were able to control and tune the lateral membrane tension of these GUVs in a range of 0.08 -1.02 mN/m. These values are well below the tension value of 2 mN/m reported in the previous study (27) and might allow monitoring ENTH-induced membrane changes including tubulation or vesiculation. We show that with this system in hand, binding and helix-0 insertion of ENTH can be monitored in a label-free fashion by determining the increase in adhesion area of the GUV. From our observations, we conclude that ENTH binding results in a substantial softening of the lipid bilayer, which can eventually result in tubule formation at low initial lateral tension.

Results and Discussion
Tuning the Adhesion of GUVs by Mg 2ϩ -Fluorescently labeled GUVs composed of DOPC/DOPE/Atto488 DPPE 3 were doped with N-(cap-biotinyl)-DOPE to attach them onto the avidin-coated glass surface (28). For specific binding of ENTH to GUVs, the natural receptor lipid PtdIns(4,5)P 2 with a concentration of 0.8 mol% was supplemented. To obtain the shape of the vesicles, they were imaged by spinning disc confocal laser microscopy (SDCLM) (Fig. 1, bottom images). Attractive forces between the GUV and the avidin-coated surface change its shape from a sphere with a radius R v to a spherical cap with a slightly increased radius of the cap denoted as R ad (Fig. 1, top images). The spherical cap geometry is indicative of the strong adhesion regime (29,30). In the limit of strong adhesion the bending energy associated with changes in the shape of the vesicle can be ignored, and the vesicle can be described as a spherical cap usually parameterized solely by the contact angle ⍜ with the surface (30,31). The adhered GUV forms a circular contact area with the functionalized glass substrate with a radius R i (Fig. 1, top image) (32). This adhesion area increases until the free energy of the vesicle is minimized. The elastic energy encompasses stretching energy, bending energy, and the entropic cost of suppressing thermal fluctuations of the contact area between the GUV and the substrate (33). Adhesion energy is obtained from the surface energy times the contact area. In the strong adhesion limit, we can safely assume that the free energy F v comprises only an elastic stretching term and an adhesion term (Equation 1), where K A is the area compressibility modulus of the bilayer, A ad is the area of the adhered vesicle, A v is the area of the free vesicle, A i is the contact area, and ␥ ad is the adhesion energy per unit area mainly originating from bond formation between a GUV and the substrate. Minimization of the free energy assuming constant volume requires the shell to be laterally dilated to allow generation of a finite-sized adhesion zone (31). Naturally, stretching of the bilayer inevitably generates lateral tension. Therefore, controlling the adhesion area permits to adjust membrane tension.
To modify the contact area of a GUV, we made use of the fact that divalent cations increase the membrane-surface interaction, i.e. it was found that a strong interaction between phosphatidylcholine and Mg 2ϩ promotes spreading of vesicles to planar supported bilayers (34 -36). To adjust the Mg 2ϩ concentration, GUVs formed in sucrose solution (298 mosmol/kg) were added to a sucrose-containing buffer solution (2 mM HEPES, sucrose, pH 7.4, 298 mosmol/kg) doped with different MgCl 2 concentrations. Isoosmolar conditions are required as the size of the contact area is also a function of the GUV excess area, which depends on the osmolality gradient. GUVs were allowed to bind to the avidin-coated surface and were imaged by SDCLM. Fig. 2 illustrates the increase in contact area of GUVs doped with 0.8 mol% PtdIns(4,5)P 2 as a function of Mg 2ϩ concentration. To compare GUVs of different sizes, the radius R i of the contact area was related to the radius R ad of the adhered GUV (Fig. 2). This allowed us to normalize the adhesion area to the size of the GUV and compare different ensembles. The increase of R i /R ad with Mg 2ϩ concentration (Fig. 3) suggests that Mg 2ϩ promotes the adhesion of the GUVs by interaction of the lipid head groups of the membrane with the functionalized glass surface (34,37). In addition to this interaction, it is also conceivable that Mg 2ϩ influences the biotin-avidin-interaction itself. Holmberg et al. (38) showed that the dissociation of a biotin-streptavidin bond strongly depends on the ionic strength and is almost fully prevented in the presence of Mg 2ϩ . In essence, both effects result in a larger contact area with increasing Mg 2ϩ concentration.
Estimating Lateral Membrane Tension from the Contact Area-The observed contact area of the GUV with the glass surface provides information about the lateral tension of the GUV membrane. Assuming constant volume of the GUV during the course of the experiment due to the low water permeability of the membrane (39) and isoosmolar conditions, lateral FIGURE 1. Scheme of an adhered GUV. Shown is a schematic drawing (top images) and corresponding SDCLM images (bottom images) of a GUV on an avidin-coated surface before and after adhesion. Without adhesion, the GUV geometry can be approximated as a sphere (radius R v ). After adhesion, a spherical cap with a radius of R ad and a circular contact area (with a radius of R i ) is formed. Scale bars: 10 m.
tension can be computed from the area dilatation generated by adhesion. In equilibrium, larger adhesion forces generate larger area dilatation. To calculate the lateral membrane tension from the normalized area increase, one has to consider two contributions, one from ironing out membrane undulations and one from dilating a flat membrane laterally in a two-dimensional Hookean fashion (40). Excess area stored in thermally excited membrane undulations is sacrificed first followed by lateral expansion of lipids. Equation 2 relates area dilatation to membrane tension of an adhered GUV (41), The bending rigidity of a DOPC membrane is assumed to be 21 k B T, and the area compressibility modulus K A ϭ 265 mN/m accordingly (42). As a reference state, we choose the free vesicle with tension 0 , which depends on the size of the vesicle and the osmotic pressure between inside and outside of the vesicle. Equation 2 holds, if 0 Ͻ Ͻ , and allows the estimation of the lateral membrane tension of an adhered GUV numerically. Because it is impossible to determine 0 for all vesicles used here independently, we determined 0 for a few representative GUVs by parallel plate compression in conjunction with optical microscopy (43). We found 0 to be in the range of 10 Ϫ6 -10 Ϫ4 N/m in accordance with typical values for weakly adhering GUVs on a glass substrate as obtained by reflectometric interference microscopy (44).
To calculate the tension , we need to determine the relative area dilatation after adhesion of the vesicle. A free GUV forms a sphere with an area A v . It changes its geometry upon adhesion and forms a spherical cap with an increased surface area A ad . Note that a spherical cap geometry corresponds to the limit of strong adhesion with respect to membrane undulations (29).
The normalized change in surface area ⌬A/A v can be calculated from the shape change (Equation 3), with the condition of constant volume, ⌬A/A v is only a function of R i /R ad , which is related to the contact angle between a sessile vesicle and the substrate by ⍜ ϭ sin Ϫ1 (R i /R ad ). Fig. 3 shows the R i /R ad values (n total ϭ 467) of adhered PtdIns(4,5)P 2 containing GUVs with an average radius of R v ϭ 18 Ϯ 5 m for different Mg 2ϩ concentrations. The mean ratios of R i /R ad increased from 0.46 at a Mg 2ϩ concentration of 0.5 mM to 0.63 at 3.0 mM. For GUVs lacking PtdIns(4,5)P 2 (n total ϭ 413) with an average radius of R v ϭ 18 Ϯ 5 m, R i /R ad increased from 0.19 at a Mg 2ϩ concentration of 1 mM to 0.53 at 6.0 mM. Fig. 4 shows how area dilatation and lateral tension change as a function of R i /R ad according to Equations 2, 3 and 4.
A small change in R i /R ad can result in a rather large change in lateral membrane tension (Fig. 4). At low R i /R ad , the excess area of the GUV stored in undulations compensates for the area dilation caused by adhesion on the surface (R i /R ad ϭ 0.46, ϭ 0.08 mN/m), whereas for larger R i /R ad undulations are ironed out, and the lateral membrane tension increases significantly due to the first term in Equation 2 (R i /R ad ϭ 0.63, ϭ 1.02 mN/m). The results clearly demonstrate that the lateral membrane tension of adhered GUVs can be readily adjusted within a certain regime making use of different Mg 2ϩ concentrations to control ␥ ad, the adhesion energy per unit area. This system thus enables us to study the influence of the specific binding of ENTH to PtdIns(4,5)P 2 containing GUVs as a function of a given initial lateral membrane tension.
ENTH Binding to PtdIns(4,5)P 2 -doped GUVs-ENTH can induce membrane tubulation (22,25). Recently we showed that binding of ENTH to PtdIns(4,5)P 2 -doped and slightly curved pore-spanning membranes results in a reduction of the lateral membrane tension (27). However, the initial tension of the bilayer was rather high, and it is well conceivable that tubulation is suppressed at elevated tension. To answer the question of how ENTH binding to PtdIns(4,5)P 2 containing adhered GUVs alters their shape dependent on the preadjusted lateral membrane tension, we prepared GUVs composed of DOPC/ DOPE/cap-biotin-PE/Atto488-DPPE/PtdIns(4,5)P 2 , (66.2:30: 2:1:0.8) in the presence of 0.5 mM and 2 mM Mg 2ϩ . The resulting mean membrane tensions were 0.08 mN/m and 0.52 mN/m, respectively. To these adhered GUVs, ENTH was added at a final concentration of c ENTH ϭ 1 M. Interestingly, dependent on the preadjusted lateral membrane tension, two different behaviors were observed after protein addition. At a low lateral membrane tension ( ϭ 0.08 mN/m), the addition of ENTH can indeed result in the formation of tubules, eventually leading to the disappearance of the GUV (Fig. 6). Because of the stirring of the solution, tubules appear preferentially on one side of the GUV and also elongated along this direction. Of note, even at this low lateral membrane tension, only part of the GUVs show tubulation, whereas others display an increase in adhesion area indicative of transient reduction in lateral membrane tension (see below).
It has been shown that upon binding of ENTH to PtdIns(4,5)P 2 , the protein experiences a change in structure in which helix-0 inserts into the lipid bilayer as demonstrated by EPR spectroscopy (22). This increases the surface area of the outer leaflet of the GUV membrane and induces an asymmetry in the two bilayer leaflets facilitating membrane bending by generating spontaneous curvature (26,45). In agreement with our observations of tubule formation upon ENTH binding, association of full-length epsin and ENTH to non-adhered tension-free GUVs also resulted in the formation of irregular structures such as vesicles and tubules (23,25,46,47) dependent on the experimental conditions and on the surface density of bound protein (48,49). Recently, Shi and Baumgart (49) provided experimental evidence for a mechanism whereby local membrane tension reduction mediated by endocytic proteins reduces the elastic energy required for membrane deformation and thus facilitates membrane budding and tubulation. Molecular dynamics simulations predicted that the formation of tubules requires a lower protein coverage on the membrane than the formation of vesicles (48). To calculate the protein surface coverage in this   study, we determined the dissociation constant of ENTH to PtdIns(4,5)P 2 -doped planar POPC membranes (9:1) by reflectometric interference spectroscopy, which resulted in a K D ϭ 0.8 M. This in turn means that at an ENTH concentration of 1 M, 56% of the receptor lipids are occupied. The GUVs used in this study contain 0.8% of the receptor lipid PtdIns(4,5)P 2 . Taking the protein's footprint of 16 nm 2 (50) into account and assuming a 1:1 binding of ENTH to PtdIns(4,5)P 2 results in a protein surface coverage on the membrane of 10%. This rather low protein coverage agrees with the observation that only tubules are formed and no vesicles (50,51).
However, if GUVs with a larger membrane tension of ϭ 0.52 mN/m was incubated with ENTH, in none of the cases was tubulation observed. Instead, the adhered vesicles started to flatten, and their contact area increased after ENTH binding (Fig. 7), i.e. the distance between the equatorial plane of the GUV (Fig. 7, blue line) and its contact area (Fig. 7, red line) was reduced as a function of time. The size of the adhesion area is a result of the newly established equilibrium between adhesion force and membrane tension (33). When the new equilibrium was adjusted after ENTH binding, the membrane tension again reached its previous value of ϭ 0.52 mN/m, which essentially prevents tubule formation as any deformation would require further dilation of the membrane as long as the volume is conserved.
This unique experimental approach allowed us to obtain statistically meaningful data by imaging 10s of GUVs before and after protein addition as a function of membrane tension. Normalized contact radii R i /R ad of GUVs adhered to an avidincoated substrate at three different Mg 2ϩ concentrations were imaged by SDCLM before and after incubation with ENTH for 1 h to ensure equilibrium conditions. Fig. 8 clearly shows that for each Mg 2ϩ concentration the ratio of R i /R ad is significantly larger after ENTH addition (Fig. 8, gray boxes) than before (Fig.  8, white boxes). For low membrane tension, where also tubulation can occur, only those GUVs with no indication of tubulation were used for statistical analysis.
Under the assumptions that the adhesion energy per unit area remains the same and the initial tension is restored after insertion of the proteins' helix-0, an increase in R i /R ad can only be explained by an area increase ␦A ENTH upon helix-0 insertion or by a decrease of the area compressibility modulus K A,ENTH  (for details see "Appendix"). First, we estimate the expected increase in area ␦A ENTH due to helix insertion. Therefore, we take the area of helix-0 of ENTH into account, which is ϳ10 -30% (ϭ 1.6 -4.8 nm 2 ) of the ENTH surface projection. Employing the determined dissociation constant K D and the known PtdIns(4,5)P 2 concentration assuming a PtdIns(4,5)P 2 area of 0.7 nm 2 (52) in the GUVs, this translates into a maximal expected relative area increase of 1-3%. Next, we compare this estimate with the experimentally obtained relative area change deduced from the measured changes in R i /R ad before and after ENTH binding. Under the assumption that the adhesion energy per unit area of the GUV and its volume remain constant, a relative area increase of 83% is calculated (see "Appendix"). The computed relative area increase is independent of the initial tension. Such a tremendous area increase is unrealistic and contradicts the theoretically expected relative area increase. Thus, we conclude that ␦A ENTH does not explain the significant change in R i /R ad . Instead, a substantially reduced area compressibility modulus K A might explain the observation. Neglecting in this case the area increase of ␦A ENTH due to insertion alone and calculating the altered K A values, one finds that K A reduces from the initial value of K A ϭ 265 mN/m to 145 mN/m independent of the initial lateral membrane tension.
We speculate that the insertion of helix-0 of ENTH into one leaflet of the membrane is responsible for the observed reduction in K A . To confirm this hypothesis, we performed the same GUV experiments at 2 mM Mg 2ϩ but replaced wild type ENTH with the mutant ENTH L6E. This mutant binds Ins(1,4,5)P 3 with the same affinity constant as the wild type but has a reduced binding affinity to short chain diC8PtdIns(4,5)P 2 and does not induce tubules in vesicles (23,25). The result depicted in Fig. 8 demonstrates that the change in R i /R ad is significantly smaller than that observed for wild type ENTH, suggesting that helix-0 is largely responsible for the observed change in K A . A reduction in K A directly translates into a reduction of the bending rigidity , i.e. ENTH binding accompanied by helix-0 inser-tion reduces the energetic cost for bending and thus tubule formation.
Membrane softening has been shown to occur upon interaction of several small peptides with membranes and has been ascribed to perturbations on the membrane organization and to a direct modulation of by membrane thinning (53)(54)(55). Settles et al. (53) showed that Sar1p, a protein involved in vesicle trafficking, dramatically lowers the rigidity of lipid bilayer membranes. This membrane softening does not rely on the imposition of strong local curvature and is a profoundly different mode of action than those typically ascribed to intracellular curvature-associated proteins. For ENTH, elasticity theory predicts curvature generation based on membrane insertion of helix-0 (56), which can occur on the local scale. In our setup, membrane softening due to helix-0 coupled to the local curvature of the membrane led either to tubule formation, if this process is faster than the adjustment of the new equilibrium state of the GUVs, or to an increase in R i /R ad if tubule formation is slower. In contrast, GUVs with a larger initial tension do not become floppy enough so that local curvature did not lead to tubule formation. Instead, the GUVs reached the new equilibrium state upon membrane softening, observable as an increased R i /R ad . It is conceivable that this is a kinetic effect, i.e. the GUV spreads faster regaining its initial tension than tubules can be formed. As the amount of bound ENTH is largely given by the mole fraction of the receptor lipid PtdIns(4,5)P 2 , the reduction in K A , is independent of the initial lateral membrane tension as expected. Fig. 9 proposes a model based on our findings. Insertion of helix-0 into a prestressed vesicle (1) relaxes tension until either tubulation sets in at low initial tension (2Ј) or a new equilibrium state is adopted (3) by generating a larger adhesion area to maintain tension.
Our results are in line with a recent study of Shi and Baumgart (49). They have shown by micropipette aspiration experiments on single GUVs that a larger membrane tension of 0.25 mN/m strongly inhibits N-BAR-induced shape instability, which is required for the formation of tubules (49). N-BAR domains are found in endocytic proteins and are one of the best studied examples of proteins that sense and induce membrane FIGURE 8. Changes in R i /R ad in response to ENTH binding as a function of initial membrane tensions. R i /R ad of GUVs adhering on an avidin-coated glass surface measured at different Mg 2ϩ concentrations before and after ENTH addition (c ENTH ϭ 1 M). For each condition 52-117 GUVs were measured obtained from at least two independent preparations; black diamonds represent the mean; n total ϭ 556. ***, p Ͻ 0.001; ns, not significant. curvature and whose binding involves also the insertion of parts of the protein into the membrane (57). Molecular dynamics simulations support the findings of Baumgart and coworkers (58) showing that a membrane tension in the range of 0.2 mN/m and larger increases the free energy barrier for the formation of tubules from a planar membrane and inhibits polymerization of N-BAR domains (59). Our results as well as those of others are also in line with the notion that exo-and endocytic pathways are regulated by lateral membrane tension in vivo. Hypoosmotic conditions resulted in a membrane tension of 4.2 mN/m and completely suppressed endocytosis in K562 cells (16). The endocytic activity in 2H3 cells decreased after increasing the membrane tension by osmotic swelling, whereas membrane tension dropped if the cells were stimulated to secrete (10).
Conclusions-We have shown that it is possible to adjust the lateral membrane tension of GUVs adhered to an avidincoated glass surface by varying the Mg 2ϩ concentration in solution. In contrast to micropipette aspiration, where the lateral tension can be precisely adjusted, the presented method has the advantage that 10s of GUVs can be imaged simultaneously allowing for high statistics to be obtained. The influence of protein binding on the morphology of the adhered GUVs as well as on the membrane's mechanical properties can be easily inspected by confocal microscopy without the need for protein labeling. Based on our results, we propose that the insertion of helix-0 of ENTH into the membrane is largely responsible for tubule formation at low lateral tension, whereas it mainly softens the membrane at higher membrane tensions. Our findings contribute to the overall discussion about how helix-0 insertion modulates membrane curvature and mechanical properties of the membranes. We suggest that the reduction of bending rigidity of membranes plays a pivotal role in all forms of endocytosis.

Experimental Procedures
Preparation of GUVs-Lipids (Avanti Polar Lipids, Alabaster, AL) were dissolved in CHCl 3 in the desired ratio and deposited on indium tin oxide-coated cover slips (V ϭ 50 l, c ϭ 0.8 mg/ml). After solvent removal under reduced pressure for 3 h at room temperature, GUVs were formed by electroformation ( Preparation of Avidin-coated Glass Slides-Glass slides (Ibidi GmbH, Planegg, Germany) were treated with a mixture of H 2 O/NH 3 /H 2 O 2 (5:1:1, v/v/v) for 20 min at 70°C before incubation with an avidin solution (c ϭ 10 M) dissolved in phosphate-buffered saline (PBS, 137 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , 1.8 mM KH 2 PO 4 , pH 7.4) for 30 min at room temperature, and the glass slides were thoroughly rinsed with ultrapure H 2 O. After adsorption of avidin and rinsing with buffer, the avidin-coated glass surface was passivated by incubation with bovine serum albumin (10 mg/ml) for 30 min. Afterward, the glass slides were transferred into sucrose-containing buffer (2 mM HEPES, sucrose, pH 7.4, O final ϭ 298 mosmol/kg) with the desired Mg 2ϩ concentration.
ENTH Expression-ENTH (wild type and L6E mutant) was expressed and purified as described previously (27). ENTH of rat epsin1 (residues 1-164) fused with a glutathione S-transferase tag was cloned into the vector pGEX-6P-1. After protein expression in BL21 cells for 3 h at 37°C, the cells were lysed using Emulsiflex C3 and spun at 140,000 ϫ g for 30 min at 4°C in a Beckman Ti45 rotor. The supernatant was bound to glutathione beads for 30 min. Afterward, the beads were washed 6 times with 20 mM HEPES, 150 mM NaCl, 2 mM dithiothreitol, 2 mM EDTA, pH 7.4, with 2 washes at 700 mM NaCl in between. After cleavage of the glutathione S-transferase tag using Pre-Scission proteases, the cleaved proteins were further purified by gel filtration (Superdex 200). The protein was transferred into sucrose-containing buffer with the corresponding Mg 2ϩ concentration using Amicon Ultra 4 ml Filters (NMGG10000, Merck, 3 ϫ 30 min, 4000 ϫ g, 4°C; Allegra™ X-22R, Rotor SX4250, Beckman Coulter, Krefeld, Germany).
GUV Data Evaluation-To automatically determine the z-position of the contact area of the GUV, a custom-written Matlab-script was used. An intensity profile was measured at the GUV center, and the center positions were determined by fitting Gaussian functions to the profile. The two maxima define the positions of the GUV contact area and the upper membrane of the GUV. The contact radius R i was calculated from the circular contact area that was measured by threshold analysis. To determine the radius R ad of the spherical cap after adhesion, the circular equatorial plane of each adhered GUV was determined.
Reflectometric Interference Spectroscopy-Reflectometric interference spectroscopy measurements were performed using a custom-built setup equipped with a NanoCalc-2000 spectrometer (OceanOptics Germany GmbH, Ostfildern, Germany) (60). SiO 2 wafers with an oxide layer of 5 m were incubated in a solution of NH 3 and H 2 O 2 (H 2 O/NH 3 /H 2 O 2 , 5:1:1) for 20 min at 70°C. The time course of bilayer formation after the addition of small unilamellar vesicles obtained by sonication (61) was monitored until a stable optical thickness was reached. After rinsing with buffer, the resulting membrane was treated with a bovine serum albumin solution (1 mg/ml) to prevent nonspecific protein adsorption. After again rinsing with buffer, ENTH was added, and spectra were recorded every 2 s and the optical thickness determined (60).

Appendix
The free energy F v of a vesicle adhering to a surface in the strong adhesion limits is given by Equation 1. Minimizing the free energy F v leads to the following equilibrium condition, with ϭ sin Ϫ1 (R i /R ad ) (31). Adhesion energies larger than 10 Ϫ4 J/m 2 were calculated for vesicles adhering to an avidin-coated glass surface based on the vesicle geometry using Equation 5 and K A ϭ 0.265 N/m (30,42). Adhesion energies of 1.25⅐10 Ϫ4 J/m 2 , 3.32⅐10 Ϫ4 J/m 2 , 4.78⅐10 Ϫ4 J/m 2 , and 7.70⅐10 Ϫ4 J/m 2 at Mg 2ϩ concentrations of 0.5, 1, 2, and 3 mM, respectively, were determined in accordance with the assumption of the limit of strong adhesion.
An increase in R i was measured as a result of ENTH binding, and we assume that helix-0 insertion into the bilayer enlarges the vesicle's surface area by ␦A ENTH . Equation 1 can then be rewritten to, ␦A ENTH,ad denotes the area change of an adhered vesicle due to ENTH helix-0 insertion, whereas ␦A ENTH,v is the corresponding change in area of a free vesicle. Rearranging Equation 6 assuming ␦A ENTH,ad Х ␦A ENTH,v Х ␦A ENTH yields, with an apparent area compressibility modulus K A,ENTH after ENTH binding of, Assuming a constant adhesion energy ␥ ad before and after ENTH addition, K A,ENTH can be obtained from, where ENTH is the equilibrium contact angle after ENTH addition. A mean apparent area compressibility modulus of K A,ENTH ϭ 0.145 Ϯ 0.009 N/m after ENTH addition was calculated for 3 initial lateral membrane tensions. The necessary change in vesicle area caused by binding of ENTH (␦A ENTH ) required for a reduction of K A ϭ 0.265 N/m to the apparent area compressibility modulus K A,ENTH ϭ 0.145 N/m was then calculated using Equation 9. An area increase caused by helix-0 insertion of ENTH of 83% would be required to explain the observed reduction in the apparent area compressibility modulus while keeping K A constant. This is unrealistic, as the relative area increase caused by the insertion of the ENTH helix was approximated to be 1-3% based on the protein's crystal structure. Therefore, the insertion of the ENTH helix is believed to cause the observed change of the area compressibility modulus neglecting the small change in area as a result of helix-0 insertion. The decrease in K A translates into a reduction of the bending rigidity, i.e. a softening of the membrane as observed after insertion of helical peptides into lipid bilayers (53)(54)(55).