Correlation between adhesion strength and phase behaviour in solid-supported lipid membranes

phase transitions observed in the layer of adsorbed lipid vesicles/membranes. It paves the way to explore structural changes on more complex biointerfaces by acoustic-based sensors.


Introduction
Lipid vesicles are self-assembled structures customarily used as model systems for cell membrane basic studies [1,2], as nanocontainers for bio-reactions [3] and in biotechnology applications such as drug delivery or biosensors [4,5].When supported on solid-surfaces, they might form intact supported vesicle layers (SVLs) or eventually rupture into planar supported lipid bilayers (SLBs).The latter are the result of spontaneous adsorption of small (diameter ≤ 200 nm) vesicles onto solid surfaces.The geometry of SVLs captures the volume to area ratio of the vesicles, strength of adhesion, membrane bending properties and osmotic stress within the supported layer making SVLs useful biomimetic platforms to probe membrane deformation.The latter plays an important role in biological processes such as adhesion, budding, lipid membrane exchange, fission and fusion [6][7][8].SVLs are mimics to endosomes or exosomes, which are important for chemical transport and intercellular communication [9,10]; however, these systems are optically inaccessible and the experimental investigation of their deformation is not straightforward.Adsorbed vesicles onto inorganic surfaces serve also as model systems relevant to biocompatibility studies.The shape of a vesicle upon adsorption to a surface is determined by the interplay of adhesion, bending and geometrical constraints.This interplay is theoretically studied starting from a simple model in which the membrane experiences a contact potential arising from the attractive surface.Let us recall the free energy F expression of an adsorbed vesicle in terms of a simple model taking into account the adhesion energy, the local bending energy term and the geometrical constraints [6,11]: The first two terms depend on the local bending modulus κ, on the Gaussian curvature modulus κ G and on C 1 , C 2 , and C 0 denoting the two principal curvatures, and the (effective) spontaneous curvature, respectively, and dA being an infinitesimal membrane area element.The third term is the adhesion free energy with W being the strength of adhesion and A c the contact area of the membrane and the surface.The last two terms represent the volume (V) and area (A) constraints with corresponding Lagrange multipliers P and Σ.
Vesicle deformation upon adhesion depends strongly on the lipid organization of the vesicle membrane, which in turn is intimately linked to their phase behaviour [12][13][14][15].The study of vesicles within such a small size range typically requires surface-sensitive techniques, since vesicles with diameter ≤ 200 nm are optically inaccessible structures.Moreover, the study of these phenomena by traditional calorimetric approaches is often hindered by the instability of SUV dispersions, which tend to fuse into larger LUVs and sediment over long-term measurements [16,17].
In this context, quartz crystal microbalance with dissipation monitoring (QCM-D) has recently emerged as a versatile technique to detect and characterize phase transformations of solid-supported lipid membrane geometries, namely SLBs and SVLs [18].QCM-D is an acousticbased, label-free technique used extensively in bio-interfacial science and has particularly contributed to the understanding of the kinetics of adsorption and formation of supported lipid bilayers [19][20][21][22] and probing their interactions with relevant biomolecules [23,24].QCM-D is highly sensitive to mass and energy dissipation changes at the solidlipid layer-liquid interface.In the particular case of adsorbed vesicles, it can detect changes in their geometry (shape) and membrane conformation when these systems undergo phase transitions [25][26][27].In this work we address the question whether and how the supported membrane shape affects the phase transitions in solid-supported membranes.Vesicles of dipamitoylphosphatidylcholine (DPPC) are chosen, since DPPC is a saturated phospholipid ubiquitous in eukaryotic cells and a well-known lung surfactant, whose main phase transition is well-characterized in bulk [28,29].Two solid surfaces bearing different adhesion levels are used, polycrystalline Au, an interesting transducer material owing to its good thermal and electrical conductivities, plasmon resonance and biocompatibility [30], and SiO 2 which bears a negative charge and additional electrostatic interactions [31][32][33] in determining adhesion [34,35].The correlation between surface wettability and membrane phase transitions has been explored as a function of vesicle size and adsorption temperature.Numerical calculations based on free energy minimization of Eq. ( 1) provide complementary information on the adsorbed vesicle systems.

Vesicle preparation
DPPC lipid in powder form was first dissolved in spectroscopic grade chloroform, and the solvent was evaporated under a mild flow of nitrogen in a round bottomed flask.The lipid film was kept under low pressure overnight to remove any traces of remaining solvent.The film was then hydrated with HEPES buffer to 1 mg/mL under continuous stirring in a temperature-controlled water bath at 55 °C (well above the melting temperature of DPPC T m ~41.5 °C).Large unilamellar vesicles (LUVs) and small unilamellar vesicles (SUVs) were formed by extrusion through filters with different pores sizes (100 nm or 30 nm) for a fixed number of passes.LUVs were formed after 25 passes through a 100 nm-pore filter, while SUVs were formed after 25 passes through a 100 nm-pore filter followed by 15 passes through a 30 nm-pore filter.

Dynamic light scattering
Vesicle effective sizes and polydispersity were determined by dynamic light scattering (DLS) (Malvern Zetasizer Nano ZS, Malvern, UK).The obtained mean diameters and polydispersity indexes of the samples used are displayed in Table 1.

Quartz crystal microbalance with dissipation monitoring (QCM-D)
A Qsense E4 instrument (Gothenburg, Sweden) monitoring the frequency and dissipation changes, Δf and ΔD was used.Q-sense E4 also enables heating or cooling temperature scans in the range between 15 °C and 65 °C.AT-cut quartz crystals with Au and SiO 2 coating (diameter 14 mm, thickness 0.3 mm, quoted surface roughness < 3 nm, and resonant frequency 4.95 MHz) were used.The Au-coated quartz sensors were cleaned with a 5:1:1 mixture of Milli-Q water (resistance of 18.2 MΩ cm at 25 °C), ammonia and hydrogen peroxide, and were UV-ozone treated with a UV-ozone cleaner (Bioforce Nanosciences, Germany) for 20 min, followed by rinsing in Milli-Q water and drying with N 2 .SiO 2 -coated quartz sensors were cleaned in a solution of sodium dodecyl sulfate (2% SDS) for several hours and UV-ozone treated for 20 min, followed by rinsing in Milli-Q water and drying with N 2. The changes in Δf/n and in ΔD were monitored at four different overtones (from 3rd to 9th).The lipid vesicles were inserted into the QCM-D cells with a flow rate of 50 μL/min.Vesicle adsorption experiments were carried out at two different temperatures, at 16 °C and at 50 °C, where DPPC is in phases possessing different bending rigidities, the gel phase and the liquid-disordered phase, respectively.The temperature stability is in the order of ±0.02 °C around the set value.First, a baseline with pure HEPES buffer was established and afterwards lipid vesicles were injected over the Au-coated or SiO 2 -coated sensor chips.After reaching a stable supported membrane layer the pump was switched off and the ensemble was left to stabilize for several hours.Subsequent temperature scans, upon heating and cooling, were performed at a rate of 0.4 °C/min, maintaining a 60 min stabilization time between successive temperature ramps.Measurements were repeated at least three times to check the reproducibility of the results.

Contact angle measurements
Contact angle (CA) measurements were carried out using an Attension ThetaLite from Biolin Scientific (Sweden) based on the sessile drop method.A small drop (3 μL) of Milli-Q water or diiodomethane was deposited onto clean, UV-ozone treated Au-coated or SiO 2 quartz surfaces, and the shape of the drop formed on the surface was analysed.The contact angle of the 3 μL droplet of either ultrapure water or diiodomethane was determined over a time period of 10 s using a recording speed of 20 frames/s and, afterwards, the average of several drops was calculated.The CA was measured at several points, and an average value was extracted.All contact angles were measured at a room temperature.Surface free energies of UV-ozoned Au and SiO 2 surfaces γ sv (polar γ sv p and dispersive γ sv d parts) were determined based on the Owens, Wendt, Rabel and Kaelble method [36] and are included in Table 2. Details on calculations are included in the Supplementary material.3. Results and discussion

Vesicle adsorption and layer formation
Fig. 1 shows the Δf and ΔD responses (represented for the third overtone) during vesicle adsorption and supported lipid layer formation on SiO 2 and Au at temperatures well below (16 °C) and above (50 °C) the melting temperature of DPPC.The mechanistic scenario of the observed Δf and ΔD changes is governed by a delicate balance between the adhesive energy from lipid-surface interactions (which tends to maximize the contact area between the vesicle membrane and the surface) and the opposing effect of bending the bilayer [37].
When adsorbed onto Au, a monotonic frequency decrease (mass increase) and dissipation increase can be observed reaching constant nonzero Δf and ΔD plateau values.Such time-dependent responses provide evidence that oxidized Au facilitates non-ruptured vesicle adsorption towards the formation of acoustically non-rigid vesicle layers with saturated coverage.
Although the vesicle adsorption profiles are similar for small and large vesicles when adsorbed in both the gel and liquid disordered phases, it is worth exploring the size-and temperature-dependent differences.After initial adsorption both Δf and ΔD plateau values clearly increase (in absolute value) with increasing vesicle size.For LUVs on Au at T < T m , an overshoot behaviour was observed in the dissipation signal but not on the frequency signal.This peak has been seen in previous works and was ascribed to vesicles 'rocking and rolling motion' [38].
Larger vesicles carry more trapped aqueous buffer (larger frequency shifts) and thus are softer structures (larger dissipation shifts).When adsorbed at T > T m the same trend with vesicle size was observed, although the plateau values were significantly smaller.At T < T m the DPPC bilayer envelope of the adsorbed vesicles is in the gel phase and its bending modulus κ ~10 • 10 −19 J [39] renders the vesicle membrane more stiff.Above T m , the bilayer is in the liquid disordered phase and the modulus attains about a ten times smaller value of κ ~1 • 10 −19 J.This makes the membrane of vesicles softer and more deformable upon adsorption with larger contact area, yielding a smaller number of vesicles for similar surface coverage.
When adsorbed onto SiO 2 , a completely different pattern of behaviour is observed below and above T m .At T < T m , monotonic frequency and dissipation changes take place reaching constant non-zero Δf and ΔD plateau values.The plateau values follow the same trend with vesicle size as observed in the case of Au.However, the reached plateau values are smaller on SiO 2 , indicating that the stronger SiO 2 adhesion might favour vesicle deformation, induce the formation of transient pores [40], local vesicle rupture events and formation of small bilayer patches (from the ruptured vesicles).It is worth noting that QCM-D is very sensitive to hydrodynamic (wet) mass and the local, partial formation of SLBs might be masked by the adsorption of vesicles on top or in between the bilayer patches [41].At T > T m an initial monotonic adsorption of vesicles is observed until a critical surface coverage is reached (minimum in Δf and maximum in ΔD), followed by vesicle fusion and rupture to form SLB [21,22,37,41].The surface-vesicle interactions on SiO 2 are stronger than on Au and the SiO 2 -adsorbed vesicles deform to a greater extent with higher contact area and higher membrane lateral tension, making them more prone to pore formation or rupture and fusion [42] and therefore less stable.Note that lipid bilayers in the gel phase can sustain higher lateral tensions and therefore rupture appears at higher values of critical tension (i.e.lysis tension) [43].
Au has been oxidized thus is highly hydrophilic and its isoelectric point is around 4.5 to 5 [44].The isoelectric point of SiO 2 is around 2.5 [45], therefore both Au and SiO 2 are negatively charged at the current experimental conditions (pH = 7.4), SiO 2 being even more than Au.This confers Au and SiO 2 an attractive potential to support intact vesicles of zwitterionic lipids such as DPPC.The lateral tension arising by attractive forces between the adsorbing vesicles and the surface does not exceed the threshold for permanent membrane rupture in the case of Au.As a matter of fact, the polar component of the surface free energy in Au is half the value calculated for SiO 2 (see Table 2), while the dispersive component is larger, in agreement with the large Hamaker constant of Au [46].In addition, the formation of transient membrane pores [40]  cannot be excluded, especially at T > T m , since in the liquid disordered phase the lipid bilayer area stretching modulus is 3 to 4 times smaller than in the gel phase [47].The theoretically predicted attraction of zwitterionic vesicles (PC headgroup) to negatively charged surfaces (Au or SiO 2 ) is accompanied by more perpendicular average orientation of the zwitterionic head groups [31][32][33], i.e. to more tightly packed lipids, which may favour the gel phase of lipids in the adhered part of the vesicle membrane.
The extent of vesicle deformation upon adsorption was estimated following the approach introduced by Tellechea et al. [38].This method consists in plotting −ΔD/Δf ratio vs −Δf for all overtones during initial adsorption (low vesicle surface coverage), which typically shows a linear decrease over a large range of frequency shifts.Extrapolation of this linear decrease to a frequency-independent intercept with the −Δf axis (where overtones intersect) provides a value of the thickness of the adsorbed vesicle layer h referred to as Sauerbrey thickness: h = − ΔfC/ρ, where C = 18 ng/cm 2 Hz and ρ = 1 g/cm 3 is the density of the film [44].This approach assumes a complete surface coverage at the end of the adsorption process, where the presence of trapped buffer has been diminished to occupy only the void spaces between densely packed vesicles (the −ΔD/Δf ratio is close to zero and the −Δf intercept values were the same on the extrapolation of a linear regression) [38,48,49].Though this approach might not provide a correct absolute value of the real thickness it provides valuable qualitative information.Sauerbrey layer thickness data are included in Table 3 and it is noteworthy that the obtained values are systematically smaller than those obtained for vesicles in bulk by DLS measurements.For LUVs and SUVs ruptured on SiO2 at T > T m , the thickness obtained are 28 nm and 23 nm, respectively, and this overestimation is likely related to the fact that layers are not fully homogeneous and they might bear some degree of oligolamellarity, especially in the case of ruptured LUVs.The -Δf intercept values have been obtained from the average of four overtones.From the Saurbrey thickness values, the extent of vesicle deformation Δd (in percent) was calculated as the relative change of size upon adsorption as compared to the original size of vesicles dispersed in buffer, Δd ¼ ðd DLS −hÞ d DLS Â 100.In Fig. 3, the extent of vesicle deformation is displayed for all cases.
The time-independent ΔD vs Δf curves provide additional information on the structural properties of the vesicle adsorbed layers as a function of vesicle size and temperature, reflecting the interplay between bending, adhesion and steric contributions in the vesicle adsorption process.The ΔD-Δf curves in Fig. 4 exhibit steeper slopes with increasing vesicle size, indicating a larger increase in energy dissipation per adsorbing vesicle.At T > T m , LUVs vesicles are prone to larger deformation and are less efficiently packed (so-called steric effect) [50], resulting in a greater viscoelastic contribution per adsorbing vesicle both on Au and SiO 2 , the dissipation being smaller in the latter as a result of an increased adhesion interaction and lateral tension by the SiO 2 surface.When adhered onto SiO 2 , the fate of adsorbed vesicles depends strongly on temperature; intact but less dissipative vesicle layers are formed as compared to Au at T < T m , while a re-entrant behaviour in the ΔD-Δf curves is observed at T > T m .This results from the combination of three factors i) the increased lateral tension with increased surface adhesion, ii) the decreased bilayer bending modulus and iii) decreased critical tension for rupture upon increasing temperature, resulting in an easier vesicle fusion and rupture.

Calculations of vesicle shapes upon adsorption
Fig. 5 shows the axisymmetric shapes of the adsorbed vesicles corresponding to minimal free energy given by Eq. ( 1).In order to better illustrate the effect of the different contributions involved in vesicle adhesion, the vesicle shapes have been calculated for values of reduced volume v = 6π 1/2 V/A 3/2 [11,51] and parameter w = WR s 2 /κ, covering a broad range, namely v = 0.54, 0.75, 0.85, 0.95 and w = 0.4, 6.4, 64 and 640.A detailed description of the numerical procedure for the determination of the vesicles shapes [52] is included in the Supplementary material.For a constant value of the reduced volume v, the calculated shapes depend on the competition between the bending and adhesion contributions.At large w adhesion dominates and favours a large contact area.At small w the obtained shapes of free (non-adhered) membrane parts of the vesicles are more undulated (Fig. 5A and B), since adhesion is smaller and the bending energy takes over.Conversely, at large w, the shapes of the free non-adhered membrane parts become increasingly more spherical (Fig. 5C and D) to maximize the contact area, i.e., the shapes at large w resemble more and more a part of the spherical surface.For very large w (small bending modulus, large W and large vesicles) the shapes of the adhered vesicles approach the limiting shape composed of free (non-adhered) part of the membrane which is spherical, and another flat adhered part (Fig. 5D).As it can be seen in Fig. 5D, the limiting shapes depend only on the value of the reduced volume v   [53].The largest part of the vesicle mass and volume is in this case distributed closer to the sensor surface.The obtained nearly limiting shapes for very large values of w (strong adhesion limit) are predominantly determined by the tendency of the vesicles to achieve the maximal possible reduced contact area between the vesicle membrane and the surface (A c /A) at given reduced volume v.In this regime the influence of the bending energy is very weak and the limiting shapes would be the same also for the initially weakly adsorbed prolate vesicle shapes.It is worth noting that QCM-D experiments were carried out at a rather large vesicle concentration thus the packing efficiency of adhered vesicles and the vesicle contact area can be reduced due to steric effects [50].
In order to better visualize the adhesion strength W behind the results presented in Fig. 5, we have calculated W = wκ/R s 2 for κ ~10 • 10 −19 J at room temperature (i.e. in the gel phase) and for two values of R s : 70 nm (LUVs) and 30 nm (SUVs), relevant to this work.For w = 64 (corresponding to the shapes of panel C in Fig. 5) we obtain the values W ~15 mN/m (LUVs) and W ~70 mN/m (SUVs), corresponding to the range of measured surface energies of water on Au and SiO 2 surfaces given in Table 2.
The first (local bending) energy term in Eq. ( 1) is scale invariant for C 0 = 0 [11], i.e., it is not dependent on the size of the vesicle.Since the second (adhesion) energy term in Eq. ( 1) is not scale invariant, it becomes increasingly more important for larger vesicles, i.e. larger w since parameter w is proportional to R s 2 (see also Eq. ( 12) in Supplementary material).As a consequence, the adhesion is more probable and energetically favourable for larger vesicles.Thus, the size of the vesicle plays an important role in the process of vesicle adhesion, as shown also in the experiments presented in this work (Table 3).The parameter w is proportional to R s 2 making LUVs more prone to adhere on the larger contact area and deform more.Their shape is closer to the limiting shape not only because they are larger, but also because they are strongly adhered (see Fig. 5D).The parameter w is on other hand inversely proportional to the bending modulus κ and since κ is smaller for T > T m , adhesion becomes stronger for T > T m and the vesicles shapes are closer to limiting shapes (see Fig. 5D).Moreover, w is larger for SiO 2 surface than for Au because of the larger strength of adhesion W of the SiO 2 surface.At constant reduced volume v the thickness (height) of the adhered vesicles should be increasing with w for large values of v (see Fig. 5).However, this effect was not observed in the QCM-D experiments for the surface with the largest adhesion strength SiO 2 .Indeed, on the SiO 2 surface the reduced vesicle volume v is decreased most likely due to the formation of transient pores [40,54] and/or increase of the vesicle area due to membrane "hidden" pool of lipids, conserved in the form of membrane nanotubular protrusions [55], which is released due to increased membrane lateral tension in the strong adhesion regime.Regarding the formation of the transient pores, it is more plausible to think that they are formed at T > T m .This is because in the liquid disordered phase the lipid bilayer area stretching modulus is smaller than in the gel phase [47] and the lipid bilayer can sustain also lesser critical lateral tension [43] needed for the membrane rupture and formation of transient pores.Accordingly, it can be seen in Table 3 that the thickness h of the adhered vesicles is substantially decreased at T > T m , and, as expected, the effect is more pronounced for the SiO 2 surface.

Phase transitions
Fig. 6 shows an overview of the temperature dependence of the firstorder derivative of Δf shifts (3rd overtone) upon the first heating run for all LUVs and SUVs vesicles adsorbed at T < T m and T > T m on Au and SiO 2 surfaces.Upon heating, lipid bilayers change from a stiffer gel phase to a softer liquid-disordered phase.As shown in recent works [18,[25][26][27], these changes are reflected as anomalies in both frequency and dissipation shift signals and, in particular, in their first-derivatives with respect to temperature that display clear extrema.The anomalies are governed by the interplay between changes in thickness, stiffness and, in the case of large adsorbed vesicles, by the presence of hydrodynamic channels, which change the shape of the adsorbed vesicles [26].
The shape of the dΔf(T)/dT obtained from QCM-D measurements is reminiscent of that of the isobaric heat capacity C p (T) from calorimetry during a melting transition [29,56].In this respect, it is instructive to draw an analogy between these two well-differentiated experimental techniques.Calorimetry is well-established and based on changes in thermal properties when vesicles undergo a phase transition, in this case, the main transition (heat absorption or release).On the other hand, QCM-D is based on changes in the viscoelastic properties of adsorbed vesicles as a result of changes in bilayer thickness, rigidity and vesicle shape.The calorimetric signal yields a maximum in C p corresponding to a finite jump in enthalpy H(T) along the first order lipid melting transition.The size of this jump scales in this case with the mass; the larger the mass, the larger the heat absorbed or released.In order to draw conclusions on the size of the anomalies observed in dΔf(T)/dT let us comment different aspects of Fig. 6 regarding surface free energy, vesicle size and adsorption temperature.

Main phase transition of SVLs on Au surfaces
We start by supported vesicle layers formed on Au exposed to UVozone, where no global rupture of vesicles to form SLBs was observed.
The main transition appears in all cases as a clear, single-peak anomaly and it takes place at the expected temperature range as compared to vesicles in bulk by calorimetric measurements [29,56,57].In some cases, a second maximum deviating from the linear, regular behaviour, can be observed at lower temperatures, which we ascribe to the pretransition.The appearance of this maximum will be discussed later.In order to obtain a more quantitative picture of temperature and size effects on the main phase transition, relevant parameters such as the area below the peak, the height of the maximum, peak width at half maximum, ΔT 1/2 , and the transition temperature (temperature corresponding to the peak maximum) have been extracted in each case and are displayed in Fig. 7.It is observed that the size of the peak, as well as the area below scale with effective vesicle size.Large adsorbed vesicles carry more trapped aqueous buffer and are more deformable structures, thus, stronger changes in Δf and ΔD occur along the main transition.First-order derivatives of the dissipation can be found in Fig. S2 of the Supplementary material.The width of the peak at half maximum is related to the cooperativity of the transition, i.e., measure of the degree of intermolecular cooperation during the main transition.
ΔT 1/2 decreases as the size of the adsorbed vesicle layer, indicating that the melting takes place in a narrower temperature range as adsorbed vesicles deform less on the solid surface.ΔT 1/2 increases as size decreases to a greater extent when vesicles are adsorbed T > T m .For vesicles adsorbed at T < T m where deformation is very small, the transition temperature T m shifts to lower values when decreasing size, since the higher curvature in SUVs decreases the lateral pressure [29,58].In addition, adsorbed layers of LUVs might be a combination of mostly unilamellar and some oligolamellar vesicles.Lamellarity results in increased cooperativity during the membrane melting owing to the strengthened bilayer-bilayer interactions [59].In general, the T m values fall within the expected temperature range, although they are slightly higher than those observed for multilamellar vesicles (MLVs) and LUVs given the rather fast scanning rate (0.4 °C/min).At T > T m , the extent of vesicle deformation is larger, they lie closer to the surface, presumably because of the decreased reduced volume v (see Fig. 5) and interactions with the substrate broaden the transition and increase the transition temperature.

Main phase transition of SVLs and SLBs on SiO 2 surfaces
Before analysing the phase transition behaviour, let us first recall the type of layers formed on SiO 2 surfaces.At T < T m , DPPC vesicles were adsorbed and deformed to a larger extent as compared to Au; however, no global rupture and thus no SLB formation was observed.At T > T m , both LUVs and SUVs are adsorbed, fused and ruptured to formed SLBs.As it can be observed in Fig. 6, the phase transition behaviour of these layers is greatly affected by the strong interactions between SiO 2 and the vesicles, resulting into films lying closer to the surface.For SVLs formed at T < T m , the transition is greatly broadened and takes place in a double-peak manner.Fig. 8 presents a closer view on the double peak melting anomaly.Both peaks have been fitted to Gaussians in an effort to decouple their relative sizes.Decoupling effects have been experimentally observed for SLBs and double SLBs using AFM [60,61], DSC [62], neutron reflectometry [63] and single particle tracking [64].They are ascribed to the stronger interaction between the lipid head groups of the proximal leaflet and the solid surface inducing a more tightly packed lipid distribution [31][32][33], which may favour the gel phase of lipids in the adhered part of the vesicle membrane [33] and/or the highly confined and orientationally ordered water layer between the proximal leaflet and the solid surface [31,32].As a matter of fact, the viscosity of the interfacial water layer is 10 4 times larger than bulk water [64].
In the present work, apart from in DPPC SLBs on SiO 2 , we also observe two peaks in the transitions of DPPC SVLs formed on SiO 2 .For the latter, the double peak is attributed to a decoupling effect in the melting between the lower part of the vesicle that is closer to the surface and the upper part at the vesicle-buffer interface.The low temperature peak takes place at a similar temperature to the melting of vesicles adsorbed on Au (T m ~42 °C).Hence, it confirms that this anomaly corresponds to the part of the vesicle envelope which is less affected by the solid surface.
The global size of both peaks scales with the effective adsorbed vesicle layer thickness.The individual peak sizes at the third overtone show an opposite pattern of behaviour between SVLs and SLBs.For SVLs on SiO 2 are greatly deformed and a large part of their area is in contact with the surface, while the remaining part melts like in bulk.As we shall see in the following section, SVLs formed on SiO 2 are quite unstable upon thermal cycling due to their largely deformed shape.In SLBs the peaks are less pronounced than in SVLs since the former are thinner and stiffer layers.Figs.S3 to S6 in the Supplementary material provide a complete overview of the transitions observed for all the systems in terms of frequency and dissipation at two overtones.The peak shapes show a complex behaviour for SLBs formed from LUVs owing to the fact that these layers might be a combination of multilayers and single bilayers.SLBs formed from SUVs precursors should be mostly single bilayers and show double peaks which are significantly broad.We aim to further explore the shapes of these two peaks in a follow up work by carrying out a purely systematic work.

Reversibility of the main transition and the pretransition
The stability of the main transition upon successive heating and cooling cycles has been also investigated.Fig. 9 displays an example of In some cases, a broader transition at lower temperatures, attributed to the pretransition, is observed upon heating (see Fig. 6).The pretransition typically takes place between the ripple phase and the liquid-disordered phase, the former being linked to the formation of periodic ripples on the membrane surface [66].Ripples with periodicity ranging from 12 nm to 16 nm appear along the so-called stable ripple phase [67], subject to variations as a function hydration and thermal history [68][69][70].The stable ripple phase is formed upon heating the sample from the gel phase to the liquid phase; a metastable ripple phase typically appears on cooling from the liquid crystalline phase to the gel phase [70].From a biological viewing point, pretransition attracts interest arising from its potential to drive the membrane protein assembly via the so-called 'orderphobic' effect [71].The question whether the pretransition takes place in solid-supported small vesicles has not been yet clearly answered; note that its existence could be hampered by enhanced curvature or even by the presence of the solid support.Its study on solid-supported membranes is limited to AFM measurements of lipid multilayers [72] and SLBs [73][74][75].For single phospholipid SLBs, the formation of ripples is precluded due to lateral stress from the solid substrate, while in the presence of tris(hydroxymethyl) aminomethane (Tris) in a buffer solution, the ripples reappear [73].The mechanisms behind this phenomenon are not fully understood.To our knowledge, the pretransition in small solid-supported vesicles is restricted to systems supported on silica beads by differential scanning calorimetry [76], where no pretransition could be observed.In our experiments, when the formed layers are intact vesicles, the transition is more clearly visible for LUVs.The latter, as explained above, are a combination of mostly unilamellar and few oligolamellar structures and possess enough free area for the pre-transition to take place.Small vesicles are mostly unilamellar and in the absence of lamellar stacks (geometrical constraints) the pretransition is expected to be weaker and overlap with the main transition, which is already shifted towards lower temperatures for small vesicles [29].For SLBs, the pretransition is very clear for those formed from adsorbed LUVs that ruptured into thin rigid layers.It is thus tempting to state that, despite being rigid and thin, those SLBs consist of several bilayers (multilamellar) where the pretransition can take place and be detectable.
Unlike the main transition, the pretransition is not detected upon cooling.The irreversibility of the pretransition has been observed by Tenchov et al. using time-resolved X-ray diffraction [77].After reaching the liquid disordered phase upon heating, the formation of ripples upon the following cooling might be strongly hindered and require a very long time, thus undercooling occurs.

Conclusions
We have examined the adhesion and phase behaviour of DPPC vesicles onto two types of surfaces, SiO 2 and Au bearing different adhesion levels.Relevant parameters associated with vesicle adhesion, such as vesicle radius and bending modulus, were also varied.
On a given surface, vesicle deformation is promoted for large vesicles at temperatures above their melting (small bending modulus).SiO 2adsorbed vesicles deform to a greater extent than their counterparts on Au, resulting in higher contact area and higher membrane lateral tension, making them more prone to pore formation or rupture and fusion.Numerical calculations based on free energy minimization illustrate the interplay of bending and adhesion contributions into the vesicle shapes of different sizes and bending modulus on surfaces bearing different degrees of adhesion.
The temperature derivatives dΔf/dT and dΔD/dT show clear signatures of the phase transitions during heating and cooling runs.The main transition is reversible and the transition peak size scales with vesicle size.When adsorbed on Au at T < T m , larger vesicles display a more cooperative transition, whereas the transition temperature T m shifts downwards for smaller vesicles, reflecting that a larger curvature decreases the lateral pressure.At T > T m , the extent of vesicle deformation is larger, the interactions with the substrate broaden the transition and increase the transition temperature.When adsorbed onto SiO 2 the main transition is greatly broadened and appears as a double-peak anomaly.The peak size appears larger for intact vesicles than for planar bilayers.The double peak can be explained as a decoupling effect in the melting between the lower part of the membrane that is closer to the surface and the upper part at the membrane-buffer interface.The stronger interaction between the lipid head groups of the proximal leaflet and the solid surface induces a more tightly packed lipid distribution.This combined with the highly confined and orientationally ordered water layer make the adhered part of the membrane melt at a higher temperature than in bulk.

Fig. 1 .
Fig. 1.Left column: the frequency (top panel) and dissipation shifts (bottom panel) are shown for vesicles adsorbed at 16 °C.Right column: the corresponding frequency (top panel) and dissipation shifts (bottom panel) are shown for vesicles adsorbed at 50 °C.Black colour: SUVs on SiO 2 ; red colour: SUVs on Au; blue colour: LUVs on SiO 2 ; green colour: LUVs on Au.

Fig. 2
displays the extrapolated Sauerbrey thickness for LUVs adsorbed at 16 °C and 50 °C onto SiO 2 and Au surfaces.The corresponding figure for SUVs adsorbed at 16 °C and 50 °C onto SiO 2 and Au surfaces is included as Fig. S1 in the Supplementary material.

Fig. 3 .Fig. 4 .
Fig. 3. Extent of vesicle deformation upon adsorption on Au and SiO2 surface above and below the melting temperature of LUVs (left panel) and SUVs (right panel).The red colour corresponds to vesicles adsorbed above melting and blue to the ones below melting.

Fig. 5 .
Fig. 5.The shapes of the adsorbed oblate vesicles obtained by the minimization of the free energy given by Eq. (1), determined for different values of reduced volume v and parameter w.Each row represents a different value of w: 0.4 (row A), 6.4 (row B), 64 (row C) and 640 (row D).The spontaneous curvature C 0 was selected to be zero.The calculated shapes of nonadsorbed vesicles corresponding to w = 0 and the same reduced volumes v are shown in Fig. S10 of the Supplementary material.

Fig. 6 .Fig. 7 .
Fig. 6.Temperature dependence of dΔf/dT (3rd overtone) for LUVs and SUVs adsorbed at T > T m and T < T m on Au-and SiO2-coated quartz surfaces.Black colour refers to supported membranes formed on Au and blue colour to supported membranes formed on SiO 2 .The inset at the bottom right panel shows a magnification of the peak for clarity.

Fig. 8 .
Fig. 8. Close view of the temperature dependence of dΔf/dT (3rd overtone) for SVLs and SLBs formed from precursor LUVs and SUVs adsorbed at T > T m and T < T m on SiO2-coated quartz surfaces.Black, red and green solid lines correspond to gaussian multiple peak fitting results.

Fig. 9 .
Fig. 9. dΔf 3 /dT vs temperature for the 3rd overtone upon successive heating and cooling runs for large DPPC vesicles adsorbed at T < T m .Left panel: LUVs adsorbed on Au, right panel: LUVs adsorbed on SiO 2 .

Table 1
Hydrodynamic mean diameters and polydispersity indexes (PI) obtained by DLS for the DPPC vesicle dispersions used in this work.The number of performed measurements per sample is n = 4.

Table 3
Sauerbrey thickness h and corresponding extent of vesicle deformation Δd values.