Vesicle Adsorption and Phospholipid Bilayer Formation on

Mar 16, 2010 - chemistry, we studied two cases: (1) the bare gold surface at the ... applications utilizing patterned lipid bilayers and highlight the...
10 downloads 0 Views 2MB Size
Download PDF
J. Phys. Chem. B 2010, 114, 4623–4631

4623

Vesicle Adsorption and Phospholipid Bilayer Formation on Topographically and Chemically Nanostructured Surfaces Indriati Pfeiffer,*,† Sarunas Petronis,‡ Ingo Ko¨per,† Bengt Kasemo,‡ and Michael Za¨ch*,‡ Max Planck Institute for Polymer Research, Ackermannweg 10, D-55128 Mainz, Germany, and Department of Applied Physics, Chalmers UniVersity of Technology, SE-41296 Gothenburg, Sweden ReceiVed: August 27, 2009; ReVised Manuscript ReceiVed: January 27, 2010

We have investigated the influence of combined nanoscale topography and surface chemistry on lipid vesicle adsorption and supported bilayer formation on well-controlled model surfaces. To this end, we utilized colloidal lithography to nanofabricate pitted Au-SiO2 surfaces, where the top surface and the walls of the pits consisted of silicon dioxide whereas the bottom of the pits was made of gold. The diameter and height of the pits were fixed at 107 and 25 nm, respectively. Using the quartz crystal microbalance with dissipation monitoring (QCM-D) technique and atomic force microscopy (AFM), we monitored the processes occurring upon exposure of these nanostructured surfaces to a solution of extruded unilamellar 1-palmitolyl-2-oleoyl-sn-glycero-3phosphocholine (POPC) vesicles with a nominal diameter of 100 nm. To scrutinize the influence of surface chemistry, we studied two cases: (1) the bare gold surface at the bottom of the pits and (2) the gold passivated by biotinamidocaproyl-labeled bovine serum albumin (BBSA) prior to vesicle exposure. As in our previous work on pitted silicon dioxide surfaces, we found that the pit edges promote bilayer formation on the SiO2 surface for the vesicle size used here in both cases. Whereas in the first case we observed a slow, continuous adsorption of intact vesicles onto the gold surface at the bottom of the pits, the presence of BBSA in the second case prevented the adsorption of intact vesicles into the pits. Instead, our experimental results, together with free energy calculations for various potential membrane configurations, indicate the formation of a continuous, supported lipid bilayer that spans across the pits. These results are significantly important for various biotechnology applications utilizing patterned lipid bilayers and highlight the power of the combined QCM-D/AFM approach to study the mechanism of lipid bilayer formation on nanostructured surfaces. Introduction Supported phospholipid bilayers (SPBs) constitute a versatile model system of the cell membrane and have been used for various biotechnological applications ranging from biosensors,1-3 immunoassays,4 and biomolecule-based microelectromechanical systems (bio-MEMS)5,6 to studies of signal transduction pathways in cell biology.7,8 The chemical and physical properties of SPBs closely resemble those of the cell membrane, and SPBs thus provide a near-native environment for immobilization templates of receptor proteins, thereby maintaining their biorecognition sites and function. As previously reported, the two most common ways to form SPBs on smooth and relatively flat hydrophilic surfaces (SiO2, Si3N4, glass, and mica) are by utilizing Langmuir transfer techniques (Langmuir-Blodgett or Langmuir-Scha¨fer) or by exploiting the adsorption and spontaneous or induced rupture of small unilamellar lipid vesicles (SUVs).9,10 The latter process has been thoroughly investigated using different surface analytical techniques such as surface plasmon resonance (SPR),11,12 quartzcrystalmicrobalancewithdissipationmonitoring(QCM-D),13,14 ellipsometry,15 and electrochemical impedance spectroscopy (EIS).16,17 Using QCM-D, the effects of temperature, lipid chemistry, and surface chemistry on SPB formation on SiO2 or mica-coated quartz crystals have been identified. This was done * Authors to whom correspondence should be addressed. E-mail: [email protected] (I.P.), [email protected] (M.Z.). Tel.: +49-(0)6131-379566 (I.P.), +46-(0)31-7723368 (M.Z.). † Max Planck Institute for Polymer Research. ‡ Chalmers University of Technology.

by introducing a solution of vesicles made of a single type of lipids or a mixture of zwitterionic or charged lipids.14,18 The advantage of using QCM-D among other techniques mentioned previously is the ability of this method to distinguish the spontaneous rupture of adsorbing vesicles from the adsorption of intact vesicles.13 Nevertheless, to follow the kinetics of the vesicle-to-bilayer transformation process in real time and to understand the underlying mechanisms including adsorption and spontaneous or induced rupture of lipid vesicles to form an SPB, QCM-D needs to be combined with other, complementary surface analytical techniques. These include SPR or imaging techniques such as atomic force microscopy (AFM).19 AFM allows vesicle adsorption and structural rearrangements on the substrate to be observed with nanometer spatial resolution.20,21 Based on the results obtained by combined QCM-D/SPR and AFM imaging and supported by Monte Carlo simulations,22 it has been proposed that the energetically unfavorable edges of bilayer patches, as formed by the rupture of an ensemble of vesicles, interact and promote rupture of adjacent intact vesicles adsorbed on a silicon dioxide (SiO2) surface. This process leads to the autocatalytic formation of an SPB once a critical vesicle coverage on the surface has been reached. Furthermore, by combining the results of QCM-D and AFM, it was possible to determine the actual surface coverage at the critical vesicle coverage to be (33 ( 2%),19 which demonstrates the power of using multitechnique approaches to study SPB formation. Beyond Si-derived surfaces, vesicle adsorption and SPB formation has also has been investigated on Au19,23,24 and TiO2 surfaces.25,26 Au is an attractive material for electrical biosensor

10.1021/jp908283g  2010 American Chemical Society Published on Web 03/16/2010

4624

J. Phys. Chem. B, Vol. 114, No. 13, 2010

systems because it is conductive and can function directly as an electrode. As reported by Reimhult et al.,19 vesicles were adsorbed intact but flattened on the surface of Au, with a widthto-height ratio of approximately 3:2. This observation suggests the presence of an attractive interaction between the vesicles and the Au, although the strength of the interaction is not sufficient to lead to vesicle rupture. In contrast, on TiO2, which constitutes the surface layer of frequently used Ti-based implant materials, the adsorption and rupture of vesicles to form a bilayer could be observed when using negatively charged vesicles in the presence of Ca2+ ions.26,27 In this case, the high electrostatic repulsion exhibited by TiO2 prevented vesicles from adsorbing onto the surface in the absence of Ca2+. However, the addition of Ca2+ ions helped to overcome this energy barrier and promote vesicle adhesion to the surface. Thus, these examples convincingly highlight the important role of surface chemistry in the adsorption of vesicles and their eventual transformation into an SPB. In addition to surface chemistry, surface topography also affects the SPB formation process.28 As we have recently reported, the presence of well-defined nanoscale topography on silicon dioxide surfaces significantly alters the kinetics of the vesicle-to-bilayer transformation.29 By combining QCM-D and AFM, we observed that the influence of surface topography depends on the perimeter length of the topographical features (pits) and the vesicle size. For 30- and 100-nm vesicles, vesicles that adsorb close to or on the edges of the pits were hypothesized to rupture spontaneously or to be in a state of high deformation and thus to be more susceptible to rupture, respectively. In the latter case, this resulted in accelerated SPB formation, whereas in the former case, the overall kinetics was hampered.29 A similar topographical effect was also reported by Okazaki et al.,30 who observed rapid vesicle fusion due to the interaction between the vesicles and the edges of micropatterned polymeric lipid areas created on a glass substrate. Currently, many biotechnological applications utilize surfaces that are created specifically to have combined surface chemistry and nanotopographical features. Applications include the development of selective patterning methods for biomolecules (such as protein and DNA arrays) for biosensor and immunoassay purposes31,32 and lipid bilayer patterning for cell attachment and tissue engineering studies.33,34 Therefore, the question regarding the effect of combined surface chemistry and nanotopographical features on vesicle adsorption and the transformation into a bilayer on such surfaces needs to be addressed. In this article, we report on the adsorption of POPC vesicles and their transformation into bilayers on pitted surfaces that were specifically designed to have two different surface chemistries: Au at the bottom of the pits and SiO2 everywhere else (see Figure S1A in the Supporting Information). As previously described, Au and SiO2 display different interactions with the vesicles, which makes this combined surface chemistry an interesting model system. We studied two cases: (1) introduction of vesicles onto the bare nanostructured Au-SiO2 and (2) introduction of vesicles onto nanostructured Au-SiO2 where biotinamidocaproyl-labeled bovine serum albumin (BBSA) had been selectively immobilized on the Au areas prior to vesicle adsorption in a controlled way31 (Figure S1B in the Supporting Information). BBSA was earlier shown to prohibit or strongly reduce the adsorption of vesicles to Au surfaces. This should lead to a distinctively different lateral rearrangement of lipid vesicles on the BBSA-modified surface and thus is expected to modify the bilayer formation process markedly as compared to that on the bare Au-SiO2 surface.

Pfeiffer et al. Materials and Methods Materials. All buffers used in the experiments contained 10 mM tris(hydroxymethyl)aminomethane (Tris) and 100 mM NaCl, both obtained from Sigma-Aldrich GmbH, Munich, Germany. The pH of the buffer was adjusted to 8.0 by addition of HCl. Biotinamidocaproyl-labeled bovine serum albumin (BBSA) was purchased in the form of lyophilized powder from Sigma-Aldrich GmbH, Munich, Germany. The phospholipid 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) in the form of lyophilized powder and polycarbonate membranes with 100-nm pore size were purchased from Avanti Polar Lipids, Alabaster, AL. The Au-coated QCM-D sensor crystals with 5 MHz resonant frequency were obtained from Q-Sense AB, Gothenburg, Sweden. Suspensions of polystyrene particles with a 107-nm diameter were acquired from Interfacial Dynamics Corporation, Eugene, OR. Poly(diallyl dimethyl ammonium chloride) (PDDA; 20% w/w, MW ) 200000-350000) and poly(sodium 4-styrenesulfonate) (PSS; MW ) 70000) were purchased from Sigma-Aldrich Sweden AB, Stockholm, Sweden. Chloroform was obtained from Merck KgaA, Darmstadt, Germany, and sodium dodecyl sulfate (SDS) was obtained from Sigma-Aldrich GmbH, Munich, Germany. Protein Preparation. The stock solution of BBSA was prepared by dissolving the protein in milli-Q water to reach a concentration of 1 mg/mL. The solution was then distributed into Eppendorf tubes, each containing 50 µL protein aliquots. These aliquots were kept in a freezer at -8 °C and were thawed immediately prior to use. Vesicle Preparation. Five milligrams of POPC lipids were dissolved in chloroform in a round-bottomed flask. The chloroform was evaporated using a slight overpressure of N2 gas for 1 h while a thin dried lipid film formed on the wall of the flask. Then, 1 mL of buffer was added to the dry lipid film to achieve a final concentration of 5 mg/mL. Using the extrusion method to form 100-nm vesicles, the solution was then passed back and forth 21 times through a polycarbonate membrane with the corresponding pore size. The size distribution of the resulting vesicles in solution was 160 ( 40 nm as determined using an ALV dynamic light scattering system equipped with a krypton ion laser (λ ) 647.1 nm, Spectra Physics KR2025). Nanofabrication on Pitted Surfaces. The colloidal lithography technique was applied to create SiO2 pitted surfaces on Au-coated QCM-D sensor crystals (Figure S1A in the Supporting Information). Two types of polyelectrolytes, PDDA and PSS, both at a concentration of 2% w/w, were used to create a triple layer of PDDA/PSS/PDDA on the sensor crystals. Thereafter, a polystyrene particle suspension (0.2% v/v, average particle diameter ) 107 nm) was deposited on top of the polyelectrolyte layer and left to adsorb for 60 s. The excess colloidal particles were rinsed away with milli-Q water, and the surface was blowdried with N2. Using an AVAC HVC-600 thin film deposition system, a thin adhesive layer of Ti (1 nm) was deposited on top of the sensor crystals covered with a sparse polystyrene particle mask, followed by electron-beam evaporation of a 24-nm-thick SiO2 film. The polystyrene particles were then removed from the surface by tape stripping, followed by thorough cleaning with isopropanol and milli-Q water in an ultrasonic bath. The homogeneity and the coverage of the resulting pitted surfaces were then characterized by AFM (Dimension 3100, Digital Imaging), yielding around 12% pit coverage.35 Preparation of Surfaces. Prior to each measurement, the QCM-D sensor crystals with pitted surfaces were cleaned by immersing them for 2 days in 0.4% w/w SDS solution and then

Vesicle and SPB Formation on Nanostructured Surfaces applying 5 min of sonication in SDS. This cleaning protocol was required to completely remove BBSA that had adsorbed on the bottom of the pits. Thereafter, the crystals were subjected to two 10-min treatments in a UV ozone chamber (FHR UVOH 150 LAB), followed by rinsing with milli-Q water and blow drying. This treatment formed a well-defined oxide layer on the surfaces and removed organic contaminants, which significantly increased the wettability of the surfaces. QCM-D Experiments. QCM-D measurements were performed using a Q-Sense D-300 instrument (Q-Sense AB, Gothenburg, Sweden). The theoretical details of this technique are described elsewhere.36,37 A clean nanostructured sensor crystal was mounted in the measurement chamber, which was immediately filled with buffer. The temperature of the instrument was set to 22 °C, and the frequency and dissipation baselines were allowed to stabilize. The buffer in the measurement chamber was then replaced by a solution containing 0.2 mg/ mL vesicles. The adsorption of vesicles onto the surfaces was indicated by changes in frequency and dissipation, which were monitored as a function of time until they reached stable values. Thereafter, the measurement chamber was rinsed with buffer to remove excess vesicles. In the second case considered in this study, a 10 µg/mL BBSA solution was introduced into the measurement chamber for 20 min, and the chamber was rinsed thoroughly with buffer prior to the addition of a vesicle solution with the same concentration as in the previous case. The measurements were repeated more than five times in both cases and gave highly reproducible results. AFM Experiments. AFM measurements were carried out using a PicoSPM microscope with a large-area scanner (both from Agilent/Molecular Imaging Inc., Palo Alto, CA) and MSCT-AUNM MicroLever silicon nitride tips (Veeco Europe, Dourdan, France). All cantilevers had a spring constant below 30 pN/nm (according to specifications). A customized, open fluid cell with an O-ring seal against the substrate was used to image the bare, nanostructrured QCM-D sensor crystals in constant-force contact mode at low imaging forces. All subsequent liquid exchanges (BBSA exposure if applicable, vesicle addition, rinsing with buffer) were carried out directly in the AFM fluid cell, and the surfaces were imaged in buffer after each step. The only exception was the end result of the second case, which we characterized by transferring the sensor crystals from the QCM-D apparatus to the AFM fluid cell after completion of a QCM-D experiment. Care was taken to keep the surface wet during the entire transfer. Images were treated and analyzed using commercial software (SPIP version 4.6.0, Image Metrology Inc., Lyngby, Denmark). Unless otherwise specified, images were plane-fitted using the “max. flatness tilt” function of SPIP, and histogram alignment was then applied. IGOR Pro software (WaveMetrics Inc., Lake Oswego, OR) was used to plot cross sections and to determine the exact positions of histogram peaks using Gaussian fits. Results and Discussion In a previous work,29 we showed that the kinetics of the POPC vesicle-to-bilayer transformation on nanostructured SiO2 surfaces is strongly dependent on the ratio between the vesicle size and the size of the nanotopographical features (pits). In this study, we have chosen to keep these two parameters constant (100nm nominal diameter at a ∼1:1 ratio), because these values have been shown to be associated with an enhancing effect of surface topography, that is, to accelerate the bilayer formation process of vesicles adsorbed on SiO2-SiO2 nanostructured surfaces.29 Thus, we focused our investigation on the effect of chemical

J. Phys. Chem. B, Vol. 114, No. 13, 2010 4625

Figure 1. (A) QCM-D curves showing changes in frequency and dissipation (∆f, ∆D) as a function of time and monitoring the kinetics of 100-nm vesicle adsorption and lipid bilayer formation on three different surfaces: flat Au (∆f, black; ∆D, blue), nanostructured SiO2-SiO2 (∆f, light gray; ∆D, purple; from ref 29), and nanostructured Au-SiO2 (∆f, dark gray; ∆D, violet). (B) Schematic illustration summarizing our understanding of the adsorption of vesicles and their transformation into a supported bilayer on a Au-SiO2 surface as a function of time: At t , tmin, intact vesicles adsorb simultaneously on both the SiO2 and Au surfaces. After some time, at t < tmin, vesicles start to rupture on the SiO2 surface as a result of a combination of vesicle-surface interactions, vesicle-vesicle interactions, and the influence of SiO2 pit edges, which enhance vesicle deformation and make the vesicles susceptible to rupture. For t > tmin, the formation of a bilayer on the SiO2 surface is almost completed, and vesicles continue to adsorb onto the Au surface. Finally, at t . tmin, a complete bilayer has formed on the SiO2 surface, and vesicle adsorption on Au has reached saturation.

nanostructuring on vesicle adsorption and bilayer formation on such surfaces. The QCM-D technique was utilized to follow the vesicle adsorption and bilayer formation process in real time, whereas AFM was employed to image and to characterize the state of the substrate at selected stages of the experiment (to visualize the bare substrate, the substrate after BBSA modification, and the substrate after extensive vesicle exposure). Kinetics of Vesicle Adsorption and Bilayer Formation on the Bare Au-SiO2 Nanostructured Surfaces. Figure 1A shows QCM-D graphs acquired upon exposure of unstructured Au, nanostructured SiO2-SiO2,29 and nanostructured Au-SiO2 surfaces to a vesicle solution. By means of the QCM-D technique, the adsorption of vesicles on smooth SiO2 and Au surfaces was investigated thoroughly in our previous work,19 as well as in the works of other groups.14,38 Briefly, the adsorption of 100-nm POPC vesicles on unstructured SiO2 is characterized by a continuous decrease in frequency (and increase in dissipation) until a critical vesicular coverage has been reached (tmin, fmin). Around the critical coverage, the strong interaction between vesicles and SiO2, in combination with vesicle-vesicle interactions, triggers the rupture of vesicles (t > tmin) and leads to the autocatalytic formation of a complete

4626

J. Phys. Chem. B, Vol. 114, No. 13, 2010

Pfeiffer et al.

Figure 2. AFM topography image of a nanostructured Au-SiO2 substrate after prolonged exposure to a vesicle solution, followed by rinsing with buffer and imaging in buffer. The scan size and height scale of the main image are 2 µm and 24 nm, respectively. Whereas the top SiO2 surface appears smooth, with no intact vesicles present, different types of features are observed inside the pits. The inset is a magnification of the area marked by a black square in the main image; it illustrates three different types of features found inside the pits. Pit i contains two clearly discernible, intact vesicles, whereas the vesicles in pit ii seem to be (semi)fused. The feature in pit iii is interpreted as a large vesicle, which results from the fusion of two or more smaller vesicles. The lateral and height scales of the inset are 500 and 22 nm, respectively.

bilayer on the surface (teq, feq). For zwitterionic lipids such as POPC, the typical feq value is -(27 ( 1) Hz. Upon nanostructuring the SiO2 surface with 107-nm pits, we observed a QCM-D response similar to the response on unstructured SiO2, yet with a reduced critical coverage that occurred earlier than for an unstructured surface (f min,SiO2 pits < fmin,SiO2 flat and tmin,SiO2 pits < tmin,SiO2 flat).29 In the case of unstructured Au, the interaction between the zwitterionic vesicles and the surface is not strong enough to cause vesicle rupture, and the adsorption of intact, yet deformed vesicles is observed instead. This condition is depicted by QCM-D curves as a monotonic decrease in frequency until equilibrium is reached (feq,Au; not shown in Figure 1A). Because the measured QCM-D frequency shift includes the adsorption of mass-coupled water, the value of feq,Au is highly dependent on the size distribution and the amount of vesicles adsorbed on the Au surface, and thus, it varies somewhat from experiment to experiment. In Figure 1A, the shape of the QCM-D curves for vesicle adsorption on Au-SiO2 nanostructured surfaces clearly shows the influence of both Au and SiO2 chemistries. Whereas the initial response is similar to the behavior observed on a nanostructured SiO2-SiO2 surface, the QCM-D curves in the region t > tmin are reminiscent of those measured on a plain gold surface. As observed for SiO2-SiO2, the Au-SiO2 curves also display a (local) frequency minimum after an initial frequency decrease. As opposed to the former case, the frequency increase after the minimum, however, is only of short duration and is followed by a frequency decrease toward an equilibrium frequency (feq,Au-SiO2). The frequency shift at the minimum for Au-SiO2 (-45 Hz) is much smaller than the corresponding frequency shift observed for an unstructured SiO2 surface (ca. -80 Hz). We attribute this large difference of 35 Hz to the facilitated rupture of vesicles that are located in the vicinity of the pit edges and that therefore exhibit an increased degree of deformation, as we have shown to be the case for SiO2-SiO2 pits.29 We note that tmin for Au-SiO2 is very similar to tmin for SiO2-SiO2, whereas fmin is somewhat smaller in the former case (fmin,Au-SiO2 ≈ 0.9fmin,SiO2-SiO2). In principle, this affords two possible explanations: (i) fewer vesicles adsorb into the pits of Au-SiO2 as compared to SiO2-SiO2 and/or (ii) the same number of vesicles are adsorbed into the pits in both cases, but these vesicles are

sensed differently by the QCM-D technique in the two cases.39 We cannot exclude one or the other possibility based on the experimental data presented here. We would like to point out, however, that the large difference between unstructured and nanostructured surfaces cannot (exclusively) be explained by the latter mechanism. Whereas the equilibrium frequency shift for Au-SiO2 (approximately -51 Hz) is much more negative than the value for unstructured and nanostructured SiO2 surfaces (approximately -27 Hz), it corresponds rather well to the weighted average for plain SiO2 and plain Au (approximately -55 Hz).40 Our hypothesis is thus that vesicle adsorption on the nanostructured Au-SiO2 surface is governed by two concurrent processes with different time scales, namely (i) the fast, initial adsorption of intact vesicles and subsequent vesicle-to-bilayer transformation on the large SiO2 top surface and (ii) the slow adsorption of intact vesicles onto the Au surface at the bottom of the pits. This hypothesis is supported by AFM images of the Au-SiO2 surface taken after the vesicle adsorption and bilayer formation process was completed (cf. Figure 2). The upper SiO2 surface appears smooth and is covered by a bilayer as confirmed by force spectroscopy; force curves taken in these areas typically showed a characteristic jump of 4 nm, which occurred when the AFM tip penetrated the bilayer at a sufficiently large load.41,42 No intact vesicles were observed on the upper SiO2 surface. Furthermore, we found that the pits were not empty but rather were occupied by features that, in some instances, were reminiscent of several individual, intact vesicles (cf. feature i in Figure 2). A novel piece of information provided by AFM is that the vesicles in the pits seem to have a tendency to fuse, as indicated by features ii, which we interpret as two interconnected vesicles, and iii, which we believe is a larger vesicle resulting from the fusion of two (or more) smaller vesicles. Feature iii is, in fact, representative for the majority of the pits. Although AFM does not allow for a detailed analysis of the structure of these features, we note that the height difference from the top surface to type iii features varies slightly from experiment to experiment and is sensitive to the imaging force. As the imaging force is increased (yet avoiding tip-induced vesicle rupture or bilayer penetration), the height difference increases to a value of roughly 21 nm, which is 4 nm smaller than the original pit depth. This is consistent with a situation in which the top SiO2

Vesicle and SPB Formation on Nanostructured Surfaces surface is covered with a bilayer and the Au surface at the bottom of the pits is covered by two bilayers belonging to a maximally flattened vesicle. This large deformation is caused by the AFM tip, together with the attractive interaction from the pit walls. In most images, we found very few pits with a depth of 29-30 nm, indicating that no lipid material was present inside the pits and confirming that there was indeed a bilayer on the SiO2 top surface. To conclude this section, we schematically summarize our picture of vesicle adsorption and bilayer formation on the bare Au-SiO2 nanostructured surface in Figure 1B. Significant Alteration of Vesicle Adsorption and Bilayer Formation on a Au-SiO2 Nanostructured Surface by Prior Adsorption of BBSA. To mimic the actual conditions of biotechnological applications that often require biological patterning, we also studied the case where BBSA was introduced onto the Au-SiO2 surface prior to vesicle adsorption. As shown previously, when BBSA is introduced onto Au-SiO2 under certain controlled conditions (of concentration and exposure time), the protein preferentially adsorbs on Au, and the SiO2 remains uncovered (observed Au vs SiO2 selectivity higher than 45:2).31 In this study, we used BBSA to passivate Au, that is, to prevent vesicle adsorption onto the Au surface at the bottom of the pits. The injection of a 10 µg/mL BBSA solution onto the Au-SiO2 nanostructured surface was monitored by QCM-D and showed a decrease in frequency and an increase in dissipation of -1.5 Hz and 0.3 × 10-6, respectively (see inset in Figure 3A). Using this frequency shift, we estimated the coverage of BBSA on the Au surface to be around 20 BBSA molecules per pit. After BBSA adsorption and rinsing with buffer, a vesicle solution with the same concentration as in the previous case (0.2 mg/mL) was applied to the surface. Figure 3A shows the QCM-D response of the vesicle adsorption and bilayer formation process on the BBSA-modified nanostructured Au-SiO2 surface, which differs significantly from the response for bare nanostructured Au-SiO2 as discussed above. The difference is particularly clear for the frequency and dissipation values at the end of the process, which were -32 Hz and 1.7 × 10-6, respectively, in the case with BBSA. These values are close to the frequency and dissipation shifts observed for complete bilayer formation on a homogeneous, flat SiO2 surface (-27 Hz, 0.5 × 10-6). From the low dissipation value, we directly infer that only a few vesicles remain intact on the BBSA-modified surface, in clear contrast to the case without BBSA, where intact vesicles were observed inside the pits. To further analyze the altered kinetics due to BBSA modification, we compare the QCM-D traces of BBSA modified Au-SiO2 with bare Au-SiO2 nanostructured surfaces and with our previously reported results on homogeneous SiO2-SiO2 nanostructured surfaces (having the same pit size and coverage) in Figure 3B. Whereas the QCM-D curves of BBSA-modified Au-SiO2 surfaces are similar to those of SiO2-SiO2 nanostructured surfaces in many respects (similar tmin and teq), the frequency minimum (fmin) is different for the three cases. The value of fmin was less negative for BBSA-modified Au-SiO2 (-40 Hz) than for the bare Au-SiO2 nanostructured surfaces (-45 Hz), which we attribute to the BBSA preventing vesicle adsorption on the Au surface. This observation indicates that not all buffer contained inside the pits is included in the QCM reference signal; if that were the case, we would expect the frequency shifts for these two cases to be identical.39 Interestingly, the

J. Phys. Chem. B, Vol. 114, No. 13, 2010 4627

Figure 3. (A) QCM-D curves (∆f, black; ∆D, blue) characterizing vesicle adsorption and bilayer formation on a BBSA-modified Au-SiO2 surface. The error bars for fmin ((8%) and feq ((6%) are minimum and maximum values obtained from eight independent measurements. The inset (i) is an enlargement of the frequency and dissipation response for BBSA adsorption. (B) Comparison of QCM-D curves (∆f, black; ∆D, blue) of vesicle adsorption and bilayer formation on BBSAmodified Au-SiO2 (solid lines), bare Au-SiO2 (2), and nanostructured SiO2-SiO2 (9) during the first 8 min after the introduction of phospholipid vesicles onto the surfaces at t ) 0 min.

frequency shift of the BBSA-modified Au-SiO2 surface at the second inflection point (-28 Hz) was very similar to feq observed on a smooth SiO2 surface (-27 Hz). However, instead of stabilizing at this value, the BBSA-Au-SiO2 curve showed a very slow decrease in frequency and stabilized at -(32 ( 2) Hz, where the error margins mark minimum and maximum values obtained by repeating the experiment eight times on different sensor surfaces. Because we cannot conclusively explain this behavior by analyzing the QCM-D response only, we employed AFM in an attempt to reveal the mechanism of vesicle adsorption and bilayer formation on BBSA-modified Au-SiO2 surfaces. Figure 4A shows the topography of a Au-SiO2 nanostructured surface that had been first exposed to BBSA for 25 min and then to vesicles for around 1 h, as imaged in buffer at five different imaging forces ranging from ca. 100 to 700 pN. To compensate for the varying background due to changes in the imaging force, we have used the histogram alignment function of SPIP to shift all portions of the image to a common background level in Figure 4B. Whereas the pits are only weakly visible for the smallest imaging force, the (apparent) pit depth increases successively as the imaging force is increased. To illustrate this behavior in more detail, we have plotted the cross sections of six randomly chosen, individual pits (thin black lines), as well as the average profile (bold red line), for each imaging force (Figure 4C-G). These

4628

J. Phys. Chem. B, Vol. 114, No. 13, 2010

Pfeiffer et al.

Figure 4. (A) AFM topography image of a nanostructured, BBSA-modified Au-SiO2 substrate after prolonged exposure to a vesicle solution, followed by rinsing with buffer and imaging in buffer at five different imaging forces. The scan size and height scale of this image are 2.5 µm and 84 nm, respectively. The varying image background, which is due to different imaging forces, has been removed in image B, which is shown with a height scale of 50 nm. (C-G) Cross sections across six randomly chosen pits for each imaging force. In addition to the individual cross sections (symbols and black lines), the average profile is shown in red. (H) Apparent pit depth as a function of imaging force, F, showing an initial increase with increasing F. The experimental data points (open circles) follow a power-law dependence for F < 400 pN, as indicated by the solid line, which represents a least-squares fit to the respective data points. For larger forces, the apparent pit depth saturates at 23 nm (dash-dotted line), which is the pit depth that was measured after BBSA adsorption but prior to vesicle exposure.

graphs show that the apparent pit depth increases monotonically with increasing imaging force up to imaging forces of around 350 pN. For imaging forces larger than 350 pN, the pit depth assumes a constant value of ca. 23 nm, which corresponds very well to the pit depth that we measured after BBSA adsorption and prior to vesicle adsorption (cf. Figures S2 and S4 in the Supporting Information). This behavior is in marked contrast to the situation found for pitted surfaces prior to vesicle and BBSA adsorption, where an imaging force of e100 pN is large enough to faithfully reproduce the full pit depth. In particular, we found that the pit depth is not or only weakly dependent on the imaging force for the bare and BBSA-coated pitted surfaces, respectively (cf. Figures S3 and S4, respectively, in the Supporting Information). Based on these observations and consideration of the equilibrium QCM-D frequency shift, our hypothesis is that vesicle adsorption on BBSA-treated Au-SiO2 surfaces ultimately leads to the formation of a continuous bilayer membrane that spans across (most of) the pits.43 To corroborate our hypothesis, we have plotted the apparent pit depth, H, as a function of imaging force, F, in Figure 4H. In the most general case, the elastic response of a pit-spanning membrane upon the action of an AFM tip consists of three major contributions, namely, bending, stretching, and lateral tension.44 The third contribution is completely negligible in our case, that

is, for a solvent-free, supported membrane. For our pit geometry and loading conditions, the contribution from bending as calculated using thin plate theory is small as well, and we thus base our analysis on stretching as the governing contribution (in the terminology of Komaragiri et al.,45 we are in the region of nonlinear membrane behavior). The stretching contribution to F, as measured in the center of the pit, is given by44,45

( )

F(H) ) g(V)RpitEt

H Rpit

3

and thus

H)

(

Rpit3 g(V)RpitEt

)

1/3

F1/3 ) CstretchingF1/3

where Cstretching is a constant; Rpit denotes the radius of the pit; E and t are the elastic modulus and the thickness of the bilayer membrane, respectively; V is the Poisson ratio; and g(V) ) π/3 for V ) 0.33. Taking Rpit ) 53.5 nm and assuming reasonable values for the bilayer thickness (t ) 4 nm) and the elastic modulus of the

Vesicle and SPB Formation on Nanostructured Surfaces

J. Phys. Chem. B, Vol. 114, No. 13, 2010 4629

Figure 5. (A) Four potential membrane configurations resulting from exposure of the BBSA-treated, nanostructured Au-SiO2 substrate to vesicles with a similar size to the pit size. (B) Possible mechanism of pit-spanning bilayer formation based on the rupture and unfolding of an edge-adsorbed vesicle as initiated by adsorption of another vesicle in its vicinity.

lipid bilayer (E ) 40 MPa),46 we expect that Cstretching ) 2.4 nm/pN1/3 in our case. Fitting all experimental data points with F e 350 pN to a power law, we obtain

H (nm) ) 3.4[F (pN)]0.3 Considering that several of the involved parameters are rather uncertain, the fitted values for Cstretching and in particular for the power-law exponent are in good agreement with theory. The observed deviations are attributed to the fact that the boundary conditions used for the theoretical derivation (clamping at the pit edge) might not be completely applicable to our experimental situation and that our measurements were obtained under the combined action of normal and lateral forces, whereas the theoretical model takes only normal (point) forces into account. Note that the model necessarily fails once H approaches the effective depth of the pits; the ultimate pit depth (Hmax ≈ 23 nm) that we measured for forces F g 350 pN is consistent with the spanning membrane being entirely pushed down into the pit (but not being penetrated by the AFM tip). We finally note that the equilibrium frequency shift obtained with QCM [-(32 ( 2) Hz] is also consistent with our hypothesis of a spanning bilayer. As mentioned above, not all of the buffer volume contained in the pits is included in the reference frequency signal. If the pits are, however, covered by a bilayer membrane, all liquid inside the pits is forced to follow the oscillation of the crystal. Consequently, we expect the closing of the pits to be associated with a frequency shift, which explains why the equilibrium frequency shift for a spanning bilayer is somewhat larger than what one typically measures for a supported bilayer membrane on a flat SiO2 surface [-(27 ( 2) Hz]. The contribution of all “trapped liquid” inside the pits should generate a frequency shift of around 20 Hz.47 Yet, we detected only a 5 Hz difference between the spanning bilayer and supported bilayer scenarios, implying that about 75% of the buffer in the pits is already included in the QCM-D reference signal and only 25% gives rise to an additional frequency shift as the pits are covered with a spanning bilayer. Although the formation of pore-spanning membranes has previously been achieved through Langmuir-Blodgett deposition,48 painting of lipids,49,50 or spreading of vesicles with a size largely exceeding the size of the pores,51,52 this is, to the best of our knowledge, the first report documenting pit-spanning

membrane formation through spreading of extruded unilamellar vesicles (EUVs) where pit size and vesicle size are similar to one another. In such a situation, bilayer formation inside the pits cannot be avoided by size exclusion, as was the case in previous reports, and we therefore attempt to speculate on the mechanism underlying pit-spanning membrane formation. First, we checked whether the formation of a pit-spanning bilayer was energetically favorable in our case or whether other configurations would imply a smaller free energy state of the system. To this end, we compared the free energy of four hypothetical cases (cf. Figure 5A), namely (I) a pit-spanning membrane, (II) a bilayer membrane that ends abruptly at the edges of the pits (i.e., neither the pit walls nor the bottom of the pits are covered), (III) a bilayer that covers the pit walls but not the bottom of the pits, and (IV) a bilayer that covers both the pit walls and the bottom of the pits. Comparing configuration II to configuration I, one directly notes that FII > FI because of the necessary introduction of a pore into the bilayer membrane, requiring an energy of ∆FII-I ) FII - FI ) 2πRpitΣ. Here, Fn denotes the free energy, Rpit is the pit radius, and Σ is the membrane line tension. Configuration I is thus (thermodynamically) preferred to configuration II. Configuration III is associated with three contributions to the free energy not present in I, namely, bending of the bilayer membrane across the pit edge, introduction of a “free” bilayer edge at the bottom of the pit, and adhesion to the pit wall. Whereas the latter contribution leads to a reduction of the free energy, the former two contributions increase it. The free energy difference between configurations I and III can be expressed as

∆FIII-I ≈ (κ/2) I dS (2H)2 + 2πRpitΣ - 2πRpitHpitWA where κ is the membrane bending rigidity; dS is an area element; H denotes the mean curvature of the membrane according to H ≡ 1/2(1/R1 + 1/R2), with R1 and R2 being the two (local) principal radii of curvature of the membrane; and WA is the adhesion energy. For our pit geometry, the integral can be evaluated as 2

I dS (2H)

)

(

1 1 + Rpit Redge

)I 2

dS ≈

( ) 1

Redge

2

(π2RpitRedge) ) π2

Rpit Redge

4630

J. Phys. Chem. B, Vol. 114, No. 13, 2010

where we have used the relation R1 ) Rpit . Redge ) R2 and, thus, 1/Rpit , 1/Redge. Using typical values for the membrane parameters,53 namely, κ ) 2.5 × 10-20 J, Σ ) 10-11 J m-1, and WA ) 10-3 J m-2, we found that ∆FIII-I is positive for Redge < 1.3 nm and negative for larger Redge values. Configuration I is thus a thermodynamically more likely configuration than configuration III if the pit edges are sharp enough. Note that the values for the membrane parameters κ, Σ, and WA reported in the literature vary widely, and the threshold value of 1.3 nm should therefore be seen as a rough guideline only. Nevertheless, we argue that the nanofabrication approach used here produces sharp pit edges, thus favoring configuration I over configuration III. Assuming that the adhesion energy between a BBSA-covered gold surface and a phospholipid bilayer is small (WA ≈ 0) and using similar arguments as above, we found that ∆FIV-I ) FIV - FI behaves similarly to ∆FIII-I, that is, configuration I is preferred for sharp pit edges (Redge < 1.5 nm in this case). Note that the pit depth, Hpit, is an important parameter for cases III and IV. Whereas the adhesion energy gain increases linearly with Hpit, neither the bending energy nor the interfacial energy is affected by Hpit. For a given set of parameters, there will thus be a critical pit depth beyond which configurations III and IV become more favorable than configuration I. To summarize, configuration I (i.e., a pit-spanning bilayer) is the thermodynamically most likely configuration of all configurations considered here, as long as the pit edges are sufficiently sharp and the pit depth does not exceed a critical value. The remaining open question is then how the pit-spanning membrane is formed. Comparing the cases with and without BBSA, we first note that an essential feature in the formation of a pit-spanning membrane is the exclusion of vesicle adsorption into the pits. As opposed to earlier reports where the vesicle size significantly exceeded the aperture size, there exists the possibility of vesicles adsorbing across the pit edges in our case. We showed in our earlier work that such vesicles are more susceptible to rupture than vesicles adsorbed onto a flat SiO2 surface, although spontaneous rupture was not observed for the present combination of vesicle size and pit size.29 Our hypothesis is that, once such an edge-adsorbed vesicle interacts with other vesicles, which adsorb to the top SiO2 surface in its vicinity, rupture of these vesicles occurs (cf. Figure 5B). Because the total lipid content of an edge-adsorbed vesicle is more than large enough to cover the entire area of a pit once the vesicle unfolds and because there is no strong thermodynamic driving force for the resulting bilayer to adhere to the pit walls, a pit-spanning membrane is formed. Given the enhanced susceptibility of edgeadsorbed vesicles to rupture, bilayer formation should predominantly start at the pit edges, and the mechanism described above is thus expected to be the governing pathway of pit-spanning membrane formation. An alternative pathway to pit-spanning membrane formation is probably possible through a two-stage process when, in the first stage, the lipid bilayer patches cover the top SiO2 surface, bend into the pits, and merge into an intact bilayer adhering to parts of or the entire pit surface (configuration IV in Figure 5A). In the later stage, because of a weak interaction with BBSA at the bottom of the pits, the bilayer relaxes into the energetically more favorable pit-spanning configuration (configuration I in Figure 5A). Conclusions We have shown that the introduction of Au-SiO2 surface chemistries in combination with nanotopography (pits) signifi-

Pfeiffer et al. cantly changes the kinetics and the mechanism of extruded POPC vesicle adsorption and bilayer formation compared to the well-known supported lipid bilayer formation process on smooth SiO2 and Au surfaces. Although the pit edges promoted rapid bilayer formation, the influence of both Au and SiO2 chemistries on the vesicle adsorption process was clearly visible as two concurrent processes with different time scales: (i) the fast, initial adsorption of intact vesicles and subsequent vesicleto-bilayer transformation on the large SiO2 top surface and (ii) the slow adsorption of intact vesicles onto the Au surface at the bottom of the pits. When the Au-SiO2 nanostructured surface was additionally passivated with BBSA prior to the introduction of vesicles onto the surface, the formation of a continuous bilayer that spanned over the pits was observed. This finding was verified by characteristic changes in the apparent pit depth as a function of AFM imaging force and supported by the equilibrium frequency shift observed by QCM-D, as well as free energy calculations for different potential membrane configurations. Our measurements indicate that the interpretation of QCM-D frequency shifts as measured on pitted surfaces is not quite straightforward because a significant and a priori unknown fraction of the liquid contained inside the pits is already included in the reference frequency signal. Acknowledgment. The authors thank Elisabeth Briand and Isabel Van de Kerre for fruitful discussions. Financial support from VINNOVA BioNano IT program (Grant 2003-00656) and the Alexander von Humboldt foundation is gratefully acknowledged. Supporting Information Available: Schematic illustration of the processing steps required to produce Au-SiO2 nanostructured surfaces using colloidal lithography (Figure S1A). Schematic overview of the two approaches presented in this study (Figure S1B). Image histograms extracted from AFM topography images of a Au-SiO2 nanostructured surface before and after BBSA adsorption (Figure S2). AFM topography images of a bare Au-SiO2 nanostructured surface and the same surface after BBSA adsorption for a series of different imaging forces (Figures S3 and S4, respectively). This material is available free of charge via the Internet at http://pubs.acs.org. References and Notes (1) Jonsson, M. P.; Jonsson, P.; Dahlin, A. B.; Hook, F. Nano Lett. 2007, 7, 3462. (2) Dahlin, A.; Zach, M.; Rindzevicius, T.; Kall, M.; Sutherland, D. S.; Hook, F. J. Am. Chem. Soc. 2005, 127, 5043. (3) Salafsky, J.; Groves, J. T.; Boxer, S. G. Biochemistry 1996, 35, 14773. (4) Larsson, C.; Bramfeldt, H.; Wingren, C.; Borrebaeck, C.; Hook, F. Anal. Biochem. 2005, 345, 72. (5) Sapuri-Butti, A. R.; Butti, R. C.; Parikh, A. N. Langmuir 2007, 23, 12645. (6) Schuy, S.; Janshoff, A. J. Colloid Interface Sci. 2006, 295, 93. (7) Groves, J. T. Angew. Chem., Int. Ed. 2005, 44, 3524. (8) Groves, J. T.; Dustin, M. L. J. Immunol. Methods 2003, 278, 19. (9) Barenholz, Y.; Gibbes, D.; Litman, B. J.; Goll, J.; Thompson, T. E.; Carlson, F. D. Biochemistry 1977, 16, 2806. (10) Sackmann, E. Science 1996, 271, 43. (11) Keller, C. A.; Glasmastar, K.; Zhdanov, V. P.; Kasemo, B. Phys. ReV. Lett. 2000, 84, 5443. (12) Tawa, K.; Morigaki, K. Biophys. J. 2005, 89, 2750. (13) Keller, C. A.; Kasemo, B. Biophys. J. 1998, 75, 1397. (14) Richter, R. P.; Berat, R.; Brisson, A. R. Langmuir 2006, 22, 3497. (15) Faiss, S.; Schuy, S.; Weiskopf, D.; Steinem, C.; Janshoff, A. J. Phys. Chem. B 2007, 111, 13979. (16) Drexler, J.; Steinem, C. J. Phys. Chem. B 2003, 107, 11245. (17) Janshoff, A.; Galla, H. J.; Steinem, C. Biological applications of solid supported membranes on gold surfaces: Quartz crystal microbalance and impedance analysis. In Planar Lipid Bilayers (BLMs) and Their Applications; Tien, H. T., Ottova-Leitmannova, A., Eds.; Membrane Science

Vesicle and SPB Formation on Nanostructured Surfaces and Technology Series; Elsevier Science: Amsterdam, 2003; Vol. 7, Chapter 36, pp 991-1016. (18) Reimhult, E.; Hook, F.; Kasemo, B. Langmuir 2003, 19, 1681. (19) Reimhult, E.; Zach, M.; Hook, F.; Kasemo, B. Langmuir 2006, 22, 3313. (20) Dufrene, Y. F.; Boland, T.; Schneider, J. W.; Barger, W. R.; Lee, G. U. Faraday Discuss. 1998, 111, 79. (21) Li, M.; Chen, M.; Sheepwash, E.; Brosseau, C. L.; Li, H.; Pettinger, B.; Gruler, H.; Lipkowski, J. Langmuir 2008, 24, 10313. (22) Dimitrievski, K.; Reimhult, E.; Kasemo, B.; Zhdanov, V. P. Colloids Surf. B: Biointerfaces 2004, 39, 77. (23) Ekeroth, J.; Konradsson, P.; Hook, F. Langmuir 2002, 18, 7923. (24) Sofou, S.; Thomas, J. L. Biosens. Bioelectron. 2003, 18, 445. (25) Reimhult, E.; Hook, F.; Kasemo, B. J. Chem. Phys. 2002, 117, 7401. (26) Rossetti, F. F.; Bally, M.; Reviakine, I.; Falconnet, D.; Michel, R.; Textor, M. Biophys. J. 2005, 88, 7A. (27) Rossetti, F. F.; Bally, M.; Michel, R.; Textor, M.; Reviakine, I. Langmuir 2005, 21, 6443. (28) Swain, P. S.; Andelman, D. Phys. ReV. E 2001, 63. (29) Pfeiffer, I.; Seantier, B.; Petronis, S.; Sutherland, D.; Kasemo, B.; Zach, M. J. Phys. Chem. B 2008, 112, 5175. (30) Okazaki, T.; Morigaki, K.; Taguchi, T. Biophys. J. 2006, 91, 1757. (31) Svedhem, S.; Pfeiffer, I.; Larsson, C.; Wingren, C.; Borrebaeck, C.; Hook, F. ChemBioChem 2003, 4, 339. (32) Wang, Z. Z.; Wilkop, T.; Cheng, Q. Langmuir 2005, 21, 10292. (33) Groves, J. T.; Boxer, S. G. Acc. Chem. Res. 2002, 35, 149. (34) Mossman, K. D.; Campi, G.; Groves, J. T.; Dustin, M. L. Science 2005, 310, 1191. (35) Hanarp, P.; Sutherland, D. S.; Gold, J.; Kasemo, B. Colloids Surf. A: Physicochem. Eng. Aspects 2003, 214, PII S0927. (36) Hook, F.; Rodahl, M.; Brzezinski, P.; Kasemo, B. Langmuir 1998, 14, 729. (37) Rodahl, M.; Hook, F.; Fredriksson, C.; Keller, C. A.; Krozer, A.; Brzezinski, P.; Voinova, M.; Kasemo, B. Faraday Discuss. 1997, 229. (38) Chot, N. J.; Kanazawa, K. K.; Glenn, J. S.; Frank, C. W. Anal. Chem. 2007, 79, 7027. (39) Note that the QCM-D technique records frequency changes as compared to a reference frequency (“baseline”), which, in the present case, is the oscillation frequency of the nanostructured surface in buffer prior to vesicle adsorption. The interesting question is the extent to which the buffer contained inside the pits contributes to the measured reference frequency. In the extreme case where all buffer inside the pits follows the oscillation of the crystal (and thus is included in the reference measurement), subsequent lipid adsorption into the pits would not give rise to any significant frequency change because the mass densities of buffer and a fully hydrated bilayer

J. Phys. Chem. B, Vol. 114, No. 13, 2010 4631 are very similar; only (lipid) mass added above the level of the top surface would be sensed. In the other extreme case where the buffer inside the pits does not influence the reference signal at all, lipid adsorption into the pits would produce the same frequency change as adsorption of an equal amount of lipids onto the top surface. The most likely case is expected to be intermediate to these two extremes; in such a case, the observed frequency change upon lipid adsorption into the pits would depend on the exact spatial location of the vesicles, as well as their size and deformation (which is known to be different for Au and SiO2). (40) The weights correspond to the relative surface coverage of the respective chemistry: 0.88feq,SiO2 + 0.12feq,Au ) 0.88 × 27 Hz + 0.12 × 263 Hz ) 55 Hz. (41) Franz, V.; Loi, S.; Mueller, H.; Bamberg, E.; Butt, H. J. Colloids Surf. B: Biointerfaces 2002, 23, 191. (42) Liang, X. M.; Mao, G. Z.; Ng, K. Y. S. Colloids Surf. B: Biointerfaces 2004, 34, 41. (43) Although AFM imaging revealed the majority of pits (>98%) to appear as in Figure 4A, that is, covered with a spanning bilayer, we observed intact vesicles to be present in a few pits (