Langmuir 2007, 23, 12293-12300
12293
Adhesion of Giant Vesicles Mediated by Weak Binding of Sialyl-LewisX to E-Selectin in the Presence of Repelling Poly(ethylene glycol) Molecules Barbara G. Lorz,† Ana-S. Smith,*,‡ Christian Gege,§ and Erich Sackmann† E22 Institut fu¨r Biophysik, Technische UniVersita¨t Mu¨nchen, D-85748 Garching, Germany, II. Institut fu¨r Theoretische Physik, UniVersita¨t Stuttgart, D-70550 Stuttgart, Germany, and Fachbereich Chemie, UniVersita¨t Konstanz, D-78457 Konstanz, Germany ReceiVed June 20, 2007. In Final Form: August 17, 2007 Prior to establishing tight contact with the endothelium, cells such as leukocytes or cancer cells use the recognition between sialyl-LewisX ligands and E-selectin receptors to establish weak, reversible adhesion and to roll along the vessel wall. We study the physical aspects of this process by constructing a mimetic system that consists of a giant fluid vesicle with incorporated lipid-anchored sialyl-LewisX molecules that bind to E-selectin that is immobilized on the flat substrate. The vesicles also carry a certain fraction of repelling PEG2000 molecules. We analyze the equilibrium state of adhesion in detail by means of reflection interference contrast microscopy and find that the adhesion process relies purely on the formation of one or more adhesion domains within the vesicle-substrate contact zone. We find that the content of ligands in the vesicle must be above 5 mol % to establish specific contacts. All concentrations of sialyl-LewisX above 8 mol % provide a very similar final state of adhesion. However, the size and shape of the adhesion domains strongly depend on both the concentrations of E-selectin (0-3500 molecules/µm2) and PEG2000 (0-5 mol %). At 3500 E-selectin molecules/µm2 and small concentrations of PEG2000, the vesicle-substrate contact is maximized and fully occupied by a single adhesion domain. At concentrations of 5 mol %, PEG2000 completely impedes the specific binding to any substrate. Lastly, an increase in the adhesion strength is observed in systems with identical compositions if the reduced volume of the vesicles is larger.
Introduction Glycoproteins and glycolipids play an important role in many cellular processes,1-6 examples of which include the ovum fertilization, embryo development, and immunological responses.7 In the case of the latter, white blood cells migrate from the bloodstream to specific sites of inflammation or injury where molecular changes on the surface of the blood vessels take place.8 Several adhesion and signaling events are evoked by the inflammatory stimulus and determine which type of leukocyte will migrate. In all cases, the first step of the immunological response is the attachment (or tethering) of circulating leukocytes to the vessel wall by the formation of adhesion contacts. These serve to slow down the motion of leukocytes in the direction of flow and promote the rolling of leukocytes along the vessel wall. This early adhesion is mediated by calcium-dependent lectin receptors of the selectin family (P- and E-selectin)9 located on the membrane of the endothelium cells. These receptors interact * Corresponding author. E-mail:
[email protected]. † Technische Universita ¨ t Mu¨nchen. ‡ Universita ¨ t Stuttgart. § Universita ¨ t Konstanz. (1) Brunk, D. K.; Hammer, D. A. Biophys. J. 1997, 72, 2820-2833. (2) Dong, C.; Cao, J.; Struble, E. J.; Lipowsky, H. H. Ann. Biomed. Eng. 1999, 27, 298-312. (3) Vogel, J.; Bendas, G.; Bakowsky, U.; Hummel, G.; Schmidt, R. R.; Kettmann, U.; Rothe, U. Biochim. Biophys. Acta 1998, 1372, 205-215. (4) Greenberg, A. W.; Brunk, D. K.; Hammer, D. A. Biophys. J. 2000, 79, 2391-2402. (5) Blackwell, J. E.; Dagia, N. M.; Dickerson, J. B.; Berg, E. L.; Goetz, D. J. Ann. Biomed. Eng. 2001, 29, 523-533. (6) Bhatia, S. K.; King, M. R.; Hammer, D. A. Biophys. J. 2003, 84, 26712690. (7) Gabius, H.-J.; Gabius S., Eds. Glycosciences: Status and PerspectiVes, 1st ed.; Chapman and Hall: London, 1997. (8) Alon, R.; Dustin, M. L. Immunity 2007, 26, 17-27. (9) McEver, R. P.; Moore, K. L.; Cummings, R. D. J. Biol. Chem. 1995, 270, 11025-11028.
with glycoprotein counter ligands, exposing a fucosylated and sialylated tetrasaccharidessialyl-LewisX (sLeX)ssituated on the leukocyte.10 The sLeX-selectin pairs associate very quickly and with affinity sufficient to tether cells over distances of up to 100 nm before rupture. The formation of adhesion contacts is regulated by a polymeric layer (glycocalix) that, by extending from the surface of cells into the extracellular space, sterically prevents the unwanted contacts. Upon the establishment of the rolling motion along the endothelium cell surface, leukocytes become strongly attached close to the site of infection due to interactions with receptors of the integrin family. Strong adhesion is followed by the penetration of the white blood cell from the blood stream to the tissue. Similar types of interactions to those described above are necessary for the growth of blood-borne metastasis.11 Here, tumor cells must first leave the cancer tissue and breach the vessel wall. Just as is in the case of white blood cells, tumor cells adhere somewhere else to the vessel wall after circulating in the bloodstream. Here, they penetrate the vessel and come out through the endothelial surface, which is then followed by proliferation and metastatic tumor growth. Again, the initial adhesion contacts and the rolling of the cells are mediated by sLeX-E-selectin recognition. In an effort to combat metastasis, liposomes equipped with specific ligands such as lipid-anchored sLeX are used to block endothelial E-selectin, thus competitively inhibiting the initial step of tumor cell adhesion.12-15 Compared to free ligands, it (10) Moore, K. L.; Eaton, S. F.; Lyons, D. E.; Lichenstein, H. S.; Cummings, R. D.; McEver, R. P. J. Biol. Chem. 1994, 269, 23318-23327. (11) Oku, N.; Koike, C.; Tokudome, Y.; Okada, S.; Nishikawa, N.; Tsukada, H.; Kiso, M.; Hasegawa, A.; Fujii, H.; Murata, J.; Saiki, I. AdV. Drug DeliVery ReV. 1997, 24, 215-223. (12) Kessner, S.; Krause, A.; Rothe, U.; Bendas, G. Biochim. Biophys. Acta 2001, 1514, 177-190.
10.1021/la701824q CCC: $37.00 © 2007 American Chemical Society Published on Web 10/06/2007
12294 Langmuir, Vol. 23, No. 24, 2007
Lorz et al.
adhesion process, the vesicle membrane is enriched with a fraction of repelling molecules (lipid-anchored PEG2000). After control of the system composition is gained, the relevant physical effects that influence the sLeX-E-selectin-mediated adhesion are measured by varying the content of the vesicle and the substrate. Materials and Methods
Figure 1. Schematic view of the investigated model system. The vesicle membrane is functionalized with specific ligands (sialylLewisX-lipids) and lipopolymers mimicking the glycocalix (PEGlipids). The glass substrate consists of a layer of aminosilanes to which E-selectin receptors are physabsorbed. The remaining free surface is made passive with proteins of fat-free milk. The equator of the vesicle is indicated with a dotted line, and the contact zone with the substrate can be associated with the darkly shaded region.
was found that liposomes have the advantage of forming clusters of bonds promoting the efficiency of the inhibition.16 Relatively recently, Zeisig and collaborators studied the inhibiting properties of vesicles carrying repelling molecules such as PEG2000 and/ or different types of lipid-anchored sLeX in vitro.15 Those vesicles with sLeX motives had much more inhibitory potential than the pure membrane vesicles or those vesicles with only PEG2000. However, if sLeX was embedded within the PEG2000 layer with a short anchor, then the inhibition of the resulting vesicles was reduced with respect to vesicles with only short-anchored sLeX (and no PEG2000). The most efficient inhibition was provided by vesicles carrying sLeX at the distal end of the PEG2000 chain and some additional PEG2000. In such vesicles, the PEG layer, which is somewhat shorter than PEG2000-sLeX motives, apparently had no restrictive influence on the binding of sLeX. Enhanced inhibition compared to vesicles with short-anchored sLeX took place probably because of the high flexibility of the sLeX-PEG2000 construct.15 These qualitative observations can be quantitatively understood from the analysis of the thermodynamic equilibrium of a system consisting of a vesicle adhering to a substrate by the formation of sLeX-E-selectin bonds. Building on experiences in mimicking cell adhesion,17-24 we construct our model system from giant fluid lipid vesicles containing lipid-anchored sLeX-molecules as specific ligands for E-selectin receptors physabsorbed on a flat substrate (Figure 1). To provide all elements essential for the (13) Mastrobattista, E.; Storm, G.; van Bloois, L.; Reszka, R.; Bloemen, P. G. M.; Crommelin, D. J. A.; Henricks, P. A. J. Biochim. Biophys. Acta 1999, 1419, 353-363. (14) Vodovozova, E. L.; Moiseeva, E. V.; Grechko, G. K.; Gayenko, G. P.; Nifant’ev, N. E.; Bovin, N. V.; Molotkovsky, J. G. Eur. J. Cancer 2000, 36, 942-949. (15) Zeisig, R.; Stahn, R.; Wenzel, K.; Behrens, D.; Fichtner, I. Biochim. Biophys. Acta 2004, 1660, 31-40. (16) Stahn, R.; Zeisig, R. Tumour Biol. 2000, 21, 176-186. (17) Ra¨dler, J. O.; Feder, T. J.; Strey, H. H.; Sackmann, E. Phys. ReV. E 1995, 51, 4526-4536. (18) Nardi, J.; Bruinsma, R.; Sackmann, E. Phys. ReV. E 1998, 58, 63406354. (19) Kloboucek, A.; Behrisch, A.; Faix, J.; Sackmann, E. Biophys. J. 1999, 77, 2311-2328. (20) Guttenberg, Z.; Lorz, B.; Sackmann, E.; Boulbitch, A. Europhys. Lett. 2001, 54, 826-832. (21) Goennenwein, S.; Tanaka, M.; Hu, B.; Moroder, L.; Sackmann, E. Biophys. J. 2003, 85, 646-655. (22) Cuvelier, D.; Nassoy, P. Phys. ReV. Lett. 2004, 93, 228101. (23) Limozin, L.; Sengupta, K. Biophys. J., doi:10.1529/biophysj.107.105544. (24) Puech, P.-H.; Feracci, H.; Brochard-Wyart, F. Langmuir 2004, 20, 97639768.
Substrates. Preparation. To deposit E-selectin proteins on a glass substrate, the cover slides (Merck, Germany) are first sonicated in methanol and then cleaned following the first step of the RCA method;25 the cover slides are kept for 30 min at 60 °C in a 1:1:5 (v/v) solution of H2O2/ NH4OH/ H2O. The slides are then rinsed thoroughly with ultrapure water (Millipore, France), dried at 70 °C, and kept in a vacuum chamber for at least 12 h. To hydrophobize the substrates, a layer of aminosilanes is deposited26 by immersing the slides for 4 min at 60 °C into a 1% solution of (3-aminopropyl)triethoxysilane in preheated water-free toluene (both from Fluka, Switzerland). To remove unbound aminosilanes, the slides are rinsed five times with toluene, and then dried under N2. Contact angles of water droplets on the slides are determined as a control measurement for successful coating. These cover slides make up the bottom of the measuring chamber (volume of ∼900 µL). To close the chamber, a Teflon frame is pressed onto the cover slide by a metal ring. A recombinant form of human E-selectin, which consists of the extracellular part of the natural protein (Calbiochem), is physabsorbed4 on the aminosilane substrate in the measuring chamber. To ensure consistency of data, a single batch of E-selectin is used for all presented experiments. (Some variation of E-selecin activity is found between different batches.) Proteins are diluted to concentrations of 1-5 µg/mL in buffer solution (10 mM Hepes, 100 mM NaCl, 1 mM CaCl2, 1 mM NaN3 at pH 7.25 and 210 mOsm). A protein solution (900 µL) is incubated for 2 h at room temperature, and the whole chamber is gently mixed on a shaking platform. After being rinsed three times with buffer solution, the substrates are incubated for another hour at room temperature with a 3% solution of fat-free dry milk (Bio-Rad, Hercules) in buffer. The fat-free milk covers the remaining free area of the substrate and thus prevents unspecific adhesion to the substrate. Rinsing the substrates six times with buffer completes the preparation. Characterization. The density of E-selectin molecules on the substrate is estimated using a method described by Brunk and Hammer.1 E-selectin substrates are prepared with solutions of several different protein concentrations (0, 1, 1.5, 2.5, and 5 µg/mL) and incubated with a monoclonal antibody (mouse anti-human E-selectin, Chemicon) at a concentration of 5 µg/mL for 1 h. After the samples are carefully rinsed three times with buffer solution, a rhodaminelabeled secondary antibody (Chemicon), which binds to the first monoclonal antibody, is added and incubated for an additional hour. The samples are rinsed again with buffer, and their fluorescence intensity is determined immediately with an Axiovert 200 inverted microscope (Zeiss, Germany) using a neofluar objective (antiflex, 63×, oil immersion, NA ) 1.3). Excitation is performed with a high-pressure mercury lamp and a fluorescence filter transmitting at 542 nm. The fluorescence signal is detected above 590 nm with a cooled 12-bit camera (Orca-ER, Hamamatsu, Japan), and the data are processed with real-time imaging software.27 To correlate the average fluorescence intensity with a surface density, a calibration measurement is performed. For this purpose, solid-supported membranes of 1-palmitoyl-2-oleoyl-sn-glycero-3phosphocholine (POPC; Avanti Polar Lipid) with 0, 1850, and 3800 molecules/µm2 of N-(6-tetramethylrhodaminethiocarbamoyl)-1,2dihexadecanoyl-sn-glycero-3-phosphoethanolamine (TRITC-DHPE, Chemicon) are prepared and measured in the exactly same way as described for the E-selectin surfaces. A comparison of fluorescence (25) Kern, W.; Puotinen, D. A. RCA ReV. 1970, 31, 187-206. (26) Siqueira Petri, D. F.; Wenz, G.; Schunk, P.; Schimmel, T. Langmuir 1999, 15, 4520-4523. (27) Keller, M.; Schilling, J.; Sackmann, E. ReV. Sci. Instrum. 2001, 72, 36263634.
sLeX-E-Selectin-Mediated Vesicle-Substrate Adhesion
Langmuir, Vol. 23, No. 24, 2007 12295
Figure 3. Thermotropic characterization of the vesicle membrane. The temperature dependence of the heat capacity is investigated with microcalorimetry. The absolute fraction of the sLeX-lipid is displayed. The beginning and end of the calorimetric band define the solidus and the liquidus line, respectively. For temperatures underneath the solidus line, the system is in a crystalline phase. Above the liquidus line, the system is in a fluid state, whereas in between, there is a coexistence of crystalline and fluid phases.
Figure 2. Estimation of E-selectin density on the substrate. E-selectin substrates are prepared from solutions with different protein concentrations and incubated with a monoclonal antibody. After incubation of a secondary antibody, marked with a fluorophore, different fluorescence intensities are observed depending on protein concentration (bottom row of snapshots). Fluorescence intensities are calibrated with lipid bilayers containing known proportions of the same fluorophore used for the E-selectin experiments (top row of snapshots). Average fluorescence intensities of all samples are compared, and the surface density of E-selectin is estimated in the graph. The error bars originate from averaging several pictures taken from the same sample. intensities provides the absolute number of E-selectins on the substrate as shown in Figure 2. The clearly visible increase in fluorescence intensity in the samples corresponds to a growing E-selectin concentration on the substrate, as expected from the preparation procedure. These values are obtained under the assumption that exactly one first antibody binds to each E-selectin and that exactly one second fluorophore-labeled antibody binds to each first antibody.4 Despite the constraints of the analysis, the obtained densities of E-selectin are in good agreement with the measurements of L-selectin coverage on glass substrates.4 This procedure does not determine the real density of E-selectins on the substrate but only the density of the protein that is active after physiabsorption. This is simply because the antibody binding constitutes direct proof of the E-selectin activity. However, it is expected that some inactive protein is present on the substrate because the preparation does not allow for establishing a preferred orientation of E-selectin, which will result in some fraction of the protein turning the binding area toward the glass and thus being unavailable for binding, either to antibodies or to sLeX. Giant Vesicles. Preparation. The matrix of giant vesicles was prepared from an equimolar mixture of 1,2-dimyristoyl-sn-glycero3-phosphocholine (DMPC) and cholesterol (both from Avanti Polar Lipids). In addition, specific molecules were embedded into this membrane (bending stiffness κ ≈ 100kBT):28 (i) 8 or 15 mol % with respect to DMPC of the naturally occurring sialyl-LewisXglycosphingolipid29 Neu5AcR(2f3) Galβ(1f4)[FucR(1f3)]GlcNAc β(1f3)Galβ(1f4)Glcβ(1f0)Cer (sLeX-lipid) (synthesis described by Gege and co-workers)30 serving as ligands for E-selectin and (ii) 0-5 mol % (with respect to DMPC) of repelling molecule 1,2dimyristoyl-sn-glycero-3-phosphoethanolamine-N-polyethyleneglycol (DMPE-PEG2000, Avanti). This lipid-coupled poly(ethylene glycol) has approximately 45 monomers and a Flory radius of 34.4 (28) Sackmann, E. In Handbook of Biological Physics; Lipowsky, R., Sackmann, E., Eds.; Elsevier Science: Oxford, England, 1995; Vol. 1, Chapter 5. (29) Pauvala, H. J. Chem. Biol. 1976, 251, 7517-7520. (30) Gege, C.; Oscarson, S.; Schmidt, R. R. Tetrahedron Lett. 2001, 42, 377380.
Å. The latter is in the same size range as the sLeX radius of gyration. A set of experiments is performed with shorter DMPE-PEG750 composed of approximately 15 monomers (Flory radius of 17.8 Å). Giant vesicle preparation follows the method of electroswelling.31,32 First, the lipid mixture is dissolved in chloroform and spread onto indium tin oxide electrodes. After the electrodes are completely dry, a condenser is formed in the swelling chamber and filled with 170 mOsm sucrose solution. An ac field of 1 A and 10 Hz is applied for 2 h at a temperature of 40 °C. Characterization. The incorporation of sLeX-lipid into the vesicle matrix is verified by microcalorimetry (VP-DSC, Microcal Inc.).33 For these measurements, multilamellar vesicles are first prepared and then ultrasonificated, which finally results in small unilamellar vesicles.34 Solutions of these small vesicles containing different amounts of sLeX-lipid (5, 10, 20, and 30 mol % with respect to DMPC) are prepared at a concentration of 500 µg/mL. The solution (500 µL) is injected into the calorimeter using pure buffer as a reference. The observed temperature range studied extends from 5 to 40 °C, and the heating rate is 15 °C/h. With increasing percentage of sLeX-lipid, the calorimetric band in the heat capacity broadens. Figure 3 shows the solidus and liquidus lines, which indicate the beginning and the end of the phase transition. The characteristics of the two lines suggest a cigar-shaped phase diagram, as found for a similar lipid mixture of dipalmitoylphosphatidylcholine and lactosylceramid.35 In that system, additional neutron scattering experiments led to the interpretation of the phase behavior as a real mixture of the components. Furthermore, recent studies of the phase behavior of DMPC vesicles containing different types of lipidanchored sLeX show that the compositions used herein are indeed real mixtures.36 All experiments are carried out at room temperature, above the solidus line. Furthermore, the additional large fraction of cholesterol in the vesicle plays a dual rolespartially disrupting the order in the liquid-crystalline phase and hence changing the transition temperature but at the same time making the layer even more fluid. Cholesterol also probably induces some order in the fluid phase, forcing the chains to be extended because it itself is a rodlike molecule sitting in the chain region. Consequently, we set the system in a regime where the vesicle membrane consists of the lipid chains that are still ordered, but the diffusion of lipid is the same as in the fluid phase. (31) Dimitrow, D. S.; Angelova, M. I. J. Electroanal. Chem. 1988, 253, 323336. (32) Albersdo¨rfer, A.; Feder, T.; Sackmann, E. Biophys. J. 1997, 73, 245257. (33) Plotnikov, V. V.; Brandts, J. M.; Lin, L.-N.; Brandts, J. F. Anal. Biochem. 1997, 250, 237-244. (34) Zantl, R. Flu¨ssigkristalle aus DNA und kationischen Lipidmembranen: Struktur und Phasenverhalten. Ph.D. Thesis, Technische Universita¨t Mu¨nchen, 2001. (35) Maggio, B.; Ariga, T.; Sturtevant, J. M.; Yu, R. K. Biochim. Biophys. Acta 1985, 818, 1-12. (36) Gege, C.; Schneider, M. F.; Schumacher, G.; Limozin, L.; Rothe, U.; Bendas, G.; Tanaka, M.; Schmidt, R. R. ChemPhysChem 2004, 5, 216-224.
12296 Langmuir, Vol. 23, No. 24, 2007 Reduced Volume of Vesicles. Two effects arise from the different solutions used as buffer in the measuring chamber and as the swelling solution for the vesicles. The mass density of the inner solution (sucrose, 170 mOsm) is about 50 µg/mm3 larger than the mass density of the outside buffer (10 mM Hepes, 100 mM NaCl, 1 mM CaCl2, 1 mM NaN3 at pH 7.25 and 210 mOsm). Therefore, vesicles precipitate to the bottom of the measuring chamber, enabling close contact of sLeX in the vesicles with the E-selectin on the substrate. The difference in osmolality of 40 mOsm between the buffer and swelling solution ensures that the vesicles deflate by reducing their volume and generate excess membrane area that can be used for the formation of the contact zone. The preparation provides vesicles with reduced volume in the range of 0.75 to ∼1, which by the definition of reduced volume means that the resulting volume is 75 to ∼100% of the volume of the initially spherical vesicle, whereas the surface area remains the same. This spread of the reduced volumes in the sample may partially be a product of some destabilization of the vesicles upon insertion into a buffer with higher osmolality. However, the equilibration of the difference in the osmotic pressure is fast in comparison to time relevant to the formation of the first adhesion clusters. So even if osmotically induced budding takes place, it is before the onset of adhesion. The adhesion process itself generates tension that can, in some cases, build sufficiently to induce pore openings or budding, which in turn changes the reduced volume of the vesicle. However, this can be relatively easily detected by following the global size of the vesicle. Also, if there is an exchange between the inner and outer solution, the contrast of the contact zone in the RICM changes. The later effect has been attributed to the modification of the refractive index of the inner buffer.23 Reflection Interference Contrast Microscopy (RICM). The adhesion properties of several variations of the adhesion system are probed using reflection interference contrast microscopy,37,38 a technique that allows for a reconstruction of membrane conformation in the contact zone of the vesicle and the substrate.39 The microscope used belongs to the Axiomat series of Zeiss (Germany). Pictures are recorded with a CCD camera (C4880-80, Hamamatsu) and processed with imaging software.40 The contour of an adhered vesicle can be reconstructed by applying a piecewise-defined arccos transformation to the interference pattern perpendicular to the contact line up to a height above the substrate of about 1 µm (see below) with a height resolution of 5 nm. The lateral resolution is 250 nm. Evaluation of the Effective Adhesion Strength. An analysis of RICM images is performed to extract the size of the specifically adhered area in the contact zone between the vesicle and the substrate. The adhesion area is identified as nonfluctuating dark patches32 found at a height of less than 30 nm above the substrate. To enable the comparison of the adhesion plate of vesicles of different sizes, the adhesion areas are normalized by the vesicle radius (Figure 1). The latter is obtained from the bright field measurements. The adhesion strength W is calculated following the model of Bruinsma.41 This model has previously proven to be very valuable. However, because this model is built on elastic arguments, it does not provide details of the binding or of the number of formed bonds. However, measuring the macroscopic contact angle R and the capillary length λ from the RIC micrographs enables the calculation of the tension γ by use of γ ) κ/λ2. Here, κ is the bending stiffness of the membrane, and for the given mixture it is taken to be 100kBT. The effective adhesion strength can finally be directly extracted by applying the law of Young-Petit: W ) γ(1 - cos R). (37) Curtis, A. S. J. Cell Biol. 1964, 20, 199-215. (38) Ploem, J. S. In Mononuclear Phagocytes in Immunity, Infection and Pathology; van Furth, R., Ed.; Blackwell Scientific Publications: Oxford, England, 1975. (39) Wiegand, G.; Neumaier, K. R.; Sackmann, E. Appl. Opt. 1998, 37, 68926905. (40) Schilling, J.; Sackmann, E.; Bausch, A. R. ReV. Sci. Instrum. 2004, 75, 2822-2827. (41) Bruinsma, R. Proc. NATO AdV. Inst. Phys. Biomater. 1995, 322 of NATO ASI, 61.
Lorz et al.
Figure 4. Reconstruction of the vesicle contour in the vicinity of the substrate, First, a line profile h(x) is drawn perpendicular to the contact line of the vesicle in the RIC micrograph (top). The intensity is recorded as a function of the distance from the contact point (bottom left), and an arccos transformation is applied (bottom right). Both the macroscopic contact angle R and the capillary length λ are obtained from the reconstructed vesicle shape. Both R and λ can be obtained from the contour of the vesicle in the vicinity of the substrate.18 The latter can be reconstructed from the RICM pattern with 250 nm lateral and 10 nm vertical resolution (Figure 4). For the calculation of the adhesion strength, the mean tension is used (an average value of several line profiles obtained all around the adhesion plate). Evaluation of the Reduced Volume of the Vesicle. We also show that adhesion depends on the reduced volume of the vesicle. To ensure experiments with constant reduced volume, the bright field pictures of adhering vesicles are collected regularly during vesicle adhesion to the substrate. Only vesicles whose overall size did not change between the beginning and the end of the experiment were considered for data analysis. For similar reasons, if a change in a contrast in the RIC micrograph took place, then data were not used for the analysis. To determine the reduced volume, first the effective adhesion strength has been determined for each vesicle within the model of Bruinsma as described in the previous section. Using the scaling w ) WR2/κ (where R is taken to be the bright field radius and κ )100kBT), W is transferred into dimensionless units of the SeifertLipowsky model. For this effective adhesion strength, a family of shapes with different reduced volumes has been calculated by solving the shape problem of Seifert and Lipowsky.42 From these shapes, the ratio of the adhesion zone size and the radius of the vesicle at the equator have been theoretically determined. This ratio is, for fixed adhesion strengths, a unique function of the reduced volume. In the final step, the reduced volume of the vesicle has been determined by evaluating the equivalent ratio from the experimental data and comparing with the theoretical findings. Thereby, the size of the adhesion zone is measured in RICM whereas the radius of the vesicle at the equator is obtained from the bright field picture. Such a procedure gives an estimate of the reduced volume within the error of ∆V ) 0.05.
Equilibrium of Adhesion We study the thermodynamic equilibrium of vesicle-substrate adhesion in a system where ligands (sLeX) and repellers (PEGlipopolymer) are mobile in the vesicle membrane whereas the E-selectin molecules are completely immobile on the substrate. This situation is the experimental analogue of the recently (42) Seifert, U.; Lipowsky, R. Phys. ReV. A 1990, 42, 4768-4771.
sLeX-E-Selectin-Mediated Vesicle-Substrate Adhesion
presented theoretical framework for the thermodynamic equilibrium of the specific adhesion of vesicles.43,44 In these works, Smith and Seifert succeed in bridging microscopic bond formation with the deformation of the enclosed vesicle shape, with a finite number of active molecules. As a result, they provide the number of formed bonds and the effective adhesion strength produced by these bonds as a function of the system parameters, which could, in principle, result in a direct quantitative comparison between the theory and the current experiments. However, because at this stage we are not able to measure the density of bonds within a dark adhesion domain, we do not fit the data with this model. Instead, we use the independent measurement of the effective adhesion strength provided by the model of Bruinsma41 and discuss observed trends with those expected in the thermodynamic limit of Smith and Seifert. The adhesion process in the current system can be divided in two distinct steps. The first, a gravity-dominated step, consists of the approach of the vesicle to the substrate and the creation of a contact zone between the vesicle and the substrate. This is manifested by a formation of a strongly fluctuating, usually circular pattern in the RIC micrograph. The second step begins with the formation of the first specific bonds, which in the case of sLeX-E-selectin can take considerable time (between a few minutes and 1 h). The formation of bonds is associated with the formation of one or a few small adhesion domains appearing as dark patches inside the contact zone on the RIC micrographs. These domains are nuclei for larger adhesion domains that grow by expansion at the edge. As suggested by Mammen and coworkers45 the formation of nuclei is a signature of the cooperation between several bonds that have to bind almost simultaneously to make the first stable adhesion domain, inferring that a single bond is not capable of sustaining strong shape fluctuations. Previous experiences were limited to systems in which adhesion was mediated by ligand-receptor pairs of large binding affinity. For example, in the integrin-RGD-mediated adhesion, where the binding affinity of the pair is about 15 kBT, even moderate concentrations of adhesion molecules result in very fast spreading of the vesicle as soon as it approaches the substrate, circumventing the described first step.19,21 However, very small concentrations promote the formation of small domains.20 The use of the much weaker sLeX-E-selectin binding obviously slows down the process considerably and allows the distinction between the two steps. The binding affinity of the sLeX-E-selectin pair has been estimated in our previous work by the use of the competitive binding of an antibody with a known binding affinity. From that work, we have estimated the binding affinity of the sLeXE-selectin binding pair to be between 5kBT and 7kBT. For this reason, one must expect larger concentrations to be necessary for promoting the adhesion, which indeed is the case. Apart from being slow, the growth of domains is irregular. Parts of the domain edge line move occasionally very rapidly in one direction and then stand still until restarting in another direction, resulting in a rather inhomogeneous adhesion domain. The time for a step in directional growth occurs on a time scale of seconds and is relative to the time necessary to establishing the equilibrium of the entire system. The latter is assumed to be achieved if no binding events are observed for substantial intervals of time (a time scale of minutes). In the following text, we study the equilibrium of the system as a function of the system parameters: the contents of sLeX and (43) Smith, A.-S.; Seifert, U. Phys. ReV. E 2005, 71, 061902. (44) Smith, A.-S.; Seifert, U. Soft Mater. 2007, 3, 275-289. (45) Mammen, M.; Choi, S.-K.; Whitesides, G. M. Angew. Chem., Int. Ed. 1998, 37, 2754-2794.
Langmuir, Vol. 23, No. 24, 2007 12297
PEG in the vesicle and the density of E-selectin on the substrate. We also consider the reduced volume of the vesicle. The role of the reduced volume has not been addressed directly in studies of specific vesicle adhesion thus far. From a theoretical point of view, except for some vesicle simulations,52 only the Smith and Seifert model43 couples the reduced model with the tension directly because the shape equation is an integral part of it. By this, we cover all relevant parameters of the system and provide a framework for controlled sLeX-E-selectin adhesion. Variation of the sLeX Component at Constant PEG and E-Selectin Concentrations. On time scales accessible via experiments, molar fractions below 0.05 sLeX in the vesicle membrane do not mediate specific adhesion, irrespective of the concentrations of PEG and E-selectin. However, although differences in the dynamics of the process occur, the variation of the sLeX content between 8 and 15 mol % does not affect the equilibrium adhesion area and the adhesion strength. Thus, no distinction between data associated with these two concentrations will be made in the remainder of the manuscript. Both of the above concentrations provide an abundance of sLeX with respect to E-selectin, and the adhesion equilibrium is limited by the number of receptors or repelling molecules. A similar effect was recently suggested in the work of Smith and Seifert where this situation is referred to as receptor-dominated equilibrium.43,44 This model predicts that for a given size of a contact zone the number of formed bonds as a function of the ligand content saturates very quickly to its limiting value, with the latter being proportional only to the number receptors on the substrate. Similarly, the effective adhesion strength in this regime, in the first-order expansion, is proportional only to the binding strength and the receptor surface coverage. The ligand concentration is present only in the second (logarithmic) term, which plays a dominant role in the experimentally inaccessible regions of very low and very high ligand contents. Interestingly enough, the concentrations necessary for specific adhesion appear to be in agreement with the sLeX content in leukocytes. Specifically, at 5 mol % sLeX in a vesicle, the surface density is about 50 000 molecules/µm2, whereas the density of sLeX on the surface of leukocytes is similarly about 50 000 molecules/µm2.46 Variation of E-Selectin Coverage at Constant PEG Content. The variation of adhesion properties for the E-selectin density on the substrate was probed for a PEG2000 content of 1 mol % in the vesicles. Microscopic images of the adhesion domains (dark patches in micrographs, Figure 5) show different behavior depending on the receptor density on the substrate: At zero E-selectin concentration, no dark patches are observed, indicating that the patches are indeed the result of specific sLeX-E-selectin binding. At low concentrations of E-selectin, several adhesion domains typically coexist with very inhomogeneous and highly curved edges. For medium receptor concentration, regions of the nonadhered membrane still commonly occur within domains. However, the contact line is more regularly shaped, and the whole adhesion plate is more uniform. An increase in homogeneity becomes more pronounced with increasing E-selectin density. At higher receptor density, the adhesion domain is homogeneous, surrounded by a rather smooth edge. There is only one adhesion domain that expands through the entire contact zone. Finally, at high coverage the contact zone itself, together with the adhesion domain, has reached its maximum size. This maximum is limited only by the reduced volume of the vesicle. (46) Symington, F. W.; Hedges, D. L.; Hakomori, S. J. Immunol. 1985, 134, 2498-2506.
12298 Langmuir, Vol. 23, No. 24, 2007
Figure 5. Adhesion as a function of E-selectin surface density at constant 1 mol % PEG2000 content. Micrographs of the adhered vesicles are shown in the top row. The numbers underneath the micrograph denote the number of E-selectin molecules per square micrometer. The graph shows the normalized adhesion domain area as a function of the surface coverage. A strong increase at low coverage is followed by saturation at high coverage. (Inset) Adhesion strength increases over the whole observed range of surface density. The lines are guides for the eye. The vertical error bar represents the error that arises from the image analysis necessary to extract the specifically bound area.54 The horizontal error bar emerges from the uncertainty in the determination of the density of E-selectins on the substrate.
In the receptor-dominated regime, one expects the number of formed bonds to first increase and then saturate upon increasing the E-selectin surface coverage.43 These expectations are indirectly confirmed by the increase and saturation of the normalized adhesion-domain area as function of the concentration of E-selectin (Figure 5). At a surface coverage above 2500 E-selectin/ µm2, it is true not only that the entire contact zone is occupied by an adhesion domain but also that the contact zone itself is maximized. At this point, the vesicle contact zone and the adhesion area are limited by the geometric properties of the deflated vesicles, and the vesicle assumes the shape of a spherical cap.42,43 The adhesion strength of the same vesicles, calculated with the Bruinsma method,41 is presented in the inset of Figure 5. The average adhesion strength increases almost linearly with the growing surface density of E-selectin, even in the region of saturated adhesion area. This is again in good agreement with the predictions of Smith and Seifert.43,44 In the latter model, for the system in ligand-dominated equilibrium (abundance of ligands), the effective adhesion strength is predicted to have an exactly linear dependence on the increased concentration of receptor surface coverage for a fixed size of the contact zone area. This result emerges from balancing the entropic cost of immobilizing a ligand when forming a bond with the gain in enthalpy when this bond is formed in a system with a finite numbers of ligands (inset in Figure 5). As expected from the composition of the experimental system, at least qualitative agreement is obtained with these theoretical observations. In the other receptor-dominated regime, the effective adhesion strength saturates or can even decrease upon increasing the density of receptors on the substrate. However, to reach this regime, one
Lorz et al.
would need to increase the density surface coverage considerably, which, if performed by physabsorption, may produce very uneven substrates. Variation of PEG Content at Constant E-Selectin Coverage. Placing any polymeric chain between the membrane and the substrate geometrically restricts its gyration, which in turn gives rise to repulsion. Such behavior is very well described by the mean-field Flory theory, which shows that the strength of the repulsion depends on elements such as the radius of gyration of the polymer, the density of chains, the strength of the attractive van der Waals potential, and, in particular, the height of the membrane above the substrate. This kind of repulsion, exhibited by molecules such as PEG2000, has been used extensively to compete with the ligand-receptor attraction and thus control the overall strength of adhesion.18-21 The general understanding arising from these previous works is that the repelling molecules are squeezed out of adhesion domains, which in turn become phase-separated agglomerates of ligand-receptor bonds. In this way, the size of the domains is limited by the lateral osmotic pressure produced by repelling molecules at the edge of the domain, which is driven by the concentration difference inside and outside the domain.47 Increasing the content of repelling molecules is then expected to exert a larger osmotic pressure at the edge of the domains, resulting in smaller and smaller domains overall. In the current system, elements of such behavior are clearly seen when the adhesive properties of vesicles are explored by varying the content of PEG2000 at high E-selectin surface density (3250 ( 500 E-selectin/µm2). A significant decrease in the normalized area of adhesion domains can be observed when the PEG2000 content is increased to 3 mol %. Rather than forming large domains typical of very low PEG2000 content, the adhesion is now reduced to small domains and a certain number of adhesion points. At a lipopolymer content of 5 mol %, the vesicles exhibit no specific adhesion. The corresponding effective adhesion strength shows an almost linear decrease in adhesion strength in response to increasing polymer content (between 0 and 5 mol % PEG2000, see the inset of Figure 6). As can be seen in Figure 6, adhesion domains form also at 0% PEG2000, and unlike in previous systems, a change in the PEG2000 content from 0 to 1 mol % influences the size of the adhesion domain very little (Figure 6). This means that PEG2000 cannot be solely responsible for the formation of domains but that other effects play a major role. For example, the domain size is very sensitive to the E-selectin concentration on the substrate (as previously discussed and shown in Figure 5, where the PEG2000 content is 1 mol %), but fluctuations could also drive the bond aggregation. Furthermore, as will be discussed in the following text, additional repulsive effects arise from polymeric chains (other than PEG2000) present in the system. To allow interactions with receptors, active ligand motives (sLeX) are anchored to a spacer, which itself is a polymeric chain residing above the vesicle membrane.30 Placed between the membrane and the substrate, such a construct provides exactly the same repulsion as any other polymeric chain (e.g., PEG2000). Consequently, every free ligand in the contact zone also acts as a repelling molecule. Becasue the radius of gyration of the lipidanchored sLeX is about the same as the radius of gyration of PEG2000, the repulsion strengths of the two species are very similar. This repulsive action of ligands was much less important in the previous vesicle-substrate systems because the ligands were (47) Bruinsma, R.; Behrisch, A.; Sackmann, E. Phys. ReV. E 2000, 61, 42534267.
sLeX-E-Selectin-Mediated Vesicle-Substrate Adhesion
Figure 6. Evolution of adhesion properties with increasing polymer content on a substrate with 3250 ( 500 µm-2 E-selectin and 8 or 15 mol % sialyl-LewisX-lipid. The normalized adhesion area is constant for 0 and 1 mol % PEG 2000 and then suddenly decreases (as also clearly visible in the micrographs on the top). At 3 mol % PEG2000, pointlike adhesion is observed, and at 5 mol %, no adhesion area is detected. (Inset) Adhesion strength decreases monotonically with increasing PEG2000 concentration. A set of data for vesicles with same content but with PEG750 are shown with crosses. The lines serve as guides for the eye. The error bar is the uncertainty in extracting the specifically bound area.54
shorter, present in much smaller concentrations, and, for the most part, participated in specific adhesion (as a result of the high binding affinity).18-21 Indeed, the high binding affinity in these systems implied that the attraction completely dominated the repulsive attributes of the ligands, allowing them to find their binding partners immediately (also circumventing the slow stages of adhesion). At 0% PEG2000, even at small concentrations of ligands, the vesicles would break because of the adhesion-induced lysis tension. In the current system, the combination of the high concentration of sLeX in the vesicles with the low sLeX-Eselectin binding affinity implies the presence of many unbound sLeX molecules in the contact zone,43 which enables the observation of specific adhesion in the absence of PEG2000. In this case, the equilibrium adhesion emerges from balancing not the concentration gradient but the chemical potential of ligands inside and outside the domain. It has been shown recently that the effect of repelling molecules embedded in vesicles with a finite total area is largest if the repelling strength of an individual polymer is zero, such that the polymer can freely explore its conformational space.44 Although counterintuitive, this result is easy to understand from a simple thermodynamic point of view. If the contact zone is very close to the substrate, then the repulsion of a single polymer is strong. Consequently, the balance between entropy and repulsion in the finite geometry of a vesicle implies that repelling molecules must leave the contact zone and, on average, populate the part of the vesicle where the membrane is far away from the substrate. However, if the distance between the substrate and the membrane in the contact zone is large, then the repulsion of the single chain is small, and it is entropically favorable for it to remain in the contact zone. It is our hypothesis that the regulation of the domain sizes and the adhesion strength by PEG2000 is not the result of only the lateral osmotic pressure at the edge of the domain, which acts as an excluded area. Because the vesicle is a finite system, it also
Langmuir, Vol. 23, No. 24, 2007 12299
depends upon the competition between sLeX and repelling molecules for the area within the contact zone. The concentrations of PEG2000 and sLeX in the contact zone are not fixed but are interdependent. From the entropic point of view, a free sLeX molecule has the same probability of being in the contact zone as PEG2000 because of their very similar radii of gyration. In the current experiments, the free membrane in the contact zone averages about 80 nm above the substrate, whereas the adhesion domain is at about 30 nm. Consequently, PEG2000 is expected to be completely expelled from the domain but can reside in the remainder of the contact zone. Increasing the fraction of PEG2000 while keeping the fraction of sLeX constant necessarily results in a net decrease in the concentration of sLeX molecules in the contact zone. As a result, the probability that an sLeX motive finds an E-selectin molecule decreases, which in turn reflects the number of formed bonds and the effective adhesion strength. However, this also increases the previously discussed osmotic pressure contribution of PEG2000 at the edge of a bondaggregated domain placed within the contact zone.47 Our hypothesis is supported by measurements of the adhesive properties of vesicles containing a shorter lipopolymer (PEG750). Its radius of gyration is about half that of the lipid-anchored sLeX or PEG2000. In the presence of PEG750, the effective adhesion strength does not decrease for lipopolymer concentrations between 0 and 2 mol %. This is most likely because the area used by these polymers outside of the domain is much smaller, but there is a possibility that polymers are sufficiently small to reside inside the adhesion domains. In either case, the concentration of free sLeX is not affected, and the lateral osmotic pressure built by PEG750 is too small to reduce the adhesion strength noticeably. The above conclusion is supported by data that can be inspected in the inset of Figure 6. Here, even though a large spread of points at 2 mol% PEG750 is evident, the mean effective adhesion strength remains unchanged between 0 and 2 mol% PEG750. Furthermore, this spread is induced by a very large difference in the reduced volume in data, the effect of which will be discussed in detail in the following section. The vesicle with the lower effective adhesion strength was determined to have a reduced volume of 0.75 whereas the one with the higher effective adhesion strength has a volume of 0.98. Both of these volumes are quite far from the average 0.85 reduced volume, inducing the discussed spread in the effective adhesion strength. Influence of Reduced Volume. It is generally accepted that the membrane tension has a major impact on the adhesion of flat membranes48 and vesicles49 to the substrate. The reduced volume is intimately related to the tension in the vesicle membrane, which makes the preparation-controlled volume an important parameter. Furthermore, the amplitude of thermal undulations is related to the tension and thus to the reduced volume. Consequently, the membrane fluctuates much more in strongly deflated vesicles than in almost spherical ones. Because ligandreceptor binding is influenced by the membrane undulations at the single-bond level, the role of the reduced volume is expected to be even more relevant than in the case of the nonspecific adhesion. Although in experiments this parameter can be regulated during preparation, the role of the reduced volume has not yet been studied systematically. Previously, the reduced volume could be determined only with confocal microscopy in complementary measurements, which can lead to considerable errors,50,51 and in the case of (48) Helfrich, W.; Servuss, R.-M. Il NuoVo Cimento 1984, 3, 137-151. (49) Seifert, U. Phys. ReV. Lett. 1995, 74, 5060-5063. (50) Nardi, J. Ph.D. Thesis, 1998, Technische Universita¨t Mu¨nchen. (51) Guttenberg, Z. Ph.D. Thesis, 2001, Technische Universita¨t Mu¨nchen.
12300 Langmuir, Vol. 23, No. 24, 2007
Figure 7. Influence of the reduced volume. The adhesion strength increases for increasing reduced volume up to V ) 0.9. Above this value, the dependency apparently switches. Different symbols denote different compositions of the system; both receptor density on the surface and polymer content and length are changed. Measuring points that belong to systems of equal composition are connected with dotted lines.
nonspecific adhesion.52 Because the preparation of vesicles often leads to a certain range of reduced volumes in a sample, the average was then found, and the volume was taken as a constant in the adhesion experiments. In this article, by using the theoretical reconstruction of global vesicle shapes, we deduce the reduced volume of the adhered vesicles directly from the microscope images for each vesicle. Although our current procedure is subject to considerable quantitative errors, some qualitative trends can be seen in Figure 7, where the adhesion strength is determined for vesicles with different reduced volumes in five different samples. Within a given sample in a range of reduced volumes between 0.75 and 0.9, more deflated vesicles show lower average adhesion strength. The adhesion strength W depends on the contact angle R and the average membrane tension γ following the law of YoungPetit: W ) γ(1 - cos R). Thus, the shift in the adhesion strength with increasing reduced volume can be either a geometrical effect imposed by a change in the contact angle or an effect of the average membrane tension. Data analysis reveals that the dependencies of the average membrane tension γ and the adhesion strength on the reduced volume have the same form, which implies that the shift in the adhesion strength is mainly a result of a change in the membrane tension. Similar effects have recently been measured for weakly adhered vesicles interacting nonspecifically with a flat substrate52 and in combination with data arising from simulations used for the determination of the membrane bending rigidity. When the reduced volumes exceed 0.9, the adhesion strength apparently decreases with increasing reduced volume. However, as the reduced volume increases, the vesicles become more spherical, diminishing their contact area with the substrate. Eventually, the adhesion area becomes so small that it is hard to get accurate values for the adhesion strength by use of the Bruinsma model. For similar reasons, the shape reconstruction becomes increasingly difficult. The errors arising from the analysis procedure may thus be too large to make definite conclusions in this regime.
Conclusions In this work, we have quantitatively characterized the entire parameter space controlling the equilibrium adhesion state of a system that consists of a substrate exposing immobilized E-selectin that binds to lipid-anchored sLeX ligands diffusing in (52) Gruhn, T.; Franke, T.; Dimova, R.; Lipowsky, R. Langmuir 2007, 23, 5423-5429.
Lorz et al.
deflated vesicles. The latter also contain PEG-repelling molecules to mimic the role of the glycocalix. We find that a very different range of parameters characterize this system than those associated with the vesicle adhesion mediated by other ligand-receptor binding pairs such as biotin-avidin, biotin-streptavidin, or integrin-RGD. In the latter systems, there was not more than 1 mol% active protein in the vesicle. The specific vesicle adhesion was established very promptly because of the very strong binding affinity (15kBT-25kBT), which is an order of magnitude more than in our system. In the sLeX-E-selectin system, we observe a highly cooperative binding behavior of sLeX that, because of its relatively weak binding affinity (about 5kBT) to E-selectin, has to be present in abundance (5-15 mol %) to mediate the specific adhesion of strongly fluctuating vesicles. A strong variation of the adhesion state with the surface density of E-selectin (0-3500 molecules/µm2) has been found, both in the presence and in the absence of repelling molecules. Repelling molecules of type PEG2000 are found to regulate adhesion when present in vesicles at concentrations of less than 5 mol %. At higher concentrations, the PEG2000 molecules are found to impede adhesion completely. These results are attributed to the successful competition of PEG2000 with sLeX for the area in the contact zone. (The radius of gyration of PEG200 is very similar to that of anchored sLeX.) The much shorter PEG750 molecules are found not to have any impact on adhesion. Last, it is found that the reduced volume has a non-negligible effect on the adhesion strength of specifically adhered vesicles. In general, the adhesion strength grows in response to higher reduced volumes, an effect that is attributed to the change in the intrinsic tension of the vesicles. Apart from providing a better understanding of multicomponent vesicle-substrate adhesion, the system presented herein provides a solid foundation for a variety of problems that can be addressed once the equilibrium conditions are correctly mapped. For example, this system has already been successfully used for modeling de-adhesion mechanisms. In particular, progress was achieved in studies of adhesion under force53 and the inhibition of sLeX binding by competitive inhibitors such as antibodies for E-selectin.54 Furthermore, because vesicles of very similar compositions (although smaller in size) have been used as a potential means to combat the blood-borne metastasis of cancer,12-16 the results of this work should provide a better understanding of the inhibition mechanism and allow for tuning the vesicle composition so as to obtain the best impact. However, for a full understanding of the vesicle adhesion process, the dynamical aspects of the interplay between the polymeric repulsion, specific adhesion, and intrinsic tension must be addressed, an undertaking that is currently in progress. Acknowledgment. We benefited from discussions with U. Seifert and K. Sengupta. B.G.L. and E.S. were funded by the SFB 563-C4. A.-S.S. was funded by DFG SE 1119/2-1 and is grateful to S. Tomic´ and T. Vuletic´ of the Institute for Physics, Croatia, for their hospitality during the preparation of this article. We thank W. Feneberg and D. Smith for their critical reading of the manuscript. LA701824Q (53) Smith, A.-S.; Lorz, B. G.; Goennenwein, S.; Sackmann, E. Biophys. J. 2006, 90, L52-L54. (54) Smith, A.-S.; Lorz, B. G.; Seifert, U.; Sackmann, E. Biophys. J. 2006, 90, 1064-1080.
