pubs.acs.org/Langmuir © 2010 American Chemical Society
Single-Step Process to Reconstitute Cell Membranes on Solid Supports M. D. Mager and N. A. Melosh* Department of Materials Science and Engineering, Stanford University, Stanford, California 94305 Received February 8, 2010. Revised Manuscript Received February 25, 2010 A new technique is presented to create supported lipid bilayers from whole cell lipids without the use of detergent or solvent extraction. In a modification of the bubble collapse deposition (BCD) technique, an air bubble is created underwater and brought into contact with a population of cells. The high-energy air/water interface extracts the lipid component of the cell membrane, which can subsequently be redeposited as a fluid bilayer on another substrate. The resulting bilayers were characterized with fluorescence microscopy, and it was found that both leaflets of the cell membrane are transferred but the cytoskeleton is not. The resulting supported bilayer was fluid over an area much larger than a single cell, demonstrating the capacity to create large, continuous bilayer samples. This capability to create fluid, biologically relevant bilayers will facilitate the use of high-resolution scanning microscopy techniques in the study of membrane-related processes.
Introduction There is increasing evidence that the lipid bilayer comprising the plasma membrane plays an active role in cell signaling and regulation.1 Often, as in the case of lipid rafts, this involvement is governed by submicrometer heterogeneities in the chemical and physical nature of the bilayer.2 Advanced characterization techniques such as atomic force microscopy (AFM)3,4 and near-field optical microscopy5,6 have been adapted to study these structures, but these methods offer the highest resolution only when using surface-supported samples.7,8 Because of this requirement, many studies have been performed on artificial bilayers assembled on inorganic solid supports. These simple bilayer models are typically formed with three or fewer lipid components, offering some biological functionality within a less-complex system than a natural membrane.9 Although this methodology removes some of the complicating influences of an intact cell surface, it is not clear whether the behavior observed is biologically relevant, especially for complex phase-separating mixtures of several lipid species.10 Bridging the gap between convenient sample geometry and biological relevance, model bilayers can be formed using natural lipid extracts rather than artificial lipids.11,12 Typically, the lipids for these samples are obtained via solvent or detergent extraction and purification. Once concentrated, these lipids can be reconstituted into bilayer form on the sample surface, most commonly *Corresponding author. E-mail:
[email protected]. (1) Rohrbough, J.; Broadie, K. Nat. Rev. Neurosci. 2005, 6, 139–150. (2) Hering, H.; Lin, C. C.; Sheng, M. J. Neurosci. 2003, 23, 3262–3271. (3) Engel, A.; Muller, D. J. Nat. Struct. Biol. 2000, 7, 715–718. (4) Dufrene, Y. F.; Lee, G. U. Biochim. Biophys. Acta 2000, 1509, 14–41. (5) Burgos, P.; Yuan, C.; Viriot, M. L.; Johnson, L. J. Langmuir 2003, 19, 8002– 8009. (6) Hwang, J.; Gheber, L. A.; Margolis, L.; Edidin, M. Biophys. J. 1998, 74, 2184–2190. (7) Hollars, C. W.; Dunn, R. C. Biophys. J. 1998, 75, 342–353. (8) Tokumasu, F.; Jin, A. J.; Feigenson, G. W.; Dvorak, J. A. Biophys. J. 2003, 84, 2609–2618. (9) Brian, A. A.; McConnell, H. M. Proc. Natl. Acad. Sci. U.S.A. 1984, 81, 6159– 6163. (10) Munro, S. Cell 2003, 115, 377–388. (11) Koynova, R.; MacDonald, R. C. Biochim. Biophys. Acta 2007, 1768, 2373– 2382. (12) Seto-Young, D.; Chen, C. C.; Wilson, T. H. J. Membr. Biol. 1984, 84, 259– 267.
Langmuir 2010, 26(7), 4635–4638
with Langmuir-Blodgett (LB)13 or vesicle rupture (VR)14,15 deposition. Given the multistep nature of the extractionconcentration-reconstitution protocol, these processes are timeconsuming and require substantial starting material. These limitations preclude the use of a subconfluent monolayer of cultured cells as the lipid source. Furthermore, there have been concerns raised about whether detergent-based techniques alter the lipid phase behavior through contamination or incomplete extraction.16 In the present work, we demonstrate a simple alternative method for transferring membrane components from the cell surface to a solid support. In contrast to LB or VR deposition, this method incorporates lipid extraction and redeposition into a single process. Lipids from cell surfaces are transferred using a modification of the bubble collapse deposition (BCD) technique, whereby a lipid monolayer is “inked” onto the surface of an air bubble and then redeposited as a bilayer when the bubble is shrunk.17 In previous work, we have shown that artificial bilayers can be used as the inking source. In this case, the source for inking is the intact cell membrane rather than an artificial bilayer. We demonstrate that brief physical contact is sufficient to extract the membrane components onto a clean bubble surface and that these lipids can be redeposited as a fluid bilayer onto a clean silica surface for further study. Erythrocytes, also known as red blood cells (RBCs), were chosen as a test case because they lack organelles and it is therefore simple to track the source of the lipid material. However, because this technique does not rely on any chemically specific interactions we expect that it would be generally applicable to a broad range of cell types. We also characterize which membrane components are transferred, showing that lipids are extracted from both membrane leaflets but that cytoskeletal elements are not transferred to the resulting supported bilayer. The trans-membrane protein band 3 also is not transferred with this process. Through the use of (13) Osborn, T. D.; Yager, P. Biophys. J. 1995, 68, 1364–1373. (14) Keller, C. A.; Glasmastar, K.; Zhdanov, V. P.; Kasemo, B. Phys. Rev. Lett. 2000, 84, 5443–5446. (15) Schonherr, H.; Johnson, J. M.; Lenz, P.; Frank, C. W.; Boxer, S. G. Langmuir 2004, 20, 11600–11606. (16) Shogomori, H.; Brown, D. A. Biol. Chem. 2003, 384, 1259–1263. (17) Mager, M. D.; Melosh, N. A. Langmuir 2007, 23, 9369–9377.
Published on Web 03/05/2010
DOI: 10.1021/la100583f
4635
Letter
Mager and Melosh
Figure 1. Schematic representation of the inking and deposition process. (A) A clean air bubble is blown at the end of a submerged needle. (B) This bubble is brought into contact with a layer of surface-supported cells. The high-energy air-water interface draws lipids off of the cell surface and into a monolayer on the bubble. (C) The monolayer-coated bubble is moved from the substrate covered with cells to another clean substrate. (D) The bubble touches the silica surface of this new substrate. (E) Air is slowly withdrawn from the bubble, shrinking it and eventually causing the monolayer to buckle and fold back on itself as a bilayer. (F) Once the completed bilayer patch is formed, the bubble snaps free of the surface and the needle can be removed from the solution.
photobleaching measurements, we show that these cell-derived bilayers are in continuous fluid contact over several tens of micrometers, an area much larger than the surface of a single cell. This result indicates that the deposited patch is composed of lipid extracted from several cells. These large, surface-supported bilayer patches will allow future high-resolution studies of cellderived lipid behavior.
Experimental Section Materials. Erythrocytes were collected from a healthy adult
donor, isolated from other blood components, and frozen at -80 °C in a mixture of 40% glycerol until use. Freezing, thawing, and removing the glycerol were performed according to standard protocols.18 Labeling of the cell membrane was performed by adding 20 μL of Texas Red DHPE (Molecular Probes) at 1 mg/mL in ethanol to a 1 mL sample of erythrocytes in 0.9% NaCl. Cells were then centrifuged and rinsed three times to remove excess dye. For control experiments, pure POPC was used as obtained from Avanti Polar Lipids. Phalloidin-conjugated Alexa Fluor 488, Annexin V-conjugated Alexa Fluor 488, FITC-anti CD59, and eosin isothiocyanate were all purchased from Molecular Probes. Silicon wafers with 100 nm of thermally grown oxide were from University Wafer. Sample Preparation. RBCs were deposited by sedimentation onto ∼1 cm2 silicon chips. Prior to this deposition, the chips were cleaned with acetone, methanol, isopropanol, and a 10 min exposure to UV/ozone. A 50 μL droplet of labeled RBC solution was placed on the chip surface and allowed to sit for 10 min. The surface was then gently rinsed with phosphate-buffered saline (PBS) (pH 7.4), leaving behind a monolayer of RBCs adsorbed to the chip surface. These remaining cells were intact and unlysed. In contrast, when deposition was performed on a surface coated with polylysine the cells adhered strongly and ruptured, consistent with the literature.19 Samples used for deposition were cleaned as above but with the addition of 1 min of oxygen plasma exposure. Deposition was performed with previously published protocols for BCD.17 Briefly, a syringe pump was used to blow a 1 μL air bubble at the end of a submerged needle. This bubble was brought into contact with the layer of adsorbed RBCs and carefully dragged a (18) Harmening, D., Pittiglio, D. H., Baldwin, A. J., Sohmer, P. R., Eds. Modern Blood Banking and Transfusion Practices, 5th ed.; F. A. Davis: Philadelphia, 1983. (19) Hategan, A.; Law, R.; Kahn, S.; Disher, D. E. Biophys. J. 2003, 85, 2746– 2759.
4636 DOI: 10.1021/la100583f
few millimeters across the surface to achieve complete inking of the bubble surface. This monolayer-coated bubble was then touched to the deposition substrate, and air was withdrawn at a rate of 0.2 μL/min until the bubble snapped free of the surface, leaving a supported bilayer behind. A schematic illustration of this process is given in Figure 1. Unless otherwise noted, all experiments were performed in PBS at 25 °C. Labeling. To label the cytoskeleton, samples were incubated with 100 diluted stock solution of Phalloidin-Alexa Fluor 488 at 25 °C for 20 min and then rinsed extensively with PBS. Phosphatidylserine labeling was performed by applying a 15 diluted stock solution of Annexin V-Alexa Fluor 488 for 20 min in a labeling solution consisting of 10 mM HEPES, 150 mM NaCl, 2 mM CaCl2, 5 mM KCl, and 1 mM MgCl2 adjusted to pH 7.4, followed by rinsing with PBS. Band 3 labeling was performed by adding 0.05 mg/mL eosin isothiocyanate and incubating for 60 min in PBS. Prior to CD59 labeling, the sample was incubated with 1 mg/mL BSA in PBS to decrease nonspecific interactions. The sample was then exposed to 10 diluted anti-CD59-FITC for 45 min, followed by extensive rinsing in PBS. Microscopy. All fluorescence microscopy was performed on an upright fluorescence microscope (Zeiss Axioskop) with a mercury lamp using a 63 water immersion lens with a numerical aperture of 1.0. All microscopy measurements were performed at 25 °C. For staining experiments, appropriate dichroic filter sets (Zeiss no. 9 and no. 14) were used for the respective dye. Images were acquired with a 1388 pixel1040 pixel CCD (Zeiss Axiocam) and were processed using ImageJ. Fluorescence recovery after photobleaching (FRAP) measurements were initiated with a 20 s bleach pulse from an unattenuated 100 W mercury lamp. To minimize unintended photobleaching during the recovery period, a mechanical shutter was used to acquire images of 50 ms integration time once every 5 s. Data from monitoring the recovery were fit to the derivation of Soumpasis20 with a modification for the nonzero bleach time as detailed previously.21 This fitting was performed with a least-squares algorithm using Igor Pro software.
Results and Discussion We evaluated several aspects of the lipid bilayer produced from erythrocyte lipids by fluorescence microscopy. The first step was (20) Soumpasis, D. M. Biophys. J. 1983, 41, 95–97. (21) Mager, M. D.; Almquist, B.; Melosh, N. A. Langmuir 2008, 24, 12734– 12737.
Langmuir 2010, 26(7), 4635–4638
Mager and Melosh
to confirm that the deposited membrane was a single lipid bilayer and not a monolayer or multilayer. Given the same dye species and concentration, the fluorescence intensity of the erythrocytederived patches was the same as that of patches produced with BCD from pure POPC, which have been shown previously to be single bilayers. Control experiments in which the same procedure was performed without preloading the bubble surface with lipids resulted in no deposition of any fluorescent structures. We further confirmed this identity by performing a cobalt quenching assay. The addition of 50 mM CoCl2 decreased the fluorescence intensity in the deposited patch by approximately 50%, and because cobalt quenching relies on direct physical collision, this drop indicates that half of the dye molecules were accessible to the surrounding solution, as is the case for supported bilayers.22 The lipid deposition was optimized by varying the bilayer deposition rate, the substrate cleaning procedure, and the solution conditions. We found a bubble shrinkage rate of 0.2 μL/min to be optimal, much slower than the 1-5 μL/min optimum for a pure POPC bilayer. Using this shrinkage rate resulted in the formation of a supported lipid bilayer from cell membrane components in a total time of about 3 min, including inking, transfer, and redeposition. Typically, successful deposition resulted in a bilayer patch that was several tens of micrometers on a side. The large patch area is significant because it demonstrates that the material for deposition is derived from more than one RBC. The membrane of a single cell would not provide sufficient lipids to cover an area of this size, implying that during the inking process the lipids from a number of cells are mixed into a large fused monolayer on the bubble surface. Altering the solution pH over the range of 6-9 did not significantly alter the success rate or the size of the deposited patch nor did changing the salinity over the range of 0-500 mM or adding up to 5 mM CaCl2. Although UV/ozone cleaning was sufficient for the adhesion of intact RBCs on the donor chip, we found that using oxygen plasma cleaning on the deposition chip, a more aggressive procedure, gave a higher success rate and larger continuous deposition areas. We performed FRAP measurements to characterize the diffusion coefficient of the bilayer. Lateral fluidity is one of the most important characteristics of a supported lipid bilayer because a low diffusion coefficient can indicate defect sites and a completely immobile bilayer will not exhibit biologically relevant behavior.23 The results of this experiment are shown in Figure 2. With a 30-μm-diameter bleach spot, the recovery occurred on a timescale of around 5 min. When analyzed quantitatively,20 this recovery curve gives a diffusion coefficient of 1.3 ( 0.1 μm2/s, consistent with high-quality fluid-phase supported bilayers. Thus, it is clear that one of the most important biological properties of the cell membrane, its lateral mobility, was preserved when the lipids were removed from the cell surface and reconstructed into a bilayer on the chip surface. To determine whether lipids are transferred from both leaflets of the membrane bilayer or just from the outer leaflet, we labeled both the intact cells and the deposited bilayer with Annexin V, which specifically binds to phosphatidylserine (PS). PS is localized to the cytosolic leaflet of erythrocytes (except during apoptosis) and is thus a simple marker of asymmetry.24 Annexin staining on the source erythrocytes was negative, confirming that the cells were intact and nonapoptotic. The deposited bilayer stained strongly for phosphatidylserine as shown in Figure 3, indicating (22) Morris, S. J.; Bradley, D.; Blumenthal, R. Biochim. Biophys. Acta 1985, 10, 365–372. (23) Saxton, M. J. Biophys. J. 1982, 39, 165–173. (24) Koopman, G.; Reutelingsperger, C. P. M.; Kuijten, G. A. M.; Kehnen, R. M. J.; Pals, S. T.; Oers, H. H. J. V. Blood 1994, 85, 1415–1420.
Langmuir 2010, 26(7), 4635–4638
Letter
Figure 2. FRAP measurements of an erythrocyte-derived bilayer patch. (A) A 20 s bleach pulse from the unattenuated mercury lamp bleaches an area that is approximately 30 μm in diameter. (B) After 4 min, the intensity within this area has recovered, indicating lateral fluidity. (C) Plotting the recovery data (gray squares) and applying a theoretical fit (solid line) yields a calculated diffusion coefficient of 1.3 ( 0.1 μm2/s. This value is consistent with other measurements in the literature of supported bilayers consisting of partially saturated lipids.
Figure 3. Determination of phosphatidylserine (PS) content. BCD was used to produce two samples, one inked from erythrocyte membranes and one inked from an artificial POPC bilayer. Both samples were double labeled with a generic membrane probe (octadecyl rhodamine, red) and a PS-specific probe (AlexaFluor488 Annexin V, green). Only the cell-derived patch showed PS labeling. Previous control experiments demonstrated that the source erythrocytes were not apoptotic (Supporting Information) and maintained their cytosolic PS localization. Together, these results indicate that the inking process draws lipids from both leaflets of the plasma membrane.
that during the inking process both leaflets of the cell membrane are incorporated into the bubble-supported monolayer. As mentioned previously, lipids from many individual cells are fused and mixed on the surface of the bubble to form a continuous monolayer. This mixing is likely the reason that bilayer asymmetry was not preserved in the redeposited bilayer. As a negative control, the Annexin tag was also applied to a BCD-produced bilayer of pure POPC, which has only phosphatidylcholine headgroups and thus was not labeled. We next performed a series of experiments to determine if additional nonlipid membrane components are transferred in this process, beginning with the cytoskeleton. Erythrocytes are unique in that, rather than possessing a 3D actin cytoskeleton, they have a 2D spectrin cytoskeleton that exists as a sheet that is densely anchored to the plasma membrane.25 The deposited bilayers were (25) Alberts, B.; Bray, D.; Lewis, J.; Raff, M.; Roberts, K.; Watson, J. D. Molecular Biology of the Cell; Garland Publishing: New York, 1994.
DOI: 10.1021/la100583f
4637
Letter
Figure 4. Characterization of lipid and cytoskeleton components. (A) Fluorescence micrograph of the edge of the deposited bilayer patch (right) and the remains of two erythrocytes that were deposited beyond this edge. The sample was double labeled for lipid (red) and actin (green). (B) Line trace taken from the dashed line in A showing that actin is present in significant concentrations only in the intact cells. The cytoskeleton is not transferred into the deposited bilayer patch.
incubated with fluorescently tagged phalloidin, which binds to actin filaments. The results of this experiment are shown in Figure 4, including a patch of intact erythrocyte as a positive control. The BCD bilayer patch showed near-background levels of phalloidin labeling, indicating that the lipid components of the membrane were transferred without measurable cytoskeletal components. In comparison, the intact cell showed high levels of phalloidin labeling. There are a few likely explanations for why the spectrin network may not have been transferred. First, the cytoskeleton is not as amphiphilic as lipid molecules are. Therefore, the larger decrease in surface energy may have driven the bubble surface to become saturated with lipids to the exclusion of spectrin. Alternatively, because the cytoskeleton is extensively cross-linked, it may not have been able to undergo sufficient molecular rearrangement to adsorb to the bubble and subsequently redeposit on the sample surface. These hypotheses are supported by the fact that we would often observe relatively intact cell remains deposited beyond the periphery of the bilayer patch. These remains stained strongly with phalloidin, indicating that they still contained a high concentration of cytoskeletal components. Thus, we can conclude that although some small fraction of the spectrin network may have become incorporated into the BCD bilayer patch, the cytoskeleton as such was not transferred. Further labeling experiments were performed to determine whether noncytoskeletal membrane proteins are transferred along with the lipids. One common erythrocyte membrane protein is band 3, a transmembrane antiporter. Labeling with eosin
4638 DOI: 10.1021/la100583f
Mager and Melosh
isothiocyanate showed that, within the current detection limits, band 3 was not transferred along with the membrane lipids (Supporting Information). This result can be explained by considering the mechanism of this process, which requires a splitting of the two leaflets of the bilayer into an air-supported monolayer. Band 3, which crosses the bilayer completely, would likely not be incorporated properly following this splitting. We also stained for CD59, a GPI-linked protein on the erythrocyte surface (Supporting Information). Although CD59 is associated with only a single lipid molecule, these experiments similarly did not detect any protein in the supported bilayer. One possibility is that the protein was transferred, but was denatured in the process by mechanical contact with the chip surface. Because the fluorescent probe was based on an antibody, it might not bind to a denatured version of its target. We are continuing to examine whether modifications to this procedure could result in the transfer of intact protein or whether there are other proteins that might survive the process. Future experiments in which we directly examine the surface of the inked bubble will also help to establish whether the uptake or deposition is the limiting step in the transfer of membrane proteins. In conclusion, we have demonstrated a simple, rapid system for removing membrane components from intact cell surfaces and reconstituting them into a supported lipid bilayer. This source material was obtained from a single layer of cells, demonstrating that the technique would be appropriate to use with common cell culture conditions. The resulting bilayer patches are much larger than a single cell and thus represent the fused components of several cells. The bilayer is laterally mobile over these large areas, with a diffusion coefficient typical of supported fluid bilayers. Compositional analysis reveals that this supported bilayer is formed from a mixture of both the cytosolic and extracellular leaflets of the cell membrane. Neither the cytoskeletal spectrin network nor membrane proteins are transferred along with the lipid components. Results thus far indicate that membrane proteins are not transferred intact. Further work with highsensitivity microscopy tools such as total internal reflection microscopy will help to characterize more fully the membrane components that may be transferred in trace quantities. In the future, combining this procedure with advanced probe techniques such as AFM and scanning optical methods will allow the study of nanoscale chemical and physical structures in cell-derived lipid bilayers. Acknowledgment. We thank B. Almquist for assistance with erythrocyte handling and preparation. Supporting Information Available: Fluorescence micrographs showing the staining of cell-derived bilayers for CD59 and band 3. This material is available free of charge via the Internet at http://pubs.acs.org.
Langmuir 2010, 26(7), 4635–4638
