Formation and Dynamics of Supported Phospholipid Membranes on a

Jan 26, 2009 - We have studied and modeled the morphology and dynamics of fluid planar lipid bilayer membranes supported on a textured silicon substra...
0 downloads 0 Views 2MB Size
2986

Langmuir 2009, 25, 2986-2993

Formation and Dynamics of Supported Phospholipid Membranes on a Periodic Nanotextured Substrate James H. Werner,*,† Gabriel A. Montan˜o,† Anthony L. Garcia,‡ Nesia A. Zurek,†,§ Elshan A. Akhadov,†,| Gabriel P. Lopez,*,‡ and Andrew P. Shreve*,† Center for Integrated Nanotechnologies, Materials Physics and Application DiVision, Los Alamos National Laboratory, Los Alamos, New Mexico 87545, and Department of Chemical and Nuclear Engineering, Center for Biomedical Engineering, 209 Farris Engineering Center, UniVersity of New Mexico, Albuquerque, New Mexico 87131 ReceiVed July 14, 2008. ReVised Manuscript ReceiVed December 20, 2008 We have studied and modeled the morphology and dynamics of fluid planar lipid bilayer membranes supported on a textured silicon substrate. The substrate is fabricated to have channels on its surface that are a few hundred nanometers across, with a channel depth of a few hundred nanometers perpendicular to the plane of observation. Using atomic force microscopy and quantitative fluorescence microscopy, we have shown that the bilayer assemblies conform to the underlying nanostructured substrate. As far as dynamics is concerned, when observed over length scales exceeding the dimensions of the nanostructured features, the macroscopic diffusion is anisotropic. However, the macroscopic anisotropy is well simulated using models of diffusion on the nanostructured surface that consider the lipids to diffuse homogeneously and isotropically on the supporting substrate. Consistent with previous observations on less well characterized or less periodic nanostructures, we find that the nanostructured substrate produces an effective anisotropy in macroscopic diffusion of the conformal membrane. More importantly, we demonstrate how quantitative analysis of dynamics probed by larger-scale fluorescence imaging can yield information on nanoscale thin-film morphology.

Introduction Supported planar lipid bilayers are important model systems enabling the control, organization, and study of membranes and membrane-associated proteins.1,2 In addition to fundamental biophysical investigations, supported bilayers are useful scaffolds for many practical applications, such as electrophoretic separations3-6 or protein recognition arrays.7,8 For many of these applications, there is often a need to physically segregate and isolate sections of planar lipid bilayers on nanometer to micrometer length scales. To this end, patterning methods directly taken from microelectronics, such as microcontact printing and stamping techniques9-13 or photolithographic methods,14-16 have * Corresponding authors. E-mail: [email protected]. E-mail: gplopez@ unm.edu. E-mail: [email protected]. † Los Alamos National Laboratory. ‡ University of New Mexico. § Current address: Department of Molecular, Cellular and Developmental Biology, University of Colorado, Boulder, Colorado 80309. | Current address: Nanomaterials Sciences Department, Sandia National Laboratories, Albuquerque, New Mexico 87185.

(1) Tanaka, M.; Sackmann, E. Nature 2005, 437, 656–663. (2) Chan, Y.-H. M.; Boxer, S. G. Current Opin. Chem. Biol. 2007, 11, 581– 587. (3) van Oudenaarden, A.; Boxer, S. G. Science 1999, 285, 1046–1048. (4) Olson, D. J.; Johnson, J. M.; Patel, P. D.; Shaqfeh, E. S. G.; Boxer, S. G.; Fuller, G. G. Langmuir 2001, 17, 7396–7401. (5) Daniel, S.; Diaz, A. J.; Martinez, K. M.; Bench, B. J.; Albertorio, F.; Cremer, P. S. J. Am. Chem. Soc. 2007, 129, 8072–8073. (6) Nabika, H.; Takimoto, B.; Murakoshi, K. Anal. Bioanal. Chem. 2008, 391, 2497–2506. (7) Groves, J. T. Curr. Opin. Drug DiscoVery DeV. 2002, 5, 606–612. (8) Kung, L. A.; Kam, L.; Hovis, J. S.; Boxer, S. G. Langmuir 2000, 16, 6773–6776. (9) Groves, J. T.; Boxer, S. G. Acc. Chem. Res. 2002, 35, 149–157. (10) Hovis, J. S.; Boxer, S. G. Langmuir 2000, 16, 894–897. (11) Hovis, J. S.; Boxer, S. G. Langmuir 2001, 17, 3400–3405. (12) Lenz, P.; Ajo-Franklin, C. M.; Boxer, S. G. Langmuir 2004, 20, 11092– 11099. (13) Sapuri-Butti, A. R.; Butti, R. C.; Parikh, A. N. Langmuir 2007, 23, 12645– 12654. (14) Yee, C. K.; Amweg, M. L.; Parikh, A. N. AdV. Mater. 2004, 16, 1184– 1189.

seen widespread use for containing and controlling lipid domains in two spatial dimensions. Whereas methods to control and characterize the interaction of planar lipid bilayers with nanostructured and microstructured surfaces in two dimensions are rather well established, a handful of recent experimental investigations have also begun to address the interaction of planar lipid bilayers with substrates that demonstrate texture or patterning in three spatial dimensions on nanometer to micrometer length scales.17-23 For example, Parthasarathy, Yu, and Groves17 studied the phase segregation of lipid/cholesterol mixtures in double bilayer assemblies formed on a solid support that was spatially patterned with micrometerscale rounded stripes that protruded several hundred nanometers from the surface. This study discovered that phase segregation was critically dependent upon the local curvature. In another recent study that explored the interaction of planar lipid bilayers with nanostructured 3D surfaces, Lopez and co-workers18 found that mixtures of long- and short-chain lipids in an appropriate concentration ratio formed micrometer-scale bilayer assemblies (bicelles) capable of spanning hundred-nanometer-sized troughs on nanostructured silicon. (15) Yee, C. K.; Amweg, M. L.; Parikh, A. N. J. Am. Chem. Soc. 2004, 126, 13962–13972. (16) Howland, M. C.; Sapuri-Butti, A. R.; Dixit, S. S.; Dattelbaum, A. M.; Shreve, A. P.; Parikh, A. N. J. Am. Chem. Soc. 2005, 127, 6752–6765. (17) Parthasarathy, R.; Yu, C. H.; Groves, J. T. Langmuir 2006, 22, 5095– 5099. (18) Zeineldin, R.; Last, J. A.; Slade, A. L.; Ista, L. K.; Bisong, P.; O‘Brien, M. J.; Brueck, S. R. J.; Sasaki, D. Y.; Lopez, G. P. Langmuir 2006, 22, 8163– 8168. (19) Davis, R. W.; Flores, A.; Barrick, T. A.; Cox, J. M.; Brozik, S. M.; Lopez, G. P.; Brozik, J. A. Langmuir 2007, 23, 3864–3872. (20) Ro¨mer, W.; Steinem, C. Biophys. J. 2004, 86, 955–965. (21) Schmitt, E. K.; Nurnabi, M.; Bushby, R. J.; Steinem, C. Soft Matter 2008, 4, 250–253. (22) Weiskopf, D.; Schmitt, E. K.; Klu¨hr, M. H.; Dertinger, S. K.; Steinem, C. Langmuir 2007, 23, 9134–9319. (23) Worsfold, O.; Voelcker, N. H.; Nishiya, T. Langmuir 2006, 22, 7078– 7083.

10.1021/la802249f CCC: $40.75  2009 American Chemical Society Published on Web 01/26/2009

Phospholipid Membranes on a Substrate

For both fundamental research concerning lipid organization and for applications such as trans-membrane protein separation technologies,5 it is important to understand the interactions between textured substrates and the adhered lipid assemblies. Although lipid membranes often conform to underlying substrate features,17,24-27 several recent reports have shown that for certain lipid compositions or with specific morphologies or functionalization of textured substrates, lipid membranes can be formed that are suspended across surface features.18,20-23 To distinguish such cases, one desires characterization methods that pertain to lipid morphology on multiple length scales both in and out of the plane of observation. Here, we demonstrate various methods of characterizing lipid bilayers and their adherence to simple 3D-nanostructured substrates. One of the most important findings of this work is the demonstration that quantitative fluorescence recovery after photobleaching28 (FRAP) measurements can be used to gain information about the underlying surface nanomorphology. Here, and in a small number of previous reports,29-32 the recovery of fluorescence following a radially symmetric bleach is anisotropic in nature, with one axis recovering faster than another. Thus far, the anisotropy in recovery has been attributed to an underlying nanostructure in a primarily qualitative fashion. The measurements reported here are much more quantitative, providing a direct link between measured diffusional anisotropy and the underlying substrate morphology. In particular, a simple geometric relation between the observed anisotropy and the underlying nanostructure is presented. Moreover, this work uses detailed Monte Carlo simulations to model the recovery. These simulations take into account both how the underlying 3D nanostructure perturbs the initial bleached population from a 2D Gaussian and how these nanometer scale features appear when imaged at visible wavelengths, physical effects neglected in previous studies of anisotropic diffusion on nanostructured surfaces29,30 or particular cell lines.31,32 In addition to FRAP measurements of the bilayer conformity, atomic force microscopy (AFM) confirmed the invagination of the planar lipid bilayers into the nanostructured surface. Both scanning electron microscopy and light diffraction were used to characterize and directly measure the underlying nanostructured silicon substrate. For the membrane compositions, surfaces, and environmental conditions used here, all measurement methods demonstrate that the lipid bilayer conforms to the underlying nanostructured surface. Even more importantly, the experimental and modeling techniques presented here should be generally useful in the analysis of the interaction of phospholipid assemblies with other nanotextured substrates on length scales spanning micrometers down to molecular scales ( 0) or flat (b ) 0) unit cell and scales as the square of the distance, as appropriate for diffusional behavior. Using the dimensions of the surface determined by SEM (a ≈ 360 nm, b ≈ 320 nm, c ≈ 160 nm), eq 5 yields Dapp equal to ∼0.2*D. Thus, the ratio obtained from this simple model, (D/ Dapp), is ∼5, in reasonable agreement with the measured diffusion ratio, 4.3 ( 0.2. Nevertheless, there are several approximations used in this simple analysis of the FRAP data that could influence the validity of this comparison, so a more detailed analysis is considered next. One may question whether eqs 1-4 adequately model the diffusion on these surfaces. For example, can the concentration profile on these surfaces following a bleach be approximated as a Gaussian? Because molecules on the vertical portions of the surface all have approximately the same probability of being bleached, the underlying concentration of fluorescent molecules is inherently “banded” at the beginning of the recovery. Moreover, the number of fluorescent lipids observed in the x-y plane is a periodic function along the direction perpendicular to the grooved surface (it is larger at every wall of the grooved surface) and is not clamped at a fixed value c∞. However, because the surface is imaged with visible wavelengths (excitation light of ∼600 nm) through a 0.45 numerical aperture (NA) objective, spikes in fluorescent lipid concentration are effectively smoothed out by diffraction. Furthermore, the finite size of the CCD pixels (each pixel in the FRAP image corresponds to a box of ∼0.6 µm sides) also blurs the underlying spatial periodicity. Equations 1-4 take into account neither the non-Gaussian nature of the initial bleach nor the fact the underlying concentration of fluorescent lipids is nonuniform in one spatial dimension. We note that, in principle, much higher resolution optical images of the samples studied here could be obtained using high numerical aperture objectives and such images might provide more information on the detailed structure. However, in any case, full resolution of the features on this surface would be challenging. Furthermore, the methods presented here are also applicable to materials with even smaller features, whereas direct optical imaging methods would reach their resolution limit at approximately the feature sizes studied here. To assess the impact of the above issues on the measured data, we constructed a simulation to model the macroscopically observed response expected from a bleach occurring on the nanogrooved surface. Figure 5a shows what a fluorescent image of the bilayer surface would look like at infinite spatial resolution following a bleach from a Gaussian excitation source. The vertical white stripes in this image are the walls of the grooved surface and reflect the fact that these regions have a large number of fluorescent lipids per unit distance in the lateral direction. Figure 5b shows what this surface would look like when viewed through a 0.45 NA objective with the CCD camera. This image is the convolution

2992 Langmuir, Vol. 25, No. 5, 2009

Werner et al.

Figure 5. (a) Simulation of the fluorescence image from a grooved surface observed with infinite spatial resolution. (b) Surface convolved with a Gaussian point spread function to account for diffusion and pixilation. (c) Simulated bleached image on an “unfolded” surface (Figure 1, bottom) at infinite spatial resolution.

of Figure 5a with a Gaussian point spread function (PSF) that has a standard deviation of 0.5 µm to account for diffraction and CCD pixilation. Lastly, Figure 5c shows what the underlying concentration profile looks like on an unfolded surface following the Gaussian bleach. The unfolded surface has a uniform lipid concentration away from the bleach center, but the bleach itself has a banded structure reflecting the presence of the vertical walls of the grooves. (Note that this latter image is at infinite spatial resolution and has not been convolved with a PSF.) Our simulations start from the image shown in Figure 5c and use finite time steps and a finite spatial mesh to evolve the concentration profile away from the t ) 0 value. In our simulation, each spatial grid point (mesh size) is 20 nm × 20 nm. This mesh size is smaller than the features of the grooved surface and, for convenience, is also chosen as a common divisor of all dimensions (a ) 360 nm, b ) 320 nm, c ) 160 nm). The number of fluorescent lipids in a given mesh point (i, j) is denoted by Nij, with the fluorescence intensity obtained from this point being directly proportional to the number of fluorescent lipids. The number of fluorescent lipids at mesh point i, j following a time step ∆t is given by

Nij(t + ∆t) ) Nij(t) + p[Ni+1j(t) + Ni - 1j(t) + Nij+1(t) + Nij - 1(t)] - 4pNij(t) (6) In eq 6, p is the probability of diffusion of a lipid from one mesh to its nearest neighbor in a time step of ∆t. The number of

fluorescent lipids at the i, j mesh point at t + ∆t, Nij(t + ∆t), is the number there at time t plus the influx of molecules from this mesh point’s nearest neighbors minus the efflux of molecules leaving the mesh point as a result of diffusion. The probability of diffusion out of the grid point in time ∆t, p, is directly related to the diffusion coefficient of the lipids as p ) D*∆t/(∆x)2. These simulations take a finite time step, ∆t, of 80 µs, assume a diffusion coefficient of 1 µm2/s, and, as mentioned above, have a mesh size of ∆x ) 20 nm. This places the probability of diffusion across a given mesh boundary, p, at 0.2 per 80 µs time step. Our simulations diffuse the lipids using the above master equation on the unfolded surface (e.g., Figure 5c). The surface is then refolded and blurred by a Gaussian PSF to arrive at an image such as that shown in Figure 5b. This procedure creates a series of FRAP images spaced apart in time by 80 µs. We analyze the simulated image every 1000th frame (i.e., every 0.08 s) using the models described in eqs 1-4. The top panel of Figure 6 shows the simulated FRAP image obtained at t ) 1.04 s after the bleach (frame 13 000). Below the image are plots of the standard deviations obtained from elliptical Gaussian fits (eq 3) to the images as a function of time. The slopes of these lines yield two times the diffusion coefficient for that axis (eq 4). The linear fit to the fast axis of the Gaussian yields 0.9803 ( 0.0005 µm2/s as the measured diffusion coefficient whereas the slow diffusion axis has an apparent diffusion coefficient of

Phospholipid Membranes on a Substrate

Langmuir, Vol. 25, No. 5, 2009 2993

these assumptions is exactly true. If we assume that the substrate can be approximated by a square well but allow the bilayer to be nonconformal, then the measured diffusion ratio (4.3) puts a limit on the penetration depth of bilayer invagination into the grooves. This analysis yields a penetration depth of ∼280 nm, indicating that the bilayer drapes down substantially but does not quite reach the bottom of the invagination, leaving an ∼40 nm gap at the bottom (assuming that this interpretation is correct). Alternatively, if we consider the bilayer to be conformal but relax the perfectly rectangular structure in the analysis, then the discrepancy between the measured anisotropy ratio and that predicted by the simple geometric model (eq 4) might be explained by simply rounding off the corners of the square well. If we replace the corner of all of the features (mesas and valleys) by quarter-circle arcs with one common radius of curvature, R, then the unfolded distance of the unit cell (Figure 1, bottom) becomes a + 2b + c + (2π - 8)R. For our measured anisotropic diffusion ratio (D/Dapp) of 4.3, an analysis similar to that of eq 5 yields a value of R ) 47.6 nm. This value for the radius of curvature is in reasonable agreement with the average sharpness of the features as estimated from an SEM image (Figure 1) and is also of the same order of magnitude that the minimum radius of curvature a POPC bilayer (∼15 nm)36 can obtain.

Conclusions

Figure 6. (Top) Simulated FRAP image 1.04 s after a bleach. (Bottom) Fits of the simulated data using eqs 3 and 4. The fast axis of the Gaussian yields an apparent diffusion coefficient of 0.98 µm2/s whereas the slow diffusion axis has an apparent diffusion coefficient of 0.19 µm2/s.

0.193 ( 0.001 µm2/s. The simulated data (Figure 6) is in good agreement with the measured anisotropic FRAP data (Figure 4). Moreover, the ratio of the fast to the slow diffusion coefficient on the synthetic data is ∼5.07, in very good agreement with the ratio of ∼4.98 obtained using the simple approximation set forth in eq 5. We also note that the simulation methods presented here are useful for the analysis of FRAP results for more complicated textured surfaces where simple analytical expressions may not be readily available. The agreement between the ratio for the diffusion coefficients obtained from the synthetic data (5.07) is closer to the value expected from the simple predictive model (∼4.98) than either is to the measured data (∼4.3 ( 0.2). Both the simple predictive model and the simulated data assume that the substrate is a perfect rectangular structure and that the lipid bilayer adheres and conforms perfectly to the surface when, realistically, neither of

We have used a combination of AFM and quantitative fluorescence microscopy to demonstrate the formation of POPC bilayers conformed to a nanotextured surface. One of the more important findings of this work is a demonstration of how quantitative analysis of anisotropic diffusion on the micrometer scale can yield information concerning the underlying nanoscale morphology. In addition, this work lays the foundation for future studies directed at how nanotextured surfaces with tunable physical or chemical properties might be used to influence membrane structure, dynamics, and functions. We are planning future studies to address the role of substrate dimensions, the influence of lipid composition, the possible generation of fluid floating layers (in contrast to previously demonstrated bicellederived suspended membranes,18 which are essentially laterally immobile), and the use of constructs for the study of biological processes such as cell-surface interactions. Acknowledgment. This work was performed, in part, at the Center for Integrated Nanotechnologies, a U.S. Department of Energy, Office of Basic Energy Sciences user facility at Los Alamos National Laboratory (contract DE-AC52-06NA25396) and Sandia National Laboratories (contract DE-AC0494AL85000). Funding for this work was provided by the National Science Foundation’s PREM (DMR-0611616) and NIRT (CTS0404124) programs. The facilities of the NSF-sponsored National Nanotechnology Infrastructure Network node at the University of New Mexico were used for the nanostructure fabrication. LA802249F (36) Cremer, P. S.; Boxer, S. G. J. Phys. Chem. B 1999, 103, 2554–2559.