Hyungsoon
Im‡
,
Nathan J.
Wittenberg‡
,
Antoine
Lesuffleur‡
,
Nathan C.
Lindquist
and
Sang-Hyun
Oh
*
Department of Electrical and Computer Engineering, University of Minnesota, Minneapolis, Minnesota 55455, USA. E-mail: sang@umn.edu; Fax: +1 612 625 4583; Tel: +1 612 625 0125
First published on 13th October 2010
Integration of solid-state biosensors and lipid bilayer membranes is important for membrane protein research and drug discovery. In these sensors, it is critical that the solid-state sensing material does not have adverse effects on the conformation or functionality of membrane-bound molecules. In this work, pore-spanning lipid membranes are formed over an array of periodic nanopores in free-standing gold films for surface plasmon resonance (SPR) kinetic binding assays. The ability to perform kinetic assays with a transmembrane protein is demonstrated with α-hemolysin (α-HL). The incorporation of α-HL into the membrane followed by specific antibody binding (anti-α-HL) red-shifts the plasmon resonance of the gold nanopore array, which is optically monitored in real time. Subsequent fluorescence imaging reveals that the antibodies primarily bind in nanopore regions, indicating that α-HL incorporation preferentially occurs into areas of pore-spanning lipid membranes.
In order to interface lipid membranes with biosensors, many groups have employed supported lipid bilayers (SLBs) or black lipid membranes (BLMs).4 In a SLB, the membrane is formed on a solid substrate, whereas in a BLM, the membrane is suspended over a small aperture in a solid substrate. SLBs have been employed in a number of biochemical and biophysical studies on the fundamental properties of lipid membranes and can be formed by different methods.5–9 Because SLBs are formed directly on a solid substrate with a thin layer of water (1–2 nm) between the membrane and the substrate, steric hindrance makes incorporation of integral proteins challenging, and those proteins that are inserted can be denatured.10,11 In order to overcome this issue, some groups have tethered lipid bilayers to the underlying substrate using linker molecules, or formed a polymer cushion between the membrane and the substrate.12 While these strategies have proven useful for incorporating proteins into lipid bilayers for biosensing,13 the lipid membrane is still only accessible from the top side, limiting their utility for studying transmembrane signaling processes. An alternative approach, suspending membranes over apertures in a substrate in the form of a BLM where both sides of membrane can be accessed, has also been employed for interrogating membranes and membrane-associated proteins. Studies on the electrical properties of lipid membranes and reconstituted ion channels have made extensive use of BLMs, even though the stability of BLMs is notoriously poor. To enhance the long-term stability of BLMs, some groups have employed nanometer-sized apertures for electrical and physical property measurements of membranes and proteins.14–17 This configuration allows for the incorporation of transmembrane proteins into the pore-spanning regions of the membrane; accordingly the ionic transport properties of the membrane proteins can be investigated.18 Similarly, in the present study we demonstrate that lipid bilayer membranes formed by vesicle rupture6 over periodic nanopore arrays in a free-standing metal film can be used for surface plasmon resonance (SPR) sensing.
Recently, a new class of SPR biosensors based on metallic nanostructures has been demonstrated to overcome the limitations of the currently commercialized technology.24–28
Of particular interest are the SPR sensors based on the extraordinary optical transmission (EOT) effect through periodic nanopore arrays in metallic films.29–31 The periodic array of nanopores can act as a diffraction grating and convert incident light into SP waves at resonance wavelengths, thereby creating a series of intense peaks in the optical transmission spectra. Following an initial demonstration of using the EOT effect for biosensing,32 several groups have shown the potential of this platform for label-free kinetic SPR detection and high-throughput imaging.33–40 While the focus of previous work has mainly addressed the optical design, sensitivity and the development of multiplexing capability, the unique potential of the nanopore geometry, i.e. the ability to naturally integrate SPR biosensing with nano-BLMs and proteins integrated therein, has not been explored. Dahlin et al. demonstrated a glass-supported phospholipid bilayer formed at the bottom of nanopores that are randomly distributed in a 20 nm-thick gold film on a glass substrate for localized surface plasmon resonance (LSPR) biosensing.41,42 However, the LSPR effect in randomly distributed single nanopores results in a broad extinction spectrum that is weaker than the EOT effect through a periodic nanopore array.31 Also, the LSPR sensors have a limited probing range because the LSPR decay length is about 20–50 nm. In contrast, in the visible range, the sensors based on propagating SP waves such as periodic nanopore arrays can probe about 100–300 nm from the metallic surface.43
In this work, we present periodic nanopore arrays in free-standing Au/Si3N4 films as SPR biosensors to detect the incorporation of a transmembrane protein, α-hemolysin (α-HL), into a pore-spanning membrane and also to detect subsequent antibody binding. α-HL is a water-soluble peptide monomer (33.2 kD) secreted from the pathogenic bacteria Staphylococcus aureus that binds to the plasma membranes of many mammalian cell types.44 Upon binding to the membrane the monomers freely diffuse about the membrane and self-assemble to form a heptameric conductive transmembrane pore.45 The crystal structure of α-HL was determined by Song et al. and shows a membrane-penetrating β-barrel and a mushroom shaped cap domain that protrudes approximately 70 Å above the membrane surface.46 Because α-HL has been so well characterized, it is a commonly used model for studies of transmembrane proteins as well as for nanopore-based biosensing.47,48
Fig. 1 (a) Concept of a nanopore array in free-standing Au/Si3N4 films for sensing in a suspended lipid membrane environment. The optical transmission is modulated by the presence of a lipid membrane, formed by vesicle rupture, and the subsequent molecular binding on the nanopores array. Part of the lipid membrane is suspended over the nanopores, better mimicking a natural cell membrane. (b) 3-D finite-difference time-domain (FDTD) calculation of the electric field intensity on resonance (787 nm) shows the plasmonic field enhancement at the edges and corners of the nanopores, extending into the membrane region. |
Fig. 1b shows three-dimensional finite-difference time-domain (FDTD) calculations of the field intensity at a transmission resonance (787 nm) in the region of a nanopore array (500 nm pitch). Strong field intensities due to SPs can be observed over the nanopores and near their edges. The intense, highly confined plasmonic fields probe the local refractive index, suggesting high sensitivity in and around the edges of the nanopores49–51 where the pore-spanning lipid membrane forms. The SP fields are, therefore, able to detect the presence of molecules binding to the membrane suspended above the pore.
Fig. 2 (a) A process flow for making a free-standing nanopore array chip integrated with microfluidic channels. (b) A picture of the device on the microscope stage. A single-channel PDMS flow cell was attached on the top surface to inject analytes, and the reservoir on the backside of the chip was sealed by a 300 μm thick PDMS membrane to keep both sides of the pore-spanning lipid immersed in water. |
A polydimethylsiloxane (PDMS, Sylgard) microfluidic channel and a thin PDMS membrane sealed the top and bottom sides of nanopore chip, allowing both sides of the chip, the nanopores, and the lipid membrane to be immersed in aqueous solution. After forming a pore-spanning lipid bilayer via vesicle rupture, analytes were injected through the top-side microfluidic channel for real-time binding kinetic measurements as shown in Fig. 2b. For EOT spectral measurements, a tungsten-halogen lamp illuminated the nanopore arrays from the top-side through a 50× microscope objective. The transmitted light was then collected with a fiber-optic spectrometer.36
Fig. 3a shows a scanning electron microscope (SEM) image of the backside of the free standing Au/Si3N4 film. The bottom reservoir is created by anisotropic wet etching of Si with KOH. Fig. 3b shows a cross-sectional SEM image of the nanopore array milled through the Au/Si3N4 suspended layers. The array consists of 16 × 16 nanopores with the pore diameter of 200 nm and periodicity of 500 nm.
Fig. 3 SEM images of a fabricated free-standing nanopore array chip. (a) The backside of the free-standing film with a bottom reservoir. (b) Cross-sectional image of the nanopore arrays in a free-standing Au/Si3N4 film (colorized for clarity). (c) Cross-sectional image of the nanopore array after the ALD encapsulation process. Inset: The top surface and vertical sidewall of each nanopore is uniformly coated with a 20-nm-thick silica to promote vesicle rupture. |
The Au sensing surface was then coated with a 20 nm-thick silica layer as shown in Fig. 3c using atomic layer deposition (ALD),54 which deposits a conformal layer covering nano-sized structures with precise thickness control. The silica layer uniformly covered the top surface and the sidewall inside each nanopore, which promoted vesicle rupture on its hydrophilic surface, allowing the formation of the suspended lipid bilayer over the nanopore array. In general, propagating SP-based periodic nanopore arrays show sensing ranges of about 100–300 nm in the visible wavelengths, several times longer than LSPR based sensors with randomly distributed nanopores.55 Therefore, the presence of the 20 nm thick silica layer does not severely degrade the EOT detection sensitivity.56
The vesicle rupture pathway, while somewhat dependent on the lipid composition and the nature of the substrate, proceeds through the following general steps. First, a small number of vesicles adsorb to the surface. Then when a critical number of vesicles populate the surface, they begin to rupture, forming a SLB. Consequently, SLB grows by rupture of additional vesicles at the edge of the supported bilayer, driven by hydrophobic interactions between the SLB edge and adjacent vesicles.57 A variety of substrates, such as glass and SiO2 are compatible with vesicle rupture.57,58 Noble metal surfaces can also be used, provided they have chemical functionalization that promotes vesicle/surface interactions and subsequent rupture.59 The vesicles used in this study averaged 361 nm in diameter, determined by dynamic light scattering. The rupture of a vesicle this size leads to a planar membrane disc that is 722 nm in diameter, which is more than sufficient to cover an individual nanopore.
After lipid membrane formation, fluorescence recovery after photobleaching (FRAP) was used to confirm the formation of a continuous lipid membrane on the substrate. FRAP is capable of quantifying the two dimensional lateral diffusion of lipid components on the membrane and is widely used to characterize supported or suspended lipid membranes.
For our experiments, a drop of vesicle solution was placed on the nanopore array surface and the substrate was placed in a humidified box for at least 1 h to allow vesicle rupture and membrane formation. After exposure to the fluorescently-labeled (Rho-PE) vesicle solution and surface washing, a homogenous fluorescent layer was readily apparent indicating that vesicle rupture had occurred and a membrane layer had been formed. The surfaces of the sample were kept immersed in aqueous solution throughout the course of the experiment. In the FRAP experiments, a small circular area of membrane (∼21.5 μm diameter) was photobleached with an intense laser beam at a wavelength of 405 nm. Diffusion of unbleached Rho-PE back into the previously bleached area indicated a continuous membrane layer. Fig. 4a depicts a FRAP experiment over an array of 200 nm holes with 500 nm periodicity. The nanopores are not visible in the fluorescence images, but their presence was confirmed by observing the sample in transmission mode. Four fluorescence images of the sample show the array before photobleaching, right after photobleaching, at 40% recovery and at maximum recovery. Immediately after bleaching a dark spot is evident; the fluorescence of which begins to recover immediately and nears full recovery (∼ 90%) after approximately 50 s, indicating that a continuous membrane is present. Fig. 4b presents the recovery curves for two different experimental conditions: an egg PC membrane over a nanopore array and an egg PC membrane supported on a flat silica surface. Qualitatively, the recovery curves from the two conditions are quite similar, indicating that the presence of the nanopore array does not significantly change the continuity or fluidity of the membrane. Quantitatively, from the recovery curve we are able to calculate the diffusion coefficient D for lipids in the membrane, using the equation D = R2/4τD, where R is the radius of the photobleached spot, and τD is the characteristic diffusion time.60,61 We determine τD from fits of the recovery curves as described by Soumpasis.60 For membranes over nanopore arrays and supported on flat silica, the diffusion coefficients were calculated to be 1.99 ± 0.14 μm2 s−1 and 1.94 ± 0.14 μm2 s−1, respectively which are in the range of previous FRAP results on silica surfaces.42,62,63 The similarity between D values suggests that a pore-spanning lipid membrane is formed over the nanopore array, and that the pores do not limit lipid diffusion.64
Fig. 4 Fluorescence recovery after photobleaching (FRAP) of lipid membranes on flat silica and over nanopore arrays. (a) Frames from a FRAP experiment on an egg PC lipid membrane over a nanopore array. The time of recording for each frame is indicated in the lower left corner. The scale bar represents 10 μm. (b) Average FRAP recovery curves for an egg PC membrane over a nanopore array (red squares) and an egg PC membrane over flat silica (blue dots). |
Recently Jönsson and coworkers performed FRAP analysis on membranes spanning random arrays of 80 nm-diameter wells in a silica substrate.64 Their theoretical analysis of diffusion coefficients in a well-spanning membrane suggests that the apparent diffusion coefficient should be 1.1-times the diffusion coefficient for a membrane totally supported by underlying layer of silica. In reality, they measured an apparent diffusion coefficient that was nearly identical (0.99-times) to the diffusion coefficient of a supported membrane on silica. They suggest that the slight discrepancy between theory and experiment may be due to a small amount of the membrane conforming to the nanowell wall. The experimental parameters presented here are quite similar to those presented by Jönsson et al. In our work, 12% of the surface area of a unit cell of the nanopore array is occupied by a nanopore. The diffusion coefficient calculated for a membrane over a nanopore array is experimentally equivalent or even a bit larger than the diffusion coefficient of a fully supported membrane, which agrees well with the theoretical and experimental results in Jönsson et al., indicating that a pore-spanning membrane is formed. Without the formation of a pore-spanning lipid membrane, the effective diffusion coefficient would be smaller than the value for a SLB on flat silica.64
Additional confirmation of pore-spanning membrane formation was provided by imaging the array from the side opposite that which the lipid membrane is formed (Supplementary Fig. S2†). Fluorescence in the shape of a nanopore array is visible from this viewpoint. Because the chip is opaque, fluorescence excitation and emission can only pass through the nanopores. Fluorescence visible from this viewpoint can only come from fluorophores over the array in the form of a pore-spanning membrane. Taken in combination with FRAP data, this provides strong evidence that the nanopores are covered with a lipid membrane and that the membrane is continuous.
Fig. 5 (a) Transmission spectra change before (black line) and after the formation of pore spanning lipid membrane (red line), after the formation of α-HL pore on the lipid membrane (green line), and after the binding of anti-α-HL (blue line). (b) Zoomed in spectra marked by dashed line in panel (a) where spectral shifts were monitored. (c) Real-time kinetic measurements of α-HL and anti-α-HL. For the positive control, 1 μM of α-HL was injected and incubated for an hour, and then 300 nM of anti-α-HL was injected for the binding with anti-hemolysin while negative control was performed with the same solutions on a lipid-free surface. (d) Spectral shift from the binding of α-HL and anti-α-HL with three negative controls: negative 1 is a lipid-free environment where pore-spanning membrane is replaced by a monolayer of APTES; negative 2 is a lipid-free environment where pore-spanning membrane is replaced by a layer of BSA; negative 3 is the same pore-spanning lipid environment as the positive control, but with the absence of α-HL; negative 4 is the same pore-spanning lipid environment as the positive control, but α-HL is replaced by BSA. |
The spectral shifts from the nanopore arrays were observed at the minima (Fig. 5b) in the transmission spectra marked by the dashed circle in Fig. 5a. After forming the pore-spanning lipid membrane, the resonance wavelength red-shifted. The position shifted further after transmembrane proteins (α-HL) were incorporated into the lipid membrane and subsequent binding of biotinylated anti-α-HL to the α-HL. The real-time kinetics were measured by monitoring the spectral shift of the transmission as a function of time while molecules bind to the surface.
Fig. 5c shows real-time kinetic measurements of α-HL binding on the membrane surface and subsequent binding of anti-α-HL with α-HL. For a positive control, 1 μM of α-HL was injected and incubated for an hour, sufficient to form transmembrane pores.65 Then 300 nM of anti-α-HL was injected to bind with α-HL after PBS washing. A first negative control was performed with the same solutions on a lipid-free surface where the nanopore surface was coated by a monolayer of 3-aminopropyl-triethoxysilane (APTES). By subtracting the spectral response of the negative control from that of the positive control, it is possible to eliminate experimental artifacts such as bulk refractive index changes and non-specific binding.40
To confirm the incorporation of α-HL with the lipid membrane and specific binding between α-HL and anti-α-HL, the spectral shifts obtained from the positive control were compared with, in all, four negative controls: (1) a lipid-free environment with APTES from Fig. 5c; (2) a lipid-free environment where the pore-spanning membrane is replaced by a layer of BSA; (3) the same pore-spanning lipid environment as the positive control, but with the absence of α-HL; (4) the same pore-spanning lipid environment as the positive control, but α-HL is replaced by BSA. As shown in Fig. 5d, spectral shifts from those three negative controls were less than 20% of the shift in the positive control for both α-HL and anti-α-HL binding, indicating that α-HL can be incorporated only with the lipid membrane environment and that specific binding of anti-α-HL can occur only with the presence of α-HL on the membrane surface.
One of the distinguishing advantages of SPR biosensing compared to other sensing techniques is that it enables the measurement of the binding affinity of protein–protein interactions from their real-time label-free association and dissociation binding kinetic curves.19
Fig. 6a shows real-time kinetic measurements with different concentrations (50, 100, and 200 nM) of anti-α-HL. After 5 min of baseline with a PBS solution, an anti-α-HL solution was injected to the channel with a constant flow rate of 30 μL min−1. The association of binding was measured for 15 min and then the channel was washed by PBS solution with the same flow rate of 30 μL min−1 for dissociation. A nonlinear least squares analysis method66 was used to determine the dissociation constant, KD, from the kinetic measurements, giving a value of (1.9 ± 1.0) × 10−8 M. This value is fairly close to the previously reported value with monoclonal anti-α-HL binding.67 We also determined the limit of detection to be 26 nM from the signal at 3 times the standard deviation of baseline noise.
Fig. 6 (a) Real-time kinetic measurements of anti-α-HL binding with different concentrations (50, 100, and 200 nM). The antibody solutions were injected after 5 min baseline with a PBS solution. Association curves were measured for 15 min with the flow rate of 30 μL min−1, and then, the surface was washed by the PBS solution for dissociation curves. (b) Real-time kinetic measurements of 50 nM streptavidin labeled with R-Phycoerythrin (R-PE). The three curves correspond to the three different concentrations of anti-α-HL antibodies from Fig. 4a. (c) Fluorescent images after binding of 50 nM streptavidin-R-PE onto 100 nM anti-α-HL. The negative control is performed without α-HL. |
For further confirmation of the specific binding of anti-α-HL with α-HL on the pore-spanning lipid membrane, 50 nM of streptavidin-R-phycoerythrin (SAPE) was injected to bind with the biotinylated anti-α-HL used in Fig. 6a.
Fig. 6b shows the binding kinetics of 50 nM of SAPE to the membrane-bound anti-α-HL antibodies with the three different concentrations from Fig. 6a. Each curve shows different maximum signal value, Rmax, which corresponds to the maximum capacity of immobilized ligand that binds with analytes in the injected solution. It indicates that the amount of anti-α-HL bound on the surface is different, since Rmax values have a linear relation to the concentration of anti-α-HL injected in Fig. 6a (Supplementary Fig. S3†).
After consequent binding of 100 nM of anti-α-HL and 50 nM of SAPE on the lipid membrane surface, the fluorescent images were taken as shown in Fig. 6c. Positive and negative controls are with and without α-HL on the membrane surface, respectively. The images are shown with the same intensity scale. While the negative control shows partial dim spots due to non-specific binding, the positive control shows strong contrast of fluorescent intensity between the inside and outside of the array. This means that more α-HL was incorporated on the nanopore array where pore-spanning membranes were formed over the nanopores and more anti-α-HL and SAPE bind to those areas.68 These α-HL and anti-α-HL kinetic measurements and subsequent binding with SAPE demonstrate successful formation of the pore-spanning lipid membrane incorporated with a transmembrane protein on free-standing nanopore arrays that are able to detect antibody binding events in real-time using an SPR sensing technique.
For FRAP, 1% (w/w) 1, 2 dimyristoyl-sn-glycero-3-phosphatidylethanolamine-N-(lissamine Rhodamine B sulfonyl) ammonium salt (Rho-PE) was included with the Egg PC. To form the vesicles, chloroform solutions with the proper amounts of egg PC and Rho-PE were placed in a glass vial and the chloroform was evaporated under vacuum for 3 h. The dry lipid film was then rehydrated with NaTrisEDTA (10 mM NaCl, 10 mM Tris, 10 mM EDTA, pH = 8) and refrigerated overnight. The rehydrated lipid suspension was vortex mixed then sonicated for 10 min in a room temperature water bath. The resulting vesicles had a diameter of 361 ± 215 nm, as determined by dynamic light scattering. The vesicles were then transferred to NaTrisCa buffer (10 mM NaCl, 10 mM Tris, 10 mM CaCl2, pH = 8) at a total lipid concentration of 200 μg mL−1. NaTrisCa buffer has been employed by other groups to promote vesicle rupture on silica surfaces.42
Analysis of FRAP image sequences was done with ImageJ software version 1.42q. In each frame of the image sequence the fluorescence intensity of the bleached spot was averaged and fractional fluorescence recovery was calculated as described by Axelrod et al.73 and the recovery curve was fit as described by Soumpasis60 using Matlab. From the fit of the recovery curve, the characteristic diffusion time was determined. The diffusion coefficient (D) was calculated from the average of three experiments with the equation D = R2/4τD, where R is the radius of the bleached spot and τD is the characteristic diffusion time.
Footnotes |
† Electronic supplementary information (ESI) available: Details of fabrication process for making a free-standing nanopore array chip integrated with microfluidic channels and supplementary figures. See DOI: 10.1039/c0sc00365d |
‡ These authors contributed equally. |
This journal is © The Royal Society of Chemistry 2010 |