Therapeutic Micro-Nano Technology BioMEMs - Tejlal Desai & Sangeeta Bhatia

FIGURE 17.1. Schematic of a supported membrane on a patterned substrate. Continuous bilayer membrane coats the substrate while patterns of barrier materials impose boundaries on the fluid membrane. A thin layer of water between the membrane and substrate helps to preserve lateral fluidity of membrane components. Adapted from [74].

hybrid live cell—supported membrane interfaces offer great scientific and technological potential as a means of forming functional interfaces between living and nonliving systems.

In recent years, there has been rapid development of techniques and methodologies to physically pattern and manipulate supported membranes. These generally combine membrane self-assembly processes with various conventional and novel forms of hard and soft lithographic techniques (Figure 17.1) [12, 13]. The resulting barrage of patterned supported membrane structures and devices, which have recently emerged, provide a wealth of opportunities for the integration of biological functionality into microand nano-devices.

In the following, an overview of the current supported membrane technology is sketched. This begins with a brief examination of the salient physical characteristics of membranes, such as fluidity and phase separation properties. The degree to which these various properties can be preserved in a variety of supported membrane configurations is discussed. Next, the practical aspects of supported membrane fabrication and patterning are reviewed. Capabilities and limitations are discussed in parallel, in an effort to provide a usable reference for the design and implementation of supported membrane systems and devices. Lastly, several representative applications are mentioned. These include specific examples illustrating the biological functionality of supported membranes as well some more bioanalytical embodyments. While these examples are by no means intended to constitute a comprehensive survey of uses of supported membrane systems, they are intended to provide a sense of the current state of the field. Ultimately, the implementation of supported membrane technology into devices is in the early stages of development, and it is hoped that this review provides background information for many new applications of this emerging technology.

17.2. PHYSICAL CHARACTERISTICS

A definitive feature of supported membranes is the long-range lateral mobility of membrane lipids. The natural fluidity of free lipid bilayers is preserved in the supported membrane configuration. This distinguishes supported membranes from the vast repertoire of immobile

FIGURE 17.2. Fluorescence recovery after photobleaching (FRAP) experiment on a supported membrane. First panel: Initially uniform fluorescence from the yet unbleached membrane. Second panel: Bleach pattern shortly after exposure. Subsequent panels: Time sequence of images illustrating diffusive mixing of the bleach pattern with nearby unbleached regions of membrane. Adapted from [80].

surface coatings and thin films, such as silane monolayers, polymer layers, adsorbed proteins etc. The consequences of lateral mobility are multifold. At the most basic level, free movement of lipids enables the membrane to react to the presence of proteins, charges, and physical forces in a dynamic and responsive manner. This characteristic ability of fluid membranes to reorganize upon interaction with external perturbations is proving to be paradigmatic in the functionality of live cell membranes [14–17]. Lateral rearrangement of membrane components, in turn, enables higher levels of behavior, ranging from basic physical chemical phenomena of membranes, such as miscibility phase separation (raft formation) [18–20], to bulk transport, such as membrane microelectrophoresis [21–26]. Preservation of lateral fluidity in supported membranes tremendously enhances their biological functionality as well; this important aspect of the technology is discussed in greater detail below.

The fluidity of supported membranes can be quantitatively characterized by a number of optical techniques. Fluorescence recovery after photobleaching (FRAP) is one of the most common ways of measuring molecular diffusion coefficients in membranes [27, 28]. FRAP measurements rely on doping a small quantity of fluorescent probe molecules, usually covalently coupled to lipids, into the membrane. A brief burst of intense excitation light is projected onto the membrane, photobleaching probe molecules within the exposure zone. This bleach pulse must form a spatial pattern on the sample, which then provides the basis to monitor diffusive mixing over time. An example of a photobleached pattern in a supported membrane at various stages of diffusive recovery is illustrated in Figure 17.2. The first panel of the sequence depicts uniform fluorescence from probe molecules prior to the bleach. The second panel depicts the bleach pattern shortly after exposure. In this example, excitation light was projected through a photomask to produce the pattern. For a basic FRAP measurement, a single bleach spot will suffice and the field stop aperture on conventional fluorescence microscopes is frequently used for this purpose [29]. Subsequent panels in Figure 17.2 illustrate the transformation of the bleach pattern by diffusive mixing. Quantitative analysis of fluorescence intensity in a diffusing pattern can yield accurate measurements of the diffusion coefficient for the fluorescently labeled species. Measurements of lipid diffusion coefficients in supported membranes typically range from 1–10 μm2s−1 with similar rates observed for membrane-linked proteins. In recent years, fluorescence correlation spectroscopy (FCS) and single particle or molecule tracking have been gaining popularity as means of assaying molecular mobility in membranes [30–39]; these techniques can be readily applied to supported membranes as well.

Looking beyond the mobility of individual lipids, phase separation and the collective mobility of phase-separated domains is also an important structural aspect of membranes. Innumerable studies of cell membrane rafts and raft-like phase separated domains in model membranes have emerged recently [3–6]. Miscibility phase transitions can also be seen in supported bilayers, however influences of the supporting substrate can be significant [19, 40, 41]. Substrate influences generally immobilize transmembrane proteins in supported membranes as well, although individual protein functionality is not necessarily destroyed [29].

In efforts to minimize the influence of the supporting substrate on membrane structure, a variety of polymeric substrate materials have been investigated [9]. These have included polyacrylamide [42], polyethylenimine (PEI) [43, 44], dextran [45], trimethylsilylcellulose (TMSC) [46–48], chitosan [49], and hyaluronic acid [50]. Additionally, several tethering strategies involving silane-polyethyleneglycol-lipid [51], lipopolymers consisting of poly(ethyloxazoline-co-ethyleneimine) (PEOX-PEI) with alkyl-chain side groups [52, 53], and streptavidin protein coupling between biotinylated lipid and biotin-derivatized substrates [54], have been developed to stabilize the membrane polymer interface. Tethering of membranes directly to a solid substrate, without intervening polymer, has also been employed in efforts to control the aqueous layer between the membrane and substrate [55–57]. This collection of polymer-supported membrane systems offers a range of differing physical characteristics. Notably, the silane-polyethyleneglycol-lipid tethered polymer-supported membrane can allow significant lateral diffusion of integral membrane proteins (cytochrome b5 and t-SNARE) [51, 58]. However, no one system has yet emerged as clearly superior and new strategies continue to be developed.

One such alternative involves using a conventional supported membrane as the foundation on top of which a second membrane is deposited (Figure 17.3). The second membrane can be deposited by rupture of a giant unilamellar vesicle (GUV) onto a preformed supported membrane [59, 60] or by successive transfer of four monolayers from the airwater interface [61, 62]. These supported membrane junctions have utility in the study of membrane-membrane interactions [63], such as occur at intercellular interfaces, but for the moment we focus on the unique physical characteristics of the upper membrane in the junction for purposes of comparison with membranes supported on solid substrates.

Resolution of membrane structure on surfaces can be achieved using a combination of fluorescence techniques, which provide real-time imaging of membrane topographical patterns. Intermembrane fluorescence resonance energy transfer (FRET) occurs between membranes, which have been doped with complementary fluorescent probes, when the intermembrane spacing is comparable to the Forster¨ distance for the probe pair ( 5 nm). Quantitative analysis of FRET efficiency provides measurement of intermembrane spacing with sub-nanometer precision in closely spaced membrane junctions [59, 60]. Measurements of intermembrane spacings beyond the range of FRET can be achieved using fluorescence interference contrast microscopy (FLIC) [19, 59, 64, 65]. This technique exploits the spatial intensity variation within an optical standing wave to modulate the fluorescence intensity of probes as a function of their position along the optical axis, which is perpendicular to the interface in this configuration. FLIC can resolve topographical structures extending hundreds of nanometers from the primary plane with nanometer resolution. Reflection interference contrast microscopy (RICM) can provide similar topographical information as FLIC, and has been successfully applied to supported membrane systems [66–68]. RICM requires transparent substrates whereas FLIC requires reflective substrates, which may or

FIGURE 17.3. Schematic illustrating two types of supported membrane junctions. Adapted from [19].

may not be transparent. For a review of optical techniques for imaging membrane surface topography, see reference [69].

The most notable feature of the upper membrane in a supported membrane junction is the existence of two distinctly different states of adhesion to the lower bilayer membrane. The first, referred to as Type 1, is characterized by uniform intermembrane FRET, which indicates intermembrane separation distances within a few nanometers. In the second state (Type 2), large intermembrane spacings ( 50 nm) are maintained by a balance between Helfrich (entropic) repulsion [70] and occasional adhesion sites that pin the two membranes together. No intermembrane FRET is visible in Type 2 junctions, however FLIC reveals large-scale thermal undulations. FLIC images of the two junction types are illustrated in Figure 17.4. The existence of multiple states in membrane junctions reveals some of the range of possibilities for the association of membranes with surfaces. At present, Type 2 junctions have only been formed by rupture of GUVs onto supported membranes. Presumably the extremely weak interaction between the two membranes in this configuration precludes monolayer transfer techniques. Further development of secondary membrane deposition methods is required before these systems can be implemented with the full range of supported membrane patterning techniques described below.

The physical properties of membranes in the different types of junction have been compared using a phase-separating mixture of phosphatidylcholine (PC), cholesterol, and sphingolipid. Below the miscibility transition temperature, the mixture separates into coexisting liquid phases, which can be observed directly by fluorescence microscopy. Miscibility

A C

B D

FIGURE 17.4. Fluorescence interference contrast images (FLIC), A and C, along with corresponding height profiles, B and D, for Type 1 and Type 2 membrane junctions systems. Adapted from [19].

phase transition temperatures are unaffected by the state of adhesion. This indicates that close apposition to another membrane surface in Type 1 junctions does not substantially affect mixing-demixing thermodynamics at the molecular level. However, collective motion of phase separated domains in the two junction types differ substantially. Domains in the upper membrane of Type 2 junctions exhibit rapid Brownian motion while similar domains in Type 1 junctions and primary supported membranes remain nearly fixed in place. In contrast, lateral diffusivities of individual molecules are of similar magnitude in each of these three configurations. These results reveal the influence of the separation distance between a membrane and a supporting substrate on the size-scaling characteristics of diffusion coefficient. Correspondingly, non-equilibrium size distributions of phase-separated domains can be stabilized in supported membranes and Type 1 junctions relative to free membranes, despite rapid diffusion of individual molecules. It is important to consider these facts in the design and study of supported membranes containing phase separating mixtures of lipids.

17.3. FABRICATION METHODOLOGIES

Supported membranes can be most easily formed by spontaneous adsorption and fusion of small unilamellar vesicles (SUVs) with an appropriate substrate [7, 11 , 12, 29]. The

vesicle fusion process is quite general and accommodates a wide range of lipid and protein compositions. Clean silica surfaces have proven to be one of the best substrates although a variety of polymeric substrates can also be used [9], as discussed above. Other methods of membrane deposition include monolayer transfer from an air-water interface [71] and membrane spreading [72], both of which can also yield high quality supported membranes. Alternatively, detergent-solubalized membrane proteins, in one case G-protein coupled receptors (GPCRs), and lipid have been assembled into supported membranes from solution on carboxylated dextran surfaces with modified alkyl groups [73]. For membranes assembled on silica by the vesicle fusion process, a variety of forces including electrostatic, van der Waals, and hydration are known to tightly trap the freely supported membrane in a plane, separated from the solid surface by a nanometer layer of water [8]. This water layer prevents the substrate from interfering with the responsive membrane bilayer structure, thus preserving physical attributes of the membrane such as lateral fluidity. The membrane patterning techniques and applications described below have been implemented with silica-supported membranes, unless otherwise noted.

The fluid nature of supported membranes presents intrinsic difficulties with respect to the formation of membrane patterns on surfaces. A number of solutions have now been developed that enable membrane patterning into a wide range of configurations [12, 13]. One general theme revolves around the use of solid-state patterns on the substrate to impose structure onto the supported membrane [74]. This strategy is quite simple: materials that do not readily support continuous membranes can function as barriers to lateral fluidity when patterned onto a membrane compatible substrate. A wide range of materials including metals (Au, Al, Cr, Ti, etc.), metal oxides (Al2O3, TiO2), and some polymers (including proteins patterned by microcontact printing [75]) have proven to function as effective barriers. Lipid vesicles readily adsorb onto a wide range of surfaces, including many barrier-forming materials. Formation of supported membranes is a two-step process in which vesicles first adsorb on the surface and subsequently fuse together to form a single, continuous supported membrane [76–78]. The distinctive characteristic that renders a material a good barrier to lateral diffusion in supported membranes is the tendency to inhibit the fusion step of supported membrane formation. Thus, although lipids are frequently present on the surface of barrier materials in the form of adsorbed, but unfused vesicles, long-range lateral diffusion is arrested.

Grids of barriers can partition a membrane into an array of isolated fluid corrals. In this configuration, membrane within each corral is fully connected and fluid. However, the barriers prevent mixing between separate corrals. This is the basis of the membrane microarray [74, 79]. Several strategies for filling each corral of the array with a different membrane composition have been explored. These include photolithographic modification [80], microcontact printing [81, 82], and direct micro-deposition [83, 84]. An array of membrane compositions within a grid of diffusion barriers is illustrated in Figure 17.5. In this example, a light pattern was projected onto the membrane in registry with the corrals. Each corral thus receives an individual dose, specified by the exposure area, and is photochemically altered accordingly. Diffusive mixing within each corral leads to a uniformly blended composition; thus a continuous composition scale can be achieved with a binary exposure.

Membranes can be directly deposited by microcontact printing [81, 82]. In this strategy, a polymeric stamp such as poly-(dimethylsiloxane) (PDMS) is used to deposit sections of

FIGURE 17.5. Array of 10 × 10μm fluid membrane corrals of differing compositions. A. Original membrane array prior to photolithographic patterning. B. Photomask. C. Projected illumination image on membrane array. D. Resulting array of photochemically altered membrane corrals. Adapted from [80].

membrane with geometry defined by the topography of the stamp. The stamp is “inked” with membrane by vesicle fusion to the freshly oxidized PDMS. A supported membrane forms on the PDMS. This membrane can then be transferred as a unit to a silica substrate by bringing the stamp into contact with the substrate. Membranes are thus literally printed. An important point, which is well illustrated by this method, is that material barriers on the substrate are not required to maintain separation between patches of membrane. If deposited with sufficient separation, separated patches of membranes tend to remain nearl fixed in place and will not mix. In this case, the barrier consists of an uncoated region of the original substrate between patches of supported membrane. Microcontact printing is particularly convenient for applications in which a small number of distinct membrane compositions are needed; for example, to characterize the way cells discriminate between two membrane types. This technique is easily implemented by hand. However, scale-up to automated printing would be challenging due to the sensitivity of the membrane transfer process to applied force. The total force needed for good transfer is pattern dependent.

The ability to print supported membrane patterns that partially cover a substrate allows for subsequent deposition of a second membrane composition into the uncoated regions of the substrate. By combining this strategy with prepatterned grids of diffusion barriers, composition arrays consisting of numerous blending ratios of the two original membrane compositions can be created [81, 82]. The starting substrate is prepatterned with a grid of diffusion barriers. The first membrane composition is then printed onto the substrate in registry with the grid. Different amounts (area) of this first membrane are contained within the various corrals. The second membrane is then deposited by blanket vesicle fusion over the whole substrate, filling in the exposed areas. Diffusive mixing within individual corrals produces uniformly blended compositions of the two membranes in ratios defined by the fractional areas of each within the corral.

Intermembrane adhesion and molecular-exchange provides a mechanism of altering the composition of fully preformed supported membranes. The adsorption and fusion of SUVs (25–250 nm diameter) is generally self-limiting at a single bilayer. However, when there is sufficient adhesive interaction between two membranes, fusion followed by diffusive mixing of membrane components can occur. This process is widespread in biological systems where selectivity and regulation is mediated by proteins such as the SNAREs [85]. Selective intermembrane exchange with supported membranes has been achieved using electrostatic complementarity between the delivery membrane and the target membrane [86]. The method has been implemented with SUVs and membrane-coated silica beads serving as the delivery agents. An attractive aspect of using membrane-coated beads is that they can be recaptured to allow retrieval of molecular components from supported membranes. These methodologies can also be implemented within microfluidic device environments.

Membrane patterns similar to those achievable by microcontact printing, though with somewhat higher spatial resolution, have been produced using a polymer lift-off technique [87]. In this process, a thin film of parylene is first patterned on the substrate, photolithographically. A membrane is then deposited, uniformly, over the substrate. The parylene thin film can then be peeled off, leaving membrane patterns stenciled onto the bare substrate. This variation of membrane patterning is expected to leave particularly clean regions between the membranes, and its application in live cell experiments suggests this is the case [87]. A drawback of this technique in its present form is that only one membrane composition may be patterned.

A very general method, suitable for the formation of large scale patterns and arrays of membranes, is by direct deposit. Though tedious when done manually, it is possible to fill the corrals of a membrane array by directly depositing droplets of vesicle suspension [83, 84]. Commercial robotic arraying systems, originally developed for DNA array fabrication, can be adapted to print membrane arrays.

An elegant example of membrane patterning, which combines microfluidic flow channels with grids of diffusion barriers, was introduced by Kam and Boxer [88, 89]. Two or more vesicle suspensions are allowed to flow down a channel under partial mixing conditions. Vesicles fuse with the substrate and form a supported membrane, the composition of which represents the particular blend of vesicle compositions that was present above that region of the surface. Grids of diffusion barriers on the substrate effectively bin the captured membrane, resulting in a permanent composition array. The fact that this method produces membrane arrays inside of microfluidic systems is a significant advantage in light of recent advances in microfluidics [90, 91].

17.4. APPLICATIONS

17.4.1. Membrane Arrays

The ability to create arrays of discrete fluid membrane corrals on surfaces rapidly led to application of membrane array technology to drug discovery [13, 92]. In this context, the technology offers two levels of advantage. At one level there is the intrinsic scaling benefit common to all array technologies. DNA and protein arrays, for example, have found a niche based largely on scaling benefit [93–96]. New scientific capabilities, such as comprehensive

gene expression or proteome profiling, are enabled by the sheer quantity of information that can be collected with the array. Supported membrane systems, in arrays or as individual membranes, additionally offer greatly enhanced levels of biological functionality. This latter point is of critical importance in drug discovery, and has fueled the development of supported membranes for a number of applications.

The G protein-coupled receptors (GPCRs) are a class of membrane proteins well known to be a rich source of drug targets [97]. Correspondingly, they have been among the most desired targets for integration in supported membrane systems and a number of strategies have been developed [57, 73, 98]. Although detailed structural data on the state of the GPCR membranes in these examples is not available, the reported ligand binding data illustrate that some of the surface adsorbed protein remains functional nonetheless. Incorporation of large transmembrane proteins into uniform and fully fluid supported membranes has been achieved previously [29]. Thus it is likely that the GPCR technology will continue to advance.

An array based immunoassay for monitoring the binding of protein to fluid membranes has been introduced in a microfluidic format [90, 99]. In one configuration, each microchannel of the array is coated with the same membrane. Different protein concentrations can then be run down each channel while binding is measured by fluorescence. Protein binding was monitored with a total internal reflection fluorescence (TIRF) system. Conventional fluorescence imaging systems could be used as well by including a rinse step. Alternatively, this configuration is also compatible with surface plasmon resonance (SPR) detection methods, which offer the added advantage of detection without the need to fluorescently label the protein of interest. By imaging an array of channels simultaneously, an entire binding curve can be read out in a single measurement.

17.4.2. Membrane-Coated Beads

Membranes supported on colloidal silica or polymer particles have proven to be an effective format for bioanalytical applications. Membranes can be assembled on the surface of silica beads by the vesicle fusion procedure and are essentially equivalent to planar supported membranes [100, 101]. The bead format is readily compatible with high-throughput screening and is a commercial product [102, 103]. The colloidal behavior of a population of membrane coated beads offers intriguing methods of detection and readout [104]. The behavior of a colloidal system is driven by the pair interaction potential between particles. In the case of membrane-derivatized silica beads, the pair potential is dominated by membranemembrane interactions. Hence, the collective phase behavior of the system is responsive to details of the interaction between membranes. The strength of this interaction can be tuned by adjusting the membrane composition. Positioning the system near a phase transition sensitizes it to small perturbations of the membrane surface. Thus the collective phase behavior serves as a cooperative amplifier that produces a readily detectable response from a small number of molecular events on the membrane surface. For example, protein binding to membrane-associated ligand at densities as low as 10−4 monolayer (corresponding to10 molecules per interface) has been observed to trigger a phase transition.

A typical colloidal phase transition triggered by protein binding to membrane surface ligand is depicted in Figure 17.6. To perform the assay, membrane-derivatized beads are dispersed, underwater, where they settle gravitationally onto the underlying substrate and

FIGURE 17.6. Protein binding-triggered colloidal phase transition of membrane-coated beads. A. Time sequence of images depicting the transition from a condensed to a dispersed colloidal phase, triggered by addition of. Transitions were triggered only when the appropriate ligand was also incorporated into the membrane. B. Corresponding plots of g(r ) for the time sequence in A. Adapted from [104].

form a two-dimensional colloid. The beads exhibit free lateral diffusion and the system behaves as an ergodic fluid. For highly sensitive assays, the membrane composition is tuned so that the colloid weakly condenses, as seen in the first panel of 6A. The condensed distribution is dynamic, with individual beads continuously evaporating and recondensing into clusters. However, the overall structure is invariant. Protein binding to membrane surfaces generally triggers a condensed to dispersed phase transition. Figure 17.6 depicts a time sequence of a phase transition triggered by addition of protein at t = 0 s. These experiments were performed with 300 μl solution in 5 mm round wells of a 96-well plate. Within 30 s of adding a drop of protein solution to the top of the well, uniform disruption of the condensed phase was discernable. Within 60 s, the colloid attained a dispersed distribution. Individual bead mobility is unaffected by protein binding. Exposure