Bioelectronic silicon nanowire devices using functional membrane proteins

Abstract

Modern means of communication rely on electric fields and currents to carry the flow of information. In contrast, biological systems follow a different paradigm that uses ion gradients and currents, flows of small molecules, and membrane electric potentials. Living organisms use a sophisticated arsenal of membrane receptors, channels, and pumps to control signal transduction to a degree that is unmatched by manmade devices. Electronic circuits that use such biological components could achieve drastically increased functionality; however, this approach requires nearly seamless integration of biological and manmade structures. We present a versatile hybrid platform for such integration that uses shielded nanowires (NWs) that are coated with a continuous lipid bilayer. We show that when shielded silicon NW transistors incorporate transmembrane peptide pores gramicidin A and alamethicin in the lipid bilayer they can achieve ionic to electronic signal transduction by using voltage-gated or chemically gated ion transport through the membrane pores.

Biological systems use a combination of ion gradients, flows of small molecules, membrane electric potentials, and even light to achieve an astonishingly effective control over signal transduction that is still largely unmatched by manmade devices. To gain this level of sophistication nature has evolved a vast arsenal of highly specific receptors, active and passive ion channels (1, 2), photo-activated proton pumps, and ion pumps (3). Utilization of these components in electronic circuits could achieve seamless integration of biological and manmade structures (4) that would enable superior biosensing and diagnostics tools (5), advanced neuroprosthetics (6), and more efficient computers (7). Previous attempts at integrating biological systems with microelectronics range from the early works on capacitive stimulation of cells (8) to monitoring neuronal activity with field-effect transistors (FETs) (9) to a recent example of using nanowire (NW) transistor arrays to follow neuronal signal propagation (6). Nanomaterials that have characteristic dimensions comparable to the size of biological molecules open up the possibility of such integration at an even more localized level. Some early examples include using carbon nanotubes as carriers for transporting intracellular proteins and DNA (10) and using silicon NWs (SiNWs) as gene delivery vehicles for mammalian cells (11). Other work has taken advantage of the superior electronic properties of NWs to create specific electronic detectors for a variety of biomolecules (5).

An important step toward building bionanoelectronic interfaces would involve functional integration of nanomaterials with membrane proteins. Lipid membranes occupy a special place in the hierarchy of the cellular structures as they represent important structural and protective elements of the cell that form a stable, self-healing, and virtually impenetrable barrier to the ions and small molecules (12). However, a lipid membrane is also a nearly universal matrix that can house a virtually unlimited number of protein machines that perform a large number of critical recognition, transport, and signal transduction functions in the cell (13). Despite some initial work (14–16), the possibilities of using lipid membranes in nanoelectronic devices remain virtually untapped. We have incorporated lipid bilayer membranes into SiNW transistors by covering the NW with a continuous lipid bilayer shell that forms a barrier between the NW surface and solution species. We show that when this “shielded wire” structure incorporates transmembrane peptide pores it enables ionic to electronic signal transduction by using voltage-gated and chemically gated ion transport through the membrane pores.

Results and Discussion

We have built our bionanoelectronic devices by using a microfabricated SiNW transistor platform. In these devices, a SiNW is connected to a pair of metallic source and drain electrodes (Fig. 1 A and B). We used semiconducting p-doped SiNWs with the diameters in the 20- to 40-nm range (Fig. 1C) synthesized by catalytic chemical vapor deposition (CVD) procedures that have been well-described in the literature (17). High quality of the starting NW material is important for achieving high performance of the transistor devices, and transmission electron microscopy (TEM) images (Fig. 1C) showed that the NWs were crystalline with a thin layer of native oxide on the surface. We then fabricated NW transistors by depositing SiNWs on the surface by using a flow-alignment procedure and then connecting them to the source and drain electrodes by using conventional photolithography. Good isolation of the electrodes is critical for successful operation of these devices in liquid; therefore we passivated the source and drain electrodes by coating them with a conformal 80-nm-thick layer of Si₃N₄ (see Methods for details). The finished chips then were mounted in a fluid cell that featured polydimethylsiloxane (PDMS) microchannels for delivery of test solutions and a reference gate electrode inserted in the solution channel. As-fabricated SiNW transistors show Ohmic contacts, on–off ratios of ≈10⁴, and transconductances of 100–500 nS (Fig. 1D), comparable to the parameters reported for similar devices operating in liquids (18).

Fig. 1.

Bionanoelectronic devices incorporating lipid-coated SiNWs. (A) Device schematics showing a NW connected to microfabricated source (S) and drain (D) electrodes. (Insets) The configuration of the lipid bilayer and a pore channel placed in the bilayer membrane. (B) An SEM micrograph of the NW transistor showing a NW bridging the source and drain electrodes. (Inset) A photograph of the device chip covered with a PDMS flow channel. (C) TEM micrograph of the as-synthesized SiNW. (D) A typical IV characteristic of the SiNW transistor in fluid. (E) Cyclic voltammetry curves measured for an uncoated SiNW device (black line) and a device coated with the lipid bilayer (red line). Fe(CN)₆ solution (10 mM) was used as a redox agent. (F) Time traces of the normalized conductance of an uncoated SiNW transistor (black line) and SiNW transistor coated with the lipid bilayer (red) as the pH of the solution in the fluid cell was changed from 6 to 9.

The key procedure for building our bionanoelectronic device was the formation of the lipid bilayer on the surface of the SiNW. The hydrophilic negatively charged native oxide present on the NW surface in solution makes NWs particularly attractive as a template for supporting lipid bilayer formation. We have reported that unilamellar vesicles fuse onto a SiNW surface producing a conformal lipid bilayer coating (19). Scanning confocal fluorescence microscopy images (see SI Appendix) showed that 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC) lipid bilayers form a continuous coatings on the NWs, which for the surface-bound NWs is likely to be “omega”-shaped (Fig. 1A), and for the NWs suspended over a TEM grid hole it should resemble a core-shell NW-lipid structure. Significantly, in-plane lipid molecule diffusion coefficients measured by fluorescence recovery after photobleaching (FRAP), ≈8 μm²/s on suspended wires and ≈5 μm²/s on substrate-bound NWs, were comparable to the values reported for the supported lipid bilayers on flat glass substrates (20). These results indicate that the lipid bilayers on SiNWs are highly fluid, which is a vital property that enables incorporation of biologically active structures into the membrane. The second key property of the lipid bilayer is its ability to shield the underlying NW surface from the solution species. Indeed, cyclic voltammetry measurements using K₄Fe(CN)₆ as a redox probe showed that formation of the lipid bilayer on the NW surface reduced the limiting current by 85–95% relative to the uncovered NW device (Fig. 1E).

To show that our devices can detect selective and specific transport of ions through the lipid membrane pore, we have exploited the sensitivity of the NW electrical response to the proton concentration in the solution. As the solution pH changes, charging of silanol groups at the silicon oxide layer on the surface of p-type SiNWs leads to changes of the depletion region in the SiNW channel that then affects the source-drain current at a given gate voltage (21). Bare NW FETs showed a pronounced increase in conductance when the pH of the fluidic environment around the NWs was changed from 6 to 9 (Fig. 1F), which corresponded to the average pH sensitivity of 50–100 nS/pH in the pH range of 5 to 9. The observed response kinetics is not limited by the device speed; rather it reflects the kinetics of solution mixing in the fluid cell. Indeed, relative conductance traces measured in a particular experiment typically collapse on a single curve shape when normalized to the same initial and final levels, which strongly argues that our kinetics are limited by the reagent delivery.

Note that protons are much smaller and much more diffusive ionic species than most redox species, including Fe(CN)₆⁴⁻; hence they present a much more stringent test to the shielding capacity of the lipid bilayers. Still, lipid bilayer formation on the SiNW surface led to an ≈10 times decrease in the FET response to the pH changes in the fluid cell. These data indicate that the DOPC bilayer membrane blocks proton transport between the fluid environment outside the lipid bilayer and the hydration layer situated between the inner leaflet of the bilayer and SiNW surface.

The final element of our bioelectronic device platform is a membrane pore channel incorporated in the lipid bilayer. The first device example that we report incorporates gramicidin A pores. Gramicidin A is a short helical polypeptide from Bacillus brevis, which can dimerize in the lipid membrane to form a transmembrane channel (Fig. 2A) that allows passage of small monovalent cations, while being impermeable to anions (22). The ionic conductance of the gramicidin channel can also be selectively blocked by divalent ions (22) such as Ca²⁺ that can bind near the mouth of the pore (Fig. 2 A and B). Recent experiments demonstrated that ionic conductance of gramicidin A channels could be used to gate macroscopic electrochemical transistors (23). Similarly, incorporation of gramicidin A channel into the lipid bilayer NW FET device leads to a dramatic recovery of the pH response of the NW FET device (Fig. 2C), indicating that functional ion channels were formed in the lipid bilayer coating the NW. These data also confirm the long-range lateral fluidity of our lipid bilayer membranes, which is a prerequisite for the successful transleaflet dimerization of gramicidin A. Moreover, addition of 1 mM Ca²⁺ to the solution dramatically alters the device behavior: the magnitude of the response to pH change drops by ≈60% (Fig. 2C). These results are consistent with reports that show a 20–50% decrease in conductance of gramicidin A channels in the presence of Ca²⁺ ions (24). Our measurements demonstrate that our bionanoelectronic device platform can successfully convert chemical signals (pH change) into an electrical signal and that device efficiency can be regulated by using the same ligand-gating mechanism used by the biological systems.

Fig. 2.

Ligand-gated operation of devices incorporating gramicidin A pores. (A and B) Schematic showing proton transport in the bilayer incorporating a gramicidin A pore in the absence (A) and presence (B) of Ca²⁺ ions. (C) Time traces of normalized conductance of the SiNW device recorded as the solution was changed from pH 5 to 7 for an uncoated NW device (red trace), a device coated with lipid bilayer incorporating gramicidin A pores (blue trace), and a device coated with the lipid bilayer incorporating gramicidin A pores in presence of Ca²⁺ ions (black trace). Dashed vertical line indicates the time when the pH of the fluid cell input stream was switched from the lower to the higher value.

Another exciting possibility is to use the intrinsic electronic functionality of the device itself to control ion transport through an ion channel in the lipid membrane surrounding the NW. We demonstrated this gating mechanism by using devices that incorporated self-inserting voltage-gated alamethicin (ALM) pores. ALM is a peptide antibiotic from the fungus Trichoderma viride that is often used to mimic nerve cell action potential across artificial membranes. ALM forms ion channels in lipid bilayer membranes (Fig. 3A) by spontaneous insertion and aggregation of 4–6 individual ALM helices into a helix bundle (22). Although some debate about the exact mechanism of the voltage-gated transport in ALM pores remains (25–27), research indicates that at positive applied transmembrane voltages on the insertion side (cis-side) ALM forms barrel-staved functional open pores big enough for small monovalent cations to diffuse through, whereas these pores do not form at zero bias or negative bias (25). Some evidence (28) indicates that at negative membrane potentials ALM helices do not span the entire transmembrane thickness, yet at the positive membrane potentials the helices tilt enough to penetrate the membrane completely and form a channel.

Fig. 3.

Voltage-gated operation of the devices incorporating ALM pores. (A) Schematics showing the mechanism of voltage-gated proton transport in self-assembled ALM pores in the lipid bilayer. (B) Time traces of normalized conductance of the SiNW device held at gate bias of 0V recorded as the solution was changed from pH 6 to 9 for the uncoated nanowire (blue trace), coated nanowire (black trace), and the coated NW device incorporating ALM pores. (C) Time traces of a similar experiment recorded at gate bias of 0.15 V. Vertical dashed lines indicate the time when the pH of the fluid cell input stream was switched from the lower to the higher value.

We exploited this behavior in our devices by using the electric field applied by the device to open and close the ALM pores in the lipid bilayer. Indeed, at zero applied gate bias, ALM pores were “turned off” and the device response was nearly identical to the response of the NW coated with a pure lipid bilayer (Fig. 3B). However, when we applied a positive gate voltage of 150 mV to the device, we obtained a dramatically different response (Fig. 3C): the device showed a strong response to the pH change of up to 50% of the response observed for the bare NW, indicating that the ALM pores had been turned on. These results demonstrate that our device platform is not only capable of using biological components as functional parts of the device, but it also can control the functionality of the biological molecules, bringing the functional integration of the biological and nanoelectronic components on a new level.

It is tempting to assign the observed levels of signal change to the differences in shielding of the NW surface by the lipid bilayer with open ALM pores compared with the uncoated NW. In reality, the situation is likely more complicated. Ionization of the silanol groups on the surface of the NW produces immobile charges that interact with the mobile charges present on both sides of the lipid membrane and establish an electrical potential on the membrane in accordance with the Donnan membrane equilibrium (29). This potential also influences the ionization equilibrium of the NW surface, causing the effective pH on the inner side of the bilayer to be lower than on the outer side. This situation is routinely observed in biological systems where Donnan equilibrium causes the pH inside negatively charged lipid vesicles to be lower than pH of the outside solution (30).

We can check the consistency of this interpretation by using it to estimate the potential on the membrane and compare it with the measured response of the NW transistor. The Donnan potential on the membrane is given by the Nernst equation:

where F is a Faraday constant, and [Cl]_i(o) denotes the concentration of mobile chloride anions on the inner (outer) surface of the membrane respectively, and [Cl]_i is given by:

where [M] is the effective concentration of the immobile anions on the inner side of the membrane, and C₀ is the concentration of the background electrolyte on the outside of the membrane (see SI Appendix for the full derivation of Eq. 2).

We can estimate the concentration of immobile ions, [M], by using the calibration data on the NW pH response (31). Conductance change values measured for proton transport through ALM pores (Fig. 3B) correspond to the NW surface charge density increase of 0.2 electron/nm². If we assume that average thickness of the water layer between the NW surface and the lipid bilayer (32) is ≈1 nm, then this value corresponds to an effective concentration of immobile ions of 350 mM. For the 100-mM background electrolyte concentration used in our experiments, Eqs. 1 and 2 estimate the membrane potential as −34 mV.

If we use the IV characteristic of the NW transistor (Fig. 1D) to convert the measured conductance shifts into the effective gate voltage shifts, then we can estimate that the presence of the membrane with ALM pores changes the effective gate voltage shift by −33 mV, the value that corresponds extremely well with our estimate of the Donnan membrane potential.

Device architecture that comprises a NW device coated with a lipid bilayer incorporating functional membrane proteins could serve as a versatile platform for building bionano interfaces that could enable direct conversion of biological signals into electronic impulses. Although we have used SiNW devices and proton transport as a proof-of-concept example of such a platform, it is possible to use other nanomaterials and transport of other species to create different ion-sensitive designs. The lipid bilayers provide a matrix for a virtually unlimited number of transmembrane proteins that can provide different functionalities. This biomimetic device platform could thus enable new applications in biosensing, bioelectronics, neuroscience, and medicine.

Methods

NW Synthesis.

Crystalline SiNWs were grown using the vapor liquid solid synthesis with gold colloidal particles as catalysts. Gold nanoparticles (20–30-nm; Ted Pella, Inc.) were dispersed onto a silicon wafer with 250-nm-thick oxide layer. The oxide layer was incubated in 0.1% aqueous solution of Poly-l-lysine (Ted Pella, Inc.) to promote adhesion of gold nanoparticles to the surface. After deposition of the gold nanoparticles, the substrate was rinsed with deionized (DI) water, dried with nitrogen, and cleaned in oxygen plasma to get clean catalyst surfaces. The NWs were grown at temperatures of 420–460 °C at 100 Torr. Boron-doped SiNWs were grown by flowing 31 standard cubic cm per min (sccm) of 10% silane in helium (Voltaix) and 3 sccm of 100 ppm diborane in helium (Voltaix) for 30 min.

Device Fabrication and Characterization.

After NW synthesis, the grown NWs were dispersed in ethanol by sonication for 5 s. The wires were then deposited onto silicon wafers with a 250-nm-thick dry-oxide layer grown at 1,200 °C. Poly-l-lysine functionalization was used to promote adhesion of the NWs to the substrate. The NW solution was flown onto the functionalized substrate in PDMS microfluidic channels on the desired areas of the substrate. After an oxygen plasma cleaning step, the deposited wires were annealed at 200 °C for 10 min to enhance adhesion. Metal contact regions were defined by photolithography, followed by e-beam deposition of Ti (10 nm)/Pt (70 nm) or Ni (80 nm). The contact regions were cleaned with oxygen plasma (100 sccm, 30 W), and the native oxide was etched away in these regions with a 5:1 buffered oxide etch for 10 s. An 80-nm conformal layer of stoichiometric silicon nitride was deposited onto the metal contacts by plasma enhanced CVD at 100 °C. After liftoff, the devices were further annealed in forming gas at 450 °C for 3 min to ensure good SiNW/metal contacts. The transistor performance in solution was tested in a DI water environment, with the solution gate voltage swept between −0.5 and 0.5 V by a leakage-free 3 M KCl Ag/AgCl reference microelectrode (Warner Instruments). All dc device measurements were done with a Keithley 2602 digital sourcemeter, and ac measurements were performed with a Stanford Research SR8650 lock-in amplifier at frequencies of 19 or 23 Hz.

Lipid Bilayer Formation and Characterization.

Unilamellar vesicles of DOPC lipid were prepared by sonication of a solution of 2 mg/mL DOPC (Avanti Polar Lipids) in buffer. Vesicle fusion onto SiNWs was allowed for 24 h. The thickness of resulting bilayers was measured to be 4.4 nm on flat oxide surfaces by surface plasmon resonance (SPR). Home-built systems were used to characterize the lipid bilayers by fluorescence recovery after photobleaching, cyclic voltammetry, and SPR.

Ion Channel Incorporation in Lipid Bilayers.

Gramicidin A (2 mg/mL; Biochemika; >90% HPLC) in 200-proof ethanol was mixed with DOPC in chloroform followed by solvent evaporation with N₂ and buffer hydration. Unilamellar vesicles of DOPC/gramicidin mol ratio of 100:1 were formed by sonication. These vesicles were then fused onto SiNW FET devices. ALM (Sigma; >90% HPLC) was incorporated into bilayers by spontaneous insertion by exposing pristine NW-bilayer structures to a solution of 5 mg/mL ALM for 30 min.

Electrical Measurements.

Rapid exchange of fluids around the NW was achieved by bonding PDMS microfluidic channels on top of the device chip and by using suction from suitable fluid reservoirs by a syringe pump. Different pH solutions were prepared in 100 mM KCl and 5 mM PBS by the addition of NaOH or HCL to adjust the pH. All pH values were measured by a calibrated benchtop pH meter. All reagents were purchased from Sigma–Aldrich with the highest purity available.

Other Information.

For confocal fluorescence microscopy images and FRAP data for suspended and surface bound SiNWs, SPR studies of lipid bilayer formation, and derivation of Eqs. 1 and 2 see SI Appendix.