Biological-to-electronic interface with pores of ATP synthase subunit C in
silicon nitride barrier
J. E. M. McGeoch I M . W . McGeoch 2 D . J . D . Carter 3 R. F. Shuman 4 G. Guidott i 1
1Department of Molecular and Cellular Biology, Harvard University, Cambridge, Massachusetts, USA 2pLEX LLC, Brookline, Massachusetts, USA
3Research Laboratory of Electronics, Massachusetts Institute of Technology, Cambridge, Massachusetts, USA
4Micrion, Peabody, Massachusetts, USA
Abstract--An oscil lator pore is identif ied that generates intermittent, large amplitude, ionic current in the plasma membrane. The pore is thought to be composed of 10-12 units of subunit c of ATP synthase. Pore opening and dosing is a co-operative process, dependent on the release, or binding, of as many as six calcium ions. This mechanism suggests a more genera/method of co-operative threshold detection of chemical agents via protein modification, the output being directly ampl i f ied in a circuit. Here the authors describe steps in the development of a sensor of chemical agents. The subunit c pore in a l ip id bi layer spans a nanometer-scale hole in a sil icon nitride barrier. Either side of the barrier are electrolyte solutions and current through the pore is ampl i f ied by circuitry. The techniques of laser ablation, electron beam l i thography and ion beam mi l l ing are used to make successively smaller holes to carry the l ipid patch. Holes of diameter as smal l as 20 nm are engineered in a sil icon nitride barrier and protein activity in l ip id membranes spanning holes as smal l as 30nm in diameter is measured. The signal-to-noise ratio of the ionic current is improved by various measures that reduce the effective capacitance of the barrier. Some l imits to scale reduction are discussed.
Keywords--A TP synthase subunit C, Sil icon nitride barrier, Ion channel, Biological-to- electronic interface
Med. Biol. Eng. Comput., 2000, 38, 113-119
J
1 Introduction
ENCOMPASSED IN the cell membrane, ion channels 'sense' their ligands at nano- to micro-molar concentrations in less than a second, the essence of the sensing being a change in ion current on ligand interaction. Vital physiological functions such as nerve conduction and movement are controlled by their activity and this has led to the development of biological weapons that exploit ligand-ion channel interactions, in principle, highly sensitive, fast response sensors can be made by making an ion channel function in electronic hardware, allowing detection of a ligand before it reaches concentrations harmful to man. Most ion channels are unsuitable candidates for electronic sensors because they will not function for hours out of the intact cell at environmental temperature, and their measurable signal is low. Here we record the experimental techniques that we employed to make a rugged ion channel pore, subunit c of ATP synthase (JUNGE et al., 1997; MCGEOCH and GUIDOTTI, 1997) function in a silicon nitride barrier that could be incorporated into a sensor
Correspondence should be addressed to Dr J. E. M. McGeoch; emaih [email protected] First received 9 August 1999 and in final form 9 November 1999
Medical & Biological Engineering & Computing 2000, Vol. 38
device. At 70mV the pore produces a large 100pAmp current that oscillates, allowing changes in the oscillation frequency and amplitude to constitute a clear sensor signal. Where ruggedness is not required, any ion channel may be placed in the sensor device to make a screening module for new drugs.
Subunit c of ATP synthase assembles into pores that can generate an oscillating current of frequency between 0.1 and 700Hz (MCGEOCH and GUIDOTTI, 1997) with a high conduc- tance for cations, in the range of 1 nS. The 7.6kD molecule is also tough and durable, withstanding room temperature in an environment of organic solvents for months without degrading. The subunit c protein is present in the ATP synthase complex in mitochondria (JUNGE et al., 1997) and in the neural plasma membrane (MCGEOCH and GUIDOTTI, 1997; MCGEOCH and PALMER, 1999). In both locations and in all mammals investi- gated (human, cow, sheep, dog, mouse), it has the same 75 amino acid sequence (DYER and WALKER, 1993). in the mitochondrion, the transmembrane part of ATP synthase includes a pore multimer of subunit c that contains possibly 12 of the subtmits (ELSTON et al., 1998). The number of subtmits present in re-constituted pores is not yet known, but is expected to be in the same range, based on the high conductance. There is a primary regulation of the subunit c pore by calcium (MCGEOCH and PALMER, 1999; MCGEOCH and MCGEOCH,
113
1994). its conductance is also modulated at a very sensitive level by polyamines such as nicotine. Based on its known sensing properties, subunit c appears to have the potential to be altered via protein engineering to sense a wider group of compounds.
ion channels have been routinely assayed by patch clamping since the technique was pioneered by SAKMANN and NEHER (1983). As an apparatus, the patch-clamp glass pipette (Fig. 1) has excellent properties for the measurement of the pico-ampere currents of ion channels that have been re-constituted in lipid membranes, its micron-sized aperture is drawn by fusion and stretching of a capillary in a swift, low-technology operation, and the glass barrier material has very good insulating proper- ties. The geometry enhances the impedance of the measurement and reduces capacitance through minimisation of the area of the thin glass barrier between the electrolyte volumes. However, in order to incorporate pores of the subunit c ofATP synthase in a small sensing device, we have explored more compact alter- native methods of mounting the ion channel protein. A successful device has to meet the following criteria:
(i) It must be simple and inexpensive to manufacture. (ii) The lipid membrane must be stable in the presence of
shock and vibration. (iii) The device must be compact, preferably with the pore-
containing barrier on the same substrate as an integrated circuit amplifier.
(iv) A deformable chamber material must be selected for the enclosure of the sensor device to ensure tight coupling between the barrier holding the pore and the electrolyte solutions.
(v) The enclosure design must provide air access to one electrolyte solution and have provision for a separate current amplifier component in the case of non-silicon- based barriers.
We focused on the use of sub-micron holes in planar barriers, testing several barrier materials and methods of forming holes. In view of item (iii) above we emphasised the use of silicon, modified to the oxide or nitride where insulation was needed.
2 Design of interface
In this section, we list some of the design considerations. The lipid bilayer has a thickness of approximately 5 nm, and may span a hole of much greater diameter. A conventional tip-dip artificial lipid patch has a diameter of 1-2 gm and is relatively fragile (Fig. 1 ). The pressure differential that will rupture a patch of this diameter is of the order of one tenth of one atmosphere (below). Smaller hole diameters support a greater pressure differential, producing a more rugged device, but the question arises as to the practical minimum hole size for protein function. One limit is provided by the elasticity of the surrounding lipid.
2.1 Limit due to lipid compressibili O,
The free subunit c pore itself, if it consists of a ring of twelve subunits by analogy with its expected structure within the ATP synthase complex, has an outer diameter of approximately 6 nm (JUNGE et al., 1997; ELSTON et al., 1998). The pore conductance is large (MCGEOCH and GUIDOTTI, 1997) and, based on the size of central channel that must exist to provide this conductance, the pore outer diameter could vary between 6 nm (closed) and 7 nm (open). As the walls of the hole are relatively rigid (Fig. 2), a certain radial extent of lipid is required around the pore to elastically accommodate its movement upon opening and closing. This extent can be estimated from the energetics of pore opening and the compressibility of lipid, if it is assumed that the lipid remains planar. Less energy would be required to push aside the lipid if an out-of-plane motion also took place, but the movement would be slower and would involve dissipative fluid motion of the two electrolytes. Based on the elastic properties of lipid (EVANS and NEEDHAM, 1987; NEEDHAM and NUNN, 1990) and the likely opening energy of a few tenths of an electron volt, the minimum membrane diameter for energetically allowed pore opening subject to the planar constraint appears to lie in the range 10-30nm, as detailed in the following paragraph.
electrolyte
Fig. 1
10 cm
Qlass //I// \ pipette /[H/ signal e
'patch' ~ f ~ ; ~
I # i i i #
I amplifier ~ electrolyte / head J
hal [ electrode
ground protein electrode pore
lipid membrane
Classic patch clamp apparatus for assaying current through the subtmit c pore. There is a giga-ohm seal between the glass pipette wall and the lipid bilavel; allowing the measurement o f pAmp cltr/'e/tt throltgh the pore. A cltr/'e/tt o f l OOpAmps is obtained through one pore when a voltage o f TO m V is applied to the pipette
sub-lOO-nrn hole protein pore /
I i:i,r n Ura:n ' l a
electrolyte
~J
rigid I ~ wall
. . . . ion ~ _ ~ j pore
Fig. 2 General concept o f a protein pore mounted in a nanosca[e- diameter lipid membrane (a) and geometo,' for calculation of the effect o f lipid compressibility (b)
114 Medical & Biological Engineering & Computing 2000, Vol. 38
With reference to Fig. 2, in which r~ is the outer radius of the closed pore (complex), situated anywhere within a hole of radius a, 6 is the outer radial expansion of the pore on opening, W is the flee chemical energy driving the opening transition, and K is the lipid's elastic areal expansion modulus, we can show that, for the channel to be able to open, the hole radius must be greater than
/
ami n = rc~/1 + 2~K62/W, the radius for which the flee energy
is equal to the energy stored in compression of the lipid. From work by other authors (EVANS and NEEDHAM, 1987; NEEDHAM and NUNN, 1990), K may vary from 100 to 2000 dyne/cm. The free energy of opening is not known for the subunit c pore, but for the sake of discussion an energy substantially in excess of the thermal energy can be assumed. Although pore opening begins with the unbinding of probably six calcium ions, each held with a binding energy of 0.5 eV (MCGEOCH and PALMER, 1999), the radial movement of the subunits is probably driven by electro- static repulsion of the negatively charged regions that are formed as calcium leaves. Most of the available energy is used to effect a conformational change, leaving a net free energy, W, that will be definitely above thermal, but may only amount to a few tenths of an electron volt. For K = 1000; W = 0.3 eV; r~ = 2.8 nm; 6 = 0.5nm (above), we find ami,, = 16nm, or a minimum hole diameter of 32 nm.
2.2 Sztstainable pressure differential
Theoretically, the maximum sustainable pressure differential is given by AP = 2TL~s/a, where a is the hole radius and TLV s is the linear rupture strength of the lipid. For example, a bilayer of pure stearoyloleoylphosphatidylcholine (SOPC) has TL,LS = 5.7dyne/cm (NEEDHAM and NUNN, 1990), rising to 30 dyne/cm in a membrane with a mixture of 37% SOPC + 63% cholesterol. A hole of diameter 100 nm spanned by pure SOPC
6 will support a maximum differential of 2.3 × 10 dyne/cm-, or 2.3 atmospheres. In the case of the mixture, the maximum differential is 12 atmospheres. In practice, the lipid has to adhere to the wall of the hole more energetically than to itself, at an arbitrary angle of intersection, for this theoretical result to apply. The strong adherence of lipid to glass electrodes has been documented (COREY and STEVENS, 1983). In the case of the silicon nitride barriers discussed below, which are of silicon-rich composition, there are many disordered 'dangling' silicon bonds (MAKINO, 1983), which present to the lipid a surface that has chemical similarity to that of glass. The apparently good adhesion of lipid to silicon-rich silicon nitride in the present experiment is consistent with this explanation.
2.3 Access impedance
We also have to consider the access impedance for ionic current as it traverses the bore of the hole in order to reach the lipid membrane. Too high an access impedance will decrease the measured pore current and in principle could contribute to measurement noise. As an example, for 30 nm diameter holes and typical hole lengths of 0.25 to 1 ~tm, the impedance of the bore of the hole, with a typical electrolyte of resistivity 100 ~cm, ranges from 0.35 to 1.4G~, becoming comparable to that presented by the pore itself (which conducts 100pA at 100mV). These access resistances, although high, do not contribute significant measurement noise (WONDERLIN et al., 1990) because the lipid membrane itself is very small in area and has a correspondingly small capacitance (at 0.8~tF/cm2). However, the barrier capacitance can be considerable, and in conjunction with the voltage noise from the amplifier, it contributes the dominant current noise in this type of design, as discussed below.
2.4 Capture probability
A last factor to be considered is the probability of capturing a protein pore into this size of hole. This is discussed below in relation to the results.
3 Materials and methods
In this section we consider the materials for the barrier, the methods to make nanometer holes in the barrier, the chamber for barrier tests, and the enclosure for the whole sensor.
3.1 Barrier material
The barrier material has to be a very good insulator that also adheres to membrane lipids. Traditional patch clamping uses borosilicate glass, and we successfully employed glass in laser- drilled barriers. We found that (silicon rich) silicon nitride also adhered well to lipid, which allowed entry into the large repertoire of silicon processing techniques, including etching of small holes using lithography. Crystalline silicon was laser drilled and used to test pore function, but did not have suffi- ciently high resistivity. Polyvinylidene fluoride (PVDF) was used in laser hole-drilling tests and also patched well. Other polymers will be tested in future work, in which ion beam milling of the polymer will be explored.
3.2 Methods for making holes in barrier
With the goal of manufacturing sub-micron holes, possibly as small as 30 nm in diameter, a number of techniques were tested, including laser drilling, electron beam lithography and ion beam milling, in the first of these techniques, material is removed by melting and evapouration, in the lithographic process, removal is by controlled chemical etch. The ion beam technique removes material by sputtering (McCRACKEN, 1975), whereby each incident ion deposits sufficient energy to dislodge several tens of surface atoms in a very localised evapouration event. The results from these will be discussed.
3.2.1 Laser-drilling: Holes of minimum diameter 1 ~tm were made in 160-~tm plane glass cover slip substrates using a focused, pulsed KrF laser (248.5 nm)*. A high-quality laser beam and f/10 optics allowed a focal spot of about 3 ~tm diameter, so that holes of diameter much smaller than this would not have been anticipated. However, optical guiding occurred via glancing reflection as the hole was drilled (by pulsed laser ablation) so that the final holes typically had an entrance diameter of 5-10 ~tm and an exit diameter of 1 ~tm. Using the same technique, 1-2-~tm exit holes were formed using a 400-~tm-thick silicon substrate, and similarly sized holes were obtained through a 40-~tm-thick PVDF membrane. Typical entrance and exit apertures for the three materials are shown in Fig. 3. There is considerable roughness around the aperture that results from local melting and fracture as the laser-produced shock wave breaks through.
3.2.2 Electron beam lithography: Smaller holes, down to 50nm in diameter, were made through 0.25-1.1 ~tm silicon nitride membranes using standard techniques of electron beam patterning. 400-~tm-thick silicon wafers with a <100> orienta- tion were coated with low-stress silicon-rich silicon nitride by low pressure chemical vapour deposition (LPCVD) (MAKrNO,
*Laser hole drilling performed by JPSA Associates, Nashua, NH
Medical & Biological Engineering & Computing 2000, Vol. 38 115
glass
silicon
PVDF
31.8 nm
entrance exit
r/~.: ~'-, ~ j ? ~ ~-~ " ~ ......
.... i
Fig. 3 Scamling electron micrographs o f entrance and exit holes" drilled bv laser
1983; SEKIMOTO et al., 1982). Film thicknesses of 0.25 and 1 gm were deposited. Polymethylmethacrylate (PMMA) was spun on the front side of the wafers to be patterned with e- beam lithography. This PMMA was exposed at 50 keV with a VS-2A e-beam writer, forming an array of exposed spots. The PMMA was developed in 1:2 methyl isobutyl ketone:iso- propyl alchohol (MIBK : IPA) to form an array of holes in the resist. Reactive ion etching (RIE) in CF 4 transferred the holes in the resist to the front-side nitride. The resist was then stripped.
Free-standing silicon nitride membranes containing holes were then formed by patterning the reverse side of the wafer, followed by etch processes. Photoresist was spun and patterned on the backside of the wafer to form windows for the membrane etch and channels for die separation. The photoresist pattern was transferred into the backside nitride using RIE in CF 4. The wafer was then etched in KOH, which etches anisotropically with crystalline direction. The end result was a nanoscale hole within a free-standing silicon nitride membrane typically 4 mm x 4 mm, situated centrally within a 12mm x 12mm silicon frame. A scanning electron micrograph of one of the smaller holes made by this technique is shown in Fig. 4a.
3.2.3 Ion beam mill ing o f holes: ion beam milling can remove material by sputtering to a precision of approximately 5 nm (STEWART and CASEY, 1997). The depth-to-diameter ratio of cylindrical holes is limited to approximately 5. Free-standing 4 mm x 4 mm silicon nitride membranes of thickness 1.1 gm within 12 mm x 12 mm frames were mounted within a focused ion beam machine (Micrion). Gallium ions at an energy of 50keV were focused to a spot of diameter less than 20 nm. Penetration of the nitride occurred in the deposition range exceeding 7.7nC/gm 2. To penetrate the 1.1 gm membranes, oblique milling was employed. When penetration occurred, the diameter of the entrance hole was approximately 250 nm.
116
Fig. 4 Electron-beam lithographic hole image with 250-rim silicon nitride barrier (a) and ion-milled exit hole image with 1. l %tm silicon nitride barrier (b)
Atomic force and scanning electron microscopy showed that holes of exit diameter as small as 31 nm could be achieved (Fig. 4b). Further milling led to a range of exit diameters between 31 nm and 100 nm.
3.3 Test chamber f o r barrier funct ion
A test cell was designed for the characterisation of protein pores in silicon barriers. The barriers had a standard outer dimension of 12mm×12mm, for ease of handling. Electrolyte solutions were held by surface tension within central wells in the upper and lower components of the cell (Fig. 5a). Chlorided silver electrodes contacted each of the solutions. These electrodes were in turn contacted to the signal and ground leads of a List EPC7 patch clamp amplifier. The upper electrode was contacted gently by a spring-loaded pin. Upon assembly the solutions were held by surface tension against the barrier as shown in Fig. 5b. The patch clamp amplifier was programmed to apply a voltage between the electrolytes, and the resulting current through the hole, or protein pore, was monitored. The test cell was positioned on a temperature stage under a binocular microscope and was main- tained at 25 =C during most tests.
Medical & Biological Engineer ing & C o m p u t i n g 2000, Vol. 38
side view of test cell prior silver electrode stud to assembly /
silicon die (12 mm x 12 m m ) ~ ~ e l e c t r o l y t e droplet
lO0-nr~ hol'e j > " ~" ' silicon nitride membrane
silver wire electode 'bottom' electrolyte droplet a
detail of contact region after " "
Fig. 5 [l[ttstratio/t of teflon test cell for the measurement o f pore current in various barrier configurations (a) and enlarged view o f electrolyte geometo' in contact with barrier (b)
3.4 Test conditions
The electrolyte solutions were: top (mM) 50 NaC1, 1.1 EGTA, 7+
1 CaC12 (free [Ca- ] = 200 nM), 5 Hepes, pH 7.6; bottom (mM) 10 NaC1, 15 MgC12, 1.3 EGTA, 1 CaC12 (free [Ca 2+] = 300 nM), 5 Hepes, pH 7.6. A mixture of lipid and protein (subunit c of ATP synthase) was prepared as described (MCGEOCH and GUIDOTTI, 1997). Small quantities of this mixture were added to the bottom electrolyte droplet before assembly and several minutes were allowed to elapse while the lipid formed a visible surface film. The silicon die was gently lowered by forceps onto the bottom solution droplet. The top section of the test cell was then lowered gently until it rested on the base. Approximately 50% of assembly attempts resulted in a high-impedance seal across the hole in the membrane, indicating that a lipid membrane now spanned the hole. Voltages of either polarity up to 200 mV were applied across the lipid membrane in order to activate and measure the conductance of the protein pore.
4 Test resul ts
4.1 Barrier capaci tance as the dominant source o f current noise
The two techniques for the generation of sub-100-nm holes both become difficult for initial barrier thicknesses much greater than 1 gm. Such thin membranes are quite capacitative and this capacitance can contribute excessive current noise unless the
active area of the barrier is restricted. For example, a 1-gm silicon nitride barrier was initially used in the test cell described above with an electrolyte contact area of 0.1 cm 2, giving a capacitance of 660 pF and a root-mean-square noise current of 6pA at 3 kHz bandwidth. The current noise of our List EPC7 amplifier scaled as the square of the measurement bandwidth (in accordance with WONDERLIN et al., 1990) obeying i , , s ( t ' )= 909CBf 2, where C B was the barrier capacitance (farads) a n d f was the amplifier bandwidth setting. Reduced barrier capaci- tance was achieved in several ways.
4.1.1 Reduced barrier area: Our first recourse was to reduce the diameter of the electrolyte wells, but greater precision was then required in the initial filling of the well, to prevent spreading of the electrolyte on assembly. An additional complication was the reduction in surface tension due to the presence of the lipid, with greater spreading of the electrolyte than anticipated.
4.1.2 hwreased barrier thickness: After working with 250 nm-thick silicon nitride membranes, the capacitance was decreased by a factor of four through a change to a 1 ~tm thickness of membrane, ion beam milling technique was improved to allow holes as small as 30 nm in diameter to be drilled through the thicker membrane.
4.1.3 Reduced silicon nitride area: As noted, the silicon nitride membrane dimensions were 4 m m x 4 m m , and the electrolyte contact area was a small fraction of this. Experi- ments with a smaller ( 1.5 mm x 1.5 mm) silicon nitride mem- brane did not reduce the noise as hoped, due to contact of the electrolyte with the angled silicon walls adjacent to the nitride. The doped silicon conducted sufficiently well to electrically contact the top electrolyte with a relatively large area of the silicon nitride membrane (Fig. 5b), maintaining a high capa- citance.
4.1.4 hTsulation: The nitride barrier was insulated with a 2- ~tm-thick low-dielectric constant insulating layer'~ which was spun onto the plane side of the die (in contrast to the generic top-side insulation shown in Fig. 5b). it was not possible to spin such a layer onto the upper side because of the angled silicon walls. Using an ion beam, a 10 ~tm x 10 ~tm area of the insulator was excavated to reveal nitride, and ion beam milling was used as before to create nanoscale holes, in practice, the cleaning solvent acetone degraded the resist layer, preventing re-use of the resist-insulated barrier.
A more permanent silicon oxide insulation layer was used in a further experiment. This was fabricated via the series of steps described in Fig. 6. The angled <111> silicon walls were oxidised in high-temperature steam to a depth of 1.5 ~tm. At the same time, the excess silicon in the silicon nitride membrane was assumed to have oxidized throughout its 1.0 ~tm thickness. This barrier design had much reduced capacity and low noise even with electrolyte completely filling the central well. Two membrane dimensions were fabricated, 400 ~tm square and 700 ~tm square. For the case of the 400-~tm-square membrane, the estimated contribution of membrane capacity was 9pF, whereas the capacity across the barrier via electrolyte contact with the angled walls was estimated to be 44 pF. The root-mean- square current noise in a 6 kHz bandwidth estimated from Section 4.1 above was 2 pA, and the measured root-mean- square noise was 3 pA. The corresponding figures for the 700 ~tm square membrane were 26pF and 61 pF, leading to an
-~Shipley 1800 series resist
Medical & Biological Engineering & Computing 2000, Vol. 38 117
silicon die sub-100-nm hole
400 i~m or 700/am
\
a
12 mm
|1
(ii)
membrane
above formula were 30 nm and 52 nm, in rough agreement with the SEM image of an exit hole drilled under the same conditions, shown in Fig. 4b.
4.3 Ion channel observations
In ion channel observations under standard test conditions (above), the smallest exit holes occasionally patched to produce essentially no current, and did not always show ion channel events. More frequently, typical conductance events were observed (approximately 0.25-1nS, characteristic of the subunit c pore), but quasi-regular current oscillations were not observed with exit holes less than 50 nm in diameter. Typical current traces from holes of exit diameter nominally 100 nm and 50 nm are shown in Fig. 7. The conductances of 0.44 nS and 0.18 nS respectively are at the lower end of the 0.25-7 nS range observed in a classical patch clamp experiment (MCGEOCH and GUDOTTI, 1997). Note that the (root-mean-square) noise current is approximately 6pA (bandwidth 3 kHz), generated by the capacitance of the relatively large electrolyte contact area in the un-insulated design used in these measurements. With the resist-insulated design a root-mean-square noise of 2 pA was measured at the same bandwidth, while the oxide-insulated design (Fig. 6) had a noise of 3 pA at double the bandwidth (6 kHz).
(iv)
(v)
oxide
Fig. 6
(vi)
Stages in the fabrication o f an oxide-insulated silicon barriel: (a) Plait o f the barriel; containing either a 400 ~tm x 400 ~tm or 700 ~tm x 700 ~tm silicon nitride membrane. Not drawn to scale. (b) Steps in barrier fabrication: (i) Start with a 500- ~tm-thick Si wafer coated with 1.0%tm SiN~ (grown by LPCVD): 60 Spin photoresist and pattern for the membrane and scribe line formation: 6ii) Reactive ion etch (RIE) the silicon nitride in a CF 4 plasma: 6v) Wet etch the silicon in KOH, exposing <111> clTstal planes and also removing the photoresist: (v) Oxidise to a depth o f l .5 ~tm on the <111> planes. This also oxidises the silicon-rich silicon nitride: (vi) Cleave into die and pelform ion beam milling to generate a sub-lOO-nm hole
estimated noise of 3 pA, compared to a measured noise of 4 pA. Electrolyte contact with the walls of the well makes possible a much more compact device.
5 Discussion We have begun the development of a sensor in which a robust,
high-conductance ion pore (subunit c ofATP synthase) is placed in a lipid membrane spanning a nanoscale hole in an insulating barrier. To improve the ruggedness of the assembly, the area of the lipid bilayer was reduced as much as possible.
In 0.25-1 ~tm silicon-rich silicon nitride barriers, we fabri- cated single holes of exit diameter down to 31 nm and obtained signals from individual pores in lipid bilayers spanning the holes. The background noise on these signals was higher than the noise in a classical glass electrode experiment, due to the higher capacitance of the barrier. Barriers thicker than 1 ~tm, which would have had lower capacitance, could not be used because of either undercutting during the lithographic etch stage, or the difficulty of removal of sputtered atoms from the hole
100 p A ~
100 ms
(expandeidl ~ t ~ ~ , k , ~ ] ~ i ~ d ~
20 ms a
4.2 Conductance measurements on holes without lipid
The conductance of ion-milled holes of nominal 30 nm exit diameter was measured. As the holes were approximately conical, the resistance of a hole was given by R : p L / z r lr 2, where p is the resistivity of the electrolyte, L the length of the hole, and r 1, r 2 the radii at each end. Using 50 mM NaC1 solution of resistivity 187f~cm (25C), resistances of 172Mf~ and 100 Mf~ were measured for two of the channels with the smallest exit holes. Considering the measured entrance diameter of 250nm in each case, the exit diameters calculated from the
10 ms b
Fig. 7 Current waveforms, sodium ion current through a fluctuating subtmit c pore. The current zero coincides with the horizontal time bal: (a) l O0-nm-diameter hole in silicon nitride, voltage 198 m E (b) nominal 50-rim-diameter hole, voltage 180 m V
118 Medical & Biological Engineering & Computing 2000, Vol. 38
during ion beam milling. The capacitance was therefore mini- mised by using 1 pm barriers, restricting the electrolyte contact area, and either partially covering the nitride and surrounding silicon with an insulating material or modifying the bare silicon surfaces to oxide.
Pore conductances observed with the smaller holes were typically at the lower end o f the range of previous observations (MCGEOCH and GUIDOTTI, 1997). i t is possible that variable numbers o f subtmits can join to form the pore, and statistically, when there is so little protein in the sub-100-nm patch, the largest pore contains less subtmits than in the case o f a larger patch.
The occasional absence o f channel electrical activity was probably the result o f using patches so small that no intact pores were present in them. in our standard re-constituted preparation from bovine brain (MCGEOCH and GUIDOTTI, 1997), there are typically between 20 and 100 functional pores in a 0.5 p m - l - p m - d i a m e t e r classical patch. Scaling down to 30 - 50-nm diameter, a factor of roughly 100-1000 in area, the occasional absence o f electrical activity in a patch is consistent with a lack o f any pores within the patch.
in conclusion, some o f the challenges in the development of a simple, rugged, protein-to-electronics interface have been addressed. Future work will explore the containment o f the electrolyte solutions on a silicon barrier, and also the potential use o f a low-capacitance polymer barrier in a manufacmrable configuration with integrated electrolyte containment.
Ackllowledgmell~This research was supported by the US Gov- ernment under a grant from the the Office of Naval Research and the Defence Advanced Research Projects Agency, Information Technology Office, N00014-97-1-0653. We wish to thank J. P. Sercel of JPSA for help with laser-drilling; J. Carter and M. K. Mondol of MIT for help with lithography; Micrion for focused ion beam milling, J. A. Golovchenko of Harvard Physics Dept and H. I. Smith of the Research Laboratory of Electronics of M.I.T. for discussions.
References
COREY, D. P., and STEVENS, C. F.: 'Science and technology of patch- recording electrodes' in SACKMANN, B. and NEHER, E. (1983): 'Single channel recording' (Plenum Publishing Corp., New York), pp. 53-68
DYER, M. R., and WALKER, J. E. (1993): 'Sequences of members of the human gene family for the subunit c of mitochondrial ATP synthase', Biochem. J., 293, pp. 51-64
ELSTON, T., WANG, H., and OSTER, G. (1998): 'Energy trax~sduction in ATP synthase', Nature, 391, pp. 510-513
EVANS, E., and NEEDHAM, D. (1987): 'Physical properties of surfac- taxlt bilayer membranes: thermal transitions, elasticity, rigidity, cohesion and colloidal interactions', J. Phys. Chem., 91, pp. 4219-4228
JUNGE, W., LILL, H., and ENGELBRECHT, S. (1997): 'ATP synthase: an electrochemical transducer with rotary mechanics', Trellds Bio- chem. Sci., 22, pp. 420-423
MAKINO, T. (1983): 'Composition and structure control by source gas ratio in LPCVD SiNx', J Electrochem. Soc., 130, pp. 450-455
MCCRACKEN, G. M. (1975): 'The behaviour of surfaces under ion bombaxdment', Rept. Prog. Phys., 38, pp. 241-327
MCGEOCH, M. W., and MCGEOCH, J. E. M. (1994): 'Power spectra and cooperativity of a calcium-regulated cation channel', Bioph).'s. J., 66, pp. 161-168
MCGEOCH, J. E. M., and GUIDOTTI, G. (1997): 'A 0.1-700 Hz current through a voltage-claxnped pore: candidate protein for initiator of neural oscillations', Braill Research, 766, pp. 188-194
MCGEOCH, J. E. M., and PALMER, D. N. (1999): 'Ion pores made of mitochondrial ATP synthase subunit c in the neuronal plasma membrane and Batten disease', Molec. Gelwt. Metab., 66, pp. 387-392
NEEDHAM, D., and NUNN, R. S. (1990): 'Elastic deformation and failure of lipid bilayer membranes containing cholesterol', Biophys. J., 58, pp. 997-1009
SAKMANN, B. and NEHER, E. (1983): 'Single channel recording' (Plenum Publishing Corp., New York)
SEKIMOTO, M., YOSHIHARA, H., and OHKUBO, T. (1982): 'Silicon nitride single-layer X-ray mask', J Vac. Sci. Tech., 21, pp. 1017- 1021
STEWART, D. K., and CASEY, J. D. (1997): 'Focused ion beams for micromachining and microchemistry', in Handbook of Microlitho- graphy, Micromachining, and Microfabrication, (S.RI.E. Optical Engineering Press, Vol. 2, Bellinghaxn, WA, USA.) pp. 153-195
WONDERLIN, W. E, FINKEL, A., and FRENCH, R. J., (1990): 'Optimiz- ing planar lipid bilayer single-channel recordings for high resolution with rapid voltage steps', Bioph).'s. J., 58, pp. 289-297
Author's biography
JULIE E. M. MCGEOCH received a BSc Honours Degree in Biochemistry from the University of Southampton in 1969 and PhD from the Univer- sity of Southampton Medical School in 1973. She has performed research at University College Medical School, London and at the University of Oxford. She is currently a research associate at Harvard University in Cambridge, Massachusetts. Her research interests have focused on membrane
proteins, notably the sodium potassium ATP ase and cation conducting ion channels. She is currently studying the 3D structure of the subunit C ion pore of the ATP synthase complex and fabricating a sensor by placing the pore in silicon- and polymer-based haxdwaxe.
Medical & Biological Engineering & Computing 2000, Vol. 38 119
