Electrodeposition and Bipolar Effects in Metallized Nanopores and

Jan 9, 2015 - Finally, we use metallized nanopores modified with homocysteine for the detection of insulin. We show that adsorption of the protein to ...
1 downloads 0 Views 1MB Size
Article pubs.acs.org/ac

Electrodeposition and Bipolar Effects in Metallized Nanopores and Their Use in the Detection of Insulin Agnieszka Rutkowska,† Kevin Freedman,†,‡ Justyna Skalkowska,†,§ Min Jun Kim,∥ Joshua B. Edel,† and Tim Albrecht*,† †

Department Department § Department ∥ Department ‡

of of of of

Chemistry, Imperial College London, South Kensington, SW7 2AZ, London, United Kingdom Chemical and Biological Engineering, Drexel University, Philadelphia, Pennsylvania 19104, United States Chemistry, Warsaw University of Technology, Noakowskiego 3, 00-664 Warsaw, Poland Mechanical Engineering and Mechanics, Drexel University, Philadelphia, Pennsylvania 19104, United States

S Supporting Information *

ABSTRACT: Solid-state nanopore devices with integrated electrodes are an important class of single-molecule biosensors, with potential applications in DNA, RNA, and protein detection and sequence analysis. Here we investigate solid-state nanopore sensors with an embedded gold film, fabricated using semiconductor processing techniques and focused ion beam milling. We characterize their geometric structure in three dimensions on the basis of experimental conductance studies and modeling as well as transmission electron microscopy imaging and tomography. We used electrodeposition to further shrink the pores to effective diameters below 10 nm and demonstrate how bipolar electrochemical coupling across the membrane can lead to significant contributions to the overall pore current and discuss its implications for nanopore sensing. Finally, we use metallized nanopores modified with homocysteine for the detection of insulin. We show that adsorption of the protein to the chemically modified nanopores slows down the translocation process to tens of milliseconds, which is orders of magnitude slower than expected for conventional electrophoretic transport.

S

transistors, with a gate electrode embedded inside the nanopore and well-insulated from the solution, the charge transfer impedance (Zct) at the electrode/solution interface, is large, formally infinitely large in the limit of a perfectly insulated gate. The surface charge distribution inside the channel, and the rectifying properties of the pore, may be controlled by the solution pH, ionic strength, and the applied gate voltage.25−29 On the other hand, if the embedded electrode is not perfectly insulated, then it can take part in electrochemical processes (finite Zct). It constitutes an additional current source or sink that can affect the electric field at the pore, induce apparent rectification effects, or be used for further modification of the pore surface.22,30 Depending on the potential applied to the embedded electrode, Zct may vary from very large to very small, offering a way to manipulate ion transport in the pore and the translocation process itself. Finally, the potential drop across the pore, due to the transmembrane bias Vbias, may lead to electrochemical activity at the embedded electrode, even if the latter is not connected to any external circuitry. Such effects are well-known in the field of bipolar electrochemistry and have

olid-state nanopores have been successfully employed as versatile sensing devices for detection of single analytes such as DNA,1,2 proteins,3 and gold nanoparticles.4 The major technological goal in the field has been the construction of fast and cheap nanopore-based next-generation DNA sequencers.5,6 To achieve this, a number of challenges, including (i) reduction of translocation velocity, (ii) increase of spatial resolution and sensor specificity, and (iii) incorporation of the nanopores into more complex electronic platforms, need to be addressed.7−10 At the fundamental level, this implies an in-depth understanding of the physical and chemical properties of the nanopore, e.g., local electrostatics,11 ion current rectification,12 the effect of chemical modification, and molecular interactions with the analyte molecule.21 In addition to “conventional”, chip-based nanopore sensors with insulating silicon dioxide,13 silicon nitride,14 alumina,15 or polymer16 membranes, more complex device geometries and structures have been used in this context recently, e.g., glass nanopipets and nanopore transistors.17−23 In high ionic strength solution, the current−voltage relation is typically symmetric, i.e., the nanopore displays Ohmic behavior. Nonlinear, rectifying behavior occurs at low ionic strength and may indicate an asymmetric geometry and charge distribution inside the pore, hydrodynamic effects, or additional current sources in the cell.24 For example, in nanopore © 2015 American Chemical Society

Received: November 6, 2014 Accepted: January 9, 2015 Published: January 9, 2015 2337

DOI: 10.1021/ac504463r Anal. Chem. 2015, 87, 2337−2344

Article

Analytical Chemistry been exploited for local, “contact-less” electrodeposition and other applications.31−36 By employing a three-electrode configuration (Figure 1A),22 we provide a detailed and comprehensive analysis of the

(Supporting Information). The inner diameter of the pore, D100 (100 nm below the top of the pore), is likely to be smaller than D0 due to the shape of beam and material build-up at the nanopore edges during the drilling process.38 We characterize this aspect by TEM cross-sectional analysis and pore conductance modeling based on the hourglass model, see below. We initially determined the total resistance of the cell, R0, from the slope of the current−voltage curve (1 M KCl aqueous solution; two Ag/AgCl electrodes). In principle, R0 is comprised of solution contributions to the resistance of the electrochemical cell (the access resistance Racc) as well as the resistance of the pore channel itself (Rpore). The models developed for calculating Racc are discussed in detail in the Supporting Information.39,40 For long cylindrical, uncharged nanopores with a high aspect ratio (diameter Dcyl ≪ L), Dcyl may be calculated from eq 1:24

Figure 1. (A) Schematic of the electrochemical setup with the nanopore device immersed in the gold electroplating ECF64 solution. RE/CE is the combined counter-/reference electrode (either Ag or Ag/AgCl, see main manuscript), and WE is the working electrode (either the metallized nanopore electrode or the second Ag or Ag/ AgCl electrode). (B) Close-up of the nanopore geometry. L is the nanopore length, and D is the nanopore diameter at L = 0 nm, 50 nm, 100 nm, 150 nm, and 200 nm.

Dcyl =

4ρL πR 0

(1)

where ρ is the solution resistivity and L is the total thickness of the membrane, which is 215 nm according to the fabrication design and 199 nm according to the TEM data. Thus, we use the average value of 207 nm throughout. Using this simple model, we obtained Dcyl values between 5 and 54 nm, Table S1 in the Supporting Information, which averages to 20 ± 15 nm. This is significantly smaller than the average D0 (Table S1 in the Supporting Information), suggesting that the pores are indeed narrower in the center of the membrane. Since the etch rate during the drilling process is expected to be materialdependent, we first wanted to check whether the composite nature of the membrane has a significant effect on the pore geometry. TEM tomography was used to form three-dimensional reconstructions of the pores and was analyzed by taking cross sections of the pore in the z-plane (in-plane to the membrane, Figure 2A−C) as well as the x-plane (normal to the membrane,

electroactivity and functionality of metallized nanopores. Specifically, we investigate electrodeposition into very small pores with diameters below 20 nm and characterize the pores before and after the deposition process, using transmission electron microscopy (TEM) and electrochemical techniques. TEM imaging confirmed that subsequent deposition steps result in a metal deposit inside the nanopore, thus decreasing its effective diameter. As expected, the process is less efficient for smaller nanopores (diameter < 10 nm) and the deposit is nonuniform, most likely due to mass transport effects. For solutions containing redox-active components, we indeed observe bipolar coupling between the top and bottom compartments, due to the effect of Vbias on the local potential of the Au electrode. The transmembrane ion current is enhanced and can even deviate from Ohmic behavior. To the best of our knowledge, this is the first time the effect is discussed in the context of nanopore sensing. Finally, the metallized nanopores are used for the detection of insulin to evaluate their sensing properties. The molecular weight of monomeric insulin is ∼5.7 kDa and its diameter is approximately 2 nm. At pH 7.4, the protein is hexameric, while at pH 1.6 (in hydrochloric acid) and pH 2 (acetic acid) it forms dimers and monomers, respectively.37 Under the latter conditions, we were unable to detect insulin with unmodified metallic or SiN-based nanopores, presumably due to the high speed of translocation and the small size of the protein. However, after modifying the Au surface with homocysteine (Hcy), we observed rectification effects in the I/V curves and were able to detect insulin transport across the pore. This observation was unexpected but can be rationalized based on the effect of Hcy immobilization on the local electrostatics of the nanopore and the adsorption behavior of the protein.

Figure 2. TEM cross-sectional images of the top (A, E), middle (B, F), and bottom sections (C, G) of solid-state nanopores before and after electrodeposition (milling time, 100 s; acceleration voltage, 30 kV; beam current, 1 pA). The cross-sectional slice thicknesses were 5 nm and taken from the following z-heights (measured from the bottom edge of the pore): 50 nm (bottom), 100 nm (middle), and 150 nm (top). Parts D and H show the side view of the respective nanopores. The scale bar is 50 nm.



RESULTS AND DISCUSSION Nanopore Fabrication and Geometry. Nanopores were milled with a focused ion beam (FIB) and the outer diameter D0 was determined from scanning electron microscopy (SEM) micrographs. The nanopore size was strongly dependent on the milling time, beam voltage, and current, cf., Table S1

Figure 2 D). The drilling time was 100 s in this case, see Table S2 in the Supporting Information for other fabrication conditions. Panels E−G and H show a device fabricated under the same conditions but after electrodeposition of gold into the pore channel, vide inf ra. The pore narrows from D0 = 133 nm at the top to D100 = 105 nm in the center, and then 2338

DOI: 10.1021/ac504463r Anal. Chem. 2015, 87, 2337−2344

Article

Analytical Chemistry

Figure 3. (A) CV of the gold electroplating solution ECF64 at concentration 1:10 (v/v) at the embedded Au WE vs Ag quasi-RE (scan rate, 100 mV/s). The arrow indicates the initial scan direction. (B) Transmembrane ion current in 1:10 (v/v) gold electroplating solution before electrodeposition (black trace) and after three subsequent electrodeposition steps. (C) Transmembrane ion current in 1 M KCl aqueous solution before the electrodeposition (black trace) and after a set of three subsequent electrodeposition steps, followed by cleaning of the nanopore devices in oxygen plasma. The transmembrane ion current measurements in panels B and C were performed with a floating Au electrode (cf. Device no. 3, Table S1 in the Supporting Information).

widens again to about D200 = 118 nm at the bottom; the channel is thus approximately hourglass shaped. Because of relatively large thickness of the devices, we were not able to resolve individual layers of the membrane, but based on the nanopore design, the 30 nm thick gold electrode is positioned ∼35 nm from the bottom of the nanopore (cf. Figure 1A and D150 in Table S2 in the Supporting Information). The channel walls appear to be relatively smooth with the top and bottom diameters larger than the middle section, similar to an hourglass geometry. In the literature, tomography has been performed on bare silicon nitride as well as on gold-coated silicon nitride (thickness, 50 nm in both cases),41,42 showing very different pore profiles than those reported here. The bare silicon nitride pores have parabolic pore profiles, very rounded and symmetric. In comparison, the thicker, multilayer membranes used here show less curvature.46,47 We then characterized the pore geometry in situ using an hourglass model for the resistance of the channel.43,44 On the basis of the measured total resistance R0, the outer diameter D0 (from SEM), and the length of the channel L (from TEM and chip design), we estimated the pore diameter in the center of the channel, D100; the narrowest part of the channel is also the most sensitive region for translocation detection. With an average D0 = 74 ± 22 nm, we obtain D100 = 14 ± 15 nm, which confirms that the pores are indeed significantly narrower inside the channel, broadly in line with our TEM results. The values for D100 also help to estimate the current blockage during translocation of an analyte, as we show below. Pore Shrinking by Electrochemical Deposition. Some of the “as drilled” nanopores were further modified using electrodeposition inside the pore channel. For this purpose, we used nanopores with measured resistances between 15 and 60 MΩ, which correspond to an effective Dcyl ranging from approximately 20 to 40 nm, according to eq 1. Figure 3A shows a typical cyclic voltammogram (CV) of the embedded Au electrode in aqueous ECF64 solution (1:10 v/v), prior to electrodeposition experiments (scan rate 100 mV/s; initial potential Ei = 0 V, negative sweep direction). For comparison, Figure S2 in the Supporting Information shows a CV at a similar device with a bare Au microelectrode in the ECF64 stock solution. ECF64 is an alkaline solution, in which gold undergoes a reduction process at around −0.7 V vs Ag quasi-RE, most likely through a decomposition of the gold sulphite ion and gold adsorption.45 At potentials below −0.9 V, oxygen is reduced, which we confirmed in separate control experiments with an Au ultramicroelectrode (diameter, 50 μm; data not shown). In the positive potential range (> 0.1 V), gold

and water oxidation processes take place, where the former proceeds via sulphite passivation of the electrode surface.50 These electrochemical features thus confirm that the embedded Au layer can operate as a functional electrode with a well-defined potential. All electrodeposition experiments were performed by stepping the potential from 0 V, where no redox reactions take place at the Au layer, to −0.7 V, i.e., the gold reduction potential. This potential was then held for 5 s. After each deposition step, the transmembrane ion current across the nanopore was measured in the electrodeposition solution (Au electrode disconnected, Vbias ± 0.5 V). Additionally, for some nanopores, after three subsequent deposition steps the devices were thoroughly cleaned and the transmembrane ion current across the nanopore was measured in both 1 M KCl and ECF64 solution, Figure 3C. While the conductivity of each solution is slightly different (measured 10.9 S/m for 1 M KCl and 3.8 S/m for the electrodeposition solution), we would have expected the qualitative behavior. However, as shown in Figure 3B, the current−voltage characteristics were qualitatively different, in that the initially linear current−voltage trace becomes nonlinear after the first electrodeposition step. The transmembrane currents are significantly higher at positive bias than at negative bias (rectification ratio ∼1.5) and the effect is apparent also after subsequent electrodeposition steps. On the other hand, in the KCl solution the current−voltage trace remains linear (rectification ratio ∼1), with the slope decreasing with the number of deposition steps. The latter is expected, if Au deposition indeed occurs inside the pore channel, thereby narrowing the pore and decreasing the pore conductance. The distinct rectification effect observed in the electrodeposition solution is therefore unlikely to be due to an asymmetric geometry and/or charge distribution inside the pore, since otherwise it would also occur in KCl solution. An electrostatic origin is also unlikely due to the high ionic strength of both solutions. On the other hand, we have shown previously that electrochemical activity inside the pore channel can induce asymmetry in the transmembrane current−voltage curve, so perhaps a similar mechanism is at play here.22,33 In support of this hypothesis, we found that the rectification decreases after exposing the nanopores to oxygen plasma, which passivates the gold surface via oxide formation, and reduces the electroactivity of the surface. We will return to this point further below. That electrodeposition does take place on the Au inside the pore channel was confirmed by TEM studies on similar devices 2339

DOI: 10.1021/ac504463r Anal. Chem. 2015, 87, 2337−2344

Article

Analytical Chemistry

Figure 4. (A) Left: Schematic of the potential distribution inside the nanopore, taking into account bipolar effects at the metal film. The total channel length is L = LAu + 2Lins, where Lins is the thickness of an insulating layer (taken to be the same on both sides). ΦM and Φs are the Coulomb potentials in the metal and solution, respectively. η is the overpotential, relative to the field-free case (Vbias = 0 V). Right: Illustration of a modified equivalent circuit taking into account both ion- and bipolar currents. (B) Measured pore current as a function of Vbias, for different concentrations of ferric/ferrous hexacyanide. (C) Limiting redox current for different concentrations of ferric/ferrous hexacyanide (cf. Device no. 16 in Table S1 in the Supporting Information).

since the overpotential required to induce a local reduction reaction at the Au electrode is too large (at least, for the Vbias values used here). The same considerations apply at Vbias < 0 and, in a symmetric system, no rectification would occur. In our devices, however, the Au layer is located toward the bottom of the pore channel. Hence, the local reduction overpotential η at the bottom and the top of the Au layer, at Vbias > 0 and Vbias < 0, respectively, are not the same, and the induced redox currents differ in magnitude, depending on the bias direction. It should be noted that, for a floating Au film and if not counterbalanced by an oxidation process, continued electrodeposition would charge the Au film increasingly positive. This would slow down the deposition reaction and increase the rate of any oxidation process that may be present. This further raised the question, whether induced electrochemical reactions could lead to a bipolar, electric coupling between the top and the bottom reservoir, for example, if both compartments contained an electrochemically active species.31−36 In that case, for Vbias > 0, reduction would be induced at the bottom edge of the Au layer (η < 0) and oxidation would be expected to take place at the top (η > 0). To a first approximation, this conductance channel is set up in parallel to the ion current path across the pore and presumably best described in terms of a charge transfer impedance Zct containing contributions from both mass transport and surface reaction kinetics. In order to quantify this effect, we used [Fe(CN)6]3−/ [Fe(CN)6]4− as a reversible redox couple at equal concentrations cb in each compartment (cb = 0.1, 1, and 10 mM in 1 M KCl) and recorded the transmembrane current as a function of Vbias, Figure 4B. Superimposed on a linear increase in the ion current, we observed an additional current contribution that was dependent on Vbias; the sigmoidal shape is reminiscent of a CV for [Fe(CN)6]3−/[Fe(CN)6]4− recorded at an ultramicroelectrode in the limit of hemispherical diffusion. However, in the present case, no potential was applied to the Au electrode and that the recorded current is the transmembrane current (not the one passing through the Au/solution interface). The limiting current, calculated from the average reduction and oxidation currents at Vbias > 0.2 V, changes linearly with cb, which indicates that it is directly related to the presence of the redox couple, Figure 4C. The limiting current at cb = 0.01 M amounts to 0.9 nA (Vbias > 0.2 V), corresponding to a value of Zct ≈ 0.1 V/0.9 nA ≈ 110 MΩ, if we take η ≈ 0.1 V (implying that the entire Vbias drops

as the ones shown above, Figure 2E−H. Gold electrodeposition was performed on an array consisting of three rows of nanopores drilled at 30 kV and 1 pA beam voltage and current, respectively, and time ranging from 5 to 100 s as illustrated in Figure S1 as well as Section 4 in the Supporting Information. Electrochemical deposition at the nanopore array was performed following the same conditions as for single nanopore experiments. The deposit is most clearly seen in the bottom cross section of the pores (panel G), which is consistent with the pore design (the Au layer is located toward the bottom part of the membrane, Figure 1A). Less material is deposited in the smaller pore, compared to the larger pore, which suggests that mass transport into the pore channel may be rate-limiting in the deposition process. The fact that the electrodeposit is localized on one side of the pore channel (indicated with arrows in Figure 2G, H), rather than homogeneously distributed on the entire Au surface, may point in the same direction (provided the entire Au surface is electrochemically active).46 Table S3 in the Supporting Information summarizes our TEM results for the nanopore devices after electrodeposition. Induced Electrochemical Activity in Metallized Nanopore Devices. Returning to the discussion of the apparent rectification effects shown in Figure 3B, it should be borne in mind that the Vbias is applied between the two Ag/AgCl electrodes, while the Au membrane electrode is disconnected during the I/Vbias scan. The total potential drop in solution is Vbias, but it is largest across the pore channel itself (since Racc ≪ Rpore, vide supra). Since the redox equilibrium at the solution/ Au interface is given by the potential drop between the metal and the solution, ΦM − Φs, the change in the solution potential along the pore channel, Φs(z), results in a local overpotential η(z). The latter depends on Vbias and the geometry of the channel, as illustrated in section 7 of the Supporting Information for a simplified model geometry. Interestingly, for a positive bias applied to the bottom compartment (Vbias > 0 in our notation), the overpotential η at the bottom of the Au layer is in fact negative and reduction reactions may be induced locally. This appears to be the case in the experiments shown in Figure 3B, where the reduction of the Au complex opens up a current path between the Au electrode and the positively biased Ag/AgCl WE in the bottom compartment (Figure 1A). This induced, additional current path determines the total current measured at WE, which explains why the observed current remains essentially unchanged in subsequent electrodeposition steps. It does not occur to the same extent in 1 M KCl solution, 2340

DOI: 10.1021/ac504463r Anal. Chem. 2015, 87, 2337−2344

Article

Analytical Chemistry over the metal layer). This is comparable to R0, measured in the absence of the bipolar contribution, and indicates that the ion current through the pore and the bipolar electronic current are similar in magnitude. For lower values of cb, Zct is larger and the bipolar contribution decreases. We also estimated the bipolar contribution using a simplified model for the limiting current. Since η is largest at the top and bottom edges of the conductive layer and the pore approximately circular, we used the expression for the (mass transport limited) current of a ring microelectrode, eq 3, and the known outer and inner pore radius b = D0/2 and a = D100/2 from above.47 Iring = nFDrdx cb

current and bipolar current modulation in the translocation signal. Bipolar currents also affect the overall resistor noise of the sensor. For example, for the simplified equivalent circuit shown in Figure 4A with Rpore and Zct in parallel, the mean square currents add, at least in the absence of correlation. For multielectrode devices,8,9 bipolar coupling can result in cross-talk between otherwise independent sets of electrodes, namely, when the local (transversal) potential drop is sufficiently high to induce electrochemical activity. Finally, it should be noted that the presence of redox-active species is not necessarily deliberate. Dissolved oxygen or protons, the continuous growth of an oxide layer on the electrode, or the dissolution of the metallic layer can also sustain a (quasi-) steady-state and thus an electronic current across the pore. In this context, it is interesting to speculate whether at least some of the current enhancement observed in nanofluidic channels made of metallic carbon nanotubes is in fact due to bipolar effects.53,54 Surface Modification Inside the Pore and the Detection of Insulin. Finally, we tested whether the metallized nanopores could be used for the detection of small proteins. In this context, controlling the potential of the embedded electrode and hence (indirectly) protein adsorption inside the pore channel could have interesting prospects for increasing the sensing capabilities of the nanopore (e.g., the specificity, controlling the residence time of the protein in the sensing region). On the other hand, the electroactivity of the membrane electrode, including the bipolar effects discussed above, could affect the sensing performance of the device. We took insulin as a model system, which should be in a predominantly dimeric state and positively charged under the conditions used here (insulin concentration 0.1 mg/mL in 0.1 M KCl + 0.04 M HCl, pH 1.6).55 Initially, we used bare metallized pores with resistances of approximately 50 MΩ (determined in 1 M KCl) corresponding to Dcyl ≈ 20 nm (e.g., R0 = 66 MΩ, D0 = 102 nm, and D100 = 10 nm). However, no events could be detected in these experiments or in controls with SiN nanopores with similar Dcyl in either bias direction. Such behavior has been observed previously and implicated with the very fast translocation of proteins, in combination with insufficient temporal resolution of the experiment.56 However, enhancing protein adsorption inside the nanopore has been shown to be viable strategy for increasing the translocation time.3,21,57 Therefore, we modified the inside of the pore with homocysteine (Hcy). Hcy binds to the Au to form a strong Au-thiol bond and exhibits two additional functional groups, namely, an amino- and carboxyl group at the α-carbon. The presence of the HCy also increases the charge transfer impedance of the metal/solution interface, thereby suppressing any Faradaic effects induced by the transmembrane bias.64 The pKa values of the three protonatable groups are 2.22 (carboxyl), 8.87 (amino), and 10.86 (thiol), the latter being attached to the Au surface.63 The pI value is approximately 5.5, neglecting the effect of surface attachment on the individual pKa values; the surface charge of the layer at pH 1.6 is positive.58 Zhang et al. found that the orientation of the molecule on the surface also depends on the applied potential.59 This raises interesting prospects toward switching the sensing capabilities of the sensor but was not investigated further at this stage here. In any case, the presence of the Hcy in the pore clearly alters the charge transport and sensing properties of the pore (Dcyl ∼

π 2(a + b)

(

a+b

ln 16 b − a

)

(3)

where n is the number of transferred electrons per redox event (here n = 1) and F Faraday’s constant. Drdx is the diffusion coefficient of the redox species. Using cb = 0.01 M, and Drdx = 7.26 × 10−10 m2/s for ferri- and 6.67 × 10−10 m2/s for ferrohexacyanide,48 we obtained values for Iring between 0.05 and 0.17 nA, for the smallest and the largest pore in Table S1 in the Supporting Information (in terms of D0). This is in acceptable agreement with the observed experimental value of 0.9 nA, given the rather crude approximation for the channel and electrode geometry inside the pore. We also ignore migration effects in our estimate, which are likely to bring the calculated values yet closer to the experimental ones. In some cases, we observed a different kind of current− voltage behavior, as shown in Figure S4 in the Supporting Information, namely, strong rectification with larger currents at Vbias < 0 (compared to the Vbias > 0 branch). It was observed in the same concentration range as in Figure 4B and the rectification ratio depended on cb. The reason for this is not immediately obvious, especially since the diffusion coefficients of ferro- and ferricyanide differ only by about 9%.48 The unmodified pores were not rectifying in redox-inactive KCl solutions. One possible explanation may be the formation of a Prussian Blue-like film from the ferri-/ferrohexacyanide in solution, which are likely to change the surface charge distribution in the channel (thus leading to rectification), and are also known to have electrocatalytic properties, both for oxidation and reduction reactions.49,50 Similar to the electrodeposition experiments discussed above, the local overpotential at the Au film may not be symmetric with regards to an inversion of Vbias, hence resulting in redox currents of different magnitude. We will address these aspects in future work. Implications of Bipolar Effects for Nanopore Sensing. Bipolar coupling may affect the properties of nanopore sensors and, more generally, electronically conductive nanofluidic channels in various ways. It is well-known that ion currents may be reduced or enhanced during the translocation, depending on the analyte, the solution conditions, and the surface properties of the electrode.51,52 Similar effects would be expected in the presence of bipolar currents across the membrane. As demonstrated above, they can increase the measured, open-pore current, which would decrease the signalto-noise ratio in single-molecule detection experiments for a given current modulation during translocation. Adsorption of an analyte to the metal surface could lead to temporary blocking of the bipolar current path (or in rare cases potentially to catalytic enhancement) and to a decrease (increase) in the observed current. The result would be a convolution of ion 2341

DOI: 10.1021/ac504463r Anal. Chem. 2015, 87, 2337−2344

Article

Analytical Chemistry

molecular volume Λ of the translocating species, according to eq 4.64

20 nm). In contrast to the unmodified pores, Hcy-modified nanopores showed a departure from Ohmic behavior and displayed negative rectification of the ionic flow (larger currents at Vbias < 0). This is most likely rooted in the asymmetric charge distribution inside the pore.60 Interestingly, in the presence of insulin in the cell, we were also able to detect translocation events of the protein, as shown in Figure 5 (data from 10 devices). The inset in Figure 5A

Λ≅

ΔIb(Heff )2 σVbias

(4)

where ΔIb is the peak amplitude, Heff is the effective length of the nanopore, and σ is the solution conductivity. With ΔIb = 110 pA, Heff = L = 207 nm, σ = 10.9 S/m, and Vbias = 0.4 V, this gives Λ ∼ 1080 nm3 or a spherical radius of ∼6.4 nm. This compares well with the hydrodynamic radius of 5.5 nm reported for insulin dimers65 and seems to suggest that the majority of translocation events are not due to protein aggregates. While we cannot rule out that at least some of the observed events correspond to proteins that entered the pore, but then leave again in the opposite direction (hence do not truly translocate from one compartment into the other), the relatively high applied voltage renders this scenario unlikely. Finally, given that both Hcy layer and insulin possess a positive net charge under the conditions used, the mode of interaction between the two inside the pore warrants further discussion. Zhang et al. studied the adsorption effect of insulin in porous charged membranes in 10 mM KCl at acidic pH 3.3 and physiological pH 7.4, at which insulin carries a positive and negative charge, respectively.66 In both cases, insulin adsorption increased in membranes with opposite charge, most likely due to electrostatic interactions between the protein and membrane inner walls. The diffusivity of insulin increased, when these interactions were repulsive. In our experiments, the observation of translocation events and hence a slowing down of pore transport would suggest that there is indeed an attractive interaction between the protein and the pore walls. Both being positively charged in solution, it seems that hydrophobic or van der Waals interactions between the protein and the Hcy are dominant, perhaps facilitated by the rather high ionic strength employed in our studies.

Figure 5. Scatter plot (A) and histogram (B) of the ion current change and dwell time of ion current blockage events for 0.1 mg/mL insulin in 0.1 M KCl, 40 mM HCl. The experiments were performed at −400 mV. Note that there is only one cluster detected for the capture events. The inset in part A shows the ion current trace as a function of time before and after the addition of the insulin solution (Device no. 8 in Table S1 in the Supporting Information).



CONCLUSIONS We have fabricated solid-state nanopore sensors with an embedded metal film and characterized their geometric structure and functional properties in detail on the basis of conductance studies, TEM imaging, tomography, as well as modeling. We used electrodeposition to shrink the nanopores to effective diameters below 10 nm and found that, under the conditions used, the deposition process is most likely mass transport controlled, taking into account the confinement in the pore channel and the structure of the deposit. TEM studies showed that the deposit is relatively rough and localized, rather than smoothly distributed over the metal surface inside the pore. We identified and quantified bipolar electrochemical effects as an additional contribution to the observed pore current, which arise from the strong, localized electric field across the pore. These may lead to apparent, redox-induced current enhancement, rectification and may affect the magnitude, shape, and polarity of translocation events detected in the current−time trace. Finally, we modified the Au surface on the inside of the pore channel with Hcy, leading to a significant change in the current−voltage characteristics of the pore as well as its sensing performance. Only after modification we were able to detect the translocation of dimeric insulin and our data further suggest that the hydrophobic, rather than electrostatic interactions

shows the ion current trace as a function of time before and after adding the insulin solution to the “cis” chamber. Before adding insulin, we observed a steady open-pore current, with a noise level below 50 pA (peak-to-peak). After the addition of insulin (0.1 mg/mL in 0.1 M KCl, 40 mM HCl (pH 1.6)), translocation events were detected, indicating the presence of the protein in the sensing region of the pore but only at Vbias < 0 (applied to the “trans” chamber), cf. Figure 5.61 This suggests that translocation is driven by electrophoresis, rather than electroosmosis (proteins were only added to the “cis” chamber). 62 We were unable to distinguish different populations in the translocation data, e.g., from the simultaneous presence of monomeric and dimeric insulin. The average dwell time of the capture events is 130 ms (most probable dwell time ∼32 ms) and average event current 102 ± 23 pA, as shown in the scatter plot and histogram in parts A and B of Figure 5, respectively. The dwell time is orders of magnitude larger than what is expected for unperturbed electrophoretic transport across the pore and points toward specific interactions of insulin with Hcy.63 Crowding effects, wherein proteins aggregate and block the entrance of the pore, are another possibility in principle but were not observed with the similarly sized, unmodified pores. That the observed events indeed correspond to individual proteins is further supported by an estimate of the excluded 2342

DOI: 10.1021/ac504463r Anal. Chem. 2015, 87, 2337−2344

Article

Analytical Chemistry

reaction) to −0.7 V (Au reduction potential vs Ag RE), and holding at the reduction potential for 5 s. After each step the conductance across the nanopore was measured in 1:10 (v/v) ECF64. After a set of three chronoamperometric steps, the samples were cleaned by exposing to oxygen plasma for 5 min each side and the conductance was measured in 1 M KCl vs Ag/AgCl RE. The sequence was repeated three times in total. Insulin Detection. A volume of 400 μL of 100 mM KCl, 40 mM HCl aqueous solution was added to both compartments of the cell, and the cell was connected to the patch clamp amplifier (Axopatch 200B, Molecular Devices, CA) using two Ag/AgCl wires as the WE and RE/CE in the cis and trans chambers, respectively. A volume of 125 μL of 1 mg/mL of insulin stock solution (Bovine pancreas insulin, Sigma-Aldrich) in 100 mM KCl, 40 mM HCl was added to the cis chamber. Ionic current traces were recorded at −400 mV, 5 kHz lowpass filter frequency and 10 kHz sampling rate and analyzed using Matlab 2014a.

between the protein and the Hcy layer govern the residence time of the protein inside the pore channel.



METHODS Fabrication of Metallized Nanopores. Metallized nanopore devices were fabricated following the protocol described elsewhere.22 Briefly, the multilayered platforms (Figure 1A) were produced using (100) silicon wafers with 30 nm SixNy layers on either side. The gold layer was electron beam sputtered on the top side (30 nm Au with 5 nm Ti adhesive layer), followed by anisotropic KOH wet etching of silicon. Finally, the top side was coated with 150 nm SiOx using plasma enhanced chemical vapor deposition (PE-CVD). Nanopore milling was performed using dual beam FIB/SEM as described before.22 A combination of beam current and exposure time were studied to optimize the nanopore size for the gold electrodeposition studies. TEM Imaging. The TEM/STEM used for this study was a JEOL JEM 2100 which was operated at 200 kV. Tomography was performed using a high tilt tomography holder capable of −60 to 60 deg tilting. Tilting procedures were typically performed at smaller angles (±35 deg) due to the high aspect ratio of the pores. The tilt series was accomplished using the SerialEM software and reconstructed using IMOD and Chimera. Cross-sectional slices were taken in the x and zplanes to visualize changes in the internal profile of the pore. The nanopore dimensions before and after the electrodeposition were calculated using ImageJ 1.48v. Since single axis tilting was performed, the pore profile was most accurately reconstructed in the x-axis. For this reason, the diameters reported here were measured perpendicular to the axis of tilting. Solution Preparation. All aqueous solutions were prepared using ultrapure (18.2 MΩ) water. The 1 mM Hcy solution was prepared in 50% ethanol. The gold electroplating solution, ECF64 (Metalor) aqueous solution was prepared in a 1:10 volummetric ratio of the stock solution to water (v/v) and contained 16.5 g/L gold in ammonium gold sulphite. The supporting electrolyte in ECF64 was ammonium sulphite (70 g/L). Electrochemical Setup. All electrochemical measurements were performed in a polyether ether ketone cell, consisting of two reservoirs filled with aqueous solutions as shown in Figure S1A in the Supporting Information and described in more detail elsewhere.22 Either a pristine silver wire or a silver wire with silver chloride coating (prepared in-house) were used as quasi-RE and CE for electrodeposition in ECF64 and conductance studies in other aqueous solutions and Hcy, respectively. Ag quasi-RE/CE was used in all experiments with ECF64 solution, since AgCl is unstable in ECF64 (complexation with amine). The WE was either a second Ag/AgCl electrode in the opposite compartment (for conductance studies) or the gold microelectrode (for the electrodeposition experiments). All electrochemical measurements were performed using a CH Instruments bipotentiostat, CHI 760C. Prior to electrochemical deposition experiments, the samples were cleaned in acetone and ethanol and exposed to oxygen plasma (5 min each side).22 The electrochemically active surface area was determined by differential normal pulse voltammetry based on the redox response of surfaceimmobilized 6-ferrocenyl 1-hexanethiol, as outlined in section 8 of the Supporting Information. Electrochemical deposition was performed by stepping the potential from 0 V (no Faradaic



ASSOCIATED CONTENT

S Supporting Information *

Nanopore diameter for different milling parameters; background for calculating the total resistance, R0, of metallized nanopores; electrochemistry at a gold microelectrode in the deposition solution; electrochemical deposition at nanopore array; bipolar effects in devices with metallized pores; rectification effect for surface modified Au pores; theoretical considerations around the bipolar effect in a metallized pore; and electroactive area of the membrane electrode. This material is available free of charge via the Internet at http://pubs.acs.org.



AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS This work was supported in part by the European Research Council Starting Investigator Grant (J.B.E.), the Biological and Biotechnological Sciences Research Council (Grant BB/ L017865/1, J.B.E., A.R.), and Whitaker Fellowship (K.F.).



REFERENCES

(1) Venkatesan, B. M.; Bashir, R. Nat. Nanotechnol. 2011, 6, 615. (2) Japrung, D.; Nadzeyka, A.; Peto, L.; Bauerdick, S.; Albrecht, T.; Edel, J. B. Biophys. J. 2012, 102, 205A. (3) Japrung, D.; Dogan, K.; Freedman, K.; Nadzeyka, A.; Bauerdick, S.; Albrecht, T.; Kim, M. J.; Jemth, P.; Edel, J. B. Anal. Chem. 2013, 85, 2449. (4) Goyal, G.; Freedman, K. J.; Kim, M. J. Anal. Chem. 2013, 85, 8180. (5) Ivanov, A. P.; Instuli, E.; McGilvery, C. M.; Baldwin, G.; McComb, D. W.; Albrecht, T.; Edel, J. B. Nano Lett. 2011, 11, 279. (6) Ivanov, A. P.; Albrecht, T.; Edel, J. B.; Freedman, K. J.; Kim, M. J. ACS Nano 2014, 8, 1940. (7) Dekker, C. Nat. Nanotechnol. 2007, 2, 209. (8) Siwy, Z. S.; Howorka, S. Chem. Soc. Rev. 2010, 39, 1115. (9) Bahrami, A.; Dogan, F.; Japrung, D.; Albrecht, T. Biochem. Soc. Trans. 2012, 40, 624. (10) Miles, B. N.; Ivanov, A. P.; Willson, K. A.; Dogan, F.; Japrung, D.; Edel, J. B. Chem. Soc. Rev. 2013, 42, 15. (11) Heins, E. A.; Siwy, Z. S.; Baker, L. A.; Martin, C. R. Nano Lett. 2005, 5, 1824. 2343

DOI: 10.1021/ac504463r Anal. Chem. 2015, 87, 2337−2344

Article

Analytical Chemistry (12) Siwy, Z. S. Adv. Funct. Mater. 2006, 16, 735. (13) Chang, H.; Kosari, F.; Andreadakis, G.; Alam, M. A.; Vasmatzis, G.; Bashir, R. Nano Lett. 2004, 4, 1551. (14) Li, J.; Stein, D.; McMullan, C.; Branton, D.; Aziz, M. J.; Golovchenko, J. A. Nature 2001, 412, 166. (15) Chansin, G. A. T.; Mulero, R.; Hong, J.; Kim, M. J.; DeMello, A. J.; Edel, J. B. Nano Lett. 2007, 7, 2901. (16) Kalman, E. B.; Sudre, O.; Vlassiouk, I.; Siwy, Z. S. Anal. Bioanal. Chem. 2009, 394, 413. (17) Jiang, Z. J.; Stein, D. Phys. Rev. E 2011, 83, 031203. (18) Wei, R.; Gatterdam, V.; Wieneke, R.; Tampe, R.; Rant, U. Nat. Nanotechnol. 2012, 7, 257. (19) Rutkowska, A.; Edel, J. B.; Albrecht, T. ACS Nano 2013, 1, 547. (20) Siwy, Z.; Heins, E.; Harrell, C. C.; Kohli, P.; Martin, C. R. J. Am. Chem. Soc. 2004, 126, 10850. (21) Sa, N.; Baker, L. A. J. Am. Chem. Soc. 2011, 133, 10398. (22) Gibb, T. R.; Ivanov, A. P.; Edel, J. B.; Albrecht, T. Anal. Chem. 2014, 86, 1864. (23) Gong, X.; Patil, A. V.; Ivanov, A. P.; Kong, Q.; Gibb, T.; Dogan, F.; deMello, A. J.; Edel, J. B. Anal. Chem. 2014, 86, 835. (24) Ayub, M.; Ivanov, A.; Instuli, E.; Cecchini, M.; Chansin, G.; McGilvery, C.; Baldwin, G.; McComb, D.; Edel, J. B.; Albrecht, T. Electrochim. Acta 2010, 55, 8237. (25) Ai, Y.; Liu, J.; Zhang, B. K.; Qian, S. Anal. Chem. 2011, 82, 8217. (26) He, Y. H.; Tsutsui, M.; Fan, C.; Taniguchi, M.; Kawai, T. ACS Nano 2011, 5, 8391. (27) He, Y.; Tsutsui, M.; Fan, C.; Taniguchi, M.; Kawai, T. ACS Nano 2011, 5, 5509. (28) Yen, P. C.; Wang, C. H.; Hwang, G. J.; Chou, Y. C. Rev. Sci. Instrum. 2012, 83, 034301. (29) Karnik, R.; Fan, R.; Yue, M.; Li, D. Y.; Yang, P. D.; Majumdar, A. Nano Lett. 2005, 5, 943. (30) Albrecht, T. ACS Nano 2011, 5, 6714. (31) Fleischmann, M.; Ghoroghchian, J.; Rolison, D.; Pons, S. J. J. Phys. Chem. 1986, 90, 6392. (32) Arora, A.; Eijkel, J. C. T.; Morf, W. E.; Manz, A. Anal. Chem. 2001, 73, 3282. (33) Mavré, F.; Chow, K.-F.; Sheridan, E.; Chang, B.-Y.; Crooks, J. A.; Crooks, R. M. Anal. Chem. 2009, 81, 6218. (34) Fosdick, S. E.; Knust, K. N.; Scida, K.; Crooks, R. M. Angew. Chem., Int. Ed. 2013, 52, 10438. (35) Duval, J. F. L.; van Leeuwen, H. P.; Cecilia, J.; Galceran, J. J. Phys. Chem. B 2003, 107, 6782−6800. (36) Duval, J. F. L.; Minor, M.; Cecilia, J.; van Leeuven, H. P. J. Phys. Chem. B 2003, 107, 4143−4155. (37) Nielsen, L.; Frokjaer, S.; Brange, J.; Uversky, V. N.; Fink, A. L. Biochemistry 2001, 40, 8397. (38) Spinney, P. S.; Howitt, D. G.; Smith, R. L.; Collins, S. D. Nanotechnology 2010, 21, 375301. (39) Hille, B. J. Gen. Physiol. 1968, 51, 199. (40) Hille, B. Prog. Biophys. Mol. Biol. 1970, 21, 1. (41) Wei, R.; Pedone, D.; Zürner, A.; Döblinger, M.; Rant, U. Small 2010, 6, 1406. (42) Prabhu, A. S.; Freedman, K. J.; Robertson, J. W.; Nikolov, Z.; Kasianowicz, J. J.; Kim, M. J. Nanotechnology 2011, 22, 425302. (43) Kowalczyk, S.; Grosberg, A. Y.; Rabin, Y.; Dekker, C. Nanotechnology 2011, 22, 315101. (44) Willmott, G. R.; Smith, B. G. Nanotechnology 2012, 23, 088001. (45) Gold: Science and Applications; Corti, C., Holliday, R., Eds.; CRC Press: Boca Raton, FL, 2010. (46) Paunovic, M.; Schlesinger, M. Fundamentals of Electrochemical Deposition, 2nd ed.; Wiley-Interscience: Hoboken, NJ, 2006. (47) Liljeroth, P.; Johans, C.; Slevin, C. J.; Quinn, B. M.; Kontturi, K. Electrochem. Commun. 2002, 4, 67. (48) Konopka, S. J.; McDuffie, B. Anal. Chem. 1970, 42, 1741−1746. (49) de Tacconi, N. R.; Rajeshwar, K.; Lezna, R. O. Chem. Mater. 2003, 15, 3046−3062. (50) Itaya, K.; Hida, I.; Neff, V. D. Acc. Chem. Res. 1986, 19, 162− 168.

(51) Smeets, R. M. M.; Keyser, U. F.; Krapf, D.; Wu, M. Y.; Dekker, N. H.; Dekker, C. Nano Lett. 2006, 6, 89−95. (52) Lan, W.-J.; Kubeil, C.; Xiong, J.-W.; Bund, A.; White, H. S. J. Phys. Chem. C 2014, 118, 2726. (53) Liu, L.; Yang, C.; Zhao, K.; Li, J.; Wu, H.-C. Nat. Commun. 2013, 4, No. 2989. (54) Liu, H.; He2, J.; Tang, J.; Liu, H.; Pang, P.; Cao, D.; Krstic, P.; Joseph, S.; Lindsay, S.; Nuckolls, C. Science 2010, 327, 64−67. (55) Kivlehan, F.; Lanyon, Y. H.; Arrigan, D. W. Langmuir 2008, 24, 9876. (56) Plesa, C.; Kowalczyk, S. W.; Zinsmeester, R.; Grosberg, A. Y.; Rabin, Y.; Dekker, C. Nano Lett. 2013, 13, 658. (57) Zhang, J.; Welinder, A. C.; Chi, Q.; Ulstrup, J. Phys. Chem. Chem. Phys. 2011, 13, 5526. (58) Nekrassova, O.; Lawrence, N. S.; Compton, R. G. Talanta 2003, 60, 1085. (59) Zhang, J.; Demetriou, A.; Welinder, A. C.; Albrecht, T.; Nichols, R. J.; Ulstrup, J. Chem. Phys. 2005, 319, 210. (60) Siwy, Z. S. Adv. Funct. Mater. 2006, 16, 735. (61) Fologea, D.; Ledden, B.; McNabb, D. S.; Li, J. Appl. Phys. Lett. 2007, 91, 053901. (62) Firnkes, M.; Pedone, D.; Knezevic, J.; Döblinger, M.; Rant, U. Nano Lett. 2010, 10, 2162. (63) Niedzwiedzki, D. J.; Grazul, J.; Movileanu, L. J. Am. Chem. Soc. 2010, 132, 10816. (64) Talaga, D. S.; Li, J. J. Am. Chem. Soc. 2009, 131, 9287. (65) Oliva, A.; Farina, J.; Llabrés, M. J. Chromatogr., B: Biomed. Sci. Appl. 2000, 749, 25. (66) Zhang, S.; Tanioka, A.; Saito, K.; Matsumoto, H. Biotechnol. Prog. 2009, 25, 1115.

2344

DOI: 10.1021/ac504463r Anal. Chem. 2015, 87, 2337−2344