Peptide aptamers are short peptide sequences presented and conformationally constrained in a robust, inert protein scaffold. They are selected in vivo from large libraries (as many as 10⁹ unique peptide sequences), yielding binders with high and very specific affinities for selected targets 1718. The three-dimensional conformational constraint of the inserted peptide applied by the protein scaffold greatly increases the affinity of the aptamer for the target over that of an unconstrained peptide sequence 17181920. They are distinguished from similar protein or peptide-based strategies by being selected in eukaryotic cells for binding to targets from large libraries using a yeast two-hybrid approach 2122. This maximizes the likelihood that the target protein is folded in the same way, and undergoes the same posttranslational modifications, as in its native environment, providing reliability and confidence in the specificity of detection.

Traditionally, the Escherichia coli protein thioredoxin A (TrxA) has been used as a scaffold to present peptide aptamers 17. However, many TrxA-based peptide aptamers are not expressed stably, which limits their use. In addition, the expression of TrxA in mammalian cells can lead to detectable phenotypes 2324, indicating that the bacterial protein scaffold is able to interact with mammalian proteins, which would inevitably increase the background signal in any microarray format. We recently developed a new biologically inert protein scaffold (Stefin A triple mutant (STM)) that can present a wide range of different peptide sequences and has a simplicity and robustness that allows the production technique to be generic for all targets 25. Here, we use peptide aptamers in STM to demonstrate label-free electronic transduction of biorecognition events occurring between human cyclin-dependent protein kinases (CDKs) as target proteins in solution and specific peptide aptamers immobilized on gold electrode arrays. Immobilization of the STM scaffold is achieved by the introduction of a single cysteine residue at the amino terminus, enabling attachment to gold electrodes via S-Au bonds (see 26 for further details). This cysteine residue is the only cysteine present in the scaffold and, in three dimensions, is located at the opposite side to the peptide insert (Figure 1a), thus separating the peptide insert from the surface.

In this study, we used two different peptide aptamers displayed by cysteine-modified STM, STM_pep2 and STM_pep9, where the subscripts refer to two different peptide inserts that both recognize cyclin-dependent kinase 2 (CDK2), but where only STM_pep9 binds to cyclin-dependent kinase 4 (CDK4). Both CDK2 and CDK4 belong to a group of proteins involved in the regulation of the cell cycle; they are functionally related, yet have less than 46% sequence identity. STM_pep2 and STM_pep9 were generated by insertion of oligonucleotides encoding the CDK-interacting peptide sequences derived from the thioredoxin-based peptide aptamers of Colas et al. 17 into restriction sites in the open reading frame encoding the STM protein scaffold 2527. The binding of CDK2 and CDK4 to the peptide aptamers was confirmed in vivo using yeast two-hybrid interaction analysis (25 and PKF, SJ, DE, SL, AGD and CW, unpublished work) and in vitro by fluorescence resonant energy transfer spectroscopy (FRET, Figure 1b) and dual polarization interferometry (DPI) 27. The latter experiments showed that the affinity between surface-immobilized STM_pep9 and CDK2 in a complex biological mixture compares well to typical values of K_D for surface-immobilized antibodies 27. These in vitro techniques also confirmed that the performance of the peptide aptamers is not affected by tethering the STM scaffold to a surface.

Electronic, label-free, on-chip detection of the peptide aptamer-target interactions is based on monitoring local changes in the impedance of the electrochemical double layer, which forms above the surface of a metal electrode when it is submerged in an electrolyte 28. Any perturbation of this double layer, for instance by attachment of proteins to the electrode, alters the electrical properties of the layer. For example, the complex electrical impedance Z(ω) is a measure of the extent to which the charge transfer to and from the electrode is impeded by the surface-immobilized proteins. Hence, Z(ω) depends on the density, thickness and internal structure of the protein layer, and any alteration in this layer, such as the binding of a molecular target, potentially results in a measurable change of Z(ω) 29. Z(ω) can be determined from the response of the system, i.e. the electrochemical current I, upon applying an ac electrochemical potential Φ of frequency ω to the electrode. Changes in Z(ω) manifest themselves in changes of the absolute impedance |Z(ω)| and its phase φ(ω), that is, the phase difference between Φ and I. We note that while |Z(ω)| scales with the electrode surface area, φ(ω) is independent of the electrode area, and changes in φ(ω), Δφ(ω), therefore provide a reliable and reproducible measure of changes at the electrode surface, which will not be affected by variability in electrode surface area.

We used electrochemical impedance spectroscopy (EIS) to determine |Z(ω)| and φ(ω) as a function of frequency, ω. The measurements were performed in the presence of a redox probe (K₃Fe(CN)₆^4-/3-) 29. Gold electrodes functionalized with STM_pep9 or STM were exposed to 45 μl of a solution containing about 200 ng/μl of purified, recombinant CDK2 (rCDK2) expressed in E. coli and subsequently rinsed with deionized water to remove any excess rCDK2.

Figure 2a,b show φ(ω) for STM_pep9- and STM-functionalized electrodes, respectively, both before and after exposure to purified rCDK2. A shift in φ is observed upon rCDK2 binding to STM_pep9, whereas no change was detected in the case of STM. This shift is more obvious when plotting the difference in the phase, Δφ, before and after exposure to rCDK2 (Figure 2c). Whereas a pronounced peak in Δφ is measured for STM_pep9, no change in Δφ is observed for STM. These results demonstrate that Δφ provides a method to detect binding of the targets to the probe molecules immobilized on a gold electrode.

To determine the concentration dependence of the phase shift, Δφ(c), gold electrodes were functionalized with STM_pep9 and were exposed to 50 μl of phosphate buffer containing a range of concentrations of purified baculoviral CDK2 between 25 pM and 100 nM. The electrodes were subsequently rinsed in phosphate buffer to remove any excess CDK2 before φ(ω) was measured. The results are shown in Figure 2d from which a sensitivity limit of around 50 pM (approximately equal to 1.5 ng/ml CDK2) can be determined, which is in the clinically relevant range 7. The phase shift is linear on a logarithmic concentration scale over more than three orders of magnitude. The solid line in Figure 2d represents a linear fit to the data.

In biologically and medically relevant samples, the proteins of interest are typically only present at low concentration and in complex mixtures of similar biological molecules. To assess the suitability of our sensing technique for the detection of proteins in such samples, we prepared gold electrodes functionalized with STM_pep9 and STM, and exposed them to 35 μl of a whole-cell lysate of CDK2-expressing yeast cells. Following exposure to yeast lysate, the devices were thoroughly washed to remove any non-specifically bound material.

The phase φ(ω) of the complex impedance measured for the different devices is shown in Figure 3. Whereas a distinct shift in φ(ω) is observed between 1 and 10³ Hz for STM_pep9 exposed to CDK2 lysate (Figure 3a,c), there is no change in φ(ω) across the whole frequency range investigated for STM exposed to the different aliquots of the same lysate (Figure 3c). Given that STM_pep9 and STM differ only in the presence (or absence) of the peptide aptamer insert, the dramatic difference in impedance characteristics following exposure to the lysate must be related to an interaction with STM_pep9 through the peptide insert. To confirm that this response is related to the formation of the CDK2-STM_pep9 complex, rather than to binding with other species in the lysate, we exposed a series of STM_pep9 functionalized electrodes to lysate of yeast cells that were not expressing CDK2 (Figure 3b). The lack of a shift in phase following exposure to this CDK-free yeast lysate (see Figure 3c) confirms the specific affinity of STM_pep9 for CDK2. The protein complexity of a yeast lysate (comprising proteins expressed from approximately 4,000 open reading frames at any given time) compares favorably with that of medically relevant biofluids, such as urine (1,400 individual proteins 30), and serum or plasma (with estimates of 1,175 31 to 3,700 32 individual proteins). These results demonstrate the ability of our peptide aptamer sensor to unambiguously detect target-aptamer binding from such samples.

Many future applications in biology (such as systems analysis of protein interactions) and medicine (such as personalized medicine protocols) will rely on the ability to investigate large numbers of proteins simultaneously. Very often the available sample volume is likely to be limited, particularly in the case of biopsies from patients. Therefore, high-density arrayed systems – that is, arrays with very small and very closely packed sensors – are required. Conventional technologies for generating protein arrays are generally based on dot-printing strategies with resolutions of the order of 0.1 mm or more. High-density arrays, with submicrometer or even micrometer feature sizes, are beyond the scope of these approaches. To enable the immobilization of different peptide aptamers on different electrodes spaced only a few micrometers apart, we developed a process inspired by our previous work 16. This enables the selective functionalization of individual microelectrodes of an array with resolutions at least an order of magnitude better than that achieved by conventional techniques.

Selective functionalization of the microelectrodes with different peptide aptamers was achieved through the molecular masking process illustrated in Figure 4. The microelectrode arrays comprised ten individually addressable gold microelectrodes of 20 μm width and separated by 15 μm, which were first coated with a methyl-terminated poly(ethylene-glycol)₆-thiol (mPEG, Polypure, Oslo, Norway) layer that prevents nonspecific binding of proteins during microelectrode functionalization. The thiol-modification of the mPEG not only allows the spontaneous formation of a molecular monolayer on the gold microelectrodes through the Au-S bond, but also provides a means for removing the masking layer from any individual microelectrode through reductive cleavage of this bond 3334. The quality of the resulting mPEG layers was verified using water contact-angle measurements and X-ray photoelectron spectroscopy, and the effectiveness of protein inhibition was confirmed by fluorescence spectroscopy (data not shown). After formation of the mPEG layer, the arrays were soaked for 1 hour in deionized water to remove residual ethanol and to form a water layer around the mPEG, which is believed to be crucial in the inhibition of protein binding 35.

The mPEG molecular mask was selectively removed from an individual microelectrode by applying an electrochemical potential of -1.4 V versus Ag/AgCl for 120 seconds using a potentiostat and identical buffer conditions to those used in EIS measurements (Figure 4b). Owing to the small spacing between the microelectrodes, the electric fields generated during desorption may conceivably influence the electrochemical potential of neighboring microelectrodes, potentially disturbing the blocking layer. To prevent this while allowing scalability to smaller microelectrode geometries, a second potentiostat was used to hold the potential of the neighboring microelectrodes at -0.2 V versus Ag/AgCl during the desorption process. The efficacy of desorption is monitored by cyclic voltammetry (Figure 5a). Having desorbed the mPEG molecular mask, the bare gold microelectrode can be functionalized with the desired protein by incubating the device in 35 μl protein solution overnight in a sealed humid environment (Figure 4c). The adsorption of the protein, and the effectiveness of the mPEG monolayers for masking deposition on protected microelectrodes, is confirmed using cyclic voltammetry and EIS. This process can be repeated to functionalize further microelectrodes with different proteins (Figure 4d).

Both STM_pep9 and STM_pep2 show an affinity for CDK2, whereas only STM_pep9 showed an affinity for CDK4 (Figure 1b). We exploit this difference in functionality to demonstrate our multielectrode array sensor's ability to discriminate between binding events occurring on differently functionalized microelectrodes, fabricated on a single device. Two separate, nominally identical array devices with adjacent individual microelectrodes functionalized with the two different peptide aptamers, STM_pep9 and STM_pep2, were challenged with either CDK2- or CDK4-expressing lysate. The EIS results are shown in Figure 5b,c. Shifts in φ(ω) are observed for both the STM_pep9- and STM_pep2-functionalized microelectrodes following exposure to the CDK2-expressing yeast lysate, as expected from the FRET data (Figure 1b). Conversely, on exposure to CDK4-expressing yeast lysate, a shift in φ(ω) of similar magnitude was only observed for the STM_pep9 functionalized microelectrode. The lack of response following exposure of STM_pep2-functionalized microelectrodes to CDK4 indicates the high selectivity of the functionalization process. In all cases, φ(ω) remains constant for those microelectrodes coated with mPEG, confirming the efficacy of the inhibiting layer.

The peptide aptamers were generated and immobilized on gold electrodes as described in 2627. The yeast lysate was prepared as detailed in 1827. Purified activated baculoviral CDK2 was sourced from Cell Signaling Technology (Danvers, MA).

STM and STM_pep9 were labeled with Alexafluor 488 TFP ester (Molecular Probes, Invitrogen, Carlsbad, CA). Ten microliters of protein (approximately 2 mg/ml) were diluted with 10 μl 1 M NaCO₃ aqueous solution, after which 5 μl Alexafluor dissolved in dimethyl sulfoxide (DMSO) was added. This mixture was left to incubate for 1 h, after which 1 μl Tris HCl was added to bind unbound dye. The solution was then filtered through a protein-desalting column (Pierce, Rockville, IL) to remove excess fluorophores. The CDK lysates were also dyed in the same way but using Alexafluor 555 NHS ester (Molecular Probes).

Each electrode array structure comprises two opposing rows of five individually addressable electrodes, each 20 μm wide and separated by 15 μm. The electrode structure was fabricated using double lift-off photolithography as follows. N-doped silicon (100) substrates 30 μm thick with a thermally grown 500 nm oxide (Compart Technology, Peterborough, UK) were cleaned in piranha solution (70% H₂SO₄:30% H₂O₂) at 100°C for 5 min. The wafer was baked at 150°C for 60 min before polymethyl methacrylate (PMMA, MW 950 kDa, 4% in anisole) was spun on at 1,500 rpm for 40 sec. Following a 90-min bake step at 150°C, Shipley 1805 photoresist was spun onto the PMMA at 5,000 rpm for 40 sec and baked on a hotplate at 115°C. The wafer was then cleaved, and individual sections were exposed under 8 mW/cm² ultraviolet light (UV) in a mask aligner for 4 sec and developed in MF319 photoresist developer for 60 sec. The chip was rinsed in deionized water (Millipore, 18.2 Ωcm) and transferred to a UV ozone cleaner where it was exposed to deep UV for 16 min to expose the PMMA and harden the upper photoresist layer. Following this treatment, the PMMA was developed in 3:1 isopropanol:methyl isobutyl ketone for 20 sec and rinsed in isopropanol before the exposed surface was cleaned in 1:5 HCl:H₂O. Finally, 15 nm Cr and 50 nm Au were evaporated onto the wafers, and lifted off using acetone for 30 min. Individual devices were then scribed out, and the wafer cleaved. The devices were mounted in ceramic packages with Araldite, wire bonded and cleaned in piranha solution before functionalization.

Gold electrodes were functionalized with peptide aptamers by exposure to 35 μl of either STM_pep2, STM_pep9 or STM in a PBS buffer (pH 7.3) for 18 h at room temperature. The devices were subsequently exposed to the solution containing the target proteins and then rinsed with deionized water (18.2 Ωcm, Millipore) to remove any excess target solution. Coating the gold electrodes with mPEG was achieved by immersing the devices in a 10 mM ethanolic solution for 96 h.

EIS measurements on individual gold electrodes, where |Z(ω)| and φ(ω) are determined as a function of frequency between 0.1 Hz and 500 kHz, were carried out using a custom built three-electrode electrochemical cell with an electrolyte volume of 50 ml, and a Princeton Applied Research VSP bipotentiostat/impedance spectrometer in a common-electrode-to-ground configuration. The gold electrode was used as the working electrode, and a 50-mm-long high-purity Pt wire (Agar Scientific, Stanstead, UK) was fully immersed in the electrolyte and used as the common electrode. The reference electrode was a Ag/AgCl double-junction ceramic wick electrode (VWR, Lutterworth, UK). The electrolyte consisted of 100 mM phosphate buffer pH 7.0 containing 10 mM K₃Fe(CN)₆^4-/3- redox probe which decreases the impedance of the system by increasing the measurable quantities and hence improves the sensitivity of the system 29. All electrochemical potentials are reported against a Ag/AgCl reference electrode. Errors in the accuracy of the phase-shift measurements can be determined from the cumulative error of the response to CDK-free phosphate buffer and the measurement error revealed through repeating the readings; the overall error in the phase shift is estimated to be ± 0.5°.

Concentration dependence and sensitivity limits were measured by exposing the gold electrodes, functionalized with STM_pep9, to phosphate buffer solutions containing various concentrations of commercial purified activated CDK2 (Cell Signaling Technology). The STM_pep9-functionalized electrodes were placed in 100 mM phosphate buffer containing 10 mM redox probe (K₃Fe(CN)₆^4-/3-), and the impedance measurements were taken repeatedly to confirm stability. The redox buffer was then removed, and the chip rinsed in phosphate buffer without redox probe, and incubated in pure phosphate buffer for 60 min, after which impedance measurements were repeated to confirm the zero response of the surface. The gold electrode was then exposed for 1 h to 50 μl phosphate buffer containing the required concentration of CDK. Final impedance measurements were taken after the gold electrodes were rinsed in phosphate buffer, transferred to the measurement buffer (phosphate buffer containing the redox probe), and the system allowed to equilibrate. All buffers were filtered through a 0.22 μm filter before use. Each measurement was repeated three times to confirm the reading, and repeated sets of measurements were made at 30-min intervals to ensure stability. The accuracy in the concentration of the CDK solutions was determined by the accuracy of the pipettes used and the repeated dilution steps from an initial 100 ng/μl solution. Errors in the accuracy of the phase-shift measurements were calculated as above and the corresponding error bars are indicated on Figure 2d.

The efficacy of the desorption of the molecular masking layer and the subsequent adsorption of the proteins was monitored using cyclic voltammetry (CV). These experiments were carried out in the same buffer conditions as the EIS measurements (100 mM phosphate buffer pH 7.7 with 10 mM K₃Fe(CN)₆^4-/3- redox probe) and at scan rates of 62 mV/s. The CV curve of the mPEG-coated gold electrode is shown in Figure 5a. Upon desorption of this layer, the peak separation on the voltammogram decreases from 425 mV to 100 mV, typical of a gold surface with this redox probe 29. Following the adsorption of cysteine-modified STM_pep2 onto the exposed gold electrode, the peaks in the CV curve decrease and the peak separation increases, indicating the successful binding of the peptide aptamer to the surface.