Catalytic activity of glucose oxidase after dielectrophoretic immobilization on nanoelectrodes

Dielectrophoresis (DEP) is an AC electrokinetic effect that is proven to be effective for the immobilization of not only cells, but also of macromolecules, for example, antibodies and enzyme molecules. In our previous work, we have already demonstrated the high catalytic activity of immobilized horseradish peroxidase after DEP. To evaluate the suitability of the immobilization method for sensing or research in general, we want to test it for other enzymes, too. In this study, glucose oxidase (GOX) from Aspergillus niger was immobilized on TiN nanoelectrode arrays by DEP. Fluorescence microscopy showed the intrinsic fluorescence of the immobilized enzymes flavin cofactor on the electrodes. The catalytic activity of immobilized GOX was detectable, but a fraction of less than 1.3% of the maximum activity that was expected for a full monolayer of immobilized enzymes on all electrodes was stable for multiple measurement cycles. Therefore, the effect of DEP immobilization on the catalytic activity strongly depends on the used enzyme.


INTRODUCTION
Dielectrophoresis (DEP) is the manipulation of particles in an inhomogeneous, alternating electric field due to differences in polarizability between the medium and the particles. DEP can be used to manipulate, classify, and sort cells [1]. Recently, more and more examples of DEP electric field in a buffer of low conductivity. There is some qualitative evidence for conserved binding ability of DEP-immobilized antibodies [10] and catalytic activity of immobilized horseradish peroxidase (HRP) [11]. Recently, we reported on a quantitative study of the activity of HRP immobilized by DEP on nanoelectrode arrays and demonstrated that up to 45% of the maximally expected activity due to geometric considerations was reached by the permanently immobilized HRP. The result was interpreted as a full monolayer of active, but randomly oriented HRP on the electrodes [12].
Because of the small size of the molecules, an effective protein DEP experiment requires extremely sharp electrode geometries or strong fields to create a DEP force that is strong enough to overcome Brownian motion and electrohydrodynamic streaming effects [13]. Reported experiments on enzymes immobilized by DEP are therefore still scarce. The mechanism of permanent adsorption to the electrodes beyond the time of field application is also unclear. Still, there are other examples of permanent protein immobilization in electric fields which work on comparably blunt electrodes, for example, the electrophoretic deposition of active enzymes in nonuniform electric fields with asymmetric AC waveforms. One approach is the immobilization on platin wire for the use in biosensors, which was demonstrated for glucose oxidase (GOX), βgalactosidase, and catalase [14][15][16]. Another interesting study describes the activity of desoxyribonuclease I after immobilization on a titanium implant assisted by AC electrophoretic deposition [17]. There is also extensive research on the adsorption of proteins on flat electrodes in homogenous electric fields, which was in some cases explained by electrostatic interactions [18,19]. In a recent study, the enzyme lysozyme is adsorbed to an electrode and its polarization was studied in an alternating electric field. Furthermore, lysozyme adsorption onto a lysozyme monolayer assisted by alternating electric fields was optimized, taking into account dielectric properties of the enzyme [20]. As adsorption took place under the influence of a strong direct current (DC) offset and in a homogeneous electric field, the mechanism of protein attraction and immobilization is expected to be very different from DEP and the results cannot be simply transferred to protein DEP.
In this study, the activity of GOX after immobilization on nanoelectrode arrays by DEP is evaluated. The same TiN nanoelectrode arrays were already used to immobilize active HRP. The geometry of the electrodes allows to produce a sufficiently high field gradient for protein DEP and is one of the sharpest we are aware of (d = 20 nm). The chip surface is also more chemically inert and causes fewer unwanted side reactions than comparable metal electrode types [12]. As HRP is an extraordinarily stable enzyme [21,22] and as there is growing evidence for the denaturation and inactivation of enzymes in AC electric fields [23,24] the effect of the immobilization procedure has to be studied on further enzymes. As an interesting example, GOX was chosen because of the enzyme's moderate activity and stability as well as for its high relevance in biosensing applications. It catalyzes the oxidation of β-d-glucose by O 2 to d-gluconolactone and H 2 O 2 [25]. As shown in Equations (1) and (2), the product H 2 O 2 can be detected in a coupled enzyme assay using HRP and the colorless substrate Amplex Red, which reacts to fluorescent resorufin [26]. The latter reaction is well tested for enzyme assays in small volumes using a fluorescence microscope and suitable for the sensitive detection of the produced H 2 O 2 in coupled enzyme assays [12,27].
Another beneficial feature of GOX is its intrinsic fluorescence at visible wavelengths caused by the tightly bound cofactor flavin adenine dinucleotide (FAD) [28,29], which is used here for the label-free and spatially resolved detection of immobilized GOX. All other chemicals were purchased in analytical grade and used without further purification. Phosphate buffer was prepared by mixing 100 mM KH 2 PO 4 and 100 mM K 2 HPO 4 until the desired pH of 7.5 and 6.0 was reached and sterile filtered using a 0.22 µm syringe filter. Amplex Red (10-acetyl-3,7-dihydroxyphenoxazine; Cayman Chemical) was dissolved in DMSO and stored in aliquots at −20 • C. BSA (bovine serum albumin, AppliChem), PEG (polyethylene glycol, M ≈ 20 000 g/mol, Merck), and glucose (Merck) were dissolved in dH 2 O to give final concentrations of 1%, 2%, and 1 M, respectively. The solutions were filtered, aliquoted, and stored at −20 • C.

Electrodes
TiN ring electrode chips were fabricated by the IHP GmbH (Leibniz Institute for Innovative Microelectronics, Frankfurt Oder, Germany). The fabrication process is illustrated in [12]. The planar TiN ring electrodes have an outer diameter of 500 nm and are 20 nm wide. They are embedded in and filled with SiO 2 (Figure 1). There are 4 arrays on each chip. Each array consists of 6256 regularly arranged electrodes. Cross-shaped areas without electrodes facilitate orientation on the chip (see Figure 3A later). The electrodes of all arrays are electrically connected by an underlying metal layer and set to the same electric potential.

Dielectrophoresis
TiN ring electrode chips were mounted to microscope slides. The contact patch of the electrode arrays was electrically connected to a socket with a thin copper wire and carbon-based conductive glue (Leit-C, Plano). A cover glass with a conductive indium tin oxide layer (ITO, 70-100 Ω, SPI Supplies) served as the counter electrode that was also connected to the socket with a wire and conduc-tive glue. Enzyme solution of 2.3 µL was placed between the electrode array and ITO coated cover glass ( Figure 1). The HRP solution (c = 4 mg/mL, σ = 45 µS/cm) was used without further dilution. The GOX solution (c = 6 mg/mL, σ = 65 µS/cm) was diluted in dH 2 O to 3 mg/mL. A silicon spacer of 100 µm thickness (ELASTOSIL Film 2030 250/100, Wacker) with a hole (d = 4 mm) separated electrodes and counter electrodes, held the counter electrode in place, and prevented evaporation of the enzyme solution. An AC electric field with amplitude of 7 V RMS and a frequency of 10 kHz was applied for 10 min. In order to ensure comparability, the same settings as in previous reports were chosen. Voltage and frequencies in the same range were proven as effective to immobilize very different protein samples, including HRP [11,12], BSA [30], and also virus material [3] or even small dye molecules [4]. From theoretical considerations and empirical evidence, attraction to the electrode is expected for any protein if the frequency is below 1 MHz [9]. Alternating currents were generated and monitored as shown in Ref. [12]. After DEP, counter electrode and spacer were removed. The chip was shortly rinsed with dH 2 O and dried in a nitrogen stream. If required, the electrode arrays were microscopically imaged. Before the enzyme activity assays were started, the chip was incubated three times for 5 min with phosphate buffer and, expect for the experiment with HRP and short measurement cycles, another 5 min with phosphate buffer supplemented with 0.1% PEG and 0.025% BSA.

On-chip enzyme activity assays
For GOX activity assays using the reactions shown in Equations (1) and (2), a reaction solution consisting of 50 mM phosphate buffer pH 7.5, 50 mM Glucose, 0.1% PEG, 0.025% BSA, 1.52 × 10 −5 mg/mL HRP, and 25 µM Amplex Red was prepared. HRP and Amplex Red were added shortly before use. The SiO 2 surface in contact with the reaction solution was kept minimal in order to reduce the spontaneous generation of H 2 O 2 by contact electrification [32]. The solutions of 0.75 µL were placed onto the electrode arrays. The droplet was covered with a polyvinylchloride-acetate copolymer coverslip (product 2225-1, Ted Pella, Inc.) holding a 200 µm thick silicone spacer (ELASTOSIL Film 2030 250/200, Wacker). Because the errors of the used pipette for volumes <1.25 µL is high, a bright-field image of the droplet was taken to determine the exact reaction volume from the area occupied by the reaction solution. The increase in fluorescence intensity was observed using the 60× objective, filter set Cy3, LED 550 nm at 20% power, and 50 ms exposure time. To avoid unnecessary exposure leading to photooxidation [33], the luminous field diaphragm was closed as far as possible and images were acquired at 2.5 min intervals, using the microscope's acquisition software. GOX concentrations were calculated using the linear increase in resorufin fluorescence occurring between 20 and 30 min after initiation of the reaction on chip. A calibration curve for the GOX concentration was determined by adding known amounts of a GOX solution to the reaction solution before placement onto the chip. Before measurements with GOX were performed, three blank values were measured on each chip. HRP activity was measured on chip as reported in Ref. [12]. The reaction solution consisted of 26.5 µM Amplex Red, 88.2 µM H 2 O 2 , and 0.1% PEG in 100 mM phosphate buffer pH 7.5. The mixes of 1.5 µL were placed on a glass coverslip equipped with a 100 µm silicon spacer. The coverslip with reaction solution was placed onto the chip, which was immediately transferred to the microscope. The reaction was monitored with the 60× objective, closed luminous field diaphragm, filter set Cy3, and LED 550 nm (100%). Images were acquired automatically in 10 s intervals for 2 min. For calculating the HRP concentration, the slope in resorufin concentration within the first 60 s of the observed reaction was determined.

2.6
Fluorescence spectroscopy Spectra were measured using the fluorescence spectrometer PerkinElmer LS 55 in a quartz cuvette with a pathlength of 1 cm (105.253-QS, Hellma).

F I G U R E 2
Fluorescence spectra of 0.3 mg/mL glucose oxidase (GOX) in phosphate buffer pH 6 compared to the same buffer without enzyme. Excitation spectra: emission detected at 525 nm. Emission spectra: excitation at 450 nm.

Intrinsic fluorescence Of GOX immobilized by DEP
GOX was immobilized on TiN ring electrode arrays by the application of an AC electric field ( Figure 1). To detect the immobilized enzyme on the electrodes, the intrinsic FAD fluorescence of GOX was used. For the correct choice of emission and excitation filter for the fluorescence microscope, fluorescence spectra of GOX in solution were measured by the fluorescence spectrometer first. The maximum fluorescence excitation of GOX was at 450 nm and the fluorescence emission maximum at 525 nm ( Figure 2). Despite the high GOX concentration of 0.3 mg/mL, the fluorescence intensity was only about eight times higher than the signal which was detected in the buffer spectra. The apparent fluorescence of the buffer can be explained by Raman scattering of water. The highest intensity of the Raman peak of water is in the region of 3400-3500 cm −1 [34]. For excitation at 450 nm, this results in an emission peak at 533 nm, which corresponds well to the peak present in the buffer spectrum. Control measurements with dH 2 O in disposable cuvettes resulted in the same peak. To partly compensate for the low emission, an excitation filter which allowed to excite the fluorescence of GOX with two LEDs simultaneously (435 and 460 nm) was used in the microscope. For lower background by Raman scattering, an objective for imaging dry samples in air was used.
For the immobilization of GOX on TiN ring electrodes, the setup and parameters which had been established for HRP were applied (U = 7 V RMS ; f = 10 kHz; t = 10 min) F I G U R E 3 Bright-field and fluorescence micrographs of TiN ring electrodes with dielectrophoretically immobilized glucose oxidase (GOX). (A) Bright-field image of chip 1; (B) fluorescence micrograph of GOX on chip 1 after incubation without field; (C) fluorescence micrograph of GOX on chip 1 after immobilization by dielectrophoresis (DEP). Exposure time: 1 s. All images were taken in air. Similar results were obtained on all testes chips. [12]. After immobilization by DEP, a low, nevertheless detectable increase in fluorescence intensity on the electrodes compared to the background of the surrounding SiO 2 surface was found ( Figure 3C). Incubation of the same GOX solutions on electrode chips without any applied electric field resulted in no difference in fluorescence intensity between TiN electrodes and SiO 2 ( Figure 3B). Therefore, the fluorescence increase on the electrodes can be attributed solely to GOX that was immobilized by AC electrokinetic effects.
The fluorescence of FAD in GOX is low because of strong quenching by four aromatic amino acid residues [35,36]. The intrinsic fluorescence can reportedly be detected only for high concentrations of GOX [28,37]. Furthermore, the strong, but not covalently bound cofactor will be lost when GOX is denatured [29]. Nevertheless, the intrinsic fluorescence of GOX was successfully used to detect the immobilized enzyme ( Figure 3). This suggests that a rather high amount of protein was immobilized and at least a fraction of the immobilized GOX enzymes is not denatured. The fluorescence intensity on the electrodes is higher at the edge of the array than at electrodes located closer to the center. This indicates that more GOX is immobilized on electrodes closer to the edge than at the center of the arrays. A similar distribution was observed for BSA, polystyrene beads, and virus material on similar electrode geometries [30,38,3]. It is a result of the electrokinetic effects acting on the molecules. The DEP force that is attracting polarizable particles toward the electrodes is proportional to the square of the gradient of the electric field. Field simulations have shown that the gradient of the electric field is highest at the corners and outer rows of the array [30]. Besides DEP, electrohydrodynamic forces are acting on the particles or molecules. AC electroosmosis and electrothermal effects generate a fluid flow across the electrodes [13,39]. The flow direction is parallel to the electrode surfaces and from outside to inside the array at the edges at the array, but upward at the center of the array, favoring immobilization at the edges over the array center [3]. The combination of all effects leads to the observed gradient of fluorescence intensity, indicating an increasing amount of immobilized particles toward the edge of the array (Figure 3).
No immobilized GOX was detected after DEP with U = 3.5 V RMS , f = 10 kHz, and t = 10 min. The DEP force acting on the particles is proportional to the applied electric field squared. A field strength lower than a certain threshold value will cause the DEP force to be overcome by other effects, like Brownian motion, electroosmotic, and electrothermal flow [13]. This value will vary depending on the investigated type of particle and on electrode geometry. The volume of 3.5 V RMS is reported to be optimal for the immobilization of HRP in cylinder electrodes [11], but apparently is too low to immobilize GOX on ring electrodes.

Activity of GOX immobilized by DEP
To evaluate the activity and stability of GOX immobilized by DEP, a coupled enzyme assay was used ( Figure 4) 1.9 mg/L (56 µM) in a microchannel of 20 µm height [32].
To omit the high background at the start of the incubation time, only the linear increase in resorufin concentration between 20 and 30 min after application of the reaction solution was analyzed. A calibration curve was calculated from on-chip reactions containing known amounts of GOX in solution. The limit of quantification (LOQ) of the approach was calculated from blank measurements and samples containing low GOX concentrations. Equation (3) derived from Ref. [41] resulted in an LOQ of 0.07 pg GOX on the chip. As the blank values varied, three blank measurements were performed on each chip on its day of use. From those, a limit of detection (LOD) for the chip was calculated following Equation (4) as suggested in Ref. [42].

LOD =̄b lank + 3 ⋅ blank (4)
y blank is the mean of blank measurements, σ blank is the standard deviation of blank measurements, and σ low is the standard deviation of the lowest tested concentration.
The optimized GOX assay was applied to TiN ring electrode chips after incubation with GOX without field and to the same set of chips after immobilization of GOX by DEP. The incubation time for both procedures was 10 min. Incubation without field led to a detectable, but low GOX activity on the chip equaling <0.1 pg free GOX, which decreased to below the LOD in the course of 2-3 measurement cycles ( Figure 5; Table 1). The GOX activity after incubation without field can be attributed to nonspecific adsorption of enzyme to the SiO 2 surface of the chip, which is not permanent and leaches completely during a few measurement cycles. A similar behavior was reported for HRP incubated on the chip [12].
After DEP, the initial activity of immobilized GOX on the chip equaled 0.1-0.7 pg of free GOX. The activity of immobilized GOX was low, but higher than after nonspecific adsorption on all chips. It decreased strongly with each measurement and washing cycle ( Figure 5). After 6 washing cycles, GOX activity was still significantly higher than blank measurements on the chips, but below the LOQ on all chips. It can be concluded that active GOX can be immobilized on TiN ring electrodes. However, the activity of the permanently immobilized enzyme is smaller than that of 0.07 pg GOX in solution.
The expected amount of immobilized GOX for the case of a complete monolayer of active GOX molecules on all electrodes can be calculated from the surface area of the electrodes and the space occupied by each GOX molecule. Determining the surface coverage experimentally is not feasible with standard methods, for example, capacitive measurements, spectroscopic ellipsometry, or cyclic voltammetry [43][44][45], because the geometry and especially the low surface area of the used electrodes do not allow to detect significant signals by these means. Atomic force microscopy of immobilized BSA on a related electrode geometry and at very similar conditions showed full coverage of the electrodes [30], which is therefore also presumed for GOX on TiN ring electrodes. Additional layers of proteins immobilized by DEP are possible [30], but only the uppermost layer of immobilized enzyme molecules is expected to be fully active. The surface area of all electrodes on the chip is 7.36 × 10 −10 m 2 [12]. According to adsorption experiments followed by a quartz microbalance and crystallographic data, a densely packed monolayer of GOX molecules consists of 4.6-4.7 × 10 −12 mol/cm 2 [46,47]. The molecular weight of GOX is 160 kDa. Consequently, 5.53 pg of GOX would fit onto the TiN ring electrode arrays of one chip. After 3-6 measurement cycles, the activity of immobilized GOX was smaller than the activity of 0.07 pg free GOX ( Figure 5). An exact amount of permanently immobilized, active GOX cannot be calculated because only a few measured activities were above the LOQ. Still, it is <0.07 pg or <1.3% of the maximum activity expected for a full monolayer of active GOX. On all tested chips, the retained catalytic activity of GOX after immobilization is therefore at least one order of magnitude lower than that reported before for HRP [12].

Comparison with DEP-immobilized HRP
In Ref. [12] it was already shown that HRP is highly active after immobilization by DEP: Initially, up to 13 pg of active HRP were detected after DEP. During repeated measurement cycles, the activity decreased. Equations (5)-(7) were used for exponential fits of the resulting activity curves.
where N is the number of measurement and washing cycle; m 1 is the temporarily adsorbed HRP; k 1 is the washing rate of m 1 ; m 2 is the permanently immobilized HRP; m 3 is the inactivation or desorption rate of m 2 .
The best option for all successful immobilization experiments was Equation (7). The empirical model was interpreted as representing two fractions of immobilized HRP: m 1 is immobilized only temporarily, for example, by adsorption to SiO 2 or to permanently immobilized HRP, and is washed off exponentially with ongoing measurement cycles. m 2 is immobilized permanently and m 3 represents losses in a linear manner. The loss in activity may be caused by desorption, deactivation, changes in conformation or aggregation. The activity of the permanently immobilized fraction equaled 0.65-0.85 pg free HRP, which corresponds to up to 45% of the activity that was maximally expected for a full monolayer of HRP [12]. As the catalytic activity of HRP in solution is higher than that of GOX and the background activity was lower in the HRP assay, HRP activities were measurable in 2 min per measurement cycle, whereas each GOX activity measurement took 30 min. Therefore, the results for HRP and GOX stability on the chip cannot be compared directly: It is unclear whether inactivation or leaching of the enzymes is caused by the incubation in the reaction solution and, hence, dependent on the incubation time, or whether it is caused by the rinsing and drying steps between measurements and as such dependent on the number of measurement cycles.
To answer this question, the determination of HRP activity was now repeated in two variations ( Figure 6). HRP was again immobilized by DEP on TiN ring electrode chips. On chip 4, the activity was measured exactly as in Ref. [12] ( Figure 6A). Each activity measurement took 2 min. After short rinsing of the reaction solution and drying under an N 2 stream, the next measurement was started. On chip 5, HRP was immobilized in the same way. However, between the activity measurements the arrays were incubated for 27 min in 1.5 µL washing buffer which had the same composition as the GOX reaction buffer, except for the lack of substrates and HRP. Consequently, the measurement cycle for HRP on chip 5 was as long as the measurement cycle for GOX. Before the first measurement, HRP on chip 5 was incubated in washing buffer for 5 min additionally to the 3 × 5 min washing step in phosphate buffer, which was also the case for GOX measurements.
Short measurement cycles on chip 4 resulted in similar activities as reported in Ref. [12] ( Table 2). The initial activity equaled 4.5 pg HRP in solution and was reduced in each measurement cycle but was still higher than the LOQ of 0.07 pg after 8 measurements. The best fit of the data was again achieved with Equation (7). The activity of permanently immobilized HRP equaled 0.82 pg of dissolved HRP and was reduced by 0.08 pg in each measurement cycle. This result confirms the previously reported measurements.
With extended washing steps, the initial activity of HRP on chip 5 was reduced in each washing cycle as well ( Figure 6B). The activity was measurable until cycle 8. The best fit according to the corrected R 2 value was still the fit following Equation (7). The fit result implies an even higher activity of permanently immobilized HRP that equaled 1.9 pg HRP in solution (Table 2). But the resulting equation represented the data less convincing ( Figure 6B). Especially later measurements resulted in higher activities than assumed in the model equation. This indicates that the activity loss in extended measurement cycles follows a more complicated mechanism. Nevertheless, it is evident that immobilized HRP remains more active and more stable after DEP than GOX in an equivalent experimental protocol.
For the activity measurement with immobilized GOX, the exponential fits were omitted because the number of F I G U R E 6 Measured activity of horseradish peroxidase (HRP) on chip after passive adsorption (grey dots) and dielectrophoresis (DEP) (orange squares) on TiN ring electrodes. Green line: limit of quantification (LOQ). Red line: fit following Equation (7). (A) Standard procedure with short rinsing between measurements on chip 4, remeasured according to Ref. [12]. (B) Additional incubation in washing buffer for 5 min before first measurement and 27 min between measurements on chip 5. data points above the LOQ did not allow to perform reliable fits for multiple parameters. Instead, an upper limit of GOX activity of the permanently immobilized enzyme was found. It equaled the activity which would be expected for 0.07 pg GOX in solution. There are many possible reasons for the comparatively low activity of immobilized GOX. First of all, the high stability of HRP has to be taken into account. An electric field applied to the enzyme solution should cause ohmic heating. However, due to the extremely high surface-to-volume-ratio, heat will be TA B L E 2 Results from fits following Equation (7) for dielectrophoresis (DEP) immobilization and measurement cycles with and without additional incubation in washing buffer

Chip 4 5
Washing procedure before first measurement dissipated very efficiently. Following Baffou et al., a rough estimate based on geometry, field distribution, and buffer properties gives a local temperature increase of well below 10 −5 K, ruling out any impact by ohmic heating [48]. According to Bekard and Dunstan [23] the AC electric field itself can cause protein unfolding, denaturation and aggregation by friction forces provoked by the movement of the protein in the electric field [23]. As GOX is an example for enzymes with a rather flexible structure [44], this may be more of a concern for GOX than for HRP. It is known that living cells can be stretched in DEP systems [49]; hence, similar forces are to be expected also on proteins. For an efficient spatial manipulation of a protein its DEP associated potential energy should exceed thermal energy at that temperature [9]. If the difference between temperature at DEP field application and that needed for protein denaturation is small, the additional electrical energy input is expected to have a significant impact on the protein's structure and, hence, on its function. This means that even without any temperature increase DEP manipulation will have more impact on thermolabile molecules. The studied HRP is very thermostable with an active half-life of 117 min at 60 • C and 13-35 min at 80 • C [21,22], whereas the active half-life of GOX amounts to only 13 min at 60 • C, if no stabilizing compounds are provided [29]. As a result, the function of GOX has to be expected to be more affected by dielectrophoretic manipulation than that of HRP. The repeated drying of the surface may have contributed to inactivation or aggregation of the enzyme. Not only the rate of possible enzyme denaturation, but also the amount of immobilized protein affects the activity on the chip and can differ for the studied enzymes. For immobilization under DC conditions, the flexibility of the protein structure has an impact on how much protein will be adsorbed and whether the protein structure will be changed during adsorption, whereat the flexibility of GOX is again a dis-advantage [44,50]. The pI of the enzyme is also important for the adsorption caused by an applied potential [19,51].
As the pI values strongly differ with pI = 4.2 for GOX and pI = 8.8 for HRP [51], the effect of the same surface may be the opposite for GOX as compared to HRP. Notably, an electric potential applied to the surface can not only be used for immobilization but may even cause desorption or repulsion of proteins [52,53,18]. Although the immobilization of both enzymes on the electrodes was proven by fluorescence microscopy, the adsorption mechanism on the electrode surface under AC conditions is unclear and the amount of immobilized enzyme can be quite different for GOX and HRP. To distinguish whether the difference in enzyme activity is caused by GOX inactivation or by differences in the adsorption behavior of the enzymes, quantitative measurements of the total amount of immobilized protein, for example, by atomic force microscopy [30], would be necessary. Another limiting factor for the activity of the immobilized enzyme is their stability after the application of the electric field. As GOX is a homodimer and the cofactor FAD is not covalently bound, the lifetime of the immobilized enzyme will be limited by dissociation [54,29]. Dissociation is no reason of concern for monomeric HRP.
To make DEP immobilization of different enzymes applicable in research and biosensing, the method needs further optimization. Amplitude and frequency of the AC signal affect the amount of immobilized protein and may not be optimal [11,20]. Asymmetric wave forms may also be considered, as their application enabled the deposition of active enzymes even at less sharp electrode geometries [14,17]. Testing alternative electrode materials can lead to the immobilization of more protein as well [18]. Additives like NaCl, K 2 SO 4 , or lysozyme would enhance the stability of GOX [29], tough at the same time may cause new problems. Higher conductivities promote ohmic heating and electrothermal flow [55], whereas lysozyme is very likely immobilized by DEP as well and may compete with GOX for surface occupation. Therefore, crosslinking or encapsulation may be a more promising approach to stabilize the enzymes before immobilization [56,57].

CONCLUDING REMARKS
AC electric fields have been used to immobilize two enzyme species from aqueous solutions onto arrays of submicrometer sized ring electrodes of only 20 nm width. For convenience such experiments are often described and interpreted as being based on DEP. This might be true for geometrical dimensions of electrodes and particles in the higher micrometer or millimeter range. However, DEP theory calls for higher field gradients when smaller objects have to be manipulated, even more for the successful action on molecules in solution. Such high fields and field gradients lead to additional forces by, for example, AC electroosmotic flow originating in the electrodes' electric double layer, and buoyancy due to temperature gradients [13,39,55]. In a similar arrangement the immobilization pattern of virus particles being positioned primarily on the rim of electrodes suggests indeed DEP as the main force directed toward the electrodes [3]. With the present setup using thin rings instead of massive cylindrical electrodes the actual area covered by the electric double layer is reduced by 85%. This should also reduce the AC electroosmotic flow as well as heating leaving DEP as the main force. To prove the presence of the enzyme without any labeling, it was detected by means of its intrinsic FAD fluorescence. Fluorescence microscopy images of the electrode arrays showed that GOX was indeed immobilized at the location of the electrodes. Furthermore, activity measurements have shown that the activity of the immobilized enzyme equaled up to 0.7 pg of GOX in solution. Most of the enzyme was immobilized only temporarily or was inactivated in the course of further measurement cycles. The activity of permanently immobilized GOX was equal to or less than the activity of 0.07 pg GOX in solution. This corresponds to <1.3% of the activity of the amount of GOX which would fit onto the electrode surfaces of the array. It was confirmed that the activity of HRP immobilized in the same way is higher and more stable, even when measurement cycles were extended to match the conditions of GOX activity measurements. Possible causes for these differences are structural changes of the immobilized molecules induced by the electric field or by surface interactions, or just smaller amounts of immobilized molecules. Either way, this study shows that for a sensible DEP induced immobilization of molecules from solution their physico-chemical properties have to be taken into account in more detail.

A C K N O W L E D G M E N T S
We gratefully acknowledge funding by the European Regional Development Fund (ERDF) and by the Brandenburg Ministry of Science, Research and Cultural Affairs (MWFK) within the framework StaF. We also thank Christian Wenger from the Leibniz Institute for Innovative Microelectronics for providing the TiN ring electrode chips.
Open access funding enabled and organized by Projekt DEAL.

C O N F L I C T O F I N T E R E S T S TAT E M E N T
The authors have declared no conflict of interest.

D ATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.