Effect of carbon microfiber materials on sensitivity of adenosine and hydroxyadenine at carbon microfiber sensors

2.1 Reagents and methods

All chemicals were of the analytical reagent grade and were used as received. 6-amino-1H-purine-2,8-dione i.e., (2,8-dihydroxyadenine) and adenosine were obtained from Sigma-Aldrich, St. Louis, MO. All other chemicals were obtained from Fisher Scientific (Pittsburgh, PA). Solutions were prepared in doubly deionized water daily before experiments. Phosphate buffer, pH 7.4, contained sodium phosphate monobasic monohydrate (NaH₂PO₄.H₂O) and dibasic anhydrous (Na₂HPO₄) at a total concentration of 31 mM was used. The pH of solutions was adjusted with NaOH or HCl before experiments.

Adenosine (ADO) and 2,8-dihydroxyadenine (2,8-DHA) were prepared before the experiments in 31 mM phosphate buffer (pH 7.4). Potassium ferricyanide was prepared in 0.5 M KCl (pH 6.0). All determinations were performed at room temperature. The percentage of CO₂ in air of 0.03% and of oxygen of 21% was not changed and the solutions were not purged with gases. The sensitivity of each analyte was determined from the slopes of the calibration curves. At least 8 points for each calibration curve and 4 determinations at each concentration were obtained at three different electrodes. The results obtained at different CFEs were pooled. All other measurements were repeated at least in triplicate. The reported results reflect the reproducibility of the measurements and the reproducibility of fabrication of carbon fiber sensors.

2.2 Instrumentation

The instrumental set up has been described previously (Kathiwala et al., 2010). Bioanalytical Systems Electrochemical Analyzer (BAS-100, West Lafayette, IN) with a home-built preamplifier was interfaced to a PC and was used in slow scan cyclic voltammetry. A two-electrode configuration was used, with CFE as working and SCE as reference. In slow scan voltammetry, the potential scan rate was (50 mV s⁻¹); the potential window was −0.2 to 0.5 V for ferricyanide.

Fast-scan cyclic voltammetry (FSV) was performed with signal averaging and with digital background subtraction (Brajter-Toth et al., 2000). Briefly, a triangular waveform was applied with a function generator (Universal Programmer, Model 175, EG&G Princeton Applied Research, Princeton, NJ, USA) to a SCE in a two electrode configuration. Current at CFE was converted into voltage and amplified by a home-built current-to-voltage converter and was recorded and stored by a digital oscilloscope (LeCroy Model 9310, Chestnut Ridge, NY, USA). A copper mesh Faraday cage was used to minimize noise.

The solutions were injected into ca. 80 μL electrochemical cell with a syringe and a permanent electrode-solution contact was maintained. In FSV, the background current was recorded in buffer first, before each analytical signal was recorded, and was signal averaged and stored by the digital oscilloscope. The analytical signal was obtained immediately afterward, and was signal averaged and stored by the digital oscilloscope in a separate channel. The stored background current was subtracted from the stored analytical signal using the digital oscilloscope. The background subtracted analytical signal was transferred from the digital oscilloscope to a personal computer. ADO and 2,8-DHA signals were recorded in the same potential window as the background current from −1.0 to 1.5 V, under the same experimental conditions. Number of cycles in FSV was 250 (Hsueh and Brajter-Toth, 1996). In FSV the potential scan rate was 500 Vs⁻¹. Data obtained from FSV were analyzed with Origin (Microcal Software, Inc. Northampton, MA) using a PC.

2.3 Electrodes

A saturated calomel electrode (SCE) was used as a reference electrode and carbon fiber microdisk electrodes (CFMEs) were used as working electrodes. Fabrication of carbon fiber microdisk electrodes has been described (Brajter-Toth et al., 2000; Bravo et al., 1998).

PAN T650 (PAN I), Pitch P25 (PT) (Cytec Engineered Materials, Greenville, SC), and PAN HCB (PAN II) (Textron Specialty Materials, Lowell, MA) fibers were used to fabricate the microdisk working electrodes (carbon assay % is 94, 99.5 and 99+, respectively). The fiber (ca. 3 cm long) was connected to copper wire (ca. 6 cm long, 0.3 mm diameter, Fisher Scientific, Pittsburgh, PA) with electrically conductive silver epoxy (E4110 part A, Epoxy Technology, Billerica, MA) mixed with 40 wt.% of curing agent (hardener) (E4110 Part B).

Approximately 1 cm of the fiber and wire was covered by the epoxy mixture, and the assembly was cured at room temperature for 48 h. The copper wire end of the electrode assembly was drawn into a 200 mL micropipette tip (5 cm long) so that ca. 1 cm of the fiber was drawn in. The pipette tip was filled with a hot (80–90 °C) mixture of epoxy (EPON Resin 828, Miller-Stephenson Chemical Co., Danbury, CT) with 10 wt.% m-phenylenediamine flaked hardener (DuPont Specialty Chemicals, Wilmington, DE). To obtain a good quality seal, the resin was cured for 72 h at room temperature. The sealed electrodes were further cured at 90 °C for 1 h to improve their mechanical properties.

The cross sectional surface of the fiber was exposed by touch-polishing with 600 grit SiC polishing paper (Mark V Laboratory, East Granby, CT), mounted on a polishing wheel (Ecomet I, Buehler Ltd, Evanston, IL) with doubly distilled water as a lubricant.

Electrodes were held at approximately the right angle to the surface of the polishing paper in order to obtain a disk-shaped surface.

Because of the well-known redox properties and limited adsorption (Abou El-Nour and Brajter-Toth, 2000), potassium ferricyanide was used in the determination of the apparent electrode radius, r. The radius was determined from the limiting current of ca. 5 mM $\mathit{Fe}{(\mathit{CN})}_6^{3-}$ in 0.5 M KCl by cyclic voltammetry at 50 mV s⁻¹, from 0.5 to −0.2 V, using Eq. (1) for steady-state disk current (Bard and Faulker, 2004).

The limiting current of ferricyanide was additionally measured to verify the reproducibility of fabrication of carbon fiber microdisk electrodes from log-plot analysis of the i–E curves of ferricyanide and from the value of E^0′ determined from E^1/2 (Abou El-Nour and Brajter-Toth, 2000).

Nanostructured microdisk electrode surfaces were obtained by an electrochemical pretreatment (ECP) method, as previously described (Kathiwala et al., 2010).

Briefly, this involves continuous potential cycling from −1.0 to 1.5 V, at 10 V s⁻¹ for 30 min in 31 mM phosphate buffer, pH 7.4. Electrodes that produced fast kinetics of $\mathit{Fe}{(\mathit{CN})}_6^{3-}$ , determined from the analysis of E vs. log[(i_L − i)/i] plots (Abou El-Nour and Brajter-Toth, 2000) and well-developed steady-state voltammograms were used for FSV measurements of ADO and 2,8-DHA at 500 V s⁻¹. Electrodes, which had excessively noisy background current at 500 Vs⁻¹ and produced irreproducible background subtracted signals, were discarded.

The CFE quality was additionally confirmed from the measurement of the electrode capacitance (Bravo and Brajter-Toth, 1999; Brajter-Toth et al., 2000).

Apparent specific electrode capacitance, C (μF cm⁻²), was determined before and after ECP, and after FSV measurements, from the background current at 0.75 V (Bravo and Brajter-Toth, 1999) in 31 mM phosphate buffer pH 7.4 at 10 V s⁻¹ . Eq. (2) was used to calculate the apparent electrode capacitance. It was assumed that the background current at 0.75 V results predominantly from double-layer charging, with limited contributions of surface faradaic reactions (Brajter-Toth et al., 2000).

2.4 Fast scan voltammetry

FSV measurements were performed as described before (Hsueh et al., 1997). The sequential acquisition of background and analyte signal, in the same potential window and at the same potential scan rate, has been optimized to minimize irreversible surface reactions (Brajter-Toth et al., 2000; Kathiwala et al., 2008, 2010). This produces reproducible signals and stable electrodes that can be reused, and allows accurate digital background subtraction (Brajter-Toth et al., 2000; Kathiwala et al., 2008, 2010).

Signal averaging was used and the number of scans recorded was a compromise between high signal-to-noise (S/N) ratio and measurement time (Hsueh et al., 1997). The reported values are averages of measurements for at least four electrodes. Geometric area was used to obtain normalized values of anodic peak current of probes and background current at electrodes.

2.5 Analytical determinations

The sequential acquisition of the background and the analyte currents, in the same potential window and at the same potential scan rate, facilitates accurate digital background subtraction (Brajter-Toth et al., 2000; Hsueh et al., 1997). This procedure was repeated for all calibrations in the measurement ranges. ADO and 2,8-DHA were determined in 31 Mm phosphate buffer, pH 7.4, by FSV at 500 V s⁻¹.

Sensitivity is reported as the average slope of the calibration curve for at least 8 concentrations and 4 determinations at each concentration at three different carbon microfiber electrodes. Data were processed using Origin software (Origin Lab, Northampton, MA).

3.1 Electrochemical measurements of carbon fiber electrodes

3.1.1

3.1.1 Response to ferricyanide

Electrochemical pretreatment (ECP) methods (Takmakov et al., 2010; Brajter-Toth et al., 2000; Kathiwala et al., 2008, 2010) are used for etching the carbon fiber electrode surface and consequently forming surface nanocracks and edge nanostructures. These methods can lead to irreversible side reactions such as electrolysis of water, production of carbon dioxide and the formation of non-conducting surface oxides. These reactions can contribute to irreversible behavior of carbon fiber electrodes (Brajter-Toth et al, 2000; Kathiwala et al., 2008, 2010; Goss et al., 1993). However, solution composition, potential window, potential scan rate and waveform, used during surface nanostructured formation by ECP, can be optimized to control these reactions (Brajter-Toth et al, 2000; Bravo et al., 1998).

Electron transfer kinetics of ferricyanide are determined by electrode surface structure and are used to detect electrode activity (McCreery, 2008; Brajter-Toth et al, 2000). Fig. 1 shows cyclic voltammograms of 5 mM $\mathit{Fe}{(\mathit{CN})}_6^{3-}$ at PAN I, PAN II and PT carbon microfiber electrodes in 0.5 M KCl. Slow electron transfer kinetics of ferricyanide are typically observed at polished and non-treated carbon fiber electrodes (Fig. 1, solid line), in spite of carbon fiber material. This is apparent from the slow rise in ferricyanide current and the poorly-defined limiting current plateau of ferricyanide (Fig. 1, solid line). On contrary ferricyanide kinetics are fast at electrodes modified by ECP (Fig. 1, small dash). This is apparent from higher E_1/2 values and well-defined plateaus of the i–E curves (Fig. 1, small dash). As mentioned before (Brajter-Toth et al, 2000; Kathiwala et al., 2008, 2010) nanostructured carbon microfiber electrodes formed by electrochemical nanostructured of the surface by ECP have a stable background current and the electrode surface can be easily renewed by FSV.

Close

Figure 1

Cyclic voltammetry of different microfiber electrodes in 5 mM K₃Fe(CN)₆, 0.5 M KCl, 50 mV s⁻¹, 31 mM phosphate buffer pH 7.4, before (solid line) and after (dotted line) electro-chemical pretreatment of electrode surface.

Export to PPT

However, after the ECP of the electrode surface, differences in response to ferricyanide are observed at the nanostructured microdisk electrodes made from different carbon fiber materials.

After ECP as shown in Fig. 1 (small dash), at PAN I carbon fiber electrodes, the limiting current of ferricyanide decreases while it increases at PAN II and PT carbon microfiber electrodes.

These differences can be attributed to the carbon fiber material properties since the fabrication conditions of electrode surface are the same. Electrodes from the two types of PAN fibers (I and II) have a similar geometric radius (Table 1). However, carbon assay fibers are relatively low 94% for PAN I while it equals 99.5% in case of PAN II fibers, also the electrical conductivity of PAN I is lower than that of PAN II (667 and 909 S cm⁻¹, respectively) (Laffont et al., 2004).

The conductivity, material disorder and crystalline structure of PAN-based carbon microfibers have been confirmed by different methods of analysis (Kim et al., 2004). From these analyses it was found that lower carbon assay indicates more amorphous structure of PAN I fibers. Less ordered fiber structure facilitates oxidation (Pamula and Rouxhet, 2003) of PAN I carbon fiber electrode surfaces during surface nanostructured formation by ECP.

Great ability of less crystalline carbon fibers to oxidation has been verified by X-ray photoelectron spectroscopy (Pamula and Rouxhet, 2003) and hence greater surface oxide coverage is expected. Ferricyanide access to oxide rich surfaces can be repressed (Cheng and Brajter-Toth, 1996) as observed here. Consequently, PAN I, PAN II and PT carbon microfiber electrodes, although fabricated by the same method, will have different surface structures and chemistry, which results from different properties of the respective carbon fiber materials.

The electron transfer kinetics of ferricyanide at polished and non-treated PAN I carbon fiber electrodes are faster than at electrodes made from PAN II and PT carbon fibers, as indicated by more positive E_1/2 values (Table 1) and lower slope of E vs. log[(i_L − i)/i] plots for ferricyanide at these microelectrodes. Faster kinetics at polished and non-treated PAN I electrodes are consistent with greater material disorder of PAN I carbon fiber material (Cheng and Brajter-Toth, 1996). Kinetics of ferricyanide is even faster at these electrodes after the surface is nanostructured by ECP. This is verified by the positive shift in E_1/2 and the lower slope values of log-plots that were obtained from the slow scan voltammetry i − E curves of ferricyanide which had slope close to 59 ± 0.1 mV at all electrodes, that were made from different carbon fiber materials which is expected for a one-electron reversible reaction (Bard and Faulker, 2004). After ECP, E_1/2 values are close to E^0′ of 210 mV (vs. SCE) which has been reported for $\mathit{Fe}{(\mathit{CN})}_6^{3-/4-}$ in 0.5 M KCl (Bard, 1928). Thus, electrode kinetics of ferricyanide, after ECP of the electrodes, is independent of the electrode carbon fiber material (Whitson et al., 1973).

3.1.2

3.1.2 Background current and electrode capacitance

It was reported before that high electrode background current reflects high electrode surface area of nanostructured carbon fiber electrodes (Brajter-Toth et al., 2000; Kathiwala et al., 2008, 2010). Fig. 2 confirms this at electrodes made from different carbon fiber materials. It is observed from Fig. 2 that the lowest background current is shown at PAN I electrodes, which is due to their small geometric radius. The specific capacitance calculated from the normalized background currents using Eq. (2) is similar at electrodes fabricated from PAN I and PAN II materials and is higher at PT electrodes, which have a larger geometric radius. (Table 1).

Close

Figure 2

Voltammetric background current at different carbon microfiber electrodes before (solid line) and after (dotted line) electrochemical pretreatment at pH 7.4 in 31 mM phosphate. Scan rate 10 V s⁻¹, 50 cycles averaged.

Export to PPT

Electrodes from the two types of PAN fibers have similar surface roughness factors due to their similar (apparent) specific capacitance (Table 1), although carbon content and electrical conductivity of PAN II fibers are higher than those of PAN I fibers.

The differences in surface roughness reflect the differences in the properties of different carbon fiber materials such as carbon assay, surface area, and density.

3.2 Electrode reaction of ADO and 2,8-DHA

The oxidation of adenosine proceeds in a mechanism involving a 6e⁻, 6H⁺ process to give 2-hydroxyadenosine and 2,8-dihydroxyadenosine and then the corresponding diimine Scheme 1. .

Close

Scheme 1

Reaction mechanism proposed for the electrooxidation of adenosine Goyal et al. (1991)

Export to PPT

Also, 2,8-dihydroxyadenine is oxidized in a 2e⁻, 2H⁺ step to give the corresponding diimine Scheme 2.

Close

Scheme 2

Proposed electrochemical oxidation pathway of 2,8-dihydroxyadenine Kathiwala et al. (2010).

Export to PPT

Formation of the diimine and its short half-life (∼50 ms) in case of various purines at different electrodes has already been documented in the literature (Wopshall and Shain, 1967; Brown and Large, 1964). The oxidation potentials of hydroxypurines have been found to be less positive in comparison with simple purines (Lingane, 1966).

It appears that 2,8 dihydroxyadenosine and 2,8-DHA undergo 1e⁻, 1H⁺ oxidation to give a radical species which rapidly dimerizes with another similar moiety to form dimer as is shown in Schemes 1–2. Oxidation of 8-hydroxypurines in a 1e⁻, 1H⁺ step to form a variety of dimers has been documented in the literature (Furniss et al., 1994; Swern, 1971).

Also, for adenosine it was reported before (Goyal et al., 1991) that one of the ribose units of the initial O–O dimer is removed due to steric hindrance in the dimer during its formation. Hence, one of the ribose units is hydrolyzed to give another dimer (Scheme 1).

As shown in Fig. 3, oxidation of 10 μM ADO results in a well-developed oxidation peak observed at ca. 0.98 V and a shoulder at ca. 0.65 V in the case of PAN I microfiber electrode. The observed peaks shifted to more positive potentials at ca.1.0 V and 0.7 V, respectively in the case of PAN II microfiber electrode, while the oxidation of ADO at a PT microfiber electrode results in one broad large peak at ca. 0.97 V and a shoulder at 0.68 V.

Close

Figure 3

Fast scan cyclic voltammograms of 10 μM Adenosine at different carbon microfiber electrodes, in 31 mM phosphate buffer pH 7.4, 500 V s⁻¹, 250 cycles, results after background subtraction.

Export to PPT

On the other hand, from Fig. 4, oxidation of 10 μM 2,8-DHA produces one broad large peak at ca. 0.90 V, small peak at 0.52 V and a shoulder at 0.35 V at PAN I microfiber electrode. Multiple small peaks were observed in the case of PAN II and PT microfiber electrodes. At PAN II small peaks were observed at ca. 1.0, 0.87, 0.71, 0.57 and 0.23 V, respectively as well as a shoulder at 0.41 V. In PT microfiber electrode the multiple small peaks shifted to less positive potentials and appeared well-developed at ca. 0.8, 0.63, 0.47, 0.32, 0.15 V, respectively.

Close

Figure 4

Fast scan cyclic voltammograms of 10 μM 2,8-dihydroxyadenine at different carbon microfiber electrodes, in 31 mM phosphate buffer, pH 7.4, 500 V s⁻¹, 250 cycles, results after background subtraction.

Export to PPT

Peaks observed at ca. 0.98 V in ADO voltammograms and at ca. 0.8 V were measured to obtain ADO and 2,8-DHA calibration curves, respectively (Abou El-Nour and Brajter-Toth, 2000; Kathiwala et al., 2010).

Current traces between 0 and 0.1 V and ca. 1.2 V (Figs. 3 and 4) are likely due to products resulting from the incomplete background subtraction.

Reduction of diimine is likely responsible for the reduction peaks between ca. −0.27 V and −0.36 V Figs. 3 and 4. The reduction of diimine followed by reoxidation process is quite fast as shown by the near reversibility of the oxidation/reduction peaks in the voltammograms of ADO and 2,8-DHA.

Large surface area of nanostructured carbon fiber electrodes favors adsorption. Peak potentials for the forward and reverse processes are close, suggesting adsorption of both ADO and 2,8-DHA. The proton-coupled electron transfer of ADO and 2,8-DHA at pH 7.4 can aid interactions with surface oxides. High analytical sensitivity of ADO and 2,8-DHA reflects high adsorption behavior at carbon microfiber electrodes which consecutively reveals structure of the electrode surfaces (Maldonado et al., 2006; O’Mullane et al., 2008). High anodic peak currents of ADO and 2,8-DHA at 500 V s⁻¹ indicate adsorption-controlled processes.

The relative magnitudes of the peak currents for ADO at different electrodes are PAN I ≈ PAN II > PT indicating more adsorption at PAN electrodes (Fig. 3). 2,8-DHA adsorbs more than ADO at PAN I electrodes as the magnitudes of the peak currents at different microfiber electrodes are in the order PAN I > PAN II > PT (Fig. 4). The results show that 2,8-DHA adsorbs highly at more disordered PAN I than at PAN II and PT electrodes.

Adsorption of 2,8-DHA at PAN electrodes is higher than that of ADO at the same electrodes this may be related to the steric hindrance due to the existence of ribose unit in the ADO structure.

3.3 Electrode sensitivity

Low adsorption and sensitivity of ADO and 2,8-DHA at pitch microfiber electrodes (Figs. 3 and 4 and Table 1), are expected as a result of small change in electrode surface results from ECP of these microfiber electrodes (Table 1). Although the two types of PAN electrodes have similar surface areas after ECP, PAN II electrodes are significantly more sensitive in ADO and 2,8-DHA measurements. This may be due to high material disorder (defects) of the surface of PAN I which facilitates over oxidation of the electrode surface, generating greater surface oxide coverage and subsequently less adsorption of ADO and 2,8-DHA.

The higher sensitivity of PAN II electrodes matches lower defects and higher electrical conductivity of PAN II fibers (909 S cm⁻¹) than PAN I (667 S cm⁻¹).

Slower kinetics at PAN I than at PAN II electrodes reflect higher material defect and expected higher degree of surface oxidation of PAN I.

Methods of carbon fiber electrode surface treatment/reactivation used here, which likely produce defects and cation-exchanged surface oxides, ultimately lead to high sensitivity in ADO and 2,8-DHA determinations by fast-scan voltammetry (Table 1).

ADO and 2,8-DHA have faster kinetics at more ordered, and presumably less oxidized PAN II carbon fiber electrodes which correlate with more adsorption and greater sensitivity. Thus, structure and composition of carbon fiber materials influence surface interactions of the microfiber electrode.