Anal. Chem. 1997, 69, 3552-3557
Direct Electrochemical Detection of Purine- and Pyrimidine-Based Nucleotides with Sinusoidal Voltammetry Pankaj Singhal and Werner G. Kuhr*
Department of Chemistry, University of California, Riverside, California 92521
Sinusoidal voltammetry was employed to detect both purine- and pyrimidine-based nucleic acids. Adenine and cytosine, representing these two classes of nucleic acids, could be measured with submicromolar detection limits at a copper electrode under these conditions, where the sensitivity for adenine was much higher than that for cytosine. Detection limits for purine-containing nucleotides [e.g., adenosine 5′-monophosphate (AMP), adenosine 5′-diphosphate (ADP), and adenosine 5′-triphosphate (ATP)] were on the order of 70-200 nM using this method. These detection limits are achieved for native nucleotides and are over 2 orders of magnitude lower than those found with UV absorbance detection. Submicromolar detection limits were also obtained for pyrimidine-based nucleotides, which could also be detected with high sensitivity due to the presence of a sugar backbone that is electroactive at the copper surface. This detector is not fouled by the nucleotides and may be used for the sensitive detection of analytes eluting continuously in a flowing stream, i.e., from a chromatography column or an electrophoresis capillary. The sensitive measurement of DNA is of primary importance due to its preeminent biological significance. The knowledge of the structure of DNA and its interactions with other biological compounds like proteins and other small molecular weight compounds can lead to advances in pharmacology and also to the prevention of many diseases such as cancer, sickle-cell anemia, and cystic fibrosis.1-5 Generally, the sequence of a DNA molecule is determined after it is digested by using a restriction endonuclease, to break down the large DNA strand into component fragments and/or nucleotides. The nucleotide mixture obtained after this digestion is separated by chromatographic or electrophoretic methods and detected after labeling with radioisotopes or fluorescent tags.6 Alternatively, new technologies employing exonucleases, which cleave off the terminal nucleotide, allow the sequence of DNA to be determined sequentially using flow cytometry.7-9 Detection and identification of each successive nucleotide as it is released from exonuclease digestion in this manner can provide the sequence information for that DNA strand. (1) Mikkelson, S. R. Electroanalysis 1996, 8, 15-19. (2) Conner, B. J.; Reyes, C. M.; Morin, C.; Itakura, K.; Teplitz, R. L.; Wallace, R. B. Proc. Natl. Acad. Sci. U.S.A. 1983, 80, 278. (3) Landegren, U.; Kaiser, R.; Caskey, C. T.; Hood, L. Science 1988, 242, 229. (4) Millan, K. M.; Mikkelsen, S. R. Anal. Chem. 1993, 65, 2317-2323. (5) Millan, K. M.; Saraullo, A.; Mikkelsen, S. R. Anal. Chem. 1994, 66, 29432948. (6) Smith, L. M.; Brumley, R. L., Jr.; Buxton, E. C.; Giddings, M.; Marchbanks, M.; Tong, X. Methods Enzymol. 1996, 271, 219-237.
3552 Analytical Chemistry, Vol. 69, No. 17, September 1, 1997
Many detection schemes have been reported for the quantitative analysis of nucleotides and DNA. Simple optical detection schemes, such as UV absorbance10 and fluorescence detection, have been used extensively when coupled with separation methods.7,8,11-13 Unfortunately, UV methods suffer from their poor sensitivity (detection limits in the 10-4-10-5 M concentration range). Fluorescence methods, while much more sensitive, rely on either the intercalation of a fluorophore within double-stranded DNA or the derivatization of oligonucleotides with a fluorophore, which leads to sample contamination and multiple labeling problems.14 Direct electrochemical detection of the nucleobases, adenine and guanine, is possible at mercury,15-20 gold,21 copper,22 and carbon23,24 electrodes, where these bases were oxidized at extremely positive potentials.23 Although these methods are sensitive for nucleic acid bases, high backgrounds and irreversible adsorption of larger molecules led to poorer sensitivity for this approach for the analysis of nucleosides, nucleotides, and DNA. In particular, mercury,15,20,25 gold,21 and carbon20,24,26 surfaces were completely fouled by the adsorption of oligonucleotides and DNA strands. Several investigators subsequently exploited this adsorptive tendency of nucleic acid bases, oligonucleotides, and DNA to obtain very sensitive electrochemical detection schemes. These schemes involved the adsorption of the nucleic acid onto the electrode surface in order to concentrate them and then employed (7) Goodwin, P. M.; Johnson, M. E.; Martin, J. C.; Ambrose, W. P.; Marrone, B. L.; Jett, J. H.; Keller, R. A. Nucleic Acids Res. 1993, 21, 803-806. (8) Huang, Z.; Petty, J. T.; O’Quinn, B.; Longmire, J. L.; Brown, N. C.; Jett, J. H.; Keller, R. A. Nucleic Acids Res. 1996, 24, 4202-4209. (9) Petty, J. T.; Johnson, M. E.; Goodwin, P. M.; Martin, J. C.; Jett, J. H.; Keller, R. A. Anal. Chem. 1995, 67, 1755-1761. (10) Holden, M.; Pirie, N. W. Biochim. Biophys. Acta 1955, 16, 317-321. (11) Kapuscinski, J.; Skoczylas, B. Anal. Biochem. 1977, 83, 252-257. (12) Le Pecq., J. B.; Paoletti, C. Anal. Biochem. 1966, 17, 100-107. (13) Markovits, J.; Roques, B. P.; Le Pecq, J. B. Anal. Biochem. 1979, 94, 259264. (14) Karsten, U.; Wollenburger, A. Anal. Biochem. 1977, 77, 464-470. (15) Palecek, E. Nature 1960, 188, 656-657. (16) Palecek, E. Anal. Biochem. 1980, 108, 129-138. (17) Palecek, E.; Postbieglova, I. J. Electroanal. Chem. 1986, 214, 359-371. (18) Palecek, E.; Boublikova, P.; Jelen, F. Anal. Chim. Acta 1986, 187, 99107. (19) Palecek, E.; Fojta, M. Anal. Chem. 1994, 66, 1566-1571. (20) Palecek, E. Electroanalysis 1996, 8, 7-14. (21) Pang, D.-W.; Qi, Y.-P.; Wang, Z.-L.; Cheng, J.-K.; Wang, J.-W. Electroanalysis 1995, 7, 774-777. (22) Lin, H.; Xu, D. K.; Chen, H. Y. J. Of Chromatogr., A 1997, 760, 227-233. (23) Nakahara, T.; Okuzawa, M.; Maeda, H.; Hirano, M.; Matsumoto, T.; Uchimura, H. J. Chromatogr. 1992, 15, 1785-1796. (24) Kafil, J. B.; Cheng, H. Y.; Last, T. Anal. Chem. 1986, 58, 285-289. (25) Palecek, E. In Topics in Bioelectrochemistry and Bioenergetics; Milazzo, G., Ed.; John Wiley & Sons: New York, 1983; Vol. 5, pp 65-157. (26) Wang, J.; Cai, X.; Jonsson, C.; Balakrishnan, M. Electroanalysis 1996, 8, 20-24. S0003-2700(97)00333-8 CCC: $14.00
© 1997 American Chemical Society
stripping voltammetric procedures to analyze the adsorbed analyte.16-18,20,24-28 DNA has also been electrochemically detected after hybridization to complementary strands immobilized on the electrode surface. Electroactive moieties can be used in several different ways to label double-stranded DNA. Intercalators bind internally in the double-stranded DNA formed at the surface of the electrode, allowing detection of the increased current at the electrode surface due to these species.4,5 Alternatively, electrostatic binding of cationic species can occur after intercalation or external binding of the electroactive molecule to DNA, where it can be monitored electrochemically or with electrogenerated chemiluminescence.29 All these methods, however, work on a batch process level, since they require the adsorption of nucleic acids and/or their components on the electrode surface for a long period of time (tens of seconds to 10-15 min). Therefore, they are not suitable for flow-through detection schemes, such as those that can be coupled to separation methods like liquid chromatography and capillary electrophoresis. Since all nucleotides contain a ribose sugar (present in addition to the base and a phosphate ester), it should be possible to detect nucleotides through the oxidation of the sugar backbone. One of the most common electrochemical methods for the detection of sugars utilizes the oxidation of the sugar moiety at a metal electrode with pulsed amperometric detection (PAD).30 Surprisingly, there have been no reports of the electrochemical measurement of nucleotides using PAD. Alternatively, the amperometric detection of sugars has also been reported at copper electrodes at dc potentials via an electrocatalytic mechanism.31-37 The use of a copper electrode also minimizes the possibility of fouling of the electrode surface, since the Cu(II) layer is soluble in high-pH buffer, and thus the oxidation of sugars and amines does not cause any fouling of the copper surface. Neither PAD nor dc amperometric methods provide qualitative information about the identity of the molecule. Traditionally, this can only be accomplished with voltammetric measurements. Unfortunately, voltammetric techniques give poorer detection limits even compared to UV absorbance detection due to the high background charging currents observed when the electrode surface is scanned. Therefore, conventional voltammetric methods have not been very useful for nucleotide analysis. Sinusoidal voltammetry (SV), an analog of which was originally described as oscillographic polarography by Heyrovsky in 1943,38 uses a large amplitude sinusoid as the excitation wave form. While the work of Heyrovsky and contemporaries used a sinusoidal current and monitored the resulting potential on an oscilloscope, there are many similarities to the response observed with linear scan potential scans used in “modern” cyclic voltammetry. Indeed, Palacek utilized oscillographic polarography for (27) Wang, J. E. A. Anal. Chem. 1996, 68, 2629-2634. (28) Palecek, E. Anal. Biochem. 1988, 170, 421-431. (29) Xu, X. H.; Bard, A. J. J. Am. Chem. Soc. 1995, 117, 2627-2631. (30) Johnson, D. C.; LaCourse, W. R. Anal. Chem. 1990, 62, 589A-597A. (31) Colon, L. A.; Dadoo, R.; Zare, R. N. Anal. Chem. 1993, 65, 476-481. (32) Kano, K.; Ueda, T. J. Electroanal. Chem. 1994, 372, 137-143. (33) Luo, P.; Zhang, F.; Baldwin, R. P. Anal. Chim. Acta 1991, 244, 169-178. (34) Prabhu, S. V.; Baldwin, R. P. Anal. Chem. 1989, 61, 2258-2263. (35) Vassilyev, Y. B.; Khazova, O. A.; Nikolaeva, N. N. J. Electroanal. Chem. 1985, 196, 127-144. (36) Singhal, P.; Kawagoe, K. T.; Christian, C. N.; Kuhr, W. G. Anal. Chem. 1997, 69, 1662-1668. (37) Xie, Y.; Huber, C. O. Anal. Chem. 1991, 63, 1714-1719. (38) Heyrovsky, M.; Micka, K. In Electroanalytical Chemistry: A Series of Advancements; Bard, A. J., Ed.; Marcel Dekker, Inc.: New York, 1967; Vol. 2, pp 193-256.
the analysis of the nucleobases in DNA at mercury electrodes in 1960.15 Recently, we reported an implementation of this methodology which, combined with analysis in the frequency domain,39 effectively decouples the background charging current from the faradaic current in the frequency domain.36,40 This is accomplished by capitalizing on the inherent difference between charging and faradaic currents. The background or charging currents are mostly linear and, therefore, are present primarily at the fundamental excitation frequency. The faradaic currents are essentially nonlinear at fast scan rates and thus have significant components even in the higher harmonics. By utilizing a sinusoidal excitation wave form, the charging current can be effectively isolated from the faradaic current signal at higher harmonics; therefore, sinusoidal voltammetry was shown to be more sensitive than most traditional electrochemical methods. Sinusoidal voltammetry was employed to detect both purineand pyrimidine-based nucleic acids. Adenine and cytosine, representing these two classes of nucleic acids, could be measured with submicromolar detection limits at a copper electrode under these conditions, where the sensitivity for adenine was much higher than that for cytosine. Detection limits for purinecontaining nucleotides [e.g., adenosine 5′-monophosphate (AMP), adenosine 5′-diphosphate (ADP), and adenosine 5′-triphosphate (ATP)] were on the order of 70-200 nM using this method. These detection limits are achieved for native nucleotides and are over 2 orders of magnitude lower than those found with UV absorbance detection. Submicromolar detection limits were also obtained for pyrimidine-based nucleotides, which could also be detected with high sensitivity due to the presence of a sugar backbone, which is electroactive at the copper surface. This detector is not fouled by the nucleotides and may be used for the sensitive detection of analytes eluting continuously in a flowing stream, i.e., from a chromatography column or a electrophoresis capillary. EXPERIMENTAL SECTION Reagents. The water used was deionized and then passed through a Milli-Q water purification system (Millipore Corp., Bedford, MA). The nucleotides and bases (98-99%, Sigma Chemical Co., St. Louis, MO) and sodium hydroxide (ACS Grade, Fisher Scientific, Fair Lawn, NJ) were used as received. All experiments were done with 0.10 M sodium hydroxide as the running electrolyte. Stock solutions (0.01 M) of the nucleotides and cytosine were prepared in deionized water. Adenine solution was always freshly prepared in high-pH running electrolyte as it does not dissolve in water. Subsequent dilutions were made whenever required using the running electrolyte. Copper Microelectrodes. The fabrication of 20 µm Cu microelectrodes has been described elsewhere.36 Prior to use, the electrodes were polished with a 1-µm diamond polish, followed by sonication in water. No electrochemical activation was performed in an effort to minimize the occurrence of background faradaic processes at the electrode surface. The potential at the electrode was cycled under the experimental conditions for ∼1 h prior to the collection of data to achieve a stable background response. (39) Smith, D. E. In Electroanalytical Chemistry: A Series of Advances; Bard, A. J., Ed.; Marcel Dekker: New York, 1966; Vol. 1, pp 1-155. (40) Cullison, J. K.; Kuhr, W. G. Electroanalysis 1996, 8, 314-319.
Analytical Chemistry, Vol. 69, No. 17, September 1, 1997
3553
Instrumentation. The flow injection analysis (FIA) apparatus has been described previously36,41 and includes a pneumatic actuator (Rheodyne, Model 5701) controlled via a solenoid valve (Rheodyne kit, Model 7163). This cell was designed to match the internal diameter of the FIA tubing (0.75 mm) in order to minimize diffusional broadening of the analyte as it was transported to the microelectrode. Finally, the flow rate (0.5 mL/min) was controlled by gravity flow by maintaining a height difference of 25 cm between the running electrolyte container and the flow cell. The volume of sample injected into the flow stream was determined by the flow rate and the length of the injection. Typically, an injection time of 60 s was used, producing an injection volume of 500 µL. This injection protocol allowed the electrode to see the full concentration of the injected sample without dispersion or dilution and obtain a flat-top response. Sinusoidal voltammetry36 and slow-scan cyclic voltammetry42 were performed as described previously. The specific conditions used for these experiments are described here. Sinusoidal voltammetry was performed by digitally generating a 2-Hz sine wave (exactly 1.95 Hz, 0.05-0.55 V vs Ag/AgCl) with software provided by Axon Instruments (SineVolt; Axon Instruments Inc., Foster City, CA). This excitation wave form was filtered with a four-pole low-pass filter having a 3-dB point at a frequency 3 times the fundamental frequency using a Cyberamp (Model 380, Axon Instruments Inc.). The filtered excitation wave form was supplied to the Cu electrode through a three-electrode potentiostat (Geneclamp, Axon Instruments Inc.). The current output of the potentiostat was filtered with a four-pole low-pass filter having a 3-dB point at 200 Hz to avoid and then by a second four-pole filter with a Cyberamp set at 40 Hz to further minimize noise [at 2 times the maximum frequency component analyzed (20 Hz)]. The current was sampled digitally with a 12-bit A/D (1200A, Axon Instruments Inc.) at a rate of 500 Hz using an 80486 IBMcompatible personal computer. Leakage was avoided by sampling a wide bandwidth (over 10 000 points). Unless otherwise specified, two sinusoidal cycles were obtained in a single scan, and 240 such scans were collected for one FIA measurement. Acquiring a larger number of scans increased the resolution at the lower frequencies, and a longer sampling time minimized artifacts due to convolution with the window function of the data. Background subtraction was performed continuously as follows. A background signal was acquired digitally prior to each FIA experiment and then converted back into an analog signal which was subtracted from all subsequent current measurements prior to digitization of the instantaneous signal. This was done to minimize the low-frequency components associated with the background at a copper electrode, to increase the dynamic range for the measurement of the signal due to nucleotides. The time domain data acquired in this manner were converted into the frequency domain with Fourier transform methods using commercial software (Matlab 4.2.c.1, The Mathworks, Inc., Englewood Cliffs, NJ). The protocol for analyzing frequency spectra has been described previously.36 Briefly, frequency spectra of the signal (only) was obtained simply by digital subtraction of a residual background vector from the instantaneous current vector. Time course data were obtained through the digital equivalent of lock-in (41) Kristensen, E. W.; Kuhr, W. G.; Wightman, R. M. Anal. Chem. 1987, 59, 1752. (42) Nowall, W. B.; Kuhr, W. G. Anal. Chem. 1995, 67, 3583-3588.
3554
Analytical Chemistry, Vol. 69, No. 17, September 1, 1997
amplification. Successive 512 point segments of the data were Fourier transformed sequentially into the frequency domain, generating the magnitude and phase information of each frequency element. Since all of the Faradaic information is contained within the harmonics of the excitation wave form,36,39,40,43 only these frequency elements were examined. Phase selectivity was obtained by taking the projection of the instantaneous current on the background-subtracted signal. Finally, the phase-resolved projections of each segment were low-pass filtered as a function of time. The time course information was generated after averaging 10 such projections together by using a moving average smooth (essentially moving boxcar integration) as a low-pass filter. RESULTS AND DISCUSSION Electrochemical detection of nucleotides and DNA at mercury and carbon surfaces employs adsorption of these analytes at the electrode surface and then the subsequent use of some voltammetric stripping procedure. This is necessitated because all these analytes are strongly adsorbed on mercury and carbon surfaces. A copper surface obviates the necessity of any stripping procedure because nucleotides are not found to adsorb on the copper surface in the present study under the excitation conditions used. The electroactive response of nucleotides on copper can be explained by the mechanism proposed by various researchers for the oxidation of sugars and amines on copper.35,37,44,45 The presence of a Cu(II) layer on the surface is hypothesized to be responsible for the electrocatalytic oxidation of these compounds.31,32 In contrast to platinum and gold surfaces,30,46 the Cu(II) layer is soluble in high-pH buffer, and thus the oxidation of sugars and amines does not cause any fouling of the copper surface. These properties of copper make it amenable for the analysis of nucleotides, which are comprised of amine-containing bases and a sugar backbone. Initial studies were done with the standard procedure of slowscan cyclic voltammetry,44 in order to determine the redox characteristics of these nucleotides. But, slow-scan measurements at 10 mV/s led to significant adsorption of the nucleotides on the copper surface. Relatively faster scanning was found necessary to keep the electrode surface clean, illustrating that a scanningtype approach was needed for the continuous detection of nucleotides. Figure 1 shows a cyclic voltammogram for AMP at a copper electrode at a scan rate of 2 V/s. The peak after +400 mV indicates that the nucleotide is being oxidized in a potential region similar to that known for smaller carbohydrates. However, the sensitivity for detection is poor, and the peak is not sharp. Consequently, it is hard to determine the Epeak for the oxidation wave, which makes it harder to make any quantitative calculations with cyclic voltammetry. Also, this indicates that the larger and more complex structure of the nucleotides might be responsible for kinetic limitations for its oxidation on the copper surface. This hypothesis was confirmed when we did the preliminary experiments with sinusoidal voltammetry. Initial tests with 10-Hz excitation frequency (optimal for the detection of simple sugars)36 had less of the signal distributed into the higher harmonics, indicating the kinetics are definitely slower for the oxidation of (43) Pettit, F. R. Fourier Transforms in Action; Chartwell-Bratt Ltd.: Sweden, 1985. (44) Prabhu, S. V.; Baldwin, R. P. Anal. Chem. 1989, 61, 2258-2263. (45) Marioli, J. M.; Kuwana, T. Electrochim. Acta 1992, 37, 1187-1197. (46) LaCourse, W. R.; Johnson, D. C. Anal. Chem. 1993, 65, 50-55.
Figure 1. Subtracted cyclic voltammograms at a copper electrode. Scan rate 2.0 V/s. (A) Blank; (B) 1 mM AMP.
Figure 2. Sinusoidal voltammetric detection of AMP at 10 Hz with FIA. Excitation frequency at the copper electrode was 10-Hz sine wave, 0.05-0.55 Vpeak-peak vs Ag/AgCl. Injected sample 100 µM AMP. (A) Background-subtracted signal in the frequency domain at time (t) 125 s. Signals are shown at optimized phase angles, which are listed above the response at each harmonic. (B) Time course for the FIA injection of 100 µM AMP at the fifth harmonic (50 Hz, φ ) -52.6°). A total of 300 scans were collected for the full FIA measurement. NaOH (0.1 M) running electrolyte. Flow rate 0.5 mL/min.
nucleotides as compared to simple mono- and disaccharides. Figure 2A shows the background-subtracted frequency response for AMP, obtained with an excitation frequency of 10 Hz. There is a very small signal apparent in the higher harmonics at this excitation frequency. Under these excitation conditions, fifth harmonic has the best signal to noise characteristics. Figure 2B illustrates the time course information for this harmonic. The detection limits for AMP are in the low-micromolar range with the 10-Hz excitation frequency. If kinetics were the limiting factor for the poorer sensitivity for nucleotides compared to that for sugars, a lower excitation frequency should result in much more signal being distributed into the higher harmonics. Another concern was that AMP also
contains the adenine base, and so it was possible that some signal was being contributed by the amine present on this purine base,23,47 as amines are also oxidized at a copper surface. Figure 3 shows the background-subtracted frequency response for the base adenine and three different adenine-containing nucleotides obtained with the slower excitation frequency of 2 Hz. It shows that adenine does give a significant signal at the copper surface, but its frequency spectrum is remarkably different from the response obtained due to the adenine-based nucleotides. This indicates that adenine itself is electroactive, but the response of the nucleotides is a convolution of the signal due to adenine and the pentose sugar present in the nucleotides. Frequency spectra for the three nucleotides (i.e., AMP, ADP, and ATP) show a similar pattern among themselves, indicating that the kinetic parameters involved are similar for these three nucleotides. The second (4 Hz), fourth (8 Hz), and the fifth (10 Hz) harmonics of the fundamental frequency show a significant amount of signal, even though the background current at these harmonics is very small (first trace in Figure 3A). This shows that the slower frequency of 2 Hz is much more suitable for the slower kinetics of nucleotides. The flow injection time course information for these harmonics can be isolated and optimized. Figure 4 shows the time course of separate injections of 100 µM each of adenine, AMP, ADP, and ATP, where the current response shown is the phase-optimized response at the fifth harmonic. The response at this harmonic is very sensitive (S/N > 2000), and the injection profile shows that the copper surface is not fouled by the oxidation of these nucleotides (since there is no evidence of tailing). The response from 20 successive injections of AMP showed a standard deviation less than 7%, indicating that there is no significant degradation in the electrode response. This indicates that the nucleotides are probably reacting in a manner similar to that of primary alcohols, amines, and carbohydrates at the copper surface. The frequency spectra of the different nucleic acids can be used to qualitatively identify each species. Since our hypothesis was that the response is a convolution of the signal due to the base and the sugar backbone, we predicted that the signal due to a pyrimidine-based nucleotide would be different (and smaller) from that of adenine-containing nucleotides. This should be true because pyrimidine bases themselves are much harder to oxidize than purine bases at most electrode surfaces. Figure 5 shows the response due to cytosine, a pyrimidine base, and a cytosinecontaining nucleotide, cytosine 5′-monophosphate (CMP). As predicted, the cytosine spectrum (Figure 5A) is very different from that of adenine (Figure 3A) and has much lower signal intensity at the higher harmonics. In contrast, the frequency spectrum of CMP (Figure 5C) still has a large signal intensity at the higher harmonics at the copper surface due to the pentose sugar present in its backbone. The frequency domain response of CMP is significantly different from that due to AMP (Figure 3B), even though they both have the same pentose sugar present. This difference can be explained on the basis of the different frequency domain response of their component bases. These frequency spectra are extremely reproducible in both the magnitude and phase angle of the signal at each harmonic. The standard deviation observed for the magnitude (7%, n ) 3) and phase angle (4%, n ) 3) of each harmonic component for adenine demonstrates the reproducibility of the frequency spectrum. This qualitative (47) Wang, J.; Chen, L.; Chicharro, M. Anal. Chim. Acta 1996, 319, 347-352.
Analytical Chemistry, Vol. 69, No. 17, September 1, 1997
3555
Figure 3. Background-subtracted frequency spectra for different adenine-containing compounds. Sample concentration 100 µM each (A) Residual background at copper electrode after analog subtraction (0), adenine (9), (B) AMP, (C) ADP, and (D) ATP at different harmonics. Signal spectra shown correspond to signal obtained at time (t) 95 s in the FIA injections shown in Figure 4. Background spectra shown in (A) corresponds to t ) 10 s in the FIA injections. Sinusoidal excitation with 2-Hz sine wave, 0.05-0.55 Vpeak-peak vs Ag/AgCl. The background frequency domain information is given for a copper electrode under these voltammetric conditions; thus it can be compared in general to all the different samples analyzed. However, each analyte response was optimized with respect to the specific background obtained in a FIA experiment. FIA injection 60 s. Flow rate 0.5 mL/min. NaOH (0.1 M) running electrolyte.
Figure 4. Time course information isolated at the fifth harmonic (10 Hz) for adenine-containing compounds in the FIA experiment corresponding to Figure 3. (A-D) Adenine, AMP, ADP, and ATP (100 µM each), respectively. All responses are phase optimized to the values given in Figure 3.
information could be very useful in DNA sequencing applications. Difference in signals due to different bases and/or nucleotides can be capitalized to identify the nucleotides derived from the exonuclease digestion of a DNA strand. In order to determine the best frequency and phase angle for the continuous measurement of the sample, all the harmonic components can be inspected simultaneously for signal to noise calculations. Figure 6 shows the flow injection time course for 10 µM AMP at all the harmonics measured (first to tenth). By 3556 Analytical Chemistry, Vol. 69, No. 17, September 1, 1997
Figure 5. Frequency and Time course information for cytosine and CMP (100 µM each). Background-subtracted frequency spectra for (A) cytosine and (C) CMP in an FIA experiment. Optimized phase angles at each harmonic are depicted above the respective response. (B) and (D) are the isolated time course information at the fifth harmonic (10 Hz) for cytosine and CMP, respectively, for the same FIA experiment. All FIA conditions are same as in Figure 3.
inspecting all the harmonics for AMP, we determined that the fifth harmonic has the highest sensitivity. The detection limit for AMP at this harmonic was 200 nM. The signal was linear (r2 ) 0.974) between the detection limit and 100 µM (the highest concentration tested). Similar calculations were done for all the nucleotides tested. The limits of detection for adenine-based
Figure 6. Sensitive detection of nucleotides with sinusoidal voltammetry in an FIA experiment. Sample 10 µM AMP. FIA injection 60 s. All FIA conditions are same as in Figure 3. Time course information is depicted at the first to tenth harmonic of the excitation frequency. These data are used to determine the harmonic component that is most sensitive for the detection of an analyte. Signals at each harmonic are depicted at the optimized phase angle as obtained by the inspection of the frequency spectra obtained in Figure 3A for AMP analysis.
nucleotides, namely, ADP and ATP at the fifth harmonic were found to be 130 and 70 nM, respectively. Cytosine-based CMP had a higher detection limit of 1.1 µM at the fifth harmonic. However, lower detection limits of 770 and 150 nM were observed for CMP at the fourth and the third harmonics, respectively. This further highlights the differences between the frequency spectra of adenine-containing nucleotides and cytosine-containing nucleotides. These differences in the frequency domain response can be used for their identification. Similar results can be expected for guanine- and thymine-based nucleotides as the detection approach is applicable to compounds containing a nucleobase and a sugar backbone, which are present in all nucleotides. CONCLUSIONS We report the development of a sensitive electrochemical detector for the analysis of native nucleotides in a flow injection analysis format. The detection of various nucleotides is based on the oxidation of their sugar backbone and on the nucleobase at a copper surface. The detection of nucleotides does not foul the electrode surface, and the detector can be used for continuous analysis of nucleotides in flow streams. The limits of detection for the adenine nucleotides are on the order of 100 nM, which is 2 orders of magnitude lower than the commonly used UV absorbance detectors. It is possible to differentiate the adenine-
(purine base) containing nucleotide from cytosine- (pyrimidine base) containing nucleotide based on the difference in the frequency domain response due to the two bases. This is not possible with other electrochemical sensors, which cannot qualitatively differentiate between the different nucleotides. Thus, this detection approach could have significant applications in the field of DNA sequencing. Also, this detector can be easily coupled to separation methods, thereby enabling the sensitive analysis of digestion products of DNA being separated by liquid chromatography and capillary electrophoresis. Experiments are underway in our laboratory currently to investigate these possibilities. ACKNOWLEDGMENT We appreciate the introduction to Matlab given by Sohail Nadimi in the computer science department at UCR. The authors also acknowledge Dr. Sharon Neal for help with optimization of the code in Matlab. Dr. Narasaiah Dontha was very helpful in discussions relating to DNA sensors. Received for review March 27, 1997. Accepted June 25, 1997.X AC970333N X
Abstract published in Advance ACS Abstracts, August 1, 1997.
Analytical Chemistry, Vol. 69, No. 17, September 1, 1997
3557
