Chromatographic Detection Using Tris (2, 2 '-bipyridyl) ruthenium (III

Chevron Science Center, University of Pittsburgh, Pittsburgh, Pennsylvania 15260, and Sandia .... (E. M. science) was titrated with 0.5 M glacial acet...
0 downloads 0 Views 112KB Size
Anal. Chem. 1999, 71, 1504-1512

Chromatographic Detection Using Tris(2,2′-bipyridyl)ruthenium(III) as a Fluorogenic Electron-Transfer Reagent Steven J. Woltman,† William R. Even,‡ and Stephen G. Weber*,†

Chevron Science Center, University of Pittsburgh, Pittsburgh, Pennsylvania 15260, and Sandia National Laboratories, MS 9403, P.O. Box 969, Livermore, California 94550-0969

Tris(2,2′-bipyridyl)ruthenium can be excited to fluorescence by visible light (λabs 454 nm, λem 607 nm) when in the M(II) oxidation state, but not in the M(III) state. A novel chromatographic detection method using the nonfluorescent M(III) form of the complex as a postcolumn fluorogenic reagent is demonstrated. The M(III) form is a powerful oxidizing agent (E° ) 1.27 V vs NHE, 1.05 V vs Ag/AgCl). The M(III) reagent is generated on-line from the M(II) form of the complex by a highly efficient porous carbon electrode and then reacted briefly with chromatographic effluent; the M(II) created by electron transfer from oxidation-susceptible analytes is then detected by fluorescence. The fluorescence detector can be calibrated for number of electrons transferred by injection of either M(II) or an oxidative standard such as ferrocyanide. It is hoped that this redox-based detection scheme will provide an alternative to electrochemical detection. Among the advantages are freedom from surface fouling and the potential for extremely low detection limits. The scheme was applied to detection of the peptide dynorphin A and several of its fragments. Dynorphin A contains tyrosine at the N-terminus (position 1) and tryptophan in position 15; these amino acid residues are susceptible to oxidation and peptides containing them can be detected on that basis. Flow injection testing of the model compounds TyrGly-Gly-Phe-Leu and Gly-Gly-Trp-Gly indicated that tyrosine transferred ∼1 electron to the M(III) reagent and that tryptophan transferred ∼4 electrons. Similar results were obtained from the chromatographic runs. Dynorphin A and six dynorphin A fragments containing the Nterminal tyrosine were detected easily at 100 nM concentration (14 pmol) using laser-induced fluorescence. As expected, one fragment that did not contain tryptophan or tyrosine was not detected. A mass detection limit of 80 fmol was estimated for the tyrosine-containing fragments. Electrochemical detection for liquid chromatography, although capable of sensitive and selective detection of redox-active species,1 has limitations imposed by surface fouling and drift in the potential †

University of Pittsburgh. Sandia National Laboratories. (1) Chen, J. G.; Woltman, S. J.; Weber, S. G. Adv. Chromatogr. 1996, 36, 273. ‡

1504 Analytical Chemistry, Vol. 71, No. 8, April 15, 1999

of the reference electrode. Frequent operator intervention is often required to keep such a system operating, and certain samples may deactive the electrode surface before completing a single analysis. In addition, some analytes exhibit slow surface kinetics that make electrochemical detection impractical. A reagent that is rendered fluorescent by electron transfer would have the potential to provide the broad applicability of electrochemical detection with the sensitivity and low-maintenance operation of fluorescent detection. Osmium and ruthenium complexes having diimine-type ligands, e.g., tris(bipyridyl) complexes, represent a potential family of such redox fluorescence reagents.2 The M(II) state of such complexes can be excited to fluorescence by visible light, while the M(III) state cannot. Indeed, detection based upon chemiluminescence, using ruthenium(III)tris(bypyridine) as a reagent, has been demonstrated.3-7 Oxidation of the analyte, e.g., an amine, by the M(III) form of the complex produced a luminescent excited state M(II)*. This method has the drawbacks of widely varying efficiency depending on the analyte and the optical signal being limited to the intrinsic luminescence produced by the reaction. Oxygen quenching of the fluorescence of ruthenium complexes has been used as the basis for oxygen sensing.8,9 Photoelectrolanalytical chemistry with Ru(bpy)3 has been demonstrated, also.10-13 However, we are unaware of any existing detection scheme that combines analyte oxidation via the M(III/II) couple of these complexes with fluorescence detection of the M(II) created by reduction of M(III). The goal of the work reported here is fluorescence detection, based upon electron transfer to M(III) from an analyte to form M(II), and detection of M(II). In such a detection scheme, the mass transport and surface electron-transfer kinetics of an electrode are replaced by the kinetics of electron transfer in solution. Potential control is by choice of fluorescent metal (2) Kalyanasundaram, K. Coord. Chem. Rev. 1982, 46, 159. (3) Lee, W. Y.; Nieman, T. A. Anal. Chem. 1995, 67, 1789. (4) Shultz, L. L.; Stoyanoff, J. S.; Nieman, T. A. Anal. Chem. 1996, 68, 349. (5) Skotty, D. R.; Lee. W. Y.; Nieman, T. A. Anal. Chem. 1996, 68, 1530. (6) Lee, W. Y.; Nieman, T. A. Anal. Chim. Acta 1996, 334, 183. (7) Ridlen, J. S.; Klopf, G. J.; Nieman, T. A. Anal. Chim. Acta 1997, 341, 195. (8) Preininger, C.; Klimant, I.; Wolfbeis, O. S. Anal. Chem. 1994, 66, 1841. (9) Velasco-Gracia, N.; Valencia-Gonzalez, M. J.; Diaz-Garcia, M. E. Analyst 1997, 122, 1405. (10) Weber, S. G.; Morgan, D. M.; Elbicki, J. M. Clin. Chem. 1983, 29, 1665. (11) Morgan, D. M.; Elbicki, J. M.; Weber, S. G. Anal. Chem. 1985, 57, 1746. (12) Berry, B. F.; Weber. S. G. J. Electroanal. Chem. 1986, 208, 77. (13) Kuhn, L. S.; Weber, A.; Weber, S. G. Anal. Chem. 1990, 62, 1631. 10.1021/ac981181e CCC: $18.00

© 1999 American Chemical Society Published on Web 03/16/1999

complex rather than by potentiostatic control relative to a reference electrode. The system is expected to exhibit high sensitivity and also stability, with a constant sensitivity not dependent on phenomena on the surface of an electrode. In addition, if the number of electrons in the redox reaction is known and is independent of solution conditions, then the potential for absolute calibration of the detection system exists, requiring only injection of a 1-electron standard such as ferrocyanide or the fluorescent M(II) complex itself. EXPERIMENTAL SECTION Chemicals. Except where noted, all chemicals were used as received. YGGFL, GWGG, and dynorphin A fragments 1-6, 1-7, 1-8, 1-9, 1-13, 1-10, 1-17, and 2-13, all from Sigma (St. Louis, MO), were dissolved in deionized water to make individual stock solutions of approximately millimolar concentration. These were frozen immediately (-20 °C). Aliquots of thawed stock solutions were subsequently combined and diluted in deionized water to concentrations in the micromolar or nanomolar range. LC-grade acetonitrile was from EM Science. Water was purified with a Milli-Q system (Millipore). Spectrophotometric-grade trifluoroacetic acid (Sigma-Aldrich) was received in sealed glass ampules. Reagent-grade sodium perchlorate (Aldrich) was recrystallized once from methanol to remove chloride. Chromatographic mobile phases and the electrolyte solution for RuII(bpy)3 oxidation were filtered under vacuum through 0.45-µm hydrophilic nylon membrane filters. Tris(2,2′-bipyridyl)ruthenium(II) chloride hexahydrate (Aldrich) and tris(2,2′-bipyridyl)osmium(II) perchlorate (gift of Adrian Michael laboratory) were dissolved in acetonitrile to make stock solutions of approximately millimolar concentration. The stock solutions were syringe-filtered through 0.2-µm PTFE filters (Whatman Inc., Clifton, NJ) and added to filtered electrolyte solution to give the desired concentrations. The electrolyte solution for RuII(bpy)3 was 0.1% TFA, 0.1 M NaClO4 dissolved in either 50/50 acetonitrile/water or 100% acetonitrile. Stock solutions of potassium ferrocyanide trihydrate (Fisher Scientific) were made daily in helium-sparged water at ∼0.1 M and diluted in helium-sparged mobile phase shortly before use. For porous electrode testing, ferrocyanide was dissolved in a nonacidic aqueous electrolyte solution of 3% 2-propanol and 0.1 M of NaClO4. For testing the oxidation by RuIII(bpy)3, the solvent for ferrocyanide was water. All buffer solutions were aqueous. For pH 0.5-3.0 buffers, sodium monochloroacetate (Adrich), dichloroacetic acid (Aldrich), and trichloroacetic acid (Fisher) were combined to provide a concentration of 0.25 M each. The pH of the mixture was adjusted with concentrated perchloric acid and 50% w/v (19 M) sodium hydroxide solution (both J. T. Baker). Liquid drawn from the upper portion of 50% NaOH is essentially carbonate-free due to the insolubility of sodium carbonate in this solution. For acetate buffers, 0.5 M, pH 4.0-5.5, a 0.5 M solution of sodium acetate (E. M. science) was titrated with 0.5 M glacial acetic acid (J. T. Baker). Maleate buffers, 0.5 M, pH 6.1-6.9, employed maleic acid (Fisher) which had been recrystallized once from water and titrated with 50% NaOH solution. For borate buffers, 0.5 M, pH 9.0-10.5, 0.125 M sodium tetraborate decahydrate (J. T. Baker) was titrated with concentrated perchloric acid.

Construction Materials. PTFE Teflon tubing (0.012-in. i.d, 30-gauge o.d.), shrink-fit PTFE tubing, and two-layer PTFE/FEP Teflon shrink-melt tubing were all from Small Parts, Inc. (Miami Lakes, FL). PTFE tubing (0.004-in. i.d., 0.16-in. o.d.) was from ColeParmer Inc. (Vernon Hills, IL). FEP tubing (0.02-in. i.d., 1/16 o.d.) was from Upchurch Scientific (Oak Harbor, WA). Woven porous Teflon tubing (3.5-µm pore size, 1.0-mm i.d.) was from International Polymer Engineering (Tempe, AZ). Platinum wire (0.125- and 0.05-mm diameter) temper “as drawn”, platinum tubing (0.8-mm i.d, 1.0-mm o.d.), and silver wire (1.0-mm diameter) were all from Goodfellow, Inc. Nafion ion-exchange polymer tubing (various diameters) was from Perma-Pure, Inc. (Tom’s River, NJ). Teflon filters (50 mm, 5-6-µm pore size) were from Cole-Parmer Instrument Co. Catalog No. E-06221-20). Omnifit Catalog No. 1010 1.5-mm PTFE union was from Chrom Tech, Inc. (Apple Valley, MN). An adjustable heat gun, Steinel model HL 1800E, obtained from Small Parts, Inc., was used to heat various types of Teflon tubing. Temperature settings used were as follows: 4.0, to soften PTFE tubing for stretching; 3.8, to cause two-layer PTFE/FEP shrinkmelt tubing to contract; 2.5, to cause single-layer shrink-fit tubing to contract; 1.5, to cause expanded FEP tubing to contract. These settings correspond to temperatures in a range roughly from 100 (setting 1.5) to 380 °C (setting 4). Note: Excessive heating of Nafion is to be avoided. Optical fiber (440-µm fiber diameter, 510-µm jacket diameter) was from Fiberguide (Stirling, NJ). Polyimide-coated silica capillary (700-µm i.d., 850-µm o.d.) was from Polymicro Technologies (Phoenix, AZ). The carbon foam was Sandia RF 2000, an experimental material derived from a resorcinol-formaldehyde polymer foam precursor pyrolyzed at 2000 °C.14,15 The precursor foam is formed by polymerization of monomers in the water phase of an oil-in-water emulsion. The material has a duplex pore structure; the walls between macroporous 4-10-µm cavities are pierced by microporosities of much smaller size range, yielding a very large specific surface area of >400 m2/g. Note: Contact with Sandia National Laboratory to request carbon materials can be made through the authors. Porous Carbon Working Electrodes. The porous working electrodes were packed-bed electrodes consisting of carbon particles in Nafion tubing, with PTFE tubing fluid connections and platinum electrical contacts. The platinum electrical contacts were located on the downstream side of the electrode so that fluid flow would tend to push the carbon onto the platinum. The fluid connections were held and sealed inside the Nafion by external compression fittings. The ends of small diameter PTFE tubing used for fluid connections were welded to 1/ -in.-o.d. tubing using shrink-melt tubing to adapt to standard 16 HPLC fittings. Frits made of small disks of PTFE filter material were used to prevent the carbon powder from escaping from the electrode. Two different electrodes, of approximately 1 µL and 100 nL internal volume, were used. These were constructed using Nafion tubing of 0.039- and 0.013-i.d., respectively, with length of packed bed ∼1 mm in each case. The large electrode was dry-packed. (14) Even, W. R.; Gregory, D. P. MRS Bull. 1994, 19, 29. (15) Crocker R. W..; Hunter M. C.; Yang N. C.; Headley T. J.; Even, W. R J. Non-crystalline Solids 1995, 186, 191.

Analytical Chemistry, Vol. 71, No. 8, April 15, 1999

1505

The small electrode was slurry-packed at 2-3 psi, using a syringe of carbon powder suspended in 2-propanol. The large electrode was reinforced with a sheath of 1-mm-i.d. woven PTFE tubing to improve resistance to back pressure. Compression fittings were two pieces of a bisected Omnifit PTFE union for the large electrode, and 0.02-in.-i.d., 1/16-in.-o.d. FEP tubing for the small electrode. The FEP tubing was expanded with a sewing needle; moderate heating caused it to shrink, providing compression. To strengthen the small electrode, shrink-melt tubing was used to secure the FEP tubing to the PTFE fluid connections. The completely assembled three electrode cells used small blocks of reticulated vitreous carbon (Electrosynthesis Co., Amherst, NY) as auxiliary electrodes. The filling solution was 0.1 M NaClO4, with 0.1% trifluoroacetic acid (TFA) and acetonitrile, either 50 or 100%, for generation of M(III) bipyridyl complexes. Either an Ag/AgCl reference or a nonaqueous reference using an Ag wire in 0.01 M AgNO3/0.1 M NaClO4/acetonitrile was used. The potential of the nonaqueous reference electrode was measured at 325 mV vs the aqueous reference electrode. Instrumentation. (a) Pumps and Injectors. Model 510, M6000, and 600 MS HPLC pumps were from Waters Associates (Milford, MA). Flow rates for the model 510 and M6000 pumps were set with a model 680 gradient controller. The 600 MS pump is a quaternary gradient pump and has a controller equipped with a SILK flow smoothing algorithm. A four-pump UltraPlus system, from MicroTech Scientific (Sunnyvale, CA) was used for low-flowrate, microbore separations. The MicroTech system included a dynamic mobile-phase mixer and a four-channel vacuum degasser. For flow injection work or for supplying postcolumn reagent, a coil of 0.0025-in.-i.d., 1/16-in.-o.d. PEEK capillary tubing was used to obtain back pressure (several hundred psi or higher). The model 510 and M6000 Waters pumps were equipped with aftermarket pulse dampeners upstream of this capillary coil. Mobile phases were contained in polyethylene bottles. For the Waters pumps, mobile phases were degassed by sparging with helium. No-Ox low-permeability tubing (Alltech Associates, Deerfield, IL) was used to connect the reservoirs to the pumps. For the MicroTech system, No-Ox tubing was used between the degasser and the pumps. All injections were performed with model 7125 injectors from Rheodyne, Inc. (Cotati, CA) with loop volumes from 20 µL to 10 mL, depending on the experiment. (b) Electrochemical Equipment. The potential of the porous electrodes was controlled using either an RDE-3 potentiostat (Pine Instruments, Grove City, PA) or a CV-1A potentiostat (BAS Inc., West Lafayette, IN). Electrochemical detection downstream of the scrubber employed a CC-5 thin-layer detection cell with dual LC4C potentiostats (BAS, Inc.). A dual glassy carbon working electrode with a thin (0.0005 in.) Teflon gasket was used with the CC-5 detector. (c) Fluorescence Detector. A Varian Fluorichom detector, model 430020-02, (Varian Associates, Walnut Creek, CA) was used for all fluorescence detection. It was found necessary to remove the stainless steel frits and to replace the original stainless steel tubing in the flow cell block with new PEEK tubing to prevent loss of the reactive M(III) form of Ru(bpy)3. For flow injection analysis, and analysis of 1 µM dynorphin A fragments, the original configuration was used a 2-mm-diameter quartz flow cell with halogen light bulb excitation. Optical filters (Corion, Inc. Franklin, 1506 Analytical Chemistry, Vol. 71, No. 8, April 15, 1999

MA) were 500-nm short pass for excitation and 600-nm long pass for emission. (d) Axial Illumination Flow Cell. For microbore chromatography of 100 nM dynorphin A fragments, the 2-mm flow cell of the Varian fluorescence detector was replaced with a capillary flow cell using axial laser illumination. A 30-mW variable-power argon ion laser, model 2201-30 BL, Cyonics/Uniphase (San Jose, CA) was coupled to a 440-µm optical fiber using a microscope objective adaptor, model M-F-91-C1-T, Newport Optics (Irvine, CA). The optical fiber was used to illuminate axially a 700-µm-i.d. silica capillary from which the polyimide coating had been removed. The capillary was held in the flow cell block vertically, with the optical fiber entering from the top, and an inlet connection at the bottom made from 0.004-in.-i.d. PTFE tubing. A tee fitting at the top of the capillary held the optical fiber and permitted egress of the mobile phase. The laser beam was not optically filtered. The 600-nm long-pass filter was used to filter the fluorescent emission. The illuminated volume of the flow cell was calculated to be 3 µL. (e) Data Collection. EZChrom chromatography software (Scientific Software, Inc, San Ramon, CA) was used to collect data. The export function of the program was used to generate ASCII files which were imported into Excel. Evaluation of Porous Electrodes by Voltammetry. Voltammograms for oxidation of ferrocyanide and RuII(bpy)3 from M(II) to M(III) were obtained by monitoring the composition of the effluent from the porous electrodes rather than the current signal of the porous electrode. Electrochemical detection was used initially, and finally flurorescence detection for Ru(II) bpy3. Small inner diameter connecting tubing between the porous electrode and the electrochemical detector was used to prevent interaction between the two potentiostats. To avoid pump contamination and facilitate changes in reagent composition, in early experiments a 10-mL loop of M(II) in electrolyte solution was injected through the porous electrode, providing several minutes supply of the reagent. In later experiments, the RuII(bpy)3 reagent was supplied using a pump reserved for that purpose. Flow Injection Analysis by Analyte-Generated RuII(bpy)3 Fluorescence. (a) Reversion of RuIII(bpy)3 to RuII(bpy)3 as a Function of pH. A 10-mL injection loop was used to supply RuII(bpy)3 to the larger, 1-µL electrode. A second pump with an injector equipped with a 1.5-mL loop was used to inject buffers of various pH values. Both pumps operated at 600 µL/min. The two streams were combined in a tee, and the combined flow was monitored with the fluorescence detector (2-mm flow cell). Degree of reversion was calculated by the ratio of the steady-state fluorescence signal reached during buffer flow to the RuII(bpy)3 signal that had been measured before initiating oxidation with the porous electrode. (b) Detection of Ferrocyanide, YGGFL, and GWGG by Fluorescence following Electron Transfer. The setup was the same as for the RuIII(bpy)3 reversion experiment, but with a 20µL PEEK loop on the second injector for injection of ferrocyanide and peptide solutions, which were prepared in 5% 2-propanol. Duplicate injections were performed at concentrations from 0 to 10 µM for ferrocyanide and 0 to 5 µM for the peptides. The final concentration of trifluoroacetic acid was 0.05%, corresponding to a pH of roughly 2.3. The response of ferrocyanide, a 1-electron

standard, was used to determine apparent number of electrons transferred (napp) for the two peptides, by dividing peptide peak area by predicted 1-electron peak area obtained from a linear regression of the ferrocyanide values. Chromatography. Separation of dynorphin A fragments at 1 µM concentration used a 2.1 × 50 mm C18 column (Mac-Mod Analytical, Chadds Ford, PA) with a water/acetonitrile gradient: solvent A, 0.1% aqueous TFA; solvent B, 0.1% TFA/60% acetonitrile. Flow rate, 0.25 mL/min. Composition, 5-80% B, 0-40 min. Postcolumn reagent, 600 µL/min of 20 µM Rubpy3Cl2/0.1%TFA/ 0.1 M NaClO4/acetonitrile. The large (1-µL) porous electrode was used to convert M(II) to M(III). Injections were performed with a 20-µL PEEK loop, supplying 20 pmol of each peptide (1 µM). The detector response was calibrated by several injections of ferrocyanide. For dynorphin A fragments at 100 nM, the modified fluorescence detector with axial laser illumination was used. A 1 × 50 mm C18 column was used for the separation, with a water/ acetonitrile gradient: solvent A, 0.1% TFA/1% 2-propanol/water qs; solvent B, 0.1% TFA/1% 2-propanol/acetonitrile qs. Gradients used were 50 µL/min, 5-90% B, 0-60 min; and 50 µL/min, 3080% B, 0-60 min. The Ru(bpy)3 postcolumn reagent, 5 µM Rubpy3Cl2/0.1%TFA/0.1 M NaClO4/acetonitrile qs, was supplied at 100 µL/min. The small porous electrode was used to convert M(II) to M(III). The detector response was calibrated by several injections of RuII(bpy)3. Peptide injections using the 1-mm column employed a relatively large 140-µL PEEK loop. This is discussed in the next section. RESULTS AND DISCUSSION Generation of MIII(bpy)3 with Porous Electrode. Highly efficient generation of M(III) from M(II) is necessary in order to minimize the fluorescent background. The reactivity of the M(III) form limits its stability, however. A fast reaction with hydroxide ion can be minimized by the use of acidic pH. A slower reaction with water also occurs, and water will be present in all but the most rigorously dried solvents. Ideally, the M(III) would be generated at high efficiency immediately before it was needed. This can be done using a porous carbon electrode. An electrode of this general type was demonstrated earlier. That design featured a plug of carbon foam derived from polyacrylonitrile (PAN) in a glass capillary, exiting into a large reservoir. High efficiency was demonstrated using ferrocyanide and ferrocene.16 A modified design, based upon high surface area carbon powder packed in a Nafion tube, has been developed (Figure 1a). The design is modeled after the electrochemically modulated LC column by Porter et al.17,18,19 The use of Nafion improves potential control by directing the ionic current in a radial direction through the walls of the tube rather than axially from the exit of the tube. The fragile PAN-derived carbon foam was replaced with a harder powdered material derived from resorcinol-formaldehyde polymer. The particular carbon material used, Sandia RF 2000,14,15 has low flow resistance and low retentiveness although the surface area is very high (see Experimental Section). The material is also (16) Davis, B. K.; Weber, S. G. Anal. Chem. 1994, 66, 1204. (17) Deinhammer, R. S.; Shimazu, K.; Porter, M. D. Anal. Chem. 1991, 63, 1894. (18) Deinhammer, R. S.; Ting, E. Y.; Porter, M. D. J. Electroanal. Chem. 1993, 362, 295. (19) Deinhammer, R. S.; Ting, E. Y.; Porter, M. D. Anal. Chem. 1995, 67, 237.

Figure 1. (a) Sketch of porous carbon working electrode: Pt, platinum wire ending in coil; CF, compression fitting; (b) dual pump, dual-injector system used for flow injection analysis. Both pumps and both injectors were used for testing RuIII(bpy)3 reversion as a function of pH, and for response of the peptides YGGFL and GWGG. PE, porous electrode. For efficiency testing of the porous electrode, only pump 1 was used. (c) Chromatographic analysis system used for fluorescence detection of dynorphin A fragments. The porous scrubber electrode on pump 2 was used in the microbore/LIF analysis of 100 nM dynorphin peptides, but not for initial separation of 1 µM dynorphin peptides.

resistant to migration of fines. The carbon powder/Nafion electrodes have low dead volumes and can withstand moderate pressures, on the order of 100-200 psi. Both large (1-µL) and small (100-nL) electrodes were used. The large electrode was used at 500-600 µL/min, linear velocity ∼1 cm/s, and the small electrode at 100 µL/min, linear velocity ∼2 cm/s. The Nafion-encased porous electrodes are intended for use primarily as analytical synthetic electrodes rather than as coulometric detectors. For this reason, the conversion efficiency of these electrodes was tested by monitoring the effluent from the porous electrode rather than the current from the porous electrode itself. M(II) was monitored downsteam of the porous electrode with either a thin-layer amperometric detector or a fluorescence detector (Figure 1b). Efficiency was calculated based upon disappearance of the M(II) signal. To be confident that a decrease in the downstream signal was caused by conversion to M(III), rather than by adsorption, the downstream M(III) concentration was also monitored electrochemically. When expressed as voltammograms, the downstream amperometric or fluorescence signals were plotted as a function of applied potential on the upstream porous electrode. Oxidation of ferrocyanide, OsII(bpy)3, and RuII(bpy)3 indicated that high conversion efficiency was possible with high applied potential. Disappearance of the M(II) signal correlated extremely well with onset of the M(III) signal for all three compounds in a number of repeated runs. Small irreproducible shifts in baseline prevented us from distinguishing between, e.g., 99 and 99.5% efficiency. We have measured efficiencies as high as 100% (or occasionally higher, due to baseline shifts) for all three test Analytical Chemistry, Vol. 71, No. 8, April 15, 1999

1507

Figure 2. Linear sweep voltammogram of Os(bpy)3(ClO)4 oxidation. Large (1-µL) porous electrode, 600 µL/min of 0.1% TFA/0.1 M NaClO4/50% acetonitrile, sweep rate 100 mV/min. The potential on the porous electrode was swept as indicated in the top panel. The middle panel shows the signal from a downstream dual electrochemical detector used to detect simultaneously both the M(II) starting material and the M(III) product. A loop of unoxidized M(II), porous electrode at +400 mV, is also shown. The bottom panel shows the M(II) and M(III) signals plotted as a function of applied potential on the porous electrode.

compounds. We are certain that efficiencies reliably reached 98% for all three test compounds, with both the 1-µL and 100-nL electrodes. The true values are probably much closer to 100%. A linear sweep voltammogram showing conversion of OsII(bpy)3 from M(II) to M(III) is shown in Figure 2. The original time-dependent amperometric signals, monitored with a chromatographic data collection program, are provided to illustrate the way in which the experiment was conducted. The electrochemical signals show the disappearance of M(II), and also the appearance of M(III), relative to the signal obtained from a separate injection of unconverted M(II). In Figure 3, a cyclic voltammogram of RuII(bpy)3 is shown, obtained from the disappearance of the M(II) signal at a downstream amperometric detector. Also in Figure 3, a linear sweep voltammogram obtained from the disappearance of the M(II) signal at a downstream fluorescence detector is shown. The choice of which M(II) complex to use depends on the analyte’s redox potential, which may be pH dependent. The equilibrium potentials for 1-electron, 1-proton oxidations of the amino acids tryptophan and tyrosine are known to be pH dependent and these values are influenced only slightly by incorporation into a peptide. At pH 2, literature values20 obtained by cyclic voltammetry are 0.93 V (vs Ag/AgCl) for tryptophan and 1.00 V for tyrosine. In our laboratory, we have measured half(20) Harriman, A. J. Phys. Chem. 1987, 91, 6102.

1508 Analytical Chemistry, Vol. 71, No. 8, April 15, 1999

Figure 3. (a) Hydrodynamic cyclic voltammogram of Ru(bpy)3Cl2 oxidation. Electrochemical detection of M(II) starting material downstream of porous electrode: large (1-µL) electrode; 500 µL/min of 0.1% TFA/0.1 M NaClO4/50% acetonitrile/50% water; sweep rate 100 mV/min. (b) Hydrodynamic voltammogram of Ru(bpy)3Cl2 oxidation: fluorescence detection of M(II) starting material; small (100-nL) electrode, 100 µL/min of 0.1% TFA/0.1 M NaClO4/acetonitrile; sweep rate 25 mV/min.

wave potentials by flow injection, using a coulometric electrode. An aqueous mobile phase, intended to approximate separation conditions, was used. In this mobile phase, 0.1% TFA/3% 1-propanol/36% acetonitrile, (pH ∼2), E1/2 values were 1.06 V vs Ag/ AgCl for the tyrosine-containing peptide YGGFL and 0.99 V vs Ag/AgCl for the tryptophan-containing peptide GWGG. In ref 20, the equilibrium potential of tyrosine was found to have a slope of ∼60 mV/pH below pH 10, while tryptophan did not exhibit pH dependence below pH 4. The M(III/II) equilibrium potentials of Os(bpy)3 and Ru(bpy)3 are pH independent. These values, +0.66 and +1.05 V vs Ag/AgCl, bracket a substantial portion of the M(III/II) equilibrium potentials for diimine complexes of ruthenium and osmium,21-23 analogous to the available potential range of an electrochemical detector. The higher potential of the M(III/ II) couple of Ru(bpy)3 makes it the choice for the task at hand, i.e., detection of tryptophan- and tyrosine-containing peptides in an acidic mobile phase. Given the chemical irreversibility of tryptophan and tyrosine oxidation and the energetically favorable, “downhill” nature of the electron transfer, an excess of RuIII(bpy)3 should oxidize these compounds completely, given sufficient time. Assuming rapid (21) Juris, A.; Balzani, V.; Barigelletti, S.; Campagna, S.; Belser, P.; Von Zelewsky, A. Coord. Chem. Rev. 1988, 84, 85. (22) Kober, E. M.; Sullivan, B. P.; Dressick, W. J.; Caspar, J. V.; Meyer, T. J. J. Am. Chem. Soc. 1980, 102, 7383. (23) Kober, E. M.; Caspar, J. V.; Lumpkin, R. S.; Meyer, T. J.; J. Phys. Chem. 1986, 90, 3722.

Figure 4. Reversion of RuIII(bpy)3 to RuII(bpy)3 as a function of pH. The buffers used are indicated on the figure. Large (1-µL) electrode. The M(II) form (10 µM) in 0.1% TFA/0.1 M NaClO4/50% acetonitrile was pumped through the porous electrode at 600 µL/min and oxidized to M(III) at +1325 mV vs Ag/AgCl. A second pump supplied buffers at 600 µL/min. Both RuII(bpy)3 and buffers were injected from large loops rather than pumped from reservoirs. Reaction time after mixing the two streams was ∼2 s. In the absence of buffer, M(III) was supplied at 98% efficiency.

mixing, the time scale would depend on the concentrations of the reactants. Reversion of RuIII(bpy)3 to RuII(bpy)3 as a Function of pH. The mixing time in our detection system was ∼2 s. Given published rate constants24 for spontaneous reduction of RuIII(bpy)3 in water, it should be possible to extend the use of this reagent from pH 2 to higher pH values. In the immediate application, dynorphin A fragments, the lower equilibrium potentials of tryptophan and tyrosine oxidation at higher pH would provide a higher driving force for oxidation by RuIII(bpy)3. Mixing RuIII(bpy)3 with a flow of buffer downstream of the porous electrode provided a means of simulating postcolumn reaction conditions. The results (Figure 4) indicated that, with the existing system, RuIII(bpy)3 reagent would be most suitable below pH 4. This was unexpected. The half-life of RuIII(bpy)3 in water was estimated from ref 24 to be 82 ( 0.3 min at pH 4 and 4.8 ( 1.8 min at pH 8 (reported uncertainties are 95% confidence intervals). The lifetime of RuIII(bpy)3 mixed with buffers was much shorter than this, on the order of seconds. A qualitative test to confirm the stability of RuIII(bpy)3 was performed, using a relatively high concentration of 50 µM. At this concentration, the colors of both the M(II) and M(III) forms are highly visible. When the effluent from a point immediately downstream of the porous electrode was collected in a test tube, the pale green RuIII(bpy)3 product (in 0.1% TFA, 50/50 acetonitrile/water) was easily distinguishable from the bright yellow-orange RuII(bpy)3 starting material. The pale green was observed to be stable for ∼10 min. The color reverted almost instantly to yellow-orange when a drop of pH 10 buffer was added. The acetate, maleate, and borate buffers used for pH control have all been used successfully in conjunction with electrochemical detection in this laboratory. The reason for the lower than expected RuIII(bpy)3 stability in these buffers is unknown. Ac(24) Creutz, C.; Sutin, N. Proc. Natl. Acad. Sci. U.S.A. 1975, 72, 2858.

Figure 5. Flow injection analysis of potassium ferrocyanide ([), YGGFL (0), and GWGG (O). Top panel shows calibration curves for all three compounds. Lower panel shows appparent number of electrons, napp, for the two peptides using a linear regression of ferrocyanide peak area as a 1-electron standard. RuIII(bpy)3 was synthesized as in Figure 3. Test compounds, dissolved in 5% 2-propanol, were injected using a 20-µL loop and 600 µL/min of 5% 2-propanol.

celeration of reduction by water or hydroxide ion by a catalytic effect of some buffer component is one possibility. Oxidizable impurities are another possibility; the high reactivity of RuIII(bpy)3 produces a general sensitivity to electron-donating species. Flow Injection Analysis of the Peptides YGGFL and GWGG. Dynorphin A has two oxidizable amino acid residues, tyrosine at the N-terminus and tryptophan in position 14. The peptide leu-enkephalin, Tyr-Gly-Gly-Phe-Leu or YGGFL, corresponds to residues 1-5 of dynorphin A. The peptide Gly-Trp-GlyGly, or GWGG, was chosen as a model compound for the oxidation of tryptophan in the interior of a peptide. Potassium ferrocyanide was also injected as a 1-electron standard to permit an estimate of napp for tryptophan and tyrosine oxidation by the ratio of peak areas; we assume that the ferrocyanide reaction has gone substantially to completion. Results (Figure 5) indicated that fluorescence detection based on liberation of RuII(bpy)3 from oxidation of these residues by RuIII(bpy)3 was feasible. The value of napp was estimated at 0.83 ( 0.06 (all points) for YGGFL and 4.7 ( 0.5 (below 2.5 µM) for GWGG under these conditions (uncertainties are 95% CI). Electron yields for these two peptides were also measured coulometrically by flow injection through a porous electrode in a similar mobile phase (0.1% TFA/30% Analytical Chemistry, Vol. 71, No. 8, April 15, 1999

1509

acetonitrile). Values obtained were 2.38 ( 0.04 for YGGFL and 2.36 ( 0.15 for GWGG. The residence time in the porous electrode was used for coulometry was estimated to be 2 s, based on flow rate and cell volume. For both tryptophan and tyrosine, an initial 1-electron, 1-proton oxidation forms a radical25,26 which can either dimerize or undergo a second 1-electron oxidation.27,28 The 2-electron oxidation products are then susceptible to solution chemical reactions, producing products that can undergo further oxidation. Electrochemical studies19 found that tryptophan yielded 2 electrons on a short (voltammetric) time scale and g4 electrons at longer times (bulk electrolysis), with a number of alternative reaction pathways postulated. For the coulometric electrochemical oxidations, we obtained a yield of ∼2 electrons for both peptides. For estimates of napp from solution reactions with RuIII(bpy)3, we apparently obtained a minimum number of electrons (∼1) transferred for YGGFL, and a maximum number (>4) for GWGG. The results indicate the unsurprising conclusion that oxidation mechanisms on electrode surfaces and in solution can differ. Chromatography of Dynorphin A Fragments. The gradient separation method was adapted from literature methods.29-31 The elution order of the peptide set was established using UV absorbance detection at 210 nm. (a) Chromatography of 1 µM Fragments Using Fluorescence Detection. At the time that this chromatography was done, only the large 1-µL (0.039-in.-i.d. Nafion) electrode design had been tested, and one of these was used to generate RuIII(bpy)3. The original 2-mm flow cell of the Varian detector was used. The filtered halogen light bulb excitation source produced a high background when used at maximum sensitivity, regardless of whether Ru(bpy)3 reagent was flowing or not. Nonetheless, detection was quite feasible. A 20-µL loop supplied 20 pmol of each peptide. The resulting chromatogram is shown in Figure 6. Six of the dynorphin A fragments contain only one oxidizable amino acid residue, the N-terminal tyrosine. One of the fragments (2-13) contains neither tryptophan nor tyrosine and was not detected. Dynorphin A itself contains both tryptophan and tyrosine and has a relatively large peak area. Using the ferrocyanide calibration of the electrode response, napp values were calculated. The first three peaks, which are cleanly separated, have values of 0.95, 0.94, and 1.06 electrons, average 0.98 ( 0.17 (95% CI). The dynorphin A peak has a value of 4.61 electrons. These values are reasonably consistent with the earlier flow injection results using model compounds (0.83 ( 0.06 for YGGFL and 4.7 ( 0.5 for GWGG). A notable feature of this chromatogram is that the peak area is proportional to the number of electrons transferred. In contrast, chemiluminescence methods utilizing RuIII(bpy)3 exhibit sensitivities that vary from one compound to another.3-7 In addition, the (25) Prutz, W. A.; Butler, J.; Land, E. J. Int. J. Radiat. Biol. 1983, 44, 183. (26) Jovanovic, S. V.; Harriman, A.; Simic, M. G. J. Phys. Chem. 1986, 90, 1935. (27) Tsai, H.; Weber, S. G. Anal. Chem. 1992, 64, 2897. (28) Nguyen, N. T,; Wrona, M. Z.; Dryhurst, G. J. Electroanal Chem. 1986, 199, 101. (29) Goldstein, A.; Fischli, W.; Lowney, L.; Hunkapiller, M.; Hood, L. Proc. Natl. Acad. Sci. U.S.A. 1981, 78, 7219. (30) Tachiabana, S.; Araki, K.; Ohya, S.; Yoshida, S. Nature 1982, 295, 339. (31) Spampinato, S.; Canossa, M.; Bachetti, T.; Campana, G.; Murari, G.; Ferri, S. Brain Res. 1992, 580, 225.

1510 Analytical Chemistry, Vol. 71, No. 8, April 15, 1999

Figure 6. (a) Chromatography of eight 1 µM dynorphin A fragments with fluorescence detection. The fluorescence detector used the original 2-mm flow cell and halogen light bulb excitation: column, 2.1 × 50 mm reversed-phase C18; solvent A, 0.1% TFA/water; solvent B, 0.1% TFA/60% acetonitrile; gradient 5-80% B in 0-40 min, 0.25 mL/min; postcolumn reagent, 10 µM Ru(bpy)3Cl2/0.1% TFA/ 0.1 M NaClO4/acetonitrile, 600 µL/min; oxidized to M(III) at +1600 mV vs Ag/AgCl. (b) Enlargement of selected region of chromatogram in (a). The sequence of dynorphin A is depicted above the chromatogram. The two electroactive amino acid residues are indicated with asterisks. The sequences of eluted peptides are indicated on the chromatogram.

baseline during the gradient is remarkably flat; electrochemical detection during gradient runs often exhibits large shifts in baseline. (b) Axial Illumination Laser-Induced Fluorescence (LIF) Modification of Varian Fluorichrom Detector. This arrangement, described in the Experimental Section, improved sensitivity significantly. The laser light is largely confined within the walls of the capillary by total internal reflection.32-34 The background

Figure 7. Suppression of background using a porous electrode as a scrubber for the mobile phase, expressed as background fluorescence signal as a function of applied potential on the mobile-phase scrubber. Axial illumination laser-induced fluorescence setup: Mobile phase, 0.1% TFA/2% 2-propanol/10%acetonitrile/90% water, 50 µL/ min; postcolumn reagent, 5 µM Ru(bpy)3Cl2/0.1% TFA/0.1 M NaClO4/ acetonitrile, 100 µL /min; oxidized at 1300 mV vs Ag/Ag+. The large (1-µL) electrode was used to scrub mobile phase and the small (100nL) electrode was used to generate RuIII(bpy)3. Low laser power was used; at high laser power the maximum background signal was off scale.

signal was much lower, despite using all available argon ion laser lines, unfiltered. (c) Suppression of LIF Background by Porous Electrode. It was discovered that, for the LIF detection apparatus, scrubbing of the chromatographic mobile phase with a second porous electrode was essential. In the absence of precolumn scrubbing, a postcolumn stream of 5 µM RuIII(bpy)3, 100 µL/min, spontaneously reverted almost 100% to the M(II) form when mixed with the chromatographic mobile phase (50 µL/min). To avoid destruction of the tryptophan and tyrosine redox activity in the dynorphin A fragments, the scrubber electrode was installed upstream of the injector (Figure 1c). The electrode used for scrubbing was the 1-µL woven-PTFE-encased electrode. The woven PTFE shroud provided some resistance to column back pressure (∼450 psi for the 1-mm C18 column used). A voltammogram of the suppression of RuII(bpy)3 fluorescent background as a function of Eapplied on the precolumn scrubber is shown in Figure 7. The mean of the E1/2 values for the forward and reverse waves, 0.71 V vs Ag/AgCl, corresponds closely to the standard potential of the Fe(III/II) couple. If the fluorescent background was indeed due to stoichiometric reduction of RuIII(bpy)3 by some species of iron in the mobile phase, then clearly it was Fe(II) and not Fe(III) that was responsible. Metal ions are known to be leached from metal chromatographic pumps, and much of the iron is present as Fe2+.35 The indirect voltammetry, obtained by converting the electrochemical activity of a nonfluorescent species (Fe2+) in a flowing (32) Taylor, J. A.; Yeung, E. S. Anal. Chem. 1992, 64, 1741. (33) Abbas, A. A.; Shelly, D. C. J. Chromatogr. 1993, 631, 133. (34) Abbas, A. A.; Shelly, D. C.; J. Chromatogr., A 1995, 691, 37. (35) Franlin, G. O. Am. Lab. 1985, 9, 65.

Figure 8. Chromatography of 100 nM dynorphin A fragments with axial illumination laser-induced fluorescence detection. Mobile phase was scrubbed as in Figure 6, Eapp +900 mV vs Ag/AgCl: column, 1 × 100 mm reversed-phase C18; solvent A, 0.1% TFA/1% 2-propanol/ water; solvent B, 0.1% TFA/1% 2-propanol/acetonitrile; gradient, 5-90% B in 0-60 min, 50 µL/min; postcolumn reagent as in Figure 6, 100 µL/min.

stream into a fluorescence signal, is intriguing and as far as we know, novel. Useful voltammetry of compounds present in trace amounts is possible with this approach, at concentrations far below those accessible by conventional voltammetry. (d) Chromatograms Using LIF Detection. Chromatography with axial illumination LIF detection (Figure 8) employed a 1-mm column and a large (140-µL) injection loop. The use of such a large loop, which is comparable to the void volume of the column, is made possible by gradient elution. At the beginning of the run, compounds preconcentrate at the head of the column and begin to elute when a sufficient organic content has been reached during the gradient run. The use of this loop is in line with previous practice in this laboratory and was intended to resemble the analytical method that ultimately will be used for biological samples containing dynorphin fragments. Comparison of sensitivity between runs using injection loops of different sizes can be based on moles of peptide injected rather than concentration. The relatively flat baseline during the gradient resembles that of the earlier separation of 1 µM dynorphin A fragments. A flow rate of 50 µL/min was used in the 1-mm column, with a postcolumn reagent flow rate of 100 µL/min. The small, 100-nL porous electrode was used to generate RuIII(bpy)3 in the postcolumn reagent stream. Dynorphin A and fragments, 100 nM (14 pmol) each, were initially separated using 5-90% solvent B, 0-60 min. The electron yields, estimated by RuII(bpy)3 flow injection, were 0.61 ( 0.19 (95% CI) for the first five peaks and a minimum of 3.4 for the dynorphin A peak, which was off scale. The S/N ratio for the first three peaks of the 1 µM (20 pmol) dynorphin separation was ∼60; for the corresponding peaks from the 100 nM (14 pmol) dynorphin separation, the S/N ratio was ∼500. A detection limit of 80 fmol (S/N ) 3) can be estimated from this initial investigation of the redox fluorescence technique. A detection limit of 6-40 fmol has been determined for electrochemical Analytical Chemistry, Vol. 71, No. 8, April 15, 1999

1511

detection of electroactive peptides in our group.36 Improvements in detection equipment are expected to result in significantly lower detection limits for this fluorescence technique. GENERAL COMMENTS ON THE REDOX FLUORESCENCE METHOD Comparison to Fluorescent Reactive Labels. Existing fluorescence detection techniques37 used in conjunction with separation employ reactive labels such as OPA or NDA or, in rare cases, native fluorescence (e.g., tryptophan or tyrosine fluorescence). Unlike reactive labeling, the redox fluorescence technique described in this paper does not involve formation of a covalent adduct of the analyte. Reactive labeling is typically performed as a precolumn reaction, and the separation properties of the analyte will then be changed by the tagging process. Redox fluorescence detection can be used in conjunction with existing separation procedures developed using chemically unmodified analytes. Because the chemical processes causing onset of fluorescence in the two methods are fundamentally different, the selectivities of fluorescent labeling and redox fluorescence are different. Fluorescent labels react with specific functional groups of the analyte that are suitable for the specific coupling reaction, such as an analyte with a primary amine reacting with o-phthalaldehyde. The redox fluorescence method has a selectivity almost identical with electrochemical detection, relying upon electron transfer between the analyte and the reagent. It is anticipated that the ultimate sensitivity of laser-induced detection using redox fluorescence will be of the same order as results obtained using fluorescent labeling, with quantum efficiency being a significant influence. The quantum efficiency (Φ) of RuII(bpy)3 is 0.042, and among similar diimine complexes of ruthenium and osmium values of Φ range from 0.005 to 0.4 Stability of Porous Electrodes. In long-term use of porous carbon electrodes to scrub mobile phases, and to oxidize RuII(bpy)3, no decrease in efficiency has been observed. All electrode failures observed have been due to rupture of the Nafion tubing under excessive pressure rather than fouling. It should be noted that, in the configuration used, injected samples do not pass through the porous electrodes and also that the postcolumn reagent and mobile phases are clean and contain an organic solvent (acetonitrile). These conditions minimize the likelihood of fouling. Use of the electrodes for electrochemical scrubbing of, for example, biological samples might cause eventual fouling of the electrodes. The measures used with commercial porous electrodes to prevent fouling, such as guard cells, would be advisable in this case. Anticipated Performance with Real-Sample Analysis. The dynorphin peptide mixture used in this study was a relatively small set of pure compounds. Real samples, including biological samples, (36) Chen, J. G.; Weber, S. G. Anal. Chem. 1995, 67, 3596. (37) Tao, L.; Kennedy, R. T. Trends Anal. Chem. 1998, 17, 484.

1512 Analytical Chemistry, Vol. 71, No. 8, April 15, 1999

environmental samples, and organic reaction mixtures, may present additional difficulties. Minor susceptibility to fluorescence quenching by oxygen and some transition metals is to be expected. The main anticipated limitation is that detection using this particular reagent may be limited to acidic conditions, perhaps pH 4 or below. In predicting selectivity and sensitivity to electrondonating interferences with real samples, the best guide is to assume the RuIII(bpy)3 reagent will be similar to an electrochemical detector operating at ∼+1100 mV vs Ag/AgCl. A solid electrode operating at this potential in aqueous solution would also be limited to somewhat acidic conditions due to background current from reactions on the surface of the electrode. The same methods used to improve selectivity in electrochemical detection, including screening electrodes, sample cleanup, and modification of the separation, should be effective with the redox fluorescence method. Gradient separation, more easily used with redox fluorescence detection than with electrochemical detection, should facilitate resolution of analytes from interferences. As noted in the introduction, Ru(bpy)3 is one member of a large family of related complexes. Values for oxidation potential, absorption and emission wavelengths, quantum efficiency, and fluorescence lifetime vary considerably within the group. Optimization for a particular analysis, including operation under more alkaline conditions, will include choosing a suitable fluorescent complex. CONCLUSION The feasibility of an analytical method utilizing fluorescence following electron transfer has been demonstrated. The high reactivity of the reagent used, RuIII(bpy)3, is both an advantage and a disadvantage. On one hand it permits detection of compounds having high oxidation potentials. On the other hand, a general sensitivity to electron-donating interferences will occur. Additional experiments to characterize the system are desirable. However, the essential features of such an analytical system have been established: efficient generation of M(III) reagent, removal of interfering redox activity from the sample stream, reaction of typical analytes with the M(III) reagent to produce M(II), fluorescence detection based upon M(II), and calibration of the detector by injecting either a one-electron oxidation standard or M(II). ACKNOWLEDGMENT Initial evaluation of reversed-phase gradient separation of dynorphin A and fragments using UV absorbance detection was performed by Wanlin Xia. Parts of this work were supported by U.S. DOE Contract AC04-94AL85000. We thank NIH for financial support through Grant GM-44842. Received for review October 29, 1998. Accepted February 6, 1999. AC981181E