Anal. Chem. 1995,67, 541-551
This Research Contribution is in Commemoration of the Life and Science of 1. M. Kolthoff (1894- 1993).
Rotating Ring-Disk Electrode Study of Copper([[) Complexes of the Model Peptides Triglycine, Tetraglycine, and Pentaglycine Steven J. Woltman, Melinda R. Alward, and Stephen 0. Weber*
University of Pittsburgh, Pittsburgh, Pennsylvania 15260
The reversible electrochemistry of the Cu(II)/Cu(III) couple was investigated for the copper(II) complexes of triglycine (G3),tetraglycine (G4),and pentaglycine (G5)in alkaline solution using a rotating *-disk electrode W E ) . The study was motivated by the need to elucidate electrochemical processes occurring in dual electrode postcolumn detection of peptides. The disk electrode served as the anode and the ring electrode as the cathode. The electrode was used in both linear sweep voltammetry and constant potential, varied rotation speed modes. Redox waves for two generic forms of the complexes, Cu(1I)-NNNN and Cu(1I)-"NO, were identified, with respective Ell2 values of -0.45 and 0.7 V. It was found that the G3 complex underwent an ECE-like process at the anode that magnified the anodic signal and suppressed the cathodic signal. The G4 and 6 5 complexes were subject to two CE processes, the C reactions being (i) deprotonation of Cu(II)-"NO to Cu(I1)-NNNN and (ii) loss of hydroxide ion by Cu(I1)-NNN(0H-), a variant of Cu(I1)-"NO. An additional C reaction, dissociation of a carbonate variant of Cu(I1)-"NO, occurred in 0.2 M carbonate buffer. Visible absorbance measurements assisted in assignment of these forms. Measurements of diffusion coefficients of the complexes were performed by Taylor-Aris laminar flow axial dispersion measurements. The analytical implications for these findings are discussed. Recently, a selective chromatographic detection scheme for peptides has been introduced, based on the reversible electrochemistry of the Cu(II)/Cu(III) couple in polydentate peptidecopper complexes.lP2 The typical embodiment of this detection scheme is sequential dual electrode detection: an upstream anode oxidizes copper00 to copper(III), and a downstream cathode reduces the copper(I1I). The purpose of the present study is to characterize the electrochemistry of copper complexes of the model peptides triglycine, tetraglycine, and pentaglycine (G3, G4, and G5) in support of electrochemicaldetection work. The copper ("biuret") complex of a given peptide develops through successive forms as a function of pH and the various pKa (1)Warner, A. M.; Weber, S. G. Anal. Chem., 1989,61,2664. (2)Chen, J. G.;Woltman, S. J.; Weber, S. G., J. Chromatogr., submitted. (3) Billo, E. J. Inorg. Nucl. Chem. Lett. 1974,10, 613. 0003-2700/95/0367-0541$9.00/0 0 1995 American Chemical Society
values for the deprotonations of the peptide amido nitrogens.4-6 Complexation by the peptide backbone begins at the amine terminus, if it is free, and successively coordinates to a changing set of carbonyl oxygens and deprotonated amido nitrogens with increasing pH. The geometry of the complex is square planar, with axial coordination by water molecules in the absence of other axial ligands. The coordinated copper(II) can be oxidized chemically or electrochemically to copper(III); the stability of the copper(III) state varies considerably depending on the identity of the coordinating ligands. The deprotonated amido nitrogens are strong donor ligands; in general, the more of these nitrogens that coordinate the copper, the lower the oxidation potential of the Cu(II)/Cu(III) couple and the longer the lifetime of the Cu(I1I) The identity of the coordinating ligands can often be deduced from the absorbance maximum of the electronic spectrum of the coordinated copper.3 Longer absorption wavelength correlates with weaker ligands, more positive oxidation potential, and lower Cu(I1I) stability.5 At present, standard reaction conditions for the postcolumn reaction of peptides with Cu(II) following their separation by HPLC are excess copper tartrate in pH 9.8,0.2 M carbonate buffer. Under these conditions, a tripeptide complex will typically coordinate copper0 with the ligands "NO: an amine nitrogen, two deprotonated amido nitrogens, and terminal carboxylate oxygen. Peptides of four residues and longer will form Cu(II)NNNN or an equilibrium mixture of Cu (II)-NNNN and Cu(II) "NO, where 0 is carbonyl oxygen; CuOD-NNNO forms Cu(II)-NNNN by deprotonation.6 Scheme 1 illustrates this reaction. The pKa of the reaction shown in Scheme 1is 9.15 for la, G4 (and we have conlimed that by potentiometric titration) and 8.04 for lb, G5 (5). The rotating ring-disk electrode W E ) is particularly suited to modeling dual electrode detection. The flow pattern introduced by the rotation of the electrode draws analyte species first over the central disk electrode and then over the concentric ring electrode. In this study, the disk electrode served as the anode and the ring electrode as the cathode. The RRDE was used in two modes: linear potential sweep hydrodynamic voltammetry and varied rotation speed at constant potential. In the latter mode, (4) Margerum, D. W. Pure Appl. Chem. 1983,55(1),23.
(5) Bossu, F.P.; Chellappa, K L;Margerum, D. W. J. Am. Chem. SOC.1976, 99,2195. (6)Margerum, D.W.; Wong, L. F.; Bossu, F. P.; Chellappa, K L.;Czarneck, J. J.; Kirksey, S. T., Jr.; Neubecker, T. A. Bioinorg. Chem. 1976,281. Analytical Chemistry, Vol. 67,No. 3,February 1, 1995 541
Scheme 1
.do
-H+
O\,N
cases of the fraction of electroactive and nonelectroactive forms. Visible spectra have provided qualitative indications of the distribution and identities of different forms of the complexes. Direct determinations of D for the copper(II) complexes of GB, G4, and Gs have been made using the axial dispersion of a concentrated spike of analyte during prolonged laminar flow through an open tube of small diameter, a hydrodynamic technique pioneered by Taylor and A r i ~ . ~ Briefly, l J ~ provided that
AN,
6 1 a R=Ob R=NHCH2COO.
246
ur$
the disk electrode signal obtained for a simple, mass transportlimited process with no complicating equilibria will obey the Levich e q ~ a t i o n : ~ , ~
I = i/C* = 0.620nFAD2'3v-1/6w1/2
(1)
where i (current) is in mA, C* (concentration) is in M, D (analyte diffusion coefficient) is in cm2/s, IJ (kinematic viscosity) is in cm2/ s, o (rotation velocity) is in rad/s, n is number of electrons transferred per reaction, and F is the Faraday constant. The geometric disk area, A, is 0.2475 cm2 for the electrode used in this study. A fraction of a stable product of the reaction at the disk electrode will reach the ring electrode surface. The theoretical value of this fraction, known as the efficiency, N, is rotation speed independent and has the value 0.37 for the particular electrode employed here. An unstable product will have a value of N less than 0.37. The values given for A and N are both manufacturer's values. The concentration of the electroactiveform of the analyte, C, may or may not equal the bulk concentration of analyte, i.e., an equilibrium can exist in solution between electroactive and nonelectroactive forms. Depletion of the electroactive form by oxidation will then perturb the equilibrium near the electrode surface, with some amount of the nonelectroactiveform converting to the electroactive form to restore equilibrium concentrations (a CE process). If this relaxation process is fast with respect to the time scale of the measurement, then the apparent concentration of the analyte will be the total bulk concentration of the electroactive and nonelectroactive forms. If the rate of relaxation is relatively slow, then the apparent concentration of the analyte will be less than its total bulk concentration. The time scale of an RRDE measurement is determined by the rotation speed. The higher the rotation speed, the shorter the residence time of the analyte near the electrode surface. Provided that C is known, if n is known, D can be determined, and vice versa. However, generally neither n nor D is known with certainty a priori. In the present study, some means of clarifying the situation are available. Often, n can be assigned a specgc integer value from knowledge of the reaction occurring at the electrode. For the Cu(II)/Cu(IIr) couple in peptide biuret complexes that lack electroactive amino acid residues (Tyr, Trp, and Cys), c* can be assumed to equal bulk concentration and n to equal 1 with a high degree of confidence in some cases (but not all,as shall be seen). RRDE measurements over a wide range of rotation speeds have permitted estimates to be made in some (7) Albery, W. J.; Hitchman, M. L Ringdisk Electrodes; Clarendon Press: Oxford, 1971; p 16. (8) Adams, R N. Mass Transfer by Forced Convection. In Electrochemishy at Solid Electrodes; Bard, A J., Ed.; Marcel Dekker: New York, 1969.
542 Analytical Chemistry, Vol. 67, No. 3, February I , 7995
>
10
(24
and
then
where uro/D is the Peclet number (dimensionless), Dt/r,2 is dimensionless time, D, is the total axial dispersion coefficient, ro is the radius of the tube, t is time elapsed from injection, and u is the average linear velocity of the solvent flowing in the tube.13 Dax is determined from the variance of the axial distribution of the analyte peak emerging from the end of the tube, and the expression for D, is then solved for D. Direct measurement of n by quantitative bulk oxidation of copper-peptide complexes in a high efficiency flow ele~trode'~ is a subject for future work. We have based a strategy for detection of peptides on the electrochemistry of the copper-peptide backbone electrochemistry, yet there has been no in-depth study of this electrochemistry at electrodes. As the chemistry is applied in a dual electrode electrochemical detector, the rotating ring-disk is the appropriate method for the investigation. We have thus undertaken to investigate the rotating ring-disk behavior of the tri-, tetra-, and pentaglycine complexes of Cu(ID in basic solution. The most noticeable feature of the chemistry of these complexes, as expressed through the voltammetry, is the presence of multiple forms of the complexes in equilibria that are slow enough to be resolved on the W E time scale. EXPERIMENTAL SECTION
Peptides (G3, G4, and Gs; all from Sigma), ACS grade copper sulfate pentahydrate,boric acid, potassium nitrate, perchloric acid, nitric acid (Fisher), sodium hydroxide, and sodium bicarbonate (Mallinckrodt) were used without further purification. All solutions were made in Milli-Q house-deionized water. Carbonate buffer, 0.2 M, was prepared as a solution of sodium bicarbonate (at the stated concentration) and sodium hydroxide in 2:l molar ratio. Borate buffer, 0.05 M, was initially prepared as a solution (9) Youngblood, M. P.;Chellapa, K L.; Banninster, C. E.; Margerum, D. W. Inorg. Chem. 1981,20, 1742. (10) Bard, A J.; Faulkner, L. R Electrochemical Methods-Fundamentals and Applications; Wiley: New York, 1980. (11) Taylor, G.I. PYOC.R. SOC.London 1953,AZ19, 196. (12) Aris, R Proc. R. SOC.London 1956,A235, 67. (13) Weber, S. G.;Cam, P. W. The Theory of the Dynamics of Liquid Chroamafogrophy. In h e m , Molecules and Methods; Hirschielder, J. O., et al., Eds.; Wiley: New York, 1989. (14) Davis, B. R; Weber, S. G. Anal. Chem. 1994,66, 1204-1207.
of 0.05 M boric acid and 0.25 M mo3. Boric acid (0.05 M) was included in carbonate solutions used at pH 8-9. The pH of carbonate buffer was adjusted with 50%NaOH solution and with either concentrated hydrochloric acid or concentrated perchloric acid. Borate buffer pH was adjusted with 50% NaOH and perchloric acid only. Unbuffered copper-peptide solutions used for spectroscopy were pH-adjusted with dilute NaOH and dilute perchloric acid. The pH values of all solutions were checked with a meter immediately before use. The meter was standardized daily. The rotating ring-disk apparatus consisted of a DT-29 glassy carbon/glassy carbon ring-disk electrode, ASR-2 analytical rotator, and RDES potentiostat (all from Pine Instrument Co., Grove City, PA). A platinum grid counter electrode and BAS Ag/AgCl (3 M NaCl) reference electrode were used. All potentials are reported relative to the reference. Before each run, the electrode was wet polished with 0.05 pm y-alumina powder and rinsed with a stream of deionized water from a wash bottle. The electrode was then placed in the rotator and conditioned in buffer for 1 min at 800 mV disk, 0 mV ring. The potentiostat signals (disk and ring) were filtered with a Wave-Tek Model 852 filter (WaveTek, San Diego, CA) set for low pass with 11 Hz cutoff frequency. A DT2802 chromatography board (Data Translation, Inc.) was used to interface the filtered analog signal to a DTK personal computer. RRDE data were collected using EZChrom chromatography software. Data collection was triggered manually while the sweep generator was simultaneously switched on. To plot hydrodynamic voltammograms, EZChrom files were converted to ASCII format using the data export utility and then processed in STATA statistical software (Computing Resource Center, Santa Monica, CA). Room temperature was 25 i 1 "C during all RRDE experiments. For hydrodynamic voltammetry, the disk potential was swept linearly from 0 mV to the maximum potential (900 or 1000 mv), using the sweep generator of the RDES potentiostat. For 900 mV maximum potential, a sweep rate of 600 mV/min was used; for 1000 mV maximum potential, 500 mV/min was used. The ring electrode was maintained at a constant low potential during each run: 0 mV at pH 9.8 and above and 100 mV at lower pH values to suppress the high cathodic background current that otherwise occurs. To maintain steady state behavior, a minimum rotation speed of 150 rpm was used for voltammetry. Copperpeptide solutions for RRDE experiments were prepared by combining aqueous solutions of peptide and copper sulfate in a 1:l molar ratio. After thorough mixing, addition of either dilute sodium hydroxide solution or alkaline buffer proceeded dropwise with mixing until the characteristic copper-peptide complex color developed; the solution was then diluted to volume (100 or 200 mL). Concentration was 100 pM in all cases. During HPLC analysis of biological samples, bioactive peptide concentrations reaching the detector are typically orders of magnitude less than this; however, the RRDE cannot give an acceptable signal to noise ratio at concentrations much below 100 pM. This concentration models actual analytical conditions as closely as is practical. RRDE rotation speed studies were conducted with the disk maintained at constant potentials, identified by voltammetry as being in limiting current regions. In each run, the electrode was rotated initially at a moderate speed, and then potential was applied to the disk. The non-Faradaic charging current was permitted to
decay 1min for first wave (500-600 mv) potentials and 1.5 min for second wave (800-900 mv) potentials. The rotation speed was then varied in a random sequence. Data collected during the initial period of current decay were ignored. The rotation speed was maintained at each value long enough to ensure that the signal had stabilized, ranging from seconds at the highest rotation speed (10 OOO rpm; 32.36 (rad/s)"? to tens of seconds at the lowest (10 rpm; 1.02 (rad/s)l/?. Ten to fifteen different rotation speeds were trpically applied during each run. The lowspeed pulley wheel of the rotator was used below 100 rpm. At the highest rotation speeds (6000-10 000 rpm) , a considerable amount of air entrainment occurred. However, this did not perturb the hydrodynamic behavior of the electrode, provided that the electrode was positioned off-center in the beaker to prevent stagnation of bubbles under the electrode. The system was tested for hydrodynamic linearity from 10 to 10 OOO rpm using 0.200 mM ruthenium (I10 hexaammine as a reversible oneelectron standard. The complex was dissolved in 0.2 M carbonate, pH 9.8, to model the standard experimentalconditions. Argon sparging was used to remove dissolved oxygen. The complex was reduced at the disk (Eapp = -280 mv) and re-oxidized at the ring (Eapp = 0 mv). The disk and ring data yielded linear Levich plots (9= 0,9998, n = 58) over the entire rotation speed range. Spectroscopic measurements employed an HP 8452A diode array spectrophotometer with HP 8953lA operating software. Solutions for spectroscopy (10 mL) were prepared as for the RRDE (above) but at approximately 1 mM concentration. pH was adjusted in a small beaker, and aliquots of solution were transferred to the cuvette. To obtain full-strength buffer concentrations for spectroscopy, copper-peptide solutions were combined 1:l with doublestrength buffers. Spectra of copper-peptide solutions in 1 cm glass cuvettes were measured over 350-820 nm in general scanning mode at the maximum integration time (25 s), using deionized water as a solvent blank in all cases. The peak finder function was used to identify the absorbance maxima. Determination of diffusion coefficients by laminar flow axial dispersion was according to the procedure of Taylor.ll A 2 M length of stainless steel tubing of 0.508 mm i.d. was connected between a ConstaMetric I11 pump (LDC, Riviera Beach, FL) and a Waters Model 440 absorbance detector; detection was at 254 nm. The tubing was jacketed for temperature control. The apparatus was passivated with 1 M nitric acid to prevent solute adsorption. Data were collected using a Metrabyte Vaunton, MA) DAS20 board installed in an IBM PC. The resulting ASCII files were processed for graphical output using STATA Dopamine (22.7 mM) in 0.9% w/v NaCl was used as a diffusion coefficient standard. The value obtained was (7.16 f 0.24) x loW6cm2 s-l, which is comparable to the value of (7.5 f 0.3) x cm2 s-l determined by Adams;15uncertainties are 95%confidence intervals. The copper-peptide solutions were prepared by first dissolving the peptides in deionized HzO, adding a few drops of concentrated NaOH if necessary for complete solvation. Solid CuSOc5HzO was added to form a 1:l complex of copper and peptide. The solutions were diluted with either carbonate buffer (0.2 M NaHC03,O.l M NaOH) or dilute borate buffer (5 mM borate, 25 mM KN03)so that the final peptide concentration was approximately 5 mM. The pH was adjusted with concentrated NaOH and perchloric acid to 9.8. The solutions were then filtered and kept frozen when not in use. Injections were made using a sample volume of 5 pL, and the mobile phase was of the same buffer composition as the Analytical Chemistty, Vol. 67,No. 3,February 1, 1995
543
Table 1. Copper(ll/lll) Anodic Half-Wave Potentials in Borate Buffer
peptide, voltammetric wave G3
G4, first wave G4, first wave G4, second wave
G4, first wave G5, first wave
G5, second wave Gs, first wave
0.4 I
PH
Ei/z vs Ag/AgcL v
9.80 9.15 9.80 9.15 11.60 8.04 8.04 9.80
0.71 0.45 0.45 0.72 0.43 0.43 0.71 0.45
/
0.4
/
0.3
, c
I
0.2 0.1
0.0 I 0.0
1
i
I
1
1
I
l
l
l
0.5
I
1.o
Anode Potential vs. Ag/AgCI, V
1
0.3 -
$.
-
1
u'J! nc
'
O,I0 0.08
/ / ,
0.2,'/
0.1 -
1 1
n0.0 n 1 Y."
I
-
0.0
I
1
1
1
1
1
I
I 1
1
1
0.5
I 1
002
1
1.o ".Y"
Anode Potential vs. Ag/AgCI, V
1 ,
/
l
0.0
l
l
l
l
I
l
l
i
0.5
1
l
1.0
Anode Potential vs. AgIAgCI, V
Figure 2. Voltammograms of 100 pM 1:l GdCuS04 in 0.05 M borate/0.25 M KN03, at pH 8.05 (-) and pH 9.80(- - -), both at 10 000 rpm, with a sweep rate (anode only) of 0.5 Vlmin. (a, top) Anode. (b, bottom) Cathode. The cathode was maintained at 80 mV during the pH 8.05run to suppress background current.
-
RESULTS AND DISCUSSION
0.04
1
0.02
Anode Potential vs. AgIAgCI, V
Figure 1. Voltammograms of 100 pM 1:l GdCuS04 in 0.05 M borate/0.25 M KN03, pH 9.15 (-), 9.80 (- - -), and 11.6 (- - -). Rotation speed was 10 000 rpm in each case, with a sweep rate (anode only) of 0.5 Vlmin. (a, top) Anode. (b, bottom) Cathode. The high anode current at pH 11.6has been truncated from approximately 1 mNmM to scale the figure conveniently.
sample solutions. Four to six determinationsof D for each peptide were obtained at 25 "C at a flow rate of approximately 0.03 mW min. The linear velocity was calculated from tube length and the transit time of the maximum peak height. Typical peak width was on the order of 400 s. The data were analyzed using the exponentially modified Gaussian (EMG) function of Peakfit nonlinear curvefitting software (landel Scientiiic,San Rafael, CA). 544
Analytical Chemistry, Vol. 67, No. 3, February 1, 1995
For reasons described below, electrochemical data obtained using 0.05 M borate buffer are considered more representative of simple peptide-copper complex behavior than data obtained in 0.2 M carbonate buffer. Results described were obtained using such a weak borate buffer unless otherwise indicated. High rotation speed voltammograms of G4 and Gs copper complexes, taken within a pH unit of literature pKavalues, exhibit two oxidation waves; E1/2values are approximately 0.45 and 0.7 V vs Ag/AgCl, respectively (Table 1). Cu(II)-G3 at pH 9.80 exhibits a single wave at -0.7 V, matching the second wave potential of the longer peptides. The first wave is assigned to 2 and the second to 1, which is chemically similar to the G3 complex. Corresponding cathodic waves at the ring electrode represent reduction of Cu(I1D to Cu(ID. Titration of G4 (Figure 1) or Gs (Figure 2) to more alkaline conditions progressively eliminates the second wave as the first wave grows to comprise the entire signal. It is unusual to find individual waves for acid and base forms of a molecule in aqueous solution, because proton transfers to heteroatoms are generally fastg However, the large molecular change that accompanies this proton transfer causes the rate constant to be rather low: at pH 10.8, the observed relaxation rate constant (stopped-flow) is 74 f 5 s - ~ . ~
Table 2. Estimated Equilibrium Percentages for Doubly and Trlply Deprotonated Copper(l1)-Peptlde Species Based on Literature Equlllbrlum Constants519
peptide
doubly deprotonated
PH
hydroxide equilibrium CU(II)-NNNN (OH-)
triply deprotonated
Cu(II)-NNNN 44.5 73 89
G3
9.80
100
G4 G4 G4
9.15 9.80 11.60
50 18
0.3
50 82 99.7
G5 G5
8.04 9.80
50 1.7
50 98.3
Above pH 9, the anode signal increases at applied potentials exceeding 0.8 V. The phenomenon occurs to a greater extent when the peptide complex is present than in blank solutions; the increased current is therefore not removed by background subtraction. The ring signal does not increase in concert with this jump in anode current. We have not analyzed this process yet. Cu(I1)-"NO Deprotonation CE Process. When the more easily oxidized 2 is selectively oxidized at a low potential (0.4-0.6 V) in the presence of 1, a CE process takes place, as shown in Scheme 2.
5.5 9 11
0.2
0.1
0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Scheme 2
Anode Potential vs. AgIAgCI, V
-H+,k,
l
x
0.10 -
2
2&2+
Q
+e-
z
l deprotonates to form additional 2 as the population of 2 is depleted by oxidation to Cu(I1I)-NNNN, or 2+. The ability of 1 to feed the supply of 2 and ultimately yield current is higher the longer the time given for the C reaction. Thus the process is the most evident at low rotation velocities. At high enough rotation velocities,the C reaction is of little influence, and the current arises from the initial concentration of 2 in the system. At pH 9.80, Cu(II)-Gd @Ka9.15) will be 82%triply deprotonated and 18%doubly deprotonated (Table 2). Voltammograms in pH 9.80 borate buffer (Figures 3 and 4,solid lines) exhibit a ratio of the iirst and second wave heights corresponding to the equilibrium concentrations of 1 and 2 only at the maximum rotation speed (10 000 rpm). To equate wave height with concentration requires that the diffusion coefficients of the two species be the same. While we do not know that they are, it seems a not unreasonable working assumption. At lower rotation speeds, the second wave is reduced considerably, virtually disappearing at 150 rpm. The residence time at lowest rotation speed permits quantitative deprotonation of 1 to 2 at the lower potential; no additional current is then obtained by increasing the applied potential. These processes are more easily appreciated by looking at Levich plots and replots of current-rotation speed data. For a CE process, the disk electrode current is the following function of rotation speed:1°
Z = i/C* = nFAD/[1.61D"3v"6w-"2
+ D"2/(Kk'/2)l
(3)
f
0.05-
M
0.00-I I I I I 1 I I l 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Anode Potential vs. AgIAgCI, V
Figure 3. Comparison of voltammograms obtained from 100 ,uM 1:l G&uS04 in pH 9.80,0.05 M borate/0.25 M KN03 (-) vs pH 9.80,0.2 M carbonate (0).(a, top) Anode. (b, bottom) Cathode. From bottom curve to top curve in each case, rotation speeds are 150, 1000,and 10 000 rpm.
+
provided that (D/v)'l3 > pK,). Recall that the yintercept of the linear portion of these should correspond to the Levich slope of a system without the C reaction (initely fast). The actual slopes of Levich plots of the pH 11.6 disk and ring data, 10-100 rpm, are indicated on the plots by the circled points. Good agreement can be estimated visually. The vertical offset of the pH 11.6 anode function relative to the circled point is due to a slight excess anode current at pH 11.6; the transformed points for this reason do not precisely intercept the value for the Levich slope obtained from the same data points. At higher Id (higher w ) , the mass transport outruns the C step, so the current is expected to reach a constant value corresponding to the amount of electroactive form of the complex present in solution. These constant values, calculated on the basis of the pKa and representing free Cu(II)-NNNN in the bulk solution, are indicated for each pH value by dashed lines. Again, the agreement is satisfying. Anode E(CE), Process. A voltammogram of Cu(ID-G3 at pH 9.80 exhibits a distinct cathode wave at 10 000 rpm. At 150 rpm, the cathode signal is virtually zero, while the corresponding anode current is larger than expected for oneelectron oxidation Figures 7 and 8). This is consistent with an ECElike process at the disk (anode). The first E reaction is oxidation of Cu(II) to Cu(I1I) at the disk, the C reaction is homogeneous loss of the Cu(III)-G3 complex by an unidentified reaction, and the second E reaction is oxidation of resulting electroactive product or products, also at the disk. This is evidenced by excess anode current and depressed cathode current at lower rotation velocities. In view of the indefinite number of electrons transferred in this process (Figure 9), the process may be represented as EKE),, rather than ECE; the x denotes an indefinite series of homogeneous/heterogeneous reaction steps. The extrapolated Levich plot of Cu(II)-G4, pH 11.6, 10-100 rpm, can be used as an approximate one-electron standard over the 10-10 000 rpm experimental range. The slope of the plot is a weak, 2/3 power function of the d&sion coefficient, minimizing differences between the G3 and G4 signals from this source. napp for the G3 anode signal is then estimated as the ratio of the Cu(II)-G3 anode signal to the corresponding anode signal of the one-electron standard. This is approximately 4-5 electrons at the Analytical Chemistry, Vol. 67,No. 3,February 1, 1995
547
0.010
1
t
0.002
o.ooo~, 0.0
,
,
0.1
0.2
Id,
0.001
0.3
1
0.1
!
0.01' 0
,I 0.3
I
I
I
10
20
30
IO",
(radls)'"
Figure 7. Levich plots of anode data for 100 pM 1:1 GdCuS04 in pH 9.80, 0.05 M borate/0.25 M KNO3 (0)and 0.2 M carbonate (A). Regression lines of Cu(II)-G4 data, pH 11.6, 10-100 rpm in borate (-) and carbonate(- - -) are shown. Inset: voltammograms of 100 pM 1:l GJCuSO4 in pH 9.80, 0.05 M borate/0.25 M K N 0 3 at rotation speeds of (from lowest curve to highest) 150,2000, and 10 000 rpm.
mNmM
u
I
$
0.05
0.000
0.00
0.05
0.10
I,, mNmM
Figure 6. Transformed plots of rotation speed-dependentcurrent for 100 pM 1:1 GdCuS04 in 0.05 M borate/0.25 M KN03; first wave, 550 mV to anode (all), 0 mV to cathode at pH 9.80 and 11.6, 80 mV to cathode at pH 9.15. (a, top) Anode. (b, bottom) Cathode. 0, pH 9.15; +, pH 9.80; 0, pH 11.6. The circled points indicated by arrows are the slopes of Levich plots using pH 11.6 data points in the 10100 rpm low rotation speed range and represent an idealized intercept for all three data sets. The dashed lines correspond to theoretical concentrations of free, triply deprotonated (NNNN) Cu(ll)-G4 obtained by multiplyingthe idealized intercept point by fractions obtained from literature equilibrium constants: 44.5% at pH 9.15, 73% at pH 9.80, and 89% at pH 11.6.
longest accessible time scale and approaches 1 electron at the shortest accessible time scale (Figure 9). The tripeptides GGF, GFG, and FGG exhibit electrochemical behavior virtually identical to that of G3, suggesting that this is a generic behavior of tripeptide-copper complexes (unpublished data). Voltammograms of Cu(II)-Gd and Cu(II)--G5 at pH = pKa exhibit two anodic waves and two corresponding cathodic waves at high rotation velocities. At lower rotation velocities, the second anodic waves, though relatively smaller, are still present. The second cathodic waves have vanished, however. The total anodic current at the second limiting current regions of Gd and G5 is in excess of one-electron behavior, but to a lesser extent than Cu(II)-G3 (Figure 5a, inset). There is not a corresponding sup pression of the cathode current in the second wave region, however (Figure 5b), suggesting that this phenomenon is not the same as the putative Cu(II)-G3 ECE-like process. 548 Analytical Chemistry, Vol. 67,No. 3, February 1, 1995
0.00 0
,
10
,
20
1
30
IO"', (radh)"' Figure 8. Levich plots of cathode data for 100 pM 1:l GdCuS04 in pH 9.80, 0.05 M borate/0.25 M K N 0 3 (0) and 0.2 M carbonate (A). Regression lines of Cu(ll)-G4 data, pH 11.6, 10-100 rpm in borate (-) and carbonate (- - -) are shown. Inset: voltammograms of 1OOpM 1:l GdCuS04 in pH 9.80, 0.05M borate/0.25 M K N 0 3 at rotation speeds of (from lowest curve to highest) 150, 2000, and 10 000 rpm.
CE/CCE Process in Carbonate Buffer. Carbonate, a component of the standard buffer, attenuates the absolute electrochemical sensitivity of Gq and Gg biuret complexes and magnifles the second wave relative to the first (Figures 3 and 4). The phenomenon is most pronounced at lower pH (Figure 4, insets). Studies of Cu(II) -Gq protonation via stopped-flow techniquesg have indicated slower kinetics for deprotonation of the NNNO form in the presence of carbonate than in several other buffering systems. It was inferred by the authors that this was due to participation by carbonate in the NNNO form of the complex as an equatorial ligand for copper. The visible absorption spectra of Cu(II)-Ga and Cu(II)-Gs are affected substantially by the presence of carbonate in a concentration-dependent manner (Figures 10 and 11). Borate buffer has virtually no effect on the spectra at the concentration used. Peptide biuret complexes exhibit broad d-d electronic transition peaks spanning the entire visible spectrum. The empirical
B
/P
//
I .
I*
+
w*6 1
I
I
I
0
10
20
30
8
9
a", (radls)'"
12
11
PH
Figure 9. Apparent number of electrons used by Cu(II)-G3, obtained by dividing the Cu(ll)-G3 rotation speed-dependent points in Figure 5 by the corresponding borate and bicarbonate Cu(IJ)-G4 regression lines that represent conditional one-electron standards. Lower curve is from cathode data; upper curve is from anode data.
600
10
i
Figure 11. pH dependence of molar absorptivity for 0.877 mM 1 :1 GdCuSO4 (-) and 0.762 mM 1:l GdCuS04 (- - -) in deionized water (u),0.0 M boratd0.2 M KN03 (A),0.1 M carbonate (0) and , 0.2 M carbonate (0).
N (terminal amine)
(4.53 f 0.04)
N (deprotonated peptide)
(4.85f 0.07)
0 (carboxylate)
(3.42 f 0.10)
io3 x
lo3
io3
0 (carbonyl, OH-,or H20) (3.01 f 0.03) x lo3 axial ligand replacing H,O
500i1 9
l
10
I
, 11
,
, 12
PH Figure 10. pH dependence of Amax for 0.877 mM 1:l GdCuS04 in deionized water (U), 0.05 M borate/0.25 M KN03 (A),0.1 M carbonate (0)and , 0.2 M carbonate (0).Regions A, B, and C represent the wavelength ranges predicted by the Billo correlation for ligation by NNNO (0= carbonyl), NNNO (0= carboxylate), and NNNN, respectively. The widths of these ranges are based upon root mean square combination of the reported uncertainties (SD)of the individual ligand contributions.
correlation by Bill03 approximately relates the identity of the four equatorial ligands in a copper-peptide complex to the frequency u of the absorbance maximum in the following manner:
where nl-n4 are the numbers of each category of equatorial ligand and V i are individual ligand contributions to Vobsd in cm-l; naxis the number of axial ligands. The correlation requires that certain conditions be met: that the complex be strictly square planar, that axial water molecules are complexed to the copper, and that donor groups be present in the axial positions if water is not. Also, ring strain is not taken into account. The peptides used in this study are believed to meet these requirements. The four equatorial ligand and one axial ligand categories and corresponding values of u f SD are as follow:
(1 x lo3 each)
In Figure 10, bands indicating expected ranges for,1 (nm) based on the Billo correlation are indicated for three relevant combinations of ligands. In the absence of carbonate, Cu(II)Gs and CuOD-Gd at pH = pK, have absorption maxima at wavelengths consistent with onehalf "NO (carbonyl) ligation and onehalf NNNN; titration to more alkaline conditions results in absorption maxima consistent with 100%NNNN. Similarly, at pH = pKa,the molar absorptivities are -100 M-l cm-l, somewhat below the typical value of -150 M-'cm-l for copperm-peptide complexes; a reasonable explanation is that the spectra are evenly divided between two broad, overlapping peaks for NNNO(carbonyl) and NNNN, centered perhaps 50 nm apart. At pH values well above the pKavalues, E has the expected value of -150 M-l cm-l. Carbonate has the effect of displacing the spectra toward longer wavelength absorbance maxima in the NNNO(carboxy1ate) r e gion. Molar absorptivities are also reduced, suggesting the existence of more than one absorbing species. This is consistent with carbonate participation to form an electrochemically inaccessible form of the Cu(II)-"NO complex, indicated hereafter by Cu(II)-NNN(COa). The specific form of carbonate involved, e.g., HCO3- or C032-, is unknown. CuOI)-G3 exhibits a similar, though weaker spectral response to carbonate (Table 4). There is a minor shift in 1- but a substantial decrease in c. Electrochemical behavior appears to reflect this response to carbonate ( F i r e s 7 and 8). Although the essential ECE-like behavior is present in either carbonate or weak borate buffer, the signal at higher rotation velocities is attenuated in carbonate relative to borate. In summary, two electroactive forms of the copper-peptide complexes have been identified: Cu (II)-NNNN (Ell2 = 0.45 V) and Cu(lI)-NNNO (Ellze 0.7 V). Two nonelectroactive forms have also been identified: NNN(0H-) and NNN(CO3). Analytical Chemistry, Vol. 67, No. 3, February 1, 1995
549
Table 4. Absorption Maxima and Molar Aborptivities of Cu(ll) Complexes of GJand G5 in Various Medla
medium peptide
PH
deionized water
borate, 0.05 M
carbonate, 0.1 M
carbonate, 0.2 M
GJ. 1.12 mM G5, 0.762 mM
9.80 8.04 (PK,) 9.80
554 (153) 534 (104) 510 (150)
554 (152) 528 (109) 510 (153)
556 (134) 567 (66) 510 (125)
560 (114) 584 (64) 516 (119)
Table 5. Equilibrium and Rate Constants Determined for Reactions of Cu(ll) Complexes 0
voltammetric
peptide
medium
G4 G5 Gsb
pH 9.8 carbonate pH 8.05 carbonate pH 9.05 carbonate
second, at ring
G4 G4 G5 G5
pH 9.15 borate pH 9.15 borate pH 8.05 borate pH 8.05 borate
first, at disk first, at ring first, at disk first, at ring
G4' G4' Gsb Gsb
pH 11.6 borate pH 11.6 borate pH 9.80 borate pH 9.80 borate
first, at disk first, at ring first, at disk first, at ring
K
wave
and 0
5
kb, S-l
ki, s - ~
Carbonate Dissociation 2.8a O.3la 3.4
second, at ring second, at ring
0.82 (0.89)c 0.62 (0.89)c 1.5 (l.O)d 1.1(1.0)d
10 f 5 4.3 f 3 40 f 15
14 f 7 18 f 13 50 i 19
3.6 2 14 f 10 10 f 4
23 f 4 (27)' 72 f 6 (27)c 20 6 13 f 2
13 2 44 f 4 (16)c 8.0 f 2 6.1 f 0.7
10 2 (11)' 27 f 2 (11)' 12 f 4 7.0 f 0.8
28 f 8 (7)c 39 10 (7)' 35 f 8 32 f 5
5.2 f 2 6.7 +. 2 4.9 f 1 5.7 f 0.9
22 f 7 32 f 8 30 f 7 26 f 4
Deprotonation
"NO
(I
k , s-l
4
*
*
*
Hydroxide Dissociation 4.3 (8)c 4.8 (8)c 6.1 4.6
*
Adjusted for the NNNO/NNNN equilibrium. No distinct horizontal region at high rotation speeds. Reference 9. Reference 5.
For Gd and Gs complexes oxidized in carbonate buffer, the equilibrium concentration of the free NNNO form can be augmented by the single CE process of carbonate dissociation, while the NNNN form is augmented by the series processes of Cu(II) -"NO deprotonation (CE) and Cu(ID -NNN(C03) dissociation followed by Cu(II)-"NO deprotonation (CCE). Recalling also the hydroxide equilibrium of Scheme 3 and representing structure 3 by NNN(OH-), we represent the processes occurring in the oxidation of the NNNN form at the first anodic wave as follow. C, Cu(II)-NNN(COd
-
Cu(II)-NNNO
C, Cu(II)-NNNO -Cu(IJ)-NNNN C, Cu(II)-NNN(OH-)
E, Cu (II) -NNNN
-
-
(7)
Cu(II)-NNNN
Cu (III) -NNNN
(6)
+ e-
(8) (9)
Note that reactions 7 and 8 are parallel. At the second anodic wave region, there is the additional process
E: Cu(II) -"NO
-
Cu(IIr>-"NO
+ e-
(10)
In weak borate buffer, reactions 7-9 occur at the first limiting current region and 7-10 in the second. The sum of processes 7-10 increases the size of the first wave relative to the second at lower rotation speeds. At all rotation speeds, the signal at the second limiting current region will be the true mass transportlimited current, except to the extent that the kinetic effect of reaction 8 renders the Cu (IO -NNNN form unavailable at shorter time scales. Reaction 8 is believed to be the C process responsible for the mild curvature of Levich plots of Cu(II)-G* and Cu(II)G5 at pH >> pK,. 550 Analytical Chemistry, Vol. 67,No. 3, February 1, 1995
In 0.2 M carbonate buffer, the nonelectroactive Cu(II) -NNN(C03) must undergo two successive chemical reactions, 6 and 7, to supply Cu(lI)-NNNN but only reaction 6 to supply Cu(II)"NO. At the second limiting current region, a substantial portion of Cu(II)-NNNO released from the carbonate complex will be oxidized to Cu(I1I)-NNNO before deprotonation to Cu(ID-NNNN can occur. A portion of Cu(II)-NNNN (e.g., -11% for G4) exists in the hydroxide form Cu(II)-NNN(0H-), independently of P H ; ~ the C reactions must occur before this portion of the peptide complex can be oxidized. The sum of processes 6-10 exaggerates the magnitude of the second limiting current (EappZ-Z 800 mv) relative to the first (Eapp= 600 mv) while attenuating the sensitivity over the entire potential range. At the longest residence times ( w Z-Z 10 rpm), the second wave is not detectable, but the first wave sensitivity is still lower in carbonate buffer than in borate buffer. This can be attributed to a slow carbonate dissociation step followed by a fast deprotonation step. The spectroscopic evidence suggests that at pH values well above the pKafor NNNO deprotonation,Cu(II)-NNN(C03) and Cu(II)NNNO are largely absent; reaction 8 is then responsible for a larger proportion of the total CE behavior. Quantitative Evaluation. Three CE processes have been described, wherein the C reactions are (i) deprotonation of Cu(ID -"NO to Cu(II)-NNNN, (ii) dissociation of Cu(II) -NNN(OH-) to Cu(II)-"NO, and (iii) dissociation of Cu(II)NNN(C03) to Cu(II)-"NO. To calculate equilibrium and rate constants for the three processes, informationwas extracted from the Z / W ' / ~ vs Z plots as indicated by the example in Figure 12. Distance B in the figure is proportional to the equilibrium concentration of the electroactive form of the complex in solution, and distance A is proportional to the equilibrium concentration of the nonelectroactive form. The location of the horizontal line indicating distance B is established by the horizontal region at
1
boo4
0.001
j
0.000 0.00
\
6
0.02
0.04
0.06
Ir, mNmM
Figure 12. Illustration of method for estimation of equilibrium and rate constants, using ring (cathode) data for 100 p M 1:i GdCuS04 in 0.05 M borate/0.25 M KN03, pH 8.05. These conditions are suitable for estimates of k (rate constant) for the deprotonation of the NNNO form to NNNN. The distances A and B represent the relative proprtions of Cu(II)-NNNO and Cu(ll)-NNNN, respectively: K= HA The set of points chosen for linear regression is indicated by the arrow.
the end of the curve. In some cases a clearly level region was not identitiable; in this case it was assumed that the lowest point at the end of the curve is the best estimate of B, with the determinate error introduced by this situation duly noted. A linear region of the plot was identified by inspection, and the points were regressed to obtain a slope and intercept. The linear region chosen in Figure 12 is indicated by the arrow. The intercept establishes A + B; K = B/A. Calculation of k follows from rearrangement of the equation for the line; the expression k = slope 1.6W2results. An average value of 5 x cm2/s was used for D in calculation of k . The l/3 power dependence minimizes error propagated from uncertainty in D. Finally, k b = k / ( K + 1) and kr = &. In order to isolate the three C processes as much as possible, three suitable combinations of pH, potential, and buffer medium were chosen from the available data: (i) deprotonation, reaction 7, borate buffer at pKa for the NNNN/NNNO equilibrium, first wave data only; (ii) hydroxide dissociation, reaction 8, borate buffer at pKa >> pH, &-stwave data; and (i) carbonate dissociation, reaction 6, carbonate buffer at pH somewhat above pK,, second wave, ring data only. The values obtained for constants related to the three CE processes are summarized in Table 5. Corresponding literature value^^^^ are enclosed in parentheses. The major contribution to the uncertainties indicated (95% confidence interval) was the (15)Nagy, G.; Rice, M. E; Gerhardt, G. A; Hierl, P.M.; Adams, R N. Neuroscience 1985, 15, 891-902. (16)Tsai, H. Y.; Weber, S.G., in preparation.
uncertainty in the slopes of the regressed lines. The rate constants estimated are on the order of 101-102. Corresponding relaxation times are on the order of seconds. The data are in agreement with literature values when checks are possible. In general, and especially for the hydroxide dissociation process, the data are internally consistent. Values of kf are for the dissociations of H+, OH-, or carbonate/ bicarbonate. It is remarkable that all of the values fall in such a small range. It is possible that the mechanisms of the reactions are essentially associative, with the slow step-similar in all cases-involving rearrangement of the peptide backbone. Analytical Implications. The importance of the processes described above to electrochemical peptide detection are strongly time scale dependent At long time scales and high pH, the signalattenuating effects of the CE processes are minimized. The effect of the ECElike anode process for tripeptide complexes will be maximized at long time scales, however, magnifying the anode signal and suppressing the cathode signal. An equivalence between the RRDE and electrochemicaldetector time scales can be calculated on the basis of mass flux to the electrode surface.16 For example, detection using a commercial thin-layer ECD working under microbore HPLC conditions (13 pm spacer, 0.07 mL/min flow rate) occurs on a time scale corresponding to the minimum RRDE rotation rate (-10 rpm). Standard HPLC conditions correspond to higher rotation speeds, e.g., 20 pm spacer, 1.5 mL/min flow rate corresponds to -400 rpm. Three model complexes have been rather completely characterized. Based on this understanding and the relationship between the RRDE and the electrochemical detector, signal behavior can be predicted. There is one lurking uncertainty remaining. We have found a modest but statistically signiticant difference between diffusion coefficients measured by two methods. At the same time, there is, under some conditions, a small amount of anodic current that is not accompanied by cathodic current. It is still possible, then, that a small fraction of any of these complexes exists in a form that is only oxidized irreversibly in a multipleelectron transfer at high potentials. ACKNOWLEDGMENT We would like to thank the NIH for their financial support through Grant GM-44842. Scient@ Parentage of the Author. S. G. Weber, Ph.D. under W. Purdy, Ph.D. under D. N. Hume, Ph.D. under I. M. Kolthoff. Received for review July 13, 1994. Accepted November
19,1994.B AC940698X @Abstractpublished in Advance ACS Abstracts, December 15, 1994.
Analytical Chemisfry, Vol. 67, No. 3, February 1, 1995
551
