Automated Flow Injection Gradient Technique for Binding Studies of

binding sites) were calculated using the Scatchard model. The concentration gradient was calibrated by injecting. ANS in the stream, and the binding e...
0 downloads 0 Views 110KB Size
Download PDF
Anal. Chem. 1999, 71, 2541-2550

Automated Flow Injection Gradient Technique for Binding Studies of Micromolecules to Proteins Using Potentiometric Sensors: Application to Bovine Serum Albumin with Anilinonaphthalenesulfonate Probe and Drugs Maria E. Georgiou,† Constantinos A. Georgiou,‡ and Michael A. Koupparis*,†

Laboratory of Analytical Chemistry, Department of Chemistry, University of Athens, 157 71 Athens, Greece, and Chemistry Laboratory, Department of Sciences, Agricultural University of Athens, 75 Iera Odos, 118 55 Athens, Greece

An automated flow injection (FI) gradient technique is described for the binding study of the potentiometric probe 1-anilino-8-naphthalenesulfonate (ANS) to bovine serum albumin (BSA). Using a single-channel FI system with a mixing chamber and a flow ANS electrode, the binding parameters (binding constant and number of binding sites) were calculated using the Scatchard model. The concentration gradient was calibrated by injecting ANS in the stream, and the binding experiment was performed by injecting ANS-BSA solution in the carrier solution of equal albumin concentration. The equations describing the concentration gradient and the corresponding electrode potential curve are presented. A systematic study of the factors affecting the complexation equilibrium and the electrode response was performed. For the ANS binding to BSA, two binding classes were determined with binding constants of (2.1 ( 0.3) × 105 and (3.3 ( 0.8) × 103 M-1 and 3.8 ( 0.6 and 10 ( 2 binding sites per class, respectively, at 27 ( 1 °C, in 0.10 M phosphate pH 7.4. Competitive binding experiments of sulfamethoxazole, salicylate, azapropazone, ketoprofen, and tolmetin to albumin were also performed by monitoring ANS binding inhibition (decrease of apparent binding constant). This technique takes advantage of FI gradients and direct potentiometry and utilizes the total information contained in FI peaks, providing fast and accurate binding information in a wide range of concentration ratios. Specific noncovalent binding of micromolecules to proteins is of great importance in pharmacology, biopharmaceutics, biochemistry, immunochemistry, and related fields. The study of protein interactions with micromolecules and the determination of related parameters is an attractive analytical problem. It is based on monitoring the change of a physicochemical property of the protein-micromolecule system upon binding either directly (direct technique) or after separation of the bound and free † ‡

University of Athens. Agricultural University of Athens.

10.1021/ac981019b CCC: $18.00 Published on Web 05/22/1999

© 1999 American Chemical Society

micromolecule (indirect technique).1 Among the direct techniques, spectrophotometry and fluorometry are extensively used and are superior to the indirect techniques (equilibrium and dynamic dialysis, ultrafiltration, gel filtration) because they do not disturb the binding equilibrium upon separation. Whenever no significant change of a physicochemical property is caused upon binding, the so-called competitive technique is used. This technique utilizes a probe molecule (spectrophotometric or fluorometric) that is displaced or inhibited by the micromolecule under study. The fluorescence probe 1-anilino-8-naphthalenesulfonate (ANS) has been proved one of the most successful probes for the investigation of drug binding to proteins.2-5 The general disadvantages of the fluorescence technique, direct or competitive, are the increased inherent fluorescence of the biological samples at high protein concentration, the quenching of fluorescence caused by the matrix, and the difficulty in determining the experimental coefficients of fluorescence yield of the various protein complex species. A novel direct technique based on potentiometry with ion-selective electrodes (sensors) has been developed in this laboratory and applied to protein6,7 and cyclodextrins8 binding studies. This technique has the advantages of direct and continuous monitoring of the free ionic micromolecule in a wide concentration range without any interference of proteins. However, this technique can be employed only for ionic drugs for which a successful sensor can be developed. Therefore a competitive potentiometric probe was highly desirable. The construction of an ANS-selective electrode was achieved in this laboratory and used for competitive drug-protein potentiometric studies.9,10 (1) Connors, K. A. Binding Constants: The Measurement of Molecular Complex Stability; John Wiley & Sons: New York, 1987. (2) Jun, H. W.; Luzzi, L. A.; Ma, J. K. H. J. Pharm. Sci. 1975, 64, 493-497. (3) Santos, E. C.; Spector, A. A. Mol. Pharmacol. 1974, 10, 519-528. (4) Varela, A. S.; Macho S. M. I.; Minones, J. J. Pharm. Sci. 1992, 81, 842844. (5) Muller, N.; Lapicque, F.; Drelon, E.; Netter, P. J. Pharm. Pharmacol. 1994, 46, 300-304. (6) Christopoulos, T.; Diamandis, E. Anal. Chem. 1990, 62, 360-367. (7) Valsami, G. N.; Macheras, P. E.; Koupparis, M. A. Pharm. Res. 1991, 8, 888-892. (8) Sideris, E.; Valsami, G.; Koupparis, M.; Macheras, P. Pharm. Res. 1992, 9, 1568-1574. (9) Angelakou, A.; Valsami, G.; Koupparis, M.; Macheras, P. J. Pharm. Pharmacol. 1993, 45, 434-438.

Analytical Chemistry, Vol. 71, No. 13, July 1, 1999 2541

In the aforementioned techniques, the preparation of a series of mixtures, with constant concentration of one of the components (usually protein) and varying concentrations of the other, is required. These mixtures may be prepared and measured in a batch mode or a manual stepwise titration mode. The complete study of the binding phenomenon requires a wide range of concentration ratios. Over the past few years, the flow injection (FI) technique has been successfully used for the automation of chemical equilibria studies (manipulation of reactants and measurement of the analytical signal). In these applications, concentration gradients were created either by flow rate variation11-13 or through the use of an external mixing chamber.14-19 Using a highdispersion FI system, a very reproducible and wide concentration gradient of the injected sample into the carrier stream can be achieved.20,21 The use of the FI gradient technique in the spectrophotometric study of the complexation of micromolecules to cyclodextrins has been proposed by our group.22 In this work, we extend the use of the automated FI gradient technique for the potentiometric study of binding of micromolecules to proteins. For this purpose, a flow-through ANS electrode has been constructed and used for the determination of binding parameters of the ANS-bovine serum albumin (BSA) complex. Using ANS as a potentiometric probe molecule, the binding study of several drugs (sulfamethoxazole, salicylate, azapropazone, ketoprofen, tolmetin) with BSA was performed using the so-called inhibition competitive approach. The derivation of the appropriate equations for the calibration of the concentration gradient and the calculation of binding parameters (constants and number of sites) from potential FI peaks is described. The effect of flow rate, calibration stability of the potentiometric sensor, and BSA concentration on binding parameters has been studied. EXPERIMENTAL SECTION Apparatus. The single-channel homemade FI system depicted in Figure 1 consists of an Ismatec MPN-8 peristaltic pump, a Rheodyne 5020 low-pressure injection valve with pneumatic actuator, a pH/pion electrometer unit (EU-200-30, Heath Schlumberger), an 80386 personal computer, an Advantech PCL-718 interface card for data acquisition and control featuring a 12-bit, 60-kHz AD converter, and an Advantech PCLD-786 solid-state relay card (controlling the pneumatic actuator). Teflon tubing (0.8 mm i.d.) was used, and the 1.0-mL mixing chamber with magnetic (10) Angelakou, A. T.; Sideris, E. E.; Valsami, G. N.; Koupparis, M. A.; Macheras, P. E. J. Pharm. Sci. 1994, 83, 1150-1154. (11) Marcos, J.; Rios A.; Valcarcel, M. Anal. Chem. 1990, 62, 2237-2241. (12) Marcos, J.; Rios, A.; Valcarcel, M. Anal. Chim. Acta 1993, 283, 429-438. (13) Marcos, J.; Rios, A.; Valcarcel, M. Anal. Chim. Acta 1995, 308, 152-158. (14) Miller, J. N.; Abdullahi, G. L.; Sturley, H. N.; Gossain, V.; McCluskey, P. L. Anal. Chim. Acta 1986, 179, 81-90. (15) Lopes da Conceicao, A. C.; Simoes Goncalves, M. L. S.; Correia dos Santos, M. M. Anal. Chim. Acta 1995, 302, 97-102. (16) Turner, D. R.; Correia dos Santos, M. M.; Coutinho, P.; Lurdes Goncalves, M.; Knox, S. Anal. Chim. Acta 1992, 258, 259-267. (17) Vithanage, R. S.; Dasgupta, P. K. Anal. Chem. 1986, 58, 326-330. (18) Tyson, J. F. Analyst 1987, 112, 527-529. (19) Echols, R. T.; Tyson, J. F. Analyst 1995, 120, 1175-1179. (20) Ruzicka, J.; Hansen, E. H. Flow Injection Analysis, 2nd ed.; John Wiley & Sons: New York, 1988; Chapter 2. (21) Valcarcel, M.; Luque de Castro, M. D. Flow-Injection Analysis, Principles and Applications; Ellis Horwood Ltd.: New York, 1987; Chapter 8. (22) Georgiou, M. E.; Georgiou C. A.; Koupparis, M. A. Anal. Chem. 1995, 67, 114-123.

2542 Analytical Chemistry, Vol. 71, No. 13, July 1, 1999

Figure 1. Flow injection system used in binding studies with albumin: C, carrier stream; P, peristaltic pump (flow rate 5.00 mL/ min); IV, injection valve (sample loop 200 µL); MC, mixing chamber; E, wall-jet electrode; W, waste. (a) and (b) control and data acquisition lines, respectively.

stirrer was constructed from Plexiglas. A homemade software package, developed in C language, provides modules for (a) control of the pump and sample valve and potential data acquisition and storage to disk for the FI binding experiments, (b) retrieval of stored FI curves, and (c) pretreatment of raw experimental data and export to the Statgraphics statistical package of STSC Inc. or the Scientist nonlinear curve fitting program of Micromath Scientific Software. The flow-through ANS sensor unit, was constructed from Plexiglas in our laboratory in the wall-jet configuration described elsewere.23 This configuration ensures a relatively low dead volume of the detector. The PVC electrochemical membrane of the ANS sensor was prepared by entrapping the ANS liquid exchanger (tetraheptylammonium-ANS ion pair in p-nitrocymene) in PVC polymer as described in detail elsewere.7 The internal reference electrode was a silver wire coated with AgCl, and the internal reference solution was a mixed 0.010 M ANS0.10 M NaCl solution saturated with AgCl, the external reference electrode was a Ag/AgCl electrode (Orion 90-01), and the salt bridge consisted of KCl saturated solution. The electrodes were connected to the pH/pion electrometer unit, and the electrode potential was fed to the data acquisition card. All experiments were carried out in a temperature-controlled ((1 °C) laboratory. Reagents. 1-Anilino-8-naphthalenesulfonate ammonium salt was obtained from Fluka. Bovine serum albumin (analytical grade, molecular mass 67 000) was obtained from Serva. Sulfamethoxazole was kindly donated by American Cyanamid Co. Sodium salicylate, azapropazone, ketoprofen, and tolmetin were obtained from local manufacturers and were used without further purification. Working ANS solutions prepared were as follows: (a) 0.0100, 0.0200, 0.0500, and 0.100 M for gradient calibration and (b) 1.00 × 10-5-1.00 × 10-2 M for ANS electrode calibration. For direct binding experiments, mixed ANS-BSA solutions containing 0.0200, 0.0300, 0.0500, 0.0800, 0.100, or 0.200 M ANS and 0.100, 0.500, 1.00, 1.50, or 2.25% (m/v) BSA (1.50 × 10-5-3.36 × 10-4 M) were prepared to be used for injection, and BSA solutions of identical concentrations were prepared to be used as carrier solutions. For competitive binding experiments, mixed ANSBSA-drug solutions containing 5.00 × 10-4 or 1.00 × 10-3 M studied drug, 1.50 × 10-5 M BSA, and 0.0500 M ANS were prepared to be used for injection, while ANS-BSA solutions of (23) Christopoulos, T. K.; Diamandis, E. P. Analyst 1987, 112, 1293-1298.

is the lowest concentration of the linear range of the electrode response curve (quantitation limit), Vs and Vg are the sample and mixing chamber volumes (µL), respectively, Q is the flow rate (µL/s), and ∆tb is the peak base width (i.e., t1 - t0, where t0 and t1 correspond to the entrance and exit of the injected ANS from the mixing chamber, respectively). The minimum value of the dispersion coefficient corresponds to tmax and is predicted by eq 4:

Dmin ) 1/[1 - exp(-Vs/Vg)]

Figure 2. Calibration of dispersion gradient by injecting 200 µL of a 0.0500 M ANS solution in phosphate buffer, 0.10 M pH 7.4. Flow rate 5.00 mL/min. Along the FI peak, RSD values of 10 consecutive injections are shown. The precision achieved was better than 1.5% RSD for potential readings along the ascending part of the FI peak.

identical concentrations were prepared to be used as carrier solutions. All solutions were prepared in 0.10 M phosphate buffer adjusted to pH 7.4. Gradient Calibration. To calibrate the concentration gradient, 200 µL of an 0.0500 M ANS solution prepared in phosphate buffer (0.10 M, pH 7.4) is injected in pure buffer carrier solution. As ANS is an anion a negative FI peak is obtained across the stable baseline potential (Eb) (Figure 2) and dispersion coefficients, Dt, are calculated along the ascending part of the FI peak according to eq 1, where [L]0,cal is the concentration of the injected ANS

Dt ) [L]0,cal/10(Et,cal-E0)/S

(1)

(ligand) solution and Et,cal is the potential at time t. E0 and S are the constant term and the slope, respectively, of the ANS electrode response curve (eq 2) that is obtained just before calibration by

E ) E0 + S log[L]

(2)

continuously pumping standard ANS solutions in the range of 1.00 × 10-5-1.00 × 10-2 M in phosphate buffer. Assuming no reaction, good stirring in the mixing chamber, negligible mixing/dispersion in all other manifold components, and relatively fast response of the ANS sensor in comparison to the flow rate of the carrier stream, the concentration of an injected solution along the FI peak is described by eq 3 for the descending

[L]t ) [L]0[1 - exp(-Qt/Vg)], t0 < t e tmax

(3)

[L]max ) [L]0[1 - exp(-Vs/Vg)]

(4)

tmax ) Vs/Q

(5)

[L]t ) [L]0[exp(Vs/Vg) - 1] exp(-Qt/Vg), tmax < t e t1 (6) ∆tb ) (Vg/Q) ln{[exp(Vs/Vg) - 1][([L]0/[L]ql) - 1]} (7) part, eqs 4 and 5 for its maximum, and eq 6 for the ascending part. In these equations, [L]0 is the concentration of ANS in the injected sample, [L]t is the transient concentration at time t, [L]ql

(8)

Assuming that the Nernst equation (eq 2) is valid to an [L]max ANS concentration (highest concentration of the linear range of the electrode response curve), the maximum value of [L]0 is selected according to eq 9, which is derived from eq 4:

[L]0,max ) [L]max/[1 - exp(-Vs/Vg)]

(9)

The maximum value of D corresponds to a point just before t1, at which the concentration reaches the [L]ql. Therefore

Dmax ) [L]0,max/[L]ql

(10)

From eqs 8 and 10, it is concluded that the range of dispersion coefficients that can be generated in an FI potentiometric system depends on Vs, Vg, the linear range of the response curve, and the flow rate in conjunction with the response rate of the potentiometric sensor. Differentiating eq 6 gives the rate of concentration change in the gradient range (tmax - t1):

d[L]/dt ) [L]0K1(-Q/Vg) exp(-Qt/Vg)

(11)

where K1 ) exp(Vs/Vg) - 1. Therefore, the rate of concentration change along the sample bolus in the given FI system depends on the initial concentration of ANS injected, the flow rate, and time. Substituting the Nernst equation (eq 2) in eq 6 the equation for E-t profile is obtained:

Et ) Econst + S log[L]0 - 0.4343SQt/Vg

(12)

and by differentiation

dE/dt ) -0.4343SQ/Vg

(12a)

Therefore an exponential concentration gradient (eqs 6 and 11) results in a linear potential gradient (eqs 12 and 12a, Figure 2) within the linear range of the Nernst equation. It is well-known that the response of membrane electrodes is not instantaneous. For a stepwise change ∆C corresponding to a potential change ∆E∞, the output signal as a function of time will be given by the equation

∆Et ) ∆E∞[1 - exp(-t/T)]

(13)

Analytical Chemistry, Vol. 71, No. 13, July 1, 1999

2543

where T is the response time constant of the electrode.24 A ramptype change dE/dt (eq 12a) will give a delayed similar ramp-type output signal, distorted at the initial part.24 Direct Binding Experiment. A BSA carrier solution of concentration PTot ) 1.50 × 10-4 M prepared in phosphate buffer is used. An [L]0 ) 0.0500 M ANS solution prepared in protein carrier solution is injected in the carrier stream. Thus, total (free plus bound) BSA concentration (PTot) remains constant along the bolus of the injected solution of the BSA-ANS complex and a gradient of total ANS concentration (free and BSA bound) is created, disturbing the binding equilibrium. In this way, a great number of “mixed solutions” with constant protein and varying ANS total concentration are prepared. The number of these solutions is limited only by the speed of the data acquisition system, the precision of the pumping system, and the precision, dynamic range, and response rate of the potentiometric sensor. Each “mixed” solution corresponds to a specific time, so one could state that these “mixtures” are separated from each other not by glassware but by time. In each of these “mixtures” total ANS concentration ([L]t,Tot) that equals free ANS concentration (Ft) plus bound ANS concentration (Bt) is calculated through the predetermined Dt values (eq 1) and the concentration of the injected ANS ([L]0), according to eq 14. The free ANS concentration is

[L]t,Tot ) [L]0/Dt ) Ft + Bt

(14)

calculated (eq 15) using the ANS-ISE calibration equation and the

Ft ) 10(Et-E0)/S

(15)

voltage reading Et at the corresponding FI curve point. Then the Bt, Ft data are fitted by nonlinear regression to the Scatchard equation25 (eq 16) using the Scientist nonlinear curve-fitting m

Bt )

∑(n K F /1 + K F )P i

i t

i t

Tot

(16)

i)1

program of Micromath Scientific Software and the Statgraphics statistical program of STSC Inc., where m, ni, and Ki denote, respectively, the number of distinct classes of independent and noninteracting binding sites, the number of sites of the ith class, and the ANS-BSA complex (intrinsic) association constant of the ith class. It has been well established4,7,9,26 that bovine and human serum albumin present two independent binding sites for ANS and thus eq 16 simplifies to

Bt n1K1Ft n2K2Ft ) + PTot 1 + K1Ft 1 + K2Ft

(17)

Competitive Binding Experiment. A mixed carrier solution containing albumin and drug in concentrations of PTot ) 1.50 × (24) Efstathiou, C. E.; Koupparis, M. A.; Hadjiioannou, T. P. Ion-Selective Electrode Rev. 1985, 203-259. (25) Scatchard, G. Ann. N. Y. Acad. Sci. 1949, 51, 660-672. (26) Bagatolli, L. A.; Kivatinitz, S. C.; Aguilar, F.; Soto, M. A.; Sotomayor, P.; Fidelio, G. D., J. Fluoresc. 1996, 6, 33-40.

2544 Analytical Chemistry, Vol. 71, No. 13, July 1, 1999

10-4 M and [I]0 ) 5.00 × 10-4 or 1.00 × 10-3 M, respectively, prepared in phosphate buffer is pumped through the FI manifold. The injected ANS (0.0200, 0.0300, or 0.0500 M) is prepared in the mixed carrier solution. Hence, constant PTot and [I]0 concentrations are maintained throughout the bolus of the injected solution containing the mixed complexes of BSA with ANS and inhibitor (drug) and a gradient of total ANS concentration is created, disturbing the binding equilibria. In these inhibition binding experiments, total and free ANS concentration in each of the “mixed solutions” created in the flow is also calculated through eq 14 and eq 15, respectively. Data are fitted to eq 16 and eq 17 to calculate the apparent binding parameters in the presence of the drug studied. RESULTS AND DISCUSSION Dispersion Evaluation. Since dispersion coefficients depend not only on the physical characteristics of FI gradient system but also on diffusion coefficients of injected solutes, the viscosity of both injected solution and carrier stream was studied.27-29 The viscosities of a 1.50 × 10-4 M BSA, a 0.0500 M ANS, and a mixed 1.50 × 10-4 M BSA-0.0500 M ANS solution were measured at 25 °C and found to be identical (relative differences less than 6.0%). In addition, the introduction of the external mixing chamber minimizes viscosity effects.30 To cut the cost related to BSA consumption when used in the carrier stream, inverse binding experiments (creation of a protein concentration gradient in the ANS carrier) were also conducted. These experiments were not successful as the dispersion calibration achieved by monitoring the gradient of a micromolecule (ANS) is not accurately valid for the gradient of a macromolecule (BSA). In addition, to use the simple equation form of the Scatchard model (eq 16) a constant protein concentration is required. Potential readings of standard ANS solutions during electrode calibration are obtained under steady-state conditions (i.e., solutions are continuously pumped through the FI system). The values of E0 and S thus determined are subsequently used for gradient calibration (eq 1) and in binding experiments for the determination of free ANS (eq 15). To test the validity of this approach, it must be proved that the detector’s response rate is adequate to follow the concentration change rates under the conditions used so that the recorded potential at each point along the acceding part of the FI curve (working gradient range) equals the steady-state potential of a solution of the same ANS concentration. To check this, the dispersion coefficients obtained by injecting solutions of various ANS concentrations (0.0100, 0.0200, 0.0500, and 0.100 M) in buffer carrier were compared. As shown in Figure 3, the potential gradients (E-t) in the ascending part of the FI peaks are linear up to the potential of the quantitation limit (Eql) of the electrode, corresponding to [L]ql ) 8.00 × 10-6 M and parallel with a potential offset given by the term S log[L]0. This proves the validity of eq 12 as fitting of data results in E ) (149.3 ( 0.2) (27) Vanderslice, J. T.; Rosenfeld, A. G.; Beecher, G. R. Anal. Chim. Acta 1986, 179, 119-129. (28) Betteridge, D.; Cheng, W. C.; Dagless, E. L.; David, P.; Goad, T. B.; Deans, D. R.; Newton, D. A.; Pierce, T. B. Analyst 1983, 108, 17-32. (29) Bakalyar, S. R.; Olsen, K.; Spruce, B.; Bragg, B. G. Technical Note 9; Rheodyne Inc., Cotati, CA, 1988. (30) Gisin, M.; Thommen, C.; Mansfield, K. F. Anal. Chim. Acta 1986, 179, 149-167.

Figure 3. Determination of dispersion coefficients by injecting 200 µL of a (a) 0.0100, (b) 0.0200, (c) 0.0500, and (d) 0.100 M ANS solution in phosphate buffer, 0.10 M pH 7.4. (a′-d′) are the respective dispersion coefficients (Dt) calculated through eq 1. Table 1. Calibration Stability of the ANS Flow Potentiometric Sensor ta (min)

S (mV/pC)

E0 (mV)

r

0 40 80 120

-57.0 ( 0.4 -58.3 ( 0.3 -58.2 ( 0.4 -59.3 ( 0.5

40 ( 1 35 ( 1 37 ( 1 33 ( 1

0.9999 0.99993 0.99991 0.9999

mean

-58.2 ( 0.9

36 ( 3

a

During a set of gradient calibration and binding experiments with 7.5 × 10-5-3.36 × 10-4 M BSA.

+ (0.992 ( 0.002)t, r ) 0.9998, for CANS ) 0.100 M (Figure 3d); E ) (166.1 ( 0.1) + (1.004 ( 0.001)t, r ) 0.9999, for CANS ) 0.0500 M (Figure 3c); E ) (189.4 ( 0.1) + (0.998 ( 0.002)t, r ) 0.9999, for CANS ) 0.0200 M (Figure 3b); and E ) (209.9 ( 0.2) + (0.969 ( 0.002)t, r ) 0.9999, for CANS ) 0.0100 M (Figure 3a). By fitting the constant term of these equations to E ) Econ + S log[L]0, the slope of the electrode is found to be -60.3 ( 1.8 mV/pC (r ) 0,9991), which is close to the -58.2 ( 0.9 mV/pC value experimentally determined (Table 1) and a further proof of the validity of eq 12. The dispersion coefficients for the four different concentrations are identical for Dt values up to 1200 while for the concentrated solutions (Figure 3c′ and d′) they are identical for Dt values up to 7000 (correlation with zero intercept and a slope of 1.048 ( 0.002 with r ) 0.9998). In the Dt range of 1200-7000, values determined by injecting dilute ANS solutions (Figure 3a′ and b′) are lower than the actual ones due to the deviation from linearity at lower concentrations in the calibration curve of the electrode. For example, for CANS ) 0.0100 M and Dt ) 1300, the transient concentration in the flow gradient is 0.0100/1300 ) 7.7 × 10-6 M, which is below the electrode quantitation limit. So to use a potential reading for gradient calibration, a check is performed by the software to ensure that it is lower (anion-selective electrode) than the potential at the quantitation limit. A convenient ANS concentration for gradient calibration was the 0.0500 M and the working gradient range was Dmin ) 20 to Dmax ) 1200. The effect of flow rate on the precision of Dt values was studied by performing calibration experiments using various flow rates from 0.5 to 8.2 mL/min and measuring RSD values of Et,cal

corresponding to a Dt range from 20 to 1200. The percent relative standard deviations of Et,cal values of 10 injections for each flow rate were 0.46-1.4, 0.39-1.1, 0.32-1.1, 0.31-1.5 (see Figure 2), and 0.95-4.1% for the flow rates of 0.5, 2.4, 3.6, 5.0, and 8.2, mL/ min, respectively. These results prove that the ANS sensor provides reproducible potential readings even at high flow rates. The flow rate chosen for routine measurements was 5.0 mL/min as a compromise of short experiment time (150-180 s/injection) and excellent reproducibility. It should be stated that for the 0.50 mL/min flow rate 20 min is required for completing a single injection (recording of the FI peak and washout of the injected ANS solution). Performance Characteristics of the Electrode. ANS-ISE responds only to the free ANS ions since the high molecular size of ANS-BSA complexes hinders the participation of these complexes in the ion-exchange equilibrium with the electrode’s membrane. This is an inherent advantage of ISEs in the determination of free ions in the presence of proteins (e.g., determination of free Ca2+ in serum). The calculation of binding parameters requires predetermined values of both the slope (S) and the constant term (E0) of the Nernst equation. The stability of the calibration curve was tested during a 2-h set of gradient calibration and binding experiments that involved carrier solutions of 7.50 × 10-5, 2.24 × 10-4, and 3.36 × 10-4 M in BSA. The results from four calibration curves obtained by pumping standard ANS solutions in the range of 1.00 × 10-5-1.00 × 10-2 M are presented in Table 1. As shown, S values are fairly constant, while E0 values vary slightly due to a gradual adsorption of albumin on the electrode PVC membrane. To compensate for this, it was necessary to calibrate the potentiometric sensor between sequential batches of dispersion calibration binding experiments. In this way, the calibration curve stability was checked and the most recent calibration curve was used in calculations. Because of the great number of data points provided by the FI gradient, it is possible to determine E0 values along with the binding parameters using nonlinear fitting of data to a combination of eqs 1, 14, 15, and 17. By using this approach, the values of the binding parameters were found identical to those obtained using the predetermined E0 value. Therefore, E0 values do not affect binding parameter estimation to the same extend as S and this is in accordance with published results from binding experiments using ion-selective electrodes in batch mode.6 The electrode membrane was changed every 3-4 days to avoid the effect of BSA adsorption to the membrane and on S and E0 values. After each renewal of the membrane, it was conditioned by pumping an ANS solution in the concentration range 1.0 × 10-4-1.0 × 10-2 M in phosphate buffer for 2-4 h. This conditioning was necessary to achieve a slope value close to the theoretical (-59 mV/pC at 25 °C) and minimize electrode drift. At 27 ( 1 °C, S values ranged from -59 to -56 mV/pC during the membrane’s use period (3-4 days). The lowest concentration of linear response under flow conditions was found to be 8.00 × 10-6 M, which is higher than that under stirring batch conditions (1.20 × 10-6 M).7,9 However, the linear range of the flow-through electrode is expanded to lower concentrations (1.00 × 10-6 M) in the presence of 1.50 × 10-4 M BSA as was observed in previous binding studies with ISEs.6,9 Analytical Chemistry, Vol. 71, No. 13, July 1, 1999

2545

Table 2. Flow Rate Effect on Binding Parameters of ANS to BSAa Q (mL/min)

K1 (M-1 × 10-5)

K2 (M-1 × 10-3)

n1

n2

0.5 2.4 3.6 5.0 8.2

1.5 ( 0.1 1.6 ( 0.1 1.7 ( 0.2 1.65 ( 0.08 2.5 ( 0.2

4.4 ( 0.4 2.3 ( 0.2 4.6 ( 0.8 3.0 ( 0.2 4.2 ( 0.8

4.5 ( 0.1 4.0 ( 0.1 4.2 ( 0.2 4.0 ( 0.1 5.2 ( 0.2

10.0 ( 0.3 10.1 ( 0.1 10.8 ( 0.6 9.5 ( 0.2 13.0 ( 0.9

a Results are mean values of five binding experiments at 27 ( 1 °C. A mixed 0.100 M ANS-1.50 × 10-4 M BSA solution was injected in a 1.50 × 10-4 M BSA carrier.

Table 3. Effect of BSA Concentration on Calculated Binding Parametersa Figure 4. Scatchard plot for ANS to BSA binding, calculated at 27 ( 1 °C by injecting a mixed 0.0800 M ANS-1.50 × 10-4 M BSA solution in a BSA carrier of the same concentration, rt ) Bt/PTot. Solutions buffered by 0.10 M phosphate, pH 7.4.

Electrode drift can be minimized by including the lowest concentration of linear response ([L]ql ) 8.00 × 10-6 M ANS) in the carrier stream during gradient calibration. Thus, a very stable baseline potential of Eb can be achieved, but in this case a modified equation for gradient calibration (eq 18) is needed. However, using

Dt ) ([L]0,cal - [L]ql)/(10(Et,cal-Eb)/S - 1)[L]ql

(18)

this approach the dynamic range of the FI gradient was decreased, resulting in a restriction of the useful range of Dt values without substantial improvement of precision. Thus the approach of injecting ANS in buffer carrier was finally used. The drugs studied for competitive binding were tested for a possible interference to the ANS electrode by determining the potentiometric selectivity coefficient using the mixed solutions method in a batch-mode experiment. The values found, 3.1 × 10-4 (sulfamethoxazole), 2.0 × 10-3 (salicylate), 3.4 × 10-3 (azapropazone), 1.0 × 10-3 (ketoprofen), and 3.4 × 10-4 (tolmetin), show no interference from these drugs. Direct Binding Studies of ANS to BSA. The experimental data obtained (free ANS concentrations) can be used with either the stepwise stoichiometric model or the independent site-oriented model (Scatchard model).25 Since the latter is widely used in the literature (values of ANS-BSA binding constants for comparison are available only with this model) and it is very useful in understanding the site-specific binding of drugs to proteins, we chose also to use the Scatchard model in this study. The Scatchard plot for ANS binding to BSA (rt/Ft vs rt, where rt ) Bt/PTot) shown in Figure 4 indicates that ANS is bound at two classes of binding sites. Nonlinear data fitting to eq 17 proved that there are two distinct classes of binding sites. This is in accordance with a recent work26 in which two distinguishable and independent classes of binding sites for ANS at subdomain IIA and IIIA of bovine albumin have been found using phase and modulation fluorescence spectroscopy. The flow rate chosen as described previously was 5.0 mL/ 2546 Analytical Chemistry, Vol. 71, No. 13, July 1, 1999

PTot (M)

K1 (M-1 × 10-5)

K2 (M-1 × 10-3)

n1

n2

1.50 × 10-5 7.50 × 10-5 1.50 × 10-4 2.24 × 10-4 3.36 × 10-4

0.89 ( 0.04 1.3 ( 0.1 1.41 ( 0.09 1.5 ( 0.1 1.6 ( 0.1

1.7 ( 0.2 3.2 ( 0.5 2.0 ( 0.2 1.5 ( 0.3 1.7 ( 0.1

4.9 ( 0.1 3.8 ( 0.2 4.4 ( 0.1 3.8 ( 0.2 4.3 ( 0.1

11.8 ( 0.8 6.0 ( 0.2 11.6 ( 0.4 10.0 ( 0.8 12.2 ( 0.4

a Results are mean values of five concequtive experiments at 27 ( 1 °C. A mixed 0.0500 M ANS-BSA solution was injected in a BSA carrier of the same concentration.

min. To further check reproducibility in the presence of BSA and examine the effect of flow rate on the binding parameters, a study was carried out using various flow rates and results are presented in Table 2. As is shown, the value of the binding constant for the first main class of binding sites (K1) that embodies most of the binding information is practically independent from the flow rate except at the high rate of 8.2 mL/min. As the binding of ANS to BSA is fast, the overestimation at high flow rate is explained by the limited electrode response rate to follow the continuous reestablishment of the complexation equilibrium during the gradient process (gradual decrease of ANS total concentration) at this high change rate (eq 11). Thus, the free ANS concentration is underestimated, resulting in overestimation of the binding constant. This incomplete response of the electrode cannot be compensated by the experimental nature of the gradient calibration since the high rate of concentration change depends on both the flow rate Q and the ANS concentration (eq 11). Another factor that has been studied is the BSA concentration used. Results for this study are shown in Table 3. The only K1 value that slightly declines is that obtained using a 1.50 × 10-5 M BSA solution. This can be explained as the low concentration of BSA provides a low concentration of binding sites. The concentration of 1.50 × 10-4 M was chosen for subsequent studies as a compromise of low BSA consumption and provision of excess of binding sites. After the optimization of experimental parameters, several binding experiments were conducted using ranging ANS concentrations and the calculated binding parameters are shown in Table 4. The results reveal that ANS is strongly bound to the first binding class of BSA, which consists of three binding sites. The withinrun precision (expressing the good fit to the Scatchard model) and the between-run precision using the same or different ANS

Table 4. Results from ANS-BSA Binding Studies by the FI Techniquea [L]0b (M × 102)

K1 (M-1 × 10-5)

K2 (M-1 × 10-3)

n1

n2

2.5 ( 0.1 2.5 ( 0.1 2.0 ( 0.1 2.3 ( 0.2 2.0 ( 0.1 1.7 ( 0.1 2.0 ( 0.1 2.0 ( 0.2 2.5 ( 0.2 1.7 ( 0.2 1.9 ( 0.2 1.8 ( 0.2

3.9 ( 0.5 4.2 ( 0.3 3.6 ( 0.4 2.6 ( 0.3 3.1 ( 0.2 2.2 ( 0.3 3.1 ( 0.2 3.3 ( 0.3 4.2 ( 0.8 4.6 ( 0.8 2.9 ( 0.3 2.2 ( 0.3

4.3 ( 0.1 3.7 ( 0.1 4.0 ( 0.1 3.4 ( 0.1 3.2 ( 0.1 3.7 ( 0.1 3.2 ( 0.1 3.8 ( 0.1 5.2 ( 0.2 4.2 ( 0.2 3.5 ( 0.1 3.8 ( 0.4

8.8 ( 0.4 8.8 ( 0.2 8.6 ( 0.3 9.2 ( 0.3 11.6 ( 0.3 14.1 ( 0.7 11.6 ( 0.3 9.0 ( 0.3 13.0 ( 0.9 10.8 ( 0.6 8.6 ( 0.2 8.5 ( 0.3

2.1 ( 0.3 14

3.3 ( 0.8 24

3.8 ( 0.6 16

2.00 3.00 3.00 3.00 5.00 5.00c 5.00 8.00 10.0 10.0 10.0 20.0 mean ( SD % RSD

A:1.00c B:1.00c,d

10 ( 2 20

Estimates Using the ANS Sensor and the Batch Potentiometric Technique 2.93 ( 0.08 3.9 ( 0.2 3.84 ( 0.05 3.51 ( 0.04 14.3 ( 0.9 3.39 ( 0.02

7.3 ( 0.1 3.2 ( 0.1

a Conditions: BSA 1.50 × 10-4 M; pH 7.4; temperature 27 ( 1 °C. Correlation coefficients calculated for nonlinear fits range from 0.9991 to 0.9999. Within-day RSD in Et values ranged from 0.9 to 2.6%. b ANS injected concentration. c BSA 3.36 × 10-4 M. d The Scatchard model with three binding classes was used. K3 found, 520 ( 92 M-1; n3, 1.1 ( 0.8.

Table 5. Literature Values of ANS-BSA and HSA Binding Parameters at 25 °C K1 (M-1 × 10-5)

K2 (M-1 × 10-5)

n1

n2

ref

Determined by Fluorometry in Phosphate Buffer 17.0a 11.0b 9.1c 70 ( 7d

3.1 ( 0.1 4.8 ( 0.7

2.9 0.32 ( 0.02

3.00 ( 0.10

Determined by Batch Potentiometry in 0.10 M Phosphate Buffer, pH 7.4 6.5 ( 1 0.21 ( 0.05 2.84 ( 0.32 3.16 ( 0.22 [K3 ) (0.54 ( 0.14) × 103 M-1, n3 ) 13.7 ( 1.8] 7.53 ( 0.59 0.34 ( 0.05 2.7 ( 0.2 2.8 ( 0.2 [K3 ) (0.8 ( 0.1) × 103 M-1, n3 ) 4.7 ( 1.0]

2 32 33 4

9 7

a Phosphate buffer, 0.050 M pH 7.5; θ, 23 °C. b Phosphate buffer, 0.050 M pH 7.5; θ, 27 °C. c Phosphate buffer, 0.020 M pH 7.5; θ, 25 °C. d HSA, pH 7.5; θ, 25 °C.

initial concentrations are adequate for binding studies. For comparison, a batch potentiometric study was also performed using the ANS-ISE (titration of BSA with ANS-BSA mixed solution)7-9 and the same lot of BSA reagent. The results obtained and shown in Table 4 are in good agreement with those obtained using the FI gradient technique. The ANS-BSA or human serum albumin (HSA) binding has already been studied by fluorometry and batch potentiometry, and binding parameters found in bibliography are presented in Table 5. The differences shown are attributed to the type of protein (BSA, HSA), the lot of protein reagent, and the Scatchard model used (one, two, or three binding classes). The bibliography values for the binding with BSA (especially those obtained with batch potentiometry) are of the same order of magnitude with those

Figure 5. Saturation plots obtained by injecting a 1.50 × 10-4 M BSA solution containing (a) 0.0100, (b) 0.0800, (c) 0.100 M, and (d) 0.200 M ANS in a BSA carrier of the same concentration at 27 ( 1 °C. Solutions buffered by 0.10 M phosphate, pH 7.4.

determined by the new FI potentiometric method. The batch potentiometric technique is based on the addition of small amounts of mixed ANS-BSA solution to a BSA solution of identical concentration (titration). Using the developed FI method, a mixed ANS-BSA is injected to a BSA carrier stream of the same concentration. So total ANS concentration is gradually reduced in the FI procedure (following the backward direction of complexation equilibrium), while in the batch technique it is gradually increased (following the forward direction of the complexation equilibrium). In the case of the batch potentiometric technique, high concentrations of the ANS micromolecule are achieved revealing a third binding class (Tables 4 and 5). However, this third binding class could be an artifact of the fitting algorithm as it serves to “facilitate" data fitting. Using the FI technique, the saturation of albumin with ANS can also be studied. As shown in Figure 5 from the rt vs log(Ft) plot, for ANS injected concentrations of up to 0.200 M, BSA is not saturated in FI experiments. This is in accordance with the work of Honore31 that there is no saturation of the protein with micromolecules. Total ANS concentration achieved in this flow gradient using the optimized parameters ranged from 8.30 × 10-6 to 1.00 × 10-2 M in a Dt range of 20-1200 obtained in the time interval from 15 to 120 s. Competitive Binding Studies. Drugs studied by the competitive binding experiments were the nonsteroid antiinflammatory drugs sulfamethoxazole, salicylate, azapropazone, ketoprofen, and tolmetin. In these experiments, drug competes with probe (ANS) for the limited binding sites in BSA and apparent ANS-BSA binding constants of lower values are thus determined. Figure 6 shows FI peaks in the presence of salicylates. As the concentration of salicylates increases, ANS is displaced from BSA resulting in stepper FI peaks. The values of apparent binding constants shown (31) Honore, B. Pharmacol. Toxicol. 1990, 667 (Suppl. II), 1-26.

Analytical Chemistry, Vol. 71, No. 13, July 1, 1999

2547

Table 6. Binding Constants of ANS-BSA in the Presence of Increasing Concentrations of Five Drugs at 27 ( 1 °Ca [L]0 (M × 102) sulfamethoxazole 5.00 salicylate 2.00 5.00 azapropazone 3.00 ketoprofen 3.00 tolmetin 3.00

[I]0 (M × 103)

K1 (M-1 × 10-5)

K2 (M-1 × 10-3)

n1

n2

0 0.500 1.00

2.0 ( 0.1 1.7 ( 0.1 1.3 ( 0.1

3.1 ( 0.2 3.4 ( 0.2 3.4 ( 0.2

3.2 ( 0.1 3.2 ( 0.1 3.2 ( 0.1

9.9 ( 0.3 9.3 ( 0.1 9.8 ( 0.2

0 0.500 1.00 0 0.500 1.00

2.4 ( 0.1 1.36 ( 0.08 0.90 ( 0.03 2.0 ( 0.1 1.36 ( 0.09 0.89 ( 0.07

3.8 ( 0.5 4.1 ( 0.4 3.4 ( 0.3 3.1 ( 0.2 3.0 ( 0.2 2.1 ( 0.2

4.3 ( 0.1 4.5 ( 0.2 4.8 ( 0.1 3.2 ( 0.1 3.2 ( 0.1 3.5 ( 0.2

8.8 ( 0.4 8.7 ( 0.2 9.1 ( 0.2 11.6 ( 0.3 12.0 ( 0.3 13.5 ( 0.6

0 0.500 1.00

2.0 ( 0.1 0.40 ( 0.03 0.30 ( 0.03

3.6 ( 0.4 2.8 ( 0.3 2.0 ( 0.4

4.0 ( 0.1 5.4 ( 0.2 6.2 ( 0.4

8.6 ( 0.3 13 ( 3 19 ( 3

0 0.500 1.00

2.5 ( 0.1 1.4 ( 0.1 1.0 ( 0.2

4.2 ( 0.3 2.8 ( 0.2 2.0 ( 0.2

3.7 ( 0.1 2.3 ( 0.1 2.3 ( 0.2

8.8 ( 0.2 10.1 ( 0.2 12.0 ( 0.3

0 0.500 1.00

2.3 ( 0.2 2.2 ( 0.2 1.3 ( 0.2

2.6 ( 0.3 2.5 ( 0.2 1.5 ( 0.2

3.4 ( 0.1 2.7 ( 0.1 2.5 ( 0.1

9.2 ( 0.3 9.7 ( 0.2 10.9 ( 0.4

a Binding parameters obtained by injecting a 1.50 × 10-4 M BSA solution containing ANS and drug in a carrier stream of the same BSA and drug concentration. Solutions buffered by 0.10 M phosphate, pH 7.4. Estimates are mean values of four consecutive experiments. Correlation coefficients of all data sets ranged from 0.9991 to 0.9998. [L]0 is the injected probe (ANS) and [I]0 the drug concentration.

Figure 6. FI peaks acquired during competitive binding experiments of salicylate-BSA using the ANS probe. Injection of a 0.0200 M ANS-1.50 × 10-4 M BSA-salicylate solution in a BSA-salicylate carrier of the same concentrations: (a) 0, (b) 5.00 × 10-4, and (c) 1.00 × 10-3 M salicylate. Solutions buffered by 0.10 M phosphate, pH 7.4.

in Table 6 depend on the relative strength of binding of drug and ANS to BSA and the relative concentrations of drug and ANS. The change in Scatchard plots of ANS-BSA binding with increasing tolmetin concentration is shown in Figure 7. The results shown in Table 6 were obtained using the same flow rate (5 mL/ min) as in the direct binding experiment. Since the kinetics of competitive binding experiments are usually slower than those in direct binding experiments, a lower flow rate (3.6 mL/min) was also used resulting in practically the same results. For example, for salicylate K1 values were (1.45 ( 0.06) × 10-5 M-1 for 3.6 mL/min and (1.36 ( 0.09) × 10-5 M-1 for 5.0 mL/min. A higher rate (8.2 mL/min) resulted in a higher value of binding constant, (2.1 ( 0.3) × 10-5 M-1. HSA has been studied in detail and has been used in more studies of drug binding compared to BSA. The two macromol2548 Analytical Chemistry, Vol. 71, No. 13, July 1, 1999

Figure 7. Scatchard plots for determining apparent binding parameters of ANS to BSA at 27 ( 1 °C in the presence of tolmetin by injecting a 0.0500 M ANS-1.50 × 10-4 M BSA-tolmetin solution in a BSA-tolmetin carrier of the same concentrations: (a) 0, (b) 5.00 × 10-4, and (c) 1.00 × 10-3 M tolmetin at 27 ( 1 °C. Solutions buffered by 0.10 M phosphate, pH 7.4.

ecules are similar; hence their behavior does not differ greatly. However, differences in binding classes have been reported.10,34 It is well-known that HSA has two main binding sites: the warfarin (32) Hsu, P.-L.; Ma, J. K. H.; Jun, H. W.; Luzzi, L. A. J. Pharm. Sci. 1974, 63, 27-31. (33) Chatelain, P.; Matteazzi, J.-R.; Laruel, R. J. Pharm. Sci. 1994, 83, 674676. (34) Panjehshahin, M. R.; Yates, M. S.; Bowmer, C. J. Biochem. Pharmacol. 1992, 44, 873-879.

site (I) and the benzodiazepines and indoles site (II).5,35,36 According to some researchers, albumin is not a strictly rigid molecule28 so its tertiary structure and binding sites could change upon binding with micromolecules. The binding site II of HSA and the respective one in BSA may differ in the amino acid sequence; hence differences in tertiary structure and binding to micromolecules are anticipated. To comment on results obtained in competitive binding experiments, one should know whether the probe is bound to binding site I or II. However, there is some confusion in the literature as to whether ANS is a probe for site I or site II.30,37 Sulfamethoxazole is weakly bound to BSA as indicated by the small decrease in the K1 value (Table 6) when sulfamethoxazole concentration is increased. As shown in Table 6, the most pronounced effect of the competition of drug and the probe is in the K1 value, while K2 and the number of binding sites n1 and n2 do not seem to be affected in the presence of the drug. It has been reported9,38 that sulfonamides do not induce allosteric effects (n1 and n2 are not altered) when bound to BSA or plasma proteins and that in competitive binding studies using ANS, only the highaffinity class of binding sites is probed. Using eq 19, the K1 binding

K1 ) (K1,ANS - (K1,ANS,app)/K1,ANS,app[I]0)

(19)

constant of sulfamethoxazole to BSA is derived,39 where K1,ANS is the binding constant of the probe (ANS) to the first class of binding sites and K1,ANS,app is the apparent binding constant in the presence of the drug. [I]0 is the drug concentration that in the FI procedure is kept constant. The determined K1 of sulfamethoxazole is 425 ( 12 M-1. Reported values of K1 for sulfamethoxazole binding to BSA are (1.5 ( 0.7) × 103 M-1, by potentiometry in phosphate buffer 0.10 M, pH 7.4 at 25 °C,9 and 1.6 × 103 M-1 (n1 ) 2.9) by fluorometry, in phosphate buffer 0.050 M, pH 7.4 at 27 °C.32 The value determined in this work is of the same order of magnitude, and the difference with literature values can be attributed to differences in BSA used. Equation 19 is not valid when the inhibitor (drug) is bound to more than one class of binding sites and/or at different classes than the probe (ANS); thus it cannot be used for the other four drugs. The combination of Scatchard model equations that describes the ANS-BSA and drug-BSA binding could lead to a complex nonlinear model suitable for calculating the binding parameters of the drug, but this is beyond the goals of this work. Salicylates are known to bind to both binding sites (I and II),30,40 stronger to binding site I.27 For binding to BSA, values reported for K1 ranged from 0.2 × 105 to 1.8 × 105 M-1 with n1 ) 1.4-2.6 and for K2 from 0.9 × 104 to 1.0 × 104 M-1 with n2 ) 8.6-21.5.30,35 (35) Sjoholm, I.; Ekman, B.; Kober, A.; Ljungstedt-Pahlman, I.; Seiving, B.; Sjodin, T. Mol. Pharmacol. 1979, 16, 767-777. (36) Herve, F.; Urien, S.; Albengres, E.; Duche, J.-C.; Tillement, J.-P. Clin. Pharmacokinet. 1994, 26, 44-58. (37) Sudlow, G.; Birkett, D. J.; Wade, D. N. Mol. Pharmacol. 1975, 11, 824832. (38) Moriguchi, I.; Wada, S.; Nishizawa, T. Chem. Pharm. Bull. 1968, 16, 601605. (39) Essassi, D.; Zini, R.; Tillement, J. P. J. Pharm. Sci. 1990, 79, 9-13. (40) Aarons, L. J.; Schary, W. L.; Rowland, M. J. Pharm. Pharmacol. 1979, 31, 322-330.

The results of our work indicate that salicylate inhibits ANS from the first BSA binding site and to a greater extent than sulfamethoxazole. Azapropazone is known to bind strongly to the first binding class (I).30,41,42 Results presented in Table 6 show the decrease of K1 in the presence of the drug, while K2 is less affected. The small increase in the number of binding sites with increasing drug concentration may be due to allosteric effects induced by the drug molecule upon binding. Literature values for azapropazone binding to BSA are for K1 7.8 × 105 M-1 with n1 ) 0.5 ( 0.7 and K2 (1.9 ( 0.4) × 104 M-1 with n2 ) 26.3 ( 3.2.30 Ketoprofen is a drug that binds stronger to site II than I.43 The K1 value for binding to BSA is (10.8 ( 2.0) × 105 M-1 with n1 ) 8.8 ( 0.2.30 The decrease of K1 and K2 of ANS, as well as n1 shown in Table 6, may be due to some sort of interaction between distinct binding sites of the two molecules. ANS may also be bound to binding site II, where ketoprofen is first bound and subsequently displaced by the drug. The decrease in n1 value shows some kind of tertiary structure change due to drug binding. According to previous studies, tolmetin is known to have a different binding site on the protein molecule, not binding to site I or II.41,44 From Table 6 it is shown that binding constants of ANS with BSA remain the same in the presence of 5.00 × 10-4 M tolmetin; thus we could also assume that the binding sites of tolmetin and ANS on BSA are different. However, changes are observed upon the addition of 1.00 × 10-3 M tolmetin. Changes in BSA tertiary structure caused by tolmetin binding may account for this observation. Results of competitive binding experiments indicate that ANS binds stronger to binding site I (interaction with azapropazone, salicylate, and sulfamethoxazole) and less to site II (interaction with ketoprofen) and has binding sites different from those of tolmetin. Results of Table 6 show that for higher drug concentrations the inhibition of ANS binding is, as expected, greater. In studies of competitive binding, a concentration ratio of drug to protein that is commonly used is 1:1. This is approximately the case for drug concentrations of 5.00 × 10-4 M (protein concentration 1.50 × 10-4 M). It is generally accepted that in this concentration ratio the various kinds of interaction (competitive binding, positive synergy, independent binding) are studied better. When this ratio is greater than 1:1, that is the case for the drug concentration of 1.00 × 10-3 M, the probe is always displaced by the high concentration of the drug when the drug binding sites are saturated.

CONCLUSIONS The proposed technique allows the full automation of binding studies of proteins with ionic micromolecules based on the highly (41) Diana, F. J.; Veronich, K.; Kapoor, A. L. J. Pharm. Sci. 1989, 78, 195199. (42) Kober, A.; Sjoholm, I. Mol. Pharmacol. 1980, 18, 421-426. (43) Lapicque, F.; Muller, N.; Payan, E.; Dubois, N.; Netter, P. Clin. Pharmacokinet. 1993, 25, 115-125. (44) Matsuyama, K.; Sen, A. C.; Perrin, J. H. J. Pharm. Pharmacol. 1987, 39, 190-195.

Analytical Chemistry, Vol. 71, No. 13, July 1, 1999

2549

precise FI concentration gradient monitored by the potentiometric sensor of the micromolecule ion. The flow rate of the carrier must be adequate to allow establishment of the complexation equilibrium and completeness of the electrode response. A binding experiment can be accomplished in 2-3 min providing a great number of data to be fitted in the Scatchard model or the stepwise stoichiometric model. The ANS ion can be used as a potentiometric probe in competitive binding studies for other micromolecules (e.g., drugs) provided that no interference is caused by the ionic drugs.

2550 Analytical Chemistry, Vol. 71, No. 13, July 1, 1999

ACKNOWLEDGMENT We gratefully acknowledge support from the ministry of Industry, Energy and Technology, General Secretariat of Research and Technology of Greece.

Received for review September 14, 1998. Accepted March 12, 1999. AC981019B