Anal. Chem. 1989, 6 1 , 1227-1230
a relatively large memory is essential for the inversion of the matrix. Four directions of exponentially expanding grids (Figure 8) were used to increase the accuracy. From the principle that the region of steeper concentration gradients has to provide more grids than the other regions, four directional expanding grids were selected as follows: f i t , NZEL grids from 2 = 0 to 2 = L' starting with a DZEL grid size, where 0 < L ' < L; second, NZSUB grids from 2 = L to 2 = L'starting with a DZSUB grid size; third, NREL grids from R = 1to R = 0 starting with a DREL grid size; fourth, NRGL grids from R = 1to R = RG starting with a DRGL grid size. The above formulation, including automatic two-dimensional exponential grid generation, was coded in FORTRAN and executed on either a CDC 6000 dual cyber computer (Control Data Corp.) or a Macintosh I1 (Apple Computer, Inc). T o obtain one data point in Figure 5, took 35 s on the CDC 6000 and 390 s on the Macintosh 11. The needed number of bytes of memory largely depended on the size of band matrix and inverted band matrix, i.e., the number of memory bytes = 8.2.n.(nR + 2) or 8.2.n.(nz + 2), where 8 bytes were used for each floating point number. Once the steady-state concentration profiles were obtained, the current at the disk electrode was easily calculated by the
1227
summation of the normal components of the concentration gradient.
LITERATURE CITED Bard, A. J.; Fan, F.4. F.; Kwak, J.; Lev, 0. Anal. Chem. 1989, 67, 132-138. Newman, J. J . Electrochem. Soc. 1966, 713, 501-502. Saito, Y. Rev. Polarogr. 1868, 75, 177-187. Feldberg, S. W. In Electroanalytical Chemistry; Bard, A. J., Ed.; Marcel Dekker: New York, 1969; Vol. 13. Feldberg, S. W. In Computers in Chemistry and Instrumentation; Mattson, J. S., Mark, H. B., MacDonald, H. C., Jr., Eds.; Marcel Dekker: New York, 1972; Vol. 2. Huebner, K. J.; Thornton, E. A. The Finite Element Methw' for Engineers, 2nd ed.;Wliey-Interscience: New York, 1982. Rao, S. S. The Finite Element Method in Engineering; Pergamon Press: New York, 1982. J o s h T.; Pletcher, D. J . Electroanal. Chem. 1974, 4 9 , 172-186. Feldberg, S. W. J . Electmanal. Chem. 1981, 727, 1-10, Bard, A. J.; Crayston, J. A.; Klttlesen, G. P.;Varco Shea, T.; Wrighton, M. S. Anal. Chem. 1886, 58. 2321-2331. Aoki, K.; Osteryoung, J. J . Electroanal. Chem. 1081, 722, 19-35. Shoup, D.; Szabo, A. J . Electroanal. Chem. 1082, 140, 237-245. Hubbard. A. T.; Anson, F. C. In €lectroanalytical Chemistry; Bard, A. J., Ed.; Marcel Dekker: New York, 1970; Vol. 4.
RECEIVED for review December 9, 1988. Accepted March 1, 1989. The support of this research by the Robert A. Welch Foundation is gratefully acknowledged.
Continuous Liquid-Phase Piezoelectric Biosensor for Kinetic Immunoassays Kenneth A. Davis*,* Department of Pediatrics, RD-20, University of Washington, Seattle, Washington 98195
T. Richard Leary National Biosensor, 20212 Densmore North, Seattle, Washington 98133
The reactlons of lmmunoglobuilns wHh protein A and, subsequently, of antlbodies to these lmmunoglobullns were continuously monitored by a plezoeiectrlc biosensor. AT-cut crystals, with a fundamental resonant frequency of 10 MHz, were mounted In a Plexlglas holder, and one slde was directly coated with proteln A. Upon exposure to solutions contalning rabbit or human IgG the resonant frequency was followed contlnuously and its decrease due to the binding of IgG to proteln A observed. Subsequent addition of sheep antihuman IgG to the now immobilized human IgG caused a specific 3-fold further decrease In resonant frequency. We observed a frequency change of approxlmately 1 Hr for each 10 ng of added immunoglobulin. Decreaslng the pH to 3 released the bound IgG but not the proteln A and permitted reuse of the crystal for further IgG binding.
Increasing attention is being paid to the development of nonelectrode biosensors, especially those that can be used to determine clinically important molecules (1,2). These sensom are constructed by immobilizing a selective binding surface 'Present address Becton Dickinson Monoclonal Center, 2375
Garcia Ave., Mountain View, CA 94043.
0003-2700/89/0361-1227$01.50/0
to a transducer. The surface can be an absorptive organic fii, an immobilized antigen, an immobilized antibody against a specific antigen, or other proteins with specific binding sites. Selective binding of a molecule to the absorptive surface causes the transducer to change one or more of its fundamental signals. The piezoelectric quartz crystal is such a transducer and is commonly used as a low-cost frequency standard. The crystal oscillates a t a very specific resonant frequency when placed in an appropriate circuit. The frequency of this resonance can be changed by adding mass to the surface of the crystal. Some biological applications of quartz crystal microbalances have been reported. Quartz crystals have been coated with antibodies to the pesticide parathion and have been used to specifically detect parathion in gases that are passed over the crystal ( 3 ) . Shons et al. ( 4 ) used antigen-coated crystals exposed to specific antisera and demonstrated a mass increase after washing and drying. Similarly, Muramatsu et al. (5) have detected specific mass changes of air-dried crystals coated with monoclonal antibodies to the yeast Candida albicans before and after incubation with suspensions of C. albicans. Recently, Muramatsu et al. (6) have reported the coupling of protein A to 9-MHz AT-cut crystals with (7-aminopropy1)triethoxysilane. They measured IgG binding by observing the change in resonant frequency of a crystal immersed in distilled 0 1989 American Chemical Society
1228
ANALYTICAL CHEMISTRY, VOL.. 61, NO. 11, JUNE 1, 1989
CRYSTAL
Flgure 1. Apparatus: (A) AT-cut quartz crystal with gold electrodes: (B)cut-away view of detector cell.
water before and after exposure to these immunoglobulins. Their end-point frequency measurements, in distilled water, obviated the necessity of carefully drying specimens as reported in their previous study ( 5 ) . The behavior of piezoelectric crystals in liquid has been studied in some detail (7-9) and the mass sensitivity in solution found to agree well with the value predicted by the Sauerbrey equation (10). However, few liquid-phase biological applications of piezoelectric crystals have been reported. A piezoelectric quartz surface acoustic wave (SAW) device has been used to detect IgG in solution using immobilized antibody (11) with a reported detection limit of only 13 Kg of antigen. Thompson et al. (12) have attempted to detect human IgG with goat antihuman IgG immobilized on an AT-cut crystal. However, they observed frequency shifts opposite in direction to those predicted from experience in the gas phase. In this paper we report the use of an oscillator circuit, designed to compensate for the degraded quality of the crystal's efficiency (8)in an energy-absorbing medium (13), to oscillate an AT-cut piezoelectric quartz crystal in an aqueous environment for the detection of protein-protein interactions. The crystal frequency is continuously monitored in phosphate-buffered saline (PBS),permitting observations of the kinetics of these biologically interesting reactions from inception. The detector's response is consistent with what has been observed with gas-phase detectors, and it allows an estimation of sensitivity and the initial kinetics of these interactions. The implications of this detector for biological assays are discussed.
EXPERIMENTAL SECTION Quartz piezoelectriccrystals (AT-cut) 1.2 cm in diameter with a fundamental frequency of 10 MHz were purchased from Standard Crystal Corp. (El Monte, CA) with gold electrodes (about 0.22 cm2)vacuum-deposited on either side and having 180" flags. Magnet wire leads were attached to the crystal with silver-filled epoxy (BA-2902, Tra-Con, Medford, MA). The crystals were powered by an oscillator circuit constructed as described by Simpson (13). Clamping crystals between two Plexiglas blocks with Neoprene O-ring seals, as shown in Figure 1, allowed the addition of liquid to the upper surface of the crystal. O-ring seals around bolts stabilizedthe Plexiglas holder and minimized crystal breakage. All analyses were performed at room temperature, approximately 25 "C. Chromatographically purified protein A and cyanogen bromide activated Sepharose were obtained from Sigma Chemical Corp. (St.Louis MO). Rabbit y-globulin, affinity purified, was obtained
from Miles Laboratories (Naperville, IL), and partially purified sheep IgG (30 mg of protein/mL) from an animal immunized against human IgG was obtained from Kent Laboratories, Inc. (Redmond, WA). The sheep IgG preparation was assayed for antibody specifically directed against human IgG by adsorption to a protein A column presaturated with human IgG. The above column bound approximately 30% of the partially purified antiserum, and therefore, the concentration of specific antibody was estimated at 9 mg/mL. Solvents were of reagent grade, and all other reagents were obtained from Sigma Chemical Corp. Protein A was coupled to activated Sepharose by standard techniques. Protein concentration was determined by the dye binding method of Bradford (14). Protein A Coating. The quartz crystals were soaked for 20 min in 1.2 N NaOH, rinsed with distilled HzO, and soaked 5 min in 1.2 N HCl. Then 100 NLof concentrated HCl was placed in the center of the electrode to be coated, with care taken to avoid contact with attached wires. After 2 min the crystal was thoroughly rinsed with H20 followed by 95% ethanol and dried at 200 "C for 20 min. Crystals were removed from the oven immediately before mounting in a Plexiglas chamber (Figure 1). Protein A (0.1 mg) in 0.2 mL of 50 mM potassium phosphate buffer, pH 7.2, was added to the chamber holding crystal, as soon as possible after mounting, and incubated for varying lengths of time. In later experiments the same amount of protein A was added in one-half the volume of phosphate buffer, and an equal volume of 0.1 M sodium acetate buffer, pH 4.5, containing 5% NaCl was then added to bring the pH of the mixture to near the isoelectric point of protein A (pH 5.5) and a standard incubation time of 2 h adopted. Reactions in Chamber. The protein A coated crystals were washed several times with PBS, and then PBS containing 5 % bovine serum albumin (BSA) was added to block any nonspecific protein binding sites. After equilibration, about 20 min, a small volume (1-2 pL) of rabbit or human IgG was introduced with mixing. The increase of crystal mass, as IgG complexed with protein A, was followed as a change in the resonant frequency of the crystal.
RESULTS AND DISCUSSION Gold surfaces labeled with biospecific ligands such as lectins, antibodies, and hormones have been used as probes in immunohistochemistry (15), in light microscopy (16), and in transmission and scanning electron microscopy (17). Goldprotein A complexes are highly stable, have an association constant of los M-l, and are believed to be Van der Waals in nature (18). In our present studies we have taken advantage of this phenomenon to develop piezoelectric crystals having protein A bound to the vapor deposited gold electrodes. Details of conditions required to achieve stable coating are listed above. We were not able to follow the binding of protein A to the crystal since the signal was complicated by the very large frequency shift associated with the addition of protein A in buffer to a dry crystal, a key condition to obtaining protein A tightly bound to the prepared gold electrode. After the crystal was coated with protein A and washed successively with PBS and distilled water, the frequency of the dried crystal decreased approximately 200 Hz. Figure 2 shows the reaction of protein A coated crystals with varying amounts of rabbit IgG. Prior to addition of IgG to the sensor the nonspecific binding sites on the electrodes were blocked by the addition of a large excess of BSA (5 mg/mL final concentration). Subsequent addition of affinity purified rabbit IgG (5 wg) to these crystals (Figure 2A) caused a frequency shift of approximately 250 Hz. A second addition of IgG ( 5 wg) caused some additional frequency decrease. This further binding is apparently of low affinity and is reversed when the solution over the crystal is replaced by PBS and BSA. Allowing this buffer to remain in contact with the crystal for 1 h did not cause any further loss of protein, but a third addition of IgG once again caused the frequency to decrease. The IgG response is specific for protein A. When IgG is added
ANALYTICAL
A
B
CHEMISTRY, VOL. 61, NO. 11, JUNE 1, 1989
1229
C Pi
ANTI I g G
t
0SA I
I
2
I
TIME
(hours)
Flgure 2. Frequency response of protein A coated crystals: (A) BSA block of nonspecific binding followed by the addition of 5 pg of rabbit
IgG, PBS plus BSA wash, and subsequent readdiiion of IgG; (B) BSA block followed by addions of 1.3and 5.4 pg of rabbit IgG;(C) control, BSA block of uncoated crystal followed by the addition of 5 pg of rabbit IgG.
to a crystal that has not been coated with protein A but is blocked with BSA, there is no response (see Figure 2C). In parts A and B of Figure 2 BSA blocking caused frequency decreases of 80 and 125 Hz, respectively. Time appears to be a factor in determining the amount of protein A bound by a crystal. In experiment 2A the crystal was preincubated 45 min with protein A whereas in 2B only 15 min. In Figure 2C BSA blocking in the absence of a protein A precoat caused a frequency decrease of over 270 Hz. This suggests a protein A coating of 70% and 54% of saturation in parts A and B of Figure 2, respectively. In Figure 3, protein A coating was carried out longer and at pH 5.5 (near the isoelectric point of protein A) and was much more complete as determined by the lack of significant frequency change upon addition of BSA. Figure 3 shows that a protein A coated crystal subsequently saturated with human IgG specifically reacts with antibodies to human IgG. In this experiment protein A coated crystals were exposed to human IgG and allowed to react approximately 1h while the resonant frequency was followed; then the crystal was washed with PBS and covered again with PBS containing 5 mg/mL BSA (0.2 mL). As a control, plasma from an un-immunized sheep was added. This caused a small, stepwise drop in resonant frequency of about 12 Hz. In similar control experrments (not shown) addition of goat antiserum against human IgM also had no significant effect. The addition of sheep IgG from an animal immunized against human IgG caused a dramatic decrease in resonant frequency. Mixing of the solution over the crystal during this reaction caused sharp increases in the rate of frequency change. This suggests that diffusion is an important factor. This is not surprising, as the liquid depth over the crystal was approximately 0.5 cm. Thus it will be important to modify future chamber designs to permit thinner liquid layers or continuous mixing. Addition of more sheep antihuman IgG caused a further frequency decrease, indicating that all the binding sites were not saturated. The frequency change upon addition of human IgG to protein A (470 Hz)and the larger change observed after saturation with antihuman IgG (1400 Hz) suggest an average binding of three molecules of antibody per molecule of antigen. This observation is not surprising, as the sheep antihuman IgG is polyclonal and would be expected to recognize multiple determinants on the human IgG molecule. Figure 3 also demonstrates dissociation of the protein A-IgG complex at low pH. After incubation with a second aliquot of IgG for over 1h the reaction buffer was replaced with one at pH 3 (0.1 M sodium acetate). The resonant frequency
1
I
2 TIME
3 (hours)
4
5
6
Figure 3. Frequency response of a protein A coated crystal to the addition of human IgG followed by sheep antihuman IgG with recycling following dissociation at pH 3. A protein A coated crystal was covered with 0.2 mL of 0.5% BSA in PBS (B) and then 1 I.IL of human IgG (50 pg) added. After approximately 1 h the crystal was washed four times with PBS and again covered with B. Then 5 pL of sheep plasma was added, followed by 1 pL of sheep antihuman I@ (9 pg). The frequency was monitored, the solution on the crystal was mixed several times, and then an additional 1 pL of sheep antihuman IgG was added. After reacting for over 1 h, the solution was replaced with 0.2 mL of 0.1 M sodium acetate buffer, pH 3.0, and the initial steps of the experiment were repeated.
increased very rapidly, presumably as IgG dissociated from the protein A bound to the electrode. Within 3 min the frequency was up to the level observed prior to the initial addition of human IgG. Replacement of the pH 3 buffer with PBS plus BSA followed by addition of a fresh aliquot of human IgG caused a decrease in frequency to within 2 Hz of that observed after the first such addition. Washing with PBS plus BSA and a subsequent addition of sheep antihuman IgG again caused a rapid decrease in frequency, demonstrating the reversibility of this system. How long the protein A can remain at pH 3 without suffering irreversible damage is being studied. Figure 3 shows the results of the addition of subsaturating amounts of antibody. In this experiment the initial addition of 9 pg of sheep antihuman IgG was not saturating, and an apparent end point was reached after a frequency decrease of approximately 910 Hz. The response of this system is a frequency shift of approximately 1 Hz per 10 ng of protein added. The Sauerbrey equation (19) predicts that for a 10MHz quartz resonator in air the deposition of 1ng of material on a surface area of 0.22 cm2 will cause a frequency decrease of 1 Hz. The nearly 10-fold discrepancy between the theoretical and observed frequency shifts may simply reflect incomplete binding or be due, at least in part, to the binding of antibody to IgG attached to nonresonating surfaces such as the electrode ”flags”, the bare quartz, or the chamber sides. Experiments with radiolabeled IgG should better address this question. One interesting result of these experiments is the demonstration that even the third layer of protein on the crystal surface causes a rapid and sensitive response. It should also be noted that the response we observe is in the direction of and of similar magnitude to that of the classical microgravimetric signal. Some decreased sensitivity upon addition of a third protein layer may partially explain the discrepancies of stoichiometry observed above. Recently Thompson et al. (9, 12), using an AT-cut piezoelectric crystal, measured an increase in frequency, “contrary to expectation for a classical microbalance response”, as IgG
Anal. Chem. 1989,6 1 , 1230-1235
1230
complexes with antigen. They explain this as being due to “changes in surface interactions during the immunochemical reaction”. With our system we do not observe an increase in frequency, under similar conditions, even after several hours. The above papers report the use of “thin” coatings of approximately 50 pm of acrylamide or (aminopropy1)silane attached to the sensor surface. Preliminary work by us using thiopropyl and aminopropyl silanes suggested that crystals coated with greater than 2000-5000 A of silane (less than 1% of the Bbove coatings) would not resonate in an aqueous environment using the Hager circuit (13). The observations in this present report may be the result of the use of much thinner coatings. In another report surface acoustic wave (SAW) crystals were used, in liquid phase, to follow antibody-antigen interactions ( I 1). Although these authors report significant background problems, the SAW-type crystal did behave in a classical fashion, and they observed a decrease in resonant frequency as antigen complexed with antibody, in accord with our observations. The ability of the system used in this communication to continuously monitor frequency during protein-protein interactions permits observation of not only the extent but the kinetics of these interactions, especially when diffusion effects are minimized. The present conditions, however, are not unlike those routinely used for enzyme-linked immunosorbent assays (ELISA) and similar assays. The results reported above clearly demonstrate that piezoelectric crystals can be used for continuous monitoring in liquid and are of significant potential use as an analytical tool, especially for biological macromolecules. The detectable concentrations of proteins are already at levels that could make this technology useful in the clinical laboratory and doctor’s office. The crystals are inexpensive, and detection depends upon a fundamental property, mass. Therefore, additional reagents such as radioactive, fluorescent, or enzyme labels are unnecessary. Sensitivity can be increased, within limits, by decreasing the surface area of the electrode and, even more, by increasing the fundamental frequency of the crystal. Arrays
of electrodes on a single crystal (20) would permit the simultaneous measurement of several parameters in a single sample. Analysis of smaller molecules should be possible by displacement of similar molecules bound to large carriers similar to many standard competitive assays in use today. Finally, a multiple analysis system should by possible using protein A with a pH cycling system to regenerate a fresh reactive surface. We believe this to be an exciting and versatile technology for the clinical laboratory. Registry No. Quartz, 14808-60-7.
LITERATURE CITED (1) Vadgarna, P.; Davis, G. Clin. Chem. 1085, 42,333-345. (2) Owen, V. M. Ann. Clin. Biochem. 1085, 22,559-564. (3) Ngeh-Ngwainbi, J.; Foley, P. H.; Kuan, S. S.; Guilbault, G. G. J. Am. Chem. SOC. l08& 108, 5444-5447. (4) Shons, A.; Dorman, F.; Najarian. J. J. Biomed. Mater. Res. 1072, 6 , 565-570. (5) Muramatsu, H.; Kajlwara, K.; Tarniya, E.; Karube, I. Anal. Chim. Acta 1986, 188, 257-261. (6) Muramatsu, H.; Dicks, J. M.; Tarniya, E.; Karube, I. Anal. Chem. 1087, 59, 2760-2763. (7) Kanarawa, K. K.; Gordon, J. G. Anal. Chin?. Acta 1085, 175, 99. (8) Hager, H. E. Chem. Eng. Commun. 1086, 43,25-38. (9) Thompson, M.; Dhaliwal, G. K.; Arthur, C. L.; Calabrese, G. S. I€€€ Trans. Ultrasonics , Ferroelectrics, Freq. Cont. 1087, UFFC-34, 127-135. (IO) Bruckenstein, S.; Shay, M. Electrochim. Acta 1085, 3 0 , 1295-1300. (11) Roederer, J. E.; Bastiaans, G. J. Anal. Chem. 1983, 55, 2333-2336. (12) Thompson, M.; Arthur, C. L.; Dhaliwal, G. K. Anal. Chem. 1086, 5 8 , 1206-1 209. (13) Simpson, R. L. Ph.D. Dissertation, University of Washington, Seattle, WA, 1985. (14) Bradford, M. M. Anal. Biochem. 1076, 72, 248-254. (15) Tolson, N. D.; Boothroyd, B.; Hopkins. C. R. J . Microsc. (Oxford) 1081, 123,215-226. (16) Roth, J. J. Histochem. Cytochem. 1082, 30,691-696. (17) Horisberger, M. CellBiol. 1070, 36, 253-258. (18) Horisberger, M.; Clerk, M.-F. Histochemistry 1085, 82,219-223. (19) Sauerbrey. G. Z. Phys. 1050, 155, 206. (20) Carey, W. P.;Beebe, K. R.; Kowalski, B. R. Anal. Chem. 1087, 59, 1529.
RECEIVED for review January 12, 1988. Accepted March 3, 1989.
Determination of Total Mercury in Water and Urine by a Gold Film Sensor following Fenton’s Reagent Digestion Liu Ping and Purnendu K. Dasgupta* Department of Chemistry and Biochemistry, Texas Tech University, Lubbock, Texas 79409-1061
+
Fenton’s reagent (Fe(1I) H202)Is utlllzed for the digestion of environmental water samples and urlne. Following the digestion, whlch converts organlc forms of Hg to inorganic Hg(II), Hg(0) is liberated by borohydride reductlon and measured by a conductometric gold flhn sensor. Ouantltative recovery of Hg from samples spiked with mercuric chloride, methylmercury(I I ) chloride, and phenylmercury(I I ) acetate was attainable in the presence of naturally occurrlng suspended matter and humlc and fuivlc acids as well as 3 % NaCi. The dlgestion is performed at moderate pH (3-4) and temperature ( S O “C) and does not use large amounts of any reagent. Excellent agreement Is shown for reference water, wastewater, and urine standards. The limit of detection, facilitated by the low blank value, Is 500 pg of Hg or 10 ng/L for a 50-mL sample.
INTRODUCTION An estimated (1.42 f 1.28) X lo9 g of mercury is annually mobilized into the environment from fossil fuel combustion alone ( 1 ) . Mercury is unique in many of its physicochemical properties and occupies a singularly important place in the present state of human technological existence. Yet almost without exception, mercury and its compounds are toxic. Not surprisingly, the literature on the environmental and physiological occurrence of mercury is extensive. The last major bibliography, compiled more than a decade ago, contained over 4000 citations (2). A more recent review summarizes current work on the determination of mercury in water and wastewater (3).
Routine analytical needs exist for measuring the total mercury content of environmental and physiological samples,
0003-2700/89/0361-1230$01.50/0 1989 American Chemical Society
