Anal. Chem. 1998, 70, 4624-4630
Automated Amplified Flow Immunoassay for Cocaine Christian G. Bauer,† Arkadi V. Eremenko,† Andrea Ku 1 hn,† Konrad Ku 1 rzinger,‡ Alexander Makower,† and Frieder W. Scheller*,†
Institute of Biochemistry and Molecular Physiology, University of Potsdam, Im Biotechnologiepark, D-14943 Luckenwalde, Germany, and Lab Diagnostics, Boehringer Mannheim GmbH, Nonnenwald 2, D-82377 Penzberg, Germany
An amplified flow immunoassay (AFIA) was developed for cocaine, which combines a noncompetitive immunoenzymometric assay (IEMA) with an on-line detection of the enzyme label alkaline phosphatase (ALP) by a substraterecycling biosensor. In the IEMA, the analyte cocaine first binds to a labeled polyclonal anti-cocaine antibody. Then, the excess labeled antibody is separated on an affinity column that contains a perfusion chromatography carrier modified by immobilized cocaine. The unbound complexes of the analyte cocaine with the ALP-labeled antibody are detected postcolumn. The detector senses phenol produced by ALP from phenyl phosphate. As detector, an amperometric substrate-recycling biosensor was used, which consists of a Clark-type oxygen electrode covered by tyrosinase and pyrroloquinoline quinonedependent glucose dehydrogenase. The lower limit of detection is 380 pM (38 fmol) for cocaine. The sampling rate is 26/h. Cocaine could be detected from “real samples” with an imprecision of (10% (n ) 3) and with a recovery of 49 ( 3% for various concentrations. AFIA is generally important as a new approach for the fast detection of picomolar concentrations of haptens. The fast detection of illegal drugs is an important analytical task. Smuggling of drugs such as cocaine is a worldwide problem and this causes a demand for screening techniques that can provide good sensitivity and fast results repetitively. Detection methods for cocaine include separation techniques such as high-perfomance liquid chromatography (HPLC), gas chromatography/mass spectrometry (GC/MS), and antibodybased approaches. HPLC1-3 is normally combined with solidphase extraction and reaches a detection limit of 3 nM cocaine. Various MS techniques4-8 can achieve similar detection limits. †
University of Potsdam. Boehringer Mannheim GmbH. (1) Muztar, J.; Chari, G.; Bhat, R.; Ramarao, S.; Vidyasagar, D. J. Liq. Chromatogr. 1995, 18, 2635-45. (2) Barat, S. A.; Kardos, S. A.; Abdelrahman, M. S. J. Appl. Toxicol. 1996, 16, 215-9. (3) Tagliaro, F.; Antonioli, C.; Debattisti, Z.; Ghielmi, S. Marigo, M. J. Chromatogr., A 1994, 673, 207-15. (4) DelaTorre, R.; Ortuno, J.; Gonzalez, M. L.; Farre, M., Cami, J.; Segura, J. J Pharm. Biomed. Anal. 1995, 13, 305-12. (5) Muddiman, D. C.; Gusev, A. I.; Martin, L. B.; Hercules, D. M. Fresenius J. Anal. Chem. 1996, 68, 103-10. ‡
4624 Analytical Chemistry, Vol. 70, No. 21, November 1, 1998
But both techniques rely on sophisticated and expensive instruments. The antibody-based techniques, however, are inherently less complicated. These are microtiter plate immunoassays9-11 and flow immunosensors.10,12-16 But these antibody-based techniques for the measurement of cocaine are either not fast9-12 or not sensitive,15,16 or do not allow frequent repetitive measurements.13,14 So there is still the need for a fast and sensitive method of cocaine detection. The literature provides examples of fast flow injection immunoassays (FIIA) for haptens.17-30 With signal response times from (6) Cardenas, S. Gallego, M.; Valcarcel, M. Rapid Commun. Mass Spectrom. 1996, 10, 631-6. (7) Cone, E. J.; Hillsgrove, M.; Darwin, W. D. Clin. Chem. 1994, 40, 1299305. (8) Wang, W. L.; Darwin, W. D., Cone E. J. J. Chromatogr., B 1994, 660, 27990. (9) Spiehler, V.; Fay, J.; Fogerson, R.; Schoendorfer, D.; Niedbala, R. S.; Clin. Chem. 1996, 42, 34-8. (10) Ziegler, T., Eikenberg, O.; Bilitewski, U.; Grol, M. Analyst 1996, 121, 11925. (11) Chen, P.; Watts, D. S.; Tai, H. H. Res. Commun. Substance Abuse 1994, 15, 71-82. (12) Devine, J. P.; Anis, N. A.; Wright, J.; Kim, S.; Eldefrawi, A. T.; Eldefrawi, M. E. Anal. Biochem. 1995, 227, 216-24. (13) Ogert, R. A.; Kusterbeck, A. W.; Wemhoff, G. A.; Burke, R., Ligler, F. S. Anal. Lett. 1992, 25, 1999-2019. (14) Yu, H.; Kusterbeck, A. W.; Hale, M. J.; Ligler F. S.; Whelan, J. P. Biosens. Bioelectron. 1996, 11, 725-34. (15) Goerlach-Graw, A.; Carstensen, C. A. TIAFT/SOFT Joint Congress, Tampa, FL, 1994. (16) Watanabe, K.; Okada, K.; Oda, H.; Furuno, K.; Gomita, Y.; Katsu, T. Anal. Chim. Acta 1995, 316, 371-5. (17) Palmer, D. A.; Mark, E.; Miller, J. N.; French, M. T. Analyst 1994, 119, 943-7. (18) Palmer, D. A.; Miller J. N. Anal. Chim. Acta 1995, 303, 223-30. (19) Oosterkamp, A. J.; Villaverde-Herraiz, M. T.; Irth, H.; Tjaden, U. R.; van der Greef, J. Anal. Chem. 1996, 68, 1201-6. (20) Osipov, A. P.; Zaitseva, N. V.; Egorov, A. M. Biosens. Bioelectron. 1996, 11, 881-7. (21) Reeves, S. G.; Rule, G. S.; Roberts, M. A.; Edwards, A. J.; Durst, R. A. Talanta 1994, 10, 1747-53. (22) Chiem, N.; Harrison, D. J. Anal. Chem. 1997, 69, 373-8. (23) Schmalzing, D.; Wassim, N.; Xian-Wei, Y. Mhatre, R., Regnier, F. E.; Afeyan, N. B.; Fuchs, M. Anal. Chem. 1995, 67, 606-12. (24) Wilmer, M.; Trau, D. Renneberg, R.; Spener, F. Anal. Lett. 1997, 30, 51525. (25) Ogert, R. A.; Kusterbeck, A. W.; Wemhoff, G. A.; Burke, R.; Ligler, F. S. Anal. Lett. 1992, 25, 1999-2019. (26) Yu, H.; Kusterbeck, A. W.; Hale, M. J.; Ligler, F. S.; Whelan, J. P. Biosens. Bioelectron. 1996, 11, 725-34. (27) Narang, U.; Gauger, P. R.; Ligler, F. S. Anal. Chem. 1997, 69, 1961-4. (28) Kronkvist, K.; Lo ¨vgren, U.; Svenson, J.; Edholm, L. E.; Johansson, G. J. Immunol. Methods 1997, 200, 145-53. 10.1021/ac971388s CCC: $15.00
© 1998 American Chemical Society Published on Web 10/07/1998
1 to 15 min, these are fast compared to microtiter plate immunoassays. Most of the FIIA use competitive,17-24 displacement,25-28 or noncompetitve formats.29,30 Competitive FIIA requires at least 5 min of total assay time for analyte concentrations of less than 10 nM. Competitive assays are most sensitive when the analyte-antibody reaction is allowed to proceed to equilibrium.31 There are a number of sensitive assays that allow fast sampling rates,22,23,28,32,33 but much of the signal response time is spent waiting for the binding of analyte by a limited amount of antibodies. An additional drawback of a typical competitive assay is the time-consuming task of regenerating the column between each measurement.17,18,20,21,24 The harsh regeneration conditions can decrease the binding capacity and hence the lifetime of the column.18,24,34,35 The regeneration of column can be avoided by using postcolumn detection of the nonbound fraction of the label; this also accelerates and stabilizes the assay.19,28,32 Displacement assays are inherently faster than competitive assays. They can be as sensitive as competitive assays when antibodies with medium affinities are employed. This, in turn, severely limits the lifetime of the column.25,26 The solution to resaturate the antibody column with labeled hapten after each injection doubles the assay time.28 Displacement assays as well as competitive or sandwich assays obviously show markedly increased sensitivity when performed in capillaries.27,36,37 Noncompetitive immunoassays, which use an excess of labeled antibodies vs analyte, are more sensitive than competitive or displacement assays, because, theoretically, one molecule of label is provided for every molecule of analyte. Noncompetitive immunoassays rely exclusively on fast association reactions and, thus, achieve binding between analyte and antibodies faster than displacement or competitive assays where dissociation contributes to signal generation. Prior to label detection, these noncompetitive immunoenzymometric assays (IEMA) use an affinity column with immobilized analyte to capture the excess labeled antibodies. The analyte complex of the labeled antibodies passes the column. This affinity purification takes several minutes using conventional chromatography.29,30,33 The perfusion chromatography used in this paper allows separations with 10-fold speed, maintaining high capacity and high resolution.38 We achieved separations with less than 15-s contact time on a high-capacity BZE-affinity perfusion column employing a peristaltic pump.39 IEMA saves additional (29) Arefyev, A. A.; Vlasenko, S. B.; Eremin, S. A., Osipov, A. P.; Egorov, A. M. Anal. Chim. Acta 1990, 237, 285-9. (30) Hara, T.; Hakamura, K.; Satomura, S.; Matsuura, S. Anal. Chem. 1994, 66, 351-4. (31) Standefer, J.C Saunders, G. C. Clin. Chem. 1978, 24, 1903-7. (32) Locasio-Brown, L.; Martynova, L.; Christensen, R. G.; Horvai, G. Anal. Chem. 1996, 68, 1665-70. (33) Gunaratna, P. C.; Wilson, G. S. Anal. Chem. 1993, 65, 1152-7. (34) Nilsson, M.; Hakanson, H.; Matthiasson, B. Anal. Chim. Acta 1991, 249, 163-8. (35) Nilsson, M.; Hakanson, H.; Matthiasson, B. Process Control Qual. 1992, 4, 37-45. (36) Kaneki, N.; Xu, Y.; Kumari, A.; Halsall H. B., Heineman, W. R.; Kissinger P. T. Anal. Chim. Acta 1994, 287, 253-8. (37) DeFrutos, M.; Paliwal, S. K., Regnier, F. E. Anal. Chem. 1993, 63, 215963. (38) Paliwal, S. K.; DeFrutos, M.; Regnier, F. E. Methods Enzymol. 1996, 270, 133-51. (39) Eremenko, A. V.; Bauer, C. G.; Makower, A.; Kanne, B.; Baumgarten, H.; Scheller, F. W. Anal. Chim. Acta, in press.
time because the regeneration of the analyte-coupled affinity column is not needed as frequently.29,33,40 In summary, the noncompetitive IEMA possesses a number of distinct advantages for the fast, sensitive, and repetitive detection of a hapten analyte like cocaine. Fast analysis by an enzyme-labeled immunoassay requires an ultrasensitive detection method to keep the substrate incubation time short. A good means to achieve this ultrasensitivity is signal amplification by substrate recycling.41-49 Recently, we demonstrated a zeptomole-detecting biosensor for alkaline phosphatase using an amplified phenol biosensor that quantitates the enzymatic hydrolysis of phenyl phosphate.50 The aim of this study was to develop a fast and sensitive, automated, and technically simple screening method for the repetitive analysis of cocaine. Our approach is called amplified flow immunoassay (AFIA) and consists of three steps (Figure 1): first, the immunorecognition, where cocaine binds to the alkaline phosphatase (ALP)-labeled antibody (pAb-ALP). Second, excess pAb-ALP is removed by a cocaine-modified affinity column. Third, the cocaine-pAb-ALP complex is detected in the column effluent. This ALP detection includes the enzymatic hydrolysis of phenyl phosphate to phenol and the amplified detection of the generated phenol. The phenol detector is a Clark-type oxygen electrode covered by an enzyme layer containing tyrosinase and pyrroloquinoline quinone (PQQ)-dependent glucose dehydrogenase (GDH). When phenol is present, it is oxidized by tyrosinase via catechol to o-quinone. Oxygen is consumed concomitantly, leading to a transient decrease in the oxygen baseline current. This oxygen consumption is amplified by recycling the tyrosinase-substrate catechol. GDH reduces o-quinone back to catechol, thus amplifying the signal ∼350-fold. In this paper, we describe the instrument, its performance to measure phenol, alkaline phosphatase, and cocaine, its optimization, and stabilization for prolonged multiple measurements. Photometrical and amperometric detection of this cocaine immunoassay are compared. Last, we tested this method for matrix blanks and applied it to the measurement of cocaine in real samples. (40) Irth, H.; Oosterkamp, A. J.; van der Welle, W.; Traden, U. R.; van der Greef, J. J. Chromatogr. 1993, 633, 65-72. (41) Wollenberger, U.; Schubert, F.; Pfeiffer, D.; Scheller, F. W. Trends Biotechnol. 1993, 11, 255-62. (42) Johansson, A.; Ellis, D. H., Bates, D. L.; Plumb, A. M., Stanley, C. J. J. Immunol. Methods 1986, 87, 7-11. (43) Stanley, C. J.; Cox, R. B.; Cardosi, M. F.; Turner, A. P. F. J. Immunol. Methods 1988, 112, 153-61. (44) Kronkvist, K.; Wallentin, K.; Johansson, G. Anal. Chim. Acta 1994, 290, 335-42. (45) Fischer, M.; Harbron, S.; Rabin, B. R. Anal. Biochem. 1995, 227, 73-9. (46) Bier, F. F.; Ehrentreich-Fo ¨rster, E.; Makower, A.; Scheller, F. W. Anal. Chim. Acta 1996, 328, 27-32. (47) Scheller F. W.; Makower, A.; Ghindilis, A. L.; Bier, F. F., Fo¨rster, E.; Wollenberger, U.; Bauer, C.; Micheel, B.; Pfeiffer, D.; Szeponik, J.; Michael, N.; Kaden, H. In Biosensor and Chemical Sensor Technology; Rogers, K. R., Mulchandani, A., Zhou, W., Eds.; ACS Symposium Series 613; American Chemical Society: Washington, DC, 1995; pp 70-81. (48) Makower, A.; Eremenko, A. V.; Streffer, K.; Wollenberger, U.; Scheller, F. W. J. Chem. Technol. Biotechnol. 1996, 65, 39-44. (49) Bier, F. F.; Ehrentreich-Fo¨rster, E.; Bauer, C. G.; Scheller, F. W. Fresenius J. Anal. Chem. 1996, 254, 861-5. (50) Bauer, C. G.; Eremenko, A. V.; Ehrentreich-Fo ¨rster, E.; Bier, F. F.; Makower, A.; Halsall, H. B.; Heineman, W. R.; Scheller, F. W. Anal. Chem. 1996, 68, 2453-8.
Analytical Chemistry, Vol. 70, No. 21, November 1, 1998
4625
EXPERIMENTAL SECTION Chemicals. Mushroom tyrosinase (monophenol monooxygenase, EC 1.14.18.1), cocaine hydrochloride, 2-(N-morpholino)ethanesulfonic acid (MES), and tris(hydroxymethyl)aminomethane (Tris) were bought from Sigma (Deisenhofen, Germany). Bovine serum albumin (BSA), alkaline phosphatase (EC 3.1.3.1., EIA grade), maleimidohexanoyl-N-hydroxysuccinimidester (MHS), succinimidyl-acetylthiopropionate (SATP), and POROS 50 OH were obtained from Boehringer Mannheim (Mannheim, Germany). Poly(vinyl alcohol) 05/20 (PVA) was bought from Serva (Heidelberg, Germany). Sodium chloride and magnesium dichloride came from VEB Laborchemie (Apolda, Germany). pNitrophenyl phosphate (PNPP), phenyl phosphate, glucose, and sodium dihydrogen phosphate were purchased from Merck (Darmstadt, Germany). Sephacryl S-300 HR was from Pharmacia (Uppsala, Sweden). All other chemicals were of analytical grade and used without further purification. Water was deionized by MilliQ (Millipore, Eschborn, Germany). Buffers were saturated with air. GDH (EC 1.1.99.17), PQQ, and benzoylecgonine-1,8-diamino3,4-dioxaoctane (BZE-DADOO) were a gift from Boehringer Mannheim. The polyclonal anti-cocaine antibody (pAb) was developed at Boehringer Mannheim. 5-Norbornene-2,3-dicarboximido carbonochloridate (Cl-COONB) was a donation from the Institute of Molecular Pharmacology (Berlin, Germany). Synthesis of Alkaline Phosphatase-Labeled Polyclonal Anti-BZE Antibody (pAb-ALP). Two slightly different immunoconjugates were used in this study. Both were synthesized by Boehringer Mannheim with a modification of Ishikawa and Klein.51,52 The polyclonal anti-benzoylecgonine antibody from sheep was isolated by ammonium sulfate precipitation followed
by DEAE-Sepharose chromatography. The immunoreactive fraction was isolated by an affinity purification on a BZE-Spherosil column using a fractionated acidic elution after washing with PBS. The antibody (pAb) was eluted with 1 mM HCl (Ka ) 7.4 × 108 M-1 for BK 241 B1) or with 1 M propionic acid (Ka ) 3.9 × 109 M-1 for BK 227 1). The affinity constant Ka was determined on a BIAcore using standard procedures. The antibody-alkaline phosphatase conjugates (pAb-ALP) were prepared from activated components. IgG and ALP were activated by maleimide residues (MHS) and by protected SH groups (SATP), respectively. The ALP was cross-linked with the activated antibody. The conjugate was dialyzed against 50 mM triethanolamine, 150 mM NaCl, 1 mM MgCl2, 0.1 mM ZnCl2, pH 7.6, and then fractionated by gel chromatography on Sephacryl S-300 HR in the same buffer at 4 °C. It was stored at 4 °C after addition of 10 mg/mL BSA and 3 M NaCl. The activities of the resulting pAb-ALP fractions (in Boehringer ALP units) used in this study were 55 units/mL (0.8 mL of 0.17 mg/mL BK 241 B1 corresponding to 0.56 µM IgG) and 101 units/mL (0.8 mL of 0.32 mg/mL BK 227 1 corresponding to 1.06 µM IgG). Synthesis of BZE-POROS Affinity Support. The synthesis was done as described earlier.39 The affinity support contained 7.1 µmol of BZE/mL of POROS. Preparation of Stock and Standard Solutions. Phenol standards were diluted in buffer 1 (25 mM Tris, pH 8, 1 mg/mL MgCl2, 0.1 M NaCl, 0.05% Kathon CG) on the day of use from aqueous stock solutions (1 mM or 10 µM, stored at -18 °C). The 1 µM ALP stock solution was prepared in 0.1 M Tris, pH 7.6, 1 mg/mL MgCl2, 1 mg/mL BSA and was stored at 4 °C. The ALP and the pAb-ALP were diluted shortly before measurement in buffer 1 with 0.05% Tween 20 and kept on ice during measurements for up to 15 h. Cocaine standards were prepared daily and diluted in buffer 1 containing 0.05% Tween 20 additionally. Preparation of Biosensor Membranes. Tyrosinase (8300 units/mg) and GDH (350 units/mg) were coentrapped in a PVA membrane. GDH (2.2 mg) was mixed with 21 µL of 0.5 mM PQQ in 0.1 M Mes, pH 6, and 1 mg tyrosinase was dissolved in 21 µL of PBS (0.1 M phosphate, 0.15 M NaCl, pH 6.5). Both solutions were combined with 38 µL of a 20% PVA/water solution, mixed and spread on an acrylic plate (4-mm diameter). The enzyme/ PVA solution was cross-linked for 30 min under a UV lamp (254 nm, type N-15 K, Kurt Benda, Wiesloch, Germany), removed from the plate and stored at 4 °C. Each membrane contained 19.2 units of GDH and 205 units of tyrosinase. Apparatus and Procedure. AFIA used the following procedure for cocaine measurements (Figure 2): (i) Two loops were filled with 100 µL of a selected sample (10-port selector S1) and 25 µL of reagent pAb-ALP. For injection, both loops were connected (injector I2) and injected (injector I1). (ii) The samplereagent-sample zone mixed and reacted during transport to the affinity column in buffer 1. (iii) The excess reagent pAb-ALP was separated on a cocaine-modified POROS-affinity column (AC, 8 × 2.0 mm i.d.) from cocaine complexes of pAb-ALP. (iv) The detection of ALP took place in a continuously merging flow of reagents. First, substrate (buffer 2) in an alkaline buffer (buffer 3) was added to the column effluent and passed a 250-µL open
(51) Ishikawa, E.; Imagawa, M.; Hashida, S.; Yoshitake, S.; Hamaguchi, Y.; Uemo, T. J. Immunoassay 1983, 4, 209-327.
(52) Klein, C.; Josel, H. P., Herrmann, R.; Maier, J.; Ertl, H.; Oberpriller, H.; Hilpert, R.; Binder, F.; Ritter, J. European Patent Application EP 0650045 A2, April 26, 1995.
Figure 1. Principle of AFIA.
4626 Analytical Chemistry, Vol. 70, No. 21, November 1, 1998
Figure 2. Flow system for AFIA. Abbreviations: S, 10-port selector; M, magnetic three-way valve; black triangle, peristaltic or air pump, 1 channel, I, 6-port injector; BT, bubble trap; AC, affinity column; OTR, open-tube reactor; FR, flow restrictor; T, temperature control; D, biosensor flow cell; P, potentiostat; R, chart or disk recorder. For details, see Experimental Section. Table 1. Buffers and Flow Rates Used in AFIA buffer 1 2 3 4 air
buffer composition
flow rate
25 mM Tris, pH 8, 1 mg/mL MgCl2, 0.1 M NaCl, 0.05% Kathon CG 1 mM phenyl phosphate, 1 mg/mL BSA in water 1 M Tris 0.5 M phosphate, pH 6, 20 mM glucose, 0.05% Kathon CG filtered with 0.02-µm filter
600 µL/min 150 µL/min 150 µL/min 900 µL/min 20 mL/min
tube reactor (OTR). Then, the phenol-containing solution was neutralized (buffer 4) and sprayed (air) through a temperature control (T) into the detector (D). The electrode current is amplified by a potentiostat (P, - 600 mV vs Ag/AgCl, Glukometer, Academy of Sciences, GDR) and recorded (R, chart recorder, type BD 112, Kipp & Zonen, or disk recorder, L2200, Linseis, Selb). Buffers and flow rates are described in Table 1. Only the injection procedure (i) was controlled automatically by a Trace 2000 FIA (Trace GmbH, Braunschweig, Germany). The two peristaltic pumps (Minipuls 3, Abimed-Gilson, Langenfeld, Germany) ran continuously. The pump for the autosampler could be switched between “on” (drawing reagents) and “off” (circulating water) by means of the two magnetic three-way valves (M1, M2). AFIA contained some precautions against air bubbles: two bubble traps (BT), that wasted bubbles with ∼10% of the solution, a flow restrictor (FR), and the stabilized measurement of oxygen by temperature-controlled spraying. The temperature control used was a 50-cm steel tube (0.8 mm i.d.) sandwiched between two aluminum blocks of 1 cm thickness. Comparison of Optical and Amperometric Detection of Cocaine. The detection part of AFIA was substituted by a fraction
collector for this experiment (Jaytee 5512 fraction collector, Jaytee Biosciences, Herne Bay, U.K.). The pAb-ALP peaks (600 µL) were collected for 75 s after injection in vessels that contained 50 mL of buffer 1 with 2 mg/mL BSA. The fractions collected were measured photometrically after overnight incubation (405 vs 640 nm, microtiterplate reader 340 ATTC, SLT, Crailsheim, Germany) mixing 25 µL of the fraction with 100 µL of 0.1 M Tris, pH 9, 1 mg/mL MgCl2 and 1.25 mg/mL PNPP. For the amperometric measurement, 100 µL of the fractions were injected manually (V7 injector, Pharmacia, Uppsala, Sweden) in the separated detection part of the flow system. To compensate for dilution and deactivation of ALP, all flow rates were set to one-third of those used in AFIA. Preparation and Measurement of Real Samples. Real samples were provided by Securetec GmbH (Ottobrunn, Germany) and pretreated in our laboratory according to their procedure. The samples were collected by a Drugdec I apparatus, a hand-held air cleaner (Black & Decker VP 321) (250 L/min) with Securetec filter brush. The filter membrane was removed from the filter brush and incubated at room temperature for 15 min with 1 mL of buffer 1. Then, the membrane was squeezed a few times with a pipet tip and a second milliliter of the same buffer was added. The filter was mixed for 1 min on a vortex mixer, incubated for another 45 min, and mixed again (1 min). The suspension was centrifuged for 3 min (Microfuge E, Beckman, Glenrothes, U.K.). Dilutions of the supernatant were injected into the automated flow system. Securetec provided one set of samples that were not spiked with cocaine: (1) unused filter brush as control, (2-4) trunk, car, and engine compartment, (7) laboratory floor, 30-60-s collection time, (8) surface of a suitcase, collected for 15 s, (9) same as (8), but the suitcase was touched with cocaine-contaminated hands. For sample (5) and (6) the filter was used directly, the armpit of a test person was wiped three times with a wet filter (5) and another filter was wetted in the mouth of a test person (6). For a set of quantitative cocaine measurements, samples of dust from a 1-m section of a copper-covered wall close to a main road were collected for 15 s with Drugdec I: (1) unused filter brush as control, (2) cocaine-free dust sample, and (3-7) dust samples spiked with an aqueous cocaine solution and dried. RESULTS AND DISCUSSION Measurement of Phenol and Alkaline Phosphatase. The system performance regarding the intermediate analytes phenol and alkaline phosphatase was studied first. Phenol could be measured linearly from 31 nM to 3 µM (R ) 0.9995, data not shown). Alkaline phosphatase could be measured between 0.98 and 100 pM after 25-s substrate incubation in an open-tube reactor. (R ) 0.9999, data not shown.) The lower detection limits (>3 standard deviations, SD) were similar to detection limits published earlier for the same sensor principle,48,50 if the different dilution and incubation time is taken into account. The measurement of alkaline phosphatase was free of blank signals, because the phenol formed by spontaneous hydrolysis was continuously measured at the detector. So the detector response indicated only the additional phenol generated enzymatically. The alkaline phosphatase label could be measured sensitively enough that a stop flow incubation was not necessary to read the results of AFIA. Analytical Chemistry, Vol. 70, No. 21, November 1, 1998
4627
Figure 3. Dependence of signal on the time for immunorecognition. A total of 1 nM cocaine and pAb-AlP (BK 241, 1:2000) were incubated. A 1-nA blank was subtracted. For details, see Experimental Section.
Baseline fluctuations and drifting encountered with an earlier flow system50 were eliminated in this flow system. The bubbles that developed mainly at the peristaltic pumps were reduced by a flow restrictor and removed by two bubble traps. More important, the flow was mixed with a 100-fold excess of air over buffer and sprayed through a temperature control immediately before the measurement of oxygen (modified after Riedel53). Spraying stabilized the oxygen measurement for automatization. Spraying, i.e., diluting the analyte by air, does not decrease the signal height severely (85% for spray-FIA) nor influence the speed of the signal. The signals indicated only a 6-fold dilution of sample in the flow system, when an equilibrium signal was compared to the transient response for 100 µL of phenol. Measurement of Cocaine. The development of signal with time for immunorecognition was studied first (Figure 3). For this experiment, the autosampler contained a mixing tee to fill equal volumes of both sample and reagent pAb-ALP into one 100-µL loop. The time between stop and injection of the mixture of sample and reagent was taken as the time of immunorecognition (Figure 1). The signal of AFIA increased with time. At the start of the time scale, a small signal had already developed, because the reagents mixed ∼20 s before the injection loop was filled (time zero). The reaction reaches equilibrium after 2 min. The immunorecognition time was set to 30 s in the following experiments (80% of the equilibrium signal). Cocaine could be quantified between 0.38 and 3.2 nM (Figure 4). The lower detection limit of 0.38 nM was calculated from the nonlinear fit (>3 SD, 4-fold injections). The cocaine calibration curve of AFIA is a typical IEMA calibration curve. The signal increased with increasing cocaine concentration, and the increase was sigmoidal as expected for bivalent binders such as pAb-ALP. One would expect a linear dose-response curve using monovalent Fab fragments.54 The small dynamic range of the assay is due to the highly diluted pAb-ALP (1.1 nM IgG BK 227 1). Two identical blanks were measured before and after the calibration curve (Figure 4). The affinity column retained 60% of the labeled polyclonal antiserum (BK 227 1) and 50% of BK 241 B1 (data not shown). Changes in the flow rate did not influence the binding of antibodies,39 indicating that the background signal was caused either by inactive antibodies or by free enzyme label and not by slow association kinetics. This is in good agreement with observations by Irth et al.40 (53) Riedel, K.; Neumann, B.; Scheller, F. Chem. Ing. Tech. 1992, 64, 518-28. (54) Freytag J. W., Lau, H. P., Wadsley, J. J. Clin. Chem. 1984, 30, 1494-8.
4628 Analytical Chemistry, Vol. 70, No. 21, November 1, 1998
Figure 4. Automatic cocaine calibration curve of AFIA. A total of 0.38 nM cocaine could be detected after 75 s. The next sample could be introduced after 135 s. A 3.8-nA blank were subtracted.
The signal response time of AFIA was 75 s after injection of cocaine. This includes calculated contact times of 30 s for the immunorecognition, 5 s for the affinity separation, 2 s for the substrate reaction (25 µL open tube reactor here) and 38 s for the transport to the detector and the development of its maximal signal. The next sample was injected after 135 s. The sampling rate (26 assays/h) was limited by the response time of the phenol biosensor membrane. The fast response time of the immunoassay was made possible (i) by a high-affinity antiserum that allowed binding of the sample in less than 2 min, (ii) by a high-speed and high-capacity perfusion chromatography for affinity separation of label and analyte-label-complex, (iii) by an on-line detector for the enzyme label that responded after a few seconds of substrate incubation time, and (iv) by an injection technique that used transport time for immunorecognition and mixed sample and reagents before dilution by transport. The performance of AFIA can be compared to a manually operated cocaine immunoassay with off-line detection based on the same reagents.39 A preincubated mixture of cocaine and pAbALP was injected there, and the detection limit of 0.5 nM cocaine in this mixture was similar to that found here. The response time to cocaine was 1 h there39 and decreased to 75 s for AFIA. This 50-fold acceleration was achieved by avoiding dilution by on-line detection and integrating the multistep manual protocol into one automated procedure with reduced times for immunorecognition and label detection. Comparison of Optical and Amperometric Detection. The optical detection of the enzyme label by p-nitrophenyl phosphate and the amperometric by the detection part of AFIA have been compared for a dose response of cocaine (Figure 5). Both detection principles of alkaline phosphatase gave superimposable results, when the loss of enzyme conjugate by adsorption was minimized. ALP had to be incubated overnight for the optical detection but could be measured amperometrically after 50 s. The cocaine detection limit in this experiment is higher than that in Figure 4 because of a different injection technique. The dose response for AFIA was independent of the detection principle, indicating a limitation by the affinity of the antibody (apparent affinity to benzoylecgonine, 3.9 × 109 L mol-1). This is in agreement with the observations that dilutions of pAb-ALP between 1:1000 and 1:50000 did not influence the cocaine detection limit. Only the ratio between signal and blank decreased with dilution (data not shown).
Table 2. Non-Cocaine-Spiked Samples of Different Origin Injected in Various Dilutionsa dilution sample
Figure 5. Comparison of optical and amperometrical detection of enzyme label alkaline phosphatase in a cocaine immunoassay. The results and nonlinear fit functions of two biosensors (X over open circles, straight lines, phenyl phosphate) and an optical reference (solid diamonds, dashed line, p-nitrophenyl phosphate) were normalized to baseline. The cocaine immunoassay was measured using conjugate pAb-ALP (BK 227 1, 1:1000).
Figure 6. Comparison of eight cocaine calibration curves used for calibration of AFIA over 6 h. All measurements of one calibration curve have been set to the time of injection of the first blank. All data result from single injections of sample. For details see text.
Stability and Precision of AFIA in Multiple Measurements. Figure 6 shows eight calibrations of AFIA over 6 h. The data result from single injections of three cocaine concentrations between 1 and 10 nM. The calibrations started (solid circles) and concluded (open circles) with a control of the blank. These calibrations were the basis for the quantitative measurement of real samples described below. Most articles about immunosensors only provide a first calibration curve of a freshly prepared sensor. But for the measurement of real samples, stability and precision of the subsequent calibration curves are even more essential. The signal for repetitive cocaine measurements decreased 6.2 ( 0.2%/h for all cocaine concentrations. The same detector could be used for continuous measurements over 2 days. The decrease in signal for the label ALP, however, does not change the detection limit for cocaine in Figure 6. AFIA could perform more than 200 consecutive cocaine assays without regeneration of the affinity column. Regeneration was not necessary, because the 50-µL affinity column contains a 13 millionfold molar excess of BZE over the amount of pAb-ALP injected per cocaine assay. This is in good agreement with earlier findings that a column could be used for more than 2000 injections over 6 months without regeneration.39 In addition, the affinity support showed excellent stability due to the urethane bond between BZE and POROS. No leakage was observed, and the column material was still active after an 18-month storage in buffer.
unused filter (1) trunk (2) car (3) engine compartment (4) sweat (5) sputum (6) laboratory control (7) suitcase cocaine-free (8) suitcase cocainecontaminated (9) cocaine calibrations
1:10 105 103 97 100 170 170 247 256
1:100 103 97 109 97 100 106 146 128
1:1000 105 94 94 103 100 106 111 97 209
1:10 000 1:100 000
111
106
0 nM 1 nM 3.2 nM 10 nM 100 131 193 247
a The results are indicated as a percentage of the blank (100%). The means of the cocaine calibration curves measured between the samples are given at the bottom of the table.
The precision of AFIA was high (Figure 6). The standard deviations of linear regressions for each cocaine concentration were between 3.4 and 1.6% (n ) 8). The key to this precision was the two bubble traps and the injection technique. The bubble traps avoided asymmetric dispersion of peaks that would cause up to 5% deviation in the injected concentration of sample or reagent. The injection technique eliminated the consequences of imprecise timing by the windows-based software that controls the valves. In automated FIA, it can be problematic to recognize malfunction. The AFIA is constructed in such a way that the imprecise positioning of the valves (selector, injectors) can be easily recognized by characteristic baseline deviations. In addition, the hardware-related loss of sample has been minimized to 1-5% of the assays by minimizing the number of valves and the number of commands per assay. AFIA is 40 times more sensitive than the displacement flow immunosensor for cocaine.25,26 Both methods detect cocaine in similar times, 60 s compared to 75 s. But the lifetime of the displacement column is limited to 32 h or 5-15 cocaine injections.26 AFIA is the most sensitive and one of the fastest immunoassays for cocaine. Measurement of Real Samples. Two groups of samples were examined: samples of different origin that were not spiked with cocaine (Table 2) and cocaine-spiked samples of street dust (Table 3). All samples were measured in the following automatic scheme. It started with a cocaine calibration curve (0, 1, 3.2, 10, and 0 nM, single injections, Figure 6). Then, a row of three 10fold dilutions of the samples was injected two or three times with a control of the buffer blank in between. This was repeated for each sample and ended with a final cocaine calibration curve. For non-cocaine-spiked samples (Table 2), the blank has been set to 100% and the results were expressed as percentage of signal to the mean of adjacent blanks. Nondisturbed blank values from matrix samples were 101 ( 4.5% (n ) 18, data from Table 2). The control (1) and the samples that originate from a car (2-4) did not deviate from buffer blank. The cocaine assay would be undisturbed when these samples were injected as a 10-fold dilution. Sweat and sputum (5, 6) allow undisturbed cocaine measurements when injected as a 100-fold dilution. The positive Analytical Chemistry, Vol. 70, No. 21, November 1, 1998
4629
Table 3. Cocaine-Spiked Samples in a Matrix of Street Dust Collected by a Drugdec Ia
1 2 3 4 5 6 7
result (nM)
dilution factor
measd amt
spiked amt
recovery (%)
0 0 0 0.74 ( 0.05 0.7 ( 0.06 0.78 ( 0.07 0.68 ( 0.09
10 10 10 10 100 1000 1000000
