Automated Microarray System for the Simultaneous Detection of

automated analysis of 10 antibiotics in milk is presented, using multianalyte immunoassays with an indirect com- petitive ELISA format. Microscope gla...
54 downloads 0 Views 362KB Size
Download PDF
Anal. Chem. 2004, 76, 646-654

Automated Microarray System for the Simultaneous Detection of Antibiotics in Milk Bertram G. Knecht,† Angelika Strasser,‡ Richard Dietrich,‡ Erwin Ma 1 rtlbauer,‡ † ,† Reinhard Niessner, and Michael G. Weller*

Institute of Hydrochemistry, Technische Universita¨t Mu¨nchen, Marchioninistrasse 17, D-81377 Mu¨nchen, Germany, and Institute for Hygiene and Technology of Food of Animal Origin, Ludwig-Maximilians-Universita¨t Mu¨nchen, Scho¨nleutnerstrasse 8, D-85764 Oberschleissheim, Germany

A parallel affinity sensor array (PASA) for the rapid automated analysis of 10 antibiotics in milk is presented, using multianalyte immunoassays with an indirect competitive ELISA format. Microscope glass slides modified with (3-glycidyloxypropyl)trimethoxysilane were used for the preparation of hapten microarrays. Protein conjugates of the haptens were immobilized as spots on disposable chips, which were processed in a flow cell. Monoclonal antibodies against penicillin G, cloxacillin, cephapirin, sulfadiazine, sulfamethazine, streptomycin, gentamicin, neomycin, erythromycin, and tylosin allowed the simultaneous detection of the respective analytes. Antibody binding was detected by a second antibody labeled with horseradish peroxidase generating enhanced chemiluminescence, which was recorded with a sensitive CCD camera. All liquid handling and sample processing was fully automated, and one analysis was carried out in milk within less than 5 min. The detection limits ranged from 0.12 (cephapirin) to 32 µg/L (neomycin). Penicillin G could be detected at the maximum residue limit (MRL); the detection limits for all other analytes were far below the respective MRLs. The PASA system proved to be the first immunochemical biosensor platform having the potential to test for numerous antibiotics in parallel, such being of considerable interest for the control of milk in the dairy industry. In the treatment of bovine mastitis, antibiotics are widely used and improper application can lead to the contamination of milk at the farm level. For the consumer, antimicrobial substances in food hold the risk of undesirable health effects;1-3 furthermore, a widespread incidence of antibiotics can increase the occurrence of bacterial resistance against these substances.4 In the dairy industry, antimicrobial residues may inhibit starter cultures for * Corresponding author. E-mail: [email protected]. † Technische Universita ¨t Mu ¨ nchen. ‡ Ludwig-Maximilians-Universita ¨t Mu ¨ nchen. (1) Paige, J. C.; Tollefson, L.; Miller, M. Vet. Hum. Toxicol. 1997, 39, 162169. (2) Werner, L. L.; Bright, J. M. J. Am. Anim. Hosp. Assoc. 1983, 19, 783-790. (3) Peters, J. H.; Gordon, G. R.; Lin, E.; Green, C. E.; Tyson, C. A. Anticancer Res. 1990, 10, 225-229. (4) Brady, M. S.; White, N.; Katz, S. E. J. Food Prot. 1993, 56, 229-233.

646 Analytical Chemistry, Vol. 76, No. 3, February 1, 2004

cheese and yoghurt production.5 These reasons make it important to effectively control antibiotic residues in milk and therefore, regulatory authorities have enacted maximum residue limits (MRLs) for a number of antiinfective agents in milk. Within the European Union, MRLs are laid down in the commission regulation No. 2377/90.6 Other countries have similar regulations.7,8 For the detection of antibiotic residues in milk, microbiological, chromatographic, and immunochemical methods as well as receptor- and enzyme-based tests have been established.9,10 Microbiological tests are commonly applied in dairies and in survey studies. However, these tests are time-consuming (2-3 h), lack sensitivity for diverse groups of antibiotics, and do not allow substance identification. Receptor- and enzyme-based assays permit fast operation and are therefore used at the dairy plant level. These tests are usually easy to handle but allow the detection of only one group of antibiotics, in most cases the β-lactam antibiotics. Chromatographic methods are expensive and thus restricted to confirmatory purposes.8-10 Immunochemical methods such as ELISAs operated in microtitration plates (MTP) are also time-consuming, but a previously described test enables the simultaneous semiquantitative detection of different antimicrobials.11 Since the use of Biacore’s surface plasmon resonance technology (SPR) for the detection of sulfamethazine in milk12 was published in 1995, this method has been applied for the detection of a number of other antibiotics, too.13-19 Recently the use of four different antibodies for the simultaneous detection of (5) Suhren, G. Kieler Milchwirtsch. Forschber. 1996, 48, 131-149. (6) Commission Regulation (EEC) No. 2377/90, Off. J. Eur. Communities 1990, L224, 1-8. (7) MacNeil, J. D.; Ellis, R. L. In Chemical Analysis for Antibiotics Used in Agriculture; Oka, H., Nakazawa, H., Harada, K.-I., MacNeil, J. D., Eds.; AOAC International: Arlington, VA, 1995; pp 2-13. (8) Di Corcia, A.; Nazzari, M. J. Chromatogr., A 2002, 974, 53-89. (9) Aerts, M. M. L.; Hogenboom, A. C.; Brinkman, U. A. T. J. Chromatogr., B 1995, 667, 1-40. (10) Boison, J. O.; MacNeil, J. D. In Chemical Analysis for Antibiotics Used in Agriculture; Oka, H., Nakazawa, H., Harada, K.-I., MacNeil, J. D., Eds.; AOAC International: Arlington, VA, 1995; pp 77-119. (11) Strasser, A.; Dietrich, R.; Usleber, E.; Ma¨rtlbauer, E. Anal. Chim. Acta 2003, 495, 11-19. (12) Sternesjo, A.; Mellgren, C.; Bjorck, L. Anal. Biochem. 1995, 226, 175181. (13) Mellgren, C.; Sternesjo, A. J. AOAC Int. 1998, 81, 394-397. (14) Gaudin, V.; Maris, P. Food Agric. Immunol. 2001, 13, 77-86. (15) Gaudin, V.; Fontaine, J.; Maris, P. Anal. Chim. Acta 2001, 436, 191-198. (16) Haasnoot, W.; Verheijen, R. Food Agric. Immunol. 2001, 13, 131-134. (17) Haasnoot, W.; Loomans, E.; Cazemier, G.; Dietrich, R.; Verheijen, R.; Bergwerff, A. A.; Stephany, R. W. Food Agric. Immunol. 2002, 14, 15-27. 10.1021/ac035028i CCC: $27.50

© 2004 American Chemical Society Published on Web 12/23/2003

aminoglycosides with a Biacore 3000 system has been described.20 Today, kits for the Biacore sensor to test for sulfamethazine, sulfadiazine, streptomycin, and chloramphenicol are available. Besides SPR, a number of other biosensors for the rapid detection of antibiotics have been described.21-25 The system Parallux, commercially available from Idexx Laboratories (Westbrook, MA), is a rapid test working with competitive solid-phase fluorescence immunoassays in four channels, making it possible to use four antibodies simultaneously within one test.26,27 Although most of these sensors offer sufficient performance, most of them allow only the use of one antibody or receptor protein. Two systems enable the use of four antibodies in parallel, but the number of analytes still remains very limited. Despite the multitude of tests described, no method is available allowing a rapid, simultaneous detection of all relevant classes of antimicrobials, even though this would be of great interest for the dairies and the milk industry. For other applications, multianalyte systems allowing a higher degree of parallelization have already been developed. Especially microarray technologies enable the integration of a large number of different antibodies within one test. Biosensors using planar waveguide technology in combination with fluorescence detection have been described for the analysis of biohazardous agents28,29 and water pollutants.30 An alternative approach is a bead-based assay in combination with fluorescence detecting flow cytometers as offered by Luminex Corp. (Austin, TX), BD Biosciences (San Jose, CA), Bio-Rad Laboratories (Hercules, CA), and Qiagen GmbH (Hilden, Germany). Microsphere-based immunoassays have demonstrated their ability to simultaneously detect up to 16 analytes.31-33 The parallel affinity sensor array (PASA) working with automated chip-based ELISAs and a chemiluminescence reaction monitored by a charge-coupled device (CCD) camera as detection originally was developed in our laboratory for the simultaneous detection of contaminants in drinking water34,35 and has also demonstrated its applicability in the field of allergy (18) Ferguson, J. P.; Baxter, G. A.; McEvoy, J. D. G.; Stead, S.; Rawlings, E.; Sharman, M. Analyst 2002, 127, 951-956. (19) Gustavsson, E.; Bjurling, P.; Sternesjo, A. Anal. Chim. Acta 2002, 468, 153159. (20) Haasnoot, W.; Cazemier, G.; Koets, M.; van Amerongen, A. Anal. Chim. Acta 2003, 488, 53-60. (21) Setford, S. J.; Van Es, R. M.; Blankwater, Y. J.; Kroger, S. Anal. Chim. Acta 1999, 398, 13-22. (22) Delwiche, M.; Cox, E.; Goddeeris, B.; Van Dorpe, C.; De Baerdemaeker, J.; Decuypere, E.; Sansen, W. Trans. ASAE 2000, 43, 153-159. (23) Zhi, Z. L.; Meyer, U. J.; Van den Bedem, J. W.; Meusel, M. Anal. Chim. Acta, 2001, 442, 207-219. (24) Agui, L.; Guzman, A.; Yanez-Sedeno, P.; Pingarron, J. M. Anal. Chim. Acta, 2002, 461, 65-73. (25) Khaldeeva, E. V.; Medyantseva, E. P.; Imanaeva, N. A.; Budnikov, G. K. J. Anal. Chem. 2002, 57, 1097-1102. (26) IDEXX Laboratories, http://www.idexx.com/Dairy/Products/Parallux, 2003. (27) Huth, S. P.; Warholic, P. S.; Devou, J. M.; Chaney, L. K.; Clark, G. H. J. AOAC Int. 2002, 85, 355-364. (28) Rowe-Taitt, C. A.; Hazzard, J. W.; Hoffman, K. E.; Cras, J. J.; Golden, J. P.; Ligler, F. S. Biosens. Bioelectron. 2000, 15, 579-589. (29) Taitt, C. R.; Anderson, G. P.; Lingerfelt, B. M.; Feldstein, M. J.; Ligler, F. S. Anal. Chem. 2002, 74, 6114-6120. (30) Barzen, C.; Brecht, A.; Gauglitz, G. Biosens. Bioelectron. 2002, 17, 289295. (31) Fulton, R. J.; McDade, R. L.; Smith, P. L.; Kienker, L. J.; Kettman, J. R. Clin. Chem. 1997, 43, 1749-1756. (32) Carson, R. T.; Vignali, D. A. A. J. Immunol. Methods 1999, 227, 41-52. (33) Park, M. K.; Briles, D. E.; Nahm, M. H. Clin. Diagn. Lab. Immunol. 2000, 7, 486-489. (34) Winklmair, M.; Schuetz, A. J.; Weller, M. G.; Niessner, R. Fresenius J. Anal. Chem. 1999, 363, 731-737.

diagnosis.36,37 Randox system Evidence (Randox Laboratories Ltd., Crumlin, U.K.) uses a setup in principle similar to the PASA, but it is designed for multianalyte detection with high sample throughput. After a broad spectrum of polyclonal and monoclonal antibodies against antibiotics had been developed during the last years,11,38,39 which has been extended by commercially available antibodies, an improved PASA system for the simultaneous detection of antibiotics was designed. To allow the immobilization of haptens on the chip instead of antibodies, an indirect competitive immunoassay format was chosen. This format was expected to result in chips with increased durability. With regard to long-term availability, only monoclonal antibodies were integrated. The microarray is conceived as a univariate system where each antibody selectively binds to one hapten. In contrast, multivariate microarrays use overlapping cross-reactivities of several haptens to different antibodies, which allow the identification of crossreacting analytes by their signal pattern.34 Univariate systems, however, permit easier assay optimization for each antibody, and the calibration and evaluation is less complicated. Furthermore, the assays are more suitable for the screening of a broad spectrum of compounds with a limited number of antibodies.40 EXPERIMENTAL SECTION General Setup. A scheme of the PASA system, which was developed in collaboration with Atto-tec GmbH (Siegen, Germany), is depicted in Figure 1. The flow cell, in which the disposable chips are inserted, is placed in a dark chamber at the top of the instrument. A magnified view of the flow cell with a volume of ∼100 µL is shown on the left-hand side of Figure 1. The signal generation by enhanced chemiluminescence combined with CCD camera detection offers a highly sensitive detection in a robust construction without the need of an expensive excitation source or additional optics. The lens (Linos MeVis-C 35/1.6, LINOS Photonics GmbH, Go¨ttingen, Germany) and the CCD detector (MX 916, Starlight Xpress Ltd., Holyport, U.K.) are situated beneath the flow cell inside the case. The CCD chip (Sony ICX083AL SuperHAD) is working in a 2 × 2 binned mode and offers 376 × 290 superpixels with 23.2 × 22.4 µm size and a quantum efficiency of 65% at 520 nm. The CCD chip is thermoelectrically cooled to 30 ° C below ambient temperature to lower the thermal background. The image scale β is ∼2.9. The fluid handling is accomplished by eight 1-mL syringe pumps (Cavro XE 1000, Tecan Systems, San Jose, CA), which are arranged at the front side of the case. Two of the pumps manage the rinsing of the flow cell and allow a nearly permanent flow by an alternating action. The pumps, the flow cell, and the waste bottle are connected with 1/16 in. (∼1.56 mm) o.d. and 0.02-0.04 (35) Weller, M. G.; Schuetz, A. J.; Winklmair, M.; Niessner, R. Anal. Chim. Acta 1999, 393, 29-41. (36) Fall, B. I.; Niessner, R.; Schedl, M.; Weller, M. G. in Proc. 2. BioSensor Symposium, Tu ¨ bingen. 2001, http://w210.ub.uni-tuebingen.de/dbt/volltexte/ 2001/376. (37) Fall, B. I.; Eberlein-Konig, B.; Behrendt, H.; Niessner, R.; Ring, J.; Weller, M. G. Anal. Chem. 2003, 75, 556-562. (38) Ma¨rtlbauer, E.; Usleber, E.; Schneider, E.; Dietrich, R. Analyst 1994, 119, 2543-2548. (39) Dietrich, R.; Usleber, E.; Ma¨rtlbauer, E. Analyst 1998, 123, 2749-2754. (40) Weller, M. G. Anal. Bioanal. Chem. 2003, 375, 15-17. (41) Haasnoot, W.; Cazemier, G.; Du Pre, J.; Kemmers-Voncken, A.; BienenmannPloum, M.; Verheijen, R. Food Agric. Immunol. 2000, 12, 15-30.

Analytical Chemistry, Vol. 76, No. 3, February 1, 2004

647

Figure 1. Scheme of the PASA system. The flow cell is at the top of the case, the lens and the CCD camera are inside, and the pumps are at the front. An enlarged scheme of the flow cell is depicted on the left-hand side. The camera and flow cell are shown out of the case for better visibility.

in. (0.5-1.0 mm) i.d. PTFE tubing. The sample loop for the preincubation step has a volume of 1.1 mL. A PC program allows the control and flexible automation of the system with customized test protocols. The components were chosen with regard to obtaining a high performance at a moderate price. Reagents and Chemicals. Cloxacillin monoclonal antibody (mAb) clone 1F7, sulfadiazine mAb 2G6, sulfamethazine mAb 4D9, and streptomycin mAb 4E2 were prepared as described previously.38,39 Penicillin G mAb 27-4C6-B8 was purchased from Maine Biotechnology Services (Portland, ME). Cephapirin mAb CH 2025, gentamicin mAb CH 2032, neomycin mAb CH 2021, erythromycin mAb CH 2012, and tylosin mAb CH 2023 were bought from Silver Lake Research (Monrovia, CA). TNT mAb A/1.1.1, which was used for the reference spots, was received from Strategic Diagnostics (Newark, DE). HRP-labeled horse anti-mouse antibody (PI-2000; lot N0904) was purchased from Vector (Burlingame, CA). For the optimization of the tests and the determination of calibration curves, all antibody solutions were prepared with phosphate-buffered saline (PBS; 80 mM potassium phosphate buffer with 145 mM NaCl, pH 7.6) containing 0.5% casein. For the quantification of samples, the antibodies were diluted with fetal calf serum/H2O 1:1 (Sigma-Aldrich, Taufkirchen, Germany) for stabilization. Pierce SuperSignal ELISA Femto Maximum sensitivity substrate (Pierce, Rockford, IL) served as enhanced chemiluminescence substrate. The washing buffer was a PBS buffer containing 0.05% Tween 20 and 0.25% casein. Penicillin G sodium salt, cloxacillin sodium salt monohydrate, cephapirin sodium salt, sulfadiazine, sulfamethazine, streptomycin sulfate, gentamicin sulfate, carboxymethoxylamine, glucose oxidase (GOx), and ovalbumin (OVA) were purchased from Sigma-Aldrich. Neomycin B trisulfate sesquihydrate (Vetranal), erythromycin A dihydrate (Vetranal), and tylosin hemitartrate dihydrate (Vetranal) were from Riedel-de Hae¨n (Seelze, Germany). N-(3-Dimethylaminopropyl)-N′-ethylcarbodiimide hydrochloride (EDC), N-hydroxysuccinimide (NHS), and N,N′-dicyclohexylcarbodiimide were purchased from Fluka (Buchs, Switzerland). 2,4,6-Trinitrobenze648

Analytical Chemistry, Vol. 76, No. 3, February 1, 2004

nesulfonic acid was from Ferak (Berlin, Germany). The hapten BSA conjugates for cephapirin, gentamicin, neomycin, erythromycin, and tylosin were delivered by Silverlake Research together with the antibodies. Organic solvents were purchased from VWR (Darmstadt, Germany). Stock solutions (1 g/L), standard solutions, and spiked samples were prepared daily in pasteurized market milk (fat content 3.5%) and 1 M NaOH for the stock solutions of the sulfonamides, respectively. The molecular structures of the 10 analytes are depicted in Figure 2. Hapten Conjugate Synthesis. Synthesis of the Penicillin G Ovalbumin Conjugate. The penicillin ovalbumin conjugate was synthesized using a solution of 20 mg of penicillin G sodium salt, 6.45 mg of NHS, and 26.7 mg of EDC hydrochloride in 1 mL of PBS/water 1:10. After the mixture was stirred in an ice bath for 25 min, a solution of 10 mg of OVA in 1 mL of PBS/water 1:10 was added. The temperature of the solution was slowly increased until it reached room temperature and stirred for another 3 h. Excess reagents of conjugation were removed with gel chromatography (PD-10 column, Pharmacia Amersham, Uppsala, Sweden). Synthesis of the Trinitrophenyl Ovalbumin Conjugate. A trinitrophenyl conjugate was used for the reference spots. A 75-mg sample of OVA was dissolved in 8.5 mL of 7 mM sodium carbonate buffer (pH 9.6) and cooled in an ice bath. Six portions of 20 µL of a solution of 1 M 2,4,6-trinitrobenzenesulfonic acid in water were added under stirring. Resulting yellow precipitate was removed by centrifugation and the supernatant was dialyzed against PBS/ water (1:5). Synthesis of the Sulfadiazine Glucose Oxidase Conjugate. Sulfadiazine was first derivatized with succinic anhydride according to an earlier description.41 A mixture of 1 g of sulfadiazine and 0.5 g of succinic anhydride was refluxed for 1.5 h in anhydrous ethanol containing 10 µL of anhydrous pyridine. The mixture was then evaporated to dryness and refluxed in 10.5 mL of a mixture (42) Schnappinger, P.; Usleber, E.; Ma¨rtlbauer, E.; Terplan, G. Food Agric. Immunol. 1993, 5, 67-73.

Figure 2. Structures of the 10 analytes determined in this study.

of ethanol and water (4:3, v/v) for 30 min. After evaporation to dryness, the residue was dissolved again in 10.5 mL of ethanol/ water (4:3) and filtered through a glass filter. The filtrate was allowed to crystallize overnight at -18 °C. The crystals were collected by filtration and washed with a cold (-10 to -20 °C) mixture of ethanol/water (4:3). The collected crystals were dried and stored at -18 °C until used. Coupling to GOx was performed with carbodiimide. Succinylated SDA (2 mg) was dissolved in 0.4 mL of 0.05 mol/L tris buffer/dioxane (1:1), followed by the addition of 1 mg of EDC in 0.02 mL of water. The mixture was added to 10 mg of GOx in 0.5 mL of PBS (0.5 mmol/L) and stirred for 2 h at room temperature. Finally, the mixture was dialyzed against three changes of PBS (5 L). For the synthesis of streptomycin GOx conjugate, streptomycin was derivatized with carboxymethoxylamine and the obtained streptomycin oxime was coupled to GOx as described earlier.42 Sulfamethazine GOx was prepared by derivatization of SMA with succinic anhydride and coupling with carbodiimide as recently described.11 Cloxacillin GOx was prepared by the activated ester method as previously published.11,43 All other conjugates were purchased from Silverlake Research. Chip Production and Microarraying. Conventional microscope glass slides (Superior, Marienfeld, Lauda-Ko¨nigshofen, Germany) were used as solid phase. The slides were cleaned at room temperature by applying washing steps of 30 min, using 2-propanol, boiling water, and a 1:1 mixture of methanol and hydrochloric acid (37%) in sequence. After the slides were immersed for 5 min in sulfuric acid (95-97%), they were rinsed (43) Usleber, E.; Lorber, M.; Straka, M.; Terplan, G.; Ma¨rtlbauer, E. Analyst 1994, 119, 2765-2768.

thoroughly with HPLC grade water and methanol, and finally they were dried under a stream of nitrogen. The modification of the glass surface was achieved with a solution of 1% (3-glycidyloxypropyl)trimethoxysilane (GOPS; Sigma-Aldrich, Taufkirchen, Germany) in toluene (pa) for 16 h at room temperature. After washing steps with toluene, 2-propanol, and methanol, the glass slides were dried under nitrogen and stored in a desiccator. The application of the hapten conjugates was accomplished with a noncontact arraying system (GeSiM, Grosserkmannsdorf, Germany) that consisted of a piezoelectric nanoliter pump (Standard pipet SPIP 1853), a positioning system, a controller (multi-dos), and a liquid handling system (Cavro pumps). A standard 96-well MTP served as reagent reservoir. The parameters for the piezopipet were U ) 85 V, f ) 100 Hz, and t ) 100 µs. For each spot, five droplets were dispensed, which is equal to a volume of ∼5 nL. The obtained spots had a diameter of ∼350 µm. Lots of up to 18 chips were prepared in one pass. The hapten conjugates were dissolved in PBS throughout the study. Although PBS as a solvent does not lead to a maximum binding of the conjugates, the haptens were less sensitive toward hydrolysis compared to buffers of higher pH. This is especially important for β-lactam antibiotics. The typical concentration of the conjugates in PBS for spotting was ∼50 mg/L. For some conjugates, it was helpful to add 25-50 mg/L casein to the spotting solution to reduce the donut effect (higher intensities at the edge of the spots compared to the center) and to obtain more homogeneous spots. The surface was blocked with a solution of 2% casein in PBS for 1 h at room temperature. After rinsing the chips with PBS, they were stored in PBS at 4 °C until use. Analytical Chemistry, Vol. 76, No. 3, February 1, 2004

649

Figure 3. Scheme of the flow cell and the active area of the chip. Reference spots (trinitrophenyl conjugate) and analyte conjugate spots are immobilized each with seven replicates in a row. The CCD exposures show the active area of slides. From left to right, exposures of a negative control, of sample 1 and of sample 5 (see Table 3) are depicted. Compounds contained in samples 1 and 5 are marked with a star. The concentrations in sample 5 were 16 times higher than in sample 1. The affected spots showed decreasing spot intensities with increasing concentration from left to right, while other spots remained unchanged.

Test Protocol. After the disposable chip had been inserted into the flow cell, all steps of the indirect ELISA were carried out automatically. The protocol of the procedure that was used consistently for all tests is summarized as follows: After a rinsing step (15 s), sample (800 µL) and antibody solution (400 µL) were mixed and preincubated in the sample loop (15 s). Typical concentrations of the antibodies were 90-350 µg/L. For the incubation with the chip, the first 800 µL of this mixture was rapidly pumped into the cell. To accelerate the reaction kinetics, 200 µL of the remaining mixture was slowly pumped into the cell in intervals (equivalent to a flow rate of 3.2 µL/s, total incubation time 70 s). After a rinsing step with washing buffer (30 s), the same incubation scheme was carried out with the peroxidaselabeled secondary antibody. The two components of the chemiluminescence substrate were pumped into the cell simultaneously with two pumps (400 µL each, 5 s). The signal of the light emission was integrated for 30 s. The tests were carried out with spiked pasteurized whole milk without further sample preparation. After each test, a washing program (150 s) was used to rinse the sensor and especially the sample syringe pump and sample loop with washing buffer. After samples with high concentrations of antibiotics, this washing program was executed twice. The quantification of spiked samples and the blank tests were carried out with one row of spot replicates for every hapten conjugate as shown in Figure 3. At the beginning of a test series, a calibration was carried out using five chips of one lot (18 chips). Three standard samples containing each analyte in a concentration in the range of the midpoint of the according calibration curve were processed. Together with a blank test and a standard sample with high concentrations of each analyte, it was possible to 650

Analytical Chemistry, Vol. 76, No. 3, February 1, 2004

calculate a sigmoidal calibration curve for each analyte. With the remaining 13 chips of the lot, samples could be tested. Data Evaluation. The background originating from the CCD dark signal was recorded during each series of tests and subtracted prior to the evaluation. The spot intensities were determined by the program SPOT,44 which was originally created for the analysis of DNA microarrays. The mean of the intensities of 13 pixels was calculated for each spot. Every hapten conjugate was immobilized with seven spot replicates, enabling us to calculate a 30% trimmed mean of the seven spot intensities and thus obtain statistically robust data.45 A row of reference spots was immobilized parallel to each row of analyte spots. The intensities of analyte spots were put in relation to the reference spots to improve reproducibility. The standard errors of the trimmed mean are indicated in the diagrams as error bars. The calibration curves were fitted according to the four-parameter logistic equation with Origin 6.0 (Microcal Software Inc., Northhampton, MA). For the quantification of samples, the calibration curve for the standard samples was fitted with a fixed parameter for the slope in the logistic equation, which should be ∼1 for monoclonal antibodies. Based on the assumption of a coefficient of variation of 10% at zero analyte concentration and the usual 3σ definition, the detection limit was defined as the concentration, which is equivalent to 30% inhibition (IC30). The limit of the working range at high concentrations was defined as an inhibition of 85% (IC85), where the approximately linear range of the curve fades out. (44) Jain, A. N.; Tokuyasu, T. A.; Snijders, A. M.; Segraves, R.; Albertson, D. G.; Pinkel, D. Genome Res. 2002, 12, 325-332. (45) Wilcox, R. R. Fundamentals of Modern Statistical Methods; Springer: New York, 2001; pp 141-166.

Figure 4. PASA calibration curves for 10 analytes in milk (7 spot replicates). Error bars indicate the standard error of the 30% trimmed mean.

RESULTS AND DISCUSSION Chip Surface. For the immobilization of the haptens on the chip surface, direct coupling of haptens at appropriate groups on a modified glass surface would be desirable. Experiments on the

immobilization of cloxacillin NHS ester on aminosilanized slides demonstrated that the coupling itself is possible; however, the chemical blocking of the unreacted amino groups (for example, Analytical Chemistry, Vol. 76, No. 3, February 1, 2004

651

Table 1. Midpoints, Working Range, and Maximum Residue Limits (MRL) for All Used Antibodies/ Antibiotics after Optimization

analyte

midpoint IC50 (µg/L)

working range (µg/L)

MRL (µg/L)

penicillin G cloxacillin cephapirin sulfadiazine sulfamethazine streptomycin gentamicin neomycin erythromycin tylosin

7.3 ( 0.9 0.64 ( 0.05 0.26 ( 0.02 7.7 ( 1.2 11.1 ( 1.8 11.6 ( 1.5 24.8 ( 3.0 75 ( 23 0.83 ( 0.05 2.2 ( 0.4

3.30-41.3 0.29-3.63 0.12-1.45 3.49-43.5 4.93-63.0 5.06-66.0 12.1-141 31.8-427 0.36-4.70 0.95-12.6

4 30 60 100 100 200 100 1500 40 50

Table 2. Coefficients of Variation (CV, n ) 10) for the Spot Intensities of Ten Blank Testsa

evaluation with

mean without ref CV (%)

trimmed mean without ref CV (%)

trimmed mean with ref CV (%)

penicillin G cloxacillin cephapirin sulfadiazine sulfamethazine streptomycin gentamicin neomycin erythromycin tylosin

13.4 5.0 6.2 4.0 6.0 14.7 10.5 5.9 15.0 9.3

11.9 4.7 5.7 5.7 6.5 8.5 10.3 5.8 16.1 9.5

8.0 4.0 3.4 2.3 2.0 7.1 6.5 2.0 11.7 5.6

a Comparison of three evaluation methods: mean of all spot replicates, 30% trimmed mean of all spot replicates, and 30% trimmed mean of spot replicates normalized by reference spots.

with succinic anhydride) turned out to be unsatisfactory. Another possibility is to conjugate haptens to proteins first and then immobilize the conjugates. This is generally used for coating MTPs. This method has the advantage that conjugates which have proven to work in MTPs can also be used on the chip surface, which facilitates the transfer from ELISA to the chip platform. Glass silanized with GOPS has reactive epoxy groups on the surface that can covalently bind to amino groups of the conjugate, but adsorption of the proteins at the hydrophobic surface is also likely.46 GOPS silanized slides turned out to be good substrates for this application. Test Performance. Beside a sufficient sensitivity, the major aim of the work was to obtain a rapid test. Compared to earlier assays (unpublished work), the incubation times in the flow cell could be reduced almost to half by introducing a continuous slow flow during the incubation steps and by reduction of the height of the cell from 0.50 to 0.25 mm. Thus, an assay time of 4 min, 40 s could be achieved, which is exceptionally short for an immunoanalytical technique employing enzyme-labeled reagents. The washing cycles, however, could not be accelerated as much as the incubation time and therefore set limits to a further reduction of the total cycle time. Nevertheless, a reduction of the total assay time to 3 min seems to be achievable. Another improvement would be the direct labeling of the primary antibodies, allowing omission of the second incubation step. The chips were blocked with a casein solution, which leads to a very low background (nonspecific binding) of the CCD exposures. This can be seen from the low intensities at high analyte concentrations of all calibration curves in Figure 4. Sensitivity and Calibration Curves. Table 1 lists test midpoints, working ranges, and MRLs6 for the antibiotics tested after optimization of the assays. Due to the high sensitivity and the narrow working range, the MRL concentration was outside the working range for most analytes, thus providing a clear positive readout but limiting quantification. For penicillin G, however, the detection limit was close to the MRL. As far as data for comparisons are available, the sensitivity reached with the antibodies on the chip is slightly better than in the MTP, although the assay time was reduced drastically. As known from experiments in MTPs, most of the antibodies show cross-reactivities (46) Fall, B. I. Dissertation, Technische Universita¨t Mu ¨ nchen, Munich, 2003; pp 99-102.

652 Analytical Chemistry, Vol. 76, No. 3, February 1, 2004

against related antibiotics so that additional analytes might be detectable with the shown system. For example, the streptomycin antibody will also bind to dihydrostreptomycin. Cross-reactivities can especially be expected for the β-lactam antibiotics and the sulfonamides but have not been examined in detail so far. Figure 4 illustrates characteristic calibration curves for the 10 analytes selected for this study. Reproducibility of Negative Controls. Particularly for disposable chips, the precision of the test is an important criterion. Table 2 lists the coefficients of variation (CVs) of 10 negative controls determined for all 10 antibodies. For a comparison, the results for three different evaluation methods are indicated. Using the mean of absolute intensities of seven spot replicates without correction by reference spots, as noted in the left column of Table 2, led to the highest CV for most analytes. The 30% trimmed mean deletes the two highest and the two lowest spot intensities and calculates the mean of the remaining three spot intensities. This procedure made the system insensitive to missing spots, which may result from failures during the spotting procedure or local interferences caused by bubbles in the flow cell. The evaluation including reference spots in every second row (Figure 3) improved the reproducibility, and except for erythromycin, the CVs of the negative controls were lower than 10%; for five analytes, they were even lower than 5% as indicated in the right column of Table 2. The evaluation with only one column of reference spots for all analytes (not shown) did not lead to a comparable improvement in the precision of the tests. It turned out that the closer the reference spots are located to the analyte spots, the lower the coefficients of variation of the measurements. We suspect that inhomogenities on the chip surface are the main reason for these variations. Quantification of Samples. The influence of parameters such as temperature, humidity, and variations in the dispensed volume of the piezopump made it difficult to produce chips in different lots with the same properties. Additionally, variations in the ambient temperature from one day to another also have an influence on the assays. Therefore, it was necessary to perform a calibration for every lot of chips or every series of tests. A useful calibration was achieved by probing three chips with three antigen concentrations in the range of the midpoint of the corresponding dose-response curve, one chip with a negative sample and one

Table 3. Comparison of Spiked Concentration and Test Results of Milk Samples (µg/L). Midpoints (IC50) of the Calibration Curves, Working Ranges, and Maximum Residue Limits (MRL) analyte

penicillin G

cloxacillin

cephapirin

sulfadiazine

sulfamethazine

IC50 working range

14.0 6.4-79

3.03 1.3-17

0.56 0.25-3.2

23.2 11-132

28.5 13-161

MRL sample 1 sample 2 sample 3 sample 4 sample 5 sample 6 sample 7 sample 8 sample 9 sample 10

4

30

10

100

100

spiked

found

spiked

found

spiked

found

spiked

found

spiked

found

4 8 16 32 64 0 0 0 0 0