Detection of Staphylococcus aureus enterotoxin B at femtomolar levels with a miniature integrated two-channel surface plasmon resonance (SPR) sensor

Detection of Staphylococcus aureus enterotoxin B at femtomolar levels with a miniature integrated two-channel surface plasmon resonance (SPR) sensor

Biosensors and Bioelectronics 17 (2002) 573 /584 www.elsevier.com/locate/bios Detection of Staphylococcus aureus enterotoxin B at femtomolar levels ...

639KB Sizes 0 Downloads 56 Views

Biosensors and Bioelectronics 17 (2002) 573 /584 www.elsevier.com/locate/bios

Detection of Staphylococcus aureus enterotoxin B at femtomolar levels with a miniature integrated two-channel surface plasmon resonance (SPR) sensor Alexei N. Naimushin a,c, Scott D. Soelberg a,c, Di K. Nguyen a,c, Lucinda Dunlap a,c, Dwight Bartholomew b, Jerry Elkind b, Jose Melendez b, Clement E. Furlong a,c,* a

Department of Medicine, University of Washington, Seattle, WA 98195, USA b Texas Instruments, Inc, Dallas, TX 75265, USA c Department of Genome Sciences, University of Washington, Seattle, WA 98195, USA Received 19 June 2001; received in revised form 22 November 2001; accepted 9 January 2002

Abstract Surface plasmon resonance (SPR) biosensors offer the capability for continuous real-time monitoring. The commercial instruments available have been large in size, expensive, and not amenable to field applications. We report here an SPR sensor system based on a prototype two-channel system similar to the single channel SpreetaTM devices. This system is an ideal candidate for field use. The two-channel design provides a reference channel to compensate for bulk refractive index (RI), non-specific binding and temperature variations. The SPR software includes a calibration function that normalizes the response from both channels, thus enabling accurate referencing. In addition, a temperature-controlled enclosure utilizing a thermo-electric module based on the Peltier effect provides the temperature stability necessary for accurate measurements of RI. The complete SPR sensor system can be powered by a 12V battery. Pre-functionalized, disposable, gold-coated thin glass slides provide easily renewable sensor elements for the system. Staphylococcus aureus enterotoxin B (SEB), a small protein toxin was directly detectable at sub-nanomolar levels and with amplification at femtomolar levels. A regeneration procedure for the sensor surface allowed for over 60 direct detection cycles in a 1-month period. # 2002 Elsevier Science B.V. All rights reserved. Keywords: Biosensor; Immunosensor; Surface Plasmon Resonance (SPR); Biological Warfare (BW) agent

1. Introduction Many biological and chemical agents including bacteria, algae, fungi, viruses and toxins are capable of adversely affecting humans and animals. The effects range from dizziness, vomiting and diarrhea to memory loss, paralysis, and death (Franz et al., 1997). Human or animal poisoning can result from drinking water or ingestion of foods not properly inspected or treated for the presence of bacteria or toxins. Food or water supplies can also be contaminated intentionally in terrorist acts or in deployment of biological warfare * Corresponding author. Present address: Division of Medical Genetics, University of Washington, PO Box 357720, Seattle, WA 98195-7720, USA. Tel.: 1-206-543-1193; fax: 1-206-543-0754. E-mail address: [email protected] (C.E. Furlong).

agents (BWA) during military operations. Despite the 1972 Biological Weapons Convention, some countries appear to be continuing the development and production of BWA (Christopher et al., 1997; Franz et al., 1997; Zilinskas, 1997). Terrorist attacks with BWA, such as the release of sarin in Nagano prefecture in June 1994 (Yoshida, 1994), in the Tokyo subway system in March 1995 (Suzuki et al., 1995), and the recent anthrax cases in the United States represent significant and continuing concerns (Holloway et al., 1997). To address these concerns, sensitive, easy to use and inexpensive portable biosensors capable of providing continuous monitoring and rapid detection of various BWA need to be developed. Surface Plasmon Resonance (SPR) sensors offer several advantages over other detection methods, with which they are commonly compared (Paddle, 1996). (1)

0956-5663/02/$ - see front matter # 2002 Elsevier Science B.V. All rights reserved. PII: S 0 9 5 6 - 5 6 6 3 ( 0 2 ) 0 0 0 1 4 - 3

574

A.N. Naimushin et al. / Biosensors and Bioelectronics 17 (2002) 573 /584

SPR sensors are able to provide continuous real-time monitoring. A large number of negative samples (i.e. not containing the analyte of interest) can be processed without resetting or changing the sensor. (2) Following analyte detection, the sensor can usually be reset by removing the target analyte with a low pH wash step (this work; Blanchard et al., 1990; O’Brien et al., 2000). (3) Direct detection with SPR sensors does not continuously consume reagents, unlike majority of other detection procedures, such as enzyme linked imunosorbent assays (ELISA) (Emanuel et al., 2000; Koch et al., 2000; Rowe et al., 1999; Yu et al., 1998), light addressable potentiometric sensors (LAPS) (Choi et al., 1998; Lee et al., 2000; Uithoven et al., 2000), array biosensors (Rowe et al., 1999; Rowe-Taitt et al., 2000; Wadkins et al., 1998), immunomagnetic separation electro-chemiluminescence and fluorescence procedures (IMS-ECL and IMS-FCL) (Yu, 1996, 1998; Yu et al., 1998, 2000) or rapid chromatographic assays (RCA) (O’Brien et al., 2000), all of which require the additional analytes used in sandwich assays. In addition to a direct detection mode, SPR biosensors can also work in an amplification mode to further lower the detection limit. This is done using antibodies for amplification at the cost of reagent consumption. (4) Compactness of the SPR sensors manufactured by Texas Instruments (TI), (Dallas, TX) makes them prime candidates for field use, especially when outfitted with a temperature control module and a reference channel. (5) Another advantage of SPR biosensors comes from the sequential manner in which samples are analyzed. The step-wise detection procedure allows for an estimation of the concentration of the BW agent in question. Large concentrations of BW agent that pose immediate danger to human life are detected directly in less than 15 min without amplification steps. A subsequent amplification step (or steps) allow detection of lower concentrations of Staphylococcus aureus enterotoxin B (SEB), down, to femtomolar levels. SPR sensors can be divided into two major classes, those that use wavelength interrogation and those that use angle interrogation. For a particular effective refractive index (RI) at the sensor surface, the minimum in the reflectance associated with the SPR signal depends on the angle of incidence and the wavelength of the interrogating light. SPR sensors using wavelength interrogation (Homola et al., 1999; Jorgenson and Yee, 1993) keep the angle of incidence fixed and monitor spectral changes, while angle interrogation sensors such as the BIACORE instruments manufactured by Biacore AB (Uppsala, Sweden) (Biacore, 1995) and the SpreetaTM sensor system manufactured by TI (Elkind et al., 1999; Woodbury et al., 1998) work at a fixed wavelength and employ photodetectors that allow tracking the angle of reflectance minimum.

S. aureus enterotoxin B (SEB), a small protein toxin (Easmon and Adlam, 1983; Franz et al., 1997), was selected for these studies as a typical small protein toxin. SEB, a 28.4 kDa protein toxin, is one of a group of five major serological types of related proteins with molecular weights ranging from 26 to 29.6 kDa. SEB is the most heat stable of this group and it is also resistant to the proteolytic enzymes of the gastrointestinal tract and low pH. If ingested, SEB can cause nausea, vomiting, diarrhea, and anaphylactic shock. After aerosol exposure, symptoms consist of sudden onset of fever, chills, headache, and cough. The fever may last several days, and the cough may persist for up to 4 weeks. Very high exposure levels may lead to pulmonary edema. SEB is an incapacitating toxin, but it is rarely lethal. The detection and quantification of SEB in buffer, as well as in more complex solutions, including milk, urine, and seawater was demonstrated using the miniature, integrated, temperature-controlled, portable, two-channel SPR sensor system.

2. Materials and methods

2.1. Materials Gold-coated borosilicate glass slides (15 /4 /0.2 mm, Erie Scientific, Erie, PA) were used for gold binding peptide (GBP) binding activity assays, while 23 /6/0.2 mm glass slides were used in the twochannel SPR sensor. The slides were first coated with 2 nm of chromium then 50 nm of gold. Both metal films were deposited by thermal evaporation (Varian model 3118, base pressure less than 2/107 Torr). The gold deposition rate was 0.8 /1.0 nm/s, and the chromium deposition rate was 0.1 /0.2 nm/s. SEB and rabbit anti-SEB IgG antibodies were purchased from Toxin Technology, Inc (Sarasota, FL); goat anti-rabbit antibodies were from Kirkegaard & Perry Laboratories, Inc (Gaithersburg, MD); monoclonal mouse anti-SEB antibodies were generous gifts of Dr Tom O’Brien at Tetracore, LLC (Rockville, MD); 1Ethyl-3-(3-dimethylaminopropyl) carbodiimide-HCL (EDC) was from Pierce (Rockford, IL); sulfo-Nhydroxysuccinimide (S-NHS) was from Fluka (Milwaukee, WI); ethanolamine was from Sigma (St. Louis, MO). All buffers used in the experiments described below were prepared using double glass-distilled water and were filtered through 0.22 mm filters from Millipore (Bedford, MA) before use. The following buffer solutions were used in this work:

A.N. Naimushin et al. / Biosensors and Bioelectronics 17 (2002) 573 /584

Trypsin buffer (10 mM Tris, 10 mM CaCl2, pH 8.0). TTBS buffer (10 mM Tris, 150 mM NaCl, 0.1% Tween 20, pH 8.0). PKT-50 buffer (10 mM KH2PO4, 50 mM KCl, 1% Triton X-100, pH 7.0).

2.2. Miniature integrated two-channel SPR sensor and temperature controller The two-channel sensors used in this work are prototype SPR sensors developed by TI. Their construction is very similar to the single channel SpreetaTM sensor described earlier, (Elkind et al., 1999; Melendez et al., 1997, 1996; Woodbury et al., 1998) except that the sensor is somewhat wider to allow incorporation of a second flow channel and a second detector. A schematic diagram of the two-channel sensor is shown in Fig. 1. An AlGaAs light emitting diode (LED) with a wavelength of 830 nm is enclosed within an absorbing apertured box. It illuminates the gold-coated thin glass slide over a distribution of angles after passing through a polarizer, which selects the transverse magnetic (TM) polarized component. The transverse electric component cannot produce surface plasmon oscillations, and therefore, would only contribute to the background. Optical coupling of the glass slide to the epoxy prism of the sensor is achieved with index matching liquid (#5040, Cargille, Cedar Grove, NJ). The light beam, after being reflected from the gold coating of a slide and a mirror on top of the sensor, reaches two independent linear 256pixel photodiode (PD) arrays. The optical substrate provides the protective encapsulation of all system components (Elkind et al., 1999). A flow cell consists of a 6 mm thick Teflon block with inlets and outlets for

575

each channel and a 0.4 mm thick silicone rubber gasket with two side-by-side laser-cut chambers (1 /16 mm each) that generates flow cells for the two sensor channels. Each flow cell volume is /6 ml. The Teflon block is held in place by six screws. The response of the PD array is digitized by a 12-bit analog to digital (A/D) converter, then transferred to a computer via an RS-232 interface. A monitoring and analysis program provides the user interface for displaying and analyzing the sensor system data. The software provides continuous monitoring and analysis of the SPR curve, sensor and flow cell temperatures, calculates statistical noise in the system and displays effective refractive index versus time.

2.3. Temperature control Keeping the SPR sensor at a constant temperature is very crucial for accurate detection of target analytes. Changes in temperature cause changes in RI of aqueous solutions; dRI/dT :/1 /104 per 8C (CRC, 1978). Since changes of the RI that correspond to the detection of the most dilute samples are 10 5 or even less, temperature stability to better than 0.1 8C for the duration of detection is necessary. Simple corrections for temperature changes (10 4 RIU/8C), which can be software implemented, tend to overcorrect the temperature effect. Detailed analysis of the effects of temperature on sensitivity of the detector, intensity, wavelength of the light source, sensor geometry, and RI of aqueous solutions can be found elsewhere (Chinowsky, 2000). This complicated problem is beyond the scope of this manuscript, however, it can be easily dealt with using a temperature-controlled housing. To keep the tempera-

Fig. 1. Schematic diagram of the two-channel integrated miniature SPR sensor with a detachable flow cell. On the left, the frontal view shows two channels laser-cut in the silicone rubber gasket. On the right is the side view of the sensor with the flow cell attached, showing the functional components of the sensor.

576

A.N. Naimushin et al. / Biosensors and Bioelectronics 17 (2002) 573 /584

ture of the sensor and the flow cell constant, they were enclosed in an aluminum box, which was machined for optimal contact with the sensor. A thermoelectric (TE) module (CP 1.4-71-10L, Melcore, Trenton, NJ) and a temperature controller (HTC-3000, Wavelength electronics, Bozeman, MT) with a feedback loop were used to keep the temperature of the enclosure constant. Since our experiments were conducted in the laboratory, where the temperature never changed by more that 9/ 2 8C, further insulation was not necessary. However, if the experiments are performed in the field, it is recommended that the sensor enclosure be insulted for optimal temperature stability. This will not only provide better temperature control when the ambient temperature is significantly different from that of the sensor, but it will reduce the power consumption by the TE module. The temperature controller can be powered by DC voltage between 5 and 12.5 V, which allows the use of a standard 12 V battery for field applications. 2.4. Pre-functionalization of the sensor slide surfaces Sensor surface functionalization is a multi-step process that prepares the surface for specific detection applications. Gold binding peptide (GBP)-alkaline phosphatase (AP) fusion protein was prepared and assayed as described previously (Woodbury et al., 1998), except that gold-coated glass slides (described above) were used in place of the gold beads for assays of GBP-AP binding to the gold surfaces. GBP binds tightly to the gold sensor surface, forming a foundation layer to which the recognition elements (antibodies) are covalently attached. Prior to each experiment, gold-coated slides were cleaned and pre-functionalized according to the following protocol. (1) The gold-coated slides were cleaned with ethanol using a lens paper to remove fingerprints, oily residues and dust particles. (2) The slide was further cleaned with an airbrush propelling n propanol as a solvent through a small nozzle and then air-dried. (3) For further cleaning each slide was then placed in a 1.5 ml Eppendorf tube containing 1 M NaOH and 0.5% Triton X-100 and shaken for 45 min on an orbital shaker. (4) They were thoroughly rinsed twice with H2O and then with PKT-50. (5) Next, the slides were incubated in 20 mg/ml GBP-AP solution in PKT-50 buffer for 1 h on an orbital shaker. (6) They were then rinsed with PKT-50 followed by trypsin buffer. (7) Slides were then incubated in 50 mg/ml trypsin for 30 min on an orbital shaker. Trypsin was used to cleave alkaline phosphatase from the gold binding peptide domain (Woodbury et al., 1998). (8) The slides were then rinsed with trypsin buffer. (9) The slides were next incubated in 50 mg/ml trypsin inhibitor for 20 min on an orbital shaker. (10) Slides were then rinsed in PKT-50 buffer. (11) Steps 5/10 were repeated and the slides were stored in PKT-50 buffer at 4 8C until used. The trypsin

step is used to remove the AP domain, bringing the analytes closer to the sensor surface and also reducing non-specific binding to the AP domain. The resulting GBP foundation layer also pacified the gold surfaces to reduce non-specific binding. 2.5. Sensor assembly and initialization A small droplet of index-matching liquid was applied to the sensor surface to optically couple the disposable gold-coated slide to the sensor. The flow cell was held in place with six screws that were tightened until the gasket provided a good seal between the Teflon block and gold slide. The SPR sensor was initialized in 15% sucrose solution. The initialization procedure prepared the sensor for refractive index measurements by: (1) setting the LED intensity and detector arrays integration times to maximize the sensor signal without saturating the output at any pixel, (2) looking at the detector background with the LED off, and (3) correcting for pixel to pixel response non-uniformity. Initialization can be performed in air, however, since drying the foundation layer may denature the proteins, initialization of prefunctionalized slides was performed in sucrose solution. Choosing a 15% solution, which has a RI of /1.35, insured that the SPR minimum for that solution did not interfere with the subsequent measurements. The initialization procedure was followed by calibration, in which the response of two channels was calibrated with respect to each other (see below). 2.6. Sensor functionalization For specific analyte detection, antibodies for the target antigen were covalently attached to the GBP foundation layer. The carboxyl groups on the GBP were activated using standard EDC, S-NHS chemistry (Grabarek and Gergely, 1990; Staros et al., 1986). EDC (100 mg) and S-NHS (27.5 mg) were dissolved in 5 ml H2O and flowed for 30 min through the sensor flow cells. The system was then briefly flushed with H2O (2 min). To slow the hydrolysis of active esters present after the addition of EDC and S-NHS, 10 mM NaAc at pH 5.0 was used as a buffer. The antibodies were coupled to the GBP foundation layer by flowing a 10 mg/ml solution in 10 mM NaAc buffer through the system for 1 h. Residual active esters were then blocked with a brief flush of 1M ethanolamine solution, pH 8.0 (10 min). The sensor was then equilibrated with the matrix in which detection was to take place. 2.7. Detection protocols 2.7.1. Direct detection Prior to each measurement, including initialization and calibration, the sensor was warmed up for a

A.N. Naimushin et al. / Biosensors and Bioelectronics 17 (2002) 573 /584

minimum of 15 min. Flow in all experiments was provided by a peristaltic mini-pump (#138761, Fisher Scientific, Hampton, NH) at a rate of /40 ml/min. The pump was modified by addition of a rigid vinyl faceplate that accepted two pump tubes. Each channel in the twochannel SPR sensor played a crucial role. The sensing channel was dedicated to detection of the target analyte, while the reference channel functioned as a control channel. The control channel had either GBP or an antibody specific for an analyte other than SEB. The output containing SEB was collected in a waste container containing 0.05% sodium hypoclorite. After a direct detection experiment, the anti-SEB active surface was regenerated by a 10 min wash in 100 mM glycine pH 2.0. This procedure removes more than 90% of the bound SEB. More than 50 regeneration cycles could be performed, while retaining 80% of the binding activity. 2.7.2. Amplification protocols Detection of SEB at lower levels required either one or two amplification steps. After sensor exposure to the low concentration of target analyte, the sensor was washed with appropriate buffer, then polyclonal rabbit anti-SEB antibodies were flowed through the sensor at 10 mg/ml to allow the antibodies to bind to the captured SEB analyte. For detection of SEB at femtomolar levels, the binding of rabbit anti-SEB antibodies was followed by binding goat anti-rabbit antibodies to the captured rabbit anti-SEB antibodies. Since goat anti-rabbit antibodies were used in the second amplification step, mouse monoclonal anti-SEB antibodies were used as capture antibodies. Regeneration procedures allowed the reuse of these sensors for primary detection only, since a small fraction of SEB molecules remained bound, even after the regeneration step.

577

sensor, flowing over both channels simultaneously. With each new sucrose solution, the RI value corresponding to a particular solution was entered in the program. The software performed least-squares fit of a quadratic function of the form: RI an (PNn )2 bn (PNn )cn

(1)

where n is the channel number, PN is the pixel number of the minimum in the SPR curve dip, and an , bn and cn are the adjustable parameters. Eq. (1) might have different coefficients for the two channels for at least two reasons: (a) the two SPR curves might have different dip profiles, therefore, forming a minimum at a different pixel number, (b) the PD arrays are not necessarily mirror images of one another with respect to the symmetry axis of the sensor, i.e. they can be slightly shifted with respect to one another. The calibration curves for the two SPR sensor channels are shown in Fig. 2. Measured SPR response (pixel position of the minimum in reflected light intensity) was plotted vs. the bulk index of refraction for six sucrose solutions of the indicated concentrations. These curves were used to synchronize the response of the two channels. The result of the calibration of the two-channel SPR sensor is shown in Fig. 3. The response of both channels was recorded continuously for six solutions containing from 1 to 6% sucrose. After the calibration, the values of the RI in two channels match to within 2% of the RI change. The sucrose solutions were chosen so that their RI’s covered the range of RI change due to surface antigen or antibody binding in typical detection experiments.

2.8. SPR sensor calibration Before any measurements could be carried out with the two-channel sensor system, it was critical to establish that both channels responded identically to the changes in RI or, alternatively, they could be mutually calibrated through the appropriate software manipulations to yield identical responses to the same change in RI’s. It is noteworthy that while the absolute value of the RI is not crucially important in most detection protocols, the relative change is. If one channel responds twice as much as the other, one to the same bioagent binding, the observer might come to an invalid conclusion that there is twice as much target analyte in the first channel, that there is significant non-specific binding, or there is no analyte present. To avoid this problem, the calibration function was incorporated into the software controlling the two-channel sensor. To calibrate the sensor, six aqueous sucrose solutions ranging from 1 to 6% were sequentially run through the

Fig. 2. Calibration curves for the two SPR sensor channels. Measured SPR response (pixel position of the minimum in reflected light intensity) vs. the bulk index of refraction of the six sucrose solutions, containing from 1 to 6% sucrose, in contact with the SPR surface. These curves were used to synchronize the response of the two channels.

578

A.N. Naimushin et al. / Biosensors and Bioelectronics 17 (2002) 573 /584

Fig. 3. Response of the two SPR sensor channels, shown in black and grey after performing calibration. The SPR response was recorded continuously for six solutions containing from 1 to 6% sucrose. Following the calibration, the values of the RI in two channels matched to within 2% of the RI change.

3. Results and discussion 3.1. Direct detection of SEB in TTBS To test the lowest detection limit (LDL) for direct detection of SEB by the two-channel SPR sensor, experiments were performed with solutions of various SEB concentrations in TTBS buffer. With the direct detection method, no amplification analytes were used, so the sensor could function continuously for hours at a time until a detection event was observed. When used in this format, the SPR biosensor functioned as a monitor. Detection of nine different concentrations of SEB in TTBS using the integrated SPR biosensor with the reference channel is illustrated in Fig. 4. The sensing channel was functionalized with affinity purified polyclonal rabbit anti-SEB antibodies. The reference channel was functionalized with anti-dinitrophenol (DNP) antibodies. The reference channel was used to compensate for non-specific binding, bulk refractive index changes and minor temperature fluctuations. Initially, the sensor was equilibrated in TTBS buffer to establish a baseline. At time t/10 min, solutions of various concentrations of SEB in TTBS were flowed through both channels. At time t/70 min, both channels were again washed with TTBS buffer. In the order of increasing initial slope, the curves correspond to the reference channel and 0.2, 0.5, 1, 2, 3, 10, 25, 50 and 75 nM SEB through the sensing channel, respectively. Samples were analyzed in random order with regeneration steps in between binding experiments. Three independent experiments were performed for each SEB concentration. Plots of SEB binding at concentrations of 1 nM or less were linear for at least the first 30 min. This linearity confirms the additive nature of the SPR response to the surface

Fig. 4. Direct detection of various concentrations of SEB in TTBS buffer by the integrated dual channel SPR sensor system. The sensing channel was functionalized with affinity purified polyclonal anti-SEB antibodies, while the reference channel was functionalized with affinity purified polyclonal anti-DNP antibodies. The reference channel was used to compensate for non-specific binding, bulk refractive index changes and minor temperature fluctuations. Initially, the sensor was equilibrated in TTBS buffer. At time t  10 min, solutions of various concentration of SEB in TTBS were introduced in both flow channels. At time t 70 min, both channels were washed with TTBS buffer again. In the order of increasing initial slope, the curves correspond to the reference channel and 0.2, 0.5, 1, 2, 3, 10, 25, 50 and 75 nM SEB, respectively.

binding. Therefore, this type of biosensor system can detect even smaller amounts of SEB if it is continuously exposed to small concentrations for extended periods of time. Solutions with higher SEB levels, especially those above 10 nM, saturate the surface in less than 5 min. To calculate the amount of time required to detect a certain concentration of SEB, or alternatively, to determine the lowest concentration that can be detected in an allocated time the initial slopes of the binding curves were analyzed. Changes in the apparent refractive index per min. (initial slopes of the binding curves shown in Fig. 4) were plotted versus SEB concentration (Fig. 5). Error bars for the three binding experiments for each SEB concentration were calculated as qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi N3 ¯ 2 =(N 1); where i is the number of a 9 ai1 (xi  x) particular experiment, N /3 is the total number of experiments at each SEB concentration, and x¯ is the average binding rate of the three experiments. The inset shows a linear dependence of the binding rate versus SEB concentration at SEB concentrations below 4 nM. The slope of this linear fit is /3 /106 RIU/min per nM. This relationship, which depends on such parameters as flow cell design, flow rate, antibodies used, etc, can be used to extrapolate the lowest detectable SEB concentration or the amount of time required to detect certain concentration for a specific sensor configuration. Pre-averaging of 32 SPR spectra yielded /4/106 RIU root-mean-squared (R.M.S.) short-term noise at /0.3 Hz. If we take three standard deviations (S.D.) as

A.N. Naimushin et al. / Biosensors and Bioelectronics 17 (2002) 573 /584

579

written as a weighted average of RI of adsorbed layer, na, and bulk RI, nb, with ca and cb as their appropriate weight coefficients. neff ca na cb nb

Fig. 5. Values of the initial slopes of the binding curves (change in apparent refractive index per min.) shown in Fig. 4 plotted vs. SEB concentration. The inset shows the linear dependence of the binding rate vs. SEB concentration at low SEB concentrations. This binding rate, which depends on such parameters as flow cell design, flow rate, antibodies used, etc., can be used to calculate the lowest detectable SEB concentration for a specific sensor configuration or the amount of time required to detect a certain concentration of SEB.

the cut-off limit, then, in 1 min, 4 nM and higher concentrations of SEB can be detected. If a longer detection time is allowed, then 16 point averaging can be performed lowering the R.M.S. short-term noise to / 1 /106 RIU and reducing the effective sampling rate to 1 per min. With this averaging, 70 pM and higher concentrations of SEB can be detected in 15 min. This represents more sensitive detection than obtained with an array biosensor (10 ng/ml or 350 pM in 14 min) (Rowe et al., 1999; Rowe-Taitt et al., 2000; Wadkins et al., 1998), a fiber optic sensor using chemiluminescence (Starodub et al., 1999) or fluorescence (Anderson et al., 2000), and a flow cell immunosensor (3 nM in 15 min) (Koch et al., 2000). In addition, the SPR sensor can run continuously for hours, sampling solutions until it encounters one with target analyte present. Unlike discrete sampling methods, SPR sensor systems do not consume antibody reagents in direct detection mode and can function effectively as stand-alone modules. 3.2. Surface coverage calculations SPR sensors are extremely sensitive to a change of refractive index within a few hundred nanometers of the gold surface. Most proteins have a refractive index near 1.6 (Jung et al., 1998), which differs significantly from that of water (1.33). Capture of the target analyte by antibodies bound to the sensing surface changes the apparent RI due to solution displacement by analytes of higher refractive index. A simple formalism (Jung et al., 1998; Liedberg et al., 1993; Lukosz, 1997) can be used to calculate the minimum detectable amount of bound protein for known detection limit of a sensor. An SPR sensor measures an effective refractive index that can be

(2)

To calculate coefficients ca and cb, the decay of the intensity of the evanescent electromagnetic field has to be taken in the account. The evanescent field decays away exponentially as exp(/z /ld) (Kurosawa et al., 1986; Liedberg et al., 1993) and light intensity decays as exp2(/z /ld)/exp(/2z/ld), where z is the height above the sensor surface and ld is the characteristic decay length of the evanescent electromagnetic field. If the adsorbed layer has a thickness da and the flow chamber extends beyond the ld then, Eq. (2) can be written as an integral (Jung et al., 1998).     2 2z neff  n(z)exp dz ld ld

g

(3)

0

For a bilayer system, where a small protein is bound to the surface, n (z) /na for 0B/z B/da and n(z )/nb for da B/z B/. In this case Eq. (3) reduces to:      2da 2da neff na 1exp nb exp (4) ld ld Eq. (4) can be rewritten as:    2da neff nb (na nb ) 1exp ld

(5)

If da /ld, which is the case with a single layer of adsorbed protein, then Eq. (5) can be rearranged to solve for the thickness of the adsorbed protein layer:   l (neff  nb ) da  d (6) 2 (na  nb ) Using Eq. (6) one can calculate the lowest detectable thickness of bound protein layer as well as detection limit in g/mm2 of sensor surface and surface density. We will use ld /307 nm, na /1.57 (Jung et al., 1998), nb / 1.33, and the SPR sensor detection limit of 4/106 (3 /S.D.). Using these values, one obtains dmin /2.6 pm. Densities of most proteins in aqueous buffers are / 1.3 g/ml (Darnell et al., 1990). This translates into minimal detectable surface coverage of /3.3 pg/mm2. Using molecular weight of SEB, this number translates into 7.1 /107 SEB molecules per mm2. At this surface coverage the average distance between two proteins is /120 nm, which is quite reasonable and well below saturation. The largest signal achieved in a direct binding experiment was 4 /104 RIU. At that satura9 tion level there should be 7.1 / /10 SEB molecules per 2 mm , and the average distance between them would be /12 nm. A 150 kDa protein, such as an antibody, has a linear dimension on the order of 6 /7 nm. Thus, the SPR

580

A.N. Naimushin et al. / Biosensors and Bioelectronics 17 (2002) 573 /584

The food processing industry is interested in detection of SEB in milk and food samples, while civilian and military antiterrorist monitoring programs are interested in detection of BW agents in aerosols and other matrices. These environments are much more complex than simple buffers, they contain many other biological molecules and organisms and other debris. To investigate the performance of the SPR biosensor in complex media, seawater, milk and urine were selected as representative examples. It was claimed (Rowe-Taitt et al., 2000) that label-free methods, such as SPR, are susceptible to problems in complex samples. We set out to test these claims with the reference compensated twochannel SPR sensor. Upon introduction of complex samples, there was an instant change in SPR signal in both channels corresponding to the bulk RI change due to switching to different media (e.g. Fig. 6, at 10 min). The detection of three different concentrations of SEB in urine is illustrated in Fig. 6. The sensor was functionalized the same way as for direct detection in TTBS. Initially, the sensor was equilibrated in urine. Adsorption of various biomolecules from urine onto the sensor surfaces caused the apparent refractive index to increase on both channels. The non-specific background rise in signal was subtracted from both channels for clarity of explanation and ease of comparison to direct detection of SEB in TTBS. In the order of increasing initial slope of SEB detection signal the curves corre-

spond to: the reference channel and the sensing channel, exposed to: 0.5, 1 and 10 nM SEB solution in urine, respectively. At time t/10 min, solutions of urine spiked with various concentrations of SEB were introduced in both flow channels. The rapid initial drop in the refractive index was due to dilution of urine with the SEB solutions. While the introduced volume of SEB was rather small, the SPR technique was still sensitive to minute changes in refractive index. Even 10 ml diluted in 10 ml will cause a noticeable change in refractive index, if the refractive indices of the two solutions differ by 0.01 RIU. The detection of SEB in urine resulted in / 50% decrease in the binding response relative to that recorded when using TTBS buffer. This may be explained in part by partial blocking of antibody active sites with other molecules and in part by reduced antibody affinity due to sub-optimal binding conditions. To confirm the detection of 0.5 nM SEB in urine, we performed amplification with anti-SEB antibodies. On an average, IgG antibodies (molecular weights of /150 kDa) are more than five times larger than SEB. If polyclonal antibodies are used for this amplification step, more than one antibody can bind per single captured SEB molecule. If an average of three antibodies bind per each SEB molecule, one can expect a 15fold increase in the signal. Fig. 7 illustrates direct detection of 0.5 nM SEB in urine followed by an amplification step with polyclonal anti-SEB antibodies. The sensor functionalization was identical to the previous experiment. Initially, the sensor was equilibrated in urine. At time t/10 min the solution of urine spiked with 0.5 nM of SEB was introduced in both flow

Fig. 6. Direct detection of various concentrations of SEB in urine by the two-channel integrated SPR sensor system. The sensor functionalization was the same as for Fig. 4. Initially, the sensor was equilibrated with urine. At time t  10 min, solutions of urine spiked with various concentrations of SEB were introduced through both flow channels. In the order of increasing initial slope, the curves correspond to the reference channel and 0.5, 1 and 10 nM SEB, respectively (all curves were corrected for non-specific binding by subtracting RI change of the reference channel).

Fig. 7. Direct detection of 0.5 nM SEB in urine followed by one amplification step. The black and gray lines represent the sensing and the reference channels, respectively. The sensor functionalization was the same as for Fig. 4. Initially sensor was equilibrated in urine. At time t 10 min, solution of urine spiked with 0.5 nM of SEB was introduced through both flow channels (reference and sensing). At time t  55 min, both channels were washed with urine for 10 min after which urine spiked with 10 mg/ml of anti-SEB antibody was flowed over both channels.

signal obtained by saturating the surface with SEB indicates /30% surface coverage with active antibodies. 3.3. Direct detection of SEB in complex media

A.N. Naimushin et al. / Biosensors and Bioelectronics 17 (2002) 573 /584

channels. At time t/55 min, both channels were washed with unspiked urine for 10 min, after which urine spiked with 10 mg/ml of anti-SEB antibody was introduced to both channels. The direct detection signal was indeed amplified approximately 15-fold. Continuous baseline drift due to non-specific binding of crude sample components was observed in both channels. In this figure, the drift was intentionally not corrected for to show the magnitude of the non-specific binding. If the SPR sensor system is employed as a remotely operated system, occasional pulses of amplifying antibodies can be introduced. Direct detection of 1 nM SEB in seawater is shown in Fig. 8. Seawater was collected from the Puget Sound, Bainbridge Island, Washington, and filtered through a 50 mm frit to remove large particulate matter. All proteins and most microorganisms remained in the filtrate. The sensor was equilibrated in the filtered seawater, then at t/10 min, 1 nM SEB solution in seawater was flowed through both channels. There was a bulk RI change upon introduction of the SEB solution in seawater. The bulk RI was then cancelled out on returning to unspiked seawater. The magnitude of the SPR signal was reduced in comparison with detection in TTBS buffer due to the same factors described above. Direct detection of SEB in milk proceeded the same as detection in urine and seawater, (data not shown). These direct detection experiments in complex sample matrices demonstrate the effectiveness of the reference-compensated SPR biosensor with the temperature control module.

Fig. 8. Direct detection of 1 nM SEB in seawater. The black and gray lines represent the sensing and the reference channels, respectively. Seawater was collected from Puget Sound and filtered through a coarse 50 mm frit. Sensor functionalization was the same as for Fig. 4. Initially, the sensor was equilibrated in seawater, then at time t  10 min, 1 nM SEB solution in seawater was introduced through both flow channels. The bulk RI change observed at t  10 and  70 min were due dilution of high RI seawater with lower RI SEB solution.

581

3.4. Signal amplification for the detection of SEB To lower the detection limit, the SPR sensor system was used with an amplification step(s). This approach can also be used to confirm the presence of analyte or further categorize detected analyte (if, for example, there are several species). Using a monoclonal antibody for a specific epitope allows for the differentiation of subspecies of captured analytes. Fig. 9 illustrates the detection of 20 pM (0.56 ng/ml) SEB in TTBS via a single step amplification with rabbit polyclonal antiSEB antibodies. First, the sensor was exposed to 20 pM solution of SEB for 60 min. The input flow of a 20 pM solution of SEB did not produce noticeable changes in either of the two channels, so the data are not included in the figure. To remove any SEB not captured by antibodies both channels were washed with TTBS for 10 min. At time t /20 min, 10 mg/ml of rabbit polyclonal anti-SEB antibody in TTBS was flowed through both channels. At time t/60 min, both channels were again washed with TTBS to remove any non-specifically bound antibodies. The specific response produced a 6 /105 apparent RI change, while the non-specific binding was initially /1.5 /105 but returned to baseline following the thorough washing. Although the solution of 20 pM SEB was too dilute to be detected directly, it could be readily detected when secondary anti-SEB antibodies were introduced. Higher concentrations of SEB required less incubation time. Using 3 /106 RIU/min per nM as the binding rate of SEB in step one, and a factor of 15

Fig. 9. Detection of 20 pM SEB in TTBS with one amplification step. The black and gray lines represent the sensing and the reference channels, respectively. The sensor functionalization was the same as for Fig. 4. The flow of 20 pM solution of SEB did not produce noticeable changes in either of the channels, so the data are not included in the figure. Following the flow of 20 pM SEB solution, both channels were washed with TTBS, then at time t  20 min, TTBS with 10 mg/ml of rabbit polyclonal anti-SEB antibodies was flowed over both channels. At time t  60 min, both channels were again washed with TTBS. The specific response produced  6  10 5 change in the apparent RI, while non-specific binding was  1.5 10 5 and was washed to zero following a thorough washing.

582

A.N. Naimushin et al. / Biosensors and Bioelectronics 17 (2002) 573 /584

amplification achieved with polyclonal antibodies, one can calculate the primary incubation time necessary to detect a specific concentration of SEB. Given the noise characteristics of this particular sensor, the initial exposure time can be cut to 5/10 min and the amplification step to 5 min, therefore, reducing overall detection time to 15 min. The LDL obtained with one amplification step was comparable to or better than methods, such as ELISA (O’Brien et al., 2000; Paddle, 1996; Rowe et al., 1999; Starodub et al., 1994), which allows for the detection of concentrations of 0.5 /1 ng/ ml (17.5 pM) albeit in 4 h. Other detection methods and their respective LDL include: bidifractive grating (1 ng/ ml or 35 pM in 15 min) (O’Brien et al., 2000), and an impedance sensor (0.4 ng/ml or 14 pM, time not stated) (DeSilva et al., 1995). Detection of 50 pM SEB in urine is shown in Fig. 10. The sensor functionalization was performed the same as for Fig. 4. A flow of urine spiked with 50 pM solution of SEB did not produce noticeable changes in either the sensing or the reference channel (data not shown). After flowing urine spiked with 50 pM SEB solution through both channels, they were washed with unspiked urine, then at time t/10 min, urine spiked with 10 mg/ml of rabbit anti-SEB antibody was flowed over both channels. There is a characteristic drop in the effective RI at that time due to dilution of urine with SEB solution. At time t/45 min, both channels were again washed with urine. The specific response produced a change of / 4 /105 RIU, while non-specific binding was zero. This time the jump in RI corresponded to switching from urine spiked with SEB to unspiked urine.

Fig. 10. Detection of 50 pM SEB in urine with one amplification step. The black and gray lines represent the sensing and the reference channels, respectively. The sensor functionalization was the same as for Fig. 4. Flow of 50 pM solution of SEB did not produce noticeable changes in either of the channels (data not shown). After flowing 50 pM SEB solution through both channels, they were washed with urine, then at time t  10 min, urine spiked with 10 mg/ml of rabbit anti-SEB antibody was flowed over both channels. At time t  45 min, both channels were again washed with urine. The specific response produced a change of  4 10 5 RIU, while non-specific binding was zero.

Amplification strategies utilizing high molecular weight secondary and tertiary detection agents can result in dramatic improvement in LDL when employing SPR detection. Fig. 11 demonstrates detection of 100 fM (2.8 pg/ml) SEB in TTBS. Sensor functionalization was the same as described above, except that mouse monoclonal anti-SEB antibodies were used as the capture antibodies in place of the rabbit polyclonal antibodies. The use of a different capture antibody was necessary, since goat anti-rabbit antibodies were used in the second amplification step. This precludes the use of polyclonal rabbit anti-SEB as capture antibodies. At 100 fM, SPR signal from neither SEB nor the first amplification step with rabbit polyclonal anti-SEB antibodies was observed. Only the second amplification step with goat anti-rabbit antibodies is shown. The initial 1-h SEB binding step and 30 min first amplification step are not shown. There was a rapid bulk refractive index change upon introduction of the solution containing 10 mg/ml goat anti-rabbit antibodies. This change due to the bulk RI change cancels out upon returning to TTBS. Goat anti-rabbit antibodies were allowed to bind for 30 min, however, the signal was clearly observable after 10/15 min. A major advantage of the two-channel miniature SPR sensor over the single-channel device is on-chip reference capability. The use of the reference channel provided confidence that the amplification steps were indeed providing amplification specific to the target analyte. At this level of detection, SPR rivals the two

Fig. 11. Detection of 100 fM SEB in TTBS buffer with two amplification steps. The black and gray lines represent the sensing and the reference channels, respectively. The sensor functionalization was the same as for Fig. 4, except that monoclonal anti-SEB antibodies were immobilized on the sensor surface in place of polyclonal antibodies. At such low concentrations of SEB, neither SEB nor the amplification step with rabbit polyclonal anti-SEB antibodies was observable. Only the second amplification step with goat anti-rabbit antibodies is shown. There was a rapid bulk refractive index change upon introduction of the solution containing 10 mg/ml goat anti-rabbit antibodies and another, in the opposite direction upon washing with TTBS. Goat anti-rabbit antibodies were allowed to bind for 30 min, however, the specific signal was clearly observable after 10 /15 min.

A.N. Naimushin et al. / Biosensors and Bioelectronics 17 (2002) 573 /584

other methods that show femtomolar detection of SEB. Light addressable potentiometric sensors (LAPS) (Choi et al., 1998; Lee et al., 2000; Uithoven et al., 2000) were shown to detect 5 pg/ml (180 fM) in 1 h. Two other techniques, IMS-ECL and IMS-FCL (Yu, 1996, 1998; Yu et al., 1998, 2000) allow low picomolar and even femtomolar detection of SEB by utilizing IMS and concentration of antigen with antibody coated magnetic beads. IMS-ECL was reported to detect 0.5 /1 pg/ml (17 /35 fM) in 30 min, and IMS-FCL could detect 100 pg/ml (3.6 pM) in the same amount of time. The authors believe that IMS played the major role in these detection methods. IMS and concentration of the sample prior to SPR measurements, should allow further improvement of the LDL. Gold conjugated antibodies are known to produce a large SPR signal (Lyon et al., 1998) and can be used to further lower the LDL.

4. Conclusions We have tested prototype miniature two-channel integrated SPR sensors with temperature stabilization for detection of SEB in buffer solutions as well as in complex media, including seawater, urine and milk. Detection was performed either directly or with amplification step(s). These sensor systems demonstrate the potential for field use and laboratory applications in monitoring, detection and identification of biological agents, and characterization of intermolecular interactions. They provide: (1) portability due to their small size and 12 V power requirement, (2) LDL equal to or better than 70 pM in direct detection mode, LDL of /5 pM with one amplification step and /100 fM with two amplification steps, (3) reference capability that eliminates false-positive signals and compensates for slow rates of drift, non-specific binding, and minor fluctuations in temperature, (4) continuous monitoring of the environment and (5) temperature stability via a temperature controller with a TE module. All of these factors make the miniature integrated SPR sensor systems prime candidates for field use. Disposable sensor modules with target prices of around 1 dollar can be pre-functionalized for specific detection applications, significantly reducing the operational expenses of such sensors. If the improvements observed in going from prototype single channel devices to production units can be used as an example, it can be assumed that production series multi-channel sensors will have lower noise levels than our current prototype modules.

Acknowledgements The authors are very thankful to Dr Thomas O’Brien (Tetracore LLC) for providing monoclonal antibodies

583

to SEB. We thank Dr John Quinn (TI) for critically reading this manuscript. This work was supported in part by the Department of Defense contract #DAAD13-C-0032, by Grant 61-9265 from Washington State SeaGrant, and by Grant 66-0618 from Center for Process Analytical Chemistry, University of Washington.

References Anderson, G.P., King, K.D., Gaffney, K.L., Johnson, L.H. 2000. Multi-analyte interrogation using the fiber optic biosensor. Biosens. Bioelectron. 14 (10 /11), 771 /777. Biacore AB, 1995. BIAcore 2000 instrument handout. Blanchard, G.C., Taylor, C.G., Busey, B.R., Williamson, M.L. 1990. Regeneration of immunosorbent surfaces used in clinical, industrial and environmental biosensors. Role of covalent and non-covalent interactions. J. Immunol. Methods 130 (2), 263 /275. Chinowsky, T.M., 2000. Optical multisensors based on surface plasmon resonance. Ph.D. Dissertation, University of Washington, Seattle. Choi, K., Seo, W., Cha, S., Choi, J. 1998. Evaluation of two types of biosensors for immunoassay of botulinum toxin. J. Biochem. Mol. Biol. 31 (1), 101 /105. Christopher, G.W., Cieslak, T.J., Pavlin, J.A., Eitzen, E.M., Jr 1997. Biological warfare. A historical perspective. J. Am. Med. Assoc. 278 (5), 412 /417. CRC 1978. CRC Handbook of Chemistry and Physics. CRC Press, West Palm Beach, FL. Darnell, J.E., Lodish, H., Baltimore, D., 1990. Molecular cell biology, Scientific American Books: Distributed by W.H. Freeman, New York, pp. 1105. DeSilva, M.S., Zhang, Y., Hesketh, P.J., Maclay, G.J., Gendel, S.M., Stetter, J.R. 1995. Impedance based sensing of the specific binding reaction between Staphylococcus enterotoxin B and its antibody on an ultra-thin platinum film. Biosens. Bioelectron. 10 (8), 675 /682. Easmon, C.S.F., Adlam, C. 1983. Staphylococci and Staphylococcal Infections. Academic Press, London/New York, p. 2v. Elkind, J.L., Stimpson, D.I., Strong, A.A., Bartholomew, D.U., Melendez, J.L., 1999. Integrated analytical sensors: the use of the TISPR-1 as a biosensor, Sens. Actuat. B (Chemical) 182 /190. Emanuel, P.A., Dang, J., Gebhardt, J.S., Aldrich, J., Garber, E.A., Kulaga, H., Stopa, P., Valdes, J.J., Dion-Schultz, A. 2000. Recombinant antibodies: a new reagent for biological agent detection. Biosens. Bioelectron. 14 (10 /11), 751 /759. Franz, D.R., Jahrling, P.B., Friedlander, A.M., McClain, D.J., Hoover, D.L., Bryne, W.R., Pavlin, J.A., Christopher, G.W., Eitzen, E.M., Jr 1997. Clinical recognition and management of patients exposed to biological warfare agents. J. Am. Med. Assoc. 278 (5), 399 /411. Grabarek, Z., Gergely, J. 1990. Zero-length crosslinking procedure with the use of active esters. Anal. Biochem. 185 (1), 131 /135. Holloway, H.C., Norwood, A.E., Fullerton, C.S., Engel, C.C., Jr, Ursano, R.J. 1997. The threat of biological weapons. Prophylaxis and mitigation of psychological and social consequences. J. Am. Med. Assoc. 278 (5), 425 /427. Homola, J., Lu, H.B., Nenninger, G.G., Yee, S.S., Campbell, C.T., 1999. Novel approach to multichannel SPR sensing. Proc. SPIEInt. Soc. Opt. Eng. 3857 (Chemical Microsensors and Applications II) 198 /206. Jorgenson, R.C., Yee, S.S. 1993. A fiber-optic chemical sensor based on surface plasmon resonance. Sens. Actuat. B 12 (3), 213 /220. Jung, L.S., Campbell, C.T., Chinowsky, T.M., Mar, M.N., Yee, S.S. 1998. Quantitative interpretation of the response of surface

584

A.N. Naimushin et al. / Biosensors and Bioelectronics 17 (2002) 573 /584

plasmon resonance sensors to adsorbed films. Langmuir 14 (19), 5636 /5648. Koch, S., Wolf, H., Danapel, C., Feller, K.A. 2000. Optical flow-cell multichannel immunosensor for the detection of biological warfare agents. Biosens. Bioelectron. 14 (10 /11), 779 /784. Kurosawa, K., Pierce, R.M., Ushioda, S., Hemminger, J.C. 1986. Raman scattering and attenuated-total-reflection studies of surface-plasmon polaritons. Phys. Rev. B: Condens. Matter 33 (2), 789 /798. Lee, W.E., Thompson, H.G., Hall, J.G., Bader, D.E. 2000. Rapid detection and identification of biological and chemical agents by immunoassay, gene probe assay and enzyme inhibition using a silicon-based biosensor. Biosens. Bioelectron. 14 (10 /11), 795 /804. Liedberg, B., Lundstroem, I., Stenberg, E. 1993. Principles of biosensing with an extended coupling matrix and surface plasmon resonance. Sens. Actuat. B B11 (1 /3), 63 /72. Lukosz, W. 1997. Integrated-optical and surface-plasmon sensors for direct affinity sensing. Part II: anisotropy of adsorbed or bound protein adlayers. Biosens. Bioelectron. 12 (3), 175 /184. Lyon, L.A., Musick, M.D., Natan, M.J. 1998. Colloidal Au-enhanced surface plasmon resonance immunosensing. Anal. Chem. 70 (24), 5177 /5183. Melendez, J., Carr, R., Bartholomew, D., Taneja, H., Yee, S., Jung, C., Furlong, C. 1997. Development of a surface plasmon resonance sensor for commercial applications. Sens. Actuat. B B39 (1 /3), 375 /379. Melendez, J., Carr, R., Bartholomew, D.U., Kukanskis, K., Elkind, J., Yee, S., Furlong, C., Woodbury, R. 1996. A commercial solution for surface plasmon sensing. Sens. Actuat. B B35 (1 /3), 212 /216. O’Brien, T., Johnson, L.H., III, Aldrich, J.L., Allen, S.G., Liang, L.T., Plummer, A.L., Krak, S.J., Boiarski, A.A. 2000. The development of immunoassays to four biological threat agents in a bidiffractive grating biosensor. Biosens. Bioelectron. 14 (10 /11), 815 /828. Paddle, B.M. 1996. Biosensors for chemical and biological agents of defense interest. Biosens. Bioelectron. 11 (11), 1079 /1113. Rowe, C.A., Tender, L.M., Feldstein, M.J., Golden, J.P., Scruggs, S.B., MacCraith, B.D., Cras, J.J., Ligler, F.S. 1999. Array biosensor for simultaneous identification of bacterial, viral, and protein analytes. Anal. Chem. 71 (17), 3846 /3852. Rowe-Taitt, C.A., Golden, J.P., Feldstein, M.J., Cras, J.J., Hoffman, K.E., Ligler, F.S. 2000. Array biosensor for detection of biohazards. Biosens. Bioelectron. 14 (10 /11), 785 /794.

Starodub, N.F., Arenkov, P.Y., Starodub, A.N., Berezin, V.A. 1994. Construction and biomedical application of immunosensors based on fiber optics and enhanced chemiluminescence. Opt. Eng. (Bellingham Wash.) 33 (9), 2958 /2963. Starodub, N.F., Dibrova, T.L., Shirshov, Y.M., Kostyukevych, K.V. 1999. Development of the myoglobin sensor based on the surface plasmon resonance. Ukr. Biokhim. Zh. 71 (2), 33 /37. Staros, J.V., Wright, R.W., Swingle, D.M. 1986. Enhancement by Nhydroxysulfosuccinimide of water-soluble carbodiimide- mediated coupling reactions. Anal. Biochem. 156 (1), 220 /222. Suzuki, T., Morita, H., Ono, K., Maekawa, K., Nagai, R., Yazaki, Y. 1995. Sarin poisoning in Tokyo subway. Lancet 345 (8955), 980. Uithoven, K.A., Schmidt, J.C., Ballman, M.E. 2000. Rapid identification of biological warfare agents using an instrument employing a light addressable potentiometric sensor and a flow-through immunofiltration-enzyme assay system. Biosens. Bioelectron. 14 (10 / 11), 761 /770. Wadkins, R.M., Golden, J.P., Pritsiolas, L.M., Ligler, F.S. 1998. Detection of multiple toxic agents using a planar array immunosensor. Biosens. Bioelectron. 13 (3 /4), 407 /415. Woodbury, R.G., Wendin, C., Clendenning, J., Melendez, J., Elkind, J., Bartholomew, D., Brown, S., Furlong, C.E. 1998. Construction of biosensors using a gold-binding polypeptide and a miniature integrated surface plasmon resonance sensor. Biosens. Bioelectron. 13 (10), 1117 /1126. Yoshida, T. 1994. Sarin poisoning in Matsumoto, Nagano prefecture. J. Toxicol. Sci. 19 (3), APP85 /APP88. Yu, H. 1996. Enhancing immunoelectrochemiluminescence (IECL) for sensitive bacterial detection. J. Immunol. Methods 192 (1 /2), 63 / 71. Yu, H. 1998. Comparative studies of magnetic particle-based solid phase fluorogenic and electrochemiluminescent immunoassay. J. Immunol. Methods 218 (1 /2), 1 /8. Yu, H., Ahmed, H., Vasta, G.R. 1998. Development of a magnetic microplate chemifluorimmunoassay for rapid detection of bacteria and toxin in blood. Anal. Biochem. 261 (1), 1 /7. Yu, H., Raymonda, J.W., McMahon, T.M., Campagnari, A.A. 2000. Detection of biological threat agents by immunomagnetic microsphere-based solid phase fluorogenic- and electro-chemiluminescence. Biosens. Bioelectron. 14 (10 /11), 829 /840. Zilinskas, R.A. 1997. Iraq’s biological weapons. The past as future? J. Am. Med. Assoc. 278 (5), 418 /424.