Rapid detection of Escherichia coli O157:H7 spiked into food matrices

Analytica Chimica Acta 584 (2007) 66–71

Rapid detection of Escherichia coli O157:H7 spiked into food matrices夽 Lisa C. Shriver-Lake ∗ , Stephanie Turner, Chris R. Taitt Center for Bio/Molecular Science and Engineering, Naval Research Laboratory, 4555 Overlook Avenue, SW, Washington DC 20375, United States Received 25 July 2006; received in revised form 26 October 2006; accepted 7 November 2006 Available online 12 November 2006

Abstract Food poisoning causes untold discomfort to many people each year. One of the primary culprits in food poisoning is Escherichia coli O157:H7. While most cases cause intestinal discomfort, up to 7% of the incidences lead to a severe complication called hemolytic uremic syndrome which may be fatal. The traditional method for detection of E. coli O157:H7 in cases of food poisoning is to culture the food matrices and/or human stool. Additional performance-based antibody methods are also being used. The NRL array biosensor was developed to detect multiple antigens in multiple samples with little sample pretreatment in under 30 min. An assay for the specific detection of E. coli O157:H7 was developed, optimized and tested with a variety of spiked food matrices in this study. With no sample pre-enrichment, 5 × 103 cells mL−1 were detected in buffer in less than 30 min. Slight losses of sensitivity (1–5 × 10−4 cell mL−1) but not specificity occur in the presence of high levels of extraneous bacteria and in various food matrices (ground beef, turkey sausage, carcass wash, and apple juice). No significant difference was observed in the detection of E. coli O157:H7 in typical culture media (Luria Broth and Tryptic Soy Broth). © 2006 Elsevier B.V. All rights reserved. Keywords: Escherichia coli O157:H7; Biosensor; Immunosensor; Food pathogen

1. Introduction Thousands of hours of human suffering and economic losses occur each year due to food poisoning. Typical symptoms include vomiting, nausea, abdominal cramps, and diarrhea [1]. An estimated 73,000 infections occur each year in which 2–7% of those cases have the severe complication called hemolytic uremic syndrome (HUS) [2]. The typical method used to determine the causative agent of an infection is through culturing of a stool sample which takes several days in which to obtain the results [2]. One of the major causative agents of painful abdominal cramps and bloody diarrhea is the bacterium, Escherichia coli O157:H7 [1,2]. It is typically associated with beef, sprouts, lettuce, milk and juice. While improvements in rapid identification after ingestion are important for treatment, it is more important to prevent infection. One way to do this is to iden-

夽 The views expressed here are those of the authors and do not represent those of the U.S. Navy, the U.S. Department of Defense or the U.S. Government. ∗ Corresponding author. Tel.: +1 202 404 6045; fax: +1 202 767 9594. E-mail address: [email protected] (L.C. Shriver-Lake).

0003-2670/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.aca.2006.11.021

tify food products contaminated with E. coli O157:H7 prior to ingestion, preferably before it is distributed to grocery stores, restaurants and manufacturing facilities. In addition to culturing, performance-based methods employing antibodies and DNA are becoming more popular with some tests being commercially available. The DNA-based tests, which are the most specific and sensitive methods available, are routinely used as a confirmatory assay [3–5]. In some cases, immunomagnetic beads are used in the sample pretreatment step. Recently, molecular beacons have been incorporated into DNA-based assays to generate a signal upon a binding event [4]. The disadvantages of the DNA-based methods are (1) the length of time to obtain results (up to 48 h) due to extensive sample pretreatment including enrichment and extraction, and (2) the requirement for highly trained staff to perform the assay and analyze the results. In addition, complex matrices such as food can contain polymerase chain reaction (PCR) inhibitors such as humics which must be removed prior to analysis. Immunoassays may not be as sensitive/specific as DNA assays but they do achieve low levels of sensitivity while avoiding many of the disadvantages of the DNA-based assays. Immunoassay-based methods employ antibodies to provide the

L.C. Shriver-Lake et al. / Analytica Chimica Acta 584 (2007) 66–71

specificity but use many different techniques to identify the antibody–antigen interaction. The traditional method is the enzyme-linked immunosorbent assay (ELISA) which is based on enzyme amplification to generate a colored product. Other methods based this method are referred to as ELISA-based [4,6,7]. In addition to spectrometric detection, electrochemical detection of an electroactive enzyme with square wave voltammetry has been employed. [8]. With these methods, 103 to 105 cells mL−1 have been detected without enrichment and down to 1 CFU mL−1 with enrichment. Enrichment can only be used with organisms that can be grown and are usually reported as CFU’s whereas the immunoassays without enrichment can also detect dead and damaged bacteria and are reported as cells mL−1 . As with the DNA assays, an immunomagnetic separation step can be used to enrich the sample, provide additional specificity, and separate the organism from complex matrices prior to analysis [9–12]. However, in contrast to PCR, separation from complex matrices is not an absolute requirement. Some immunoassay methods use changes in optical properties including transmitted light, surface plasmon or acoustic waves. Changes in mass due to antibody–antigen binding cause changes in the surface plasmons or acoustic wave properties that are proportional to the antigen present [4,13,14]. In addition, fluorescent labels have been used to provide the optical signal in several methods [6,15–17]. Flow cytometry assays using fluorescently labeled antibodies bound to bacteria have been developed [18,19]. Yamaguchi et al. [19] were able to detect E. coli O157:H7 at 103 cells mL−1 in food matrices following extensive pretreatment (removal of lipids and protein and centrifugation at 12,000 × g). Hibi et al. [18] reduced the time to 2 h for the flow cytometery assay (sample preparation and assay) by using immunomagnetic beads as a method for sample pretreatment. A subset of these fluorescent methods employ the evanescent wave (EW) to selectively excite fluorophore-tagged “tracer” molecules (primarily antibodies) using a “sandwich” assay. These EW-based methods have been used on both optical fibers and planar waveguides [3,20]. The array biosensor is an EW-based detector using a planar waveguide and was developed to detect multiple targets in several samples simultaneously in less than 30 min. Assays have been developed using the array biosensor for toxins and bacteria with detection limits in pg mL−1 and 104 to 105 cells mL−1 ranges, respectively. Stripes of capture molecules (antibodies, sugars, peptides,) are patterned on the surface of a microscope slide [21–24]. Multiple samples are then applied to the patterned slide, using flow channels oriented perpendicular to the stripes of capture molecules such that each sample is exposed to all immobilized capture molecules. For bacteria and large molecules, a sandwich immunoassay format is employed. After the sample is exposed to the slide surface, a solution containing fluorescently labeled tracer antibody is run over in the same direction as the sample [25–31]. For small molecules, the fluorescent tracer is added to the sample for a single-step competitive immunoassay [32,33]. The location and fluorescent intensity identify the target and concentration. This report describes a single analyte assay for E. coli O157:H7 developed for the array biosensor. The study included

67

the optimization of antibodies and assay times as well as the effect of food matrices and other bacteria on specificity. The ultimate goal is to incorporate this method into a multi-analyte, multi-sample assay. 2. Methods 2.1. Material Goat anti-E. coli O157:H7, heat-killed E. coli O157:H7, Salmonella typhimurium, and Listeria monocytogenes positive controls were obtained from KPL Inc. (Gaithersburg, MD). Other anti-E. coli O157:H7 antibodies included a monoclonal and a goat polyclonal from Biodesign (Saco, ME), and a monoclonal from Fitzgerald (Concord, MA). The positive control antibodies (biotin-rabbit anti-chicken IgY and Cy5-labeled chicken IgY) were purchased from Jackson ImmunoResearch (West Grove, PA). Mercaptopropyl-trimethoxy silane (MTS), n-succinimidyl 4-maleimidobutyrate (GMBS), toluene, phosphate buffered saline (PBS), phosphate buffered saline with Tween 20 (PBST) were obtained from Sigma/Aldrich/Fluka (St. Louis, MO). NHS-LC-biotin was purchased from Pierce Chemical Co. (Rockford, IL). The fluorophore AlexaFluor 647 carboxylic acid, succinimidyl ester (AF647) was obtained from Invitrogen/Molecular Probes (Carlsbad, CA). E. coli K-12 was graciously supplied by Linda Powers (Utah State University). All bacteria were inactivated prior to arrival at NRL. All assays were performed without enrichment; results are therefore reported as cells mL−1 . Difco Tryptic Soy Broth (TSB), yeast extract, and tryptone was purchased from BD Biosciences (San Jose, CA). Luria Broth (LB) was prepared by dissolving 5 g yeast extract, 10 g NaCl (Sigma), and 10 g tryptone in 1 L distilled water. 2.2. Antibody labeling (biotin and fluorophore) Several antibodies were evaluated for their suitability as either a capture and/or tracer antibody. Capture antibodies were biotinylated for 30 min with NHS-LC-biotin (Pierce, Rockford, IL) at a 5:1 molar ratio in 0.5 M borate buffer, pH 8.5, at room temperature. Biotin-labeled antibodies were separated from the free biotin using a Bio-Rad P-10 column. For the tracer antibodies, AF647 labeling was performed according to Anderson and Nerurkar [34]. After separation of the free fluorophore from the antibody fraction, the antibody concentration and the dyeto-protein ratio were calculated from the absorbance at 280 and 650 nm. 2.3. Preparation of antibody-patterned waveguides NeutrAvidin-coated glass microscope slides were prepared as described in Ligler et al. [25]. Briefly, the slides were cleaned with immersion in a 10% KOH in methanol solution for 30 min, extensively rinsed with 18 m water (Millipore, Milford, MA) and dried with nitrogen. The slides were then incubated in a solution of 2% MTS in toluene. After 1 h, the slides were rinsed with toluene and dried with nitrogen. The slides were then exposed to

68

L.C. Shriver-Lake et al. / Analytica Chimica Acta 584 (2007) 66–71

2 mM GMBS in ethanol for 30 min. The slides were again rinsed with water, placed in 30 ␮g mL−1 NeutrAvidin in PBS, and incubated overnight at 4 ◦ C. The next day the NeutrAvidin-coated slides were rinsed and stored with PBS at 4 ◦ C, until used. To create stripes of capture antibodies, a 12-channel PDMS template was placed over the NeutrAvidin surface of the slide. A full description can be found in Golden et al. [21]. The biotinylated capture antibody was injected into each channel with a 1-mL syringe. The normal pattern for the 12 channels in this study was channels 1 and 12 were the positive controls (antichicken IgY), channel 7 was the negative control (buffer), and channels 2–6 and 8–11 were biotinylated anti-E. coli O157:H7 antibodies (10 ␮g mL−1 ). 2.4. Preparation of spike samples Heat-killed E. coli O157:H7 from KPL was reconstituted in 50:50 glycerol/water (v:v) at least one day prior to use. Stock E. coli was added to PBST containing 1 mg mL−1 BSA (PBSTB) to generate buffer samples in the concentration range of 1 × 102 to 5 × 106 cells mL−1 . Ground beef or turkey sausage were homogenized with PBSTB at a 1:3 (w:v) ratio in a blender for 1 min. For carcass wash samples, PBSTB (100 mL) was added to a bag containing a whole chicken (∼5 lbs). The bag was rocked rapidly for 2 h at room temperature and the liquid, which contained fats, blood, and chicken juices was removed and frozen until use. Apple juice was neutralized as follows: 2 mL of 10 × PBSTB and 60 ␮L 10N NaOH were added to 18 mL of apple juice. Stock E. coli were added to the appropriate food matrices to create 10-fold dilution samples from 5 × 103 to 5 × 106 cells mL−1 and incubated a minimum of 2 h at room temperature prior to analysis [29]. The meat slurry mixture was clarified by centrifugation (5 min, 2000 rpm) to remove large particles to prevent clogging of the fluidics. The supernatant was used for analysis. No centrifugation was required for the carcass wash and apple juice. Since assays may be run after enrichment, typical culture media (TSB and LB) were also tested. TSB was prepared according to the Difco’s instructions. Concentrations of E. coli (5 × 102 to 5 × 105 cells mL−1 ) were spiked into the prepared media and incubated at room temperature for 2 h. Since the stock E. coli was heat-killed, there was no cell growth in the culture media during the 2 h interaction period. To test specificity, several bacteria often found in similar foodstuffs were co-spiked with the E. coli at 2–5 × 107 cells mL−1 with the various E. coli concentrations in PBSTB: Salmonella and Listeria. E. coli K-12 was tested as a non-pathogenic E. coli. 2.5. Array biosensor The array biosensor was developed to analyze multiple samples for multiple analytes. The sensor has been described in detail in Golden et al. [21]. A 635-nm diode laser illuminates one end of a microscope slide (away from patterned region) and the image of the patterned region is captured by a CCD aligned 90◦ to the illumination with the appropriate filters for AF647.

2.6. Assay A sandwich immunoassay was performed on the waveguide surface by applying sample and tracer antibodies through PDMS flow cells oriented perpendicular to the stripes of capture antibodies [21]. This allowed each sample to be exposed to all capture antibodies. To accomplish this, the slide with the attached flow cell was connected to an Instec 24-channel pump with tubing extending from the exit end of each channel. Syringe barrels (1 mL), serving as reservoirs for the various assay fluids, were attached to the entry end of each channel. The channels were first washed with 1 mL PBSTB at the maximum flow rate. Next, 0.8 mL sample was applied to each channel for 15 min at 0.1 mL min−1 with recirculation of the sample. This was followed by another rapid wash with 1 mL PBSTB. The fluorescent tracer antibody (0.4 mL containing 10 ␮g mL−1 fluorescently labeled anti-E. coli O157:H7 and 100 ng mL−1 Cy5-chicken IgY in PBSTB) was applied to the channels for 4 min, followed by a final rapid wash with 1 mL PBSTB. The slide was dried and imaged with the array biosensor. 2.7. Data collection and analysis Using the CCD camera’s software (SpectraSource), an image of the slide was captured in the flexible image transport system (FTS) digital format. The image intensity was converted to average fluorescence intensity per square of capture antibody using a program developed at NRL [35]. Basically, the program identifies the location of the square and background areas next to the square. From the average fluorescence intensity of the assay square, the average background intensity value (from either side of the square) was subtracted. All values were imported into Microsoft Excel. The average mean intensity for each spot after background subtraction was used for the analyses. 3. Results and discussion The first task was the selection of appropriate antibodies to provide the best sensitivity and range of detection. An advantage of the array biosensor is the ability to screen multiple combinations of capture/tracer antibodies simultaneously. In this study, concentrations of 2, 10, and 50 ␮g mL−1 capture antibody were patterned and exposed to 103 to 107 cells mL−1 spiked into PBSTB and a blank. The tracer antibody concentration was held constant at 10 ␮g mL−1 . The combination of biotin-KPL goat anti-E. coli O157:H7 (capture) and AF647-KPL goat anti-E. coli O157:H7 (tracer) gave the largest net increase (data not shown) and was selected to use for the remaining studies. It was also determined that the length of time for the sample to be exposed to the slide surface should be increased from 8 to 15 min to improve detection sensitivity. Further increases in sample exposure time may offer additional sensitivity, but were not tested, as we desired to keep the total assay time under 30 min. The limit and range of detection were first determined in the buffer PBSTB. Fig. 1 shows the dose–response curve in buffer with the inset enlarging the limit of detection (LOD) region. The mean values for each concentration were fitted to

L.C. Shriver-Lake et al. / Analytica Chimica Acta 584 (2007) 66–71

Fig. 1. E. coli O157:H7 dose–response curve in buffer. For each point the mean and S.E.M. for a minimum of 14 spots are shown.

an asymmetric sigmoidal curve (Fig P software, BioSoft, Cambridge, United Kingdom). The results are shown in Table 1. The LOD, calculated as the lowest concentration tested giving a signal that is three standard deviations above the blank, was 5 × 103 cells mL−1 . There was no leveling off at the highest concentration tested (5 × 106 cells mL−1 ).

69

Fig. 2. E. coli O157:H7 dose–response curve in the presence of high levels of extraneous bacteria. () buffer; (䊉) Listeria; () E. coli K-12; * S. typhimurium. For each point the mean and S.E.M. for a minimum of 18 spots are shown.

To determine the specificity of the assay, high levels of S. typhimurium, L. monocytogenes, and E. coli K12 (2–5 × 107 cells mL−1 ) were spiked into PBSTB samples containing the various concentrations of E. coli O157:H7. The dose–response curves for the various bacteria and controls were

Table 1 Limits of detection and equations for curves Treatment

LOD (cells mL−1 )

Curve

r2

PBSTB

5 × 103

y = −24.5 + (53375.5/(1 + ((x/1.8e6)0.97 ))

0.997

Other bacteria L. monocytogenes E. coli K-12 S. typhimurium

1 × 104 5 × 104 5 × 104

y = 492.0 + (27208.9/(1 + ((x/129447.9)1.31 )) y = 2175.2 + (16598.1/(1 + ((x/106631.5)2.14 )) y = 1033.7 + (29965.6/(1 + ((x/617045.9)0.82 ))

0.999 0.959 0.991

Food matrices Carcass wash Turkey sausage Apple juice Ground beef

1–5 × 104 1–5 × 104 1–5 × 104 1–5 × 104

y = 333.5 + (34078.7/(1 + ((x/750669.9)0.98 )) y = 951.4 + (22338.4/(1 + ((x/102660.8)1.04 )) y = 3743.9 + (13602.9/(1 + ((x/21316.0)0.80 )) y = 867.4 + (24729.6/(1 + ((x/202530.9)1.07 ))

0.999 0.999 0.989 0.999

Culture media Luria Broth Tryptic Soy Broth

1 × 104 1–5 × 104

y = 0.05x + 1002.7 y = 0.03x + 32.18

0.987 0.999

Fig. 3. Image of E. coli O157:H7 detection in buffer and food matrix.

70

L.C. Shriver-Lake et al. / Analytica Chimica Acta 584 (2007) 66–71

fitted with the asymmetric sigmoid and are shown in Fig. 2 with the curve results shown in Table 1. In all cases, there was a reduction in assay sensitivity in the presence of these contaminating bacteria. The excess non-relevant bacteria may bind non-specifically to the surface thereby decreasing the number of sites available for the E. coli O157:H7 to bind. This nonspecific binding by these other species are not detected by the anti-E. coli tracer; however, they would be detectable in a multianalyte assay format, since tracer antibody cocktail would be directed against all species present. While analysis in buffer is important, analysis in food samples with minimal sample preparation is the goal. Several food matrices (ground beef, turkey sausage, chicken carcass wash, and apple juice) were spiked with various concentrations of E. coli O157:H7. Fig. 3 is a representative image showing both the buffer control and matrix samples on a single slide. Fig. 4 shows the dose–response curves in food matrices with the buffer control samples. Limits of detection in the different matrices were similar but slightly higher than the buffer at 1–5 × 104 cells mL−1 . This demonstrates the importance of doing control standard curves in the matrix of interest if quantitation is required. To achieve the regulatory goal of one bacterial cell per milliliter, preconcentration with a method such as immunomagnetic separation (IMS) may prove to be very useful. With E. coli doubling times ranging from 15 to 20 min depending on the media, approximately 4–5 h would be required to grow from one bacterium per milliliter to the array biosensor detection limit. If culturing is already being performed, direct testing from culture at intermediate times would be useful in reducing the time needed and provide another application for the array biosensor. To determine if the array biosensor method could also be used with samples from culture media, E. coli O157:H7 detection was determined in LB and TSB matrices. Results from the dose–response curves are shown in Fig. 5 and Table 1. Comparison of the PBSTB samples on the same slides as the culture media samples indicated no significant difference in the curves.

Fig. 5. E. coli O157:H7 dose–response curve in buffer and culture media. () none; (䊉) LB; () TSB. The E. coli concentrations for these curves are 10fold lower than shown in the two previous figures. For each point the mean and S.E.M. for a minimum of 36 spots are shown.

4. Conclusion In this study, it has been demonstrated that the array biosensor can be employed for the detection of E. coli O157:H7 in a variety of matrices and in the presence of high levels of extraneous bacteria. The assay was completed in less than 30 min with minimal sample preparation. The limit of detection without sample concentration or enrichment is 5 × 103 cells mL−1 in buffer and 1–5 × 104 cells mL−1 in most other matrices. If quantitation is desired, this study indicated the need to develop standard curves in the appropriate matrices. To achieve the one cell per milliliter level desired by regulatory agencies, preconcentration with immunomagnetic separation would be advantageous. As demonstrated, direct detection of cells in culture media can be performed and would be useful if microbiological culture is being performed for confirmation. The next step is the incorporation of this assay into a multi-analyte format with the array biosensor. Acknowledgement The authors thank Dr. Linda Powers for her generous gift of E. coli K-12. This work was funded by the Food and Drug Administration (FDA). References

Fig. 4. E. coli O157:H7 dose–response curve in buffer and various food matrices. () buffer; (䊉) carcass wash; () turkey sausage; () apple juice; * ground beef. For each point the mean and S.E.M. for a minimum of 18 spots are shown.

[1] CDC, Foodborne Illness, 2006, www.cdc.gov/ncidod/dbmd/diseaseinfo/ foodborneinfections g.htm. [2] CDC, Escherichia coli O157:H7, 2006, www.cdc.gov/ncidod/dbmd/ diseaseinfo/escherichiacoli g.htm. [3] J.M. Simpson, D.V. Lim, Biosens. Bioelectron. 21 (2005) 881. [4] A.K. Deisingh, M. Thompson, J. Appl. Microbiol. 96 (2004) 419. [5] C.-F. Wu, J.J. Valdes, W.E. Bentley, J.W. Sekowski, Biosens. Bioelectron. 19 (2003) 1. [6] C. Nyquist-Battie, L.E. Frank, D. Lund, D.V. Lim, J. Food Prot. 67 (2004) 2756. [7] R.A. Abuknesha, F. Darwish, Talanta 65 (2005) 343. [8] A.G. Gehring, S.-I. Tu, J. Food Prot. 68 (2005) 146.

L.C. Shriver-Lake et al. / Analytica Chimica Acta 584 (2007) 66–71 [9] X.-L. Su, Y. Li, Anal. Chem. 76 (2005) 4806. [10] I.H. Boyaci, Z.P. Aguilar, M. Hossain, H.B. Halsall, C.J. Seliskar, W.R. Heineman, Anal. Bioanal. Chem. 382 (2005) 1234. [11] M.L. Ochoa, P.B. Harrington, Anal. Chem. 77 (2005) 5258. [12] D.P. Chandler, J. Brown, D. Call, R.S. Wunschel, J.W. Grate, D.A. Holman, L. Olson, M.S. Stottlemyre, C.J. Bruckner-Lea, Int. J. Food Microbiol. 70 (2001) 143. [13] A. Subramanian, J. Irudayaraj, T. Ryan, Biosens. Bioelectron. 21 (2006) 998. [14] E. Berkenpas, P. Millard, M. Pereira da Cunha, Biosens. Bioelectron. 21 (2006) 2255. [15] T.R. DeCory, R.A. Durst, S.J. Zimmerman, L.A. Garringer, G. Paluca, H.H. DeCory, R.A. Montagna, Appl. Environ. Microbiol. 71 (2005) 1856. [16] J.-a.A. Ho, H.-W. Hsu, Anal. Chem. 75 (2003) 4330. [17] S. Tu, M. Golden, P. Cooke, G. Paoli, A. Gehring, J. Rapid Meth. Aut. Mic. 13 (2005) 269. [18] K. Hibi, A. Abe, E. Ohashi, K. Mitsubayashi, H. Ushio, T. Hayashi, H. Ren, H. Endo, Anal. Chim. Acta 573/574 (2006) 158. [19] N. Yamaguchi, M. Sasada, M. Yamanaka, M. Nasu, Cytometry A 54 (2003) 27. [20] G.P. Anderson, B.M. Lingerfelt, C.R. Taitt, Sens. Lett. 2 (2004) 18. [21] J. Golden, L. Shriver-Lake, K. Sapsford, F. Ligler, Methods 37 (2005) 65.

71

[22] N.V. Kulagina, M.E. Lassman, F.S. Ligler, C.R. Taitt, Anal. Chem. 77 (2005) 6504. [23] M.M. Ngundi, C.R. Taitt, S.A. McMurry, D. Kahne, F.S. Ligler, Biosens. Bioelectron. 21 (2006) 1195. [24] C.A. Rowe-Taitt, J.J. Cras, C.H. Patterson, J.P. Golden, F.S. Ligler, Anal. Biochem. 281 (2000) 123. [25] F.S. Ligler, C.R. Taitt, L.C. Shriver-Lake, K.E. Sapsford, Y. Shubin, J.P. Golden, Anal. Bioanal. Chem. 377 (2003) 469. [26] K.E. Sapsford, A. Rasooly, C.R. Taitt, F.S. Ligler, Anal. Chem. 76 (2004) 433. [27] K.E. Sapsford, C.R. Taitt, N. Loo, F.S. Ligler, Appl. Environ. Microbiol. 71 (2005) 5590. [28] L.C. Shriver-Lake, F.S. Ligler, IEEE Sens. J. 5 (2005) 751. [29] L.C. Shriver-Lake, Y.S. Shubin, F.S. Ligler, J. Food Prot. 66 (2003) 1851. [30] L.C. Shriver-Lake, C.R. Taitt, F.S. Ligler, J. AOAC Int. 87 (2004) 1498. [31] C.R. Taitt, Y.S. Shubin, R. Angel, F.S. Ligler, Appl. Environ. Microbiol. 70 (2004) 152. [32] M.M. Ngundi, S.A. Qadri, E.V. Wallace, M.H. Moore, M.E. Lassman, L.C. Shriver-Lake, F.S. Ligler, C.R. Taitt, Environ. Sci. Technol. 40 (2006) 2352. [33] M.M. Ngundi, L.C. Shriver-Lake, M.H. Moore, M.E. Lassman, F.S. Ligler, C.R. Taitt, Anal. Chem. 77 (2005) 148. [34] G.P. Anderson, N.L. Nerurkar, J. Immunol. Meth. 271 (2002) 17. [35] K.E. Sapsford, Z. Liron, Y.S. Shubin, F.S. Ligler, Anal. Chem. 73 (2001) 5518.