Quantitative and simultaneous detection of four foodborne bacterial pathogens with a multi-channel SPR sensor

Quantitative and simultaneous detection of four foodborne bacterial pathogens with a multi-channel SPR sensor

Biosensors and Bioelectronics 22 (2006) 752–758 Quantitative and simultaneous detection of four foodborne bacterial pathogens with a multi-channel SP...

440KB Sizes 8 Downloads 52 Views

Biosensors and Bioelectronics 22 (2006) 752–758

Quantitative and simultaneous detection of four foodborne bacterial pathogens with a multi-channel SPR sensor Allen D. Taylor a , Jon Ladd a , Qiuming Yu a , Shengfu Chen a , Jiˇr´ı Homola a,b , Shaoyi Jiang a,∗ a

b

Department of Chemical Engineering, University of Washington, Box 351750, Seattle, WA 98195, USA Institute of Radio Engineering and Electronics, Academy of Sciences, Chaberska 57, 18251 Prague, Czech Republic Received 4 November 2005; received in revised form 23 February 2006; accepted 1 March 2006 Available online 25 April 2006

Abstract We report the quantitative and simultaneous detection of four species of bacteria, Escherichia coli O157:H7, Salmonella choleraesuis serotype typhimurium, Listeria monocytogenes, and Campylobacter jejuni, using an eight-channel surface plasmon resonance (SPR) sensor based on wavelength division multiplexing. Detection curves showing SPR response versus analyte concentration were established for each species of bacteria in buffer at pH 7.4, apple juice at native pH 3.7, and apple juice at an adjusted pH of 7.4, as well as for a mixture containing all four species of bacteria in buffer. Control experiments were performed to show the non-fouling characteristics of the sensor surface as well as the specificity of the amplification antibodies used in this study. The limit of detection (LOD) for each of the four species of bacteria in the tested matrices ranges from 3.4 × 103 to 1.2 × 105 cfu/ml. Detection curves in buffer of an individual species of bacteria in a mixture of all four species of bacteria correlated well with detection curves of the individual species of bacteria alone. SPR responses were higher for bacteria in apple juice at pH 7.4 than in apple juice at pH 3.7. This difference in sensor response could be partly attributed to the pH dependence of antibody–antigen binding. © 2006 Elsevier B.V. All rights reserved. Keywords: Foodborne pathogens; Salmonella typhimurium; Listeria monocytogenes; Campylobacter jejuni; Escherichia coli O157:H7; Surface plasmon resonance (SPR)

1. Introduction Bacteria exist in almost every facet of nature. While many are beneficial to humans and essential to the balance of the environment, a number of bacteria are pathogenic. It is estimated that approximately 76 million illnesses, 325,000 hospitalizations, and 5000 deaths occur in the United States each year due to foodborne pathogens (Mead et al., 1999). The four bacterial pathogens Salmonella spp., Listeria monocytogenes (L. monocytogenes), Campylobacter jejuni (C. jejuni), and Escherichia coli (E. coli) O157:H7, have been estimated to account for approximately 67% of food-related deaths (Mead et al., 1999). The economic impact of these four pathogens in the United States is estimated to cost up to US$ 5.4 billion annually for foodborne cases (Buzby et al., 1996). To reduce the occurrence of food-related illnesses and death, there is a need for fast, sen-



Corresponding author. Tel.: +1 206 616 6509; fax: +1 206 543 3778. E-mail address: [email protected] (S. Jiang).

0956-5663/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.bios.2006.03.012

sitive, and reliable detection methods to identify contaminated foods. Conventional culturing methods using selective media, biochemical methods, and serological methods are the current standard for the detection of bacterial pathogens. Culturing methods are reliable, but are time consuming, typically requiring 3–7 days to obtain results (Alocilja and Radke, 2003; USDA/FSIS, 1998). Detection time for bacteria has been reduced to 8–48 h by the adoption of methods such as polymerase chain reaction (PCR) and enzyme-linked immunosorbant assays (ELISA) (Alocilja and Radke, 2003). While PCR and ELISA have improved detection time, other biosensors can perform faster detection while maintaining high sensitivity and specificity. Several biosensor technologies have been applied to the detection of bacteria (Abdel-Hamid et al., 1999; Deng et al., 1996; Ivnitski et al., 1999; Libby and Wada, 1989; Plomer et al., 1992; Schneider et al., 1997). Recently, an immunosensor array based on the optical method of total internal reflection fluorescence was applied to the simultaneous detection of bacteria (Taitt et al., 2004a). The surface plasmon resonance (SPR) sensor is another optical method

A.D. Taylor et al. / Biosensors and Bioelectronics 22 (2006) 752–758

which can be scaled up to detect many analytes simultaneously using an array. Compared to most biosensor technologies, SPR has the advantage of providing real-time and label-free detections for the direct and continuous monitoring of bioanalytes. SPR sensors have been demonstrated to detect a wide range of pathogens, including protein toxins (Homola et al., 2002), small molecule toxins (Yu et al., 2005), and bacteria (Fratamico et al., 1998). Since each of these analytes has a relatively different size, their detection usually require different methods. In this study, we used a novel multi-channel SPR biosensor, based on wavelength division multiplexing (Homola, 2003), to simultaneously and quantitatively detect four foodborne pathogens, Salmonella choleraesuis (S. choleraesuis) serotype typhimurium, L. monocytogenes, C. jejuni, and E. coli O157:H7 in buffer and apple juice at pH 3.7 and 7.4. 2. Materials and methods 2.1. Materials 2.1.1. Reagents Streptavidin and phosphate buffered saline (PBS) (0.01 M phosphate, 0.138 M sodium chloride, 0.0027 M potassium chloride, pH 7.4) were purchased from Sigma–Aldrich (St. Louis, MO). Oligo (ethylene glycol) (OEG) alkanethiol (HS–(CH2 )10 –(OCH2 )4 –OH) was purchased from ProChimia (Gdansk, Poland). Biotinylated (BAT) alkanethiol (HS–(CH2 )15 –(OCH2 )3 –biotin) was a gift from Buddy Ratner’s group at the University of Washington. Safeway brand apple juice was purchased at a Safeway grocery store (Seattle, WA). 2.1.2. Bacteria Stock cultures of E. coli O157:H7 (ATCC 700728) and S. choleraesuis serotype typhimurium (ATCC 700720) were purchased from the American Type Culture Collection (ATCC, Manassas, VA). Heat-killed L. monocytogenes and heat-killed C. jejuni were purchased from Kirkegaard & Perry Laboratories (Gaithersburg, MD).

753

2.1.3. Antibodies Polyclonal rabbit antibody (PAb) and biotinylated polyclonal rabbit antibody for C. jejuni, L. monocytogenes, E. coli, and Salmonella species were purchased from Biodesign International (Saco, ME). 2.2. SPR sensor In this study, we used a custom-built SPR sensor based on the Kretschmann geometry of the attenuated total reflection (ATR) method and wavelength modulation (Fig. 1a). A SF14 glass chipcoated with a gold SPR-active film is optically matched to a SF14 glass prism with immersion oil (Cargille, Cedar Grove, NJ). An acrylic flow cell with a laser cut 50 ␮m thick Mylar gasket is mechanically pressed to the functionalized chip creating eight sensing channels arranged as shown in Fig. 1a. A multichannel peristaltic pump and Teflon tubing are used to address each sensing channel individually. A collimated polychromatic light beam is directed through the prism to the gold-coated substrate where it excites surface plasmons at the metal–dielectric interface. As shown in Fig. 1c, the light reflected from the thin metal interface meets the surface twice at different angles of incidence, producing a SPR at different wavelengths. Both SPR wavelengths are monitored with one spectrophotometer (Fig. 1b). The SPR wavelength for each of the eight flow channels is monitored by four spectrophotometers. Since the two angles of incidence are fixed, the wavelength of light absorbed by the SPR is directly related to the refractive index of the dielectric film in contact with the thin metal surface. A SPR sensorgram records the shift of the resonant wavelength as a function of time, and it can be used to quantify the amount of captured analyte. 2.3. Surface functionalization Glass chips are coated with a 2 nm adhesion-promoting chromium film and then a 55 nm surface plasmon-active gold film, both deposited by e-beam evaporation in vacuum. The bare gold surface is cleaned by washing with absolute ethanol,

Fig. 1. (a) Schematic of the eight-channel SPR sensor based on wavelength division multiplexing, (b) normalized reflectivity as a function of wavelength for two different angles of incidence, and (c) SPR wavelength division multiplexing of one beam of light being reflected onto the gold surface at two different angles of incidence (α = 52◦ and β = 55◦ ) by a custom-made prism coupler.

754

A.D. Taylor et al. / Biosensors and Bioelectronics 22 (2006) 752–758

drying with nitrogen, stripping organic contaminants in an UV Ozone cleaner for 20 min, washing the surface with 18.2 M cm DI water and absolute ethanol, and drying with nitrogen. The surface is subsequently functionalized with the formation of a mixed self-assembled monolayer (SAM) by incubating the cleaned substrate in ethanol with 10 ␮M OEG alkanethiol and 90 ␮M BAT alkanethiol overnight at room temperature. The BAT is used to target streptavidin for subsequent biotinylated antibody binding, while the OEG creates a non-fouling background. The 1:9 ratio of BAT:OEG was used to maximize streptavidin binding, while minimizing non-specific binding. After the formation of the mixed SAM, the substrate is removed from the solution and washed with absolute ethanol. 2.4. Patterning antibodies on sensor surface Within each of the eight flow channels, a baseline with PBS is established by recording the SPR response for 10 min at a flow rate of 10 ␮l/min. A solution of streptavidin (50 ␮g/ml) is flowed at 10 ␮l/min for 10 min. Then, PBS is again flowed for 10 min to wash the surface before functionalization with a commercially biotinylated PAb. Biotinylated PAb (1 mg/ml) is flowed at 10 ␮l/min for 1 min followed by a PBS wash for 10 min. Each of the flow channels can be functionalized with a different antibody, enabling the simultaneous detection of multiple analytes. 2.5. Sample preparation Stocks of E. coli and S. choleraesuis were stored frozen at −80 ◦ C in Luria-Bertani (LB) broth containing 16% glycerol. The bacteria were resuscitated by inoculating 5 ml of sterile LB broth in a test tube using an inoculating loop to scrape bacteria from the frozen stock and incubating 16–20 h at 37 ◦ C in a reciprocal shaking water bath. Samples of overnight E. coli cultures were pelletized in a centrifuge at 10,000 × g for 10 min and the supernatant was discarded. The samples were then resuspended in 5 ml of PBS, pelletized again at 10,000 × g for 10 min, and resuspended in 5 ml of PBS. The washed bacteria were enumerated by plating 10-fold serial dilutions on LB agar plates and incubating at 37 ◦ C. The bacteria were then heat-killed by heating to 90 ◦ C for 1 h, aliquoted, and frozen at −20 ◦ C for subsequent detection. L. monocytogenes and C. jejuni were purchased heat-killed and lypholized and were resuspended in PBS for treatment and detection. The bacteria were prepared for the detection by inserting an ultrasonication rod (Branson Sonifier 450) at 90% power for 15 min with the sample sitting in an ice bath. The samples were serial diluted in PBS or apple juice to produce samples for analysis. It was previously shown that the limit of detection (LOD) for bacteria using a SPR sensor is not only dependent on the sensitivity of the instrument and the specificity and affinity of the surface chemistry, but also on the sample preparation method (Taylor et al., 2005). Detergent lysing bacteria improved the LOD of E. coli O157:H7 by two orders of magnitude compared to untreated (live) bacteria and one order of magnitude compared to heat-killed bacteria. The detergent lysing method changes the diffusion and hydrodynamics of bacteria by breaking the

Fig. 2. Sensorgram showing (a) binding of streptavidin to an OEG:BAT mixed SAM on Au, (b) binding of biotinylated PAb to streptavidin, (c) direct detection of an analyte, and (d) secondary amplification using PAb.

cells, which improves the LOD of E. coli using a SPR sensor. Ultrasonication is a method commonly used for lysing bacteria to obtain proteins from inside the bacteria and it is a non-intrusive alternative to detergent lysing. In this work, ultrasonication of heat-killed bacteria improved the LOD to values comparable to the LOD of detergent-lysed samples (data not shown). 2.6. SPR detection of bacteria Fig. 2 shows the sensorgram from the multistep protocol used to detect bacteria with the SPR sensor. The detection of samples containing diluted bacteria in PBS or apple juice was performed by first establishing a baseline with PBS at a flow rate of 50 ␮l/min. The detection samples are then flowed for 20 min at 50 ␮l/min, followed by a wash of PBS for 10 min. A sandwich assay was performed to amplify the direct response by flowing 50 ␮g/ml of PAb in PBS for 10 min at 10 ␮l/min, followed by a 10 min wash with PBS. The reliability of detections using the sensing instrument is improved by using a reference channel. A reference channel is created by following the same protocol as a detection channel with the only difference being buffer flowed in place of an analyte-containing solution. After amplification antibody is flowed over the sensor surface, the signal from the reference channel is subtracted from the signal in the detection channel. This compensation removes the portion of sensor responses that could otherwise be attributed to temperature shifts or nonspecific binding to the sensor surface. 3. Results and discussion 3.1. Calibration of SPR sensor with spectral discrimination of flow channels The eight-channel SPR sensor relies on two angles of incidence to spectrally discriminate the SPR response of two flow channels with one spectrometer. To compare SPR responses from flow channels with the two angles of incidence of our instrument, it is necessary to calibrate the bulk and surface refractive index sensitivities. We established the bulk refractive index calibration by comparing a refractive index step change caused by switching from DI water to PBS. Fig. 3a shows the

A.D. Taylor et al. / Biosensors and Bioelectronics 22 (2006) 752–758

755

resistant to protein adsorption, it is still difficult to produce surfaces that completely resist bacterial adhesion (Qian et al., 2002). Thus, it is desirable to use a secondary antibody specific to the target species of bacteria to verify any direct detection. In this work, each of the four antibody-immobilized sensor surfaces was tested for non-specific bacterial adhesion by flowing a mixture of the three non-target species of bacteria, each at a concentration of 107 cfu/ml. Fig. 4 shows that for each antibodyimmobilized sensor surface, non-specific adhesion of the nontarget species of bacteria produced sensor responses ranging from 0 to 0.41 nm. The subsequent exposure of each channel to the target amplification antibody using the same protocol as for a detection resulted in no net SPR shift. This indicates that the secondary antibody is specific and further validates the direct detection. In the analysis of all SPR experiments, the bacterial detection response is compensated using the non-specific binding response of the amplification antibody as a reference. The detection of bacteria in apple juice was demonstrated in this work. When apple juice is flowed over the sensor surface, there is non-specific adsorption of some components in apple juice (e.g. polyphenols) onto the sensor surface (data not shown). A subsequent exposure of the sensor surface to target amplification antibody resulted in no net SPR shift, indicating that the non-specific adsorptions from apple juice do not affect the responses seen for the detection of bacteria in apple juice. 3.3. Detection of four bacterial pathogens simultaneously Fig. 3. Sensorgrams showing (a) bulk refractive index sensitivity calibration and (b) surface refractive index sensitivity calibration for the two angles of incidence (52◦ and 55◦ ) of the SPR sensor.

bulk refractive index calibration for the two angles of incidence. The bulk refractive index sensitivity ratio was determined to be 2.93 with an error of ∼5% of the response. The surface refractive index calibration was determined by comparing the response from the binding of streptavidin to the BAT surface. Binding of streptavidin to BAT is the first step of functionalizing the sensor surface. All of the flow channels are treated identically (flow, concentrations, and substrate film thickness) with the exception of the two angles of incidence. The sensorgram in Fig. 3b represents the SPR response of biotin–streptavidin binding in the flow channels at two angles of incidence. A surface refractive index sensitivity ratio of 1.92 was determined, with an error of ∼8%. The experimentally obtained surface refractive index sensitivity ratio is ∼66% of the bulk refractive index sensitivity ratio and it agrees well with theoretical predictions (Homola et al., 2001). 3.2. Control of non-specific adsorption of proteins and bacteria Both the OEG and BAT components of the mixed SAM contain ethylene glycol units, which have been shown to minimize non-specific binding of proteins (Ladd et al., 2004). While significant progress has been made in creating sensor surfaces

A successful biosensor must discriminate a target analyte from a complex mixture. The multi-channel SPR biosensor used in this work can simultaneously discriminate multiple target analytes from complex mixtures. The capability of the SPR sensor to simultaneously detect multiple species of bacteria is shown using a sample containing 5 × 105 cfu/ml E. coli, 5 × 105 cfu/ml S. choleraesuis, 107 cfu/ml C. jejuni, and 5 × 105 cfu/ml L. monocytogenes in PBS. These concentrations were chosen to provide a similar magnitude of SPR response for each of species of bacteria in their specifically functionalized flow channel. In Fig. 5 we show the detection of each species of bacteria using a secondary antibody. Amplification using secondary antibodies allows us to achieve a lower detection limit while assuring specific detection. Our control experiments, shown in Fig. 4, demonstrated the specificity of these secondary antibodies, which were tested after the exposure of a sensor surface to three non-target species of bacteria at concentrations of 107 cfu/ml. Therefore, a detection performed on a mixture containing all four species of bacteria using secondary antibody amplification is assured to be specific. 3.4. Detection of four bacterial pathogens in buffer Detection curves were obtained for each of the four foodborne bacterial pathogens (E. coli, S. choleraesuis, C. jejuni, and L. monocytogenes). Fig. 6 compares the detection curves for each species of bacteria in buffer to those of a mixture of all four species of bacteria in buffer at equal concentrations. The

756

A.D. Taylor et al. / Biosensors and Bioelectronics 22 (2006) 752–758

Fig. 4. Control experiments demonstrating that the sensor surfaces are relatively non-fouling to non-target species of bacteria and the amplification antibodies do not bind to non-target species of bacteria: (a) anti-E. coli O157:H7 (EC) sensor surface, (b) anti-Salmonella choleraesuis (SC) serotype typhimurium sensor surface, (c) anti-Campylobacter jejuni (CJ) sensor surface, and (d) anti-Listeria monocytogenes (LM) sensor surface.

presence of the non-target species of bacteria does not interfere with the detection of the target species of bacteria. We calculate LOD for each species of bacteria from the data shown in Fig. 6 by interpolation or extrapolation based on three standard deviations of the baseline noise (Thomsen et al., 2003), which is 0.09 nm of SPR shift for our SPR sensor. The calculated LOD

for E. coli in PBS is 1.4 × 104 cfu/ml, S. choleraesuis in PBS is 4.4 × 104 cfu/ml, C. jejuni in PBS is 1.1 × 105 cfu/ml, and L. monocytogenes is 3.5 × 103 cfu/ml. Previous studies using a multi-analyte fluorescence-based sensor established the LOD for Salmonella enterica serovar typhimurium, L. monocytogenes and C. jejuni at 8 × 104 , 2.4 × 104 and 9.7 × 102 cfu/ml, respectively (Sapsford et al., 2004; Taitt et al., 2004a,b). These studies used antibodies from the same source for the detection of Salmonella and L. monocytogenes and obtained similar detection limits. The detection of C. jejuni was determined using different antibodies, which could contribute to the higher LOD obtained in this study. 3.5. Detection of bacterial pathogens in apple juice

Fig. 5. Reference compensated SPR sensorgrams showing the simultaneous detection of four foodborne bacterial pathogens in a mixture (solid lines) (i.e. 5 × 105 cfu/ml E. coli O157:H7 (EC), 5 × 105 cfu/ml S. choleraesuis (SC) serotype typhimurium, 107 cfu/ml C. jejuni (CJ), and 5 × 105 cfu/ml L. monocytogenes (LM)) compared to three non-target bacteria as control (dotted lines).

To determine the effect of an apple juice matrix on the detection of foodborne bacterial pathogens, samples of apple juice at the native pH of 3.7 and samples of apple juice at an adjusted pH of 7.4 were spiked with bacteria. Serial dilutions of spiked samples in apple juice were analyzed using the SPR sensor with a standard sandwich assay. The detection curves for E. coli, S. choleraesuis, C. jejuni, and L. monocytogenes in apple juice at pH of 3.7 and 7.4 were obtained. Results are shown in Fig. 6. Fig. 6 shows that the LOD for each of the four species bacteria in apple juice at pH of 3.7 and 7.4 and in PBS ranges from

A.D. Taylor et al. / Biosensors and Bioelectronics 22 (2006) 752–758

757

Fig. 6. SPR resonant wavelength shift vs. concentration of bacteria for the detection of (a) E. coli O157:H7, (b) Salmonella choleraesuis serotype typhimurium, (c) Campylobacter jejuni, and (d) Listeria monocytogene in various matrices. Samples contain either only one bacterial species in buffer (pure PBS), all four species of bacteria at equal concentrations in buffer (mixed PBS), one bacterial species in apple juice at adjusted pH 7.4 (pure apple juice 7.4), or one bacterial species in apple juice at native pH 3.7 (pure apple juice pH 3.7). The error bars shown for pure PBS data represent standard deviations for at least five detections at each concentration.

3.4 × 103 to 1.2 × 105 cfu/ml. Adjusting the pH of apple juice from the native pH 3.7 to 7.4 increased the sensor response to bacteria, but had little effect on the LOD. No trend was observed when comparing the detection of bacteria in PBS at pH 7.4 to that in apple juice at the same pH. Nyquist-Battie et al. previously reported a comparison study on the detection of E. coli O157:H7 in different detection matrices using ELISA. Their work showed a variance in the LOD from 103 to 104 to 105 cfu/ml as the detection matrix was changed from PBS to apple juice at pH 7.4 to apple juice at pH 3.9 (Nyquist-Battie et al., 2004). In this work, the LOD is almost constant for E. coli O157:H7 in apple juice at pH 3.7, apple juice at pH 7.4, and PBS. However, the sensor responses to E. coli O157:H7 from both studies followed the same trend. The sensor response was the highest for detections in PBS and the lowest for detection in apple juice at its native pH. The differences in the LOD between these two studies could be partly attributed to the difference in sensor technologies used (SPR versus ELISA). While our SPR is a label-free technique, the components of apple juice produced background fluorescence which interfere with their specific fluorescence detection (Nyquist-Battie et al., 2004). Nyquist-Battie et al. also suggest that components in

apple juice could interfere with antibody–antigen interactions. Their study showed that by reducing polyphenols in apple juice using a refining agent or changing pH, LODs could be improved. We studied the pH effect in buffer by comparing the detection of E. coli O157:H7 and S. choleraesuis at a concentration of 107 cfu/ml in PBS at a range of pH values between 3.0 and 9.0. Results showed that the SPR response was dependent on pH (data not shown). According to Gibas et al. (2000), the specificity of antibody–antigen binding is affected by several factors including surface shape complementarity, electrostatic interactions, hydrogen bonding, hydrophobic packing, and interfacial solvent interactions. The pH dependence of antibody–antigen binding can be attributed to subtle structural changes induced by the protonation/deprotonation state of the acidic and basic amino acids in both the paratope and epitope regions (Gibas et al., 2000). 4. Conclusions In this work, we demonstrate the simultaneous and specific detection of each of the four foodborne bacterial pathogens, E. coli O157:H7, S. choleraesuis serotype typhimurium, Listeria monocytogenes, and C. jejuni in both buffer and apple juice

758

A.D. Taylor et al. / Biosensors and Bioelectronics 22 (2006) 752–758

using a multi-channel SPR sensor based on wavelength division multiplexing. Detections of these bacteria were compared for individual species of bacteria in buffer, apple juice at native pH 3.7, apple juice with pH adjusted to 7.4, and in a mixture containing all four species of bacteria in buffer. The non-fouling characteristics of the sensor surface were evaluated and no net response was observed when secondary antibody was flowed across a sensor surface exposed to non-target species of bacteria. These results show the specificity of the amplification antibodies used in this study. The detection of individual species of bacteria from a mixture of four species of bacteria correlated well with the detection of only an individual species of bacteria, thus indicating that the non-target species of bacteria did not interfere with the specific detection. The detection of bacteria in apple juice at the adjusted pH of 7.4 produced an increase in the sensor response over the detection in apple juice at the native pH 3.7. The difference in sensor response in apple juice at pH 3.7 and pH 7.4 could be partly attributed to the pH dependence of antibody–antigen binding. Detection curves from apple juice differed from those obtained in buffer, but the LOD are similar for all cases. Thus, standard curves must be determined for each sample matrix if quantifiable results are desired. Acknowledgements We thank Maxi Boeckl in Buddy Ratner’s group at the University of Washington for synthesis of the BAT. This project was supported by a grant from the U.S. Food and Drug Administration (FD-U-002250). Allen D. Taylor was partially supported by the Graduate Opportunities and Minority Achievement Award Fellowship at the University of Washington. References Abdel-Hamid, I., Ivnitski, D., Atanasov, P., Wilkins, E., 1999. Biosens. Bioelectron. 14, 309–316.

Alocilja, E.C., Radke, S.M., 2003. Biosens. Bioelectron. 18, 841–846. Buzby, J.C., Roberts, T., Lin, C.-T., MacDonald, J., 1996. Agricultural Economic Report No. 741. Economic Research Service, U.S. Department of Agriculture, Washington, DC. Deng, L., Bao, L.L., Yang, Z.Y., Nie, L.H., Yao, S.Z., 1996. J. Microbiol. Methods 26, 197–203. Fratamico, P.M., Strobaugh, T.P., Medina, M.B., Gehring, A.G., 1998. Biotechnol. Tech. 12, 571–576. Gibas, C.J., Jambeck, P., Subramaniam, S., 2000. Methods-Companion Methods Enzymol. 20, 292–309. Homola, J., 2003. Anal. Bioanal. Chem. 377, 528–539. Homola, J., Dostalek, J., Chen, S.F., Rasooly, A., Jiang, S.Y., Yee, S.S., 2002. Int. J. Food Microbiol. 75, 61–69. Homola, J., Lu, H.B.B., Nenninger, G.G., Dostalek, J., Yee, S.S., 2001. Sensors Actuat. B-Chem. 76, 403–410. Ivnitski, D., Abdel-Hamid, I., Atanasov, P., Wilkins, E., 1999. Biosens. Bioelectron. 14, 599–624. Ladd, J., Boozer, C., Yu, Q.M., Chen, S.F., Homola, J., Jiang, S., 2004. Langmuir 20, 8090–8095. Libby, J.M., Wada, H.G., 1989. J. Clin. Microbiol. 27, 1456–1459. Mead, P.S., Slutsker, L., Dietz, V., McCaig, L.F., Bresee, J.S., Shapiro, C., Griffin, P.M., Tauxe, R.V., 1999. Emerg. Infect. Dis. 5, 607– 625. Nyquist-Battie, C., Frank, L.E., Lund, D., Lim, D.V., 2004. J. Food Prot. 67, 2756–2759. Plomer, M., Guilbault, G.G., Hock, B., 1992. Enzyme Microb. Tech. 14, 230–235. Qian, X., Metallo, S.J., Choi, I.S., Wu, H., Liang, M.N., Whitesides, G.M., 2002. Anal. Chem. 74, 1805–1810. Sapsford, K.E., Rasooly, A., Taitt, C.R., Ligler, F.S., 2004. Anal. Chem. 76, 433–440. Schneider, B.H., Edwards, J.G., Hartman, N.F., 1997. Clin. Chem. 43, 1757–1763. Taitt, C.R., Golden, J.P., Shubin, Y.S., Shriver-Lake, L.C., Sapsford, K.E., Rasooly, A., Ligler, F.S., 2004a. Microb. Ecol. 47, 175–185. Taitt, C.R., Shubin, Y.S., Angel, R., Ligler, F.S., 2004b. Appl. Environ. Microbiol. 70, 152–158. Taylor, A.D., Yu, Q.M., Chen, S.F., Homola, J., Jiang, S.Y., 2005. Sensors Actuat. B-Chem. 107, 202–208. Thomsen, V., Schatzlein, D., Mercuro, D., 2003. Spectroscopy 18, 112– 114. USDA/FSIS, 1998. Microbiology Laboratory Guidebook. Yu, Q.M., Chen, S.F., Taylor, A.D., Homola, J., Hock, B., Jiang, S.Y., 2005. Sensors Actuat. B-Chem. 107, 193–201.