ARTICLE pubs.acs.org/ac
Carbon Nanoparticles as Detection Labels in Antibody Microarrays. Detection of Genes Encoding Virulence Factors in Shiga Toxin-Producing Escherichia coli Patricia S. Noguera,†,‡ Geertruida A. Posthuma-Trumpie,† Marc van Tuil,† Fimme J. van der Wal,§ Albert de Boer,§ Antoine P. H. A. Moers,† and Aart van Amerongen*,†,|| †
)
Biomolecular Sensing & Diagnostics, Wageningen UR Food & Biobased Research, P.O. Box 17, 6700 AA Wageningen, The Netherlands ‡ Centro de Reconocimiento Molecular y Desarrollo Tecnologico, Departamento de Química, Universidad Politecnica de Valencia, Camino de Vera, s/n, 46020 Valencia, Spain § Division of Infection Biology, Central Veterinary Institute, part of Wageningen UR, P.O. Box 65, 8200 AB Lelystad, The Netherlands Laboratory of Organic Chemistry, Wageningen University, P.O. Box 8026, 6700 EG Wageningen, The Netherlands
bS Supporting Information ABSTRACT: The present study demonstrates that carbon nanoparticles (CNPs) can be used as labels in microarrays. CNPs were used in nucleic acid microarray immunoassays (NAMIAs) for the detection of different Shiga toxin-producing Escherichia coli (STEC) virulence factors: four genes specific for STEC (vt1, vt2, eae, and ehxA) and the gene for E. coli 16S (hui). Optimization was performed using a Box Behnken design, and the limit of detection for each virulence factor was established. Finally, this NAMIA using CNPs was tested with DNA from 48 field strains originating from cattle feces, and its performance was evaluated by comparing results with those achieved by the reference method q-PCR. All factors tested gave sensitivity and specificity values higher than 0.80 and efficiency values higher than 0.92. Kappa coefficients showed an almost perfect agreement (k > 0.8) between NAMIA and the reference method used for vt1, eae, and ehxA, and a perfect agreement (k = 1) for vt2 and hui. The excellent agreement between the developed NAMIA and q-PCR demonstrates that the proposed analytical procedure is indeed fit for purpose, i.e., it is valuable for fast screening of amplified genetic material such as E. coli virulence factors. This also proves the applicability of CNPs in microarrays.
I
n drug discovery and life science research both DNA and, increasingly, protein microarrays are crucial tools.1 Even though the application of microarrays as a diagnostic tool is very promising2 due to the high potential of protein microarrays, they are still not commonly used in the regular diagnostic field.3 There are several reasons for the limited presence of such microarrays: the lack of sufficient biological recognition elements (e.g., antibodies), their limited sensitivity and specificity, the lack of integrated systems that include fluid handling, sample preparation, and signal processing, and inacceptable background signals. As indicated by Hartmann and co-workers,4 there are many problems that still have to be overcome to develop validated in vitro diagnostic systems using protein microarrays. A problem that has to be solved in multiplexed protein microarrays is its automation, because this would increase performance, robustness, and reliability of the assays. To increase the applicability and reduce costs of protein microarrays, we investigated the use of carbon nanoparticles (CNPs) as signal labels in combination with a conventional flatbed scanner. The acquired image of the developed microarray is evaluated with image analysis software, and the pixel gray volume r 2011 American Chemical Society
data (PGV) of the spots generated by the CNPs is quantified. The use of PGV of carbon nanoparticles in data processing following digitization by a CCD camera and image analysis was first shown in 1994 when comparing a simple one-step lateral flow immunoassay (LFIA) with a radioimmunoassay.5 The high level of agreement between these techniques showed the possibilities of using this combination (CNP-image analysis) as a diagnostic tool. Moreover, in 2001 L€onnberg and Carlsson6 described the use of a conventional flatbed scanner to digitize carbon lines. In a recent literature survey conducted by FIND Diagnostics regarding the sensitivity of different LFIAs,7 the sensitivity of the carbon label was calculated to be in the low picomolar range, even when assays were judged by visual inspection.8,9 The position occupied by CNPs in the Gordon and Michels7 ranking list of nanoparticle sensitivity further underpins the potential use of these particles as signal labels, also in microarrays. Received: July 15, 2011 Accepted: September 21, 2011 Published: September 21, 2011 8531
dx.doi.org/10.1021/ac201823v | Anal. Chem. 2011, 83, 8531–8536
Analytical Chemistry A common approach to detect the presence/absence of pathogenic organisms is direct detection using antibodies;10 13 the accuracy of the results is linked to the quality of the antibodies used and is limited by availability of specific antibodies. Alternatively and often more sensitively, pathogens can be detected by using specifically amplified genetic material. If tagged primers are applied in the PCR procedure, the amplicons can be bound by anti-tag antibodies.14 In this way the result of the PCR is visualized by an antibody microarray. This not only permits the (simultaneous) detection of a wide range of pathogens15,16 but also the discrimination between pathogenic and nonpathogenic strains.17 To get proof of concept for the use of CNPs as signal labels in antibody microarrays, we studied an application in which the antigens consisted of double labeled DNA amplicons (tag/biotin) from different E. coli genes. The development of a microarray test that would detect different E. coli genes was considered of great importance because, even though many strains are harmless to humans, some E. coli strains can cause severe illnesses. Highly pathogenic strains are Shiga toxin-producing E. coli (STEC, also called verotoxigenic E. coli (VTEC) or enterohemorrhagic E. coli (EHEC)).18 A recent outbreak19,20 with a new serotype of EHEC in Germany and other West-European countries showed the risk of such an infection. Virulence factors of the classical STEC strains are cytotoxins (verotoxins) encoded by the genes vt1 or vt2, the outer membrane protein intimin encoded by the gene eae and involved in the attachment to intestine epithelial cells, and the enterohemolysin encoded by the gene ehxA, as reviewed.18 Preliminary studies to detect E. coli virulence factors in lateral flow format using CNPs have been published recently21 and showed that there is a limitation in the number of genes that can be simultaneously detected in a lateral flow test (five test lines = five genes). This limitation can be easily overcome using antibody microarrays because many spots can be printed on one microarray. The first attempt to use CNPs in miniarrays (nine spots) was very promising,22 allowing the detection of amplified DNA material in a single step. In this work, a proof of principle assay for the application of CNPs in nucleic acid microarray immunoassays (NAMIAs) was developed for the detection (presence or absence) of a combination of virulence factors in STEC (vt1, vt2, eae, and ehxA) and the gene coding for E. coli 16S (hui). The developed microarray was tested with DNA from 48 field strains originating from cattle feces, and the performance of the method was evaluated by comparing the results with those achieved by q-PCR.
’ EXPERIMENTAL SECTION Chemicals. H3BO3, NaN3, and Tween20 were purchased from Merck (Darmstadt, Germany) and Na2B4O7 3 10H2O and bovine serum albumin (BSA) from Sigma-Aldrich Chemie BV (Zwijndrecht, The Netherlands). Buffers. Phosphate-buffered saline (PBS): 0.01 M phosphate, 0.15 M NaCl, pH 7.4; borate buffer (BB): 100 mM (and 5 mM) pH 8.8 (from 100 mM solutions of H3BO3 and Na2B4O7 3 10H2O); washing buffer (WB): 5 mM BB, 1% (w/v) BSA); storage buffer (SB): 100 mM BB, 1% (w/v) BSA; and incubation buffer: 100 mM BB, 1% (w/v) BSA, 0.05% (v/v) Tween20. NaN3 was added to PBS and 100 mM BB to a final concentration of 0.02% (w/v). Antibodies. Polyclonal antibodies against digoxigenin (α-DIG) (Roche, Penzberg, Germany), 2,4-dinitrophenol (α-DNP), Texas Red (α-TxR) (Invitrogen Corporation, Camarillo, CA), and fluorescein isothiocyanate (α-FITC) (AbD-Serotech,
ARTICLE
Oxford, UK) were used. The α-FITC antibody was used against FAM (carboxyfluorescein) as label in the primer. Also a monoclonal antibody against cyanine dye 5 (α-Cy5) (Sigma-Aldrich Chemie BV, Zwijndrecht, The Netherlands) and a biotinylated antirabbit antibody (Pierce Biotechnology, Rockford, IL) were used. The latter was used as positive control. E. coli Strain. EDL 933 strain was used as control, selected for having all the genes coding for the targeted virulence factors. DNA isolation was performed as described,21 with 5 108 cfu/mL corresponding to a DNA concentration of 831 ng/μL. PCR. Primers used in the amplification of the genes coding for four STEC virulence factors (vt1, vt2, eae, and ehxA) and a 16S control specific for E. coli (hui) were used in a preceding study21 and were taken from the literature23,24 (cf. Table S-1 (Supporting Information) for sequences). Forward primers were labeled with specific tags, TxR (vt1), FAM (vt2), DIG (eae), DNP (ehxA), or Cy5 (hui) whereas all reverse primers were labeled with biotin (Eurogentec, Seraing, Belgium). To achieve an amplification reaction in the shortest time possible with economical equipment, PCR reactions were carried out with a PIKO Thermal cycler (Finnzymes, Espoo, Finland) using extra thin vessels, which allow an increase in heat transfer speed, and Phire Hot Start DNA Polymerase (Finnzymes, Espoo, Finland). DNA isolates were diluted 1:20 before amplification. All amplifications were carried out individually and performed as described21 using 0.25 mM dNTPs, 0.9 μM of each primer, 0.4 μL polymerase, 2 μL template, 1x Phire Reaction Buffer, and 1.5 mM MgCl2 in a total volume of 20 μL and the same 30 min amplification program. Positive and negative controls were included in all runs. Amplified DNA was used directly or stored at 20 °C until use. q-PCR. q-PCR was used as reference method in this work using the same primer sequences and reaction conditions as described for the nucleic acid lateral flow immunoassay (NALFIA) test format PCR.21 Carbon Conjugate. Carbon nanoparticles (CNPs) (Special Black SB4, Evonik Degussa, AG, Frankfurt, Germany) were labeled with neutravidin (NA) biotin-binding protein (Pierce Biotechnology). The labeling procedure is very simple and involves the incubation of CNPs with NA and the removal of unbound protein by washing.21 The final homogeneous suspension contains 0.2% (w/v) carbon conjugate. Slide Manufacturing. Nitrocellulose (NC) FAST16 slides (5 5 mm pads) (Whatman’s-Hertogenbosch, The Netherlands) were used. PBS dilutions of antibodies (50 800 μg/mL) were printed on the NC pads using a noncontact S3 spotter (Scienion AG, Berlin, Germany); two drops of about 350 pL were printed for every spot. Slides were dried overnight at 37 °C and used immediately or stored desiccated at room temperature until use. NAMIA Optimization. To determine the optimum amount of antibody printed, NC pads were printed with serial dilutions of each of the antibodies and incubated with 1 μL of each specifically labeled PCR product, 5 μL of the 0.2% (w/v) carbon conjugate, and 69 μL of incubation buffer. To optimize the incubation step (i.e., optimum amount of reagents and incubation time), the response surface methodology (RSM) was used. Of the different possibilities,25,26 the Box Behnken design was applied; Statgraphics Centurion XVI (StatPoint Technologies, Warrenton, VA) was used. All optimizations were performed using EDL 933 DNA diluted 1:20 as PCR template, and the optimal conditions were used in all following tests. Cross-Reactions and Analytical Performance. Cross-reactions were studied by testing all amplified genes individually, and 8532
dx.doi.org/10.1021/ac201823v |Anal. Chem. 2011, 83, 8531–8536
Analytical Chemistry
ARTICLE
Figure 1. (From left to right) Scheme of the detection strategy: a double-labeled amplicon (tag/biotin) is sandwiched between an immobilized tagspecific antibody and a carbon nanoparticle coated with (neutr)avidin, distribution of the antibodies printed on the microarray, and a typical example of four microarrays for ehxA, eae, hui, and vt1 (real size of one microarray is 3 mm 3 mm).
for the evaluation of the analytical performance of the NAMIA slides developed, DNA of E. coli strain EDL 933 was subjected to 10-fold serial dilutions and divided into aliquots. Aliquots were used in the PCR amplification procedure, and the resulting amplicons were analyzed in three independent experiments. For comparison, three aliquots were analyzed using the reference method (q-PCR). Evaluation of the Results. Microarrays were scanned using a conventional flatbed scanner Epson 3200 Photo scanner (Seiko Epson, Nagano, Japan). The PGV of positive spots were obtained using image analysis software (TotalLab, Nonlinear Dynamics, Newcastle upon Tyne, UK). Microsoft Excel and SigmaPlot 11 (Systat Software, Inc., San Jose, CA) were used for subsequent data analysis. Field Samples. To test field samples, DNA was obtained from 48 E. coli isolates originating from cattle feces.27 The isolated DNA was subjected (in duplicate) to the 30 min PCR amplification, and all factors amplified were mixed and used in a single well of the microarray slide. The result achieved for each single sample (i.e., the presence or absence of a virulence factor) was compared to the result achieved by q-PCR. When results were identical, it was considered as a true positive (or negative), and when they differed it was considered a false positive (or negative). With these values, reliability parameters such as sensitivity, specificity, positive (and negative) predictive values, efficiency, and the kappa coefficient (k) were calculated to evaluate the performance of the NAMIA slides.28 31 The mathematical equations used to calculate these parameters are given in Table S-2 (Supporting Information), and they were attained using an Excel spreadsheet specifically developed for evaluating diagnostic tests.32
’ RESULTS AND DISCUSSION The principle of the detection is depicted in Figure 1 (left) where double-labeled amplicons (specific tag/biotin) were sandwiched between specific tag antibodies immobilized on NC covered slides and CNPs functionalized with NA. The selected pattern printed in this work is depicted in Figure 1 (middle). This distribution makes use of less than 40% of the total NC microarray surface, and it has several advantages: there are five spots (repetitions) per antibody, and the orientation is easily recognizable, allowing visual verification of the quality of the microarray. Moreover, a visually positive/negative response
could be easily achieved directly or by means of a magnifying glass/microscope. NAMIA slides were printed with serial dilutions of each of the antibodies and were incubated with 1 μL of each PCR product and 5 μL of the 0.2% carbon conjugate suspension. As expected, spot intensity increased with increasing amounts of antibody until a maximum plateau value was reached; (not shown). For each antibody (and the control), the selected optimal concentration to be printed on the NC slides was 200 μg/mL for α-TxR, α-DIG, α-DNP, and biotinylated IgG (positive control), and 50 μg/mL for α-FITC, and 800 μg/mL for α-Cy5. To perform the optimization, the Box Behnken design was selected. This design is characterized by adjusting factor levels to three equally space intervals (coded 0, (1), and it has the advantage that it does not contain combinations where all factors are at their highest or lowest levels. Factors studied were amount of PCR product (1, 3, and 5 μL of each amplified gene), amount of carbon (2, 6, and 10 μL), and incubation time (0.5, 2, and 3.5 h). (Table S-3 (Supporting Information) shows in detail the experimental design applied). For all amplicons, two of the factors tested accounted for the ability to predict the intensity (P < 0.05): amount of carbon conjugate and incubation time; the general expression achieved was intensity (PGV) = A + B[amount of carbon] + C[time] (where A, B, and C are constants; values given in Table S-4, Supporting Information). These results are easily visualized in three-dimensional graphs; as shown in Figure 2, variations in the amount of PCR product do not significantly affect the intensity values, while there is a significant effect of incubation time and the amount of carbon. The final step of RSM is to find the optimal set of experimental parameters that produce the most adequate results. Because this methodology was applied to all virulence factors, a compromise solution had to be reached, having in mind that a good signal had to be achieved (around 105 PGV) using a reasonable amount of reagents (no extreme values). Therefore, the selected final conditions for the incubation reaction were a mixture of 1 μL of each PCR product (sample), 5 μL of carbon NA conjugate, and 65 μL of incubation buffer incubated for 1 h at room temperature. When microarrays were used with individual genes, no cross-reactivity was found (see Figure 1, right). Following PCR, a total of three separate NAMIAs were performed on each of the E. coli dilutions. Maximum intensity values achieved were around 104 PGV. PGV data was transformed 8533
dx.doi.org/10.1021/ac201823v |Anal. Chem. 2011, 83, 8531–8536
Analytical Chemistry
ARTICLE
Figure 2. Surface models of microarrays based on the Box Behnken design and developed for DNP (left) and DIG (right). Left: Signal intensity vs amount of carbon and PCR product. Right: Signal intensity vs amount of carbon and time.
Table 2. Comparison of Results between NAMIA and q-PCR Achieved with Isolates from Cattle Fecesa reference method (q-PCR) vt1
vt2
eae
ehxA
hui
pos neg pos neg pos neg pos neg pos neg
a
Figure 3. NAMIA performance for the different E. coli virulence factors. Average ( SD (average of three microarrays with five spots).
Table 1. Detection Limit Achieved by NAMIA and q-PCR for Different E. coli Virulence Factors detection limit (cfu/mL) TxR/vt1
FAM/vt2
DIG/eae
DNP/ehxA
Cy5/hui 1.7 104 8.4 101
NAMIA
3.3 10
2.2 10
5.3 10
4.0 10
5
q-PCR
8.4 10
5.9 10
8.4 10
8.4 10
3
4 2
4 5
4 3
to percentage of the maximum signal, showing differences in the performance achieved for each amplified gene (Figure 3). For all systems, the regression values for the adjustment of these data to a four-parameter logistic curve was larger than 0.95, and the LOD for most of the virulence factors (Table 1) was around 104 cfu/ mL (0.09 ng/μL), which would correspond to 20 cfu in the PCR tube. It is noteworthy that when comparing sensitivities achieved in microarrays and lateral flow strips21 using the same CNPs, microarrays are equally (same order of magnitude, vt2 and ehxA) or more sensitive (1 order of magnitude lower, vt1, eae, hui) than the one achieved in lateral flow. Finally, when LODs of NAMIA were compared with those of q-PCR, in most cases the LOD
NAMIA pos
17
0
30
0
19
0
31
1
47
0
neg
3
28
0
18
4
25
1
15
0
1
pos: number of positive samples; neg: number of negative samples.
achieved by q-PCR was 6 to 200 times lower, i.e., more sensitive, than for NAMIA with, as the sole exception, vt2; in that case, the LOD achieved in q-PCR was more than 25 times higher, i.e., less sensitive (Table 1). DNA of each of the 48 cattle feces isolates was amplified by PCR, and the resulting amplicons were mixed and used with NAMIA tests. The qualitative result achieved for each sample (presence/absence) was compared to the result achieved by q-PCR, and they are summarized in Table 2. Both methods considered one of the 48 samples as negative for all genes, indicating that no E. coli was present or that DNA was of low quality; they also classified alike all samples for vt2 and hui. Reliability parameters of the NAMIA developed are indicated in Table 3. For all factors tested, sensitivity and specificity values were very high (higher than 0.80 and in many cases equal to 1.00). In most cases, positive and negative predictive values (i.e., how well a sample is classified) were higher than 0.90. Moreover, efficiency values (i.e., correct answers) were 0.92 or higher, meaning that more than 90% of the answers were the same as the reference procedure. The value of the kappa coefficients indicates that there is an almost perfect agreement (k > 0.8) between NAMIA and the reference method used (q-PCR) for vt1, eae, and ehxA, and a perfect agreement (k = 1) for vt2 and hui. This excellent agreement between the microarray developed using CNPs and the reference system demonstrates that the proposed analytical procedure is indeed fit for purpose, i.e., fast screening of amplified genetic material such as for E. coli virulence factors. This STEC-specific NAMIA could be even further improved by developing a multiplex PCR amplification for the five genes. 8534
dx.doi.org/10.1021/ac201823v |Anal. Chem. 2011, 83, 8531–8536
Analytical Chemistry
ARTICLE
Table 3. Reliability Parameters for Different E. coli Virulence Factors in NAMIAa vt1
vt2
eae
ehxA
hui
sensitivity
0.85
1.00
0.83
0.97
1.00
specificity positive predictive value
1.00 1.00
1.00 1.00
1.00 1.00
0.94 0.95
1.00 1.00
negative predictive value
0.90
1.00
0.86
0.94
1.00
efficiency
0.94
1.00
0.92
0.96
1.00
kappa coefficient (k)
0.87
1.00
0.83
0.91
1.00
a
Field samples (48) were tested in NAMIA, and the results were compared to those obtained with q-PCR.
When NAMIA and q-PCR are compared, the latter system could have the advantage that (in most cases) the LOD is lower than NAMIA and that the identification is performed in one step. However, it has also several drawbacks when compared to NAMIA: the first one is the price of the equipment; the second one is that specific knowledge is needed to understand and interpret the results; the third one is the need for a laboratory setting. When the number of factors that can be analyzed are compared, only a limited number of fluorescent dyes are available for q-PCR. Multiplexing is complicated by the spectral overlap of the various dyes, allowing only a limited combination of dyes in one run (generally up to six). In our recent work using NALFIA,21 the same restrictions about the number of factors to be analyzed holds, although the NALFIA method does not require specialized personnel and expensive equipment and is faster. However, even though in this work NAMIA slides have been developed for the detection of five factors, these can be easily increased by introducing other tag-antibody combinations to generate a comprehensive assay that also includes relevant genes of other pathogenic E. coli, e.g., the outbreak strain responsible for many cases of hemolytic uraemic syndrome and accompanying fatalities in Germany, May/June 2011.33 The final number of genes that can be detected is only limited by the number of available tagantibody combinations and the growing chance of cross-reactive and background signals upon substantially increasing (>25) the number of discrete target spots in the assay. The test, as developed here, can be used in any food chain that is appropriate to find the source of such an outbreak. A swab of potentially contaminated food, if necessary followed by a short enrichment step, is sufficient for this test. The only step needed to adapt would be the amplification, where the specific primers labeled with discriminating tags would have to be used. If specialized software dedicated to manage, quantify, and derive results from the microarray experiments would be used, the speed of analyzing the microarrays would be greatly improved. Initial experience with such software (results not shown) reveals that the research tool as presented in this work can be easily developed into a diagnostic tool that could be commercially applicable.
’ CONCLUSIONS There are many advantages of using carbon nanoparticles as labels in diagnostic assays, such as price, sensitivity, and their very high signal-to-noise ratio.7 Moreover, the ease of the labeling procedure of proteins onto CNPs is another remarkable advantage. The incubation reaction between the CNPs and the label (here NA) can be performed under simple laboratory settings. In this work it has been demonstrated that the use of CNPs in
microarrays (with a preceding 30 min PCR reaction step) is a system that is feasible and reliable. We showed the specific identification of four E. coli virulence genes (vt1, vt2, eae, and ehxA) of classical STEC strains and a marker specific for E. coli (hui) in approximately 90 min (once DNA is isolated). The results of this approach, a 30 min amplification followed by a 1 h incubation step on the microarray, can be obtained by automated data processing following the digitization of the spots by flatbed scanning or digital photography (10 min total processing time). The system developed allows manufacturing of printed arrays with multiple anti-tag antibodies that can be used with only one set of CNPs (as a neutravidin conjugate), showing that the proposed analytical procedure is indeed fit for purpose and demonstrating the applicability of CNPs for microarrays. The system is generic because the specificity of the assay is in the combination of discriminating tags with specific primers; by changing primer sequences to amplify genes of other targets (e.g., other pathogenic microorganisms), a NAMIA with another specificity can be readily developed.
’ ASSOCIATED CONTENT
bS
Supporting Information. Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*E-mail:
[email protected].
’ ACKNOWLEDGMENT This work was partially supported by the Generalitat Valenciana (BEST/2009/026), the Universidad Politecnica de Valencia (PAID-00-09-2837), and the Dutch Ministry of Agriculture, Nature and Food Quality (Strategic research program Food Safety, Monitoring and Detection KB-06-005). The authors thank Dr. Eva Møller Nielsen at the Danish Veterinary Institute (Copenhagen, Denmark) for providing E. coli control strains and Dr. Lutz Geue (FriedrichLoeffler-Institut, Wusterhausen, Germany) and Dr. D€ orte D€opfer (School of Veterinary Medicine, University of Wisconsin, Madison, WI) for field isolates. ’ REFERENCES (1) Timlin, J. A. In Methods in Enzymology; Kimmel, A., Brian, O., Eds.; Academic Press: Amsterdam, The Netherlands, 2006; Vol. 411, pp 79 98. (2) Uttamchandani, M.; Neo, J. L.; Ong, B. N. Z.; Moochhala, S. Trends Biotechnol. 2009, 27, 53–61. (3) Dieterle, F.; Marrer, E. Anal. Bioanal. Chem. 2008, 390, 141–154. (4) Hartmann, M.; Roeraade, J.; Stoll, D.; Templin, M. F.; Joos, T. O. Anal. Bioanal. Chem. 2009, 393, 1407–1416. (5) van Amerongen, A.; van Loon, D.; Berendsen, L. B. J. M.; Wichers, J. H. Clin. Chim. Acta 1994, 229, 67–75. (6) L€ onnberg, M.; Carlsson, J. Anal. Biochem. 2001, 293, 224–231. (7) Gordon, J.; Michel, G. Clin. Chem. 2008, 54, 1250–1251. (8) Koets, M.; Sander, I.; Bogdanovic, J.; Doekes, G.; van Amerongen, A. J. Environ. Monit. 2006, 8, 942–946. (9) van Dam, G. J.; Wichers, J. H.; Ferreira, T. M. F.; Ghati, D.; van Amerongen, A.; Deelder, A. M. J. Clin. Microbiol. 2004, 42, 5458–5461. (10) Anjum, M. F.; Tucker, J. D.; Sprigings, K. A.; Woodward, M. J.; Ehricht, R. Clin. Vaccine Immunol. 2006, 13, 561–567. (11) Delehanty, J. B.; Ligler, F. S. Anal. Chem. 2002, 74, 5681–5687. 8535
dx.doi.org/10.1021/ac201823v |Anal. Chem. 2011, 83, 8531–8536
Analytical Chemistry
ARTICLE
(12) Gehring, A. G.; Albin, D. M.; Bhunia, A. K.; Reed, S. A.; Tu, S. I.; Uknalis, J. Anal. Chem. 2006, 78, 6601–6607. (13) Gehring, A. G.; Albin, D. M.; Reed, S. A.; Tu, S. I.; Brewster, J. D. Anal. Bioanal. Chem. 2008, 391, 497–506. (14) Kostrzynska, M.; Bachand, A. Can. J. Microbiol. 2006, 52, 1–8. (15) Sergeev, N.; Distler, M.; Courtney, S.; Al-Khaldi, S. F.; Volokhov, D.; Chizhikov, V.; Rasooly, A. Biosens. Bioelectron. 2004, 20, 684–698. (16) Wang, L.; Li, P. C. J. Agric. Food Chem. 2007, 55, 10509–10516. (17) Wu, C. F.; Valdes, J. J.; Bentley, W. E.; Sekowski, J. W. Biosens. Bioelectron. 2003, 19, 1–8. (18) Nataro, J. P.; Kaper, J. B. Clin. Microbiol. Rev. 1998, 11, 142–201. (19) Askar, M.; Faber, M. S.; Frank, C.; Bernard, H.; Gilsdorf, A.; Fruth, A.; Prager, R.; H€ohle, M.; Suess, T.; Wadl, M.; Krause, G.; Stark, K.; Werber, D. Eurosurveillance 2011, 16, pii=19883. (20) Frank, C.; Faber, M. S.; Askar, M.; Bernard, H.; Fruth, A.; Gilsdorf, A.; H€ohle, M.; Karch, H.; Krause, G.; Prager, R.; Spode, A.; Stark, K.; Werber, D. Eurosurveillance 2011, 16, pii=19878. (21) Noguera, P.; Posthuma-Trumpie, G.; van Tuil, M.; van der Wal, F.; de Boer, A.; Moers, A.; van Amerongen, A. Anal. Bioanal. Chem. 2011, 399, 831–838. (22) van Amerongen, A.; Koets, M. In Rapid methods for biological and chemical contaminants in food and feed, 1st ed.; Wageningen Academic Publishers: Wageningen, The Netherlands, 2005; pp 105 126. (23) Huijsdens, X. W.; Linskens, R. K.; Mak, M.; Meuwissen, S. G. M.; Vandenbroucke-Grauls, C. M. J. E.; Savelkoul, P. H. M. J. Clin. Microbiol. 2002, 40, 4423–4427. (24) Nielsen, E. M.; Andersen, M. T. J. Clin. Microbiol. 2003, 41, 2884–2893. (25) Bezerra, M. A.; Santelli, R. E.; Oliveira, E. P.; Villar, L. S.; Escaleira, L. A. Talanta 2008, 76, 965–977. (26) Ferreira, S. L.; Bruns, R. E.; Ferreira, H. S.; Matos, G. D.; David, J. M.; Brandao, G. C.; da Silva, E. G.; Portugal, L. A.; dos Reis, P. S.; Souza, A. S.; dos Santos, W. N. Anal. Chim. Acta 2007, 597, 179–186. (27) Geue, L.; Segura-Alvarez, M.; Conraths, F. J.; Kuczius, T.; Bockemuhl, J.; Karch, H.; Gallien, P. Epidemiol. Infect. 2002, 129, 173–185. (28) Eurachem/CITAC. In Eurachem/CITAC Conference Discussion Paper, 2003 ed., 2003; Vol. September, p 22. (29) Ellison, S. L. R.; Fearn, T. TRAC, Trends Anal. Chem. 2005, 24, 468–476. (30) Gilchrist, J. M. J. Clin. Epidemiol. 2009, 62, 1045–1053. (31) Landis, J. R.; Koch, G. G. Biometrics 1977, 33, 159–174. (32) Mackinnon, A. Comput. Biol. Med. 2000, 30, 127–134. (33) EHEC O104:H4, Robert Koch Institute, 2011. URL: http:// www.rki.de/cln_178/nn_217400/EN/Home/EHECO104,templateId= raw,property=publicationFile.pdf/EHECO104.pdf.
8536
dx.doi.org/10.1021/ac201823v |Anal. Chem. 2011, 83, 8531–8536
