Detecting and genotyping Escherichia coli O157:H7 using multiplexed PCR and nucleic acid microarrays

International Journal of Food Microbiology 67 Ž2001. 71–80 www.elsevier.nlrlocaterijfoodmicro

Detecting and genotyping Escherichia coli O157:H7 using multiplexed PCR and nucleic acid microarrays Douglas R. Call ) , Fred J. Brockman, Darrell P. Chandler EnÕironmental Microbiology, Pacific Northwest National Laboratory, P.O. Box 999, MS P7-50, Richland, WA 99352 USA Received 27 June 2000; received in revised form 1 September 2000; accepted 17 December 2000

Abstract Rapid detection and characterization of food borne pathogens such as Escherichia coli O157:H7 is crucial for epidemiological investigations and food safety surveillance. As an alternative to conventional technologies, we examined the sensitivity and specificity of nucleic acid microarrays for detecting and genotyping E. coli O157:H7. The array was composed of oligonucleotide probes Ž25–30 mer. complementary to four virulence loci Žintimin, Shiga-like toxins I and II, and hemolysin A.. Target DNA was amplified from whole cells or from purified DNA via single or multiplexed polymerase chain reaction ŽPCR., and PCR products were hybridized to the array without further modification or purification. The array was 32-fold more sensitive than gel electrophoresis and capable of detecting amplification products from - 1 cell equivalent of genomic DNA Ž1 fg.. Immunomagnetic capture, PCR and a microarray were subsequently used to detect 55 CFU mly1 Ž E. coli O157:H7. from chicken rinsate without the aid of pre-enrichment. Four isolates of E. coli O157:H7 and one isolate of O91:H2, for which genotypic data were available, were unambiguously genotyped with this array. Glass-based microarrays are relatively simple to construct and provide a rapid and sensitive means to detect multiplexed PCR products; the system is amenable to automation. q 2001 SOCIETY. Published by Elsevier Science B.V. All rights reserved. Keywords: DNA microarray; Enterohemorrhagic; Food safety; Shiga-like toxin; Intimin; Hemolysin

1. Introduction Enterohemorrhagic ŽEHEC. bacteria, including Escherichia coli O157:H7, are capable of causing significant illness and represent a serious public health threat worldwide ŽBell et al., 1994; Allerberger et al., 1996; Stewart and Flint, 1999.. Numer-

) Corresponding author. Department of Veterinary Microbiology and Pathology, Washington State University, P.O. Box 647040, Pullman, WA 99164-7040, USA. Tel.: q1-509-335-6030; fax: q1-509-335-8529. E-mail address: [email protected] ŽD.R. Call..

ous molecular tools have been employed to detect and classify EHEC bacteria for epidemiological investigations, clinical diagnoses, or routine surveillance of food products ŽJohnson et al., 1995; Brewster and Mazenko, 1998; Fratamico and Strobaugh, 1998; Oberst et al., 1998; Seo et al., 1998; Tortorello et al., 1998; Pyle et al., 1999; Sharma et al., 1999., with a significant number relying upon multiplexed polymerase chain reaction ŽPCR . techniques ŽFratamico et al., 1995; Gannon et al., 1997; Louie et al., 1998; Paton and Paton, 1998; Fagan et al., 1999.. Recently developed fluorogenic exonuclease assays Že.g. TaqMane PCR., in particular, hold tremendous promise for rapid, specific and automated detection

0168-1605r01r$ - see front matter q 2001 SOCIETY. Published by Elsevier Science B.V. All rights reserved. PII: S 0 1 6 8 - 1 6 0 5 Ž 0 1 . 0 0 4 3 7 - 8

72

D.R. Call et al.r International Journal of Food Microbiology 67 (2001) 71–80

of food-borne pathogens ŽOberst et al., 1998; Sharma et al., 1999.. PCR is a powerful analytical tool, but conventional detection methods impose some important limitations on the overall efficacy or practical utility of PCR techniques. For example, with food safety applications most of the aforementioned methods still require a lengthy sample pre-enrichment step Ž6–48 h.. This practically negates the benefit of ‘fast’ PCR Žor other. detection technologies and precludes true automation at the point of use. PCR Žconventional or TaqMane. amplifies only a limited portion of the genome, and has practical limits to the number of targets that can be simultaneously amplified. Gel electrophoresis is typically used for post-PCR detection, but it is cumbersome, fairly insensitive, and difficult to automate. TaqMane PCR improves overall specificity, detection limits and confidence of nucleic acid assays, but is even more restricted in its multiplex capabilities due to overlapping spectra Žexcitation or emission. of common fluors. Hence, there is a continued need for improved sample preparation, detection methods and detection systems to accommodate the stringent demands of pathogen surveillance in food safety applications. Nucleic acid microarrays represent the latest development in multi-gene detection technology ŽSchena, 2000.. Well known for their applications in whole-genome expression profiling ŽLockhart et al., 1996; Heller et al., 1997., microarrays can also be used to detect and identify multiplexed PCR products ŽGuo et al., 1994.. In the former case, microarrays are constructed from large cDNAs or genes ordered on a solid support Že.g. glass slide., with the intent of identifying complementary nucleic acid AtargetsB from the sample. In the latter case, microarrays are constructed from short oligonucleotides for the purpose of sequencing, identifying single nucleotide polymorphisms or minor sequence variants in a target population. Within the context of food safety and pathogen surveillance, the advantage of microarrays over conventional detection techniques lies in the ability to simultaneously detect the presence of distinct PCR products based on interproduct DNA sequence differences. The objectives of this study were to construct and evaluate a lowdensity microarray for efficient detection and genotyping of E. coli O157:H7 following single or multi-

plexed PCR reactions. We also investigated the possibility of coupling immunomagnetic capture with PCR and microarrays as a precursor to automated sample processing in the absence of sample pre-enrichment ŽChandler et al., 2000..

2. Materials and methods 2.1. PCR templates All bacterial isolates were obtained from the American Type Culture Collection ŽATCC, Manassas, VA, USA. and were maintained in trypticase soy broth and agar ŽLife Technologies, Rockville, MD, USA.. Unless otherwise indicated, genomic DNA was prepared from cell culture using a standard phenol:chloroform extraction ŽSambrook et al., 1989.. Five strains of E. coli O157:H7 were selected, of which ATCC 43894 was previously described as having genes for both Shiga-like toxins I and II Ž stx1rstx2.. Other strains had alternative genotypes for the Shiga-like toxins ŽATCC 43888, yry; ATCC 43889, yrstx2; ATCC 43890, stx1ry .. The phenotype for ATCC 51657 included resistance for tetracycline and we used this strain to generate DNA template for most PCR reactions. One non-O157 strain ŽATCC 51435, O91:H2. was previously shown to have the gene for Shiga-like toxin II Žyrstx2.. Data on genotypes for Shiga-like toxins was provided by ATCC. A negative control strain ŽDH5a; Life Technologies, Gaithersburg, MD. was also tested with the array. 2.2. Probes and targets All oligonucleotides were purchased from Midland Certified Reagent ŽMidland, TX, USA.. Most microarray probes were purchased with 3X or 5X amine modifications, although our method for attaching probes to glass slides does not require these modifications ŽCall et al., 2001.. PCR primers and oligonucleotide targets were biotinylated Ž5X . and purified by gel filtration. Although all primers were biotinylated for our experiments, the microarray probes were designed to capture the forward strand of the PCR product. Consequently, only the forward primer required biotin modification.

D.R. Call et al.r International Journal of Food Microbiology 67 (2001) 71–80

A 36-mer oligonucleotide ŽQC. and complementary sequence were adapted from Lamture et al. Ž1994. ŽTable 1. for use as a positive hybridization control for hybridization and detection. Probes and primers targeting specific virulence loci included a probe for the promoter region of the attachment and effacing gene Ž eaeA. of E. coli O157:H7 ŽOberst et al., 1998.. This sequence Ž25-mer. was originally designed for TaqMane PCR ŽPerkin Elmer, Foster City, CA, USA., with the authors demonstrating presumptive specificity for O157:H7. In some cases, we printed a biotinylated complement to the eaeA probe as a positive control for the ELF signal generation system Žbelow.. Additional probes and primers were designed for stx1, stx 2, and for hemolysin A Ž hlyA. ŽTable 1.. PCR products were limited to - 105 bp in total length, both to maximize PCR efficiency and the final molar concentration of PCR products. In most cases, final PCR conditions were 1 = PCR buffer and 1–2 units HotStarTaqe ŽQiagen, Valencia, CA. with 4 mM MgCl 2 , 200 nM each dNTP, 300 nM each primer and 50–250 ng DNA in a total volume of 50 ml. Three cycle PCR was used to generate products Ž958, 15 s denature; 588, 45 s annealing;

73

728, 45 s extension; 35–40 cycles; Perkin Elmer 9600 thermal cycler.. Identical reaction conditions were used for both single and multiple product amplification Žfour loci.. For most experiments, PCR products were hybridized to arrays without further purification or concentration. 2.3. Array construction Twelve-well, Teflon masked slides ŽErie Scientific, Portsmouth, NH, USA. were washed with 2% Micro w cleanser ŽInternational Products, Burlington, NJ, USA. followed by successive 15-min acid baths Ž3 N HCl and 3 N H 2 SO4 .. Slides were then briefly washed in deionized H 2 O followed by drying with compressed nitrogen. A subset of acid washed slides were derivatised with 2% 3-Glycidoxpropyltrimethoxysilane Žepoxy-silane; Aldrich, Milwaukee, WI, USA. in methanol for 15 min. These slides were then rinsed in methanol and dried before printing with an unmodified eaeA probe. These latter slides were used for assessing template sensitivity and for hybridizing PCR products from immunomagnetic capture experiments Žsee below.. Regardless of the surface preparation, oligonucleotide probes were di-

Table 1 PCR primers and microarray probes used for the pathogen array Locusa

Oligomer b

Sequence 5’–3’

Annealing ŽC8.

eaeA

Forward Reverse Probe Forward Reverse Probe Forward Reverse Probe Forward Reverse Probe Target Probe

CAA-TTT-TTC-AGG-GAA-TAA-CAT-TG AAA-GTT-CAG-ATC-TTG-ATG-ACA-TTG TCA-AGA-GTT-GCC-CAT-CCT-GCA-GCA-A AAG-CCG-GAA-CAG-TTC-TCT-CAG-C CTC-CTT-CCC-GTT-GTT-TTC-TCA-G TTC-ATC-AAG-AGC-CAT-GCC-TGA-TAA-AGC-AAT TCT-TAT-CTG-GAT-TTA-ATG-TCG-C TCA-GCT-GTC-ACA-GTA-ACA-AAC-C AAC-ATC-GCT-CTT-GCC-ACA-GAC-TGC-GTC-AGT TTA-TAC-CAC-TCT-GCA-ACG-TGT-C AAC-TCC-ATT-AAC-GCC-AGA-TA CCA-GTG-AGT-GAC-GAC-TGA-TTT-GCA-TTC-CGG CCA-CCA-CCC-AAC-CCC-ACC-ACC-ACA-CCA-CCA-CCA-CAA TTG-TGG-TGG-TGG-TGT-GGT-GGT-GGG-GTT-GGG-TGG-TGG

60.0 64.0 76.0 68.0 66.0 90.0 60.0 64.0 98.0 64.0 56.0 98.0 NA NA

hlyA

stx1

stx 2

QC a

Product Žbp.

Genbank c

104

U32312

96

X94129

89

M19437

96

Z37725

Genetic locus that is targeted by the described PCR primers and probes: eaeA Žintimin., hlyA Žhemolysin A., Shiga-like toxin I Ž stx1. and Shiga-like toxin II Ž stx 2.. QC is a oligonucleotide sequence from Lamture et al. Ž1994.. X b Forward and reverse refer to the orientation of the PCR primers. Forward primers must have a 5 biotin to be compatible with the X X antisense probes described in this table. Probes require a 3 or 5 amine modification if they are to be covalently bound to an epoxy-silane surface. c Accession numbers for Genbank sequences that were used to design primers and probes.

74

D.R. Call et al.r International Journal of Food Microbiology 67 (2001) 71–80

luted in print buffer Ž50 mM NaOH, 0.01% sodium dodecyl sulfate. to a final concentration of 80–100 mM and spotted in triplicate onto prepared slides using an Affymetrix 417 Arrayer ŽSanta Clara, CA, USA.. Each spot was 150 mm in size. Printed slides were baked 30 min at 1308C in a vacuum oven and stored long term at 48C. 2.4. Hybridization and detection Five microliters of PCR amplicons were added to 31 ml hybridization buffer Ž150 mM Na–citrate; 5 = Denhardt’s w0.001% Ficoll, 0.001% polyvinylpyrrolidone, 0.001% bovine serum albuminx., heat denatured Ž988C for 2 min. and rapidly chilled to 48C. In some cases, samples were spiked with QC targets Ž0.57 mM. before denaturing. PCR products were incubated 2 h within individual wells on Teflon masked slides Žeach well defined an independent microarray. and all incubations were either at 238C, 378C, or 508C in a humidified chamber, as indicated in the appropriate figure legends. Incubation temperatures were arbitrarily selected for convenience Ž238C., for accelerated reaction kinetics Ž378C. or to illustrate compatibility with increased hybridization stringency Ž508C.. After this incubation, hybridization solution was aspirated and slides were washed in 1 = SSC Ž150 mM NaCl, 15 mM Na–citrate, pH 7.0.. Streptavidin-alkaline phosphatase ŽAMDEX, Amersham, Piscataway, NJ, USA. was incubated Ž30 min. on the arrays at 1:500 dilution in hybridization buffer. Slides were then rinsed in 1 = ELF wash buffer A ŽELF-97 w mRNA in situ hybridization kit; Molecular Probes, Eugene, OR, USA. and ELF-97 w substrate was incubated on the slides for 60 min Ž1:100 ELF developing buffer C.. Slides received a final wash Ž150 mM NaCl, 0.1% Tween-20, 100 mM Tris HCl, pH 8.0. and were rinsed in deionized water. After air drying, slides were illuminated with UV Ž290–365 nm. and fluorescent emissions Ž520 q nm. were captured as optical density units ŽOD. with a Fluor-Se MultiImager ŽBio-Rad, Hercules, CA, USA.. The imager was equipped with a 28–200 mm DL Hyperzoom macro lens ŽSigma, Rodermark, Ger¨ many. that was fitted with a q1 close-up lens. Images were quantified with Phoretix array software ŽVer. 1.00, Phoretix International, Newcastle, UK..

For statistical purposes, each well was considered an independent replicate. Average ODs were calculated from three replicate spots in each well, but standard errors were expressed from the average of independent wells. 2.5. Array sensitiÕity We evaluated the sensitivity of the array for detecting O157:H7 based on both total copy number applied to the microarray, and initial template concentration in the PCR reaction. For the former, 20 PCR reactions for the eaeA locus were pooled, ethanol precipitated and quantified by spectrophotometry ŽSambrook et al., 1989.. Serial dilutions of PCR products from 2.5 = 10 12 to 2.4 = 10 9 copies Ž119 nM to 114 pM. were then hybridized to the array in standard hybridization buffer. Replicate dilutions were evaluated with 3% agarose ŽBioWhittaker Molecular Applications, Rockland, ME. containing 100 ng mly1 ethidium bromide. For assay or process-level sensitivity Žfrom PCR through detection., DNA extracted from ATCC 51657 was quantified by spectrophotometry and PCR reactions were performed on serial dilutions of DNA template Ž100 ng to 1 fg.. After PCR, amplification products were augmented with 7 ml 50 = Denhardt’s, 3.5 ml 20 = SSC and 9.5 ml H 2 O. After heat denaturing, 35 ml of this mixture was hybridized on each of two wells while at 508C Ž2 h.. Streptavidin and ELF-97 w incubations were at 378C Ž30 min and 1 h, respectively.. 2.6. Chicken rinsate Store-bought chicken breast was rinsed for 5 min in 20 ml Butterfield’s solution Ž250 mM KH 2 PO4 , 8 mM Na 2 SO 3 . per 100 g chicken breast. Overnight cultures of E. coli O157:H7 ŽATCC 51657. were washed in PBS-Tween Ž0.15 M NaCl, 0.05 M NaHPO4 , pH 7.4; 0.05% Tween-20. and serially diluted in chicken rinsate. Spiked carcass rinsates Ž1 ml volume. were then incubated with 20 ml magnetic beads Žanti-E. coli O157:H7; Dynal, Oslo, Norway. for 10 min at 238C. Beads were retrieved using a block magnet and rinsed twice with PBS-Tween Ž0.15 M NaCl, 0.05 M NaHPO4 , pH 7.4; 0.05% Tween-20. before being resuspended in 50 ml H 2 O and stored at y208C. Colony forming units ŽCFU.

D.R. Call et al.r International Journal of Food Microbiology 67 (2001) 71–80

were calculated by plating replicate dilution series. Efficiency of cell capture was estimated by dividing CFU remaining Žafter beads were removed. by the total CFU that were mixed with the beads. Beads and plate counts were prepared from five to six independent dilution series of E. coli. For each dilution, PCR reactions Ž eaeA primer set. were spiked with 10–20 ml of final bead suspension, and the initial PCR denaturation step served to lyse captured cells. Post-PCR reactions were adjusted to a final concentration of 1 = SSC and 5 = Denhardt’s Ž70 ml total volume.. A total of 35 ml of this mixture was hybridized in each of two wells as described above.

3. Results and discussion 3.1. Detection sensitiÕity From a process-level perspective, detection sensitivity can be defined at the level of target, template

75

or cell. Target sensitivity is the minimum number Žor molar concentration. of targets in the hybridization solution that are required to exceed the detection threshold of the sensor after hybridization and washing steps are complete. Template sensitivity is the minimum number of DNA templates required for successful PCR and microarray detection. Cell sensitivity is the minimum number of cells required to complete the entire analytical process. Each of the three process-level perspectiverdefinitions of microarray sensitivity was addressed in this study. 3.2. Target sensitiÕity Microarray detection of the 104 bp eaeA PCR product provided approximately 32-fold greater detection sensitivity than a standard agarose gel ŽFig. 1.. Under conditions employed here, approximately 2.4 = 10 9 copies of target in hybridization solution Ž114 pM. were required to exceed the minimum detection limit of the ELF signal generation system

Fig. 1. Comparison of eaeA copy number sensitivity for an ethidium bromide-stained, 3% agarose gel Žtop panel. and a replicate dilution series hybridized to a series of arrays Žbottom panel.. Within each array well, the lower row of spots represents a positive control hybridization ŽQC. and the upper row shows hybridization of the eaeA PCR product. Arrays were printed on acid-washed slides and all array incubations were at 238C. NTC s no template control.

76

D.R. Call et al.r International Journal of Food Microbiology 67 (2001) 71–80

and Fluor-Se imager. For a 100-bp PCR product, 2.4 = 10 9 copies corresponds to approximately 36.8 ng. Assuming that up to 50% of a PCR reaction can be added to a single well, then products should be visible when as little as 75 ng of product is produced by the PCR process. Assuming further that the PCR reaction is efficient Žyielding 2 mg. and equal for all primer sets, then the theoretical upper limit for this system would be 25 distinct products Ži.e., 25 primer sets multiplexed together.. Maximum microarray signal sensitivity was achieved with approximately 10 11 copies of eaeA amplicons Ž4.7 nM; Fig. 2.. If the goal was to limit the number of primer sets so that maximum signal would be generated for all products, then 300 ng of each product would be required and six to seven primer sets could be multiplexed together. In contrast to the array results, the same eaeA PCR product was not visible in an agarose gel Žand Fluor-Se imager. until 7.8 = 10 11 copies were applied to the gel ŽFigs. 1 and 2.. These results are valid for short PCR products, but as product length increases, more ethidium bromide will intercalate into the products and gel sensitivity will improve. In contrast, achieving the necessary molar concentration of longer products Že.g. 1500 bp. for array hybridization might require pooling two or more PCR reactions. This latter limitation is due to having only one biotin per hybridized strand and would probably be

alleviated by incorporating more biotin molecules into the PCR product. Despite this limitation, we used the original primers of Oberst et al. Ž1998. and successfully hybridized the resulting 633-bp products to the eaeA probe following our standard protocol Ždata not shown.. Importantly, these detection sensitivities were achieved with a non-optimal, retrofitted gel documentation system wherein microarray images were captured with a 200 mm lens, having a focal point approximately 60 cm from the slide surface. We anticipate that significant gains in sensitivity can be achieved using a proximal imaging system and, when combined with other labeling systems, the sensitivity of microarray formats should be equivalent to the degree of sensitivity that is possible with isotopically labeled oligonucleotide probes and membrane bound targets Žapproximately 10 7 copies; Sambrook et al., 1989.. The biophysical constraints and process-level Žin.efficiencies of surface hybridizations are likely to limit sensitivity of our system to less than what can be achieved for solution-phase, fluorescent detection systems Že.g., TaqMane, 10 6 fluors.. Nevertheless, the FluorSe imaging system is certainly suitable where target number is not a limiting factor, and these results demonstrate how laboratories can take advantage of microarray technologies using standard Žand relatively inexpensive. imaging equipment. 3.3. Coupled PCR and microarray sensitiÕity

Fig. 2. Quantification of eaeA copy number sensitivity experiment ŽFig. 1.. Intensity ŽOD. of array spots Žblack circles. and gel bands Žopen circles. were normalized and represent averages Ž"SEM. from two independent experiments. Microarrays were 32-fold more sensitive compared to agarose gel detection.

Detection sensitivity was also evaluated as a function of the absolute quantity of DNA template introduced into the original PCR reaction. Under optimal hybridization and detection conditions, we were able to easily detect eaeA PCR products with 10 fg of genomic DNA Žone to two cell equivalents; Fig. 3.. Only one of three samples was positive with 1 fg O157:H7 genomic DNA and 35 cycles of PCR, a result that is consistent with molecular sampling error at low concentrations of target. That is, reproducible Žandror quantitative. PCR amplification typically requires ) 100 copies of target in the reaction volume in order to counter the variance associated with molecular sampling error ŽSarkar et al., 1990; Walsh et al., 1992; Mutter and Boynton, 1995.. PCR kinetics may also explain why microarray signal

D.R. Call et al.r International Journal of Food Microbiology 67 (2001) 71–80

Fig. 3. Microarray hybridization signal Ž eaeA locus. after 35 cycles PCR and a range of starting templates Ž100 ng to 1 fg.. Average intensity Ž"SEM. was calculated from three sets of serial dilutions. At the lowest concentration Ž1 fg., only one out of three samples produced hybridization signal. Target hybridizations were at 508C and subsequent incubations were at 378C.

intensity was not appreciably affected by the starting concentration of genomic DNA in the initial PCR reaction ŽFig. 3.. That is, 35 amplification cycles was enough for all dilutions of O157:H7 template to achieve sufficient molar concentration of targets to saturate the hybridization signal on the array ŽFig. 3.. It is probable that assay sensitivity can be somewhat enhanced by employing a nested PCR strategy Že.g. Herman et al., 1995., where first stage PCR is used to generate copies of larger PCR products and gene families Že.g. ) 200 bp. and second stage PCR is used to generate shorter amplification products for array detection and sequence discrimination. In either case, the detection sensitivity of the PCRrmicroarray technology is certainly equivalent to that of TaqMane PCR detection ŽG 100 CFU mly1 . ŽOberst et al., 1998.. 3.4. Sample to microarray sensitiÕity Spike recovery experiments were used to address microarray detection sensitivity in combination with immunomagnetic separation and PCR with unenriched, unprocessed carcass rinsate. Dilutions of E. coli O157:H7 in chicken rinsate averaged 530, 168, 91, and 55 CFU mly1 . We were able to detect the eaeA gene in 100% of the 530 and 168 CFU mly1

77

dilutions Žfive independent replicates for each dilution.. Detection was less reliable for lower cell concentrations, with five of six replicates positive for 91 CFU mly1 and two of six replicates positive for 55 CFU mly1 . For these cell concentrations, the immunomagnetic bead protocol appeared to capture 95% of viable cells, although we did not assess loss of additional cells during subsequent washing steps. Reduced detection success at the lowest cell concentrations may therefore be due to loss of cells during the washing steps, or potential carryover of PCR inhibitors from the chicken rinsate. Our PCR reactions were also limited to 40% of the total bead suspension. Consequently, the reduced success at 55 CFU mly1 may be attributable to having at most the equivalent of 21 cells in the initial PCR reaction combined with the potential presence of PCR inhibitors. Adding more beads to the PCR reaction would occasionally generate excessive background on the microarrays and was therefore not pursued as a means to increase detection sensitivity. Two options are available to increase sensitivity for cells in liquid food samples. A pre-enrichment step could be used Že.g. Grant et al., 1998; Oberst et al., 1998., but this in part negates the many advantages of PCR-based systems. Alternatively, the entire analytic process could be coupled with microfluidic technology to permit concentration of low numbers of CFU by processing much larger volumes of sample ŽChandler et al., 2000.. 3.5. Genotyping pathogenic E. coli The added benefits of multiplexed sequence detection by microarray hybridization are demonstrated in Fig. 4. Whereas TaqMane PCR is currently limited to two simultaneous detection probes, four virulence loci were simultaneously amplified from O157:H7 and hybridized directly with the microarray, with no cross-hybridization of individual PCR targets with non-target probes ŽFig. 4.. This simple array was then used to evaluate the genotypes of seven strains of E. coli ŽFig. 5.. The array unambiguously and correctly genotyped the five strains known to produce either Shiga-like toxin I or II ŽATCC 51435, 43888, 43889, 43890, 43894.. In

78

D.R. Call et al.r International Journal of Food Microbiology 67 (2001) 71–80

Fig. 4. Hybridization specificity for E. coli O157:H7 virulence loci. Upper left panel illustrates electrophoretic separation of eaeA, stx1, stx 2, and hlyA PCR products. Upper right panel illustrates hybridization of a multiplex PCR reaction to an array with probes complementary for each PCR target and for QC. The lower panel illustrates probe specificity for individual primer sets when PCR products were hybridized to arrays identical to the upper right panel. All microarray incubations were at 238C.

addition, all strains of O157:H7 were positive for the eaeA and hlyA loci, as expected. We also found that the EHEC O91:H2 serotype ŽATCC 51435. harbored

a copy of the hlyA gene, a result that may be due to the presence of a virulence plasmid that is frequently associated with EHEC bacteria ŽSchmidt et al., 1995..

Fig. 5. Genotyping results for seven strains of E. coli using multiplex PCR for eaeA, stx1, stx 2, and hlyA loci Žsee Fig. 4 for spot legend.. Genotypes for ATCC 51435, 43888, 43889, 43890, and 43894 match respective genotypes for stx1 and stx 2. All microarray incubations were at 238C.

D.R. Call et al.r International Journal of Food Microbiology 67 (2001) 71–80

The O91:H2 sample was negative for eaeA and stx1, as expected. Both the negative control strain ŽDH5a . and the no-template-PCR control were negative for all loci.

79

Memorial Institute for the U.S. DOE under contract DE-AC06-76RLO 1830. References

3.6. Conclusion This study has demonstrated a sensitive microarray detection system for genotyping strains of O157:H7 and other EHEC bacteria. It is clearly feasible to expand the current array to include primers and probes for detecting many additional genes or alleles, which would provide higher levels of discrimination between isolates and a valuable tool for epidemiological investigations. While the array described herein was intended for E. coli O157:H7 detection and characterization, it is also relatively simple to add additional probes for other pathogens and distinguish among them using a single PCR primer set Že.g., McCabe et al., 1999; in progress.. This system does not require sample pre-enrichment if bacterial titers are reasonably high Ž) 100 CFU mly1 .. For lower concentrations of bacteria, the immunomagnetic cell concentration procedure employed here can be linked to an automated fluidic system ŽChandler et al., 2000. and potentially avoid any need for pre-enrichment. This is a technological advance that could significantly enhance microarray detection sensitivity by processing much larger volumes of non-enriched liquid sample. Fully integrated, near real-time biodetection systems Žwith microarray detection. for multiple food-borne pathogens are therefore a near-term probability ŽSarkar et al., 1990; Walsh et al., 1992; Mutter and Boynton, 1995..

Acknowledgements Jeremy Brown, Mark Stottlemyre, Jack Small, and Eileen Jutras provided assistance with experiments or interpretation. We gratefully acknowledge Genometrix ŽWoodlands, TX. for providing assistance with initial attachment and hybridization protocols. This work was supported by the U.S. Department of Energy ŽDOE. LDRD and Laboratory Technology Research ŽLTR. Programs. Pacific Northwest National Lab is operated by Battelle

Allerberger, F., Rossboth, D., Dierich, M.P., Aleksic, S., Schmidt, H., Karch, H., 1996. Prevalence and clinical manifestations of Shiga toxin-producing Escherichia coli infections in Austrian children. European Journal of Clinical Microbiology and Infectious Diseases 15, 545–550. Bell, B.P., Goldoft, M., Griffin, P.M., Davis, M.A., Gordon, D.C., Tarr, P.I., Bartleson, C.A., Lewis, J.H., Barrett, T.J., Wells, J.G., 1994. A multistate outbreak of Escherichia coli O157:H7-associated bloody diarrhea and hemolytic uremic syndrome from hamburgers. The Washington experience. Journal of the American Medical Association 272, 1349–1353. Brewster, J.D., Mazenko, R.S., 1998. Filtration capture and immunoelectrochemical detection for rapid assay of Escherichia coli O157:H7. Journal of Immunological Methods 211, 1–8. Call, D.R., Chandler, D.P., Brockman, F., 2001. Fabrication of DNA microarrays using unmodified oligonucleotide probes. Bio Techniques Žin press.. Chandler, D.P., Brockman, F.J., Holman, D., Grate, J.W., Bruckner-Lea, C.J., 2000. Renewable microcolumns for solid-phase nucleic acid separations and analysis from environmental samples. Trends in Analytical Chemistry 19, 314–321. Fagan, P.K., Hornitzky, M.A., Bettelheim, K.A., Djordjevic, S.P., 1999. Detection of shiga-like toxin Ž stx1 and stx 2., intimin Ž eaeA., and enterohemorrhagic Escherichia coli ŽEHEC. hemolysin ŽEHEC hlyA. genes in animal feces by multiplex PCR. Applied and Environmental Microbiology 65, 868–872. Fratamico, P.M., Strobaugh, T.P., 1998. Evaluation of an enzyme-linked immunosorbent assay, direct immunofluorescent filter technique, and multiplex polymerase chain reaction for detection of Escherichia coli O157:H7 seeded in beef carcass wash water. Journal of Food Protection 61, 934–938. Fratamico, P.M., Sackitey, S.K., Wiedmann, M., Deng, M.Y., 1995. Detection of Escherichia coli O157:H7 by multiplex PCR. Journal of Clinical Microbiology 33, 2188–2191. Gannon, V.P., D’Souza, S., Graham, T., King, R.K., Rahn, K., Read, S., 1997. Use of the flagellar H7 gene as a target in multiplex PCR assays and improved specificity in identification of enterohemorrhagic Escherichia coli strains. Journal of Clinical Microbiology 35, 656–662. Grant, I., Ball, H., Rowe, M., 1998. Isolation of Mycobacterium paratuberculosis from milk by immunomagnetic separation. Applied and Environmental Microbiology 64, 3153–3158. Guo, Z., Guilfoyle, R.A., Thiel, A.J., Wang, R., Smith, L.M., 1994. Direct fluorescence analysis of genetic polymorphisms by hybridization with oligonucleotide arrays on glass supports. Nucleic Acids Research 22, 5456–5465. Heller, R.A., Schena, M., Chai, A., Shalon, D., Bedilion, T., Gilmore, J., Woolley, D.E., Davis, R.W., 1997. Discovery and analysis of inflammatory disease-related genes using cDNA microarrays. Proceedings of the National Academy of Sciences of the United States of America 94, pp. 2150–2155.

80

D.R. Call et al.r International Journal of Food Microbiology 67 (2001) 71–80

Herman, L., De Block, J., Moermans, R., 1995. Direct detection of Listeria monocytogenes in 25 milliliters of raw milk by a two-step PCR with nested primers. Applied and Environmental Microbiology 61, 817–819. Johnson, R.P., Durham, R.J., Johnson, S.T., MacDonald, L.A., Jeffrey, S.R., Butman, B.T., 1995. Detection of Escherichia coli O157:H7 in meat by an enzyme-linked immunosorbent assay, EHEC-Tek. Applied and Environmental Microbiology 61, 386–388. Lamture, J.B., Beattie, K.L., Burke, B.E., Eggers, M.D., Ehrlich, D.J., Fowler, R., Hollis, M.A., Kosicki, B.B., Reich, R.K., Smith, S.R., 1994. Direct detection of nucleic acid hybridization on the surface of a charge coupled device. Nucleic Acids Research 22, 2121–2125. Lockhart, D.J., Dong, H., Byrne, M.C., Follettie, M.T., Gallo, M.V., Chee, M.S., Mittmann, M., Wang, C., Kobayashi, M., Horton, H., Brown, E.L., 1996. Expression monitoring by hybridization to high-density oligonucleotide arrays. Nature Biotechnology 14, 1675–1680. Louie, M., Read, S., Simor, A.E., Holland, J., Louie, L., Ziebell, K., Brunton, J., Hii, J., 1998. Application of multiplex PCR for detection of non-O157 verocytotoxin-producing Escherichia coli in bloody stools: identification of serogroups O26 and O111. Journal of Clinical Microbiology 36, 3375–3377. McCabe, K., Zhang, Y., Huang, B., Wagar, E., McCabe, E., 1999. Bacterial species identification after DNA amplification with a universal primer pair. Molecular Genetics and Metabolism 66, 205–211. Mutter, G.L., Boynton, K.A., 1995. PCR bias in amplification of androgen receptor alleles, a trinucleotide repeat marker used in clonality studies. Nucleic Acids Research 23, 1411–1418. Oberst, R.D., Hays, M.P., Bohra, L.K., Phebus, R.K., Yamashiro, C.T., Paszko-Kolva, C., Flood, S.J., Sargeant, J.M., Gillespie, J.R., 1998. PCR-based DNA amplification and presumptive detection of Escherichia coli O157:H7 with an internal fluorogenic probe and the 5’ nuclease ŽTaqMan. assay. Applied and Environmental Microbiology 64, 3389–3396.

Paton, A.W., Paton, J.C., 1998. Detection and characterization of Shiga toxigenic Escherichia coli by using multiplex PCR assays for stx1, stx 2, eaeA, enterohemorrhagic E. coli hlyA, rfbO111, and rfbO157. Journal of Clinical Microbiology 36, 598–602. Pyle, B.H., Broadaway, S.C., McFeters, G.A., 1999. Sensitive detection of Eschericia coli O157:H7 in food and water by immunomagnetic separation and solid-phase laser cytometry. Applied and Environmental Microbiology 65, 1966–1972. Sambrook, J., Fritsch, E.F., Maniatis, T., 1989. Molecular Cloning. 2nd edn. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, New York, NY, USA. Sarkar, G., Cassady, J., Bottema, C.D.K., Sommer, S.S., 1990. Characterization of polymerase chain reaction amplification of specific alleles. Analytical Biochemistry 186, 64–68. Schena, M. ŽEd.. 2000. Microarray Biochip Technology. Eaton Publishing, Natick, MA, USA. Schmidt, H., Beutin, L., Karch, H., 1995. Molecular analysis of the plasmid-encoded hemolysin of Escherichia coli O157:H7 strain EDL 933. Infection and Immunity 63, 1055–1061. Seo, K.H., Brackett, R.E., Frank, J.F., 1998. Rapid detection of Escherichia coli O157:H7 using immunomagnetic flow cytometry in ground beef, apple juice, and milk. International Journal of Food Microbiology 44, 115–123. Sharma, V.K., Dean-Nystrom, E.A., Casey, T.A., 1999. Semi-automated fluorogenic PCR assays ŽTaqMan. for rapid detection of Esherichia coli O157:H7 and other shiga toxigenic E. coli. Molecular and Cellular Probes 13, 291–302. Stewart, C., Flint, H. ŽEds.. 1999. Escherichia coli O157 in Farm Animals. CABI Publishing, Wallingford, England. Tortorello, M.L., Reineke, K.F., Stewart, D.S., Raybourne, R.B., 1998. Comparison of methods for determining the presence of Escherichia coli O157:H7 in apple juice. Journal of Food Protection 61, 1425–1430. Walsh, P.S., Erlich, H.A., Higuchi, R., 1992. Preferential PCR amplification of alleles: mechanisms and solutions. PCR Methods and Applications 1, 241–250.