Pathotyping Escherichia coli by Using Miniaturized DNA Microarrays

The detection of virulence determinants harbored by pathogenicEscherichia coli is important for establishing the pathotyperesponsible for infection. A sensitive and specific miniaturizedvirulence microarray containing 60 oligonucleotide probes wasdeveloped. It detected six E. coli pathotypes and will be suitablein the future for high-throughput use.

INTRODUCTION

Top

INTRODUCTION

Pathogenic Escherichia coli strains constitute a significantpublic health problem worldwide (12). In contrast to their nonpathogeniccounterparts, these strains have acquired specific virulenceattributes that allow them to cause a spectrum of human andanimal illnesses (10, 15). Numerous methods exist for the detectionof pathogenic E. coli, including geno- and phenotypic markerassays for the detection of virulence genes and their products(7, 17, 21, 23). These methods have the common drawback of screeninga relatively small number of determinants simultaneously. DNAmicroarrays offer a viable alternative due to their abilityto screen multiple markers simultaneously.

The aim of this work was to develop a simple high-throughputsystem based in a microtube (details are available from CLONDIAG,Jena, Germany) (13, 20) for pathotyping E. coli isolates sentto clinical diagnostic laboratories.

Design and validation of miniaturized virulence arrays.

Top

Design and validation of...

A miniaturized E. coli oligonucleotide virulence array was designedcontaining 39 virulence, 7 bacteriocin, and 15 control (rrland gad) gene probes (Table 1). Eighteen genes were specificto a particular E. coli pathotype, 13 were common between 2or more pathotypes, and 7 were unassigned. The design of probes/primersand the specificity were tested as previously described (1,13).

View this table:
[in this window]
[in a new window]

TABLE 1. Probes and primers used in the miniaturized microarray^a

Control strains were used to validate each probe present onthe array (Table 1). PCR amplification and sequencing, usingprimers given in Appendix 1 of the supplemental material, verifiedthe presence of the probes in control strains. The sequencedgenes showed between 92 and 100% sequence identity to the respectivetarget gene and showed 100% sequence identity to the probe andprimer regions (data not shown).

Genomic DNA was extracted from cells grown aerobically overnightat 37°C in LB broth, using a DNeasy tissue kit (catalogno. 69504; QIAGEN). One microgram of genomic DNA from each strainwas used as a template in a multiplex linear amplification andlabeling reaction with the set of 60 primers (Table 1), as previouslydescribed (1). The amplified products were added to ArrayTubesfor hybridizations performed according to the method of Ballmeret al. (1, 13).

The sequenced strains EDL933, CFT073, and E2348/69 were usedto estimate assay sensitivity to ensure strong signal intensitywith minimal nonspecific cross-hybridization. Optimization includedvarying the concentrations of genomic DNA used for labeling(2 to 0.05 µg), the primers present in the linear multiplexmix (0.135 to 0.810 µM), and the poly-horseradish peroxidase-streptavidinconjugate used for detecting hybridization (50 to 400 pg/µl).The minimal concentration of genomic DNA found to reliably detectall expected genes was 1.0 µg, while a concentration of0.135 µM per primer in the stock solution was sufficientfor the detection of target DNA (Fig. 1). The optimal concentrationof poly-horseradish peroxidase-streptavidin conjugate was foundto be 200 pg/µl; concentrations above or below this valueresulted in high background or no detectable reaction at all(data not shown).

View larger version (35K):
[in this window]
[in a new window]

FIG. 1. Optimization of the genomic concentration used in this study. The optimal concentration of genomic DNA from EDL933 used for the detection of genes on the virulence oligonucleotide miniaturized microarray chip was assessed using (a) 2 µg, (b) 1 µg, (c) 0.5 µg, (d) 0.1 µg, and (e) 0.05 µg of DNA. The six biotinylated marker spots (C) are visible in all arrays. A concentration of 0.135 µM per primer in the stock solution and 200 pg/µl of poly-horseradish peroxidase-streptavidin conjugate were used for these assays.

The spot signal intensity was derived by calculating the quantitativestaining value with IconoClust software (version 2; CLONDIAG).The data were normalized using the signal intensity of the gadprobe, and the normalized signal intensity for genes withinpositive and negative control strains was used to differentiatebetween present (signal intensity value above 0.4) and absent(signal intensity value below 0.3) genes. Genes with signalintensity values between 0.3 and 0.4 were considered ambiguous.Two replicate hybridizations were performed for each controlstrain, and the 95% confidence interval of error across replicatehybridizations was 1.6 to 3% (see Appendix 2 in the supplementalmaterial).

The specificity of each probe was estimated by comparing arraydata with PCR and sequenced data from control strains. In allcases, the virulence gene(s) known to be present within positivecontrol strains was clearly identified by array, while two negativecontrol strains, including the sequenced strain MG1655, showedthe presence of only 23S rRNA and gad genes (see Appendix 2in the supplemental material). For many positive control strains,additional virulence genes were detected (Table 2). Furthermore,PCR amplification in all control strains of five randomly chosengenes (eae, astA, ehx or hlyA, iss, and mcmA), showed 100% correlationbetween array and PCR data, indicating the probes to be highlyspecific with minimum cross-reactions (data not shown).

View this table:
[in this window]
[in a new window]

TABLE 2. Virulence determinants detected within each positive control strain

Pathotyping clinical isolates.

Top

Pathotyping clinical isolates.

A panel of 63 E. coli human and animal clinical isolates werepathotyped using the virulence miniaturized microarray (seeAppendix 3 in the supplemental material). For five strains,two hybridization reactions were performed and the 95% confidenceinterval of error between replicates was 0.9 to 5.0%. Only onehybridization reaction was performed for the remaining teststrains.

Fifty-five of the isolates hybridized to more than one virulencedeterminant and were readily designated within a recognizedpathotype, mostly matching the clinical diagnosis where available.Five isolates that harbored only the iss gene and/or microcinsand three isolates that hybridized to only control genes couldnot be pathotyped. These isolates may harbor virulence genesnot present on our array. Several isolates with novel combinationsof genes were detected and included two shigatoxigenic E. colistrains, one with senB, iss, cma, cba, and mchBCF genes andanother with astA, cdtB, and cnf genes. The most commonly detectedgene was iss, which was present in half the strains tested.Other genes which were detected in at least 10 or more isolatesincluded eae, ehx, astA, iroN, mchF, mchB, mchC, f17A (threevariants combined), f17G, mcmA, cba, cma, and prfB/papB. GenesvirF, pet, hlyE, fasA, and cfa were not detected in any testisolate (see Appendix 3 in the supplemental material).

Conclusion.

Top

Conclusion.

Several E. coli virulence arrays for genotyping have been describedpreviously (2-5, 9, 11, 18). These arrays use mostly a glassslide printed with oligonucleotide probes or PCR products fortarget genes and fluorescent Cy dyes to label DNA used for hybridization.This system is time consuming, with expensive reagents and requiresa skilled technician. In contrast, the microtube-based arraysystem used in this study has a short assay time due to an amplificationstep and inexpensive reagents and requires low technical skills,making it amenable for use in clinical diagnostic laboratories.In the future, the routine use of virulence microarrays in suchlaboratories will not only allow rapid detection and designationof the pathotypes of strains sent to diagnostic laboratoriesbut also enable emergent strains harboring novel virulence combinationsto be detected before such strains spread to become a healthproblem.

ACKNOWLEDGMENTS

We are grateful to the Enteric Reference Laboratory at VLA forthe provision of E. coli strains and in particular to KatherineSprigings and Louise Finch. We thank Elke Müller and JanaSachtschal for their assistance.

This project was funded through the VLA seedcorn fund.

FOOTNOTES

* Corresponding author. Mailing address: Department of Food and Environmental Safety, Veterinary Laboratories Agency-Weybridge, New Haw, Addlestone, Surrey KT15 3NB, United Kingdom. Phone: 44 (0)1932 359476. Fax: 44 (0)1932 357268. E-mail: m.anjum{at}vla.defra.gsi.gov.uk

Published ahead of print on 13 July 2007.

Supplemental material for this article may be found at http://aem.asm.org/.

REFERENCES

Top