Expression of Putative Virulence Factors of Escherichia coli O157:H7 Differs in Bovine and Human Infections

Strains and media.Bacteria were grown in a shaking incubator at 37°C in Luria-Bertani (LB) medium. When appropriate, ampicillin (100 μg/ml) or nalidixic acid (32 μg/ml) was added. Strain 86-24 is a well-studied clinical isolate of E. coli O157:H7 (39). Strain 86-24^nalR (3) was used to infect calves. ORN172(pIHA) contains iha on a high copy plasmid (38). Shiga toxin-producing E. coli (STEC) O103:H2 expresses enterohemolysin at high levels (34).

Clinical samples.Fecal specimens from children infected with E. coli O157:H7 were obtained from the Pacific Northwest. The Institutional Review Boards of the Children's Hospital and Regional Medical Center, Seattle, Wash., approved the use of these specimens, which were frozen immediately after collection and stored at −70°C until use.

Six conventional Holstein calves were housed in isolation units and prescreened for health status and freedom from fecal STEC/enterohemorrhagic E. coli. Half of the calves were obtained at 1 day of age, screened during week 1, inoculated on day 8, monitored for fecal shedding for 7 days, and euthanized and necropsied at 14 to 16 days of age. The remaining calves were obtained at 6 weeks of age, screened for 1 week, challenged at 7 weeks of age, monitored for fecal shedding for 7 days, and euthanatized and necropsied between 54 and 58 days of age. Calves were orally challenged with 10⁹ to 10¹⁰ CFU of E. coli O157:H7 strain 86-24^nalR and monitored for 6 to 8 days for appetite, demeanor, diarrhea, and Nal^r fecal E. coli O157:H7. At necropsy, approximately 50 g of fecal sample from the rectum was mixed in a stomacher bag, aliquoted, and frozen. The Washington State University Institutional Animal Care and Use Committee approved this research.

Positive controls.To create stool with known amounts of E. coli O157:H7, bacteria were grown in LB broth to an optical density at 600 nm of 0.6 and 10⁸ CFU were added per gram of stool donated by a healthy volunteer. To examine the sensitivity of the assay, stool was spiked with broth-grown E. coli O157:H7 diluted in phosphate-buffered saline.

RNA extraction from stool.RNA was extracted using a modification of the silica-binding method (2, 4). Buffers L6, L11, L10, and L2 and silica were prepared as described previously (2, 4). Guanidine isothiocyanate (GITC)-containing buffers were stored in the dark and used within 3 weeks after preparation. The silica pH was adjusted to exactly 2.0 using 32% HCl. Approximately 0.2 g stool was mixed with 5 ml L6 containing 0.2 g polyvinylpyrrolidone (PVP-40). For watery human samples and all bovine samples, the amount of stool was increased to 0.4 g. The mixture was vortexed thoroughly and centrifuged for 5 min at 4,300 × g and 4°C. The supernatant was transferred to fresh tubes containing 2.5 ml GITC-phenol (7.5 M GITC, 0.5% sodium dodecyl sulfate, 1 mM EDTA, 50 mM sodium acetate [NaOAc] [pH 4.0], 50% H₂O-saturated phenol) and 2.5 ml chloroform. The solution was vortexed and centrifuged (5 min, 4,300 × g, 4°C). The aqueous phase was transferred to a fresh tube containing 10 ml ethanol (EtOH) and 400 μl 3 M NaOAc, incubated for 30 min at −20°C, and centrifuged (20 min, 4,300 × g, 4°C). The precipitated nucleic acids were vacuum dried, resuspended in 3 ml L11, and split into 1-ml aliquots in microcentrifuge tubes, each containing 300 μl silica (pH 2.0) to bind DNA. The tubes were vortexed thoroughly, shaken (10 min, 4°C), and centrifuged (2 min, 2,000 × g), and the RNA-containing supernatants were transferred to a tube containing 700 μl L10 and 200 μl silica (pH 2.0) to bind RNA. Again, the tubes were vortexed thoroughly, shaken, and centrifuged under the same conditions, and the supernatants were discarded. The RNA-containing pellets were washed twice with 1 ml L2, twice with 1 ml 70% EtOH, and once with 1 ml acetone. The pellets were dried by incubating the tubes with open lids at 56°C for 5 min. To elute RNA, 200 μl H₂O was added to the tubes, which were then vortexed thoroughly and incubated (56°C for 10 min with lids closed). RNA was concentrated by adding 3 volumes EtOH, 1 volume 3 M NaOAc, and 1 μl glycogen (Roche), freezing at −80°C overnight, and centrifuging (20 min, 11,000 × g, 4°C). Pelleted RNA was washed with 70% EtOH, dried, and resuspended in a 1:1 solution of formamide and H₂O.

DNA was eliminated using Ambion's (Austin, TX) “DNA-free” DNase. Samples were treated twice following the recommendations of the manufacturer, using 1 unit of DNase per microgram of nucleic acid. RNA integrity after the DNase treatment was confirmed by visualizing rRNA after electrophoretic separation in a 0.8% agarose gel.

Primers and probes.Primers and probes were designed using Primer Express 1.5 (PE Applied Biosystems) and modified to allow approximately 10 bp between primer and probe binding sites. When possible, we used previously published primers and probes (Table 1). TaqMan probes were ordered from Applied Biosystems (Foster City, CA) and fluorescently labeled with 6-carboxyfluorescein at the 5′ end and minor groove binder/nonfluorescent quencher at the 3′ end. Primers and probes were tested using positive control (spiked) stools (described above). Real time RT-PCR was also performed on normal stool without added O157:H7 to confirm that amplification did not occur in the absence of E. coli O157:H7. To determine the sensitivity of mRNA detection, real-time RT-PCR was performed for each gene on stool from a healthy individual spiked with dilutions of broth-grown bacteria.

Real-time RT-PCR.Real-time RT-PCR was performed on 50 ng fecal RNA by using the Invitrogen SuperScript III Platinum one-step quantitative RT-PCR system. TaqMan probes (described above) were used so that the reaction contained two levels of specificity: (i) the annealing of specific primers and (ii) hybridization of the probe. For each gene, amounts of primer, probe, and MgSO₄ were optimized. These were used at the concentrations detailed in Table 2. All reactions using clinical samples were performed in triplicate using RNA from a single extraction. Reactions were run on a Rotor-Gene 2000 real-time cycler: program parameters were 55°C for 25 min, 95°C for 2 min, and 45 cycles of 95°C for 15 s and 60°C for 30 s. To confirm that RNA was free of DNA, an additional reaction was included that was identical to the RT-PCR but substituted platinum Taq for the RT enzyme mix provided in the kit. If a clinical specimen yielded no amplicon for a particular gene, the reaction was repeated. If only two of the three reactions yielded a product, these two were included in the analysis.

Standard curves.RNA was extracted from bacteria grown with shaking at 37°C to logarithmic phase (12). For each gene, real-time RT-PCR was performed on 10, 5, 1, 0.5, and 0.1 ng of RNA. Conditions were identical to those used for the clinical samples. A standard curve was generated by plotting log 10 ng RNA on the x axis and the cycle threshold (C_T) value on the y axis. The equation of the slope of the best-fit line was used to determine relative amounts for each gene for the clinical samples.

Determination of ratios of genes of interest to control gene.For each clinical sample, real time RT-PCR was performed for each gene in triplicate in a single experiment. The relative amounts of RNA were determined from the C_T values by reference to the standard curves. The average relative amount of the transcript of interest was then divided by the average relative amount of gnd to determine the ratio for each gene.

Statistical analysis.For each putative virulence gene, the ratio of mRNA to gnd mRNA for human samples was compared to that for bovine samples by using the two-tailed Mann-Whitney test. Significance was set at 5%.

Selection of genes for study.To determine if virulence genes are expressed differently when E. coli O157:H7 is excreted from humans compared to cattle, we selected several known or candidate virulence loci for expression analysis. These included LEE pathogenicity island genes eae and espA, as well as the gene encoding Shiga toxin 2 (stx₂). stx₂ was selected because, unlike stx₁ and other Shiga toxin variant genes, it is present in nearly all North American isolates (26, 39). We also included genes for which a role in human infection is uncertain: rfbE, which is involved in the synthesis of the O antigen of lipopolysaccharide (LPS), and ehxA, which encodes enterohemolysin. Finally, we included iha, which encodes Iha, an adherence-conferring molecule that is a virulence factor in uropathogenic E. coli (16, 38). Levels of expression of these genes were related to that of a constitutively expressed housekeeping gene, gnd. gnd was selected as a reference locus because it is unusually polymorphic among E. coli housekeeping genes (25), enabling the specific analysis of expression of the O157:H7 gene against a background of commensal E. coli. Moreover, it differs from the gnd genes of related E. coli O55:H7 and E. coli with the O157 rfb region, which are only distantly related to E. coli O157:H7 and are not pathogens. Thus, gnd_O157:H7 provides a target that is theoretically quite specific for the identification of pathogenic E. coli O157:H7.

RNA extraction from fecal samples.We developed an extraction method to isolate sufficiently pure RNA, free of RNases and inhibitors of enzymatic reactions. First, we employed a silica-binding method (2, 4) to isolate intact RNA as determined by separation on a 0.8% agarose gel (data not shown). However, we were unable to amplify fecal RNA obtained by this method, and we could not amplify RNA from broth-grown bacteria mixed 1:1 with fecal RNA extracted using this method. We suspected that inhibition of amplification reactions was at least in part the result of polyphenolics, which inhibit PCR (19). Polyphenolics are found in the leaves, bark, and fruit of most plants and reportedly copurify with nucleic acids during extraction from plant material and sludge (23, 37). PVP binds polyphenolics, and adding PVP-40 reduces inhibition of amplification by polyphenolics when included in an RNA extraction (15, 36, 37). Therefore, additional steps to optimize the RNA extraction protocol included adding PVP-40 in the lysis buffer, and the resulting procedure (see Materials and Methods) enabled the isolation of RNA suitable for RT-PCR from control (uninfected) human stool samples to which O157:H7 was added (data not shown).

Selection and optimization of primers and probes.Primers and TaqMan probes designed for real-time RT-PCR are shown in Table 1. Optimum concentrations of primer, probe, and MgSO₄ were determined (Table 2) using positive control stools (described in Materials and Methods). For each gene, an increase in fluorescence during real-time RT-PCR was observed for RNA from human stool spiked with E. coli O157:H7 but not for RNA from unspiked stool containing commensal E. coli (data not shown). Furthermore, gene expression was not detected in RNA from a bovine fecal sample taken before infection (data not shown), nor was it detected in bovine samples with low numbers of E. coli O157:H7 (see below). Thus, the primers and probes are specific for E. coli O157:H7 and do not amplify codefecated flora.

The sensitivity of the assay was determined by performing real-time RT-PCR on control stool spiked with serial dilutions of broth-grown E. coli O157:H7, ORN172(pIHA) (for iha), or STEC O103:H2 (for ehxA). ORN172(pIHA) and STEC O103:H2 highly express iha and ehxA, respectively, and were used because LB-grown O157:H7 expressed these genes at very low levels. The lowest dilution for which an increase in fluorescence was observed for each gene is listed in Table 2.

Analysis of human and bovine specimens.RNA was extracted from eight human and four bovine specimens. Six bovine specimens were available, but we could detect E. coli O157:H7 gene expression in only four. The remaining two specimens contained <10⁴ CFU E. coli O157:H7/g. Each gene was amplified from each RNA preparation in triplicate. For the majority of the >80 reactions performed in triplicate, the C_T values within the triplicates varied by <10%. For four reactions, the C_T values of the triplicates varied by <35%. These four were cases in which the C_T value was >30, i.e., near the limit of detection. The relative amounts of RNA were determined from the experimental C_T values by reference to the standard curves.

Average relative amounts of mRNA for each gene were then normalized to the average relative amount of gnd mRNA of the same specimen. The resulting ratios were compared between human and bovine specimens by using the two-tailed Mann-Whitney test independently for each gene. For calf no. 3, in which expression of only some genes was detected, the sample was excluded from analysis for the genes whose expression was not detected. We elected to exclude these samples rather than assign an arbitrary low number because our statistical analysis used a nonparametric method that requires that the data be rank ordered. Assigning an artificial value had the potential to denote that data point with an incorrect rank and might have invalidated the results. Table 3 indicates the ratios of expression relative to gnd.

iha and espA exhibited statistically significant differences in expression between human and bovine hosts (Fig. 1). In both cases, the genes were more highly expressed in cattle feces. In contrast, ehxA (Fig. 2) (P = 0.073) and rfbE (P = 0.11) appeared to be more highly expressed in human stool, although these results failed to achieve statistical significance.

• Open in new tab
• Download powerpoint

FIG. 1.

Genes differentially expressed in human and bovine hosts. iha (A) or espA (B) expression normalized to gnd expression is shown with specimens grouped by source. The bars represent the means for human and bovine specimens (P < 0.05).

• Open in new tab
• Download powerpoint

FIG. 2.

ehxA exhibits a trend towards differential expression. ehxA expression normalized to gnd expression is shown with specimens grouped by source. The bars represent the means for human and bovine specimens (P = 0.073).