|Home | About | Journals | Submit | Contact Us | Français|
Escherichia coli O157:H7 strains can be classified into different genotypes based on the presence of specific Shiga toxin-encoding bacteriophage insertion sites. Certain O157:H7 genotypes predominate among human clinical cases (clinical genotypes), while others are more frequently found in bovines (bovine-biased genotypes). To determine whether inherent differences in gene expression explain the variation in infectivity of these genotypes, we compared the expression patterns of clinical genotype 1 strains with those of bovine-biased genotype 5 strains using microarrays. Important O157:H7 virulence factors, including locus of enterocyte effacement genes, the enterohemolysin, and several pO157 genes, showed increased expression in the clinical versus bovine-biased genotypes. In contrast, genes essential for acid resistance (e.g., gadA, gadB, and gadC) and stress fitness were upregulated in bovine-biased genotype 5 strains. Increased expression of acid resistance genes was confirmed functionally using a model stomach assay, in which strains of bovine-biased genotype 5 had a 2-fold-higher survival rate than strains of clinical genotype 1. Overall, these results suggest that the increased prevalence of O157:H7 illness caused by clinical genotype 1 strains is due in part to the overexpression of key virulence genes. The bovine-biased genotype 5 strains, however, are more resistant to adverse environmental conditions, a characteristic that likely facilitates O157:H7 colonization of bovines.
Escherichia coli O157:H7, a food-borne zoonotic pathogen that causes hemorrhagic colitis and hemolytic uremic syndrome (HUS) in humans, is the most prevalent type of enterohemorrhagic E. coli (EHEC) in the United States (22, 52). Cattle are a major reservoir of E. coli O157:H7, and the fecal shedding prevalence on cattle farms can range from 0 to 100% (17). Colonized cattle do not exhibit clinical disease (41). It has been reported that only 10 to 100 cells of E. coli O157:H7 are sufficient to induce overt disease in humans (55) and also that <300 cells are sufficient for colonization of cattle (5). Based on the relative frequency of isolation of different genotypes of E. coli O157:H7 from cattle and humans, it has been suggested that bovine-derived E. coli O157:H7 strains vary in their ability to cause human disease (6); the basis for this variation is not known.
E. coli O157:H7 possesses unique virulence properties that mediate disease development, including Shiga toxins (Stx), the locus of enterocyte effacement (LEE) pathogenicity island, and the pO157 virulence plasmid (59). The LEE encodes a type 3 secretion system (T3SS) that mediates the formation of attaching and effacing lesions (33), while the pO157 plasmid encodes several putative virulence factors, such as an enterohemolysin (EhxA or EHEC HlyA) (27) and a type 2 secretion system (T2SS) (44). Both the LEE and pO157 have been shown to be critical for disease pathogenesis (21, 29). Shiga toxins, which are the cytotoxins responsible for renal damage in HUS (21), are encoded by genes located on lysogenic lambdoid phages that are inserted into the O157:H7 chromosome at specific locations (46). A prior study of 80 bovine isolates and 282 clinical isolates from humans with O157-associated disease demonstrated that the distribution of Stx insertion sites varied between isolate types (6). Genotypes 1 to 3, for example, predominated among clinical strains despite being isolated from cattle, while other genotypes (e.g., 5 and 7) were underrepresented among clinical strains (bovine-biased genotypes) (6). Similarly, octamer-based genome scanning of bovine and clinical isolates of E. coli O157:H7 identified three genetically distinct lineages, of which lineages I and II were isolated mostly from humans and bovines, respectively, and the intermediate lineage I/II was less well characterized based on host distribution (23). A recent study demonstrated that lineage I, I/II, and II strains belong to Stx insertion site genotypes 3, 1, and 6, respectively (25). Comparing the presence of virulence genes between E. coli O157:H7 isolates from various sources using DNA microarrays also has revealed that O157 isolates from beef cattle and humans are genetically distinct (28).
In addition to intrinsic differences, it is possible that there are differences in the expression of important virulence genes as well as variations in the degree of resistance to adverse environmental conditions between clinical and bovine-biased genotypes. To investigate this hypothesis, the exponential-phase transcriptomes of four clinical genotype 1 strains were compared to transcriptomes of four bovine-biased genotype 5 strains using microarrays. The main objective of this study was to identify gene expression differences between these two E. coli O157:H7 genotypes following growth in vitro to better understand the complex gene regulatory events important for survival in the bovine reservoir and disease pathogenesis in humans.
The eight bacterial strains used in this study were selected based on the Stx-encoding bacteriophage insertion site genotypes determined in a prior study (6). Strains representing genotypes 1 (clinical genotype) and 5 (bovine-biased genotype) were selected from among 80 bovine strains originally isolated between 1991 and 2004 as described previously (6). Although the clinical genotype 1 strains used in this study were bovine derived, their genotypes were identical to those of genotype 1 strains isolated in a prior study from humans with O157 infections (6). Four strains of each genotype were included in the microarray analyses. A previously described (30) stx2 and stx2c restriction fragment length polymorphism analysis demonstrated that all genotype 1 strains used in this study harbored stx2 alone, whereas the genotype 5 strains contained only stx2c.
Each strain, stored at −70°C in LB broth containing 10% glycerol, was inoculated into 10 ml of LB broth and grown to an optical density at 600 nm (OD600) of ~0.1 to recover cells. Cells were grown twice to stationary phase in morpholinepropanesulfonic acid (MOPS)-buffered minimal medium (pH 7.4) before transfer at a 1:30 dilution into 100 ml of Dulbecco's modified Eagle medium (DMEM) (0.45% glucose) for both RNA isolation and the model stomach assay. DMEM was chosen because it provides a controlled environment with minimum variation in growth conditions, and it also has been shown to induce expression of LEE genes (1). To minimize the confounding effect of acidic pH that develops in the stationary phase of growth in unbuffered medium, the DMEM was buffered with MOPS to pH 7.4.
To compare global gene expression profiles between genotypes 1 and 5, microarrays were hybridized in a double loop design, thereby allowing strains from one genotype to be directly compared to strains from the other genotype (Fig. (Fig.1).1). A total of four strains were selected from each genotype to represent the biological and technical replicates required for the double loop design. The four strains from each genotype represent biological replicates and, therefore, significant differences in gene expression are representative of the two distinct genotypes.
For RNA isolation, the strains were grown to exponential phase (~2.25 h; OD600, ~0.5) in DMEM, and RNA extractions were performed using a modified version of the previously described hot phenol method (7). Briefly, 5 ml of the culture was mixed with a 1/10 volume of 10% phenol-ethanol buffer to stabilize the RNA and centrifuged at 4°C (4,300 × g) for 30 min to pellet cells. The supernatant was decanted, and cell pellets were suspended in 5 ml of buffer (2 mM EDTA, 20 mM NaO-acetate, pH 5.2) before RNA extraction with hot phenol. Reverse transcription reactions and the coupling of cDNA with Cy3 or Cy5 dyes were conducted as described elsewhere (4).
cDNA hybridizations were performed according to the double loop microarray design (Fig. (Fig.1).1). As described elsewhere (4), the cDNAs were hybridized onto microarray slides printed with 6,088 open reading frames (ORFs) representing E. coli genome strains K-12 (8), EDL933 (37), and Sakai (18); 110 ORFs from the pO157 plasmid were included (E. coli oligo set version 1.0.2; Qiagen Operon Technologies). Arrays were scanned with an Axon 4000b scanner (Molecular Devices, Sunnyvale, CA) followed by image analysis using GenePix 6.0 (Molecular Devices).
Microarray data were processed as previously described (20) and fitted to a mixed microarray analysis of variance (MAANOVA) model (13). The linear model tested was Y (intensity) = array + dye + strain (clinical or bovine biased) + sample (biological replicate) + error. Significant differences in expression were determined using the Fs test in the MAANOVA with 500 random permutations to estimate the P values. This test uses a shrinkage estimator for gene-specific variance components that makes no assumption about the variance across genes (14). In addition, the q-value package in R was used to determine the false discovery rate (FDR) (48). Genes with a >1.5-fold change in expression and an FDR of <0.1 were considered significantly increased, as these parameters have been used previously to define significant differential expression (32). Additionally, significance analysis of microarrays (SAM) was used to analyze data with an FDR of 0.05.
Overrepresentation of gene sets with a common biological function in the two genotypes was determined using the Gene Set Enrichment Analysis preranked analysis program (GSEA v2.0; Broad Institute, Massachusetts Institute of Technology) (49). The gene sets were designated based on the annotation for the Sakai genome (18) available through the J. Craig Venter Institute (http://cmr.jcvi.org/tigr-scripts/CMR/GenomePage.cgi?org/=ntec03). Additionally, genes for the LEE and the glutamate decarboxylase system and acid fitness island (AFI-GAD) also were included in the analysis.
Select genes that had significantly different levels of expression between genotypes in the microarray analysis were confirmed by quantitative real-time PCR (qRT-PCR). TaqMan assays (3) were used to quantify the expression of gadA, gadB, and ler, with mdh as a reference for normalization. For all other genes, SYBR green was used as described elsewhere (20); methods for cDNA synthesis and qRT-PCR also were described previously (20). The expression level of the 16S rRNA gene was used for normalization of data, and the relative expression levels were quantified using a modified Livak method (45). The results presented are averages from four biological replicates ± the standard errors of means (SEM).
Genomic DNA was extracted with the Puregene DNA extraction kit (Gentra Systems, Minneapolis, MN) for use with the GenomeLab SNPstream system (Beckman Coulter, Fullerton, CA). Single-nucleotide polymorphism (SNP) genotyping via the SNPstream was performed using a modified version of a previously described protocol (30) according to the manufacturer's instructions. Briefly, PCR was conducted using four panels of 48-plex primers targeting 192 distinct SNP loci identified previously via the comparison of three O157:H7 genomes (30). The primers, which differed from the original protocol, were designed using the Autoprimer program (61). After purification, the PCR products were subjected to single-base primer extension reactions, which add a labeled nucleotide to the SNP site, followed by hybridization onto a 384-well SNP microarray plate. Detection and processing were performed via an SNPstream Imager (version 2.3; GenomeLab). A total of 52 of the 192 SNPs were found to be informative; these SNPs were concatenated in MEGA4 (50) to construct a neighbor-joining tree (43) using untransformed distances (p distance) and bootstrap confidence values based on 1,000 replications for examining the phylogenetic relationships between the eight strains. SNP data from reference strains representing each of the nine O157:H7 clades (30) were included in the analysis.
The model stomach system (MSS) (19) was prepared as described previously (3). Gerber turkey rice dinner baby food (30 g) was mixed with 120 ml of synthetic gastric fluid (pH 1.70), yielding a final pH of 2.5. Strains, grown to an OD600 of ~2.5 in DMEM, were inoculated into the MSS at the rate of 106 cells/ml. Contents of the MSS were stomached for 30 s, sampled, diluted, and plated onto LB agar plates every 30 min for 1.5 h to enumerate viable cells. CFU/ml from duplicate plates were averaged and converted to the log10 CFU/ml. Survival rates were calculated as the log decrease in viable cell count per 30 min, and the averages from two experimental replicates are reported.
Microarray data were submitted to the NCBI Gene Expression Omnibus database (http://www.ncbi.nlm.nih.gov/geo) and are available under accession number GSE15783.
The MAANOVA Fs test identified significantly different expression of 191 genes between the two genotypes, of which 71 were upregulated in the clinical genotype and 120 were upregulated in the bovine-biased genotype (FDR, <0.1; fold change, ≥1.5) (see Table S1 in supplemental material). Additionally, SAM identified more genes significantly differentially expressed between the two genotypes: 154 were upregulated in the clinical genotype and 238 were upregulated in the bovine-biased genotype (FDR, <0.05; fold change, >1.5) (see Table S2 in supplemental material). One hundred sixteen genes were found to be differentially expressed by both MAANOVA Fs test and SAM. Differentially expressed genes included those involved in virulence, response to stress, acid resistance, and metabolism. Overall, several important O157:H7 virulence genes were upregulated in the clinical genotype (Table (Table1),1), whereas genes related to acid resistance and stress fitness were upregulated in the bovine-biased genotype (Table (Table2).2). GSEA also identified enrichment of eight gene sets in the clinical genotype and six gene sets in the bovine-biased genotype (Table (Table3),3), thereby providing additional support for the MAANOVA and SAM results.
There was an overall increase in expression of the LEE genes in clinical genotype 1 relative to bovine-biased genotype 5 strains; expression was significantly different in 12 genes (Table (Table1).1). For example, secreted proteins encoded by espF and espG, T3SS proteins encoded by escF, sepQ, escT, and escR, and the cesD chaperone were upregulated in genotype 1 strains (Table (Table1).1). Although the remaining 29 LEE genes were not statistically significantly different between genotypes, 27 were upregulated in genotype 1 relative to genotype 5. The statistically nonsignificant result was possibly due to interstrain variation within the genotypes. To explore this further and to confirm expression differences, qRT-PCR was used to examine the expression of four important LEE genes, ler, espB, espD, and tir, that were not significantly different by microarray analysis (Fig. (Fig.2).2). More than a 2-fold increase in expression was observed for espB, espD, and tir by qRT-PCR in the clinical genotype (Fig. (Fig.2),2), a level that was similar to the microarray data for the 12 significant genes. Expression of ler was slightly lower than that of the other genes, although it still exhibited a 1.4-fold increase in clinical strains (Fig. (Fig.2).2). Additionally, GSEA confirmed the enrichment of the entire set of 41 LEE genes in the clinical genotype (Table (Table33).
The pO157 plasmid encodes a number of virulence-associated genes in E. coli O157:H7 strains. Fourteen of these genes, including ehxA (EHEC hlyA; enterohemolysin), toxB (toxin B), and 8 of the 13 genes that encode the T2SS, were significantly upregulated in the clinical genotype 1 (Table (Table1).1). The T2SS etp cluster (44) showed a 1.8- to 2.8-fold increase in expression in the clinical genotype, which was confirmed by qRT-PCR (Fig. (Fig.2).2). ehxA and toxB also were confirmed to have 2.2- and 2.1-fold increases in expression, respectively, by qRT-PCR (Fig. (Fig.2).2). Expression of some virulence genes encoded by pO157, such as ehxCBD and stcE, was not significantly different between the two genotypes.
Numerous genes that are essential for acid resistance in E. coli were significantly upregulated in the bovine-biased genotype 5 strains. This included genes that encode all three components of the GAD system: gadA, gadB, and gadC (15). In addition, the 12 AFI genes (31) had increased expression in the bovine-biased genotype, with 8 of the 12 expressed at significantly different levels (Table (Table2).2). The increased expression of GAD system genes was confirmed by qRT-PCR, which showed a more-than-10-fold increase relative to clinical genotype 1 (Fig. (Fig.2).2). Similarly, there were 3.6-, 5.6-, 6.6-, and 7.5-fold increases in expression of gadX, gadE, hdeA, and hdeB, respectively, by qRT-PCR (Fig. (Fig.2).2). The upregulation of gadX, however, was not statistically significant in the microarray analysis, although the direction of change was the same. This discrepancy was also possibly due to high interstrain variation in expression within the genotype.
Expression of dps, which is involved in protecting DNA during starvation and acid stress (12), was upregulated by 3.6-fold in the bovine-biased genotype strains. Similarly, clpA, a chaperone necessary for protein degradation by the ClpAP protease (24), showed a 2-fold increase in expression. Other stress fitness-associated genes with increased expression in the bovine-biased genotype included the cold shock protein gene cspD (60), cation transport regulator genes chaBC (35), and the universal stress protein gene uspA (11, 38) (Table (Table2).2). Moreover, expression of katG (36) and of osmC and osmY (57), the genes involved in resistance to peroxide and osmotic stress, also were upregulated (Table (Table2).2). Interestingly, the general stress sigma factor gene rpoS (58) was not differentially expressed between the two genotypes.
Several metabolic genes, including nine genes involved in the nitrogen regulatory response that are regulated by the sigma factor RpoN (40), were upregulated in the bovine-biased genotype. The nitrogen regulatory protein gene glnK and the ammonium transporter gene amtB were upregulated by 11.6- and 14.9-fold, respectively. glnD, which is involved in the posttranscriptional modification of glnK, was also upregulated in bovine-biased strains, as were genes for the nitrogen regulator I (glnG), a permease of the periplasmic glutamine binding protein (glnH), and proteins associated with glutamate biosynthesis (gltB, gltK, and gltJ) (Table (Table22).
Furthermore, nac, which encodes the nitrogen assimilation control protein, was upregulated by 12.7-fold in the bovine-biased genotype. Consequently, a number of Nac-regulated genes, including oppA, oppB, oppD, oppF, gabD, gabT, and dppA, had higher expression levels in bovine-biased strains. Other RpoN-regulated genes, such as hypA, hypB, and fdhF, also had increased expression in the bovine-biased genotype (Table (Table22).
Because the eight strains in this study were only characterized by the distribution of Stx insertion sites and represented the same multilocus sequence type (6), a more sensitive SNP genotyping method (30) was used to better understand the phylogenetic relationships of strains within and between the two genotypes. Among the four strains representing clinical genotype 1, three grouped together with a clade 8 control strain and one grouped with a clade 1 control strain (Fig. (Fig.3).3). By contrast, all four strains representing the bovine-biased genotype 5 belonged to clade 7 (Fig. (Fig.33).
Since one clinical genotype 1 strain was part of a phylogenetically distinct lineage (clade 1) relative to the other three clinical genotype 1 strains (clade 8), the microarray data were reanalyzed after excluding the data generated from the clade 1 strain. The Fs test identified significant upregulation of 400 genes in the clinical genotype and 349 genes in the bovine-biased genotype (FDR, <0.1; fold change, >1.5) in this reanalysis. All but three genes (bioB, dps, and ycaL) identified as differentially expressed in the first analysis were also differentially expressed in the second analysis. Further, 561 additional genes were identified in the second analysis, as elimination of the clade 1 strain likely reduced the within-genotype variation. Twenty-eight LEE genes, including the genes encoding intimin (eae), the translocated intimin receptor (tir), and a positive regulator of LEE (grlA), were significantly upregulated by 1.6-, 1.6-, and 1.7-fold, respectively, in the three clinical genotype strains. Similarly, pO157-encoded genes, such as ehxA, toxB, and 11 genes of the etp polycistron, which encodes a T2SS, also were upregulated in the clinical genotype. As expected, the bovine-biased genotype strains showed increased expression of GAD and AFI genes relative to the three clinical genotype strains. One additional gene that was significantly upregulated in the second analysis, but not the first analysis, was gadX. The stationary-phase sigma factor gene, rpoS, and adiY, which encodes an ARAC-like regulator of the arginine decarboxylase acid resistance system, also were upregulated in the clade 7 bovine genotype strains, as were a number of stress fitness-associated and RpoN-regulated genes.
To determine whether the increased expression of acid resistance genes in the bovine-biased genotype translates to a phenotypic difference, model stomach assays were conducted. These assays were used to directly compare the survival of both genotypes in a complex acidic environment which simulates both the human stomach and the bovine abomasum. Consistent with the microarray expression data, there was a significant difference (P = 0.003) between the survival rates of clinical and bovine-biased genotypes, as bovine-biased strains had a 2-fold increase in survival in the MSS (Fig. (Fig.4).4). The average survival rate per 30 min for the clinical genotype was −0.55 ± 0.04, whereas the bovine-biased genotype rate was −0.27 ± 0.04.
Genotyping based on Stx-encoding bacteriophage insertion sites has demonstrated that the E. coli O157:H7 strains present in the bovine reservoir are considerably more diverse than strains that cause human infections (6). Furthermore, it was suggested that some bovine-biased genotypes have reduced virulence and, hence, cause disease less frequently relative to those bovine-derived clinical genotypes that are commonly isolated from patients (6). This variation could be due to gene content differences, including allelic variation in key genes among genotypes, or due to expression differences in critical genes. Here, we describe inherent differences in the expression of important genes that provide a possible explanation for the variation in infectivity between bovine-biased and clinical genotypes. Specifically, genome-wide expression profiling using microarrays demonstrated differential expression of key virulence and stress fitness genes between the two genotypes following growth in DMEM. The microarray data were confirmed by qRT-PCR, and variations in survival rates were examined in the model stomach system. While we recognize that growth in DMEM does not mimic the complex environment of either the bovine or human gastrointestinal tract, the goal of this study was to identify genes that are inherently differentially expressed between genotypes. We suspect that similar genes will be up- and downregulated in vivo, although the magnitude of the fold change may be altered. Future studies involving exposure to epithelial cells or animal models are required for confirmation. However, as the model stomach functional assay used in this study confirmed the expression differences identified by microarray, it is likely that these in vitro results correlate well with actual situations in vivo.
One of the most important differences identified in this study was the upregulation of the LEE in clinical versus bovine-biased genotypes. The LEE is considered a critical factor in E. coli O157:H7 disease pathogenesis, as it encodes a T3SS that mediates adherence to the intestinal mucosa (33). Because strains of the clinical genotype express key LEE genes at a higher level, it is likely that these strains have an enhanced ability to adhere to the intestinal epithelium and cause the attaching and effacing lesions that initiate the disease process. In contrast, it is possible that increased expression of negative regulators of LEE suppresses the expression of important LEE genes in the bovine-biased genotype, thereby reducing adherence and subsequent disease. The increase in gadE expression in the bovine-biased genotype supports this hypothesis, as our prior study determined that GadE, the central activator of the GAD system, negatively regulates LEE in O157:H7 strains (20). Similarly, GadX, a negative regulator of LEE in enteropathogenic E. coli (47), also was upregulated in the bovine-biased genotype, as was yhiF, an AFI-encoded regulator that suppresses LEE expression in EHEC (54).
Similar to the LEE, another important factor in E. coli O157:H7 disease pathogenesis is the possession of the pO157 plasmid. The 92-kb plasmid carries genes for many putative virulence factors, including ehxA (enterohemolysin) (42), toxB (toxin B) (9), and a number of etp genes necessary for a T2SS (16); all were upregulated in the clinical genotype. The enterohemolysin encoded by ehxA is a cell-associated, pore-forming toxin that belongs to the repeats-in-toxin (RTX) family (42). Although the exact role of the enterohemolysin in O157:H7 pathogenesis is not clear, it has been shown to cause injury to microvascular endothelium (2) and induce production of the proinflammatory cytokine interleukin-1β, a serum marker of HUS (51). The increased expression of grlA, a LEE-encoded positive regulator of ehxA (42), may partly explain the upregulation of ehxA in the clinical genotype. StcE, a protease involved in the intimate adherence of EHEC to host cells, is secreted by the T2SS encoded by the etp gene cluster (16); eight of the etp genes were significantly upregulated in the clinical genotype. Similarly, toxB, which was upregulated in clinical genotype strains, also has been shown to be important for full adherence to human epithelial cells (53). Together, these results demonstrate that the clinical genotype 1 strains express factors important for adherence at higher levels, suggesting that genotype 1 strains may have an enhanced ability to adhere to host cells and thus are inherently more virulent than bovine-biased genotype 5 strains. A comparison of Stx expression was not possible using microarrays, as the two genotypes had distinct stx profiles. A previous study, however, demonstrated that the differences between Stx protein expression levels among genotypes 1 and 5 strains were insignificant (6).
During passage through the bovine gastrointestinal tract, E. coli O157:H7 has to survive a number of adverse environmental conditions, including extreme acidity in the abomasum (56), organic acid stress from volatile fatty acids in the rumen and colon, and occasional hyperosmolarity. Therefore, the capacity to withstand acidity and other environmental stressors is critical for O157:H7 strains to successfully persist in cattle. Consistent with this, we observed upregulation of acid resistance and stress fitness-associated genes in the bovine-biased genotype relative to the clinical genotype. The GAD system is the most efficient acid resistance system in E. coli (10, 26) and is essential for the survival of E. coli O157:H7 in bovines (39). All three GAD system genes (gadA, gadB, and gadC) (15) had increased expression in the bovine-biased genotype. While it is possible that both the oxidative system (i.e., the RpoS-mediated acid resistance system) and the arginine decarboxylase system are also important for acid resistance, the expression results varied in the SNP genotype (clade 8 versus clade 7) and stx insertion site genotype (clinical versus bovine biased) analyses. Only genes involved in the GAD system were differentially expressed in both analyses.
The increased expression of GAD and AFI genes, which are also involved in acid resistance (31), is possibly induced by upregulation of GadE, an essential activator of GAD and many AFI genes in E. coli O157:H7 (20). DMEM buffered with MOPS maintains a neutral pH during exponential phase and, therefore, it appears that under noninducing conditions, expression of gadE and GAD system genes is markedly higher in the bovine-biased genotype, which may enhance survivability in the bovine gastric environment (pH 2.1) (56). This inference was confirmed in the model stomach assays, where bovine-biased strains survived better than clinical genotype strains. Because the model stomach represents a complex acidic environment similar to the human stomach and bovine abomasum, the increased survivability of bovine-biased strains may also be attributable to the increased expression of stress fitness genes, such as dps, clpA, and uspA, in addition to the acid resistance genes. Together, these findings indicate that bovine-biased genotype 5 strains are more resistant to adverse environmental conditions, which likely facilitates survival in and colonization of bovines. A prior study comparing resistance to acetic acid (pH 3.3) among E. coli O157:H7 strains isolated from environmental sources and humans produced similar results (34). Specifically, human isolates were less resistant to acetic acid than isolates from bovine feces (34). This characteristic enables less-resistant strains, or genotypes, to express other factors, particularly those involved in adherence, following exposure to human cells, mainly due to the previously described (20) negative interaction between acid resistance regulators and adherence genes. Clinical genotype strains also express acid-resistant genes at a lower level, which is possibly sufficient for transmission to humans or survival in unfavorable environments.
As the nitrogen source in DMEM is glutamine, strains face an ammonia-limiting condition during growth in DMEM. This condition leads to the induction of the nitrogen regulatory response (40), and expression data from this study indicated that bovine-biased strains are more efficient at mounting this response at the transcriptional level. Ten RpoN-regulated genes involved in the response were upregulated in bovine-biased strains. These observations indicate that strains of the bovine-biased genotype are equipped with a more efficient nitrogen regulatory response system that can enhance survival under ammonia-limiting conditions. Such conditions are likely encountered in the bovine gastrointestinal tract and the external environment.
Decreasing the prevalence of O157 colonization in cattle has become increasingly significant in the current strategies to control O157 infections. In this context, identifying the genes that are critical for O157 persistence in cattle is important for developing novel prevention strategies against this pathogen. This study, along with previous studies (39), indicates that acid and stress resistance genes encompass an important set of genes that are crucial for E. coli O157:H7 survival in the environment and cattle and so may represent future targets for disease prevention. In addition, there is considerable variation in expression levels of different genes among E. coli O157:H7 strains isolated from cattle compared to strains of clinical genotype 1 and the increased expression of important virulence factors. The upregulation of key virulence components provides a possible explanation for the predominance of this genotype in clinical cases.
We dedicate this paper to the late Thomas S. Whittam.
We thank Lindsey M. Ouellette and Hans Steinsland for technical assistance, Teresa M. Bergholz for helpful suggestions for the microarray design, and Linda S. Mansfield, Galeb S. Abu-Ali, and James T. Riordan for critical review of the manuscript.
This project has been supported by federal funds from the National Institute of Allergy and Infectious Diseases, National Institutes of Health, Department of Health and Human Services, under contract no. N01-AI-30055.
Published ahead of print on 30 October 2009.
†Supplemental material for this article may be found at http://aem.asm.org/.