In order to screen large chromosomal fragments that confer virulence phenotypes, a HindIII BAC library of DNA fragments with an average size of 150 kbp from gDNA isolated from E. coli O157:H7 strain Sakai stx⁻ (Dahan et al., 2004) was generated in pCLD04541 (Jones et al., 1992) and transformed into E. coli DH10B. The BAC library was produced by Lucigen Corporation (Middleton, WI, USA) and was screened for specific sequences by Southern hybridization of colony filters made by the same company.

Type III secretion (T3S) is of central importance for EHEC and is required for colonization of the bovine host. The T3S machine is encoded by the LEE, an approx. 45 kb island inserted into the common E. coli backbone. Using a probe designed to hybridize ecs4562 (encoded within the LEE) we identified 11 BAC clones that contain ecs4562 and then screened these for clones containing the entire LEE by using PCR primers that amplified across the junctions between the LEE and common backbone (Table 1). Clone 3/A2 was found to encode the entire LEE and sequencing of the clone ends from the plasmid indicated that it contained bases 4486393–4639841 of the Sakai chromosome, a 153.4-kbp fragment containing the LEE. To determine if this region was sufficient to confer T3S on E. coli DH10B; clone 3/A2 (pBAC3/A2) was grown under T3S permissive conditions and secreted proteins isolated and analyzed for the needle filament protein EspD by western blotting. Secretion of EspD could not be detected (data not shown) in agreement with previous observations (Elliott et al., 1999). To verify that pBAC3/A2 encodes a functional copy of the LEE, a triparental mating using pBAC3/A2 as the donor, pRK2013 (Ditta et al., 1980) as helper and E. coli O157:H7 strain TUV93-0 Δ148A Nal^r as the recipient was performed. TUV93-0 Δ148A Nal^r contains an 8.93-kb deletion within the LEE (from 4677931 to 4686861 as defined for the EDL933 genome) and is deleted for LEE1, LEE2, and most of the LEE3 operon (escN onwards are intact) and is unable to export translocon or effector proteins (Figure 1; Campellone et al., 2004). Introduction of pBAC3/A2 restored T3S to TUV93-0 Δ148A Nal^r (Figure 1) indicating that this clone is likely to contain a functional LEE and that the LEE encoded by E. coli O157:H7 strain Sakai stx⁻ is not sufficient to confer detectable T3S on a non-pathogenic K-12 isolate such as DH10B. The extra-LEE elements that confer T3S in Sakai are still unclear.

E. coli O157:H7 colonize the terminal rectal tissue of cattle and replication at this site and associated mucus leads to excretion in the feces and contamination of the environment and the food-chain. High levels of colonization at the terminal rectum are correlated with high numbers of organisms shed into the environment and a higher probability of causing human disease (so called “supershedders”; Matthews et al., 2006). Colonization factors that confer adherence to terminal rectal tissues are important potential targets to prevent colonization by vaccination or more direct interventions. In an effort to identify genomic regions of E. coli O157:H7 that may confer adhesion 384 BAC clones were used in repeated rounds of adhesion and culture on bovine primary terminal rectal cells. BAC clones were incubated with primary tissue culture for 4 h at 37°C after which the monolayer was washed to remove unattached bacteria and disrupted to recover adherent bacteria. This pool of adherent bacteria was further selected by another two rounds of adherence. Selected bacteria were inoculated into DMEM for a final round of growth to amplify the amount of gDNA for labeling and hybridization.

In order to determine which BAC clones increase adherence to primary tissue culture total gDNA was isolated from the input and output pools and comparative genomic hybridization (CGH) performed. CGH allows screening of the entire output pool for regions of the genome that have been enriched by selection. This data is normalized using a LOWESS regression and can be visualized as a ratio of signal within the input and output pools for each gene (Figure 2 and Data files 1,3 in Supplementary Material). To reduce variation between individual genes and to identify large regions (containing potentially overlapping BAC clones) that are uniformly increased a sliding window of 50 genes was used to average the signal for each gene. In line with our earlier observations that the LEE did not confer T3S on DH10B, LEE containing BAC clones were not selected on primary cells and in fact were negatively selected under these conditions. A number of genomic regions were enriched above two-fold in the primary tissue culture selected pools and these regions encode adhesins or other factors known to confer attachment to cultured epithelial cells. These include pO157 (Grys et al., 2005b; Dziva et al., 2007; Ho et al., 2008), the autotransporter ehaA (Wells et al., 2008), and type 1 fimbriae (Galfi et al., 1998). Encouragingly, autotransporters that have been demonstrated not to confer adhesion to primary bovine terminal rectal cells, such as ehaB, ehaC, and ehaD were not enriched in this genome-wide screen indicating a level of specificity (Wells et al., 2008). By screening the BAC library by CGH we have also identified two novel regions that are enriched by selection on bovine primary cells, a region encompassing O-island 71 (OI-71), and a region containing OI-90 and 91. We were not able to identify a known adhesin on, or surrounding OI-71 but note that the fimbrial cluster designated loc9 (indicated by yehD in Figure 2) is encoded adjacent to OI-90. Previous work on loc9 has indicated that this gene cluster can be expressed under in vitro conditions (Low et al., 2006b) and does provide a colonization advantage in cattle (Low et al., 2006a). These enriched regions provide promising targets for further study.

The restricted colonization site of EHEC O157 in cattle means that majority of bacterial in vivo replication is limited to the rectal epithelium or mucus layer before excretion in feces. EHEC O157 should have evolved to grow efficiently in rectal mucus which is likely to require genetic regions allowing use of specific carbon sources at this site, the capacity to deal with host innate defense products and potentially the production of factors that can restrict the growth of competitive microbes. Our recent work modeling bacterial replication rates required at the terminal rectum to account for the excretion levels measured indicates doubling times between 20 and 30 min (unpublished data), toward the maximum rates established for E. coli in vitro. Doubling times measured in bovine rectal mucus diluted in Hanks buffer were between 30 and 40 min (data not shown) supporting the nutritional quality of even the diluted substrate.

To identify genomic regions that provide a selective advantage at the terminal rectum we cultured 384 BAC clones in bovine terminal rectal mucus and screened output pools by CGH. Four regions were strongly enriched by culturing in mucus and contain OIs 20, 115, 126, and 175 (indicated in (Figure 3 and Data files 2,4 in Supplementary Material). lacZ is encoded adjacent to OI-20 and is inactivated in the E. coli background DH10B. Lactose is reported to be absent from mucus, but the metabolic pathway for metabolism of lactose is induced by culturing on murine mucus (Chang et al., 2004; Fabich et al., 2008). The region adjacent to lac also contains the mhp gene cluster allowing bacterial breakdown of 3-(3-hydroxyphenyl)propionic acid. Phenylpropanoic and phenylpropenoic acids and their derivatives are common in the environment, arising as breakdown products of lignin and other plant-derived flavonoids and phenylpropanoids (Torres et al., 2003). Adjacent to this region and central to the selected fragments, is also a two component regulatory system for sensing hexose phosphate and an associated hexose phosphate transporter (z0461–z0463).

The region containing OI-175 was also enriched and contains a lesion in the restriction–modification system hsdR in DH10B. It seems unlikely that hsdR would confer a growth advantage in mucus and this region may contain novel genes involved in mucus growth. For example at the start of the selected region is gntP, a fructuronate transporter and central to the region is phosphoglycerol transferase I (mdoB).

Two regions contained genes that are known to play an important role in EHEC catabolism of mucus sugars and colonization of cattle. fucAO are encoded adjacent to OI-115 and confer a selective advantage on wild-type E. coli in mucus when cultured in the presence of an isogenic fucAO mutant (Fabich et al., 2008). A strong colonization defect has also been observed in a fucAO mutant during cattle colonization (Snider et al., 2009). Similarly, the uronate isomerase uxaC (required for hexuronate catabolism) is encoded adjacent to OI-126 and confers a growth advantage on wild-type cells (Fabich et al., 2008). This region also encodes the agaWEFA cluster that is required for catabolism of galactosamine and N-acetylgalactosamine (GalNAc) and is inactivated in commensal E. coli K-12 strains. Both sugars are present in mucus and support growth of EHEC. The presence of the agaWEFA cluster has been reported not to confer a significant advantage on wild-type cells in vitro however an agaWEFA mutant is outcompeted during cattle colonization with significantly less CFU found in terminal rectal mucus (Snider et al., 2009). Enrichment of regions containing fucAO and uxaC/agaWEFA supports the validity of our results and the ability to BAC-CGH to identify phenotypically relevant regions of the chromosome.

To further confirm our CGH data we have isolated BAC clones that contain ecs4487 (containing Sakai specific sequence within the uxaC/agaWEFA region) and determined their growth phenotype relative to the BAC vector in mucus. A clone containing the uxaC/agaWEFA region was identified, designated I12. Both I12 and the BAC vector only (DH10B pV41) strain were marked with Nal^r to allow selection. In reciprocal experiments, the Nal^r tagged I12 or pV41 strain was inoculated at equal CFU into bovine mucus with Nal^s pV41 or I12 respectively. After 24 h culturing in mucus the total CFU/milliliter increased from an average 4.4 × 10⁵–6.5 × 10⁷ (Figure 4). During these seven generations the I12 clone was able to outcompete the control vector, increasing from 42 and 43 to 74 and 72% of the total population in reciprocal experiments (p < 0.05 in both experiments using a Student’s t-test). These results confirm that the uxaC/agaWEFA region identified in our BAC-CGH screen can confer a selective advantage during growth in bovine mucus.

The bacterial strains and plasmids used in the study are described in Table 1. LB broth was also used (Oxoid). Antibiotics were included when required at the following concentrations: chloramphenicol (12.5 μg/ml), kanamycin (25 μg/ml), tetracycline (12.5 μg/ml), and ampicillin (50 μg/ml).

Bacteria were cultured in 30 ml of MEM–HEPES at 37°C (200 rpm) to an OD₆₀₀ of 0.8 unless specifically stated. The bacterial cells were pelleted by centrifugation at 4000 × g for 15 min, and supernatants were passed through filters (0.22 μm). Supernatant proteins were precipitated overnight at 4°C with 10% TCA, and separated by centrifugation at 3220 × g for 30 min at 4°C; the proteins were suspended in 100 μl of 1.5 M Tris (pH 8.8). The bacterial pellet was initially suspended in 50 μl sample buffer (Sigma#S3401) and 50 μl molecular biology grade water. Proteins were separated by SDS-PAGE using standard methods and western blotting performed as described previously (Roe et al., 2003).

Bacterial artificial chromosome library mixture (384 colonies) and pV41 vector only were separately cultured overnight in triplicate in LB. One volume of the above overnight cultures was incubated in 50 volumes of DMEM until the OD₆₀₀ reached 0.6, then adjusted to 0.4. Cultures were diluted 1/100 and then 100 μl/well was transferred into 6-well plates of primary cells. Tissue samples for primary cell culture and isolation of mucus were sourced from a local abattoir. Primary tissue culture was prepared as described previously (Mahajan et al., 2005). Parallel experiments using 1 ml of mucus (verified E. coli negative by plating on sorbitol-MacConkey agar) from the first scrap of cattle rectum tissues were also inoculated with 100 μl of bacterial culture. Tetracycline was added to mucus samples to select for the BAC library or vector, and samples were serially diluted and plated to determine CFU/ml on sorbitol-MacConkey agar at indicated times. For each experiment bacteria with primary cells were incubated for 4 h. Culture medium for primary cells was replaced with DMEM 1 h prior to the above treatment. After unattached bacteria were removed by washing twice with pre-warmed bacterial medium, bacteria from primary cell cultures were collected by mechanical disruption of the monolayer. Bacteria were inoculated and collected from primary tissue culture two further times. Bacterial cells were collected at the end of three rounds of adhesion and subjected to genomic DNA extraction for CGH.

Genomic DNA was extracted using an Invitrogen ChargeSwitch gDNA Mini Bacteria Kit (Invitrogen). Labeling was carried out using a Bioprime Plus Array CGH Genomic Labeling System (Invitrogen). Protocol for pre-hybridization, hybridization and washing was essentially as previously described for hybridization of cDNA (Roe et al., 2007). Briefly, array slides were washed twice in Wash Buffer II (0.1 × SSC and 0.1% × SDS), each time for 30 s. Slides were transferred in pre-warmed pre-hybridization buffer (0.1% BSA, 0.1% SDS and 5× SSC) for 120 min at 42°C. Slides were washed once in Wash Buffer II and twice in Wash buffer III (0.1 × SSC), and the denatured (95°C for 3 min) hybridization probe mixture (20 pmol of each labeled probe, 0.4 × Ultrahyb, 0.8 × SSC, 90 ng/μl polyA, 10 μg/μl BSA) was added to slides and hybridized overnight at 42°C. Following hybridization slides were washed in Wash Buffer I (0.1% SDS, 2× SSC), Wash buffer II and twice in Wash buffer III. Slides were dried by centrifugation and scanned using an Axon Genepix 4000A scanner (Axon Instruments, Union City, CA, USA). Slide Images were processed using Genepix software and data analyzed using Genespring GX 7.3 (Agilent). Data were normalized using a LOWESS regression. To visualize genome-wide data; a window of 25 genes on either side of each gene was used to average the signal in that region. Data plotted for each gene in Figures 2 and 3 represents this average. Regions with an average intensity above two-fold across >50 consecutive genes (on average ∼35 kb or ∼25% of the average BAC insert length) were considered significant.

For competition experiments of individual BAC clones in mucus, spontaneous nalidixic acid resistant (Nal^r) mutants were recovered for both DH10B pCLD04541 and I12 by plating onto LB Nal plates. Competition experiments were performed between DH10B pCLD04541 Nal^s and I12 Nal^r or DH10B pCLD04541 Nal^r and I12 Nal^s. Equal CFU were inoculated into bovine terminal rectal mucus and incubated at 37°C for 24 h. The start and end CFU/ml of each culture was determined by plating on both LB Tet (to select for both strains carrying the BAC vector) or LB Tet/Nal to select for the tagged strain. The relative proportion each strain was determined by subtracting the number of Nal^r colonies from the total Tet resistant population.