INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

Escherichia coli is by far the most common cause of urinary tract infection (UTI), particularly in uncomplicated cases. Strains causing these infections possess traits that distinguish them from commensal strains of E. coli and other pathogenic strains such as those causing diarrhea and meningitis. Characteristically, uropathogenic strains of E. coli are composed of a restricted number of O serogroups, produce hemolysin, P fimbriae, and aerobactin, exhibit serum resistance, and are encapsulated (6, 7, 14, 15, 18, 31, 35). The presence of these features, not found in the typical fecal strain, implies that uropathogenic strains possess a defined set of virulence determinants that allow the bacterium to colonize the urinary tract, avoid host defenses, and elicit histological damage to the uroepithelium, allowing in some cases passage of the bacterium into the bloodstream.

Indeed, such clustered sets of virulence genes, termed pathogenicity islands (PAIs), have been defined for three strains of uropathogenic E. coli: 536 (3), J96 (39), and CFT073 (17). Typically, these sequences are large (>30-kb) blocks of DNA inserted within or near tRNA genes (12, 33, 39), contain direct repeats and insertion sequences, have a GC content that differs from that of the rest of the genome, and encode defined virulence determinants (3, 19, 26). We have previously shown that such sequences are widespread among uropathogenic isolates (17). One probe from the PAI of strain CFT073 hybridized with genomic DNA from approximately 80% of acute pyelonephritis and cystitis strains but only 19% of fecal strains.

Previously in our laboratory the boundaries of a PAI were identified for E. coli CFT073, a highly virulent strain isolated from the blood and urine of a woman with acute pyelonephritis (17). In this report, we provide an analysis of the nucleotide sequence for a 57,988-bp region. Previously unrecognized open reading frames (ORFs), as well as homologs of characterized genes of other species, were found. In addition, we identified the distribution of these sequences that span the PAI among the majority of other uropathogenic strains of E. coli.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Bacterial strains. E. coli CFT073 was isolated from the blood and urine of a woman admitted to the University of Maryland Medical System for the treatment of acute pyelonephritis (27). This hly⁺ pap⁺ sfa⁺ pil⁺ strain is highly virulent in the CBA mouse model of ascending UTI (28) and is cytotoxic for cultured human renal proximal tubular epithelial cells (27). It is phenotypically positive for the production of P fimbriae, hemolysin, and type 1 fimbriae. E. coli DH5 (34) was used as a recipient for gene bank and recombinant clones.

Cosmid library. A cosmid library was constructed with partially Sau3A-digested genomic DNA isolated from E. coli CFT073. DNA was ligated into BamHI-digested pHC79. The ligation mixture was packaged in vitro with the Gigapack lambda packaging kit (Stratagene) and used to infect E. coli DH5. Transformants were selected on Luria agar containing ampicillin (200 µg/ml).

Preparation of templates for nucleotide sequencing. Three overlapping cosmid clones (8-3f, 18-2f, and 5-4a), prepared from genomic DNA from E. coli CFT073 (28) and found previously to carry a PAI (17), were used to prepare subclones for nucleotide sequencing by deletion, subcloning of specific restriction fragments, and PCR amplification of specific sequences.

Nucleotide sequencing and analysis. Double-stranded DNA was used as a template for sequencing by the dideoxy-chain termination method (36). Primers used in these studies are listed in Table 1. Reactions were run with reagents from a Prism Ready Reaction Dye Deoxy Termination kit (Applied Biosystems) in conjunction with Taq polymerase. A model 373A DNA sequencer (Applied Biosystems) was used, and sequences were determined in both directions. DNAsis software (version 2.1; Hitachi) was used for analysis of the DNA sequence for base composition, identification of ORFs and restriction sites, and other basic analyses. Apparent homologies between ORFs both outside and inside the PAI were sought in GenBank by using the Wisconsin Package (version 8.1; Genetics Computer Group, Inc.).

DNA probes and dot blot hybridization. DNA restriction fragments isolated from cosmid clones 8-3f and 5-4a and subclones 8HS9 and 5HS11B were used as gene probes to determine whether homologous sequences were present in genomic DNA preparations of E. coli strains isolated from clinical sources. E. coli CFT073 and DH5 were used as positive and negative controls. Fragments were labeled by using the Amersham enhanced chemiluminescence system. For dot blots, E. coli strains were cultured in Luria broth (80 µl) in 96-well microtiter plates. Bacterial suspensions were lysed and pipetted onto a nucleic acid transfer membrane. Samples were neutralized with Southern blot neutralization buffer (34). Hybridization was done with 11 probes encompassing the length of the PAI (Fig. 1).

View larger version (42K): [in this window] [in a new window]   FIG. 1.   Features of the pathogenicity island of E. coli CFT073. ORFs defined by nucleotide sequencing are shown as arrows. The ORF designations, below the arrows, are defined in Table 2. The direction of each arrow indicates the predicted direction of transcription. The top line of shaded boxes indicates that ORFs in this region are highly homologous to or identical with ORFs previously identified in E. coli K-12 (dark shading) (2) or uropathogenic J96 (light shading) (1, 8, 20, 21, 29, 30, 32, 40). The unshaded boxes on the second line indicate regions in which the complete nucleotide sequence was obtained in both directions. Sample sequencing was conducted only in areas not covered by unshaded boxes; these regions matched K-12 or J96 sequences with respect to restriction endonuclease sites or nucleotide sequence identity. The positions of restriction fragments or PCR products that were used as probes for DNA hybridization are shown below the ORFs. The scale at the bottom is shown in kilobases. Cosmid clone 8-3f includes ~35 kb of the PAI beginning 3 kb to the left of the depicted left junction (0 kb) and extending to the center of the pap operon (~39 kb). Subclone 8HS9 includes PAI sequences from ~10 to 17 kb. Cosmid clone 5-4a includes ~37 kb extending from the center of the hemolysin gene cluster (~23.5 kb) to 5 kb to the right of the right junction (58 kb). Subclone 5H11B extends from ~38 to 49 kb.

Nucleotide sequence accession numbers. The sequences of the ORFs within the boundaries of the left and right junctions of the PAI of E. coli CFT073 have been assigned GenBank accession no. AF081283, AF081284, AF081285, and AF081286.

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

Novel genes inside the PAI. To determine whether newly described genes were present within the boundaries of the PAI, a 61-kb region was subjected to nucleotide sequencing. In this region, which included an apparent 58-kb PAI, we have identified 44 ORFs (Fig. 1). These are listed along with homologs and their accession numbers in Table 2. Four small gaps (~2, ~1, ~1, and <1 kb) that presented difficulties in sequencing, PCR amplification, and subcloning are also identified (Fig. 1). Among the sequenced ORFs are genes that appear to be involved in iron utilization and transcriptional regulation (see below).

Notable homologs. Four ORFs inside the left junction (defined as the sequences associated with lower numbers on the E. coli K-12 linkage map), prrA, modD, yc73, and fepC, represent an apparent iron transport system. The genes are contiguous and are predicted to be transcribed in the same direction; however, this has not been demonstrated experimentally. In other systems (25), the modD gene is part of a gene cluster involved in molybdenum transport, although no specific function has been ascribed to this gene. Another iron acquisition gene homolog, comprising the R4 ORF (terminology used in Fig. 1 and Table 2), appears to encode a homolog of an exogenous ferric siderophore receptor (9).

Insertion sequences and transposons. Six ORFs, designated HP1, HP2, R15, R14, R12, and R6, are related to insertion sequences and transposons which are common features of PAIs (26). The HP1 and HP2 ORFs represent the IS600 hypothetical 31- and 11-kDa proteins (24), respectively. R14 represents the transposase from insertion sequence IS629 (22). These two insertion sequences, which are members of the IS3 family and are found elsewhere in the K-12 genome (5), were originally identified in a strain of Shigella sonnei, a species closely related to E. coli (23).

Other features of the PAI. The PAI of strain CFT073 displays two other features common to such blocks of virulence genes. First, the GC content of the sequences, not including those of K-12 origin, is 42.9%. This value is significantly different from a value of 50.8% for the E. coli genome (2). Also, there appears to be segmentation with respect to unique PAI sequences and sequences of K-12 origin. Approximately 7 kb downstream of the left junction of the PAI, there is an 8-kb sequence, identical to that found in the K-12 genome, carrying six ORFs (Fig. 1). The hemolysin gene cluster hlyCABD follows this block.

PAI sequences are associated with virulent strains. To determine whether sequences found within the junctions of the PAI are common to other uropathogenic strains of E. coli, 11 probes were isolated along the length of the PAI (Fig. 1) and used to hybridize dot blots of genomic DNA isolated from clinical isolates (Table 3). A high percentage of isolates from patients with acute pyelonephritis or cystitis reacted with the probes (mean = 81%). These proportions were significantly higher than for strains from patients with catheter-associated bacteriuria (mean = 58%) or from fecal strains (mean = 22%) (P < 0.001) indicating that genomic sequences homologous to those of PAI sequences from strain CFT073 are found in most other strains recovered from patients with cystitis or pyelonephritis.

Hybridization signatures of clinical isolates. Since uropathogenic strains clearly carry sequences that are homologous to sequences in the first CFT073 PAI, we examined which probes reacted with each isolate in the strain collection. Each strain was assigned a signature based on a positive or negative hybridization with 11 probes (Table 4). Of the 35 strains that reacted with all 11 probes, 18 were pyelonephritis-associated strains and 17 were cystitis-associated strains; no fecal strains fell into this category (P < 0.007). The 72 strains that reacted with either 10 or all 11 probes included 39 of 67 (58%) pyelonephritis-associated strains, 23 of 38 (61%) cystitis-associated strains, 10 of 49 (20%) strains from patients with catheter-associated bacteriuria, but only 1 of 27 (4%) fecal strains (P < 0.0001). Of the 12 strains that reacted with none of the probes, 11 were fecal strains and 1 was from a patient with catheter-associated bacteriuria. The 20 strains that reacted with either none or only 1 probe included only 1 of 67 (1%) pyelonephritis-associated strains, none of the cystitis-associated strains, 3 of 49 (6%) strains from patients with catheter-associated bacteriuria, but 16 of 27 (59%) fecal strains (P < 0.0001). These results indicate that PAI sequences present in E. coli CFT073 are also found in other uropathogenic strains, especially those isolated from patients suffering with pyelonephritis and cystitis. These sequences are not generally found in fecal strains and are found less frequently in isolates from patients with catheter-associated bacteriuria.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

We have characterized a 61-kb region of DNA from E. coli CFT073, a pyelonephritis- and bacteremia-associated isolate, by isolation of overlapping cosmid clones, restriction endonuclease mapping, subcloning, hybridization, and nucleotide sequencing (Fig. 1). In this region, we have identified what can be defined as a PAI that includes 44 ORFs (Table 2). PAIs are typically larger than 30 kb (3), have a GC content lower than that of neighboring DNA (19), and have gene clusters positioned near each other which contribute to a single virulence property (26). The first PAI of CFT073 is 58 kb in size (Fig. 1), has a GC content of 42.9% (compared to 50.8% in K-12 genomic DNA), and includes the genes encoding HlyA hemolysin and P fimbriae.

Distinct segmentation and insertion sequences are also common features of PAIs, which suggests that DNA rearrangements mediated by illegitimate or recA-dependent recombination have occurred or that DNA has been inserted by transposons, phage, or integrons (19, 26). This 61-kb region shows some evidence of rearrangement with the K-12 genome, as two blocks of four and six ORFs within this region closely match entries in the complete K-12 genome sequence (Fig. 1, darkly shaded boxes). Another four genes, in the hylCABD cluster, encode the HlyA hemolysin and are nearly identical to sequences from E. coli J96, a notable uropathogenic strain. Also carried on this stretch of DNA is the pap operon, comprising of 11 ORFs (papIBAHCDJKEFG) encoding one of the two P fimbriae expressed by this strain (28); this pap operon encodes a class II PapG adhesin (data not shown). Insertion elements and transposases occur six times in the region, perhaps predisposing these sequences to rearrangement. Remaining are 19 ORFs, which are also listed in Table 2, beginning at the left junction of the PAI and moving toward the right junction.

Nucleotide sequences of the first 58-kb PAI of CFT073 reveal newly described genes. These gene sequences, as assayed by DNA hybridization, are present in virulent uropathogenic strains and are generally absent in nonvirulent strains. Therefore, we postulate that these newly described genes may represent virulence determinants that contribute to the pathogenesis of cystitis and acute pyelonephritis caused by uropathogenic E. coli. As we continue our studies on the pathogenicity islands of CFT073, we will select mutants with phenotypes that may relate to virulence such as iron uptake, metabolite uptake, and transcriptional regulation. We will undertake allelic exchange mutagenesis of specific genes and test these mutants by in vitro assays and, in vivo, by using the CBA mouse model of ascending UTI (13). By creating such mutants, we hope to determine which genes contribute to the virulence phenotype of uropathogenic E. coli.

It is likely, however, that strain CFT073 contains another PAI. For two other uropathogenic strains that have been studied closely, 536 and J96, each strain was found to contain two separate PAIs. For strain 536, PAIs of 190 and 70 kb are inserted at 97 min (within leuX) and 82 min (within selC), respectively (reviewed in reference 19). For strain J96, PAIs of 110 and >170 kb are inserted at 94 min (within pheR) and 64 min (within pheV), respectively (39). Because we know that strain CFT073 contains two complete pap operons encoding distinct P fimbriae (28) and that only one operon is found in the PAI described in this report (17), it is likely that these sequences are found in a separate PAI along with the genes encoding F1C fimbriae (foc [unpublished observation]). That CFT073 has another PAI of unknown size raises the possibility that a significant number of virulence genes, not detected on the PAI described here, also contribute to the virulence of this strain.

Finally, an important question raised by these studies is whether there are distinctions between strains isolated from patients with cystitis and strains from patients with pyelonephritis. Based on hybridization of chromosomal DNA with PAI probes, we are unable to make such a distinction. With the exception of the hly hemolysin-containing probe, there were no significant differences (P > 0.2) between the percentages of cystitis and pyelonephritis isolates that reacted with each of the 11 PAI probes (Table 3). In contrast, other hybridization studies which used specific probes (e.g., pap, hly, sfa, and foc) have revealed that a higher percentage of pyelonephritis-associated isolates than of cystitis-associated isolates reacted with these probes (6). It is now clear that uropathogenic isolates generally express adhesins that aid in colonization and toxicity (hemolysin and cytotoxic necrotizing factor), but this group of strains clearly does not have a single phenotype (6). It will be interesting to see whether cystitis and pyelonephritis strains are, in the future, delineated on the basis of specific genotypes.