INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Hemorrhagic colitis is caused by a number of serotypes of Shiga toxin-producing Escherichia coli (STEC) (14). Among the clinical STEC strains that have been isolated, a subset of enterohemorrhagic E. coli (EHEC) strains has been found which carry common sets of virulence genes that encode factors for attachment to host cells, elaboration of effector molecules, and production of two different types of Shiga toxins (22). The sets of virulence genes are found in the locus of enterocyte effacement (LEE) pathogenicity island, lambdoid bacteriophages, and a large virulence-associated plasmid (8, 9, 23, 25, 26, 31, 32). Population genetic analysis of EHEC and STEC strains has shown that EHEC strains comprise two divergent lineages, termed EHEC 1 and EHEC 2, that are only distantly related but apparently experienced similar pathways of virulence gene acquisition (24, 28, 38). The EHEC 1 lineage is comprised solely of a geographically disseminated cluster of strains with related genotypes bearing O157:H7 and O157:H serotypes, while the EHEC 2 lineage is serotypically and genotypically more diverse.

The O157 serotype can be found in genetically diverse populations of E. coli, apparently as a consequence of transfer of a common allele combination in the gnd-rfb region that encodes the O157 antigen (36). Among the populations bearing the O157 serotype, only the genetically related populations in the serotype O157:H7 EHEC 1 lineage, including the derived nonmotile O157:H populations (collectively referred to as O157 EHEC), carry the combination of virulence genes that includes the genes found in the LEE island, the large virulence plasmid, and the Shiga toxin-converting phages (38). A reconstruction of the evolutionary history of this lineage suggests that it arose from transfer of the O157 gnd-rfb region into an O55:H7 enteropathogenic E. coli host bearing the LEE island and an stx2-converting prophage (11, 28, 36, 38, 39). In subsequent steps, this ancestral population lost the ability to ferment sorbitol, was lysogenized by an stx1-converting phage, and finally acquired a mutation that inactivated the uidA gene, which resulted in a loss of -glucuronidase activity (11, 39).

The contemporary non-sorbitol-fermenting, -glucuronidase-negative O157:H7 clone is the EHEC that is most frequently isolated from hemorrhagic colitis patients in the United States, Canada, Japan, and the United Kingdom. Significant numbers of nonmotile O157 EHEC strains have been recovered from samples from continental Europe and Australia (6, 15, 19, 29). The nonmotile strains found in continental Europe are sorbitol positive and -glucuronidase positive, and phylogenetic analysis suggests that they comprise a population that diverged from an ancestral intermediate of the contemporary O157:H7 clone prior to loss of the sorbitol fermentation and -glucuronidase production traits (11, 19). In contrast, nonmotile O157 EHEC strains that are commonly isolated in Australia are sorbitol negative and -glucuronidase negative (29), suggesting that they are members of a more recently derived population that diverged from the lineage after loss of the sorbitol fermentation and -glucuronidase production traits.

In addition to populations of nonmotile variants, two distinct subpopulations of the widespread non-sorbitol-fermenting, -glucuronidase-negative O157:H7 clone have been detected in the United States by high-resolution genome comparison (20). Strains belonging to these two subpopulations, termed lineages I and II, appeared to be nonrandomly distributed among the bovine and human clinical isolates examined and could be distinguished by octamer-based genome scanning (OBGS) and by restriction fragment length polymorphism analysis with lambdoid phage probes. Since these studies were conducted exclusively with United States strains, it was not clear whether one or both of the subpopulations are regional subclones or whether they resulted from a more ancestral divergence event.

To determine the ancestry of the lineage I and II subpopulations, we studied the phylogenetic relationships of representative lineage I and II O157:H7 strains from the United States to O157 EHEC strains from the geographically isolated continent Australia. The phylogeny inferred from cladistic and statistical analyses of OBGS data shows that populations of O157 EHEC strains isolated from the two continents comprise two lineages corresponding to the distribution of lineage I and II United States strains, indicating that the divergence of the two lineages was ancestral and predated the arrival of the lineages on one or both continents. Results obtained by sorting polymorphic OBGS markers relative to the inferred phylogeny of the strains are consistent with this conclusion and were used to identify sets of monophyletic markers that distinguished divergence of the lineages and emergence of three independent lineage II subpopulations. Two mechanisms of phylogeographic variation were also detected, and these mechanisms corresponded to expansion of regional subclones and accumulation of regional polymorphisms in a lineage-independent fashion. DNA sequence analysis of monophyletic markers provided a set of specific genome alterations that marked divergence of the lineages and subpopulations. Sequencing of polyphyletic markers suggested that random drift in the populations is frequently a result of common alterations that randomly create or destroy OBGS priming sties and events associated with movement and recombination among lambdoid prophages.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacterial strains and growth conditions. Strains used in this study and their known characteristics are shown in Table 1. United States isolates have been described previously (10, 13, 20, 21, 33) and were chosen because they represent the genetic diversity of the strains in the United Stares that have been examined by OBGS analysis. Strains from Australia and New Zealand were chosen on the basis of temporal and spatial representation of strain collections. The set of 15 human isolates included 2 isolates from patients with hemolytic-uremic syndrome in New Zealand and 13 clinical isolates from patients with diarrhea or hemolytic-uremic syndrome in Australia (kindly supplied by R. Robins-Browne and D. Lightfoot, University of Melbourne, Parkville, Victoria, Australia). Isolates of animal origin were obtained from bovine or ovine sources in Australia during various pre- and postharvest surveys of animals. All strains were maintained as frozen stock preparations and were minimally propagated on Luria-Bertani agar or broth.

OBGS analysis. DNA was prepared from 10-ml cultures of each strain by standard methods, and 50 ng was used for each OBGS reaction. Reactions were primed with combinations of decamer primers, each of which had a 5' AT dinucleotide followed by a skewed octamer sequence (20). The 5' AT dinucleotide, which was added to limit steric effects of the 5' fluorophore on annealing of the octamers, was also included in the sequences of the unlabeled octamers. For each reaction, a labeled primer (containing a leading strand-biased octamer) was combined with a cold primer (containing a lagging strand-biased octamer). The following primers were labeled at their 5' ends with the fluorophore IRD800 (Li-Cor, Inc., Lincoln, Nebr.): OCT3B (ATGCTGGTGG), OCT4B (ATGCTGGCGG), OCT21 (ATGCGCTGGA), and OCT22 (ATCTGCGCAA). The following unlabeled primers were used with the fluorescent primers: OCT4C (ATCCGCCAGC), OCT6C (ATGCCAGCGC), OCT12C (ATGCCGCCAG), OCT13C (ATTGCGCCAG), and OCT19C (ATCTTCCAGC). The combinations OCT3B-OCT4C, OCT3B-OCT12C, OCT4B-OCT19C, OCT21-OCT13C, and OCT22-OCT6C were used for OBGS studies. The fluorescent OBGS products were resolved by fragment analysis with Li-Cor 4200 automated DNA sequencers, and data were collected with the Global IR² system (Li-Cor, Inc.). Strain MC1061, a K-12 derivative, and a distantly related bovine STEC strain, AU133c, were included in the analyses to assess the ancestral condition of the markers. Band patterns were printed with an Alden 9315 thermal printer and scored as described previously (20).

OBGS data analysis. Binary strings, representing the presence or absence of OBGS products ranging from 200 to 1,500 bases long for each strain, were constructed in Microsoft Excel spreadsheets from gel images of products obtained with individual OBGS primer combinations. The binary strings for each of the primer combinations were then added head to tail for each strain with FORMATALL, a Perl-based program that creates composite files from fragment analysis and microarray data files and generates output files compatible with PAUP and other clustering programs (Wise and Benson, unpublished data). Genetic distance was calculated by determining the mean number of character differences from all pairwise alignments of taxa, and phylogenetic trees were resolved by neighbor joining (NJ) and statistical (bootstrap) sampling of the data using algorithms in PAUP, version 4.0 (34). Trees generated from binary strings resulting from individual primer combinations were in general agreement with the tree generated from the composite binary strings resulting from all primer combinations by both NJ and bootstrap methods.

OBGS band isolation and sequencing. Polymorphic OBGS products that indicated divergence of the lineages and clades were identified from images of gels in which the samples were loaded in phylogenetic order. Representative bands for each class were excised from automated sequencing gels in which electrophoresis was halted immediately after elution of the appropriate molecular weight markers past the scanning optical unit of the sequencer. The bands of interest were then localized in a gel by drying the gel on filter paper and scanning the dried gel with an Odyssey near-infrared fluorescence scanner (Li-Cor, Inc.). After localization and excision, fluorescently labeled DNA was eluted from a gel slice by boiling for 5 min in water, followed by phenol extraction and ethanol precipitation. The precipitated pellet was redissolved in 50 µl of water, and 10 µl was used as a template in a 20-µl OBGS reaction mixture containing the primer combination and reagents that were present in the original mother reaction mixture. The reamplification products were then electrophoresed alongside the mother reaction mixtures to check band purity and efficiency of reamplification. The reamplified products were then used as templates in a second round of reamplification performed with the following OBGS primers tailed with M13 forward (M13F) and reverse (M13R) primer sequences: M13FOCT3B (CACGACGTTGTAAAACGACATGCTGGTGG), M13FOCT21 (CACGACGTTGTAAAACGACATGCGCTGGA), M13FOCT22 (CACGACGTTGTAAAACGACATCTGCGCAA), M13ROCT4C (GGATAACAATTTCACACAGGGCCGCCAGCT), M13ROCT6C (GGATAACAATTTCACACAGGATGCCAGCGC), M13ROCT12C (GGATAACAATTTCACACAGGATGCCGCCAG), and M13ROCT13C (GGATAACAATTTCACACAGGATTGCGCCAG). Tailed PCR products were purified from agarose gels with a QIAquick gel extraction kit (Qiagen, Inc., Valencia, Calif.). Purified products were cloned into pTOPO (Invitrogen, Carlsbad, Calif.) and sequenced with T7 and T3 sequencing primers. DNA sequence analysis was performed for both strands.

Nucleotide sequence accession numbers. The GenBank accession numbers for the DNA sequences determined in this study are AF368044 (226C-3), AF368045 (226C-4), AF368046 (3B4C-5), AF368047 (226C-13), AF368048 (226C-14), AF368049 (2212C-1), AF368050 (3B12C-2), AF368051 (3B4C-2), AF368052 (2113C-1), and AY036017 (3B4C-1).

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

OBGS analysis of United States and Australian O157:H7 isolates. To examine the genetic relationships of Australian O157 EHEC strains and United States lineage I and II O157:H7 strains, we used OBGS to analyze a set of 37 isolates obtained from cattle and humans in Australia and eight representative United States isolates belonging to each of the two lineages (Table 1). The Australian strains were chosen because they represented broad ranges of temporal and spatial diversity with respect to origin and source. The United States isolates belonging to lineages I and II represented the spectrum of genetic diversity observed for strains derived from these lineages (20). Five different OBGS primer combinations were used for each strain, and OBGS products obtained with each primer combination were converted to binary characters. Binary strings for each primer combination were added head to tail, which resulted in composite binary strings consisting of 1,159 characters for each strain. Strain AU133c, a bovine-derived STEC strain that is distantly related to O157:H7 based on the distribution of OBGS polymorphisms, was included in the strain set as an outgroup.

View larger version (29K): [in this window] [in a new window]   FIG. 1.   Genetic relationships of E. coli O157: H7 strains from the United States and Australia. The dendrogram was produced by NJ analysis of binary files representing 1,159 OBGS products from each strain obtained with five OBGS primer combinations (tree length, 1,261; consistency index, 0.6772; retention index, 0.7573; 328 characters). Bootstrap percentages, derived from 1,000 replications of an NJ search, are indicated at the nodes for which the percentages are more than 50%. The tree was rooted by using strain AU133c, a bovine STEC isolate, as the outgroup. The designations of strains originating from Australia and New Zealand begin with AU and NZ, respectively. Human clinical isolates are indicated by boldface type, and their designations end with H.

Genetic markers and phylogeny of United States and Australian O157:H7 and O157:H EHEC strains. If the clades inferred from the phylogenetic analysis indeed constitute distinct subpopulations, then signatures that indicate their phylogeny and diversification should be present and observable in polymorphic OBGS products. To mine the putative signatures from the OBGS data, we developed a marker-sorting method that identified and quantified polymorphisms with monophyletic and polyphyletic distributions. This method required classification of polymorphisms into four classes, classes I to IV, based on four possible patterns of distribution relative to user-specified groups. Relative to phylogeny, the four distribution classes are shown in Fig. 2; to obtain the data, the OBGS samples were loaded in inferred phylogenetic order on the gel.

View larger version (117K): [in this window] [in a new window]   FIG. 2.   Distribution of different classes of polymorphisms. Segments of OBGS images illustrating polymorphic bands of the four distribution classes (classes I to IV) are shown in the four panels. The samples were loaded in phylogenetic order, as listed across the top. Each image segment was reduced fivefold vertically, and the molecular weights for the top and bottom regions of each image segment are indicated on the right. The image segments were obtained by using the following primer combinations: class I, OCT22-OCT6C; class II, OCT22-OCT12C; class III, OCT3B-OCT4C; and class IV, OCT3B-OCT4C. The bands of interest are indicated by arrows.

View larger version (68K): [in this window] [in a new window]   FIG. 3.   Sorting of polymorphic OBGS products by phylogeny. Binary coding of the presence or absence of bands in five OBGS reaction mixtures was rendered in phylogenetic order, and the polymorphic positions were grouped by class of distribution by using sorting functions in Microsoft Excel. Polymorphic addresses present in three or more strains were sorted and rendered as cells with different colors depending on the class of distribution pattern. Occurrences of class I polymorphisms are red, occurrences of class II polymorphisms are pink, occurrences of class III polymorphisms are green, and occurrences of class IV polymorphisms are blue. The order of strains is indicated on the left.

Polyphyletic pattern of genome variation and phylogeographic variation. Relative to the inferred phylogeny, monophyletic markers comprised only about one-fourth of the polymorphisms that were detected and accounted for the phylogenetic signal that distinguished only three of the five clades. If we excluded three class III polymorphisms and 11 class IV polymorphisms that could be explained simply by loss of an OBGS product from monophyletic clusters of strains, the remaining 97 polymorphic OBGS products were distributed among polyphyletic groups of strains. The ratio of monophyletic polymorphisms to polyphyletic polymorphisms implies that genome diversification occurred rapidly by random drift and accumulation of common alleles and that selection only periodically fixed alleles in subpopulations.

View larger version (30K): [in this window] [in a new window]   FIG. 4.   Sorting of polymorphic OBGS products by geography. Binary coding of the presence or absence of bands in five OBGS reaction mixtures was rendered in geographic order and secondarily in phylogenetic order by sorting functions in Microsoft Excel. Only the geography-specific polymorphic addresses present in three or more strains are shown. The cells containing geographic bands were rendered in different colors by conditional formatting functions according to the class of distribution relative to phylogeny (red, class I; pink, class II; green, class III; blue, class IV). The order of the strains is indicated on the left.

DNA sequence analysis of monophyletic lineage markers and clade divergence. In order to identify the nature of the OBGS polymorphisms that resulted in the inferred patterns of phylogeny, we purified, cloned, and sequenced representative monophyletic polymorphisms that distinguished the two lineages and the three lineage II clades. Bands were purified from OBGS reaction mixtures by using a two-dimensional y-axis scanner to locate the bands in gels in which migration had been stopped prior to migration of the bands past the optical unit of the sequencer. After excision, purification, and reamplification, the labeled reamplified products were then checked for purity by electrophoresis alongside the mother reaction mixture preparations from which they were originally excised (Fig. 5). Once their identities were confirmed, the reamplified products were tailed, cloned, and sequenced. The DNA sequences were then compared to the genome sequences of the lineage I EDL933 O157:H7 strain, (Fig. 1), the Sakai O157:H7 strain (17, 27), and the MG1655 K-12 strain (1a).

View larger version (54K): [in this window] [in a new window]   FIG. 5.   Purification and reamplification of the class I OBGS product. The excised class I polymorphic bands obtained with the OCT22-OCT6C primer combination were purified from the reaction products of strains 93-001 and FRIK920 and were reamplified (Reamp) with the same primer combination. The reamplification products were then electrophoresed alongside the original OBGS reaction mixtures to check for purity. The arrows indicate the positions of the reamplified bands that comigrated with the bands in the mother reaction mixture preparations from which they were excised.

DNA sequence analysis of polyphyletic genome alterations. Since clades A and B are distinguished from one another by geographic polymorphisms, we cloned and sequenced a class IV band obtained with primer combination OCT3B-OCT12C, designated 3B12C-2 (accession number AF368050), that was found only in Australian strains except clade D strains AU1808, AU1809, AU1817, and AU1822. There was no apparent band obtained with the same primer combination for the opposing pattern of distribution. The 3B12C-2 sequence aligned with a segment of the EDL933 and Sakai genome sequences spanning the folB and bacA genes, which encode dihydroneopterin aldolase (16) and a putative undecaprenol kinase that confers resistance to bacitracin at a high copy number, respectively (3). Four single-nucleotide polymorphisms (SNPs) were present in the 3B12C-2 sequence compared to the EDL933 and Sakai sequences, but six SNPs were conserved in the 3B12C-2, EDL933, and Sakai sequences compared to the MG1655 sequence, supporting the notion that unique positions in the 3B12C-2 marker were derived and that the allele was probably not acquired. Two of the derived SNPs in clade B are silent, while the other two result in an isoleucine-to-threonine substitution at position 26 of the bacA open reading frame (ORF) and a valine-to-alanine substitution at position 314 of the folB ORF. In addition to the internal SNPs, sequences obtained from the 3B12C-3 band and the EDL933 and Sakai genomes also diverged immediately upstream of the OCT3B and OCT12C binding sites, suggesting that mismatches with the AT dinucleotide spacers of the octamer primers are probably the reason why the band did not appear to be amplified from EDL933 and other United States strains or K-12 control strain MC1061 in our OBGS reaction mixtures.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Comparative genome analyses of sorbitol-negative, -glucuronidase-negative E. coli O157:H7 strains previously showed that two distinct lineages could be detected in the United States (20). In this study, we found that the divergence of these lineages was probably an ancestral event that preceded geographic dissemination. This conclusion was based on phylogenetic inferences from OBGS data, the distribution of genome alterations detected by OBGS, and the nature of alterations in representative loci that mark the pattern of descent.

Because we sorted markers using the inferred phylogeny and identified conserved polymorphisms using the same phylogeny, the apparent circularity indicates that it is formally possible that some other process is responsible for the distribution of markers. For example, class I and II polymorphisms could be a result of lateral spread of mobile elements among different populations rather than a result of a stepwise pattern of vertical descent. However, the nature of the representative monophyletic alterations that were determined argues against this hypothesis being the sole explanation for the data since the class I polymorphism and two of the three class II polymorphisms are derived states in chromosomal genes and are not linked to mobile elements. Furthermore, sequence analysis of additional class I markers, identified in our laboratory as part of a systematic identification of all conserved markers of lineage divergence, showed that each marker is in an independent locus (Kim and Benson, unpublished), implying that the markers most likely originated through stepwise descent rather than convergent evolution. Although each of the additional class I markers represents an independent event, nearly one-half of them resulted from gene acquisition events, movement of insertion sequences, and movement or recombination within prophage, implying that the conserved polymorphisms comprise a mosaic of mutation, recombination, and acquisition events fixed by one or more selective bottlenecks.

Extended model for evolution of extant O157 EHEC populations. Based on the inferred phylogeny and the evidence that vertical descent of conserved markers occurred, we propose extension of the evolutionary model for contemporary O157:H7 populations (11) to include additional steps subsequent to the emergence of the sorbitol-negative, -glucuronidase negative ancestor (Fig. 6). The most recent common ancestor of contemporary E. coli O157:H7 populations, designated A6 by Feng et al. (11), is believed to have been derived from an LEE⁺ Stx2⁺ O157:H7 ancestor, designated A3, through a stepwise sequence of events that included lysogenization by an stx1-converting phage, acquisition of the TAI locus encoding tellurite resistance and the Iha adhesin, and sequential loss of sorbitol fermentation and -glucuronidase production (11, 35). Evolution from A3 to A6 can be inferred from multilocus genotyping data, as well as from cumulative character states in the uidA alleles of A3, A5, and A6 descendants (11), culminating in a dinucleotide insertion at position 685 observed in the EDL933 and Sakai genomes that apparently inactivated the gene in A6 descendants (17, 27). Prior to these events, a nonmotile lineage diverged from A3, leading to ancestor A4 of populations of sorbitol-fermenting, -glucuronidase-positive O157:H strains that have been found in continental Europe (11, 19).

View larger version (22K): [in this window] [in a new window]   FIG. 6.   Stepwise model of the evolution of O157: H7 and O157:H EHEC strains. The model is based on that of Feng et al. (11). Each oval represents a progenitor, and evolution proceeds from left to right from ancestral to contemporary states. The A3 to A6 ancestors are designated as described by Feng et al. (11). The markers of events are indicated as follows: Sor, inability to ferment sorbitol; uidA, loss of -glucuronidase activity; TAI, acquisition of the tellurite resistance-adherence island (35). Specific polymorphisms marking the divergence of A7 and clades A to E are described in the text.

Distribution of human and bovine isolates. Although descendants of both the A6 (lineage I) and A7 (lineage II) ancestors were detected among the Australian strains, most of the strains tested, including several human clinical clade D and E strains from Australia, were derived from the A7 ancestor. This finding is in contrast to the results of OBGS analysis of human and bovine O157:H7 strains from the United States, where few human isolates were observed among lineage II strains (20). There are several possible explanations for the observed differences. First, although in these studies we used strains that are temporally and spatially diverse in terms of their origins, it is possible that the differences reflect bias in the relatively small sets of strains. A second possibility is that differences in animal production and food preparation practices between the two countries studied contribute to bias in transmission. And finally, it is also possible that clade D and E strains have virulence properties that are distinct from those of their clade C counterparts. It is worth noting that most bovine strains from the United States that have been tested by the OBGS method cluster with the clade C strains, while the United States clade E strain is closely related to the few human lineage II isolates identified in the United States studies (20). We are currently devising a multiplex assay, based on class I and II polymorphisms, which will provide a rapid method for strain classification that can be used to statistically evaluate the differential virulence hypothesis.

Generation of genome diversity. Although the small set of class III and IV polymorphisms that we sequenced does not allow quantitative conclusions, the fact that we identified alterations in two different genomic sequences and two different prophage sequences implies that variation in each sequence contributed to genome diversification. The polyphyletic alterations detected in genomic sequences appear to be SNPs that create or destroy OBGS primer binding sites, and the results of comparisons of SNPs inside these regions are consistent with the hypothesis that the alleles are derived. Among the polyphyletic alterations detected in prophage sequences, we found one example of a polymorphism that arose either from lysogenization by a unique phage or by a recombination event that generated a chimeric segment in a resident cryptic prophage. Due to the high degrees of similarity among prophages and cryptic prophages that reside in O157:H7 genomes (17, 27), the diversity of prophage contents of different strains (7, 17, 20, 27, 30), and the modular nature of phage genomes (2, 4, 5), it is difficult to determine whether polymorphic OBGS products carrying chimeric prophage sequences arise from integration and excision of lysogenic phages or by recombination. Several of the class III and IV polymorphisms can be found in common sets of polyphyletic strains, which supports the view that the polymorphisms are physically linked and arise from movement of phages into and out of the genome. On the other hand, recombination events in prophage sequences could also appear as linked sets of class III and IV polymorphisms since highly conserved modules of phage genomes from similar and diverse hosts can vary significantly in length (2, 4, 5, 12, 18). Regardless of the precise mechanism of phage-mediated variation, it seems that infection of strains by bacteriophages is common and offers significant opportunities for lysogenization and recombination events to occur since the populations occupy phage-dense environments, such as the rumen.