Escherichia coli Serotype O55:H7 Diversity Supports Parallel Acquisition of Bacteriophage at Shiga Toxin Phage Insertion Sites during Evolution of the O157:H7 Lineage

Abstract

Enteropathogenic Escherichia coli (EPEC) continues to be a leading cause of mortality and morbidity in children around the world. Two EPEC genomes have been fully sequenced: those of EPEC O127:H6 strain E2348/69 (United Kingdom, 1969) and EPEC O55:H7 strain CB9615 (Germany, 2003). The O55:H7 serotype is a recent precursor to the virulent enterohemorrhagic E. coli O157:H7. To explore the diversity of O55:H7 and better understand the clonal evolution of O157:H7, we fully sequenced EPEC O55:H7 strain RM12579 (California, 1974), which was collected 1 year before the first U.S. isolate of O157:H7 was identified in California. Phage-related sequences accounted for nearly all differences between the two O55:H7 strains. Additionally, O55:H7 and O157:H7 strains were tested for the presence and insertion sites of Shiga toxin gene (stx)-containing bacteriophages. Analysis of non-phage-associated genes supported core elements of previous O157:H7 stepwise evolutionary models, whereas phage composition and insertion analyses suggested a key refinement. Specifically, the placement and presence of lambda-like bacteriophages (including those containing stx) should not be considered stable evolutionary markers or be required in placing O55:H7 and O157:H7 strains within the stepwise evolutionary models. Additionally, we suggest that a 10.9-kb region (block 172) previously believed unique to O55:H7 strains can be used to identify early O157:H7 strains. Finally, we defined two subsets of O55:H7 strains that share an as-yet-unobserved or extinct common ancestor with O157:H7 strains. Exploration of O55:H7 diversity improved our understanding of the evolution of E. coli O157:H7 and suggested a key revision to accommodate existing and future configurations of stx-containing bacteriophages into current models.

INTRODUCTION

Enteropathogenic Escherichia coli (EPEC) is a leading cause of mortality and morbidity affecting infants in developing countries (52, 54, 68). Until the 1970s, it was also a significant pathogen in the developed world, and it was the first diarrheal E. coli recognized as a distinct pathotype (44, 56). In spite of this, only two EPEC genomes have been fully sequenced: those of the typical, model EPEC O127:H6 strain E2348/69 (30), which was isolated in 1969 in the United Kingdom, and of the contemporary, atypical EPEC O55:H7 strain CB9615 (77), which was isolated in Germany in 2003. The latter serotype has been established as a recent precursor to the virulent enterohemorrhagic E. coli (EHEC) serotype O157:H7 (18, 71), which can cause disease in all age groups. Extensive analysis of the diversity of the O157:H7 serotype has provided insight into its continued evolution (15, 34, 42, 75, 76) and has led to a search to define genetic properties that could identify more virulent strains of related O157:H7, nonmotile O157, and non-O157 EHEC (3, 4, 5, 8, 32, 63, 74).

Following the first reported outbreaks of hemorrhagic colitis linked to the O157:H7 serotype in 1982 (62), the clonal nature of O157:H7 (70) and its close evolutionary relationship to O55:H7 (71) were established. Subsequent analyses proposed and refined a stepwise evolutionary model for the emergence of O157:H7 from O55:H7 based on multilocus enzyme electrophoresis typing, as well as phenotypic and genetic analysis of existing isolates of both serotypes (18, 20, 42, 72). This stepwise model included the following: acquisition of Shiga toxin-encoding stx genes (18, 65), a well-characterized virulence factor for EHEC; phenotypic markers, such as loss of the ability to ferment sorbitol (SOR) or loss of β-glucuronidase (GUD) activity; and nucleotide changes in the uidA (or gusA) (encodes GUD) and fimA and fimH fimbrial genes (17, 50, 64). Meanwhile, analysis by multilocus sequence typing (MLST) confirmed a hypothesis of parallel evolution of both EPEC and EHEC within related serogroups via multiple acquisitions of the locus of enterocyte effacement (LEE), the large virulence plasmids EAF and pO157, respectively, and stx-containing lambda-like bacteriophages in the case of EHEC (60, 69, 73).

EPEC strains themselves constitute a diverse genetic group. Although most EPEC strains possess the LEE, insertion sites vary among different clades due to parallel horizontal transfer events (73). Moreover, EPEC isolates are classified as typical or atypical based upon the presence of the EAF plasmid, which is found only in typical EPEC (68). Typical EPEC strains cause disease in infants, with adults serving as asymptomatic carriers, whereas atypical EPEC strains cause disease in both adults and children and have animal reservoirs, traits shared by most EHEC strains (52). Beyond a diversity of insertion sites, the sequences of key LEE-carried genes vary among EPEC isolates. The LEE carries structural genes for a type III secretion system (T3SS), as well as several effectors that are translocated by the T3SS into host cells (33, 66). As elucidated using the typical EPEC pathogen E. coli O127:H6 strain E2348/69, the intimin gene eae and the translocated intimin receptor gene tir are key players involved in the formation of attaching and effacing lesions or pedestals on host cells (46, 52). EHEC O157:H7 strains also possess the LEE pathogenicity island, but due to a mutation in their tir gene that prevents phosphorylation at tyrosine 474, pedestal formation in O157:H7 also requires a non-LEE gene, tccP, encoding the Tir cytoskeleton coupling protein (6). Most atypical EPEC O55:H7 strains also possess a tccP gene, along with the variant tir_Sakai allele that is missing the tyrosine residue at position 474. Of note, some O55:H7 strains do not possess tccP and instead harbor a tir allele similar to the one from the typical EPEC strain E2348/69, designated tir_E2348/69 (29). Furthermore, among O55:H7 strains that possess tccP, differences in gene length that reflect a variable number of proline-rich repeats have been found (23).

In order to fully understand the emergence of the O157:H7 serotype from its O55:H7 precursor, including the evolution of key virulence factors, further exploration of O55:H7 diversity is necessary. Moreover, because the genetic diversity of E. coli is determined largely by mobile genetic elements (41, 53, 55), additional complete genome sequences are required in order to resolve accurate placement of multiple, nearly identical sequences representing highly conserved repeat copies within multiple mobile prophages and plasmids. To this end, we sequenced to completion EPEC O55:H7 strain RM12579 (California,1974), which was chosen due to its proximity in time and space to the earliest documented O157:H7 strain isolated in the United States, a California isolate from 1975 (CDC 2886-75) (62, 70). Comparative genomic analysis with the first completed O55:H7 genome, that of strain CB9615 (77), and with targeted genes in O55:H7 isolates from different regions around the world revealed significant genomic diversity within this serotype and suggested refinements to the O157:H7 evolutionary model.

MATERIALS AND METHODS

Bacterial strains and media.

Escherichia coli O55:H7 strain RM12579 was collected in December 1974 by the California Department of Public Health from a child less than 5 years old. Bacterial strains were stored at −80°C in Difco Luria-Bertani (LB) broth (Becton Dickinson and Company) with 20% glycerol and grown in LB broth or on LB agar (LB broth with 1.5% agar) at 37°C (Table 1). Strains were tested for the ability to utilize sorbitol by plating on sorbitol MacConkey agar with BCIG (5-bromo-4-chloro-3-indolyl-β-d-glucuronide) (Oxoid) or for GUD activity by plating on Rainbow agar O157 (Biolog), followed by incubation at 37°C for 20 to 24 h. Resistance to ampicillin and streptomycin was determined by plating on LB agar containing 100 μg/ml of ampicillin (Amresco) or 30 μg/ml of streptomycin (Sigma-Aldrich).

Table 1.

Bacterial strains

USDA namea	Alternate name(s)a	E. coli serotype	Yr isolated	Country (state or city) where isolateda	Source datah	Reference(s)
RM1239	96A13466	O157:H7	1996	USA (California)	Human, outbreak, apple juice	This studyb
RM1273	CDC 3417-86	O157:H7	1986	USA (Washington)	Human	ATCC 43888c
RM2011	DEC4A	O157:H7	1977	Argentina	Calf	59, 71
RM2012	DEC5A	O55:H7	1950	USA (New York)	Human	59, 71
RM2027	DEC5B	O55:H7	1979	USA (Florida)	Human	59, 71
RM2041	DEC4C	O157:H7	1983	Egypt	Buffalo	59, 71
RM2042	DEC5C	O55:H7	1966	USA (New Jersey)	Human	59, 71
RM2057	DEC5D	O55:H7	1965	Sri Lanka	Human	59, 71
RM2071	DEC4E	O157:H7	1988	Denmark	Human	59, 71
RM2072	DEC5E	O55:H7	1963	Iran	Human	59, 71
RM2163	CDC 5644-62	O55:H7	1962	USA (Illinois)	Human	24
RM3641	93-8160	O157:H7	1993	Canada	Female, HUS	This studyd
RM3650	91-8231	O157:H7	1991	Canada	Female	This studyd
RM3651	92-9104	O157:H7	1992	Canada	Female	This studyd
RM3654	91-8071	O55:H7	1991	Canada	Female, 2 yr, HUS	This studyd
RM4876		O157:H7	2005	USA (California)	Watershed	This study
RM5203		O157:H7	2005	USA (California)	Human	This studyb
RM6540	SO2007-0454	O157:H7	2006	USA(Minnesota)	Human	This study
RM6605	12R-2	O157:H7	2007	USA(California)	Cow	37
RM6693	SC-14445	O157:H7	2007	USA (Idaho)	Male, 5 yr	This studye
RM6649	RIMD 0509952	O157:H7	1996	Japan (Sakai)	Human, outbreak, radish sprouts	ATCC BAA-460 (26)
RM7015	37-S	O157:H7	2007	USA (California)	Cow	37
RM7054	30-S	O157:H7	2008	USA (California)	Cow	37
RM7415	F121-34	O157:H7	2008	USA (California)	Cow	This study
RM7474	F54-47	O157:H7	2008	USA (California)	Cow	This study
RM7632	G07-0632	O157:H7	2007	USA (Oregon)	Human	This studyf
RM7709	G08-0640	O157:H7	2008	USA (Oregon)	Human	This studyf
RM8428	F284-188	O157:H7	2008	USA (California)	Cow	This study
RM8457	W111-191	O157:H7	2008	USA (California)	Water, cattle trough	This study
RM8530	CDC 2886-75	O157:H7	1975	USA (California)	Female, 50 yr, HC	62, 70
RM8559	S994-18-249	O157:H7	2009	USA (California)	Soil, pasture	This study
RM12506	BB2, C523-03	O55:H7	2003	Denmark	Male, 55 yr, HC	19
RM12579	10591-12-74	O55:H7	1974	USA (California)	Human, <5 yrs, urine	This studyb
RM13616	99107(90)	O55:H7	1999	Brazil	Male, 9 mo, HC	This studyg
RM13617	21443(24)	O55:H7	2001	Brazil	Female, 6 yr, persistent diarrhea	This studyg
RM14504	4694-9-75	O55:H7	1975	USA (California)	Human, <5 yr, stool	This studyb
RM14505	2022-7-76	O55:H7	1976	USA (California)	Human, <5 yr, stool	This studyb
RM14506	5150-10-74	O55:H7	1974	USA (California)	Human, <5 yr, urine	This studyb
RM14507	7305-10-74	O55:H7	1974	USA (California)	Human, <5 yr, stool	This studyb

^aNames in boldface are used in this report.

^bStrain provided by S. Abbott or M. Janda of the California Department of Public Health.

^cStrain provided by J. Keen of the USDA Meat Animal Research Center, Clay Center, NE.

^dStrain provided by M. Bosilevac of the USDA Meat Animal Research Center, Clay Center, NE.

^eStrain provided by C. Ball of the Idaho Bureau of Laboratories.

^fStrain provided by W. Keene and R. Vega of the Oregon State Public Health Laboratory.

^gStrain provided by L. Riley of the University of California at Berkeley and D. Monteiro Girao of the Federal University of Rio de Janeiro.

^hHC, hemorrhagic colitis; HUS, hemolytic uremic syndrome.

Genome sequencing strategy.

Genome sequencing was conducted in parallel using both the 454 GS FLX Titanium (Roche 454) and SOLiD 3.0 system (Life Technologies) sequencing platforms. Shotgun and 8-kb paired-end libraries for sequencing using the 454 GS FLX Titanium system were prepared following the GS FLX Titanium General Library Preparation Method manual (April 2009) with the gel cut option and GS FLX Titanium Paired-End Library Prep-20-8kbSpan Method manual (October 2009) protocols (Roche 454). The libraries were sequenced using the FLX Genome Sequencer, resulting in totals of 277,841 and 244,699 reads that passed filters for general and paired-end libraries, respectively. DNA sequencing using the Applied Biosystems SOLiD system was performed essentially as previously described (11). However, the mate-paired library had an insert size of 2.6 kb, and the library was sequenced on a single flow cell of a SOLiD 3.0 instrument to yield nearly 4 Gb of paired 25-bp reads.

The 454 system provided 33× coverage of the genome, and the GS de novo assembler (Roche 454, version 2.3) yielded 246 contigs and 9 scaffolds using default assembly settings. An optical map of E. coli O55:H7 strain RM12579 was constructed using the Argus optical mapping system (OpGen) and used to verify the order of 454-generated scaffolds and contigs. Five scaffolds were determined to be plasmids, whereas three scaffolds corresponded to repetitive chromosomal elements (5S/23S rRNA, rhs gene, and bet and gam genes of lambda-like bacteriophages), leaving one large scaffold spanning the complete chromosome. Comparison of this large scaffold to the NotI-restricted optical map revealed one major inversion in the scaffold assembly, the breakpoints of which were determined to lie in nearly identical regions of two lambda-like bacteriophages. During the genome assembly process, the previously published E. coli O55:H7 strain CB9615 genome (NC_013941) was used to assist and verify contig placement within the scaffold assembly.

Mate-paired reads from the SOLiD system with a 2.6-kb average insert length covered the genome at 730× depth. Contigs assembled using the SOLiD system de novo accessory tools 2.0 were used to span gaps in the 454 scaffold as well as to verify rRNA regions that were not adequately covered. SOLiD reads were also used to verify chromosomal and plasmid contig sequences and to correct errors, particularly indels in homopolymeric tracts. Final gap closure was achieved using targeted PCR and Sanger sequencing. PCR was performed as described below, followed by gel extraction (Qiagen) and sequencing using BigDye Terminator v3.1 and XTerminator reaction cleanup (Applied Biosystems), and run on either an AB3130xl or AB3730xl sequencer (Applied Biosystems).

Designation of phage boundaries.

Duplication of a core attachment (att) sequence to the left (attL; integrase side) and the right (attR; tail protein side) of an integrated prophage results from lysogeny (14). In the O157:H7 Sakai strain, most phage boundaries excluded the attL, allowing the sequence of preceding tRNA or other genes to remain intact (26). However, att core regions were subsequently redefined for some Sakai prophages (2). In most cases, the original Sakai att regions were used to define phage boundaries in strain RM12579 (see Table S1 in the supplemental material). However, the redefined Sp9 att was used to define boundaries for the homologous Cp6, and boundaries for Cp1 and Cp2 were adjusted to reflect a predicted 46-bp core att region for Cp1 (ATTCGTAATGCGAAGGTCGTAGGTTCGACTCCTATTATCGGCACCA) identified in this study. Because Cp1 and Cp2 were integrated in tandem, this required an adjustment to the left side of the Cp2 phage compared to Sp2. Similarly, Cp3 boundaries were adjusted, compared to Ep3, in order to reflect a predicted 45-bp core att region (CTTCTAAGTCGTGGGCCGCAGGTTCGAATCCTGCAGGGCGCGCCA). Finally, two adjacent halves of a lambda-like prophage in strain CB9615 (Ep8 and Ep9) have been combined and named as one prophage in strain O55:H7 RM12579 (Cp8), which was flanked by a repeat sequence (GAAATCCATAA) that may represent a core att sequence.

PCR for virulence gene testing and identification of phage insertion sites.

PCR was performed on either genomic DNA (Wizard Genomic DNA purification kit; Promega) or whole-cell lysates after overnight growth in LB. PCR conditions for virulence genes were as described in publications listed in Table S2 in the supplemental material for previously published primers. For newly designed primers and all phage insertion site assays, Taq polymerase (NEB) was used for products of <5 kb and the Phusion system (Finnzymes) was used for products of >5 kb. Annealing temperatures varied as described in Table S2 in the supplemental material. Reaction conditions when Taq polymerase was used consisted of 1× ThermoPol buffer (NEB), 200 μM deoxynucleoside triphosphates (dNTPs) (Invitrogen), 400 nM each primer (Operon), 50 ng of DNA, and H₂O added to provide the correct buffer dilution. PCR cycling consisted of 95°C for 5 min, 35 cycles of 95°C for 30 s, the specific annealing temperature for 30 to 60 s, and 72°C for 1 to 2 min, and final extension at 72°C for 5 min. Reaction conditions when Phusion enzymes were used consisted of a 1× concentration of the 2× Phusion master mix with HF buffer, 400 nM each primer, 50 to 100 ng of DNA, 3% dimethyl sulfoxide (DMSO), and H₂O added to provide the correct buffer dilution. PCR cycling consisted of 98°C for 30 s, 30 cycles of 98°C for 10 s, the specific annealing temperature for 30 s, and 72°C for 4 to 6 min, and final extension at 72°C for 10 min. O55 and O157 serogroups and H7 flagellar antigens were confirmed by PCR for all strains. General stx₁ and stx₂ primers, detecting all stx₁ and stx₂ variants, were used to test all strains, and stx₁-positive strains were additionally tested with primers for subtyping stx₁ variants stx_1a, stx_1c, and stx_1d (see Table S2 in the supplemental material).

Phage insertion site analysis at previously characterized loci consisted of PCR amplification across an intact gene, disruption of which suggested the presence of a prophage or other mobile element. The presence of a lambda-like phage at either the wrbA or yehV locus was confirmed by amplification from the bacterial genome into the site-specific integrase of the prophage (see Table S2 in the supplemental material). Confirmation of the right side, or tail protein-containing end, of phages at yehV was performed using primers specific for tail fiber genes but under PCR conditions that allowed for up to 12 kb of product amplification in order to accommodate for the presence of T3SS effector genes that are sometimes located at the right side of the phage (see Table S2 in the supplemental material).

PFGE analysis.

E. coli O55:H7 strains were analyzed by pulsed-field gel electrophoresis (PFGE) according to standardized methods (61). Plugs were prepared, digested with XbaI and BlnI restriction enzymes, and analyzed on a contour-clamped homogeneous electric field (CHEF) Mapper II PFGE system (Bio-Rad), as described previously (61). The initial and final switch times were 6.76 and 35.38 s, respectively. The electrophoresis equipment was set at 6 V with an included angle of 120° and a run time of 20.5 h. The CDC universal standard, Salmonella enterica subsp. enterica serovar Braenderup strain H9812, was run in duplicate on each gel, allowing images from multiple gels to be included in the analysis. Gel images were analyzed in Bionumerics software version 6.0 (Applied Maths), and dendrograms were produced using UPGMA (unweighted-pair group method using average linkages) pairwise similarity cluster analysis. Groupings for O55:H7 strains with greater than 75% similarity were given an arbitrary profile designation; O157:H7 strains Sakai and 2886-75 were included for comparison.

MLVA analysis.

Multilocus variable-number tandem repeat analysis (MLVA) was conducted as described previously (7). Seven of the nine loci characterized for this study (CVN1, CVN2, CVN3, CVN4, CVN7, CVN14, and CVN15) were evaluated with the same primer sets originally described by Lindstedt et al. (45); two additional loci were tested using the EHC2 and EHC6 primers described by Izumiya et al. (31). Briefly, multiplex PCR was performed using fluorescently labeled primers, and the resulting products were pooled and analyzed using a 3130xl Genetic Analyzer (Applied Biosystems). The number of tandem repeats at each locus was determined from the fragment size, and a dendrogram was generated from these data in Bionumerics software using the UPGMA method. By this analysis, the O55:H7 isolates fell into two distinct groupings.

Phylogenetic analyses based on SNPs of core genes.

Single nucleotide polymorphisms (SNPs) in 31 complete and draft E. coli genomes were identified using the MUMmer software package (40). A total of 62,838 core SNPs (those in chromosomal regions that were present in all examined genomes) were compiled using a custom Perl script. Maximum-likelihood SNP trees were inferred with RaxML 7.0.4 using the rapid bootstrap procedure to infer nonparametric bootstrap values (67).

Software used for genome analysis.

Alignments of full genomes and full-length bacteriophages were conducted using Geneious software version 5.4.6 (Biomatters) with the Mauve plug-in function. Assembly of the RM12579 genome, and of the DEC5E bacteriophage at the yehV locus using contigs and Sanger sequencing, was also performed using Geneious software. Genome annotation was completed using the NCBI Prokaryotic Genomes Automatic Annotation Pipeline (PGAAP), available at http://www.ncbi.nlm.nih.gov/genomes/static/Pipeline.html. For the prophage at yehV in E. coli O55:H7 strain DEC5E, annotation was based on output from PHAST (Phage Search Tool) (available at http://phast.wishartlab.com/), GeneMarkS (available at http://exon.gatech.edu/genemarks.cgi), and tRNAscan-SE (available at http://lowelab.ucsc.edu/tRNAscan-SE/).

Nucleotide sequence accession numbers.

The complete sequence of the E. coli O55:H7 strain RM12579 genome has been deposited in GenBank under accession numbers CP003109 (chromosome) and CP003110, CP003111, CP003112, CP003113, and CP003114 (plasmids). The complete sequence of the prophage at yehV in E. coli O55:H7 strain DEC5E has been deposited in GenBank under accession number JQ347801. The E. coli DEC5A (SRX038474), -5B (SRX038475), -5C (SRX038476), -5D (SRX038473), and -5E (SRX038493) genome sequences were generated by the University of Maryland School of Medicine, Institute for Genome Sciences Genome Sequencing Center for Infectious Diseases (GSCID).

RESULTS AND DISCUSSION

General genomic features.

Escherichia coli O55:H7 strain RM12579 had a chromosome of 5,263,980 bp, plus five plasmids (Table 2). The RM12579 chromosome contained seven rRNA loci, as is typical for E. coli, and 102 tRNA genes. Three plasmids were similar to ones previously identified in E. coli (see Fig. S1 in the supplemental material). The other two plasmids, p12579_3 and p12579_5, shared a similar backbone of plasmid mobilization (mob) genes that were 88% and 92% identical to the 3.5-kb mob loci of ColE1 (NC_001371) and of p5217 (NC_011799) from strain E2348/69 (25), respectively (see Fig. S1 in the supplemental material). Antibiotic resistance loci were located on two of the plasmids: a Tn3 insertion element containing the TEM-1 beta-lactamase gene in p12579_3, that encodes ampicillin resistance (locus tag ECO55CA74_26309), and two genes encoding streptomycin-inactivating enzymes in p12579_4 (locus tags ECO55CA74_26389 and ECO55CA74_26394). Resistance of RM12579 to both ampicillin and streptomycin was confirmed experimentally.

Table 2.

General features E. coli O55:H7 strain RM12579 chromosome and plasmids

Region	Size (bp)	G+C ratio (%)	No. of:
			ORFsa	rRNA	tRNA	Prophages
Chromosome	5,263,980	50.5	4,913	7	102	12
p12579_1	94,015	48.1	106	0	2	1b
p12579_2	66,078	48.9	79	0	0	0
p12579_3	12,068	48.9	16	0	0	0
p12579_4	6,211	52.5	7	0	0	0
p12579_5	5,954	45.8	9	0	0	0
Genome	5,448,306		5,130	7	104	13

^aORFs, open reading frames.

^bp12579_1 is a lysogenic bacteriophage that remains circularized and extrachromosomal.

O55:H7 strains RM12579 and CB9615 differ primarily in prophage content.

The overall, chromosome-length, pairwise nucleotide identity of RM12579 to CB9615 was 99.9%, excluding prophages, prophage-like regions, and one 10.9-kb inversion compared to CB9615 within the “block 172” region (see below). Boundaries established for prophages (p) and prophage-like elements (pLE) in the Sakai strain of O157:H7 (Sp1-Sp18; SpLE1-SpLE6) (26) were used to define these regions in O55:H7 strain RM12579 (labeled with an initial “C” for California), with some exceptions. A few phage boundaries for strain RM12579 were adjusted as described in Materials and Methods. Accordingly, RM12579 was found to contain 12 prophages (Cp1 to Cp12), compared to 15 in CB9615 (Fig. 1; see Table S1 in the supplemental material). The prophage at yehV in strain RM12579 is novel and does not harbor stx, as does the prophage at this location in Sakai. Also, RM12579 contained one prophage-like element (CpLE4) that was not present in CB9615 and corresponded to SpLE5 in O157:H7 Sakai. A chromosome-wide comparison between these three strains (O157:H7 Sakai, O55:H7 RM12579, and O55:H7 CB9615) showed that the majority of phages were similar in placement across all three genomes and that the overall phage composition of RM12579 was intermediate between those of Sakai and CB9615 (Fig. 1; see Fig. S2 and Table S1 in the supplemental material).

Fig 1.

Comparative analysis of prophage integration sites in E. coli O157:H7 Sakai and O55:H7 strains RM12579 and CB9615. Linear representations of all three chromosomes were annotated with boundaries for prophages (black boxes) and phage-like elements (gray boxes). Phages are labeled with the nomenclature specific to each strain: Sakai, Sp1 to Sp18; California isolate RM12579, Cp1 to Cp12; E. coli strain CB9615, Ep1 to Ep15. tRNA loci are indicated by pink triangles, and rRNA loci are marked with blue triangles. Phages found only in the two O55:H7 strains are indicated with a blue rectangle (Cp3 and Ep3), and phages unique to only one of the strains are indicated with a red rectangle. Two gene loci are marked with green triangles: wrbA, the insertion site for the stx₂-containing phage Sp5, and yehV (mlrA), the insertion site for the stx₁-containing phage Sp15.

Comparative analysis of E. coli O55:H7 strains reveals multiple, genetically distinct clades.

To further explore genetic diversity within serotype O55:H7, we performed pulsed-field gel electrophoresis (PFGE) and multilocus variable-number tandem repeat analysis (MLVA) on strain RM12579 and 14 other E. coli O55:H7 isolates from around the world and across 51 years of time (Table 1; see Fig. S3 in the supplemental material). One subgroup of five isolates (not including RM12579) shared a similar PFGE pattern using both XbaI and BlnI enzymes, and because this subgroup of five isolates included the reference strain DEC5B (59, 71), it is referred to in this paper as the DEC5B subgroup.

In order to determine the prevalence of O157:H7 virulence factors within strains of the O55:H7 serotype, we tested a panel of genes, including T3SS effectors and stx, from among those currently under evaluation as markers for virulence (3, 74). For comparison, the earliest documented O157:H7 isolate, E. coli O157:H7 strain 2886-75, was tested for the presence of these virulence genes. Most genes were present either in the O157:H7 strain alone or in all strains (Table 3). All O55:H7 strains tested were found to have the tir_Sakai allele and the tccP gene, which are involved in the formation of attaching and effacing lesions by O157:H7. However, the length of the tccP gene varied among isolates, with the DEC5B subgroup having the shortest version at 732 bp and DEC5E possessing the longest tccP gene at 1,296 bp.

Table 3.

Virulence genes in O55:H7 strains and original O157:H7 isolate^e

^aDECB subgroup (as described in the text).

^bPFGE subgroups are based on strains showing >75% similarity, as shown in Fig. S3 in the supplemental material.

^cMLVA subgroups are as shown in Fig. S3 in the supplemental material.

^dnleA is located on the chromosome in O157:H7 strains and on a plasmid in O55:H7 strains.

^eAbbreviations: ND, not done; IS, insertion element IS2.

Of note, the DEC5E strain was found to have several virulence factors that are typically present in O157:H7 strains but absent in most O55:H7 strains, including the espV, iha, espX7, and espN genes (Table 3). Consistent with the positive iha result, DEC5E was shown to harbor part of the SpLE1 prophage-like element that is present in O157:H7 strains but not typically in O55:H7 strains (65, 72). Also, the espV gene located on the DEC5E prophage at yehV (see below) showed 87% pairwise nucleotide identity to a functional T3SS effector in Citrobacter rodentium (locus tag ROD_19111) (1), whereas the espV genes in the O157:H7 Sakai strain (locus tag ECs1127) and on p12579_2 (locus tag ECO55CA74_26199) are truncated.

All genetic evidence except stx prophage content supports the existing models of stepwise O157:H7 evolution.

Two currently accepted, overlapping, stepwise models of E. coli O157:H7 evolution have been proposed. Key elements of these two previous models are depicted in Fig. 2. In the first model, a series of steps from A1 to A6 was proposed (18, 20, 72). Groups A1 and A2 contained O55:H7 strains, separated by the acquisition of stx₂ by the A2 strains 3256-97 (GenBank accession number AEUA01000000) and USDA 5905 (GenBank accession number AEUB01000000). In contrast, our updated model grouped O55:H7 strains based on the genetic analyses described above, irrespective of stx status. The A5 and A6 lineages together included most SOR⁻ GUD⁻ O157:H7 pathogenic human isolates known at that time, generally possessing both stx₁ and stx₂ (Fig. 2; Table 4) (17, 18, 50). It was acknowledged that stx-containing bacteriophages can mobilize between genomes, and readers were admonished to interpret association of stx prophages within groups A1 to A6 with caution (20, 72). Nonetheless, all diagrams of the model depicted the acquisition of the stx prophages as key, stable, stepwise milestones in the evolutionary trajectory of the O157:H7 lineage (18, 20, 50, 71).

Fig 2.

Revised evolutionary model describing evolution of O55:H7 to O157:H7 with variable stx presence. A comprehensive evolutionary model is presented, encompassing elements of previous models as well as two new O55:H7 clades based on genetic differences identified in this study. The stepwise acquisition of stx-containing phages is not present in this model, based on the observation that stx₁- and stx₂-containing phages have been acquired (and possibly lost) multiple times in both O55:H7 and O157:H7 strains. Some group names from previous models have been conserved in order to preserve well-established and widely used terminology. These include the A3, A4, A5, and A6 groups originally described by Feng et al. (18) and the cluster 1, 2, and 3 designations within subgroup C (or A6) as described by Leopold et al. (42). Absent the requirement for stepwise acquisition of stx-containing bacteriophages, clusters 2 and 3 can be collapsed into one group, as indicated by the square surrounding both circular clusters. In this way, strains that otherwise would not have fit into previous models are included within this final group. Additionally, former A1 and A2 strains that preceded A3 (18) are included in the O55 group but reassorted to distinguish the DEC5B subgroup. Locations of key genetic changes in the uidA, fimA, and fimH genes as discussed in the text, the ability to utilize sorbitol (SOR), β-glucuronidase (GUD) activity, and the locations of the 11 SNPs (single nucleotide polymorphisms) discussed in Table S3 in the supplemental material are highlighted. Representative strains for each group have been placed within each square or circle, and the presence of stx phages is noted for each representative strain in red circles. Nonhuman isolates have been underlined. 1, stx₁ positive; 2, stx₂ positive; *, single known sorbitol-negative O55:H7 strain.

Table 4.

Phenotypic and genotypic characteristics of selected strains

Strain	E. coli serotype	Shiga toxin (Stx) status	Sorbitol utilization	uidA +93a	uidA insertion at +686b	fimA 16-bp deletionc	fimH binding pocket mutationd
DEC5E	O55:H7	Negative	Positive	T	No	No	N
RM12579	O55:H7	Negative	Positive	T	No	No	N
BB2	O55:H7	Stx1+	Negative	T	No	No	N
LSU-61	O157:H7	Negative	Positive	G	No	No	N
CDC 2886-75	O157:H7	Stx1+	Negative	G	GG	Present	K
Sakai	O157:H7	Stx1+ Stx2+	Negative	G	GG	Present	K

^aThe T-to-G transversion at +93 in the coding region of the uidA gene is characteristic of O157:H7 strains (19).

^bThe dinucleotide insertion (GG) at +686 in the coding region of the uidA gene disrupts the function of the β-glucuronidase protein that it encodes (52).

^cThe 16-bp deletion located 5′ of the fimA gene is present in a subset of O157:H7 strains (66).

^dThe C-to-A transversion at +467 in the coding region of the fimH gene results in an amino acid change from asparagine (N) to lysine (K) (66).

The second model maintained the overall stepwise progression of the first model, but some groups were condensed and others expanded (42, 64, 65). Specifically, all O55:H7 strains were placed into subgroup A, all O157:H⁻ strains into subgroup B, and most O157:H7 isolates into “human” subgroup C, which was further subdivided into clusters 1, 2, and 3 (Fig. 2). Transitions for all three O157:H7 clusters in this model were defined in terms of prophage acquisition, specifically requiring a gain of a truncated prophage at the yehV locus for inclusion in cluster 1 and additionally a stx₂ prophage at the wrbA locus for the transition from cluster 1 to 2. Finally, the acquisition of an stx₁-containing phage at yehV distinguished cluster 2 from cluster 3 (Fig. 2) (42, 64, 65).

Our current analysis supported all of the nonprophage background genotypic changes described in these previous models. The O55:H7 strains examined in this study were typical. All strains were GUD⁺, all but one (BB2) were SOR⁺, and the three strains that were genotyped (BB2, DEC5E, and RM12579) lacked O157:H7-specific alleles in the uidA, fimA, and fimH genes (Table 4). Additionally, we confirmed that the fimA deletion leading to subgroup C corresponded to the transition from A5 to A6 (Fig. 2) (64), as it was not found in O157:H7 strain G5101 (GenBank accession number AETX01000114) (27). Finally, we retained the fimH mutation at the transition from cluster 1 to 2, as posited in an earlier version of the model (Fig. 2) (64), although the most recent model placed the fimH division within cluster 1 in order to prioritize the stx phage requirements at cluster transitions (42).

stx-containing prophages are not stable markers within all radiations of the O157:H7 lineage.

In contrast to the analysis of nonmobile chromosomal loci, our data showed anomalies in the presence and locations of stx-containing bacteriophages compared to those in established O157:H7 evolutionary models (Table 4; see Table S3 in the supplemental material). First, O55:H7 strain BB2 was found to harbor stx₁. Second, O157:H7 strain 2886-75 harbored stx₁ but not stx₂ (Table 4). Thus, neither strain fit within the current models for O157:H7 evolution, which argued that stable acquisition of stx₂ occurred before that of stx₁ (20, 42, 43) and drew conclusions about the constrained radiation of the O157:H7 lineage based on strains categorized by phage acquisition (42, 43).

In order to identify additional O157:H7 strains that had inconsistent patterns of stx₁ and stx₂ presence, we tested 22 O157:H7 strains, half of which were environmental isolates and half of which were clinical isolates (4 stx negative and 18 carrying stx₁ only) for sorbitol utilization, GUD function, fimA deletion, and fimH mutation, as well as occupancy of stx phage insertion sites (see below). All 22 strains showed the typical SOR⁻ GUD⁻ phenotype and presence of the fimA deletion, which would have placed them within the A6 group except for their lack of the stx₂ toxin gene (Fig. 2; see Table S3 in the supplemental material). Most of the 22 strains also contained the fimH mutation as well as the 11 SNPs present in cluster 3 strains (Fig. 2; see Table S3 in the supplemental material). Of interest, strains 2886-75 and DEC3A (GenBank accession number SRX038471) were shown by PCR or sequence data, respectively, to possess the same subset of 5 of the 11 cluster 3-specific SNPs (43).

These results indicated that the presence of stx-containing bacteriophages was not a fully reliable indicator of evolutionary history in the O55:H7 and O157:H7 lineages. Further support of the unstable nature of these mobile genetic elements includes their demonstrated ability to excise in the laboratory (2, 28), spontaneous loss of both stx₁ and stx₂ documented in EHEC patients (16, 47, 49), and loss of stx₂ upon subculture of a stx₂⁺ SOR⁺ GUD⁺ O157:H7 strain isolated from a red deer in Spain (strain 290 in Fig. 2) (22). Although prophage instability had been acknowledged previously by the authors of some of the evolutionary models (20, 72), this subtlety within what is otherwise a seminal and well-accepted model has been lost. The model diagrams that have been cited and adapted subsequently by other investigators (3a, 35) accordingly included the stepwise, stable acquisition of stx₂ prior to stx₁. Based on the evidence presented in this work, we suggest a revision of the model to explicitly exclude the stepwise acquisition of first stx₂ and later stx₁ toxin genes, while retaining all other elements of the original models that support the well-established clonal nature of the EHEC 1 clade, containing both O55:H7 and O157:H7 strains (20, 60, 64). This revision may also necessitate a reanalysis of the “constrained radiation” of E. coli O157:H7 lineages (42, 43), as the clusters analyzed to support this argument may have been artificially minimized by the required presence of specific prophages instead of being based on the nonphage genetic analysis of the clonal evolution of O157:H7.

Analysis of stx-containing lambdoid phage insertion sites in O55:H7 and O157:H7 strains supports multiple insertion sites and events.

The stx-containing lambdoid bacteriophages are known to insert at multiple chromosomal locations. Variants of stx₁-containing bacteriophages have been found inserted at the yehV gene in O157:H7 and O84:H4 strains (9, 26, 57) and at the Z2577 gene in a O146:H21 strain (36). Variants of stx₂-containing phages have been found in O157:H7 strains at wrbA (stx₂) (26, 57), at argW (stx₂) and sbcB (stx_2c) in the 2006 spinach-outbreak related strains (38, 39), and at yecE (stx₂) (10). We screened our panel of O55:H7 strains, as well as O157:H7 strains 2886-75 and Sakai, for prophage insertions at these six previously characterized insertion sites (Table 5). None of these 17 strains had disrupted loci at the Z2577, argW, or sbcB gene (data not shown).

Table 5.

Shiga toxin-containing phage insertion site analysis by PCR^f

Strain	E. coli serotype	Shiga toxin (Stx) status	yehVa	wrbAb	yecEc
Sakai	O157:H7	Stx1+ Stx2+	λ-like phage (Sp15)d	λ-like phage (Sp5)e	I
CDC 2886-75	O157:H7	Stx1+	λ-like phage	λ-like phage	I
DEC5A	O55:H7	Negative	I	I	I
DEC5B	O55:H7	Negative	I	IS	D
DEC5C	O55:H7	Negative	I	I	I
DEC5D	O55:H7	Negative	I	I	I
DEC5E	O55:H7	Negative	λ-like phage	I	I
CDC 5644-62	O55:H7	Negative	I	IS	D
RM3654	O55:H7	Negative	λ-like phage	I	I
BB2	O55:H7	Stx1+	λ-like phage	I	I
RM12579	O55:H7	Negative	λ-like phage (Cp10)	I	I
RM13616	O55:H7	Negative	I	IS	D
RM13617	O55:H7	Negative	I	IS	I
RM14504	O55:H7	Negative	I	I	I
RM14505	O55:H7	Negative	I	I	I
RM14506	O55:H7	Negative	I	I	I
RM14507	O55:H7	Negative	I	IS	I

^aInsertion site of stx₁-containing phages in O157:H7 Sakai (26) and EDL933 (57). The presence of λ-like phage was verified by sequencing from the yehV gene locus (bacterial) into the yehV-specific integrase (bacteriophag), as described in the text.

^bInsertion site of stx₂-containing phages in O157:H7 Sakai (26) and EDL933 (57) and O104:H4 (48). The presence of a λ-like phage in strain RM8530 was verified by sequencing from the wrbA gene (bacterial) into the wrbA-specific integrase (bacteriophage), as described in the text.

^cInsertion site of Stx2e-encoding phage P27 from an ONT:H⁻ strain (51, 58) and Stx2-encoding phage P297 from an O157:H7 strain (10).

^dStx1-containing phage Sp15; data from published genome (57).

^eStx2-containing phage Sp5; data from published genome (57).

^fI, intact gene at the locus; D, interrupted gene at the locus; IS insertion element.

Several O55:H7 strains (DEC5E, RM3654, BB2, and RM12579) had a lambda-like phage inserted at yehV, the presence of which was confirmed by amplification into the yehV-specific integrase at the left side of the phage (Table 5). The five DEC5B subgroup strains were observed to have a disrupted wrbA gene with the simple insertion element ISEhe3 present at this site. CDC strain 2886-75 was found to have a full or partial lambda-like phage located at both the yehV and wrbA loci, which was verified in both cases by sequencing into the site-specific integrases (Table 5). Finally, the 22 stx-negative or stx₁-only O157:H7 strains were also analyzed at the same six insertion sites (see Table S3 in the supplemental material). A lambda-like phage occupied the yehV locus in 20 of 22 isolates, but the locus was intact in two clinical stx₁-positive strains (RM6540 and RM6693), suggesting that the stx₁-containing bacteriophage was not located at yehV in these two strains.

The presence of full-length prophages at yehV in two O55:H7 strains suggests multiple, parallel acquisitions of distinct phages at this locus.

The second stepwise evolutionary model discussed above (42, 43, 64, 65) had postulated that the yehV locus was intact in ancestral O55:H7 strains and that a truncated phage of approximately 30 kb, missing the centrally located stx genes, was later acquired by the O157:H7 lineage, facilitating subsequent acquisition of the full-length stx₁-containing phage in cluster 3 isolates. Indeed, the yehV locus was found to be intact in the O55:H7 CB9615 strain (77) (Fig. 1). In contrast, by completely closing the RM12579 genome, we have provided the first report and characterization of a full-length lambda-like prophage integrated at yehV in an O55:H7 strain (Cp10) (Fig. 3A). Furthermore, we were able to produce a closed assembly of the full-length prophage located at yehV in strain DEC5E by using publicly available sequence reads (Fig. 3B). Neither of these full-length prophages at yehV in O55:H7 strains contained stx genes.

Fig 3.

Analysis of modular phages at the yehV locus in two O55:H7 strains. Phages are shown as an alignment using Geneious software, in which gray segments represent exact nucleotide identity, black segments or boxes reflect nucleotide differences, and a thin line indicates a gap in the sequence alignment (although not a gap in the sequence itself, as all sequences represented are continuous). Color coding for genes was as follows: brown, integrase or excisionase; red, recombination; cyan, transcriptional regulation; yellow, replication; pink triangles, tRNA; green, lysis; blue, structural/morphogenesis; white with red outline, T3SS effector or other virulence-related protein; white with black outline, other gene with known or predicted function; gray, hypothetical gene; and white with green outline and hatch marks, transposase or transposon-related derivative. (A) The complete lambda-like phage Cp10 from RM12579 is shown in the center aligned to a similar phage located at yehV in S. boydii strain CDC 3083-94 (above) and to genomic island GE12.21, a partial sequence from an E. coli O111:H⁻ strain (below). (B) The complete lambda-like phage found in strain DEC5E at yehV is shown in the center in comparison to the stx₂⁺ phage BP-933W inserted into wrbA in E. coli O157:H7 strain EDL933 (above) and to the stx₁⁺ phage BP-4795 inserted at yehV in E. coli O84:H4 strain 4795/97 (below). The center of each phage contains distinctly different genes; BP-933W contains both tRNA and stx₂ genes, the DEC5E phage contains tRNA genes as well as tellurite resistance-related genes, and BP-4795 contains only stx₁ genes.

Previously, it was suggested that the prophage located at yehV in DEC5E was truncated (65), due in part to a failure to amplify the right side of the phage with O157:H7-specific primers. We designed new primers targeted to the right side, which were able to recognize both O157:H7 and O55:H7 sequences, and used long PCR conditions in order to accommodate the potential presence of T3SS effectors variably present at the right sides of lambda-like bacteriophages (2) (see Table S2 in the supplemental material). In this way, we were able to amplify a product from the right sides of all O55:H7 strains with a phage insertion at yehV.

Variability within the candidate O55:H7-specific region, designated block 172, further refines the evolutionary model.

In comparing E. coli O55:H7 strain CB9615 to available O157:H7 genomes, Zhou et al. (77) identified a 72-kb block of the genome, designated “block 172,” that was specific to CB9615 and not found in the sequences of available O157:H7 strains. Most of block 172 was present in O55:H7 RM12579, with one deletion and one 10.9-kb inversion (see Fig. S4 in the supplemental material). Unlike most O157:H7 strains, E. coli O157:H7 strain LSU-61 (GenBank accession number AEUC01000000) possessed block 172 in its entirety, indicating that block 172 was most likely present in the last common ancestor of both the O55:H7 and O157:H7 lineages. This also suggested that the presence of block 172 could be used to identify candidates for the hypothetical A3 intermediate, regardless of their stx status (Fig. 2).

Strain LSU-61 is a SOR⁺ GUD⁺ stx-negative O157:H7 strain isolated from a white-tailed deer in Louisiana (13). The authors of the A1 to A6 evolutionary model (20) did note that the LSU-61 strain would qualify as a member of the hypothetical A3 intermediate in their stepwise model, except for the absence of the stx₂-containing prophage (Fig. 2). Examination of available contigs confirmed that LSU-61 does possess the +93 T → G nucleotide change in uidA that is characteristic of other O157:H7 and O157:H⁻ strains but does not possess the other genotypic changes in the uidA, fimA, and fimH genes found in subsequent O157:H7 strains as described in the stepwise models (Table 4; Fig. 2). In addition, all known stx phage integration sites were found to be intact in LSU-61. It cannot be ruled out that a stx phage may have been spontaneously lost from this strain; nonetheless, LSU-61 qualified as a potential candidate for the hypothetical A3 intermediate.

Core genome SNP-based phylogenetic analysis of multiple O55:H7 and O157 genomes supported the hypothesis that LSU-61 represents a deep branch lineage that shares a precursor with the typical O157:H⁻ (A4) and O157:H7 (A5) clades (Fig. 4). The recent discovery of additional SOR⁺ GUD⁺ stx-negative O157:H7 strains in the feces of wild red deer suggested that this lineage may be specifically adapted to the cervine gut (12). Alternately, conditions within cattle (21), but not wild animals, may have supported the stable fixation of stx-containing bacteriophages in the chromosome and thereby increased the chance of humans encountering stx⁺ O157:H7 strains via this domesticated animal.

Fig 4.

Whole-genome maximum-likelihood phylogeny of O157:H7, O157:H⁻, and O55:H7 strains based on core SNPs. The tree is rooted based on the position of the outgroup, E. coli O26:H11 (not shown). Internal nodes with bootstrap values of 100 are indicated. All other nodes had bootstrap values less than or equal to 80. Relative positions of the O157:H7, O157:H⁻, and LSU-61 clades cannot be definitively determined by this analysis.

Conclusions.

Comparison of the genome sequence of the fully sequenced EPEC O55:H7 strain RM12579 to the previously completed sequence of the EPEC O55:H7 strain CB9615 illustrated the stability of the nonphage genome, given that the two strains were isolated on different continents and separated by a span of 29 years. In contrast, the differences within mobile genetic islands between the two fully sequenced strains, in combination with a targeted, complementary analysis of 14 other O55:H7 strains and 22 clinical and environmental O157:H7 strains, improved our understanding of the events leading to serotype O157:H7 emergence from the O55:H7 precursor. These analyses suggested that the DEC5B strain, in combination with five other strains, formed a distinct subgroup of O55:H7 with a divergent evolutionary path and that that both extant lineages (O55:H7 and O157:H7) share an extinct or as-yet-unobserved last common ancestor (Fig. 2 and 4).

The presence, placement, and composition of bacteriophages at previously identified insertion sites in O55:H7 and O157:H7 strains supported the assertion that stx-containing bacteriophages are not stable evolutionary markers and that their presence should not be required in defining a stepwise evolutionary model of the evolution of O157:H7 from O55:H7. These data further suggested that stx-containing bacteriophages have been integrated multiple times, in parallel, at multiple sites. Thus, we suggest that the existing models are robust in their characterization of non-prophage-associated diversity but that the requirement for stepwise acquisition of stx-containing phages should be removed in order to accommodate and eventually delineate the broader diversity of O157:H7 strains, including stx₁-only strains. Finally, analysis of the diversity present within the candidate O55:H7-specific region “block 172,” which was nonetheless found to be present in O157:H7 strain LSU-61, may permit further insights into the early emergence of O157:H7. In sum, exploration of O55:H7 diversity has improved our understanding of and suggested refinements to current models of the evolution of the deadly EHEC O157:H7 serotype.