Heteropathogenic virulence and phylogeny reveal phased pathogenic metamorphosis in Escherichia coli O2:H6

Extraintestinal pathogenic and intestinal pathogenic (diarrheagenic) Escherichia coli differ phylogenetically and by virulence profiles. Classic theory teaches simple linear descent in this species, where non-pathogens acquire virulence traits and emerge as pathogens. However, diarrheagenic Shiga toxin-producing E. coli (STEC) O2:H6 not only possess and express virulence factors associated with diarrheagenic and uropathogenic E. coli but also cause diarrhea and urinary tract infections. These organisms are phylogenetically positioned between members of an intestinal pathogenic group (STEC) and extraintestinal pathogenic E. coli. STEC O2:H6 is, therefore, a ‘heteropathogen,’ and the first such hybrid virulent E. coli identified. The phylogeny of these E. coli and the repertoire of virulence traits they possess compel consideration of an alternate view of pathogen emergence, whereby one pathogroup of E. coli undergoes phased metamorphosis into another. By understanding the evolutionary mechanisms of bacterial pathogens, rational strategies for counteracting their detrimental effects on humans can be developed. Subject Categories Microbiology, Virology & Host Pathogen Interaction

Escherichia coli virulence and phylogeny are intertwined. Each pathogenic E. coli group possesses 'signature' repertoires of virulence genes, which enable them to colonize and injure their host (Kaper et al, 2004). While exceptions exist, particularly among ETEC (Turner et al, 2006), strains within pathogroups are often phylogenetically closely related (Achtman et al, 1983;Wirth et al, 2006). Current concepts of pathogen emergence employ linear descent scenarios, where horizontal acquisition of pathogenicity islands, bacteriophages, and plasmids by non-pathogens results in phylogenetically fixed pathogroups (Dozois & Curtiss, 1999). Here, we determined the phylogenetic relationship of STEC O2:H6 isolated from patients with diarrhea to other intestinal and extraintestinal pathogenic E. coli and then characterized the O2:H6 virulence genes as well as their UPEC virulence potential. To our surprise, these analyses did not portray the linear emergence of pathogenicity by step-wise recombination events, but, instead, identified STEC O2:H6 as a 'transitional' pathogen in the process of morphing between pathogroups.
To more thoroughly analyze the phylogenetic relationships between STEC O2:H6, most closely related STEC (serotype O91:H21; ST442; Fig 1) and prototypic HUS-associated STEC, UPEC and AIEC, we used whole genome sequencing and a gene-by-gene analysis of in total 2827 open reading frames that were present in all 14 strains investigated (see supplementary Table S1). This analysis confirmed the intermediate positioning of STEC O2:H6 between the major HUS-associated STEC serotypes and UPEC. Interestingly, AIEC were grouped closely to UPEC (Fig 2).
As AIEC were also phylogenetically closely related to UPEC and to STEC O2:H6 (Figs 1 and 2), we further investigated whether STEC O2:H6 strains contain FimH belonging to the same clade as that possessed by UPEC and AIEC strains (Sepehri et al, 2009). FimH is the adhesin subunit of the type 1 pili that mediate adherence and play an essential role in the invasive ability of AIEC (Boudeau et al, 2001). Comparative sequence analysis of FimH from the sequenced STEC O2:H6 strains (supplementary Table S1) and from the UPEC (strains 536, UTI89, CFT073) and AIEC (strains LF82, LF73) reference strains showed the phylogenetic positioning of all FimH proteins of STEC O2:H6 in the S70/N78 FimH clade (supplementary Fig S1), which is typical for UPEC and AIEC (Dreux et al, 2013).
Nearly all of the identified UPEC virulence loci were also expressed by the cognate STEC O2:H6 isolates. Each of the ten STEC O2:H6 that harbored a-hlyA and cnf1 (Table 1) produced a-hemolysin on blood agar and cytotoxic necrotizing factor (CNF) 1, detectable as an~115-kDa band in an immunoblot (supplementary Fig S3). Each of the three a-hlyA-negative, cnf1-negative and clb-positive STEC O2:H6 (Table 1) produced colibactin, a hybrid polyketide-peptide cyclomodulin encoded by the clb island (Nougayr ede et al, 2006). Similar to prototypic colibactin-producing MNEC strain IHE3034 (Nougayr ede et al, 2006), each STEC O2:H6 strain arrested HeLa cells in the G2 phase of the cell cycle after 48 h and distended these epithelial cells and their nuclei, converting them into megalocytes (Fig 3). In the ten remaining clb-positive isolates, HeLa cell lysis caused by a-hemolysin confounded our attempts to study colibactin expression. Neither G2 arrest nor cell distension were elicited by a clb-negative, Stx2producing STEC O2:H27 (Fig 3), excluding a contribution of Stx to these effects.
P and S fimbriae were expressed in eight of ten and six of nine strains, harboring these respective loci (Table 1), as demonstrated by the ability of the bacteria to agglutinate human and bovine erythrocytes, respectively, in the presence of mannose (Blumer et al, 2005). Thus, the UPEC virulence loci found in STEC O2:H6 are largely functional across the collection of strains that we studied.

Analysis of urovirulence of STEC O2:H6
The ability of randomly selected STEC O2:H6 isolates to cause urinary tract infection (UTI) was tested in an experimental murine model of ascending UTI. The bacterial numbers in the bladder and the kidneys were determined 72 h after infection (Fig 6). The prototypic UPEC strain 536 and STEC O2:H6 strain 05-00787 were recovered in nearly equal numbers from the bladder tissue (1 × 10 5 -1 × 10 6 colony-forming units (CFU)/g bladder) ( Fig 6A). Bladder colonization by STEC O2:H6 strains 04-00955 and 03-08304 resulted in 10-fold higher organ loads relative to UPEC 536. In all four cases, bacterial concentrations in the bladder tissue were significantly higher than after infection with non-pathogenic E. coli K-12 strain MG1655 ( Fig 6A). All three STEC O2:H6 strains also colonized the kidneys as efficiently as UPEC strain 536 ( Fig 6B). In contrast, strain MG1655 was unable to ascend to the kidneys (Fig 6B). Consequently, the potential of STEC O2:H6 strains to cause UTI in this model is comparable to that of classic UPEC strain 536.  Table  S1). The different pathotypes are distinguished by colors of the circles and the serotypes and strain numbers (in parentheses) are given.

Discussion
The clinical significance of hybrid pathogens was clearly demonstrated by the deadly 2011 outbreak caused by E. coli O104:H4 (Bielaszewska et al, 2011;Frank et al, 2011;Karch et al, 2012). The outbreak started in Germany in May 2011 and subsequently spread to other European countries and North America, affecting in total nearly 4000 persons of which more than 900 developed HUS and 54 died . The outbreak strain uniquely combined virulence genes typical for STEC and EAEC and expressed the phenotypes that define these pathogroups including Stx2 production and aggregative adherence to intestinal epithelial cells (Bielaszewska et al, 2011;Brzuszkiewicz et al, 2011;Mellmann et al, 2011;Rasko et al, 2011), the features that likely increased its virulence. This outbreak, therefore, tragically illustrated that blended virulence profiles in enteric pathogens introduced into susceptible populations can have serious public health consequences . Moreover, this outbreak highlighted the lack of our knowledge of the basic principles of evolutionary trends of new pathogens as demonstrated by the fact that the origin of the E. coli O104:H4 hybrid and its evolutionary history remain obscure. In this context, our identification of another hybrid pathogen, STEC O2:H6 in this study, and gaining insight into its evolutionary role is of particular importance, because we can determine the phylogenetic coordinates of its evolution. Also, STEC O2:H6 is the first hybrid pathogen to demonstrate pluripotential pathogenicity in intestinal and extraintestinal milieus, as predicted by its virulence repertoire. Specifically, STEC O2:H6 occupies an evolutionary and pathogenic interface between intestinal and extraintestinal pathogenic E. coli, and several lines of evidence corroborate this assignment: sequence typing, virulence genotyping and phenotyping, and the clinical potential. First, MLST/rMLST, employing sequences of a sample of housekeeping and ribomosal genes scattered around a chromosomal backbone, agnostically placed these isolates between HUSEC and ExPEC (Fig 1). Also whole genome sequencing of STEC O2:H6, closely related HUS-associated STEC, and prototypic HUSEC, AIEC and UPEC strains positioned STEC O2:H6 separately from the other pathotypes in an intermediate position (Fig 2). Next, STEC O2:H6 contain UPEC and HUSEC virulence genotypes and phenotypes, using ascertainments independent of sequence typing, a finding that appears highly non-random. While it is possible that chance acquisition of a gene encoding Stx via phage transduction into either a commensal E. coli or an ExPEC could have resulted in STEC O2: H6, our data argue in favor of true UPEC/HUSEC heteropathogenicity. Specifically, the phylogeny places STEC O2:H6 at quite a distance from commensal E. coli, and at moderate distances from ExPEC (including UPEC) and from HUSEC (Fig 1). Hence, the STEC O2:H6 are incarnated as a phylogenetically recognizable group, and one that differs from UPEC (Figs 1 and 2). Also, an additional virulence gene found in a subset of HUSEC, saa, encoded independently on a plasmid (Paton et al, 2001) (unlike stx which is encoded on a bacteriophage that integrates into the chromosome) was present in all STEC O2:H6, in addition to the many UPEC virulence loci. This simultaneous convergence on UPEC and STEC phylogeny and virulence profiles is most simply explained by a phased transition from one group of pathogens to the other. This is also corroborated by the whole genome sequence data, where STEC O2:H6 strains represent a separate lineage (Fig 2) indicating that the core genome has . AIEC are, similar to STEC O2:H6, phylogenetically located between ExPEC (including UPEC) and intestinal pathogenic E. coli (Fig 1). Our in-depth analysis based on the whole genome sequencing and core genome analysis corroborated this fact and positioned AIEC even closer to UPEC (Fig 2). These data, together with the recently published genome sequence of the prototypic AIEC isolate LF82 (Miquel et al, 2010), demonstrate the genetic hybrid character of AIEC. However, as AIEC strains have only been associated with Crohn's disease (Darfeuille-Michaud et al, 1998;Darfeuille-Michaud, 2002) and not detected in human extraintestinal diseases, they cannot be considered, in contrast to STEC O2:H6, as heteropathogens from the clinical standpoint. The heteropathogenicity of E. coli O2:H6 has multiple implications for our conceptualization of bacterial evolution and pathogen emergence. First, and most significantly, the 'missing link' phylogeny and virulence traits of STEC O2:H6 suggest that pathogenic E. coli emerge not only by the simple linear acquisition of virulence loci by non-pathogenic E. coli strains. We speculate, based on our data, that one pathogroup gradually exchanges one suite of virulence loci for those of another pathogroup, as its core genome transitions simultaneously and in the same direction in a process of 'phased metamorphosis'. This emergence model differs considerably from a linear 'pathogenic stem cell' descent scenario because metamorphosis implies that a pathogenic E. coli in a given venue and phylogeny retains alternative virulence options. Certainly, the stepwise scenario whereby non-virulent (or less virulent) progenitors of pathogens acquire virulence genes resulting in pathogen emergence (Dozois & Curtiss, 1999) is appropriate for tightly circumscribed groups of pathogens, such as the enterohemorrhagic E. coli (EHEC) 1 clade (Leopold et al, 2009) or the recent E. coli O104:H4 outbreak strain, where, most plausibly, an EAEC acquired a Stx2-encoding bacteriophage (Bielaszewska et al, 2011;Brzuszkiewicz et al, 2011;Mellmann et al, 2011;Rasko et al, 2011). We wish to note that the latter pathogen, while possessing blended genotypes and phenotypes of EAEC and STEC, has a largely linear descent method of evolution, because it is phylogenetically closely related to an EAEC prototype strain 55989 (Brzuszkiewicz et al, 2011;Mellmann et al, 2011;Rasko et al, 2011). However, linear models might not apply across the broader E. coli genospecies, where evolution and pathogen emergence might be more trabeculated. Our data are compatible with a recent description of the complex and highly interwoven evolutionary history of E. coli (Touchon et al, 2009), including the flow of specific 'highways' of gene exchange .
Our data actually validate the concept that virulence is ordained by phylogeny as well as by genes encoding specific effector molecules (Whittam et al, 1993). Specifically, the phylogenetic positioning of the STEC O2:H6 isolates between STEC and ExPEC (including UPEC) (Figs 1 and 2) and their possession of virulence loci from both STEC and UPEC (Table 1), reflect the co-evolution of the core chromosome with the accrual of virulence traits of the two closest pathogroups. Notably, loci specific for other pathogroups, such as those common to EPEC, ETEC, EIEC, or EAEC, are not present.
Shiga toxin-producing Escherichia coli are rare among urinary strain set collections. Johnson et al (Johnson et al, 2002) screened 597 UPEC isolates, and found no STEC, even though presumably non-toxigenic E. coli O2:H6 might have been in this collection. Indeed, E. coli O2:H6 accounted for 2.9% of UPEC isolated by these investigators in another study (Johnson et al, 2005). STEC O2:H6 is also rare in fecal samples from patients with diarrhea (Pi erard et al, 1990) and absent from the HUSEC collection (Mellmann et al, 2008). The paucity of these organisms in large strain collections prompts us to speculate that possession of traits of intestinal and urinary pathogroups might reduce the ability of an organism to cause disease in either venue. Nevertheless, the heteropathogenic potential of these strains is substantiated by their isolation as the only pathogens from stools of epidemiologically unrelated patients with diarrhea in this study, and their identification as diarrheagenic in other studies (Pi erard et al, 1990), and their ability to cause UTI in an animal model. The heteropathogenic nature of STEC O2:H6 warrants additional comments. The transitional nature of phased metamorphosis differs from parallel and convergent evolution. Specifically, this model portrays pathogens evolving along a continuum, with the backbone changing in synchrony with virulence loci. Our data clearly recommend combining backbone phylogeny assignment by MLST/rMLST and whole genome sequencing with a broad spectrum virulence profiling to discern metamorphosis between E. coli pathogroups. While backbone analysis (i.e. the combination of MLST, rMLST and whole genome sequencing in this case) accomplishes phylogenetic positioning, its combination with virulence profiles provides a more textured picture of evolution, and enabled us to propose this alternate mechanism of pathogen emergence. Application of such orthogonal assessments to additional isolates in other serotypes should determine the extent to which other heteropathogenic E. coli are found among the E. coli species, an association that might be obscured by microbiologists′ focusing only on diagnostically and pathogenetically relevant virulence factors.
In summary, STEC O2:H6 is an extant 'way station' between major groups of pathogenic E. coli. It affords a unique opportunity to study pathogen emergence via pathogroup conversion, and introduces phased metamorphosis as a new evolutionary concept. Broader surveys combining backbone phylogeny and systematic virulence gene enumeration will be needed to determine if this form of pathogen emergence via transition is a generalized process.

Bacterial strains
The STEC O2, other intestinal pathogenic E. coli, and ExPEC strains used in this study are listed in supplementary Table S1. Except for an STEC O2:H29 (Tasara et al, 2008) (gift of R. Stephan, University of Z€ urich, Z€ urich, Switzerland), all the other STEC O2 strains originated in Germany and represent all STEC of this serogroup recovered in the European authors′ laboratories. HUSEC strains (Mellmann et al, 2008) and their characteristics are available at http://www.ehec.org. The MLST data and the phylogenetic groups of the EPEC, ETEC, EIEC, EAEC, AIEC, UPEC and MNEC strains, and E. coli K-12 strain MG1655 were derived from the published genome sequences at http://www.ncbi.nlm.nih.gov. E. coli strain EC93 used as a control in growth inhibition experiments has been previously described (Aoki et al, 2005) and was a gift of D. A. Low (University of California-Santa Barbara, Santa Barbara, CA, USA).

MLST, rMLST, PCR phylogrouping and whole genome sequencing
Internal fragments of seven housekeeping genes were sequenced (Wirth et al, 2006) and STs were assigned according to the E. coli MLST website (http://mlst.ucc.ie/mlst/dbs/Ecoli). Similarly, sequences of the 53 rMLST loci were determined, alleles were assigned in accordance to the rMLST database (Jolley et al, 2012) and listed in supplementary Table S2. The minimum spanning tree based on the MLST and rMLST allelic profiles (in total 60 loci, ca. 24.6 kb) was generated using the SeqSphere software version 0.9 beta (Ridom GmbH, M€ unster, Germany). Classification into ECOR phylogenetic groups A, B1, B2, and D was performed as described (Clermont et al, 2000). For whole genome shotgun sequencing of selected O2:H6 strains (see supplementary Table S1), sequencing libraries were prepared using the Nextera XT chemistry (Illumina Inc., San Diego, CA, USA) for either a 100 bp or a 250 bp pairedend sequencing run on an Illumina HiScanSQ or MiSeq sequencer in accordance to the manufacturer′s recommendations (Illumina). After quality trimming using the default parameters of the CLC Genomic Workbench software (CLC bio, Arhus, Denmark) the sequencing reads were assembled using the CLC Genomic Workbench de novo assembler (CLC bio). Gene sequences for subsequent analyses were extracted from contigs using the Ridom Seqsphere software version 0.9 beta (Ridom GmbH). For the gene-by-gene core genome analysis as described , we included all genes present in all strains analyzed (see supplementary Table  S1). The whole genome sequence reads have been deposited at ENA SRA (study accession no. PRJEB4756).

FimH sequencing
FimH protein sequences were derived from the whole genome sequences and translated using the Ridom Seqsphere software version 0.9 beta (Ridom GmbH). The Neighbor-joining tree of the FimH sequences was created using the MEGA software (Tamura et al, 2011). For comparison, the published sequences of AIEC/UPEC typical FimH alleles (Dreux et al, 2013) were included.
Sequence analysis of stx genes stx genes of six randomly selected STEC O2:H6 strains (supplementary Table S1) were amplified and Sanger sequenced; resulting sequences were analyzed using the SeqSphere software version 0.9 beta (Ridom GmbH) and homologies were sought in GenBank (http://www.ncbi.nlm.nih.gov/BLAST). The stx 2 sequence from strain 03-08304 (representative of those present in all six sequenced strains) (stx 2b subtype) was deposited in GenBank (accession no. GU126552).

Phenotype determinations
Stx titers were determined by Vero cell cytotoxicity (Bielaszewska et al, 2006), and production of a-hemolysin was determined on Columbia blood agar (Heipha, Heidelberg, Germany) after overnight incubation. Production of Saa and CNF1 was assayed by immunoblot (Paton et al, 2001). Briefly, lysates of overnight Luria-Bertani (LB) broth cultures were subjected to sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE), separated proteins were transferred to a membrane (Immobilon P, Roth, Karlsruhe, Germany) and probed with polyclonal mouse anti-Saa antibody (Paton et al, 2001) (a gift from J. C. Paton, Adelaide University, Adelaide, Australia) or monoclonal mouse antibody against CNF1 + CNF2 (clone JC4) (Abcam, Cambridge, UK). Bound antibodies were detected using alkaline phosphatase-conjugated goat anti-mouse IgG (Dianova, Hamburg, Germany). Colibactin expression was tested (Nougayr ede et al, 2006) using HeLa cells (ATCC CCL-2) maintained in Eagle minimum essential medium supplemented with 10% fetal calf serum (FCS), 2 mM Lglutamine, and 1% non-essential amino acids (Cambrex Bioscience, Verviers, Belgium). For cell cycle analysis, the cells were cultured in 12-well plates seeded with 7.5 × 10 4 cells per well. Semiconfluent monolayers were infected with overnight LB broth cultures of tested isolates diluted in interaction medium (culture medium with 5% FCS) to a multiplicity of infection (number of bacteria per cell) 100:1, and cocultured for 4 h (37°C, 5% CO 2 ). Cells were then extensively washed and incubated in full fresh medium supplemented with gentamicin (200 lg/ml) for 48 h. After harvesting, the cells were stained with propidium iodide-containing Nicoletti buffer (Bielaszewska et al, 2005a) and the DNA content of the nuclei was analyzed by flow cytometry on FACScalibur (Becton Dickinson, Heidelberg, Germany) as described previously (Bielaszewska et al, 2005a). For the cell distension assay, HeLa cells (3 × 10 4 per well) were seeded into 24-well microtiter plates. Semiconfluent monolayers were exposed for 4 h to overnight bacterial cultures as described above, the cells were then washed, and incubated in full medium with gentamicin for 48 h; morphology of native cells was examined using a light microscope (Axiovert 40; Zeiss, Jena, Germany).
Contact-dependent growth inhibition was determined (Aoki et al, 2005) using three randomly selected cdiAB-positive STEC O2:H6 strains and an cdiAB-negative STEC O2:H27 strain as inhibitors and E. coli K-12 strain MG1655 transformed with pBluescript KS II(+) (Stratagene, La Jolla, CA, USA) encoding ampicillin resistance as a target strain. Briefly, a log-phase LB broth culture of each inhibitor strain was mixed (ratio 50:1) with a stationary culture of E. coli MG1655/pBluescript KS II(+) and cocultured (37°C, 225 rpm) for 1.5, 3, 4.5, and 6 h. At each time point, growth of the target strain incubated with each inhibitor and alone (growth control) was determined (CFU/ml) by plating 10-fold dilutions of the cultures on LB agar with ampicillin (100 lg/ml) (to which all inhibiting bacteria were susceptible).
To determine if the isolates expressed Vat, CHO-K1 cells (ACC 110; German collection of microorganisms and cell cultures, Braunschweig, Germany) were seeded in 24-well plates (4 × 10 3 cells per well) in Ham′s F12 medium with 10% of FCS (Cambrex). The cells were incubated with sterile overnight culture supernatants of the strains for 24 h (37°C, 5% CO 2 ) and vacuoles were sought by microscopy (Axiovert 40).

Experimental murine model of ascending UTI
Mice experiments were performed according to the guidelines for the Care and Use of Laboratory Animals in compliance with German regulations (Tierschutzgesetz). Permission for this study was provided by the regional government (AZ 55.2-2531.01-53/09). Mice (C57BL/6; female, 6-7 weeks old) were obtained from Charles River Laboratories (Sulzfeld, Germany) and kept under specific pathogen-free conditions. For murine infection, the bacterial strains were grown overnight under aerated conditions in LB medium, harvested by centrifugation and resuspended in sterile saline to a concentration of 1 × 10 10 to 1 × 10 11 CFU/ml. Groups of 8-15 C57BL/6 mice were transurethrally inoculated with 5 × 10 8 -5 × 10 9 CFU of strain 536 (UPEC, positive control), STEC O2:H6 strains 05-00787, 04-00955, or 03-08304, or non-pathogenic E. coli K-12 strain MG1655 (negative control) as previously described (Hagberg et al, 1983). In each case, animals were sacrificed 72 h after infection. The bladder and kidneys were removed under sterile conditions, rinsed extensively with 0.9% NaCl and homogenized mechanically in 1 ml of 0.025% Triton-X 100, 0.9% NaCl. The number of bacteria was quantified by scoring CFU after overnight culture at 37°C on LB agar plates. The results were expressed as CFU/g of tissue. Statistically significant differences in colonization levels (P < 0.05) were assessed using Graphpad Prism 5 software. If the bladder challenge data were normally distributed, they were analyzed using an unpaired one-tailed t-test. Kidney infections were analyzed using the unpaired, non-parametric one-tailed Mann-Whitney test. The paper explained Problem Classic evolutionary theory teaches that pathogens arise from nonpathogens by horizontal acquisition of virulence genes. This theory implies a unidirectional path to virulence, and a non-pathogen to pathogen gradient. Hybrid pathogens that combine virulence traits of different pathogroups can cause severe diseases in humans but their evolutionary history is poorly understood. Specifically, we do not know if they are 'weaponized' by the step-wise acquisition of virulence loci, or if pathogens transition from one pathogroup to another by multilocus emergence.

Results
Shiga toxin-producing E. coli (STEC) O2:H6 are phylogenetically positioned between intestinal pathogenic STEC and uropathogenic E. coli (UPEC). They possess and express virulence factors associated with both STEC and UPEC. This hybrid causes both diarrhea and urinary tract infection. STEC O2:H6 are thus heteropathogens that occupy an evolutionary and pathogenic interface between intestinal and extraintestinal pathogenic E. coli. They are in a transitional evolutionary and virulence state.

Impact
The phylogeny and virulence potential of STEC O2:H6 compels us to propose a novel evolutionary concept whereby one pathogroup of E. coli undergoes phased metamorphosis into another. Improved understanding of evolutionary mechanisms of human pathogens could inspire novel strategies to predict pathogen emergence, and counteract their detrimental effects on human hosts.