Future perspectives, applications and challenges of genomic epidemiology studies for food-borne pathogens: A case study of Enterohemorrhagic Escherichia coli (EHEC) of the O157:H7 serotype

Abstract

The shiga-toxin (Stx)-producing human pathogen Escherichia coli serotype O157:H7 is a highly pathogenic subgroup of Stx-producing E. coli (STEC) with food-borne etiology and bovine reservoir. Each year in the U. S., approximately 100,000 patients are infected with enterohemorrhagic E. coli (EHEC) of the O157:H7 serotype. This food-borne pathogen is a global public health threat responsible for widespread outbreaks of human disease. Since its initial discovery in 1982, O157:H7 has rapidly become the dominant EHEC serotype in North America. Hospitalization rates among patients as high as 50% have been reported for severe outbreaks of human disease. Symptoms of disease can rapidly deteriorate and progress to life-threatening complications such as Hemolytic Uremic Syndrome (HUS), the leading cause of kidney failure in children, or Hemorrhagic Colitis. In depth understanding of the genomic diversity that exists among currently circulating EHEC populations has broad applications for improved molecular-guided biosurveillance, outbreak preparedness, diagnostic risk assessment, and development of alternative toxin-suppressing therapeutics.

EHEC

The pathovar definition of enterohemorrhagic Escherichia coli (EHEC) is based on key virulence factors such as carriage of both the eae and shiga toxin (stx) genes encoded on the Locus of Enterocyte Effacement (LEE) and Stx-converting prophages. The genomic comparison of EHEC to other E. coli elucidated the close relatedness of the O157 and O55 serotypes.^1,2 A wealth of genomic data has now become available for E. coli O157:H7 isolates and allows examination of this EHEC pathotype in molecular detail, thus providing the foundation for novel surveillance, detection and diagnostic strategies. Analysis within EHEC isolates must be conducted on both a fine and genome-wide scale using whole genome sequence-based synergistic approaches in order to assess the content of conserved and laterally acquired genes.

Epidemiology and Public Health Relevance of E. coli O157:H7

The shiga-toxin-producing non-sorbitol fermenting, and β-glucuronidase negative Escherichia coli O157:H7 serotype is a highly pathogenic subgroup of Shiga-toxin-producing E. coli (STEC) with a food-borne etiology and a bovine reservoir^3,4 that poses a serious human health concern as well as a significant financial burden due to product recalls and related incalculable societal costs. The significant differences in host prevalence, transmissibility, and virulence phenotypes among strains from bovine and human sources are of major interest to the public health community and livestock industry.⁵ Cattle are recognized as a main reservoir of Shiga-toxin-producing E. coli O157:H7 (STEC), and supershedders are of particular concern for the spread and transmission within the human food chain.^6,7 E. coli O157:H7 is the major cause of hemorrhagic E. coli disease in North America causing 100,000 human infections each year in the US alone.^8-10 An estimated 15–20% of people infected with E. coli O157:H7 present with gastrointestinal indications severe enough to require hospitalization.¹¹ In such cases, symptoms may progress to life-threatening complications, such as Hemolytic Uremic Syndrome (HUS).^12-15

EHEC Genomes

E. coli is one of the most intensively studied bacterial species, and a rich comprehensive collection of genomes of different pathovars has become available since the release of the first complete genomes of strains EDL933 ¹⁶ and Sakai.¹⁷ Genomes of pathogenic EHEC are distinguished from non-pathogenic E. coli representatives by the lateral acquisition (∼20%) of EHEC-specific DNA through multiple independent horizontal transfer events, which are integrated non-randomly and accumulated in the O-islands.¹⁸ The genomic hallmarks of this in particular pathogenic pathovar are its abundance of lambdoid-like phage pool, of including Stx-converting prophages, and carriage of the lineage-specific pO157 virulence plasmid.^19,20 More recently, the sequencing of a number of isolates from the same food-borne outbreak revealed the genomic heterogeneity that is now acknowledged to exist and cautions against the use of a single reference genome to define a particular serovar, pathovar, or species.^21,22

Genome Plasticity

Despite the fact that multiple genomes of this lineage have become available, many of which were contributed by our group,^21,23,24 the genomic diversity that exists among currently circulating populations of this genetically highly homogenous lineage is poorly understood. The ability to accurately track outbreak strains to the contaminated source is central to public health communities, but current epidemiological and molecular investigations rely on only a limited set of genetic markers and thus often fail to distinguish individual isolates. This serotype was first identified in 1982 in conjunction with human disease during a hamburger outbreak in Michigan, although a single O157:H7 strain that predates this outbreak has been recovered.²⁵ The rapid emergence of the O157:H7 serotype to the now dominant enterohemorrhagic E. coli (EHEC) serotype, causing widespread and potentially lethal outbreaks of human illness, highlights the need to critically assess the degree of genomic plasticity and to evaluate the extent to which different O157:H7 genotypes contribute to disease severity and outcome in infected patients. Surveillance data have demonstrated a high prevalence of E. coli O157:H7 among cattle, but a relatively low incidence of human infection. This supports the hypothesis that only a subset of, rather than all, E. coli O157:H7 inhabiting cattle is responsible for the majority of human disease.⁶ Comparative phylogenomic analyses of bovine and clinical isolates identified genome signatures in bovine dominated lineage II strains and allowed us to conjecture the detected bovine genotypes with altered pathogenic potential, bovine niche adaptation, and differentiation of bovine super shedders.^7,21,23

Pre- and Post-Genomic Typing Methodologies for EHEC Outbreak Investigations

Genetic heterogeneity among O157:H7 EHEC strains has been documented using a variety of molecular methodologies. Pre- and post-genomic technologies include, but are not limited to, methodologies such as multilocus sequence typing (MLST),²⁶ octamer and PCR-based genome scanning,^27,28 phage typing,^7,29,30 multiple-locus variable-number tandem repeat analysis (MLVA),^31,32 microarrays,³³ microarray-based comparative genome hybridization (CGH),³⁴ nucleotide polymorphism assays,^21,35,36 pulsed-field gel electrophoresis (PFGE),³⁷ subtractive hybridization^38,39 and whole-genome mapping.^21,23,40,41 Molecular typing has revealed divergence of E. coli O157:H7 into 3 lineages: lineage I and I/II strains are commonly associated with human disease, while lineage II strains are overrepresented in the asymptomatic bovine host reservoir.^23,34

The E. coli O157:H7 Mobilome

The size and genomic location of laterally acquired genomic regions in EHEC are crucial in assessing the Shiga toxin virulence state, as it is determined by Stx allele prevalence and the respective chromosomal insertions sites of the converting prophages.^21,23,40,41 Assays pioneered by Tarr et al. have been developed to determine the toxin type and its relevance in the evolution of E. coli O157.⁴² Comparative map analysis reveals valid biological markers to trace evolution, and prophage-typing assays assist molecular epidemiology outbreak investigations in determining the stx virulence type.²¹ These techniques revealed that O157 strains partitioned into 3 different, but interlinked, lineages, and genetic evidence suggests that distribution of the genotypes of these lineages, such as polymorphisms in Shiga toxin subtypes and synergistically acting virulence factors, are correlated with phenotypic differences of major biological relevance in virulence, host ecology, and epidemiology.

Transition From Targeted to Whole Genome-Based Typing Approaches

PFGE has served as the “gold standard” for the last 25 y in epidemiological investigations of EHEC O157:H7. The XbaI PFGE digest patterns of EHEC O157:H7 give insights into the genome structure and show changes in restriction patterns anchored in spontaneous genomic rearrangements or recombination of mobile elements,^43,44 though the underlying genetic composition of the isolates remains unknown by the very nature of this technology. A refined classification system that relies on a PCR-based assay and interrogates the repeat length at 6 genic and intergenic chromosomal loci, i.e., the “lineage-specific polymorphism assay” (LSPA), ultimately separated E. coli O157:H7 into the aforementioned lineages.⁴⁵ Both MLST and PFGE schemes and databases, for example, exist for E. coli ^46,47 as a first line for inclusion/exclusion criteria in examining E. coli O157:H7 isolates within the narrow time window of a food-borne outbreak. Yet, with the shift of technology to whole genome sequence type based approaches, the investigation of outbreaks of E. coli ^48-53 has clearly achieved unprecedented phylogenetic resolution of closely related outbreak isolates, and, moreover, sequence-based methods offer powerful, standardized and automated subtyping tools for molecular epidemiology investigations.

Whole Genome Sequence Typing/Shortcoming of Single Colony Isolates and Benefits of Population Sequencing

Whole genome analysis of very similar E. coli O157:H7 isolates must be used to assess both the conserved and laterally acquired gene complements. There is no gold standard, and the favored methodologies in achieving short- or long-term resolution power largely depend on the types of scientific enquiries and data sets being analyzed. With potentially lethal and widespread outbreaks, a large-scale and in-depth survey of both genic and genomic polymorphisms is essential for obtaining crucial insights into pathogenome evolution, i.e., the relation of E. coli O157:H7 disease-pathotypes to other E. coli pathogens. The application of these newer higher resolution technologies will help elucidate the evolutionary paths that generate a fully virulent E. coli O157:H7 isolate from an evolutionary related E. coli O55:H7 ancestor,⁵⁴ or perhaps more recent progenitors, thus allowing trends and dynamics of pathogenome evolution to be followed at a high level of phylogenetic accuracy and resolution.^{7,21,35,55,56} While the current molecular markers and typing assays used by public health microbiology laboratories may be adequate for routine surveillance and identification of E. coli O157:H7, these approaches lack the discriminatory power and the polymorphic markers necessary for studying the relatedness of strains of unknown provenance. With the increased use of next generation sequencing (NGS) technologies, whole genome sequence typing (WGST) approaches, for example use of single nucleotide polymorphisms (SNPs), have more and more gained popularity. SNP typing not only provides stable long-term genetic markers to study evolution but also offers higher phylogenetic resolution. SNP discovery approaches have yielded thousands of high-quality SNPs as critical bases for the development of a refined phylogenomic framework.^9,21,57 Canonical SNPs can be implemented in an efficient and cheap typing assay that surpasses classical technologies in terms of phylogenetic accuracy and resolution. SNP typing, in particular, allowed elucidation of the evolutionary origin and emergence of pathogenic O157:H7 lineages and revealed the genetic relationships of the intermediate non-motile O157:H(-) and ancestral EPEC O55:H7 serotypes.⁹ A SNP-based typing assay that detects SNPs in 96 loci has been applied to >500 clinical E. coli O157 strains, resulting in a refined LSPA-6 lineage classification that not only further separated EHEC isolates into 9 distinct clades but also correlated the frequency and distribution of Stx-converting prophages with varying types of clinical sequelae.^35,58 Manning et al.,³⁵ for example, used SNP analysis to identify a unique group (clade 8) of hypervirulent E. coli O157:H7 strains found in the United States.^21,58-60

Pathogen Definition Bias

The definition of a pathogen and its adverse pathogenic potential is unfortunately often skewed by anthropogenic biases. Virulence factors must be assessed within their ecological context, i.e., as survival factors, and might rather mediate better adaptation to distinct ecological niches, thus in turn gaining secondary access into the human food production and consumption chain. For example, cellular appendages, such as fibers, are only prevalent in subset of E. coli and promote adhesion to green leaf vegetables, without rendering the curli positive group more pathogenic to humans, per se, but adhesion does mediate access to the human food chain as prerequisite in causing widespread outbreaks of human disease.^8,61 A more detailed understanding of the complex interplay of the pathogen, host, and environment offers potential for developing new protection strategies. The host background and genetics in the infected patients such as age, immune status or animal reservoir, though these underlying individual genetic components are often not available to the investigator, are crucial factors in the severity of the disease manifestation. When, for example, the lack of vascular receptors in cattle was observed, this suggested reasons for why these animals were Stx tolerant.⁶²

What is the degree of genomic heterogeneity found in a bacterial outbreak? Are there hypervirulent clades of E. coli O157:H7 that can be delineated from phylogenomic or functional studies? It has been noted that distinct E. coli O157:H7 virulence genotypes are associated with and contribute to the progression of severe clinical forms of disease in infected patients. The E. coli O157:H7 outbreak in 2006 from ingested spinach captured the attention of both the public health and lay communities. Rather than studying the dissemination of the few established lineage-specific polymorphic markers for this study, we performed a WGST survey on a lineage-wide scale to assess the genome plasticity and to resolve the very limited marker base. This outbreak, like others associated with green leafy vegetables, resulted in an increased number of individuals being admitted to healthcare facilities. The spinach outbreak caused 199 illnesses across 26 states and illnesses traced to ingestions of fresh spinach contaminated with E. coli O157:H7, and this outbreak claimed the lives of 2 elderly women and a 2-year-old child. Among the ill, 51% were hospitalized and in 16% of the cases, infection progressed to HUS and kidney failure. The increased morbidity observed for this outbreak suggested that a more virulent strain of E. coli O157:H7 was responsible for these illnesses. These isolates of the most recent outbreaks have been examined by genome sequencing as part of the epidemiological and microbiological examination and identified specific-gene content in likely hypervirulent clades,^{21,59,60,58-60} though it is still largely unknown how the bacterial genotype relates to clinical disease manifestation and severity.

Genomics-Guided Outbreak Investigations

Enriched mutational database

Mounting evidence suggests that genotypic differences between these lineages, such as polymorphisms in Shiga toxin subtypes and synergistically acting virulence factors, are correlated with phenotypic differences in virulence, host ecology, and epidemiology.²³ The growing O157:H7 database of genomic variants serves as a critical resource for molecular epidemiological investigations, allowing one to reconstruct the evolutionary emergence of E. coli O157:H7 and to ascertain whether particular variant genotypes reflect the presence of highly pathogenic clones within EHEC O157:H7 populations.⁶³ This resource is critical to partition precisely this globally emerging pathogen into subgroups distinguished by different genotypic, clinical characteristics, and phenotypic traits. This again highlights the crucial need for the development of more stable genetic biomarkers from the ever-accumulating genomic information in the post-genomics era. The refined frameworks have begun to show distinct phylogeographic genotype profiles in strains from North America, Europe (UK), Australia and Asia.^21,23,35 Definitive conclusions regarding strain and source attribution, however, will require the need for a robust genomic database that captures the diversity of extant E. coli O157 strains from temporally- and geographically-diverse settings, a database that is well curated and fortified with informative clinical, environmental, and epidemiological metadata.⁶⁴ Garnering this information efficiently in order to wage effective infectious disease research in today's genomic world calls for the convergence and cooperation of many scientific disciplines.

Future directions

The decrease in cost and increase in higher-throughput sequencing capabilities has made possible the large-scale sequencing of significant numbers of isolates associated with outbreak and environmental settings. We envision that in the near future it will be possible to routinely, and in a cost-effective manner, sequence complete genomes of clinical bacterial isolates, thereby enabling molecular epidemiology at a system-wide scale and with previously unprecedented resolution. Several research groups have applied genome sequence based high-resolution genotyping in an attempt to associate genetic changes with noted increases in virulence and/or phylogenetic markers to identify and track outbreak strains down to the individual strain level. For the study of the 2006 spinach outbreak we had sequenced, at the time, what was considered a rather large number of isolates (#25) to reconstruct the “anatomy of the outbreak”,²¹ and through comparative phylogenomics identified robust mutational and structural markers in the core and the mobilome, providing an exceptional level of phylogenetic resolution and accuracy and established a panel of markers that distinguished outbreak isolates from clinical sources and those derived from animal reservoirs. This type of study demonstrated how the assessment of the genome variability allows identification to a very fine scale, even among genomically very similar isolates, typically observed within a single point source foodborne outbreak of disease. Clearly, much has changed since the release of the first 2 EHEC genomes in the mid 90s or even the studies from only a few short years ago.^16,17 Next generation sequencing technology has matured, thus enabling deep sampling and phylogenomic analysis of ever-larger bacterial pathogen populations and metagenomic samples. Now, tens or even hundreds of bacterial isolates can be sequenced rapidly, thus obviating the reliance on a single archival strain or even a few archetypical outbreak strains that may not be fully representative of the population of bacteria responsible for the disease outbreak. So too NGS makes possible large-scale comparative genome analyses between and among multiple outbreak-associated isolates; analyses that can reveal multiple, unique molecular markers within the particular pathogen of interest that permits better strain attribution. A comparison of multiple genomes from clinical and animal reservoirs has been extremely valuable in our understanding of the extant populations of E. coli O157:H7, revealing the subtle, yet very important, genomic plasticity in this genetically homogenous lineage and helping to identify regions among this pathovar and across other EHEC serotypes that may be associated with increased virulence. In taking into account clinical presentation and gene prevalence and distribution,^21,58-60 comparative and functional phylogenomics has ushered in the dawning of genomic epidemiology.

Epilogue

We have restricted our discussion to E. coli O157:H7, although the genomic principles described herein have general implications for examining outbreaks of disease caused by other pathogens, as evidenced, for example, by genomic interrogation of multiple isolates of methicillin-resistance Staphylococcus aureus in a health care setting,⁶⁵ Pseudomonas aeruginosa in a hospital-associated incident,⁶⁶ Vibrio cholerae in the Haitian cholera outbreak,²² or enteroaggregative Stx-positive E. coli O104:H4 in the German food-borne outbreak.^48-51,53 In the NGS era we have transitioned from sequencing prototype isolates from a group of pathogens to sequencing large numbers of isolates of the same pathogen in an effort to understand population structures of pathogens associated with human health.

For E. coli O157:H7, the enriched mutational database resources generated by NGS and attendant technologies will provide a robust foundation to better associate genotypic group profiles and EHEC virulence phenotypes within particular strains or population of strains. Despite the fact that multiple genomes of this lineage have become available in the genomics era,^21,23,24 biological insights into the mechanism of disease suffer due to a deficiency of markers for accurate typing and genotype(s) to phenotype association. High-resolution phylogenomic approaches offer the accuracy and resolution ^7,21,35,56 necessary to explore the dynamics of pathogenome evolution. Although the current molecular markers and typing assays used by public health microbiology laboratories may be adequate for routine surveillance and identification of E. coli O157:H7, these approaches lack polymorphic markers and discriminatory power necessary to study the relatedness of strains of unknown provenance, which becomes of particular importance when investigating outbreak strains that form tight clonal complexes, containing a paucity of genetic polymorphisms. Sequence-based information is critical in linking a particular strain to an outbreak of human disease and in tracing the implicated strain to its ultimate source. Early outbreak isolates often elude epidemiological detection, yet timely strain- and source-attribution are critical in limiting the spread and transmission of a pathogen and in enacting effective risk management strategies. Such information, for example, enables to track and differentiate separate, but temporally coincident, outbreaks and to trace and type isolates as members of a particular outbreak population^.21,67 Moreover, genomic data better defines the pathogenic strain of interest, thus facilitating development of a better and more effective epidemiological schema.

The aim of the collection of data sets gathered by independent research groups is to provide a more complete and refined picture of the E. coli O157:H7 pathogenome evolution and its pathovar-specific characteristics and pathogenic potential in relation to the E. coli species biology. Yet, without coupling biological function with informatics, it will not be possible to understand the dynamic variability in gene content nor the activities among identified O157 genotypes, e.g., in phage mobilization and resulting stx production as a major driver of E. coli O157:H7 genome evolution and pathogenesis. Though a widely accepted host model system for virulence studies is still lacking,^68,69 applications of technologies like DNA microarray analysis for gene profiling and differentiation of strains and phenomics for characterization of a pathogen's physiological and metabolic capabilities allows one to evaluate genetic predictions and to correlate isolate-specific gene content and mutations with outbreak-associated physiological and virulence capabilities.

This enriched marker base is crucial in formulating a molecular-guided surveillance and outbreak investigation strategy aimed at reducing human morbidity and mortality rates of hemorrhagic E. coli (Fig. 1).

Figure 1.

Genomic epidemiology outbreak investigation. Sequence data can be transmitted into timely and informed actionable countermeasures. Sequencing speed achieved by current NGS technology enables to obtain complete sequences, functional annotations and genomic comparison from multiple samples within hours. Outbreak investigation needs a close collaboration between different scientific disciplines and constant exchange of meta-, experimental- and in-silico derived data sets. After recognition of an outbreak, the investigation would simultaneously initiate the collection of isolates for bacterial cultivation and DNA preparation. This is the foundation for both a whole array of experimental biochemical assays and DNA-based technologies, such as genome sequencing and whole genome sequence typing and discovery in the genomic backbone and mobilome .

In the future the enriched databases of mutational and structural makers will allow the examination of the pathogenicity of these isolates and the true impact of novel combinations of virulence factors will be identified. Although the described phylogenomic principles are geared toward E. coli O157:H7, the developed approaches are readily translatable and have been deployed to other emerging infectious E. coli pathotypes.^49,70 The E. coli outbreak in Europe in 2011 shows the full potential in the use of genomics to examine outbreak isolates in near real time^48-51,53 promoting the concept of genomic epidemiology.

In the near future, genomics will allow the refinement of boundaries and evolutionary relationships among EHEC pathovars and allow one to reassess the degree of genomic plasticity by deploying high-resolution whole genome-based sequence typing methodologies on growing sets of samples. It is thus critical that cultures be properly maintained and curated within publicly funded repositories and made available to the scientific community at-large. So too must proper attention be paid to maintaining the attendant metadata that accurately describes the geographical and temporal derivation of environmentally- and clinically-derived strains with proper ascription of their genotypic and phenotypic traits. Such penchant for detail will not only benefit present and future researchers but also prove invaluable in criminal and public health investigations.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Acknowledgments

This article is based upon work at the University of Texas at San Antonio supported by the Army Research Office of the Department of Defense under Contract No. W911NF-11-1-0136. The authors would like to thank Sara SK Koenig for critically reading the manuscript.

Funding

This work received support from the South Texas Center of Emerging Infectious Diseases (STCEID), Department of Biology and Computational System Biology Core at the University of Texas at San Antonio.