Characterisation of pathogenic Escherichia coli associated with food and irrigation water in South Africa

Diarrheagenic Escherichia coli (DEC) has been implicated in foodborne outbreaks worldwide and additionally associated with childhood stunting in the absence of diarrhoea. DEC are pathogenic E. coli that cause human gut gastrointestinal infections. They are categorized based on how they elicit disease into groups called pathotypes. Infection is extraordinarily common, but the routes of transmission have not been determined. This work categorized into three phases 1, 2 and 3 aimed at characterizing pathogenic E. coli previously isolated from food sources and irrigation water in South Africa. In phase 1, the study characterized 205 E. coli strains previously isolated from producer distributor bulk milk (PDBM)(n=118), irrigation water (n=48), irrigated lettuce (n=29) and street vendor coleslaw (n=10) in South Africa. Enteropathogenic E. coli (EPEC), enterotoxigenic E. coli (ETEC), enteroaggregative E. coli (EAEC) and diffusely adherent E. coli (DAEC) were sought. The study used PCR and partial gene sequencing for all 205 strains while 46 out of 205 that showed poor resolution were subsequently characterized using cell adherence (HeLa cells). PCR and partial gene sequencing of aatA and/or aaiC genes confirmed EAEC (2%, 5 out of 205) as the only pathotype. Strains from Genbank showing ?80% nucleotide sequence similarity with those used in this study based on possession of the aaiC gene included 10 clinical and 5 food strains. On the other hand, all strains from GenBank (n=22) showing high nucleotide sequence similarity (?80%) to those from this study based on possession of the aatA gene were clinical. Cluster analysis of sequenced EAEC strains with E. coli strains in GenBank showing ?80% nucleotide sequence similarity based on possession of aaiC and aatA generated distinct clusters of strains separated predominantly based on their source of isolation (food source or human stool). This suggested a potential role of virulence genes in source tracking. EAEC 24%, 11 out of 46 strains (PDBM=15%, irrigation water=7%, irrigated lettuce=2%) was similarly the predominant pathotype followed by strains showing invasiveness to HeLa cells, 4%, 2 out of 46 (PDBM=2%, irrigated lettuce=2%), using cell adherence. All strains that were not characterized based on possession of virulence genes and cell adherence assays were designated as non-DEC. These non-DEC strains were subsequently used as the basis for characterization in phase 2. In phase 2, the study assayed non-DEC strains from environmental sources in South Africa for phenotypes that may be associated with intestinal dysfunction (ID). DEC strains were also used. The strains had previously been isolated from PDBM, irrigated lettuce, street vendor coleslaw and irrigation water. In-vitro assays included; biofilm formation (n=38), extracellular polymeric substance (EPS) formation (n=38), cytotoxic activity (n=10), disruption of tight junctions and induction of Interleukin 8 (IL-8) on polarized T-84 cells (n=20). The number of strains tested for each assay differed, depending on prior molecular and phenotypic characterization that signalled potential pathogenicity. Subsequently, all strains having data points for all analyses were used to compute Principal Component Analysis (PCA) plot curves to infer associations amongst test strains. Biofilm formation varied based on pathotype (DEC and non-DEC) and source of isolation suggesting these two factors influence persistence within a defined environmental niche. Additionally, DEC isolated from irrigated lettuce had significantly higher (p?0.05) biofilm formation in both media compared to all strains including DEC standard controls suggesting irrigated lettuce as a potential source of persistent pathogenic strains. All strains were able to form EPS suggesting ability to form mature biofilms. Of the (60%, 6 out of 10) strains showing cytotoxic activity, most (83%, 5 out of 6 strains) were non-DEC isolated from food sources. Mean percentage reduction in initial TER (a measure of intestinal disruption) in all test strains, was comparable (53.5 to 73.8%) to that observed in the standard DEC. Additionally, IL-8 induction from strains isolated from PDBM (139pg/mL), irrigation water (231.93pg/mL) and irrigated lettuce (152.98pg/mL) was significantly higher (p?0.05) than in the commensal strain aafa. PCA categorized strains based on sources of isolation and showed potential for use in source tracking. This study shows that non-DEC strains along the food chain possess characteristics that may lead to ID. However further investigations with a larger collection of strains may provide a clearer link to these observations.