DNA microarray-based assessment of virulence potential of Shiga toxin gene-carrying Escherichia coli O104:H7 isolated from feedlot cattle feces

Escherichia coli O104:H4, a hybrid pathotype reported in a large 2011 foodborne outbreak in Germany, has not been detected in cattle feces. However, cattle harbor and shed in the feces other O104 serotypes, particularly O104:H7, which has been associated with sporadic cases of diarrhea in humans. The objective of our study was to assess the virulence potential of Shiga toxin-producing E. coli (STEC) O104:H7 isolated from feces of feedlot cattle using DNA microarray. Six strains of STEC O104:H7 isolated from cattle feces were analyzed using FDA-E. coli Identification (ECID) DNA microarray to determine their virulence profiles and compare them to the human strains (clinical) of O104:H7, STEC O104:H4 (German outbreak strain), and O104:H21 (milk-associated Montana outbreak strain). Scatter plots were generated from the array data to visualize the gene-level differences between bovine and human O104 strains, and Pearson correlation coefficients (r) were determined. Splits tree was generated to analyze relatedness between the strains. All O104:H7 strains, both bovine and human, similar to O104:H4 and O104:H21 outbreak strains were negative for intimin (eae). The bovine strains were positive for Shiga toxin 1 subtype c (stx1c), enterohemolysin (ehxA), tellurite resistance gene (terD), IrgA homolog protein (iha), type 1 fimbriae (fimH), and negative for genes that code for effector proteins of type III secretory system. The six cattle O104 strains were closely related (r = 0.86–0.98) to each other, except for a few differences in phage related and non-annotated genes. One of the human clinical O104:H7 strains (2011C-3665) was more closely related to the bovine O104:H7 strains (r = 0.81–0.85) than the other four human clinical O104:H7 strains (r = 0.75–0.79). Montana outbreak strain (O104:H21) was more closely related to four of the human clinical O104:H7 strains than the bovine O104:H7 strains. None of the bovine E. coli O104 strains carried genes characteristic of E. coli O104:H4 German outbreak strain and unlike other human strains were also negative for Shiga toxin 2. Because cattle E. coli O104:H7 strains possess stx1c and genes that code for enterohemolysin and a variety of adhesins, the serotype has the potential to be a diarrheagenic foodborne pathogen in humans.

Introduction Escherichia coli O104:H4, a hybrid pathotype possessing genes characteristic of enteroaggregative E. coli (EAEC) and Shiga toxin (Stx)-producing E. coli (STEC), was responsible for a large foodborne outbreak of hemorrhagic colitis and hemolytic uremic syndrome in Europe, mainly Germany, in 2011. The O104:H4 serotype was also isolated from patients suffering from bloody diarrhea in the Republic of Georgia in 2009 and comparative genomics revealed that they are nearest neighbors to 2011 German outbreak strains. However, there were structural and nucleotide differences in plasmid and prophage profiles and in antimicrobial resistance between German and Georgian strains [1]. Because Stx-negative EAEC O104:H4 had been previously isolated from humans in African countries, it was suggested that the hybrid pathotype has evolved from EAEC O104:H4 by the uptake of Stx-carrying bacteriophage [2][3][4][5]. Escherichia coli O104:H4 serotype has not been detected in cattle, unlike other major STEC serotypes that cause human illnesses [6][7][8]. However, cattle harbor O104 serotypes other than H4 in the gut and shed them in the feces [6,9,10]. A stx2-carrying O104:H21 serotype, possibly of cattle origin, was implicated in an outbreak of hemorrhagic colitis associated with consumption of raw milk in Helena, Montana in 1994 [11]. We have conducted a study with cattle feces collected from a number of feedlots (n = 29) to determine the prevalence of O104 serogroup and reported that serotype O104:H7 was the most commonly isolated stx-carrying O104 serotype [9]. The O104:H7 serotype is reported to be associated with sporadic diarrheal cases in humans [12,13], however, there has been no evidence to suggest that cattle were the source of human infections. Not much is known about the virulence potential of the strains of E. coli O104:H7 of cattle origin, particularly in relation to human clinical strains of O104:H7. Therefore, the objectives of our study were to analyze the gene content to assess the virulence potential of cattle O104:H7 strains using FDA-ECID DNA microarray and to compare their virulence gene profiles with that of human STEC O104:H7 (clinical strains), O104:H21 (milk-associated Montana outbreak strain) and O104:H4 (German outbreak) strains.

Materials and methods
Kansas State University Institutional Biosafety Committee and Institutional Animal Care and Use Committee approved this study.

Microarray assay and data analysis
The strains were subjected to a custom Affymetrix DNA microarray developed by the FDA for identification and characterization of E. coli [14]. The array was designed using 368 E. coli and Shigella sequence sets to identify 55,918 annotated open reading frames. The array incorporates 41,932 probe sets, which includes 54 closed chromosomes, 47 closed plasmids, and 267 whole genome shotgun sequences. The array can be used to identify 152 O types, 54 H types, 4 stx1 subtypes, 8 stx2 subtypes, 48 eae subtypes, and other virulence genes. The microarray assay was performed according to the protocol described by Patel et al. 2016. Briefly, each strain was grown overnight in Luria broth (Sigma-Aldrich Co., St. Louis, MO) at 37˚C. Total DNA was extracted from 1 ml of the culture and fragmented by a brief digestion with RQ1 RNase-Free DNase I. The 3' end of the digested DNA was labelled with biotin, hybridized to FDA-ECID array and incubated for 16 h at 45˚C. The presence or absence of each gene was determined using Robust MultiArray Averaging (RMA) method and MAS5.0 algorithm. Robust MultiArray Averaging (RMA) summarized probe set intensity data is provided in S1 Excel File. In addition to bovine and human O104:H7 strains, other O104 strains of different serotypes (H2, H4, H8, H12, H21, H23) from our culture collection were also included in the phylogenetic analysis, performed using FDA-ECID gene differences data from all the probesets. The tree was developed using Neighbor net algorithm using Neighbor joining method (SplitsTree4). Pearson correlation coefficients (r) were determined to compare the bovine O104:H7 and human (O104:H7, O104:H4, O104:H21) strains. Pair wise scatter plots were also generated to visualize the gene level differences between the strains. Hierarchical cluster dendrogram was generated by array probe set differences for 241 strains including the diarrheagenic E. coli strains from our culture collection representing the broad diversity across the E. coli species.

Phylogenetic relationship
To gain perspective on the importance of O104 with regard to pathogenicity potential, we analyzed its relationship to the observed diversity in the E. coli species using the pan genome array. E. coli segregate into six observable phylogroups (A, B1, B2, CL, D, E, and F) as reported previously [14]. All strains analyzed in this study belonged to the B1 phylogroup ( Fig 3A). However, O104:H4 strains of the German and Georgian outbreaks of 2011 and 2009 [15], respectively (Fig 3A noted with asterisks), are distant from either the clinical or bovine O104: H7 strains (denoted in red and green boxes, respectively). The most closely related strain at a pan genome level, to four out of five clinical strains in this study was the O104:H21 strain isolated from bovine milk samples in Montana. The other clinical strain clustered with the bovine strains and this collective group was most closely related to a stx1 and 2 positive O146:H21 diarrheagenic E. coli strain.
As a corollary, we compared all O104 strains in our collection belonging to different serotypes independently to measure genomic networking and presumed recombination (Fig 2) which separated into 8 general groups. Similar to the hierarchical clustering method, one clinical O104:H7 strain (2011C-3665) clustered with the bovine O104:H7 STEC strains showing some parallel recombinatory paths. Likewise, O104 Georgian and German strains were distinctly separate groups. Regardless, given the instances of virulence markers noted above  (Table 1) and observed relationships to known pathogens, O104:H7 is an important serotype to monitor for significant human pathogenicity potential.

Discussion
Several molecular methods are available for the genetic characterization of foodborne pathogens, including Shiga toxin-producing E. coli. Molecular techniques, such as DNA microarray and whole genome sequencing (WGS), allow genome-wide characterization of the organisms. The limitation of the microarray is that it allows the detection of only specific gene targets, unlike WGS, which enables the detection of novel genes in the target organisms. However, WGS is time-consuming and laborious compared to microarray [14,16]. Custom DNA microarray detecting multitude of genes, designed using published whole genome sequences of E. coli and Shigella, enable rapid genome-scale analysis of pathogenic E. coli [14,[17][18][19].
All the bovine O104:H7 strains investigated in this study were negative for stx2, which is more frequently associated with human illnesses, particularly hemolytic uremic syndrome, than stx1 [20]. The bovine O104:H7 strains and one of the five human O104:H7 strain were positive for stx1 of subtype 1c. Other STEC serogroups carrying stx1c have been associated with diarrhea and in a few cases serious complications in human patients [21]. For example, STEC O78:Hcarrying stx1c was reported to be the causative agent of HUS in a 2-week old boy in Finland [22]. Only human clinical O104:H7 (3/5) and both human outbreak strains (O104:H4 and O104:H21) carried stx2. However, stx2 subtype was not identified in our study, because the array cannot differentiate the stx2 subtypes [14]. All O104:H7 strains, both bovine and human, similar to O104:H4 and O104:H21 outbreak strains [9,11,23], were negative for intimin (eae), an adhesin involved in the attachment to the host cells [24]. However, O104:H7 carried other adhesins, such as IrgA homologue adhesion (iha), mannose-specific adhesin (fimH), and long polar fimbriae (lpfA), which are involved in adhesion of bacteria to the host epithelial cells [25][26][27]. These adhesins may be involved in the attachment of the bacteria to the host cells in the absence of intimin. Long polar fimbriae (lpfA), which is also present in German outbreak strain (O104:H4), has been reported to be involved in initial attachment and colonization of the intestine [26]. Enterohemolysin (ehxA) was present in four of the bovine strains and in all human O104:H7 strains, except one, and in Montana outbreak strain (O104:H21). Previous studies have reported the association of ehxA with diarrhea and hemolytic uremic syndrome [28]. Cheng et al. (2015) have reported that hemolysin encoded by ehxA in E. coli O157:H7 is involved in cytotoxicity of macrophages and release of IL-1β in humans [29]. All bovine O104: H7 strains and the German outbreak strain were positive for terD, which codes for tellurite resistance protein. Shiga toxin-producing E. coli serogroups such as O26, O45, O111 and O157 carry tellurite resistance gene [30]. Tellurite resistance gene has been reported to be involved in the survival of bacteria by conferring protection against host defense [31].
Although there was some similarity in the virulence gene profile of human and bovine O104 strains, the key virulence genes, such as stx2 responsible for more serious infections, including hemolytic uremic syndrome, was absent in bovine strains, but present in human clinical and outbreak strains. Escherichia coli O104:H4 responsible for food-borne outbreak in Germany has been reported to be evolved by the uptake of stx2 encoding bacteriophages [32]. Similarly, bovine O104:H7 strains could evolve into highly pathogenic strains by acquiring other virulence genes responsible for human illnesses by gain and loss of genomic islands, prophages, and plasmids. All the strains investigated in the present study belong to B1 phylogroup. Most E. coli strains isolated from beef cattle carrying virulence genes (hlyA, stx1, stx2, eae) belonged to this phylogroup [33]. Phylogenetic analysis of STEC strains has revealed that majority (70%) of them belonged to B1 phylogroup [34].
Based on the scatterplots, splits tree, cluster dendrogram and Pearson correlation coefficients; all the bovine O104:H7 strains were closely related to each other, and one of the five human clinical strains (O104:H7; 2011C-3665) included in the study was also more closely related with the bovine strains. The close genetic relatedness of the human clinical strain (O104:H7; 2011C-3665) and bovine O104:H7 strains is further explained by fewer probe set differences (3,344 to 4,024) between them, they also showed similarity in their virulence gene profile. These findings indicate that the human clinical strain (2011C-3665) and the bovine strains most likely shared similar environment. This suggests that cattle feces could potentially be a source of O104:H7 strains associated with diarrhea in humans. Additionally, H-type appears to be more important than O-type as an attribute in relating all of the clinical and bovine strains included in this study based on the phylogenetic analysis. Similar finding was reported on O104 STEC strains based on pulsed-field gel electrophoresis [13]. Ju et al. (2012) hypothesized that E. coli serogroups carrying same flagellar antigens share common ancestors based on phylogenetic analysis of whole genome sequences of non-O157 E. coli serogroups [35]. We compared bovine and human O104:H7 strains to human outbreak strains of O104 belonging to different serotypes (O104:H4 and O104:H21) to determine similarities. However, our study revealed that the outbreak strains are distantly related to bovine and human O104: H7 strains, except for a few virulence genes common among all the strains.
In conclusion, microarray-based analyses of genes of bovine O104:H7 strains isolated from cattle feces suggest the potential of the serotype to be a diarrheagenic human pathogen. Because cattle harbor these strains and shed them in their feces, O104:H7 strains also have the potential to be foodborne pathogens to humans.