Sequence Variations in the Flagellar Antigen Genes fliC H25 and fliC H28 of Escherichia coli and Their Use in Identification and Characterization of Enterohemorrhagic E. coli (EHEC) O145:H25 and O145:H28

Enterohemorrhagic E. coli (EHEC) serogroup O145 is regarded as one of the major EHEC serogroups involved in severe infections in humans. EHEC O145 encompasses motile and non-motile strains of serotypes O145:H25 and O145:H28. Sequencing the fliC-genes associated with the flagellar antigens H25 and H28 revealed the genetic diversity of the fliC H25 and fliC H28 gene sequences in E. coli. Based on allele discrimination of these fliC-genes real-time PCR tests were designed for identification of EHEC O145:H25 and O145:H28. The fliC H25 genes present in O145:H25 were found to be very similar to those present in E. coli serogroups O2, O100, O165, O172 and O177 pointing to their common evolution but were different from fliC H25 genes of a multiple number of other E. coli serotypes. In a similar way, EHEC O145:H28 harbor a characteristic fliC H28 allele which, apart from EHEC O145:H28, was only found in enteropathogenic (EPEC) O28:H28 strains that shared some common traits with EHEC O145:H28. The real time PCR-assays targeting these fliC H25[O145] and fliC H28[O145] alleles allow better characterization of EHEC O145:H25 and EHEC O145:H28. Evaluation of these PCR assays in spiked ready-to eat salad samples resulted in specific detection of both types of EHEC O145 strains even when low spiking levels of 1–10 cfu/g were used. Furthermore these PCR assays allowed identification of non-motile E. coli strains which are serologically not typable for their H-antigens. The combined use of O-antigen genotyping (O145wzy) and detection of the respective fliC H25[O145] and fliC H28[O145] allele types contributes to improve identification and molecular serotyping of E. coli O145 isolates.


Introduction
The ability to produce Shiga (Vero) toxins (Stx) was found to be associated with more than 472 different serotypes (O:H types) of Escherichia coli [1]. Many of these Shiga toxin-producing E. coli (STEC) strains are part of the intestinal flora of domestic and wildlife animals and can thus be found in the environment and as contaminants of food. Humans can get infected with STEC by contact with excreting animals or humans, a polluted environment and ingestion of contaminated food [2]. STEC infections in humans can cause diarrheal disease but only a few number of STEC types may cause more severe illness such as hemorrhagic colitis (HC) and hemolytic uremic syndrome (HUS). These latter types of STEC are also designated as enterohemorrhagic E. coli (EHEC) [3].
EHEC were defined on the basis of the severe clinical picture of the disease they cause, their frequency in outbreaks of disease and the presence of the eae (intimin) gene encoded by the LEE (locus of enterocyte effacement), and of non-LEE effector genes located on different genomic islands in the strains [3][4][5]. At present, five EHEC ("top-five") serotypes encompassing motile and non-motile strains of O26:H11, O103:H2, O111:H8, O145:H25/H28, and O157:H7 are regarded as most important for public health in the European Union (EU) [6] and in the United States [1][2]. In addition, two other EHEC serotypes, (O45:H2 and O121:H19) were associated with severe disease in humans and therefore included in the panel of EHEC strains ("top-seven") searched routinely in meat products in the United States [2].
Real-time PCR methods have been developed for specific detection of strains belonging to the most important EHEC types and a number of such methods were described for specific detection of EHEC O145 strains [7][8][9][10][11][12][13][14]. Harmonized real-time PCR procedures developed for the countries of the EU [15] and the U.S. [16] operate in a cascade-like process where first the presence of both stx and eae genes in the samples is searched. If positive, the presence of genes encoding the O-antigen of the suspected EHEC O-serogroup is investigated. Both procedures became effective in 2012 [2].
Samples from clinical, environmental and food origin frequently contain mixtures of bacteria which may hamper the detection and isolation of suspected EHEC contaminants. Quite often, feces from ruminant animals carry mixtures of Stx-negative E. coli strains carrying the eae-gene (so called atypical EPEC) together with eae-negative STEC strains [17]. In case of fecal contamination, food may contain both, atypical EPEC and STEC strains, thus falsely indicating the presence of EHEC when analyzed for the presence of stxand eae-genes. Such samples would be subsequently investigated for the presence of one of the seven EHEC O-groups for confirmation [15][16]. However, a positive result for the O-antigen may also be misleading, as many non-EHEC strains occurring in food and clinical samples share the same O-antigen with the top seven EHEC strains [14]. In order to reduce the number of false-positive results in the initial screening of samples for EHEC, the inclusion of more EHEC-specific markers is desirable.
Investigation of flagellar (fliC) genes was previously shown to be helpful for detection of EHEC strains with serological cross-reacting H-types [18] and for further characterization of non-motile EHEC strains [19][20][21][22]. The fliC genes of almost all flagellar antigens in E. coli have been sequenced [23] and sequence diversity of fliC genes was used for identification of EHEC strains such as O157:H7 [24]. Sequence diversity of fliC genes was found among E. coli strains expressing different O-antigens but sharing the same flagellar antigenic type such as H6 and H7 strains [24][25]. fliC sequence data derived from different E. coli H7 strains were used to develop PCR methods for specific detection of the H7 antigen encoded by E. coli O55 and O157 strains [24].
In this work we describe real-time PCR methods for identification of the fliC-genes encoded by EHEC O145:H25 and O145:H28 strains. The combined use of O-antigen genotyping (O145wzy) and detection of the respective O145-fliC types H25 and H28 improves the identification and molecular O:H serotyping of EHEC O145.

Materials and Methods Bacteria
The isolates were provided from the collections of the National Reference Laboratory for E. coli (NRL E. coli) at the Federal Institute for Risk Assessment (BfR) in Berlin, Germany and from the French Agency for Food, Environmental and Occupational Health and Safety (Anses) in Maisons-Alfort, France. E. coli strains used for the experiments were previously described for their serotypes and their virulence genes [4,[26][27][28][29][30]. The reference strains belonging to E. coli O-serogroups O1-O181 [30][31] and sixty-six serologically confirmed O145 strains carrying different H-antigens [4,[26][27][28][29][30] were used to test the specificity of O145wzy PCR. E. coli strains expressing the flagellar antigens H25 or H28 were detected by H-serotyping and by nucleotide sequencing of PCR amplified fliC products as previously described [19]. Relevant characteristics of E. coli strains used for nucleotide sequence analysis of their fliC genes are listed in Table 1.

PCR detection and mapping of E. coli O-antigen and H-antigen genes
Mapping of fliC gene variants to their respective H-types was performed by PCR and analysis of restriction fragment length polymorphism of HhaI digested PCR-products (PCR/RFLP) as previously described [19,32]. Nucleotide sequence data obtained from different fliC H25 and fliC H28 genes were used for designing TaqMan PCR probes and primers for common detection of all genetic variants of fliC H25 and fliC H28 genes (Table 2 of this work, [4]) and for identification of O145-specific fliC H25 (fliC H25[O145] ) and O145-specific fliC H28 (fliC H28[O145] ) genes ( Table 2). The real-time PCR assay specific for the ihp1 gene of EHEC O145:H28 was described previously [26,33]. A TaqMan PCR probe and specific primers covering all types of E. coli O145 strains was deduced from the O145wzy (O-antigen polymerase) gene (Table 2). Realtime PCR reactions were set up as singleplex assays. TaqMan PCR probes and primers used in this work were designed with the software Primer Express V3.0 (Applied Biosystems) and are described in Table 2. The position of primers and gene probes for specific detection of O145

Nucleotide sequencing
The nucleotide sequence of the PCR products were determined by Sanger sequencing [34] and analyzed by the use of the Accelrys DS Gene software package (Accelrys Inc., USA). The nucleotide sequences of the respective products for fliC homologs were determined and have been submitted to European Nucleotide Archive (ENA). Accession numbers are listed in Table 1.

Results
Development of an E. coli O145 serogroup specific real-time PCR-assay The O145wzy PCR has been tested for its specificity on reference strains belonging to E. coli O-serogroups O1-O186 and was found to react only with the O145 reference strain (E1385, O145:NM) [30]. Also, the O145wzy PCR was positively tested on a collection of sixty-six serologically confirmed O145 strains carrying different H-antigens (Table 3). As the O145wzy PCR proved to be specific it was used to confirm results obtained by O-serotyping and for screening for the presence of O145 O-antigen genes in serologically Orough:H25 and Orough:H28 strains (see below). Characteristic nucleotide sequences associated with fliC genes of EHEC O145:H25 and O145:H28 strains The fliC sequences obtained from E. coli O145:H25 and O145:H28 strains revealed characteristic differences when compared to non-O145:H25 and non-O145:H28 serotypes. For further examination, the complete fliC nucleotide sequences of both EHEC O145:H25 and EHEC O145: H28 strains were determined. The H25 flagellin encoding genes derived from two E. coli O145:H25 strains CB12641 (GenBank Accession LN555738) and CB12671 (GenBank Accession LN555739) had each a length of 1332 bp and were identical between both strains. BLAST search revealed 100% and 99% sequence identity with partial fliC sequences of O145 strains 4392/97 (GenBank Accession AJ566340.1) and 2839/98 (GenBank Accession AJ566341.1) respectively [22] and 98% identity with the partial fliC H25 sequence (GenBank Accession AY250007) of the O15:H25 type reference strain HW26 [23]. Nucleotide sequences identical to those obtained from CB12641 and CB12671 were found in whole genome sequencing (WGS)-contigs deposited at GenBank (Accession numbers JHMZ01000105.1, JHHD01000016.1, JASO01000012.1, AIAX01000332.1 and AIAQ01000055.1). Among the WGS-contigs deposited at GenBank two are of unknown serotype, one is O145:H25 and one is O177:H25.
The differences found in the fliC H25 and fliC H28 genes of EHEC O145 strains compared to non-O145 strains carrying the same flagellar antigens prompted us to develop real-time PCR assays to improve the identification of these clinically important EHEC O145 types.   Table 2). The fliC sequences derived from these fliC H25[O145] gene PCR-reacting strains were compared to those of representative non-reacting strains such as N234 (O15:H25) and CB11499 (O103: H25). A cluster analysis of the fliC H25 coding sequences is presented in Fig 3. A cluster of genetically closely related strains is formed by fliC H25[O145] gene PCR reacting strains (O2, O100, O103, O145, O165, O172 and O177). The fliC H25 sequences of PCR non-reacting strains N234 and CB11499 were found genetically more distant (Fig 3).
In detail, the fliC sequence of the E. coli O177:H25 strain was identical with those of investigated EHEC O145:H25 strains. The fliC sequences of the O165:H25 and the O172:H25 strains differed from O145:H25 by one nucleotide exchange at the same position within the sequence (C/T, position 1134). The fliC sequence of the O100:H25 strain showed one nucleotide exchange at another position (G/A, position 110). These nucleotide exchanges had no effect on the derived amino-acid (aa) sequence (S1 Fig). On the other hand, changes in the amino acid composition of the H25 flagellins compared to that of O145:H25 were observed for the genetically more distant strains O2:H25 (4 aa), O15:H25 (7 aa) and O103:H25 (8aa) (S1 Fig). The specificity of the fliC H25[O145] PCR assay was further tested on a panel of strains including reference strains encompassing H-types H1 to H56. None of these strains gave a positive result with the fliC H25[O145] PCR (data not shown).

Development of a fliC real-time PCR for detection of EHEC O145:H28 strains
The real-time PCR developed from the fliC H28[O145] sequence was tested on 70 strains identified as H28 positive by serotyping or nucleotide sequence typing of fliC genes and by a generic fliC H28 PCR targeting common regions of the fliC H28 genes. The strains belonged to 14 different O groups, as well as to ONT and Orough strains. The results are summarized in Table 5. The fliC H28[O145] PCR reacted with all 35 O145:H28 and two Orough:H28 strains which were positive for the O145wzy gene. Except for four O28:H28 strains, no other strain expressing flagellar type H28 reacted in the fliC H28[O145] PCR. In order to explore the nature of the observed reaction with O28:H28 strains we determined the nucleotide sequence of the fliC gene of a representative strain (CB9651, O28:H28, GenBank LN649615). The sequence had the same length and was identical to that found with O145:H28 strains.
A cluster analysis of fliC H28 sequences derived from fliC H28[O145] PCR reacting (E. coli O145:H28 and O28:H28) and non-reacting (E. coli O132:H28 and O152:H28) strains revealed two genetic clusters of strains that were distinguishable by the fliC H28[O145] PCR (Fig 4). None of the E. coli H-type reference strains H1-H56 reacted in the fliC H28[O145] PCR except for strain 5306-56 (O26:H46) which gave a positive reaction with a cycle threshold (CT) value of 21.3. The reaction was supposed to be fliC specific as another H46 strain from our collection (CB13742, Or:H46) reacted equally well (CT value of 21.9) in the PCR. By comparing the corresponding regions in the EHEC fliC H28[O145] genes and the fliC H46 gene of strain 5306-56 (GenBank AY250024, 1719bp), only small differences were found (one mismatch in the backward primer and one mismatch in the probe sequence). The sequence similarity in the target region could explain the PCR cross-reaction with flagellar type H46 strains. Comparison of the fliC gene sequences of five E. coli H28 strains belonging to serogroups O145 (CB12513 and   Orough:H28 9 8 0 a) All 70 strains reacted positive with the generic fliC H28 PCR described in Table 2  fliC Genes of EHEC O145:H25 and O145:H28 RM12581), O28 (CB9651), O132 (N87) and O152 (SE11) with that of strain 5306-56 (O26: H46) revealed that the latter is genetically more distant from all these (Fig 4).
Evaluation of the O145:H25 and EHEC O145:H28 real-time PCR detection systems on artificially contaminated mixed salad samples Two random 25-g portions of the salad used for spiking experiments were analyzed for their bacterial microflora. The total mesophilic counts on Standard I agar (25°C) were 2.06-3.26 x 10 6 cfu/g. Lactose-negative and lactose-positive Enterobacteriaceae were detected in quantities of 9.0 x 10 4 -1.89x10 5 cfu/g after growth on VRB agar for 22h at 37°C. Portions of salad (25g) were spiked with high (100-1000 cfu/g) and low (1-10 cfu/g) quantities of the E. coli strains described in Table 6

Discussion
Shiga-toxin producing E. coli O145 are known as the causative agents of severe illness such as hemorrhagic colitis (HC) and hemolytic uremic syndrome (HUS) worldwide. EHEC O145 encompass motile and non-motile strains of serotypes O145:H25 and O145:H28. While EHEC O145:H28 strains have been described for many years, EHEC O145:H25 were identified more recently as a second important type among serogroup EHEC O145 strains and were isolated from human patients [22,[39][40].
A number of different real-time PCR assays were developed for detecting EHEC O145 strains. Some of these methods are based on the presence of the ihp1 gene [8,9,14,26,41] which is specific for O145:H28 strains [26]. However, the O145 ihp1 PCR-assay does not react with O145 strains carrying other H-antigens [26] including EHEC O145:H25 strains (Table 3, this work). Other real-time PCR methods have used E. coli O145 O-antigen specific wzx or wzy sequences as targets for specific detection of these strains ( [7,10,12], this study). In contrast to the ihp1-based PCR which detects only O145:H28 and shows cross-reactions with some strains belonging to other O-serogroups [26], the O145 wzy and O145 wzx PCR assays are potentially suitable to detect all variants of serogroup O145 strains including both EHEC types O145:H25 and O145:H28. This was demonstrated using the O145 wzy PCR, which was specific, giving no cross reactions with strains belonging to O-groups other than O145 (this work). Many of the so far described real-time PCR assays use combinations of different genes for a more specific detection of EHEC O145 and other EHEC strains. For example, assays targeting both eae-γ1 (intimin) and fliC H28 sequences were developed for specific detection of EHEC O145:H28 strains [13,14]. However, these assays are not suitable for specific detection of O145:H25 strains which carry fliC H25 and the eae-β genes [22].
In this study we have employed the sequence diversity of fliC H25 and fliC H28 genes for developing diagnostic assays for detection of both EHEC O145:H25 and EHEC O145:H28 strains. Molecular typing of fliC genes allows further characterization of non-motile (NM) E. coli strains which are serologically not-typable for their H-antigens [18][19][21][22]. This was found to be important for characterization of EHEC O145:H28 strains, which were represented by a high number of non-motile variants in our study (Tables 4 and 5, this work).
FliC genotyping was also found useful to discriminate clearly between serologically cross-reacting flagellar antigens H8 and H40 and genetic subtypes of flagellar antigen H8 were identified and associated with EHEC O111 and STEC strains belonging to other O-groups [18]. Analysis of the fliC genes encoding the H7 antigen in E. coli strains belonging to different Ogroups revealed ten different alleles and PCRs specific for the fliC gene carried by EHEC O157: H7 and O157:HNM were developed for DNA-based typing [24][25]. Isolates of classical EHEC strains belonging to serogroups O26, O111, O145 and O157 are frequently non-motile and their H-type can thus only be determined by molecular characterization [19,22,25,[42][43][44]. Moreover, EHEC belonging to serogroup O145 were found to split into two serotypes, O145:H25 and O145:H28 that possess different alleles of the eae-gene [22]. Flagellar-types H25 and H28 are widely spread among E. coli strains belonging to different Ogroups and pathotypes ( [30,40,44], this work).
In order to better characterize EHEC O145:H25 and EHEC O145:H28 strains we have developed real time PCR-assays targeting the fliC alleles present in these strains. The fliC H25 genes present in O145:H25 were found very similar to those of some other STEC and EPEC strains such as O2, O100, O165, O172 and O177 pointing to their common evolution but were different from fliC H25 genes of a multiple number of E. coli serotypes including STEC, EPEC and apathogenic strains. The high similarity between the EHEC O145 and E. coli O2, O100, O165, O172 and O177 fliC H25 genes (between 100% and 98.87% identity) prevents the design of a real-time PCR assay that would allow a complete discrimination.
Evaluation of the specific real-time PCR assays for EHEC O145:H25 and O145:H28 strains with artificially contaminated ready-to eat salad samples resulted in specific detection of both types of EHEC O145 strains even when inoculated in low quantities (1-10 CFU/g) after overnight enrichment. The presence of a dense natural microflora (> 10 6 CFU/g salad) did not affect detection of the inoculated EHEC strains.
To our knowledge, the genetic diversity of fliC H28 and fliC H25 has not been previously studied in E. coli. The fliC H28 and fliC H25 alleles described in this work contribute to a better characterization of the flagellar antigens of E. coli. Also, the combination of a new O145 wzy PCR with the fliC H25[O145] and fliC H28[O145] PCR assays improves the identification and characterization of EHEC O145 strains.