From 1996 through 2006, 524 EHEC were isolated as the only pathogens from fecal samples of epidemiologically unrelated patients with HUS (1 strain per patient). The isolation was achieved by using previously described procedures (7). The isolates were confirmed as E. coli by API 20 E (bioMérieux, Marcy l’Etoile, France) and serotyped (8) by using antisera against E. coli O antigens 1–181 and H antigens 1–56. In all nonmotile isolates from serogroups O26, O103, O111, O145, and O157, fliC genes were genotyped (9,10). MLST was performed as described previously (6) with small modifications (11). Phylogenetic analyses were based on allelic data that used the BURST algorithm (12) to achieve a more robust interpretation of the clustering and to reduce the influences by the effects of the recombination, which are widespread in E. coli (6). In addition, the stringent definition of clonal complexes (CCs), with which strains sharing at least 6 identical alleles are grouped into the same CC, was applied. The minimum spanning tree was generated from the allelic profiles by using Shigella dysenteriae strain M1354 (ST243, by using data from http://web.mpiib-berlin.mpg.de/mlst/dbs/Ecoli) as outgroup ( Appendix Figure).

Among 524 EHEC isolated from HUS patients, 355 (67.7%) belonged to serotypes O157:H7/H^– (249 were non–sorbitol-fermenting EHEC O157:H7/H^– and 106 were sorbitol-fermenting EHEC O157:H^–), and 169 (32.3%) belonged to 34 non-O157 serotypes. Because the phylogeny of E. coli O157:H7/ H^- has been extensively studied and is well established (5,13), we focused on MLST analysis of the 169 non-O157 strains, which represent all non-O157 EHEC serotypes from German HUS patients during the study period. We performed MLST analysis of only a subset of 10 O157 strains as controls.

MLST analysis of 169 non-O157 EHEC isolates distinguished 29 different sequence types (STs), which clustered into 10 CCs and 12 singletons (Table 1). The predominant ST was ST21, which consisted of 43 isolates (25.4% of non-O157 EHEC), followed by ST29 (30 isolates, 17.8%), ST32 (30 isolates, 17.8%), ST17 (15 isolates, 8.9%), and ST16 (14 isolates, 8.3%) (Table 1). These 5 STs included 78.1% of all HUS-associated non-O157 EHEC. The remaining 21.9% (n = 37) of the non-O157 EHEC strains belonged to 24 other STs that comprised only 1–3 strains; 14 of these STs were found only once (Table 1). Among the 10 CCs identified, CC29 was the most frequent. It comprised 89 strains of 5 STs, corresponding to 60.1% of 148 strains that could be assigned to a CC and to 52.7% of all non-O157 EHEC (Table 1). After CC29, CC32 and CC20 were the most frequently identified CCs (Table 1).

The predominant serotypes identified among the 169 non–O157 HUS-associated EHEC were O26:H11/H– (n = 72; 42.6%), O145:H28/H– (n = 32; 18.9%), O111:H8/H– (n = 14; 8.3%), and O103:H2/H^– (n = 14; 8.3%). The nonmotile strains within these serogroups shared the H antigen–encoding fliC gene with the motile strains that expressed the respective H antigen (Table 1). These 8 serotypes together constituted 132 (78.1%) of the non-O157 EHEC associated with HUS, whereas the other 37 strains (21.9%) belonged to 26 different serotypes, 17 of which contained only a single isolate (Table 1).

The most frequent serotypes including O26:H11/H–, O103:H2/H–, O111:H8/H–, and O145:H28/H– clustered into the 5 most prevalent STs (Table 1). However, not all isolates of the same serotype always belonged to the same ST (Table 1). One example is serotype O26:H11/H^– (fliC_H11), which was the most common non–O157 EHEC associated with HUS and clustered into 4 STs as single-locus variants (Table 1). Each of four O rough (OR) strains (2 OR:H11, and 1 each OR:H2 and OR:H^–), none of which could be successfully serotyped, was matched by its ST to an O typeable strain, indicating a recent conversion from the smooth to the rough strain form.

The relationships among members of the different STs and CCs are demonstrated in the Appendix Figure. Within the serogroup O111, 14 isolates belonging to serotypes O111:H8 and O111:H^– (fliC_H8) were ST16 (CC29). In contrast, the EHEC O111:H10 isolate with ST43 (CC10) shared none of the 7 MLST loci with the O111:H8/H^– strains, indicating that EHEC O111 causing HUS originate from 2 different clonal sources. Similar differences were observed between EHEC O145:H25 (ST342)/O145:H^– (fliC_H25) (ST659) and O145:H28 (ST32). Whereas ST659 is a single-locus variant of ST342, both allelic profiles differ in all loci from ST32.

The combination of MLST analysis and serotyping enabled us to establish the HUSEC collection. This collection comprises 41 EHEC isolated from HUS patients in Germany, which includes all 36 EHEC serotypes (O157 and non-O157) isolated from HUS patients and all 31 STs identified within these serotypes (Table 2). The strains included in this HUSEC collection were reserotyped and characterized for their stx genotypes and the presence of the eae gene (Table 2). Phenotypic characteristics and additional properties such as putative virulence determinants are available at www.EHEC.org.

Most (81.1%) of the non-O157 EHEC clustered into 3 CCs and belonged to a limited number of serotypes. These strains were recovered independently from different regions in Germany over an 11-year period. For the remaining strains, epidemiologic support is not as strong, and the clonal analysis demonstrated that their chromosomal backgrounds are highly divergent from those of CC29, CC32, and CC20. In 14 STs, we have only 1 isolate. In these cases, excluding concurrent or recent infection by E. coli O157 serologically is even more important. This exclusion was not always possible because patients’ serum for the investigation of immunoglobulin M anti-O157 lipopolysaccharide antibodies is frequently not available. However, at least some of these strains might represent emerging clones in the human population, such as O111:H10 (10), O113:H21 (14), and O121:H19 (15). Thus, strains of these serotypes included in our HUSEC collection can be used in future studies as a reference to compare EHEC isolated in other countries from HUS patients. This would allow timely discovery of the emergence of new non-O157 clones associated with HUS and the virulence traits that they contain (www.ehec.org).