Atypical enteropathogenic Escherichia coli (aEPEC) are one of the most frequent intestinal E. coli pathotypes isolated from diarrheal patients in Brazil. Isolates of aEPEC contain the locus of enterocyte effacement, but lack the genes of the bundle-forming pilus of typical EPEC, and the Shiga toxin of enterohemorrhagic E. coli (EHEC). The objective of this study was to evaluate the phylogeny and the gene content of Brazilian aEPEC genomes compared to a global aEPEC collection.
Single nucleotide polymorphism (SNP)-based phylogenomic analysis was used to compare 106 sequenced Brazilian aEPEC with 221 aEPEC obtained from other geographic origins. Additionally, Large-Scale BLAST Score Ratio was used to determine the shared versus unique gene content of the aEPEC studied.
Phylogenomic analysis demonstrated the 106 Brazilian aEPEC were present in phylogroups B1 (47.2%, 50/106), B2 (23.6%, 25/106), A (22.6%, 24/106), and E (6.6%, 7/106). Identification of EPEC and EHEC phylogenomic lineages demonstrated that 42.5% (45/106) of the Brazilian aEPEC were in four of the previously defined lineages: EPEC10 (17.9%, 19/106), EPEC9 (10.4%, 11/106), EHEC2 (7.5%, 8/106) and EPEC7 (6.6%, 7/106). Interestingly, an additional 28.3% (30/106) of the Brazilian aEPEC were identified in five novel lineages: EPEC11 (14.2%, 15/106), EPEC12 (4.7%, 5/106), EPEC13 (1.9%, 2/106), EPEC14 (5.7%, 6/106) and EPEC15 (1.9%, 2/106). We identified 246 genes that were more frequent among the aEPEC isolates from Brazil compared to the global aEPEC collection, including espG2, espT and espC (P<0.001). Moreover, the nleF gene was more frequently identified among Brazilian aEPEC isolates obtained from diarrheagenic patients when compared to healthy subjects (69.7% vs 41.2%, P<0.05).
The current study demonstrates significant genomic diversity among aEPEC from Brazil, with the identification of Brazilian aEPEC isolates to five novel EPEC lineages. The greater prevalence of some virulence genes among Brazilian aEPEC genomes could be important to the specific virulence strategies used by aEPEC in Brazil to cause diarrheal disease.
Atypical EPEC (aEPEC) is one of the most frequent diarrheagenic Escherichia coli pathotypes isolated from patients in Brazil and is associated with diarrheal outbreaks. This study is the first to sequence the genomes of a collection of aEPEC isolates from a South American country, Brazil, and compare their phylogenetic relationships and gene content with a global collection of aEPEC. This approach identified Brazilian aEPEC genomes in previously characterized EPEC/EHEC phylogenomic lineages and resulted in the identification of five novel EPEC phylogenomic lineages, designated EPEC11 to EPEC15. We also observed that virulence genes, such as espG2, espT and espC were more frequently identified among the Brazilian aEPEC genomes, demonstrating potential differences in the virulence repertoire of this pathogen in Brazil.
Citation: Hernandes RT, Hazen TH, dos Santos LF, Richter TKS, Michalski JM, Rasko DA (2020) Comparative genomic analysis provides insight into the phylogeny and virulence of atypical enteropathogenic Escherichia coli strains from Brazil. PLoS Negl Trop Dis 14(6): e0008373. https://doi.org/10.1371/journal.pntd.0008373
Editor: Ruifu Yang, Beijing Institute of Microbiology and Epidemiology, CHINA
Received: November 6, 2019; Accepted: May 7, 2020; Published: June 1, 2020
Copyright: © 2020 Hernandes et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the manuscript and in its Supporting Information files.
Funding: This study was funded by federal funds from the National Institute of Allergy and Infectious Diseases, National Institutes of Health, Department of Health and Human Services (grant number U19AI110820 to DAR) and São Paulo Research Foundation - FAPESP (grant numbers: 2015/26207-6 and 2016/17584-3 to RTH). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
The enteropathogenic Escherichia coli (EPEC) pathotype is divided into two subgroups: typical EPEC (tEPEC) and atypical EPEC (aEPEC) . The main difference between these two groups is the presence of the EPEC adherence factor plasmid (pEAF) encoding the bundle forming pilus (BFP) in tEPEC, and its absence in aEPEC [1, 2]. A previous epidemiological study demonstrated that tEPEC was a significant cause of moderate to severe diarrhea and increased mortality in infants . In recent decades aEPEC has emerged as a frequent cause of diarrheal disease in children and adults, and has been linked to several diarrheal outbreaks [2, 4–12]. In particular, aEPEC isolates of serotypes O2:H16, O33:H34, O39:H9, O108:H- and ONT:H19 were implicated in five distinct diarrheal outbreaks in São Paulo State, Brazil, in 2013 .
The main virulence feature of EPEC is its ability to produce the histopathological attaching and effacing (A/E) lesions on the surface of infected host cells . The A/E lesion is characterized by the destruction of the microvilli brush border, intimate bacteria-cell adherence and formation of pedestal-like structures rich in F-actin and other cytoskeleton elements, which causes intestinal epithelial cells to have a reduced absorptive capacity [13, 14]. The 41 genes associated with this phenotype are located in a pathogenicity island (PAI) termed the locus of enterocyte effacement (LEE) . The LEE region encodes all the structural components of the type 3 secretion system (T3SS), as well as seven effector proteins, the intimin outer membrane protein, the intimin translocated receptor protein (Tir), chaperones, and transcriptional regulators [15–18]. Although enterohemorrhagic E. coli (EHEC) similarly contain the LEE and produce the A/E lesion on infected epithelial cells, EHEC also contain the stx genes, encoding the Shiga toxin, which is not present in EPEC . Besides the T3SS effectors located in the LEE region, effectors encoded outside of LEE, known as non-LEE encoded effectors, have emerged as important accessory virulence factors in A/E pathogens [19, 20]. These effectors are responsible for inducing several modifications in the host cells during infection, including inhibition of apoptosis, interference with inflammatory signaling pathways, invasion, tight junction disruption and phagocytosis [19, 20].
Previous comparative genomic studies have demonstrated that EPEC exhibit considerable genomic diversity, with representatives in at least 10 different phylogenomic lineages [21, 22]. This is most likely due to the acquisition of the LEE region and its stable retention by genomically diverse lineages of E. coli . These prior studies have focused on investigating the genomic diversity of EPEC isolates from locations in Europe, Africa, and Asia [22, 23]. While aEPEC poses a significant diarrheal burden in South America [2, 9, 24], to our knowledge there have been no previous comparative genomics studies that have investigated the genetic diversity of aEPEC from Brazil or other countries in South America. Thus, the goal of this study was to analyze the genetic relatedness and virulence gene content of Brazilian aEPEC isolates compared to a globally distributed collection of aEPEC.
DNA extraction, genome sequencing, and assembly
Genomic DNA of the 106 aEPEC isolates from Brazil (S1 Table) was extracted from overnight cultures using the Sigma GenElute bacterial genomic DNA kit (Sigma-Aldrich; St. Louis, MO). The genomes were sequenced using paired-end libraries on the Illumina HiSeq 4000. The 150 bp Illumina reads were assembled using SPAdes v.3.7.1 , and the final assemblies were filtered to contain contigs that were ≥500 bp in length and had ≥5X kmer coverage.
A single nucleotide polymorphism (SNP)-based phylogenomic analysis was performed on the 106 Brazilian aEPEC (S1 Table), 221 global aEPEC publicly available from previous studies [22, 23] (S2 Table, S1 Fig), and a collection of 48 diverse E. coli and Shigella genomes (S3 Table). A total of 179,202 SNPs were detected relative to the completed genome sequence of the phylogroup F laboratory isolate E. coli IAI39 (accession number: NC_011750) using the In Silico Genotyper (ISG) . A maximum-likelihood phylogeny with 100 bootstrap replicates was generated using RAxML v.7.2.8  with the GTR model of nucleotide substitution, the GAMMA model of rate heterogeneity, and visualized using FigTree v.1.4.2 (http://tree.bio.ed.ac.uk/software/figtree/).
In silico Multilocus Sequence Typing (MLST) and serotyping
The seven loci of the multilocus sequence typing (MLST) scheme (adk, gyrB, fumC, icd, mdh, purA, and recA) developed by Wirth et al.  were identified in each of the genomes listed in S1 and S2 Tables using BLASTN, and MLST sequence type (ST) were determined by querying the PubMLST database (https://pubmlst.org).
The molecular serotype of each genome analyzed was determined using Serotype Finder 1.1 (https://cge.cbs.dtu.dk/services/SerotypeFinder/) with the default settings of an 85% identity threshold and 60% minimum alignment length .
Large-Scale BLAST Score Ratio (LS-BSR) analysis
The 106 Brazilian aEPEC and additional A/E E. coli genomes (S1 to S3 Tables) analyzed in this study were compared by BLASTN LS-BSR analysis as previously described . Briefly, the predicted protein-encoding genes of each genome that had ≥90% nucleotide sequence identity and ≥90% alignment length were assigned to gene clusters using CD-HIT v.4.6.4  (S1 Data Set). The predicted protein function of each gene cluster was determined with an ergatis-based  in-house annotation pipeline . The presence of gene clusters with significant similarity in each of the five novel phylogenomic lineages identified in this study (EPEC11, EPEC12, EPEC13, EPEC14 and EPEC15) were defined by using an LS-BSR value ≥0.9; while the absence in the remaining A/E genomes were defined by using an LS-BSR value ≤0.4.
Additionally, we examined the occurrence of genes that were more common among the Brazilian aEPEC isolates when compared to the isolates from outside of Brazil, as well as among Brazilian and global aEPEC identified in the same phylogenomic lineage. The presence of each gene cluster was defined as having a BLASTN LS-BSR value ≥0.8, and differences in genes frequency among the two group of aEPEC studied (Brazilian versus Global) were tested by using the Chi-square test with Yates correlation and two-tailed Fisher's exact test.
In silico detection of virulence factor-encoding genes and intimin subtyping
TBLASTN LS-BSR was used to detect the presence of known EPEC virulence factor-encoding genes in the 106 Brazilian aEPEC genomes, as previously described [22, 34]. The virulence factor-encoding genes investigated included the repertoire of Non-LEE effectors present in the typical EPEC prototypes E2348/69  and B171 ; since many of them has been associated with the EPEC pathogenesis , as well as espT  and espV , first indentified and functionally characterized in the outbreak-associated aEPEC E110019. In addition, we searched for genes reponsible for encoding adhesins (efa1/lifA, iha, paa, toxB, cah, ehaA, ehaB, ehaC, ehaD, ehaG, and ehaJ) and toxins (astA, pic, pet ehxA, espP, sepA, sigA, and espC) that haven been frequentelly deteceded among aEPEC isolates, demontrated to contribute to aEPEC pathogenesis and/or previously associated with diarrheal disease [8, 39–42]. The intimin subtypes [43, 44] were identified in the 106 aEPEC genomes analyzed using BLASTN LS-BSR as described above.
Characteristics of the Brazilian aEPEC genomes
A total of 106 aEPEC isolates obtained from previous studies performed in Brazil [5, 8, 9, 45] were sequenced. This collection of Brazilian aEPEC were isolated over a span of seven years from 2009 to 2016, and were associated with multiple clinical presentations, including 17 isolates from subjects with asymptomatic infection, 76 from patients with diarrhea, and 13 from diarrheal patients investigated in five distinct diarrheal outbreaks (S1 Table). The average genome size of the Brazilian aEPEC genome assemblies was 4,927,030 bp (range 4,551,111 to 5,442,868), and the average GC content was 50.53% (range: 50.13% to 50.78%), which is consistent with previous studies of sequenced E. coli genomes [22, 46] (S1 Table).
The majority of the aEPEC Brazilian genomes were identified in phylogroup B1
The 106 Brazilian aEPEC genomes were identified in phylogroup B1 (47.2%, 50/106), followed by phylogroups B2 (23.6%, 25/106), A (22.6%, 24/106), and E (6.6%, 7/106) (Fig 1, Table 1). The 221 global aEPEC genomes analyzed had a similar phylogroup distribution with phylogroup B1 (55.7%, 123/221) being the most common, followed by phylogroups A (22.6%, 50/221) and B2 (11.3%, 25/221), while the aEPEC genomes of phylogroups E (9.5%, 21/221) and D (0.9%, 2/221) were the least common (Table 1). While the phylogroup distributions of Brazilian aEPEC were similar to those of the global aEPEC genomes, there were significantly more phylogroup B2 genomes identified among the Brazilian aEPEC (23.6%, 25/106) compared with the global collection (11.3%, 25/221) (Table 1) (P<0.05). The reasons for a greater prevalence of phylogroup B2 aEPEC in Brazil compared with the global collections are not clear at this time and may be representative of an isolation or sequencing bias.
Phylogenomic analysis of atypical EPEC genomes (labeled in blue), compared to typical EPEC (labeled in green), EHEC (labeled in red) and reference E. coli and Shigella genomes (labeled in black). Red circles indicate the 106 Brazilian aEPEC genomes. The clades in the phylogeny are colored by phylogenomic lineage, and the E. coli phylogroups are indicated by the letters: A, B1, B2, D, E, and F. An asterisk (*) indicates the five novel EPEC phylogenomic lineages identified in this study. The nodes with bootstrap values >80% are labeled with a black circle.
Brazilian aEPEC genomes were identified in previously described EPEC/EHEC phylogenomic lineages
Previous genetic studies focusing on understanding the evolution of EPEC have demonstrated that isolates from this pathotype can belong to at least ten distinct EPEC phylogenomic lineages, which have been designated EPEC1 to EPEC10 [21, 47, 48], in addition to unclassified EPEC isolates, that exist outside of these defined EPEC lineages. We observed that 59.4% (63/106) of the aEPEC Brazilian genomes could be classified into previously defined phylogenomic lineages: EPEC2 (1.9%, 2/106), EPEC4 (1.9%, 2/106), EPEC5 (2.8%, 3/106), EPEC6 (5.7%, 6/106), EPEC7 (6.6%, 7/106), EPEC9 (10.4%, 11/106), EPEC10 (17.9%, 19/106), as well as a small proportion in EHEC lineages, EHEC1 (4.7%, 5/106) and EHEC2 (7.5%, 8/106) (Fig 1, Table 1).
The greatest number of Brazilian aEPEC genomes were identified in the EPEC10 lineage (17.9%, 19/106). The majority of the Brazilian aEPEC in this lineage were MLST ST10 (78.9%, 15/19), including seven aEPEC isolates with serotype O2:H16 (36.8%, 7/19). We also observed a significantly greater number of Brazilian aEPEC genomes than global aEPEC (10.4% vs 0.9%, P<0.001) in the EPEC9 lineage (Table 1). In this lineage we observed eight Brazilian aEPEC of serotype O33:H34 (72.7%, 8/11). Of note, the Brazilian aEPEC genomes identified with serotypes (sequence types) O2:H16 (ST10) and O33:H34 (ST1951), were obtained from diarrheal outbreaks that occurred in Brazil in 2013 (4 isolates O2:H16 and 4 isolates O33:H34, each serotype from a distinct outbreak), as well as from sporadic cases of diarrhea that occurred in Brazil over several years (3 isolates O2:H16 and 4 isolates O33:H34) (S1 Table), while in the global aEPEC collection, only one isolate, from Mali (isolate 200135), was identified with serotype O33:H34 (ST2346); and aEPEC of serotype O2:H16 were not detected (S2 Table). The differences observed in the frequency of aEPEC of serotypes O2:H16 (6.6% vs 0.0%, P<0.001) and O33:H34 (7.5% vs 0.5%, P<0.001), among the Brazilian vs Global aEPEC collections, may be due to these serotypes being associated with Brazilian diarrheal outbreaks.
Brazilian aEPEC identified in the lineages EHEC1 and EHEC2 had the serotypes O55:H7 (100.0%, 5/5) and O26:H11 (75.0%, 6/8), respectively. The five O55:H7 were ST335 (Table 2) and were most closely related to the O157:H7 of the EHEC1 lineage (Fig 1). The six aEPEC with serotype O26:H11 in the EHEC2 lineage were one of two MLST sequence types: ST21 (50.0%, 3/6) or ST29 (50.0%, 3/6) (Table 2). The number of aEPEC genomes identified in the EHEC2 lineage was significantly higher among the global collection of aEPEC than the Brazilian aEPEC (7.5% vs 16.7%, P<0.05) (Table 1).
Designation of five novel EPEC phylogenomic lineages
Phylogenomic analysis of Brazilian aEPEC genomes along with aEPEC genomes from a global distribution, led to the identification of five novel phylogenomic lineages with five or more genomes in each, which we have designated EPEC11 to EPEC15 (Fig 1). The EPEC11 to EPEC14 lineages were identified in phylogroup B1, whereas EPEC15 was identified in phylogroup A (Fig 1). Among the aEPEC Brazilian genomes analyzed, we observed that 28.3% (30/106) were identified in one of the five novel EPEC lineages. The distribution of these isolates is as follow: EPEC11 (14.2%, 15/106), EPEC12 (4.7%, 5/106), EPEC13 (1.9%, 2/106), EPEC14 (5.7%, 6/106) and EPEC15 (1.9%, 2/106) (Table 1). It is important to highlight the number of Brazilian aEPEC genomes identified in the EPEC11 lineage (14.2%, 15/106) (Table 1), which are second in abundance after the Brazilian aEPEC isolates identified in the EPEC10 lineage (17.9%, 19/106) (Table 1). The Brazilian aEPEC identified in EPEC11 had eight distinct serotypes (O35:H19, O71:H19, O88:H25, O108:HNT, O123/O186:H19, O126:H19, O160:H19 and ONT:H19), harbored intimin subtype epsilon2 (73.3%, 11/15) or eta (26.7%, 4/15), and the majority (86.7%, 13/15) were ST517 (Tables 2 and 3). Moreover, we observed that the aEPEC archetype isolate E110019, from an outbreak that occurred in Finland , was present in the newly designated EPEC14 lineage (Fig 1), along with two Brazilian aEPEC isolates (IAL5152 and IAL 5153) which were serotype O39:H9 and obtained from an outbreak (S1 Table). It is important to emphasize that the number of Brazilian aEPEC genomes identified in EPEC14, a novel EPEC lineage, is statistically different when comparing with aEPEC from outside of Brazil (5.7% vs 0.9%, P<0.05) (Table 1). Additionally, 12.3% (13/106) of the Brazilian aEPEC genomes identified in phylogroups B1 (5 isolates), B2 (6 isolates) and E (2 isolates) were not classified into any of the described EPEC or EHEC phylogenomic lineages (Tables 1 and 2).
Identification of genes that are unique among the novel EPEC phylogenomic lineages
We used LS-BSR to identify gene content that is present in one group of isolates and lacking in another. First, we examined the existence of genes that were present in each of the five novel EPEC lineages (EPEC11, EPEC12, EPEC13, EPEC14 and EPEC15) and absent in all other A/E E. coli genomes included in this study. A total of 936 gene clusters were found exclusively, at a variable frequency, in the five novel EPEC phylogenomic lineages as follows: 196 (EPEC11), 181 (EPEC12), 241 (EPEC13), 140 (EPEC14), and 178 gene clusters (EPEC15) (Table 4). However, only one of these gene clusters was identified in all of the genomes within its associated lineage and absent from all other genomes analyzed. The EPEC11 lineage harbors a single gene cluster (termed CYCE01000006.1_37), that encodes a hypothetical protein, present in all 45 (100.0%) aEPEC genomes from this phylogenomic lineage and absent in all other A/E genomes analyzed (Table 4). A summary of the LS-BSR analyses can be found in Table 4, and all genes exclusively present in each of the five novel phylogenomic lineages are listed in S4 Table.
Differential prevalence of gene content between Brazilian aEPEC compared to a global collection of aEPEC
Since so few genes were exclusively present or absent in the novel phylogenomic lineages we wanted to examine if there were any genes that were more statistically prevalent among the Brazilian aEPEC isolates when compared to the isolates obtained from locations outside of Brazil. This analysis of the gene content led to the identification of 246 genes that were statistically more frequent (P<0.001) among the Brazilian aEPEC genomes than aEPEC from other geographic origins (Brazilian aEPEC vs Global aEPEC). Included among these genes were two T3SS-effectors: espG2, gene cluster 128_11_7_153 (20.8% vs 6.8%), and espT, gene clusters 148_12_48_29 (21.7% vs 3.6%) and CYEQ01000057.1_8 (17.9% vs 3.2%). The espC gene, encodes an autotransporter protein , represented by gene clusters 1024_15_1_31 (27.4% vs 7.2%) and 276_12_73_9 (26.4% vs 8.1%) was also significantly more prevalent among the Brazilian aEPEC than the global aEPEC (Table 5).
Conversely, 446 genes were statistically less frequent (P<0.001) among the Brazilian aEPEC when compared to the global aEPEC from outside Brazil, including genes encoding a type VI secretion ATPase (49.1% vs 69.7%), the T3SS-effectors nleA (36.8% vs 59.7%), ospB (24.5% vs 44.3%) and espW (7.5% vs 25.3%), and the IrgA homolog adhesin iha (6.6% vs 22.6%) (Table 5). Curiously, genes encoding proteins that mediate antimicrobial resistance were also statistically less frequent (P<0.001) among the Brazilian aEPEC genomes compared to the aEPEC genomes from outside Brazil, as follows: blaTEM-1 (17.0% vs 38.9%) encoding a TEM-1 beta-lactamase, sul2 (15.1% vs 40.7%) encoding a sulfonamide resistant dihydropteroate synthase, aph(6)-Id (11.3% vs 38.9%) encoding an aminoglycoside phosphotransferase, and tetA (7.5% vs 24.0%) encoding an antibiotic efflux pump (Table 5). Other genes that were differentially detected among Brazilian and global aEPEC have predicted functions associated with central metabolism and transcriptional regulation, as well as numerous hypothetical proteins (S5 Table).
Virulence genes identified among the Brazilian aEPEC that exhibit lineage specific distributions
We also investigated the differences in gene content between Brazilian and global aEPEC isolates identified in the same phylogenomic lineage. Statistically different prevalences (P<0.001) were observed in five phylogenomic lineages (EHEC1, EHEC2, EPEC7, EPEC10 and EPEC11) between gene content of the groups of aEPEC genomes analyzed (S6 Table). These included the toxB gene, which was first characterized in the pO157 plasmid of EHEC O157:H7 . Interestingly, toxB was detected in four Brazilian aEPEC (50.0%, 4/8) that were serotype O26:H11 and identified in the EHEC2 lineage, while the EHEC2 isolates analyzed from outside Brazil lack this gene (S6 Table). In addition, the autotransporter-encoding gene sepA (85.7% vs 5.9%) and the T3SS-effector gene espK (71.4% vs 0.0%) were statistically more prevalent (P<0.001) among the EPEC7 Brazilian aEPEC genomes than the EPEC7 aEPEC from outside Brazil (S6 Table).
We also observed that genes from the PAI OI-122 were statistically (P<0.001) different among Brazilian aEPEC compared to global aEPEC in two phylogenomic lineages: EPEC7 and EPEC10 (S6 Table). The OI-122 is a 23-kb chromosomal PAI that consists of three distinct modules, and has been associated with virulence in EHEC and EPEC [52, 53]. Virulence factor-encoding genes can be found in each one of these three modules, as follows: module 1: pagC; module 2: espL2, nleB and nleE; and module 3: efa1 and efa2 (in O157:H7 EHEC EDL933), or efa1/lifA (in tEPEC E2348/69) [35, 52, 54]. All seventeen global aEPEC identified in the EPEC7 lineage contained genes of module 2 and module 3, while the Brazilian aEPEC isolates lack these genes (S6 Table). In contrast, the pagC (module 1) and efa1/lifA (module 3) genes were detected in seven (36.8%, 7/19) Brazilian aEPEC identified in the EPEC10 lineage (serotype O2:H16), and in none of the global aEPEC (S6 Table).
NleF-encoding gene is associated with aEPEC diarrheal disease in Brazil
Since the Brazilian aEPEC isolates studied here were obtained from patients with diarrhea (89 isolates) as well from healthy subjects (17 isolates), we decided to compare if any of the 43 genes encoding virulence factors investigated was associated with their ability to cause illness in the human host. The number of virulence factor-encoding genes detected in aEPEC isolates obtained from patients with diarrhea ranged from 3 to 23, whereas in those from healthy subjects ranged from 5 to 18 (Fig 2). Of importance, the nleF gene was significantly more prevalent among aEPEC isolates obtained from diarrheagenic patients than from healthy subjects (69.7% vs 41.2%, P<0.05) (S7 Table).
The virulence factor-encoding genes were detected in each of the 106 Brazilian aEPEC genomes analyzed in this study using large-scale BLAST score ratio (LS-BSR) analysis. Each row is a different aEPEC genome, grouped by phylogenomic lineage, and each column represents a single gene (listed in the S7 Table). The presence or absence of all predicted protein-encoding genes in the 106 Brazilian aEPEC genomes is indicated by the BSR values represented in the heatmap generated by using the heatmap2 function of gplots v.3.0.1 in R v.3.3.2. Colors of the heat map indicate virulence genes that were detected with significant similarity (blue), with divergent similarity (gray) or were absent (white) in each of the Brazilian aEPEC genomes analyzed. On the far right red circles indicate aEPEC isolates obtained from patients with diarrhea, while green circles indicate aEPEC isolates obtained from healthy subjects.
As prior studies have focused on investigating the genomic diversity of EPEC isolates from locations in Europe, Africa, and Asia [22, 23], the genomic diversity of aEPEC from Brazil or other countries in South America remains largely unexplored. Thus, in this study we describe the genomic diversity of a collection of 106 human aEPEC isolates from Brazil. We also considered the genomic diversity of the Brazilian aEPEC compared to a global collection of aEPEC isolates to gain insight into how Brazilian aEPEC may differ in their evolutionary relatedness and virulence gene content compared with aEPEC from outside of Brazil. By examining the prevalence of known EPEC virulence factors, we determined that the T3SS non-LEE encoded effectors espG2  and espT [37, 56], and the autotransporter gene espC  were statistically more prevalent among aEPEC from Brazil compared to the global aEPEC from outside Brazil. Of particular interest is EspT, which was functionally characterized in aEPEC archetype isolate E110019 from one of the more several diarrheal outbreaks associated with aEPEC as there were 650 individuals in a school and 137 associated household members in Finland that developed diarrhea . EspT is responsible for triggering the formation of lamellipodia and membrane ruffles, through the activation of Rac-1 and Cdc42 , thus facilitating the bacterial invasion into non-phagocytic cells . Although the occurrence of invasive aEPEC in the Brazilian settings have already been reported , no correlation between this phenotype and the presence of the espT gene has been observed to date. EspG2 is also of importance as it is involved in microtubule disruption . EspC, a member of the autotransporter family, is involved in EPEC-mediated cell death and induces both apoptosis and necrosis in epithelial cells . In tEPEC archetype isolate E2348/69 the espC and espG2 (orf3) are located in the same PAI, termed integrative element (IE) 5 or EspC PAI [35, 59], which may explain the simultaneous dissemination of these two genes among the Brazilian aEPEC isolates. In addition, we observed that nleF was the only virulence factor-encoding gene that demonstrated a statistical association with the isolates from diarrheal disease cases when compared to Brazilian aEPEC isolates from asymptomatic individuals. NleF is a Non-LEE effector, secreted to the host cells by the T3SS, that binds to caspase-4, -8, and -9 inhibiting the catalytic activity of these caspases in vitro, and therefore, prevents apoptosis in HeLa and Caco-2 cells .
We also observed that the Brazilian aEPEC differ in their gene content at the phylogenomic lineage level, when compared with aEPEC from other geographic origins. In particular, the toxB gene of pO157 from O157:H7 EHEC strains, and efa1/lifA of the PAI OI-122 [54, 61], were more prevalent among Brazilian aEPEC identified in the EHEC2 and EPEC10 phylogenomic lineages, respectively. The proteins encoded by the toxB and efa1/lifA genes are associated with the ability of EHEC strains to adhere to epithelial cells. Of note, the EHEC adherence factors efa1 (Z4332) and efa2 (Z4333) genes, located in the third module of the PAI OI-122 from EHEC O157:H7 strains , can encode proteins with amino acid sequences that are identical to the amino acids 1 to 433 (99.0% of identity) and 437 to 711 (100.0% identity) of the Efa1 adhesin identified in the EHEC O111:H- . A previous study, using random transposon mutagenesis, demonstrated that the EHEC O157:H7 Sakai strain, harboring a transposon insertion in the efa1 gene, was significantly less adherent than the wild type strain in assays performed with Caco-2 cells . The nucleotide sequence of the efa1 gene, from the EHEC O111:H- strain, is 99.0% identical to the lifA gene (also termed efa1/lifA) from the tEPEC E2348/69, that has been demonstrated to inhibit lymphocyte proliferation and lymphokine production . A clinical study, that evaluated stool samples from three distinct children with diarrhea associated with an EHEC isolate with the serotype O157:HNM, demonstrated that EHEC may lose the stx2 gene during trafficking in the host . This line of evidence suggests that some of the aEPEC isolates in our study, which were genomically related to EHEC lineages and/or harbored virulence genes associated with this pathotype may in fact be EHEC that have lost the Shiga toxin genes during trafficking in the host or environment, or during culture in the clinical laboratory.
Overall, this study demonstrated the aEPEC from Brazil are genomically diverse, with representatives in nearly all of the E. coli phylogroups and previously described EHEC and EPEC phylogenomic lineages, which is similar to what has been previously described for aEPEC from outside of Brazil. The Brazilian aEPEC genomes described in this study not only provide insight into the genomic diversity of aEPEC in Brazil, but also allowed the description of five novel EPEC phylogenomic lineages. We also identified important differences in gene content, including known virulence associated genes, between Brazilian aEPEC compared to a global collection of aEPEC from outside Brazil. Further investigation of the genomic diversity and prevalence of select virulence genes among additional aEPEC from Brazil as well as other locations in South America may provide insight into the greater prevalence of some of these lineages of aEPEC in association with diarrheal illness in Brazil.
S1 Data. Nucleotide sequence of the gene clusters in fasta format.
S1 Fig. World map indicating the countries from which the globally distributed aEPEC collection was obtained.
Atypical EPEC from the global collection was obtained from all five continents, mainly from Africa, Asia, and Europe. In parentheses is the number of aEPEC isolates obtained from each country. This world map was constructed using the free-software Q-GIS.
S1 Table. Brazilian atypical EPEC (aEPEC) strains sequenced in the present study.
S2 Table. Global collection of atypical enteropathogenic Escherichia coli (aEPEC) strains from outside Brazil.
S3 Table. Escherichia coli and Shigella reference genomes included in the phylogenomic analysis.
S4 Table. LS-BSR identification of gene clusters that are unique in the five novel phylogenomic lineage identified in this study.
S5 Table. Gene clusters differentially detected in the Brazilian aEPEC genomes in comparison with aEPEC from outside Brazil.
S6 Table. Gene clusters differentially detected among aEPEC from Brazil and from other origins by phylogenomic lineage.
We thank Waldir Pereira Elias, from Instituto Butantan, for providing the atypical EPEC isolate BA92.
- 1. Trabulsi LR, Keller R, Tardelli Gomes TA. Typical and atypical enteropathogenic Escherichia coli. Emerg Infect Dis. 2002; 8(5):508–13. pmid:11996687.
- 2. Hernandes RT, Elias WP, Vieira MA, Gomes TA. An overview of atypical enteropathogenic Escherichia coli. FEMS Microbiol Lett. 2009; 297(2):137–49. pmid:19527295.
- 3. Kotloff KL, Nataro JP, Blackwelder WC, Nasrin D, Farag TH, Panchalingam S, et al. Burden and aetiology of diarrhoeal disease in infants and young children in developing countries (the Global Enteric Multicenter Study, GEMS): a prospective, case-control study. Lancet. 2013; 382(9888):209–22. pmid:23680352.
- 4. Acosta GJ, Vigo NI, Durand D, Riveros M, Arango S, Zambruni M, et al. Diarrheagenic Escherichia coli: Prevalence and Pathotype Distribution in Children from Peruvian Rural Communities. Am J Trop Med Hyg. 2016; 95(3):574–9. pmid:27382080.
- 5. Dias RC, Dos Santos BC, Dos Santos LF, Vieira MA, Yamatogi RS, Mondelli AL, et al. Diarrheagenic Escherichia coli pathotypes investigation revealed atypical enteropathogenic E. coli as putative emerging diarrheal agents in children living in Botucatu, Sao Paulo State, Brazil. APMIS. 2016; 124(4):299–308. pmid:26752102.
- 6. Hao R, Qiu S, Wang Y, Yang G, Su W, Song L, et al. Quinolone-resistant Escherichia coli O127a:K63 serotype with an extended-spectrum-beta-lactamase phenotype from a food poisoning outbreak in China. J Clin Microbiol. 2012; 50(7):2450–1. pmid:22553233.
- 7. Moreno AC, Filho AF, Gomes Tdo A, Ramos ST, Montemor LP, Tavares VC, et al. Etiology of childhood diarrhea in the northeast of Brazil: significant emergent diarrheal pathogens. Diagn Microbiol Infect Dis. 2010; 66(1):50–7. pmid:18508227.
- 8. Vieira MA, Dos Santos LF, Dias RC, Camargo CH, Pinheiro SR, Gomes TA, et al. Atypical enteropathogenic Escherichia coli as aetiologic agents of sporadic and outbreak-associated diarrhoea in Brazil. J Med Microbiol. 2016; 65(9):998–1006. pmid:27412254.
- 9. Ori EL, Takagi EH, Andrade TS, Miguel BT, Cergole-Novella MC, Guth BEC, et al. Diarrhoeagenic Escherichia coli and Escherichia albertii in Brazil: pathotypes and serotypes over a 6-year period of surveillance. Epidemiol Infect. 2019; 147:1–9. pmid:30229714.
- 10. Estrada-Garcia T, Lopez-Saucedo C, Thompson-Bonilla R, Abonce M, Lopez-Hernandez D, Santos JI, et al. Association of diarrheagenic Escherichia coli Pathotypes with infection and diarrhea among Mexican children and association of atypical Enteropathogenic E. coli with acute diarrhea. J Clin Microbiol. 2009; 47(1):93–8. pmid:19020055.
- 11. Cohen MB, Nataro JP, Bernstein DI, Hawkins J, Roberts N, Staat MA. Prevalence of diarrheagenic Escherichia coli in acute childhood enteritis: a prospective controlled study. J Pediatr. 2005; 146(1):54–61. pmid:15644823.
- 12. Robins-Browne RM, Bordun AM, Tauschek M, Bennett-Wood VR, Russell J, Oppedisano F, et al. Escherichia coli and community-acquired gastroenteritis, Melbourne, Australia. Emerg Infect Dis. 2004; 10(10):1797–805. pmid:15504266.
- 13. Nataro JP, Kaper JB. Diarrheagenic Escherichia coli. Clin Microbiol Rev. 1998; 11(1):142–201. pmid:9457432.
- 14. Kaper JB, Nataro JP, Mobley HL. Pathogenic Escherichia coli. Nat Rev Microbiol. 2004; 2(2):123–40. pmid:15040260.
- 15. McDaniel TK, Jarvis KG, Donnenberg MS, Kaper JB. A genetic locus of enterocyte effacement conserved among diverse enterobacterial pathogens. Proc Natl Acad Sci U S A. 1995; 92(5):1664–8. pmid:7878036.
- 16. Gaytan MO, Martinez-Santos VI, Soto E, Gonzalez-Pedrajo B. Type Three Secretion System in Attaching and Effacing Pathogens. Front Cell Infect Microbiol. 2016; 6:129. pmid:27818950.
- 17. Jerse AE, Yu J, Tall BD, Kaper JB. A genetic locus of enteropathogenic Escherichia coli necessary for the production of attaching and effacing lesions on tissue culture cells. Proc Natl Acad Sci U S A. 1990; 87(20):7839–43. pmid:2172966.
- 18. Kenny B, DeVinney R, Stein M, Reinscheid DJ, Frey EA, Finlay BB. Enteropathogenic E. coli (EPEC) transfers its receptor for intimate adherence into mammalian cells. Cell. 1997; 91(4):511–20. pmid:9390560.
- 19. Dean P, Kenny B. The effector repertoire of enteropathogenic E. coli: ganging up on the host cell. Curr Opin Microbiol. 2009; 12(1):101–9. pmid:19144561.
- 20. Wong AR, Pearson JS, Bright MD, Munera D, Robinson KS, Lee SF, et al. Enteropathogenic and enterohaemorrhagic Escherichia coli: even more subversive elements. Mol Microbiol. 2011; 80(6):1420–38. pmid:21488979.
- 21. Hazen TH, Donnenberg MS, Panchalingam S, Antonio M, Hossain A, Mandomando I, et al. Genomic diversity of EPEC associated with clinical presentations of differing severity. Nat Microbiol. 2016; 1:15014. pmid:27571975.
- 22. Hazen TH, Sahl JW, Fraser CM, Donnenberg MS, Scheutz F, Rasko DA. Refining the pathovar paradigm via phylogenomics of the attaching and effacing Escherichia coli. Proc Natl Acad Sci U S A. 2013; 110(31):12810–5. pmid:23858472.
- 23. Ingle DJ, Tauschek M, Edwards DJ, Hocking DM, Pickard DJ, Azzopardi KI, et al. Evolution of atypical enteropathogenic E. coli by repeated acquisition of LEE pathogenicity island variants. Nat Microbiol. 2016; 1:15010. pmid:27571974.
- 24. Araujo JM, Tabarelli GF, Aranda KR, Fabbricotti SH, Fagundes-Neto U, Mendes CM, et al. Typical enteroaggregative and atypical enteropathogenic types of Escherichia coli are the most prevalent diarrhea-associated pathotypes among Brazilian children. J Clin Microbiol. 2007; 45(10):3396–9. pmid:17670930.
- 25. Bankevich A, Nurk S, Antipov D, Gurevich AA, Dvorkin M, Kulikov AS, et al. SPAdes: a new genome assembly algorithm and its applications to single-cell sequencing. J Comput Biol. 2012; 19(5):455–77. pmid:22506599.
- 26. Sahl JW, Beckstrom-Sternberg SM, Babic-Sternberg J, Gillece JD, Hepp CM, Auerbach RK, et al. The In Silico Genotyper (ISG): an open-source pipeline to rapidly identify and annotate nucleotide variants for comparative genomics applications. bioRxiv. 2015.
- 27. Stamatakis A. RAxML-VI-HPC: maximum likelihood-based phylogenetic analyses with thousands of taxa and mixed models. Bioinformatics. 2006; 22(21):2688–90. pmid:16928733.
- 28. Wirth T, Falush D, Lan R, Colles F, Mensa P, Wieler LH, et al. Sex and virulence in Escherichia coli: an evolutionary perspective. Mol Microbiol. 2006; 60(5):1136–51. pmid:16689791.
- 29. Joensen KG, Tetzschner AM, Iguchi A, Aarestrup FM, Scheutz F. Rapid and Easy In Silico Serotyping of Escherichia coli Isolates by Use of Whole-Genome Sequencing Data. J Clin Microbiol. 2015; 53(8):2410–26. pmid:25972421.
- 30. Sahl JW, Caporaso JG, Rasko DA, Keim P. The large-scale blast score ratio (LS-BSR) pipeline: a method to rapidly compare genetic content between bacterial genomes. PeerJ. 2014; 2:e332. pmid:24749011.
- 31. Fu L, Niu B, Zhu Z, Wu S, Li W. CD-HIT: accelerated for clustering the next-generation sequencing data. Bioinformatics. 2012; 28(23):3150–2. pmid:23060610.
- 32. Orvis J, Crabtree J, Galens K, Gussman A, Inman JM, Lee E, et al. Ergatis: a web interface and scalable software system for bioinformatics workflows. Bioinformatics. 2010; 26(12):1488–92. pmid:20413634.
- 33. Galens K, Orvis J, Daugherty S, Creasy HH, Angiuoli S, White O, et al. The IGS Standard Operating Procedure for Automated Prokaryotic Annotation. Stand Genomic Sci. 2011; 4(2):244–51. pmid:21677861.
- 34. Hazen TH, Kaper JB, Nataro JP, Rasko DA. Comparative Genomics Provides Insight into the Diversity of the Attaching and Effacing Escherichia coli Virulence Plasmids. Infect Immun. 2015; 83(10):4103–17. pmid:26238712.
- 35. Iguchi A, Thomson NR, Ogura Y, Saunders D, Ooka T, Henderson IR, et al. Complete genome sequence and comparative genome analysis of enteropathogenic Escherichia coli O127:H6 strain E2348/69. J Bacteriol. 2009; 191(1):347–54. pmid:18952797.
- 36. Ogura Y, Abe H, Katsura K, Kurokawa K, Asadulghani M, Iguchi A, et al. Systematic identification and sequence analysis of the genomic islands of the enteropathogenic Escherichia coli strain B171-8 by the combined use of whole-genome PCR scanning and fosmid mapping. J Bacteriol. 2008; 190(21):6948–60. pmid:18757547.
- 37. Bulgin RR, Arbeloa A, Chung JC, Frankel G. EspT triggers formation of lamellipodia and membrane ruffles through activation of Rac-1 and Cdc42. Cell Microbiol. 2009; 11(2):217–29. pmid:19016787.
- 38. Arbeloa A, Oates CV, Marches O, Hartland EL, Frankel G. Enteropathogenic and enterohemorrhagic Escherichia coli type III secretion effector EspV induces radical morphological changes in eukaryotic cells. Infect Immun. 2011; 79(3):1067–76. pmid:21189318.
- 39. Afset JE, Bruant G, Brousseau R, Harel J, Anderssen E, Bevanger L, et al. Identification of virulence genes linked with diarrhea due to atypical enteropathogenic Escherichia coli by DNA microarray analysis and PCR. J Clin Microbiol. 2006; 44(10):3703–11. pmid:17021100.
- 40. Abreu AG, Bueris V, Porangaba TM, Sircili MP, Navarro-Garcia F, Elias WP. Autotransporter protein-encoding genes of diarrheagenic Escherichia coli are found in both typical and atypical enteropathogenic E. coli strains. Appl Environ Microbiol. 2013; 79(1):411–4. pmid:23104414.
- 41. Scaletsky IC, Aranda KR, Souza TB, Silva NP, Morais MB. Evidence of pathogenic subgroups among atypical enteropathogenic Escherichia coli strains. J Clin Microbiol. 2009; 47(11):3756–9. pmid:19759223.
- 42. Gomes TA, Hernandes RT, Torres AG, Salvador FA, Guth BE, Vaz TM, et al. Adhesin-encoding genes from shiga toxin-producing Escherichia coli are more prevalent in atypical than in typical enteropathogenic E. coli. J Clin Microbiol. 2011; 49(9):3334–7. pmid:21795517.
- 43. Lacher DW, Steinsland H, Whittam TS. Allelic subtyping of the intimin locus (eae) of pathogenic Escherichia coli by fluorescent RFLP. FEMS Microbiol Lett. 2006; 261(1):80–7. pmid:16842363.
- 44. Xu Y, Bai X, Zhao A, Zhang W, Ba P, Liu K, et al. Genetic Diversity of Intimin Gene of Atypical Enteropathogenic Escherichia coli Isolated from Human, Animals and Raw Meats in China. PLoS One. 2016; 11(3):e0152571. pmid:27031337.
- 45. Abe CM, Trabulsi LR, Blanco J, Blanco M, Dahbi G, Blanco JE, et al. Virulence features of atypical enteropathogenic Escherichia coli identified by the eae+ EAF-negative stx- genetic profile. Diagn Microbiol Infect Dis. 2009; 64(4):357–65. pmid:19442475.
- 46. Rasko DA, Rosovitz MJ, Myers GS, Mongodin EF, Fricke WF, Gajer P, et al. The pangenome structure of Escherichia coli: comparative genomic analysis of E. coli commensal and pathogenic isolates. J Bacteriol. 2008; 190(20):6881–93. pmid:18676672.
- 47. Lacher DW, Steinsland H, Blank TE, Donnenberg MS, Whittam TS. Molecular evolution of typical enteropathogenic Escherichia coli: clonal analysis by multilocus sequence typing and virulence gene allelic profiling. J Bacteriol. 2007; 189(2):342–50. pmid:17098897.
- 48. Tennant SM, Tauschek M, Azzopardi K, Bigham A, Bennett-Wood V, Hartland EL, et al. Characterisation of atypical enteropathogenic E. coli strains of clinical origin. BMC Microbiol. 2009; 9:117. pmid:19490652.
- 49. Viljanen MK, Peltola T, Junnila SY, Olkkonen L, Jarvinen H, Kuistila M, et al. Outbreak of diarrhoea due to Escherichia coli O111:B4 in schoolchildren and adults: association of Vi antigen-like reactivity. Lancet. 1990; 336(8719):831–4. pmid:1976876.
- 50. Stein M, Kenny B, Stein MA, Finlay BB. Characterization of EspC, a 110-kilodalton protein secreted by enteropathogenic Escherichia coli which is homologous to members of the immunoglobulin A protease-like family of secreted proteins. J Bacteriol. 1996; 178(22):6546–54. pmid:8932311.
- 51. Tatsuno I, Horie M, Abe H, Miki T, Makino K, Shinagawa H, et al. toxB gene on pO157 of enterohemorrhagic Escherichia coli O157:H7 is required for full epithelial cell adherence phenotype. Infect Immun. 2001; 69(11):6660–9. pmid:11598035.
- 52. Karmali MA, Mascarenhas M, Shen S, Ziebell K, Johnson S, Reid-Smith R, et al. Association of genomic O island 122 of Escherichia coli EDL 933 with verocytotoxin-producing Escherichia coli seropathotypes that are linked to epidemic and/or serious disease. J Clin Microbiol. 2003; 41(11):4930–40. pmid:14605120.
- 53. Vieira MA, Salvador FA, Silva RM, Irino K, Vaz TM, Rockstroh AC, et al. Prevalence and characteristics of the O122 pathogenicity island in typical and atypical enteropathogenic Escherichia coli strains. J Clin Microbiol. 2010; 48(4):1452–5. pmid:20181917.
- 54. Perna NT, Plunkett G 3rd, Burland V, Mau B, Glasner JD, Rose DJ, et al. Genome sequence of enterohaemorrhagic Escherichia coli O157:H7. Nature. 2001; 409(6819):529–33. pmid:11206551.
- 55. Smollett K, Shaw RK, Garmendia J, Knutton S, Frankel G. Function and distribution of EspG2, a type III secretion system effector of enteropathogenic Escherichia coli. Microbes Infect. 2006; 8(8):2220–7. pmid:16781180.
- 56. Bulgin R, Arbeloa A, Goulding D, Dougan G, Crepin VF, Raymond B, et al. The T3SS effector EspT defines a new category of invasive enteropathogenic E. coli (EPEC) which form intracellular actin pedestals. PLoS Pathog. 2009; 5(12):e1000683. pmid:20011125.
- 57. Pacheco VC, Yamamoto D, Abe CM, Hernandes RT, Mora A, Blanco J, et al. Invasion of differentiated intestinal Caco-2 cells is a sporadic property among atypical enteropathogenic Escherichia coli strains carrying common intimin subtypes. Pathog Dis. 2014; 70(2):167–75. pmid:24339197.
- 58. Serapio-Palacios A, Navarro-Garcia F. EspC, an Autotransporter Protein Secreted by Enteropathogenic Escherichia coli, Causes Apoptosis and Necrosis through Caspase and Calpain Activation, Including Direct Procaspase-3 Cleavage. MBio. 2016; 7(3). pmid:27329750.
- 59. Mellies JL, Navarro-Garcia F, Okeke I, Frederickson J, Nataro JP, Kaper JB. espC pathogenicity island of enteropathogenic Escherichia coli encodes an enterotoxin. Infect Immun. 2001; 69(1):315–24. pmid:11119520.
- 60. Blasche S, Mortl M, Steuber H, Siszler G, Nisa S, Schwarz F, et al. The E. coli effector protein NleF is a caspase inhibitor. PLoS One. 2013; 8(3):e58937. pmid:23516580.
- 61. Burland V, Shao Y, Perna NT, Plunkett G, Sofia HJ, Blattner FR. The complete DNA sequence and analysis of the large virulence plasmid of Escherichia coli O157:H7. Nucleic Acids Res. 1998; 26(18):4196–204. pmid:9722640.
- 62. Nicholls L, Grant TH, Robins-Browne RM. Identification of a novel genetic locus that is required for in vitro adhesion of a clinical isolate of enterohaemorrhagic Escherichia coli to epithelial cells. Mol Microbiol. 2000; 35(2):275–88. pmid:10652089.
- 63. Tatsuno I, Kimura H, Okutani A, Kanamaru K, Abe H, Nagai S, et al. Isolation and characterization of mini-Tn5Km2 insertion mutants of enterohemorrhagic Escherichia coli O157:H7 deficient in adherence to Caco-2 cells. Infect Immun. 2000; 68(10):5943–52. pmid:10992506.
- 64. Klapproth JM, Scaletsky IC, McNamara BP, Lai LC, Malstrom C, James SP, et al. A large toxin from pathogenic Escherichia coli strains that inhibits lymphocyte activation. Infect Immun. 2000; 68(4):2148–55. pmid:10722613.
- 65. Mellmann A, Lu S, Karch H, Xu JG, Harmsen D, Schmidt MA, et al. Recycling of Shiga toxin 2 genes in sorbitol-fermenting enterohemorrhagic Escherichia coli O157:NM. Appl Environ Microbiol. 2008; 74(1):67–72. pmid:17981936.