Prospective genomic characterization of the German enterohemorrhagic Escherichia coli O104:H4 outbreak by rapid next generation sequencing technology.

Prospective Genomic Characterization of the German
Enterohemorrhagic Escherichia coli O104:H4 Outbreak
by Rapid Next Generation Sequencing Technology
Alexander Mellmann1., Dag Harmsen2*., Craig A. Cummings3., Emily B. Zentz4, Shana R. Leopold1,
Alain Rico5, Karola Prior2, Rafael Szczepanowski2, Yongmei Ji3, Wenlan Zhang1, Stephen F. McLaughlin3,
John K. Henkhaus4, Benjamin Leopold1, Martina Bielaszewska1, Rita Prager6, Pius M. Brzoska3, Richard L.
Moore4, Simone Guenther5, Jonathan M. Rothberg7, Helge Karch1
1 Institute of Hygiene, University Mu¨nster, Mu¨nster, Germany, 2Department of Periodontology, University Mu¨nster, Mu¨nster, Germany, 3 Life Technologies, Foster City,
California, United States of America, 4OpGen, Gaithersburg, Maryland, United States of America, 5 Life Technologies, Darmstadt, Germany, 6 Robert Koch Institute,
Wernigerode Branch, Wernigerode, Germany, 7 Ion Torrent by Life Technologies, Guilford, Connecticut, United States of America
Abstract
An ongoing outbreak of exceptionally virulent Shiga toxin (Stx)-producing Escherichia coli O104:H4 centered in Germany,
has caused over 830 cases of hemolytic uremic syndrome (HUS) and 46 deaths since May 2011. Serotype O104:H4, which
has not been detected in animals, has rarely been associated with HUS in the past. To prospectively elucidate the unique
characteristics of this strain in the early stages of this outbreak, we applied whole genome sequencing on the Life
Technologies Ion Torrent PGMTM sequencer and Optical Mapping to characterize one outbreak isolate (LB226692) and a
historic O104:H4 HUS isolate from 2001 (01-09591). Reference guided draft assemblies of both strains were completed with
the newly introduced PGMTM within 62 hours. The HUS-associated strains both carried genes typically found in two types of
pathogenic E. coli, enteroaggregative E. coli (EAEC) and enterohemorrhagic E. coli (EHEC). Phylogenetic analyses of 1,144
core E. coli genes indicate that the HUS-causing O104:H4 strains and the previously published sequence of the EAEC strain
55989 show a close relationship but are only distantly related to common EHEC serotypes. Though closely related, the
outbreak strain differs from the 2001 strain in plasmid content and fimbrial genes. We propose a model in which EAEC
55989 and EHEC O104:H4 strains evolved from a common EHEC O104:H4 progenitor, and suggest that by stepwise gain and
loss of chromosomal and plasmid-encoded virulence factors, a highly pathogenic hybrid of EAEC and EHEC emerged as the
current outbreak clone. In conclusion, rapid next-generation technologies facilitated prospective whole genome
characterization in the early stages of an outbreak.
Citation: Mellmann A, Harmsen D, Cummings CA, Zentz EB, Leopold SR, et al. (2011) Prospective Genomic Characterization of the German Enterohemorrhagic
Escherichia coli O104:H4 Outbreak by Rapid Next Generation Sequencing Technology. PLoS ONE 6(7): e22751. doi:10.1371/journal.pone.0022751
Editor: Niyaz Ahmed, University of Hyderabad, India
Received June 29, 2011; Accepted July 6, 2011; Published July 20, 2011
Copyright: � 2011 Mellmann 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.
Funding: Funding for this work was provided by the BMBF (no. 01KI1020 to AM, DH, HK) and the Medical Faculty Muenster (grant BD9817044 to AM). The
funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing Interests: Dr. Cummings, Mr. Rico, Dr. Ji, Mr. McLaughlin, Dr. Broska, Dr. Guenther, and Dr. Rothberg are all employees of Life Technologies, Inc. Ms.
Zentz, Mr. Henkhaus, and Dr. Moore are all employees of OpGen, Inc. Dr. Harmsen is a part-time employee of the RIDOM Bioinformatics company. While these
authors are company employees, this does not alter their adherence to all the PLoS ONE policies on sharing data and materials. All other authors have declared
that no competing interests exist.
* E-mail: dharmsen@uni-muenster.de
. These authors contributed equally to this work.
Introduction
Enterohemorrhagic Escherichia coli (EHEC) are a pathogenic
subgroup of Shiga toxin (Stx)-producing E. coli (STEC) that cause
human disease including diarrhea, bloody diarrhea and the
hemolytic uremic syndrome (HUS) [1]. After ingestion of the
pathogen and a subsequent incubation period of 2 to 3 days
patients (most frequently children) develop watery diarrhea that is
typically accompanied by abdominal pain. Bloody diarrhea ensues
after a 2–4-day interval in about 80% of cases. Approximately one
week (range 3 to 13 days) after the onset of diarrhea, 10% to 15%
of patients (data for children under 10 years of age) develop HUS
[2,3]. Since early May 2011, there has been an outbreak of
multidrug-resistant EHEC O104:H4 in Germany [4]. As of June
24, 2011, 834 cases of HUS and 2,967 non-HUS cases were
reported by the German Robert Koch Institute (RKI); 30 of the
HUS cases and 16 of the non-HUS cases resulted in death [5].
Furthermore, 100 additional infections have been identified in 12
other European countries and even in the United States and
Canada [6]. While still ongoing at the time of publication, this is
already the largest outbreak caused by EHEC in Germany and the
largest outbreak of HUS worldwide. Extensive efforts to identify
the source implicated contaminated sprouts, which was later
confirmed by isolation of the outbreak strain from the sprouts on
June 12th [7]. Historically, E. coli O104:H4 has been associated
with very few HUS cases [8,9]. To date in Germany only one of
588 EHEC strains isolated from HUS patients in the National
Consulting Laboratory for HUS and the Reference Laboratory for
PLoS ONE | www.plosone.org 1 July 2011 | Volume 6 | Issue 7 | e22751

Enterobacteriaceae of the RKI belongs to serotype O104:H4, and this
strain was isolated in 2001 [9]. The E. coli O104:H4 isolate from
this German HUS case is included in the HUS-associated E. coli
(HUSEC) collection [9]. This collection presently contains 42
representative, well-characterized EHEC strains that cover the
entire serotype and phylogenetic spectrum of HUS-associated
EHEC in Germany [9]. Among several different Stx types present
in members of the HUSEC collection, Stx2 is most common.
Here, we prospectively use whole genome based methods,
including Ion Torrent sequencing [10] and Optical Mapping,
during an ongoing outbreak to characterize and compare the E.
coli O104:H4 outbreak strain with the historical EHEC O104:H4
isolate (01-09591) and EAEC O104:H4 strain 55989, isolated in
Central Africa in the late 1990s [11]. These analyses enabled us to
propose an evolutionary model for the emergence of the present
German outbreak strain, and to identify sequences that are specific
to the current outbreak strain.
Results
Description of the outbreak
Major events in the time course of the outbreak and the
genomic elucidation efforts are shown in Figure 1. On May 19, the
RKI noted an increased frequency of HUS and bloody diarrhea
cases in northern Germany, predominantly among adults [5]. By
retrospective analysis this outbreak had begun in early May 2011.
The outbreak was reported for the first time to the European
Center for Disease Prevention and Control (ECDC) on May 22.
On May 25, the RKI and Federal Institute for Risk Assessment
(BfR) issued a statement that warned against consuming
cucumbers, leaf lettuce, and tomatoes, and the ECDC informed
all European countries about the German EHEC outbreak,
defined as such based on disease characteristics. One day later,
STEC were detected on Spanish cucumbers by PCR, though
further investigations ruled out any link to the current O104:H4
outbreak. On June 5, sprouts were suspected as the outbreak
source by epidemiological evidence. This was confirmed by
detection of EHEC O104:H4 five days later [7]. The efforts to
elucidate the source of the outbreak by epidemiological investiga-
tions of cases and food supply chains were coordinated by the
German Federal Institutions RKI and BfR; microbiological
investigations were also coordinated by us as the German National
Consulting Laboratory for HUS, a part of the Institute of Hygiene.
Although standard methods (stool broth enrichment, subsequent
culture on ESBL agar and CT-SMAC, and stx PCR) enabled
identification of the index isolate, only molecular methods
including gnd sequencing, MLST, and fliC genotyping, enabled
us to recognize that the outbreak isolates were very closely related
to strain 01-09591 from the HUSEC reference strain collection
(The outbreak strain and 01-09591 are collectively referred to here
as the ‘HUSEC041 complex’). Subsequently, we used Life
Technologies and OpGen next-generation genomic technologies
in a prospective manner for the first time ever during an ongoing
outbreak [12].
Relatedness of outbreak and reference isolates
Whole genome optical maps were created for four outbreak
isolates (including LB226692) from four different German cities
and two historical reference strains 02-03885 (HUSEC037) and
01-09591 (HUSEC041). Comparison of the optical maps
demonstrated that all four outbreak strains are identical within
the limits of Optical Mapping resolution, suggesting that the
outbreak is likely to be clonal and single-sourced (Fig. 2).
Moreover, alignments of optical maps to an in silico digest of
EAEC 55989 showed strong similarity of this strain to the
outbreak strains and 01-09591.
Genomic characterization of the EHEC O104:H4 outbreak
strain
Sequencing on the Ion Torrent PGMTM sequencer was
completed within 62 hours, leading to the public release of the
draft assembly of outbreak strain LB226692 on June 3 (Fig. 1,
Table S1). Sequence data of the closely related historical isolate
01-09591 was also generated while the outbreak was still
occurring. Genome assemblies based on the PGMTM reads
showed that both of these HUS-causing strains (LB226692 and
01-09591) carry genes typically found in two different E. coli
pathotypes, specifically EAEC and EHEC. Genome wide
phylogenetic analysis based on core chromosomal ORFs
(n= 1,144) demonstrated the close relationship of the LB226692
and 01-09591 strains to the previously sequenced EAEC strain
55989 (NCBI acc. no. NC_011478), and indicated that these
strains are only distantly related to the commonly isolated EHEC
serotypes (Fig. 3). However, unlike typical EAEC strains, both
LB226692 and 01-09591 have an stx2-harboring prophage
integrated in wrbA, which is also the integration site for stx2-
phages in EHEC O157:H7 outbreak strains EDL933 [13] and
Sakai (RIMD 0509952) [14]. The wrbA gene of EAEC 55989 is
not occupied by a prophage. Furthermore, the IrgA homologue
adhesin encoding gene (iha), which is responsible for adherence to
epithelial cells and has been found in eae-negative STEC [15], is
present in all three strains. In contrast to the two HUSEC041
complex strains, 55989 does not harbor the tellurite resistance
encoding genes (ter). These characteristics led to the development
of a rapid PCR-based test of stx2, O104 lipopolysaccharide (LPS)
gene (rfbO104), H4 flagellin-encoding gene (fliCH4), and terD for the
detection of the HUSEC041 complex [16].
Plasmid profiling demonstrated that LB226692 and 01-09591
each harbor two large plasmids (Fig. 4; 83 and 90 kb and 75 and
95 kb, respectively). Sequence analysis shows that the smaller
plasmid of LB226692 contains aggregative adherence fimbriae
type I (AAF/I) but lacks the EAEC heat-stable enterotoxin
encoding gene, astA. The larger plasmid is an incompatibility
group I1 (IncI1) plasmid with high similarity to pEC_Bactec
(NCBI acc. no. GU371927) that harbors TEM-1 and CTX-M-15
beta-lactamase genes [17]. The large plasmid of strain 01-09591
appears to be closely related to the IncI1 family plasmid pSERB1
(NCBI acc. no. AY686591) from EAEC strain C1096 [18]. This
strain (01-09591) also has a TEM-1 beta-lactamase that is located
on a sequence contig carrying a number of genes encoding
plasmid functions; we propose that this locus is also carried on the
larger plasmid. The smaller plasmid of 01-09591 is an EAEC
plasmid containing AAF/III, which is astA-positive, and closely
related to the plasmid from EAEC 55989. VitekH 2 and E-testH
resistance testing indicates that both strains (LB226692 and 01-
09591) have a TEM-1 phenotype. The ESBL genotype (CTX-M-
15) of strain LB226692 was also phenotypically confirmed.
In addition to their plasmid content, the current and historical
HUS strains differed slightly in chromosomal gene content. Most
of the regions of difference were in prophage and insertion
elements. One of the prophage regions, present only in 01-09591,
harbored a gene encoding a homologue of the EmrE multidrug
resistance efflux pump. The strains also differed in their
complement of fimbrial operons. LB226692 lacks part of ybgOPQ,
encoding a putative fimbrial adhesin, and 01-09591 lacks the fmlA
gene encoding a major fimbrial subunit. Both of these operons are
intact in EAEC 55989.
Genomic Characterization of the O104:H4 Outbreak
PLoS ONE | www.plosone.org 2 July 2011 | Volume 6 | Issue 7 | e22751

Phenotypic acid resistance testing demonstrated a high number
of surviving cells for both strains at pH 2.5 (580 and 540 colony
forming units [CFU] per ml liquid culture, LB226692 and 01-
09591, respectively), even higher than the number observed for
EHEC O157:H7 strain EDL933 (480 CFU/ml). The gene rpoS,
which is an important regulator for stress response, including acid
resistance [19], was found to be identical with an intact reading
frame in strains LB226692 and 01-09591.
Discussion
Rapid next-generation genomics technologies facilitated pro-
spective whole genome characterization in the early stages of this
deadly outbreak. Previously, Optical Mapping has been used in
retrospective outbreak investigations to examine isolate relatedness
and distinguish outbreak from background samples [20]. Devel-
opment of the rapid, automated ArgusTM Optical Mapping
System now allows this technique to be used in real time outbreak
investigations. In this study, de novo, whole-chromosome Optical
Maps were created for six organisms in just over two days (Fig. 1).
The Ion Torrent PGMTM sequencing platform [10] makes whole-
genome sequencing of microbial isolates in the early stages of an
outbreak possible. Using this platform, an outbreak isolate and related
strains were sequenced and assembled within two to three days
(Fig. 1). A draft assembly of an independent isolate from the same
outbreak was simultaneously produced on the PGMTM sequencer
over a similar time frame [21]. In addition, by combining whole
genome-based next-generation genomics technologies from the
laboratory with sophisticated software solutions (e.g., Geographical
Information Systems and space-time cluster analysis), highly specific
[12] and sensitive real-time monitoring of infectious diseases and
early-warning outbreak detection is achievable [22,23].
The current German outbreak strain is unusual in several
aspects. First, Stx-producing serotype O104:H4 are rarely isolated
worldwide from HUS patients. In this situation, the availability of
the HUSEC collection greatly facilitated the phylogenetic
grouping, providing guidance about the virulence profile [9].
Second, there is no evidence of zoonotic origin, in contrast to Stx-
producing E. coli O104 with H antigens H7, H12 and H21 [24].
Third, the chromosomal backbone is similar to EAEC of serotype
O104:H4, which does not cause such severe diseases like HUS. It
is also lacking intimin (eae), which is responsible for adhesion to
epithelial cells in typical EHEC, though it does possess the iha,
which has been described in other pathogenic E. coli lacking
intimin [25,26]. Finally, there have been no previously reported
outbreaks caused by this serotype, though another Stx-producing
Figure 1. Events timeline of German EHEC O104:H4 outbreak. Major events relating to the outbreak epidemiology (below arrow) and those
relating to genomic elucidation efforts (above arrow) are noted separately in the graph. Lines within the arrow indicate single day progression, with
the date noted every 5th day. Events span from early May 2011 to early June 2011. Times are noted in Central European Time (CET). Abbreviations:
BfR = Bundesinstitut fu¨r Risikobewertung (Federal Institute for Risk Assessment, Germany), BGI = Beijing Genomics Institute (People’s Republic of
China), ECDC = European Center for Disease Prevention and Control (Sweden), HPA = Health Protection Agency (United Kingdom), HUS =
hemolytic uremic syndrome, LT = Life Technologies Group, PGMTM = Ion Torrent Personal Genome MachineTM, RKI = Robert Koch Institute
(Germany), ST = multilocus sequence type, UKE = University Hospital Hamburg (Germany), UKM = University Hospital Muenster (Germany), WGS =
whole genome sequencing.
doi:10.1371/journal.pone.0022751.g001
Genomic Characterization of the O104:H4 Outbreak
PLoS ONE | www.plosone.org 3 July 2011 | Volume 6 | Issue 7 | e22751

EAEC (Serogroup O111) has been linked to an outbreak of HUS
in France [27]. Due to its hybrid pathogenicity characteristics,
Brzuszkiewicz and colleagues assigned a new pathotype ‘Entero-
Aggregative-Haemorrhagic Escherichia coli (EAHEC)’ [3]. While
this is an accurate description, we believe that grouping of the
current outbreak strains into the pathotype ‘EHEC’ is appropriate
as it reflects the major clinical attributes and the follows the
precedent set by other eae-negative EHEC (e. g. O91:H21,
O113:H21, O121:H19) [28].
An evolutionary model of the origin of the present outbreak
strain (Fig. 5) proposes that a hypothetical Stx-producing E. coli
O104:H4 with an EAEC genetic background gave rise to both
HUSEC041 strains (LB226692 and 01-09591) and the EAEC
strain 55989. In principle, there are two evolutionary models
possible: (i) the ‘common ancestor model’ proposes a hypothetical
O104:H4 progenitor and (ii) the ‘linear ancestry model’ suggests
that all EHEC O104:H4 originated from the prototypical EAEC
55989. Inclusion of the sequence information of the historical
EHEC O104:H4 strain (01-09591) provides additional informa-
tion supporting the common ancestor model. Acquisition of Stx-
encoding genes in strains with an unoccupied insertion site has
been previously shown [29], however, loss of several genes and
genomic islands is more likely and occurs frequently [30,31].
Therefore, EAEC 55989 (E. coli O104:H4, ST678) appears to be
recently derived from a progenitor Stx-producing E. coli O104:H4.
Only 24 out of a total of 1,144 core genes vary in primary
sequence between these strains (Fig. 5). Moreover, the existence of
an intact stx integration site at wrbA in EAEC 55989 and the
presence of iha, which is often adjacent to tellurite resistance genes
[31], further corroborates the model of descent from the putative
HUSEC041 progenitor. The HUSEC041 complex is terD positive,
but we recently showed that the ter island can frequently be lost via
complete or internal deletions in ter-harboring O island [31]. Such
profound chromosomal changes can occur during the brief period
that EHEC passes through the human gastrointestinal tract
leading to gains and losses of virulence determinants, which may
account for the multiple loci differences and plasmid gain and loss
seen in this model [1]. In almost all major EHEC serotypes stx2
loss has been described, leading to enormous difficulties in the
diagnosis of these pathogens [30]. Both isolates of the HUSEC041
complex had 87 (01-09591) and 95 (LB226692) unique core
genome alleles which possibly can be attributed to the additional
time to acquire mutations since divergence from the common
progenitor as these were isolated in 2001 and 2011, respectively
(Fig. 5). To visualize the common ancestor model in the
phylogenetic tree, we have incorporated the hypothetical
O104:H4 progenitor in figure 3.
Optical Mapping data using four outbreak strains, 01-09591
and EAEC 55989 also corroborates the common ancestor model
(Fig 6). If it is assumed that the hypothetical common O104:H4
ancestor contains the shared genomic regions of the outbreak
strains, then LB226692 evolved by three insertion events, 01-
09591 by one insertion event, and EAEC 55989 by six insertion
events. All of these events would be unconstrained, meaning that
the nature and location of the insertions leading to one strain
would be independent of the insertions that lead to the other
strains. According to the linear model, following the time of
isolation, EAEC 55989 leads to 01-09591, which then leads to
LB226692. In this model, EAEC 55989 loses six regions and gains
one new genomic region to give rise to 01-09591. Then, 01-09591
loses this recently gained region and gains three additional new
regions leading to the current outbreak strain. Although the total
number of genomic changes is the same in both models, the linear
model imposes an additional constraint of the gain and subsequent
loss of the same genetic region.
Interestingly, the isolate from 2001 retained the 75 kb plasmid
from the progenitor encoding AAF/III fimbriae (also present in
EAEC strain 55989 in a different size) and acquired a 95 kb
plasmid encoding type IV pilus system and TEM-1. In contrast,
the outbreak isolate LB226692 acquired two new plasmids, one
encoding AAF/I fimbriae (83 kb) and another encoding TEM-1
and CTX-M-15 beta-lactamases (90 kb), and lost the AAF/III
fimbriae encoding plasmid. Whereas the ESBL phenotype is an
additional characteristic of the outbreak strain, the remaining
antibiotic susceptibility profile is similar to the isolate from 2001
(http://www.ehec.org). Finally, the high acid resistance of the
HUSEC041 complex strains may facilitate survival during passage
through the acidic environment of the stomach, thus contributing
to the high pathogenicity of the current outbreak strain.
In conclusion, rapid next-generation technologies facilitated
prospective whole genome characterization in the early stages of
an outbreak. In the future, these technologies will help to make
informed decisions about treatment, prevention, and source
tracking.
Materials and Methods
Patients and strains
The outbreak isolate LB226692 and the historic isolates 01-
09591 (HUSEC041; serotype O104:H4) were selected for whole
Figure 2. Optical Map similarity cluster of German EHEC
O104:H4 outbreak. De novo whole genome optical maps from EHEC
O104:H4 outbreak strains (‘LB’ prefix) and historical O104:H4 and
O104:H21 strains (01-09591 and 02-03885) were created using the
ArgusTM Optical Mapping System with the NcoI restriction enzyme. An
in silico genomic map of the reference strain 55989 was created in
MapSolverTM by importing genomic sequence data from NCBI (acc. no.
NC_011478) and applying the NcoI restriction pattern. Optical maps and
in silico maps were compared using the default MapSolverTM
parameters and clustered using UPGMA based on the resulting pairwise
distance metrics. Scale represents percent difference. Strain name,
serotype, country, city, and year of isolation are indicated.
doi:10.1371/journal.pone.0022751.g002
Genomic Characterization of the O104:H4 Outbreak
PLoS ONE | www.plosone.org 4 July 2011 | Volume 6 | Issue 7 | e22751

Genomic Characterization of the O104:H4 Outbreak
PLoS ONE | www.plosone.org 5 July 2011 | Volume 6 | Issue 7 | e22751