LEEways: tales of EPEC, ATEC and EHEC


E-mail infekt@uni-muenster.de; Tel. (+49) 251 8356466; Fax (+49) 251 8356467.


Intestinal pathogenic Escherichia coli are a major cause of worldwide morbidity and mortality. Currently seven intestinal pathovars are recognized causing a wide range of intestinal disorders that are sometimes associated with severe and even lethal complications. The arsenal of virulence factors is used to subvert cellular functions of the host thereby enhancing adaptation, virulence and pathogenicity. Virulence factor profiles are largely the result of the acquisition of mobile genetic elements such as prophages and pathogenicity islands. A group of highly adapted intestinal pathogenic E. coli that are characterized by the induction of ‘attaching-and-effacing (A/E) lesions’ have acquired a decisive pathogenicity island, the ‘locus of enterocyte effacement – LEE’ by horizontal gene transfer. This review focuses on recent advances in our understanding of A/E E. coli. It highlights novel functions of effector proteins, addresses the LEE flanking regions where additional genetic elements such as the LifA/Efa1 region have been identified, and points to implications for diagnostics and therapy due to the putative interconversion of A/E E. coli during infection.


Escherichia coli is a remarkable versatile organism exhibiting astonishing genome plasticity (Croxen and Finlay, 2010; Tenaillon et al., 2010). Even in the non-pathogenic E. coli K-12 strain used in many laboratories an estimated 17.6% of its coding sequence have been acquired from other species. The sequenced E. coli isolates exhibit a core genome of approximately 2200 genes and a pan-genome of around 13 000 genes (Rasko et al., 2008). Due to the acquisition of pathogenicity islands and other mobile genetic elements, the genomes of pathogenic E. coli can be up to 1 Mb larger than of their commensal relatives and could encode about 5000 genes. This leaves ample opportunity for substantial genetic diversity resulting in different virulence factor profiles. Recent genome analysis indicated that acquisition of genes as well as loss of genes has contributed to the emergence of the pathogroups in E. coli (Croxen and Finlay, 2010).

At present nine distinct pathovars have been recognized in E. coli. Further to the pathovars of uropathogenic (UPEC) and neonatal meningitis/sepsis-causing E. coli (NMEC), seven intestinal (diarrhoeagenic) pathovars – enteropathogenic E. coli (EPEC), atypical enteropathogenic E. coli (aEPEC or ATEC), enterotoxigenic E. coli (ETEC), enteroinvasive E. coli (EIEC), enteroaggregative E. coli (EAEC), enterohaemorrhagic E. coli (EHEC) and diffusely adhering E. coli (DAEC) – have been described. In addition, there are other human pathovars such as adherent invasive E. coli (AIEC) that has been associated with Crohn's disease (Rolhion and Darfeuille-Michaud, 2007; Lapaquette et al., 2009). Commonly, pathovars are defined based on specific virulence factor profiles and their characteristics and manifestations of pathogenicity in a given host. E. coli pathotypes tend to be clonal groups that express O (somatic) and H (flagellar) antigens that define their serogroup (O-type) or serotype (O- and H-type). Lately, intermediate strains have been described that based on their arsenal of virulence factors cannot be attributed to a recognized pathovar which exemplifies the ongoing changes in pathogenic E. coli strains (Müller et al., 2007).

This review addresses recent advances in our understanding of the LEE-harbouring E. coli pathovars: EHEC, ATEC and EPEC. As by no means this can be comprehensive I do apologize to all colleagues whose excellent work could not be included due to space limitations.

Enteropathogenic E. coli (EPEC)

In the early 1940s E. coli has been implicated with human diarrhoeal disease and these strains (O111) have been termed ‘enteropathogenic’E. coli (EPEC). In 1978, studies with human volunteers by Myron Levine using theEPEC O127 : H6 strain E2348/69 clearly demonstrated the intestinal pathogenic potential of EPEC strains. Based on these studies EPEC E2348/69 became the prototype EPEC strain. E2348/69 harbours the core ‘locus of enterocyte effacement – LEE’ whereas almost all other isolates investigated to date contain quite extended 5′ and 3′ flanking regions (e.g. Gärtner and Schmidt, 2004; Müller et al., 2009). The typical ‘attaching-and-effacing (A/E) lesions’ in concert with pedestal formation that are characteristic for LEE-harbouring pathogens (Fig. 1) were only identified in subsequent investigations. A consensus definition of EPEC was accepted during the 2nd International Symposium on EPEC in São Paulo in 1995.

Figure 1.

Pedestal formation by atypical enteropathogenic Escherichia coli (ATEC) strains. Scanning (SEM) and transmission (TEM) electron microscopy of ATEC strains following HeLa or Int-407 cell infection (3 h).
A. SEM of pedestal formation by the ATEC strain 3431-4/86 (O8 : H-).
B. TEM of ATEC strain B6 (O26 : K60) 3 h of infection/HeLa cells.
C. TEM of ATEC strain B6 (O26 : K60) 6 h of infection/HeLa cells.
D. TEM of ATEC strain B6 (O26 : K60) 3 h of infection/INT-407 cells.
B–D. Polymerized actin is clearly visible underneath adherent bacteria.

EPEC are diarrheagenic Escherichia coli that produce characteristic histopathology known as attaching and effacing (A/E) on intestinal cells and that do not produce Shiga, Shiga-like, or verotoxins. Typical EPEC of human origin possess a virulence plasmid known as EAF (EPEC adherence factor) plasmid that encodes localized adherence on cultured epithelial cells mediated by the Bundle Forming Pilus (BFP), while atypical EPEC do not possess this plasmid. The majority of EPEC strains fall into certain well-recognized O : H serotypes such as O26; 55, 86, 111, 114, 119, 125, 126, 127, 128, 142 and 158.

Following these basic definitions various subgroups based on evolutionary considerations and the presence of certain isotypes particularly of the eae gene product ‘intimin’ have been proposed (e.g. Ito et al., 2007). Currently, 27 eae variants encoding distinct intimin types and subtypes have been established (e.g. Yamamoto et al., 2009). The complete genome sequence of the EPEC E2348/69 prototype strain was reported in 2009 (Iguchi et al., 2009) and – among other insights – provided a clear picture of the core LEE.

Atypical enteropathogenic E. coli (ATEC)

Atypical enteropathogenic E. coli (aEPEC, ATEC) by definition are EPEC that lack the EAF plasmid including the EAF-encoded bundle-forming pili and the per (plasmid-encoded regulator)-mediated regulation of T3SS-dependent virulence factors. This results in a distinct adherence phenotype (ATEC: mostly diffuse adherence versus EPEC: localized adherence) and a moderate pathogenicity. As ATEC strains have been isolated from diarrhoea patients as well as healthy individuals their pathogenic potential has long been controversial. However, ATEC strains harbour the LEE and secrete and inject effector proteins into target cells (Hernandes et al., 2009). More recent epidemiology clearly demonstrated that (some) ATEC strains can cause prolonged diarrhoea and are a quite frequent cause of diarrhoea in children (Afset et al., 2006; Scaletsky et al., 2009; Spano et al., 2009). The heterogeneity of ATEC strains is reflected by the more than 200 O-serogroups that have been identified so far (Tennant et al., 2009). Interestingly, most do not belong to the classical EPEC O-serogroups and many have been designated Ont (non-typeable). The most common O-serogroups are O51>O145>O26, O55 and O111. These observations suggest that ATEC represent most likely not a subgroup of EPEC generated by the putative loss of the EAF plasmid.

Enterohaemorrhagic E. coli (EHEC)

In a seminal discovery, Riley et al. (1983) identified the – at the time – rare E. coli O157 : H7 as the causative agent in an outbreak of bloody diarrhoea. Other researchers demonstrated the production of verotoxin (Shiga toxin) (see Johannes and Römer, 2010) and Karmali et al. reported the link between O157 strains and the haemolytic-uraemic syndrome (HUS) (Karmali et al., 2010). O'Brien and colleagues demonstrated the presence of homologues of the Shigella dysenteriae 1 toxin in these E. coli strains (O'Brien et al., 1984). These strains have been denoted enterohaemorrhagic E. coli (EHEC) and are predominantly distinguished from EPEC strains by the presence of stx genes. EHEC are further subgrouped in sorbitol-fermenting SF-EHEC and non-fermenting or typical strains. Among the more than 160 serogroups identified in EHEC in addition to the classic O157 serogroup, O26, O111, O103 and O145 have been recognized as important causes of enterohaemorrhagic colitis and HUS.

The locus of enterocyte effacement – LEE

EPEC, ATEC and EHEC share the A/E lesion formation as one of the most characteristic traits of pathogenicity (Croxen and Finlay, 2010). These bacteria bind intimately to intestinal epithelial cells, which leads to a localized effacement of absorptive microvilli and the accumulation of host cytoskeletal proteins just beneath the attached bacteria. This results in the development of ‘pedestals’ (Fig. 1). The capacity for A/E lesion formation is encoded mainly on the LEE, a pathogenicity island of non-E. coli origin that had been identified in 1995 (McDaniel et al., 1995; Elliott et al., 1998). The core LEE harbours the genes for a type III secretion system (T3SS), regulators, chaperones and effector proteins. Reminiscent of a ‘molecular syringe’, effector proteins are injected through the T3SS needle directly into the cytoplasma of the target cell where they subvert the cells cytoskeletal and signalling machinery. Apparently, T3SSs are very successful pathogenicity modules as besides E. coli they have been found in many Gram-negative animal and plant pathogens including Salmonella, Yersinia, Shigella, Pseudomonas, Xanthomonas, etc. T3SSs are fascinating molecular machines and have been studied structurally and mechanistically in great detail by many laboratories (addressed recently, e.g. by Müller et al., 2008; Enninga and Rosenshine, 2009). Genome sequencing of the prototype O157 : H7 strain identified a secondary T3SS (E. coli type III secretion system 2 – ETT2) present in almost all commensal and pathogenic E. coli strains. The ETTS gene cluster harbours homologues of genes found in Salmonella Spi-1, -2 and -3 pathogenicity islands. However, in almost all E. coli strains the ETT2 T3SS is non-functional due to varying degrees of mutational attrition (Makino et al., 2003; Ren et al., 2004).

The genes encoding structural proteins of the T3SS are largely conserved whereas genes encoding effector proteins show substantial variability (Müller et al., 2009). Castillo et al. (2005) pointed out that the conserved T3SS gene cluster in the LEE appears to have been acquired by horizontal gene transfer while genes encoding secreted proteins are more diverse and might have been obtained by distinct events. The LEE is integrated in the chromosome at either one of the tRNA sites of selC, pheU or pheV and one further site that has not been identified yet.

LEE flanking regions

The prototype EPEC E2348/69 strain harbours the core LEE of 35.4 kb whereas subsequent studies showed that the size of the LEE due to the varying size of its flanking sequences might reach up to 110 kb in other EPEC, ATEC and EHEC isolates (Gärtner and Schmidt, 2004; Müller et al., 2009). Examples of genetic elements integrated in the flanking regions include IS elements (e.g. IS2, IS3, IS629 and IS630 homologues), prophages (e.g. CP4-44, 933L), novel effector genes (rorf0/ibe), etc. (Fig. 2). Several strains have been found to harbour IS elements on both sides of the core LEE (Müller et al., 2009). The newly identified effector protein Ibe (Buss et al., 2009) is encoded 5′ of the LEE core region. In the 3′ LEE flanking regions the lifA/efa gene encoding lymphostatin was identified regularly indicating that these regions might contribute to pathogenicity, e.g. by facilitating colonization of intestinal sites (Klapproth et al., 2005; Babbin et al., 2009; Deacon et al., 2010). The lifA region of several strains contains ent, nleA and nleB genes in the same order. The G+C content of the lifA region (ranging from 42.9% to 44.4%) differs from the E. coli chromosome (50.8%) and the LEE core region (38.4%). Therefore, a simultaneous integration of the LEE and various flanking regions as one genetic element appears rather unlikely (Müller et al., 2009).

Figure 2.

Schematic overview of flanking regions associated with LEE pathogenicity islands of EPEC, ATEC and STEC strains. This scheme depicts the varying sizes of flanking regions associated with the core LEE in human and animal pathogenic EPEC, ATEC and EHEC strains. The positions of IS elements, prophage genes, the LifA/Efa1 regions and the rorf0 (ibe) genes are indicated. The three known chromosomal insertion loci are given. (Modified with permission from Müller et al., 2009.). BSTEC: bovine STEC; REPEC: rabbit EPEC.

Pedestal formation

Actin reorganization leading to pedestal formation is regarded as one of the hallmarks of A/E E. coli pathogenicity. In contrast to intracellular pathogens A/E E. coli are unique in subverting the host actin cytoskeleton from the outside making infection with EPEC a welcomed tool for cell biologists. In EPEC this is a dynamic process resulting in considerable movement of the bacteria on the surface of epithelial cells. Many laboratories have focused on unravelling the signals and mechanisms leading to actin polymerization and pedestal formation and many host factors involved in actin assembly have been localized to pedestals (Bhavsar et al., 2007).

However, in triggering the actin assembly EPEC and EHEC show interesting differences despite the homology between the translocated intimin receptor (Tir) proteins that are essential for signal induction (Sal-Man et al., 2009). Tir is a T3SS-injected protein that following injection inserts into the cytoplasmic membrane as a ‘hairpin’ leaving its N- and C-terminus in the cytoplasma. The central part of Tir interacts with intimin thereby mediating intimate attachment of the bacteria. Interestingly, even isolated (recombinant) Tir protein is able to functionally integrate into the target cell membrane – without the need for adherent bacteria. This unexpected property of Tir might broaden the range of affected cells and potentially also enhances intimin-mediated attachment (Michgehl et al., 2006). EPEC mediate actin polymerization by triggering Arp2/3-mediated nucleation through members of the Wiskott–Aldrich Syndrome Protein family (WASP). With modifications this pathway is exploited by several pathogens (such as EPEC, Vaccinia virus, Shigella flexneri, Listeria monocytogenes, Rickettsia) (Bhavsar et al., 2007; Sal-Man et al., 2009). EPEC activate the Nck pathway that is triggered by phosphorylation of Tir-Y474 by host kinases (Bommarius et al., 2007). This creates a docking site for Nck that in turn mediates activation of WASP proteins. The Src-family kinase c-Fyn was found to be essential for TirY474 phosphorylation following its transient activation induced by intimin–Tir clustering in lipid rafts (Hayward et al., 2009). Recently, it has been reported by the Frankel group that – prior to pedestal formation – the WxxxE effector Map (Kenny et al., 2002; Bulgin et al., 2010) induces transient filopodia formation before robust actin polymerization is triggered (Berger et al., 2009). A second Nck-independent pathway requires Y454 (in EPEC E2348/69) to be phosphorylated (Campellone and Leong, 2005). This facilitates binding to a second adaptor protein that, however, remains elusive and needs to be identified.

Despite the homology between EHEC and EPEC Tir, the EHEC-Tir protein is not interchangeable with EPEC-Tir. For pedestal formation in EHEC the non-LEE-encoded effector protein EspFu (E. coli secreted protein F-like protein from prophage U, also known as TccP) is essential. A 12-amino-acid sequence in the C-terminus of EHEC-Tir recruits EspFu to the sites of EHEC adherence (Allen-Vercoe et al., 2006; Campellone et al., 2006). EspFu belongs to the growing family of effector proteins that are encoded outside the LEE (non-LEE-encoded effector proteins – NLEs) but nevertheless are injected into target cells by the T3SS of the LEE. Apparently, EspFu activates N-WASP by direct binding to the autoinhibitory segment of the GTPase-binding domain (GBD) of N-WASP (Sallee et al., 2008; Campellone, 2010). Members of the IRSp53 family – involved in membrane and actin dynamics – directly interact with EspFu and the Tir 454-463 domain which is essential for EspFu-dependent actin assembly (Weiss et al., 2009). Genetic and functional loss of IRSp53 blocks actin assembly and show this protein to be the missing host factor linking Tir and EspFu in EHEC pedestal formation. This is a novel mechanism of activation restricted to the activation of N-WASP with thus far no counterpart in the host cell. Whether IRSp53 is actually the missing link in the Nck-independent second pathway in EPEC infection needs to be seen. Furthermore, Whale et al. (2006) identified several strains with non-O157 EHEC, EPEC and ATEC serogroups, which simultaneously use the TccP and the Nck pathway for induction of actin remodelling – again emphasizing the versatility of these pathogens.

There has been remarkable progress in recent years in our understanding of the intricate molecular mechanisms evolved by A/E E. coli to subvert the host cell cytoskeleton by specifically inducing actin polymerization from their extracellular location. However, the physiological significance of pedestal formation and subversion of the actin network in infection are not really understood. How these perturbations might serve for the benefit of the pathogen has not been resolved besides more general statements that disturbing the cytoskeleton of the target cells affect cell shape, motility and signalling and thereby enhances pathogenicity. Indeed, several reports demonstrated no direct correlation between actin polymerization, adhesion, pedestal formation, effacement of brush border villi, pathogenicity and bacterial colonization in vivo (Bai et al., 2008; Sal-Man et al., 2009). Nevertheless, investigating pedestal formation by A/E E. coli has certainly generated new insights into mechanisms of actin polymerization.

LEE- and non-LEE-encoded effectors

Among the 41 open reading frames of the LEE (EPEC E2348/69) (Elliott et al., 1998; Iguchi et al., 2009) only three are genes encoding translocator proteins (EspA, EspB, EspD) and seven encode effector proteins (Tir, Map, EspB, EspF, EspH, EspZ and EspG). In other A/E E. coli strains, additional LEE-encoded effectors have been identified such as the ‘IQGAP1-binding effector’ protein Ibe (rOrf0). Ibe could be identified mostly in EHEC, less frequently in ATEC followed by EPEC and appears to enhance ruffle formation in target cells (Buss et al., 2009). The classic LEE-encoded effectors have been reviewed recently (e.g. by Kenny et al., 2002; Dean and Kenny, 2009).

Recent studies have shown that the effector repertoire of A/E E. coli pathogens is much larger than previously thought and by no means restricted to LEE-encoded proteins. Following the first non-LEE-encoded effector NleA/EspI in 2004 many additional NLEs have been identified (for review see Dean and Kenny, 2009). In fact, a recent study addressing the T3SS-dependent secretome in the A/E model organism Citrobacter rodentium by employing SILAC (stable isotope labelling with amino acids in tissue culture) identified all previously known effectors and several novel ones. Addressing the effector repertoire of the O157 : H7 (Sakai) EHEC strain, > 60 putative effector genes were identified and 39 proteins were confirmed as translocated effectors (Tobe et al., 2006). These studies exemplified the vast repertoire of injected factors in A/E E. coli that are directly involved in manipulating the host cell.

Functional studies on the effects of translocated effector proteins in A/E E. coli have demonstrated that (probably all) effectors are multifunctional proteins that each manipulate the host cell in several ways. EspB was the first effector shown to be multifunctional as it doubles as translocator being part of the injection pore and as an effector protein in the cytoplasma of the target cell (e.g. Hamada et al., 2010). Interestingly, many translocated effectors show overlapping functions and team-up for cooperative activities – a theme that has been summarized as ‘functional redundancy’ (Dean and Kenny, 2009). Well-described examples are EspF and Map or NleA that affect cellular junctions and induce apoptosis, or the effacement of microvilli that depends on activities of Map, Tir and EspF (Dean et al., 2006; Guttman et al., 2006; Thanabalasuriar et al., 2009). EspH and EspJ team-up to antagonize phagocytosis and opsono-phagocytosis of EPEC by macrophages (Marchès et al., 2008; Dong et al., 2010).

Even though some effectors have been well studied there are still surprises. Just recently, EspF – previously associated with tight-junction disruption and apoptosis – was shown to be the first bacterial protein to target and disrupt the nucleolus through a mitochondria-based mechanism that is effective late in EPEC infection (Dean et al., 2010a). NleE – previously associated with polymorphonuclear transepithelial migration – and NleB – associated with C. rodentium colonization of murine intestine – were found to interfere with innate immune responses by inhibiting NF-κB activation by blocking the transfer of NF-κBp65 into the nucleus (Nadler et al., 2010; Newton et al., 2010).

The activities of the effector proteins investigated thus far clearly show that A/E E. coli modulate a plethora of host cell functions. Intriguingly, there are already numerous examples that these pathogenic bacteria simultaneously co-inject effector proteins to moderate and counterbalance exceedingly destructive activities. If effectors such as EspF induce apoptosis (Dean et al., 2006), another effector such as NleH inhibits apoptosis (Hemrajani et al., 2010). Likewise, Tir induces pedestal formation while EspM inhibits pedestal formation. However, EspM induces quite dramatic changes in the cytoskeletal architecture of epithelial cells and disorients tight junctions. Surprisingly, this does not affect barrier function and survival of the cells (Simovitch et al., 2010). This ‘balanced pathogenic impact’ of A/E E. coli is further illustrated by recent studies addressing the functions of EspGs and EspZ. The two EPEC effectors EspG and EspG2 that induce a detaching phenotype in host cells by activating the host cell protease calpain were found to be kept in check by the co-delivered Tir protein. Lack of Tir protein results in rapid host cell loss and necrosis indicating that this activity of the Tir protein is needed to maintain epithelial integrity. This study not only identifies an additional function of the Tir protein but also provides a further example of balanced pathogenicity (Dean et al., 2010b). Recently, the host binding partners and function for the LEE-encoded effector EspZ/SepZ were identified (Shames et al., 2010). These authors could show that EspZ enhances the phosphorylation of focal adhesion kinase (FAK) and AKT during EPEC infection thereby contributing to host cell survival mechanisms.

A/E E. coli pathogens use their portfolio of LEE-encoded and non-LEE-encoded effector proteins to subvert and modulate cellular and barrier properties of the host in a well-controlled manner. In addition to the well-documented hierarchy in the regulation of LEE-encoded gene expression (Mellies et al., 2007; Humphries et al., 2010; Wang et al., 2010) there is now growing evidence that – following injection – there appears to be a second level of control as these processes might even be functionally equilibrated among the co-injected effectors in the host cell cytoplasma to ensure balanced pathogenicity. The interactions between the expanding repertoire of injected virulence factors will provide fascinating insights into the subtle balance in the interaction with target cells, will provide additional tools to study signalling processes in the host cell, and, furthermore, might identify putative targets for intervention.

Pathogroup discrimination and putative implications for diagnostics and therapy

The relationship and putative interconversion of the three A/E E. coli pathogroups has been studied by several laboratories (e.g. Mellmann et al., 2009). There are many differences between EPEC and EHEC isolates beyond the expression of phage-borne Stx and its subtypes, such as the presence of pathogenicity islands, the adaptation to different hosts, and their complex interactions with host barriers resulting in either asymptomatic or overt infections. ATEC strains represent a rather heterogenous group and therefore conclusive characteristics beyond the lack of the EAF plasmid distinguishing ATEC strains from classical EPEC and the lack of stx genes to set them apart from EHEC/Shiga toxin-producing E. coli (STEC) are missing. Indeed, some ATEC O26 strains have been isolated that carry virulence and fitness modules of EHEC (Bielaszewska et al., 2007). A recent study describing stx-/EAF- A/E E. coli strains isolated from patients with bloody diarrhoea based on multi-locus sequence typing (MLST), serotype and virulence profiles identified these isolates as EHEC (Bielaszewska et al., 2008) that have lost the stx genes during infection. These strains were consequently termed ‘EHEC-LST’. The closer relatedness of (at least a fair number of) ATEC strains to EHEC rather than to EPEC is further emphasized by studies of strains isolated from cattle and sheep – the common host of EHEC (Cookson et al., 2010). Loss and gain of genes during E. coli infections is a common event in the human intestine (Mellmann et al., 2009) and may result in the interconversion of EHEC to apparent ATEC strains. In addition, transfer of virulence factors through mobile elements and horizontal gene transfer further complicates the characterization of distinct intestinal E. coli pathogroups. This has been exemplified by the identification of ‘intermediate strains’ expressing a profile of indicator virulence factors that does not allow for a reliable assignment of a distinct pathovar (Müller et al., 2007). Therefore, for reliably assessing the virulence potential of a given isolate, the identification of a few ‘indicator virulence factors’ is surely not enough. Here, a considerable extension of the tested matrix of ‘diagnostic’ genes might be a first improvement. In addition, in ATEC strains the serogroup should be determined as well. Identification of a common EHEC serotype could be indicative of a potential risk for the development of HUS.

In summary, recent studies addressing A/EE. coli have underlined the enormous plasticity of the E. coli genome and have again clearly shown that bacterial chromosomes are not static and fixed entities. Moreover, the detailed studies on the molecular mechanisms of effector interactions with host cells have led to fascinating insights into the intricate subversion strategies developed by these seemingly common organisms resulting in a well-balanced pathogenicity.


Work in the author's laboratory has been funded by grants from the DFG (SFB293, SFB629, Graduate School GRK1409), the Interdisciplinary Center for Clinical Research (IZKF) Münster (SchMA2/027/08), and a grant from the EU Network ERA-NET PathoGenoMics (No. 0313937C). Electron microscopy by Lilo Greune (Infectiology, Münster) is gratefully acknowledged.