EspP consists of a 55 aa N-terminal signal peptide (SP), the secreted passenger domain (PD, amino acids 56–1023) and the C-terminal β-domain, also called translocator, which is 277 amino acids in length [1] (see Figure 1). This structural organization is seen in all SPATEs. The serine protease motif is located in the PD (see 2.4), while the β-domain is responsible for efficient autotransport through the outer bacterial membrane (see 2.3). The C-terminal part of the PD harbors the approximately 30 aa linker-region that connects the β-domain and the PD and is necessary for folding and stability of the β-domain [51,52,53]. In addition, the autochaperone motif (AC) at the passenger C-terminus is essential for folding of the β-helical structure of the passenger domain [54,55].

At least seven different pathways exist for the protein transport in Gram-negative bacteria [56]. Autotransport, also referred to as Type V pathway, is probably the most widely distributed mechanism and is divided into classical autotransport (also known as Type Va), the two-partner secretion system (Type Vb), and the trimeric autotransporters (TAA, Type Vc) as well as the two recently described autotransporter Types Vd and Ve [57]. Briefly, the transport through the inner membrane (IM) is mediated by the N-terminal signal peptide via the Sec-machinery in the Type V pathway and the C-terminal β-domain is inserted into the outer membrane (OM) and facilitates secretion of the passenger domain into the extracellular space. EspP is transported by the Type Va secretion system, and it has to be noted that biogenesis via Type Va secretion apparently shows some diversity in SPATE. We will therefore limit the discussion to novel mechanistic details of EspP biogenesis, which not necessarily fully applies to other SPATE.

EspP has an unusually long SP which is found in about 10% of autotransporters and which has been shown to facilitate posttranslational targeting [58]. The C-terminal part resembles a typical SP with canonical N, H, and C regions (termed N2, H2, and C) while the N-terminal part harbors a unique sequence motif (N1 and H1) that is highly conserved among SPATEs from E. coli and Shigella sp. [59,60]. The N-terminal extension is not required for efficient translocation across the IM. However, modifications of the SP lead to misfolding of EspP in the periplasm and inhibition of translocation across the OM. The SP transit via the Sec pathway is relatively slow and it is believed that this is due to tethering of the PD to the IM [61] and a slow dissociation from the IM probably due to an unusual conformation of the SP [58]. EspP interacts with several chaperones in the periplasm, namely SurA, Skp, DegP, FkpA, and the Bam complex [7,8,9]. A direct protein-protein interaction of SurA, DegP [8] and FkpA [9] has been observed with the unfolded but not with the native folded PD of EspP, and it has been shown that surA and skp mutants moderately reduce secretion while in a degP mutant translocation was abolished [8]. Interestingly, DegP acts as a chaperone and as serine endoprotease that degrades unfolded proteins. Interaction with chaperones probably prevents misfolding and keeps the PD in a loosely folded translocation-competent state [8,53] and probably prevents DegP-mediated degradation. Skp binds to the PD before translocation while SurA binds during the transit across the OM through a pore inside the β-domain [10]. The β-domain is highly conserved among SPATEs [20] and consists of a 12-stranded β-barrel, a short N-terminal α-helix, and a short linker loop [62] (Figure 2).

The β-barrel domain shows considerable tertiary structure in the periplasm and during membrane integration [53], which might also be achieved by interactions with periplasmic chaperones. As with the PD, SurA, DegP [8], and Skp [10] interact with the β-domain. In addition, the Bam complex catalyzes protein assembly into the OM [7]. Ieva et al. suggested that Skp interacts with EspP on the periplasmic side as it translocates the IM and transfers it to SurA. SurA subsequently binds to BamA, which together with BamB and BamD interacts with the β-domain. The lipoprotein components of the Bam complex facilitate the assembly into the OM. Notably, secretion of the PD is initiated before β-domain assembly is complete [10]. The PD is translocated in a C- to N-terminal direction across the β-barrel domain [7]. When a C-terminal ~17 kDa segment is exposed to the cell surface folding of the PD is initiated. This traps the PD outside the cell and also provides at least part of the energy necessary for the translocation process. Folding might be the rate-limiting step in EspP biosynthesis [63]. After complete translocation of the PD to the extracellular milieu, the PD is cleaved from the β-domain within the pore of the β-barrel in the OM [10,62,64]. Passenger domains can be released from the β-domain by extracellular proteases or—as in the case of EspP—autocatalytically. EspP is cleaved between Asn¹⁰²³ and Asn¹⁰²⁴, a cleavage site that is highly conserved among SPATEs [55,64]. In the first mechanistic study, it has been suggested that the carboxyl group of Asp¹¹²⁰ activates Asn¹⁰²³ within the β-barrel pore [64]. The Asn¹⁰²³ amide group mediates a nucleophilic attack on the scissile bond between Asn¹⁰²³ and Asn¹⁰²⁴ thereby forming a succinimide that is further hydrolyzed into Asn and iso-asparagine via a cyclic intermediate [64] (Figure 3b). Asparagine cyclization is in general slow in peptides, underlining the catalytic mechanism of this reaction. Notably, residues important for autoproteolysis are conserved in all SPATEs suggesting a common cleavage mechanism. Barnard et al. demonstrated that the conserved residues E¹⁰²¹–L¹⁰³² (which are part of the α-helix that spans the PD-β-domain junction) constrain the active site asparagine to conformations favorable for cyclization. In addition to steric effects, positively charged residues in the α-helix interact with negatively charged residues on the barrel wall facing the lumen. This leads to formation of a salt bridge that is essential for the cleavage reaction [65]. Mutations inside the α-helix abolished autocatalytic cleavage whereas passenger translocation of EspP was not affected [66]. However, in Hbp/Tsh comparable mutations impaired both cleavage and translocation [67]. Interestingly, a recent study has demonstrated that in Hbp/Tsh the active Asn¹¹⁰⁰ (which is equivalent to Asn¹⁰²³ in EspP) is not activated by the respective Asp¹¹⁹⁷ residue [68]. Instead, Tyr¹²²⁷ and Glu¹²⁴⁹ position a water molecule in proximity to Asn¹¹⁰⁰, resulting in deprotonation and activation of this residue [68]. A similar model has been proposed recently for EspP, where the nucleophilicity of the active site asparagine is enhanced by proton abstraction via water molecules that are positioned by the conserved acidic residues Asp¹¹²⁰, Glu¹¹⁵⁴ and Glu¹¹⁷² [65]. The resulting oxyanion intermediate is stabilized by protonated Glu¹¹⁷² [65]. The newly formed N-terminus of the β-barrel (Asn¹⁰²⁴) interacts with an acidic cluster and the α-helix and the linker loop block access from the periplasmic space, which probably protects the bacterium from uncontrolled influx and efflux [62]. Whether the recently identified translocation and assembly module (TAM) that contributes to secretion of the autotransporters Ag43, EhaA, and p1121 [11] also is involved in secretion of EspP remains to be investigated. The complete secretion of EspP is illustrated in Figure 3.

The espP gene has been first described on the large plasmid of O157:H7 EHEC strain EDL 933 [1] and since then, distribution and prevalence in different E. coli populations has been investigated intensively. In general, espP is strongly associated with STEC, EHEC and atypical enteropathogenic E. coli (EPEC) and has been detected only very rarely in other E. coli pathotypes and, to the best of our knowledge, not in commensal E. coli [18,69,70]. The espP gene was found in one study in 3.6% of the analyzed EAEC strains [18] and in 1.4% of typical EPEC [70]. Generally speaking, highest prevalences of espP have been seen in classical EHEC serotypes such as O157:H7, O26:H11/NM, or O145:H25/H28/NM [49,50,71,72,73,74]. In addition to pO157, espP has been described on the virulence plasmids pO113 [22], and pO26-Vir [75]. Interestingly, espP is not found in sorbitol-fermenting (SF)O157:NM and sequence analysis of the respective virulence plasmid has revealed that pSFO157 encodes the sfp fimbriae gene cluster instead of espP [76]. Similarly, Nagano and coworkers showed that β-glucuronidase (GUD⁺) NSF O157:H7 also do not harbor espP [77]. The respective large plasmid in these strains (pO157_2) also lacks espP [78].

Besides the variable distribution of espP in virulence plasmids, different espP alleles have been described. Brunder et al. first showed seven different restriction types for espP PCR products in a limited strain collection indicating a certain degree of heterogeneity [79]. A systematic analysis of 98 espP-positive EHEC and STEC strains from 56 different serotypes identified four different alleles, namely espPα, espPβ, espPγ, and espPδ [49]. The four alleles have been found in 17, 16, 15, and eight serotypes, respectively. Notably, the highly virulent serogroups O157, O26, O111, and O145 all exclusively harbored espPα. The encoded proteins EspPα and EspPγ are secreted (see 2.2) and proteolytically active (see 2.4) while EspPβ is either not secreted or proteolytically inactive, and EspPδ is secreted but proteolytically inactive [49], demonstrating significant functional differences of espP alleles on protein level. Based on the allele-specific PCR developed in this study, Khan et al. investigated 121 espP-positive strains belonging to 61 different serotypes. The serotypes O157:H7/NM, O26/H11, and O145:NM again all contained espPα and serotypes with lower virulence harbored espPβ or espPγ, espPδ has not been found in the strain collection [50]. Furthermore, espPα was more prevalent in human isolates (84%) than in environmental isolates (47%), while espPγ was more present in the environment (40%) than in humans (11%) [50], indicating that the environmental reservoir of espP-positive strains might differ on subtype-level. It has been suggested that further espP alleles might be found due to additional recombination events [49]. In accordance, Bielaszewska et al. described a new espP allel, espPε, that was found in one E. coli O91:H8 and 49 of 77 O91:H14/Hnt/NM [80].

Several studies have investigated the correlation of espP in human isolates with disease [72,74,81] and suggested that there is no significant correlation between espP and clinical outcome, while others found differences in prevalence of the espP gene in clinical isolates and environmental samples [82]. These inconsistent results underline that it is not possible to assign the virulence potential of espP without taking into account the functional differences of EspP subtypes. Accordingly, Khan et al. found the espP gene in 65% of isolates from patients with bloody diarrhea or HUS and in 67% of isolates from patients with watery diarrhea or without any symptoms at all. On subtype level however, all espP-positive strains from HUS patients harbored espPα, indicating a potential role of this subtype in EHEC pathogenicity [50]. It has however to be kept in mind that EHEC quite likely have developed different pathways of host injury.

Concerning the different animal and environmental reservoirs, espP is widely distributed and shows a certain association with cattle reservoirs. Toszeghy et al. found the espP gene in 22 of 101 bovine isolates but in none of 26 and 19 isolates of chicken and turkey, respectively [83]. Horcajo et al. found espP in 60.3% of EHEC and atypical EPEC strains in cattle but just in 11.7% of strains from sheep and goat [84]. In accordance, several other studies showed the association of the espP gene with bovine isolates with a prevalence of 23.8% to nearly 100% [73,81,84,85,86,87,88,89,90,91,92]. Furthermore, espP has been found in E. coli isolates from food samples like raw-milk, cheese [82,93], meat products [94,95], drinking water, tea [94], and ground beef [96] as well as in environmental isolates, e.g. from soil samples, ground, cattle water troughs, and feeders [50,82,97,98]. As indicated, reservoirs of espP-positive strains might differ on subtype-level. However most studies have only determined the presence of espP and not performed subtype-specific analysis.

The passenger domain of EspP harbors a serine protease function as seen by the typical serine protease sequence motif GDSGSPLF and by the fact that it is inhibited by addition of phenylmethanesulfonylfluoride (PMSF), a specific serine protease inhibitor, but not by addition of ethylenediaminetetraacetic acid (EDTA), an inhibitor of metalloproteases [1]. The catalytic triade has been determined by site-directed mutagenesis and consists of His¹²⁷, Asp¹⁵⁶, and Ser²⁶³ [2].

At present, only a limited number of substrates has been identified. These are coagulation factor V, porcine pepsin A [1], apolipoprotein A-I [26], complement factor C3/C3b and C5 [99], and EHEC-hemolysin [100]. In addition, a casein-based assay has been used to determine proteolytic activity of the homologue PssA [3]. Furthermore, EspP has been shown to cleave the synthetic peptides Suc-Ala-Pro-Leu-pNA (Suc, succinic acid, pNA, para-nitroaniline) and Arg-Arg-pNA [45]. However, the latter could not be reproduced by our own studies and also no EspP substrate has been described that is cleaved C-terminal to Arg. Several physiological relevant proteins have been shown to resist EspP-mediated degradation namely human IgA, a human myeloma IgA1 preparation, bovine serum albumin, alpha-2-macroglobulin, transferrin, lactoferrin, pepsinogen [1], spectrin, bovine submaxillary mucin [45] as well as complement factors H and I [99].

The cleavage of complement factors has been studied in more detail. Orth et al. demonstrated that EspPα cleaves complement factors C3/C3b and C5 by incubation of purified proteins and human serum with EspPα and incubation of purified complement proteins with EHEC culture supernatants. Factor C3b was degraded into three major fragments with molecular masses of 42, 38, and 37 kDa, respectively, while the degradation of factor C5 into three fragments with molecular masses of 39, 37, and 33 kDa was found to be less pronounced. C5-depleted serum that was supplemented with purified C5 preincubated with EspP showed significantly reduced complement activation in all three activation pathways (classical pathway, alternative pathway, mannan-binding lectin pathway) [99]. The complement system plays a crucial role in the host response of the innate immune system and downregulation of complement activation might consequently impair host defense in EHEC infections. It has been suggested that complement downregulation might protect EspPα-secreting EHEC as well as host cells from opsonization, complement-mediated lysis, and inflammatory events [99]. Cleavage of C3 has been shown for several other bacterial proteases. These include dentilisin from Treponema denticola [101], interpain A from Prevotella intermedia [102], extracellular gelatinase from Enterococcus faecalis [103], streptococcal pyrogenic exotoxin B from Streptococcus sp. [104], as well as proteases from Porphyromonas gingivalis [105], and Pseudomonas aeruginosa [106]. Cleavage of C5 has been shown by the 56-kilodalton protease from Serratia marcescens [107], gingipain-R and gingipain-K from Porphyromonas gingivalis [108], and streptococcal C5a peptidase from Streptococcus sp. [109]. A potential role of EspPα is underlined by the finding that specific antibodies have been isolated from sera of HUS patients [1,3].

Another potentially relevant substrate in vivo might be coagulation factor V (FV), which is present in plasma and in α-granules of platelets. FV plays a role in hemostasis and can act either as procoagulant or as anticoagulant (reviewed in [110]). Brunder et al. have suggested that this degradation could lead to decreased coagulation possibly leading to prolonged bleeding and increased hemorrhage in the gastro intestinal tract that can be observed during EHEC infection [1]. Together with further virulence factors such as EHEC-Hly [111] this might lead to improved supply of iron via hemoglobin. The ability to degrade FV is widely distributed within the SPATE family. Pet and Pic from EAEC, Sat from UPEC, and EspC from EPEC also cleave this protein. Also, Tsh from avian pathogenic E. coli was found to degrade FV when a purified protein was used but not in human plasma [45]. Unfortunately, functional investigations have not been performed so far.

Apolipoprotein A-I (apo A-I) has been reported to be cleaved by EspPα and EspI, another SPATE, which is also found in EHEC [26]. However, degradation has not been investigated on a functional level. Apo A-I is known to be involved in lipid binding and transport [112]. More interestingly, Concha et al. suggested that during bacterial infection apo A-I could release antimicrobial peptides from HDL (high density lipoprotein) particles after proteolysis [113]. Furthermore, apo A-I is a negative acute phase protein with anti-inflammatory properties [114] that can directly bind lipopolysaccharides (LPS) [115] and is capable of inactivating endotoxins and decreasing LPS-induced cytokine release [116,117]. Therefore, degradation of apo A-I by EspPα might either activate or inhibit its antimicrobial function and could interfere with inflammatory processes during EHEC infection. Again, experimental data in model systems concerning the functional implications of apo A-I cleavage are missing.

In addition, it has been demonstrated that EspPα cleaves EHEC hemolysin (EHEC-Hly) in its free and vesicle-bound form [100]. EHEC-Hly belongs to the repeat-in-toxin (RTX) family [118] and is able to lyse erythrocytes and lymphocytes [118,119] and injures microvascular endothelial cells [120]. Functional characterization was performed using purified proteins, supplementation of purified EspPα to culture supernatants of EHEC-Hly expressing strains, and coexpression of both proteins [100]. Cleavage of EHEC-Hly occurs in the hydrophobic domain which is important for the cytolytic activity of RTX toxins [121] and consequently leads to loss of hemolytic activity. In a cellular infection model using human brain microvascular endothelial cells (HBMEC), coexpression of EspPα and EHEC-Hly in recombinant E. coli resulted in basal HBMEC cytolysis whereas pronounced dose-dependent cytolysis was observed in controls [100]. It has been proposed that pathogenic E. coli might be able to regulate their virulence phenotypes by interference of effector molecules [100]. While cleavage and inactivation has been, to the best of our knowledge, only described so far for EspP and EHEC-Hly, other RTX toxins are activated by proteolytic cleavage. For example, various proteases of Vibrio cholerae cleave and activate the El Tor cytolysin/hemolysin [122]. Another example is the proteolytic activation of Shiga toxins by eukaryotic proteases [123,124].

Khan et al. solved the 2.5 Å crystal structure of the EspP passenger domain using the proteolytically inactive mutant S263A [125]. It consists of a large β-helical stalk and a globular subdomain that harbors the catalytic triad (see Figure 4). The overall fold of EspP is typical for known autotransporters and resembles that of E. coli hemoglobin protease (Hbp) [126] and Haemophilus influenza immunoglobulin A protease (IgAP) [127].

The globular subdomain shows similarity to the chymotrypsin family. However, the S1 subsite of EspPα is narrower and shallower as well as slightly more hydrophobic than that of bovine chymotrypsin. Therefore, Khan et al. suggested that EspPα cleaves after smaller and more hydrophobic amino acids than chymotrypsin, which cleaves after tyrosine, tryptophane, phenylalanine, leucine, and methionine. This suggestion is in good accordance with the cleavage sites that have been determined so far where EspPα mainly cleaves after Leu [1,45,99,100]. When compared to Hbp and IgAP, the active site cleft of EspPα is slightly wider and much deeper indicating different substrate specificities. Furthermore, it is more exposed than that of Hbp and much more exposed than that of IgAP, leading to the suggestion that EspPα interacts with a larger surface of substrates or with larger substrates in general [125]. This is in accordance with the finding that EspPα is able to cleave porcine pepsin A but not its precursor pepsinogen [1] suggesting that not only the amino acid sequence but also the three-dimensional structure might be important for substrate recognition. This might be amongst other effects mediated by the β-helical stalk as proposed by Khan et al. [125].

Taken together, EspPα seemingly shows high substrate specificity. There are only a few known human substrates, which are related to blood coagulation and immune response. Cleavage of these substrates might contribute to the severity of EHEC infections. Also, EspP might be involved in regulation of E. coli virulence as shown by cleavage of EHEC hemolysin. Cleavage predominantly occurs after leucine at position P1 and other small hydrophobic amino acids at position P2.

Cytotoxicity of EspP is discussed controversially. Incubation of Vero cells with culture supernatants of PssA secreting E. coli strains resulted in cytotoxic effects as analyzed by fluorescence microscopy with phalloidin staining. Dependent on incubation time the cells showed defects in cell-cell junctions, retraction of cell bodies, and loss of stress fibers (5 h), with disruption of the actin cytoskeleton in about 30% of cells, loss of cell-cell junctions, detachment from the substratum (10 h), and cell death at later time points [3]. It was suggested that these effects might be due to apoptosis, however cytotoxic pathways have not been analyzed. In contrast, Dutta et al. were not able to reproduce any cytopathic effects on Vero cells after incubation with 0.5 and 1 μM EspP for 5–30 h. In addition, no cytotoxic effects have been shown for the epithelial cell line HT-29 and HEp-2 in this study [45]. In our own studies, we neither found any cytotoxic effects in epithelial HT-29 cells nor in endothelial HBMEC after incubation for 30 h or 48 h with up to 1 μM EspPα (Brockmeyer, unpublished data). A recent study used the epithelial HeLa cells and observed binding of EspPα after 6 h of incubation in a dose-dependent manner [129]. Prolonged incubation (24 h) with EspPα at higher concentrations lead to cell rounding in this study. Since EspPα did not induce LDH release, Xicohtencatl-Cortes et al. stated that it showed cytopathic but not cytotoxic effects. Again, potential underlying mechanisms have not been analyzed.

Initial motivation to analyze the potential of EspPα to contribute to cellular adherence or biofilm formation was raised by the finding that pO157 is required for full adherence of E. coli to Henle 407 intestinal and HEp-2 epithelial cells [130] and influences the colonization of the bovine terminal rectum [131]. In addition, pO157 is important for biofilm formation and adherence to the epithelial T84 cells [132]. Puttamreddy et al. generated a transposon insertion library in E. coli strain EDL933 and identified 51 distinct genes or intergenic regions responsible for biofilm formation. Among the genes responsible for full biofilm potential is espPα which also contributed to adherence of T84 cells [133].

A potential mechanism for biofilm formation and adherence mediated by EspPα has been published recently by Xicohtencatl-Cortes et al. [129]. Under certain conditions, EspPα is able to form macroscopic rope-like structures via oligomerization. These ropes are up to 2 cm in length, and show pronounced stability against mechanical stress, high temperature, and detergents [129]. Under laboratory conditions, the ropes formed a substratum for bacterial biofilm formation and were able to bind exogenously added bacteria [129]. Bacteria associated with the ropes were protected from antimicrobial drugs, indicating a potential protective role for the bacterial population within ropes. Like monomeric EspPα, ropes also bind to HeLa cells [129]. Whether the rope-like structures are formed during infection is still elusive.

Studies regarding espP gene expression are rare and mostly descriptive. Ebel et al. observed enhanced EspP levels in EHEC culture supernatants when bacteria were grown in lysogeny broth (LB) medium rather than in minimal essential medium (MEM), indicating that a medium rich in amino acids leads to higher EspP production [134]. In addition, more EspP is produced at 37 °C than at 20 °C [1,134] and pH 7 and pH 9 are favorable compared to pH 5 at which EspP production is nearly completely abolished [1]. Based on the methodology used by Ebel et al. and Brunder et al. it is not possible to differentiate between effects on gene expression or protein secretion. Enhanced secretion under slightly alkaline conditions indicates that EspP might act on the human colon or in the circulation. Indeed, we have recently shown a potential interaction of EspP with the human intestine [100]. Incubation of EHEC in contact with human intestinal epithelial HCT-8 cells lead to upregulation of espP expression up to more than 35-fold [100].