Type 1 Fimbriae, Curli, and Antigen 43: Adhesion, Colonization, and Biofilm Formation

PER KLEMM^1* AND MARK SCHEMBRI²

[SECTION EDITOR: JAMES P. NATARO]

Posted November 15, 2004

¹Microbial Adhesion Group, Center for Biomedical Microbiology, BioCentrum-DTU, Technical University of Denmark, Bldg. 301, Technical University of Denmark, DK-2800 Lyngby, Denmark, and ²Department of Microbiology and Parasitology, School of Molecular and Microbial Sciences, Bldg. 76, The University of Queensland, Brisbane, Qld 4072, Australia

*Corresponding author. Phone: +45 45252506, Fax: +45 45932809, E-mail: This e-mail address is being protected from spambots. You need JavaScript enabled to view it

INTRODUCTION

In nature, most bacteria live attached to or in close association with surfaces. It has been estimated that at least 90% of all bacteria in the environment reside attached to a surface, and many of these form sessile communities called biofilms (34). Biofilm formation can be divided up in different stages, as follows. (i) Attachment of the bacteria to the surface in question. After the initial attachment, several possibilities are open. If the bacterium is attached to a eukaryotic cell it can sometimes initiate (ii) invasion of that cell. Alternatively, it can follow the path that ultimately leads to (iii) biofilmformation. This usually takes place via an initial expansion step resulting in microcolony formation and subsequently formation of a young biofilm. The final step is maturation into a mature biofilm, i.e., a complex three-dimensional structure. In this chapter we are primarily concerned about the first step in these developments, namely, bacterial attachment to surfaces. Bacterial attachment is a prelude to both invasion and biofilm formation. Without this crucial initial step the other developments will simply not happen.

Bacterial attachment to surfaces is mostly mediated by specific bacterial adhesins, surface structures that typically recognize specific molecular motifs in a lock-and-key fashion. This enables the bacterium to target to a specific surface, for example, a specific tissue like the bladder. This phenomenon is often referred to as tissue tropism. Escherichia coli and Salmonella strains can express a plethora of adhesins with different receptor specificities (80, 143). Individual strains are generally capable of expressing several different adhesins that can target to different receptor molecules. This provides the bacterium with a choice of potential target surfaces. Emerging evidence suggests that the expression of adhesins is often coordinated through inter system cross talk, offering flexibility in niche selection (133, 136). Here we will describe three examples of bacterial adhesins, each of which belongs to a different subgroup and follows different strategies for surface presentation and adhesin exposure. These are type 1 fimbriae, very long stiff rod-like organelles; curli, amorphous fluffy coat structures; and finally antigen 43, short outer membrane structures with a simple assembly system.

TYPE 1 FIMBRIAE

Occurrence

Type 1 fimbriae were originally characterized and defined by virtue of their ability to cause mannose-sensitive agglutination of a variety of eukaryotic cells (40, 41, 42). Type 1 fimbriae constitute the most widespread adhesive organelle among members of the Enterobacteriaceae. Closely related variants of these organelles have been reported to be present on E. coli, Klebsiella spp., and Shigella spp., whereas type 1 fimbriae of Salmonella and Citrobacter spp. constitute a second serological group (3). In E. coli∼80% of all strains have the capacity to express type 1 fimbriae. Arguably, the fact that these adhesive organelles are so common is indicative of important role(s) in the ecology of these bacteria.

Structure and Biosynthesis

A typical type 1-fimbriated bacterial cell has 100 to 500 fimbriae arranged peritrichously on the surface, each with a width of 7 nm and a length varying between 0.2 and 2 μm (Fig. 1). A type 1 fimbria has a tubular structure with a ∼2-nm-diameter hollow core (23, 56). The bulk of the organelle consists of a major building component, FimA (23, 73, 82). FimA monomers are noncovalently associated head-to-tail and organized in a right-helical structure with 3.1 subunits per revolution and a subunit pitch distance of 2.3 nm (23, 27, 56). Adjacent turns of the helix are connected via three binding sites, making the organelle rod rather stiff (56). At the tip of the structure a short, ~16-nm-thin, 2-nm-wide fibrillum is located (69). According to current knowledge the tip fibrillum is made of minor fimbrial components, namely, the FimF, FimG, and FimH proteins (56, 69). FimH is the type 1 fimbrial adhesin (81), and FimF and FimG act as adaptors for integration of the adhesin into the organelle structure (2, 69, 77, 82, 124). The subunits FimG and FimF connect FimH to the FimA rod, the sequential orientation being FimA–FimF–FimG–FimH (56). In addition to a tip fibril position, FimH has been reported to be interspersed along the fimbrial shaft (2, 3, 81). The detailed three-dimensional structure of FimH has been elucidated by X-ray crystallography (27). According to this the protein is folded into two domains, an N-terminal adhesive domain (residues 1 through 156) and a C-terminal organelle-integration domain (residues 160 through 279) linked by a tetra-peptide loop. An isolated truncated version of the N-terminal domain of FimH retains its adhesive faculty (129).

Fig. 1Electron micrograph of E. coli expressing type 1 fimbriae.

Fimbrial biosynthesis takes place by chaperone/usher-assisted assembly (147). FimC is a periplasmic chaperone, while FimD is an usher protein that forms a pore in the outer membrane (74, 76). The individual subunits are secreted into the periplasm via the general secretary pathway and bound by a chaperone to prevent premature self-assembly (143). The chaperone-subunit complex is then targeted to the usher in the outer membrane. Here, the subunit is translocated and incorporated into the base of the growing organelle (148). Biogenesis proceeds via mechanisms called donor strand complementation and donor strand exchange (11, 27). Fundamentally the same principle is employed in Lego building blocks. According to this, a deep hydrophobic groove present on the top of each subunit (except FimH) interacts with a peptide strand present at the base of the next subunit. The minor components, notably FimH, have been demonstrated to be of paramount importance for the initiation of organelle formation and determine organelle length in a dose-dependent manner (27, 77, 124).

Genetic Organization and Regulation of Transcription

The fim genes responsible for type 1 fimbriation are located at 98 min on the E. coli chromosome (24, 46) and at 15 min in the Salmonella chromosome (126). Neither cluster is associated with any known pathogenicity islands. The E. coli fim gene cluster contains nine genes which encode the structural components, the system-specific transport system, and regulatory genes (74, 76, 77, 79, 107). Expression of type 1 fimbriae is phase variable due to a flip-flop-type control system based on an invertible 314-bp DNA switch located immediately upstream of the major subunit gene (1). A promoter located in the switch drives the expression of seven of the fim genes (105). On or off orientation of the switch results in a fimbriated or bald phenotype, respectively. Two tyrosine-class recombinases, FimB and FimE, catalyze the inversion of the switch (75). The FimB recombinase displays bidirectional switching activity (off-to-on and on-to-off) while FimE primarily catalyzes on-to-off inversion (48, 75). FimE can also catalyze off-to-on inversion but only at very low efficiency (144, 145) or as a result of mutations that alter its functional activity (138). Expression of the fimB and fimE genes is driven by individual promoters (105). A number of global regulators including integration host factor (IHF) (22), the leucine-responsive regulatory protein (Lrp) (21), and H-NS (105, 106, 134) also affect inversion of the fim switch. Regulation of fim gene expression was recently reviewed in depth (20). The primary regulation of the fim gene product levels takes place at the transcriptional level (136). Fine-tuning of the organelle structure and composition is modulated by a number of posttranscriptional factors such as different mRNA stability, ribosomal binding sites, and codon usage.

Role as Adhesin

Type 1 fimbriae confer binding to a variety of eukaryotic cells by virtue of their capacity to recognize mannosides (42, 91, 161). The FimH adhesin is responsible for the adhesive properties of type 1 fimbriae and confers a lectin-like binding to various mannosides (81). Although FimH-mediated binding in general is sensitive to the presence of d-mannose, it has become increasingly clear that there exists significant heterogeneity among type 1 fimbriae. The FimH adhesins from different genera and species vary in their binding affinity towards defined oligomannose motifs. Striking differences were reported between the specificity of type 1 fimbriae from E. coli and Salmonella (44). Work by Sokurenko and others has clearly demonstrated that binding differences are due to alterations in the primary structures of different FimH vaiants (139, 140, 141). Other reports, although not contradicting this tenet, indicate that the affinity of FimH can be modified by the fimbrial filament on which it is presented (90, 149). It appears that changes in FimH specificity, whether caused by sequence variation in FimH or by influence from filament presentation, strongly affect target specificity and thereby tissue tropism. FimH variants from commensal isolates have weak affinity for monomannose residues but have strong affinity for complex mannose structures such as terminally exposed trimannose units. On the other hand, most uropathogenic isolates express FimH variants that have strong affinity for both types of targets (139). Minor amino acid sequence alterations in FimH can affect receptor recognition profile profoundly (114, 135, 139, 141, 142). The extent of the FimH target range can also be gathered from the fact that variants of this versatile adhesin are able to bind protein targets such as laminin and collagen (83, 114). The adaptable nature of the FimH receptor affinity was also investigated by creation of FimH libraries based on random mutagenesis of the FimH lectin domain. These studies revealed that FimH can exhibit a vast range of affinities for various targets and that even single amino acid alterations can change receptor specificity (135). The recent discovery of shear-force-induced binding enhancement of FimH further underlines the surprising flexibility of this adhesin. Computer simulations suggest that force-induced conformational changes are responsible for the phenomenon (150). Furthermore, regions in FimH distant from the receptor binding site seem to be involved. It is conceivable that force-induced conformational changes and binding enhancement can actually explain why the same FimH adhesin presented by different fimbrial filaments can adapt different receptor affinities.

Role in Biofilm Formation

Early studies of type 1 fimbriae revealed a connection between expression of these organelles and the ability to form a pellicle during growth in aerobic static broth, i.e., a bacterial surface film on the air-water interface (42). Pellicle formation was observed to be sensitive to d-mannose derivatives indicative of FimH involvement (100), which was subsequently verified by fimH-knockout mutations (59). Under static liquid conditions, type 1 fimbriae-assisted pellicle formation provides a dramatic selective advantage (in the order of 1:10⁶) when comparing Fim⁺ and Fim^– cells, arguably because of facilitated access to atmospheric oxygen (100, 144). In this connection it is important to note that members of the Enterobacteriaceae do have a life outside a mammalian host. Arguably, pellicle formation can provide a selective advantage in vivo under static liquid conditions, i.e., ponds, pools, etc. However, it is also conceivable that pellicle formation provides a means of host-to-host transfer: when mammals defecate in ponds and other stagnant water bodies, the excreted bacteria can subsequently form a pellicle on the surface by virtue of type 1 fimbriae expression. In turn the pellicle will provide a rich opportunity for transfer to a new host when other animals come to quench their thirst and drink from the surface.

A pellicle is essentially what we would now call an air-water biofilm. Seen in this light, it was hardly a surprise that manipulation of the function or expression of type 1 fimbriae turned out to interfere with biofilm formation on various abiotic surfaces such as plastics and glass (115, 131, 132). In line with the lessons learned from fimbriae involvement in pellicle formation, lesions that affected the fim genes or auxiliary functions such as periplasmic disulfide bridge formation were observed to reduce biofilm formation of E. coli on abiotic surfaces (49, 115). Again, the FimH adhesin seemed to play a prominent role since methyl mannoside was reported to inhibit biofilm formation (115).

In nature, bacteria often face brutal hydrodynamic flow-shear forces, for example in the urinary tract. We recently isolated FimH variants from a random mutant library that had the capacity to promote biofilm formation under hydrodynamic flow conditions (132). Some FimH variants were capable of biofilm enhancement even in the presence of methyl mannoside, and some promoted biofilm formation but were incapable of binding to mannoside targets. This suggests that mannose binding and biofilm enhancement can be independent properties of FimH. Several of the biofilm-promoting FimH variants were subsequently shown to confer autoaggregation of cells (127). Interestingly, similar phenotypes were also observed in wild-type variants of FimH originating from uropathogenic E. coli (UPEC) strains; i.e., they conferred biofilm formation and autoaggregation. However, these variants were always capable of binding to mannosides and all of them were sensitive to methyl mannoside (127, 132). Apparently there is strong selection in natural isolates for conservation of mannose recognition (to provide binding to host targets) and the additional ability to promote biofilm formation. Autoaggregation seems to be a trait acquired in, for example, some UPEC strains. Again, these observations underline the highly adaptive nature of the FimH adhesin.

Role in Pathogenesis

As previously mentioned, the ability to express type 1 fimbriae is widespread and a conserved trait throughout the Enterobacteriaceae and underlines the importance of this adhesin in the ecology of these bacteria. In animal studies, type 1 fimbriae seem to play an important role in bacterial dissemination among litter mates. It was found that lesions affecting expression of the fim genes resulted in a dramatic decrease in transmission of E. coli among rats (19). This indicated a critical role for type 1 fimbriae-assisted oropharyngeal colonization in the fecal/oral cycle of enteric bacteria.

The excellent qualities of FimH in assisting bacterial adhesion and biofilm formation also hinted at a role for type 1 fimbriae in bacterial pathogenesis. Several studies in the 1990s strongly implicated type 1 fimbriae as virulence factors in urinary tract infections. Urinary tract infections by UPEC strains affect humans (and other mammals) and account for several million cases annually. Type 1 fimbriae are expressed by more than 90% of all UPEC strains, and they bind to the uroepithelial surface (5, 32, 33). It was demonstrated in a mouse model that type 1-fimbriated strains generally caused more severe infections than their Fim-negative counterparts and that inactivation of the fimH gene in a highly virulent UPEC strain rendered it virtually avirulent (33). While P fimbriae recognize kidney glycolipid receptors, type 1-fimbriated E. coli recognize uroplakins Ia and Ib, two major high-mannose-type glycoproteins of urothelial cells (161). Arguably, such binding would facilitate bacterial colonization of the bladder and cystitis. Further evidence implicating Fim involvement in cystitis included the observations that FimH-mediated binding to bladder cells conferred invasion of these cells, probably via FimH recognition of CD48 with ensuing internalization (10, 91, 96), and ultimately led to formation of intracellular pod-like bacterial aggregates (5). This phenomenon may account for the high amount of disease recurrence in many patients despite antibiotic treatment. Also, systemic vaccination with FimH component vaccines significantly reduces UPEC infection of the bladder in mice and monkeys (85, 86).

The fact that variants of FimH differ, often dramatically, in their affinity for target motifs, with accompanying change in tissue tropism, indicates that some variants have evolved and adapted to suit the need of specific pathogenic strains (e.g., UPEC strains). The first step in the colonization of the urinary tract by pathogenic E. coli is FimH-mediated binding to bladder epithelium. FimH variants from UPEC strains tend to exhibit a high affinity for monomannose targets, and this faculty enables binding to host cells in the urinary tract even in the presence of soluble molecular decoys such as Tamm-Horsfall glycoprotein (111). The affinity for trimannose units allows E. coli to bind, for example, buccal epithelial cells even in the presence of soluble mannose-containing molecules of saliva (139). This could be important in the transient colonization of the oropharyngeal mucosa that is thought to be a critical step for host-to-host transfer of E. coli (19). Type 1 fimbriae seem to be involved primarily in bladder infection and less in kidney infection, judging from adhesion affinity as well as in vivo expression (54, 68).

In summary, the FimH protein presented by type 1 fimbriae seems to be a highly versatile adhesin fulfilling a diverse spectrum of roles ranging from pellicle and biofilm formation to being a bona fide virulence factor in UPEC strains, where it plays important roles in the manifestation of cystitis.

CURLI

Occurrence

Curli are thin, coiled, aggregative, amyloid-like fibers first identified on the surface of E. coli (103) and also produced by Salmonella (30) as well as other Enterobacteriaceae including Shigella, Citrobacter, and Enterobacter spp. (125, 164). In Salmonella these organelles are referred to as thin aggregative fimbriae (Tafi) (29, 30, 120). Other designations previously used include GVVPQ and SEF17 fimbriae (30, 31). In this review we will refer to these structures collectively as curli.

Expression of curli has been reported for most enterohemorrhagic, enterotoxigenic, and sepsis strains of E. coli; strains of enteroinvasive E. coli and enteropathogenic E. coli (EPEC) do not appear to produce these fibers (12, 101). Curli are also produced by avian pathogenic E. coli (118). Some E. coli K-12 laboratory strains such as MC4100 and MG1655 do not express curli despite the fact that they contain functional copies of the relevant genes (30, 57, 101). This has typically been attributed to the presence of mutations that affect the regulation of curli gene expression. Most salmonellae produce curli, the best studied being Salmonella enterica serovar Enteritidis (29) and S. enterica serovar Typhimurium (123). Curli genes in Shigella spp. are typically nonfunctional (125).

Structure and Biosynthesis

Curli organelles are approximately 6- to 12-nm-wide fibers of various lengths and appear as a tangled and amorphous matrix extending 0.5 to 1 μm from the cell surface (26) (Fig. 2). Curli characteristically bind Congo red dye, allowingfor an initial rapidscreen for their presence on the cell surface (57). These fibers also bind to a variety of human serum and tissue proteins (6, 12, 57, 103, 104, 137).

Fig. 2Electron micrograph of immunogold-labeled sepsis E. coli expressing curli fibers. Bacterial cells were probed with CsgA-specific antibody.

Source: Reproduced with permission from reference 15.

Curli fibers are composed primarily of subunits of a major component, CsgA (also referred to as curlins). CsgB is the only additional structural component that has been identified within these fibers and is present along the filament in minor amounts (16). These two proteins are nearly identical in size and share 49% amino acid similarity (58). Both proteins are required for fiber polymerization, yet possess very different biochemical properties. The assembly of curli fibers occurs outside the cell via a process referred to as the extracellular nucleation pathway (143). Curli growth proceeds by the addition of subunits principally to the distal end of the growing fiber (58). The CsgA major subunit is secreted from the cell and polymerized on the surface, assisted by the nucleator protein CsgB (16, 58). Evidence for this observation was obtained through the construction and analysis of mutant derivatives. When colonies of CsgB^– CsgA⁺ and CsgB⁺ CsgA^– mutants are grown in close proximity to one another, precipitation of curli subunits occurs on the side of the CsgB⁺ CsgA^– colony facing the CsgB^– CsgA⁺ colony (58). The CsgEFG proteins appear to be involved in export of the curli subunits and are required for fiber formation and assembly (57). CsgG is an outer membrane-located lipoprotein that protects the CsgA and CsgB subunit proteins from proteolysis and serves as a curli assembly platform. Deletion of csgG results in abrogation of curli fiber formation (88). The precise roles of CsgE and CsgF are less well understood. Recent work suggests a chaperone-like function for CsgE and a nucleation function for CsgF in the fiber assembly process (26). Cells containing a mutation in the csgE gene are more susceptible to CsgA degradation, are affected in Congo red binding, and are partially defective in their ability to nucleate extracellular CsgA subunits. Likewise, a csgF mutant possesses a nucleation defect that indicates a role for CsgF in the extracellular polymerization of CsgA (26).

Genetic Organization and Regulation of Transcription

The genes encoding curli are referred to as csg (curli synthesis genes) in E. coli and agf (thin aggregative fimbriae) in Salmonella. They are highly conserved with regard to operon structure, sequence, and mechanism of regulation (29, 37, 57, 120). In E. coli these genes are clustered in two divergent operons; the csgBA operon encodes the structural components of curli, and the csgDEFG operon encodes genes necessary for regulation, transport of subunits, and fiber stabilization. Both operons are required for the production of functional curli (57). An additional open reading frame located downstream of csgBA (i.e., csgC) may also constitute part of the curli gene cluster.

Control of curli expression in E. coli and Salmonella involves a complex network of regulatory proteins and is affected by a range of environmental conditions (25). Curli expression occurs typically at temperatures below 30^oC, at low osmolarity, and in stationary phase (57, 92, 101). The transcriptional factor CsgD belongs to the LuxR family and is a positive activator of csgBA (57). Environmental conditions such as temperature, oxygen tension, starvation, osmolarity, iron, and pH regulate csgD expression (51, 116, 123, 146). This control is mediated by several global regulators including rpoS (101, 123), crl (7, 118), adrA (122, 165), hns (101), ihf (50), ompR (123, 156), mlr (25), and cpxR (39). A complex model demonstrating direct binding of IHF, H-NS, and OmpR to the csgDEFG-csgBA promoter region in response to oxygen tension has been proposed (50). Six binding sites for OmpR were identified (D1 through D6); binding to D1 alone activates csgD transcription under both microaerophilic and aerobic conditions. Expression of csgD is enhanced under microaerophilic conditions by binding of IHF, which prevents phosphorylated OmpR from binding to D2 and/or D3 through D6. Under aerobic conditions, phosphorylated OmpR binds at all six sites and represses csgD transcription, possibly by preventing RNA polymerase binding. Binding of H-NS within this region also represses csgD transcription. although the mechanism is not understood (50). Stationary-phase induction of curli synthesis is dependent on the σ factor RpoS (101). However, transcription of csgD in the absence of H-NS is RpoS independent (6), suggesting that H-NS might selectively repress σ⁷⁰-dependent transcription of csgD.

Role as Adhesin

Curli possess a broad protein-binding capacity that contributes to enhanced virulence. A large number of human proteins interact with curli including laminin (101), fibronectin (103), plasminogen (137), human contact phase proteins (12), and MHC class I molecules (104). It is generally assumed that the CsgA subunit is responsible for mediating these interactions.

Two protein-binding regions of the CsgA subunit capable of activating the contact system have been identified, one in the NH₂-terminal region and one in the COOH-terminal region (102). These regions possess a β-sheet conformation that is characteristic of other amyloid fibrils and may also participate in the polymerization of CsgA subunits.

Role in Biofilm Formation

Curli formation promotes two fundamental processes associated with biofilm formation: initial adhesion and cell-to-cell aggregation. A role for curli in the colonization of inert surfaces has been demonstrated (9, 156). In E. coli, mutations in the ompR gene (such as the ompR234 mutation) stimulate biofilm growth by activating the transcription of csgD (156). This mutation can even stimulate curli production in K-12 strains such as MG1655 that do not normally express these fibers (116). The mutated OmpR234 protein contains a leucine-to-arginine shift at position 43 and exerts a positive effect on curli synthesis (156). A model postulating how curli production is controlled during biofilm growth has been proposed (116). According to this,increase in curli synthesis (e.g., during conditions of low osmolarity) occurs via transcriptional activation of the csgD promoter by OmpR (or, more efficiently, by OmpR234). Alternatively, repression of curli synthesis occurs in response to Cpx pathway activation (via RpoS) during conditions including high osmolarity, curlin accumulation, or a combination of both. The csgDEFG-csgBA promoter region contains sequences with high similarity to the proposed binding site for CpxR (113). Direct binding of CpxR to both the csgD and csgB promoters results in the down-regulation of curli expression (116).

In Salmonella and certain isolates of E. coli, curli fibers promote a specific aggregative colony morphology referred to as the rdar morphotype, a regulated multicellular behavior that mediates properties such as cell aggregation in liquid culture and pellicle formation at the air-liquid interface (15, 123, 153, 165). The rdar morphotype is associated with the formation of an extracellular matrix consisting of curli and cellulose and additionally promotes biofilm formation on abiotic surfaces. A third component, resembling an anionic polysaccharide distinct from colanic acid, may also be associated with this extracellular matrix (160). The coordinate expression of both cellulose and curli is controlled by the CsgD regulator (4, 9, 117, 123, 165). Over 90% of S. enterica serovar Typhimurium and S. enterica serovar Enteritidis strains from human disease, food, and animals express the rdar morphotype at growth temperatures below 28^oC (121).

Role in Pathogenesis

Several properties inherent in the production of curli fibers suggest they contribute to bacterial virulence. The expression of curli is frequently observed on virulent isolates of E. coli and Salmonella. As already mentioned, curli mediate binding to a large repertoire of human proteins and are associated with biofilm formation. Curli production enhances adherence to various intestinal epithelial cell lines and contributes to virulence in mice (155). Curliated E. coli O157 strains contain promoter alterations that enhance curli expression and are associated with increased pathogenesis in mice and invasion of Hep-2 cells (154). Curli also contribute to virulence in avian pathogenic E. coli by promoting adherence to intestinal cells (84).

Severe sepsis and septic shock are frequently caused by gram-negative bacteria, and several factors suggest a significant role for curli during E. coli sepsis. First, E. coli strains isolated from patients with sepsis frequently express curli and these patients often also possess antibodies to CsgA (12, 15). Second, the assembly of components of the contact phase system on the surface of curliated bacteria triggers proinflammatory and procoagulatory cascades (such as the induction of proinflammatory cytokines) that may contribute to these disease states (67). Third, curliated E. coli sepsis bacteria are highly pathogenic when injected into mice or rats and induce symptoms often associated with severe sepsis, such as bleeding, lung disorders, and fall in blood pressure (17, 67, 112). Finally, high-affinity binding of curli to fibronectin is associated with invasion of eukaryotic cells (53). Curli genes from virulent septicemic E. coli isolates mediate enhanced internalization in comparison to their K-12 counterparts, suggesting differences in structure or regulation of expression between the loci originating from the two strain types (53).

ANTIGEN 43

Occurrence

In 1980 Diderichsen characterized a locus on the E. coli K-12 chromosome that affected a number of surface properties, including colony morphology and autoaggregation of bacterial cells. The latter could be visualized macroscopically as flocculation and settling of cells from static liquid suspensions (Fig. 3); hence the name flu was originally coined for the corresponding genetic locus (38). In an independent study, a major E. coli outer membrane antigen was investigated by virtue of its aggregative properties and termed Antigen 43 (Ag43) (108). Only later was Ag43 unambiguously identified as the product of the flu gene (60, 65). In line with the gene product name Ag43, the term agn43 is also frequently used for the corresponding gene instead of flu.

Fig. 3Settling, aggregation, and biofilm formation properties of E. coli expressing Ag43. (A) Settling of cells from static liquid suspensions after overnight growth. (B) Phase-contrast microscopy demonstrating the aggregation phenotype conveyed by Ag43. (C) Biofilm formation of Ag43-negative (i) and -positive (ii) cells on a glass slide after 24 h of growth in glucose minimal medium using a continuous flow chamber system.

Ag43 is present in up to ∼50,000 copies per cell (109). Ag43 is found in most E. coli strains, and, interestingly, it is expressed by many pathogenic strains. A survey of EPEC and UPEC strains showed that 77% and 60%, respectively, of these were capable of Ag43 expression (110). Although E. coli K-12 only has one copy of the flu gene, many wild-type strains (including many EPEC, EHEC, and UPEC strains) possess duplex or multiple copies (up to four) of the gene (78, 119, 152). The flu genes and corresponding gene products vary considerably at the sequence level (60 to 100% identity) when copies from the same strain and from different strains are compared (78). Although the K-12 version of Ag43 does not seem to confer attachment to mammalian cells, it is noteworthy that the close relatives of Ag43, the AIDA-I and TibA proteins of certain EPEC and ETEC strains, both are bona fide adhesins. They confer strong bacterial attachment to mammalian cells and exhibit ∼30% sequence identity with Ag43 of E. coli K-12 (13, 87). It is conceivable that the sequence variability observed within the Ag43 family will also be reflected in functional variability, and some variants may have developed the ability to attach to mammalian cells.

Structure and Biosynthesis

Ag43 is a member of the family of autotransporter proteins. These are characterized by the fact that all information required for traverse of the bacterial membrane system and final routing to the surface resides in the protein itself. Autotransporter proteins possess some general features: an N-terminal signal sequence, the passenger (α) domain that is secreted to the cell surface, and a transporter (β) domain that forms a β-barrel pore, which assists the passenger domain in gaining access to the surface. Many autotransporters are virulence factors in gram-negative bacteria (63, 64). Ag43 is produced as a 1,039-amino-acid preprotein. An N-terminal signal peptide (52 amino acids) directs translocation across the cytoplasmic membrane to the periplasm. Ag43 consists of two subunits, α and β, comprising 499 and 488 amino acids, respectively. The α subunit (Ag43^α) or passenger domain is presented on the cell surface via β-subunit-assisted crossing of the outer membrane. Subsequently, or during this transfer, the peptide bond linking the two domains is cleaved; all available evidence indicates that this happens by autocatalytic action (65). A sequence, LADSGAAVSGT, resembling a consensus sequence of an aspartyl protease active site is located in Ag43^α (65). However, we recently mutagenized this sequence and correct processing still occurred, ruling out the potential role of this sequence in Ag43 processing (our unpublished results). Meanwhile, this observation does not in any way exclude Ag43 autocatalytic cleavage, and several other possible candidates for internal catalytic sites exist. The β subunit has all the hallmarks of a β-barrel porin and, in line with other autotransporters, presumably forms a pore through which Ag43^α gains access to the surface (65). The α subunit remains attached to the cell surface via noncovalent interactions with the β subunit, but can easily be detached by brief heat treatment (65). The simplicity of the autotransporter mode of secretion is also reflected in the amenability by which Ag43 can be expressed in a wide range of gram-negative bacteria (70, 71). In all the different gram-negative bacteria in which we have expressed Ag43, we have observed correct processing into the α and β moieties; this also supports the notion that Ag43 processing is autocatalytic.

The nearest homologue to Ag43^α with a known tertiary structure is P.69 pertactin, an autotransporter toxin from Bordetella pertussis that exhibits ∼20% amino acid sequence identity and 35% similarity to Ag43^α. This protein is folded in a right-handed parallel β-helix (43). Based on this, the tertiary structure of Ag43^α was modeled (78). Ag43^α has 18 internal repeats of approximately 19 residues between position 53 and 450. Each repeat is predicted to form a β-helix rungconsisting of three β strands and three turns. The backbone folds up in a helical fashion with β strands from adjacent rungs stacking in a parallel orientation. The buried cylindrical core is predominantly composed of hydrophobic side chains. A β-helix model of featuring 18 rungs can be fashioned along these lines. The bottom of this long helix would fit into the β barrel formed by the Ag43^β domain, whereas the tip would protrude ∼10 nm from the cell surface and be free to interact with Ag43 from neighboring cells. Together with AIDA-I, TibA, and pertactin, Ag43 seems to form a subfamily of the autotransporter proteins in which the passenger domains all have repetitive sequence motifs and likely extended β-helix structures.

Autotransporter proteins are generally highly similar with respect to the structure of the β module, whereas they differ substantially in their α module (63). Generally, the β modules of the autotransporter proteins have between 10 and 18 (an even number) membrane-spanning amphiphatic β sheets (64). In accordance with these general features, the β module of Ag43 was modeled to consist of an 18-stranded β-barrel module (65).

Genetic Organization and Regulation of Transcription

The flu locus was reported to map on the chromosome at ∼43.5 min near his (38). From early on, expression of the flu gene was observed to be phase variable, and a genetic locus, mapping in the 89-min region, was shown to be involved in this phenomenon (38). Since Diderichsen's pioneer studies, a quite detailed picture of the genetics and regulatory mechanisms of flu has emerged. The 3-kb flu gene has been sequenced from several E. coli strains of both K-12 and wild-type ancestry. In addition to the copy found at ∼43.6 min (on E. coli K-12 reference strain MG1655), other copies can be present, for example, on pathogenicity islands, as in the case of UPEC strain CFT073 (159). Like type 1 fimbriae, expression of Ag43 is phase variable, with switching rates of ∼10^-3 per cell per generation under normal laboratory growth conditions. Ag43 phase transition is readily visualized by colony morphological changes, i.e., frizzy ↔ non-frizzy (61). However, the regulatory systems are quite different, and flu exhibits phase-variable expression due to the concerted action of the Dam methylase (positive regulation) and the cellular redox sensor OxyR (negative regulation). This model was originally proposed by Henderson et al. (62) and experimental evidence from several groups has proven it correct. The regulatory element mapping in the 89-min region turned out to be oxyR or mor (28, 62, 158). Lesions in oxyR result in strong constitutive expression of flu (38, 65, 130). The OxyR arm of the oxidative stress response primarily senses peroxides and monitors the cellular thiol-disulfide state in the cytoplasm (8, 163). The OxyR regulon encompasses more than 30 genes that are either activated or repressed. According to the prototypic two-state model, OxyR can exist in two principal forms, reduced or oxidized, due to reversible disulfide bond formation (163). The two forms have different conformations and recognize different DNA motifs (151). The oxidized form of OxyRgenerally acts as an activator. Several independent studies have suggested that the flu gene is primarily repressed by the reduced form of OxyR (55, 65, 130, 133). Conversely to oxyR mutants, strains carrying knockout mutations in the Dam methyl transferase encoding dam are locked in the off state and do not express Ag43 (110). Expression of the flu gene is driven from a sigma-70 promoter (60, 157). The promoter region contains three tightly spaced GATC sites overlapping with OxyR binding sites (55, 60). Dam methylation of these sites abrogates OxyR binding and OxyR-mediated repression of flu; conversely, bound OxyR prevents methylation (55, 157). Studies of OxyR interaction with flu DNA showed that methylation of any two (of three) GATC sites in the promoter region was necessary and sufficient to block binding of the repressor (157).

Diderichsen reported that cells that were able to fluff were nonfimbriated and vice versa (38). This initial observation was later followed up by Hasman et al. (60), who demonstrated that the presence of fimbriae sterically blocks Ag43-mediated cell aggregation. It was subsequently shown that Ag43 and the type 1 class of fimbriae interact not only at the physical level but also at the transcriptional level through molecular cross-talk. Synthesis of type 1 and related fimbriae abrogates Ag43 expression; this effect is not observed in an oxyR mutant (133). The major fimbrial subunit, FimA, constitutes the most abundant phase variable surface protein in E. coli with up to 5 × 10⁵ copies. In contrast to many other surface proteins including Ag43, FimA contains a cysteine bridge. During fimbria biosynthesis and export to the surface, disulfide bridge formation takes place at a very high level. Conceivably, this could affect the cell's thiol-disulfide status; thus fimbriae expression per se constitutes a signal transduction mechanism that affects expression of the flu gene, arguably via the oxidoreductive state of OxyR (133). Also, addition of a simple reducing agent, dithiothreitol, to cells negatively affects Ag43 expression (130, 133).

Interestingly, expression of P and type 1 fimbriae also seems to be coordinated through intersystem cross-talk via PapB, in which P fimbriation dominates (162). Thus the expressions of P and type 1 fimbriae and Ag43 are highly coordinated processes with a hierarchic structure. P fimbration is dominant to type 1 fimbriation, and fimbriation in general is dominant to Ag43 expression. The fimbrial phase "on" allows bacterial attachment to various epithelial targets, and the differential expression of type 1 and P fimbriae through cross-talk allows differential target and niche selection and cell invasion. The fimbrial phase "off" allows expression of Ag43, which promotes enhanced microcolony and biofilm formation (see later).

Role as Adhesin

Ag43 expression was early on observed to correlate with cell-to-cell aggregation, fluffing, and settling of cells from static liquid suspensions. This faculty was later demonstrated to be based on intercellular Ag43-to-Ag43 recognition (60, 71). Thus, Ag43 is exceptional in being a self-recognizing adhesin; i.e., both receptor recognition and receptor target are provided in the same polypeptide. As previously mentioned, E. coli is capable of autoaggregation, and several different systems other than Ag43 are also able to independently produce this phenotype. As previously discussed, curli-mediated autoaggregation occurs via intercellular fiber precipitation. Bundle-forming pili (BFP) are a type IV class of fimbriae, produced by EPEC strains, that emanate from the cell surface and align along their longitudinal axes to form bundles of filaments (52). Expression of BFP mediates two phenotypes thought to play a role in colonization: autoaggregation in liquid cultures and localized adherence on tissue culture cell monolayers (18). In enteroaggregative E. coli strains, two flexible 2- to 3-nm-wide fimbrial types, designated aggregative adherence fimbriae I and II (AAF/I and AAF/II), have been identified (35, 97). The aggregative adherence phenotype is distinguished by prominent autoagglutination of bacterial cells to each other (98). Variants of the FimH adhesin of type 1 fimbriae have also been shown to promote cell aggregation in laboratory and wild-type background strains (127, 132). In contrast to other aggregating systems, the self-recognizing Ag43 adhesin is anchored directly to the outer membrane. Thus Ag43-mediated aggregation results in a more intimate cell-cell contact than seen with systems in which the intercellular interactions are based on polymeric structures that reach far out from the bacterial surface, i.e., fimbriae and curli.

Several lines of circumstantial evidence implicate the α subunit of Ag43 to be involved in autoaggregation. First, cells depleted in α but not β subunit do not autoaggregate. Second, addition of crude preparations of Ag43^α to Ag43-expressing cells reduces autoaggregation significantly. Finally, cells treated with antiserum specific for the α subunit do not autoaggregate (our unpublished data). In a recent study on wild-type variants of Ag43 that we expressed in a defined K-12 background, three out of nine investigated Ag43 variants did not confer aggregation (78). However, these versions were expressed in normal quantities on the surface and gave rise to frizzy colonies, indicating that cell aggregation is not involved in the characteristic frizzy phenotype associated with Ag43. By employing a combination of linker insertion mutagenesis and domain swapping between aggregating and nonaggregating versions of Ag43, we pinpointed the region responsible for autoaggregation to be located within the N-terminal one-third of the passenger domain (78). Ag43 aggregation was strongly affected by pH. Aggregation was optimal at neutral and weakly acidic pH and abolished at pH values of <3 and >10. Also, high NaCl concentration inhibited autoaggregation. Taken together, the data indicate that ionic interactions between charged basic and acidic side chains play a role in Ag43-Ag43 self-recognition. The N-terminal segment of Ag43^α contains numerous basic and acidic amino acids.

In addition to self-recognition, Ag43 has also been reported to confer weak binding to certain human cell line cells such as HEp-2 cells (66, 110). In this context it should be mentioned that a recent investigation could not confirm any affinity towards HeLa cells when the aggregation-positive Ag43 variant from the EDL933 EHEC strain was investigated (152). Interestingly, the same Ag43 variant, also called Cah (short for calcium-binding Ag43 homologue) was found to be capable of binding Ca²⁺ (152).

The presence of the flu gene is thought to be highly conserved in E. coli, and previous studies have shown that many wild-type clinical isolates indeed contain multiple copies of this allele (119). Our recent data indicate that the capacity of these duplicated genes to encode intact versions of Ag43 has been maintained. The inability of some Ag43 variants to promote cell-cell aggregation suggests that these novel variants may possess an alternative function. Gene duplication provides a source for diversification; the extra gene copy can acquire a new function, or the two copies can evolve by dividing up the ancestral functions. The fact that the open reading frames in the duplicated genes are maintained points to the development of new functional properties, as extraneous genes of no beneficial function are known to be gradually lost over time (45). It might therefore be the case that some versions of Ag43 have been selected and optimized for other functions than Ag43-self recognition, for example as adhesins to mammalian cells.

Role in Biofilm Formation

A biofilm mode of growth offers many advantages to a bacterial population compared to planktonic growth, such as a high level of resistance against predation and various antimicrobial agents. It has been hypothesized that phase-variable regulation of bacterial adhesins may be an important theme in virulence and biofilm formation (66, 115). It can be argued that it is in the interest of a bacterial species to maintain a fraction of the population that are primed to initiate biofilm formation under distinct conditions. In that way bacteria are always capable of immediately initiating biofilm formation when appropriate conditions are met, but they do not need to spend excessive amounts of energy by priming the cells of the entire population for such tasks. The phase-variable expression of Ag43 fits neatly into this picture.

The attachment of bacteria to a surface often results in proliferation into more complex microcolony structures. Indeed, bacterial aggregation and microcolony formation can be seen as a prelude to biofilm formation. In line with type 1 fimbriae and curli, Ag43 is implicated in aggregation and microcolony formation (61). It was therefore not too surprising that several groups have shown that Ag43 expression enhances biofilm formation of E. coli on various abiotic surfaces (Fig. 3) (36, 71, 72, 152). Recently, in a microarray-based study on global gene expression in E. coli biofilms, it was demonstrated that Ag43 expression is specifically upregulated during sessile growth when compared to both exponential and stationary planktonic cultures (131). It is also interesting to note that a survey of the biofilm-forming faculty of a diverse group of Ag43 variants showed that all of them enhanced biofilm growth, albeit with different efficacy (78). The ability of Ag43 to augment biofilm formation was recently shown to be sterically blocked by bulky surface structures such as capsules (128). It remains to be determined whether or not capsule expression is coordinated with Ag43 expression.

Nonindigenous expression of Ag43 in many gram-negative bacteria (for example, Pseudomonas fluorescens) was demonstrated to enhance biofilm formation (71, 72). It has also been shown that Ag43 can be used as a tool for forcing different bacteria together in biofilms and thereby a means to determine species composition in biofilms (71, 72).

Role in Pathogenesis

Many bacteria, notably pathogens, are known to have the ability to form aggregates in vitro and in vivo. Examples include a diverse range of bacteria of both gram-positive and -negative origin such as Bordetella pertussis (94), Mycobacterium tuberculosis (95), Staphylococcus aureus (93), and Streptococcus pyogenes (47). Such aggregates are known to be able to resist various host defenses, e.g., complement attack and phagocytosis, more efficiently than solitary bacteria (14, 99). In a study on S. pyogenes it was reported that mice infected with coaggregated cells more frequently developed abscesses, indicating that virulence is enhanced by aggregation (99). These observations lend strong support to the notion that aggregation is an important virulence mechanism. The formation of aggregates usually takes place through autoaggregation of cells. In a few cases the underlying molecular mechanism is known, and often self-recognizing surface proteins are responsible. A well-studied example of this phenomenon is the autoaggregation of S. pyogenes through intercellular interactions between pairs of protein H (47). On this background it can be argued with some justification that the ability of Ag43 to enhance bacterial aggregation and biofilm formation in itself constitutes a virulence trait since these phenotypes are often related to enhanced virulence. Recently, however, Ag43 was shown to be expressed in vivo during formation of intracellular bacterial aggregates or pods in bladder cells (5). It was suggested that the bacteria were in a quiescent state in these aggregates and were out of reach for killer cells and the immune defenses of the host. This interesting observation implicates Ag43 as a virulence factor in UPEC strains.

The protection provided by Ag43-mediated aggregation was also underlined in another series of experiments addressing the role of Ag43 in protection against oxidizing agents. It was speculated that the tight packing of cells in cellular aggregates connected with Ag43 expression could provide a mechanism for reducing local oxygen concentrations, thereby protecting the cells from damage caused by oxidizing agents. We found strong support for this hypothesis in the fact that Ag43 expression and ensuing cell-to-cell aggregation was observed to be concomitant with a high degree of protection against H₂O₂ killing (130).

A picture of the Ag43 family of autotransporters is slowly emerging. Several lines of evidence implicate this family in bacterial virulence, as follows. (i) Many, but not all, Ag43s are self-recognizing adhesins conferring cell aggregation and biofilm formation, both of which are defensive phenotypic traits related to increased virulence. (ii) Some Ag43s seem to confer a low level of adhesion to mammalian cells. (iii) Ag43 expression undergoes phase variation controlled by OxyR and Dam, not only in K-12 but also in wild-type strains (119). Phase variation has notoriously been linked to many bacterial virulence factors (66); additionally, Dam is known to regulate the expression of numerous virulence factors (89). It is hoped that future research will unravel more aspects of this group of autotransporter proteins.

CONCLUDING REMARKS

Type 1 fimbriae, curli, and Ag43 are structurally different bacterial surface structures and follow completely different strategies for surface display and assembly. However, despite these differences they have several functional traits in common: (i) they are adhesins; (ii) members of all three families promote bacterial aggregation; (iii) they confer biofilm formation; (iv) members of all three families are linked with enhanced virulence; and (v) their expression promotes distinct colony morphology types. In effect, in spite of their physical differences, these surface structures are able to fulfil many of the same functions and are prime examples of convergent functional evolution. Meanwhile, despite these similarities each system has its own unique properties. For example, only type 1 fimbriae can bind to mannosides, and while Ag43 function is shielded by capsule type 1, fimbriae are not. Future work will undoubtedly identify additional virulence properties associated with these surface structures.

Acknowledgments

We thank Zhao Bian for providing us with an electron micrograph of a curliated cell. This work was supported by the Danish Technical Research Council (26-02-0183), the Danish Natural Sciences Research Council (21-01-0296), the Danish Medical Research Council (22-03-0462), and the Australian National Health and Medical Research Council (301163).