Immunoglobulin domains in Escherichia coli and other enterobacteria: from pathogenesis to applications in antibody technologies


  • Gustavo Bodelón,

    1. Department of Microbial Biotechnology, Centro Nacional de Biotecnología, Consejo Superior de Investigaciones Científicas (CNB-CSIC), Campus de Cantoblanco UAM, Madrid, Spain
    Search for more papers by this author
  • Carmen Palomino,

    1. Department of Microbial Biotechnology, Centro Nacional de Biotecnología, Consejo Superior de Investigaciones Científicas (CNB-CSIC), Campus de Cantoblanco UAM, Madrid, Spain
    Search for more papers by this author
  • Luis Ángel Fernández

    Corresponding author
    • Department of Microbial Biotechnology, Centro Nacional de Biotecnología, Consejo Superior de Investigaciones Científicas (CNB-CSIC), Campus de Cantoblanco UAM, Madrid, Spain
    Search for more papers by this author

Correspondence: Luis Ángel Fernández, Centro Nacional de Biotecnología, CNB-CSIC, Campus UAM, Cantoblanco, Darwin 3, Madrid 28049, Spain. Tel.: +34 91 585 48 54; fax: +34 91 585 45 06; e-mail:


The immunoglobulin (Ig) protein domain is widespread in nature having a well-recognized role in proteins of the immune system. In this review, we describe the proteins containing Ig-like domains in Escherichia coli and enterobacteria, reporting their structural and functional properties, protein folding, and diverse biological roles. In addition, we cover the expression of heterologous Ig domains in E. coli owing to its biotechnological application for expression and selection of antibody fragments and full-length IgG molecules. Ig-like domains in E. coli and enterobacteria are frequently found in cell surface proteins and fimbrial organelles playing important functions during host cell adhesion and invasion of pathogenic strains, being structural components of pilus and nonpilus fimbrial systems and members of the intimin/invasin family of outer membrane (OM) adhesins. Ig-like domains are also found in periplasmic chaperones and OM usher proteins assembling fimbriae, in oxidoreductases and hydrolytic enzymes, ATP-binding cassette transporters, sugar-binding and metal-resistance proteins. The folding of most E. coli Ig-like domains is assisted by periplasmic chaperones, peptidyl–prolyl cis/trans isomerases and disulfide bond catalysts that also participate in the folding of antibodies expressed in this bacterium. The technologies for expression and selection of recombinant antibodies in E. coli are described along with their biotechnological potential.


Since the initial report of the three-dimensional structure of a human antibody fragment (Poljak et al., 1973), the immunoglobulin (Ig) fold has been found in an increasing number of proteins with diverse biological functions and without an apparent sequence identity (Bork et al., 1994; Halaby & Mornon, 1998; Otey et al., 2009). Initially, the Igs and the proteins involved in the immune response sharing the same fold were classified under the term ‘immunoglobulin superfamily’ (IgSF; Williams, 1984; Williams & Barclay, 1988). Currently, the IgSF is recognized as one of the largest families in vertebrate genomes, and an analysis of the human genome revealed that the Ig-like domain has the widest representation of any protein domain (Lander et al., 2001), accounting for more than 2% of the total human genes and one of most common structural motifs found in more than 80 protein superfamilies (Srinivasan & Roeske, 2005). The Ig-like domain is amply distributed in nature, and it is present in vertebrates and invertebrates, plants, fungi, parasites, bacteria, and viruses (Halaby & Mornon, 1998). While the general functional role of Ig-like domains is related to binding or molecular recognition processes, the specific reactions mediated by these domains vary widely. According to their functional characteristics, members of the IgSF are classified into eight groups (Halaby & Mornon, 1998): molecular transport, morphoregulation, cell phenotype markers, cell adhesion molecules, virus receptors, shape recognition and toxin neutralization, viral and bacterial molecules and other functions including regulation of gene transcription, cell migration or death cell. The heterogeneity of this classification reflects the ample functional diversity of the IgSF.

In bacteria, the first examples of Ig-like domains were found in the crystal structures of the proteins PapD of Escherichia coli (Holmgren & Branden, 1989) and Cyclodextrin glycosyltransferase of Bacillus circulans (Hofmann et al., 1989), thus demonstrating the presence of the Ig-like fold in both Gram-negative and Gram-positive bacteria. The success and the prevalence of the Ig-like domain can be explained by its structural and functional properties such as its stability and resistance to proteolysis. Functionally, it can interact through its different faces and across domains to form homodimers and heterodimers or tandem linear arrays of Ig-like domains. Also, the presence of a single exon coding for most IgSF domains provides the genetic basis for duplication and diversification.

This review summarizes the structural and functional properties of the proteins and organelles containing Ig-like domains in E. coli and related enterobacterial species, highlighting their role during cell host and tissue adhesion, invasion, or other steps of the infection process. The folding and assembly of these E. coli Ig-like proteins and organelles is discussed, along with the chaperones and protein machineries involved. In addition, we review the expression of heterologous Ig domains in E. coli, focusing on the current technologies for expression and selection of small antibody fragments and full-length Ig molecules in this bacterium, given their biotechnological significance for the development of therapeutic antibodies.

General structural properties and classification of Ig and Ig-like domains

The functional properties conferred by the Ig-like domain rely on the structural characteristics of the molecule that allow a high degree of interaction specificity and diversity (Amit et al., 1986). The Ig-like domain has a three-dimensional structure based on what is called the Ig fold, which is composed of about 70–100 amino acid residues in anti-parallel β-strands designated by letters A, B, C, D, E, F, G in order of appearance in the sequence and organized in two β-sheets that are packed against each other in a β-sandwich (Fig. 1). The β-strands are composed by an alternation of hydrophobic and hydrophilic amino acids with the hydrophobic side chains pointing toward the interior of the molecule. Traditionally, two conserved cysteine residues separated by 55–75 amino acids, and the so-called ‘invariant’ tryptophan residue located within 10–15 residues C-terminal to the first cysteine were two hallmarks that allowed a putative identification of an Ig-like domain at the primary sequence level (Williams & Barclay, 1988). However, some IgSF domains lack these important features but still adopt an Ig-like fold (Vaughn & Bjorkman, 1996).

Figure 1.

Topology of Ig domains. Topology diagrams and strand nomenclature for the β-sheets of four variants of the Ig fold: variable (V set), constant-1 (C1 set), constant-2 (C2 set) and intermediate (I set). The front and back sheets are depicted in red and blue, respectively, and the presence of a disulfide bond between B and F strands is indicated as a yellow line. The complementarity-determining regions (CDRs) in the V set are also indicated.

Several variants of the Ig-like fold general architecture are found in nature and are classified structurally according to the number of strands, the presence of conserved sequence signatures and the length of the loops interconnecting the β-strands. However, and despite that the Ig-like domains are defined by a common topology and connectivity similar to those of Igs, their amino acid sequences can share as low as 10% sequence identity (Halaby et al., 1999). As an example, the Ig-like structures of the N-terminal domain of the bacterial protein PapD (pdb code: 3dpa), the Fibronectin type-III (FnIII) receptor (pdb code: 1ten), the human growth hormone receptor (pdb code: 3hhr) and the Ig constant (C) 2 domain (pdb code: 7fab) share the same β-sheet domain topology, confirmed by their crystallographic data, although they do not have significant sequence similarity (Vaughn & Bjorkman, 1996). This enormous sequence variability results in a wide range of structural variants responsible for the diversity of function exhibited by members of the IgSF (Chattopadhyay et al., 2009).

The classification of Ig and Ig-like domains has evolved since the first description of the IgSF. Originally, the domains containing an Ig fold were divided into three different topological subtypes V, C1, and C2 (Williams & Barclay, 1988). The structures grouped in the V subtype that includes the variable (V) domain of Igs are generally composed of a back sheet formed by the G, F, C, C′, and C″ strands, and a front sheet formed by the B, E, and D strands. The A strand is shared between the two sheets. The CB, C′C″, and FG strands are connected by loops which correspond to complementarity-determining regions (CDRs) of the V domains of Igs (Fig. 1). The C1 subtype, which includes the constant (C) domain of Igs, contains structures with a back sheet formed by G, F, and C strands and a front sheet formed by A, B, E, and D strands. Protein domains were assigned to the V and C1 subtypes based on the topology of known structures. However, the structural characteristics of the C2 subtype were based on structural predictions. Subsequent structure determinations redefined the C2 domain into the C2 and I subtypes (Fig. 1; Harpaz & Chothia, 1994), and revealed new S (switched) and H (hybrid) subtypes (Bork et al., 1994). Finally, analyzing the sequence and structure of 52 3D structures, Halaby and coworkers classified the Ig-like domains into the C1, C2, C3, C4, V, I, H, and FnIII subtypes (Halaby et al., 1999).

An important structural feature of the classical Ig fold is the presence of a conserved intra-domain disulfide bond connecting the β-strands B and F of opposite β-sheets (Lesk & Chothia, 1982; Fig. 1), which in the case of the Ig domains of the heavy (H) and light (L) chains of antibodies is formed by the conserved Cys residues at framework positions H22/H92 and L23/L88, respectively (Williams & Barclay, 1988; Worn & Pluckthun, 2001). Whereas this disulfide bond is remarkably well conserved in the V domain of the germline genes of all antibodies (Proba et al., 1998), the number and the location of disulfide bridges in Ig-like structures vary, and when they exist, may connect two strands in the same sheet, a strand and a loop or two loops (Halaby et al., 1999). Both the intra-domain disulfide bond (Proba et al., 1998; Worn & Pluckthun, 1998; Hagihara et al., 2007), and a cluster of hydrophobic amino acid residues that is formed by the packing of the four B, C, E, and F β-strands of the common hydrophobic core (Williams & Barclay, 1988; Bork et al., 1994; Fowler & Clarke, 2001) are responsible for the stability and the folding of Ig and Ig-like domains.

Origin and evolution of Ig domains

The existence of the Ig-like fold in functionally distinct and phylogenetically distant molecules, such as enterobacterial fimbriae, plant cytochromes or vertebrate antibodies, raises important questions as to when and how the IgSF arose. From an evolutionary standpoint, it is not clear whether the IgSF domains evolved from a common ancestor by vertical or horizontal gene transfer or are the consequence of the drift of independent sequences toward a favorable folding topology (Halaby & Mornon, 1998; Barclay, 2003). It has been suggested that the robust framework of the Ig-like fold is simply the result of an energetically favorable folding (Shapiro et al., 1995). Indeed, the number of possible stable configurations of globular proteins may be limited, and thus, a considerable number of known proteins adopt one of 10 favorable ‘superfold’ configurations (Orengo et al., 1994, 1997), one of which is the Ig fold (Steiner, 1996).

The fact that many of the IgSF members do not show significant sequence similarity to one another makes very difficult to distinguish between convergent or divergent mechanisms of evolution at such low levels of sequence identity (Klein & Nikolaidis, 2005). Despite of this, it is generally believed that the IgSF derives from a single common ancestor, and examples of putative evolutionary relationships by vertical descendent between bacterial and eukaryotic Ig-like domains can be found in the literature (Bateman et al., 1996; Stevens, 2008).

Interestingly, the structure of the FnIII Ig-like domain of certain carbohydratases (e.g. chitinases) from soil bacteria and that of PapD chaperone is so similar to the eukaryotic FnIII sequences that it has been proposed that these domains were initially acquired from eukaryotes and subsequently spread by horizontal transfers between distantly related bacteria (Holmgren & Branden, 1989; Bork & Doolittle, 1992). Contrary to this, information from bacterial genomes suggested that the FnIII domain might have had its origins in bacteria (Aravind et al., 2003). Along with this hypothesis, the bacterial FnIII domains of ApaG proteins, which are found in a wide variety of bacterial genomes and do not show significant sequence similarity with the FnIII domains of carbohydratases or PapD, may in fact have a bacterial origin (Cicero et al., 2007).

Although the origin and the phylogenetic relationship among the different Ig-like types are controversial, various hypotheses have been proposed to explain the evolution of a primordial domain into its variants by gaining or losing β-strands. Depending on the authors, the original domain might be either the V domain (Williams & Barclay, 1988) or the C2 and I domains (Smith & Xue, 1997; Teichmann & Chothia, 2000). Smith and colleagues proposed that the V and the C1 types appear to be derived from the C2 type, either directly or through the type-I. The authors speculated that the C2 set might be the primordial domain because a more variable and simpler domain is likely to be more ancient (Smith & Xue, 1997). Teichmann and Chothia suggested that the IgSF was originated from the I set because this molecule has structural characteristics that combine structural features of V and C sets (Teichmann & Chothia, 2000). To our knowledge, the phylogeny of the FnIII, C3 and C4 types, which are generally agreed to be in an outgroup position relative to the V, C1, C2 and I sets (Klein & Nikolaidis, 2005), has not been addressed to date. Finally, it has been proposed that both mechanisms of convergent and divergent evolution might explain the IgSF: convergence of unrelated domains toward a simple and stable fold and divergence within each subtype (Halaby et al., 1999). In case of divergent evolution of the domain, it has been suggested that the role of the primordial Ig-like domain was to mediate cell–cell homotypic interactions and, with gene duplication and divergence, developed into more complex heterotypic interactions. It also remains to be elucidated whether the primordial IgSF domain originated in cytosolic proteins or membrane proteins (Tilson & Rzhetsky, 2000; Barclay, 2003).

Interestingly, Ig-like domains have also been identified in certain bacteriophage proteins and bioinformatic analysis of all sequenced phage genomes have revealed that Ig-like domains are widely distributed in phage genomes, as insertions into the coding sequences of surface-exposed structural proteins of tailed double-stranded (ds) DNA bacteriophage particles (Caudovirales). Strikingly, no Ig-like domains were found in either ssDNA, RNA phages, or in other classes of dsDNA phages (Fraser et al., 2006). The 68 phage Ig-like domains discovered in 41 genomes belonging to the three families of Caudovirales (Siphoviridae, Myoviridae and Podoviridae), were classified into three distinct sequence families of Ig-like domains (I set, FNIII and Big2) defined in the PFAM database (Finn et al., 2008). The phage Ig-like domains appear to be present only in structural proteins such as tail fiber, baseplate wedge initiator, major tail, major head or outer capsid proteins. Although the precise function of these domains is unknown, several lines of evidence suggest that they may play a role in phage infection mediating processes of attachment to host cell (Fraser et al., 2006, 2007; Sathaliyawala et al., 2010). As the members of these bacteriophage families posses very limited sequence similarities to one another, it has been proposed that the presence of similar Ig-like domains among different virus families is attributable to horizontal gene transfer. Fraser and collaborators also hypothesized that Caudoviridae, which infect both Gram-negative and Gram-positive bacteria, might have been an important vector in the spread of Ig-like domains through diverse species of bacteria (Fraser et al., 2006).

Ig-like domains in E. coli and enterobacteria

Ig-like domains have been reported in a good number of E. coli and enterobacterial proteins, as will be reviewed in the following sections. To find known E. coli and enterobacterial proteins bearing an Ig-like fold we performed bibliographic and bioinformatic searches to screen current protein databases. We searched in the structural classification of proteins (SCOP, version 1.75; Andreeva et al., 2008) and in the conserved domain protein family (Pfam; Finn et al., 2008) databases. Next, an estimate of the number of E. coli and enterobacteria protein entries containing Ig-like domains was obtained performing a taxonomic search of the accession numbers of the Pfam domains in the Uniprot knowledgebase (UniProtKB) database (UniProt Consortium, 2010).

In Table 1 is represented the Pfam code and accession number of the Ig-like domains identified along with representative examples of polypeptide domains with Ig-like fold. For each Pfam domain, the current number of Uniprot entries found in E. coli and enterobacteria is indicated. In addition, a protein example of each category, with its PDB code, the presence or absence of disulfide bonds, function and localization is indicated. As seen in Table 1, the Ig-like domain is amply distributed among E. coli and enterobacteria, being present in proteins with different functions (such as fimbrial and afimbrial adhesins, chaperones, transporters, and enzymes among others) and not circumscribed to a particular subcellular location. Ig-like domains are found in proteins in different cellular compartments such as cytoplasm, inner and outer membrane (IM and OM), periplasm and attached to the cell surface or secreted to the extracellular medium. The presence of a disulfide bond is not a requisite for the folding of all Ig domains and is not a conserved feature among enterobacterial Ig-like domains (Table 1). As could be expected given the reducing environment of the E. coli cytoplasm (Ritz & Beckwith, 2001), the Ig-like domains from cytoplasmic enzymes (C-Wzt, β-galactosidase and the Glycogen branching enzyme) lack disulfide bonds.

Table 1. Immunoglobulin domains in proteins from Escherichia coli and other Enterobacteria
Organelles and proteinsIg-like domainUniprot entries (E. coli/Enterobacteriaa)Protein example (Uniprot accession)PDBDisulfide bond in Ig-like domainFunctionLocalization
  1. a

    Uniprot entries updated at December 2011.

  2. b

    Big_2 (PF02368) also includes DsDNA viruses (Caudovirales, 63 entries) and two enterobacterial phages.

  3. c

    Number of sequences obtained from PFAM database and not from Uniprot database. No Uniprot entries yet available.

Fimbrial adhesins
Pilin domainFimbrial (PF00419) FimH-C4366/8102FimA (P04128)2JTYYesAdhesionCell surface
PapG_C (PF03628) PapG-C55/59
Fimbrial_K88 (PF02432) FaeG88/259
Fimbrial_CS1 (PF04449) Cfae-C79/190
Adhesin_Dr (PF04619) DraE60/60
AfaD (PF05775) AfaE/D36/112
Antig_Caf1 (PF09255) Caf12/29
Saf-Nte_pilin (PF09460) SafA0/59
Lectin domain (two-domain adhesins)FimH_man-bind (PF09160)513/589FimH (P08191)1UWFYesAdhesionCell surface
Fim-adh_lectin (PF09222)12/21
PapG_N (PF03627) PapG-N49/51
Cbld (PF07434) Cfae-N72/150
Non-fimbrial adhesinsBig_1 (PF02369)656/926Intimin (P19809)1F00NoAdhesionCell surface
Big_2 (PF02368)b640/884
Invasin_D3 (PF09134)5/47Invasin (P11922)1CWVNoInvasionCell surface
Fimbrial usher
Plug domainsUsher (PF00577)1938/4108FimD (P30130)3RFZNoTranslocation of fimbrial subunitsOM
C-domainsPapC_C (PF13953)964/1669cFimD (P30130)3RFZYesTranslocation of fimbrial subunitsOM
Fimbrial chaperonesPili assembly_C (PF02753)1947/4071PapD (P15319)3DPAYesFolding of fimbrial subunitsPeriplasm
Pili assembly_N (PF00345)2206/4630PapD (P15319)3DPANo
Superoxide dismutases and oxidorreductasesCopper/zinc superoxide dismutase Sod_Cu (PF00080)209/427SodC (P0AGD1)1ESOYesPhysiologic response to oxygen toxicityPeriplasm
Thiol–disulfide reductaseNDDsbD (P33655)1L6PYesElectron transferIM (periplasmic side)
DsbD N-terminal D4-A124 domain
Sugar-binding enzymesGlycogen branching enzyme431/969GlgB (P07762)1M7XNoGlycogen biosynthetic processCytoplasm
CBM_48 (PF02922)
O-Glycosyl hydrolase family 2633/854LacZ (P00722)1DP0NoCarbohydrate metabolic processCytoplasm
Glyco_Hydro_2 (PF00703)
O-Glycosyl hydrolase family 184/40ChiA (P07254)1EDQYesChitin degradationExtracellular
Chitinase A_N (PF08329)
Glucans biosynthesis protein GNDOpgG (P33136)1TXKNoBiosynthesis of glucansPeriplasm
OpgG C-terminal P397-V512 domain
ATP-binding ABC-transportersMaltose transport system aNDMalF (P02916)2R6GNoMaltose uptakeIM (periplasmic side)
MalF(P2) N93-K275 domain
Wzm/Wzt transport systemNDWzt (Q47591)2R5ONoO-polysaccharide exportIM (cytoplasmic side)
C-Wzt M264-Q431 domain
Other Ig-like containing proteinsChitin binding proteinNDCbp21 (O83009)2BEMYesChitin degradationExtracellular
Copper resistance protein169/408PcoC (Q47454)1IX2NoCopper iron bindingPeriplasm
CopC (PF04234)

Structure and function of bacterial Ig-like domains

Fimbrial adhesins, chaperones and ushers

The Ig-like domain is frequently found among fimbrial adhesins, and interestingly, it is also found in the components that constitute their specific secretory and assembly pathway, the chaperone-usher (CU) pathway (Dodson et al., 1993; Waksman & Hultgren, 2009). For recent review see (Thanassi et al., 2012). Fimbrial adhesins are proteinaceus filamenteous appendages assembled on the bacterial surface that mediate attachment to host cells and tissues (Proft & Baker, 2009). The word fimbriae (fimbria in singular) is generally used to refer to protein threads or fibers located on the bacterial surface. These adhesive organelles are made of noncovalently linked pilin subunits that are constituted by the Ig-like structure. Interestingly, the polymerization of these Ig-like subunits on the surface of the bacteria enables the adhesin to reach the target located on eukaryotic cells or abiotic surface. The morphology of these organelles can vary from thin flexible polyadhesive fibrils (2 nm of diameter) to thick rigid monoadhesive organelles (7–10 nm of diameter). Consequently, fimbriae have been divided into two main structural and functional families (Zavialov et al., 2007; Proft & Baker, 2009): the ‘adhesive pilus’ and the ‘nonpilus adhesins’ (Supporting information, Table S1).

The ‘adhesive pilus’ (pili in plural) family comprises fimbriae that have a small number of different subunits at various stoichiometries and typically displaying only one specialized adhesive domain (adhesin) on the tip of the pilus fiber. The most extensively characterized examples of this family are the E. coli Type 1 and P-pili from uropathogenic E. coli (UPEC) strains. The Type 1 pilus (Fig. 2) is a 6.9 nm wide and 1–2 μm long helical rod formed by a right-handed helical array of 500–3000 copies of the main structural pilin subunit FimA connected via FimF to a 3 nm wide tip fibrillum containing FimG and the adhesin FimH (Krogfelt et al., 1990; Hahn et al., 2002). The P pilus is a 6.8 nm wide and several micrometers long right-handed helical cylinder composed of c. 1000 copies of the major structural protein, PapA, attached to the OM by a minor structural protein, PapH, and terminated by a 2–3 nm wide tip fibrillum containing the PapG adhesin and the three minor pilin proteins PapE, PapF and PapK (Kuehn et al., 1992; Bullitt & Makowski, 1995).

Figure 2.

Fimbrial adhesins carrying Ig-like domains in Enterobacteriaceae. Schematic representation of adhesive pilus (left) and nonpilus adhesins (right) and their assembly by the CU pathway. Type 1 pili (left) and Afa adhesins of Escherichia coli (right) are examples of these two families. The subunits that constituted these fimbrial adhesins are first translocated to the periplasm through the Sec translocon located in the inner membrane (IM). Afterward, the subunits are recognized by the periplasmic chaperone and are delivered to the usher in the OM. In the figure, FimC (left) and Afa chaperone (right) are examples of the periplasmic chaperones. Because each pili subunit is folded in an incomplete Ig-like domain (down and left, exemplified by FimA) the specific chaperone donates a β-strand (G1) that stabilizes the Ig fold. This interaction is known as DSC. Once in the usher, the pili subunits suffer an exchange of the chaperone G1 strand by an N-terminal extension (Nte or G0) of the next subunit that is going to be incorporated in the fimbrial adhesin (illustrated on the left with FimA-FimA polymerization). This reaction is known as donor strand exchange (DSE).

The second fimbrial family comprises the ‘nonpilus adhesins’ that contain one or two types of subunits, being generally the main structural subunit the one implicated in adhesion. These polyadhesive fimbriae typically present amorphous or capsule-like morphology at low resolution (Zavialov et al., 2007). Relevant examples of nonpilus adhesins are the Afa/Dr adhesins from E. coli, the polymeric F1 capsular antigen of Yersinia pestis and the atypical Saf fimbriae of Salmonella enterica. Dr adhesins from UPEC and diffusely adherent E. coli (DAEC) strains, are constituted by numerous copies of a single-domain adhesin (i.e. AfaE) that assemble in a thin, flexible, filamentous polymer capped with a different subunit (i.e. AfaD) at the tip that mediates invasion of the host cell (Anderson et al., 2004ab; De Greve et al., 2007; Knight & Bouckaert, 2009; Fig. 2). The F1 capsular antigen of Y. pestis consists of high-molecular weight polymers built from a single protein subunit (Caf1) that confers, as an additional role, protection to the bacteria against phagocytosis (Zavialov et al., 2002). Saf fimbriae are composed by two subunit types, SafA and SafD. Whereas SafA forms the major pilus subunit and adhesin, SafD is classified as a putative invasin, based on sequence similarity with AfaD, and is predicted to be localized at the tip of the protein fiber (Strindelius et al., 2004; Salih et al., 2008).

Importantly, in addition to the adhesive functions that the Ig-like domain confers to fimbriae, their biogenesis also relies on the Ig-like domains found in each fimbrial subunit, in the dedicated fimbrial chaperones found in the periplasm, and in the fimbrial ushers involved in the translocation of these organelles across the OM (Sauer et al., 2004; Remaut et al., 2008; Waksman & Hultgren, 2009; Phan et al., 2011). During the biogenesis of the fimbrial adhesins, the subunits are first secreted to the periplasm, via the general secretory pathway (Sec-system; Driessen & Nouwen, 2008), where a specific fimbrial chaperone assists their folding and prevents premature assembly of the subunits (Barnhart et al., 2000; Vetsch et al., 2004). Because each fimbrial subunit presents an incomplete and unstable Ig-like fold, this fold has to be completed and stabilized by an extra β-strand donated by the chaperone, a reaction known as donor strand complementation (DSC; Sauer et al., 1999; Waksman & Hultgren, 2009). After DSC, the chaperone-subunit complexes are delivered to the OM usher, which recruits the complexes and catalyzes polymerization of subunits. Subunit polymerization is based on a concerted interaction between incomplete Ig-like domains, in this case of two consecutive fimbrial subunits, in the usher. In a reaction known as donor strand exchange (DSE) the Ig fold of the fimbrial subunit is completed by a N-terminal extension of the incoming subunit that replaces the β-strand donated by the chaperone (Fig. 2; Sauer et al., 2002; Barnhart et al., 2003; Remaut & Waksman, 2006; Vetsch et al., 2006).

As mentioned previously, Ig-like domains are also found in both the periplasmic chaperone and the OM usher. Whereas the chaperone is constituted by two complete Ig-like domains (the N and C domains; Holmgren, 1989; Choudhury et al., 1999; Waksman & Hultgren, 2009), the usher, in addition to the translocation channel, contains four complete Ig-like folds distributed among their periplasmic soluble domains (the N- terminus, C-terminus and the plug domain; Remaut et al., 2008; Yu et al., 2009; Ford et al., 2010; Phan et al., 2011). The structures and known functions of these Ig-like domains will be described later in this review.

Pilin domains and the N-terminal domain of two domain adhesins

Multiple crystal structures of chaperone–subunit complexes have been reported, FimC–FimH (Choudhury et al., 1999), PapD–PapK, PapD–PapE and PapD–PapA (Sauer et al., 1999, 2002; Verger et al., 2008), Caf1M–Caf1 (Zavialov et al., 2003, 2005) and SafB–SafA (Remaut et al., 2006). Structural data is also available for adhesive fimbrial subunit complexes FimF-FimG (Gossert et al., 2008), FimA-FimA (Puorger et al., 2011), the P pilus rod subunit PapA (Verger et al., 2007), the tip of Type 1 fimbria (Le Trong et al., 2010), and fimbrial polyadhesins subunits AfaE/DraE, DraD, DaaE, and SafA (Anderson et al., 2004ab; Pettigrew et al., 2004; Cota et al., 2006; Jedrzejczak et al., 2006; Korotkova et al., 2006; Remaut et al., 2006). These studies have revealed that the structural components of fimbrial adhesins present two general types of Ig-like folding: an incomplete Ig-like domain or ‘pilin domain’ and a complete Ig-like fold in the N-terminal domain of two domain adhesins.

The incomplete Ig-like domain is present in all the structural subunits, this is, in the pilin subunits of adhesive pili (e.g. PapA, FimA) and in nonpilus adhesin fimbriae (e.g. AfaE, DraD, DaaE, Caf1, SafA), in addition it is also found in the C-terminal domain of the tip adhesins of pili (e.g. FimH, PapG). The incomplete Ig-like fold is composed of six anti-parallel β-strands (A–F) that form a β-sandwich with two β-sheets. Hence, the seventh and last strand (G) of Ig domains is missing in these incomplete Ig-like domains (Fig. 3a). The fimbrial structural subunits, with the exception of the tip subunits, contain an N-terminal extension (Nte) peptide. This disordered tail will act as a complementing G β-strand to complete the Ig fold of the previous subunit. Strands B, E and D form one β-sheet packed against the second one in such a way that the hydrophobic side chains of the two sheets face each other. This produces, in the absent of the seventh strand, a large hydrophobic groove on the side of the pilin subunit (Choudhury et al., 1999; Gossert et al., 2008). The second β-sheet consists, in the case of the subunits from the adhesive pili, of strand A, the N-terminal donor strand G and strands F and C. In the case of the subunits of nonpilus adhesin fimbriae, this second β-sheet does not include the A strand. Independently on the fimbriae family type, the Nte extension of pilus subunits (corresponding to G strand) contain a set of well-conserved alternating nonpolar residues (named as P1–P5 residues) that will occupy the groove of the subunit in positions known as P1–P5 pockets or sites (Remaut et al., 2006; Waksman & Hultgren, 2009). The residue localized at P4 is a strictly conserved Gly, as it is the only one that can avoid steric constrains in the groove and properly adjusts at the corresponding position of P4 pocket (Gossert et al., 2008).

Figure 3.

Ig-like domains present in the subunits of fimbrial adhesins. (a) Incomplete Ig-like domain. FimA of Type 1 pili (left) and AfaE pili subunits of Dr adhesins (right) are examples of adhesive pilus and nonpilus adhesins, respectively. The incomplete Ig-like domain of these subunits is constituted by six β-strands (a–f) organized in two β-sheets (in red and in blue in the figure). A hydrophobic groove on the subunits is left because of the lack of the seventh β-strand of complete Ig domains. This hydrophobic groove is stabilized by the periplasmic chaperone or the next subunit incorporated in the fimbriae, which donate a β-strand (G1 of the chaperone and G0 of fimbriae subunits, represented in light blue). The 3D structures of FimA (pdb: 2JTY) and AfaE (pdb: 1RXL) show the incomplete Ig-like domain completed by the Nte of an incoming FimA (G0, light blue, on the left) or an incoming AfaE subunit (G0, light blue, on the right), respectively. (b) Complete Ig-like domains. Structure of FimH as an example of two domains adhesins carrying one complete Ig-like domain (lectin domain, LD) and one incomplete Ig-like domain (pilus domain, PD). The two β-sheets that constituted the Ig domains are shown in blue and red for the incomplete Ig-like domain and in orange and purple for the complete Ig-like domain. In the figure, it is showed the movement of residues Valine 155 and Glycine 115 after the interaction of the lectin domain with its receptor. Images generated with PyMOL program (Molecular Graphics Systems, LLC).

The complete Ig-like fold is present in the N-terminal domain of the tip adhesins of fimbriae and pilus (e.g. F17-GII, PapG, FimH and CfaE). These proteins are two domain adhesins consisting of a N-terminal (carbohydrate) receptor-binding domain (lectin domain) and a C-terminal pilin domain joined by a short interdomain linker. Whereas the C domain presents an incomplete Ig fold of structural pilin domains, the N-terminal domain has a complete Ig fold (De Greve et al., 2007) with variable number of β-sheets. Beside the difference in the number of β-sheets that constitutes the lectin domain, all of them encode a functional carbohydrates binding site, named sugar-binding pocket (Fig. 3b; De Greve et al., 2007). For instance, in the case of FimH, the N-terminal region consists in an elongated Ig-like fold with 11 β-strands (Choudhury et al., 1999) that mediate mannose-specific adhesion (Bouckaert et al., 2005; Le Trong et al., 2010). In the case of F17-G, the overall fold of the N-terminal domain consists in a β-sandwich with a back sheet of five anti-parallel β-strands and a front sheet of four anti-parallel strands that enable the bacteria to attach to N-acetylglucosamine (GlcNAc; Buts et al., 2003). The PapGII N-terminal domain is larger than that of FimH and F17-G and is composed by a central antiparallel β-sheet of six strands flanked by two double-stranded β-sheets on one side, and by an α-helix on the other side, bearing in this sugar-binding structure the site for Galα1-4Gal (Dodson et al., 2001).

An additional surprising feature of two domain adhesins has been recently described once the structure of the whole fimbrial tip of Type 1 fimbriae was solved (Le Trong et al., 2010). This consists on an important conformational change of the N-terminal domain of FimH when it binds to its carbohydrate receptor at the tip of the fimbria. In the absence of mannose, the N-terminal domain of FimH is maintained in a compressed conformation against the C-terminal domain, being the sugar-binding pocket present in a loose (open) form and the short interdomain linker trapped by both domains (Fig. 3b, left). However, when FimH is bound to its ligand, the sugar-binding pocket is tighten around it, causing a switch from the compressed conformation to an elongated one making the N-terminal domain to be separated from the C-terminal domain (pilin domain) and the linker to be exposed (Le Trong et al., 2010; Kisiela et al., 2011; Tchesnokova et al., 2011). This interesting phenomenon is proposed to be the basis for a mechanical force regulation of FimH adhesion (Aprikian et al., 2011) and for the allosteric regulation of the pilin domain by which FimH is maintained in a low affinity state through internals contacts (Le Trong et al., 2010). Upon interaction with the ligand and/or tensile-force, both domains are separated from each other and the lectin domain untwists and holds tightly the ligand acquiring the elongated conformation of high affinity (Fig. 3b, right).

Periplasmic fimbrial chaperones

After release from the Sec translocon, pilin subunits are recognized by a dedicated periplasmic chaperone. This interaction not only avoids their aggregation and degradation but also maintains the subunits in a polymerization-prone folding state until the complex is targeted to the OM usher (Jones et al., 1997; Vetsch et al., 2004). The structures of several periplasmic chaperones are known: PapD (P-pili; Sauer et al., 1999), FimC (Type 1 fimbria; Pellecchia et al., 1998; Choudhury et al., 1999), SfaE (S pili; Knight et al., 2002), Caf1M (F1 capsule; Zavialov et al., 2003), FaeE (F4 Fimbria; Van Molle et al., 2005), SafB (Saf Pilus; Remaut et al., 2006). These structures reveal that they consist of two Ig-like domains joined at 90° angle, which are separated by a large cleft that accommodates the bound subunit (Holmgren & Branden, 1989; Holmgren et al., 1992; Kuehn et al., 1993). The Pfam database has classified the domains as Pili assembly_N and _C. As an example, the PapD domain 1 (Pili assembly_N) is located at the N-terminus of the protein and folds as a β-sandwich of seven anti-parallel β-strands that is very similar to that of the V domain of Igs (Fig. 4). The C-terminal domain of PapD (Pili assembly_C) presents an Ig-like β-sandwich of eight anti-parallel β-strands in which the H strand is disulfide bonded to strand G. However, as the H strand is not always present in PapD homologs it should not be considered as a conserved feature of periplasmic chaperones (Holmgren et al., 1992).

Figure 4.

The periplasmic fimbrial chaperones. (a) PapD chaperone complementing one of the pilus subunits of Pap-pili, PapGII (from PDB: 3M0). The N and C-domains of PapD chaperone consist of two Ig-like domain joined at 90° angle. In the figure, the two β-sheets that constitute the Ig domains of PapD chaperone are showed in red and in blue. The complementing G1 strand of the chaperone is labeled. (b) Structure of Caf1M chaperone as an example of FGL chaperone complementing a Caf1 pilus subunit (from PDB: 1Z9S). In this case, the two β-sheets that constitute the Ig-like domain of Caf1 pilus subunit are showed in red and in blue. The complementing G1 strand and the unstructured long F1–G1 loop (dashed line) of the chaperone are labeled. Images generated with PyMOL program (Molecular Graphics Systems, LLC).

Structural data of pilin-chaperone complexes show that the interactive surfaces between the pili subunits and the chaperone are formed two specific areas which include: (1) two conserved basic residues located in the large cleft left between the two Ig-like domains of the chaperone; and (2) by the first and the seventh β-strands of the N-terminal Ig-like domain (A1 and G1, respectively). As mentioned above, the incomplete Ig fold of pilin subunits left a hydrophobic groove in the barrel between strands A and F that makes these polypeptides unstable in the periplasm unless they interact with the chaperone. In DSC the chaperone donates a motif of four alternating hydrophobic residues (P1–P4 residues) of its G1 β-strand to the subunit, thus capping the hydrophobic groove (P1–P4 pockets) and stabilizing the Ig-like structure of the subunit. This reaction produces a complete, noncanonical, Ig-like fold by which the chaperone G1 β-strand aligns parallel to the F strand of the pilin subunits (Fig. 4a and b).

On the basis of conserved structural features found in the flexible loop that connects the F1 and G1 β-strands of the chaperone N domain, Hung and collaborators classified the periplasmic chaperones into two distinct classes: the FGL (F1G1 long) and FGS (F1G1 short) chaperones (Hung et al., 1996). Interestingly, these two groups of chaperones assemble the two groups of adhesive pilus and nonpilus adhesive fimbriae, respectively (Soto & Hultgren, 1999; Waksman & Hultgren, 2009; Zav'yalov et al., 2010). FGL chaperones contain a significantly longer G1 strand and longer F1-G1 loop stabilized by a disulfide bridge between two conserved Cys residues of F1 and G1. In addition, FGL chaperones also contain a longer binding motif at the N-terminus, which extends the A1 strand by at least three residues (Zavialov et al., 2007). G1 strand of FGL chaperones contain an additional alternating hydrophobic residue (P5 residue) that inserts intermittently into a corresponding P5 pocket of the groove, which is never occupied in the case of FGS chaperones.

During subunit polymerization, the complementing G1 strand of the chaperone is replaced by the Nte extension on the incoming subunit. This reaction is known as DSE and takes place in the OM usher (Saulino et al., 1998; Thanassi et al., 1998, 2012; Sauer et al., 2004; Vetsch et al., 2006; Waksman & Hultgren, 2009). In contrast to the complementation achieved by the chaperone, the canonical Ig fold of the subunits is completed by the Nte strand in an anti-parallel orientation to the F β-strand (Fig. 2; Sauer et al., 2002). DSE is proposed to occur through a concerted ‘zip in-zip out’ mechanism that is initiated by the insertion of the Nte P5 residue into the P5 pocket of the groove of the previous subunit bound to the usher (Zavialov et al., 2003; Remaut et al., 2006). Once the P5 pocket is occupied by Nte P5 residue of the next subunit, a transient ternary chaperone-subunit-subunit complex is formed allowing a gradual displacement of the chaperone G1 strand (Remaut et al., 2006; Waksman & Hultgren, 2009). In addition to the initiating role in DSE of the P5 pocket, the accessibility of Nte to this pocket of specific subunit also marks DSE kinetics and the termination of the biogenesis of adhesive organelles (Rosen et al., 2008; Verger et al., 2008). As an example, in the P pilus system termination of the pilus depends on the last subunit PapH that lacks the P5 pocket. This subunit undergoes DSE with the last PapA subunit but is unable to undergo DSE with any other subunit, resulting in the truncation of pilus assembly (Verger et al., 2006).

Ig domains in OM fimbrial ushers

The polymerization of adhesive organelles takes place at the OM usher. This membrane protein not only recruits chaperone-subunit complexes but also coordinates the mentioned DSE reaction and translocates the growing adhesive organelle to the OM (Waksman & Hultgren, 2009). Ushers are integral OM proteins of c. 800 amino acid residues that comprise four functional domains: two periplasmic soluble N-terminal and C-terminal domains that interact with the pilus subunit-chaperone complexes, a β-barrel domain that constitutes the translocation channel and a plug domain, located in the middle of the channel, occluding the lumen in resting ushers, or underneath the translocation domain in active ushers (Fig. 5; Remaut et al., 2008; Huang et al., 2009; Phan et al., 2011). The N-terminal domain of ushers has been shown to form the initial binding site for subunit-chaperone complexes (Nishiyama et al., 2003, 2005; Ng et al., 2004) with a very fast recruiting activity (Nishiyama & Glockshuber, 2010). Using crystallographic and nuclear magnetic resonance (NMR) to characterize the N-terminus domain of FimD usher from Type 1 fimbriae (Nishiyama et al., 2005), it has been revealed that this domain is composed of a disordered N-terminal tail that becomes structured upon binding of chaperone-subunit complex, a folded core of six β-strands and a hinge segment to the β-barrel domain. Interestingly, in adhesive pilus such as Type 1 fimbriae, this binding features different affinities depending on the type of subunit-chaperone complex that parallel the relative position in the polymerized fimbriae (Nishiyama et al., 2003; Ng et al., 2004; Li et al., 2010; Nishiyama & Glockshuber, 2010).

Figure 5.

The OM usher of fimbrial adhesins. (a) Structure of FimD protein as representative of fimbrial ushers. In the 3D structure of FimD (from PDB: 3RFZ) are indicated the Ig domains present in fimbrial ushers, the C-terminus domains 1 and 2 (CDT1 and CTD2) and the plug domain. The two β-sheets that constituted the Ig domains are shown in red and in blue. In addition, the β-barrel translocation domain and the N-terminus domain containing a β-strand structure are also shown. (b) Conformational change of fimbrial ushers shown with the structure of FimD-FimH-FimC complex (from PDB: 3RFZ) and FimD alone (from PDB: 3OHN). The plug domain (in red) containing an Ig-like domain moves from inside the lumen of the channel (left, in closed state) to a position beneath the translocation domain when the channel is open (right, in open activated state). Top and side views illustrate the rotation of the plug domain outside the channel and enable to show other structures inside the lumen of the channel, the β5-6 hairpin and the α-helix. Images generated with PyMOL program (Molecular Graphics Systems, LLC).

The three-dimensional structure of the C-terminal region of PapC usher (Ford et al., 2010) and the structure of the full-length FimD usher (Phan et al., 2011) have been solved, indicating that the C-terminal regions of these ushers are composed by two Ig-like domains of seven β-strands, named CTD1 and CTD2 (Fig. 5a; Phan et al., 2011). These C-terminal domains also interact with subunit-chaperone complexes (Thanassi et al., 2002; So & Thanassi, 2006; Phan et al., 2011). In vivo and in vitro experiments carried out with deletion mutants indicate that this region is necessary for the organelle assembly (So & Thanassi, 2006; Huang et al., 2009).

Structure of the embedded OM part of PapC and FimD showed that the β-barrel domain of ushers is composed by 24 β-strands (Remaut et al., 2008; Huang et al., 2009; Phan et al., 2011). This translocation channel presents a kidney-shaped with a pore size large enough for the passage of individual folded subunit but not for any chaperone-subunit complexes (Fig. 5b; Huang et al., 2009). The plug domain is inserted into the loop connecting β6 and β7 strands and is positioned laterally inside the β-barrel gating the lumen of the pore (in resting ushers) or underneath the translocation channel (in actively translocating ushers; Fig. 5b; Remaut et al., 2008; Huang et al., 2009; Phan et al., 2011). The crystal structure of the plug domains in FimD and PapC ushers indicated that they are composed by six β-strands with a fold similar to an Ig fold (Remaut et al., 2008; Huang et al., 2009; Phan et al., 2011). Crystallization of the isolated plug domain of Caf1A usher suggested that this domain could be composed of seven β-strands as a result of the swapped dimerization of two monomers. Each monomer has the classical s-type Ig fold constituted by six β-strands and a seventh additional β-strand coming from the N-terminus of the second monomer (Yu et al., 2009). Interestingly, mutations located in the corresponding N-region of the plug domain of PapC abolish the adhesive capacity of the bacteria to eukaryotic cells, suggesting the loss of P fimbriae (Henderson et al., 2004). From in vitro and in vivo experiments it is deduced that the plug domain not only acts merely as a barrier occluding the translocation channel, but it appears to be involved in the polymerization process (Huang et al., 2009; Mapingire et al., 2009; Yu et al., 2009).

The three-dimensional structure of FimD bound to its cognate FimH-FimC substrate (Phan et al., 2011) revealed that the usher suffers an important conformational change from a closed nonactive state to an open active one (Fig. 5b). Although this conformational change implies the movement of the plug domain from the lumen of the channel to a position beneath the translocon, the mechanism by which this conformational change occurs is not yet known. However, in some of the adhesive fimbriae system it has been identified a natural activator of the usher. This is the case of the type 1 fimbriae, in which it has been shown that FimH is involved in the activation of FimD usher (Nishiyama et al., 2008). Importantly, it has been demonstrated that the lectin N-terminal domain of FimH, in addition to its adhesive capacity, is required for recognition of FimH by FimD and is essential for activation of the usher (Munera et al., 2007, 2008; Nishiyama et al., 2008). An interesting possibility is that the N-terminal domain of FimH, being the activator, could trigger the initial displacement of the plug domain to leave free space for its passage across the lumen of FimD.

Nonfimbrial OM adhesins: the family of intimins and invasins

The nonfimbrial adhesins containing Ig-like domains belong to the family of intimins in strains of enteropathogenic E. coli (EPEC), enterohaemorrhagic E. coli (EHEC), Citrobacter spp., and Hafnia alvei and invasins found in strains of Yersinia spp. Intimins and invasins are large OM proteins (OMPs) of c. 900 amino acids related to each other in terms of sequence and structure, that protrude from the bacterial membrane to mediate bacterial invasion (invasin) or adhesion (intimin) to their host cell. Invasin is encoded by the chromosomal inv gene and this polypeptide mediates bacterial entry into M cells at the Peyer's patches of the gut (Isberg et al., 1987; Marra & Isberg, 1997) by binding to β1-integrin receptors (Isberg & Tran Van Nhieu, 1994). This process triggers the internalization of the pathogen by a zipper-like mechanism (Isberg et al., 2000). Intimin is encoded by the eae gene located in the locus of enterocyte effacement (LEE; McDaniel et al., 1995), and is responsible for the intimate adhesion of EPEC, EHEC and Citrobacter bacteria to the enterocytes of the gastrointestinal tract where they produce a characteristic cytopathic effect known as an attaching and effacing (A/E) lesion (Moon et al., 1983). This event is mediated, in part, by binding of intimin to its cognate receptor termed Tir (Tanslocated intimin receptor; Kenny et al., 1997), which is a bacterial effector injected by the bacteria in the host cell plasma membrane via a type-III secretion system (Frankel & Phillips, 2008).

Intimins and invasins contain two functionally distinct regions, an amino (N) and a carboxy (C)-terminal region (Fig. 6a). The N-region comprises a signal peptide for Sec-dependent secretion across the IM, a hydrophilic periplasmic domain (Lys-M type in Gammaproteobacteria) that could tether the polypeptide to the peptidoglycan layer, and a large β-domain of about 500 amino acids that forms a β-barrel predicted to contain at least 12 amphipathic β-strands (Touze et al., 2004; Bodelón et al., 2009; Tsai et al., 2010). The β-domain participates in the secretion of the C-region to the extracellular milieu and directs both the dimerization of intimin and Tir-clustering in the cell host membrane. The reported structures of the polypeptides comprising the extracellular C- region of 497 amino acids of invasin (Inv497) from Yersinia pseudotuberculosis (Hamburger et al., 1999), the C-terminal 280 amino acids of intimin (Int280) from EPEC (Kelly et al., 1999; Luo et al., 2000) and 188 amino acids (Int188) from EHEC (Yi et al., 2010), along with structural predictions of the rest of the C-region, show a rod-like structure consisting of four (invasin D1–D4) or three (intimin D0–D3) Ig-like domains (SCOP 49373) followed by a C-type lectin-like domain (invasin D5 and intimin D4) responsible for receptor-binding and located at the C-terminal tip of the molecule (Fig. 6b). The lectin-like domain and the most distal Ig-like domain of intimins and invasins form a rigid superdomain that participates in the binding to Tir (Tir-binding domain; Batchelor et al., 2000; Yi et al., 2010) or β1-Integrins (Leong et al., 1990), respectively (Niemann et al., 2004). The Tir–binding domain of EHEC and EPEC, despite its low sequence identity (48%), adopt a very similar structure, confirming previous data in which both intimins can cross-complement in vitro (Yi et al., 2010). The cell-binding activity of intimins and invasins, depends on a conserved intra-domain disulfide bond that is present in similar places in the lectin-like domain of these proteins (Leong et al., 1993; Frankel et al., 1995; Batchelor et al., 2000). The overall topology of Inv497 D1-D3, EPEC Int280 D1-D2 and EHEC Int188 D2 Ig-like domains (there is no structural data available from intimin D0) resembles that of the IgSF type-I set. Inv497 D4 is similar to the IgSF C1 set (Hamburger et al., 1999; Kelly et al., 1999; Yi et al., 2010). The Int/Inv Ig-like domains are classified into the Pfam database as Bacterial Ig-like domains 1 and 2 (Big_1 and Big_2) and invasin_D3: The D1–D4 domains of invasin are classified as Big_1 (D1), invasin_D3 (D3) and Big_2 (D2 and D4). The D0 to D2 domains of intimin are classified as Big_1 (D0 and D1) and Big_2 (D2; Fig. 6b).

Figure 6.

Intimin and Invasin nonfimbrial adhesins. (a) Scheme of the N and C-regions of intimin and invasin showing the signal peptide (SP) for Sec-dependent secretion across the inner membrane (IM), the LysM domain, the β-domain and the secreted D domains. (b) Ribbon diagrams of the Int280 and Inv497 crystals showing the string of Ig-like domains and the C-terminal lectin-like domain. The PFAM terms for the Ig-like domains is indicated below each domain. (c) Intimin secretion to the extracellular medium starts with its translocation across the IM via the Sec translocon. In the periplasm DsbA catalyzes the formation of the disulfide bond (S-S) present in the D3 domain. The SurA pathway delivers the β-domain (Int550) to the BAM complex, where BamA is responsible for its folding and OM insertion. The alternate DegP/Skp pathway is depicted with a dashed line. The LysM domain is predicted to bind the peptidoglycan (PG) layer. Translocation of the secreted domains D0 (green oval), D1, D2 and D3 is believed to take place concomitantly with insertion of the β-domain in the OM via the BAM complex.

Interestingly, a recent study has identified a new member of the intimin/invasin family in extraintestinal pathogenic E. coli strains causing urinary tract infections and sepsis, but which is also conserved in the genomes of pathogenic E. coli strains causing enteric diseases (Nesta et al., 2012). This novel OM adhesin, called factor adherence E. coli (FdeC), is predicted to have an OM β-barrel and a large C-terminal region of about 900 amino acids containing nine surface-exposed Ig-like domains, but no evidence of a lectin-like domain. The 3D structure of a soluble FdeC fragment comprising three internal domains of the surface-exposed region confirmed their Ig-like structure (Nesta et al., 2012). It has been also demonstrated that the exposed Ig-like domains of FdeC are able to bind epithelial cells and different collagens types in vitro. FdeC is expressed in vivo during colonization of the uroepithelium and, remarkably, immunization with the exposed Ig-like domains protects from infection (Nesta et al., 2012).

A phylogenetic study of the Int/Inv family identified 69 sequence-divergent proteins present in Alpha-, Beta- and Gammaproteobacteria as well as Chlamydia, demonstrating that these adhesins comprising a β-barrel and Big motifs are not limited to enterobacteria, but they are widely distributed among other Gram negatives (Tsai et al., 2010). This tandemly arrayed Ig-like domains, typical of many eukaryotic cell adhesion molecules or cell surface receptors, is also present in adhesins of Gram-positive pathogenic bacteria that belong to the microbial surface components recognizing adhesive matrix molecules (MSCRAMMs) such as the staphylococcal adhesin binding to fibrinogen (Ponnuraj et al., 2003), or in Gram-positive pilin proteins like SpaA from Corynebacterium diphteriae (Kang et al., 2009). Also, cell surface proteins of pathogenic Leptospiral species termed Leptospiral Ig-like (Lig) proteins present a structural organization of its extracellular domain very similar to that of intimins and invasins with tandem Ig-like domains identified as Big_2 in Pfam (Matsunaga et al., 2003).

The mechanism of secretion of intimin and invasin (Int/Inv) family of proteins has not been elucidated. However, secretion of their members follows a multistep process described for most OMPs, which involves their translocation through the IM via the Sec-system, transport across the periplasm interacting with chaperones (e.g. SurA, DegP, Skp) and disulfide bond enzymes (e.g. DsbA), and insertion in the OM mediated by the β-barrel assembly machinery (BAM) complex (Fig. 6c; Ruiz et al., 2006; Knowles et al., 2009; Hagan et al., 2011; Ricci & Silhavy, 2011; Dalbey & Kuhn, 2012). It has been demonstrated that intimin requires the BAM complex for OM insertion and that SurA is the major periplasmic chaperone involved in folding of intimin β-barrel (Bodelón et al., 2009). The protease activity of DegP participates in the degradation of unassembled intimin β-barrel in the periplasm. In addition, periplasmic DsbA catalyzes the formation of the disulfide bond of the lectin-like domain of intimin (Bodelón et al., 2009), therefore indicating that this domain is at least partially folded in the periplasm prior to its translocation across the OM.

The mechanism of secretion of Int/Inv proteins may be similar to that used by autotransporters (ATs), a large superfamily of secreted virulence factors that contain an OM-anchored β-domain at the C-teminus, and a secreted ‘passenger’ domain at the N-terminus (Dautin & Bernstein, 2007; Leo et al., 2012; Leyton et al., 2012). The secreted domains of Int/Inv and ATs present different structures (a rod of Ig-like domains vs. a β-helix rod, respectively), and they are located in opposite sites of the polypeptides, but their translocation across the OM might follow similar mechanisms (Touze et al., 2004; Adams et al., 2005; Bodelón et al., 2009; Leo et al., 2012). Originally, it was proposed that these proteins could use a ‘self-translocation’ mechanism in which the β-barrel inserts in the OM forming a hydrophilic protein-conducting channel that is used for secretion of the passenger domain (Pohlner et al., 1987; Henderson et al., 1998). The crystallographic structures of ATs β-barrels show the existence of an α-helix that fills the hydrophilic lumen of the β-barrel connecting it to the ‘passenger’ domain (Oomen et al., 2004; Barnard et al., 2007; van den Berg, 2010). This α-helix is essential for the translocation process in ATs and, along with the β-barrel, constitutes a functional transport unit (Marín et al., 2010). Interestingly, a similar region is predicted in the members of the Int/Inv family after the putative twelfth β-strand (Tsai et al., 2010). However, data indicating that the Int/Inv and AT proteins can translocate Ig-like and other protein domains in, at least, a partially folded conformation, are difficult to explain in the context of the narrow hydrophilic channel provided by the β-barrels (Brandon & Goldberg, 2001; Veiga et al., 2004; Skillman et al., 2005; Purdy et al., 2007; Bodelón et al., 2009; Marín et al., 2010). Therefore, an ‘assisted-translocation’ mechanism has been proposed for AT and Int/Inv proteins, in which insertion of the β-barrel and translocation of the surface-exposed domains are both assisted by the BAM complex (Bernstein, 2007; Bodelón et al., 2009; Ieva et al., 2011; Rossiter et al., 2011; Leyton et al., 2012). Interestingly, a recent study has discovered a novel conserved, but nonessential, OM protein member of the BamA(Omp85)-family, called TamA, which interacts with an integral IM protein (TamB) to assemble a transmembrane complex that promotes efficient secretion of ATs (Selkrig et al., 2012). Whether this newly identified TAM-complex, or the essential BAM complex, forms the actual OM translocon for the Ig-like domains of Int/Inv proteins is unknown.

Enzymes with Ig-like domains in E. coli and enterobacteria

Interestingly, the Ig-like domain is not only limited to structures that mediate bacterial adhesion, and thus, enterobacteria have adopted this type of folding in proteins bearing enzymatic functions. Ig-like domains are present in enzymes involved in an ample variety of cellular processes such as anti-oxidative damage, protein folding and, biosynthesis, degradation or transport of sugars (Table 1). The majority of these enzymes are complex structures formed by different functional domains, in which the Ig-like domain usually displays binding functions. For example, some of the Ig-like domains identified in sugar-binding enzymes are carbohydrate-binding modules (CBMs) that bring catalytic domains into close proximity with their cognate substrates (Boraston et al., 2004; Guillen et al., 2010). However, the striking example of the Copper, Zinc Superoxide Dismutase, an enzyme formed by a single Ig-like domain that has catalytic activity, confirms the versatility and evolutionary success of this topology.

Superoxide dismutases and oxidoreductases

Periplasmic Cu,Zn-superoxide dismutases (SODs)

The Cu,ZnSODs are metalloenzymes present in nearly all eukaryotic cells and in a large number of bacteria (Fang et al., 1999; Battistoni, 2003) that catalyze the dismutation of superoxide to oxygen and hydrogen peroxide (McCord & Fridovich, 1969; Fridovich, 1995) protecting these organisms from oxygen-mediated free-radical damage (Bannister et al., 1987). Both eukaryotic and prokaryotic Cu,ZnSODs share a conserved structure based on an Ig-like fold (Richardson et al., 1976; Halaby et al., 1999; Khare et al., 2003; Culotta et al., 2006; SCOP 49330). Whereas eukaryotic Cu,ZnSODs are homodimeric enzymes (Tainer et al., 1982) characterized by high structural conservation (Bordo et al., 1994; Perry et al., 2010), the prokaryotic Cu,ZnSODs present larger structural variability than the eukaryotic counterparts (Pesce et al., 2000) and, interestingly, some bacterial variants were isolated as active enzymes either as monomers [e.g. E. coli (Pesce et al., 1997), S. enterica (Mori et al., 2008), Brucella abortus (Chen et al., 1995)] or as homodimers [e.g. Salmonella typhimurium (Pesce et al., 2000), Photobacterium leiognathi (Bourne et al., 1996)].

The Cu,ZnSOD of E. coli, encoded by the sodC gene, is a periplasmic polypeptide of about 15 kDa that is specifically induced in stationary phase (Benov & Fridovich, 1994; Imlay & Imlay, 1996). The crystal structure of this enzyme (Pesce et al., 1997) revealed a monomer composed by eight antiparallel β-strands with the two β-sheets of the sandwich each containing four strands (A, B, C and F and E, D, G and H) that are connected by seven loops (Fig. 7a). Two of them, termed the electrostatic and zinc loops (Tainer et al., 1982; Bordo et al., 1994), enclose the enzyme active center that is composed by one copper- and one zinc-binding sites linked via a histidine ligand (Pesce et al., 1997). Whereas Copper is at the catalytic center of the enzyme and is cyclically oxidized and reduced during the successive encounters with the superoxide anion, it is believed that Zinc greatly contributes to the structural stability of the enzyme, modulation of the redox properties of the Copper ion, resistance to denaturing agents and proteolytic enzymes (Spagnolo et al., 2004; Perry et al., 2010). The Cu,ZnSODs of E. coli (SodC) and S. enterica (SodCII) have an intra-subunit disulfide bond between two cysteines located in the S-S subloop and in strand H (Cys 50 and Cys 144 in SodC) that is required for proper stability and formation of the Zinc-binding pocket of the active site of the enzyme (Pesce et al., 1997; Mori et al., 2008). The in vivo folding of the Cu,ZnSOD of E. coli depends on DsbA, which catalyzes the formation of its disulfide bond (Battistoni et al., 1999).

Figure 7.

Ig-like domains in Escherichia coli oxidoreductases. (a) Copper/zinc superoxide dismutase SodC. Ribbon diagram of the Ig-like periplasmic SOD SodC indicating the amino acid residues that contribute to the Copper (green sphere) and Zinc (brown sphere) binding sites. (b) DsbD thiol disulfide oxidoreductase. The thioredoxin (Trx-1) transports two electrons from cytoplasmic NADPH to the transmembrane domain of DsbD (tDsbD), and across the membrane via periplasmic cDsbD to the Ig-like domain nDsbD. This polypeptide functions as a hub and delivers the electrons to periplasmic substrates (e.g. DsbC, DsbG) or to IM proteins (e.g. CcmG). The black arrows indicate the electron flow.

As many Gram-negative bacteria contain a periplasmic Cu,ZnSOD (Gort et al., 1999), it has been suggested that this enzyme plays an important role for scavenging superoxide species from the periplasm generated during aerobic growth, possibly attributable to electron leakage from the respiratory chain (Korshunov & Imlay, 2006). However, the fact that several pathogenic bacteria including B. abortus, Legionella sp., E. coli, Salmonella sp., or Neisseria meningitidis contain this enzyme, has led to the hypothesis that the Cu,ZnSOD could serve to protect these microorganisms against host defense-derived free-radical-mediated damage, thus facilitating bacterial survival within the infected host (Lynch & Kuramitsu, 2000; Battistoni, 2003; Perry et al., 2010). Interestingly, in contrast to the nonpathogenic E. coli K-12 strain that has a single sodC gene, the pathogenic EHEC O157:H7 strain contains three sodC genes. One of them is homologous to that of the K-12 strain, and the other two, which encode almost identical proteins, are embedded within the sequences of two lambdoid prophages (CP-933R and CP-933V in the EDL933 strain; D'Orazio et al., 2008). This redundancy also occurs in virulent Salmonella strains, which, besides the chromosomal sodCII gene, have a bacteriophage-encoded copy (sodCI) that contributes to Salmonella virulence (Fang et al., 1999). These and other studies have established a role of the prophage associated SodC in virulence (Ammendola et al., 2008; D'Orazio et al., 2008).

Inner membrane disulfide oxidoreductase DsbD

The E. coli DsbD is an inner membrane (IM) protein of 546 amino acids composed of a periplasmic N-terminal (nDsbD) domain (residues 1–143) with an Ig-like fold, a central transmembrane (tDsbD) domain (residues 144–418) and a C-terminal (cDsbD) periplasmic domain (residues 419–546; Chung et al., 2000; Rozhkova et al., 2004). Each of the domains contains two redox-active cysteines that are involved in the transfer of electrons from cytoplasmic NADPH, via thioredoxin-1 (Trx-1; Rietsch et al., 1997; Cho & Beckwith, 2009), to various periplasmic targets (Stewart et al., 1999; Katzen & Beckwith, 2000). In the first step, tDsbD transfers electrons from Trx-1 to cDsbD, which in turn reduces the Cys103-Cys109 disulfide of nDsbD (Chung et al., 2000; Katzen & Beckwith, 2000; Collet et al., 2002), which enables a transient disulfide bond between Cys 109 of nDsbD and Cys 461 of cDsbD leading to the formation of a mixed disulfide complex termed nDsbD-SS-cDsbD (Rozhkova et al., 2004). Next, nDsbD subsequently transfers the reducing potential to various substrates including periplasmic DsbC (Missiakas et al., 1994) and DsbG (Bessette et al., 1999; Depuydt et al., 2009), and the membrane-anchored protein disulfide oxidoreductase DsbE/CcmG involved in c-type cytochrome maturation (Fabianek et al., 1998).

The structures of the oxidized free form of E. coli nDsbD (Goulding et al., 2002) and of its mixed disulfide complex with either cDsbD (Rozhkova et al., 2004) or the periplasmic partners DsbC (Haebel et al., 2002) and DsbE/CcmG (Stirnimann et al., 2005) reveal that, unlike other Dsb proteins such as DsbA, DsbC, DsbE and cDsbD which all possess a thioredoxin-like fold, nDsbD features an Ig-like fold (SCOP 74863; Fig. 7b). This Ig-like domain of nDsbD consists of a β-sandwich formed by two β-sheets, each of which is constituted by three anti-parallel strands (β1, β2, β8 and β4, β10, β12, respectively; Goulding et al., 2002). However, in the nDsbD-DsbC complex the Ig-like domain is composed by a four-stranded β-sheet (β1, β2, β8 and β5) that packs against a three-stranded β-sheet (β4, β9 and β12; Haebel et al., 2002), indicating that the topology of the domain varies depending on whether it is bound or not to its periplasmic substrates. The structural data also shows that the active site containing Cys103 (strand β11) and Cys109 (strand β10) is located at the N-terminus of the Ig-like domain. Interestingly, the cystein-active site of the oxidized form of nDsbD is protected from illegitimate redox reactions by a cap-loop region located between strands β6 and β7 of the Ig-like domain. In contrast, the active site in complexed-nDsbD is in an open conformation exposing the catalytic sulfur of Cys109. Therefore, opening of the cap-loop could be a prerequisite for the formation of an interdisulfide bond between Cys109 of nDsbD and the catalytic Cys of its partners (Stirnimann et al., 2006; Quinternet et al., 2009).

Sugar-binding enzymes

Cytoplasmic glycogen branching enzymes (GBEs)

The biosynthesis of glycogen in bacteria requires the participation of GBEs (EC that cleave the linear α-1-4 linked glucose chain and the resulting oligosaccharide is subsequently linked in α-1-6 position to the carbohydrate chain. This branching activity increases the number of nonreducing ends, thus making glycogen more reactive to synthesis and digestion. Bacterial GBEs are classified into the glycoside hydrolase family 13 (GH13; Stam et al., 2006) that belongs to the α-amylase superfamily (Jespersen et al., 1991; Abad et al., 2002). Structurally, bacterial GBEs are multidomain enzymes containing a central catalytic A-domain with a ‘TIM’ barrel folding that is invariably present in all the members of the α-amylase superfamily, and two C- and N-terminal extensions termed C and N domains, respectively, domains (Jespersen et al., 1991; Palomo et al., 2009). The C domain is believed to protect the hydrophobic residues of the catalytic domain from contacts with the solvent, and it been suggested to be involved in substrate binding. The N domain contains a CBMs (CBM48) of about 150 amino acids (Palomo et al., 2009), which is classified into the SCOP database as having an Ig-like fold (SCOP 81282).

Some GBEs like the one from E. coli possess a long N-terminal region that contains two N domains of about 100 aminoacids termed N1 and N2, respectively (Lo Leggio et al., 2002; Palomo et al., 2009). Abad and collaborators elucidated the 3D structure of the E. coli GBE lacking its first 113 aminoacids (N113BE) and the crystal revealed that the N2 domain indeed adopts an Ig-like fold (Abad et al., 2002). Although the N1 domain is absent in the reported crystal, its Ig-like folding has been predicted elsewhere (Lo Leggio et al., 2002). The truncated N113BE shows around 60% of activity compared to the full-length enzyme, but its substrate preference and Km value were similar to the full-length protein. Amino terminal truncations of the E. coli GBE resulted in almost half reduction of enzyme activity (Hilden et al., 2000), altered branching pattern (Binderup et al., 2002) and to a gradual increase in the length of the glucan chains, indicating that this domain is involved in determining the size of the chain transferred (Abad et al., 2002; Devillers et al., 2003). It has been also reported that the putative CBM48 region of the N domain of two GBEs from two species of Deinococcus, dictates both the substrate and chain length specificity (Palomo et al., 2009).

Cytoplasmic glycosyl hydrolase family 2 (GHF2)

The O-glycosyl hydrolase superfamily (EC 3.2.1) is a widespread group of enzymes responsible for the hydrolysis and/or transglycosylation of glycosidic bonds. A classification system for glycosyl hydrolases (GH) which is available on the CAZy database, has led to the definition of 113 different protein families. The E. coli β-galactosidase, encoded by the lacZ gene, is a member of this family of enzymes and its structure revealed two Ig-like domains (Juers et al., 2000). The enzyme is a 464 kDa tetramer that is comprised of four polypeptide chains, named A–D, each of 1023 amino acids. Each monomer is made up of five domains, 1–5, being most of the enzyme active site residues within Domain 3 (residues 335–624), which has an α/β or ‘TIM’ barrel structure. The domains 2 (residues 219–334) and 4 (residues 625–725) of β-galactosidase have an identical Ig-like fold topology (SCOP 49305) composed by seven anti-parallel β-strands arranged in two β-sheets similar to that of FnIII (Jacobson et al., 1994; Juers et al., 2000). Importantly, the Ig-like Domain 2 contributes via a loop with amino acid residues to the completion of the active site within Domain 3 of the neighboring subunit (residues 272–288). Thus, monomer A donates its Domain 2 loop to complete the active site of monomer D and vice versa. The reciprocal situation also occurs between monomer B and C to give a total of four functional active sites (Juers et al., 2000; Matthews, 2005). The Ig-like Domain 4 presents low homology among different β-galactosidases and it may just function as a linker between Domains 3 and 5. The Ig-like fold of Domains 2 and 4 might not only be limited to the β-galactosidase of E. coli and, based on structural comparisons, similar domains are found in other glycosidases (Juers et al., 1999).

Extracellular chitinase A

Chitin is the second most abundant polysaccharide in nature. It is a linear insoluble polymer of β(1–4)-linked N-acetylglucosamine (GlcNAc), present in fungal cell walls, shells of crustaceans, and exoskeletons of insects (Khoushab & Yamabhai, 2010). Genes encoding chitin-binding proteins or proteins containing chitin-binding domains have been found in many organisms including viruses, bacteria, fungi, insects, higher plants, and mammals. Based on their amino acid sequences, three-dimensional structures, and molecular mechanisms of catalytic reactions the chitinases can be grouped into GH families 18 and 19. Originally, it was indicated that bacterial chitinases belong to family 18 of glycoside hydrolases (Henrissat & Davies, 1997), although some members are now classified under family 19 of GH (Kawase et al., 2004; Khoushab & Yamabhai, 2010). Serratia marcescens secretes to the extracellular medium four distinct types of chitinases (A, B and C1 and C2) and a chitin-binding protein (CBP21) lacking chitinase activity, which are encoded by the ChiA, B, C and Cbp21 genes, respectively (Horn et al., 2006). ChiA and CPB21 contain Sec-dependent N-terminal signal peptides that are cleaved off by a periplasmic peptidase during their secretion to the extracellular medium. Nevertheless, its mechanism of OM secretion is still unknown.

The chitinase A (ChiA) from S. marcescens is a 58 kDa protein that comprises three domains: (1) an N-terminal module (amino acids 24–137) with an Ig-like fold (SCOP 49233) similar to a FnIII domain, termed ChiN domain and identified in the Pfam database as Chitinase A_N, (2) a catalytic domain with (α/β) 8-barrel fold structure, and (3) a C-terminal domain with (α + β) structure (Perrakis et al., 1994). The ChiN domain is classified into the Cazy database as a CBM5. Although the overall topology and folding of ChiN and FnIII Ig-like domains is similar, they present a number of differences: ChiN strand A is shorter, there is long insertion between strands A and B, and strands F and G are linked by a cysteine bridge, which is absent in FnIII domains (Perrakis et al., 1994). The catalytic mechanism of this type of chitinases involves substrate-assisted catalysis, and although the function of the ChiN domain is not well defined, the crystallographic data suggests that it may help the enzyme to remain bound to the chitin chain and direct the terminal sugar residues toward the catalytic groove. The surface of the ChiN domain has four conserved tryptophan residues important for substrate affinity that could be involved in interacting with the loose ends of chitin chains (Perrakis et al., 1997). Two adjacently arranged tryptophans (Trp-33 and Trp-69), exposed on a continuous surface with the conserved aromatic residues of catalytic domain (Trp-245 and Phe-232), play important roles in guiding a chitin chain into the catalytic cleft to be hydrolyzed at the catalytic site (Uchiyama et al., 2001).

Periplasmic glucans biosynthesis protein G

Escherichia coli protein named OpgG is a 56 kDa periplasmic enzyme necessary for the production of glucans, which are linear-branched polysaccharides composed of 8–10 glucose units per molecule linked by β-1,2 linkages and branched by β-1,6 linkages, that constitute the bacterial envelope of many Gram-negative bacteria (Lacroix & Bohin, 2010). Although the precise function of OpgG remains unknown, under low nutrients and osmolarity growth conditions, E. coli and other Gram-negative bacteria synthesize glucans, which contribute to maintain the osmolarity of the periplasm preventing the swelling and rupturing of cytoplasmic membrane (Kennedy, 1982). It has been reported that OpgG is secreted via the Sec-system to the periplasm (Lequette et al., 2004) where it may interact with the glycosyltransferase OpgH, the other member of the bicistronic OpgGH operon, to catalyze the formation of β-1,6 glucose branches (Debarbieux et al., 1997; Hanoulle et al., 2004). The structure of OpgG showed that it is composed of two β-sandwich domains connected by a helix (Hanoulle et al., 2004). Whereas the N-terminal domain (residues 22–388), that bears the putative catalytic activity, displays a 25-stranded β-sandwich fold, the C-terminal domain (residues 401–512) has a seven-stranded Ig-like fold (SCOP 110054) formed by two β-sheets constituted by A, B, D, E and C, F and G anti-parallel strands. Folding of the OpgG Ig-like domain resembles that of the E. coli α-1,4-glucan branching enzyme (GlgB), suggesting that, as GlgB, it may act to modulate the enzymatic activity of the OpgG catalytic N-terminal domain.

Ig-like domains in ATP-binding cassette (ABC) membrane transporters

ABC transporters are ubiquitous across all cells and utilize the free energy of ATP hydrolysis to import or export a wide variety of substrates across biological membranes, ranging from small molecules such as ions, sugars or amino acids to larger compounds such as antibiotics, drugs, lipids and oligopeptides (Moussatova et al., 2008; Cuthbertson et al., 2010). Bacterial ABC transporters are involved in the uptake of nutrients (e.g. sugars and vitamins) and they also participate in the export of molecules (e. g. proteins, lipids, and oligo- and polysaccharides). Two examples of Ig-like domains have been reported ABC transporters in E. coli. One is present in the maltose transporter of E. coli and Salmonella (importer) and the other forms part of the LPS O-antigen transporter (exporter) of certain E. coli strains.

Despite their ample functional diversity ABC transporters share a general domain architecture: Two transmembrane domains (TMDs) that dimerize to form the substrate translocation pathway, coupled to two nucleotide-binding domains (NBDs), also known as ABC, that control the conformation of the TMDs through ATP-induced dimerization and hydrolysis-induced separation (Davidson et al., 2008; Procko et al., 2009). In addition to these four domains that form the core transporter, many ABC transporters have accessory domains for regulation or protein–protein interactions (Biemans-Oldehinkel et al., 2006; Kos & Ford, 2009). This seems to be the case for the Ig-like domains present in the periplasmic loop region 2 of the MalF subunit (MalF-P2) of the maltose transporter and in the C-terminal region of the Wzt protein (cWzt) of the LPS O-antigen transporter.

The periplasmic MalF-P2 Ig-like domain

The maltose transporter of E. coli and Salmonella is composed of the periplasmic soluble maltose-binding protein (MBP or MalE), encoded by malE, two integral IM proteins MalF and MalG, and two copies of the ATPase subunit MalK protein, which mediate the transport of maltose across the IM (Fig. 8a). The crystal structure of the reconstituted maltose transporter (MalFGK2) in complex with MalE (MalFGK2-E) in open conformation revealed that MalF contains a long periplasmic loop, termed (MalF-P2), connecting its transmembrane helices 3 and 4, that adopts an Ig-like fold (residues N93-K275) and contacts periplasmic MalE in a cap-like manner (Oldham et al., 2007). Although the MalF-P2 region loop is characteristic of enterobacterial MalF proteins (Tapia et al., 1999), is almost unique among bacterial maltose ABC transporters and is missing in homologous systems of Archaea (Daus et al., 2009). The MalF-P2 domain is able to interact with MalE independently of the transmembrane region of MalF (Jacso et al., 2009). Binding of the maltose-loaded MalE to MalFGK2 brings together the MalK NBDs in the cytoplasm such that ATP would promote reorientation of the MalF and MalG subunits, opening MalE and the concomitant release of maltose toward the transmembrane binding site (Oldham & Chen, 2011).

Figure 8.

Ig domains in ABC transporters. (a) The maltose importer MalFGK2. Maltose binds in the periplasm to MalE and stabilizes it in a closed conformation that interacts with the periplasmic Ig-like domain of MalF (MalF-P2). Binding of MalE to MalFGK2 (1) brings the NBDs of MalK closer such that ATP would promote a concerted motion of MalK closure (2), reorientation of the TM subunits, and opening of MalE (3). The change in conformation of MalE allows the transfer of maltose to the TM subunits binding site and positions ATP at the catalytic site for hydrolysis. Once ATP is hydrolyzed, the TMDs of MalF and MalG will reorient toward the cytoplasm, the transporter returns to the resting state and maltose completes its translocation across the membrane. (b) Export of O-polysaccharides (O-PS) by Wzt/Wzm. The synthesis of O-PS takes place at the cytoplasmic face of the IM starting with the WecA-mediated transfer of an N-acetylglucosamine-1-phosphate (yellow sphere) to form the undecaprenol pyrophosphate-linked intermediate. Next the mannosyltransferases WbdA, WbdB, and WbdC synthesize the O-PS by addition of mannose residues (green spheres) to the nonreducing end of the growing chain. WbdD finishes the extension of the nascent polymannan incorporating a methyl (O8 serotype) or a phospho-methyl (O9a serotype) residue (black sphere) to block the nonreducing terminus. The Ig-like domain of C-Wzt binds to the polymannose O-antigen substrate, possibly recognizing structural features in the nonreducing terminal modification. The export of the O-PS to the outer leaflet of the IM is initiated by ATP binding and hydrolysis in N-Wzt. Next, the O-antigen ligase WaaL, joins lipid A-core and O-antigen and the mature LPS is transported to the cell surface.

The cytoplasmic C-Wzt Ig-like domain

ABC transporters are also responsible for the export of an ample variety of glycans in cell surface glycoconjugates from bacteria. Examples include glycans from glycoproteins, teichoic acids, capsular polysaccharides, and the O-antigenic polysaccharide (O-PS) of lipopolysaccharide (LPS). LPS typically consists of three structural and functional regions: (1) the lipid A, (2) a nonrepeat core oligosaccharide, and (3) and a linear polymer termed O-polysaccharide (O-PS or O-antigen), which gives rise to about 180 O serotypes in E. coli (Raetz & Whitfield, 2002; Cuthbertson et al., 2007). During LPS biogenesis, the lipid A-core oligosaccharide and O-PS are synthesized independently at the cytoplasmic side of the IM (Raetz & Whitfield, 2002) and the two pathways converge at a ligation reaction, which transfers the O-PS from undecaprenol-PP to lipid A-core oligosaccharides at the periplasmic face of the IM. Once assembled, LPS molecules are shuttled to the OM through a process involving the LptABCDE complex (Sperandeo et al., 2009).

The synthesis of the O-PS follows two different pathways: The Wzm/Wzt ABC transporter-dependent pathway or the Wzy-dependent pathway (Raetz & Whitfield, 2002; Davidson et al., 2008; Ruiz et al., 2009). The Wzm/Wzt ABC transporter is comprised of two transmembrane (Wzm) and two nucleotide-binding polypeptides (Wzt) and translocates to the periplasm the O-PSs of E. coli serotypes O8, O9, and O9a (Fig. 8b; Clarke et al., 2011). Interestingly, in contrast to other glycan-ABC transporters that export different types of polysaccharides (Davidson et al., 2008), the E. coli Wzm/Wzt (ABC)-transporter is specific for its cognate substrate and this specificity relies in the C-terminal portion of Wzt (Cuthbertson et al., 2005).

The Wzt polypeptides from the O8 and O9a serotypes are homologous proteins of 404 and 431 amino acids, respectively, that contain two functional regions: The transmembrane N-terminal region of Wzt (N-Wzt) is involved in ATP binding and hydrolysis and is highly conserved between the two proteins. The cytosolic C-terminal region (C-Wzt) is relatively less conserved and binds specifically the O-PS chain-terminating residue, thus dictating the serotype specificity for either the O8 or O9a (Fig. 8b; Cuthbertson et al., 2005).

The structure of C-Wzt from serotype O9a reveals a dimer in which each monomer adopts an Ig-like fold that is composed by two anti-parallel β-sheets formed by the β-strands 2, 3 and 6, and 1, 4, 5, 7 and 8, respectively. The later β-sheet forms a concave groove that bears the O-PS pocket essential for binding. This carbohydrate-binding groove, conserved among O9a-like Wzt homologs, contains six critical aromatic residues, which allow substrate binding through a ring-stacking mechanism (Cuthbertson et al., 2007). It has been proposed that this binding domain introduces the polymer into the transport channel. In an alternative model, it may be required to disengage the nascent O-PS from the assembly enzymes to allow it to enter the export pathway (Cuthbertson et al., 2010).

Other Ig-like containing proteins

The extracellular sugar-binding Cbp21 protein

The Chitin-binding protein 21 (CBP21) secreted by S. marcescens (Suzuki et al., 1998), is a 21 kDa noncatalytic protein, classified as a CBM family 33 CBM in the CaZy database. Although the precise biological function of CBP21 is not clear, it has been suggested that it may play a role to enhance efficient chitin degradation. CBP21 binds to the insoluble crystalline substrate, leading to structural changes in the substrate and increased accessibility promoting hydrolysis (Vaaje-Kolstad et al., 2005a).

The structure of Cbp21 consists in an Ig-like fold (SCOP 117045) of two β-sheets (β-strands A, B, and E and β-strands C, D, F, and G, respectively) with a 65-residue ‘bud’ of three short helices located between β-strands A and B. The protein contains four cysteine residues forming two disulfide bridges, one in the loop/helical region (C41 and C49) and one joining β-strands D and E (C145 and C162; Vaaje-Kolstad et al., 2005a2005bb). The structural data showed that Cbp21 and the ChiN domain of ChiA from S. marcescens (Perrakis et al., 1994) present a similar FnIII-like fold, although the ChiN domain lacks the Cbp21 bud-like extension. In contrast to other CBMs, and ChiN domain in particular, CBP21 does not have a cluster of aromatic amino acids on the surface that is used for carbohydrate binding. Remarkably, it contains a conserved surface patch of hydrophilic residues that was shown to be essential for the disruption of the polysaccharide, as variants with single mutations on the largely polar binding surface lost their ability to promote chitin degradation while retaining considerable affinity for the polymer. Thus, the interaction of CBP21 with chitin could be primarily governed by polar interactions that may disrupt the hydrogen-binding network between individual polysaccharide chains (Vaaje-Kolstad et al., 2005ab).

The periplasmic copper resistance protein (CopC)

Copper is as a catalytic and structural cofactor for enzymes that is essential for most prokaryotic and eukaryotic organisms because it is involved in numerous biological processes (e.g. respiration, iron transport, oxidative stress protection; Puig & Thiele, 2002; Turski & Thiele, 2009). However, this metal ion can become toxic when is present free in the cell because it participates in redox reactions that result in the transfer of electrons to hydrogen peroxide with the concomitant generation of damaging hydroxyl radicals. Therefore, the levels of intracellular Copper are tightly regulated by homeostatic systems, which involve Copper acquisition, sequestration, and efflux (Banci et al., 2009). Certain bacterial strains contain extra-chromosomal operons that confer Copper resistance and allow survival of the cell under extremely high Copper levels (Zhang et al., 2006; Djoko et al., 2008). The best-characterized Copper resistance loci have been isolated from Gram-negative bacteria colonizing areas contaminated by the use of Copper salts (Munson et al., 2000). In E. coli and Pseudomonas syringae resistance to high levels of Copper is conferred by plasmid-borne clusters termed pcoABCDRSE (Brown et al., 1995) and copABCDRS (Cha & Cooksey, 1991), respectively.

The Copper resistance protein PcoC from E. coli (and its homologous CopC in Pseudomonas syringae) are periplasmic soluble monomeric proteins of about 100 amino acids (Djoko et al., 2007), constituted by seven anti-parallel β-strands having an Ig-like fold (SCOP 81969). Both PcoC and CopC are proposed to function as Copper carriers in the periplasm exchanging copper with the multicopper oxidase CopA, that converts CuI to the less toxic CuII (Djoko et al., 2008), and with the membrane-bound copper pumps CopB and CopD ((Koay et al., 2005). PcoC and CopC have two distinct (CuI) and (CuII) binding-regions (Arnesano et al., 2003; Wernimont et al., 2003). Crystallographic and spectroscopy studies of the E. coli PcoC have shown that the CuI site is constituted by a methionine rich loop located at the C-region of the protein, and the CuII site is formed both by β-strand-interconnecting loops containing histidine residues and by the N-terminal amino group located at the N-region of PcoC (Fig. 9; Drew et al., 2008). Interestingly, the CuII site in PcoC is classified as type-II2, like that of Cu,Zn-SOD, nevertheless Cu-PcoC does not exhibit superoxide dismutase activity in vitro (Huffman et al., 2002).

Figure 9.

Copper resistance protein PcoC. Ribbon diagram of the Ig-like periplasmic protein PcoC indicating the Copper (green spheres) CuI (N-region) and CuII (C-region) binding sites and the amino acid residues involved.

Chaperones involved in folding of Ig-like proteins in E. coli

General periplasmic chaperones

With a few exceptions (e.g. LacZ, GlgB, c-Wzt), most Ig-like domains of E. coli proteins are located in the periplasm or exposed to the periplasm at some point during their secretion and/or biogenesis (Table 1). These periplasmic-exposed E. coli proteins with Ig-like domains are transported across the IM by SecYEG translocon (Driessen & Nouwen, 2008), reaching the periplasm in an unfolded conformation. Folding of Ig-like proteins in the periplasm is assisted by different folding factors. There are three different types of periplasmic folding catalysts that in some cases have overlapping functions: chaperones, which bind to proteins and help to prevent undesirable off-pathway interactions or aggregation of their substrates (e.g. Skp, SurA, DegP), peptidyl–prolyl cis/trans isomerases (PPIases) that catalyze cis/trans isomerization of proline peptide bonds in proteins (e.g. SurA, FkpA, PpiA, PpiD), and protein disulfide forming enzymes and isomerases that catalyze the formation and exchange of disulfide bonds (e.g. DsbA, DsbC; Allen et al., 2009). These folding factors assist folding of mutiple proteins in the periplasm and also have a role in folding of OMPs, including intimin and fimbrial ushers, during their periplasmic transit to the OM (Justice et al., 2005; Bos et al., 2007; Bodelón et al., 2009; Knowles et al., 2009; Palomino et al., 2011). The dedicated periplasmic chaperones (e.g. PapD, FimC) of fimbrial subunits were discussed in a different section.

Some of the above mentioned periplasmic chaperones, which play an essential role for the homeostasis of the cell, are also key players in the cellular response under periplasmic and membrane stress conditions. The accumulation of aggregated protein in the periplasm leads to cell damage, and therefore, bacteria have evolved stress response pathways to sense and combat this effect. In E. coli the σE and Cpx pathways, which are induced upon extracytoplasmic stress, upregulate periplasmic chaperones such as SurA, Skp or DegP that bind unfolded or misfolded proteins to alleviate folding defects (SurA and Skp) or to proteolyze misfolded polypeptides (DegP; Duguay & Silhavy, 2004; Mogensen & Otzen, 2005).

The SurA polypeptide presents four domains: an N-terminal domain, two central parvulin-like domains (PPIase 1 and 2) observed in other PPIases and a short C-terminal domain (Bitto & McKay, 2002). The N- and the C-terminal regions bear the chaperone activity, whereas the PPIase activity resides in the PPIase domain 2 (Behrens et al., 2001). Xu and collaborators reported the crystal structure of SurA in complex with short peptides known to specifically interact with SurA and suggested that the PPIase domain 1 is responsible for substrate selection (Xu et al., 2007). SurA binds preferentially to Ar-X-Ar tripeptide motifs common in OMPs, being Ar and X any aromatic and polar residues, respectively (Bitto & McKay, 2003), and it was shown that SurA binds OMP assembly intermediates immediately after they leave the Sec translocon before signal sequence cleavage (Ureta et al., 2007).

Skp is a highly basic protein of about 17 kDa that forms a homotrimer in solution, which binds its substrate in a 1 : 1 stoichiometry. The crystal structure of the homotrimer resembles a jellyfish with tentacles defining a substrate-binding cavity that is likely to hold substrates of different sizes protecting them from aggregation (Korndorfer et al., 2004; Walton & Sousa, 2004; Walton et al., 2009). Functionally, the arms of the Skp oligomer have a net positive charge on their external side and hydrophobic patches on their internal side that are thought to bind to LPS and hydrophobic substrates, respectively (Allen et al., 2009). This periplasmic chaperone was identified in a screen for periplasmic folding factors because it was retained on an affinity column with Sepharose-bound OmpF (Chen & Henning, 1996). Interestingly, Skp was also identified as a folding factor for heterologous antibody Ig domains expressed in E. coli (Bothmann & Plückthun, 1998). Skp it is known to bind OMPs to prevent their aggregation and to escort them from the IM to the OM (Schafer et al., 1999; Harms et al., 2001). Skp was co-purified in complex with 31 envelope proteins being 19 of them β-barrel proteins, indicating that Skp exhibits a broad substrate spectrum (Jarchow et al., 2008).

DegP belongs to the high temperature requirement (HtrA) family of serine proteases and is able to perform the antagonistic functions of protein repair and degradation. It rescues and refolds aggregated or slightly misfolded proteins or proteolyzes irretrievably misfolded proteins with the aim of reducing extracytoplasmic stress and cellular damage. DegP mediates the encapsulation and/or refolding of misfolded proteins via its chaperone-like activity. Recent biochemical and structural analysis have shed light into the molecular architecture and function of DegP (Allen et al., 2009; Ortega et al., 2009; Subrini & Betton, 2009; Sawa et al., 2010). Each DegP monomer is composed of three domains: a N-terminal trypsin-like protease domain and two C-terminal PDZ domains that contain the structural features allowing DegP to recognize and tether its substrates for proteolytic cleavage. The first crystal structure reported for DegP consists in a hexamer formed by two homotrimers stabilized by interactions between the PDZ domains (Krojer et al., 2002). Subsequently, two independent studies reported that, in the presence of the substrate, the homotrimers further oligomerize via PDZ1–PDZ2 interactions into cage-like spherical structures of 12- and 24-mers that exhibit both protease and chaperone activity. Inside of the cage, the substrate protein is then subjected to either protease degradation or refolding (Jiang et al., 2008; Krojer et al., 2008).

FkpA is a member of the FKBP family of periplasmic PPIases with general chaperone activity (Horne & Young, 1995; Arie et al., 2001). The PPIase activity catalyzes the cis/trans isomerization of peptide bonds before proline (Xaa-Pro or prolyl bonds). Although this is a sequence-specific reaction, the ample specificity of the chaperone domain for its substrate allows these enzymes to act over a broad range of protein substrates. Also, catalysis of isomerization is several orders of magnitude faster than chaperone-mediated substrate delivery (Jakob et al., 2009). The crystal structure of dimeric FkpA shows a V-shaped molecule comprised of two monomers, each containing a globular C-terminal domain and a N-terminal dimerization domain linked together by a long α-helix, that functions as a flexible arm. The cleft formed between the two arms of the homodimer might accommodate the substrates (Saul et al., 2004; Hu et al., 2006). The chaperone activity of FkpA is independent of its PPIase activity. A structural and functional study of FkpA indicated that the PPIase and the chaperone activities reside in the N- and C-terminal domains, respectively, and act independently of one another (Saul et al., 2004; Hu et al., 2006). There is limited data available regarding the natural substrates of this folding factor. FkpA does not seem to have a specific affinity for OMPs, and to date it still remains unclear whether the protein is involved in OMP biogenesis (Allen et al., 2009). An in vivo target of FkpA is the toxin colicin M, which is a polypeptide produced by E. coli that kills sensitive E. coli cells. Colicin M appears to be refolded by FkpA upon its entry into the periplasm of sensistive E. coli cells (Hullmann et al., 2008). Interestingly, it was also demonstrated that FkpA improves the functional expression of antibody Ig domains in E. coli (Bothmann & Plückthun, 2000; Ramm & Plückthun, 2000).

Disulfide bond catalysts

The presence of disulfide bonds is not a conserved feature in all E. coli and enterobacterial Ig-like domains (Table 1). Nevertheless, many of them contain structural disulfide bonds that may be essential for the stability and functionality of these polypeptides. The available crystal structures of Ig-like domains from E. coli proteins reveal disulfide bonds in structural components of secreted adhesive organelles such as fimbria, pili, fibrills or capsule-like structures, some periplasmic chaperones and ushers of the CU pathway, the Cu,ZnSOD and DsbD enzymes, Chitin-hydrolase ChiA and Chitin-binding protein Cbp21. Most disulfide bonds of E. coli Ig-like domains are present in the structural components of fimbriae such as the pilin domain of major (e.g. PapA and FimA) and minor fimbrial subunits (e.g. PapE, PapH, PapK, FimF, FimG, and FimH) or the pilin subunits of polyadhesive fimbria (e.g. AfaE-III). Sequence analysis of fimbrial proteins assembled by classical CU systems showed that they contain two conserved cysteine residues that may form a noncanonical intradomain disulfide bond between the beginning of β-strand A1 and the end of β-strand B (Fig. 3a; Piatek et al., 2010). This cysteine bond possesses a unique localization (joining two adjacent β-strands) not found in any other protein family belonging to the IgSF.

Most organisms have specialized machinery to catalyze formation and isomerization of disulfide bonds (Heras et al., 2007), and the best-characterized disulfide bond (DSB) system is that of E. coli K-12 (Heras et al., 2009). A bioinformatic analysis of the cysteine content of predicted cell envelope proteins from 375 bacterial genomes showed that, although the bacterial species posses an ample diversity in the mechanisms for disulfide bond formation, Gamma- and Betaproteobacteria have DSB systems similar to that of E. coli K-12 (Dutton et al., 2008; Heras et al., 2009). In E. coli formation of disulfide bonds takes place in the periplasm and is mainly catalyzed by the DsbA oxidoreductase (Bardwell et al., 1991), which is a member of the thioredoxin (Trx) superfamily (Kadokura et al., 2003; Sevier & Kaiser, 2006) that contains a Trx domain with a redox-active cysteine pair (Cys30–Cys33; Martin et al., 1993). DsbA introduces disulfide bonds nonspecifically by donating its disulfide. It has been suggested that DsbA may have more than 300 in vivo substrates of the 700 periplasmic and membrane proteins listed in the Swiss Protein Database having two or more cysteine residues (Hiniker & Bardwell, 2004). The DsbB IM protein (Bardwell et al., 1993), which is able to generate disulfides de novo, restores the oxidizing activity of DsbA by reoxidizing its cysteine pair.

In addition to the DsbA–DsbB oxidation pathway the periplasmic space contains a DsbC-DsbD disulfide-isomerizing pathway, in which DsbC functions as the major chaperone/protein disulfide isomerase repairing incorrectly paired cysteines (Rietsch et al., 1996). a DsbC must be maintained in its reduced state for proper isomerase activity and the bacterial periplasm lacks a source of reducing equivalents, E. coli has evolved a mechanism for importing electrons from the cytoplasm that are used for reduction of DsbC in a process mediated by the IM thiol disulfide reductase DsbD (Missiakas et al., 1995). The E. coli DSB system also includes DsbG, a homolog of DsbC that acts to protect single cysteines from sulfenylation (Depuydt et al., 2009), and DsbE, which is restricted to the pathway of cytochrome c maturation (Fabianek et al., 1999). In-depth reviews regarding the E. coli and other DSB systems have been published elsewhere (Ito & Inaba, 2008; Heras et al., 2009; Inaba, 2009).

It is also becoming apparent the importance of the DSBs for bacterial pathogenesis and numerous Gram-negative pathogens either encode additional copies of their DSB genes or posses functional DSB paralogs (e.g. DsbL, DsbI) that, despite of some levels of redundancy, showed differences in substrate specificity (Bouwman et al., 2003; Totsika et al., 2009). The implications of DSBs in the assembly of functional fimbria was demonstrated in bacterial pathogens assembling different types of adhesive organelles such as P fimbriae of UPEC (Jacob-Dubuisson et al., 1994), bundle-forming pili of EPEC (Zhang & Donnenberg, 1996), plasmid-encoded fimbriae (Pef) of S. enterica (Bouwman et al., 2003), the type-IV pilus of N. meningitidis (Tinsley et al., 2004). As an example, UPEC bacteria lacking DsbA are defective in the assembly of P fimbriae owing to the inability of the PapD chaperone to fold into a conformation able to bind subunits in vivo (Jacob-Dubuisson et al., 1994). Zav′yalov and collaborators reported that capsule production was significantly decreased in an E. coli dsbA mutant carrying the Y. pestis F1 operon. As Caf1 subunit contains no cysteine residues, the most likely explanation for the observed decrease in subunit assembly is owing to an effect on the Caf1M chaperone. Two cysteine residues of Caf1M chaperone form a disulfide bond and reduction of this bond increases the dissociation constant for the Caf1M-Caf1 complex (Zavialov et al., 2007). Interestingly, the dsbA mutant strain accumulates considerably lower amount of Caf1M than the wild type, suggesting that a misfolded intermediate would be more susceptible to degradation by periplasmic proteases. Totsika and collaborators also showed that DsbL catalyzes disulfide bond formation of PapD as overexpression of DsbL in the absence of DsbA restored the production of functional P fimbriae in E. coli CFT073 dsbAB dsbLI mutant (Totsika et al., 2009). Finally, DsbA has been shown to catalyze disulfide bond formation in the C-terminal lectin-like domain of the afimbrial adhesin intimin (Bodelón et al., 2009). Intimin expressed in an E. coli dsbA mutant lacks this disulfide bond, shows lower display levels on the cell surface and has higher sensitivity to protease digestion (Bodelón et al., 2009).

Biotechnological applications of Ig proteins in E. coli

Natural E. coli proteins with Ig-like domains, such as Type 1 fimbrial subunits and intimin, have been exploited biotechnologically for the display of peptides and small protein domains on the surface of E. coli cells (Pallesen et al., 1995; Klemm & Schembri, 2000; Wentzel et al., 2001). Nonetheless, the major biotechnological breakthrough related to the expression of Ig proteins in E. coli is the development of technologies for the expression and selection of full-length antibodies and small antibody fragments in this bacterium.

Well before the discovery of a natural E. coli protein containing Ig-like domains (Holmgren & Brändén, 1989), biotechnologists had already focused their attention in the possibility of producing antibodies in E. coli. This objective was spurred by the success of an emerging recombinant DNA technology for the production of small human polypeptide hormones of biomedical interest in E. coli, such as insulin (Goeddel et al., 1979), and by the therapeutic expectations raised by the technology of monoclonal antibodies (mAbs) from murine hybridomas (Kohler & Milstein, 1975). However, expression of IgGs in E. coli was a formidable task owing to their secreted nature and large mass (c. 150 kDa), with heterotetrameric structure composed of two identical light (L) chains (c. 25 kDa) and two identical heavy (H) chains (c. 50 kDa), the presence of multiple intra- and inter-chain disulfide bonds and post-translational modifications (i.e. glycosylation; Elgert, 1998; Maynard & Georgiou, 2000; Schroeder & Cavacini, 2010). The difficulties found for producing full-length H and L chains in stoichiometric amounts and good yields to assemble functional IgGs in E. coli led scientist to focus on the expression of small antibody fragments with antigen-binding activity. Currently, it is also possible to produce and select in E. coli full-length IgGs and other antibody formats. These developments will be described in the following sections.

Structure of classical antibodies and recombinant Fab and scFv fragments

The classical IgGs of humans, mice, and most mammal species, are Y-shaped bivalent molecules with identical antigen-binding sites at the tip of the two arms (Fig. 10a). Each arm is referred to as the antigen-binding fragment (Fab), which is composed by the association of the VH and CH1 domains of the H chain with the VL and CL domains of the L chain. The actual antigen-binding site is formed of three hypervariable loops found in the VH and VL domains, also known as complementarity-determining regions (CDRs). The sequence and structural diversity of these six CDRs, juxtaposed in the folded antibody structure forming flat or concave antigen-binding surfaces, provide the capacity to recognize multiple antigenic structures (Padlan, 1996; Schroeder & Cavacini, 2010). The stalk of the Y, or crystallizable fragment (Fc), is assembled by the constant CH2-CH3 domains of the two H chains in the IgG molecule, and contains a conserved N-linked glycan at residue N297 in the CH2 domain. The Fc region mediates the effector functions of antibodies by recruiting the complement, in complement dependent cytotoxicity (CDC), and immune cells, in antibody dependent cell-mediated cytotoxicity (ADCC) and phagocytosis (ADCP; Raju, 2008; Schroeder & Cavacini, 2010; Desjarlais & Lazar, 2011). To exert these effector functions the glycosylated Fc region interacts with blood serum proteins (e.g. C1q) and specific Fcγ receptors (FcγRs) found in immune cells (e.g. Natural Killer, Macrophages; Carroll, 2008; Nimmerjahn & Ravetch, 2008). In addition, the Fc region plays an important role for the long-circulating half-life of full-length IgGs in the body owing to its recognition by the neonatal Fc receptor (FcRn) expressed in endothelial cells of the vasculature, epithelial cells of intestine, kidneys and lungs, among other organs (Roopenian & Akilesh, 2007). The FcRn is involved in the recycling of endocyted IgG molecules by antigen presenting cells (APCs) and endothelial cells, and in the transcytosis of IgGs through several epithelial and endothelial layers, including kindney podocytes (Roopenian & Akilesh, 2007).

Figure 10.

Schematic diagram of antibodies and small antigen-binding fragments. (a) Structure of classical IgGs with heavy (H) and light (L) chains indicating the crystallizable fragment (Fc) region with conserved glycosylation site (orange circle) of CH2 domain. The small antigen-binding fragments Fab and scFv derived from classical IgGs are shown on the right. (b) Structure of camelid HCAbs lacking the light (L) chains and CH1 domains. The glycosylated Fc region of HCAbs is similar to that of classical IgGs. The single-domain antibody (sdAb) fragment derived from the VH of camelid HCAbs, referred to as VHH (VH of HCAb) or nanobody, is shown on the right.

The first recombinant antibodies (rAbs) expressed in E. coli were Fab fragments (Better et al., 1988) and single-chain variable fragments (scFv), which are assembled by linking the VH and VL domains into a single polypeptide chain by means of a short flexible peptide (Fig. 10a; Bird et al., 1988; Huston et al., 1988). These formats have been the more common antibody fragments expressed in E. coli because of their simple structure and the ability to select specific binders from large combinatorial libraries of Fab and scFvs genes (Marks et al., 1991; Plückthun et al., 1996; Hoogenboom, 2005). Libraries of antibody fragments can be amplified from B-cells of diverse origin (e.g. mouse, human) after immunization or disease (‘immune libraries’) or from nonimmunized organisms (‘naïve libraries’). Alternatively, the sequence of V genes can be mutagenized in vitro to diversify the CDRs generating the so-called ‘synthetic libraries’ (Griffiths et al., 1994; Nissim et al., 1994; Aujame et al., 1997; Benhar, 2007; Zhai et al., 2011). These repertoires of antibody fragments can be displayed on the capsid of filamenteous E. coli bacteriophages (e.g. M13) for selection of high-affinity antigen-binding clones, which allows the isolation and affinity maturation of antibodies for therapeutic and in vivo diagnostic applications (Clackson et al., 1991; Hoogenboom et al., 1998; Hoogenboom, 2002; Kretzschmar & von Ruden, 2002; Bradbury et al., 2011). As discussed later in the text, full-length IgGs can also be expressed in E. coli cells and display on filamentous bacteriophages for selection of high-affinity clones (Mazor et al., 2007, 2008, 2009, 2010).

Heavy-chain antibodies and single-domain antibodies (sdAbs)

Besides Fab and scFvs other recombinant antibody fragments are currently expressed in E. coli and selected by phage display from antibody libraries. Among them, the most remarkable are the single-domain antibodies (sdAbs; Holliger & Hudson, 2005; Saerens et al., 2008; Wesolowski et al., 2009). These small fragments (c. 12–14 kDa) are based on a single V domain with full antigen-binding capacity isolated from a special class of natural homodimeric antibodies devoid of L chains called heavy-chain antibodies (HCAbs; Fig. 10b) found in camelids (e.g. dromedaries, llamas) and cartilaginous fish (e.g. sharks), where they are called IgNARs (Ig of shark new antigen receptor). The sdAb from a camelid HCAb is referred to as VHH (VH of HCAb; Hamers-Casterman et al., 1993; Muyldermans et al., 1994; Arbabi Ghahroudi et al., 1997; van der Linden et al., 2000), whereas that from an IgNAR is called VNAR (Diaz et al., 2002; Dooley et al., 2003). VHHs and VNARs are able to bind with high affinity and specificity their cognate antigens in the absence of VL domains. Although their overall Ig domain folding is similar to that of VH domains from classical antibodies (with H and L chains), the lack of a VL partner has driven the evolution of special features for antigen recognition and domain stability and solubility (Desmyter et al., 1996; Vu et al., 1997; Decanniere et al., 1999; van der Linden et al., 1999; Muyldermans et al., 2001; Stanfield et al., 2004; Flajnik et al., 2011). The CDR3 loops of VHHs and VNARs, and to a lesser extent the CDR1 loops, are longer than those of classical VHs providing novel conformations and an extra interaction surface for antigen recognition. The long CDR3 loop provides the major antigen-interaction surface in VHHs and VNARs and often protrudes from the Ig domain as a finger-like structure. This long CDR allows the recognition of epitopes in narrow clefts and concave cavities in protein structures (e.g. active sites of enzymes), which are frequently not accessible to conventional antibodies with larger size and flat or concave antigen-binding surfaces (Fig. 11a; Desmyter et al., 1996). Potent enzyme inhibitors based on VHHs binding to active sites of enzymes have been isolated (Lauwereys et al., 1998; Transue et al., 1998; Conrath et al., 2001; Desmyter et al., 2002). Interestingly, small VHHs and VNARs can also reach to the less-accessible inner regions of proteins found on the cell surface of pathogens, and not only to the outermost epitopes of these proteins, which are frequently variable among strains or during infection to evade the immune system (Stijlemans et al., 2004; Henderson et al., 2007).

Figure 11.

Comparison between conventional VH/VL and VHH antibody fragments. (a) Three-dimensional structure of chicken lysozyme recognized by an scFv (VH/VL domains) (left; PDB code 1mlc) or by a VHH domain of camelid HCAb (right; PDB code 1mel). The CDRs of VH, VL and VHHs are shown in different colors (CDR1 in orange, CDR2 in green, CDR3 in cyan). The long CDR3 loop of the VHH protrudes as a finger-like structure toward an epitope located in the cavity of the active site of lysozyme (right), whereas the flat VH/VL paratope binds to an epitope located at the exposed surface of lysozyme (left). Images generated with PyMOL program (Molecular Graphics Systems, LLC). (b) Cartoon showing the polypeptide chains of conventional VH and camelid VHHs, indicating the amino acids of framework region 2 of classical VHs (V37, G44, L45, W47) that are substituted by more hydrophilic amino acids (e.g. Y37, E44, R45, G47) in VHHs. Other distinct features shown are the different length of CDR1 and CDR3 in VHs and VHHs, the presence of the conserved canonical disulfide bond between residues C22 and C92, and the extra-disulfide bond between Cys residues of CDR1 and CDR3 in VHHs.

In addition, the protein solubility and stability of VHHs are higher than isolated VH domains from classical antibodies (van der Linden et al., 1999; Dumoulin et al., 2002). This is owing to a combination of subtle amino acid changes along their sequence (Fig. 11b), including substitutions of conserved hydrophobic residues in classical VHs (V37, G44, L45, and F47 or W47) for more hydrophilic amino acids (Y37/F37; E44/Q44; R45/C45; G47/R47/L47/S47; Muyldermans et al., 2001, 2009). In some cases, the long CDR3 loops of VHHs participate in the higher solubility of these domains by protecting exposed hydrophobic residues of the core-Ig domain from aqueous environment. Furthermore, VHHs and VNARs often contains, besides the canonical disulfide bond of Ig domains, an extra-disulfide bond connecting CDR3 and CDR1 (in camels and sharks) or CDR3 with CDR2 (in llamas) that assist in stabilizing the conformation of these CDRs and the overall stability of the domain (Govaert et al., 2011). Interestingly, the protease resistance and thermal stability of VHHs can be enhanced by the introduction of an additional disulfide bond in the core-Ig domain (A/G54C and I78C; Hussack et al., 2011).

The distinct properties of VHHs and VNARs explain their high level of expression in E. coli, increased solubility, monomeric behavior, resistance to degradation, and to heat and chemical denaturation. Given the unique biophysical and antigen-binding properties of sdAbs, these antibody fragments have become attractive molecules for diverse biotechnological applications (Harmsen & De Haard, 2007; De Marco, 2011). Remarkably, camelid VHHs also show high sequence identity with the human VH family 3 (the most commonly expressed human VH family) opening the possibility of their therapeutic application (Holliger & Hudson, 2005; Saerens et al., 2008; Vincke et al., 2009). This potential and their small nanometer size (c. 2.5 nm width and c. 4 nm height) prompted a ‘marketing’ rename of VHHs as Nanobodies (Nbs), which is now a widely accepted term for these sdAbs also in scientific literature.

There has been considerable interest in developing sdAbs based on human VH (or VL) sequences that could mimic the binding and stability properties of natural VHHs. Isolated heavy and light chains of antibodies are overproduced as Bence-Jones proteins in multiple myeloma and other human pathologies (Hilschmann & Craig, 1965; Seligmann et al., 1979; Hendershot et al., 1987; Prelli & Frangione, 1992). In fact, murine VH sdAbs against lysozyme were isolated from phage display in E. coli before the discovery of natural HCAbs in camels (Ward et al., 1989). However, most VH sequences from conventional antibodies tend to aggregate in solution and show low affinity for antigens. Introduction of ‘camel’ mutations G44E, L45R, and W47G, in a human VH3 sequence significantly increased its solubility (Davies & Riechmann, 1994). Human VH clones with low tendency to aggregate were isolated by repeated cycles of heating and cooling of phages with displayed VH sequences (Jespers et al., 2004). Based on these findings, synthetic libraries of human VHs have been developed by randomizing CDRs in ‘camelized’ and/or selected human VH sequences and used for isolation of functional sdAbs with nanomolar affinity (Davies & Riechmann, 1994, 1995; Martin et al., 1997; Reiter et al., 1999; Dumoulin et al., 2002; Holt et al., 2008; Arbabi-Ghahroudi et al., 2009). SdAbs based on human VL domains have also been reported and expressed in E. coli. These include VL domains from conventional IgGs and naive scFv libraries (Colby et al., 2004; Martsev et al., 2004; Cossins et al., 2007; Schiefner et al., 2011) and sdAbs isolated from synthetic libraries (van den Beucken et al., 2001; Holt et al., 2008).

Periplasmic expression of antibody fragments and their selection by phage display

Ab fragments (Fab, scFv and sdAbs) are commonly exported into the periplasm of E. coli cells (Fig. 12a) where disulfide bond forming and isomerization enzymes (i.e. DsbA, DsbC) and other protein chaperones (e.g. Skp, FkpA, SurA) enable their correct folding (Skerra & Pluckthun, 1991; Glockshuber et al., 1992; Missiakas & Raina, 1997; Jurado et al., 2002; Merdanovic et al., 2011; Schlapschy & Skerra, 2011). Standard expression vectors are derived from phagemids with a replication origin of high-copy number pBR/pUC-plasmids and a phage origin of replication and packaging from M13/fd filamentous phages (Qi et al., 2012). These vectors carry an inducible E. coli promoter (e.g. Ptac, Plac, PBAD, PphoA) and an N-terminal secretion signal (e.g. PelBss, OmpAss, PhoAss, STIIss) to which the gene segment encoding the Ab fragment is fused in frame (Choi & Lee, 2004). In addition, these vectors commonly incorporate epitope tags (e.g. HA, myc, 6xhis) in the C-terminus for detection and purification of the overproduced Ab fragment. Vectors for expression of scFvs and sdAbs (e.g. pHEN) have a single monocistronic transcriptional unit. Fab expression vectors export two polypeptide chains (the L and VH-CH1 chains) from either a single bicistronic transcriptional unit (e.g. pComb3) or from two independent monocistronic transcriptional units with separate promoters (e.g. pComb3X; Qi et al., 2012).

Figure 12.

Expression of antibody fragments in the periplasm of Escherichia coli cells. (a) Soluble Ab fragments, depicted as a single-domain Ab in this scheme, are exported to the periplasm with an N-terminal signal peptide (e.g. PelBss) fused to epitope tags (e.g. myc, 6xhis) at the C-terminus for immune detection and protein purification (red circle). The Ab fragment can also be fused at its C-terminus to the minor coat protein III (pIII) of filamentous phages for phage display. In this case, the Ab fragment is exposed to the periplasm but tethers to the inner membrane (IM) of E. coli by pIII. (b) Phage antibody particles assembled upon infection of E. coli cells expressing Ab-pIII fusions with a helper phage. The Ab-pIII fusion is displayed on the phage capsid (usually one copy per phage), which contains the phagemid DNA with the Ab gene.

Prolonged high-level expression of the Ab fragment at 37 °C frequently leads to protein misfolding and aggregation and induces permeabilization of the bacterial OM with leakage of periplasmic proteins, releasing a significant fraction of the overproduced Ab fragment soluble in the culture media (Skerra, 1993). Misfolding and aggregation of the Ab fragment, along with toxicity and lysis of the overproducing E. coli cells, can be minimized growing the cells at lower temperatures (16–30 °C; Somerville et al., 1994). The coexpression of protein chaperones (Skp, FkpA, SurA) and disulfide bond enzymes (DsbA, DsbC, DsbG, human PDI) in the periplasmassists the correct folding of Ab fragments and improve their final yield in E. coli (Humphreys et al., 1996; Plückthun et al., 1996; Bothmann & Plückthun, 1998, 2000; Hayhurst & Harris, 1999; Zhang et al., 2002; Friedrich et al., 2010). In addition, E. coli strains deficient in various periplasmic proteases (DegP, Prc, Spr) increase the accumulation of certain Ab fragments (Chen et al., 2004).

Phagemid vectors for Ab expression also enable the fusion of the Ab fragment (the VH-CH1 chain in the case of Fabs, the VH-VL chain of scFvs or the single V domain of sdAbs) with the minor coat protein III (pIII) of filamentous phages (e.g. M13, fd) for phage display (Fig. 12a). Frequently, between the Ab gene (with an N-terminal secretion signal and C-terminal epitope tags) and the gene III segment (encoding pIII lacking its own secretion signal) is placed an amber stop codon (TAG) that is suppressed to Glutamine in E. coli strains carrying the supE mutation. As a consequence, these vectors encode the soluble periplasmic Ab fragment in ‘wild-type’ E. coli strains (e.g. WK6, HB2151) and the hybrid protein between the Ab fragment and pIII in E. coli supE strains (e.g. TG1, XL-1Blue). The Ab moiety of the hybrid protein is exposed to the periplasm but is tethered to the IM of E. coli by the pIII. As mentioned above, the N-terminal signal peptides commonly used for Ab fragments are post-translational Sec-signals, such as PelBss, OmpAss, and PhoAss. Interestingly, highly stable protein domains need the use of co-translational SRP-signals (e.g. DsbAss) for efficient display on filamentous phages (Steiner et al., 2006).

Bacteriophages displaying the Ab fragments, also called Phabs (Fig. 12b), are produced by infecting E. coli F+ supE cells (e.g. TG1, XL1-Blue) carrying the phagemid vector, with a helper phage encoding the bacteriophage proteins (e.g. M13-K07, VCS-M13). These helper phages have a defective origin of replication and packaging that reduces incorporation of its own genome into phage particles. Thus, released phage particles contain phagemid DNA instead of helper phage DNA, linking the genotype (antibody gene) and phenotype (displayed antibody fragment) in an infective particle. Ab gene libraries packaged as Phabs are incubated with a chosen antigen and/or cell, and clones with the desired antigen-binding specificity and affinity can be recovered and amplified by infecting fresh E. coli F+ supE cells, in a process called biopanning (Clackson et al., 1991; Marks et al., 1991; Winter et al., 1994; Chames et al., 2002; Mutuberria et al., 2004; Hoogenboom, 2005). Assembled Phabs usually contain a single copy of the hybrid Ab-pIII fusion and up to 3–5 copies of the wild-type pIII encoded by the helper phage. Monovalency of Phabs results in the selection of high-affinity Ab binders (O'Connell et al., 2002). In some situations, such us panning against rare antigens on cell surfaces, phage multivalency may be desirable. Multivalent Phabs can be produced by infection of E. coli with mutant helper phages having a deletion in gene III (Rakonjac et al., 1997; Rondot et al., 2001; Baek et al., 2002; Soltes et al., 2003; Oh et al., 2007).

Expression and selection of full-length IgGs in the periplasm of E. coli

Access to libraries of human V sequences and their engineering is a major application of the phage display of Ab fragments in E. coli (Lonberg, 2008). However, given their small size and lack of the Fc region, Ab fragments miss effector functions and show a significantly reduced in vivo half-live (Roopenian & Akilesh, 2007; Carroll, 2008; Nimmerjahn & Ravetch, 2008). Hence, Ab fragments have been traditionally formatted as full-length IgGs for therapeutic applications and produced in bioreactors of mammalian cell culture systems (e.g. CHO cells) being released into the culture media as fully assembled and glycosylated IgG molecules that have followed the mammalian secretory route (Chadd & Chamow, 2001; Brekke & Sandlie, 2003, Birch & Racher, 2006; Beck et al., 2010).

Given the structural complexity of IgGs, early works that attempted their expression in E. coli only reported the production of insoluble aggregates that need to be refolded in vitro (Boss et al., 1984; Cabilly et al., 1984). Although refolding of IgGs produced in E. coli as inclusion bodies has been greatly optimized (Hakim & Benhar, 2009), it is not widely used for IgG expression. Production of fully assembled, functional, full-length IgGs in the periplasm of E. coli was not reported until 2002 (Simmons et al., 2002). The geness encoding L and H chains of a human IgG1 molecule were cloned in a plasmid vector with two independent transcriptional units controlled by independent phoA promoters. To balance the expression of L and H chains, E. coli translation initiation regions (TIR) of various strengths were placed upstream of the fusions between the N-terminal signal peptide of STII and the L and H chains. Optimal TIRs for the L and H chains were found empirically by analyzing the amount of assembled IgG in the periplasm of E. coli cells. These E. coli-produced aglycosylated IgGs were purified and compared to the glycosylated IgGs from mammalian CHO cells, demonstrating identical antigen-binding activity. In addition, they exhibited equal affinity for human FcRn receptor, whose binding does not depend on the glycosylation of the Fc region, having similar half-lives in chimpanzees. However, aglycosylated E. coli IgGs do not bind human C1q and FcYRs (Simmons et al., 2002; Reilly & Yansura, 2010) and, therefore, are not able to elicit Ab effector functions such as CDC, ADCC and ADCP. Although the absence of Ab effector functions may be desirable in some cases, such as those in which binding or neutralization of the antigen (e.g. hormone/toxin/virus) is the only therapeutic effect required for the Ab, aglycosylation could represent a major limitation for the IgGs produced in E. coli. Despite this, aglycosylated IgG antibodies obtained through protein engineering have entered clinical trials with excellent therapeutic results, tolerance, and pharmacokinetics (Jung et al., 2011). In addition, variants of therapeutic IgGs with mutations in the Fc region can, at least partially, bypass the need of glycosylation and selectively activate certain FcγRs (Sazinsky et al., 2008; Jung et al., 2011).

The therapeutic potential of aglycosylated IgGs and simplified bioprocessing triggered an interest in producing and selecting full-length human IgGs directly in E. coli. Research conducted by Dr. George Georgiou and collaborators (University of Texas) has achieved this major goal. This group reported an expression vector with a single Plac promoter and a bicistronic operon for L and H chains, both fused to the PelBss, which efficiently assembled several IgGs in E. coli (Mazor et al., 2007). In addition, based on the same vector, they developed a library screening system in which IgGs against any given antigen can be selected and directly expressed in E. coli (Mazor et al., 2007). Pools of VH and VL gene segments were cloned in this vector, replacing the VH and VL domains of a well-expressed human IgG, and selected for specific antigen binders using fluorescent sorting of spheroplasted E. coli cells, with a modification of the anchored-periplasmic expression (APEx) technology (Chen et al., 2001; Harvey et al., 2004). To this end, the library of IgGs is co-expressed with a chimeric lipoprotein in the periplasm, named NlpA-ZZ, which tethers IgGs to the bacterial IM Fig. 13a). NlpA-ZZ comprises the N-terminal signal peptide and the first six amino acids of the E. coli lipoprotein NlpA fused to a synthetic analog of the IgG-binding B domain of protein A from Staphylococcus aureus, called ZZ (Jendeberg et al., 1995). Spheroplasts obtained by OM permeabilization of E. coli cells are incubated with the antigen of interest, labeled with a fluorophore (Fig. 13b), and sorted in a cytometer. After a few cycles of selection, full-length IgGs of high-affinity for the antigen are obtained and expressed in E. coli at high levels (Mazor et al., 2007, 2008; Makino et al., 2011).

Figure 13.

Expression of full-length IgGs in the periplasm of Escherichia coli. (a) Scheme showing the periplasm of E. coli cells containing a full-length IgG tethered to the bacterial inner membrane (IM) with chimeric lipoprotein NlpA-ZZ (shown in red). (b) IgGs expressed in the periplasm can be selected for specific binders by permeabilization of the OM and incubation of the spheroplasts with a labeled antigen for fluorescence-activated cell sorting.

Screening of large IgG libraries (> 108 clones) is complex using only fluorescent cell sorting. Selection of binders can be optimized by combining phage panning with subsequent rounds of fluorescent cell sorting of E. coli spheroplast. To this end, full-length IgGs are displayed on phage particles having the ZZ domain fused to pIII (Mazor et al., 2010). Collectively, these developments enable the fast and versatile isolation and production of functional aglycosylated full-length IgGs in E. coli, or E-clonal antibodies, against an antigen of interest from combinatorial libraries of VH and VL genes. Finally, these technologies have also been used for selection of aglycosylated Fc variants that selectively bind certain FcγRs (Jung et al., 2010).

Systems for multimerization, surface display, extracellular secretion and intracellular expression of antibody fragments in E. coli

Albeit the expression of Ab fragments and full-length IgGs in the periplasm of E. coli is the most common strategy, other Ab formats and expression systems have been developed, which have potential for specialized applications.

Protein engineering provided dimerization and multimerization of Fab, scFv, and sdAbs to produce bi-, tri- and multi-valency for increased antigen avidity (apparent affinity; Cuesta et al., 2010). This has been accomplished by joining monovalent fragments by means of short peptide linkers (e.g. Diabodies; Perisic et al., 1994; Conrath et al., 2001) and stable dimerization and oligomerization domains, such as CH3 domains and Fc regions derived from natural Igs (e.g. Minibodies and Fc fusions; Hu et al., 1996; Li et al., 2000), Leucine zippers (e.g. ZIP-miniantibodies; Pack & Pluckthun, 1992), dimeric Alkalyne Phosphatase from E. coli (Furuta et al., 1998), Barnase-Barstar complex (Deyev et al., 2003), trimeric coiled coils (Fan et al., 2008; Cuesta et al., 2009), tetrameric Streptavidin (Dubel et al., 1995; Cloutier et al., 2000), and pentameric B-subunits of toxins (Zhang et al., 2004), among others. In addition these multimerization strategies enable the generation of bi- and multi-specific Ab molecules for the simultaneous recognition of several antigens. For instance, diabodies that simultaneously bind the target antigen and recruit TNFα (Schmidt & Wels, 1996), serum Igs (Holliger et al., 1997), the complement (Kontermann et al., 1997), T cells (Helfrich et al., 1998; Kipriyanov et al., 1998), or NK cells (Arndt et al., 1999). Bi-specific diabodies based on two sdAbs have also been constructed for increasing circulating half-life in vivo through binding to serum albumin (Harmsen et al., 2005; Coppieters et al., 2006; Roovers et al., 2007; Tijink et al., 2008).

Alternative expression strategies have been developed for the production of Abs in E. coli. For instance, functional scFvs have been displayed on the surface of E. coli cells anchored to the OM with the chimeric Lpp-OmpA protein (Francisco et al., 1993; Daugherty et al., 1998) and the autotransporter (AT) β-domain of the Neisseria gonorrhoeae IgA protease (Veiga et al., 1999), but the frequent aggregation associated to scFv fragments limits their translocation across the OM (Wörn & Plückthun, 1999; Veiga et al., 2004). Contrary to scFvs, VHHs have been efficiently displayed on the surface of E. coli cells with the AT β-domain of IgA protease from N. gonorrhoeae (Veiga et al., 2004), and the β-domain of EhaA from enterohemorrhagic E. coli O157:H7 (Marín et al., 2010).

Functional scFv and VHHs can be secreted to the extracellular media of E. coli cultures, at concentrations c. 0.5–5 mg L−1 in shake flasks, with the E. coli α-hemolysin (HlyA) transport system (Fernández et al., 2000; Fraile et al., 2004). In this expression system, a fusion between the Ab fragment and the C-terminal secretion signal of HlyA (Koronakis et al., 1989) is transported from the cytoplasm to the extracellular media, without a periplasmic step, through the HlyB/HlyD/TolC complex that connects the bacterial IM and OM (Koronakis et al., 2004). The Ab fragments secreted with the HlyA system do not accumulate in the periplasm of E. coli and elicit very low toxicity for the overproducing bacterium. This situation contrasts to the release of Ab fragments to the culture media owing to OM leakage and cell lysis after their Sec-dependent periplasmic overexpression in E. coli. The HlyA transport system could be particularly suited for continuous Ab fragment production in bioreactors and the delivery of Ab fragments by commensal E. coli strains in vivo (Fernández et al., 2000; Rao et al., 2005).

Another secretion strategy of interest for Ab fragments in E. coli is the use of the type-III secretion system (T3SS) form EPEC strains. T3SSs of EPEC and other Gram-negative pathogens form supramolecular protein complexes, or injectisomes, expanding the bacterial cell envelope and protruding from the cell surface with a needle-like structure and a tip translocon complex able to form a pore in the plasma membrane of eukaryotic cells (Knutton et al., 1998; Garmendia et al., 2005; Cornelis, 2006). Injectisomes allow bacteria to translocate a number of proteins (i.e. effectors) to the cytoplasm of the eukaryotic host cells, which subvert different cell functions to benefit infection (Galan & Wolf-Watz, 2006; Galan, 2009). Functional sdAbs (VHHs) have been translocated to the cytoplasm of human cells using the T3SS of EPEC (Blanco-Toribio et al., 2010). The VHHs were fused to the 20 amino acid N-terminal T3S signal from a natural EPEC effector (EspF). The fusions were injected by wild-type and attenuated EPEC strains into the cytoplasm of HeLa cells cultured in vitro at levels c. 105–106 molecules per cell (Blanco-Toribio et al., 2010). Escherichia coli bacteria remain extracellular during translocation of sdAbs to HeLa cells. Additional improvements are needed in the infecting E. coli strain for selective delivery of therapeutic sdAbs in the cytoplasm of human cells (intrabodies). The high-stability properties of VHHs make them excellent candidates to function as intrabodies to abrogate or modulate the activity of intracellular proteins (Kirchhofer et al., 2010; Peréz-Martínez et al., 2010; Vercruysse et al., 2010).

Functional Ab fragments have also been expressed in the cytoplasm of E. coli cells. As most Ig V domains contain a conserved disulfide bond that is required for the stability of the folded polypeptide (Wörn & Plückthun, 2001), their expression in the reducing conditions found in the cytoplasm of E. coli mostly produce misfolded polypeptides unable to bind the antigen. Lack of disulfide bonds in the cytoplasm of E. coli is owed to the thioredoxin and glutaredoxin pathways that keep free thiol groups in a reduced state (Ritz & Beckwith, 2001; Kadokura et al., 2003). Production of functional Fab and scFv fragments in the cytoplasm of E. coli was achieved in double mutant strains in trxB (thioredoxin reductase) and gor (glutathione oxidoreductase) that simultaneously coexpress in the cytoplasm the chaperone Skp or the disulfide bond isomerase and chaperone DsbC, both devoid of the N-terminal signal peptide (ΔssSkp and ΔssDsbC, respectively; Levy et al., 2001; Jurado et al., 2002). In E. coli trxB gor cells coexpressing ΔssSkp or ΔssDsbC, disulfide bonds are formed correctly in cytoplasmic Fab and scFv fragments. It has also been demonstrated that N-terminal fusions between thioredoxin (Trx1) and scFv fragments allow their correct folding in the cytoplasm of E. coli trxB gor cells (Jurado et al., 2006b). In this case, coexpression ΔssDsbC is not needed because Trx1 acts itself as a chaperone assisting the proper folding of the scFv (Jurado et al., 2006b). Using these strategies, immune libraries of scFvs have been screened in vivo for selection of intrabodies blocking a prokaryotic transcriptional activator (Jurado et al., 2006a) and a relaxase involved in plasmid conjugation in E. coli (Garcillan-Barcia et al., 2007). Intrabodies selected in vivo with E. coli trxB gor cells may be stabilized later for correct folding in the reducing environment of the cytoplasm of wild-type cells. Interestingly, APEx has been employed for selection of stable scFvs able to fold in reducing environments taking advantage of their expression in the periplasm of E. coli dsbA mutants (Seo et al., 2009), in which oxidation of disulfide bonds in periplasmic Ab fragments is abolished (Jurado et al., 2002). Using this methodology highly stable scFvs that fold in the periplasm of E. coli dsbA mutants were selected and shown to be active when expressed in the cytoplasm of wild-type E. coli cells.

Concluding remarks

The Ig-like domain is frequently found in E. coli and enterobacterial extracellular and cell surface proteins that function in adhesion and cell host invasion, especially structural and components of multiple fimbrial systems and members of the intimin and invasin family of OM proteins. Periplasmic fimbrial chaperones and OM ushers involved in the assembly of fimbriae also contain Ig-like domains that are essential for their biological function. But the presence of the Ig-like domain is not limited to adhesive organelles and their assembly machineries. Ig-like domains are found in enzymes with oxidoreductase and hydrolytic activities, ABC transporters, sugar-binding and metal-binding proteins. Most of these domains are exposed to the periplasm where general periplasmic chaperones, PPIases and DSB forming and isomerization enzymes participate in their folding. In the light of the ‘skills learnt’ by E. coli during evolution for expression, secretion, and folding and their endogenous Ig-like-containing proteins, it is not surprising the biotechnological succes obtained in the heterologous expression of full-length antibodies and small antibody fragments in the periplasm of this bacterium and its filamentous bacteriophages. A better understanding of the structure and the mechanisms of secretion, folding and assembly of the natural Ig-like-containing proteins of pathogenic E. coli strains will not only provide a rationale basis for the design of new anti-microbial compounds to combat infection but will help to improve current expression systems of heterologous Igs in nonpathogenic laboratory E. coli strains and to expand their potential biotechnological applications.

Note added in proof

While this manuscript was at the proof stage the crystal structures of the β-barrels of Intimin and Invasin have been reported (Fairman et al., 2012). These structures reveal β-barrels of 12 antiparallel β-strands with a peptide linker that runs from the periplasm toward the extracellular space through the internal barrel pore.


This work has been supported by Grants of the Spanish Ministry of Science and Innovation (BIO2008-05201; BIO2011-26689), the Autonomous Community of Madrid (S-BIO-236-2006; S2010-BMD-2312), CSIC (PIE 201120E049), ‘la Caixa’ Foundation, and the VI Framework Program from the European Union (FP6-LSHB-CT-2005-512061 NoE ‘EuroPathogenomics’). The authors declare that there is no conflict of interest.